Biogeography of the Weakly Electric Knifefish Gymnotus Carapo (Teleostei: Gymnotidae) in the Guianas

Biogeography of the weakly electric knifefish Gymnotus carapo (Teleostei: Gymnotidae) in the Guianas

Emma S. Lehmberg

A thesis submitted in conformity with the requirements for the degree of Master of Science Ecology and Evolutionary Biology University of Toronto

Biogeography of the weakly electric knifefish Gymnotus carapo (Teleostei: Gymnotidae) in the Guianas

Emma S. Lehmberg

Master of Science

Ecology and Evolutionary Biology University of Toronto

2015 Abstract

The electric knifefishes (Gymnotiformes) are widely distributed across South and Central

America, with the highest species concentrations occurring in the Amazon and Orinoco basins.

Riverine features such as waterfalls and rapids can cause disjunct populations to form between highland and lowland areas in these basins. The Guiana Shield provides a good model to study the genetic differences between populations precisely because it has upland and lowland areas with extant populations of Gymnotiformes. To examine genetic divergence between highlands and lowlands, mitochondrial (cytochrome b) and nuclear (S7) DNA was sequenced for members of the Gymnotus carapo species complex (Gymnotidae). Population and phylogenetic analysis indicate a distinct split between upland and lowland populations, with those species in the highlands showing greater genetic similarity to populations from the Amazon basin.

Acknowledgements

First and foremost, thank you to my advisor, Nathan Lovejoy. Always a well of information, he has been patient and supportive throughout the duration of this project. This past year has been a growing one, and I will carry the things he has taught me into the next stage of my research life.

My timely completion of my thesis would also not have been possible without the support of the entire Lovejoy lab. I owe huge thanks to Dominik Halas, for being an endless source of information about all things phylogenetic and sharing that knowledge without hesitation; Ahmed

Elbassiouny, who taught me so much in the lab and continues to be inspiring in more ways than one; Alex Van Nynatten, who taught me the “five minute” rule and always gave thought- provoking feedback; Charmaine Condy, who was always supportive and kind, especially when I needed it most; and to Matt Kolmann, who patiently answered my questions and aided in the quiet proliferation of office aquaria and their inhabitants. Thanks also to Megan McCusker,

Thanara Rajakulendran, Frankie Janzen and Michael Dobrovetsky for all the office chats and thoughtful comments.

Many people were responsible for the collections that made this project possible, not least among them Hernán López-Fernández, Devin Bloom, and William GR Crampton, as well as Matt

Kolmann. Kristen Brochu and Gabrielle Malcolm provided additional sequences that helped round out the datasets.

Thank you to my supervisory committee, Hernán López-Fernández and Maydianne Andrade.

Your feedback was invaluable and helped a great deal.

Thanks to my friends and family, who encouraged and supported me throughout my entire iii masters. This thesis is for three of them: my mother, who taught me how to look; my father, who taught me how to ask; and Noel, who first showed me the beauty of a fish.

Table of Contents

Abstract………………………………………………...………………………………………….ii

Acknowledgments………………………………………………………..………………………iii

Table of Contents……………………………………………………………………………...….v

List of Tables…………………………………………………………………………………….vii

List of Figures………………………………………………………………………………...…viii

Chapter 1 Introduction…………………………………………………………………….……....1

1 Biogeography of the Guianas………………………...……………………………….…1

1.1 Riverine Barriers and Connectivity…………………………………………...……….2

1.2 Model Species: Gymnotus carapo……………………………………………….…....4

1.3 Objectives, Hypotheses, and Predictions………………………………………….…..7

1.4 Significance……………………………………………………………………..……..8

Chapter 2 Materials and Methods…………………………………………………………..……..9

2 Taxon Sampling………………………………...………………………………...……..9

2.1 Mapping and Elevational Division………………………………………………..…10

2.2 DNA Extraction………………………………………………………………..…….10

2.3 PCR and Sequencing………………………………………………………..………..11

2.4 Sequence Alignments and Matrices………………………………………..………...13

2.5 Phylogenetic Analyses………………………………………………………….……13

2.6 Sequence Divergence Calculations…………………………………………………..15

Chapter 3 Results……………………………………………………………...…………………16 v

3 Haplotype Sharing Between Upland and Lowland Guianas…………………...………16

3.1 Biogeographic Relationships Between Upland and Lowland Gymnotus carapo……17

3.2 Continent-wide Gymnotus carapo Relationships……………………………………19

Chapter 4 Discussion…………………………………………………………………………….21

4 Upland and Lowland Division of Guianas Gymnotus carapo…...…………………….21

4.1 Biogeographic Patterns of the Guiana Shield Gymnotus carapo…………………….23

4.2 The Gymnotus carapo Species Complex…………………………………………….25

References………………………………………………………………………………………..27

List of Tables

Table 1: Members of the Gymnotus carapo species complex included in this study……………33

Table 2: Gymnotus carapo collected from the Guiana Shield, and included in this study..……..37

Table 3: Primers used for S7 and cytb in this study ……………….……………………………45

Table 4: Average pairwise distances between allopatric populations of the Gymnotus carapo species complex included in this study...... 45

vii

List of Figures

Figure 1: Profile of different elevational categories of the Guiana Shield. Altitude divisions taken from Hammond (2005)…………………………………………………..………………………40

Figure 2: The drainages and major rivers of the Guiana Shield in Guyana and Suriname………41

Figure 3: Fish assemblages found above and below the Kaieteur Falls and the Tumatumari Cataract of the Potaro River. Modified from Hardman et al (2002).……….……………………42

Figure 4: Proposed relationships between members of the Gymnotus carapo clade based on a maximum parsimony analysis of morphological and meristic characters. Figure adapted from Albert et al. (2005)……………………………..………………………………………………...43

Figure 5: Collection locations of Gymnotus carapo from the Guiana Shield.………………....44

Figure 6: Cytb gene tree showing relationships between allopatric populations of Gymnotus carapo and closely related species. …………………….…………………..……………………46

Figure 7: S7 gene tree showing relationships between populations of G. carapo and closely related species.…………………………………………………..……………………………….47

Figure 8: Tree showing relationships of G. carapo and closely related species when S7 and cytb are concatenated and run as a single dataset.…………………………….………………………48

Figure 9: *BEAST analysis of cytb and S7 data showing relationships between populations of Gymnotus carapo and closely related species included in this study……….…………………...49

viii 1

Chapter 1 Introduction 1 Biogeography of the Guianas

South America is a continent of high ichthyofaunal diversity, and it is likely that physical changes in river connections and drainage patterns have contributed to the diversification of

South American fishes (Lundberg et al. 1998). The Guiana Shield region (Guianas) of northwestern South America is famous for its striking topography. This region includes sheer- edged tabletop mountains (tepuis), which rise above a slightly lower main platform known as the pantepui, which itself is elevated above the coastal lowland river drainages (Lujan and

Armbruster 2011; McConnell 1968; Rull 2005) (Figure 1). The high elevation regions of the

Guianas exhibit considerable isolation, with resulting biological endemicity. The Guiana highlands and uplands are home to 42% of vascular plants and approximately 100 birds are exclusive to these highland areas (Berry and Riina 2005; Zyskowski et al. 2011). Throughout the upland Guianas, ichthyological sampling is sparse, but estimates of endemicity of the upper

Mazaruni, an isolated pantepui river, are placed between 67 - 95% (Alofs et al. 2014).

The Guianas region (Figure 2) is centered in northeast South America and encompasses parts of

Venezuela, Guyana, Suriname, and French Guiana, covering approximately 2 288 000 km2

(Hammond 2005). The region is bounded to the north and east by the Orinoco River, to the south by the Amazon basin, and to the east by the Atlantic Ocean. Forty-seven major rivers currently contribute to the major Guiana drainages, with the largest basins being the Orinoco and

Essequibo. The Orinoco is the second largest basin in South America, while the Essequibo basin encompasses a number of major highland rivers found in central Guyana, including the

Essequibo itself, the Cuyuni, the Mazaruni, and the Potaro. A number of smaller drainages that flow directly to the oceans have been collectively referred to as the Orinoco and Coastal drainage basins (Albert, Petry and Reis 2011). The Berbice, Commewijne, Coppename, Suriname and

Marowijne drainages are included in this group of coastal basins (Figure 2).

Geographically, the Guianas exhibit a gradient of altitudes from the tepui summits to the coastal floodplains of northeast South America. The highest peak, Pico Neblina, is approximately 3 000 meters above sea level (m-asl), with the average elevation of the pantepui at 1 000 m-asl, and the lowlands found at an average of 150 m-asl (Figure 1; Lujan and Armbruster 2011). As a result of this variation in altitude, many rivers of the Guiana Shield experience a numbers of abrupt changes in elevation (as rapids and waterfalls). With the exception of coastal drainage basins, most of the major rivers in the region originate in the highlands or uplands of the Guianas and then descend through waterfalls or rapids until they reach the coastal plains and eventually empty into the Atlantic. The topography of this area likely affects gene flow and patterns of community structure in fishes. Indeed, the aquatic habitats of the Guianas are an ideal system for examining the effects of waterfalls, rapids, and river isolation on gene flow and population differentiation of riverine fishes.

1.1 Riverine Barriers and Connectivity

Previous studies have shown that rapids represent a significant barrier to the dispersal and general movement of aquatic organisms (Torrente-Vilara et al. 2011; Rahel 2007). Hardman et al. (2002) conducted a survey that spanned two major riverine barriers on the Potaro River, as well as above and below rapids on the Essequibo (Figure 3). They found that species assemblages of fishes on either side of significant waterfalls and rapids were different,

3 suggesting that these landscape features act as riverine barriers for certain species. For example, there were nine species that only occur above Kaieteur Falls in the upper Potaro, while there are

51 species that only occur below the Falls. An additional 19 species (including Gymnotus carapo) are distributed both above and below the Falls (Figure 3).

River barriers not only affect assemblages of fishes, but can also have intraspecific effects.

Rapids and waterfalls can act as natural filters for some organisms, separating one population from another and allowing genetic diversification to take place, even on a very localized scale

(Makert, Schelly and Stiassny 2010; Lovejoy and de Araújo 2000). While gene flow across barriers may occur, it may happen in an asymmetrical fashion, with genetic diversity decreasing towards headwaters, which suggests that most gene flow occurs in a downstream fashion (Alp et al. 2012). If a barrier exists for a significant amount of time and gene flow is minimal or non- existent across it, genetic drift between populations can occur and new species may eventually arise.

The ability of fishes to disperse across barriers is poorly understood. Unlike some artificial barriers (e.g. dams), which are impenetrable, rapids may be permeable to aquatic organisms.

However, the ability of fishes to cross natural barriers is dependent on a number of factors, including life history, body size, morphology, and ecology (Makert, Schelly and Stiassny 2010;

Winemiller, Agostinho and Caramaschi 2008). Rheophilic fishes can navigate fast-flowing, high- altitude streams (Lujan and Armbruster 2011), and large doradid catfishes are known to migrate up steep natural gradients on the way to spawning grounds (Barthem and Goulding 1997), but the ability of many smaller neotropical fishes to surpass riverine barriers is yet to be thoroughly investigated.

Despite its unique geological and biological history, the biogeography of the Guianas remains understudied. As such, it represents an area that is rich with possibility for phylogeographic and biogeographic research. The region is topographically and geologically complex, with rivers interrupted by major waterfalls and rapids; for some fishes, these seem to represent a barrier to movement, though others seem to be able to surpass them. Genetic approaches should allow us to understand the potential effects of barriers on gene flow and isolation of populations.

1.2 Model Species: Gymnotus carapo

Gymnotus is a genus of electric knifefish belonging to the neotropical family Gymnotidae (Order

Gymnotiformes). Gymnotiforms are well known for their ability to generate an electric field and use this system to sense their environment (electroreception). The field is generated from a myogenically- or neurogenically-derived organ called an Electric Organ (EO). Each species produces a unique Electric Organ Discharge (EOD), and this species-specific EOD has made them an ideal model organism for examining genetic, behavioural, and morphological diversification.

Gymnotus is the most speciose genus of the Gymnotiformes, with 38 species described to date, and the expectation of more to be delineated in the future (Crampton et al. 2013; Maxime, 2013;

Rangel-Pereira 2014). Within the genus, Gymnotus carapo is the most widely distributed, and one of the most well studied species; it is found throughout the Orinoco, Amazonian, Guiana

Shield, and northeast Atlantic Brazilian drainages, as well as on Trinidad (Albert and Crampton

2003). Though G. carapo is the fifth most widely distributed fish in South America, it is not known from Central America, the western slope of the Andes, or in the rivers south of the city of

Recife, Brazil (Albert and Crampton 2003; Albert, Petry and Reis 2011). This fish generally

5 inhabits shallow water with very little current, and has been collected from small channels and floodplains. Adults are often found in the roots of woody plants and beneath floating macrophytes, and males are known to be territorial (Albert and Crampton 2003). Nesting and paternal care has been observed in populations of Gymnotus carapo found in Trinidad

(Crampton and Hopkins 2005), with mouthbrooding observed in captive populations

(Kirschbaum and Wieczorek 2002).

Currently, Gymnotus carapo is thought to include six different allopatric populations across northern South America, delineated by morphological and meristic characters. These include the

Eastern Amazonian region (EA), the Rio Branco basin (RO), the Western Amazon (WA), the

Rio Madeira basin (MD), the Parnaiba and Itapicuru basins (PI), and finally the populations found in the Orinoco basin, across the Guianas, and on Trinidad (GO) (Albert and Crampton

2003; Albert et al. 2005). Additionally, Lovejoy et al. (2010) showed that both nuclear and mitochondrial phylogenetic trees indicated a separation in allopatric populations of Gymnotus carapo. Individuals collected from the west and central Amazon basin resolve in a separate clade from members of the same species collected from the Orinoco basin, suggesting that movement of genes between these populations is low or non-existent.

Gymnotus carapo is also closely related to a number of other Gymnotus, which together form an unresolved phylogenetic group. Gymnotus arapaima, Gymnotus ucamara, Gymnotus choco, and

Gymnotus ardilai have been described as separate species based on morphology and EOD signal

(Albert and Crampton 2001; Albert and Crampton 2003; Crampton, Lovejoy and Albert 2003;

Maldonado-Ocampo and Albert 2004), but morphological and molecular analyses place them amongst G. carapo lineages. In a parsimony-based morphological phylogeny, the relationship between G. arapaima, G. choco and the allopatric populations of Gymnotus carapo is largely

6 unresolved (Albert et al. 2005; Figure 4), but G. arapaima was found to be closely related to

Western Amazonian population of G. carapo (Figure 4).

Molecular analyses also fail to resolve the monophyly of Gymnotus carapo with respect to closely related species. A Bayesian phylogeny of nuclear, mitochondrial and morphological data by Lovejoy et al. (2010) replicates the close relationship between G. ucamara, G. arapaima, and the west and central Amazonian populations of G. carapo. Nuclear and mitochondrial parsimony analyses show a more ambiguous relationship, similar to Albert et al.’s morphological analyses

(2005). G. ardilai, too newly described to be included in Albert et al.’s analyses, is included in a molecular analysis done by Brochu (2011) and is the sister lineage of G. carapo from the

Orinoco. For these reasons, it is necessary to include the close relatives of Gymnotus carapo when examining the relationships between the allopatric populations of G. carapo. Here, I will use the term “Gymnotus carapo species complex” to refer to the clade previously represented in morphological and molecular phylogenies that is inclusive of all lineages of Gymnotus carapo as well as several described and undescribed Gymnotus species that are closely related to Gymnotus carapo lineages (Albert et al. 2005; Brochu 2011; Lovejoy et al. 2010; Maxime 2013).

Gymnotus carapo is found across the lowland areas of the Guianas (Albert and Crampton 2003;

Hardman et al. 2002), and the type specimen actually originates from the lowland areas of

Suriname. It has been collected from a number of coastal tributaries and lowland rivers including the Demerara, Berbice, Commewijne, Rupununi, and the Suriname Rivers. G. carapo has also been collected from the pantepui (upland) region at varying elevations, in tributaries of the upper

Mazaruni and Potaro. Also, G. carapo was recently collected from Tafelberg tepui, an isolated

500 to 1000m elevation tepui in Suriname that is the easternmost of all the tepuis. For the upland collections, the habitats that collections have been taken from are separated from lowland rivers

7 by rapids or waterfalls that may have isolated upland and lowland populations (Alofs et al. 2014;

Hardman et al. 2002). Because of its wide distribution in the Guianas, and its presence in both upland and lowland regions, G. carapo is an ideal fish to study the effects of riverine barriers and isolation on gene flow.

1.3 Objectives, Hypotheses, and Predictions

My thesis has three main objectives that address the biogeography of the Gymnotus carapo species complex in the Guianas region.

First, given that Gymnotus carapo populations occur in upland and lowland habitats, is there haplotype sharing between these areas? I will address this by collecting sequence data from a mitochondrial gene (cytochrome b) and from a nuclear intron (S7) for multiple individuals from upland and lowland habitats. I will assess patterns of haplotype sharing between different habitats to test for gene flow or genetic isolation. Haplotype sharing is indicative of gene flow or recent dispersal across barriers; lack of shared haplotypes across barriers indicates no current gene flow between populations.

My second objective is to determine the biogeographic relationships among populations of the

Gymnotus carapo species complex in the Guianas. In particular, I will test whether upland and lowland populations from the Guiana Shield form monophyletic groups, indicative of historical or current dispersal of individuals within river drainages. Alternatively, I will test whether upland populations from different river drainages form monophyletic groups with respect to lowland populations, which would indicate historical or current dispersal between headwaters of river drainages.

My third objective is to assess species relationships and boundaries in the Gymnotus carapo species complex. I will do this by examining phylogenetic relationships and sequence divergence between species and populations in the Gymnotus carapo complex.

1.4 Significance

This project will contribute to the body of knowledge concerning neotropical fishes, particularly the Gymnotiformes, while providing a case study of biogeography and population divergence within a widespread fish species complex in the neotropics. It is one of the first ichthyological studies to examine genetic diversification across the landscape of the Guianas. My study will also contribute to longstanding investigations regarding the phylogenetic structure of the

Gymnotus carapo species complex, including its possible division into several separate species.

Chapter 2 Materials and Methods 2 Taxon Sampling

Tissues for this study were obtained from across the Guiana Shield region (Table 2; Figure 5).

Many of these tissues were collected in lowland and upland Guyana over a number of collection trips undertaken by the Royal Ontario Museum (ROM) between 1992 to 2014. WGR Crampton collected lowland tissues from several different Surinamese coastal locations. Additionally, DD

Bloom (Kansas State University) collected a single tissue of Gymnotus carapo in August 2013 from a river on Tafelberg tepui, located in eastern Suriname (locality 14; Figure 5). This is the first collection of Gymnotus carapo from the summit of a tepui, and represents an upland sample that is geographically separated from the rest of the Guiana upland samples.

Upland samples are shown in red in Figure 5. The samples come from three main sources: the upper Potaro (localities 8-11), the upper Mazaruni (localities 4-7), and from atop Tafelberg tepui

(locality 14). In all three cases, these upland localities are separated from lowland localities by rapids/waterfalls. The upper Mazaruni and tributaries experience no significant waterfalls across the pantepui, but a series of steep rapids or waterfalls separate them from the lowlands (Alofs et al. 2014), while the upper Potaro is separated from the lower reaches by the Kaieteur Falls, a 226 meter obstacle to movement. Tafelberg may experience some connection with the right arm of the Coppename river on the pantepui below, but several falls lead off the tepui to get to it.

Although the upper Potaro and upper Mazaruni samples are geographically adjacent, there is no current connection between the upper reaches of these rivers.

The lowland samples (localities 1-3, 12-13, and 15-18) were collected largely from coastal rivers in Guyana and Suriname, some of which are tributaries to major inland rivers. Two localities, 12 and 13, are further inland: 12 is from the upper reaches of the Berbice River, while 13 is the

Rupununi River, a river which runs through a lowland area found between the Pakaraima and

Kanuku mountain ranges of the Guiana Shield. It is a part of the upper Branco-Amazon drainage.

To provide geographical and phylogenetic context, I also included G. carapo samples from five of the six allopatric populations delineated in Crampton and Albert (2003), as well as samples of the closely related species G. ucamara, G. choco, G. ardilai, G. arapaima, G. mamiraua, G. n. sp. IGUA, G. n. sp. ITAP, G. bahianus, and G. sylvius, which belong to the Gymnotus carapo species complex (Table 1). Based on prior phylogenetic studies of Gymnotus (Albert et al. 2005;

Brochu 2011; Lovejoy et al. 2010), I included G. obscurus, G. varzea, G. curupira, and G. chaviro as appropriate outgroups (Table 1).

2.1 Mapping and Elevational Division

To categorize elevation for populations of G. carapo found on and around the Guiana Shield, the elevation gradient was divided into three categories: highland, upland, and lowland (Hammond

2005; Figure 1). Highland is defined as elevations over 1500 m-asl, while upland is between

1500 and 300 m-asl, and lowland is anything below 300 m-asl. To determine which category each collection fell into, latitude and longitude were plotted on a map using ArcGIS software and elevation maps acquired from DivaGIS. Drainage basin data were taken from the USGS

HydroSHEDS database. Using this information, it was determined that all G. carapo included in this study were either upland or lowland. No highland examples of G. carapo were included.

2.2 DNA Extraction

Tissue extractions were done with Qiagen DNEasy Blood and Tissue kits, and an attempt was

made to use a new kit each time a separate extraction was done to ensure the highest quality

yield of DNA from each sample. The Qiagen protocols for tissue extraction were used.

2.3 Polymerase Chain Reactions and Sequencing

A mitochondrial gene (cytochrome b) and the intron of a nuclear gene (S7) were amplified using

Polymerase Chain Reactions (PCR). The cytochrome b gene (cytb) is approximately 1100 base

pairs long, and is a component of respiratory chain complex III. It is a protein-coding gene, with

several characteristics that make it ideal for phylogenetic analyses. These include a small overall

size, a high rate of nucleotide substitution, particularly at synonymous sites, and exclusively

maternal inheritance of DNA (Graur and Li 2000). Having a high rate of nucleotide substitution

allows a fine scale examination of divergence within the species complex. Additionally, cytb has

been used to examine phylogenetic relationships in Gymnotus in previous studies because of its

phylogenetic signal (Brochu 2011; Lovejoy et al. 2010), and there is a wealth of other fish

sequences for comparison.

The nuclear ribosomal gene S7 is composed of seven exons and six introns, and is roughly 3930

base pairs long. The introns range in length from 339 base pairs to 920 base pairs, and four

through six code for a small nucleolar RNA (Cecconi et al 1996). S7 introns have been used as

markers for phylogenetic and population-level investigations. The polymorphisms noted in

introns one and two are higher than both the exons and four other introns. It is this characteristic

that makes them useful in phylogenetic and population studies (Alda et al. 2013; Bossu and Near

2009; Chow and Hazama 1998; Lavoué, Sullivan and Hopkins 2003). In some previous studies on teleosts, the first intron of S7 (hereafter, S7) has been found to be unreliable at resolving relationships between distantly related taxa (i.e. deeper nodes on a phylogeny) possibly because of fast rates of evolution (Corrigan and Beheregaray 2009). However, S7 is expected to be a suitable marker for resolving close relationships within the allopatric Gymnotus carapo populations.

PCRs for cytb were conducted using 25 µL reaction volumes made up of 2.0 µL of DNA, 14.8

µL of de-ionized water (ddH2O), 2.5 µL of Taq polymerase buffer KCl-MgCl2, 2.0 µL MgCl2,,

1.5 µL of 10 µM deoxynucleotide triphosphates (dNTPs), 1.0 µL of 10 µM forward primer, 1.0

µL of 10 µM reverse primer, and 0.2 µL of Taq polymerase. Universal vertebrate primers

GLUDG.L, cytbR, and cytbF were the most commonly used in this project. Primers specific to

Gymnotus were developed for tissues that were difficult to amplify (see Table 3 for primers and their sequences). Thermocycler conditions for cytb were conducted using the following protocol:

95°C for 30s for initial denaturation followed by 95°C for 30s to denature DNA; 50.0°C for 60s to anneal DNA, and 72°C for 90s for elongation. This cycle was repeated 35 times, followed by

300s at 72°C for further extension.

The first intron S7 marker had not previously been examined in Gymnotus. Primers designed for teleosts (Chow and Hazama 1998) were used for amplification (Table 3). PCRs for S7 were conducted using 25 µL reaction volumes of 11.875 µL ddH2O, 2.5 µL of Taq polymerase buffer

KCl-MgCl2, 1.5 µL of MgCl2, 2.0 µL of 10 µM dNTPs, 3.0 µL of forward primer and reverse primer, respectively, and 0.125 µL of Taq polymerase. Thermocycler conditions for S7 were

95°C for 30s of initial denaturation followed by 30 cycles of 95°C for 30s denaturation; 50.6 –

55.5°C for 60 annealings, and 72.0°C for 120s of elongation. An extension period of 600s followed the cycles.

All PCR products were visualized using gel electrophoresis. Products were run on a 0.8% agarose gel preloaded with 5.0 µL of Amresco EZVision in-gel stain. 5.0 µL of DNA was pipette-mixed with 3.0 µL of Thermoscientific 6x loading dye. The electrophoresis was run for

30 minutes at 80 Volts and 70 milliAmperes.

Purifications of PCR products were done with Affymetrix ExoSAP-IT according to the manufacturer protocol. Purified product was then sent to SickKids Centre for Applied Genomics for sequencing.

2.4 Sequence Alignment and Matrices

Forward and reverse sequences for each individual were imported into Geneious 6.1.7 and aligned using the highest sensitivity of the de novo assembly function. Sequences were then edited for ambiguities, and contigs were generated using the consensus between the forward and reverse sequences. Once contigs were completed, multi-sequence alignments were performed using the CLUSTALW algorithm. A gap opening cost of 10, a gap extension cost of 20, and free end gaps was imposed to increase alignment accuracy for cytb, while the S7 alignment was conducted using the default CLUSTALW parameters.

For cytb, I obtained 1152 base pairs for 113 individuals. No insertions or deletions were noted.

For S7, I obtained 821 base pairs for 51 individuals. Ten indels were observed in S7. I created three different matrices using these methods: a cytb matrix, with sequences from 113 individuals; an S7 matrix, with sequences from 51 individuals; and a concatenated dataset, in which cytb and

S7 were combined for each individual which had both sequences available, (1995 characters for

48 individuals). Using the general teleost S7 primers designed by Chow and Hazama (1998), I had a great deal of difficulty amplifying a number of G. carapo from different regions, resulting in a smaller dataset for that gene.

2.5 Phylogenetic Analyses

Alignments were exported from Geneious into Mesquite (Maddison and Maddison 2014), to convert to nexus and phylip file formats. Each of the three datasets (cytb, S7, and concatenated) was analyzed using PartitionFinder (Lanfear et al. 2012) to determine the most appropriate partition schemes and model of molecular evolution for each partition, based on the Akaike

Information Criterion (AIC). In each analysis, Gymnotus varzea was used as an outgroup.

Bayesian and maximum likelihood analyses were conducted, using MrBayes (Huelsenbeck et al.

2001), and RAxML (Stamatakis 2006), respectively. For Bayesian analyses, each analysis was run until the average standard deviation of split frequencies was below 0.01, an indication that the algorithm has settled around an optimal tree space.

For analysis of cytb, the gene was partitioned by codon position for both Bayesian and maximum likelihood analyses. The cytb analysis in MrBayes was run for 60 million generations. Codon positions one and two were assigned the GTR+G+I model, and codon position three was assigned the GTR+G model. The same partition was used for RAxML. A total of 50 searches were used to generate the best ML tree. Node support was estimated by 1000 bootstrap replicates. For Bayesian and maximum likelihood analysis of S7, the gene was not partitioned and the GTR+G model was specified. MrBayes was run for 80 million generations. For

RAxML, a total of 50 searches were used to generate the best ML tree. Node support was estimated by 1000 bootstrap replicates.

The concatenated dataset was partitioned by gene for all analyses, according to the model recommendation generated by PartitionFinder. For Bayesian inference, the concatenated dataset was run with all nucleotides assigned the HKY+G+I model, except for second codon positions of cytb, which were assigned the GTR+G model. This analysis was run for 25 million generations.

For maximum likelihood, the concatenated dataset was run with all nucleotides assigned the

GTR+G model, except for first codon positions of cytb, which were assigned the GTR+G+I model. A total of 50 searches were used to generate the best ML tree. Node support was estimated by 1000 bootstrap replicates.

*BEAST (Drummond, Suchard and Rambaut 2012) was used to implement a Bayesian species tree approach. In this analysis, I used geographically-defined lineages as proxies for species, since the species identities within the Gymnotus carapo complex are unclear. This analysis implemented using the GTR+G+I model of molecular evolution for both genes, a rate of mutation = 0.01, a relaxed clock, and the birth-death model. Three separate runs were conducted for 100 million generations, and log files were examined in Tracer v1.6 (Rambaut et al. 2014) before creating a single tree file in TreeAnnotator v1.8.1, part of the BEAST software package.

2.6 Sequence Divergence Calculations

Using PAUP* (Swofford 2003), uncorrected pairwise distances between all sequences were calculated, and the average pairwise distances were calculated for G. carapo populations and lineages defined by the phylogenetic analyses.

Chapter 3 Results 3 Haplotype Sharing Between Guianas-upland and Guianas-lowland Regions

I collected cytb sequence data from 26 individuals from upland pantepui rivers, including the

Upper Mazaruni and Upper Potaro Rivers, as well as from Tafelberg tepui in Suriname. From these 26 individuals, a total of three different haplotypes were observed (Table 2). I also sequenced 22 individuals from multiple rivers in lowland Guyana and Suriname and observed eight haplotypes in this gene (Table 2).

Within upland regions of the Guiana Shield, cytb haplotypes were shared between localities within the same river drainage. For example, individuals from localities 5, 6, and 7 within the upper Mazaruni shared haplotype B. Haplotype B is also shared with localities 8, 9, 10, and 11, all found in the upper Potaro River. The main stem of the Upper Mazaruni has a unique haplotype, A. The Gymnotus carapo from Tafelberg has a unique haplotype (I).

Lowland rivers also showed cytb haplotype sharing between localities. All G. carapo collected from the Commewijne share the same sequence (haplotype D), and this haplotype was found in the Suriname River as well. Two individuals from the Waini River share a haplotype (E), with an individual from a tributary of the same river having a unique haplotype (F). Two fish from the

Suriname River share a haplotype (G), and four fish from the Suriname, Berbice and Waini

Rivers share a haplotype (H).

Seventeen fish from the Guianas were sequenced for S7: fourteen from upland localities including the upper Mazaruni and upper Potaro (localities 4 - 8, and 10), and three from lowland

17 localities, including the Rupununi (locality 12), the Suriname river (locality 16), and the

Commewijne (locality 17) (Figure 6). Unfortunately, difficulties with PCR amplifications prevented additional sequencing of lowland G. carapo individuals. I found four S7 haplotypes.

(Table 2). One of these (haplotype aa) was found only in fishes from the upper Mazaruni, and all fishes from the upper Mazaruni exhibited this haplotype. Haplotype bb was more widespread, and was found in fishes from the upper Potaro, the Rupununi, and Tafelberg (and the west

Amazonian G. arapaima). Finally, haplotypes cc and dd were found, respectively in fishes from the lowland Suriname and Commewijne Rivers.

For both cytb and S7, I found no evidence for haplotype sharing between the pantepui and coastal regions of the Guiana Shield. However, for S7, one haplotype (bb) was shared between upland regions (upper Potaro and Tafelberg) and the lowland Rupununi region. The average pairwise sequence divergence between haplotypes from upland and lowland regions was calculated as 0.012 in cytb and 0.006 in S7 (Table 4). Phylogenetic trees (see next section) show that fishes from these upland and lowland regions of the Guianas are generally not each other’s closest relatives.

3.1 Biogeographic Relationships Between Upland and Lowland Guianas Gymnotus carapo

To examine biogeographic relationships of upland and lowland populations, gene trees were created for both cytb and S7, along with a tree generated using concatenated data, and a *BEAST species tree.

In the cytb topology (Figure 6), haplotypes from the upper Mazaruni and upper Potaro form a monophyletic clade (hereafter, the “Guianas-upland” clade). This clade is most closely related to

18 the individual from the Rupununi lowlands area. The Guianas-upland + Rupununi clade is most closely related to a group of G. carapo and several species (G. arapaima, G. n. sp. LORE, and G. ucamara) from the Amazon region. Thus, analysis of cytb indicates that G. carapo individuals from the Guiana uplands (with the exception of the Tafelberg individual) are not closely related to geographically proximate individuals from the Guiana lowlands. Instead, the Guiana-upland clade is most closely related to populations from the Rupununi and the more geographically distant Amazon regions.

Some of the patterns observed in cytb are also observed in the S7 topology (Figure 7), although this dataset was smaller and exhibited less variation. The upper Mazaruni and upper Potaro samples are closely related in a well-supported clade; however, this clade also includes the

Rupununi sample, the Tafelberg sample, and haplotypes from Gymnotus arapaima from the

Amazon region. As in the cytb phylogeny, samples from lowland rivers of the Guianas are not closely related to rivers from the uplands. However, in the S7 phylogeny, the two lowland

Guianas samples do not make up a monophyletic group.

The phylogenetic separation of uplands and lowlands Guianas samples is also evident in the concatenated phylogeny (Figure 8), and the *Beast tree (Figure 9). In both of these analyses, the

Guianas-upland clade is monophyletic, and is the sister taxon to the Rupununi sample. Also, in both these analyses, the Guianas-lowland clade is resolved. The Guianas-upland and Guianas- lowland clades are not each other’s closest relatives in these trees. The position of the Tafelberg sample differs across the concatenated and *BEAST analysis; this sample is positioned as the sister taxon to the Guianas-lowland clade in the *BEAST tree (Figure 9), but is more closely related to the Guianas-upland clade in the concatenated trees (Figure 8).

3.2 Continent-wide Gymnotus carapo Relationships

Gymnotus carapo individuals resolve as several distinct phylogenetic lineages, each associated with a geographical area. Gymnotus carapo individuals do not make up a monophyletic clade, because several allied species, which have been previously assigned to the G. carapo species group in previous morphological and molecular phylogenies, are sister taxa of some of the

Gymnotus carapo lineages.

According to the cytb phylogeny, which has the broadest sampling of populations, six distinct G. carapo lineages can be identified. These include lineages from: Guianas-upland (which may include the Rupununi sample), Amazon (including the Ucayali, and Tefé), Guianas-lowland

(which may include the Tafelberg sample), Xingu/Tapajos, Beni, and Orinoco (Figure 6). In some cases, these G. carapo lineages are most closely related to named or undescribed species of

Gymnotus that occur in the same geographical region. For example, G. carapo from the Amazon is closely related to G. ucamara, G. arapaima, G. n. sp, LORE, all of which are distributed throughout the Amazon region. In other cases, G. carapo lineages are sister taxa to species found in different geographical areas. For example, G. carapo from the upper Madeira (Beni) is closely related to G. n. sp. ITAP and G. sylvius from south and southeast coastal regions.

The level of genetic divergence between G. carapo lineages and sister taxa varies. Some lineages are quite genetically distinct (e.g., Guianas-upland, and Guianas-lowland, G. carapo Beni), and form monophyletic haplotype clades. However, in other cases genetic divergence is very low, and G. carapo individuals do not form monophyletic haplotype clades. For example, G. carapo

Ucayali do not form a distinct monophyletic lineage in relation to the closely related species G. ucamara, G. arapaima, G. n. sp, LORE.

While cytb provides the most comprehensive portrait of continent-wide Gymnotus carapo relationships, due to denser sampling of individuals, the nuclear S7 locus provides additional support for many of the cytb-based relationships. Many of the clades supported by cytb are replicated in the S7 tree, including the G. sylvius, G. n. sp. ITAP, and G. carapo Beni clade, and the G. carapo Orinoco, G. ardilai, and G. choco clade. One interesting difference between cytb and S7 is the close relationship, according to S7, between G. arapaima / G. carapo Tafelberg and the upland Guiana fishes.

Chapter 4 Discussion 4 Upland and Lowland Division of Guianas Gymnotus carapo

Gymnotus carapo does not share haplotypes between upland and lowland Guianas habitats and there is genetic divergence between the upland and lowland Guianas clades (approximately 1.2% divergence in cytb). Also, phylogenetic results show that upland and lowland clades are not each other’s closest relatives. This pattern is also evident for the single individual from Tafelberg tepui—this sample was closely related to lowland individuals (based on cytb), but it was genetically distinct, suggesting geographic isolation. These analyses also suggest that river features such as waterfalls and rapids prevent genetic admixture between geographically adjacent populations of fishes. Riverine barriers between upland and lowland areas appear to have a great influence on the movement of G. carapo individuals.

Current large-scale barriers to movement in the relevant Guiana drainages include four planation surfaces, which are areas of the earth that have been worn away by fluvial, wind or marine processes. Each of the planation surfaces in the Guianas is sharply delineated by abrupt decreases in elevation from one to the next (Hammond 2005). The four planation surfaces are separated by an average difference of 200 - 400 m, through which several major rivers included in this study descend, most notably the Mazaruni and the Potaro (McConnell 1968). The planation surfaces are responsible for the rapids and waterfalls that divide the rivers of the

Guiana Shield (Lujan and Armbruster 2011; McConnell 1968). As a result of abrupt changes in altitude, both the upper Mazaruni and the upper Potaro are physically separated from the lowlands by a series of riverine barriers. As the upper Mazaruni leaves the pantepui region, it

22 encounters a series of rapids in succession, which likely isolates the upper region (Alofs et al.

2014). The Potaro flows over the Kaieteur Falls, which represents a 226-m change in elevation from pantepui to lowlands (Hardman et al. 2002). The Potaro also crosses the Tumatumari

Cataract in the lowlands, further separating the pantepui from the lowland Guiana region.

Other studies have demonstrated the importance of river features for genetic structure in fishes.

Some fishes show fine genetic structuring in concordance with current landscape topology, while others show evidence of past drainage basins and river connections; it is common to find signatures of both (Guy, Creswell and Banks 2008; Lovejoy and De Araujo 2000; Strange and

Stepien 2007; Waters et al. 2001; Wong, Keogh and McGlashan 2004). Gene flow is dependent on the ability of fishes to navigate their environments successfully (Hutchison and Templeton

1999), and connectivity of rivers (i.e. lack of large barriers to movement) has a great influence on that. Genetic patterns of connectivity in Gymnotus carapo appear to reflect riverine barriers in the Guianas.

The separation of upland and lowland populations has implications for, and reflects, high levels of overall endemicity that have been reported from the tepui and pantepui area. In fishes, several species that are endemic to the upland regions of the Guianas have recently been described, including members of the Hypopomidae, and the Loricariidae, as well as three characiforms

(Armbruster and Taphorn 2011; Gery and Zarske 2002; Lujan 2008; Lujan et al. 2013;

Maldonado-Ocampo et al. 2014; Taphorn, López-Fernández and Bernard 2008). Many of these species are thought to have been isolated from the closest relatives for a significant amount of time (Alofs et al. 2014). Though it is currently unclear whether the endemics found in the pantepui rivers are relicts or the result of isolation and radiation, the uniqueness of fishes on the

Guiana Shield reflect faunal patterns which are seen across vertebrate taxa (Alofs et al. 2014;

Zyskowski et al. 2011).

4.1 Biogeographical Patterns of the Guiana Shield Gymnotus carapo

Gymnotus carapo from the Guianas (upland and lowland) do not form a monophyletic group, suggesting a complex relationship between rivers and populations. Upland fishes from the Potaro and Mazaruni group together consistently, suggesting a close relationship between these rivers.

Though no known connection currently exists between the headwaters of these rivers, several mechanisms could explain this pattern: headwater capture caused by tilting of the underlying basement shield, headwater course changes caused by erosion, and low relief between headwaters, allowing these areas to experience seasonal or incidental connection due to periods of increased rainfall (Lujan and Armbruster 2011). My results showing a close genetic connection between these rivers suggest that Gymnotus carapo individuals have recently moved between these two drainages.

On a broader scale, fishes belonging to the clade defined here as Guianas-uplands (upper Potaro

+ upper Mazaruni) are most closely related to a single fish collected from the Rupununi lowland region. Because only a single individual was available from the Rupununi, this pattern must be regarded as tentative. However, if accurate, it could indicate recent dispersal from the Rupununi region to the uplands of Guyana, or vice versa. Finally, the Guianas-upland + Rupununi clade is most closely related to the Amazonian fishes, which suggests that there is an existing or previous connection between these areas. The Rupununi River has been suggested as a possible portal for movement of fishes between the rivers of the Guianas and the Amazon basin (de Souza,

Armbruster and Werneke 2012; Lujan and Armbruster 2011; Hubert et al. 2007; Lovejoy and de

Araujo 2000; Willis et al. 2007).

The G. carapo collected from the Suriname and Commewijne Rivers consistently group together. The coastal-lowland clade also includes individuals that were collected from coastal rivers further north in Guyana (Figure 2). In cytb phylogenies, there is a moderately supported split between Gymnotus carapo collected from the Guyanese rivers and from the Suriname rivers, suggesting some interruption of gene flow between these areas. The close relationships of

Gymnotus carapo from along the coastal lowlands of the Guyanas may reflect corridors of movement proposed for fishes in this region (Lujan and Armbruster 2011). These authors cite two patterns of connectivity for fishes found in the coastal Guianas: the Eastern Atlantic

Connection, and the Western Connection. The Eastern Connection extends from the Amazon flume to the flume of the Essequibo River, while the Western Connection links the flume of the

Orinoco to that of the Essequibo as well. Measurements taken along the continental shelf in the

Amazonian flume show that the salinity of the water in this confluence is considerably lower than that of the surrounding Atlantic Ocean (Gibbs 1976), and though not much is known about

Gymnotus carapo’s tolerance of salinity, its presence on the Caribbean island of Trinidad would suggest that this species is capable of movement across marine barriers (Crampton and Hopkins

2005). Thus, dispersal along the coast of South America mediated by freshwater plumes from large rivers may explain the distribution of the Guianas-lowland G. carapo clade.

In the cytb phylogenies, the G. carapo collected at the top of Tafelberg tepui, an 1100m to

490m-tall isolated landform in Suriname, is sister to all coastal-lowland fishes. Tafelberg is on the boundary between the Coppename and Sarramacca watersheds of Suriname, and the rivers of the tepui drain into the right arm of the Coppename River, which runs close to the base of the

25 tepui before eventually draining into the Atlantic (Maguire 1945). The topology of the

Coppename basin, moving west to east, decreases in elevation gradually rather than abruptly

(Hammond 2005). These factors, suggest that dispersal to the summit of Tafelberg may have taken place within the basin itself. It remains to be determined whether G. carapo atop

Tafelberg represents a genetically isolated, self-sustaining population—confirming this would require additional sampling from the top and immediate surroundings of the tepui.

4.2 The Gymnotus carapo Species Complex

Phylogenetic results based on cytb and S7 indicate that the relationships of Gymnotus carapo lineages are extremely complex, complicating any efforts to delineate this widespread “species”.

Gymnotus carapo has long been a “catch-all” taxon, with many specimens collected across South

America designated as G. carapo without adequate consideration of morphological and genetic variation. However, ongoing work (Albert and Crampton 2001; Albert and Crampton 2003;

Albert et al. 2005; Crampton et al. 2013; Crampton, Lovejoy and Albert 2006) has attempted to refine understanding of G. carapo species limits, including descriptions of several new closely related species. It is clear, however, that G. carapo is not likely to represent a monophyletic lineage. Both morphological (Albert et al. 2005) and genetic studies (Lovejoy et al. 2010) indicate that some species (e.g., G. arapaima, G. ucamara) are nested within G. carapo.

A potential solution to this situation would be to recognize geographically and genetically distinct lineages of G. carapo that have sister group relationships to other species. Since the type species for G. carapo was collected in Suriname, the Guianas-lowland lineage would be designated as G. carapo sensu stricto. Other lineages, such as the Guianas-upland lineage, could be given new species names. However, this is unlikely to solve some existing problems, such as

26 the lack of reciprocal monophyly in the species of the Amazon, including G. carapo, G. arapaima, G. ucamara, and G. n. sp. LORE. The solution is also problematic because ongoing efforts have failed to find diagnostic morphological characters for several Gymnotus carapo lineages, thus species recognition would be based purely on molecular evidence.

To further resolve the relationships between the geographically distinct populations of Gymnotus carapo, more thorough sampling should be done, especially in the southern reaches of the species complex range and in Guiana Shield. These collections would further aid in generating molecular and morphological patterns of diversification. Additionally, next-generation sequence data may provide a wealth of information about not only lineages relationships, but many other aspects of the Gymnotus carapo genome.

References

Albert JS, Crampton WGR. 2001. Five new species of Gymnotus (Teleostei: Gymnotiformes) from an Upper Amazonian floodplain, with description of electric organ discharges and ecology. Ichthyological Exploration of Freshwaters. 12:241-266.

Albert JS, Crampton WGR. 2003. Seven new species of the Neotropical electric fish Gymnotus (Teleostei, Gymnotiformes) with a redescription of G. carapo (Linnaeus). Zootaxa. 54: 1– 54.

Albert JS, Crampton WGR, Thorsen DH, Lovejoy NR. 2005. Phylogenetic systematics and historical biogeography of the Neotropical electric fish Gymnotus (Teleostei: Gymnotidae). Systematics and Biodiversity. 2(4): 375–417.

Albert JS, Lovejoy NR, Crampton WGR. 2006. Miocene tectonism and the separation of cis- and trans-Andean river basins: Evidence from Neotropical fishes. Journal of South American Earth Sciences. 21(1-2): 14–27.

Albert JS, Petry P, Reis RE. Major Biogeographic and Phylogenetic Patterns. 2011. Pages 21-57 in: Albert JS, Reis RE, eds. Historical Biogeography of Freshwater Fishes. 1st ed. University of California.

Alda F, Reina RG, Doadrio I, Bermingham E. 2013. Phylogeny and biogeography of the Poecilia sphenops species complex (Actinopterygii, Poeciliidae) in Central America. Molecular Phylogenetics and Evolution. 66: 1011–1026.

Alofs KM, Liverpool EA, Taphorn DC, Bernard CR, López-Fernández H. 2014. Mind the (information) gap: the importance of exploration and discovery for assessing conservation priorities for freshwater fish. Diversity and Distributions. 20(1): 107–113.

Alp M, Keller I, Westram AM, Robinson CT. 2012. How river structure and biological traits influence gene flow: a population genetic study of two stream invertebrates with differing dispersal abilities. Freshwater Biology. 57(5): 969–981.

Armbruster JW, Taphorn DC. 2011. A new genus and species of weakly armored catfish from the upper Mazaruni River, Guyana (Siluriformes: Loricariidae). Copeia. 2011(1): 46–52.

Barthem R, Goulding M. 1997. Migration and Reproduction. 63 – 91 in The Catfish Connection. Columbia University Press.

Berry, P.E., and Riina, R. 2005. Insights into the diversity of the Pantepui Flora and the biogeographic complexity of the Guyana Shield. Biologiske Skrifter. 55: 145–167.

Brochu K. 2011. Molecular Phylogenetics of the Neotropical Electric Knifefish Genus Gymnotus (Gymnotidae, Teleostei): Biogeography and Signal Evolution of the Trans-Andean Species by Molecular Phylogenetics of the Neotropical Electric Knifefish. Masters thesis. University of Toronto. Canada.

Cecconi F, Crosio C, Mariottini P, Cesareni G, Giorgi M, Brenner S, Amaldi F. 1996. A functional role for some Fugu introns larger than the typical short ones: the example of the gene coding for ribosomal protein S7 and snoRNA U17. Nucleic Acids Research. 24(16): 3167–72.

Chow S, Hazama K. 1998. Universal PCR primers for S7 ribosomal protein gene introns in fish. Molecular Ecology. 7(9): 1255–6.

Corrigan S, and Beheregaray LB. 2009. A recent shark radiation: Molecular phylogeny, biogeography and speciation of wobbegong sharks (family: Orectolobidae). Molecular Phylogenetics and Evolution. 52: 205 – 216.

Crampton, W. G. R. and Hopkins CD. 2005. Nesting and paternal care in weakly electric fish Gymnotus (Gymnotiformes: Gymnotidae) with descriptions of larval and adult electric organ discharges of two species. Copeia. 2005(1): 48–60.

Crampton WGR, Lovejoy NR, Albert JS. 2003. Gymnotus ucamara : a new species of Neotropical electric fish from the Peruvian Amazon (Ostariophysi: Gymnotidae), with notes on ecology and electric organ discharges. Zootaxa. 18(August):1–18.

Crampton WGR, Rodríguez-Cattáneo A, Lovejoy NR, Caputi AA. 2013. Proximate and ultimate causes of signal diversity in the electric fish Gymnotus. Journal of Experimental Biology. 216(13): 2523–41.

Drummond AJ, Suchard MA, Xie D & Rambaut A. 2012. Bayesian phylogenetics with BEAUti and the BEAST 1.7 Molecular Biology And Evolution. 29: 1969-1973.

Gery J, Zarske A. 2002. Derhamia hoffmannorum gen. et sp. n. - a new pencil fish (Teleostei, Characiformes, Lebiasinidae), endemic from the Mazaruni River in Guyana. Zoology Abhandlungen. 52:35–47.

Gibbs RJ. 1973. Amazon river sediment transport in the Atlantic Ocean. Geology. 1971: 45–48.

Graur D and Li W-H. 2000. Molecular Phylogenetics. Pages 165 – 248 in Fundamentals of Molecular Evolution. Sinauer: Massachusetts.

Guy TJ, Gresswell RE, Banks, MA. 2008. Landscape-scale evaluation of genetic structure among barrier-isolated populations of coastal cutthroat trout, Oncorhynchus clarkii clarkii. Canadian Journal of Fisheries and Aquatic Science. 65(8): 1749–1762.

Hammond, D.S. 2005. Biophysical Features of the Guiana Shield. Pages 15 – 194 in Tropical Forests of the Guiana Shield. CABI Publishing: Oxfordshire, England.

Hardman M, Page LM, Sabaj MH, Armbruster JW, Knouft JH. 2002. A comparison of fish surveys made in 1908 and 1998 of the Potaro, Essequibo, Demerara, and coastal river drainages of Guyana. Ichthyological Exploration of Freshwaters. 13(3): 225–238.

Hubert N, Duponchelle F, Nuñez J, Garcia-Davila C, Paugy D, Renno J-F. 2007. Phylogeography of the piranha genera Serrasalmus and Pygocentrus: implications for the diversification of the Neotropical ichthyofauna. Molecular Ecology. 16(10): 2115–36.

Huelsenbeck JP, Ronquist F, Nielsen R, and Bollback JP. 2001. Bayesian inference of phylogeny and its impact on evolutionary biology. Science. 294: 2310-2314.

Hutchison DW, Templeton AR. 1999. Correlation of pairwise genetic and geographic distance measures: inferring the relative influences of gene flow and drift on the distribution of genetic variability. Evolution. 53(6): 1898–1914.

Kirschbaum, F., and L. Wieczorek. 2002. Entdeckung einer neuen Fortpflanzungs-strategie bei südamerikainischen Messerfischen (Teleostei: Gymnotiformes: Gymnotidae): Maulbrüten bei Gymnotus carapo. Verhalt Aquarienfisch 2(2002): 99-107.

Lanfear R, Calcott B, Ho SYW, Guindon S. 2012. PartitionFinder: combined selection of partitioning schemes and substitution models for phylogenetic analyses. Molecular Biology and Evolution. 29(6): 1695–701.

Lavoué S, Sullivan JP, Hopkins CD. 2003. Phylogenetic utility of the first two introns of the S7 ribosomal protein gene in African electric fishes (Mormyridae: Teleostei) and congruence with other molecular markers. Biological Journal of the Linnean Society. 78(2): 273–292.

López-Fernández H, Taphorn DC, Liverpool EA. 2012. Phylogenetic diagnosis and expanded description of the genus Mazarunia Kullander, 1990 (Teleostei: Cichlidae) from the upper Mazaruni River, Guyana, with description of two new species. Neotropical Ichthyology. 10(3): 465–486.

Lovejoy NR, Lester K, Crampton WGR, Marques FPL, Albert JS. 2010. Phylogeny, biogeography, and electric signal evolution of Neotropical knifefishes of the genus Gymnotus (Osteichthyes: Gymnotidae). Molecular Phylogenetics and Evolution. 54(1):278–90.

Lovejoy NR, De Araújo ML. 2000. Molecular systematics, biogeography and population structure of neotropical freshwater needlefishes of the genus Potamorrhaphis. Molecular Ecology. 9(3): 259–68.

Lujan NK. 2008. Description of a new Lithoxus (Siluriformes: Loricariidae) from the Guyana Highlands with a discussion of Guiana Shield biogeography. Neotropical Ichthyology. 6(3): 413–418.

Lujan NK, Agudelo-Zamora H, Taphorn DC, Booth PN, López-Fernández H. 2013. Description of a new, narrowly endemic South American darter (Characiformes: Crenuchidae) from the central Guiana Shield highlands of Guyana. Copeia. 2013(3): 454–463.

Lujan NK, Armbruster JW. 2011. The Guiana Shield. Pages 211 – 224 in Albert JS, Reis RE, eds. Historical Biogeography of Freshwater Fishes.

Lundberg JG, Marshall LG, Guerrero J, Horton B, Malabarba MCSL, Wesselingh F. 1998. The Stage for Neotropical Fish Diversification: A History of Tropical South American Rivers. Pages 14 – 51 in LR Malabarba, RE Reis, RP Vari, ZM Lucena, CAS Lucena, eds. Phylogeny and Classification of Neotropical Fishes. University of California.

Maddison WP, and Maddison DR. 2014. Mesquite: a modular system for evolutionary analysis. Version 3.01.

Maguire B. 1945. Notes on the geology and the geography of Tafelberg, Suriname. Geographical Review. 35(4): 563–579.

Maldonado-Ocampo JA, Albert JS. 2004. Gymnotus ardilai : a new species of Neotropical electric fish (Ostariophysi: Gymnotidae) from the Rio Magdalena Basin of Colombia. Zootaxa. 10:1–10.

Maldonado-Ocampo JA, López-Fernández H, Taphorn DC, Bernard CR, Crampton WGR, Lovejoy NR. 2014. Akawaio penak, a new genus and species of Neotropical electric fish (Gymnotiformes, Hypopomidae) endemic to the upper Mazaruni River in the Guiana Shield. Zoologica Scripta. 43(1): 24–33.

Markert JA, Schelly RC, Stiassny MLJ. 2010. Genetic isolation and morphological divergence

mediated by high-energy rapids in two cichlid genera from the lower Congo rapids. BMC Evolutionary Biology. 149(10).

Maxime, E. 2013. Phylogenetic and revisionary studies of the banded knifefishes Gymnotus (Teleostei, Ostariophysi, Gymnotiformes) with descriptions of fourteen new species. Dissertation. PhD Thesis. University of Louisiana at Lafayette. United States of America.

McConnell, RB. 1968. Planation surfaces in Guyana. The Geographical Journal. 134(4): 506– 520.

Rahel FJ. 2007. Biogeographic barriers, connectivity and homogenization of freshwater faunas: it’s a small world after all. Freshwater Biology. 52(4): 696–710.

Rambaut A, Suchard MA, Xie D and Drummond AJ. 2014. Tracer v1.6.

Rull V. Biotic diversification in the Guyana Highlands: a proposal. 2005. Journal of Biogeography. 32(6):921–927.

Stamatakis A. 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics. 22(21): 2688–90.

Strange RM, Stepien CA. 2007. Genetic divergence and connectivity among river and reef spawning groups of walleye (Sander vitreus vitreus) in Lake Erie. Canadian Journal of Fisheries and Aquatic Science. 448: 437–448.

Sullivan JP, Lavoué S, Hopkins CD. 2002. Discovery and phylogenetic analysis of a riverine species flock of African electric fishes (Mormyridae: Teleostei). Evolution. 56(3): 597– 616.

Swofford, D. L. 2003. PAUP*. Phylogenetic Analysis Using Parsimony (*and other methods). Version 4. Sinauer Associates, Sunderland, Massachusetts.

Taphorn DC, Lopez-Fernandez H, Bernard CR. 2008. Apareiodon agmatos, a new species from the upper Mazaruni River, Guyana (Teleostei, Characiformes: Parodontidae). Zootaxa. 38:31–38.

Torrente-Vilara G, Zuanon J, Leprieur F, Oberdorff T, Tedesco PA. 2011. Effects of natural rapids and waterfalls on fish assemblage structure in the Madeira River (Amazon Basin). Ecology of Freshwater Fish. 20(4): 588–597.

Waters JM, Craw D, Youngson JH, Wallis GP. 2001. Genes meet geology: fish phylogeographic pattern reflects ancient, rather than modern, drainage connections. Evolution. 55(9): 1844– 51.

Willis SC, Nunes MS, Montaña CG, Farias IP, Lovejoy NR. Systematics, biogeography, and evolution of the Neotropical peacock basses Cichla (Perciformes: Cichlidae). 2007. Molecular Phylogenetics and Evolution. 44(1): 291–307.

Winemiller KO, Agostinho AA, Pellegrini Caramaschi E. 2008. Fish Ecology in Tropical Streams. Pages 107 – 126 in Dudgeon D, ed. Tropical Stream Ecology. Elsevier/Academic Press, San Diego, CA.

Wong BBM, Keogh JS, McGlashan DJ. 2004. Current and historical patterns of drainage connectivity in eastern Australia inferred from population genetic structuring in a widespread freshwater fish Pseudomugil signifer (Pseudomugilidae). Molecular Ecology. 13(2): 391–401.

Zyskowski K, Mittermeier JC, Ottema O, Rakovic M, Shea BJO, Lai JE, Hochgraf SB, de León J, Au K. 2011. Avifauna of the easternmost tepui, Tafelberg in Central Suriname. Bulletin of the Peabody Museum of Natural History. 52(1): 153–180.

Figures and Tables

Table 1: Members of the Gymnotus carapo species complex included in this study. Catalogue Genus Species Tissue ID number Region Latitude Longitude Locality Gymnotus ucamara 1927 UF 126184 Western Amazon - - Pacaya Samiria Gymnotus ucamara 1950 UF 126184 Western Amazon - - Pacaya Samiria MZUSP Gymnotus arapaima 2002 75179 Central Amazon 3.0383 -64.85 Mamirauá Lake System MZUSP Gymnotus arapaima 2003 103219 Central Amazon 3.1225 -64.8303 Mamirauá Lake System carapo sp- MZUSP Gymnotus complex 2004 76066 Central Amazon 3.1123 64.967 Mamirauá Lake System Gymnotus n. sp. LORE 2006 UF 131129 Western Amazon - - Iquitos aquaria Gymnotus n. sp. LORE 2007 UF 131129 Western Amazon - - Iquitos aquaria MZUSP Gymnotus curupira 2009 75148 Central Amazon 3.4336 -64.7296 Rio Tefé Gymnotus mamiraua 2012 MZUP 103221 Central Amazon 3.1283 -64.8005 Mamirauá Lake System Gymnotus mamiraua 2013 MCP 29805 Central Amazon 3.1283 -64.8005 Mamirauá Lake System MZUSP Gymnotus varzea 2014 75163 Central Amazon 3.0383 -64.85 Mamirauá Lake System MZUSP Gymnotus obscurus 2017 75155 Central Amazon 3.0383 -64.866 Mamirauá Lake System carapo sp- MZUSP Gymnotus complex 2030 76066 Central Amazon 23.2261 -44.7661 Mamirauá Lake System carapo sp- Gymnotus complex 2040 UF 174335 Orinoco 7.8255 -68.923 Rio Guaratico carapo sp- Gymnotus complex 2041 UF 174335 Orinoco 7.8255 -68.923 Rio Guaratico carapo sp- UF Un-Cat- Gymnotus complex 2463 PU02 Western Amazon 5.2943 -74.498 Rio Ucayali carapo sp- UF Un-Cat- Gymnotus complex 2464 PU02 Western Amazon 5.2943 -74.498 Rio Ucayali MZUSP Gymnotus n. sp. ITAP 2558 85943 Paraná 22.2741 -47.6093 Rio Paraná MZUSP Gymnotus n. sp. IGUA 2559 85947 Paraná 22.2741 -47.6093 Rio Paraná Gymnotus n. sp. SAO JOAO 2956 uncatalogued Fluminense 22.096 -42.4146 Rio São João Gymnotus n. sp. IGUA 2957 uncatalogued Paraná 22.096 -42.4146 Rio São João Gymnotus carapo 6998 UF 180165 Northeast coastal 5.4524 -55.245 Suriname River Gymnotus carapo 6999 UF 180175 Northeast coastal 5.582 -54.2332 Commewijne River Gymnotus carapo 7001 UF 180169 Northeast coastal 5.2448 -55.1012 Suriname River Gymnotus carapo 7002 UF 180175 Northeast coastal 5.582 -54.2332 Commewijne River Gymnotus carapo 7003 UF 180173 Northeast coastal 5.5868 -54.2852 Commewijne River Gymnotus carapo 7004 UF 180173 Northeast coastal 5.5868 -54.2852 Commewijne River Gymnotus carapo 7005 UF 180169 Northeast coastal 5.2448 -55.1012 Suriname Riviere Gymnotus carapo 7006 UF 180165 Northeast coastal 5.4524 -55.245 Suriname Riviere

Gymnotus carapo 7007 UF 180173 Northeast coastal 5.5868 -54.2852 Commewijne River Gymnotus carapo 7008 UF 180169 Northeast coastal 5.2448 -55.1012 Suriname River Gymnotus carapo 7009 UF 180173 Northeast coastal 5.5868 -54.2852 Commewijne River Gymnotus carapo 7010 UF 180165 Northeast coastal 5.4524 -55.245 Suriname River Gymnotus carapo 7011 UF 180173 Northeast coastal 5.4524 -55.245 Suriname River Gymnotus carapo 7012 UF 180169 Northeast coastal 5.2448 -55.1012 Suriname River Gymnotus carapo 7013 UF 180173 Northeast coastal 5.5868 -54.2852 Commewijne River Gymnotus carapo 7014 UF 180173 Northeast coastal 5.5868 -54.2852 Commewijne River Gymnotus carapo 7016 UF 180173 Northeast coastal 5.5868 -54.2852 Commewijne River Gymnotus n. sp. ITAP 7065 GY7065 Paraná - - - Gymnotus n. sp. ITAP 7066 GY7066 Paraná - - - Gymnotus n. sp. ITAP 7067 GY7067 Paraná - - - Gymnotus n. sp. IGUA 7071 GY7071 Paraná - - - Gymnotus n. sp. IGUA 7072 GY7072 Paraná - - - Gymnotus n. sp. IGUA 7074 uncatalogued Paraná - - - Gymnotus n. sp. IGUA 7075 uncatalogued Paraná - - - Gymnotus pantanal 7076 uncatalogued Paraná - - - Gymnotus n. sp. ITAP 7084 MNRJ 31520 Paraná - - Lagoa dos Tropeiros Gymnotus n. sp. ITAP 7085 MNRJ 31520 Paraná - - Lagoa dos Tropeiros Gymnotus n. sp. ITAP 7086 MNRJ 31521 Paraná - - Corrego dos bois Gymnotus n. sp. ITAP 7087 MNRJ 31523 Paraná - - Corrego dos bois Gymnotus n. sp. ITAP 7088 MNRJ 31520 Paraná - - Lagoa dos Tropeiros AMNH Gymnotus omarorum 7092 239656 Paraná 34.8388 -55.1144 Laguna del Cisne AMNH Gymnotus omarorum 7093 239656 Paraná 34.8388 -55.1144 Laguna del Cisne Gymnotus carapo 7101 UMSS 6976 Western Amazon -10.9256 -66.01 Rio Beni Gymnotus carapo 7103 CBF 10211 Western Amazon -10.9129 -65.9969 Rio Beni MZUSP Gymnotus sylvius 7239 100267 Paraná - - Rio São Lourenço MZUSP Gymnotus sylvius 7240 100267 Paraná - - Rio São Lourenço MZUSP Gymnotus bahianus 7244 102898 Northeast Brazil - - Rio Almada MZUSP Gymnotus bahianus 7245 102898 Northeast Brazil - - Rio Almada Gymnotus chaviro 7357 MUSM 33715 Paraná - - Rio Yuruna Gymnotus ardilai 8175 IAvHP11511 Magdalena 7.10675 -73.2816 Rio de Oro Gymnotus ardilai 8186 IAvHP11510 Magdalena 7.0348 -73.2696 Rio de Oro Gymnotus choco 8209 IAvHP10646 Magdalena - - Rio Atrato Gymnotus n. sp. IGUA 8235 STRI-01591 Paraná - -58.91 Rio Paranà Gymnotus carapo 8250 DNA Only Orinoco 8.3394 -69.6661 Rio Apure 2009-07-24- Gymnotus carapo 8607 71 Western Amazon 4.8929 -73.6501 Rio Ucayali Gymnotus carapo 8624 uncatalogued Western Amazon 4.8997 -73.6474 Rio Ucayali

Gymnotus carapo 8631 no voucher Western Amazon 4.9095 -73.6622 Rio Ucayali Gymnotus carapo 8634 no voucher Western Amazon 4.901 -73.6472 Rio Ucayali Gymnotus carapo 8640 no voucher Western Amazon 3.4166 -72.75 Rio Ucayali Gymnotus carapo 8641 no voucher Western Amazon 3.4166 -72.75 Rio Ucayali Gymnotus n. sp. ITAP 8756 MNRJ 31521 Paraná - -46.2766 Piumhi River Gymnotus n. sp. ITAP 8757 MNRJ 31523 Paraná - -46.0061 Piumhi River Gymnotus carapo 8759 MNRJ 33638 Eastern Amazon - -53.7902 Xingú-Tapajós confluence Gymnotus carapo 8760 MNRJ 33641 Eastern Amazon -3.8744 -54.2358 Xingú-Tapajós confluence Gymnotus n.sp. IGUA 8764 MNRJ 33634 Paraná -10.0602 -55.4483 Xingú-Tapajós confluence Gymnotus carapo 8767 ROM 66503 Northeast coastal - - Waini River Gymnotus n.sp. IGUA 8778 MNRJ 33633 Paraná -9.9802 -55.2611 Xingú-Tapajós confluence Gymnotus carapo 9801 KU 41319 Coppename 3.9195 -56.2005 Tafelberg Gymnotus carapo 9805 ROM 83812 Essequibo 5.4755 -60.77967 Mazaruni River Gymnotus carapo 9806 ROM 83885 Essequibo 5.93614 -60.61494 Mazaruni River Gymnotus carapo 9810 ROM 87338 Northeast coastal 7.42878 -58.67671 Waini River Gymnotus carapo 9811 ROM 83885 Essequibo 5.93614 -60.61494 Mazaruni River Gymnotus carapo 9815 ROM 83714 Essequibo 5.36031 -60.37178 Mazaruni River Gymnotus carapo 9816 ROM 89937 Essequibo - - Potaro River Gymnotus carapo 9820 ROM 87381 Northeast coastal 6.73499 -58.303 Demerara River Gymnotus carapo 9819 ROM 89575 Essequibo 5.70846 -60.36099 Mazaruni River Gymnotus carapo 9821 ROM 83714 Essequibo 5.36031 -60.37178 Mazaruni River Gymnotus carapo 9947 ROM 89937 Essequibo - - Potaro River Gymnotus carapo 9948 ROM 96022 Essequibo 2.82915 -59.80895 Rupununi River Gymnotus carapo 9949 ROM 83812 Essequibo 5.4755 -60.77967 Mazaruni River Gymnotus carapo 9950 ROM 83812 Essequibo 5.4755 -60.77967 Mazaruni River Gymnotus carapo 9951 ROM 66503 Northeast coastal 7.7 -59.23333 Waini River Gymnotus carapo 9977 ROM 87338 Northeast coastal 7.42878 -58.67671 Waini River Gymnotus carapo 9978 ROM 83714 Essequibo 5.36031 -60.37178 Mazaruni River Gymnotus carapo 9983 ROM 83812 Essequibo 5.4755 -60.77967 Mazaruni River Gymnotus carapo 9989 ROM 87030 Essequibo 4.90557 -58.25014 Berbice River Gymnotus carapo 10027 ROM 94993 Essequibo 5.010888889 -59.637 Potaro River Gymnotus carapo 10028 ROM 94993 Essequibo 5.010888889 -59.637 Potaro River Gymnotus carapo 10030 ROM 94993 Essequibo 5.010888889 -59.637 Potaro River Gymnotus carapo 10031 ROM 95035 Essequibo 5.007027778 -59.63183333 Potaro River Gymnotus carapo 10032 ROM 95035 Essequibo 5.007027778 -59.63183333 Potaro River Gymnotus carapo 10033 ROM 95035 Essequibo 5.007027778 -59.63183333 Potaro River Gymnotus carapo 10034 ROM 95035 Essequibo 5.007027778 -59.63183333 Potaro River Gymnotus carapo 10035 ROM 95051 Essequibo 5.007805556 -59.63652778 Potaro River Gymnotus carapo 10036 ROM 95051 Essequibo 5.007805556 -59.63652778 Potaro River Gymnotus carapo 10037 ROM 95051 Essequibo 5.007805556 -59.63652778 Potaro River Gymnotus carapo 10038 ROM 95051 Essequibo 5.007805556 -59.63652778 Potaro River Gymnotus carapo 10039 ROM 95071 Essequibo 5.070111111 -59.65369444 Potaro River Gymnotus carapo 10040 ROM 95071 Essequibo 5.070111111 -59.65369444 Potaro River

Gymnotus carapo 10041 ROM 95088 Essequibo 5.108583333 -59.63563889 Potaro River Gymnotus carapo 10042 uncatalogued Essequibo 5.275 -59.516 Potaro River Gymnotus carapo 10043 uncatalogued Essequibo 5.275 -59.516 Potaro River

Table 2: Gymnotus carapo collected from the Guiana Shield, and included in this study. Map numbers correspond to Figure 5. Elevations were acquired by plotting collection locations in ArcGIS software using latitudes and longitudes taken in the field.

Approximate Tissue Elevation Map ID Cytb S7 Catalogue Number number Region Locality (m-asl) Haplotype Haplotype number

1 8767 Coastal Waini River 0 - 105 E - ROM 66503

1 9951 Coastal Waini River 0 - 105 E - ROM 66503

2 9810 Coastal Waini River 0 - 105 H - ROM 87338

2 9977 Coastal Waini River 0 - 105 F - ROM 87338

3 9820 Coastal Demerara River 0 - 105 D - ROM 87381

4 9806 Essequibo Mazaruni River 493 - 646 A aa ROM 83885

4 9811 Essequibo Mazaruni River 493 - 646 A aa ROM 83885

5 9819 Essequibo Mazaruni River 335 - 492 B aa ROM 89575

6 9805 Essequibo Mazaruni River 493 - 646 B aa ROM 83812

6 9949 Essequibo Mazaruni River 493 - 646 B aa ROM 83812

6 9950 Essequibo Mazaruni River 493 - 646 B aa ROM 83812

6 9983 Essequibo Mazaruni River 493 - 646 B - ROM 83812

7 9815 Essequibo Mazaruni River 493 - 646 B - ROM 83714

7 9821 Essequibo Mazaruni River 493 - 646 B aa ROM 83714

7 9978 Essequibo Mazaruni River 493 - 646 B - ROM 83714

8 10027 Essequibo Potaro River 493 - 646 B - ROM 94993

8 10028 Essequibo Potaro River 493 - 646 B bb ROM 94993

8 10030 Essequibo Potaro River 493 - 646 B - ROM 94993

8 10031 Essequibo Potaro River 493 - 646 B - ROM 95035

8 10033 Essequibo Potaro River 493 - 646 B bb ROM 95035

8 10034 Essequibo Potaro River 493 - 646 B - ROM 95035

8 10035 Essequibo Potaro River 493 - 646 B - ROM 95051

8 10036 Essequibo Potaro River 493 - 646 B - ROM 95051

8 10037 Essequibo Potaro River 493 - 646 B - ROM 95051

8 10038 Essequibo Potaro River 493 - 646 B - ROM 95051

8 10032 Essequibo Potaro River 493 - 646 B bb ROM 95035

9 10041 Essequibo Potaro River 647 - 818 B - ROM 95088

10 10039 Essequibo Potaro River 493 - 646 B - ROM 95071

10 10040 Essequibo Potaro River 493 - 646 B bb ROM 95071

not in ROM 11 10042 Essequibo Potaro River 493 - 646 B - collections

not in ROM 11 10043 Essequibo Potaro River 493 - 646 B - collections

- 9947 Essequibo Potaro River - B bb ROM 89937

12 9948 Essequibo Rupununi River 106 - 225 C bb ROM 96022

13 9989 Essequibo Berbice River 0 - 105 H - ROM 87030

Coppena 14 9801 me Tafelberg 493 - 646 I bb KU 41319

15 7001 Coastal Suriname River 0 - 105 H - UF 180169

15 7005 Coastal Suriname River 0 - 105 G - UF 180169

15 7008 Coastal Suriname River 0 - 105 H - UF 180169

15 7012 Coastal Suriname River 0 - 105 G - UF 180169

16 6998 Coastal Suriname River 0 - 105 D - UF 180165

16 7006 Coastal Suriname River 0 - 105 D cc UF 180165

16 7010 Coastal Suriname River 0 - 105 D - UF 180165

16 7011 Coastal Suriname River 0 - 105 D - UF 180173

Commewijne 17 6999 Coastal River 0 - 105 D - UF 180175

Commewijne 17 7002 Coastal River 0 - 105 D dd UF 180175

Commewijne 18 7003 Coastal River 0 - 105 D - UF 180173

Commewijne 18 7004 Coastal River 0 - 105 D - UF 180173

Commewijne 18 7007 Coastal River 0 - 105 D - UF 180173

Commewijne 18 7009 Coastal River 0 - 105 D - UF 180173

Commewijne 18 7013 Coastal River 0 - 105 D - UF 180173

Commewijne 18 7014 Coastal River 0 - 105 D - UF 180173

Commewijne 18 7016 Coastal River 0 - 105 D - UF 180173

Coast Lowland 0-300 m-asl 0-300 300-1500 m-asl 300-1500 Upland (pantepui) : Profile :of differentProfile Shield. of the Guiana categories elevational (2005). from taken Hammond divisions Altitudinal Highland (tepui) 1501-3000 m-asl 1501-3000 Figure 1

ar e n j i w ro Ma

Commewijne

ur me a n ri Su ar acca ma ra Sa

me a n e p p o C Atlantic Ocean Atlantic

Nickerie

ant n j ti n ra o C

er e er ve i R ce i rb Be

er a er ve i R ra ra me e D

sequibo er ve i R o b i u q Esse

sequibo er ve i R o b i u q Esse sequibo ainage g a n i ra D o b i u q Esse Rupununi River Rupununi Mazaruni River Mazaruni Potaro River Potaro Rio Branco Rio

Cuyuni River Cuyuni Negro g e N o i R Amazon River Amazon : The drainages and major rivers of the Guiana Shield in Guyana and Suriname. Figure 2

19 species (8%) occur throughout the drainage

51 species (26%) occur only below Kaieteur

81 species (42%) occur only below Tumatumari

33 species (17%) occur only between the falls

2 species (1%) occur only above both Tumatumari and Kaieteur

9 species (5%) occur only above Kaieteur

Tumatumari Cataract Kaieteur Falls

Upstream

Figure 3:4: Fish assemblages found above and below Kaieteur Falls and the Tumatumari Cataract on the Potaro River. From Hardman et al. (2002)

G. mamiraua

G. curupira G. obscurus G. varzea G. paraguensis G. tigre G. henni G. esmeraldas G. bahianus G. sylvius G. inaequilabiatus G. diamantinensis G. ucamara (WA) G. choco (PS) G. arapaima (WA) G. carapo (WA) G. carapo (GO) G. carapo (MD) G. carapo (PI) G. carapo (RO) G. carapo (EA)

Figure 4: Proposed relationships between members of the Gymnotus carapo clade based on a maximum parsimony analysis of morphological and meristic characters. Figure adapted from Albert et al. (2005).

0 0 0 0 0 0 0 e 0 0 0 0 0 0 0 0 d 2 4 6 8 0 2 4 0 , , , , , , , 0 0 0 u 0 1 1 1 1 2 2 2 0 0 0 , t ------i 0 4 6 8 1

0 t

0 - - - - 1 1 1 1 1 1 1 l -

2 0 0 0 0 0 0 0

1 1 1 1 8 - 0 2 4 6 8 0 2 A 0 0 0 0 , , , , , , , 2 - 0 2 4 6 8 1 1 1 1 1 2 2 e asl u l - a 17 V 18 15 16 14 Fishcollected below 300 m

. 13 3 2 1 asl(upland) are in red.Information on specific - 9 11 8 12 fromtheGuiana Shield 10

5 7 4 and rivers are found in Table 2. 6 0 0 0 0 0 0 0 e 0 0 0 0 0 0 0 0 d 2 4 6 8 0 2 4 0 , , , , , , , 0 0 0 u 0 1 1 1 1 2 2 2 0 0 0 , t ------i 0 4 6 8 1

0 - - - - 1 1 1 1 1 1 1 l -

2 0 0 0 0 0 0 0

1 1 1 1 8 - 0 2 4 6 8 0 2 A 0 0 0 0 , , , , , , , 2 - 0 2 4 6 8 1 1 1 1 1 2 2 e u l a V Gymnotuscarapo

Collectionlocations of : 5 igure F (lowland)are in orange and fishcollected above 300 m tissuescan be found in Table 2.

: Collection locations of tissues included in this study. Fish collected below 300 m-asl (lowland) are in orange and ﬁsh collected above 300 m-asl (upland) are in red. Information on speciﬁc basins Figure 5

Table 3: Primers used for S7 and cytb in this study.

Gene Primer Sequence 5' - 3' Source cytb GLUDGL CGAAGCTTGACTTGAARAACCAYCGTT Palumbi et al. 1991 cytb CYTBR CTCCGATCTTCGGATTACAAG Palumbi et al. 1991 cytb CYTBF TCYAWCATCTCAGCCTGATG Palumbi et al. 1991 cytb GLUDGL.GYM AGACCTATGACTTGAAAAACCATCGTTG This study cytb JOHNNY5.CYTBR GCTGAAGATGGGAGTTAAACCC This study S7 S7RPEX1F TGGCCTCTTCCTTGGCCGTC Chow and Hazama 1998 S7 S7RPEX2R AACTCGTCTGGCTTTTCGCC Chow and Hazama 1998

Table 4: Average pairwise distances between allopatric populations of the Gymnotus carapo species complex included in this study. Calculated using PAUP*. Guianas- Guianas- Paraná- cytb upland Lowland Amazon Orinoco Madeira Guianas-upland 0.012 0.007 0.040 0.032 Guianas-lowland 0.012 0.011 0.036 0.033 Amazon 0.006 0.011 0.040 0.030 Orinoco 0.040 0.035 0.039 0.030 Paraná-Madeira 0.032 0.033 0.030 0.030 S-7

Guianas-Highland 0.005 0.011 0.009 0.007 Guianas-Lowland 0.006 0.001 0.006 0.021 Amazon 0.011 0.001 0.011 0.027 Orinoco 0.009 0.006 0.011 0.005 Paraná-Madeira 0.007 0.021 0.027 0.005

Figure 6: Cytb gene tree showing relationships between allopatric populations of Gymnotus carapo and closely related species. Bayesian posterior probabilities and maximum likelihood bootstrapping are shown on all major clades.

9806_Gcarapo_Mazaruni 9811_Gcarapo_Mazaruni 10036_Gcarapo_Potaro 10040_Gcarapo_Potaro 10037_Gcarapo_Potaro 9947_Gcarapo_Potaro 10038_Gcarapo_Potaro 10039_Gcarapo_Potaro 9983_Gcarapo_Mazaruni 10034_Gcarapo_Potaro 9815_Gcarapo_Mazaruni 9978_Gcarapo_Mazaruni 9821_Gcarapo_Mazaruni Guianas-upland 0.83 9805_Gcarapo_Mazaruni 9950_Gcarapo_Mazaruni 9819_Gcarapo_Mazaruni 9949_Gcarapo_Mazaruni 10041_Gcarapo_Potaro 10043_Gcarapo_Potaro 10032_Gcarapo_Potaro 0.92/66 10042_Gcarapo_Potaro 10035_Gcarapo_Potaro 9816_Gcarapo_Potaro 10028_Gcarapo_Potaro 10031_Gcarapo_Potaro 10027_Gcarapo_Potaro 10030_Gcarapo_Potaro 10033_Gcarapo_Potaro 9948_Gcarapo_Rupununi 1927_Gucamara 2002_Garapaima 2007_GnspLORE 8634_Gcarapo_Ucayali 100/901.0 8640_Gcarapo_Ucayali 2463_Gcarapo_Ucayali 1950_Gucamara 2464_Gcarapo_Ucayali 8624_Gcarapo_Ucayali Amazon 8607_Gcarapo_Ucayali 8631_Gcarapo_Ucayali 2003_Garapaima 2004_Gcarapo_Tefé 2030_Gcarapo_Tefé 2006_GnspLORE 8641_Gcarapo_Ucayali 7016_Gcarapo_Commewijne 7004_Gcarapo_Commewijne 6999_Gcarapo_Commewijne 6998_Gcarapo_Suriname 7010_Gcarapo_Suriname 7011_Gcarapo_Suriname 7014_Gcarapo_Commewijne 7009_Gcarapo_Commewijne 7013_Gcarapo_Commewijne 7002_Gcarapo_Commewijne 7006_Gcarapo_Suriname 7003_Gcarapo_Commewijne 7007_Gcarapo_Commewijne 9820_Gcarapo_Demerara Guianas-lowland 100/981.0/ 9951_Gcarapo_Waini 8767_Gcarapo_Waini 9977_Gcarapo_Waini 7005_Gcarapo_Suriname 7012_Gcarapo_Suriname 7001_Gcarapo_Suriname 7008_Gcarapo_Suriname 9810_Gcarapo_Waini 9989_Gcarapo_Berbice 9801_Gcarapo_Tafelberg 7244_Gbahanius 0.92/67 7245_Gbahanius 8759_Gcarapo_XinguTapajos 8760_Gcarapo_XinguTapajos 7240_Gsylvius 7239_Gsylvius 2558_GnspITAP 8756_GnspITAP 7067_GnspITAP 7088_GnspITAP 7066_GnspITAP 8757_GnspITAP 0.65/63 7085_GnspITAP 7086_GnspITAP Paraná-Madeira 7084_GnspITAP 7087_GnspITAP 7065_GnspITAP 7101_Gcarapo_Beni 7103_Gcarapo_Beni 2040_Gcarapo_Guaratico 8175_Gardilai 8186_Gardilai 100/100 2041_Gcarapo_Guaratico Orinoco 1.0/ 8250_Gcarapo_Orinoco 8209_Gchoco 7071_GnspIGUA 7074_GnspIGUA 7072_GnspIGUA 8235_GnspIGUA 7075_GnspIGUA 8764_GnspIGUA 8778_GnspIGUA 2559_GnspIGUA 2957_GnspIGUA 2012_Gmamiraua Paraná-Southeast 2013_Gmamiraua 7092_Gomarorum 7093_Gomarorum 2956_GnspSAOJOAO 2009_Gcurupira 2017_Gobscurus 7076_Gpantanal 7357_Gchaviro 2014_Gvarzea

0.4 47

Figure 7: S7 gene tree showing relationships between populations of G. carapo and closely related species. Bayesian posterior probabilities and maximum likelihood values are shown on all geographic clades. Colours show clade membership as defined in Figure 6.

Figure 8: Tree showing relationships of G. carapo and closely related species when S7 and cytb are concatenated and run as a single dataset. A) Bayesian analysis and B) Maximum likelihood analysis. Colours show clade membership as defined in Figure 6.

Figure 9: *BEAST analysis of cytb and S7 data showing relationships between populations of Gymnotus carapo and closely related species included in this study. Colours show clade membership as defined in Figure 6.

Gymnotus_carapo_Mazaruni 0.89 Gymnotus_carapo_Mazaruni

0.87 Gymnotus_carapo_Mazaruni Guianas-upland Gymnotus_carapo_Potaro 0.78 Gymnotus_carapo_Potaro

0.86 Gymnotus_carapo_Rupununi Gymnotus_carapo_Tefé

Gymnotus_ucamara

Gymnotus_carapo_Ucayali Amazon 1.0100 Gymnotus_n_sp_LORE

Gymnotus_arapaima

Gymnotus_carapo_Tafelberg 0.73 Gymnotus_carapo_Commewijne 0.39 Guianas-lowland Gymnotus_carapo_Suriname

Gymnotus_carapo_Beni 0.99 Gymnotus_n_sp_ITAP Paraná-Madeira Gymnotus_sylvius

Gymnotus_choco 0.97 Gymnotus_ardilai Orinoco

Gymnotus_carapo_Guaratico

Gymnotus_omarorum

Gymnotus_n_sp_IGUA Paraná-Southeast Gymnotus_mamiraua

Gymnotus_varzea

0.9