A Biogeographic Perspective of Speciation Among Desert Tortoises in the Genus Gopherus

Item Type text; Electronic Dissertation

Authors Edwards, Taylor Artemus

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 04/10/2021 12:23:17

Link to Item http://hdl.handle.net/10150/556486

A BIOGEOGRAPHIC PERSPECTIVE OF SPECIATION AMONG DESERT TORTOISES IN THE GENUS GOPHERUS

Taylor Edwards

______

A Dissertation Submitted to the Faculty of the

SCHOOL OF NATURAL RESOURCES AND THE ENVIRONMENT

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

With a Major in Natural Resources

In the Graduate College of

THE UNIVERSITY OF ARIZONA

2015

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Taylor Edwards, titled A Biogeographic Perspective of Speciation among Desert Tortoises in the Genus Gopherus and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy.

______Date: 13 Feb 2015 Melanie Culver

______Date: 13 Feb 2015 Michael S Barker

______Date: 13 Feb 2015 Robert Ward Murphy

______Date: 13 Feb 2015 Adrian Quijada-Mascarenas

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

______Date: 13 Feb 2015 Dissertation Director: Melanie Culver

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of the requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that an accurate acknowledgement of the source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: ______Taylor Edwards______

ACKNOWLEDGEMENTS

I am grateful for the support of a multitude of people, not the least of which is my wife, Cori Dolan. She and our daughter Amira made the significant sacrifice of allowing me time away to complete this work.

This dissertation was truly a collaborative effort and would not have been possible without the contributions of my coauthors on the manuscripts featured in each of the appendices. I especially want to thank my graduate committee for their support and guidance, including M Nachman who greatly influenced the direction of this research.

Additional guidance and support came from M Kaplan, B Fransway, R Sprissler and my friends and colleagues at the UAGC as well as my fellow graduate students in the Culver Lab. L Johnstone, S Miller and N Merchant were invaluable in their technical assistance with computing and scripting. R Sandoval graciously provided me with office space to complete this work.

Sample collection and coordination would not have been possible without the assistance of many, many people and this international project involved the collaboration of the US Geological Survey, Tucson Herpetological Society, Arizona Game and Fish Department, and the Comisión de Ecologíca y Desarollo Sustentable del Estado de Sonora (CEDES) in Mexico. I obtained Mexican samples with assistance from CEDES and FR Méndez de la Cruz, Instituto de Biología, Universidad Nacional Autónoma de México (UNAM). Permits for Mexican samples were facilitated by Secretaría de Medio Ambiente y Recursos Naturales (SEMARNAT). Field collection of samples in Mexico would not have been possible without the help of C Melendez, P Rosen, M Vaughn, A Karl, P Woodman, L Smith, M Figueroa, Y Leon and numerous volunteers that participated in the project. K Berry, US Geological Survey graciously provided samples from California collected by K Anderson, T Bailey, R Evans, R Woodard, P Woodman, P Frank, B Henen and K Nagy. Many of the Arizona samples were provided by C Jones and the Arizona Game and Fish Department and were collected by D Grandmaison, S Goodman and P Woodman. A McLuckie granted use of samples of G. agassizii collected in Utah. B Brandner provided samples of G. berlandieri. I thank J Jarchow and the Sonoran Animal Hospital (Tucson, AZ) for their assistance in obtaining RNA samples from G. agassizii. RNA samples collected in Nevada were facilitated through J Germano, San Diego Zoo’s Desert Tortoise Conservation Center and R Averill-Murray, US Fish and Wildlife Service. Sample collection in Nevada was assisted by B Jurand, R Lamkin, S Madill, J Rose and A Dolan.

I thank J Still, J Galina-Mehlman and S Sotak at Arizona Research Laboratories and the University of Arizona Genetics Core (UAGC) for assistance with Sanger and next-

generation sequencing. J Barraza, Y Jimenez, M Christinsen and E Cox assisted with laboratory work.

Financial support came from several awards including the Fisher Scientific Scholar of Excellence Award, the Jarchow Conservation Award presented by the Tucson Herpetological Society and the David J. Morafka Memorial Research Award presented by the Desert Tortoise Council.

Finally, I thank the communities of the Tucson Herpetological Society and the Desert Tortoise Council for their continued inspiration and support.

DEDICATION

To my mom and dad, Sandy and Artemus Edwards who encouraged me to follow my passions; and to my daughter Amira, whom I hope will inherit the same joie de vivre that I was so fortunate to receive from my parents.

TABLE OF CONTENTS

ABSTRACT………………………………………………………………………..……08

SUMMARY………………………………………………………………………..……09

LITERATURE CITED…………………………………………….…………………..19

APPENDIX A…………………………………………………...………………………..……24

Testing Taxon Tenacity of Tortoises: Evidence for a geographical-selection gradient at a secondary contact zone; Published Manuscript: Ecology and Evolution

APPENDIX B …………………………………………………...………………………..……67

Shaping species with ephemeral boundaries; the distribution and genetic structure of desert tortoise (Gopherus morafkai) in the Sonoran Desert Region; Manuscript submitted for publication – in review

APPENDIX C…………………………………………………...………………………....…102

Assessing models of speciation under different biogeographic scenarios; an empirical study using multi-locus and RNA-seq analyses; Manuscript submitted for publication – in review

APPENDIX D………………………………………..………...………………………..……149

Population and Conservation Genetics; book chapter published in: Biology and Conservation of North American Tortoises. 2014. D.C. Rostal, E.D. McCoy and H.R. Mushinsky (eds). Johns Hopkins University Press, Baltimore, MD.

ABSTRACT

One of the important contributions genetic studies have made to conservation is the ability to resolve taxonomy and define relationships among populations. However, this can be complicated when species exhibit hybridization. Hybridization can be an important part of the evolutionary process and a critical component in a species ability to adapt to a changing environment. Most hybrid zones are observed at ecotones between two distinct habitats and this may be important in defining the role of hybrid zones in the evolutionary process. I examined hybridization among the three distinct lineages of desert tortoises in the genus Gopherus. An important aspect of this study system is the presence of areas of overlap between divergent lineages of desert tortoise which allowed me to test hypotheses about which forces influence these taxonomic boundaries. Specifically, I tested hypotheses about the contribution of physical vs. ecological segregation and the relative importance of isolation and gene flow in the formation of these disparate desert tortoise lineages. I used mtDNA sequence data and 25 microsatellite loci to perform Bayesian clustering, clinal analyses and habitat suitability modeling to infer population structure and influence of landscape features at each contact zone. In both instances, I observed ecological niche partitioning and limited hybridization at ecotones. I then used mtDNA and four nDNA loci to perform a multi-locus phylogenetic analysis to estimate the species tree among desert tortoise lineages and tested for ancestral admixture with RNA-seq data using demographic inference employed in the software package ∂a∂i. My results validate taxonomic distinction among all three lineages without evidence of ancestral introgression. These data suggest that despite the presence of contemporary hybridization and incomplete reproductive isolation, divergence among these lineages is consistent with species-level differentiation. By clarifying the evolutionary processes that influence the distribution of desert tortoise lineages, this study will directly inform efforts to preserve the evolutionary potential of these threatened species. Ultimately, understanding the evolutionary history of desert tortoises not only clarifies the forces that have driven speciation in this group, but it also contributes to our knowledge of the biogeographic history of the southwestern deserts and how diversity is maintained within them.

SUMMARY

I have now studied desert tortoises for more than fifteen years and I still find myself asking, “What is a desert tortoise”? Through increasing my knowledge of molecular genetics, I have found new ways to answer this question. The focus of this dissertation is directed toward the evolutionary history of desert tortoise populations over the last 6 million years and specifically the role that habitat selection has played in shaping taxonomic boundaries.

When I started this dissertation work in 2009 there was only one species of desert tortoise, Gopherus agassizii, but during the progress of this research colleagues and I defined G. morafkai which officially made a taxonomic distinction between Mojave and Sonoran lineages (Murphy et al. 2011). Now, after completing this work, I can identify the third member of this desert tortoise trichotomy, the Sinaloan lineage, as being taxonomically distinct.

I was first motivated to explore these evolutionary topics when I learned of the Black Mountain population of G. agassizii in northwestern Arizona. Here, McLuckie et al. (1999) characterized a “Mojave” population of desert tortoise at the edge of the range of the “Sonoran” population. I performed some preliminary assessments of these individuals and observed both pure Mojave G. agassizii as well as individuals that could be characterized as hybrids. I was intrigued as to how such a small population might be maintained in the presence of potential introgression.

Then, in 2005 I was fortunate to be invited to participate in a multinational, cooperative research effort focusing on crucial aspects of desert tortoise health, genetics, general biology and ecology in Mexico. This allowed me to obtain samples from throughout the entire range of the Desert Tortoise, including Sonora and Sinaloa, Mexico. My preliminary data suggested that what we were then calling Gopherus agassizii was potentially three separate lineages (Mojave, Sonoran, and Sinaloan). However, the distribution of the Sonoran- and Sinaloan-type tortoises in Mexico was yet to be defined and lacked easily identifiable geographic barriers.

I was intrigued not just by how the distributions of these different tortoise populations were maintained in the present, but also by how they evolved to be this way. Winston Churchill is attributed to saying, “The farther backward you can look, the farther forward you are likely to see” and although tortoises were likely far from his mind, I also saw that a deeper understanding of the evolutionary history of desert tortoises could help us make better choices now about what they need to persist.

It was clear that habitat played an important role in the distribution of disparate desert tortoise lineages and the principles of biogeography drove many of the hypotheses I 9

began to develop to explain their divergence; specifically, the contribution of physical vs. ecological segregation and the relative importance of isolation and gene flow in the formation of these lineages. Ultimately, my study organism became a powerful system to study theoretical aspects of speciation. In addition, by clarifying the evolutionary processes that influence the distribution of desert tortoise lineages, I hope this research will directly inform efforts to preserve the evolutionary potential of these threatened species.

Resulting Research

I applied multiple molecular methods in an empirical study to assess the contribution of gene flow under different speciation scenarios. In the Appendices of this dissertation I report on the results of this research:

In Appendix A, I focus on an anomalous population of G. agassizii that occurs east of the Colorado River in and around the Black Mountains of Arizona where it comes into contact with G. morafkai. This secondary contact zone between the two species of desert tortoises provides an opportunity to examine reinforcement of species’ boundaries under natural conditions. My objective was to describe the distribution of G. agassizii and G. morafkai where they come into contact in northwestern Arizona and to investigate the occurrence of hybridization among the parental lineages. My collaborators and I sampled 234 tortoises representing G. agassizii in California (n = 103), G. morafkai in Arizona (n = 78), and 53 individuals of undetermined assignment in the contact zone including and surrounding the Black Mountains. We genotyped individuals for 25 short tandem repeats and determined maternal lineage using mtDNA sequence data. We performed multi-locus genetic clustering analyses and used multiple statistical methods to detect levels of hybridization. We used habitat suitability models to define the properties of the contact zone and distribution of habitat specific for each species. Our data suggest a recent shared ancestry (~2,400 years) between G. agassizii populations directly across the Colorado River. In the contact zone 60% of individuals genotyped as G. agassizii, 6% as G. morafkai and 34% as hybrids, primarily of F2 or backcrossed recombinant classes. The maintenance of the hybrid zone is best described by a geographical selection-gradient model. G. agassizii and G. morafkai maintain independent taxonomic identities likely due to ecological niche partitioning.

In Appendix B, I investigate the biogeography and evolutionary history of Morafka’s desert tortoise, Gopherus morafkai, in Mexico. This project was part of a multi-national collaborative effort and our samples included 155 wild desert tortoises collected in Mexico from 2005 to 2013. We sampled the three major biomes where tortoises occur: Sonoran desertscrub, Sinaloan thornscrub, and tropical deciduous forest (TDF). We analyzed mtDNA sequences to reconstruct the matrilineal genealogy and estimate times 10

of divergence using Bayesian clustering. We inferred genetic population structure using analyses of 25 microsatellite (STR) loci. We observed a deep divergence (~5.7 Ma) between matrilines of G. morafkai and strong genetic differentiation across the STR loci. We conclude that Gopherus morafkai consists of two genetically and geographically distinct “Sonoran” and “Sinaloan” lineages. Each lineage exhibits genetic structure with isolation-by-distance. Both lineages occur in a relatively narrow zone of overlap in Sinaloan thornscrub where we detect limited introgression. We observed a sharp cline between the two lineages, which suggests that divergence may have been influenced by periods of isolation in temporary refugia, driven apart during periods of climatic flux during the Pleistocene. We propose that the shifting ecotone between Sinaloan thornscrub and Sonoran desertscrub biomes acts as an ephemeral boundary that fosters adaptations in each lineage of tortoise and results in their parapatric distribution. Despite incomplete reproductive isolation, the Sonoran and Sinaloan lineages of G. morafkai are on separate evolutionary trajectories. Conservation and management of both lineages will benefit from focusing on specific actions toward each lineage independently based on their habitat and resource needs.

Finally, in Appendix C, I build hypotheses based on the two previous studies and use this to investigate more theoretical aspects of the process of speciation. Here, my coauthors and I used mtDNA and four nDNA loci to perform a multi-locus phylogenetic analysis to estimate the species tree among desert tortoise lineages. We tested for ancestral admixture with RNA-seq data using the Diffusion Approximation for Demographic Inference employed in the software package ∂a∂i. Our results validate species-level differentiation among the three lineages without evidence of ancestral introgression. These data suggest that despite the presence of contemporary hybridization and incomplete reproductive isolation, divergence among these lineages occurred in the absence of gene flow, whether through physical allopatry or ecological niche segregation.

Interpretation of Results

Informed by this new evidence, we can infer an evolutionary scenario for this taxon based on the biogeographic history of the region. Some have proposed that many of the characteristic taxa of the Sonoran Region are tropical in origin (Riddle & Hafner 2006) and it is likely that tropical forest covered the region through the early Miocene (23.7 – 5.3 Ma). It is generally agreed upon that the region itself did not begin its drying trend until around 15-8 Ma due to episodes of vicariance, such as the formation of the Sierra Madre (Van Devender 2000). By the end of the Miocene (5.3 Ma), the physical topography was in place for the formation of the current North American deserts and their diverse biota. Here, the common ancestor of the three desert tortoise lineages may have been widespread throughout what is now the Mojave and Sonoran desert regions. This changing environment would have opened up new niches in the northern portion of 11

the ancestral range of the desert tortoise and opportunities to adapt to more arid conditions could have started the process of ecological divergence. During this same time period on the western side of the Sonoran Desert, tectonic activity between the Pacific Plate and the North American Plate resulted in the formation of the Baja Peninsula (Gastil et al. 1983). The exact timing of these events is debated, with estimates beginning from 14 Ma to 5.5 Ma, and climaxing approximately 3 Ma (Wilson & Pitts 2010). This resulted in the formation of the Sea of Cortez and it has been proposed that this waterway (called the Bouse embayment) extended north into the continent between 8-4 Ma. This waterway created a barrier between what is currently the Sonoran Desert and the Mojave Desert (summarized in Wilson & Pitts 2010). It was this geologic event that isolated the “Mojave” lineage of desert tortoise (G. agassizii) and allopatric speciation was imminent. G. morafkai had potentially already begun to segregate into tropical and arid ecotypes (Sinaloan and Sonoran lineages) during this time although much of the Sonoran region was still covered in tropical forests or desert thornscrub during the warming period of the Pliocene (5.3 to 1.8 Ma; Shreve 1942).

By the late Pleistocene, the Sonoran Desert region was probably interspersed with woodland plants originating from the north and tropical forest elements encroaching from the south, each retreating and expanding many times, but with semi-arid woodlands occupying much of the current desert lowland the majority of the time. Warm desert communities (dominated by creosote, Larrea tridentata) were restricted to areas below 300 m (Van Devender 2000). However, more tropical conditions likely fluctuated with glacial/interglacial climatic cycles (Oswald & Steadman 2011). This fluctuating environment may have influenced the formation of the thornscrub vegetative community (Riddle & Hafner 2006). Even though about 85% of regional thornscrub is in Sonora, Mexico, phytogeographers have characterized it as “Sinaloan” due to floristic affinities to tropical deciduous forests (Brown 1982). Thornscrub shares many of the structural and biotic features that occur throughout the aborescent subdivisions of the Sonoran Desert today and may be the source of many of the region’s endemic species (Van Devender, 2000). The distribution of thornscrub stands are disjunct; especially where they form the ecotone with desertscrub and a mosaic with the taller TDF. Although characteristic thornscrub maintains 100% ground cover, where it gradates into desertscrub it becomes patchy (Felger et al. 2001). The transition of TDF and thornscrub to desertscrub dominated by more xeric species is often between 200-300 m but with notable exceptions (Van Devender et al. 2000). For example, patches of thornscrub can be found in the foothills of the Sierra Madre and within the Plains of Sonora well north of Hermosillo (Shreve 1951).

These past events of global climate change affected the entire American Southwest, yet the more northern latitudes were affected to a far greater extent than the southern regions

(Meltzer & Holliday 2010). Accurate paleoenvironment reconstructions have been challenging in part due to inconsistency in the evidence of environmental change based on packrat middens, lake sediment records, speleotherms, and alluvial chronology reconstructions (Ballenger et al. 2011). However, geomorphogy likely drove the mosaic structure among tropical forest, thornscrub and semi-arid (temperate) woodland biotic communities and glacial cycling resulted in their oscillating boundaries. Some desert species show evidence of utilizing “refugia” during glacial cycles, but the timing and location has been inconsistent among taxa (Van Devender et al. 1987; Bell et al. 2010; Ceballos et al. 2010; Dominguez-Dominguez et al. 2011). For example, while modern vegetation was not established until ~5,400 BP at desert sites such as the Puerto Blanco Mountains, Arizona (Van Devender et al. 1987), Sonoran desertscrub has been established for at least the last 10,000 years on the Sonoran coastal plains on xeric east- facing slopes at Sierra Bracha, Mexico (Anderson & Van Devender 1995). In the northern part of the desert, the Lower Colorado River Valley (below 300 m) may have been a core area of desert throughout most of the Quaternary (Van Devender et al. 1987). Likewise, vegetation communities in the Gran Desierto area consisting of creosote bush (Larrea divericata) and white bursage (Ambrosia dumosa) seem to have undergone little change over the Holocene (Van Devender et al. 1987). The Pleistocene glacial/interglacial climatic cycles may have intensified the physiological barriers between tropical and arid environments (Ceballos et al. 2010) and for G. morafkai, would have further driven divergence between the Sonoran and Sinaloan lineages.

The large number of isolated, endemic mammal species in the region, “Pleistocene relicts”, supports the presence of refugia which may have acted as centers of speciation in isolated regions (Ceballos et al. 2010). In addition, mixed lineage populations of the Mexican stoneroller fish (Campostoma ornatum) in the Yaqui River basin support environmental changes during the Quaternary that would promote episodes of isolation with secondary contact in the southern portion of Sonora (Dominguez-Dominguez et al. 2011). In a phylogenetic study of six arid-adapted species of ground squirrels, Bell et al. (2010) observed evidence of recent expansion from several refugia including the Sonoran Coastal Plains. Within Xerospermophilus tereticaudus, they found evidence for both southern and northern Sonoran clades that exhibited northward movement post glaciation from an ancestral Sonoran distribution, consistent with our observed SON and SON_B mtDNA clades in G. morafkai (Appendix B). Other reptile species restricted to the Sonoran Desert exhibit similar patterns of population genetic structure that are consistent with fluctuating climatic regimes during the Pleistocene, including the Gila monster (Heloderma suspectum; Douglas et al. 2010) and rattlesnakes (Crotalus ruber and C. tigris; Douglas et al. 2006).

During the early Holocene (11,700 yrs BP) the range of G. morafkai likely extended as far east as southwestern New Mexico, which is currently dominated by Chihuahuan Desert vegetation (Van Devender et al. 1976). This refutes the theory that desert faunas may have been restricted to Mexican refugia during glacial periods (Holman 1995). An alternative explanation, however, is that Sonoran assemblages of species extended into southwestern New Mexico and Southeastern Arizona not from the west, but from the south. The western edge of the Sierra Madre currently maintains “fingers” of thornscrub that extend to the north, flanking the desertscrub communities. These vegetation corridors may have easily extended past the current international border during warmer conditions in the early Holocene. This area in the southeastern corner of Arizona is associated with the Cochise Filter Barrier (Castoe et al. 2007) and represents a point of connection between multiple Sonoran and Chihuahuan taxa. Today, there are remnants of other “strictly” Sonoran species like tiger rattlesnakes (C. tigris) around Douglas, Arizona and a small number of tortoises have been reported from this region (Germano et al. 1994). In 2009 I performed a genetic analysis on a tortoise found at San Bernardino National Wildlife Refuge, Cochise County, Arizona (collected by Bill Radke, Refuge Manager) and I found that its genetic profile was consistent with tortoises found in central Arizona (e.g. Tucson/Phoenix area). I concluded that this individual was likely a released or escaped captive (data not presented).

In contrast, other studies suggest that desert flora probably survived the last glacial period as disparate populations isolated in microsites within woodland habitats, as opposed to single refugia (Holmgren et al. 2014). This is consistent with the Gleasonian paradigm that species are distributed based on their own attributes and respond independently to environmental conditions and are not confined to a community per se (Gleason 1926). Desert tortoise populations are currently distributed in small, seemingly isolated patches throughout the Sonoran Desert and this scenario of maintaining their populations as a meta-population is consistent with patchy distribution of habitat and the strong isolation by distance genetic structure in our microsatellite data (representing recent gene flow). However, the predominant haplotype SON_01 and its associated star network is more consistent with the refugia scenario, albeit at a deeper time scale then observed for the pattern in STRs (Appendix B). Thus, multiple scenarios may be inferred depending on the time-scale under consideration.

At the northern end of the distribution of desert tortoises, in the range of G. agassizii, tortoises may have also experienced population contractions and expansions correlating with Pleistocene climate fluctuations. Bradford et al. (2003) suggested that the Amargosa River and the lower Colorado River were sources for post-Pleistocene expansions of B. punctatus into the surrounding areas post glaciation. Death Valley and the Amargosa River drainage in the northern Mojave Desert and the area of the lower Colorado River

extending northward into southern Nevada both appear to have maintained more desert- like conditions during the Wisconsin glacial period (Betancourt et al. 1990) and these areas may have acted as refugia for desert-evolved species. However, the faunal assemblage at these sites also contained numerous species common to colder regions (Spaulding 1990).

Our findings (Appendix A) are consistent with G. agassizii having potentially experienced periods of isolation in these proposed refugia. We observe much greater genetic differentiation between the Eastern Mojave Recovery Unit (RU) and the Colorado Desert RU in California than between the G. agassizii populations isolated across the Colorado River (Appendix A). We included the Eastern Mojave RU populations in our analysis as an equidistant comparison between Arizona G. agassizii population and their source populations west of the Colorado River. Our results are consistent with Hagerty et al. (2010) in that the Providence and New York mountain ranges are a strong barrier to gene flow between the Eastern Mojave RU and Chemehuevi Valley, Fenner, and other populations to the east (Appendix A). This genetic differentiation corresponds to the divergence between the northern Mojave “MOJ_B’ and the California ‘MOJ_A’ maternal clades which exhibit a deep split between these regions. However, strong segregation between MOJ_A and MOJ_B mtDNA distributions where there are not physical barriers to gene flow suggest a unique population history of northern clades and continued maintenance of genetic structure, likely via local adaptation.

Conclusions

What is particularly of interest in our comparison of models of divergence among the three desert tortoise taxa is that the genetic signature is the same under our two different proposed scenarios: an allopatric model of speciation between Mojave and Sonoran lineages and a parapatric model between Sonoran and Sinaloan lineages. All three lineages appear to have diverged at relatively the same time (5.7-5.9 Ma) and signatures of past introgression during divergence were not detectable. While opportunities for gene flow between Sonoran and Sinaloan lineages may have intermittently existed during these oscillating periods of isolation in refugia followed by secondary contact (as exists now), our results suggest that their initial differentiation occurred in isolation (geographical or ecological) and signatures of ancestral introgression were not detected by our analyses.

purged through backcrossing. Any remaining signature, even if maintained under positive selection, will contribute only a very minor portion of the existing genome (Mendez et al. 2012). Where hybridizing species exhibit ecological segregation, divergent selection acting on specific genomic regions can prevent the homogenizing effects of gene flow (Andrew & Rieseberg 2013, Chapman et al. 2013, de la Torre et al. 2014).

Each pair of desert tortoise lineages had a similar (and simultaneous) process of speciation. Divergence driven by either geography (geographic distance and physical barriers) or selection through ecological niche segregation should reduce gene flow, and both may even occur simultaneously (Endler 1977, Cooke et al. 2014). Gopherus agassizii likely first diverged as a result of isolation and the observed secondary contact zone in northwestern Arizona shows that through time it has adapted to the unique environmental conditions of the Mojave Desert (Appendix A). Such adaptation may first occur in allopatry and, thus, ecological speciation does not necessarily require sympatry (Bernardi 2013). Geographic isolation may also be a reasonable explanation for the divergence of the Sonoran and Sinaloan lineages, where an allopatric ‘speciation pump’ may have facilitated divergence (April et al. 2013). However, there is little evidence for geographic barriers that would have maintained their long-term isolation over time. An alternative explanation, and one that may better fit the geographical history of the region is that Sonoran and Sinaloan lineages first diverged into distinct ecotypes under a parapatric model and that isolation in temporary, Pleistocene refugia only reinforced the differentiation that had already occurred (Fisher-Reid et al. 2013). However, in this scenario our findings suggest that the genetic signature of ecological isolation (parapatric model) cannot be differentiated from geographic isolation (allopatric model).

We cannot exclude the possibility of anthropogenic influences on the distribution of tortoises in our study. Desert tortoises are known to have been moved by indigenous peoples, including the Seri Indians who have potentially moved tortoises between the mainland and Tiburòn Island (Felger et al. 1981; Nabhan, 2002). More recently, throughout the species’ range local people keep tortoises as pets where they interbreed in captivity and escape or are released back to the wild (Edwards et al. 2010; Edwards & Berry 2013). For a few individuals that we sampled in the wild, we observed signs of human intervention, such as paint on the shell (ALC, ALS and AED sites; Appendix B) or even a metal ring pierced through the supracaudal scute (RLP site; Appendix B). The RLP site (Rancho La Pintada) is highly accessible, not far south of the major city of Hermosillo, and is visited relatively often by people in the area to picnic and view petroglyphs. Local residents confirmed that people release pet tortoises into, or remove tortoises from the area. We characterize the habitat at this site as transitional and found late-generation hybrids present. In Hermosillo, El Centro Ecológico del Estado de Sonora (CES) maintains a large number of captive tortoises. Most of these animals were brought

in by the public, which is an indication that keeping tortoises as pets is common in Mexico, as it is in the U.S. We genotyped 20 individuals from CES (data not presented) and observed both parental types as well as F1 hybrids, suggesting that animals were moved large distances from their original capture location and then bred in captivity. Considering this information, it is possible that the detection of hybrid individuals in our study has been elevated. However, the remote locations of most of the sites and the strong ecological association we observed in our data suggest that the distribution of parental genotypes and the maintenance between them at the ecotone follows patterns expected without anthropogenic influence.

Implications

Other tortoise species with large, continuous distributions do not exhibit the deep phylogenetic structure we observe within G. morafkai between the Sonoran and Sinaloan lineages; e.g. Gopherus polyphemus (Ennen et al. 2012), Testudo graeca (Fritz et al. 2009), T. hermanni boettger (Fritz et al. 2006) and Stigmochelys pardalis (Fritz et al. 2010). In addition, these studies observed multiple intermediate haplogroups in contrast to the bifurcated tree that defines the G. morafkai mtDNA genealogy. The Chaco tortoise (Chelonoidis chilensis) of Argentina and Paraguay is perhaps the most appropriate comparison in that it has similar latitudinal range (>1,500 km), exhibits clinal variation, and occupies a variety of arid environments, including plains, deserts and semi-deserts (Fritz et al. 2012). However, the mtDNA sequence divergence in C. chilensis is ~1.37% whereas in G. morafkai, the mtDNA sequence divergence between Sonoran and Sinaloan lineages was 4.3%.

This study resolves taxonomic uncertainty of Morafka’s desert tortoise (G. morafkai) by defining the mode of speciation between the Sonoran and Sinalon lineages. Agassiz’s desert tortoise (G. agassizii) is listed as threatened under the Endangered Species Act (USFWS 1990) and Morafka’s desert tortoise (G. morafkai) has been a candidate for federal listing under the U.S. Endangered Species Act since 2010 and is considered Wildlife of Special Concern in Arizona (USFWS 2010; AGFD 2012). Mexican populations of Gopherus receive protection as threatened species (Category A "Amenazada" in NOM-059; D.O.F. 2002). By clarifying the evolutionary processes that influence the distribution of desert tortoise lineages, this study informs efforts to preserve the evolutionary potential of these species.

The opportunity to describe the process of speciation in a single system under different biogeographic scenarios helps us better understand the relative importance of various mechanisms of speciation and the rates of speciation associated with them. In this case, the process of isolation (physical or ecological) appears to be the driving factor in the differentiation of the three lineages. By focusing on a single subject within the context of 17

its environment, the findings of this study may lessen the polarization among conflicting theories of speciation and exemplify that there are potentially multiple and equally viable scenarios through which speciation occurs. Ultimately, understanding the evolutionary history of desert tortoises not only clarifies the forces that have driven speciation in this group, but also contributes to our knowledge of the biogeographic history of the southwestern deserts and how diversity is maintained within them.

LITERATURE CITED

Anderson RS, Van Devender TR (1995) Vegetation history and paleoclimates of the coastal lowlands of Sonora, Mexico - pollen records from packrat middens. Journal of Aird Environments, 30, 295–306.

Andrew RL, Rieseberg LH (2013) Divergence is focused on few genomic regions early in speciation: incipient speciation of sunflower ecotypes. Evolution, 67, 2468– 2482.

April J, Hanner RH, Dion-Cote AM, Bernatchez L (2013) Glacial cycles as an allopatric speciation pump in north-eastern American freshwater fishes. Molecular Ecology, 22, 409–422.

Arizona Game and Fish Department (2012) Arizona’s State Wildlife Action Plan: 2012- 2022. Arizona Game and Fish Department, Phoenix, Arizona.

Ballenger JAM, Holliday VT, Kowler AL, Reitze WT, Prasciunas MM, Miller DS, Windingstad JD (2011) Evidence for Younger Dryas global climate oscillation and human response in the American Southwest. Quaternary International, 242, 502–519.

Bell K C, Hafner DJ, Leitner P, Matocq MD (2010) Phylogeography of the ground squirrel subgenus Xerospermophilus and assembly of the Mojave Desert biota. Journal of Biogeography, 37, 363–378.

Bernardi G (2013) Speciation in fishes. Molecular Ecology, 22, 5487–5502.

Betancourt JL, Van Devender TR, Martin PS eds (1990) Packrat middens: the last 40,000 years of biotic change. University of Arizona Press, Tucson, AZ.

Bradford DF, Neale AC, Nash MS, Sada DW, Jaeger JR (2003) Habitat patch occupancy by toads (Bufo punctatus) in a naturally fragmented desert landscape. Ecology, 84, 1012–1023.

Brown DE (1982) Sinaloan thornscrub. Desert Plants, 4, 101–105.

Castoe TA, Spencer CL, Parkinson CL (2007) Phylogeographic structure and historical demography of the western diamondback rattlesnake (Crotalus atrox): A perspective on North American desert biogeography. Molecular phylogenetics and evolution, 42, 193–212. 19

Ceballos G, Arroyo-Cabrales J, Ponce E (2010) Effects of Pleistocene environmental changes on the distribution and community structure of the mammalian fauna of Mexico. Quaternary Research, 73, 464–473.

Chapman MA, Hiscock SJ, Filatov DA (2013) Genomic divergence during speciation driven by adaptation to altitude. Molecular Biology and Evolution, 30, 2553– 2567.

Cooke GM, Landguth EL, Beheregaray LB (2014) Riverscape genetics identifies replicated ecological divergence across an Amazonian ecotone. Evolution, 68, 1947–1960. de la Torre AR, Roberts DR, Aitken SN (2014) Genome-wide admixture and ecological niche modeling reveal the maintenance of species boundaries despite long history of interspecific gene flow. Molecular Ecology, 23, 2046–2059.

Domínguez-Domínguez O, Vila M, Pérez-Rodríguez R, Remón N, Doadrio I (2011) Complex evolutionary history of the Mexican stoneroller Campostoma ornatum Girard, 1856 (Actinopterygii: Cyprinidae). BMC Evolutionary Biology, 11, 153. doi:10.1186/1471-2148-11-153

Douglas ME, Douglas MR, Schuett GW, Porras LW (2006) Evolution of rattlesnakes (Viperidae; Crotalus) in the warm deserts of western North America shaped by Neogene vicariance and Quaternary climate change. Molecular Ecology, 15, 3353–3374.

Douglas ME, Douglas MR, Schuett GW, Beck DD, Sullivan BK (2010) Conservation phylogenetics of helodermatid lizards using multiple molecular markers and a supertree approach. Molecular Phylogenetics and Evolution 55, 153–167.

Edwards T, Jarchow CJ, Jones CA, Bonine KE (2010) Tracing Genetic Lineages of Captive Desert Tortoises in Arizona. Journal of Wildlife Management, 74, 801– 807.

Edwards T, Berry KH (2013) Are captive tortoises a reservoir for conservation? An assessment of genealogical affiliation of captive Gopherus agassizii to local, wild populations. Conservation Genetics, 14, 649–659.

Endler JA (1977) Geographic variation, speciation, and clines. Princeton University Press, Princeton, N.J. USA.

Ennen JR, Kreiser BR, Qualls CP, Gaillard D, Aresco M, Birkhead R, Tuberville TD, McCoy ED, Mushinsky H, Hentges TWB, Schrey A (2012) Mitochondrial DNA

Assessment of the Phylogeography of the Gopher Tortoise. Journal of Fish and Wildlife Management, 3, 110–122.

Felger RS, Moser MB, Moser EW (1981) The desert tortoise in Seri Indian culture. Proceedings of the Desert Tortoise Council Symposium, 1981, 6.

Felger RS, Johnson MB, Wilson MF (2001) Trees of Sonora, Mexico. Oxford University Press, New York, NY USA.

Fisher-Reid MC, Engstrom TN, Kuczynski CA, Stephens PR, Wiens JJ (2013) Parapatric divergence of sympatric morphs in a salamander: incipient speciation on Long Island? Molecular Ecology, 22, 4681–4694.

Fritz U, Auer M, Bertolero A, Cheylan M, Fattizzo T, Hundsdorfer AK, Sampayo MM, Pretus JL, Široky P, Wink M (2006) A rangewide phylogeography of Hermann's tortoise, Testudo hermanni (Reptilia: Testudines: Testudinidae): implications for taxonomy. Zoologica Scripta, 35, 531–543.

Fritz U, Harris DJ, Fahd S, Rouag R, Martinez EG, Casalduero AG, Siroky P, Kalboussi M, Jdeidi TB, Hundsdorfer AK (2009) Mitochondrial phylogeography of Testudo graeca in the Western Mediterranean: Old complex divergence in North Africa and recent arrival in Europe. Amphibia-Reptilia, 30, 63–80.

Fritz U, Daniels SR, Hofmeyr MD, González J, Barrio-Amorós CL, Široký P, Hundsdörfer AK, Stuckas H (2010) Mitochondrial phylogeography and subspecies of the wide-ranging sub-Saharan leopard tortoise Stigmochelys pardalis (Testudines: Testudinidae) - a case study for the pitfalls of pseudogenes and GenBank sequences. Journal of Zoological Systematics and Evolutionary Research, 48, 348–359.

Fritz U, Alcalde L, Vargas-Ramirez M, Goode EV, Fabius-Turoblin DU, Praschag P (2012) Northern genetic richness and southern purity, but just one species in the Chelonoidis chilensis complex. Zoologica Scripta, 41, 220–232.

Gastil G, Minch J, Phillips RP (1983) The geology and ages of islands. In: Island Biogeography in the Sea of Cortéz (eds Case TJ, Cody ML) pp. 13–15. California University Press, Berkeley, CA.

Germano DJ, Bury RB, Esque TC, Fritts TH, Medica PA (1994) Range and habitats of the desert tortoise. In: Biology of North American Tortoises (eds Bury RB, Germano DJ) (eds.), pp. 73–84. U.S. Department of the Interior National Biological Survey, Fish and Wildlife Research 13, Washington, D.C.

Gleason HA (1926) The individualistic concept of the plant association. Bulletin of the Torrey Botanical Club, 53, 7–26.

Hagerty BE, Nussear KE, Esque TC, Tracy CR (2010) Making molehills out of mountain landscape genetics of the Mojave desert tortoise. Landscape Ecology, 2, 267–280.

Holman JA (1995) Pleistocene amphibians and reptiles in North America. Oxford University Press, New York.

Holmgren CA, Betancourt JL, Peñalba MC, Delgadillo J, Zuravnsky K, Hunter KL, Rylander KA, Weiss JL, Riddle B (2014) Evidence against a Pleistocene desert refugium in the Lower Colorado River Basin. Journal of Biogeography, 41, 1769–1780.

McLuckie AM, Lamb T, Schwalbe CR, McCord RD (1999) Genetic and morphometric assessment of an unusual tortoise (Gopherus agassizii) population in the Black Mountains of Arizona. Journal of Herpetology, 33, 36–44.

Meltzer DJ, Holliday VT (2010) Would North American Paleoindians have noticed Younger Dryas age climate changes? Journal of World Prehistory, 23, 1–41.

Mendez FL, Watkins JC, Hammer MF (2012) A haplotype at STAT2 introgressed from Neanderthals and serves as a candidate of positive selection in Papua New Guinea. The American Journal of Human Genetics, 91, 265–274.

Murphy RW, Berry KH, Edwards T, Leviton AE, Lathrop A, Riedle JD (2011) The dazed and confused identity of Agassiz's land tortoise, Gopherus agassizii (Testudines: Testudinidae) with the description of a new species, and its consequences for conservation. ZooKeys, 113, 39–71.

Nabhan GP (2002) When Desert Tortoises Talk, Indians Listen. The Sonoran desert tortoise; natural history, biology and conservation (ed. by T.R. Van Devender), pp. 355–375. University of Arizona Press, Tucson, Arizona USA.

Oswald JA, Steadman DW (2011) Late pleistocene passerine birds from Sonora, Mexico. Palaeogeography, Palaeoclimatology, Palaeoecology, 301, 56–63.

Riddle B, Hafner D (2006) A step-wise approach to integrating phylogeographic and phylogenetic biogeographic perspectives on the history of a core North American warm deserts biota. Journal of Arid Environments, 66, 435–461.

Shreve F (1942) The desert vegetation of North America. The Botanical Review, 8, 195– 246.

Shreve F (1951) Vegetation of the Sonoran Desert. Carnegie Institute of Washington, Washington, DC.

Spaulding WG (1990) Vegetational and climatic development of the Mojave Desert: the last glacial maximum to the present. In: Packrat middens: the last 40,000 years of biotic change (eds Betancourt JL, Van Devender TR, Martin PS), pp. 166–199. University of Arizona Press, Tucson, AZ.

U.S. Fish and Wildlife Service (1990) Endangered and threatened wildlife and plants: Determination of threatened status for the Mojave population of the desert tortoise. Federal Register, 55, 12178–12191.

U.S. Fish and Wildlife Service (2010) Endangered and threatened wildlife and plants; 12- month finding on a petition to list the Sonoran population of the desert tortoise as endangered or threatened; proposed rule. Federal Register, 75, 78094–78146.

Van Devender TR (2000) The deep history of the Sonoran Desert. In: A Natural History of the Sonoran Desert (eds Phillips SJ, Comus PW), pp. 61–69. University of California Press, Berkeley, CA.

Van Devender TR, Moodie KB, Harris AH (1976) The Desert Tortoise (Gopherus agassizi) in the Pleistocene of the Northern Chihuahuan Desert. Herpetologica, 32, 298–304.

Van Devender TR, Thompson RS, Betancourt JL (1987) Vegetation history in the Southwest: the nature and timing of the late Wisconsin-Holocene transition. In: North America and Adjacent Oceans during the Last Deglaciation (eds Ruddiman WF, Wright JHE), pp. 323–352. Geological Society of America.

Van Devender TR, Sanders AC, Wilson RK, Meyer SA (2000) Vegetation, flora, and seasons of the Río Cuchujaqui, a tropical deciduous forest near Alamos, Sonora, México. In: The Tropical Deciduous Forest of Alamos: Biodiversity of a Threatened Ecosystem in Mexico (eds Robichaux RH, Yetman DA), pp. 36–101. University of Arizona Press, Tucson, AZ.

Wilson JS, Pitts JP (2010) Illuminating the lack of consensus among descriptions of earth history data in the North American deserts: A resource for biologists. Progress in Physical Geography, 34, 419–441.

APPENDIX A

Published Manuscript: Ecology and Evolution

Testing Taxon Tenacity of Tortoises: Evidence for a geographical-selection gradient at a secondary contact zone

Taylor Edwards1,2, Kristin H. Berry3, Richard D. Inman4, Todd C. Esque4, Kenneth E. Nussear4, Cristina A. Jones1,5 and Melanie Culver1,6

1School of Natural Resources and the Environment. The University of Arizona, Tucson, AZ 85721 USA 2University of Arizona Genetics Core, University of Arizona, 1657 E. Helen Street, Tucson, AZ 85721, USA 3U.S. Geological Survey, Western Ecological Research Center, Falcon Business Park, 21803 Cactus Avenue, Suite F, Riverside, CA 92518, USA 4U.S. Geological Survey, Western Ecological Research Center, 160 North Stephanie St., Henderson, Nevada 89074, USA 5Arizona Game and Fish Department, Nongame Wildlife Branch, Phoenix, AZ 85086, USA 6U.S. Geological Survey, Arizona Cooperative Fish and Wildlife Research Unit, University of Arizona, Tucson, AZ 85721, USA

Corresponding author: Taylor Edwards Assistant Staff Scientist University of Arizona Genetics Core 1657 E. Helen Street, room 111 Tucson, Arizona 85721 [email protected]

Abstract

We examined a secondary contact zone between two species of desert tortoise, Gopherus agassizii and G. morafkai. The taxa were isolated from a common ancestor during the formation of the Colorado River (4-8 mya) and are a classic example of allopatric speciation. However, an anomalous population of G. agassizii comes into secondary contact with G. morafkai east of the Colorado River in the Black Mountains of Arizona and provides an opportunity to examine reinforcement of species’ boundaries under natural conditions. We sampled 234 tortoises representing G. agassizii in California (n = 103), G. morafkai in Arizona (n = 78), and 53 individuals of undetermined assignment in the contact zone including and surrounding the Black Mountains. We genotyped individuals for 25 STR loci and determined maternal lineage using mtDNA sequence data. We performed multi-locus genetic clustering analyses and used multiple statistical methods to detect levels of hybridization. We tested hypotheses about habitat use between G. agassizii and G. morafkai in the region where they co-occur using habitat suitability models. Gopherus agassizii and G. morafkai maintain independent taxonomic identities likely due to ecological niche partitioning and the maintenance of the hybrid zone is best described by a geographical selection-gradient model.

Keywords: Cline, Gopherus, Hybrid Zone, Mojave Desert, Secondary Contact, Sonoran Desert

Introduction

Exploring the relative importance of isolation and gene flow in the process of speciation is made possible when predictable patterns of divergence have occurred, such as regions where sister taxa come into secondary contact. When species diverge in allopatry, speciation may be incomplete because reinforcement is lacking and reproductive isolating mechanisms may not fully develop. Thus, hybridization may occur if the genetically distinct populations come into contact. Multiple models are used to describe hybrid zones where the amount of hybridization is dependent on factors such as dispersal ability, environment, and selection (Harrison 1993; Arnold 1997). Hybrid zones often are observed at ecotones between two distinct habitats (Harrison 1993; Arnold 1997) where exogenous selection may drive the amount of hybridization. Where hybrid zones are environment- and dispersal-dependent, a cline may be observed (Endler 1977) and cline width can be suggestive of the strength of selection (Smith et al. 2013).

The presence of a secondary contact zone between two species of desert tortoise (Gopherus agassizii and G. morafkai) in northwestern Arizona provides a natural experiment for testing the tenacity of these two taxa (Hewitt 1988). G. agassizii (Agassiz’s desert tortoise) and G. morafkai (Morafka’s desert tortoise; Murphy et al. 2011) differ in distribution, morphology, seasonal activity, reproductive ecology, habitat selection, and genetics (Lamb et al. 1989; McLuckie et al. 1999; Berry et al. 2002). The divergence between the two desert tortoise species appears to be a classic example of allopatric speciation resulting from geographic isolation 4-8 million years ago (mya) by the Bouse embayment, which is now occupied by the Colorado River (Lamb et al. 1989; Avise et al. 1992; McLuckie et al. 1999; Murphy et al. 2011). Gopherus agassizii is distributed primarily in the Mojave Desert and the lower Colorado River subdivision of the Sonoran Desert, west and north of the Colorado River (Berry et al. 2002). Gopherus morafkai is found entirely east and south of the Colorado River in the Sonoran Desert region. The general consensus is that the Colorado River has been throughflowing continuously along the Arizona-California border for at least 4.3 mya (House et al. 2005; Wilson and Pitts 2010). This waterway resulted in the divergence of multiple species found on either side of the Colorado River (Wood et al. 2013).

Preliminary genetic work identified a possible population of G. agassizii east of the Colorado River in the Black Mountains of Arizona (Glenn et al. 1990) and then McLuckie et al. (1999) characterized the population as Mojavean based on mitochondrial DNA and morphometrics. This isolated population of G. agassizii has been hypothesized to have resulted from meandering or drying of the river or even as a result of historic or prehistoric human translocation (McLuckie et al. 1999). The Black Mountains and surrounding area exhibit a complex composition of flora where the Mojave and Sonoran desert ecosystems converge (McLuckie et al. 1999). The Black Mountain population of 26

G. agassizii is in proximity to multiple other G. morafkai tortoise populations and McLuckie et al. (1999) observed that tortoises characterized as Mojavean based on mtDNA but residing in the Black Mountains occupied habitat more typical of G. agassizii than of neighboring G. morafkai. Hybridization between G. agassizii and G. morafkai has been observed in captivity (Edwards et al. 2010) but has not been studied in a natural setting.

In this study, our objective was to describe the distribution of G. agassizii and G. morafkai where they come into contact in northwestern Arizona and to investigate the occurrence of hybridization among the parental lineages. We tested the hypothesis that despite an apparent lack of pre-zygotic reproductive isolating mechanisms, G. agassizii and G. morafkai maintain independent taxonomic identities based on ecological niche specialization. We employed genetic analyses to explore the extent of hybridization and used habitat suitability models to define the topographic, climatic and vegetative properties of the contact zone and specific distribution of habitat for each species.

Materials and Methods

Sample Collection

We analyzed tissue samples collected from a total of 234 tortoises representing G. agassizii in California (n = 103), G. morafkai in Arizona (n = 78), and 53 individuals of undetermined lineage in the presumed contact zone (Table 1). The California samples represent two different Federal Recovery Units (RUs; USFWS 2011): the Colorado Desert RU (n = 45) and the Eastern Mojave RU which borders the Colorado Desert RU to the west and north but is geographically separated by mountain ranges (Figure 1). Samples from the Colorado Desert RU are west of and across the Colorado River from the Black Mountains. We included Eastern Mojave RU samples in the analysis as an equidistant comparison to Colorado Desert RU sites (Figure 1).

We captured desert tortoises by hand following federal and Arizona state protocols (Averill-Murray 2000; Berry and Christopher 2001), collected <1 ml whole blood via brachial, jugular, or subcarapacial venipuncture and stored the samples in 95% EtOH, EDTA, or lithium heparin. The University of Arizona Institutional Care and Use Committee (IACUC) approved all tortoise-handling protocols (IACUC Control nos. 09- 138 and 02-120). All California samples were collected between 1990 and 2002 and were previously analyzed by Murphy et al. (2007). Arizona samples were collected between 2005 and 2010. Because tortoises are long-lived animals with long generation times (ca. 25 years; USFWS 1994), we made the assumption that major changes in population genetic structure did not occur within the sampling period.

Sample Habitat

Tortoises in the Eastern Mojave RU typically occur on flats, valley bottoms, alluvial fans, and bajadas characterized by a broad range of Mojave Desert vegetation associations (Berry et al. 2002; USFWS 2011). Tortoises from the northern portion of the Colorado Desert RU occur in similar topography and vegetation associations (USFWS 2011). Tortoises in the Black, Buck, and New Water mountains in extreme western Arizona typically occur in habitat similar to that of the Eastern Mojave RU, but can also be found in deeply incised washes. As their distribution moves south and east, tortoises in Arizona will occasionally use flats, valley bottoms and alluvial fans, but more typically occur on rocky, steep, boulder-strewn slopes and bajadas, and ridges interspersed among shallow to deeply-incised washes (Averill-Murray et al. 2002; Grandmaison et al. 2010; Riedle et al. 2008). Tortoise habitat in Arizona comprises an amalgamation of distinct vegetative communities. From west to east, these communities shift from crucifixion-thorn series, pinyon juniper and Joshua tree series in the Hualapai Foothills, into the paloverde mixed- cacti series with influence of interior chaparral and Mojave desertscrub in the Arrastra Mountains, Bonanza Wash, and Little Shipp Wash (Turner and Brown 1982). At sites in central Arizona (Harcuvar and Wickenburg mountains), tortoises occupy habitat typical of the Arizona Upland vegetative community (Turner and Brown 1982). Sites in the southwest of Arizona (New Water and Eagletail mountains) occur in the Lower Colorado River Valley subdivision of the Sonoran Desert interspersed with elements of Arizona Upland vegetation (Turner and Brown 1982).

Molecular Techniques

We isolated genomic DNA from whole blood, salvaged red blood cells or lymphatic fluid. We amplified an ~1,500 base pair (bp) portion of mitochondrial DNA (including the ND3, arginine tRNA, ND4L, and part of ND4 genes) following methods in Edwards (2003) and Murphy et al. (2007) for PCR conditions. We submitted PCR products to the University of Arizona Genetics Core for DNA sequencing on a 3730XL DNA Analyzer (Applied Biosystems, Carlsbad, California USA). We aligned a 1,109 bp sequence using CLC DNA Workbench ver. 5.7.1 (CLC bio, Denmark).

We characterized all samples for 25 loci previously described as short tandem repeats (STRs): Cm58 (Fitzsimmons et al. 1995); Goag03, Goag04, Goag05, Goag06, Goag07, Goag32 (Edwards et al. 2003); Test56 (Hauswaldt and Glenn 2005); GP15, GP19, GP30, GP55, GP61, GP81, GP96, GP102 (Schwartz et al. 2003); ROM01, ROM02, ROM03, ROM04, ROM05, ROM07, ROM10 (Edwards et al. 2011); and ROM08, ROM09 (Davy et al. 2011). We followed protocols in Edwards et al. (2003), Murphy et al. (2007) and Edwards et al. (2011) for PCR conditions and multiplexing. We analyzed electropherograms using Genemarker 1.85 (SoftGenetics, State College, Pennsylvania 28

USA). We used a novel scoring nomenclature for locus Goag05. This locus exhibits a fixed motif in G. agassizii, whereas in G. morafkai it is observed with both the fixed motif and a functional microsatellite motif (Engstrom et al. 2007). We therefore treated Goag05 as two independent loci; Goag05a and Goag05b where "a" represents the variable motif A and "b" represents the fixed motif B (Engstrom et al. 2007). The pattern of amplification was informative to our investigation of hybridization because samples could easily be identified as G. morafkai (Goag05a heterozygous), G. agassizii (Goag05a null plus Goag05b fixed) or potentially admixed (Goag05a homozygous plus Goag05b fixed) based on the amplification of this marker alone.

Multi-locus Genetic Clustering Analyses

We used STRUCTURE 2.3.4 (Pritchard et al. 2000) to define populations in our dataset and to identify admixed individuals. We ran a multi-locus STR analysis using all 234 tortoises in the study. We used an admixture model with allele frequencies correlated among populations. We tested for K = 1-8 with 10 trials per K, each run for 500,000 iterations following a burn-in period of 50,000 Metropolis-coupled Markov chains (MCMC). We used STRUCTURE HARVESTER Online (Earl and Vonholdt 2012) to evaluate STRUCTURE and DeltaK to determine the best fit of K for the data following Evanno et al. (2005).

We used GENELAND (Guillot et al. 2005) to delineate the location and shape of hybrid zones. Geneland incorporates geographic coordinates into its Bayesian model-based clustering algorithm and spatially infers genetic discontinuities between populations. Geneland assumes that input coordinates are planar such as the Universal Transverse Mercator (UTM) coordinate system. Our sampling locations fall into 2 different transverse Mercator zones: 11S and 12S (datum WGS84). We first normalized our GPS data to geographic latitude and longitude coordinates and then converted all the points to UTM zone 11S using publically available tools (S. Dutch, Univ of Wisconsin; http://www.uwgb.edu/dutchs/UsefulData/HowUseExcel.HTM). We ran GENELAND for the STR dataset only and for the STR dataset combined with mtDNA haplotype information. We performed 20 independent runs testing for K = 1–6 populations for 500,000 iterations each, thinning every 500. We took into account the putative presence of the null allele in our model choice and assumed uncorrelated allele frequencies. We used a 1000m coordinate uncertainty which is greater than the average home range of G. morafkai (13.3 ±14.72 ha; Averill-Murray et al. 2002) but less than the maximum distance an individual tortoise is known to travel (>30 km; Edwards et al. 2004). For the STR only analyses, we also employed the admixture model of Guedj and Guillot (2011).

Hybrid Characterization

We assessed the presence of hybrid individuals in our dataset several ways; first, we made a qualitative guess based on the mtDNA haplotype and the presence/zygosity of locus Goag05a. We then identified potential hybrids using the inferred admixture proportion (Q) from the STRUCTURE analysis. We also used GeneClass 2 Version 2.0.h (Piry et al. 2004) for population assignments using STR loci. We formatted our input data files to include known populations of G. agassizii and G. morafkai individuals and the individuals of unknown genetic origin residing in the contact zone (Table 1). In GeneClass 2 we employed four different methods: 2 Bayesian methods (Baudouin and Lebrun 2001; Rannala and Mountain 1997), a frequency-based method (Paetkau et al. 1995) and a probability computation using Monte-Carlo resampling (Paetkau et al. 2004). We set the assignment threshold of scores to 0.001 and for the probability computation we simulated 10,000 individuals and set the type 1 error (alpha) to 0.01. For each of the 4 methods we ran the analysis with and without locus Goag05. Finally, we ran NewHybrids (Anderson 2008) to further predict the presence of admixed individuals in the contact zone using STR loci. We used the default genotype frequency classes in NewHybrids to compute the posterior probability that a hybrid individual belongs to a specified hybrid category (F1, F2 or backcross). We performed several trial runs to test the effect of changing the allele frequency prior from Jeffreys-like to uniform and for the mixing proportions prior (without major effect on the outcome). We ran the program three times: with and without Goag05a (500,000 iterations each) and once using gene frequency priors with Goag05a (180,000 iterations). We analyzed results after a burn-in of 50,000.

Population Descriptive Statistics

After determining the extent of hybridization in our sample set, we removed all probable hybrid individuals and then reassigned previously unknown individuals to either G. agassizii or G. morafkai. We then compared descriptive statistics among each of four populations: Eastern Mojave RU, Colorado Desert RU, Arizona G. agassizii and Arizona G. morafkai. We used ARLEQUIN v.3.11 (Excoffier et al. 2005) to detect significant departures from Hardy-Weinberg expectations (HWE; Guo and Thompson 1992), assess population differentiation using analysis of molecular variance (AMOVA), pairwise FST (Weir and Cockerham 1984), and Theta(H) (Ohta and Kimura 1973). We also used ARLEQUIN to test for linkage disequilibrium (nonrandom association between loci) among all pairs of loci within each population (Slatkin and Excoffier 1996). We used

FSTAT v.2.9.3.2 (Goudet 1995) to estimate inbreeding coefficients (FIS). We used default parameters in FSTAT and ARLEQUIN for all Markov-chain tests and permutations. We used BOTTLENECK (Piry et al. 1999) to test for evidence of historical changes in effective population sizes and deviations from equilibrium conditions. We ran

100,000 replications for each of three tests, Sign Test, Standardized Differences Test and Wilcoxon Test, under I.A.M., T.P.M. and S.M.M. models.

Coalescent Analyses

We performed phylogenetic analyses on the mtDNA sequence data to establish relationships among maternal lineages. We used BEAST v1.7.5 (Drummond and Rambaut 2007) to produce a phylogenetic tree and to establish estimates of time to most recent common ancestor (TMRCA). Previous estimates of mtDNA divergence time between G. agassizii and G. morafkai have been fairly consistent at 5-6 mya (Avise et al. 1992; Lamb and Lydeard 1994; McLuckie et al. 1999; Edwards 2003). We set a prior of 5.9 ±0.5 mya, normal probability distribution, for our Bayesian analysis in BEAST based on Edwards (2003) because this estimate used the same mitochondrial locus as in this study. We ran BEAST using haplotype sequences only and included one outgroup sequence of G. berlanderi (GenBank: DQ649409.1). We selected an HKY substitution model with the gamma parameter set to 4 using MrModeltest 2.3 (Nylander 2004) and we chose a relaxed, log-normal clock and Yule process model. We ran the MCMC for 500,000,000 generations, sampling every 5,000 with a burnin of 10%. We viewed results in TRACER: MCMC Trace Analysis Tool Version v1.6.0 (Rambaut et al. 2003-2013) and used TreeAnnotator v1.7.5 (Rambaut and Drummond 2002–2013) to select the Maximum Clade Credibility tree which has the highest product of the posterior probability of all its nodes from the BEAST analysis. We used FigTree Version 1.4.0 (Rambaut 2006-2012) to visualize the tree. In addition, we used PAUP* v.4.0b10 (Swofford 2002) to reconstruct maternal phylogenies using both likelihood and parsimony optimality criterion searches to generate tree topologies for comparison with the Bayesian analyses executed with BEAST. We performed analyses only with unique haplotypes giving all characters equal weight and we defined G. berlandieri as the outgroup. We performed a heuristic search with 100,000 random addition replicates. We estimated support for inferred relationships by conducting 10,000 nonparametric bootstrap replicates. We performed maximum likelihood analyses using the HKY model of nucleotide evolution.

We used IMa2 version 2.0 (Hey 2010) to estimate gene flow and divergence times among populations. This program estimates the posterior densities of the time of divergence, population size and gene flow using an MCMC. The model assumes constant population sizes and gene exchange rates that these data inevitably violate; however, this modeling exercise was primarily employed to estimate timing of divergence between cross-river G. agassizii populations. We used the three G. agassizii populations that we defined for generating descriptive statistics and removed all probable hybrids. We defined the population tree by ((0,1):3,2):4 with Colorado Desert RU = 0, Arizona G. agassizii = 1, and Eastern Mojave RU = 2 and populations 3 and 4 representing ancestral nodes. The 31

demographic parameters used by IMa2 are scaled by the mutation rate (μ; Hey 2010; Nielsen and Wakeley 2001) which the user must provide and generation time to convert estimates to demographic units. For STRs, estimating μ posed some challenges. Microsatellite mutation rates range from 10-6 to 10-2 per generation (Schlotterer 2000) and there are no empirical data available for desert tortoises. We estimated μ across all

STR loci by using the mean θ for STRs (θ =1.5: Table 2). Using θ = 4Neμ we assumed an Ne of 10,000 (to account for the ancestral population and large geographic area of sampling) and calculated μ = 3.75x10-5 (0.0001 - 0.00001). For mtDNA, we calibrated the mtDNA mutation rate for the locus based on μ = 4.0 x 10-9 per base pair per year as per Edwards (2003). We ran IMa2 using 22 STRs that fit the program criteria plus mtDNA data. We made several initial runs to assess whether the priors and heating conditions were appropriate for the data set. We experimented with a range of independent heated chains to ensure adequate MCMC mixing (Geyer 1991). We assessed plot trend lines to ensure adequate mixing of MCMC chains and viewed marginal distributions to determine appropriate parameter settings. In our chosen model parameters for IMa2, we set the upper limit of divergence to t = 2, population size to q = 1, and migration to m = 1. We employed a geometric model of Metropolis-coupling with 100 independent heated chains and chain heating values of 0.999 to 0.3. Ultimately, we performed 4 independent runs of 800,000 genealogies each (sampling every 100 steps) after an initial burn-in of 80,000 steps and with each run starting at a different seed value. Replicated analyses converged on the same posterior probabilities. We combined genealogies from all 4 runs for analysis for a total of 32,000 sampled genealogies. We achieved informative posterior probabilities for mean times of divergence among populations and we converted the divergence time parameter (t) into years by dividing by the geometric mean of the estimated STR and mtDNA mutation rates (3.42 x 10-5) and multiplying by the generation time.

Habitat Suitability Modeling

We tested hypotheses about distribution and habitat use between G. agassizii and G. morafkai in a region where they co-occur using habitat suitability models and contributing habitat variables for both species represented in a GIS at a 1 km cell resolution. This resolution adequately represents desert tortoise habitat at regional scales (Nussear et al. 2009), and was used to derive topographic (topographic position index [TPI], surface roughness [SRF], surface texture [ST] and solar insolation [SLR]), climatic (summer maximum temperature [TSMX], winter minimum temperature [TWMN], summer/winter temperature difference [TDIFF], winter precipitation [PCPwnt], summer/winter precipitation ratio [PCPrat] and probability of 3 year drought [PCPp3yd]) and vegetative (length of growing season [DUR], photosynthetic activity [AMP] and annual change in photosynthetic activity [AMPt]) variables for the contact zone study

area (see Inman et al. 2014 for methods for topographic and climatic variables). We derived vegetation metrics from Western CONUS 250 m eMODIS RSP data downloaded from the USGS EROS Center (http://phenology.cr.usgs.gov). We defined the study area by a 150 km buffer of the convex hull around the sampling locations of the 234 sampled tortoises. We assigned each animal’s sampling location topographic, climatic and vegetative values from the habitat variables. Prior to use, we transformed each variable by mean centering to unit variance. We also assigned individuals by genotype class (G. agassizii; G. morafkai, or admixed) based on the STRUCTURE analysis Q-value and used this to relate habitat use and niche separation to genotype.

Because differences in habitat selection between G. agassizii and G. morafkai are apparent across their combined ranges, we tested the hypothesis that each genotype occurs on the landscape in different distributions with respect to habitat variables using two-sample Kolmogorov-Smirnov tests. Hypotheses included complete niche separation (as represented by our habitat variables) for the G. agassizii and G. morafkai genotype classes to niche overlap between admixed and pure individuals. Separately, we tested for habitat selection differences among the different genotype classes by creating a map of two (‘Mojave’ and ‘Sonoran’) and three (‘Mojave’, ‘Sonoran’, and ‘transitional’) landscape designations with respect to the habitat variables using a k-medoids partitioning method (Reynolds et al. 1992) in R (R Core Team, 2014) with the package cluster (Maechler et al. 2014). We tested how well sorted the genotype classes are with respect to these landscape designations. If genotype classes follow habitat use patterns similar to their respective species, we hypothesized that the landscape designations derived from habitat variables would geographically separate the genotype classes. We used Cohen's Kappa scores (Cohen 1960) for the different genotype classes to test for significance.

Finally, we map the contact zone by treating the admixed genotype class as a separate species and using species distribution modeling (SDM; Franklin 2010) to create a probability distribution map of their potential range. We employed the SDM approach because admixed individuals showed overlap with non-admixed individuals in habitat use, and were not easily separated in geographic space because admixed individuals occurred proximal to non-admixed individuals. We derived estimates of admixed genotype suitability using Generalized Additive Models (GAM) with the package mgcv (version 1.7-6; Wood 2011) in R (version 2.13.1; R Core Development Team 2011) at a spatial resolution of 1 km, and considered the 13 habitat variables along with two additional variables: distance to ecotype edge (D2Edge), and distance to Colorado River (D2River). The first, D2Edge, was calculated as the euclidean distance from the interface between the two landscape designations, which were derived from the clustering analyses. We hypothesized that admixed individuals would most likely be located

proximal to the boundary between the two landscape designations. We calculated the second layer, D2River, as the euclidean distance from the Colorado River channel, and used this to represent the vicariance between G. agassizii and G. morafkai. We log- transformed the two distance metrics prior to analyses. Pseudo-absence data were taken as a large random selection (e.g. 10,000) of the study area to characterize areas where hybrids were not present. We evaluated model fit using the Area Under the receiver operating characteristic Curve (AUC; Fielding and Bell 1997; Cumming 2000) in the package ROCR (Sing et al. 2005) and used the UnBiased Risk Estimator (UBRE; Craven and Wahba 1979) derived from model training data as a measure of model fit. This metric is appropriate for GAM models fit with REML and is analogous to the commonly used AIC metric (Wood 2006).

Results

Multi-locus Genetic Clustering Analyses

The STRUCTURE analysis using all 234 tortoises (with and without Goag05) converged on K = 2 following Evanno et al. (2005). This differentiation clearly distinguishes the two species, G. agassizii and G. morafkai. However, examination of the bar plots show that "forcing" K = 3 reveals that the 3rd identifiable population is the Eastern Mojave RU and that tortoises in the Colorado Desert RU are more closely affiliated with the Arizona G. agassizii population (Figure 2). In the spatial analysis using GENELAND the optimal number of clusters was K = 4, analyzed both with and without mtDNA, based on likelihood estimates (Figure 3). Similar to the STRUCTURE analysis, the Eastern Mojave RU constitutes a genetically distinct population and the Colorado Desert RU and Arizona G. agassizii populations cluster together based on posterior probability estimates (Figure 3A, B). In addition, a unique cluster was identified that includes locations in the Hualapai and Buck mountains and this spatially defines the contact zone where hybridization occurs at highest frequencies (Figure 3C).

Hybrid Characterization

Genotype assignments using a combination of STRUCTURE, GENELAND, GeneClass II, mtDNA haplotype and NewHybrids show that among our samples in Arizona, purebred G. agassizii is restricted to the Black and Buck mountains with a single individual observed in the Hualapai Mountains (Table 3). Comparison of the proportion of G. agassizii genotypes to G. morafkai genotypes is significantly different between the adjacent Black and Hualapai mountains (Table 3; two-tailed Fisher's exact test, p = 0.001). We identified a total of 19 individuals of apparent mixed ancestry (Tables 3 and 4); however not all individuals were identified by all four of the methods (Table 4). Most were identified as F2 mixture components based on the NewHybrids analysis.

Hybridization does not appear to be biased toward one species or gender (e.g., mtDNA bias). While most hybrid individuals occur within three mountain ranges (Buck, Hualapai, Black), the STRUCTURE and NewHybrids analyses identified one apparent backcrossed individual outside the presumed contact zone in the Arrastra Mountains (Tables 3 and 4).

Descriptive Statistics

Descriptive statistics for the four a priori defined populations (with hybrid individuals removed from analyses) exhibit a remarkable similarity; suggesting shared patterns of demographic history, including population size (θ) and allelic richness (Table 2). Among the 345 pairwise comparisons of 26 total STR loci, significant linkage disequilibria were detected between 12 and 16 locus pairs in the Eastern Mojave and Colorado Desert RUs, respectively, and for 59 and 60 locus pairs in both the Arizona G. agassizii and G. morafkai populations (Table 2). FIS and deviations from HWE were not consistent across loci and at all sites within the hybrid zone (Table 2). Using program Bottleneck, all four populations exhibited an excess heterozygosity consistent with a recent reduction in effective population size with significant p-values (p<0.05) for all three tests for the TPM and SMM models but not the IAM model (except Arizona G. agassizii was not significant for the Wilcoxon Test, TPM). The AMOVA confirms the strong differentiation between the two species, G. agassizii and G. morafkai, as well as the greater similarity of the Colorado Desert RU and Arizona G. agassizii populations.

Population differentiation (pairwise FST) between the Arizona G. agassizii population and the Colorado Desert RU and Eastern Mojave RU were 0.013 and 0.051, respectively, and between the two RUs was 0.084. Pairwise FST among G. morafkai and the Arizona G. agassizii, Colorado Desert RU and Eastern Mojave RU populations was 0.218, 0.242, 0.222, respectively.

Coalescent Analyses

We generated mtDNA sequences for 195 samples and identified 7 haplotypes, all of which had been previously described (S1). In Arizona, G. agassizii haplotypes were predominant in the Black and Buck mountains and were not found outside of the putative contact zone. The reconstructed gene tree had strong support (Figure 4) and divergence time between G. agassizii, G. morafkai at this locus was consistent with previous studies; 5.75 mya (Figure 4: Avise et al. 1992; Lamb and Lydeard 1994; McLuckie et al. 1999; Edwards 2003). Both the maximum likelihood and parsimony reconstructions were consistent with the Bayesian results for all major clades.

The IMa2 analysis estimated the TMRCA between the Colorado Desert RU and Arizona G. agassizii populations to be 2,408 yrs. (95% CI; 732-5,120) and both of these

populations from the Eastern Mojave RU population at 1.38 mya (95% CI; 1.19–1.46). Parameter estimates of population size for the Colorado Desert RU and Arizona G. agassizii populations were similar (q = 0.426 ±0.18 and 0.490 ±0.20 respectively) and both smaller than the Eastern Mojave RU (q = 0.761 ±0.11). All existing populations were estimated to be smaller than ancestral populations (q3 = 0.976 ±0.02 and q4 = 0.919 ±0.08). Migration parameter estimates were similar between the Colorado Desert RU and Arizona G. agassizii population (0.426 and 0.490 respectively); however, the distributions of posterior densities for migration parameters did not obtain levels near either the upper or the lower limit of the prior and thus may not be reliable.

Habitat Suitability Modeling

Niche separation among the genotype classes was apparent for most of the habitat variables (Figure 5), with the most drastic differences occurring between G. agassizii and G. morafkai genotype classes. In general, the covariate distributions for admixed genotype classes overlapped with both the G. agassizii and G. morafkai genotypes, though their distributions were compressed (Figure 5). All habitat variables showed significant differences between the G. agassizii and G. morafkai genotypes (S2), with summer/winter temperature difference (TDIFF) and topographic position index (TPI) showing the highest Kolmogorov–Smirnov statistics. Admixed genotypes were less separated from the other genotypes (S2). When compared to non-admixed genotypes, admixed genotypes showed significantly different distributions in multiple habitat layers, with distance to Colorado River (D2River) showing the highest Kolmogorov–Smirnov statistic (S2).

The two landscape designations resulting from the k-medoids partitioning of the habitat variables accurately predicted the geographic location of the G. agassizii and G. morafkai genotypes, and misclassified only three individuals of the non-admixed genotypes, resulting in classification that was significantly better than random (K = 0.971, z = 14.3, p < 0.0001). The admixed genotypes were predicted less accurately with three landscape designations, with only 5 of the 17 admixed genotypes identified correctly. Thus, the model with two landscape designations better explains the data than a model with three landscape designations, where admixed individuals were treated as an independent category. The probability distribution map generated from our habitat suitability modeling accurately predicted the locations where we currently observe G. agassizii in Arizona as well as the transitional sites where we observe hybridization (Figure 6). The model also predicted areas in Arizona outside of our sampling area that may be suitable for G. agassizii (Figure 6). Model performance was generally high, with an AUC score 0.945, far above that of random prediction (Fielding and Bell 1997).

Discussion

Biogeography

Our results are largely consistent with the Colorado River having acted as a geographic barrier between G. agassizii and G. morafkai, resulting in the deep divergence between these two species (Figure 4). Since the actual mutation rates of these loci cannot easily be estimated for these taxa, we rely primarily on the well-established divergence of G. agassizii and G. morafkai lineage by the vicariant influence of the Bouse inundation which now forms the Colorado River boundary between the species (Avise et al. 1992). However, mtDNA mutation rates based on this geological event are inconsistent with fossil records of divergence among other congeners and it is expected that the molecular estimates of the time of divergence within this genus are too recent (Avise et al. 1992; Bramble and Hutchinson 2014). Until a recalibration of the existing molecular clock is performed using distantly related groups, we treat our projected evolutionary rates as conservative estimates (Figure 4).

The presence of G. agassizii east of the Colorado River appears to be relatively recent and has resulted in a zone of secondary contact with its congener, G. morafkai. In our analysis, G. agassizii and G. morafkai were separated on the landscape with respect to all habitat variables, though the admixed genotype class exhibited less separation, with topographic position index (TPI) having the highest explanatory value in the separation between the two species (S2, Figure 5A). TPI classifies both slope position (i.e., ridge top, hillside, flats etc.) and landform category (i.e., bajada, wash, foothills, etc.). This variable is substantially different between the distributions of G. agassizii and G. morafkai, with G. agassizii generally occurring in areas with higher values (corresponding to alluvial fans and valley bottoms), whereas G. morafkai tend to occur in areas with lower values (corresponding to foot-hills, hillside slopes and more mountainous terrain) (Nussear and Tuberville 2014). The genetic samples reported here are subsamples of a larger geographic area that has yet to be genotyped, yet does not present apparent geographic barriers to tortoise dispersal. In fact, tortoise presence and habitat suitability modeling provide additional areas where the genotype could be either G. agassizii or hybrids of G. agassizii and G. morafkai that are east and south of the Colorado River (Figure 6). Specifically, habitat for G. agassizii is predicted to extend further north of the Black Mountains (Figure 6), although there are not many affirmative tortoise localities across Detrital Valley and toward the north end of the White Hills, Arizona (CA Jones, pers obs.).

Our IMa2 data analysis suggest a recent shared ancestry (~2,400 years) between G. agassizii populations directly across the Colorado River. The Arizona population of G. agassizii in the Black Mountains does not constitute a genetically distinct population unto 37

itself but closely resembles the cross-river, Colorado Desert RU population. The IMa2 program makes assumptions on a large number of parameters which our data obviously violated, for example, modeling constant population sizes. This is not a realistic assumption for G. agassizii and G. morafkai which likely experienced reductions of population size and therefore genetic diversity during the late Pleistocene (Edwards 2003). Despite this, we used IMa2 for modeling general patterns and thus these simplifications were necessary. We interpret our results in light of these biases. The relative proportion of shared mtDNA haplotypes and the absence of novel mtDNA haplotypes in the Arizona G. agassizii population also indicate that this small population has not been isolated with enough time for drift or mutation to drive divergence (Avise et al. 1987).

We observed much greater genetic differentiation between the Eastern Mojave RU and the Colorado Desert RU in California than between the Arizona G. agassizii populations isolated across the Colorado River. Our results are consistent with Hagerty et al. (2010) in that the Providence and New York mountain ranges are a strong barrier to gene flow between the Ivanpah Valley and the Chemehuevi Valley (Figure 1). This genetic differentiation corresponds to the divergence between the northern Mojave “MOJ_B’ maternal haplogroup and the California ‘MOJ_A’ haplogroup which exhibit a deep split between these regions (3.12 mya; Figure 4). The distribution of mtDNA haplogroups despite gene flow with isolation by distance (Murphy et al. 2007; Hagerty and Tracy 2010) suggest a unique population history of northern G. agassizii clades. Northern clades potentially diverged during extended periods of isolation from southern populations, however, the current genetic structure may also be maintained by local adaptation.

We considered potential ways that G. agassizii might have moved from west to east across the Colorado River: 1) human transport by aboriginal peoples, 2) rafting or floating across the Colorado River, and 3) geological events and movement of the Colorado River channel. The consistency of genetic diversity indices between the Black Mountain population and its adjacent populations in the Colorado RU across the river (Table 2) suggest that the hybrid zone was founded by a significant number of individuals. The crossing or isolation of G. agassizii on the east side of the river need not have been a single event and may be a result of episodic gene flow. Turtles are susceptible to human-mitigated movements (González-Porter et al. 2011) and desert tortoises in particular are known to have been moved by early humans (Schneider and Everson 1989). For example, the Chemehuevi, a once nomadic people who have occupied the Colorado River basin for thousands of years, were known to use tortoises for food (Schneider and Everson 1989) and treated the tortoises as having an aura of sacredness (Laird 1976), and consequently might have moved tortoises across the river.

We think that geological events, such as periodic cycles of aggradation, degradation and avulsion in the Lower Colorado River during the Holocene and late Pleistocene, may provide a more likely explanation of the occurrence of G. agassizii east of the Colorado River and more closely fall within the estimated timeframe for contact for the two tortoise species. During the last 10,000 years, the Lower Colorado River and surrounding areas in the vicinity of the Mojave River Valley have undergone cycles of aggradation and filling of paleovalleys, followed by episodes of degradation, and in some cases stranding of incised river valleys (Howard et al. 2008, 2011). Based on radiocarbon dating of river sediments, this part of the Colorado River was in a narrower and deeper valley 8500 years ago and the valley aggraded and widened between 8500 and 6000 years ago (Howard et al. 2011; personal communication). These regional, episodic increases in sediment supply resulted in valley-floor aggradation and incision of the Colorado River, consistent with our timing of G. agassizii cross-river shared ancestry. During one of these cycles, habitat with G. agassizii may have become stranded as the river narrowed and changed course.

In addition to tortoises, several other species of reptiles, less likely to have been influenced by human transport, exhibit patterns of species differentiation and isolation from one side of the River to another, e.g., horned lizards (Mulcahy et al. 2006), rosy boas (Wood et al. 2008), and night snakes (Mulcahy 2008). Other species show cross- river patterns of shared species, haplotypes or clades; desert iguana and chuckwalla (Lamb et al. 1992), western diamondback rattlesnake (Castoe et al. 2007), spiny lizards (Leaché and Mulcahy 2007), fringe-toed sand lizards (Gottscho et al. 2014), and western shovel-nosed snakes (Wood et al. 2014). However, the estimated timing of cross-river dispersal for most of the above species as well as several others (Wood et al. 2013) is greater than the ~2,400 year time span we discuss here for the presence of G. agassizii in the Black Mountains.

Hybridization

Many turtle species are known to hybridize (Lutterschmidt et al. 2007; Vilaca et al. 2012; Parham et al. 2013) and we documented that hybridization naturally occurs between G. agassizii and G. morafkai at a secondary contact zone in northwestern Arizona. However, we did not always have continuity of results across the different methods we used to detect hybrids (Table 4). Bohling et al. (2013) observed that STRUCTURE has the potential to misclassify a pure individual as a hybrid, and Shurtliff et al. (2014) suggests that NEWHYBRIDS is much more accurate than STRUCTURE in defining F1s, but performs poorly in identifying F2s. Nevertheless by using multiple methods we believe we have captured the full breadth of hybridization as is detectable in this dataset. One reason that different methods may not detect hybrids equally is that not all hybrid individuals exhibit the same proportion of admixture. We observed a wide variety of 39

recombinant classes including a high proportion of apparent F2s and backcrosses (Table 4). This is consistent with the idea that F1 individuals may be hard to form under natural conditions and therefore F2 and backcrossed individuals should be in greater proportion than F1 individuals in a mixed population of two taxa (Arnold 1992; Reed and Sites 1995; Arnold 1997). We also observed a disproportionate distribution of hybrid classes, with G. agassizii backcrosses dominant in the Black Mountains and G. morafkai backcrosses primarily distributed in the Hualapai Mountains (Table 4). Bimodal distributions of hybrid individuals associated with some habitat variables may reflect this differentiation in hybrid classes (Figure 5).

Within the area we predefined as the contact zone, we observed a relatively small number of individuals (n = 19) of hybrid origin. There appears to be limited distance of penetration across the contact zone from either parental or hybrid genotype class and the distribution of parentals and hybrid classes is suggestive of exogenous selection and/or assortative mating. Hybrid individuals occur at highest frequency in transitional habitat (Figures 5 and 6), but were not necessarily “bounded” there (Moore 1977). That the pattern of divergence is maintained along transitional habitat is highly suggestive of ecological segregation (Tarroso et al. 2014) and we might assume that interactions between genotype and environment determine the genetic structure of the hybrid population – e.g., selection gradient due to environmental heterogeneity (Endler 1977). While highly discriminating between the G. agassizii and G. morafkai, our topographic, climate, and vegetation variables could not differentiate as easily admixed individuals from non-admixed individuals (S2). Instead, distance to the Colorado River (D2River; S2) provided strong distinction between them, and was included in the model used to predict hybrid zones. This variable may be interpreted as the likelihood that G. agassizii and G. morafkai might come into contact with each other. The results of the k-mediods partitioning support that hybrid individuals do not occupy a unique or novel environment exclusive of parental genotypes. The geographic distribution of hybrid individuals, and thus the hybrid zone, is likely a result of proximity to phylogenetic recontact zones, and not a result of unique habitat selection by hybrids. We therefore propose a geographical selection-gradient model (GSGM) consistent with Endler (1977) to explain the distribution of genotypes at this secondary contact zone.

Theoretically a hybrid zone can remain stable through a balance of dispersal and selection against hybrid genotypes (Barton and Hewitt 1985). The role of effective population size (Ne) is important in interpreting these findings, as the predicted habitat appropriate for G. agassizii is limited east of the Colorado River in Arizona (Figure 6).

Also, because the population is isolated from the rest of its geographic range, Ne is expected to be small. With low Ne, genetic drift reduces the effectiveness of selection and populations with low Ne may exhibit greater introgression (Nichols and Hewitt 1986).

Hybrid individuals can escape selection where they are locally abundant despite having lower fitness, when introgression is constant from outside the zone (Nichols 1989). Although desert tortoises exhibit strong site tenacity, dispersal ability of a tortoise may exceed the ecotone width (>30 km; Edwards et al. 2004).

The presumed allopatric divergence between these two species suggests that opportunities for the evolution of reproductive isolating mechanisms to prevent hybridization are lacking or are minimal. Since secondary contact has been relatively recent, there may not have been enough time for this to occur. However, drivers of reproductive isolation such as Haldane’s rule (Haldane 1922) do not apply to tortoises since they do not possess heterogametic sex chromosomes and instead have temperature dependent sex determination. In contrast, mixed genotypes may have functionality in an ecotone, and complete isolation may not always be beneficial for sister taxa (Nei and Nozawa 2011).

Conservation implications

A distinct population segment (DPS) of G. agassizii was federally listed in 1990 as threatened under the U.S. Endangered Species Act based on the conservation status of desert tortoises in this part of their distribution (ESA; USFWS 1990). The DPS was defined as tortoises occurring west and north of the Colorado River (Figure 1; USFWS 1990). The isolated, Arizona population of G. agassizii in the Black Mountains and surrounding area is not currently afforded protection under the ESA like its kin across the river because the listed population was geographically delineated, however it is protected by the Arizona Game and Fish Department (AGFD 2012). Increasing development in this region of Arizona may threaten the viability of this small population of G. agassizii, in particular, the proposed realignment of state highways east of Bullhead City would pass directly through primary habitat of this population (ADOT 2014). G. morafkai is not federally listed in the USA, but became a candidate for federal listing in 2010 and is considered a Wildlife of Special Concern in Arizona ( USFWS 2010, 2012; AGFD 2012).

The identity of species is important for legally determining the lineage of a specimen, as in the case of poaching or designation of conservation easements on private lands (Haig et al. 2004). Unfortunately, G. agassizii and G. morafkai are challenging to distinguish in the field (Murphy et al. 2011). Where species identification is not easy to determine, geographical delineation may be the best indicator for management (Barrowclough et al. 2011). The existence of natural hybridization between G. agassizii and G. morafkai further complicate this issue. In the context of species conservation, it is not possible for us to determine which individuals contribute most to the evolutionary potential of the species, or more importantly, which adaptive traits will be most critical in the face of 41

environmental change. For a species to persist, it requires genetic diversity to cope with changes in its environment. With unpredictable stochastic processes, such as climate change, the individuals that have the ‘best’ adaptations may very well be the ones living on edges and in marginal habitats (Eckert et al. 2008; Hardie and Hutchings 2010; Palstra and Ruzzante 2008; Shafer et al. 2011), such as the admixed individuals we observed in this study. Thus, the prudent approach to species conservation is to preserve the entirety of genetic diversity in a species including viable hybrids or populations where the species may benefit from limited introgression. Knowing that the evolutionary potential of a species is directly related to its genetic diversity, we would do best to include the full extent of genetic variation and its maintenance, including the potential for natural hybridization, in conservation efforts.

Acknowledgements

We thank S. Goodman, D. Grandmaison and P. Woodman for assistance with sample collection in Arizona. K. Anderson, T. Bailey, R. Evans, R. Woodard, P. Woodman, P. Frank, B. Henen, and K. Nagy assisted by collecting samples in California. We are grateful for the financial support the US Geological Survey. Additional support came from Arizona Research Laboratories. Tissue samples were collected in Arizona under Scientific Collecting Permits to C.A. Jones (2002–2006) and P. Woodman (2005–2010) and in California under federal and state permits to K.H. Berry (1994–2005). Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Data Archiving

DNA sequences used in this study are available from Genbank: mtDNA: DQ649394.1, DQ649397.1, DQ649398.1, DQ649399.1, DQ649400.1, DQ649401.1, DQ649405.1, DQ649409.1; STRs: AF546890, AF546895,AF546891, AF546892, AF546889, AF546893, AF546896, AF546894, AF546888, AY317141, AY317142, AY317143, AY317144, AY317145, AY317147, FJ743597, FJ743598, FJ765736, FJ765733, FJ765734, FJ765731, FJ765732, FJ765727, FJ765735, FJ765730, FJ765726, FJ765737, HQ625078, HQ625081, HQ625080, HQ625079

Literature Cited

Anderson, E. C. 2008. Bayesian inference of species hybrids using multilocus dominant genetic markers. Philos. Trans. R. Soc. Lond. B Biol. Sci. 363:2841–2850. Arizona Department of Transportation. 2014. State Route 95: Interstate 40 to State Route 68 Tier 1 Environmental Impact Statement Study. http://www.azdot.gov/projects/far-west/sr-95-i-40-to-sr-68-tier-1-eis- study/overview. Accessed 20 October 2014. Arizona Game and Fish Department. 2012. Arizona’s State Wildlife Action Plan: 2012- 2022. Arizona Game and Fish Department, Phoenix, Arizona. 245 p. Arnold, M. L. 1992. Natural hybridization as an evolutionary process. Annu. Rev. Ecol. Syst. 23:237-261. Arnold, M. L. 1997. Natural hybridization and evolution. Oxford University Press, New York. Averill-Murray. R. C. 2000. Survey protocol for Sonoran Desert tortoise monitoring plots: reviewed and revised. Arizona Interagency Desert Tortoise Team and the Arizona Game and Fish Department, Phoenix, Arizona. 41pp. Averill-Murray, R. C., B. E. Martin, S. J. Bailey, and E. B. Wirt. 2002. Activity and behavior of the Sonoran Desert Tortoise in Arizona. Pp. 135–158 in T. R. Van Devender, ed. The Sonoran Desert Tortoise; Natural History, Biology and Conservation. University of Arizona Press, Tucson, Arizona. Avise, J. C., J. Arnold, R. M. Ball, E. Bermingham, T. Lamb, J. E. Neigel, C. A. Reeb, and N. C. Saunders. 1987. Intraspecific phylogeography - the mitochondrial-DNA bridge between population-genetics and systematics. Annu. Rev. Ecol. Syst. 18:489–522. Avise, J. C., B. W. Bowen, T. Lamb, A. B. Meylan, and E. Bermingham. 1992. Mitochondrial-DNA evolution at a turtles pace - evidence for low genetic- variability and reduced microevolutionary rate in the Testudines. Mol. Biol. Evol. 9:457–473. Barrowclough, G. F., R. J. Gutiérrez, J. G. Groth, J. E. Lai, and D. F. Rock. 2011. The Hybrid Zone Between Northern and California Spotted Owls in the Cascade– Sierran Suture Zone. Condor 113:581–589. Barton, N. H. and G. M. Hewitt. 1985. Analysis of hybrid zones. Annu. Rev. Ecol. Syst. 16:113–148. Baudouin, L. and P. Lebrun. 2001. An operational Bayesian approach for the identification of sexually reproduced cross-fertilized populations using molecular markers. Proceedings of the International Symposium on Molecular Markers for Characterizing Genotypes and Identifying Cultivars in Horticulture, 81–93. Berry, K. H. and M. M. Christopher. 2001. Guidelines for the field evaluation of desert tortoise health and disease. J. Wildl. Dis. 37:427–450. 43

Berry, K. H., D. J. Morafka, and R. W. Murphy. 2002. Defining the desert tortoise(s): our first priority for a coherent conservation strategy. Chelonian Conserv. Biol. 4:249–262. Bohling, J. H., J. R. Adams, and L. P. Waits. 2013. Evaluating the ability of Bayesian clustering methods to detect hybridization and introgression using an empirical red wolf data set. Mol. Ecol. 22:74–86. Bramble, D.M. and J.H. Hutchison. 2014. Morphology, Taxonomy, and Distribution of North American Tortoises; An Evolutionary Perspective. Pp. 1–12 in D. C. Rostal, E. D. McCoy and H.R. Mushinsky, eds. Biology and Conservation of North American Tortoises. Johns Hopkins University Press, Baltimore, MD Castoe, T. A., C. L. Spencer, and C. L. Parkinson. 2007. Phylogeographic structure and historical demography of the western diamondback rattlesnake (Crotalus atrox): A perspective on North American desert biogeography. Mol. Phylogenet. Evol. 42:193–202. Cohen, J. 1960. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20:37–46. Craven, P. and G. Wahba. 1979. Smoothing noisy data with spline functions. Numer. Math. 31:377–403. Cumming, G. S. 2000. Using between-model comparisons to fine-tune linear models of species ranges. J. Biogeogr. 27:441–455. Davy, C. M., T. Edwards, A. Lathrop, M. Bratton, M. Hagan, B. Henen, K. A. Nagy, J. Stone, L. S. Hillard, and R. W. Murphy. 2011. Polyandry and multiple paternities in the threatened Agassiz’s desert tortoise, Gopherus agassizii. Conserv. Genet. 12:1313–1322. Drummond, A. J. and A. Rambaut. 2007. BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol. Biol. 7:214 Earl, D. A. and B. M. Vonholdt. 2012. STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conserv. Genet. Resour. 4:359–361. Eckert, C. G., K. E. Samis, and S. C. Lougheed. 2008. Genetic variation across species' geographical ranges: the central-marginal hypothesis and beyond. Mol. Ecol. 17:1170–1188. Edwards, T. 2003. Desert tortoise conservation genetics, University of Arizona. Unpublished M.S. Thesis, University of Arizona, Tucson, USA. 86 pp. Edwards, T., C. S. Goldberg, M. E. Kaplan, C. R. Schwalbe, and D. E. Swann. 2003. PCR primers for microsatellite loci in the desert tortoise (Gopherus agassizii, Testudinidae). Mol. Ecol. Notes 3:589–591. Edwards, T., C. J. Jarchow, C. A. Jones, and K. E. Bonine. 2010. Tracing Genetic Lineages of Captive Desert Tortoises in Arizona. J. Wildl. Manag. 74:801–807.

Edwards, T., A. Lathrop, A. Ngo, K. Choffe, and R. W. Murphy. 2011. STR/microsatellite primers for the desert tortoise, Gopherus agassizii, and its congeners. Conserv. Genet. Resour. 3:365–368. Edwards, T., C. R. Schwalbe, D. E. Swann, and C. S. Goldberg. 2004. Implications of anthropogenic landscape change on inter-population movements of the desert tortoise (Gopherus agassizii). Conserv. Genet. 5:485–499. Endler, J. A. 1977. Geographic Variation, Speciation, and Clines. Princeton University Press, Princeton, New Jersey. Engstrom, T. N., T. Edwards, M. F. Osentoski, and E. M. Myers. 2007. A compendium of PCR Primers for mtDNA, Microsatellite, and Other Nuclear Loci for Freshwater Turtles and Tortoises. Pp. 124–141 in H. B. Shafer, N. N. FitzSimmons, A. Georges, and A. G. H. Rhodin, eds. Defining Turtle Diversity: Proceedings of a workshop on Genetics, Ethics, and Taxonomy of Freshwater Turtles and Tortoises. Chelonian Res. Monogr. 4. Evanno, G., S. Regnaut, and J. Goudet. 2005. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol. Ecol. 14:2611–2620. Excoffier, L., G. Laval, and S. Schneider. 2005. Arlequin (version 3.0): An integrated software package for population genetics data analysis. Evol. Bioinform. 1:47–50. Fielding, A., and J. Bell. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environ. Conserv. 24:38–49. Fitzsimmons, N. N., C. Moritz, and S. S. Moore. 1995. Conservation and dynamics of microsatellite loci over 300-million years of marine turtle evolution. Mol. Biol. Evol. 12:432–440. Franklin, J. 2010. Mapping species distributions: spatial inference and prediction. Cambridge University Press, Cambridge, UK. Geyer, C. J. 1991. Markov-Chain Monte-Carlo Maximum-Likelihood. Comp. Sci. Stat. 156–163. Glenn, J. L., R. C. Straight, and J. W. Sites. 1990. A plasma-protein marker for population genetic-studies of the desert tortoise (Xerobates-agassizi). Great Basin Nat. 50:1–8. González-Porter, G. P., F. Hailer, O. Flores-Villela, R. García-Anleu, and J. E. Maldonado. 2011. Patterns of genetic diversity in the critically endangered Central American river turtle: human influence since the Mayan age? Conserv. Genet. 12:1229–1242. Goudet, J. 1995. FSTAT (Version 1.2): A computer program to calculate F-statistics. J. Hered. 86:485–486. Gottscho, A. D., S. B. Marks, and W. B. Jennings. 2014. Speciation, population structure, and demographic history of the Mojave fringe-toed lizard (Uma scoparia), a species of conservation cocern. Ecology and Evolution 4:2546–2562. 45

Grandmaison, D. D., M. F. Ingraldi, and F. R. Peck. 2010. Desert tortoise microhabitat selection on the Florence Military Reservation, south-central Arizona. J. Herpetol. 44:581–590. Guedj, B. and G. Guillot. 2011. Estimating the location and shape of hybrid zones. Mol. Ecol. Resour. 11:1119–1123. Guillot, G., F. Mortier, and A. Estoup. 2005. GENELAND: a computer package for landscape genetics. Mol. Ecol. Notes 5:712–715. Guo, S. W. and E. A. Thompson. 1992. Performing the exact test of Hardy-Weinberg proportion for multiple alleles. Biometrics 48:361–372. Hagerty, B. E., K. E. Nussear, T. C. Esque, and C. R. Tracy. 2010. Making molehills out of mountains: landscape genetics of the Mojave desert tortoise. Landsc. Ecol. 26:267–280. Hagerty, B. E. and C. R. Tracy. 2010. Defining population structure for the Mojave desert tortoise. Conserv. Genet. 11:1795–1807. Haig, S. M., T. D. Mullins, E. D. Forsman, P. W. Trail, and L. Wennerberg. 2004. Genetic identification of Spotted Owls, Barred Owls, and their hybrids: Legal implications of hybrid identity. Conserv. Biol. 18:1347–1357. Haldane, J. B. S. 1922. Sex ratio and unisexual sterility in hybrid animals. Journal of Genetics 12:101–109. Hardie, D. C. and J. A. Hutchings. 2010. Evolutionary ecology at the extremes of species' ranges. Environ. Rev. 18:1–20. Harrison, R. G., ed. 1993. Hybrid zones and the evolutionary process. Oxford University Press, New York, USA. Hauswaldt, J. S. and T. C. Glenn. 2005. Population genetics of the diamondback terrapin (Malaclemys terrapin). Mol. Ecol. 14:723–732. Hewitt, G. M. 1988. Hybrid zones - Natural laboratories for evolutionary studies. Trends Ecol. Evol. 3:158–167. Hey, J. 2010. Isolation with Migration Models for More Than Two Populations. Mol. Biol. Evol. 27:905–920. House, P. K., P. A. Pearthree, K. A. Howard, J. W. Bell, M. E. Perkins, J. E. Faulds, and A. L. Brock. 2005. Birth of the lower Colorado River — stratigraphic and geomorphic evidence for its inception near the conjunction of Nevada, Arizona, and California. Pp. 357–387 in J. Pederson, and C. M. Dehler, eds. Interior Western United States, Field Guide 6. Geological Society of America, USA. Howard, K. A., S. C. Lundstrom, D. V. Malmon, and S. J. Hook. 2008. Age, distribution, and formation of late Cenozoic paleovalleys of the lower Colorado River and their relation to river aggradation and degradation. Pp. 391–410 in Reheis, M. C., R. Hershler and D. M. Miller, eds. Late Cenozoic Drainage History of the Southwestern Great Basin and Lower Colorado River Region: Geologic and Biotic Perspectives. Geological Society of America Special Paper 439. 46

Howard, K. A., D. V. Malmon, J. P. McGeehin, and P. Martin. 2011. Holocene aggradation of the lower Colorado River in Mohave Valley, Arizona. P p. 151– 152 in L. S. Beard, K. E. Karlstrom, R. A. Young and G. H. Billingsley, eds. Colorado River Evolution 2—Origin and Evolution of the Colorado River System, Workshop Abstracts: U.S. Geological Survey Open File Report 2011- 1210, (extended abstract), http://pubs.usgs.gov/of/2011/1210/ Inman, R. D., K. E. Nussear, T. C. Esque, A. G. Vandergast, S. A. Hathaway, D. A. Wood, K. R. Barr, and R. N. Fisher. 2014. Mapping habitat for multiple species in the Desert Southwest. U.S. Geological Survey Open-File Report 2014-1134, 92 p. Laird, C. 1976. The Chemehuevis. Malki Museum Press, California, USA. Lamb, T., J. C. Avise, and J. W. Gibbons. 1989. Phylogeographic patterns in mitochondrial-DNA of the desert tortoise (Xerobates-agassizi), and evolutionary relationships among the North-American gopher tortoises. Evol. 43:76–87. Lamb, T., T. R. Jones, and J. C. Avise. 1992. Phylogeographic histories of representative herpetofauna of the southwestern United States - mitochondrial-DNA variation in the desert iguana (Dipsosaurus-dorsalis) and the chuckwalla (Sauromalus- obesus). J. Evol. Biol. 5:465–480. Lamb, T. and C. Lydeard. 1994. A molecular phylogeny of the gopher tortoises, with comments on familial relationships within the Testudinoidea. Mol. Phylogenet. Evol. 3:283–291. Leache, A. D. and D. G. Mulcahy. 2007. Phylogeny, divergence times and species limits of spiny lizards (Sceloporus magister species group) in western North American deserts and Baja California. Mol. Ecol. 16:5216–5233. Lutterschmidt, W. I., S. A. Escobar, and E. D. Wilson. 2007. Multivariate analyses of shell morphology of putative hybrid box turtles. Southeast. Nat. 6:571–576. Maechler, M., P. Rousseeuw, A. Struyf, M. Hubert, and K. Hornik. 2014. cluster: Cluster Analysis Basics and Extensions. R package version 1.15.2. McLuckie, A. M., T. Lamb, C. R. Schwalbe, and R. D. McCord. 1999. Genetic and morphometric assessment of an unusual tortoise (Gopherus agassizii) population in the Black Mountains of Arizona. J. Herpetol. 33:36–44. Moore, W.S. 1977. Evaluation of narrow hybrid zones in vertebrates. Q. Rev. Bio. 52:263–277. Mulcahy, D. G. 2008. Phylogeography and species boundaries of the western North American Nightsnake (Hypsiglena torquata): revisiting the subspecies concept. Mol. Phylogenet. Evol. 46:1095–1115. Mulcahy, D. G., A. W. Spaulding, J. R. Mendelson, 3rd, and E. D. Brodie, Jr. 2006. Phylogeography of the flat-tailed horned lizard (Phrynosoma mcallii) and systematics of the P. mcallii-platyrhinos mtDNA complex. Mol. Ecol. 15:1807– 1826.

Murphy, R. W., K. H. Berry, T. Edwards, A. E. Leviton, A. Lathrop, and J. D. Riedle. 2011. The dazed and confused identity of Agassiz's land tortoise, Gopherus agassizii (Testudines, Testudinidae) with the description of a new species, and its consequences for conservation. ZooKeys 39–71. Murphy, R. W., K. H. Berry, T. Edwards, and A. M. McLuckie. 2007. A genetic assessment of the recovery units for the Mojave population of the desert tortoise, Gopherus agassizii. Chelonian Conserv. Biol. 6:229–251. Nei, M. and M. Nozawa. 2011. Roles of mutation and selection in speciation: from Hugo de Vries to the modern genomic era. Genome Biol. Evol. 3:812–829. Nichols, R. A. 1989. The fragmentation of tension zones in sparsely populated areas. Am. Nat. 134:969–977. Nichols, R. A. and G. M. Hewitt. 1986. Population-structure and the shape of a chromosomal cline between 2 races of podisma-pedestris (orthoptera, acrididae). Biol. J. Linn. Soc. 29:301–316. Nielsen, R. and J. Wakeley. 2001. Distinguishing migration from isolation: A Markov chain Monte Carlo approach. Genetics 158:885–896. Nussear, K. E., T. C. Esque, R. D. Inman, L. Gass, K. A. Thomas, C. S. A. Wallace, J. B. Blainey, D. M. Miller, and R. H. Webb. 2009. Modeling habitat of the desert tortoise (Gopherus agassizii) in the Mojave and parts of the Sonoran Deserts of California, Nevada, Utah, and Arizona: U.S. Geological Survey Open-File Report 2009–1102, 18 p. Nussear, K. E. and T. D. Tuberville. 2014. Habitat characteristics of North American tortoises. Pp. 77–84 in D. C. Rostal, E. D. McCoy, and H. R. Mushinsky, eds. Biology and Conservation of North American Tortoises. Johns Hopkins University Press, Baltimore, MD. Nylander, J.A.A. 2004. MrModeltest v2. Program distributed by the author. Evolutionary Biology Centre, Uppsala University. Ohta, T. and M. Kimura. 1973. Model of mutation appropriate to estimate number of electrophoretically detectable alleles in a finite population. Genetical Research 22:201–204. Paetkau, D., W. Calvert, I. Stirling, and C. Strobeck. 1995. Microsatellite analysis of population-structure in Canadian polar bears. Mol. Ecol. 4:347–354. Paetkau, D., R. Slade, M. Burden, and A. Estoup. 2004. Genetic assignment methods for the direct, real-time estimation of migration rate: a simulation-based exploration of accuracy and power. Mol. Ecol. 13:55–65. Palstra, F. P. and D. E. Ruzzante. 2008. Genetic estimates of contemporary effective population size: what can they tell us about the importance of genetic stochasticity for wild population persistence? Mol. Ecol. 17:3428–3447. Parham, J. F., T. J. Papenfuss, P. P. Dijk, B. S. Wilson, C. Marte, L. R. Schettino, and W. Brian Simison. 2013. Genetic introgression and hybridization in Antillean 48

freshwater turtles (Trachemys) revealed by coalescent analyses of mitochondrial and cloned nuclear markers. Mol. Phylogenet. Evol. 67:176–187. Piry, S., A. Alapetite, J. M. Cornuet, D. Paetkau, L. Baudouin, and A. Estoup. 2004. GENECLASS2: A software for genetic assignment and first-generation migrant detection. J. Hered. 95:536–539. Piry, S., G. Luikart, and J. M. Cornuet. 1999. BOTTLENECK: A computer program for detecting recent reductions in the effective population size using allele frequency data. J. Hered. 90:502–503. Pritchard, J. K., M. Stephens, and P. Donnelly. 2000. Inference of population structure using multilocus genotype data. Genetics 155:945–959. R Core Development Team. 2011. R-A language and environment for statistical computing: Vienna, Austria, R Foundation for Statistical Computing. Vienna, Austria. http://www.R-project.org/. R Core Team. 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/. Rambaut, A. 2006-2012. FigTree; Tree Figure Drawing Tool. Institute of Evolutionary Biology, University of Edinburgh, http://tree.bio.ed.ac.uk/. Rambaut, A. and A. J. Drummond. 2002-2013. TreeAnnotator v1.7.5; MCMC Output analysis. Institute of Evolutionary Biology, University of Edinburgh, http://beast.bio.ed.ac.uk/. Rambaut, A., M. A. Suchard, W. Xie, and A. J. Drummon. 2003-2013. Tracer: MCMC Trace Analysis Tool. Institute of Evolutionary Biology, University of Edinburgh, http://beast.bio.ed.ac.uk/. Rannala, B. and J. L. Mountain. 1997. Detecting immigration by using multilocus genotypes. Proc. Natl. Acad. Sci. U. S. A. 94:9197–9201. Reed, K. M. and J. W. Sites. 1995. Female fecundity in a hybrid zone between 2 chromosome races of the Sceloporus-Grammicus complex (Sauria, Phrynosomatidae). Evol. 49:61–69. Reynolds, A., G. Richards, B. de la Iglesia, and V. Rayward-Smith. 1992. Clustering rules: A comparison of partitioning and hierarchical clustering algorithms. J. Math Model. Algor. 5:475–504. Riedle, J. D., R. C. Averill-Murray, C. L. Lutz, and D. K. Bolen. 2008. Habitat use by desert tortoises (Gopherus agassizii) on alluvial fans in the Sonoran Desert, south- central Arizona. Copeia 2008:414–420. Schlotterer, C. 2000. Evolutionary dynamics of microsatellite DNA. Chromosoma 109:365–371. Schneider, J.S. and G. D. Everson. 1989. The desert tortoise (Xerobates agassizii) in the prehistory of the southwestern Great Basin and adjacent areas. Journal of California and Great Basin Anthropology 11:175–202.

Schwartz, T. S., M. Osentoski, T. Lamb, and S. A. Karl. 2003. Microsatellite loci for the North American tortoises (genus Gopherus) and their applicability to other turtle species. Mol. Ecol. Notes 3:283–286. Shafer, A. B. A., S. D. Cote, and D. W. Coltman. 2011. Hot spots of genetic diversity descended from multiple Pleistocene refugia in an alpine ungulate. Evol. 65:125– 138. Shurtliff, Q. R., P. J. Murphy, and M. D. Matocq. 2014. Ecological segregation in a small mammal hybrid zone: habitat-specific mating opportunities and selection against hybrids restrict gene flow on a fine spatial scale. Evol. 68:729–742. Sing, T., O. Sander, N. Beerenwinkel, and T. Lengauer. 2005. ROCR: visualizing classifier performance in R. Bioinformatics 21:3940–3941. Slatkin, M. and L. Excoffier. 1996. Testing for linkage disequilibrium in genotypic data using the expectation-maximization algorithm. Heredity 76:377–383. Smith, K. L., J. M. Hale, M. R. Kearney, J. J. Austin, and J. Melville. 2013. Molecular patterns of introgression in a classic hybrid zone between the Australian tree frogs, Litoria ewingii and L. paraewingi: evidence of a tension zone. Mol. Ecol. 22:1869–1883. Swofford, D.L. 2002. PAUP*. Phylogeneitc Analysis Using Parsimony (*and Other Methods). Version 4. Sinauer Associates, Sunderland, Massachusetts. Tarroso, P., R. J. Pereira, F. Martinez-Freiria, R. Godinho, and J. C. Brito. 2014. Hybridization at an ecotone: ecological and genetic barriers between three Iberian vipers. Mol. Ecol. 23:1108–1123. Turner, R. M. and D. E. Brown. 1982. Sonoran desertscrub. Pp. 181–221 in D. E. Brown, ed. Biotic communities of the American Southwest--United States and Mexico. University of Arizona Press and the Boyce Thompson Southwestern Arboretum. U.S. Fish and Wildlife Service. 1990. Endangered and threatened wildlife and plants: Determination of threatened status for the Mojave population of the desert tortoise. Federal Register 55:12178–12191. U.S. Fish and Wildlife Service. 1994. Desert Tortoise (Mojave Population) Recovery Plan. 73pp. and appendices, Portland, Oregon, USA. U.S. Fish and Wildlife Service. 2010. Endangered and threatened wildlife and plants; 12- month finding on a petition to list the Sonoran population of the desert tortoise as endangered or threatened; proposed rule. Federal Register 75:78094–78146. U.S. Fish and Wildlife Service. 2011. Revised recovery plan for the Mojave population of the desert tortoise (Gopherus agassizii), p. 222. U.S. Fish and Wildlife Service, Pacific Southwest Region, Sacramento, California. U.S. Fish and Wildlife Service. 2012. Listing priority changes in candidates. Federal Register 77:69997. Vilaca, S. T., S. M. Vargas, P. Lara-Ruiz, E. Molfetti, E. C. Reis, G. Lobo-Hajdu, L. S. Soares, and F. R. Santos. 2012. Nuclear markers reveal a complex introgression 50

pattern among marine turtle species on the Brazilian coast. Mol. Ecol. 21:4300– 4312. Weir, B. S. and C. C. Cockerham. 1984. Estimating F-statistics for the analysis of population structure. Evol. 38:1358–1370. Wilson, J. S. and J. P. Pitts. 2010. Illuminating the lack of consensus among descriptions of earth history data in the North American deserts: A resource for biologists. Progress in Physical Geography 34:419–441. Wood, D. A., R. N. Fisher, and T. W. Reeder. 2008. Novel patterns of historical isolation, dispersal, and secondary contact across Baja California in the Rosy Boa (Lichanura trivirgata). Mol. Phylogenet. Evol. 46:484–502. Wood, D.A., A. G. Vandergast, K. R. Barr, R. D. Inman, T. C. Esque, K. E. Nussear, and R. N. Fisher. 2013. Comparative phylogeography reveals deep lineages and regional evolutionary hotspots in the Mojave and Sonoran deserts. Divers. Distrib. 19:722–737. Wood, D. A., R. N. Fisher, and A. G. Vandergast. 2014. Fuzzy boundaries: Color and gene flow patterns among parapatric lineages of the western shovel-nosed snake and taxonomic implication. PLOS One DOI:10.1371/journal.pone.0097494. Wood, S.N. 2006. Generalized additive models - An introduction with R. CRC Press, Boca Raton, Florida, 392 p. Wood, S. N. 2011. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. J Roy. Stat. Soc. B. 73:3– 36.

Tables

Table 1. Sample locality information for a total of 234 desert tortoise samples representing G. agassizii in California (n = 103), G. morafkai in Arizona (n = 78), and 53 individuals of undetermined lineage in the secondary contact zone.

Population Location Sample sites Site ID n County State Category Eastern Mojave RU Ivanpah IV 33 San Bernardino California G. agassizii G. agassizii Ivanpah (site 14) IV14 22 San Bernardino California G. agassizii Shadow Valley SV 3 San Bernardino California G. agassizii Colorado Desert RU Fenner FEN 4 San Bernardino California G. agassizii Goffs G 26 San Bernardino California G. agassizii Chemhuevi CH 6 San Bernardino California G. agassizii Upper Ward Valley UWV 9 San Bernardino California South side (near Oatman Black Mountains Hwy) GS 4 Mohave Arizona Contact Zone East Bajada (Long Term Monitoring Plot) EB 13 Mohave Arizona Contact Zone West side (near Silver Creek Rd.) WBM 18 Mohave Arizona Contact Zone Buck Mountains BUCK 6 Mohave Arizona Contact Zone Hualapai Mountains Shingle Canyon HSC 7 Mohave Arizona Contact Zone Hualapai foothills HF 5 Mohave Arizona Contact Zone Arrastra Mountains AM 4 Mohave Arizona G. morafkai G. morafkai Miller Mountain Bonanza Wash BW 5 Yavapai Arizona G. morafkai Bismarck Mountain Little Shipp Wash LS 10 Yavapai Arizona between Wickenberg and G. morafkai US Highway 93 Wikieup, AZ US93 10 Yavapai Arizona G. morafkai Eagletail Mountains ET 14 Maricopa Arizona G. morafkai Harcuvar Mountains HARC 16 Yavapai Arizona New Water G. morafkai Mountains NW 4 La Paz Arizona Wickenburg G. morafkai Mountains WM 15 Yavapai Arizona

Table 2. Descriptive statistics of 26 STR loci for 4 desert tortoise populations: Eastern Mojave RU (G. agassizii), Colorado Desert RU (G. agassizii), Arizona G. agassizii and Arizona G. morafkai with all probable hybrid individuals removed from the dataset. n = number of individuals genotyped; % pairwise linkage = linkage disequilibrium (nonrandom association between loci) calculated among all pairs of loci within each population (Slatkin and Excoffier 1996); Theta(H) population parameter where θ = 4Neμ for diploids (Ohta and Kimura 1973); Hobs = observed heterozygosity; Hexp = expected heterozygosity; s.d. = standard deviation of randomization tests for Hardy-Weinberg equilibrium; and FIS = inbreeding coefficient (Weir and Cockerham 1984).

Eastern Colorado G. agassizii G. morafkai Mojave RU Desert RU (AZ) (AZ) Total n 58 45 32 80 214 Theta(H) 1.500 1.500 1.502 1.547 1.512 # of alleles 7.615 8.192 7.423 8.769 15.115 s.d. 6.494 7.441 6.307 7.881 12.087 Allelic Range 16.714 15.524 16.091 17.708 19.846 s.d. 11.718 13.299 13.494 14.505 15.712 % pairwise linkage 4.64 3.48 17.39 17.10

Hobs 0.568 0.594 0.542 0.539 s.d. 0.282 0.265 0.323 0.262

Hexp 0.627 0.617 0.591 0.623 s.d. 0.275 0.270 0.329 0.247

FIS per population 0.095 0.038 0.085 0.137 p-value 0.001 0.016 0.001 0.001

Table 3 Genotyping results for 234 desert tortoises based on 26 STR loci. Species identification based primarily on Q-values from the STRUCTURE analysis (Pritchard et al. 2000) with additional analyses used to characterize hybridization.

Location Map/Sample Code G. agassizii G. morafkai Hybrid Eastern Mojave RU IV 33 IV14 22 SV 3 Colorado Desert RU FEN 4 G 26 CH 6 UWV 9 Black Mountains GS 4 EB 9 4 WBM 16 2 Buck Mountains BUCK 2 4 Hualapai Mountains HSC 1 2 4 HF 1 4 Arrastra Mountains AM 3 1 Miller Mountain BW 5 Bismarck Mountain LS 10 US Highway 93 US93 10 Eagletail Mountains ET 14 Harcuvar Mountains HARC 16 New Water Mountains NW 4 Wickenburg Mountains WM 15

Table 4. Comparison of analysis methods used to classify 19 individuals as having a possible hybrid origin. Qualitative methods included identity of mtDNA haplotype as either MOJ (G. agassizii) or SON (G. morafkai) and presence/zygosity of locus Goag05a where the locus is typically “null” in G. agassizii and thus an admixed individual is expected to be homozygous at this locus. GeneClass 2 (Piry et al. 2004) was run for 4 different models and on datasets with and without locus Goag05 (8 total methods); proportion of methods identifying an individual as a hybrid are reported as well as whether the dominant admixture type, G. agassizii (Goag) or G. morafkai (Gomo) matches that of the maternal lineage. STRUCTURE (Pritchard et al. 2000) was used to classify individuals without prior putative population information. NewHybrids

(Anderson 2008) was used to predict specific recombinant classes (F1, F2 or backcross).

Structure MtDNA Gen Goag05 New Proportion of Location Site ID Haplotype der a Hybrids Geneclass II admixture (Q) Proportion of Geneclass/ GeneClass2 hybrid mtDNA mismatch detection Gomo Goag Black

Mountains EB MOJ_A1 M HOM F2 Yes 7 of 8 0.60 0.40 Black

Mountains EB SON_01 M HET F2 Yes 5 of 8 0.37 0.63 Black

Mountains EB MOJ_A1 M NULL F2/Bx No 2 of 8 0.12 0.88 Black

Mountains EB SON_01 F NULL F2/Bx Yes 3 of 8 0.15 0.85 Black

Mountains WBM MOJ_A01 M HOM F2 Yes 8 of 8 0.50 0.50 Black Mountains WBM MOJ_A01 F NULL Pure_Goag No 2 of 8 0.02 0.98 Buck

Mountains BUCK SON_05 M HET F2 No 4 of 8 0.66 0.34 Buck

Mountains BUCK MOJ_A01 F NULL F2 No 6 of 8 0.40 0.60 Buck

Mountains BUCK MOJ_A01 M HOM F2 No 2 of 8 0.30 0.70 Buck

Mountains BUCK MOJ_A01 F HOM F2 Yes 8 of 8 0.54 0.46 Hualapai

Mountains HF SON_05 M HET F2 - 0 of 8 0.71 0.29 Hualapai

Mountains HF SON_01 M HOM F2 No 1 of 8 0.71 0.29 Hualapai Mountains HF MOJ_A01 M HET Gomo_Bx Yes 3 of 8 0.96 0.04 Hualapai Mountains HF SON_01 F HET Gomo_Bx - 0 of 8 0.90 0.10 Hualapai

Mountains HSC MOJ_A01 F HOM F2 Yes 5 of 8 0.61 0.39 Hualapai

Mountains HSC SON_01 M HOM F2 No 2 of 8 0.76 0.25 Hualapai

Mountains HSC SON_01 F NULL F2 - 0 of 8 0.63 0.37

Hualapai

Mountains HSC SON_01 unk. HOM F2 Yes 6 of 8 0.52 0.48 Arrastra

Mountains AM SON_01 unk. HOM F2/Bx - 0 of 8 0.72 0.28

Figure Legends

Figure 1. Location of sample sites for G. agassizii (circles) in California (CA), G. morafkai (squares) in Arizona (AZ), and individuals of undetermined lineage (triangle) in the presumed contact zone (Table 1). The California samples represent two different Recovery Units (RUs); the Colorado Desert RU (FEN, G, CH and UWV) and the Eastern Mojave RU (SV, IV and IV14), within the Distinct Population Segment (DPS) for G. agassizii as defined by the Desert Tortoise Recovery Plan (USFWS 2011). The boundary of the DPS and the State border between California and Arizona is the Colorado River.

Figure 2. STRUCTRE analysis results for K = 3 (A) and K = 2 (B) using all 234 tortoises in the study. Site IDs defined in Table 1.

Figure 3. Spatially inferred genetic discontinuities between populations visualized using GENELAND (Guillot et al. 2005). Images generated using STR dataset only. Plotted using UTM planar coordinates, blue line approximates the Colorado River. Cluster A) captures G. agassizii sample sites in the Eastern Mojave RU; Cluster B) includes Colorado Desert RU and Black Mountain G. agassizii sites in Arizona; Cluster C) includes sites in the Hualapai and Buck mountains, representing the hybrid zone – note the narrow cline between clusters B and C; Cluster D) represents all other G. morafkai sites in Arizona.

Figure 4. Genealogical reconstruction of mitochondrial haplotypes (ND3, arginine tRNA, ND4L, and part of ND4) rooted with G. berlandieri outgroup. Nodes labeled with mean TMRCA as million years ago with brackets containing 95% Highest Posterior Density intervals; generated using Beast v1.7.5 (Drummond and Rambaut 2007). 50% majority-rule consensus tree bootstrap values in parentheses.

Figure 5. Covariate distributions of genotype class with respect to topographic (A, B), climatic (C) and vegetative (D) habitat variables. Genotype classes characterized as G. agassizii (GOAG), G. morafkai (GOMO) or Admixed. (Not all habitat variables analyzed are presented in figure).

Figure 6. Probability distribution map of the potential range of G. agassizii and G. morafkai where “stippled pattern” represents predicted G. morafkai (Gomo) habitat and “boxed pattern” represents predicted G. agassizii (Goag) habitat. Gray background is not predicted to be habitat for desert tortoises. Habitat suitability modeling of habitat use and niche separation of species based on the genotype of sampled individuals (G. agassizii; G. morafkai, or admixed) determined from the STRUCTURE analysis Q-values. Sampling sites include G. agassizii (circles) in California, G. morafkai (squares) in Arizona, and individuals from the secondary contact zone in Arizona (triangles). Location is an inset map of Figure 1, and is bisected east/west by the Colorado River. 57

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Supplementary Material S1. Distribution of mtDNA haplotypes (as defined by Murphy et al. 2007) among 195 G. agassizii and G. morafkai samples characterized in 3 groupings: “GOAG” - representing G. agassizii sampled from two Recovery Units (RUs) in California; “Contact Zone” – samples collected where G. agassizii and G. morafkai both occur east of the Colorado River in northwestern Arizona; “GOMO” – representing G. morafkai sampled in western Arizona.

MOJ_A01 MOJ_A04 MOJ_B01 MOJ_B02 MOJ_B03 SON_01 SON_05

Location Sample sites DQ649394.1 DQ649397.1 DQ649398.1 DQ649399.1 DQ649400.1 DQ649401.1 DQ649405.1 n Eastern Mojave RU IV 13 3 7 23

IV14 16 3 19

SV 3 3

Colorado Desert RU FEN 4 4

G 19 1 20

CH 2 1 3

UWV 4 1 5 Total "GOAG"

64 29 1 34 3 10 77

Black Mountains GS 4 4

EB 9 2 11

WBM 14 2 1 17

Buck Mountains BUCK 5 1 6

Hualapai Mountains HSC 2 4 1 7

HF 1 2 2 5 Total "Contact Zone" 35 2 1 8 4 50

Arrastra Mountains AM 4 4

Hualapai Mountains LS 7 1 8

Miller Mountain BW 2 3 5

US Highway 93 US93 5 2 7

Table S1 (cont.) MOJ_A01 MOJ_A04 MOJ_B01 MOJ_B02 MOJ_B03 SON_01 SON_05

Location Sample sites DQ649394.1 DQ649397.1 DQ649398.1 DQ649399.1 DQ649400.1 DQ649401.1 DQ649405.1 n Eagletail Mountains ET 13 13

Harcuvar Mountains HARC 11 2 13

New Water Mountains NW 3 3

Wickenburg Mountains WM 13 2 15 Total "GOMO" 55 13 68

Total All 64 1 36 4 10 63 17 195

Supplementary Material S2. Results of two-sample Kolmogorov-Smirnov tests used to compare the occurrence/distribution of tortoise genotype classes on the landscape with respect to multiple habitat variables. Genotype class comparisons between G. agassizii (GOAG), G. morafkai (GOMO) and admixed individuals (AD) and where nonAD represents a combination of non-admixed genotypes GOAG and GOMO. Topographic variables include; topographic position index (TPI), surface roughness (SRF), surface texture (ST) and solar insolation (SLR). Climatic variables include; summer maximum temperature (TSMX), winter minimum temperature (TWMN), summer/winter temperature difference (TDIFF), winter precipitation (PCPwnt), summer/winter precipitation ratio (PCPrat) and probability of 3 year drought (PCPp3yd). Vegetative variables include; length of growing season (DUR), photosynthetic activity (AMP) and annual change in photosynthetic activity (AMPt). In addition, distance to Colorado River (D2River) and distance to ectype edge (D2Edge) were specifically included to predict the potential distribution of admixed individuals. p- and D- values derived from KS tests.

Genotype class Genotype class Genotype class Genotype class Variable comparison p D comparison p D comparison p D comparison p D TPI 6 GOAG x GOMO 0.000 0.956 GOAG x AD 0.000 0.764 AD x GOMO 0.000 0.854 AD x nonAD 0.000 0.582

6 SRF GOAG x GOMO 0.000 0.939 GOAG x AD 0.000 0.764 AD x GOMO 0.000 0.831 AD x nonAD 0.003 0.457 ST GOAG x GOMO 0.000 0.917 GOAG x AD 0.000 0.704 AD x GOMO 0.000 0.661 AD x nonAD 0.003 0.449 SLR GOAG x GOMO 0.000 0.427 GOAG x AD 0.047 0.352 AD x GOMO 0.071 0.344 AD x nonAD 0.255 0.255 TSMX GOAG x GOMO 0.000 0.494 GOAG x AD 0.003 0.468 AD x GOMO 0.023 0.397 AD x nonAD 0.114 0.301 TWMN GOAG x GOMO 0.000 0.475 GOAG x AD 0.037 0.363 AD x GOMO 0.057 0.356 AD x nonAD 0.132 0.294 TDIFF GOAG x GOMO 0.000 0.980 GOAG x AD 0.000 0.837 AD x GOMO 0.000 0.902 AD x nonAD 0.000 0.638 PCPwnt GOAG x GOMO 0.000 0.773 GOAG x AD 0.000 0.527 AD x GOMO 0.000 0.574 AD x nonAD 0.023 0.375 PCPrat GOAG x GOMO 0.000 0.872 GOAG x AD 0.000 0.610 AD x GOMO 0.000 0.657 AD x nonAD 0.004 0.441 PCPp3yd GOAG x GOMO 0.000 0.870 GOAG x AD 0.000 0.566 AD x GOMO 0.000 0.598 AD x nonAD 0.006 0.431 DUR GOAG x GOMO 0.000 0.416 GOAG x AD 0.088 0.322 AD x GOMO 0.134 0.310 AD x nonAD 0.332 0.238 AMP GOAG x GOMO 0.000 0.519 GOAG x AD 0.001 0.522 AD x GOMO 0.018 0.410 AD x nonAD 0.098 0.309 AMPt GOAG x GOMO 0.001 0.269 GOAG x AD 0.271 0.257 AD x GOMO 0.148 0.304 AD x nonAD 0.715 0.176 d2River GOAG x GOMO 0.000 0.831 GOAG x AD 0.000 0.556 AD x GOMO 0.000 0.580 AD x nonAD 0.009 0.416 d2Edge GOAG x GOMO 0.000 0.402 GOAG x AD 0.161 0.289 AD x GOMO 0.139 0.308 AD x nonAD 0.397 0.226 66

APPENDIX B

Manuscript submitted for publication – in review

Shaping species with ephemeral boundaries; the distribution and genetic structure of desert tortoise (Gopherus morafkai) in the Sonoran Desert Region

Taylor Edwards1,2, Mercy Vaughn3, Philip C. Rosen1, Cristina Meléndez Torres4, Alice E. Karl5, Melanie Culver1,6 and Robert W. Murphy 7

1School of Natural Resources and the Environment. The University of Arizona, Tucson, AZ 85721 USA 2University of Arizona Genetics Core, University of Arizona, Tucson, AZ 85721 USA 3Paso Robles, CA 93446 USA 4Comisión de Ecología y Desarrollo Sustentable del Estado de Sonora, Sonora, Mexico 5P.O. Box 74006, Davis, CA 95617 USA 6U.S. Geological Survey, Arizona Cooperative Fish and Wildlife Research Unit, University of Arizona, Tucson, AZ 85721 USA 7Royal Ontario Museum, Toronto M5S 2C6, Canada

Corresponding author: Taylor Edwards University of Arizona Genetics Core 1657 E. Helen Street, room 111 Tucson, Arizona 85721 [email protected]

ABSTRACT

Aim We examine the role biogeographic features played in the evolutionary history of Morafka’s desert tortoise (Gopherus morafkai) and test whether this species exhibits a gradient of genetic diversity across its distribution or if it constitutes distinct lineages. Increased knowledge of the past and present distribution of the region’s biota provides insight into the forces that drive and maintain the biodiversity in the desert Southwest.

Location Sonoran Desert biogeographic region; Sonora and Sinaloa, Mexico and Arizona, U.S.A.

Methods We examined wild Morafka’s desert tortoises from Mexico (n = 155) and Arizona (n = 78), spanning their known distribution. We used mtDNA sequences for phylogenetic reconstruction, 25 microsatellite (STR) loci for Bayesian clustering analysis, and data from both mtDNA and STRs for clinal analyses to infer population structure and evolutionary history.

Results Gopherus morafkai is comprised of genetically and geographically distinct “Sonoran” and “Sinaloan” lineages. Within each lineage, genetic structure exhibited a pattern of isolation by distance. Both lineages occurred in a relatively narrow zone of overlap in Sinaloan thornscrub, where it transitions into Sonoran desertscrub. Limited introgression occurred at the contact zone and this appeared to be a result of secondary contact. A mosaic model of ecological segregation best described the pattern.

Main conclusions The shifting ecotone between tropical deciduous forest and Sonoran desertscrub biomes appears be an ephemeral boundary that fostered adaptations in each tortoise lineage and resulted in their parapatric distribution. The sharp cline currently observed between the two lineages suggests that divergence was further influenced by periods of isolation in temporary refugia, driven apart during periods of climatic flux during the Pleistocene. Despite incomplete reproductive isolation, the Sonoran and Sinaloan G. morafkai lineages are on separate evolutionary trajectories.

Key Words Cline, Ecotone, Haplotype, Hybrid Zone, Mosaic Model, Speciation, Testudinidae

INTRODUCTION

The biota of the Sonoran Desert is the most diverse of all North American deserts and this assemblage has been sculpted by time, climate, and adaptation. The Sonoran Desert developed gradually with the full extent of its current flora not realized until < 10,000 years ago (Riddle & Hafner, 2006). However, desert-adapted (-exapted) taxa long predate formation of the desert. Prior to and during the Pleistocene (~1.8 Ma to 11,000 yrs) the Sonoran Desert probably had a more tropical flora and fauna. Many of the desert’s characteristic taxa have tropical origins (Riddle & Hafner, 2006) and tropical forest likely covered the region through the early Miocene.

The recently described Morafka’s desert tortoise, Gopherus morafkai (Murphy et al., 2011), occurs in the Sonoran Desert region and is distributed south and east of the Colorado River, extending southward into the northerly extensions of tropical deciduous forests in southern Sonora, northern Sinaloa, and extreme southwestern Chihuahua (Woolrich-Piña 2004; Murphy et al., 2011). Across its large geographic range, Morafka’s desert tortoise occupies a variety of habitats including Sonoran desertscrub, Sinaloan thornscrub and Sinaloan tropical deciduous forests (Fritts & Jennings, 1994; Berry et al., 2002). Mitochondrial DNA (mtDNA) sampled from G. morafkai in the southern portion of the range in Mexico is deeply divergent from haplotypes in the central part of the species’ range (Lamb et al., 1989; Berry et al., 2002; Edwards et al., 2012). Although research on the population structure of G. morafkai has been conducted on U.S. populations (Edwards et al., 2004), little is known about its population genetic structure within the majority of its range in Mexico (Berry et al., 2002).

Herein, we characterize the population genetic structure of G. morafkai throughout its range and examine biogeographic features associated with its distribution. We test the hypothesis that G. morafkai maintains two genetically distinct lineages with the null hypothesis that population structure meets expectations of isolation by distance and genotype distribution forms a gradient with uninhibited introgression among populations. To test these hypotheses, we reconstruct matrilineal genealogies, test for geographic association of haplotypes and estimate the time to most recent common ancestor among matrilines. We then perform a Baysian clustering analysis on 25 microsatellite/STR loci to discern the genetic structure among populations of G. morafkai. Finally, we perform clinal analyses on both mtDNA and STR loci to determine the position and amount of introgression where lineages converge and we assess the extent of population genetic structure associated with ecological features.

MATERIALS AND METHODS

Sample Collection

We hand-captured and collected samples from 155 wild desert tortoises in Mexico from 2005–2013. We collected <1 ml whole blood in a lithium heparin tube via brachial or subcarapacial venipuncture. The University of Arizona Institutional Care and Use Committee (IACUC) approved all tortoise-handling protocols (IACUC Control no. 09- 138). Data from 78 tortoises from Arizona, USA had previously been collected for other studies (Edwards et al., 2004; Edwards et al., 2010; Edwards et al. in press) and was included in this study.

We attempted to sample the full distribution of G. morafkai in Mexico and all representative habitats where this species is known to occur (Fig. 1; Table 1). Where possible, we targeted sites where tortoises were locally known to exist. We sampled from the three major biotic community types in Mexico where tortoises occur; Sonoran desertscrub (SDS), Sinaloan thornscrub (STS), and tropical deciduous forest (TDF) (Table1). At the northern edge of the TDF distribution it becomes discontinuous where it grades into thornscrub at lower elevations. In this ecotone, tropical deciduous forest elements may still predominate on some hillsides, but the vegetation stature is reduced. The distribution of thornscrub is disjunct, especially at ecotones with desertscrub and TDF, where it forms a mosaic. The transition of TDF and thornscrub to desertscrub dominated by more xeric species is often between 200–300 m but with notable exceptions (Van Devender et al., 2000).

Molecular Techniques

We isolated genomic DNA from either whole blood, salvaged red blood cells or lymphatic fluid. We sequenced a 1,108 bp portion of mtDNA (ND3, arginine tRNA, ND4L, and part of ND4) and genotyped all samples for 25 previously described short tandem repeats (STRs/microsatellites) following Edwards et al. (in press). Fragment analysis and DNA sequencing were performed by the University of Arizona Genetics Core.

Molecular Analyses

We performed phylogenetic analyses on the mtDNA sequence data to establish relationships among maternal lineages. We used DnaSP (v.5.10.01; Librado & Rozas, 2009) to estimate nucleotide diversity, polymorphism and other descriptive statistics within and among maternal lineages (Table 2). We used BEAST v.1.7.5 (Drummond & Rambaut, 2007) to produce phylogenetic trees and to establish estimates of time to most recent common ancestor (TMRCA). We ran BEAST using haplotype sequences only and

included sister taxa G. agassizii (GenBank Accession no. DQ649394–DQ649400) plus one sequence of G. berlanderi (GenBank: DQ649409.1). Previous estimates of mtDNA divergence time between G. agassizii and G. morafkai have been fairly consistent at 5–6 mya (Avise, 1992; Lamb & Lydeard, 1994; McLuckie et al., 1999; Edwards, 2003). We set a prior of 5.9 ±0.5 mya for our Bayesian analysis in BEAST based on Edwards (2003) because that estimate used the same mtDNA locus as this study. We selected an HKY substitution model with the gamma parameter set to 4 using MrModeltest 2.3 (Nylander, 2004) and we chose a relaxed, log-normal clock and Yule process model. We ran the Metropolis-coupled Markov chain (MCMC) for 100,000,000 generations sampling every 1000, with a burnin of 10%. We viewed results in MCMC Trace Analysis Tool Version v1.6.0 (Rambaut et al., 2003-2013) and used TreeAnnotator v.1.7.5 to select the Maximum Clade Credibility tree which has the highest product of the posterior probability of all its nodes from the BEAST analysis. We used FigTree v.1.4.0 (Rambaut, 2006-2012) to visualize the tree. In addition we used PAUP* v.4.0b10 (Swofford, 2002) to reconstruct maternal genealogies using both likelihood and parsimony optimality criterion searches to generate tree topologies for comparison with the Bayesian analysis executed with BEAST. We performed analyses only with unique haplotypes with all characters of equal weight and G. berlandieri as an outgroup. We performed a heuristic search with 100,000 random addition replicates. We estimated support for inferred relationships by conducting 10,000 nonparametric bootstrap replicates. We performed maximum likelihood analysis using the HKY model of nucleotide evolution.

We used STRUCTURE v.2.3.4 (Pritchard et al., 2000) to define populations in our STR dataset and to identify admixed individuals. We ran a multi-locus STR analyses for all 233 samples. We used an admixture model with allele frequencies correlated among populations. We tested for K = 1–8 with 10 trials per K, each run for 500,000 iterations following a burn-in period of 50,000 MCMC. We used STRUCTURE HARVESTER Online (Earl & Vonholdt, 2012) to evaluate STRUCTURE results and used DeltaK to determine the best fit of K for the data following Evanno et al. (2005).

We generated descriptive statistics on a dataset of 20 populations based on locality, excluding sample sites with less than five individuals. We used ARLEQUIN v.3.11 (Excoffier et al., 2005) to detect significant departures from Hardy-Weinberg expectations (Guo & Thompson, 1992), assess population differentiation with analysis of molecular variance (AMOVA), Theta(H) (Ohta & Kimura, 1973), and to test for linkage disequilibrium (nonrandom association between loci) among all pairs of loci within each population (Slatkin & Excoffier, 1996). We used FSTAT v.2.9.3.2 (Goudet, 1995) to estimate inbreeding coefficients (FIS; Weir & Cockerham 1984). We used default parameters in FSTAT and ARLEQUIN for all Markov-chain tests and permutations.

Geographic modeling and cline analyses

We used GenGIS v.2.1.1 (Parks et al., 2009) to visualize mtDNA sequence distributions and to test for geographic association of haplotypes. GenGIS performed a Monte Carlo permutation test to determine if the fit of ordered leaf nodes at geographic locations from the tree topology was significantly greater than expected by chance alone (Parks & Beiko, 2009). We used NTSYSpc (v.2.02h, Applied Biostatistics Inc.) to perform mantel tests to assess correlation between genetic distances calculated from STRs (FST) and geographic distances among populations.

We evaluated available vegetation maps and GIS layers (Rzedowski, 1978, 2006; Brown & Lowe, 1980; INEGI 1996, 2005), recent literature and published maps on regional vegetation and floristics (Burquez & Martínez-Yrízar, 2010; Ceballos et al., 2010b; Felger et al., 2012), as well as our own observations of fauna and flora at our sample sites. We mapped the transition from Sonoran desertscrub to thornscrub (Figure 1) following Brown and Lowe (1980), which is widely used and agreed with our field observations. The transition from thornscrub to TDF in Sonora was mapped by digitizing from published maps in Burquez et al. (1999) and Felger et al. (2012), which correctly represented changes in vegetation structure and flora we found during tortoise sampling. We incorporated these boundaries into a GIS map using INEGI (2005) data that also represented the subtropical to tropical transitions within thornscrub, and computed distances to biotic community transitions for each of our sample sites in ArcGIS 10 (ESRI, Redlands, CA). Because of the north-pointed wedge-shaped distribution of “Sinaloan-type” genotypes and the patchy (though overall similarly shaped) distribution of habitat in the complex TDF-thornscrub-desertscrub transition, we did not assess geographic distance along a linear transect but instead measured the shortest distance (km) of each site to the biotic community boundaries described above, calculated from the map we assembled.

We used CFIT7 (Gay et al., 2008) to perform a geographical cline analyses to determine the position and amount of introgression where lineages converged and to determine the extent of population genetic structure associated with ecological features. We treated mtDNA as allelic data and microsatellites as quantitative traits for the analyses. For the microsatellite data, we used the q-value of each individual derived from the highest likelihood model generated by STRUCRTURE as a multilocus genotype. For mtDNA data, we assessed three different cline models: simple sigmoid model, asymmetric model and three-parts model with different positions of tails. For microsatellite genotype data, we tested multiple unimodal, bimodal and trimodal character cline models and used Akaike Information Criterion (AIC) to determine the best supported model. A bimodal distribution is expected where parental types are distinct whereas a unimodal distribution

is expected when unrestricted introgression occurs and intermediate genotypes predominate (i.e. hybrid swarm; Jiggins & Mallet, 2000). A flat or “trimodal” distribution suggests that the proportion of intermediates is relatively uniform around the center, where there was no selective disadvantage to interbreeding (Barton & Gale, 1993; Jiggins & Mallet, 2000; Gay et al., 2008). We also tested for concordance (slope) and coincidence (center) of clines between the best models selected for the microsatellite and mtDNA clade assignments.

RESULTS

We identified 18 mtDNA haplotypes unique to G. morafkai in Mexico (Appendix S1). SON (Sonoran) and SIN (Sinaloan) haplogroups differed by 47 fixed single nucleotide polymorphisms (Dxy per site = 0.049; Jukes and Cantor 1969) and an estimated TMRCA of 5.77 Ma (Fig. 2). The maximum likelihood, parsimony and Bayesian trees consistently resolved three major clades (Fig. 2). Phylogenetically, SIN formed the sister clade of G. agassizii (MOJ), native to the Mojave Desert (Fig. 2). Pairwise estimates of divergence suggested that the SON clade is equally divergent from the SIN and MOJ clades; SONxSIN, Dxy = 0.050; SONxMOJ, Dxy = 0.051; and SINxMOJ, Dxy = 0.045. Haplotype distributions were strongly associated with geography (Fig. 3; MC permutation test, p ≤ 0.005). Haplotype SON_01 was ubiquitous throughout Sonoran desertscrub habitat spanning over 850 km from Empalme, Sonora, to Kingman, Arizona (Hualapai Mountains). An uncommon (n = 7 individuals; haplogroup SON_B) matriline occurred in (but was not exclusive to) the southern part of the Sonoran Desert transitional habitat at sites OPO, CCO, RSJ, and RLP. Five fixed polymorphisms differed between

SON_01-13 and SON_B01–B03 (Dxy = 0.00574). The population on Tiburòn Island (TIB) had the ubiquitous SON_01 haplotype (n = 2), a haplotype shared with the mainland coastal sample site (SER), and unique haplotype SON_11 (n = 3) that differed from SON_01 by a single nucleotide (Table 3). Isla Tiburòn has not been connected to the mainland for at least 4,700 ybp (Richman et al., 1988). Negative values for neutrality tests for the Sonoran matriline (SON) suggested a reduction of heterozygosity owing to an expansion event, positive directional selection, or the presence of weakly deleterious alleles (Tajima, 1989) (Table 2).

The STRUCTURE analysis performed on the STR data obtained a best fit model of K = 2 following Evanno et al. (2005). Analyses using both K=2 and K=3 clearly defined “Sinaloan” and “Sonoran” genotypes at opposite ends of the geographic distribution but admixed individuals were present at transitional locations (Table 1, Figure 4; REA, RM, SUA, RAS, OPO, CCO, RLP). The distribution of genotypes was bimodal with predominant parental (Sinaloan/Sonoran) genotypes (n = 194) and relatively little admixture (17%; n = 39) (Fig. 4). Most admixed individuals represented backcrossed genotypes and only 6% (n = 13) were suggestive of being F1 hybrids (0.3–0.7 q-value 73

mixture proportion; Fig. 4). These potential F1 hybrids occurred in both sexes, but with 9 males, 2 females and 2 unknown.

The AMOVA analysis exhibited a wide range of genetic differentiation among pairwise population comparisons (Appendix S2). Among all 20 populations, 11% of the genetic variation occurred among different sample locations (Fst = 0.106, p≤0.001). Genetic distances among populations grouped according to location/genotype showed the greatest differentiation between Sinaloan populations and both Sonoran, AZ (Fst = 0.182, p≤0.05) and Sonoran, MX populations (Fst = 0.164, p≤0.05). In comparison, Sonoran AZ and MX were genetically similar (Fst = 0.027, p≤0.05). Populations categorized as being transitional exhibited disproportionate genetic affinities to each of the neighboring parental; Sonoran MX (Fst = 0.033, p≤0.05) and Sinaloan (Fst = 0.769 and, p≤0.05). Pairwise population genetic and geographic differentiation were significantly correlated (Appendix S2).

In the analysis of clinal variation, the distribution of mtDNA haplogroups best fit a simple sigmoid curve (Table 4). For the microsatellite genotype assignment, the best fit model was a bimodal distribution but we were not able to distinguish this fit as being more appropriate than a unimodal distribution using the evidence ratio for AIC. A strong coincidence of slope and concordance of center occurred between the microsatellite and mtDNA clines (Table 4). Viewing these data using the distance from each sample site to the STS/SDS ecotone boundary, it appeared that introgression between Sonoran and Sinaloan lineages was restricted to STS environments, generally within 50 km of the boundary with SDS (Figure 5). The limited introgression did not readily extend into SDS or TDF habitats.

DISCUSSION

Divergence among and within lineages

The G. morafkai Sonoran “SON” mtDNA clade of desert tortoises exhibited equal divergence from the Sinaloan “SIN” clade and the G. agassizii “MOJ” clade. Although lineage sorting best explains the paraphyletic placement of SIN as the sister to MOJ despite SON being geographically proximate to the latter; additional genomic analyses may resolve the conundrum. The mtDNA divergence observed among these lineages (4.5–5.1%) is consistent with species-level divergence in other chelonian genera (cyt b gene, 2.8–18.3%; Vargas-Ramirez et al., 2010).

Within each of the SON and SIN mtDNA clades, tree-bifurcation defines geographically isolated haplogroups (Figure 2). The southernmost SIN samples (site FUE) displays the highest mtDNA diversity (SIN_B01–05) none of which occur in other TDF/Sinaloan- type populations (Figure 3). Seemingly, haplogroup SON_B occurs only in the southern 74

portion of the range. Relative to populations in the northern part of the range of G. morafkai, the higher diversity and deeper branch lengths of southern SON haplotypes may reflect a tropical origin (Rogers & Harpending, 1992). The high frequency of younger haplotypes and negative values for neutrality tests (Table 2) is consistent with desert tortoises having undergone a recent population expansion (Avise et al., 1987). The similar but independent coalescence times (TMRCA) within mtDNA haplogroups (Figure 2; SON, SON_B, SIN_A, SIN_B, and including G. agassizii MOJ_A and MOJ_B) may reflect a prehistoric reduction of genetic diversity that coincides with Pleistocene glacial–interglacial climatic change (Morafka, 1977; Van Devender et al., 1987; Morafka & Berry, 2002). This suggests that widespread environmental conditions (Hewitt, 1996) drove population expansion in different desert tortoise lineages.

The deep divergence between maternal lineages of G. morafkai is corroborated by the genetic structure observed across STR loci, supporting two distinct Morafka’s desert tortoise lineages, here referred to as “Sonoran” and “Sinaloan”. Each lineage was strongly associated with specific habitats; tropical deciduous forest and Sonoran desertscrub. However, in transitional habitat (SDS/STS ecotone), we observed populations exhibiting mixed genotypes and spatially limited hybridization (Figs. 4 & 5).

Contact zone

Introgression between Sonoran and Sinaloan lineages is minimal and appears to be limited to the STS, which acts as an ecotone between TDF and SDS. The shared, common center in clines across neutral markers is indicative of secondary contact (Endler, 1977; Barton & Hewitt, 1985). Cline width (slope) is relative to the time since contact, intensity of dispersal and the strength of selection, where wide clines infer that reinforcement is weak (Endler, 1977; Hewitt, 1996; Smith et al., 2013). Without the influence of exogenous or endogenous selection, the slope of the cline is expected to decrease over time with the rate of decrease proportional to the dispersal intensity (Gay et al., 2008). For desert tortoises, the within lineage isolation by distance population structure (IBD) is consistent with maintaining greater than one migrant per generation among population pairs (Edwards et al., 2004). Thus, subsequent to secondary contact between lineages (assumedly post last glacial maximum) if introgression between lineages is uninhibited we expect to observe a gradient consistent with IBD across the contact zone. Instead we observed a sharp cline for both mtDNA and STR loci which indicates that introgression is inhibited by selection. Our failure to detect elevated linkage disequilibrium, FIS, or allelic richness across the hybrid zone suggests that either hybridization is not frequent, there is a high proportion of later-generation hybrids or introgressed alleles are quickly reincorporated into the existing population through backcrossing (Culumber et al., 2011; Wallace et al., 2011). Although a bimodal distribution of genotypes across the contact zone is the best-fit model for our data, we 75

cannot exclude a unimodal distribution fit of the data, which is expected in cases where reproductive isolation is weak and the center of the cline may exhibit a hybrid swarm (Jiggins & Mallet, 2000; Gay et al., 2008). A bimodal distribution better explains our data given the low levels of hybridization and because the majority of admixed individuals are later-generation hybrids and backcrosses.

Hybrid zones are often observed at ecotones between two distinct habitats (Harrison, 1993; Arnold, 1997) and this may be important in understanding the role of hybridization in the evolutionary process. The ecotone offers habitat elements of each parental group but microhabitat use of individuals in these areas has not been characterized. The maintenance of genetic differentiation between Sonoran and Sinaloan lineages along an ecotone is highly suggestive of ecological segregation (Tarroso et al., 2014), where interactions between genotype and environment may determine the genetic structure of the hybrid population (Endler, 1977). There are several models used to describe hybrid zones where the amount of hybridization is dependent on factors such as dispersal ability, environment, and selection (Harrison, 1993; Arnold, 1997). A “Tension Zone” occurs in areas of parapatry where there is no barrier to mating between two populations, but where hybrids show reduced viability or fecundity (Key 1968). This model describes a hybrid zone in continuous habitat where the amount of hybridization is driven by a balance between amount of dispersal and endogenous selection (Moore & Price 1993; Kruuk et al. 1999). However, because our observations occur at an ecotone between two distinct habitats, exodogenous selection is more likely influencing the amount of hybridization that occurs. In the “Bounded Hybrid Superiority Model”, hybrid zones occur in disturbed areas or along ecotones where parental types are less adapted and the hybrids exhibit greater fitness than the parental types (Moore 1977). Similarly, the “Evolutionary Novelty Model” hybrids may be more fit under certain conditions (Arnold 1997). The two aforementioned models where hybrids exhibit greater fitness then parental genotypes in an intermediate environment are not dependent on the dispersal ability of the organism, as defines the Tension Zone model. Our data do not indicate that hybrids are more likely to persist in the thornscrub environment where the hybrid zone occurs, as other sites in STS that are further away from the SDS boundary (RM, SUA) exhibit a greater propensity for Sinaloan-type tortoises. Where hybrid zones are environment- and dispersal-dependent a cline may be observed (Endler 1977). Cline width can be suggestive of the strength of selection where wide clines infer that reinforcement is weak (Smith et al. 2013). While our observations are consistent with the geographical-selection gradient model described by Endler (1977), the ecotone observed between Sonoran Desertscrub and Sinaloan Thornscrub environments does not form a smooth gradient but instead is very complex and patchy. We instead suggest the distribution of Sonoran and Sinaloan lineages of G. morafkai at the ecotone best follows a “Mosaic Model” (Howard, 1986), which describes patchy environments where different parental lineages have a 76

high probability of coming into contact with one another, but mating results in hybrids that are less fit. Despite each lineage remaining independent, hybrid zones can remain stable through a balance of dispersal and selection against hybrid genotypes (Barton & Hewitt, 1985, 1989; Hewitt, 1996). In addition, mixed genotypes may have functionality in an ecotone and, therefore, there may be no benefit for sister taxa to become fully isolated (Nei & Nozawa, 2011).

Biogeographic Considerations

No major geographic barriers are present to limit gene flow between the Sonoran and Sinaloan lineages which suggests they may have evolved under a parapatric model where divergence is a result of adaptation to specific ecological niches. It is suggested that many of the characteristic taxa of the Sonoran Region are tropical in origin (Riddle & Hafner, 2006) and it is likely that tropical forest covered the region through the early Miocene (23.7 – 5.3 Ma). Here the common ancestor of the three desert tortoise lineages may have been widespread throughout what are now the Mojave and Sonoran desert regions. It is generally agreed upon that the region itself did not begin its drying trend until around 15- 8 Ma due to episodes of vicariance, such as the formation of the Sierra Madre (Van Devender, 2000). As such, the physical topography was in place by the end of the Miocene (5.3 Ma) to be the starting point for what can be considered the formation of the current North American deserts and their diverse biota. This changing environment would have opened up new niches in the northern portion of the ancestral range of the desert tortoise and opportunities to adapt to more arid conditions could have started the process of ecological divergence between arid and tropical ecotypes; resulting in the parapatric distribution of Sonoran and Sinaloan lineages.

However, these desert and tropical environments likely expanded and contracted many times during the relatively recent transformation into what is now the Sonoran Desert (Van Devender 2000) and this dynamic, ephemeral system has undoubtedly influenced speciation of the flora and fauna. In addition, divergence between Sinaloan and Sonoran lineages appears to have been further driven during periods of isolation resulting from Pleistocene climate change. In the Sonoran lineage, population expansion within mtDNA haplogroups and the strong geographic association of matrilines indicates population reductions to a few, isolated refugia for G. morafkai, followed by expansions to fuse back into a contiguous distribution. The changing boundaries of the desert and subtropical habitats throughout the Pleistocene provide insight into potential refugia. During glacial periods, thornscrub may have been restricted to hillsides and riparian corridors while desertscrub was isolated on xeric slopes and in valleys. Woodlands (oak-pine-juniper) may have maintained connectivity throughout this matrix (Van Devender, 2000). In the northern part of Sonora, the Gran Desierto may have been a core area of desert throughout most of the Quaternary (Van Devender et al., 1987), although now it only 77

supports tortoises in very low densities. Further south, the central gulf coast likely maintained more xeric vegetation while surrounding areas were converted to woodlands during cooling trends (Anderson & Van Devender, 1995).

Comparative biogeography among species can provide evidence for landscape-level changes that shaped the overall diversity of the region (Flores-Villela & Martinez- Salazar, 2009). The large number of isolated, endemic mammal “Pleistocene relicts” in the region supports the presence of refugia that may have acted as centers of speciation (Bell et al. 2010; Ceballos et al., 2010a; Dominguez-Dominguez et al., 2011). In their phylogenetic study of six arid-adapted species of ground squirrels, Bell et al. (2010) observed evidence of recent expansion from several refugia including the Sonoran Coastal Plains. Both southern and northern Sonoran clades exhibited northward post glaciation movement from an ancestral Sonoran distribution. This discovery is consistent with our SON and SON_B matrilines in G. morafkai.

Conclusion

The transition from tropical to desert vegetation skirts north along the foothills of the Sierra Madre is like the black ink of a Japanese Shodo brushstroke fading into whitespace. This ephemeral boundary has likely oscillated over thousands of years during periods of climatic change, leaving patches of different vegetation types scattered among the hillsides and valleys that interweave into the desert landscape. This dynamic system has shaped the evolutionary history of Sonoran Desert species, like the desert tortoise, through population expansions, contractions, adaptation, isolation and gene flow. In desert tortoises, this has resulted in two distinct lineages, each uniquely adapted to different environments. Following the evolutionary species concept as elucidated by Wiley (1978), the Sonoran and Sinaloan lineages are on separate evolutionary trajectories but exhibit incomplete reproductive isolation. While we believe in taking a conservative approach to delimit species (Sites & Marshall, 2003), we support making a taxonomic distinction between Sonoran and Sinaloan lineages of G. morafkai. In addition, conservation and management of both lineages would benefit by focusing specific actions toward each metapopulation independently based on their habitat and resource needs.

ACKNOWLEDGEMENTS

We obtained Mexican samples through a multinational, collaborative effort including Comisión de Ecología y Desarrollo Sustentable del Estado de Sonora (CEDES) and F.R. Méndez de la Cruz, Instituto de Biología, Universidad Nacional Autónoma de México (UNAM). Permits for Mexican samples were facilitated by Secretaría de Medio Ambiente y Recursos Naturales (SEMARNAT). The collection of samples was only made possible by a dedicated field crew of over 50 volunteers over the 8 year period.

Special thanks go to K. Berry, S, Boland, Y. Leon, M. Figueroa, L. Smith and P. Woodman. Access to almost all of the collection sites in Mexico were made possible through the generosity of local land owners and we are extremely grateful for the assistance of ranch hands (vaqueros) who shared their local knowledge with us in the field. C. Jones, D. Grandmaison, and the Arizona Game and Fish Department granted permission to use the comparative data from Arizona samples. Primary funding for this project came from private donors, the Royal Ontario Museum Foundation, Natural Sciences and Engineering Research Council of Canada Discovery Grant 3148, and the U.S. Fish and Wildlife Service, Science Support Program (SSP). Additional funding and support was received from The Tucson Herpetological Society, Desert Tortoise Council, Arizona Game and Fish Department, University of Florida, University of Arizona Genetics Core, and the U.S. Geological Survey. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

REFERENCES Anderson, R.S. & Van Devender, T.R. (1995) Vegetation history and paleoclimates of the coastal lowlands of Sonora, Mexico - pollen records from packrat middens. Journal of Aird Environments, 30, 295–306. Arnold, M.L. (1992) Natural hybridization as an evolutionary process. Annual Review of Ecology and Systematics, 23, 237–261. Arnold, M.L. (1997) Natural hybridization and evolution. Oxford University Press, New York, USA. Avise, J.C. (1992) Molecular population-structure and the biogeographic history of a regional fauna - a case-history with lessons for conservation biology. Oikos, 63, 62–76. Avise, J.C., Arnold, J., Ball, R.M., Bermingham, E., Lamb, T., Neigel, J.E., Reeb, C.A. & Saunders, N.C. (1987) Intraspecific phylogeography - the mitochondrial-DNA bridge between population-genetics and systematics. Annual Review of Ecology and Systematics, 18, 489–522. Barton, N.H. & Hewitt, G.M. (1985) Analysis of hybrid zones. Annual Review of Ecology and Systematics, 16, 113–148. Barton, N.H. & Hewitt, G.M. (1989) Adaptation, speciation and hybrid zones. Nature, 341, 497–503. Barton, N.H. & Gale, K.S. (1993) Genetic Analysis of Hybrid Zones, in Hybrid zones and the evolutionary process (ed. by R.G. Harrison), pp. 13–45. Oxford University Press, New York, NY USA. Bell, K.C., Hafner, D.J., Leitner, P. & Matocq, M.D. (2010) Phylogeography of the ground squirrel subgenusXerospermophilusand assembly of the Mojave Desert biota. Journal of Biogeography, 37, 363–378. 79

Berry, K.H., Morafka, D.J. & Murphy, R.W. (2002) Defining the desert tortoise(s): our first priority for a coherent conservation strategy. Chelonian Conservation and Biology, 4, 249–262. Brown, D.E. & Lowe, C.H. (1980) Biotic communities of the southwest. Rocky Mountain Forest and Range Experiment Station, U.S. Forest Service. General Technical Report RM-78. Map. Búrquez, A. & Martínez-Yrízar, A. (2010) Límites geográficos entre selvas secas y matorrales espinosos y xerófilos: ¿qué conservar? Diversidad, Amenazas y Areas Prioritarias para la Conservación ee las Selvas Secas del Pacífico de México. Fonda Económica de Cultura (eds. G. Ceballos et al.) pp. 53–62. México, D.F. Búrquez, A., Martínez-Yrízar, A., Felger, R.S. & Yetman, D. (1999) Vegetation and habitat diversity at the southern edge of the Sonoran Desert. Ecology of Sonoran Desert Plants and Plant Communities (ed. R.H. Robichaux), pp. 36–67. University of Arizona Press, Tucson AZ. Ceballos, G., Arroyo-Cabrales, J. & Ponce, E. (2010a) Effects of Pleistocene environmental changes on the distribution and community structure of the mammalian fauna of Mexico. Quaternary Research, 73, 464–473. Ceballos, G., Espinoza, E., Bezaury, J. & Garcia, A. (eds.) (2010b) Diversidad, Amenazas y Areas Prioritarias para la Conservación ee las Selvas Secas del Pacífico de México. Fonda Económica de Cultura. México, D.F. Culumber, Z.W., Fisher, H.S., Tobler, M., Mateos, M., Barber, P.H., Sorenson, M.D. & Rosenthal, G.G. (2011) Replicated hybrid zones of Xiphophorus swordtails along an elevational gradient. Molecular Ecology, 20, 342–56. Dominguez-Dominguez, O., Vila, M., Perez-Rodriguez, R., Remon, N. & Doadrio, I. (2011) Complex evolutionary history of the Mexican stoneroller Campostoma ornatum Girard, 1856 (Actinopterygii: Cyprinidae). Bmc Evolutionary Biology, 11, 153. Drummond, A.J. & Rambaut, A. (2007) BEAST: Bayesian evolutionary analysis by sampling trees. Bmc Evolutionary Biology, 7, 214. Earl, D.A. & Vonholdt, B.M. (2012) STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources, 4, 359–361. Edwards, T. (2003) Desert tortoise conservation genetics. University of Arizona, Tucson, Arizona USA. Edwards, T., Schwalbe, C.R., Swann, D.E. & Goldberg, C.S. (2004) Implications of anthropogenic landscape change on inter-population movements of the desert tortoise (Gopherus agassizii). Conservation Genetics, 5, 485–499. Edwards, T., Jarchow, C.J., Jones, C.A. & Bonine, K.E. (2010) Tracing Genetic lineages of captive desert tortoises in Arizona. Journal of Wildlife Management, 74, 801– 807. 80

Edwards, T., Vaughn, M., Torres, M.C.M., Karl, A.E., Rosen, P.R., Berry, K.H. & Murphy, R.W. (2012) A Biogeographic perspective of speciation among desert tortoises in the Genus Gopherus: A preliminary evaluation. Merging science and management in a rapidly changing world: biodiversity and management of the Madrean Archipelago III (ed by G.J. Gottfried, P.F. Ffolliott, B.S. Gebow and L.G. Eskew), pp. 243–247. Tucson, AZ, USA. Edwards, T. & Berry, K.H. (2013) Are captive tortoises a reservoir for conservation? An assessment of genealogical affiliation of captive Gopherus agassizii to local, wild populations. Conservation Genetics, 14, 649–659. Edwards, T., Berry, K.H., Inman, R.D., Esque, T.C., Nussear, K.E., Jones, C.A. & Culver, M. (in press) Testing taxon tenacity of tortoises: evidence for a geographical-selection gradient at a secondary contact zone. Ecology and Evolution Endler, J.A. (1977) Geographic variation, speciation, and clines. Princeton University Press, Princeton, N.J. USA. Evanno, G., Regnaut, S. & Goudet, J. (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology, 14, 2611–2620. Excoffier, L., Laval, G. & Schneider, S. (2005) Arlequin (version 3.0): An integrated software package for population genetics data analysis. Evolutionary Bioinformatics, 1, 47–50. Felger, R.S., Austin, D.F., Van Devender, T.R., Sánchez-Escalante, J. & Costea, M. (2012) Convolvulaceae of Sonora, Mexico – I. Convolvulus, Cressa, Dichondra, Evolvulus, Ipomoea, Jacquemontia, Merremia, Operculina. Journal of the Botanical Research Institute of Texas, 6: 459–527. Flores-Villela, O. & Martinez-Salazar, E.A. (2009) Historical explanation of the origin of the herpetofauna of Mexico. Revista Mexicana de Biodiversidad, 80, 817–833. Fritts, T.H. & Jennings, R.D. (1994) Distribution, habitat use, and status of the desert tortoise in Mexico. Biology of North American Tortoises. Fish and Wildlife Research 13 (ed. by R.B. Bury and D.J. Germano), pp. 49−56 US Dept. of the Interior, National Biological survey, Washington, DC, USA. Fu, Y.X. & Li, W.H. (1993). Statistical tests of neutrality of mutations. Genetics, 133, 693-709. Gay, L., Crochet, P.A., Bell, D.A. & Lenormand, T. (2008) Comparing clines on molecular and phenotypic traits in hybrid zones: a window on tension zone models. Evolution, 62, 2789–806. Goudet, J. (1995) FSTAT (Version 1.2): A computer program to calculate F-statistics. Journal of Heredity, 86, 485–486. Guo, S.W. & Thompson, E.A. (1992) Performing the exact test of Hardy-Weinberg proportion for multiple alleles. Biometrics, 48, 361–372. 81

Harrison, R.G. (ed.) (1993) Hybrid zones and the evolutionary process. Oxford University Press, New York, USA. Hewitt, G.M. (1996) Some genetic consequences of ice ages, and their role, in divergence and speciation. Biological Journal of the Linnean Society, 58, 247–276. Howard, D.J. (1986) A zone of overlap and hybridization between 2 ground cricket species. Evolution, 40, 34-43. Instituto Nacional de Estadística, Geografía e Informática (1996) Conjunto nacional de usos actual des suelo y vegacion a escala 1:250,000. Serie II. DGG-INEGI, Mexico. Instituto Nacional de Estadística, Geografía e Informática (2005) Conjunto nacional de usos actual des suelo y vegacion a escala 1:250,000. Serie III. DGG-INEGI, Mexico. Jiggins, C.D. & Mallet, J. (2000) Bimodal hybrid zones and speciation. Trends in Ecology & Evolution, 15, 250–255. Jukes, T.H. & Cantor, C.R. (1969) Evolution of protein molecules. Mammalian Protein Metabolism (ed. by H.N. Munro), pp.21-132. Academic Press, New York. Key, K.H.L. (1968) Concept of stasipatric speciation. Systematic Zoology 17, 14–22. Kruuk, L.E.B., Baird, S.J.E., Gale, K.S. & Barton, N.H. (1999) A comparison of multilocus clines maintained by environmental adaptation or by selection against hybrids. Genetics 153, 1959–1971. Lamb, T. & Lydeard, C. (1994) A molecular phylogeny of the gopher tortoises, with comments on familial relationships within the Testudinoidea. Molecular Phylogenetics and Evolution, 3, 283–291. Lamb, T., Avise, J.C. & Gibbons, J.W. (1989) Phylogeographic patterns in mitochondrial-DNA of the desert tortoise (Xerobates-agassizi), and evolutionary relationships among the North-American gopher tortoises. Evolution, 43, 76–87. Librado, P. & Rozas, J. (2009) DnaSP v5: A software for comprehensive analysis of DNA polymorphism data. Bioinformatics, 25, 1451–1452. McLuckie, A.M., Lamb, T., Schwalbe, C.R. & McCord, R.D. (1999) Genetic and morphometric assessment of an unusual tortoise (Gopherus agassizii) population in the Black Mountains of Arizona. Journal of Herpetology, 33, 36–44. Moore, W.S. (1977) Evaluation of narrow hybrid zones in vertebrates. Quarterly Review of Biology 52, 263–277. Moore, W.S. & Price, J.T. (1993) Nature of selection in the northern flicker hybrid zone and its implications for speciation theory. Hybrid Zones and the Evolutionary Process (ed. by R.G. Harrison), pp. 196–225. Oxford University Press, New York, USA. Morafka, D.J. (1977) A biogeographical analysis of the Chihuahuan Desert through its herpetofauna. The Hague: Dr. W. Junk B.V. Publishers.

Morafka, D.J. & Berry, K.H. (2002) Is Gopherus agassizii a desert-adapted tortoise, or an exaptive opportunist? Implications for tortoise conservation. Chelonian Conservation and Biology, 4, 263–287. Murphy, R.W., Berry, K.H., Edwards, T., Leviton, A.E., Lathrop, A. & Riedle, J.D. (2011) The dazed and confused identity of Agassiz's land tortoise, Gopherus agassizii (Testudines, Testudinidae) with the description of a new species, and its consequences for conservation. Zookeys, 39–71. Nei, M. & Nozawa, M. (2011) Roles of mutation and selection in speciation: from Hugo de Vries to the modern genomic era. Genome Biology and Evolution, 3, 812–29. Nylander, J.A.A. (2004) MrModeltest v2. Program distributed by the author. Evolutionary Biology Centre, Uppsala University. Ohta, T. & Kimura, M. (1973) Model of mutation appropriate to estimate number of electrophoretically detectable alleles in a finite population. Genetical Research, 22, 201–204. Parks, D.H. & Beiko, R.G. (2009) Quantitative visualizations of hierarchically organized data in a geographic context. Geoinformatics, Fairfax, VA, 1–6. Parks, D.H., Porter, M., Churcher, S., Wang, S., Blouin, C., Whalley, J., Brooks, S. & Beiko, R. (2009) GenGIS: A geospatial information system for genomic data. Genome Research, 19, 1896–1904. Pritchard, J.K., Stephens, M. & Donnelly, P. (2000) Inference of population structure using multilocus genotype data. Genetics, 155, 945–959. Rambaut, A. (2006-2012) FigTree; tree figure drawing tool. Institute of Evolutionary Biology, University of Edinburgh. Rambaut, A., Suchard, M.A., Xie, W. & Drummon, A.J. (2003-2013) Tracer: MCMC Trace Analysis Tool. Institute of Evolutionary Biology, University of Edinburgh. Reed, K.M. & Sites, J.W. (1995) Female fecundity in a hybrid zone between 2 chromosome races of the Sceloporus grammicus complex (Sauria, Phrynosomatidae). Evolution, 49, 61–69. Richman, A.D., Case, T.J. & Schwaner, T.D. (1988) Natural and unnatural extinction rates of reptiles on islands. The American Naturalist, 5, 611–630. Riddle, B.R. & Hafner, D.J. (2006) A step-wise approach to integrating phylogeographic and phylogenetic biogeographic perspectives on the history of a core North American warm deserts biota. Journal of Arid Environments, 66, 435–461. Rogers, A.R. & Harpending, H. (1992) Population-growth makes waves in the distribution of pairwise genetic-differences Molecular Biology and Evolution 9, 552–569. Rzedowski, J. (1978) La Vegetacion de Mexico. Limusa, Mexico. Rzedowski, J. (2006) 1ra. Edición digital, Comisión Nacional para el Conocimiento y Uso de la Biodiversidad. México.

Sites, J.W. & Marshall, J.C. (2003) Delimiting species: a Renaissance issue in systematic biology. Trends in Ecology & Evolution, 18, 462–470. Slatkin, M. & Excoffier, L. (1996) Testing for linkage disequilibrium in genotypic data using the expectation-maximization algorithm. Heredity, 76, 377–383. Smith, K.L., Hale, J.M., Kearney, M.R., Austin, J.J. & Melville, J. (2013) Molecular patterns of introgression in a classic hybrid zone between the Australian tree frogs, Litoria ewingii and L. paraewingi: evidence of a tension zone. Molecular Ecology, 22, 1869–83. Swofford, D.L. (2002) PAUP*. Phylogeneitc analysis using parsimony (*and other methods). Version 4. Sinauer Associates, Sunderland, Massachusetts. Tajima, F. (1989) Statistical-method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics, 123, 585–595. Tarroso, P., Pereira, R.J., Martinez-Freiria, F., Godinho, R. & Brito, J.C. (2014) Hybridization at an ecotone: ecological and genetic barriers between three Iberian vipers. Molecular Ecology, 23, 1108–23. Van Devender, T.R. (2000) The Deep History of the Sonoran Desert. A Natural History of the Sonoran Desert (eds. S.J. Phillips & P.W. Comus), pp. 61–69.. University of California Press, Berkeley, CA Van Devender, T.R., Thompson, R.S. & Betancourt, J.L. (1987) Vegetation history in the Southwest: the nature and timing of the late Wisconsin-Holocene transition. North America and adjacent oceans during the last deglaciation (ed. by W.F. Ruddiman and J. H. E. Wright), pp. 323–352. Geological Society of America. Van Devender, T.R., Sanders, A.C., Wilson, R.K. & Meyer, S.A. (2000) Vegetation, flora, and seasons of the Río Cuchujaqui, a tropical deciduous forest near Alamos, Sonora, México. The tropical deciduous forest of Alamos: Biodiversity of a threatened ecosystem in Mexico (ed. by R.H. Robichaux and D.A. Yetman), pp. 36–101 University of Arizona Press, Tucson, AZ USA. Vargas-Ramirez, M., Vences, M., Branch, W.R., Daniels, S.R., Glaw, F., Hofmeyr, M.D., Kuchling, G., Maran, J., Papenfuss, T.J., Siroky, P., Vieites, D.R. & Fritz, U. (2010) Deep genealogical lineages in the widely distributed African helmeted terrapin: evidence from mitochondrial and nuclear DNA (Testudines: Pelomedusidae: Pelomedusa subrufa). Molecular Phylogenetics and Evolution, 56, 428–40. Wallace, L.E., Culley, T.M., Weller, S.G., Sakai, A.K., Kuenzi, A., Roy, T., Wagner, W.L. & Nepokroeff, M. (2011) Asymmetrical gene flow in a hybrid zone of Hawaiian Schiedea (Caryophyllaceae) species with contrasting mating systems. PLoS ONE, 6, e24845. Weir, B.S. & Cockerham, C.C. (1984) Estimating F-statistics for the analysis of population structure. Evolution, 38, 1358–1370.

Wiley, E.O. (1978) The Evolutionary Species Concept Reconsidered. Systematic Biology, 27, 17–26. Woolrich-Piña JG (2004) Geographic distribution: Gopherus agassizii. Herpetological Review 35, 1.

TABLES

Table 1. Desert Tortoise (Gopherus morafkai) sample locality information. n = number of individuals sampled. Biome descriptions; TDF = Tropical Deciduous Forest, STS = Sinaloan Thornscrub, SDS = Sonoran Desertscrub.

Site Biotic Site Location ID n community Vegetation Description Sinaloa (3 locations) Sinaloa, MX FUE 6 TDF Multiple sites; TDF and STS Alamos-El Divisadero Sonora, MX AED 6 TDF Tropical Deciduous Forest Alamos-Las Cabras Sonora, MX ALC 10 TDF Tropical Deciduous Forest Alamos-La Sierrita Sonora, MX ALS 13 TDF Tropical Deciduous Forest El Chupadero Sonora, MX EC 14 STS Ecotone: STS/SDS Monte Mojino Sonora, MX AMM 5 TDF Tropical Deciduous Forest

La Noria Sonora, MX RLN 2 TDF Ecotone: TDF/STS El Alamo Sonora, MX REA 3 TDF Ecotone: TDF/STS Moscobampo Sonora, MX RM 5 STS Ecotone: STS/SDS Suaqui Grande region Sonora, MX SUA 7 STS Tropical Deciduous Forest Arroyo Seco Sonora, MX RAS 3 STS Ecotone: STS/Semi-desert Grassland Opodepe-Las Milpas region Sonora, MX OPO 10 STS Sinaloan Thornscrub Cerro Colorado Sonora, MX CCO 8 STS SDS: Plains of Sonora La Pintada Sonora, MX RLP 11 STS Ecotone: STS/SDS San Judas Sonora, MX RSJ 8 SDS SDS: Plains of Sonora Seri region Sonora, MX SER 11 SDS SDS: Central Gulf Coast Isla Tiburòn Sonora, MX TIB 6 SDS SDS: Central Gulf Coast Bamuri Sonora, MX BAM 9 SDS Ecotone: SDS/Lower Colorado River Valley La Candelaria Sonora, MX CAN 18 SDS Ecotone: SDS/Lower Colorado River Valley Sauceda Mts region Arizona, USA SAU 11 SDS Ecotone: SDS/Lower Colorado River Valley

Tucson Mts Arizona, USA TMT 10 SDS SDS: Arizona Upland Granite Hills Arizona, USA GRH 17 SDS SDS: Arizona Upland Eagletail Mts Arizona, USA EAG 14 SDS Ecotone: SDS/Lower Colorado River Valley Harcuvar Mts Arizona, USA HAR 16 SDS SDS: Arizona Upland Little Ship Wash Arizona, USA LSW 10 SDS Ecotone: SDS/interior chaparral

Table 2. Descriptive statistics for mitochondrial DNA sequences from 213 desert tortoise (G. morafkai): n = number of samples/haplotype, π = nucleotide diversity (per site), θ = nucleotide polymorphism (per site; Theta-W), s.d. = standard deviation of estimate. Test of neutrality include: Tajima's D based on the differences between π and θ. (Tajima 1989); Fu and Li's D* derived from the number of singletons and the total number of mutations; and Fu and Li's F* derived from the number of singletons and the average number of nucleotide differences between pairs of sequences (Fu and Li 1993). Calculated using DnaSP (version 5.10.01; Librado and Rozas 2009).

Sonoran clade Sinaloan clade (SON_ and Sonoran SON_B All Mexican ALL (SIN_) SON_B) clade samples samples n samples 75 138 7 153 213 n haplotypes 8 12 3 18 20 Variable sites (of 1108) 9 15 2 69 71 Haplotype Gene Diversity 0.399 0.304 0.524 0.707 0.638

8 s.d. 0.067 0.051 0.209 0.024 0.027

π 0.00063 0.00075 0.00052 0.02414 0.0221 s.d. 0.00014 0.00019 0.00023 0.00024 0.00087

θ 0.00165 0.00246 0.00074 0.01109 0.01078 Tajima's D -1.604 -1.846 -1.237 3.657 3.178 Significance 0.10 > p > 0.05 p < 0.05 p > 0.10 p < 0.001 p < 0.01 Fu and Li's D* -1.541 -2.448 -1.296 0.663 0.423 Significance p > 0.10 p < 0.05 p > 0.10 p > 0.10 p > 0.10 Fu and Li's F* -1.848 -2.659 -1.374 2.377 2.019 Significance P > 0.10 P < 0.05 P > 0.10 P < 0.02 P < 0.05 88

Table 3. Summary statistics of 25 STR loci for 20 desert tortoise (G. morafkai) populations: n = number of individuals genotyped; Theta(H) population parameter where θ = 4Neμ for diploids (Ohta and Kimura 1973); % pairwise linkage = linkage disequilibrium (nonrandom association between loci) calculated among all pairs of loci within each population (Slatkin and Excoffier 1996); Hobs = observed heterozygosity; Hexp = expected heterozygosity; s.d. = standard deviation of randomization tests for Hardy-Weinberg equilibrium; and FIS = inbreeding coefficient (Weir & Cockerham 1984). %

# of Allelic pairwise FIS p-

Location Site Id n Theta(H) alleles s.d. Range s.d. linkage Hobs s.d. Hexp s.d. FIS value Sinaloa (3 locations) FUE 6 1.708 5.273 2.66 12.96 11.64 11.0% 0.652 0.235 0.742 0.174 0.133 0.002 Alamos-Las Cabras ALC 10 1.564 5.48 3.41 11.87 10.09 6.3% 0.601 0.303 0.640 0.261 0.063 0.058 Alamos-La Sierrita ALS 13 1.628 7.04 4.42 14.17 10.26 7.0% 0.658 0.238 0.677 0.232 0.039 0.102 El Chupadero EC 14 1.530 6.08 3.75 13.38 10.81 8.0% 0.501 0.250 0.584 0.245 0.147 0.000 Moscobampo RM 5 1.738 4.50 1.75 12.83 9.62 4.0% 0.625 0.252 0.691 0.208 0.106 0.022 Suaqui Grande region 8 SUA 7 1.517 4.95 2.48 12.52 9.82 13.0% 0.618 0.234 0.650 0.228 0.051 0.128

9 Opodepe-Las Milpas region OPO 10 1.865 6.38 3.17 14.25 10.30 14.3% 0.741 0.226 0.725 0.194 -0.022 0.778 Cerro Colorado CCO 8 1.898 6.67 3.44 15.79 11.39 5.3% 0.728 0.254 0.733 0.219 0.008 0.421 La Pintada RLP 11 1.801 7.25 3.89 16.54 13.34 5.3% 0.647 0.252 0.709 0.227 0.094 0.001 San Judas RSJ 8 1.726 5.96 3.14 14.88 12.48 5.0% 0.646 0.275 0.687 0.239 0.064 0.051 Seri region SER 11 1.664 6.33 3.85 14.08 12.07 9.0% 0.599 0.253 0.664 0.256 0.103 0.002 Isla Tiburòn TIB 6 1.608 4.08 1.86 11.33 11.22 6.3% 0.597 0.243 0.639 0.236 0.076 0.063 Bamuri BAM 9 1.630 5.74 3.21 14.65 12.71 2.7% 0.628 0.219 0.678 0.216 0.078 0.023 La Candelaria CAN 18 1.660 7.17 4.38 15.92 14.58 5.7% 0.639 0.240 0.663 0.217 0.037 0.086 Sauceda Mts region SAU 11 1.526 5.76 3.33 13.48 11.92 8.3% 0.574 0.262 0.663 0.270 0.142 0.000 Tucson Mts TMT 10 1.618 5.46 2.70 12.42 10.32 3.7% 0.601 0.275 0.644 0.252 0.066 0.046 Granite Hills GRH 17 1.508 6.52 3.95 14.00 10.98 3.7% 0.619 0.277 0.632 0.251 0.021 0.243 Eagletail Mts EAG 14 1.500 5.35 2.98 15.20 12.31 12.0% 0.587 0.290 0.622 0.229 0.058 0.062 Harcuvar Mts HAR 16 1.502 5.55 2.80 14.25 11.96 5.3% 0.576 0.275 0.645 0.210 0.111 0.001 Little Ship Wash LSW 10 1.503 4.67 2.99 14.52 13.71 4.3% 0.597 0.276 0.620 0.239 0.034 0.245

Table 4.

A. Comparison of models of clinal variation in microsatellite genotype and mtDNA clade assignment using the Akaike Information Criteria (AIC).

AICc Model Parameters Center Slope AICc weight Evidence ratio STRs unimodal 7 -36.023 -0.521 14.267 0.980 1.020 bimodal 7 -24.633 -0.566 14.227 1.000 - bimodal_no_introgression 8 14.138 -8.655 16.419 0.334 2.993 trimodal_no_introgression 14 -25.086 -3.129 30.021 0.010 96.745 trimodal 10 -25.357 -2.809 20.877 0.036 27.800

0 mtDNA

simple sigmoid 2 8.930 -10.000 3.995 1.000 - asymmetric 4 8.929 -10.000 8.158 0.125 8.015 three-parts model 6 8.930 -10.000 12.418 0.015 67.446

B. Comparison of cline models for coincidence and concordance of clines of the microsatellite and mtDNA clade assignments. Mix = shift of the means around the center.

AICc Model Parameters Center Slope Mix AICc weight Evidence ratio Unconstrained 9 - - 42.796 18.905 0.112 8.918 Center Constraint 8 9.127 - 74.128 16.705 0.337 2.969 Slope Constraint 8 - -9.996 29.903 16.705 0.337 2.969

Slope and Centre Constraints 7 8.930 -10.000 0.100 14.529 1.000 1.000

LIST OF FIGURE LEGENDS

Figure 1. Desert tortoise (G. morafkai) sample locations in Arizona, USA and Mexico. Habitat distribution estimated from (Brown & Lowe, 1980) and by digitizing published maps in Burquez et al. (1999) and Felger et al. (2012). “Alamos” label represents proximate sample sites; AED, ALC, ALS and AMM

Figure 2. Genealogical reconstruction of mitochondrial haplotypes from representative Gopherus sp. Haplogroups MOJ_A and MOJ_B sampled from G. agassizii. Haplogroups SON, SON_B, SIN_A and SIN_B sampled from G. morafkai. Texas tortoise, G. berlandieri used to root the tree. Nodes labeled with mean TMRCA as Ma with brackets containing 95% Highest Posterior Density intervals; generated using BEAST v1.7.5. 50% majority-rule consensus tree bootstrap values in parentheses; estimated from parsimony reconstruction using PAUP*.

Figure 3. MtDNA sequence distributions and geographic association of G. morafkai haplotypes. Purple shades represent “Sinaloan” haplogroup (haplotypes SIN_A01- SIN_B05). Shades of gray represent “Sonoran” haplogroup (haplotypes SON01-13). Brown shades represent southern Sonoran clade (haplotypes SON_B01-B04). Size of pie chart represents population sample size. Sites characterized by habitat type; Sonoran desertscrub (SDS), Sinaloan thornscrub (STS) and tropical deciduous forest (TDF). The distribution of haplotypes exhibited a strong association with geography (MC permutation test, p ≤ 0.005). Generated using GenGIS v.2.1.1 (Parks et al. 2009).

Figure 4. STRUCTURE multi-locus STR analysis for 233 wild caught samples from Mexico and Arizona. Best fit K based on DeltaK (Evanno et al. 2005) determined to be K = 2. Site IDs defined in Table 1.

Figure 5. Scatter plot of clinal variation in microsatellite (STR) genotype and mtDNA clade assignment. Each point represents a sampled population where MtDNA treated as allelic data (Sonoran or Sinaloan) for each individual represented and for microsatellites a multilocus genotype was estimated for each individual using the q-value generated by program STRUCTURE. Habitat distribution approximated for Sonoran desertscrub (SDS), Sinaloan Thornscrub (STS) and Tropical deciduous forest (TDF). Geographic distance was measured as the shortest distance (km) from each site to the STS/SDS ecotone boundary using the BLP GIS layer. Dotted line approximates TDF/STS ecotone boundary.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Appendix S1. Desert tortoise (G. morafkai) mtDNA Haplotype distribution. SIN haplotypes characterize the “Sinaloan” lineage while

SON haplotypes represent the “Sonoran” lineage. Average number of nucleotide substitutions per site between SON and SIN; Dxy = 0.049 (Jukes and Cantor 1969). Haplotypes labeled with ID and GenBank accession number.

MtDNA Haplotype

Biotic 516 Site ID

community

SON_01 SON_03 SON_08 SON_09 SON_10 SON_11 SON_12 SON_13

SON_05 SON_05

SIN_B01 SIN_B02 SIN_B03 SIN_B04 SIN_B05

SIN_A01 SIN_A02 SIN_A03 SIN_A04

SON_B01 SON_B02 SON_B03

KM411506 KM411507 KM411508 KM411509 KM411510 KM411511 KM411512 KM411513 KM411514 KM411515 KM411 KM411517 KM411518 KM411519 KM411520 KM411521 KM411522 KM411523

DQ649401.1 DQ649403.1 DQ649405.1

FUE TDF 2 1 1 1 1

AED TDF 3 2 1

ALC TDF 10

ALS TDF 5 7 EC

9 STS 14 1

AMM TDF 3

RLN TDF 2

REA TDF 2 1

RM STS 4 1

SUA STS 7 1

RAS STS 2 2

OPO STS 7 2

CCO STS 1 5 1 1

RLP STS 4 6 1

RSJ SDS 5 1 1

SER SDS 9 1 1

TIB SDS 2 3 1

BAM SDS 6 1 1 1

CAN SDS 15 3

SAU SDS 7

Table S1 - Cont. MtDNA Haplotype

Biotic Site ID

community

SON_01 SON_03 SON_08 SON_09 SON_10 SON_11 SON_12 SON_13

SON_05 SON_05

SIN_B01 SIN_B02 SIN_B03 SIN_B04 SIN_B05

SIN_A01 SIN_A02 SIN_A03 SIN_A04

SON_B01 SON_B02 SON_B03

KM411506 KM411507 KM411508 KM411509 KM411510 KM411511 KM411512 KM411513 KM411514 KM411515 KM411516 KM411517 KM411518 KM411519 KM411520 KM411521 KM411522 KM411523

DQ649401.1 DQ649403.1 DQ649405.1

TMT SDS 5 1

GRH SDS 13

EAG SDS 13

HAR SDS 11 2

LSW SDS 7 1

Total 57 10 1 1 2 1 1 1 1 115 1 3 1 1 1 3 5 1 5 1 1

Appendix S2. Pairwise population differentiation (Fst; below diagonal) and geographic distances (Km; above diagonal) among sampled populations. Italicized fields indicate non-significant p-values (p≥0.05). Shading designates location/genotype groupings for Mantel and AMOVA tests.

“Sinaloan” sites Sonoran/Sinaloan transition “Sonoran” MX sites “Sonoran” AZ sites FUE ALC ALS EC RM SUA OPO CCO RLP RSJ SER TIB BAM CAN SAU TMT GRH EAG HAR LSW

FUE 0 112 108 237 289 282 359 360 346 428 450 451 564 591 814 735 792 918 989 1027 ALC 0.114 0 5 151 207 189 263 278 269 341 385 391 478 509 720 633 689 824 893 929 ALS 0.085 0.001 0 151 207 189 264 278 269 342 384 389 479 510 721 635 691 826 894 930 EC 0.147 0.054 0.039 0 56 47 123 128 119 192 237 245 329 359 576 501 558 681 752 790 RM 0.124 0.099 0.087 0.102 0 41 86 72 62 139 181 191 275 304 526 458 515 629 702 741 SUA 0.085 0.029 0.039 -0.019 0.038 0 77 96 94 153 214 226 290 322 533 455 512 637 707 745 OPO 0.125 0.135 0.104 0.148 0.036 0.069 0 67 82 90 178 196 221 255 457 378 435 562 631 668 CCO 0.125 0.142 0.126 0.168 0.048 0.094 -0.004 0 21 70 120 135 204 232 457 398 453 560 633 674 RLP 0.150 0.157 0.127 0.196 0.051 0.119 0.019 0.011 0 88 120 132 219 245 473 417 473 576 650 691

00 RSJ 0.137 0.172 0.153 0.213 0.063 0.125 0.016 0.018 0.036 0 107 128 137 169 387 330 385 490 563 604

SER 0.167 0.180 0.151 0.222 0.078 0.146 0.029 0.046 0.012 0.038 0 23 152 161 404 384 435 501 579 625 TIB 0.180 0.200 0.167 0.242 0.079 0.170 0.035 0.055 0.021 0.031 -0.023 0 169 175 417 403 453 512 591 638 BAM 0.177 0.193 0.167 0.230 0.062 0.152 0.016 0.021 0.015 0.021 -0.004 -0.019 0 39 257 235 284 358 434 478 CAN 0.185 0.195 0.166 0.229 0.073 0.144 0.027 0.035 0.029 0.026 0.023 0.009 -0.002 0 243 244 288 340 418 464 SAU 0.156 0.180 0.149 0.220 0.077 0.144 0.021 0.012 0.027 0.019 0.009 -0.004 -0.012 0.031 0 146 132 105 177 221 TMT 0.182 0.189 0.165 0.234 0.086 0.158 0.045 0.063 0.044 0.039 0.015 -0.006 0.006 0.030 0.019 0 57 230 277 303 GRH 0.173 0.200 0.179 0.238 0.131 0.171 0.043 0.049 0.080 0.051 0.052 0.061 0.042 0.068 0.017 0.039 0 194 229 250 EAG 0.223 0.248 0.225 0.288 0.167 0.231 0.095 0.074 0.091 0.100 0.069 0.075 0.079 0.072 0.045 0.092 0.056 0 80 132 HAR 0.212 0.245 0.208 0.278 0.147 0.210 0.088 0.083 0.074 0.086 0.049 0.034 0.065 0.070 0.037 0.088 0.082 0.020 0 53 0.076 LSW 0.218 0.248 0.219 0.284 0.154 0.223 0.074 0.076 0.073 0.057 0.060 0.076 0.066 0.057 0.098 0.077 0.036 0.003 0

Mantel Tests results: All pops, n = 20, r2 = 0.368, p = 0.002; All Mexican samples (n = 14, r2 = 0.461, p = 0.002); Sinaloan-type only (n = 4, r2 = 0.553, p = 0.13); Mexican Sonoran-type only (n = 10, r2 = 0.461, p = 0.002); Mexican - Sonoran Desert Scrub only (n = 8, r2 = 0.379, p = 0.006); Arizona only (n = 6, r2 = 0.768, p = 0.004).

101

APPENDIX C

Manuscript submitted for publication – in review

Assessing models of speciation under different biogeographic scenarios; an empirical study using multi-locus and RNA-seq analyses

Taylor Edwards1,2, Marc Tollis3, PingHsun Hsieh4, Ryan N Gutenkunst4,5, Zhen Liu6, Kenro Kusumi3, Melanie Culver1,7 and Robert W. Murphy6,8

1The University of Arizona, School of Natural Resources and the Environment, Tucson, AZ 85721 USA 2University of Arizona Genetics Core, University of Arizona, Tucson, AZ 85721, USA 3School of Life Sciences, Arizona State University, Tempe, AZ 85287 4The University of Arizona, Department of Ecology and Evolutionary Biology, Tucson, AZ 85721 USA 5The University of Arizona, Department of Molecular and Cellular Biology, Tucson, AZ 85721 USA 6State Key Laboratory of Genetic Resources and Evolution, Kunming Institute of Zoology, Chinese Academy of Sciences, Kunming 650223, China 7Arizona Cooperative Fish & Wildlife Research Unit, USGS, University of Arizona, Tucson, AZ 85721, USA 8Centre for Biodiversity and Conservation Biology, Royal Ontario Museum, Toronto, Canada

Corresponding author: Taylor Edwards University of Arizona Genetics Core 1657 E. Helen Street, room 111 Tucson, Arizona 85721 [email protected]

102

Abstract

Evolutionary biology often seeks to decipher the drivers of speciation, and much debate persists over the relative importance of isolation and gene flow in the formation of new species. Genetic studies of closely related species can assess if gene flow was present during speciation, because signatures of past introgression often persist in the genome. We test hypotheses on which mechanisms of speciation drove diversity among three distinct lineages of desert tortoise in the genus Gopherus. These lineages offer a powerful system to study speciation, because different biogeographic patterns (physical vs. ecological segregation) are observed at opposing ends of their distributions. We used 82 samples collected from 38 sites, representing the entire species’ distribution. We generated sequence data for mtDNA and 4 nuclear loci and performed a multi-locus phylogenetic analysis in *BEAST to estimate the species tree. We also performed RNA- seq and characterized 20,126 synonymous variants from 7,665 contigs called in six individuals, representing each of the three lineages. We analyzed the RNA-seq data using the demographic inference package ∂a∂i to test the null hypothesis of no gene flow during divergence. The best-fit demographic model for the three taxa is concordant with the species tree, and the ∂a∂i analysis yielded no evidence of gene flow during divergence among any of the three lineages. These analyses suggest that divergence among the lineages occurred in the absence of gene flow, whether through physical allopatry or ecological niche segregation. Our results also validate species-level differentiation among the three lineages of desert tortoise.

Key words: allopatric, ∂a∂i, gene flow, Gopherus, parapatric, phylogenetic, transcriptome

103

Introduction

Geography, gene flow, and time strongly influence speciation, but the relative importance of these mechanisms can be difficult to quantify (Via 2009; Pinho & Hey 2010). Competing species concepts often differ fundamentally in the contribution of gene flow to the process of speciation. Recently diverged taxa facilitate studying the influence of gene flow on speciation because the signature of past introgression may persist in the genome (Pinho & Hey 2010).

It is difficult to test empirically for signatures of past introgression in natural systems due to pervasive genetic drift, genetic draft and variation in coalescence times (Hudson & Turelli 2003). Differences among gene genealogies may arise from differences in male/female dispersal, assortative mating and differential selection (Coyne & Orr 2004). Such processes result in discordance among gene tress of recently evolved species (Pollard et al. 2006; Degan & Rosenberg 2009; Zhang 2011). It is difficult to discriminate patterns of lineage sorting from patterns of past introgression because they have similar genetic signatures (McCormack et al. 2009; Payseur 2010; Pinho & Hey 2010). The competing explanations of gene tree discordance render the study of the contribution of gene flow to speciation in natural systems very challenging, although the more loci examined throughout the genome, the more likely a clear phylogenetic signal can be distinguished (Leaché & Rannala 2011).

Recent advances in biotechnology and biostatistics enable investigations into speciation. New molecular technologies and multi-locus genomic methods can fuse evolutionary history within an ecological context (Brito & Edwards 2009). Genomic approaches also allow for simultaneous exploration of differences in introgression among different parts of the genome (Payseur et al. 2004; Geraldes et al. 2006; Teeter et al. 2008; Melo- Ferreira et al. 2009). This integration of population genetic and phylogenetic perspectives promotes the creation of meaningful species trees (Degnan & Rosenberg 2009).

Explorations into the relative importance of divergence and gene flow requires identifiable patterns of speciation, such as in cases where two recently diverged populations come into secondary contact. Ecotones between two distinct habitats facilitate the testing of hypotheses on patterns of divergence. In this situation, hybridization, which may occur, can be an important part of the evolutionary process and an essential component in species’ ability to adapt to a changing environment (Barton & Hewitt 1989; Arnold 2007; Payseur 2010).

Desert tortoises (Gopherus sp.) lend themselves well to testing for the drivers of speciation and the roles played by ecology because they are recently diverged and wide-

104

ranging in multiple biomes (Fig. 1). Importantly, regions of overlap between divergent lineages occur, hybridization is ongoing and there may be signals of past introgression (McLuckie et al. 1999; Edwards et al. 2010). Ecotones define the distribution of divergent lineages and selection appears to maintain taxonomic boundaries (Edwards et al. 2015; Edwards et al. in review). Further, the three lineages in this system allow obtainment of a consensus among multiple gene genealogies, as there is greater potential to converge on an incorrect species tree when four or more taxa exhibiting discordance among gene trees (Degnan & Rosenberg 2009).

Gopherus agassizii (Agassiz’s desert tortoise) and G. morafkai (Morafka’s desert tortoise) are a classic example of allopatric speciation resulting from geographic isolation by the Colorado River (Lamb et al. 1989; Avise et al. 1992; McLuckie et al. 1999; Murphy et al. 2011). The former species occurs primarily west and north of the river in the Mojave Desert and G. morafkai ranges solely south and east of the river in the Sonoran Desert. A small population of G. agassizii occurs on the east side of the Colorado River in the territory of G. morafkai. Limited hybridization has been observed in a secondary contact zone between the two tortoises yet ecological niche partitioning maintains the species via a geographical selection-gradient (Edwards et al. 2015).

In the southern part of the range of G. morafkai, parapatry may explain the formation of genetically and geographically distinct “Sonoran” and “Sinaloan” lineages (Edwards et al. 2012). The Sonoran genotype has a large distribution throughout desertscrub in Sonora, Mexico and Arizona, USA. In contrast, the southern, Sinaloan lineage occurs solely in tropical deciduous forest and thornscrub environments (Fig. 1; Edwards et al. in review). The lineages occur sympatrically in a narrow ecotone between the two habitats and limited hybridization has been detected (Edwards et al. in review). No geographic barrier limits gene flow. Although this pattern implicates a parapatric model of speciation, these desert and tropical environments likely expanded and contracted many times during the relatively recent transformation (Van Devender 2000); this dynamic system has undoubtedly influenced speciation of the flora and fauna.

The underlying population structure of an organism is critical to making inferences about the rate and patterns of speciation. Within each of the G. agassizii (hereto referred to as the “Mojave” lineage) and the G. morafkai Sonoran and Sinaloan lineages, gene flow is geographically extensive and there is genetic structure with isolation by distance (IBD: Edwards et al. 2004; Murphy et al. 2007; Hagerty et al. 2011; Edwards et al. 2015; Edwards et al. in review). All three lineages appear to have experienced population expansions since the last glacial maximum (LGM) and populations within each lineage have low estimates of genetic differentiation: FST = 0.06 for G. agassizii; FST = 0.05 for G. morafkai Sonoran; and FST = 0.09 for G. morafkai Sinaloan (Edwards & Harrison 105

2014; Edwards unpublished data). This suggests that any sampling location within a lineage should contain roughly 90–95% of the genetic diversity of that lineage. At contact zones between lineages, if there is no species boundary then gene introgression should be expected to follow an IBD model. In contrast, for alleles that are under strong selection at species boundaries the rate of introgression should be near zero in one or both directions (Payseur 2010).

Herein, we test hypotheses on the biogeography and its influence on drivers of speciation. These involve comparisons of the well-established allopatric model of speciation observed between the Sonoran/Mojave lineages with predictions of the parapatric model between Sonoran/Sinaloan lineages. For the analyses, we use mitochondrial DNA (mtDNA), nuclear loci (nDNA) and RNA-seq data.

Material and methods

Ethics Statement

The University of Arizona Institutional Care and Use Committee (IACUC) approved all handling protocols (IACUC Control no. 09-138).

Hypotheses We test hypotheses that the three lineages of desert tortoises experienced different mechanisms of speciation. Hypothesis H-MS0 assumes that the Mojave and Sonoran lineages experienced allopatric speciation. Alternatively, H-MS1 involves parapatric speciation. Hypothesis H-SS0 assumes that the Sonoran and Sinaloan lineages experienced parapatric speciation without isolation and in the presence of gene flow.

Alternatively, H-SS1 assumes that these lineages diverged with cycling periods of isolation in refugia followed by secondary contact and with repeated periods of introgression, or, H-SS2, that they diverged without introgression during a single event of isolation (physical or ecological) followed by secondary contact.

Samples

Phylogenetic analyses/DNA sequencing employed 82 DNA samples collected from 38 sites previously used in other studies (Edwards et al. 2004; Murphy et al. 2007; Edwards et al. 2010; Edwards et al. 2015; Edwards et al. in review) representing samples from across the range of the desert tortoises (Table 1; Fig. 1). In addition, we obtained two samples of G. berlandieri from a private collection and two samples of G. flavomarginatus collected in Durango, Mexico (Morafka et al. 1994).

106

The RNA-seq data gathering used nine individuals in three flowcell lanes on the Illumina HiSeq platform. We dedicated two lanes to high-coverage sequencing of three individuals, one for each of the following lineages: G. agassizii was from a captive individual in Arizona that originated in California (Moj_A haplotype); Sonoran lineage G. morafkai was collected from a captive individual from Arizona; and Sinaloan lineage G. morafkai was obtained from just outside of Alamos, Sonora Mexico (Rancho Las Cabras; RLC). Data from these individuals were used to assemble reference transcriptomes. We used the third flowcell lane to generate low-coverage RNA-seq reads from six samples that were then mapped to the reference assembly to assess diversity within and among the three lineages. For these six samples, we hand-captured and collected samples from wild desert tortoises from the following three sites in 2013: two Sinaloan G. morafkai from the Reserva La Sierrita, Sierra de Alamos, Sonora, Mexico (Fig. 1, site ALS; tropical deciduous forest); two Sonoran G. morafkai from the Rancho San Judas north of Hermosillo, Sonora, Mexico (Fig. 1, site RSJ; Sonoran desertscrub/plains of Sonora); and two G. agassizii from Trout Canyon west of Las Vegas bordering the Eastern Mojave and Northeastern Mojave recovery units (USFWS 2011), Clark County, Nevada (Fig. 1, site TC; Mojave desertscrub with Larrea tridentata and Yucca brevifolia).

For RNA sample collection, we obtained <1 ml whole blood via brachial or subcarapacial venipuncture and mixed it with a greater than 50% volume RNA lysis/binding buffer from the Ambion RNAqueous kit (Life Technologies, Thermo Fisher Scientific Inc., Grand Island, NY). We immediately put the sample on ice and transferred the sample to liquid nitrogen for storage within 4 hrs of collection. We verified all nine RNA samples as being of the assumed lineage (and not hybrids) with subsequent DNA analyses using 25 microsatellite loci and mtDNA (Edwards & Berry 2013).

DNA Sequencing

We sequenced a 1,109 base pair (bp) portion of mitochondrial DNA (ND3, tRNAarg, ND4L, and part of ND4) following Edwards (2003) and Murphy et al. (2007). We sequenced the same fragment for G. flavomarginatus following Edwards et al. (2014). We optimized PCR conditions for six nuclear loci: BDNF, R35, and four uncharacterized loci identified by Thomson et al. (2008) derived from BAC libraries (TB02, TB07, TB53 and TB95). For PCR amplification of the nuclear loci, we used primer pairs as described by Leaché & McGuire (2006) and Thomson et al. (2008) except for locus R35, where we used an available GenBank sequence (accession number AY434646) to design Gopherus- specific PCR primers with OLIGO PRIMER ANALYSIS 6.68 (Molecular Biology Insights, Inc., Colorado Springs, CO) as follows: R35EX1_GOPH CACATACTGAATTTCCAGG, and R35EX2_GOPH GGACCTTTAAGTCATACAC. 107

We assessed optimal PCR conditions for each primer pair under 72 possible conditions by varying temperature from 52º to 64º C and MgCl2 concentration from 1.0 to 4.5 mM. We performed PCR amplification for subsequent Sanger sequencing in 30 µl reaction volumes containing 0.2 µM of each primer, 10 mM Tris–HCl (pH 8.3), 0.2 mM of each dNTP, 0.4 units of Platinum Taq (Life Technologies, Thermo Fisher Scientific Inc.), 5.0 mM KCl, 10 ng of genomic DNA, and locus-specific amounts of MgCl2 (Table 2). PCR began with an initial 3 min denaturation at 94 °C, followed by 35 cycles of 30 s at 94 °C, 30 s at the locus-specific annealing temperature (Table 2), and 90 s at 72 °C, followed by 3 min incubation at 72 °C. We submitted PCR product to the University of Arizona Genetics Core for DNA sequencing in both forward and reverse directions and followed standard protocols for the 3730XL DNA Analyzer (Applied Biosystems, Foster City, CA). We used CLC DNA WORKBENCH 5.7.1 (CLC bio, Denmark) to visually align sequences and DNASP 5.10.01 (Librado & Rozas 2009) to build fasta files and generate general descriptive statistics. We used PHASE (Stephens & Donnelly 2003) for haplotype reconstruction of diploid loci.

Phylogenetic Analysis

We reconstructed unrooted haplotype networks of nuclear loci to establish relationships among lineages using BEAST 2.1.2 (Bouckaert et al. 2014). We selected a substitution model for each locus using MRMODELTEST 2.3 (Nylander 2004); all loci fit a HKY, gamma 4 site model except TB07, which fit the GTR gamma 4 model. We chose a relaxed, log-normal clock and used the calibrated Yule model (Drummond et al. 2006). We ran the Metropolis-coupled Markov chain (MCMC) for 500,000,000 generations sampling every 5000, with a burnin of 10%.

For mitochondrial lineage reconstructions, we performed the analysis in BEAST as described above using G. flavomarginatus as the outgroup taxon to enable construction of a rooted tree. To establish estimates of time to most recent common ancestor (TMRCA) for the mtDNA locus only, we set the prior for our Bayesian analysis in BEAST for divergence time between G. agassizii and G. morafkai (Sonoran) lineages to 5.9 ±0.5 Ma based on Edwards (2003). Previous estimates of mtDNA divergence time between G. agassizii and G. morafkai have been fairly consistent at 5–6 Ma (Avise et al. 1992; Lamb & Lydeard 1994; McLuckie et al. 1999; Edwards 2003). In addition, we used PAUP* 4.0b10 (Swofford, 2002) to reconstruct maternal genealogies using both likelihood and parsimony optimality criterion searches to generate tree topologies. We compared these topologies with that derived from Bayesian analysis executed with BEAST. Analyses used unique haplotypes and all characters received equal weight. We performed a heuristic search with 100,000 random addition replicates. Support for inferred relationships was

108

estimated using 10,000 nonparametric bootstrap replicates. We performed maximum likelihood analysis using the HKY model of nucleotide evolution.

We also performed maximum-likelihood estimates using branch models of CODEML in PAML 4 (Yang 2007) to determine the mean selection pressures on different branches of the mtDNA tree. This method compared the ratio dN/dS, termed ω, where ω < 1 indicated purifying selection, ω = 1 indicated neutral selection, and ω > 1 indicated adaptive selection. First, we calculated ω under a one-ratio model in which the same ratio occurred across the tree. Next, we estimated an independent ω value for each branch under the free-ratio model.

We used the *BEAST model (Heled & Drummond 2010) for species tree estimation in BEAST using mtDNA and four of the nuclear loci (TB02, TB07, R35 and BDNF). *BEAST analyses used multi-locus data and the multi-species coalescent approach to infer species trees. We assigned individuals to putative species/lineages, which was difficult for individuals of G. morafkai that occurred along the thornscrub/desertscrub ecotone zone (Edwards et al. in review). We defined individuals with questionable genotypes based on their mtDNA haplotype as either Sinaloan or Sonoran based on Edwards et al. (in review). We used the HKY, gamma 4 site model for all loci except TB07 which we applied the GTR gamma 4 model. We used the Yule Process prior and did not set a coalescent prior. We set the population size function to linear with constant root and ran the MCMC for 500,000,000 generations with a 10% burnin for both strict and relaxed log normal clocks. We viewed results in MCMC TRACE ANALYSIS TOOL 1.6.0 (Rambaut et al. 2003-2013); both runs achieved stationary MCMC distributions and ESS values > 200. We used TREEANNOTATOR 1.7.5 (Rambaut & Drummond 2002-2012) to select the Maximum Clade Credibility tree which had the highest product of the posterior probabilities of all its nodes from the BEAST analysis, and FIGTREE 1.4.0 (Rambaut 2006– 2012) to visualize the tree. We performed a qualitative analysis on consensus trees using DENSITREE 2.2.1 (Bouckaert & Heled 2014).

Next-Generation Sequencing

We isolated total RNA from whole blood using standard protocols for the Qiagen RNeasy Kit (Qiagen, Valencia, CA). We quantified recovered RNA using a RiboGreen TBS-380 Flourometer (Turner BioSystems, Sunnyvale, CA) and assessed sample quality with an Advanced Analytics Fragment Analyzer using the High Sensitivity RNA Kit (Advanced Analytics, Ames, IA). We used the Illumina TruSeq RNA kit (Illumina, Inc., San Diego, CA) to build the cDNA library using total RNA. The kit targets polyadenylated mRNA for second strand cDNA synthesis and size-selects through enzyme-mediated fragmentation. While building the cDNA library, we applied unique

109

tags to each individual sample. We quantified each cDNA library using a Kapa Biosystems qPCR Kit (Kapa Biosystems, Inc., Wilmington, MA) specific to the Illumina Adapter sequence. We then pooled samples in equimolar concentrations for the final cDNA library. We ran the three high coverage individuals representing each of the three tortoise lineages on 2 flowcell lanes using an Illumina HiSeq 2000 and we ran the six lower-coverage individuals on a single flowcell lane using the Illumina HiSeq 2500 platform. All next-generation sequencing protocols were performed at the University of Arizona Genetics Core following standard protocols.

Transcriptome Assembly

For each library, we processed raw reads to remove sequencing adaptors, trimmed for quality score (Q > 28) and length-filtered (>37 bp) using TRIMMOMATIC 0.32 (Bolger et al. 2014). We used reads from the three high-coverage samples (Mojave, Sonoran and Sinaloan) to create de novo transcriptome assemblies for each species/lineage, as well as a combined assembly consisting of reads from all three libraries to be used as a reference. We assembled transcript contigs using TRINITY (Grabherr et al. 2011; Haas et al. 2013) with default settings. As de novo transcriptome assemblies often consist of many thousands of possibly chimeric contigs that lack clear gene content (Cahais et al. 2012), we further filtered the TRINITY output for contigs with single gene annotations. To accomplish this, we treated the TRINITY contigs as a query in a BLASTX search of mouse and chicken proteins from UniProt (Magrane et al. 2011) with an E-value cutoff of 1e-6. We then selected contigs containing unique BLAST hits to incorporate into a reference transcriptome for downstream analyses.

We followed a slightly modified protocol of De Wit et al. (2012) for mapping and variant detection. We performed the analysis using the six low-coverage RNA-seq samples and mapped these to the reference transcriptome. We used Burroughs Wheeler Aligner (BWA) 0.6.1 (Li & Durbin 2009) to generate Sequence/Alignment Map (SAM) files. We performed several trials to assess parameter sets and settled on using default parameters with assumed offset of 33, allowed for 0.005 differences between reference and query (- n), and allowed up to five differences in the seed (-k) to achieve > 67% of reads for each individual mapped to the reference transcriptome. We then converted SAM to BAM file format and removed duplicates using SAMTOOLS 0.1.18 (Li et al. 2009). We next merged all six individuals together to create a single BAM file and then followed recommendations in the De Wit et al. (2012) protocol to realign poorly mapped regions near indels using GENOMEANALYSISTK-1.0.5974 (GATK; McKenna et al. 2010). We also used GATK to detect and annotate variants and generate Variant Call Format (VCF) files (DePristo et al. 2011). We followed the recommendations of De Wit et al. (2012) and called only variant sites with a Phred scale quality of more than 30. We then performed 110

low threshold variant detection and Variant Quality Score Recalibration (VQSR) following De Wit et al. (2012) to build a Gaussian mixture model to be able to accurately distinguish true variant sites from false positives. We then parsed out only the variant sites for which we have genotype information for all individuals with a Phred quality score cutoff of 20. We used these data for all downstream analysis.

We performed Principal Components Analysis as an initial assessment of these data using SMARTPCA in EIGENSOFT 5.0.2 (Patterson et al. 2006). We then annotated these data using TRANSDECODER (Haas et al. 2013) and identified candidate coding regions within transcript sequences. TRANSDECODER generated a gff3 file which we then converted to a GTF (general transfer format) file using GFFREAD in CUFFLINKS 2.2.1 (Trapnell et al. 2010). We then used SNPDAT 1.0.5 (Doran & Creevey 2013) to generate a GTF file annotating gene sequences shared by all six individuals.

Demographic inference using ∂a∂i

We used ∂a∂i 1.6.3 (Gutenkunst et al. 2009) to fit demographic models to our six transcriptome samples. We built the folded joint allele frequency spectrum using only variants that SNPDAT unambiguously called as synonymous and that were successfully genotyped in all six individuals. To choose between our final 5- and 6-parameter models, we performed a likelihood ratio test (LRT) with a statistical correction for linkage based on the Godambe information matrix (Coffman, Hsieh, Gravel, & Gutenkunst, pers comm). Parameter confidence intervals were estimated by 100 conventional bootstraps over contigs and calculated as , where was the parameter of interest and was the standard deviation of the bootstrap results. To convert parameters from population-genetic units (scaled by ancestral population size Na) to physical units, we assumed a divergence time between G. agassizii and Sonoran G. morafkai lineages of 5.9 Ma (Edwards 2003) and a generation time of 25 yr (USFWS 1994). To convert bootstrap parameter values, we used our maximum composite-likelihood parameter values and the relation to estimate the quantity and then applied this value all bootstrap samples.

Results

We obtained usable sequences for the mtDNA and four of the six nuclear loci (BDNF, R35, TB02 and TB07; Table 2). Loci TB53 and TB95 generated electropherograms with excessive noise on repeated attempts and, thus, they were excluded. In phasing the data for nuclear loci, all samples for BDNF had a minimum pair probability of 0.996 and all were selected for downstream analyses. For R35, all but two samples were selected with a minimum pair probability of 0.942. For TB02, all samples were accepted; three pairs 111

were below a probability of 0.922 but were still included. None of these samples had unique haplotypes that were not represented in another sample. For TB07, 51 samples had 100% pair probabilities and were selected for analysis; 18 samples had pair probabilities close to 0.50 between two haplotype pairs (13 samples with the same two possible haplotype pairs were G. agassizii or Sonoran-type G. morafkai and five G. morafkai samples collected in Mexico shared the same two ambiguous haplotype pairs). Both examples represented the same ambiguity in the same SNP and each of the four haplotypes among the ambiguous pairs was represented in the other 51 samples. For this locus we choose the haplotype pair with the highest p value to use for analyses.

We assessed patterns of haplotype diversity by grouping samples by species (G. flavomarginatus, G. berlandieri, G. agassizii, G. morafkai), by lineages within G. morafkai (Sonoran and Sinaloan) and by geographic regions where mtDNA differentiation had been previously observed (Murphy et al. 2007; Edwards et al. 2015; Edwards et al. in review). Among the nuclear loci, haplotypes were both lineage-specific and found globally across species of Gopherus (Table S1, Fig. 2). Haplotype diversity, nucleotide polymorphism and nucleotide diversity estimates varied across loci and among sample-groupings and did not suggest strong trends. For example, haplotype diversity was greater in G. agassizii then G. morafkai at some loci but not others, and Tajima’s D was positive for some loci and negative for others (Table S2).

Phylogenetic analysis

Our phylogenetic reconstructions of nuclear loci lacked reciprocal monophyly such that haplotypes of G. flavomarginatus and G. berlandieri nested between or were shared with desert tortoises (Table S1, Fig. 2). The mtDNA reconstruction had strong support across analyses for distinct Mojave/Sonoran/Sinaloan lineages, however, with the Mojave and Sinaloan clades sister to the Sonoran clade (Fig. 2a). The multi-locus analyses exhibited the same topology for both strict and relaxed log normal clocks with species relationships consistent with expectations and with Sinaloan and Sonoran lineages sister to the Mojave. Estimated TMRCAs exhibited wide standard deviations (Fig. 3). Qualitative analysis using DENSITREE suggested strong concordance among iterations for the resulting tree topology but with less precision around depth of nodes (Fig. 3b).

For tests of selection on the mtDNA locus, our estimations of ω employed different models for branches of the tree (Table 3). First, assuming a uniform ω for all branches of the five species/lineages of Gopherus, ω was estimated to be 0.6192, which was significantly less than one (P < 0.05). This result suggested that this gene was under overall strong purifying selection. Next, we applied a model in which every branch had its own ω. This model was significantly more effective than one-ratio model (P < 0.05;

112

Table 3), suggesting that ω varied among the different lineages. The value of ω on the branch leading to SIN (Sinaloan lineage haplogroup) was estimated to be larger than one (1.1), indicating that positive selection may have affected this locus (Table 3).

Transcriptome assembly

We assembled 111,635,751 trimmed reads from whole blood total RNA into a combined G. agassizii and G. morafkai assembly that contained 235,412 contigs (Table 4). The blast-filtered combined assembly contained 40,341 transcripts with a contig N50 of 3,010 bp and a mean contig length of 1,957 bp. After aligning the six individuals against the combined assembly and identifying the variant alleles, we characterized 95,220 polymorphic sites for which we have genotype information for all individuals. The PCA assessment showed extremely strong clustering of individuals within each lineage and relatively equidistant differentiation among lineages (Fig. 4).

Demographic modeling

We used the allele frequency spectrum (AFS)-based inference tool ∂a∂i (Gutenkunst et al. 2009) to infer the joint demographic history of the three desert tortoise lineages from our transcriptome data. Because AFS-based demographic inference is sensitive to genotyping errors (Gutenkunst et al. 2009) and selection (Williamson et al. 2005), we considered only synonymous variants successfully called in all six individuals, yielding an AFS with 20,126 synonymous variants from 7,665 contigs.

To guide development of three-population models, we first considered simpler two- population models. Initial two-population models without gene flow (modeling allopatric speciation) consistently yielded a larger size for the Sonoran population than the others and a more recent divergence between the Sinaloan and Sonoran populations than between either of those populations and the Mojave. When we added gene flow into the models either as continuous flow (H-MS1, H-SS0, parapatric speciation with continuous contact) or as flow after a period of isolation (H-SS1, cycling periods in refugia followed by introgression during secondary contact), the maximum composite-likelihood estimates for the gene flow parameter either failed to converge or converged to zero. Thus, we found no evidence of gene flow between any pair of the three populations and reject hypotheses H-MS1, H-SS0 and H-SS1.

Based on our two-population analyses, we considered three-population models in which the Sinaloan and Sonoran populations were sisters and there was no gene flow between them. Figure 5 showed the two best-fit models among those we tested (composite log- likelihood: -887 for the 6-parameter model vs. -909 for the 5-parameter model). Qualitatively, these models produced similar allele frequency spectra and residuals when 113

compared with the data (Fig. 5C, D), but the 6-parameter was preferred in a composite likelihood ratio test (adjusted likelihood ratio 9.242, p = 0.0024, chi-squared test with d.f. = 1). To estimate parameter uncertainty while accounting for linkage among variants, we used conventional bootstrapping. Table 5 showed the confidence intervals for the parameters of our 6-parameter demographic model. Our best-fit model suggests that the Mojave and Sinaloan populations have similar effective sizes (128,000 and 150,000 individuals, respectively), but the effective size of the Sonoran population is much larger (600,084 individuals). The two divergence times in our model are also similar (Table 5), suggesting a trichotomy among these populations.

Discussion

Phylogenetic relationships

Consistent with previous findings (Edwards et al. 2012) the matrilineal genealogy depicts the Mojave and Sinaloan lineages as sisters to each other, and both nest within the Sonoran lineage (Fig. 2). This relationship may result from strong, positive selection acting on the locus and driving the Sonoran lineage away from the Mojave/Sinaloan branches, as detected in the PAML analysis. Major differences in the physiological demands on organisms might result in selective advantages. For example, Sinaloan G. morafkai lives in tropical deciduous forest and it exhibits a markedly different thermal ecology (~3 °C lower optimal activity temperature) than desert-living G. agassizii and Sonoran G. morafkai (Lara-Resendiz et al. 2015; P.R. Rosen, University of Arizona, pers. comm.).

This phylogenetic conundrum, however, is resolved by the multi-locus Bayesian species tree analysis, which reduces errors of inference due to lineage sorting (Dupuis et al. 2012). This tree depicts Sonoran and Sinaloan tortoises as sister lineages (Fig. 3), which is expected given their geographic distribution (Fig. 1). While the topology of the species tree is robust, the branch lengths are not likely representative of true divergence times. This is potentially due to the multiple loci exhibiting different mutation rates where the nuclear loci are likely causing a younger date for the desert tortoises and deeper divergence times for the basal taxa. Because the actual mutation rates of these loci cannot easily be estimated in these taxa, we rely primarily on the well-established divergence of G. agassizii and G. morafkai by vicariance of the Bouse inundation, which now forms the Colorado River boundary between the species (Avise et al. 1992). However, mtDNA mutation rates based on this geological event are inconsistent with fossil records of divergence among other congeners and the molecular estimates of divergence within Gopherus may be too recent (Avise et al. 1992; Bramble & Hutchinson 2014). Until a recalibration of the existing molecular clock is performed using distantly related groups,

114

we treat our projected evolutionary rates as conservative estimates (Fig. 2, 3). Regardless, the Sonoran, Sinaloan, and Mojave lineages are genetically distinct.

Patterns of divergence

The RNA-seq analyses generated two very strong results that test our hypotheses. First, the best-fit model in the ∂a∂i analysis for the relationship among the three taxa is concordant with the species tree (Fig. 3) but more clearly elucidates the relative divergence times among the three lineages. This suggests that the Sonoran/Sinaloan split occurred only a short time after (or simultaneous with) the divergence of the Mojave group (Table 5). Thus, the three lineages form a trichotomy with relatively equal divergence from each other (Fig. 5). The ∂a∂i analysis also clearly fails to demonstrate evidence of gene flow during divergence of the Sonoran/Sinaloan lineages. The best fit model (e.g. no migration) does not differ from that obtained for the divergence of

Sonoran/Mojave. Thus we cannot reject H-SS2 that Sonoran and Sinaloan lineages diverged without significant introgression and that divergence was driven in isolation (geographical or ecological) followed by secondary contact, as is currently observed.

Our assumption that signatures of ancient admixture between Sonoran and Sinaloan lineages would be identifiable relies on the likelihood of recurring biogeographic proximity during their evolution and observations of contemporary hybridization. Had hybridization occurred, as it does today, most outcrossed genes would have been purged through backcrossing. Any remaining signature, even if maintained under positive selection, will contribute only a very minor portion of the existing genome (Mendez et al. 2012). Where hybridizing species exhibit ecological segregation, divergent selection acting on specific genomic regions can prevent the homogenizing effects of gene flow (Andrew & Rieseberg 2013, Chapman et al. 2013, de la Torre et al. 2014).

Hybrid zones between parapatric taxa typify repeated population contractions into refugia followed by expansions during climate oscillations (Hewitt 1996). Edwards et al. (in review) performed a clinal analysis of the zone of overlap between the Sonoran and Sinaloan lineages using microsatellite and mtDNA data. They found a bimodal distribution with a strong coincidence of slope and concordance of center between clines, which is supportive of secondary contact (Endler 1977; Barton & Hewitt 1985). They conclude that the current contact zone between the Sonoran and Sinaloan lineages is a result of secondary contact after periods of isolation in Pleistocene refugia. Edwards et al. (in review) suggests that the shifting ecotone between tropical deciduous forest and Sonoran desertscrub likely acted as an ephemeral boundary providing recurring opportunities for interbreeding which may have reinforced niche segregation in each tortoise lineage. Edwards et al. (in review) proposes a mosaic model of ecological

115

segregation (Howard 1986) best describes the observed pattern. They characterize the Sonoran and Sinaloan G. morafkai lineages as having independent evolutionary trajectories despite incomplete reproductive isolation. The data presented here are consistent with these findings and the rejection of H-SS1 suggests that the periods of secondary contact did not result in significant introgression between lineages.

Each pair of desert tortoise lineages has a similar (and simultaneous) process of speciation. Divergence driven by either geography (geographic distance and physical barriers) or selection through ecological niche segregation should reduce gene flow, and both may even occur simultaneously (Endler 1977, Cooke et al. 2014). Gopherus agassizii may have first diverged as a result of isolation, under an allopatric model of speciation. The observed secondary contact zone in northwestern Arizona shows through time it has adapted to the unique environmental conditions of the Mojave Desert (Edwards et al. 2015). Such adaptation may first occur in allopatry and, thus, ecological speciation does not necessarily require sympatry (Bernardi 2013). Geographic isolation may also be a reasonable explanation for the divergence of the Sonoran and Sinaloan lineages, where an allopatric ‘speciation pump’ may have facilitated divergence (April et al. 2013). However, there is little evidence of geographical barriers that would have maintained their long-term isolation over time. Similarly, a peripatric model could be used to explain the divergence of Sonoran and Sinaloan lineages in isolation where a small population is isolated from its source population in refugia and where divergence occurs both through drift and selection. The ∂a∂i analysis, however, suggests that ancestral populations were fairly robust (Na; Table 5) and we do not observe a signature suggesting that one population was ancestrally signigificantly smaller then the other.

An alternative explanation and one that better fits the geographical history of the region is that Sonoran and Sinaloan lineages first diverged into distinct ecotypes under a parapatric model and that isolation in temporary, Pleistocene refugia only reinforced the differentiation that had already occurred (Fisher-Reid et al. 2013). However, in this scenario our findings suggest that the genetic signature of ecological isolation (parapatric model) cannot be differentiated from geographic isolation (allopatric or peripatric model).

It is challenging to distinguish between parapatric speciation and allopatric divergence followed by secondary contact (Pettengill & Moeller 2012), but examples like the apple maggot fly (Rhagoletis pomonella) show that speciation can occur without the complete geographical separation of populations (Jiggins & Bridle 2004). Although many such examples may be facilitated by chromosomal rearrangements, (Drosophila: Noor et al. 2001; Helianthus: Rieseberg 2001), it has also been shown that environmental heterogeneity is an important driver of adaptive divergence. Parapatric speciation has 116

been observed in salamanders (Plethodon cinereus; Fisher-Reid et al. 2013) and differentiation in threespine sticklebacks (Gasterosteus aculeatus) is suggested to be an adaptive response to a heterogeneous environment (DeFaveri et al. 2013). Environmental heterogeneity can even drive adaptive divergence in species where gene flow is not a limiting factor, such as island scrub-jays (Aphelocoma insularis; Langin et al. 2015).

Barriers to reproduction will eventually develop in these taxa. However, common drivers of reproductive isolation such as Haldane’s rule (Haldane 1922) do not apply to tortoises because they do not possess heterogametic sex chromosomes and instead have temperature dependent sex determination. Thus, evolutionarily distinct species may still experience interspecific hybridization (Freedberg & Myers 2012). In the continuum of the process of speciation, the consistency of differentiation across nuclear and mitochondrial loci suggests that tortoises are in the latter stages of speciation vis-a-vie reproductive isolation (Conflitti et al. 2014). Following the evolutionary species concept as elucidated by Wiley (1978), the Sonoran and Sinaloan lineages are on separate evolutionary trajectories but exhibit incomplete reproductive isolation. All three taxa (Mojave, Sonoran, and Sinaloan) are essentially equal in their divergence and exhibit ecological affiliations that maintain their independent evolutionary trajectories.

The sample set utilized for the RNA-seq analyses is small and limited to discrete populations. However, the analysis includes a massive amount of independent gene sequences that allow for the high resolution of evolutionary patterns. Our sampling strategy minimizes geographic bias through equidistant sampling, while maximizing the opportunity to detect introgression in regions of the genome. Short-read technology ensures that our analysis is not limited by the number of loci (Table 4) and we present a high level of resolution beyond what could be inferred through traditional phylogenetic analyses, as comparative analyses exemplify. Importantly, the RNA-seq data effectively tests our hypotheses, and we are confident that these results reflect the evolutionary history of the desert tortoises. Inferences based on large numbers of gene sequences and few individuals have been shown to be robust for inference of population history (Wang & Hey 2010; Lohse et al. 2011; Jones et al. 2012; Hearn et al. 2014). Our limiting the number of individuals and implementing a genomic reduction approach (RNA-seq instead of whole genomes) is a compromise. Notwithstanding, the results approximate the evolutionary relationships among these taxa. Thus, massive parallel sequencing is a cost-effective and informative strategy for working with non-model organisms (Ekblom & Galindo 2011). The reconstructed phylogeny using SNPs has limitations, and yet congruence exists between the next-generation sequencing and more traditional phylogenetic methods.

117

For species of conservation concern, like desert tortoises, it was necessary to sample only a small amount of tissue using non-lethal methods. Thus whole blood provided the most obtainable source (Meitern et al. 2014). It is difficult to assess whether the number of putative transcripts in our analysis or the number of polymorphic sites characterized in this analysis meets expectations since this study focuses on a non-model organism. There are multiple challenges in assembling de novo genomes (McGettigan 2013) and as such we expect a high likelihood of generating technical artifacts. We addressed this through careful screening (UniProt queries) and conservative filtering. Despite this, these data may still contain sequencing and assembly errors. Compared to a transcriptome assembly from whole blood RNA-seq of another non-model organism (greenfinch; Carduelis chloris) we observed almost twice as many functionally annotated transcripts (Meitern et al. 2014). However, for the purposes of this phylogenetic reconstruction, these data appear more than adequate.

Unfortunately, the small sample size of our study limits our ability to perform meaningful outlier tests, selection tests, and descriptive statistics, which could further elucidate the genomic regions driving the observed evolutionary patterns. Our research clarifies the phylogenetic relationships among these taxa and elucidates the relative contribution of isolation and gene flow in the formation of these species. It also opens up many new opportunities for studying the process of speciation in a natural setting. Further research, with the aid of an annotated genome, may investigate the presence of chromosomal rearrangements, which may better explain the mechanism by which distinct species persist despite the potential for hybridization (Kulathinal et al. 2009). Future research might also examine heterogeneity among loci in proportion to levels of divergence, which could shed additional light on the debate over the formation of ‘genomic islands’ in the process of speciation (Feder & Nosil 2010; Nachman & Payseur 2012; Cruickshank & Hahn 2014).

Because uncertainty often occurs when defining species among evolving populations (Hey et al. 2003, Leaché et al. 2009), we delineate these lineages as evolutionarily significant units by using multiple loci, multiple forms of analyses, and incorporating biogeographic and ecological information. By clarifying the taxonomic relationships among these lineages, and the processes that influence their distribution, we hope this study will directly inform efforts to preserve the species’ evolutionary potential. Ultimately, the evolutionary history of desert tortoises not only clarifies the forces that have driven their speciation, but also contributes to our knowledge of the biogeographic history of Southwest deserts and maintenance of diversity within them.

118

Acknowledgements

This international project involved the collaboration of the US Geological Survey, Tucson Herpetological Society, Arizona Game and Fish Department, and the Comisión de Ecologíca y Desarollo Sustentable del Estado de Sonora (CEDES) in Mexico. We obtained Mexican samples with assistance from CEDES and F.R. Méndez de la Cruz, Instituto de Biología, Universidad Nacional Autónoma de México (UNAM). Permits for Mexican samples were facilitated by Secretaría de Medio Ambiente y Recursos Naturales (SEMARNAT). Field collection of samples in Mexico would not have been possible without the help of C. Melendez. P. Rosen, M. Vaughn. A. Karl. P. Woodman, L. Smith, M. Figueroa, Y. Leon and the numerous volunteers that participated in the project. K. Berry, US Geological Survey graciously provided samples collected from California. Many of the Arizona samples were provided by C. Jones and the Arizona Game and Fish Department. A. McLuckie granted use of samples of G. agassizii collected in Utah. B. Brandner provided samples of G. berlandieri. We thank J. Jarchow and the Sonoran Animal Hospital (Tucson, AZ) for their assistance in obtaining RNA samples from G. agassizii. RNA samples collected in Nevada were facilitated through J. Germano, San Diego Zoo’s Desert Tortoise Conservation Center and R. Averill-Murray, US Fish and Wildlife Service. Sample collection in Nevada was assisted by B. Jurand, R. Lamkin, S. Madill, J. Rose and A. Dolan. We thank Arizona Research Laboratories and the University of Arizona Genetics Core (UAGC) for assistance with Sanger and next- generation sequencing. J. Barraza, Y. Jimenez and M. Christinsen assisted with laboratory work. L. Johnstone, S. Miller and N. Merchant were invaluable in their technical assistance with computing and scripting. P. Rosen generated the map. Early conversations with M. Nachman greatly influenced the direction of this research. Finally, we thank the communities of the Tucson Herpetological Society and the Desert Tortoise Council for their inspiration and support. Funding for this project was obtained from the Royal Ontario Museum Foundation, NSERC Discovery Grant A3148, Arizona Research Laboratories, Tucson Herpetological Society, Desert Tortoise Council, and Fisher Scientific through their Scholar of Excellence Award with matching funds received from the School of Natural Resources and the Environment, University of Arizona. RNG and PH were supported by NSF grant DEB-1146074.

References

Andrew RL, Rieseberg LH (2013) Divergence is focused on few genomic regions early in speciation: incipient speciation of sunflower ecotypes. Evolution, 67, 2468– 2482.

119

April J, Hanner RH, Dion-Cote AM, Bernatchez L (2013) Glacial cycles as an allopatric speciation pump in north-eastern American freshwater fishes. Molecular Ecology, 22, 409–422.

Arnold ML (2007) Evolution through genetic exchange. Oxford University Press, London.

Avise JC, Bowen BW, Lamb T, Meylan AB, Bermingham E (1992) Mitochondrial DNA evolution at a turtle's pace: Evidence for low genetic variability and reduced microevolutionary rate in the Testudines. Molecular Biology and Evolution, 9, 457–473.

Barton NH, Hewitt GM (1985) Analysis of hybrid zones. Annual Review of Ecology and Systematics, 16, 113–148.

Barton NH, Hewitt GM (1989) Adaptation, speciation, and hybrid zones. Nature, 341, 497–503.

Bernardi G (2013) Speciation in fishes. Molecular Ecology, 22, 5487–5502.

Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer for Illumina Sequence Data. Bioinformatics, btu170.

Bouckaert RR, Heled J, Kühnert D, Vaughan T, Wu C-H, Xie D, Suchard MA, Rambaut A, Drummond A J (2014). BEAST 2: a software platform for Bayesian evolutionary analysis. PLoS Computational Biology, 10, e1003537.

Bouckaert RR, Heled J (2014) DensiTree 2: Seeing trees through the forest. bioRxiv, http://dx.doi.org/10.1101/012401

Bramble DM, Hutchison JH (2014) Morphology, taxonomy, and distribution of North American tortoises; An evolutionary perspective. In: Biology and Conservation of North American Tortoises (eds Rostal DC, McCoy ED, Mushinsky HR). pp. 1– 12. Johns Hopkins University Press, Baltimore, MD, USA.

Brito PH, Edwards SV (2009) Multilocus phylogeography and phylogenetics using sequence-based markers. Genetica, 135, 439–455.

Brown DE, Lowe CH (1980) Biotic communities of the southwest. Rocky Mountain Forest and Range Experiment Station, U.S. Forest Service. General Technical Report RM-78. Map.

120

Búrquez A, Martínez-Yrízar A, Felger RS, Yetman D (1999) Vegetation and habitat diversity at the southern edge of the Sonoran Desert. In: Ecology of Sonoran Desert Plants and Plant Communities (ed Robichaux RH), pp 36–67. University of Arizona Press, Tucson AZ.

Cahais V, Gayral P, Tsagkogeorga G, Melo-Ferreira J, Ballenghien M, Weinert L, Chiari Y, Belkhir K, Ranwez V, Galtier N (2012) Reference-free transcriptome assembly in non-model animals from next-generation sequencing data. Molecular Ecology Resources, 12, 834–845.

Chapman MA, Hiscock SJ, Filatov DA (2013) Genomic divergence during speciation driven by adaptation to altitude. Molecular Biology and Evolution, 30, 2553– 2567.

Conflitti IM, Shields GF, Murphy RW, Currie DC (2014) The speciation continuum: population structure, gene flow, and maternal ancestry in the Simulium arcticum complex (Diptera: Simuliidae). Molecular Phylogenetics and Evolution, 78, 43– 55.

Cooke GM, Landguth EL, Beheregaray LB (2014) Riverscape genetics identifies replicated ecological divergence across an Amazonian ecotone. Evolution, 68, 1947–1960.

Coyne JA, Orr HA (2004) Speciation. Sinauer Associates, Inc. Sunderland, MA.

Cruickshank TE, Hahn MW (2014) Reanalysis suggests that genomic islands of speciation are due to reduced diversity, not reduced gene flow. Molecular Ecology, 23, 3133–3157. de la Torre AR, Roberts DR, Aitken SN (2014) Genome-wide admixture and ecological niche modeling reveal the maintenance of species boundaries despite long history of interspecific gene flow. Molecular Ecology, 23, 2046–2059.

De Wit P, Pespeni MH, Ladner JT, Barshis DJ, Seneca F, Jaris H, Overgaard Therkildsen N, Morikawa M, Palumbi SR (2012) The simple fool's guide to population genomics via RNA-Seq: an introduction to high-throughput sequencing data analysis. Molecular Ecology Resources, 12, 1058–1067.

DeFaveri J, Jonsson PR, Merila J (2013) Heterogeneous genomic differentiation in marine threespine sticklebacks: adaptation along an environmental gradient Evolution 67, 2530–2546.

121

Degnan JH, Rosenberg NA (2009) Gene tree discordance, phylogenetic inference and the multispecies coalescent. Trends in Ecology and Evolution, 24, 332–340.

DePristo M, Banks E, Poplin R, Garimella K, Maguire J, Hartl C, Philippakis A, del Angel G, Rivas MA, Hanna M, McKenna A, Fennell T, Kernytsky A, Sivachenko A, Cibulskis K, Gabriel S, Altshuler D, Daly M (2011) A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nature Genetics, 43, 491-498.

Doran AG, Creevey CJ (2013) Snpdat: Easy and rapid annotation of results from de novo snp discovery projects for model and non-model organisms. BMC Bioinformatics, 14, 45.

Drummond AJ, Ho SYW, Phillips MJ, Rambaut A (2006) PLoS Biology, 4, e88.

Dupuis JR, Roe AD, Sperling FAH (2012) Multi-locus species delimitation in closely related animals and fungi: one marker is not enough. Molecular Ecology, 21, 4422–4436.

Edwards T (2003) Desert tortoise conservation genetics, Unpublished M.S. Thesis, University of Arizona, Tucson, USA.

Edwards T, Schwalbe CR, Swann DE, Goldberg CS (2004) Implications of Anthropogenic Landscape Change on Inter-Population Movements of the Desert Tortoise (Gopherus agassizii). Conservation Genetics, 5, 485–499.

Edwards T, Jarchow CJ, Jones CA, Bonine KE (2010). Tracing Genetic Lineages of Captive Desert Tortoises in Arizona. Journal of Wildlife Management, 74, 801– 807.

Edwards T, Vaughn M, Meléndez-Torres MaC, Karl AE, Rosen PR, Berry KH. Murphy RW (2012) A biogeographic perspective of speciation among desert tortoises in the genus Gopherus: A preliminary evaluation. In: Merging science and management in a rapidly changing world: biodiversity and management of the Madrean Archipelago III, 2012 May 1-5, Tucson, AZ (eds Gottfried GJ, Ffolliott PF, Gebow BS, Eskew LG) pp. 243–247. Proceedings RMRS-P-67. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station.

Edwards T, Harrison S (2014) Population and conservation genetics. In: Biology and Conservation of North American Tortoises (eds Rostal DC, McCoy ED, Mushinsky HR) pp. 127–133. Johns Hopkins University Press, Baltimore, MD.

Edwards T, Cox EC, Buzzard V, Wiese C, Hillard LS, Murphy RW (2014) Genetic assessments and parentage analysis of captive Bolson tortoises (Gopherus flavomarginatus) inform their "rewilding" in New Mexico. PLoS ONE, 9, e102787.

Edwards T, Berry KH, Inman RD, Esque TC, Nussear KE, Jones CA, Culver M (2015) Testing taxon tenacity of tortoises: evidence for a geographical-selection gradient at a secondary contact zone. Ecology and Evolution, in press,

Edwards T, Vaughn M, Rosen PC, Meléndez-Torres C, Karl AE, Culver M, Murphy RW (in review). Shaping species with ephemeral boundaries; the distribution and genetic structure of the desert tortoise (Gopherus morafkai) in the Sonoran Desert. Journal of Biogeography, Draft manuscript for peer review.

Ekblom R, Galindo J (2011) Applications of next generation sequencing in molecular ecology of non-model organisms. Heredity, 107, 1–15.

Endler JA (1977) Geographic variation, speciation, and clines. Princeton University Press, Princeton, N.J. USA.

Feder JL, Nosil P (2010) The efficacy of divergence hitchhiking in generating genomic islands during ecological speciation. Evolution, 64, 1729–1747.

Felger RS, Austin DF, Van Devender TR, Sánchez-Escalante J, Costea M (2012) Convolvulaceae of Sonora, Mexico – I. Convolvulus, Cressa, Dichondra, Evolvulus, Ipomoea, Jacquemontia, Merremia, Operculina. Journal of the Botanical Research Institute of Texas, 6, 459–527.

Freedberg S, Myers EM (2012) Cytonuclear equilibrium following interspecific introgression in a turtle lacking sex chromosomes. Biological Journal of the Linnean Society, 106, 405–417.

123

Fujita MK, Engstrom TN, Starkey DE, Shaffer HB (2004) Turtle phylogeny: insights from a novel nuclear intron. Molecular Phylogenetics and Evolution, 31, 1031– 1040.

Geraldes A, Ferrand N, Nachman MW (2006) Contrasting patterns of introgression at X- linked loci across the hybrid zone between subspecies of the European rabbit (Oryctolagus cuniculus). Genetics, 173, 919–933.

Germano DJ, Bury RB, Esque TC, Fritts TH, Medica PA (1994) Range and habitats of the desert tortoise. In: Biology of North American Tortoises (eds Bury RB, Germano DJ) pp. 73–84. U.S. Department of Interior National Biological Survey, Fish and Wildlife Research 13.

Grabherr MG, Haas BJ, Yassour M, et al. (2011) Full-length transcriptome assembly from RNA-seq data without a reference genome. Nature Biotechnology, 29, 644– 52.

Gutenkunst RN, Hernandez RD, Williams SH, Bustamante CD (2009) Inferring the joint demographic history of multiple populations from multidimensional SNP frequency data. PLoS Genetics, 5, e1000695.

Haas BJ, Papanicolaou A, Yassour M, et al. (2013) De novo transcript sequence reconstruction from RNA-seq using the Trinity platform for reference generation and analysis. Nature Protocols, 8, 1494–512.

Hagerty BE, Nussear KE, Esque TC, Tracy CR (2011) Making molehills out of mountain landscape genetics of the Mojave desert tortoise. Landscape Ecology, 2, 267–280.

Haldane JBS (1922) Sex ratio and unisexual sterility in hybrid animals. Journal of Genetics, 12, 101–109.

Hearn J, Stone GN, Bunnefeld L, Nicholls JA, Barton NH, Lohse K (2014) Likelihood- based inference of population history from low-coverage de novo genome assemblies. Molecular Ecology, 23, 198–211.

Heled J, Drummond AJ (2010) Bayesian inference of species trees from multilocus data. Molecular Biology and Evolution, 27, 570–580.

Hewitt GM (1996) Some genetic consequences of ice ages, and their role, in divergence and speciation. Biological Journal of the Linnean Society, 58, 247–276.

124

Hey J, Waples RS, Arnold ML, Butlin RK, Harrison RG (2013) Understanding and confronting species uncertainty in biology and conservation. Trends in Ecology and Evolution, 78, 597–603.

Howard DJ (1986) A zone of overlap and hybridization between 2 ground cricket species. Evolution, 40, 34-43.

Hudson RR, Turelli M (2003) Stochasticity overrules the "three-times rule": genetic drift, genetic draft, and coalescence times for nuclear loci versus mitochondrial DNA. Evolution, 57, 182–190.

Jiggins CD, Bridle JR (2004) Speciation in the apple maggot fly: a blend of vintages? TRENDS in Ecology and Evolution, 19, 111–114.

Jones FC, Grabherr MG, Chan YF, et al. (2012) The genomic basis of adaptive evolution in threespine sticklebacks. Nature, 484, 55–61.

Kulathinal RJ, Stevison LS, Noor MAF (2009) The genomics of speciation in Drosophila: Diversity, divergence, and introgression estimated using low- coverage genome sequencing. PLoS Genetics, 5, e1000550.

Lamb T, Avise JC, Gibbons JW (1989) Phylogeographic patterns in mitochondrial DNA of the desert tortoise (Xerobates agassizii), and evolutionary relationships among the North American gopher tortoises. Evolution, 43, 76–87.

Lamb T, Lydeard C (1994) A molecular phylogeny of the gopher tortoises, with comments on familial relationships within the Testudinoidea. Molecular Phylogenetics and Evolution, 3, 283–291.

Langin KM, Sillett S, Funk WC, Morrison SA, Desrosiers MA, Ghalambor CK (2015) Islands within an island: Repeated adaptive divergence in a single population. Evolution, 69, 653–665.

Lara-Resendiz RA, Villa-Corella LC, Sinervo B, Rosen PC, Karl AE, Torres CM, de la Cruz FRM, Vaughn ML (2015) Temperature preferences of Gopherus morafkai in tropical dry forest. In: 40th annual meeting and symposium of the desert tortoise council. February 20-22, 2015, Poster. Las Vegas, Nevada.

Leaché AD, McGuire JA (2006) Phylogenetic relationships of horned lizards (Phrynosoma) based on nuclear and mitochondrial data: evidence for a misleading mitochondrial gene tree. Molecular Phylogenetics and Evolution, 39, 628–644.

125

Leaché AD, Koo MS, Spencer CL, Papenfuss TJ, Fisher RN, McGuire JA (2009) Quantifying ecological, morphological, and genetic variation to delimit species in the coast horned lizard species complex (Phrynosoma). Proceedings of the National Academy of Sciences USA, 106, 12418–12423.

Leaché AD, Rannala B (2011) The accuracy of species tree estimation under simulation: a comparison of methods. Systematic Biology, 60, 126–137.

Li H, Durbin R (2009) Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics, 25, 1754–1760.

Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R, and 1000 Genome Project Data Processing Subgroup (2009) The sequence alignment/map (SAM) format and SAMtools. Bioinformatics, 25, 2078– 2079.

Librado P, Rozas J (2009) DnaSP v5: a software for comprehensive analysis of DNA polymorphism data. Bioinformatics, 25, 1451–1452.

Lohse K, Harrison RJ, Barton NH (2011) A general method for calculating likelihoods under the coalescent process. Genetics, 58, 977–987.

Magrane M, UniProt Consortium (2011) UniProt Knowledgebase: a hub of integrated protein data. Database: The Journal of Biological Databases and Curation. doi: 10.1093/database/bar009.

McCormack JE, Huang H, Knowles LL (2009) Maximum likelihood estimates of species trees: How accuracy of phylogenetic inference depends upon the divergence history and sampling design. Systematic Biology, 58, 501–508.

McGettigan PA (2013) Transcriptomics in the RNA-seq era. Current Opinion in Chemical Biology 17, 4–11.

McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, Altshuler D, Gabriel S, Daly M, DePristo MA (2010) The genome analysis toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Research, 20, 1297–1303.

126

Meitern R, Andreson R, Horak P (2014) Profile of whole blood gene expression following immune stimulation in a wild passerine. BMC Genomics, 15, 533.

Melo-Ferreira J, Alves PC, Freitas H, Ferrand N, Boursot P (2009) The genomic legacy from the extinct Lepus timidus to the three hare species of Iberia: contrast between mtDNA, sex chromosomes and autosomes. Molecular Ecology, 18, 2643–2658.

Morafka DJ, Aguirre G, Murphy RW (1994) Allozyme differentiation among gopher tortoises (Gopherus) - conservation genetics and phylogenetic and taxonomic implications. Canadian Journal of Zoology, 72, 1665–1671.

Murphy RW, Berry KH, Edwards T, McLuckie AM (2007) A genetic assessment of the recovery units for the Mojave population of desert tortoises, Gopherus agassizii. Chelonian Conservation and Biology, 6, 229–251.

Nachman MW, Payseur BA (2012) Recombination rate variation and speciation: theoretical predictions and empirical results from rabbits and mice. Philosophical Transactions of the Royal Society B: Biological Sciences, 367, 409–421.

Noor MAF, Grams KL, Bertucci LA, Reiland, J (2001) Chromosomal inversions and the reproductive isolation of species. Proceedings of the National Academy of Sciences USA, 98, 12084–12088.

Nylander JAA (2004) MrModeltest v2. Program distributed by the author. Evolutionary Biology Centre, Uppsala University.

Patterson N, Price A, Reich D (2006) Population structure and eigenanalysis. PLoS Genetics, 2, e190.

Payseur BA (2010) Using differential introgression in hybrid zones to identify genomic regions involved in speciation. Molecular Ecology, 10, 806–820.

127

Payseur BA, Krenz JG, Nachman MW (2004) Differential patterns of introgression across the X chromosome in a hybrid zone between two species of house mice. Evolution, 58, 2064–2078.

Pettengill JB, Moeller DA (2012) Phylogeography of speciation: allopatric divergence and secondary contact between outcrossing and selfing Clarkia. Molecular Ecology, 21, 4578–4592.

Pinho C, Hey J (2010) Divergence with gene flow: models and data. Annual Review of Ecology Evolution and Systematics, 41, 215–230.

Pollard DA, Iyer VN, Moses AM, Eisen MB (2006) Widespread discordance of gene trees with species tree in Drosophila: evidence for incomplete lineage sorting. PLoS Genetics, 2, e173.

Rambaut A (2006-2012) FigTree; tree figure drawing tool. Institute of Evolutionary Biology, University of Edinburgh.

Rambaut A, Drummond AJ (2002-2012) TreeAnnotator; MCMC output analysis. Institute of Evolutionary Biology, University of Edinburgh.

Rambaut A, Suchard M, Xie W, Drummond AJ (2003-2013) Tracer: MCMC trace analysis tool. Institute of Evolutionary Biology, University of Edinburgh.

Rieseberg LH (2001) Chromosomal rearrangements and speciation. Trends in Ecology & Evolution, 16, 351–357.

Spinks PQ, Shaffer HB, Iverson JB, McCord WP (2004) Phylogenetic hypotheses for the turtle family Geoemydidae. Molecular Phylogenetics and Evolution, 32, 164–182.

Stephens M, Donnelly P (2003). A comparison of Bayesian methods for haplotype reconstruction from population genotype data. American Journal of Human Genetics, 73, 1162–1169.

Swofford DL (2002) PAUP* Phylogenetic analysis using parsimony (*and other methods). Version 4. Sinauer Associates, Sunderland, Massachusetts.

Teeter KC, Payseur BA, Harris LW, Bakewell MA, Thibodeau LM, O'Brien JE, Krenz JG, Sans-Fuentes MA, Nachman MW, Tucker PK (2008) Genome-wide patterns of gene flow across a house mouse hybrid zone. Genome Research, 18, 67–76.

128

Thomson RC, Shedlock AM, Edwards SV, Shaffer HB (2008) Developing markers for multilocus phylogenetics in non-model organisms: a test case with turtles. Molecular Phylogenetics and Evolution, 49, 514–525.

Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ, Salzberg SL, Wold BJ, Pachter L (2010) Transcript assembly and quantification by RNA-seq reveals unannotated transcripts and isoform switching during cell differentiation. Nature Biotechnology, 28, 511–515.

U.S. Fish and Wildlife Service (1994) Desert Tortoise (Mojave Population) Recovery Plan. U.S. Fish and Wildlife Service, Portland, Oregon, USA.

U.S. Fish and Wildlife Service (2011) Revised recovery plan for the Mojave population of the desert tortoise (Gopherus agassizii). U.S. Fish and Wildlife Service, Pacific Southwest Region, Sacramento, California. USA.

Van Devender TR (2000) The deep history of the Sonoran Desert. In: A Natural History of the Sonoran Desert (eds Phillips SJ, Comus PW), pp. 61–69. University of California Press, Berkeley, CA.

Via S (2009) Natural selection in action during speciation. Proceedings of the National Academy of Sciences, 106, 9939–9946.

Wang Y, Hey J (2010) Estimating divergence parameters with small samples from a large number of loci. Genetics, 184, 363–379.

Wiley EO (1978) The evolutionary species concept reconsidered. Systematic Biology, 27, 17–26.

Williamson SH, Hernandez R, Fledel-Alon A, Zhu L, Nielsen R, Bustamante CD (2005) Simultaneous inference of selection and population growth from patterns of variation in the human genome. Proceedings of the National Academy of Sciences of the United States of America, 102, 7882–7887.

Yang Z (2007) PAML 4: a program package for phylogenetic analysis by maximum likelihood. Molecular Biology and Evolution, 24, 1586–1591.

Zhang LX (2011) From gene trees to species trees II: Species tree inference by minimizing deep coalescence events. IEEE-ACM Transactions on Computational Biology and Bioinformatics, 8, 1685–1691.

129

Table 1. Desert Tortoise sample locality information. Site ID corresponds with Figure 1; n = number of individuals sampled. Biotic community descriptions: TDF = Tropical Deciduous Forest, STS = Sinaloan Thornscrub, SDS = Sonoran Desertscrub and MDS = Mojave Desertscrub. Asterisk (*) indicates sites sampled for RNA-seq (ALS samples used in both DNA and RNA analyses).

Site ID n Taxa Biotic community Location Site UT 2 Goag- Northern Mojave Desertscrub Utah, USA Near St. George TC* 2 Goag- Northern Mojave Desertscrub Nevada, USA Trout Canyon SV 2 Goag- Northern Mojave Desertscrub California, USA Shadow Valley IV 2 Goag- Northern Mojave Desertscrub California, USA Ivanpah KH 1 Goag- Western Mojave Desertscrub California, USA Kramer Hills FIC 2 Goag- Western Mojave Desertscrub California, USA Fort Irwin ORC 1 Goag- Western Mojave Desertscrub California, USA Ord-Rodman Marine Corps Air MCSH 1 Goag- Western Mojave Desertscrub California, USA Ground Combat Center Marine Corps Air MCB 1 Goag- Western Mojave Desertscrub California, USA Ground Combat Center Chocolate Mountain CM 1 Goag- Western Lower Colorado River Valley California, USA Aerial Gunnery Range FEN 1 Goag- Western Mojave Desertscrub California, USA Fenner G 1 Goag- Western Mojave Desertscrub California, USA Goffs CH 1 Goag- Western Mojave Desertscrub California, USA Chemehuevi UWV 2 Goag- Western Mojave Desertscrub California, USA Upper Ward Valley Southwest side of GS 2 Goag- Western Ecotone: SDS/MDS Arizona. USA Black Mtns West side of Black WBM 4 Admixed - MOJ/SON Ecotone: SDS/MDS Arizona. USA Mtns East Bajada Long EB 4 Admixed - MOJ/SON Ecotone: SDS/MDS Arizona. USA Term Monitoring Plot HSC 1 Admixed - MOJ/SON Ecotone: SDS/MDS Arizona. USA Hualapai Mtns HAR 2 Gomo - SON SDS: Arizona Upland Arizona. USA Harcuvar Mtns. WM 4 Gomo - SON SDS: Arizona Upland Arizona. USA Wickenburg Mtns. NW 1 Gomo - SON Ecotone: SDS/Lower Colorado River Valley Arizona. USA New Water Mtns. EAG 2 Gomo - SON Ecotone: SDS/Lower Colorado River Valley Arizona. USA Eagletail Mtns. SAU 2 Gomo - SON Ecotone: SDS/Lower Colorado River Valley Arizona. USA Sauceda Mtns. region Sugarloaf, Mazatzal SL 3 Gomo - SON SDS: Arizona Upland Arizona. USA Mtns GRH 2 Gomo - SON SDS: Arizona Upland Arizona. USA Granite Hills Saguaro National Park RK 2 Gomo - SON SDS: Arizona Upland Arizona. USA and adjacent land SP 2 Gomo - SON SDS: Arizona Upland Arizona. USA San Pedro Valley 130

Site ID n Taxa Biotic community Location Site CAN 2 Gomo - SON Ecotone: SDS/Lower Colorado River Valley Sonora, Mexico La Candelaria BAM 2 Gomo - SON Ecotone: SDS/Lower Colorado River Valley Sonora, Mexico Bamuri TIB 4 Gomo - SON SDS: Central Gulf Coast Sonora, Mexico Tiburon Island SER 2 Gomo - SON SDS: Central Gulf Coast Sonora, Mexico Seri region RSJ* 2 Gomo - SON SDS: Plains of Sonora Sonora, Mexico San Judas RM 3 Admixed - SON/SIN Ecotone: STS/SDS Sonora, Mexico Moscobampo RAS 3 Admixed - SON/SIN Sinaloan Thornscrub Sonora, Mexico Arroyo Seco Opodepe-Las Milpas OPO 4 Admixed - SON/SIN Sinaloan Thornscrub Sonora, Mexico region REA 3 Gomo - SIN Ecotone: TDF/STS Sonora, Mexico El Alamo RLN 2 Gomo - SIN Ecotone: TDF/STS Sonora, Mexico La Noria ALC 2 Gomo - SIN Tropical Deciduous Forest Sonora, Mexico Alamos-Las Cabras ALS* 2/2 Gomo - SIN Tropical Deciduous Forest Sonora, Mexico Alamos-La Sierrita FUE 4 Gomo - SIN Tropical Deciduous Forest Sinaloa, Mexico Rio Fuerte

131

Table 2. Summary of one mtDNA and four nDNA loci and their amplification conditions used for phylogenetic analysis of 86 tortoises in the genus Gopherus.

# of Length # variable MgCl2 Annealing Locus haplotypes (bp) sites (mM) temp °C Citation mtDNA Edwards 2003, Edwards (ND3/ND4) 18 1109 190 4 52 et al. 2014 BDNF 6 640 5 2 57 Leaché & McGuire 2006 R35 13 500 17 4 53.4 *Spinks et al. 2004 TB02 16 425 13 1 59 Thomson et al. 2008 TB07 13 590 10 1.5 58 Thomson et al. 2008 *Original primers from Fujita et al. (2004)

132

Table 3. Likelihood Ratio Tests for Selection Pressure estimated among mtDNA haplotypes from tortoises in the genus Gopherus.

a b Models Selection Models np lnL ω (dN/dS) P values compared A. All branches have one ω 9 -2463.7 ω = 0.62

B. All branches have the same ω=1 8 -2469.4 ω=1 A vs. B P < 0.001 C. Each branch has its own ω 15 -2454.6 variable ω by branch A vs. C P < 0.01 aNumber of parameters. bThe natural logarithm of the likelihood value.

133

Table 4. Summary of RNA-seq and reference transcriptome assembly results for three individuals representing each lineage of desert tortoise.

No. Mean contig Trinity Assembly No. Reads Transcripts N50 length (bp) G. agassizii (Mojave) 44,068,129 150,135 1,190 709 G. morafkai (Sonoran) 22,528,007 202,778 2,500 1,093 G. morafkai (Sinaloan) 45,039,615 138,380 1,994 899 Combined assembly 111,635,751 235,412 1,302 718 Combined filtered assembly -- 40,341 3,010 1,957

134

Table 5. Maximum composite likelihood parameter estimates and confidence intervals for the best-fit 6-parameter demographic model. Parameters are effective population sizes (N) in individuals and times of divergence (T) in years.

Demographic parameter Estimate 95% C.I.

Na: size of Gopherus ancestral population 336,200 328,000–344,000

Nmoj: size of contemporary Mojave population 128,400 122,000–135,000

Nsin-son: size of contemporary Sinaloan population 149,600 143,000–156,000

Nson: size of contemporary Sonoran population 600,000 548,000–668,000

Tdiv-1: time of Mojave divergence 5,900,000 5,597,000–6,183,000

Tdiv-2: time of Sinaloan and Sonoran divergence 5,650,000 5,376,000–5,967,000

135

Figure 1. Map of desert tortoise sampling locations where Site IDs correspond with Table 1. DNA samples obtained from locations marked with a circle; RNA samples obtained from sites marked with a triangle. Habitat distribution estimated from (Brown & Lowe 1980) and by digitizing published maps in Burquez et al. (1999) and Felger et al. (2012). Desert tortoise range limit from Germano (1994).

136

Figure 2. Bayesian phylogenetic reconstructions of representative nuclear and mtDNA haplotypes using BEAST; Colored branches indicate haplotypes fixed in designated lineages. A: mtDNA nodes labeled with estimated time to most recent common ancestor (TMRCA) in millions of years. Bootstrap values in parentheses estimated from parsimony reconstruction B-E: unrooted nDNA haplotype networks.

137

138

Figure 3. Estimated species tree from multilocus data using *BEAST reconstructed from mtDNA and 4 nDNA loci from 86 tortoises in the genus Gopherus; G. agassizii, Mojave lineage; G. morafkai Sonoran and Sinaloan lineages; G. berlandieri; G. flavomarginatus. Generated from 500 million iterations using a strict clock; A. Maximum clade credibility tree with common ancestor node heights with nodes representing relative TMRCA, bars corresponding with standard deviations and branches labeled with posterior probabilities, and B. DENSITREE visualization of consensus trees.

139

B. DensiTree.

140

Figure 4. Principle Components Analysis of 95,220 polymorphic sites in the reference transcriptome among six individuals representing the three lineages of desert tortoise; Mojave, Sonoran, and Sinaloan.

141

Figure 5. Best-fit demographic models and observed and predicted frequency spectra for Mojave, Sinaloan, and Sonoran desert tortoise populations. (A) Simpler model fit with the five free parameters labeled. (B) More complex model, with the 6 free parameters labeled. (C,D) The marginal spectra for each pair of populations. Row one is data, rows two and four are models, and rows three and five are Anscombe residuals of model minus data. Ns represent effective population sizes and Ts represent times of population divergence.

A C Tdiv-2

Nson Sonoran Tdiv-1

Na-div

Na-div Sinaloan

1 N Mojave 4 a-div

B Tdiv-2 N son Sonoran D Tdiv-1

Nsin-son

Nsin-son Sinaloan

Mojave Nmoj

Supporting Information

Table S1. Haplotype summary of mtDNA and 4 nDNA loci. Taxa grouped by species and geographic distribution, where n = number of individuals sequenced. Parenthetical values represent number of haplotypes assigned for individuals with ambiguous haplotype determination.

A. mtDNA Haplotypes (with GenBank accession numbers)

Taxa

519

DQ649409.1 flavomarginatus G. DQ649408.1 MOJ_A01 DQ649394.1 MOJ_B01 DQ649398.1 MOJ_B02 DQ649399.1 MOJ_B03 DQ649400.1 SON_01 DQ649401.1 SON_04 DQ649404.1 SON_05 DQ649405.1 SON_08 KM411515 SON_11 KM411518 SON_12 KM411 SON_B01 KM411521 SIN_A01 KM411506 SIN_A02 KM411507 SIN_B01 KM411510 SIN_B03 KM411512 SIN_B04 KM411513 n berlandieri G. G. flavomarginatus 2 2 G. berlandieri 2 2 G. agassizii - Western 12 12 G. agassizii - Northern 6 3 1 2 G. agassizii - Arizona 10 10 G. morafkai - Sonoran, Arizona 21 16 1 4 G. morafkai - Sonoran, Mexico 10 6 1 2 1 G. morafkai - Sonoran, ecotone 6 5 1 G. morafkai - Sinaloan, ecotone 7 6 1 G. morafkai - Sinaloan 10 4 2 2 1 1 Total 86 2 2 22 3 1 2 27 1 4 1 2 1 1 10 3 2 1 1

143

B. R35 Haplotype

1 2 3 4 5 6 7 8 9

11535

0 0 0 0 0 0 0 0

KM411530 Hap KM411531 Hap KM411532 Hap KM411533 Hap KM411534 Hap KM4 Hap KM411536 Hap KM411537 Hap KM411538 11 Hap KM411540 12 Hap KM411541 13 Hap KM411542 Taxa n Hap G. flavomarginatus 1 1 1 G. berlandieri 1 2 G. agassizii - Western 10 1 18 1 G. agassizii - Northern 6 2 2 4 4 G. agassizii - Arizona 10 3 15 1 1 G. morafkai - Sonoran, Arizona 21 30 7 3 2

4 G. morafkai - Sonoran, Mexico 9 9 6 2 1

4 G. morafkai - Sonoran, ecotone 6 4 2 3 3 G. morafkai - Sinaloan, ecotone 6 5 1 1 4 1 G. morafkai - Sinaloan 9 3 1 2 9 3 Total 79 54 7 13 3 41 8 1 16 2 1 1 11

C. BDNF Haplotype

3 4 5 6

0 0 0

02 0

KM41152 Hap KM411526 Hap KM411527 Hap KM411528 Hap KM411529 Taxa n Hap G. flavomarginatus 2 4 G. berlandieri 2 3 1 G. agassizii - Western 9 8 6 4

G. agassizii - Northern 4 1 7

144

G. agassizii - Arizona 10 7 9 4 G. morafkai - Sonoran, Arizona 21 10 32 C. BDNF (cont.) Haplotype G. morafkai - Sonoran, Mexico 9 6 12 G. morafkai - Sonoran, ecotone 6 10 2 G. morafkai - Sinaloan, ecotone 6 1 9 2 G. morafkai - Sinaloan 10 3 14 3 Total 79 23 98 7 22 8

D. TB02 Haplotype

Taxa

1 2 3 4 5 6 7 8 9

0 0 0 0

0 0 0 0 0

4 p 12 p

Hap KM411556 Hap KM411557 Hap KM411558 Hap KM411559 Hap KM411560 Hap KM411561 Hap KM411562 Hap KM411563 Hap KM411564 Ha KM411567 13 Hap KM411568 14 Hap KM411569 15 Hap KM411570 16 Hap KM411571 n G. flavomarginatus 2 1 3 G. berlandieri 2 1 1 1 1 G. agassizii - Western 8 9 3 1 2 1 G. agassizii - Northern 6 5 4 2 1 G. agassizii - Arizona 10 14 1 1 4 G. morafkai - Sonoran, Arizona 21 37 5 G. morafkai - Sonoran, Mexico 9 16 2 G. morafkai - Sonoran, ecotone 6 10 1 1 G. morafkai - Sinaloan, ecotone 6 10 2 G. morafkai - Sinaloan 10 16 2 2 Total 80 118 1 1 1 5 5 8 3 6 2 5 1 3 1

145

E. TB07 Haplotype

552

1 2 3 4 5 6 7 8 9

0 0 0 0 0 0 0

0 0

KM411543 Hap KM411544 Hap KM411545 Hap KM411546 Hap KM411547 Hap KM411548 Hap KM411549 Hap KM411550 Hap KM411551 10 Hap KM411 11 Hap KM411553 12 Hap KM411554 13 Hap KM411555 Taxa n Hap G. flavomarginatus 2 2 2 G. berlandieri 2 2 2 G. agassizii - Western 5 1 4 (3) 2 3(3) G. agassizii - Northern 2 2(2) 2(2) G. agassizii - Arizona 7 3(2) 1 4 6(2)

4 G. morafkai - Sonoran, Arizona

6 18 18(5) 13 5(5)

G. morafkai - Sonoran, Mexico 8 1 7 7 1 G. morafkai - Sonoran, ecotone 6 5(1) 1(1) 5(1) 1(1) G. morafkai - Sinaloan, ecotone 6 4 2(1) 1 5(1) G. morafkai - Sinaloan 10 4 6(3) 3 7(3) Total 66 2 2 2 47 9 4 38 2 2 2 1 4 17

146

Table S2. Overview of DNA polymorphism for mtDNA and 4 nDNA loci estimated for tortoises in the genus Gopherus. bp = locus length in base pairs; Pop = putative population; S = number of polymorphic (segregating) sites; Hn = number of haplotypes; Hd = haplotype diversity; π = nucleotide diversity (per site); θ = theta (per site). Bold indicates significance value of p ≤ 0.05.

Locus bp Pop N S Hn Hd π θ Tajima's D mtDNA 1108 G. flavomarginatus 2 0 0 0 0 0 - G. berlandieri 2 0 0 0 0 0 - G. agassizii - Mojave 28 8 4 0.378 0.0021 0.0019 0.3732 Goag- Northern 6 2 3 0.733 0.0008 0.0008 -0.0500 Goag- Western 22 0 0 0 0 0 - G. morafkai 54 63 10 0.658 0.0211 0.0125 0.0125 Gomo - Sonoran 37 9 5 0.338 0.0006 0.0091 -2.1008 Gomo - Sinaloan 17 6 5 0.640 0.0013 0.0016 -0.5749 desert tortoises 82 89 14 0.782 0.0303 0.0163 2.8676

BDNF 640 G. flavomarginatus 4 0 1 0 0 0 - G. berlandieri 4 1 2 0.500 0.0008 0.0009 -0.6124 G. agassizii - Mojave 46 2 3 0.634 0.0012 0.0007 1.2103 Goag- Northern 8 1 2 0.250 0.0004 0.0004 -1.0548 Goag- Western 38 2 3 0.661 0.0013 0.0007 1.4323 G. morafkai 104 2 3 0.414 0.0007 0.0006 0.2271 Gomo - Sonoran 72 2 3 0.393 0.0006 0.3930 -0.0291 Gomo - Sinaloan 32 2 3 0.458 0.0008 0.0008 0.0055 All desert tortoises 150 4 5 0.575 0.0012 0.0011 0.0772

R35 500 G. flavomarginatus 2 1 2 1 0.0020 0.0020 - G. berlandieri 2 0 0 0 0 0 - G. agassizii - Mojave 52 4 6 0.531 0.0026 0.0018 0.9952 Goag- Northern 12 4 4 0.788 0.0040 0.0027 1.7935 Goag- Western 40 4 5 0.318 0.0014 0.0019 -0.6266 G. morafkai 102 7 8 0.709 0.0024 0.0027 -0.2293 Gomo - Sonoran 72 6 7 0.619 0.0021 0.0025 -0.3695 Gomo - Sinaloan 30 6 7 0.740 0.0027 0.0030 -0.3111 All desert tortoises 154 7 9 0.783 0.0028 0.0025 0.2650

147

Locus bp Pop N S Hn Hd π θ Tajima's D TB02 425 G. flavomarginatus 4 1 2 0.500 0.0012 0.0013 -0.6124 G. berlandieri 4 2 4 1.000 0.0031 0.0026 1.8931 G. agassizii - Mojave 48 4 6 0.631 0.0022 0.0021 0.0390 Goag- Northern 12 2 4 0.742 0.0023 0.0016 1.2896 Goag- Western 36 4 6 0.568 0.0020 0.0023 -0.3495 G. morafkai 104 3 4 0.262 0.0006 0.0014 -0.9331 Gomo - Sonoran 72 2 3 0.225 0.0005 0.0010 -0.7345 Gomo - Sinaloan 32 2 3 0.331 0.0008 0.0012 -0.6063 All desert tortoises 152 7 9 0.402 0.0012 0.0029 -1.2941

TB07 590 G. flavomarginatus 4 4 2 0.667 0.0045 0.0037 2.0803 G. berlandieri 4 3 2 0.667 0.0034 0.0028 2.0119 G. agassizii - Mojave 28 4 6 0.741 0.0034 0.0017 2.4879 Goag- Northern 4 4 2 0.667 0.0045 0.0037 2.0803 Goag- Western 24 4 6 0.768 0.0034 0.0018 2.3530 G. morafkai 96 5 7 0.687 0.0019 0.0017 0.3516 Gomo - Sonoran 64 5 6 0.628 0.0019 0.0018 0.1073 Gomo - Sinaloan 32 2 4 0.742 0.0017 0.0008 2.0368 All desert tortoises 124 5 9 0.742 0.0027 0.0016 1.5300

148

APPENDIX D

Book Chapter, published in:

Biology and Conservation of North American Tortoises. 2014. D.C. Rostal, E.D. McCoy and H.R. Mushinsky (eds). Johns Hopkins University Press, Baltimore, MD.

Chapter 15: Population and Conservation Genetics

TAYLOR EDWARDS

J. SCOTT HARRISON

One of the principal goals of conservation is to preserve the diversity of species. Integral to achieving this goal is the conservation of genetic, ecological and morphological variation within a species and among populations. The field of conservation genetics provides tools and principles for preserving genetic diversity, which improves a species’ ability to cope with environmental change and decreases its susceptibility to extinction. In addition, genetic data can be used to estimate effective population size, migration rates and to assess the role of landscape features and environmental conditions that contribute to the evolutionary history of a species. This information can be directly applied to conservation and management.

Genetic studies contribute to our knowledge of North American tortoises (Gopherus) in a wide variety of ways, including: systematics (Lamb et al. 1989, Lamb and Lydeard 1994, Morafka et al. 1994, Spinks et al. 2004, Le et al. 2006, Thompson and Shaffer 2010), taxonomy (Murphy et al. 2011), phylogeography (Lamb et al. 1989, Ostentoski and Lamb 1995, Edwards et al. 2012, Ennan et al. 2012), population structure (Jennings 1985, Rainboth et al. 1989, Glenn et al. 1990, McLuckie et al. 1999, Edwards et al. 2004a, Schwartz and Karl 2005, Ennen et al. 2010, Fujii and Forstner 2010, Sinclair et al. 2010, Latch et al. 2011, Richter et al. 2011, Alejandra Urena-Aranda and Espinosa de los Monteros 2012, Clostio et al. 2012), management (Britten et al. 1997, Murphy et al. 2007, Hagerty and Tracy 2010, Ennen et al. 2011, Edwards and Berry 2013), ecological genetics (Hagerty et al. 2011), hybridization (Edwards et al. 2010), paternity (Moon et al. 2006, Davy et al. 2011), and forensics (Schwartz and Karl 2008). In this chapter, we first introduce some of the fundamental principles of population genetics as observed in Gopherus and show how they contribute to our understanding of tortoise ecology. We then review how genetic studies help to inform management and conservation efforts.

149

POPULATION GENETICS, GENETIC DIVERSITY, AND POPULATION GENETIC STRUCTURE

Population genetics evaluates differences in genetic variation among populations or individuals. Differences in the distribution of genetic variation can arise from a combination of processes, including natural selection, isolation, dispersal, demographic processes (e.g., changes in population size) or stochastic processes (genetic drift). Patterns in the distribution of genetic variation at different spatial and/or temporal scales can be used to test hypotheses relating to past or ongoing processes of evolutionary and ecological importance.

Quantifying the amount of genetic variation within and among populations is essential for understanding several aspects of a species ecology and conservation (Frankham 1995). For example, this information can be used to determine if a species constitutes a single, continuous population or multiple, distinct populations. Informed practices for preserving genetic variation within a species are critical for effective conservation.

When genetic variation of a species is partitioned non-randomly across its geographic range, it is called population structure. In the absence of selection, the genetic structure of a population is influenced primarily by the amount of gene flow (exchange of genetic material via immigration of individuals) among populations. The absence of structure is called panmixia, where mating between individuals is completely random. Patterns of genetic variation can be assessed using a variety of methods and inferences are dependent on the type of genetic marker being measured (Table 1).

Mitochondrial DNA (mtDNA) is a commonly used genetic marker for population studies. Unlike nuclear genes, mtDNA is maternally inherited and haploid. Because of these characteristics, mtDNA may be affected by factors such as population size reductions (bottlenecks) and isolation at a faster rate than some diploid nuclear markers. In addition, the lack of recombination in mtDNA makes it useful for deciphering the evolutionary relationships among populations relative to their geographic distribution (phylogeography). Mitochondrial markers were some of the first used in the study of genetic diversity of Gopherus species. More recently, the inclusion of nuclear genetic markers called microsatellites or short (or simple) tandem repeats (STRs) to population genetic studies of Gopherus species has greatly improved our understanding of tortoises on a population level. Because STRs have a faster rate of evolution than mtDNA, are inherited from both parents and provide a broader sampling of the genome, STR data may be used to estimate more recent levels of gene flow and demographic history.

150

The diversity of mtDNA when compared among different Gopherus species illustrates their different evolutionary histories (Table 1). A modest amount of mtDNA genetic diversity exists in the G. polyphemus (Gopher tortoise) just within Florida: nucleotide sequence divergence among sampled populations averages 1.1% and is attributed to episodes of Pleistocene marine inundation of approximately 1.2 million years ago (Ostenoski and Lamb 1995). Across the entire range of G. polyphemus, Clostio et al. (2012) observed a large amount of mtDNA sequence divergence, 2.3%, with a pronounced phylogenetic break across the Apalachicola River drainage. This genetic barrier was also supported by Ennan et al. (2012) although they observed a broad zone of geographic overlap of mtDNA haplotypes on both sides of the Apalachicola River. Thus, barriers to gene flow over the evolutionary history of G. polyphemus seem to have influenced the distribution of mtDNA variation. This pattern is hardly surprising in this wide ranging species, occupying habitats in a landscape with heterogeneous climate, vegetation, topography, and other environmental factors. In contrast to the entire species’ range, variation of mtDNA within local populations of G. polyphemus is low, a pattern which seems to be characteristic of all Gopherus species (Ostenoski and Lamb 1995).

Other Gopherus species exhibit less divergence between populations than are observed in G. polyphemus. Edwards (2003) compared the mtDNA sequence divergence of G. agassizii (Agassiz’s desert tortoise) within the western Mojave Desert to that of G. morafkai (Morafka’s desert tortoise) in the Sonoran Desert of southern Arizona. Within similar sized geographic areas, the divergence within populations of both species ranged from 0.1% to 0.2%. Edwards (2003) speculated that the low mtDNA genetic variation observed within populations of both desert tortoise species may indicate a prehistoric reduction of population size in both the Mojave and Sonoran Deserts, coinciding with the most recent major glacial-maximum, approximately 22,000 years ago.

The mtDNA of G. berlandieri (Texas Tortoise) also exhibits very little within population variability. Using the samples of Fujii and Forstner (2010), Edwards (unpublished data) estimated sequence divergence at 0.08% (n = 58, ND2 region, 509 bp). G. flavomarginatus (Bolson Tortoise) exhibits the lowest genetic diversity: Edwards (unpublished data) observed no mtDNA polymorphisms in an assessment of 78 individuals (ND2 region, 1109 bp). Urena-Aranda and Monteros (2012) sampled 76 wild G. flavomarginatus representing the entire distribution of the species in the Bolson de Mapimí in the Chihuahuan Desert, Mexico, and observed that 74 individuals exhibited a single haplotype of only two that were identified (D-loop, 842 bp). The reduced mtDNA genetic variation in G. flavomarginatus could have resulted from the species’ undergoing an extreme population bottleneck, caused by range reduction; however a lack of mtDNA

151

diversity would also be expected if the species has simply maintained a small population size over its evolutionary history.

With the exception of perhaps the G. agassizii and G. polyphemus, the full extent of genetic variation across the entire range of the different North American Gopherus species has not yet been fully described. For example, G. berlandieri samples analyzed by Fujii and Forstner (2010) were limited to Texas; no samples were collected within the species’ range in Mexico. G. morafkai exemplifies another species with yet-to-be described diversity; Lamb et al. (1989) suggested a distinct mtDNA lineage of the species in Mexico. Tortoises genotyped in southern Sonora and Sinaloa, Mexico exhibit mtDNA and STR marker divergence from the rest of G. morafkai equivalent to that between the two recognized desert tortoise species (Edwards et al. 2012). This pattern could suggest a possible trichotomy of desert tortoise lineages evolving from a single ancestral type 5–6 million years ago. From a conservation perspective, the current abundance and range of G. morafkai would be reduced dramatically by defining a distinct population or species in Mexico.

How genetic variation is partitioned within and among populations can be assessed with greater resolution using highly polymorphic genetic markers, such as STRs. This method can provide an estimate of gene flow and thus the historic and/or current connectivity of habitat across a landscape. In a range-wide study of G. agassizii using STRs, Murphy et al. (2007) estimated that 6.1% of the variation was found among different sample locations. This relatively low level of differentiation among locations suggests that gene flow occurs throughout the sampling area. In an STR study of G. morafkai, Edwards et al. (2004a) also detected weak population genetic structure, with 3.7% differentiation among tortoise localities, but the authors sampled from a smaller study area, not representative of the entire range of the species. In G. berlandieri, Fujii and Forstner (2010) estimated population differentiation of 8.3% between the north and south of the species’ range in Texas and suggested that continued gene flow has prevented strong differentiation. Within a fairly small areas such as the Kennedy Space Center, Florida (Sinclair et al. 2010) or the Camp Shelby Joint Forces Training Facility, Mississippi (Richter et al. 2011) as much as 3.0% genetic differentiation is observed among geographically isolated colonies of G. polyphemus.

All of the above estimates of population differentiation using STR markers are consistent with models of population structure in which gene flow occurs among sampling locations, but the scale at which these studies were conducted varied significantly. For example, although the estimates of population differentiation within G. polyphemus and G. morafkai may appear quite comparable, it is important to note that samples collected by Sinclair et al. (2010) and Richter et al. (2011) each covered a 152

maximum distance of only 45 km whereas the study by Edwards et al. (2004a) spanned 186 km. In a different study of G. polyphemus, Schwartz and Karl (2005) observed extremely high genetic differentiation (using STRs) across a much larger study area, spanning Florida and southern Georgia, with an average of 24% of the genetic variation partitioned among sampling localities. Similarly, in a range-wide study, Clostio et al. (2012) observed as much as 54% genetic differentiation of G. polyphemus between disparate populations separated by both the Mobile and Apalachicola Rivers.

The difference between G. polyphemus and desert tortoises may be accounted for by the mobility of the different species. Radio telemetry studies of G. morafkai confirm that individuals are capable of making long distance movements (32 km; Edwards et al. 2004b). G. polyphemus do not appear to disperse as readily, however (McRae et al. 1981, Diemer 1992, Eubanks et al. 2003). This difference may account for the much greater genetic differentiation observed among localities throughout the species’ range, relative to the other Gopherus species. Dispersal is an important characteristic to consider for a species whose habitat has been highly fragmented over the last one hundred years. Other factors that could influence gene flow via dispersal ability are sex-biased dispersal or territoriality, but these have not been determined to have an influence on genetic structure in species of Gopherus (Hagerty et al. 2011).

For most Gopherus species, genetic structure is primarily a function of the distance between populations acting as a barrier to gene flow, called isolation by distance (IBD). Edwards et al. (2004a) stated that “the desert tortoise is perhaps the ideal organism for the IBD model; one that is distributed across the landscape in patches and for which the difficulty of dispersal is a function of geography.” In G. morafkai, Edwards et al. (2004a) observed that the geographic distance among populations accounts for approximately 30% of the genetic distance measured among sampled populations. For G. agassizii, geographic distance explains approximately 68% of the genetic differentiation observed (Murphy et al. 2007; Hagerty et al. 2011). Similarly, in G. berlandieri, 50.4% of the observed genetic differentiation among populations can be attributed to geographic distance (Fujii and Forstner 2010). IBD is observed across the range of G. polyphemus, with 89% of the genetic differentiation explained by geographic distance (Clostio et al. 2012). Interestingly, the relationship between geographic distance and genetic distance in Gopherus appears to break down when observed at finer, geographic scales (<50 km: Sinclair et al. 2010, Latch et al. 2011, Richter et al. 2011). It is also important to note that several recent studies in other taxa have suggested that “ecological distance” also can have a significant impact on the distribution of genetic variation (see Manel et al. 2003, Holkit et al. 2010). Ecological distance is a measure of the least cost path distance between populations due to landscape factors. Ecological distance may be important to

153

consider when deciphering historical evolutionary processes vs. the potential consequences of recent anthropogenic fragmentation. This area of research remains to be explored in Gopherus.

The maintenance of genetic diversity across the range of a species requires careful management. Local adaptation of gene complexes may be vital for survival in some habitats. Consequently, mixing among populations via uninformed translocations and reintroductions may result in reduced survival and lower the viability of the population (called outbreeding depression). Of special concern for long-lived species is the fact that the consequences of outbreeding depression are often not expressed until the second generation (Sheffer et al. 1999). In addition, a loss of a population may represent a loss of unique genetic variants represented nowhere else, and thus fail the objectives of a comprehensive conservation plan. Knowledge of the genetic variability of a species across its range can make an important contribution to implementing successful conservation actions.

POPULATION SIZE

Genetic diversity is generated in populations through gene flow (immigration) and mutation, but its preservation is largely dependent on population size. Changes in population size affect the genetic variation of a population. A recent reduction in population size (population bottleneck) may be detected through genetic studies. Schwartz and Karl (2005) detected recent reductions in population size in several Florida populations of G. polyphemus. A significant population bottleneck was also observed in the wild population of the G. flavomarginatus (Urena-Aranda and Monteros 2012; Edwards et al., unpublished data). Unfortunately, the genetic signature of a population bottleneck is difficult to identify in tortoises because the loss of variation proceeds more slowly in long-lived organisms (Tessier et al. 2005). In G. agassizii, the genetic signature of a population bottleneck was only detected in the northern part of its range, although population declines throughout its distribution have been well documented (Murphy et al. 2007). Advances in the algorithms used to detect genetic bottlenecks are taking steps to address this limitation (see Kuo and Janzen 2003).

A reduction in population size can result in loss of genetic variation as a result of drift and reduced gene flow (Templeton et al. 1990). Small populations also increase the potential for inbreeding, which can reduce individual fitness and population viability, resulting in an even smaller population. This negative cycle is called an extinction vortex (Gilpin and Soulé 1986). Ennen et al. (2010) found western populations of G.

154

polyphemus to have low genetic diversity, which they attributed to population declines in this portion of the species’ range. The authors suggested that low genetic diversity may be a contributing factor to reduced hatchling success and that reproductive problems in these populations of G. polyphemus may be a consequence of increased inbreeding within small populations (inbreeding depression). Harrison et al. (unpublished data) found lower levels of heterozygosity and allelic diversity in small populations relative to large populations of G. polyphemus in Georgia. This finding exemplifies the effects that changes in population size can have on the genetic make-up of a population.

An important application of conservation genetics is to predict a population’s evolutionary potential. Evolutionary potential describes the ability of a population to evolve in the presence of environmental change. In the context of species conservation, it is not possible for us to determine which individuals contribute most to the evolutionary potential of the species, or more importantly, which adaptive traits will be most critical in the face of environmental change. For a species to persist in a changing environment, genetic diversity provides the foundation for adaptation. In the face of stochastic processes, such as climate change, the prudent approach to species conservation is to preserve the entirety of a species’ genetic diversity.

APPLICATIONS OF GENETIC INFORMATION TO CONSERVATION

Knowing the population structure of a species and the distribution of genetic variation allows us to implement conservation efforts that maintain natural variation and preserve a species’ evolutionary potential. We now present some examples of how knowledge gained by genetic studies has been applied to conservation efforts of Gopherus species.

For G. agassizii, genetic information was considered in the designation of new recovery units (RU’s) in the revised recovery plan (USFWS 2011). Two independent studies (Murphy et al. 2007, Hagerty and Tracy 2010) have assessed population genetic structure across the range of the species, in the context of the RU’s of the original 1994 recovery plan (USFWS 1994). Both studies combined suggest that there are as many as nine genetic subdivisions across the range of the species (Figure 1). The revised recovery plan reduced the original number of RU’s from six to five, however. The preparers of the revised recovery plan based the new RU’s primarily on geographic discontinuities that coincide with observed genetic variation and the plan considered “demographic, ecological, and behavioral considerations to be of greater importance than genetic issues alone” (USFWS 2011). While the advancement of molecular techniques makes it

155

possible to delineate differences between population subunits, an assessment of neutral genetic variability alone does not necessarily mean that those differences are biologically significant (Hedrick 1999). Decisions should not necessarily be made solely on genetics, but should also incorporate relevant ecological and natural history information in defining management units for conservation (Berry et al. 2002). Ultimately, it is up to resource managers to determine what constitutes a reasonable amount of variation in a population that lends itself to constructive management.

The maintenance of genetic diversity is best achieved by ensuring that there is no reduction in population size and natural patterns of gene flow within and among populations are maintained (Frankham 1995). Artificial habitat fragmentation as a result of anthropogenic landscape change can result in rapid genetic differentiation between remnant patches of habitat (Templeton et al. 1990, McCraney et al. 2010). This rapid differentiation is a consequence of reduction of core habitat, an increase of edge effects, increased rates of drift, and the prevention of normal patterns of gene flow (immigration). BenDor et al. (2009) simulated habitat fragmentation effects on G. polyphemus and found that fragments the landscape with as little as 10% total habitat loss, but that fragments the landscape, could reduce the rate of dispersal among sub-populations by as much as 31%. Schwartz and Karl (2005) attribute some of the genetic differentiation they observed among populations of G. polyphemus to reduced gene flow caused by anthropogenic habitat fragmentation. In a fine-scale analysis of G. agassizii, Latch et al. (2011) observed recent population subdivision, and suggested that roads may have influenced gene flow on a very local level (within a ~25 km radius). Because of the long life span of Gopherus species, however, it is generally assumed that current genetic studies can only assess historic rates of gene flow, since landscape changes have been relatively recent in relation to the generation time of tortoises (Edwards et al. 2004a, Sinclair et al. 2010). In a study of recently fragmented populations of G. polyphemus, Ennen et al. (2011) observed a greater rate of change in demographic traits than in genetic traits, which they suggest is typical of long-lived species with long generation times. Hagerty and Tracy (2010) warn that despite the inability to detect current reductions in inter-population movement via genetic data, habitat fragmentation throughout the range of G. agassizii has “likely removed all possible paths among previously connected populations.”

The potential effects of habitat fragmentation in G. morafkai were assessed by Edwards et al. (2004a). In the Sonoran Desert, tortoises occur in seemingly isolated patches of upland habitat separated by low, desert valleys (Barrett 1990). Historic connectivity of these patches, resulting from gene flow, was identified in a genetic study by Edwards et al. (2004a), and the authors suggest that movements between populations have been an important part of the species’ evolutionary history. Thus, despite the low

156

density of tortoises that occur in valley bottoms separating core populations in the foothills, these individuals are critical in maintaining gene flow (Averill-Murray and Averill-Murray 2005). Currently, landscape changes, such as roads, urban development, agriculture, and canals, likely make such movements impossible (USFWS 2010). The fact that each of these foothill habitats contains only a small population of tortoises, and considering the life history traits of tortoises (long-lived and slow to mature), recovery of a small population after a reduction in population size (from drought, disease, etc.) may require immigration from neighboring populations. Thus, persistence of tortoise populations distributed in small patches throughout the Sonoran Desert likely relies on gene flow among populations to ensure long-term viability.

When natural patterns of gene flow can no longer take place, or when a small population is suffering from the effects of reduced genetic diversity, translocation may be a viable management option. Without careful management, however, the introduction of genetically differentiated individuals into wild populations of tortoises could be detrimental. In addition to concerns about health and survivorship of translocated individuals (Tuberville et al. 2005), biologists and managers should consider the potential of outbreeding depression caused by introduction of genetically differentiated individuals (Edwards and Berry 2013). Several studies of Gopherus have detected evidence of translocated tortoises in their genetic datasets from samples collected in the wild (Schwartz and Karl 2005, Murphy et al. 2007, Fujii and Forstner 2010, Clostio et al. 2012). These undirected translocations often occur in large numbers (Glenn et al. 1990, Enge et al. 2002, Murphy et al. 2007), resulting from relocations during development or as escaped or released captives. The introduction of “pet” tortoises into wild populations is of particularly concern, considering the abundance of hybrid tortoises observed in captivity. In Arizona, Edwards et al. (2010) identified that 33% of captive tortoises from the Phoenix area were not the local G. morafkai but instead either G. agassizii, G. morafkai x G. agassizii hybrids, or even G. morafkai x G. berlandieri hybrids.

Another caveat of developing translocation and repatriation strategies or addressing the problems associated with escaped or released pet tortoises is polyandry and sperm storage (Moon et al. 2006, Davy et al. 2011). Davy et al. (2011) give a conservative estimate of 64% of female G. agassizii being polyandrous, with a resulting 57% of clutches observed that were sired by multiple males. Murphy et al. (2007) report a captive female G. agassizii that continued to produce viable clutches of eggs for 15 years following isolation from males. Polyandry and long-term sperm storage allow a single mature female to represent a source of genetic material from multiple individuals. This information should make adult females an important target of any managed translocation or repatriation effort, since they have the potential to successfully introduce

157

additional genetic diversity to the population beyond just their own. A female tortoise of unknown history (such as from captivity), even if genotypically suitable for repatriation, could potentially introduce admixed or hybrid progeny into the population after release, however (Edwards and Berry 2013).

The combination of tortoises’ being popular pets and also protected by law means that legal conflicts of ownership are inevitable. Genetic testing has been used to inform identity and placement of confiscated tortoises. In particular, the species or population identity of tortoises has important legal implications, since some species and populations are federally protected and even when held legally as pets, individuals are not allowed to be taken across state lines without permits. Schwartz and Karl (2008) used DNA assignment testing to assess the population of origin for confiscated G. polyphemus in Florida. Desert tortoises confiscated by the Arizona Game and Fish Department and of suspicious origins are genotyped so that the appropriate legal action can be taken and the tortoise can be properly placed in a suitable captive facility (Edwards 2008). Unfortunately, genotyping does not entirely solve the problem for desert tortoises. Because the geographic boundary of the federally protected portion of the range of G. agassizii is defined by the Colorado River, individuals of the species that naturally occur in the Black Mountains of Northwestern Arizona (McLuckie et al. 1999, Edwards et al. 2010) are not afforded the same legal protection as the rest of the species under the current recovery plan (USFWS 2011). Thus, law enforcement officers cannot always rule out that a confiscated G. agassizii in Arizona may have originated legally in the state.

The only captive breeding effort specifically for conservation of a Gopherus species has been for G. flavomarginatus, as part of the repatriation effort into the United States (Truett and Phillips 2009). A population of captive tortoises is being maintained by the Turner Endangered Species Fund in a semi-natural setting in part of their historic range in the Chihuahuan Desert in New Mexico. Genetic information is being used to inform mate pairing, to maximize genetic diversity in the captive population and to reduce the potential for inbreeding (Edwards et al., unpublished data). Captive breeding and repatriation can be an effective conservation strategy, but requires special consideration when applied to long-lived species like tortoises (Williams and Osentoski 2007). Small populations are more susceptible to inbreeding, which can decrease heterozygosity of individuals and lead to a reduction of fitness. Inbreeding can also lead to the expression of recessive alleles, resulting in a decrease in population viability. One way to improve genetic diversity of G. flavomarginatus captive population is to introduce new individuals into the breeding program from zoological institutions or private collections. Genetic screening of potential new animals has uncovered multiple G.

158

flavomarginatus x G. polyphemus hybrids in captivity which are ineligible for inclusion in the repatriation effort (Edwards et al., unpublished data).

SUMMARY

Understanding the population ecology and genetics of a species is fundamental to implementing effective strategies for conservation. For a broad overview of genetic applications to turtle conservation, see Shaffer et al. 2007. As previous chapters of this book have exemplified, there are several life history traits such as dispersal ability, survivorship of adults, longevity and delayed sexual maturity that are shared among species of the genus Gopherus. The ability to “share” information from one species to inform management decisions about another is sometimes necessary in making timely conservation actions. Conservation biologists and resource managers are often forced to move forward based on the best information available. As is apparent in this book, most of what we know about the genus Gopherus has come from studies of G. polyphemus and G. agassizii, despite other species’ perhaps having greater immediate threats of extinction. Application of the tools and principles of population genetics is imperative for successful management and conservation of all species in the genus.

ACKNOWLEDGEMENTS

We thank RW Murphy, KH Berry, ME Kaplan, C Dolan and an anonymous reviewer for their helpful comments to this chapter.

LITERATURE CITED

Averill-Murray, R. C., and A. Averill-Murray. 2005. Regional-Scale Estimation of Density and Habitat Use of the Desert Tortoise (Gopherus agassizii) in Arizona. Journal of Herpetology, 39(1):65-72. Barrett, S.L. 1990. Home range and habitat of the desert tortoise (Xerobates agassizi) in the Picacho Mountains of Arizona. Herpetologica 46(2):202-206. BenDor, T., Westervelt, J., Aurambout, J. P., & Meyer, W. 2009. Simulating population variation and movement within fragmented landscapes: An application to the gopher tortoise (Gopherus polyphemus). Ecological Modelling 220: 867-878. Berry, K.H., D.J. Morafka, and R.W. Murphy. 2002. Defining the desert tortoise(s): our first priority for a coherent conservation strategy. Chelonian Conservation and Biology, 4(2):249-262. Britten, H. B., Riddle, B. R., Brussard, P. F., Marlow, R., Lee, T. E., Marlow, R. O. N., & Jr, T. E. L. E. E. 1997. Genetic Delineation of Management Units for the Desert Tortoise, Gopherus agassizii, in Northeastern Mojave Desert. Copeia 1997: 523- 530. 159

Clostio RW, AM Martinez, KE LeBlanc, and NM Anthony. 2012. Population genetic structure of a threatened tortoise across the south-eastern United States: implications for conservation management. Anim. Conserv. 6: 613-625. Davy, C. M., T. Edwards, A. Lathrop, M. Bratton, M. Hagan, B. Henen, K. A. Nagy, J. Stone, L. S. Hillard and R. W. Murphy. 2011. Polyandry and multiple paternities in the threatened desert tortoise, Gopherus agassizii. Conservation Genetics 12:1313–1322. Diemer JE (1992) Home range and movements of the tortoise Gopherus polyphemus in northern Florida. J. Herpetol., 26, 158–165. Eubanks JO, Michener WK, Guyer C (2003) Patterns of movements and burrow use in a population of gopher tortoises (Gopherus polyphemus). Herpetologica, 59, 311– 321. Edwards T. 2003. Desert Tortoise Conservation Genetics. Unpublished M.S. Thesis, University of Arizona, Tucson, USA. Edwards T. 2008. Turtle trouble in Kingman, Arizona. Sonoran Herpetol. 21:62–65. Edwards, T., Schwalbe, C. R., Swann, D. E., & Goldberg, C. S. 2004a. Implications of Anthropogenic Landscape Change on Inter-Population Movements of the Desert Tortoise (Gopherus agassizii). Conservation Genetics 5: 485-499. Edwards, T., E.W. Stitt, C.R. Schwalbe, and D.E. Swann. 2004b Natural History Notes: Gopherus agassizii (Desert Tortoise). Movement. Herpetological Review 35(4):381-382. Edwards, T., Jarchow, C. J., Jones, C. A, & Bonine, K. E. 2010. Tracing Genetic Lineages of Captive Desert Tortoises in Arizona. Journal of Wildlife Management 74: 801-807. Edwards T, M Vaughn, CM Torres, AE Karl, PR Rosen, KH Berry, and RW Murphy. 2012. A biogeographic perspective of speciation among desert tortoises in the genus Gopherus: A preliminary evaluation. In Gottfried, GJ, PF Ffolliott, BS Gebow, and LG Eskew, eds. Merging science and management in a rapidly changing world: biodiversity and management of the Madrean Archipelago III. 2012 May 1-5, Tucson, AZ. Proceedings RMRS-P-67. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. Edwards T, and KH Berry. 2013. Are captive tortoises a reservoir for conservation? An assessment of genealogical affiliation of captive Gopherus agassizii to local, wild populations. Conservation Genetics 14:649-659. Enge, K.M., B.A. Millsap, T.J. Doonan, J.A. Gore, N.J. Douglass, and G.L. Sprandel. 2002. Conservation plans for biotic regions in Florida containing multiple rare or declining wildlife taxa. Florida Fish and Wildlife Conservation Commission Technical Report 20. 146 p. Ennen, J. R., Kreiser, B. R., & Qualls, C. P. 2010. Low Genetic Diversity in Several Gopher Tortoise (Gopherus polyphemus) Populations in the Desoto National Forest, Mississippi. Herpetologica 66: 31-38. Ennen JR, RD Birkhead, BR Kreiser, DL Gaillard, CP Qualls, and JE Lovich. 2011. The effects of isolation on the demography and genetic diversity of long-lived species:

160

implications for conservation and management of the gopher tortoise (Gopherus polyphemus). Herpetol. Conserv. Biol. 6:202–214. Ennen JR, BR Kreiser, CP Qualls, D Gaillard, M Aresco, R Birkhead, TD Tuberville, ED McCoy, H Mushinsky, TW Hentges, and A Schrey. 2012. Mitochondrial DNA assessment of the phylogeography of the Gopher Tortoise. Journal of Fish and Wildlife Management 3:110–122. Frankham R. 1995. Conservation Genetics. Annu. Rev. Genet. 29:305–27. Fujii, A., & Forstner, M. R. J. 2010. Genetic Variation and Population Structure of the Texas Tortoise, Gopherus berlandieri (Testudinidae), with Implications for Conservation. Chelonian Conservation and Biology 9: 61-69. Gilpin ME, Soulé ME (1986). "Minimum Viable Populations: Processes of Species Extinction". In M. E. Soulé. Conservation Biology: The Science of Scarcity and Diversity. Sinauer, Sunderland, Mass. pp. 19–34. Glenn, J. L., Straight, R. C., & Sites, jr., J. W. 1990. GlennEtAl_DTProteinMarkers_GBNaturalist1990.pdf. The Great Basin Naturalist 50: 108. Hagerty, B. E., & Tracy, C. Richard. 2010. Defining population structure for the Mojave desert tortoise. Conservation Genetics 11: 1795-1807. Hagerty BE, KE Nussear, TC Esque, CR Tracy. 2011. Making molehills out of mountain landscape genetics of the Mojave desert tortoise. Landscape Ecol. 2: 267–280. Hedrick PW. 1999. Perspective: highly variable loci and their interpretation in evolution and conservation. Evolution 53:313–318. Hokit DG, Ascunce M, Ernst J, Branch LC and AM Clark. 2010. Ecological metrics predict connectivity better than geographic distance. Conservation Genetics 11:149-159. Jennings, R. D. 1985. Biochemical variation of the Desert Tortoise, Gopherus agassizii. Unpublished M.S. Thesis, University of New Mexico, Albuquerque, NM. 64 pp. and appendices. Kuo CH, Janzen FJ. 2003. BottleSim: a bottleneck simulation program for long-lived species with overlapping generations. Molecular Ecology Notes, 3, 669-673. DOI: 10.1046/j.1471-8286.2003.00532.x Lamb, T, Avise, J. C., & Gibbons, J W. 1989. Phylogeographic Patterns in Mitochondrial DNA of the Desert Tortoise (Xerobates agassizi), and Evolutionary Relationships Among the North American Gopher Tortoises. Evolution 43: 76-87. Lamb, T, & Lydeard, C. 1994. Molecular Phylogeny of the Gogher Tortoises, with Comments on Familial Relationships within Testudinoidea. Molecular phylogenetics and evolution 3: 283-291. Latch EK, WI Boarman, A Walde, and RC Fleischer. 2011. Fine-scale analysis reveals cryptic landscape genetic structure in desert tortoises. PLoS ONE 6:e27794 Le, M., Raxworthy, C. J., McCord, W. P., & Mertz, L. 2006. A molecular phylogeny of tortoises (Testudines: Testudinidae) based on mitochondrial and nuclear genes. Molecular phylogenetics and evolution 40: 517-31.

161

Manel, S, M. K. Schwartz MK, G. P. Luikart and P. Taberlet. 2003. Landscape genetics: combining landscape ecology and population genetics. Trends in Ecology and Evolution 18(4):189-197. McCraney W. T., G. Goldsmith, D. K. Jacobs and A. P. Kinziger. 2010. Rampant drift in artificially fragmented populations of the endangered tidewater goby (Eucyclogobius newberryi). Molecular Ecology 19(16):3315-3327. Mcluckie, A. M., Lamb, Trip, Schwalbe, C. R., Mccord, R. D., Journal, S., & Mar, N. 1999. Genetic and Morphometric Assessment of an Unusual Tortoise (Gopherus agassizii) Population in the Black Mountains of Arizona. Journal of Herpetology 33: 36-44. McRae WA, Landers JL, Garner JA (1981) Movement patterns and home range of the gopher tortoise. Am. Midland Nat., 106, 165–179. Moon, J. C., McCoy, E. D., Mushinsky, H. R., & Karl, Stephen a. 2006. Multiple paternity and breeding system in the gopher tortoise, Gopherus polyphemus. The Journal of heredity 97: 150-7. Morafka, D. J., L., G. A., & Murphy, Robert W. 1994. Allozyme differentiation among gopher tortoises (Gopherus): conservation genetics and phylogenetic and taxonomic implications. Canadian Journal of Zoology 72: 1665-1671. Murphy, R. W., K. H. Berry, T. Edwards, A. Leviton, A. Lathrop, J. D. Riedle. 2011. The dazed and confused identity of Agassiz’s land tortoise, Gopherus agassizii (Testudines: Testudinidae) with the description of a new species and its consequences for conservation. Zookeys 113: 39–71. Murphy, Robert W, Berry, K. H., Edwards, Taylor, & Mcluckie, A. M. 2007. A Genetic Assessment of the Recovery Units for the Mojave Population of the Desert Tortoise , Gopherus agassizii. 6: 229-251. Osentoski MF, and T Lamb. 1995. Intraspecific phylogeography of the gopher tortoise, Gopherus polyphemus: RFLP analysis of amplified mtDNA segments. Mol. Ecol. 4:709–718. Rainboth, W. J., Buth, D. G., & Turner, F. B. 1989. Allozyme Variation in Mojave Populations of the Desert Tortoise, Gopherus agassizi. Copeia 1989: 115-123. Richter SC, JA Jackson, M Hinderliter, D Epperson, CW Theodorakis, and SM Adams. 2011. Conservation genetics of the largest cluster of federally threatened gopher tortoise colonies with implications for species management. Herpetologica 67:406–419. Schwartz, Tonia S., & Karl, Stephen a. 2005. Population and conservation genetics of the gopher tortoise (Gopherus polyphemus). Conservation Genetics 6: 917-928. Schwartz, Tonia S., & Karl, Stephen a. 2008. Population Genetic Assignment of Confiscated Gopher Tortoises. Journal of Wildlife Management 72: 254-259. Shaffer, H.B., N.N. FitzSimmons, A. Georges, and A.G.H. Rhodin. Eds. 2007. Proceedings of a workshop on Genetics, Ethics, and Taxonomy of Freshwater Turtles and Tortoises. Chelonian Research Monographs, 4:200 pp. Sheffer, R. J., P. W. Hedrick and A. L. Velasco. 1999. Testing for inbreeding and outbreeding depression in the endangered Gila topminnow. Animal Conservation 2(2):121-129.

162

Sinclair CS, PJ Dawes, and RA Seigel. 2010. Genetic structuring of gopher tortoise (Gopherus polyphemus) populations on the Kennedy Space Center, Florida, USA. Herpetol. Conserv. Biol. 5:189–195. Spinks, P. Q., Bradley Shaffer, H., Iverson, J. B., & McCord, W. P. 2004. Phylogenetic hypotheses for the turtle family Geoemydidae. Molecular phylogenetics and evolution 32: 164-82. Templeton, A.R., K. Shaw, E. Routman and S. K. Davis. 1990. The Genetic Consequences of Habitat Fragmentation. Annals of The Missouri Botanical Garden 77(1):13-27. Tessier, N., Paquette, S., and Lapointe, F.J. 2005. Conservation genetics of the wood turtle (Glyptemys insculpta) in Quebec, Canada. Canadian Journal of Zoology. 83:765–772. Thomson, R. C., & Shaffer, H. B. 2010. Sparse Supermatrices for Phylogenetic Inference: Taxonomy, Alignment, Rogue Taxa, and the Phylogeny of Living Turtles. Systematic Biology 59: 42-58. Truett , J. and M. Phillips. 2009. Beyond historic baselines: Restoring bolson tortoises to Pleistocene Range. Ecological Restoration 27(2):144-151. Tuberville, T. D., Clark, E. E., Buhlmann, K. a, & Gibbons, J. Whitfield. 2005. Translocation as a conservation tool: site fidelity and movement of repatriated gopher tortoises (Gopherus polyphemus). Animal Conservation 8: 349-358. Urena-Aranda AC, and AE Monteros. 2012. The genetic crisis of the Mexican Bolson tortoise (Gopherus flavomarginatus: Testudinidae). Amphibia-Reptilia 33:45–53. U.S. Fish and Wildlife Service. 1994. Desert Tortoise (Mojave Population) Recovery Plan. U.S. Fish and Wildlife Service, Portland, Oregon, USA. 73pp. and appendices. U.S. Fish and Wildlife Service. 2010. Endangered and threatened wildlife and plants; 12- month finding on a petition to list the Sonoran population of the desert tortoise as endangered or threatened; proposed rule. Federal Register 75(293): 78094-78146. December 14, 2010. U.S. Fish and Wildlife Service. 2011. Revised recovery plan for the Mojave population of the desert tortoise (Gopherus agassizii). U.S. Fish and Wildlife Service, Pacific Southwest Region, Sacramento, CA. 222 pp. Williams, D.A. and M. F. Osentoski. 2007. Genetic considerations for the captive breeding of tortoises and freshwater turtles. Chelonian Conservation and Biology 6(2):302-313.

163

Table 1. Descriptive statistics from within species genetic studies of the species of Gopherus.

Species Study Locality Fst Divergence (%) Heterozygosity Allelic Richness Marker Type Source

G. polyphemus

Eastern Mississippi 0.50 1.7 STR Ennan et al. 2010

Western Mississippi 0.20 2.6 STR Ennan et al. 2010

Camp Shelby, Mississippi 0.03 0.21 1.8 STR Richter et al. 2011 Kennedy Space Center, FLA 0.03 0.35-0.47 3.3 - 4.1 STR Sinclair et al. 2010

Georgia and Florida 0.24 0.42-0.44 3.0-3.4 STR Schwartz et al. 2005

Range-wide 0.01-0.54 0.23-0.67 13.4 STR Clostio et al. 2012

1 Florida 1.1 mtDNA (RFLP) Osteneski and Lamb 1995

4 Range-wide 2.3 mtDNA Clostio et al. 2012 Range-wide 0.57-0.81 1.5 mtDNA Ennan et al. 2012 G. agassizii

Range-wide 0.06 STR Murphy et al. 2007

Range-wide 0.01 - 0.16 0.64 - 0.80 8.35 STR Hagerty and Tracy 2010

Fort Irwin, CA 0.005 0.74 14.4 STR Latch et al. 2011 Western Mojave RU 0.04 0.1 - 0.2 mtDNA Edwards 2003

San Bernardino County, CA 0.04 allozymes Rainboth et al. 1989

G. morafkai

Southern Arizona 0.04 STR Edwards et al. 2004a

Southern Arizona 0.62 11.1 STR Fujii and Forstner 2010

Southern Arizona 0.06 0.1 - 0.2 mtDNA Edwards 2003

164

Table 1 (cont.)

Allelic Species Study Locality Fst Divergence (%) Heterozygosity Richness Marker Type Source

G. berlandieri

North Texas 0.47 2.0 - 6.7 STR Fujii and Forstner 2010

Southern Texas 0.53 2.0 - 9.5 STR Fujii and Forstner 2010

Texas 0.08 0.56 5.6 STR Fujii and Forstner 2010 Edwards unpublished Texas 0.08 mtDNA data

G. flavomarginatus Edwards unpublished Chihuahua, Mexico 0.5 STR data Urena-Aranda and de los Chihuahua, Mexico 0.24 mtDNA Monteros 2012 Chihuahua, Mexico 0.69 0.02-0.08 allozymes Morafka et al. 1994

165

Figure 1. Information from multiple genetic studies contributing to the delineation of recovery units (RUs) in the revised recovery plan for the Gopherus agassizii (USFWS 2011). Solid line and unique shading outlines the current RUs (USFWS 2011). Dashed lines reflect differences in the original 6 RUs (USFWS 1994) relative to current designations. Double lines delineate genetic units identified by Murphy et al. (2007) in the western Mojave Desert and by Hagerty and Tracy (2010) in the northern Mojave Desert relative to current designations: Western Mojave (WM), Southern Mojave (SM), Central Mojave (CM), Virgin River (VR), Muddy Mountains (MD), Amargosa Desert 166

(AM), South Las Vegas (SLV). Murphy et al. (2007) and Hagerty and Tracy (2010) both also identified genetic units in the eastern and northern Colorado Desert.

167