MIAMI UNIVERSITY the Graduate School

MIAMI UNIVERSITY The Graduate School

Certificate for Approving the Dissertation

We hereby approve the Dissertation

Kentaro Inoue

Candidate for the Degree:

Doctor of Philosophy

______David J. Berg, Director

______Bruce J. Cochrane, Reader

______Thomas O. Crist, Reader

______Brian Keane, Reader

______Richard C. Moore, Graduate School Representative

ABSTRACT

A COMPREHENSIVE APPROACH TO CONSERVATION BIOLOGY: FROM POPULATION GENETICS TO EXTINCTION RISK ASSESSMENT FOR TWO SPECIES OF FRESHWATER MUSSELS

by Kentaro Inoue

Species conservation is an enormously complex task, which includes identification of phenomena that affect the loss, maintenance, and restoration of biodiversity and advocate for sustaining evolutionary processes that promote all levels of biological organization. Endangered species conservation requires a comprehensive approach to evaluate the conservation status of a given species, develop optimal recovery plans, and establish quantitative recovery criteria, in order to remove the necessity of protection. In my dissertation, I demonstrate such a comprehensive approach for evaluating the conservation status of two imperiled freshwater mussel species: Cumberlandia monodonta and Popenaias popeii, and providing guidance for development of species recovery plans. I characterized novel microsatellite markers for the species in order to assess population genetic diversity and structure (Chapter 1 and 3). I assessed fine-scale population structure of C. monodonta and used ecological and genetic simulations to investigate the effects of future climate change on distributional shifts in suitable habitats and population genetic connectivity (Chapter 2). I also investigated evolutionary history and genetic structure of P. popeii and used long-term mark-and-recapture monitoring to determine population dynamics (Chapter 4 and 5). I used demographic and population genetic information acquired from the previous chapters to develop recovery strategies for these species (Chapter 6). Using a large number of polymorphic microsatellite markers for both species, I revealed that climate change during the mid-to-late-Pleistocene likely shaped current distribution and genetic structure in both species. Current genetic structure of C. monodonta is likely a consequence of connectivity of suitable habitat; however, future climate change will likely reduce connectivity across populations. Climate change during the mid-to-late Pleistocene caused regional and local population structures of P. popeii. A long-term demographic study revealed that reduced river discharge is associated with significant decreased survival. Current anthropogenic activities threaten the availability of high-quality habitat for P. popeii. Demographic parameters during early life stages interact to influence the extinction probability and population growth rates for both species. The trajectories of population genetic persistence were influenced by initial allele frequency within populations, population size, and population growth rates. I used freshwater mussels to demonstrate the importance of considering evolution and ecology of organisms when developing species recovery strategies and making conservation decisions. I hope that my dissertation encourages others to utilize ecological and genetic studies in developing recovery plans for species from this and other highly imperiled groups.

A COMPREHENSIVE APPROACH TO CONSERVATION BIOLOGY: FROM POPULATION GENETICS TO EXTINCTION RISK ASSESSMENT FOR TWO SPECIES OF FRESHWATER MUSSELS

A DISSERTATION

Presented to the Faculty of

Miami University in partial

fulfillment of the requirements

for the degree of

Doctor of Philosophy

Department of Biology

Kentaro Inoue

The Graduate School Miami University Oxford, Ohio

2015

Dissertation Director: David J. Berg

Kentaro Inoue

2015

TABLE OF CONTENTS

General Introduction ...... 1

Chapter 1: Isolation and characterization of 17 polymorphic microsatellite loci in the spectaclecase, Cumberlandia monodonta (Bivalvia: Margaritiferidae) ...... 10

Abstract ...... 10 Introduction ...... 10 Methods ...... 11 Results and Discussion ...... 12 Acknowledgements ...... 12 References ...... 13 Table ...... 15

Chapter 2: Predicting the effects of climate change on population connectivity and genetic diversity in riverine systems ...... 17

Abstract ...... 17 Introduction ...... 18 Methods ...... 20 Results ...... 25 Discussion ...... 29 Acknowledgements ...... 33 References ...... 33 Tables and Figures ...... 39 Supplementary Information ...... 49

Chapter 3: Development and characterization of 20 polymorphic microsatellite markers for the Texas hornshell, Popenaias popeii (Bivalvia: Unionidae), through next-generation sequencing ...... 55

Abstract ...... 55 Introduction ...... 55 Methods ...... 56 Results and Discussion ...... 57 Acknowledgements ...... 57

iii References ...... 58 Table ...... 60

Chapter 4: Past climate change drives current genetic structure of an endangered freshwater mussel species ...... 62

Abstract ...... 62 Introduction ...... 63 Methods ...... 66 Results ...... 71 Discussion ...... 75 Acknowledgements ...... 79 References ...... 80 Tables and Figures ...... 88 Supplementary Information ...... 98

Chapter 5: Long-term mark-and-recapture study of a freshwater mussel reveals patterns of habitat use and an association between survival and river discharge ...... 102

Abstract ...... 102 Introduction ...... 103 Methods ...... 105 Results ...... 110 Discussion ...... 112 Acknowledgments ...... 116 References ...... 116 Tables and Figures ...... 121

Chapter 6: Integrating population viability analysis and genetic assessments into endangered species recovery planning for freshwater mussels ...... 128

Abstract ...... 128 Introduction ...... 129 Methods ...... 131 Results ...... 136 Discussion ...... 139 Acknowledgments ...... 143 References ...... 143

iv Tables and Figures ...... 148 Supplementary Information ...... 154

Summary and General Conclusion ...... 160

LIST OF TABLES

Chapter 1 Table 1. Characterization of 29 microsatellite loci for Cumberlandia monodonta. Chapter 2 Table 1. Descriptive statistics for COI sequences and 16 microsatellite loci for each collection site of Cumberlandia monodonta. Population IDs correspond to Figure 1. Table 2. Results of the hierarchical analysis of molecular variance (AMOVA) in Cumberlandia monodonta. Seven rivers represent four biogeographic provinces: Tennessee-Cumberland (Clinch River), Upper Mississippi (Gasconade, Meramec, Osage, and St. Croix rivers), Ohioan (Green River), and Mississippi Embayment (Ouachita River). Table 3. Areas (km2) of binary predictions of presence and absence under current conditions and two greenhouse gas concentration scenarios in 2050 and 2070. Chapter 3 Table 1. Characterization of 20 polymorphic microsatellite loci for Popenaias popeii. Table S1. A list of bioclimatic, geographic, and landscape layers used in ecological niche models. We chose eight uncorrelated layers (<0.6 Pearson correlation coefficient; “X” in the Used column).

Table S2. Pairwise DEST (above diagonal) and FST (below diagonal) values for 16 microsatellite loci from 17 populations of Cumberlandia monodonta. Chapter 4 Table 1. Descriptive statistics for COX sequences and 18 microsatellite loci for Popenaias popeii populations in the Black River, New Mexico and the Devils River and Rio Grande, Texas. Table 2. Prior distributions of parameters for each scenario (Figure 2) and posterior parameter values for Scenario 3. The unit of time is generations, except that values in parentheses are 1000 years before present (ka) calculated with an average generation time of 8.9 years (based on 8.1 - 9.6 years for Lampsilis radiata; Chagnon & de la Cheneliere 1998).

vi Table 3. Pairwise FST (below diagonal) and DEST (above diagonal) for 18 microsatellite

loci from STRUCTURE-defined populations of Popenaias popeii. All values are statistically significantly different from 0 at P<0.05.

Table 4. Asymmetric migration rates between pairs of STRUCTURE-defined populations

estimated from (A) MIGRATE-N and (B) BAYESASS. In MIGRATE-N, values in italics are θ; other entries are migration rates (M) from column populations

into row populations. In BAYESASS, all values refer to contemporary migration

rates (mc). Values in parentheses are 95% confidence interval (or credible

interval for BAYESASS) of estimates. Table 5. Type I and Type II error rates and Bayesian posterior probabilities for each

scenario estimated using DIYABC. Table S1. Estimates of relative contributions (%) of bioclimatic variables ranked

according to the MAXENT models for Popenaias popeii and its host fish (Carpiodes carpio, Cyprinella lutrensis, and Moxostoma congestum)

populations. AUCtest is the area under the curve in the receiver operating characteristic curve from a test dataset. Chapter 5 Table 1. Results for the five most-parsimonious models for the robust design in 2006- 2008 and the Cormack-Jolly-Seber (CJS) models for annual and seasonal datasets in 1997-2012. Model parameters include survival (), migration factors (γ' = emigration, γ'' = immigration), encounter probability (p), recapture probability (c) and abundance (N). Parameters denoted in parentheses indicate no-variance (.) and variance by time (t), habitat (h) and site (s) for the robust design. Additionally for CJS models, parameters indicate variance by hydrological parameters (mean, minimum, and maximum discharge and numbers of days below the 10th and 25th percentile and above the 75th and 90th percentile of daily discharge). Asterisks (*) indicate combinations of variances. Encounter probability for CJS models is conditioned with recapture probabilities estimated from the robust design (RD).

vii Chapter 6 Table 1. Lists of base-line demographic parameters for Cumberlandia monodonta and Popenaias popeii used in Vortex simulations. Standard deviations are in parentheses. Table 2. Results of relative sensitivity analyses using standardized coefficients (z- values) from logistic regression of demographic parameter sets against probability of extinction and quasi-extinction. QE (250) = quasi-extinction at N = 250 individuals; QE (500) = quasi-extinction at N = 500 individuals. Table S1. Recovery goals in recovery plans that were approved by the US Fish and Wildlife Service for 71 mussel species listed under the US Endangered Specie Act. Table S2. Definitions of “a viable population” that mentioned in 45 recovery plans that were approved by the US Fish and Wildlife Service. Table S3. Quantitative downlisting/delisting criteria in 55 recovery plans that were approved by the US Fish and Wildlife Service. Table S4. Use of population viability analysis (PVA) and genetic assessments (Genetic) for mussel species listed under the US Endangered Species Act.

viii

LIST OF FIGURES

Chapter 2 Figure 1. Map of the central United States indicating sites where Cumberlandia monodonta were sampled (black dots). See Table 1 for sampled rivers and sample size in each site. Colored watersheds represent historic (gray) and current (blue) distributions of C. monodonta obtained from NatureServe (http://explorer.natureserve.org; accessed June 25, 2015). Figure 2. A parsimony network of COI sequences for Cumberlandia monodonta. Each circle represent unique haplotype and lines between haplotypes represent one base pair; black dots are inferred missing haplotypes. Haplotype frequency is relative to the size and number in the circle. Colors represent drainages. Lineage designations were based on Inoue et al. (2014).

Figure 3. Stacked bar plots obtained from STRUCTURE, assigning individuals into k = 2 clusters; clusters were divided by the Ouachita populations (dark gray) and the remainder of populations (light gray). Top labels are biogeographic provinces identified by Haag (2009): Tennessee-Cumberland (TN), Upper Mississippi (UM), Ohioan (OH), and Mississippi Embayment (ME). Bottom labels are a priori population assignments. Figure 4. Prediction of suitable (blue) and unsuitable (light gray) habitats for Cumberlandia monodonta identified using ecological niche models (ENMs) under “current” bioclimatic conditions (interpolations of observed data from 1950 to 2000) and projections of near future climatic in 2050 (average for 2041-2060) and 2070 (average for 2051-2080) under low greenhouse gas concentration scenarios (IPCC 5th Assessment). Low greenhouse gas concentration was based on representative concentration pathways 2.6 (RCP2.6). Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs. Figure 5. Prediction of suitable (blue) and unsuitable (light gray) habitats for Cumberlandia monodonta identified using ecological niche models (ENMs)

ix under “current” bioclimatic condition (interpolations of observed data from 1950 to 2000) and projections of near future climatic in 2050 (average for 2041-2060) and 2070 (average for 2051-2080) under high greenhouse gas concentration scenarios (IPCC 5th Assessment). High greenhouse gas concentration was based on representative concentration pathway 8.5 (RCP8.5). Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs.

Figure 6. Results of CDPOP to compare between observed (empirical) and projected th DEST at 1000 year. Simulations were (A) no climate change model, (B) RCP2.6 model, and (C) RCP8.5 model. Dotted lines represent 1:1

relationship. Points above the line indicate that DEST increases in the future (population differentiation increases) and points below the line indicate that

DEST decreases in the future. Red circle represents pairwise DEST between OR

populations and all the other populations; blue circle represents pairwise DEST between CR/SC populations and GR, GRN, MR, and OSG populations; and

green clusters represent pairwise DEST between GRN population and CR, GR, MR, OSG, and SC populations.

Figure 7. Results of CDPOP simulations projecting trajectories of genetic diversity (i.e., total number of alleles and expected heterozygosity) over 1000 years under different climate change scenarios. Simulations were based on the no-climate- change model (A and D), the RCP2.6 model (i.e., low greenhouse gas concentration scenario; B and E), and RCP8.5 model (i.e., high greenhouse gas concentration scenario; C and F). Colored lines represent extant populations. Areas of light gray showed 95% confidence intervals of each population. Figure S1. (A) Log-likelihood [Ln P(k)] and (B) ∆k for each k over the 10 replicates in

STRUCTURE. Error bars are standard deviation. Figure S2. Potential suitable habitat for Cumberlandia monodonta identified using ecological niche models (ENMs) using both “current” bioclimatic conditions (interpolations of observed data from 1950 to 2000) and landscape variables,

x and projections of four near-future climate scenarios (low and high greenhouse gas concentration in 2050, average for 2041-2060; and in 2070, average for 2051-2080; IPCC 5th Assessment). Low and high greenhouse gas concentrations were based on representative concentration pathways (RCP2.6 and RCP8.5, respectively). Gradient colors represent the probability of suitable habitat in percentage. Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs. Figure S3. Potential suitable habitat for Cumberlandia monodonta identified using ecological niche models (ENMs) using “current” bioclimatic conditions (interpolations of observed data from 1950 to 2000), and projections of four near-future climate scenarios (low and high greenhouse gas concentration in 2050, average for 2041-2060; and in 2070, average for 2051-2080; IPCC 5th Assessment). Low and high greenhouse gas concentrations were based on representative concentration pathways (RCP2.6 and RCP8.5, respectively). Gradient colors represent the probability of suitable habitat in percentage. Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs. Chapter 4 Figure 1. Map of the Rio Grande drainages in the southwest USA indicating major tributaries, reservoirs, and Popenaias popeii sampling sites (circles). Colors correspond to the parsimony network of COX sequences in Figure 3. The magnified inset map shows Black River sampling locations.

Figure 2. Four demographic scenarios tested using DIYABC: 1) all populations have constant size over time; 2) a small number of individuals founded the

ancestral Black River population followed by population expansion at te2; 3) a large number of founders colonized the ancestral Black River population

followed by a population bottleneck at te2; and 4) Rio Grande populations

experienced population bottleneck at te2 followed by population expansion at

xi te1, in addition to Scenario 3. All scenarios assume divergence between the

ancestral BR and RG populations at t2, divergence between BR-u and BR-d at

t1, and STRUCTURE-defined populations at the present time (t0). Timing of events is shown at right. See Table 2 for detailed labels. Figure 3. Parsimony network of COX sequences for Popenaias popeii. Each circle represents a unique haplotype; lines between haplotypes represent single- base-pair changes; black dots are inferred extinct or unsampled haplotypes. Haplotype frequency is relative to the size and number in the circle. The most common haplotype is shared between BR (n = 185) and RG (n = 10) individuals. Circle colors represent localities (Black River = dark gray; Devils River = light gray; Rio Grande = white).

Figure 4. Bar plots obtained from STRUCTURE, assigning individuals into k = 2 and k = 3 clusters. For k = 2, clusters consisted of BR (light gray) and RG (black) individuals. For k = 3, an additional division was observed in the BR populations (light gray, upstream sites; medium gray, downstream sites). Site name abbreviations at the bottom of the plot are a priori population assignments. Figure 5. Potential distribution of Popenaias popeii identified using ecological niche modeling under current bioclimatic conditions (1950 to 2000) and projections back to two paleoclimatic periods (last interglacial, 120 – 140 ka; last glacial maximum, 21 ka). Models included major rivers of the Rio Grande watershed in the USA and Mexico. Black dots on the Present map represent occurrence points included in the ENMs. Scale bars in the bottom-left corners represent 200 km.

Figure S1. Relationships between pairwise genetic differences (ST for mtDNA

sequences and DEST for microsatellites) and geographic (river) distances among localities. Overall relationships are shown in A and D. Overall relationships were decomposed into within river (B and E) and among rivers (C and F). Colors show relationships within the Black River localities (black dots), within the Rio Grande localities (white dots), and among the Black River and Rio Grande localities (grey dots).

xii Figure S2. (A) Log-likelihood [Ln P(k)] and (B) ∆k for each k over the 10 replicates in

STRUCTURE. Error bars are standard deviation. Figure S3. Potential distribution of three primary host fish species (Carpiodes carpio, Cyprinella lutrensis, and Moxostoma congestum) identified using ecological niche modeling under present climatic conditions (1950 to 2000) and projections of two paleoclimatic (last interglacial, 120 – 140 ka; last glacial maximum, 21 ka). Models included major rivers of the Rio Grande watershed in the United States and Mexico. Black dots on the Present map represent occurrence points included in the ENMs. Scale bars on the left-bottom corner showed 200 km. Chapter 5 Figure 1. Regional map (top-left corner) and the Pecos River and its perennial tributaries in Eddy County, New Mexico. Life history (LH) sites are shown as black dots. The study area for distance sampling is highlighted in grey. Figure 2. (a) Annual and seasonal survival and (b) annual finite rate of population growth (λ) of Popenaias popeii over a 15-year period estimated from the most parsimonious mark-and-recapture model, and (c) total population size in the Black River estimated from λ and total abundance for 2012 estimated from distance sampling. In panel (a), the dotted line with circles represents the seasonal dataset and the solid line with diamonds represents the annual dataset. In panel (b), the dotted line indicates λ = 1 and asterisks (*) represent λ values significantly different than 1. Error bars are 95% confidence intervals. Figure 3. Survival of Popenaias popeii plotted against minimum monthly discharge of the Black River. The line represents the best-fit regression line. Error bars are 95% confidence intervals. Figure 4. Mussel distributions along transect lines for riffle (dark grey) and pool (light grey) habitats. Number of mussels found per distance category from the bank for riffle and pool habitats is shown in bar graphs. Cumulative frequencies for riffle (solid line) and pool (dashed line) habitats are shown in line graphs. The

xiii vertical dotted line indicates the boundary between the river channel and riverbank. Figure 5. Monthly climatic and hydrological trends after removing seasonal effects (grey lines) in southeast New Mexico and the Black River. Climatic measures include mean monthly temperature (a), total monthly precipitation (b) and Palmer hydrological drought index (PHDI; c) from 1895-2012, while the hydrological measure is mean monthly discharge of the Black River (d) from 1948-2012. Black lines are smooth splines (smoothing parameter = 0.5). Asterisks (*) indicate statistical significance (P < 0.05) of Mann-Kendall trend tests. Figure 6. Conceptual diagram of the effects of climate change and increases in anthropogenic activities on aquatic invertebrate survival. Arrows indicate directional effects, and symbols along the arrows represent positive (+) and negative (-) effects. Chapter 6 Figure 1. Heat map showing the probability of extinction over 300 years relative to the number of juveniles-per-female and juvenile mortality rate based on Vortex population simulations of Cumberlandia monodonta (a) and Popenaias popeii (b). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles-per-female and juvenile mortality rate, with 100 replicate simulations of 300 years each. Note that the scale on x-axis varies between species. Figure 2. Heat map showing finite rate of population growth () relative to number of juveniles-per-female and juvenile mortality rate based on Vortex population simulation of Cumberlandia monodonta (a) and Popenaias popeii (b). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles-per-female and juvenile mortality rate, with 100 replicate simulations of 300 years each. Note that the scale on x-axis varies between species. Figure 3. Heat map showing the probability of genetic diversity persistence (a and c:

mean allelic richness, AR; b and d: mean expected heterozygosity, HE) relative

xiv to initial population size and mean finite rate of population growth () based on Vortex population simulation of two populations of Cumberlandia monodonta: upper Mississippi population (a and b) and lower Mississippi population (c and d). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles per female and initial population size, with 100 replicate runs per scenario. Figure 4. Heat map showing the probability of genetic diversity persistence (a and c:

mean allelic richness, AR; b and d: mean expected heterozygosity, HE) relative to initial population size and mean finite rate of population growth () based on Vortex population simulation of two populations of Popenaias popeii: Rio Grande population (a and b) and Black River population (c and d). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles per female and initial population size, with 100 replicate runs per scenario.

DEDICATION

To my wife, my mom, my dad, and my cats

xvi

ACKNOWLEDGEMENTS

My dissertation projects could not have been done without the support of numerous people. First and foremost, I would like to thank four people who greatly supported my PhD life and my professional development as a scientist. Dr. Dave Berg, my major advisor, was an amazing mentor who spent countless hours discussing evolution, conservation, and all of the various aspects of one’s life as a PhD student. He was encouraging and patient throughout my entire time at Miami. His confidence in me allowed me to pursue a number of different research projects. Brian Lang has been a constant supporter, confidant, and friend over the years. Without him, I could not have finished the half of my dissertation projects conducted in New Mexico. I will miss his delicious food in camp on the Black River and his corny jokes. Dr. John Harris introduced to, and taught me everything about, mussels when I was an undergraduate. He always gave me encouragement and inspiration. And finally, I want to thank Dr. Alan Christian who provided me a great opportunity to work on freshwater mussels when I was an undergraduate. This introduction to these fascinating creatures, and his own experiences at Miami, led me to pursue my PhD with Dave. I would like to thank Dr. Bruce Cochrane, Dr. Tom Crist, Dr. Brian Keane, and Dr. Rich Moore for agreeing to serve on my committee and supporting my research. They were more- than-generous with their precious time and expertise, and offered their encouragement and wisdom on my dissertation. My dissertation projects were made possible because of the efforts of previous researchers. Dr. Emy Monroe and Dr. Curt Elderkin helped with sampling and genetic analyses and provided me insight into understanding Cumberlandia monodonta, a most peculiar creature. Dr. Greg Moyer and Ashantyé Williams helped with initial screening of microsatellite markers. Dr. Todd Levine conducted host suitability experiments for Popenaias popeii and initiated mark- and-recapture studies, essential components of my dissertation. I thank my colleagues and friends in the Berg Lab and at Miami University: Nicole Adams, Jeff Moore, Mohammed Al-Saffar, Ashley Walters, Dr. Makiri Sei, Cayla Morningstar, Danielle Holste, Trevor Williams, Kristina Taynor, Emily Robinson, David Moulton, Alyssa McQueen, and Megumi Sugita, for field and/or laboratory assistance. Additionally, I thank Dr. Andor Kiss and Xiaoyun Deng at the Center for Bioinformatics and Functional Genomics for

xvii their extensive help with DNA sequencing, microsatellite genotyping, and other molecular analyses. I would like to extend thanks to many people who assisted me with field work and sample collection: Susan Oetker (US Fish and Wildlife Service), Steve McMurray and Scott Faiman (Missouri Department of Conservation), John Caldwell (New Mexico Department of Game and Fish), Mike Davis and Bernard Sietman (Minnesota Department of Natural Resources), Don Hubbs (Tennessee Wildlife Resources Agency), Dr. Jess Jones (Virginia Tech University), Steve Ahlstedt (US Geological Survey), Bill Posey (Arkansas Game and Fish Commission), Chad Lewis (Lewis Environmental Consulting), Leroy Koch (US Fish and Wildlife Service), Dr. Monte McGregor and Jacob Culp (Kentucky Department of Fish and Wildlife Resources), Dr. Lyuba Burlakova and Dr. Alexander Karatayev (Buffalo State University), and Tom Miller (Laredo Community College). Additionally, I thank Melissa Youngquist (Miami University), Cecilia Franz Berg (Miami University), Tamara Smith (US Fish and Wildlife Service), Steve McMurray, Scott Faiman, Berg lab members, and my committee members for their helpful comments on drafts of manuscripts. Lastly and most importantly, I thank my amazing wife, Melissa, and my parents. Melissa has always supported, encouraged, inspired, and loved me. She is the one who supported me the most mentally and emotionally during my PhD life. Finally, even though my parents never knew what their son was doing in a foreign country, they were always supportive in every endeavor that I undertook. My dissertation was supported financially by the New Mexico Department of Game and Fish, the US Fish and Wildlife Service, the Department of Biology, the Graduate School (via a Dissertation Scholarship), and the Graduate Student Association.

xviii General Introduction

Global biodiversity is currently facing a crisis of escalating species extinction and habitat loss as a consequence of human-induced disturbances. When species are threatened with extinction, it is necessary to develop strategies for species conservation in order to prevent extinction and recover species to the point at which protection is no longer necessary. Both governmental and non-governmental conservation organizations work to assess the conservation status of species, subspecies, and populations in order to achieve these goals (Scott et al. 2005; Tear et al. 2005). The effort by these organizations provides the foundation for making informed conservation decisions and species recovery planning, and ultimately, conserves global biodiversity at all levels of biological organization, from genetic diversity to ecosystem diversity. However, endangered species conservation is an enormously complex task, which includes identification of phenomena that affect the loss, maintenance, and restoration of biodiversity and advocate for sustaining evolutionary processes that promote genetic, population, species, and ecosystem diversity (Neel et al. 2012). Since the Endangered Species Act (ESA) of 1973 was enacted, more than 1000 animal and plant species in the USA that are listed as endangered or threatened have had recovery plans implemented (USFWS 2015). Conservation of endangered species requires a comprehensive approach to evaluate conservation status of a given species, develop optimal recovery plans, and establish quantitative recovery criteria, in order to remove the necessity of protection. Such an approach includes, but is not limited to, population genetic assessment to infer intra-population genetic diversity and population genetic structure (Frankham 2010), understanding of demography and population dynamics to estimate population size and vital rates (Neel et al. 2012), and extinction risk assessment to evaluate conservation strategies and recovery criteria (Akçakaya & Sjögren-Gulve 2000). For example, a comprehensive approach to the development of recovery criteria for the endangered Mexican wolf (Canis lupus baileyi) revealed that decreasing dispersal rates greatly increased extinction risks in small populations (Carroll et al. 2014). The authors developed population viability analyses (PVA) integrated with habitat, genetic, and demographic data and recommended population connectivity goals to conservation planners. Such comprehensive approaches—integrating genetic and demographic assessments to model demographic and genetic trajectory into the future—require detailed estimates of species-specific variables (Reed et al. 2002); however, these data are often only available for taxa that are well-studied.

1 In this dissertation, I demonstrate such a comprehensive approach for evaluating the conservation status of two imperiled freshwater mussel species and provide guidance for development of species recovery plans (Figure 3). I focus on two freshwater mussel species, Cumberlandia monodonta and Popenaias popeii (Bivalvia: Unionoida) as case studies. Freshwater mussels are amongst the most endangered groups of animals in the world (Lydeard et al. 2004; Strayer et al. 2004), yet conservation efforts are hindered by a lack of ecological and evolutionary information (Strayer et al. 2004). Among ~300 recognized species in North America, over 70% are presumed extinct, endangered, threatened, or proposed for listing (Williams et al. 1993). Most freshwater mussels possess a complex life cycle in which larvae (glochidia) are obligate parasites of vertebrate hosts. Cumberlandia monodonta was listed as endangered under the ESA in 2012 (USFWS 2012) due to severe decline in the number of occurrences; the recovery plan for this species has not yet been developed as of April 2015. This species was distributed throughout the Mississippi River system in historic times, including the mainstems of the Mississippi, Ohio, and Tennessee rivers and 41 tributaries. However, the species currently occurs in a few locally abundant populations scattered across approximately 20 streams within the historic range (Figure 1; USFWS 2012). The life history of the species has been extensively studied (reviewed in Butler 2002); however, its host species is still unknown. Popenaias popeii is currently a candidate for listing under the ESA (Priority 8; USFWS 2014). This species is endemic to the Rio Grande drainage in the USA and Mexico, and coastal Gulf of Mexico drainages in northern Mexico; however, currently P. popeii populations in the USA are restricted to a few localities in the mainstem of the Rio Grande and two of its tributaries (Figure 2; Carman 2007). Previous studies have identified reproductive characteristics, including the timing of reproduction and its host species (Smith et al. 2003; Levine et al. 2012); however, demographic and genetic characteristics have not yet been assessed. Berg et al. (2008) drew attention to the importance of developing conservation strategies for freshwater mussels that integrate theory from population biology and conservation genetics. Although biology of adult freshwater mussels is relatively well understood, information regarding population genetics and biology of early life stages (i.e., glochidia and juveniles) is often limited. In order to inform conservation and management of these species, my dissertation describes a series of studies to assess distribution-wide population genetic diversity and structure, demography and population

2 dynamics, and extinction risks in order to inform potential recovery strategies for these imperiled species (Figure 3). To assess population genetic diversity and structure, highly variable genetic markers such as microsatellites are necessary for high-level resolution of genetic structure within and among extant populations of the species. In Chapters 1 and 3, I characterize novel microsatellite markers for C. monodonta and P. popeii, respectively. While I employ a conventional enrichment library technique for C. monodonta, I use a de novo sequencing approach via next-generation sequencing for P. popeii. These microsatellite markers are used for population genetics studies of C. monodonta (Chapter 2) and P. popeii (Chapter 4). Chapter 2 uses multiple genetic markers (i.e., mitochondrial DNA [mtDNA] gene sequences and microsatellite loci developed in Chapter 1) to examine population genetic diversity and structure of C. monodonta throughout its range. A previous study revealed that C. monodonta is composed of two genetically distinct populations within the current range, which diverged during the Pleistocene glaciation (Inoue et al. 2014). I assess fine-scale population structure of C. monodonta using data from this previous study (Inoue et al. 2014) supplemented with additional populations, and then investigate the effects of future climate change on distributional shifts in suitable habitats and population genetic connectivity. I develop an ecological niche model (ENM) and a forward-time genetic simulation to accomplish these aims. In Chapter 4, I use mtDNA gene sequences and microsatellite loci developed in Chapter 3 to investigate historic demography and current population genetic structure of P. popeii throughout its range. Because current population fragmentation is likely due to anthropogenic disturbances (Carman 2007), I test the hypothesis that among-drainage population divergence occurred with European settlement of the region. I use ENMs with current climatic conditions to predict suitable environments for P. popeii and project the model into the past based on mid-to- late-Pleistocene climatic scenarios. In Chapter 5, I use long-term mark-and-recapture monitoring to determine population dynamics of P. popeii over a 15-year period. I then use adaptive distance-sampling methods to estimate density and total population size of P. popeii in the Black River, New Mexico. I seek to (1) investigate differences in demographic parameters in various microhabitats, (2) evaluate effects of hydrological cycles on demography, and (3) estimate total abundance of P. popeii in the Black River taking into account heterogeneous micro- and macrohabitats.

3 Finally in Chapter 6, I use demographic and population genetic information acquired from the previous chapters and previously published data (e.g., Baird 2000) to develop recovery strategies and criteria for C. monodonta and P. popeii. I integrate population genetic simulations into population viability analysis (genetic PVA) to infer extinction risks and population growth rates by varying a series of demographic parameters and predict the persistence of genetic diversity over time. Furthermore, I review endangered species recovery plans for listed mussel species to determine current trends in the use of genetic PVA in recovery planning. Through my dissertation, I demonstrate a comprehensive approach to conservation biology for imperiled freshwater mussels integrating population genetics with demography in order to develop objective, measureable goals for quantitative recovery strategies and criteria. Results of my dissertation provide essential information not only for developing recovery strategies for imperiled species, but also insight into evolution of stream-dwelling organisms inhabiting the same regions as these mussels.

REFERENCES Akçakaya HR, Sjögren-Gulve P (2000) Population viability analyses in conservation planning: an overview. Ecological Bulletins, 48, 9-21. Baird MS (2000) Life history of the spectaclecase, Cumberlandia monodonta Say, 1829 (Bivalvia, Unionoidea, Margaritigeridae) M.S. thesis, Southwest Missouri State University, Springfield, Missouri. Berg DJ, Levine TD, Stoeckel JA, Lang BK (2008) A conceptual model linking demography and population genetics of freshwater mussels. Journal of the North American Benthological Society, 27, 395-408. Butler RS (2002) Status assessment report for the spectaclecase, Cumberlandia monodonta, occurring in the Mississippi River system (U.S. Fish and Wildlife Service Region 3, 4, 5, and 6), p. 69. The Ohio River Valley Ecosystem Team, Asheville, NC. Carman SM (2007) Texas hornshell Popenaias popeii recovery plan, p. 57. New Mexico Department of Game and Fish, Santa Fe, NM. Carroll C, Fredrickson RJ, Lacy RC (2014) Developing metapopulation connectivity criteria from genetic and habitat data to recover the endangered Mexican wolf. Conservation Biology, 28, 76-86.

4 Frankham R (2010) Challenges and opportunities of genetic approaches to biological conservation. Biological Conservation, 143, 1919-1927. Inoue K, Monroe EM, Elderkin CL, Berg DJ (2014) Phylogeographic and population genetic analyses reveal Pleistocene isolation followed by high gene flow in a wide-ranging, but endangered, freshwater mussel. Heredity, 112, 282-290. Levine TD, Lang BK, Berg DJ (2012) Physiological and ecological hosts of Popenaias popeii (Bivalvia: Unionidae): laboratory studies identify more hosts than field studies. Freshwater Biology, 57, 1854-1864. Lydeard C, Cowie RH, Ponder WF, Bogan AE, Bouchet P, Clark SA, Cummings KS, Frest TJ, Gargominy O, Herbert DG, Hershler R, Perez KE, Roth B, Seddon M, Strog EE, Thompson FG (2004) The global decline of nonmarine mollusks. BioScience, 54, 321- 330. Neel MC, Leidner AK, Haines A, Goble DD, Scott JM (2012) By the numbers: how is recovery defined by the US Endangered Species Act? BioScience, 62, 646-657. Reed JM, Mills LS, Dunning Jr. JB, Menges ES, McKelvey KS, Frye R, Beissinger SR, Anstett M-C, Miller P (2002) Emerging issues in population viability analysis. Conservation Biology, 16, 7-19. Scott JM, Goble DD, Wiens JA, Wilcove DS, Bean M, Male T (2005) Recovery of imperiled species under the Endangered Species Act: the need for a new approach. Frontiers in Ecology and the Environment, 3, 383-389. Smith DG, Lang BK, Gordon ME (2003) Gametogenetic cycle, reproductive anatomy, and larval morphology of Popenaias popeii (Unionoida) from the Black River, New Mexico. Southwestern Naturalist, 48, 333-340. Strayer DL, Downing JA, Haag WR, King TL, Layzer JB, Newton TJ, Nichols SJ (2004) Changing perspectives on pearly mussels, North America's most imperiled animals. BioScience, 54, 429-439. Tear TH, Kareiva P, Angermeier PL, Comer P, Czech B, Kautz R, Landon L, Mehlman D, Murphy K, Ruckelshaus M, Scott JM, Wilhere G (2005) How much is enough? The recurrent problem of setting measurable objectives in conservation. Bioscience, 55, 835- 849.

5 USFWS (US Fish and Wildlife Service). 2012. Endangered and threatened wildlife and plants; determination of endangered status for the sheepnose and spectaclecase mussels throughout their range, final rule. Federal Register 77, 14914-14949. USFWS (US Fish and Wildlife Service). 2014. Endangered and threatened wildlife and plants; review of native species that are candidates for listing as endangered or threatened; annual notice of findings on resubmitted petitions; annual description of progress on listing actions; proposed rule. Federal Register 79, 72450-72497. USFWS (U.S. Fish and Wildlife Service). 2015. Summary of listed species, listed populations and recovery plans. Available at https://ecos.fws.gov/tess_public/pub/Boxscore.do. Williams JD, Warren ML, Jr., Cummings KS, Harris JL, Neves RJ (1993) Conservation status of freshwater mussels of the United States and Canada. Fisheries, 18, 6-22.

Figure 1. Map of the central United States indicating historic and current distribution of Cumberlandia monodonta. Colored watersheds represent historic (gray and blue) and current (blue only) distributions obtained from NatureServe (http://explorer.natureserve.org; accessed June 25, 2015).

Figure 2. Map of the Rio Grande drainage in the southwest United States and northern Mexico indicating historic and current distribution of Popenaias popeii. Colored watersheds represent historic (gray and blue) and current (blue only) distributions obtained from NatureServe (http://explorer.natureserve.org; accessed June 25, 2015).

8 Cumberlandia monodonta Popenaias popeii

Genetics Demography Genetics Demography

Chapter 1: Chapter 3: Microsatellite loci Microsatellite loci development development

Chapter 5: Published data Demography & (e.g., Baird 2000) population dynamics

Chapter 2: Chapter 4: Population genetics Phylogeography & & genetic simulations Population genetics

Chapter 6: Population viability analysis integrated with genetic assessments

Figure 1. Outline for the dissertation chapters.

9 Chapter 1: Isolation and characterization of 17 polymorphic microsatellite loci in the spectaclecase, Cumberlandia monodonta (Bivalvia: Margaritiferidae)

Inoue K, Moyer GR, Williams A, Monroe EM, Berg DJ (2011) Isolation and characterization of 17 polymorphic microsatellite loci in the spectaclecase, Cumberlandia monodonta (Bivalvia: Margaritiferidae). Conservation Genetics Resources, 3, 57-60.

ABSTRACT We isolated 29 microsatellite loci from Cumberlandia monodonta, a freshwater mussel species that has experienced population declines throughout its range. Seventeen loci were polymorphic, with 3–13 alleles, observed heterozygosity values of 0.375–1.00, and 38% of alleles found in more than one population. These loci should be useful for describing population genetic diversity, which will facilitate ongoing conservation efforts for C. monodonta.

Keywords: Cumberlandia monodonta, spectaclecase, Margaritiferidae, microsatellite primers

INTRODUCTION Freshwater mussels (families Margaritiferidae and Unionidae) are among the most endangered groups of animals in North America (Lydeard et al. 2004). Cumberlandia monodonta (Say 1829), the spectaclecase, was historically widespread and abundant throughout the Mississippi River system (Williams et al. 2008). However, habitat alteration has led to the extirpation and/or fragmentation of numerous populations. Distribution extent has declined as populations have been extirpated, and many of the remaining populations have been reduced in size (Watters et al. 2009). Several recent surveys have found only single live specimens (Harris et al. 2009). Extirpation, drastic population reduction, and continuing decline has led to candidate status for this species under the US Endangered Species Act (USFWS 2009). Phylogeographic analysis of C. monodonta using allozymes and mitochondrial DNA sequencing showed a lack of genetic structure among putative populations (Monroe 2008); sequencing of the nuclear internal transcribed spacer region showed a lack of within—population variation (Elderkin 2009). These results suggest that C. monodonta has high gene flow among populations and/or that post-Pleistocene dispersal throughout the Mississippi River basin has

10 homogenized populations (Monroe 2008). Therefore, detection of variation among populations will require the use of highly variable genetic markers such as microsatellite loci. Since Eackles and King (2002) first isolated and characterized microsatellite loci from an endangered mussel species, such loci have been isolated from other endangered freshwater mussels (e.g., Geist et al. 2003) and used to examine levels of genetic variation and population structure (e.g., Jones et al. 2006). Such information uncovers patterns of genetic structure, while also suggesting strategies for conserving genetic variation and developing effective population management practices. The goal of our study was to develop and characterize microsatellite markers from C. monodonta for future population genetics studies in order to inform conservation and management of this species.

METHODS We extracted genomic DNA from mantle tissue frozen at −80°C, using standard phenol– chloroform extraction with ethanol precipitation. Whole genomic DNA was quantified using a nano-spectrophotometer. We employed Genetic Identification Services (GIS; Chatsworth, CA, USA) for the development of microsatellite loci. Microsatellite libraries were enriched for two trinucleotide motifs (AAC, ATG) and three tetranuceotide motifs (AAAC, CATC, TAGA). Of the 192 sequence clones provided by GIS, 29 contained microsatellite loci with enough flanking sequence for primer design. Primers were developed using GENEIOUS v4.8 (Drummond et al. 2010). Twenty-three of 29 primer pairs produced reliable PCR amplification in individuals collected from the Gasconade and Meramec rivers, MO, the Clinch River, TN, and the St. Croix River, WI. We performed amplifications in a 20 μL reaction volume using PCR Master Mix (QIAGEN, Inc.), 0.2 μM of forward and reverse primers (Table 1), and 20–40 ng of DNA template. PCR conditions consisted of: initial denaturing at 94°C for 2 min, followed by 35 cycles at 94°C for 30 s, annealing at 52–55°C for 30 s, extension at 72°C for 1 min, and final extension at 72°C for 5 min. For problematic primer pairs, we also used a touchdown PCR method (Don et al. 1991) with annealing temperature decreasing 1°C every cycle from 60 to 52°C and 35 annealing cycles at 52°C (Table 1). We performed fragment analyses on ABI 3130 or 3730 Genetic Analyzers with LIZ600 size standard (Applied Biosystems, Inc.). We used

PEAKSCANNER v1.0 (Applied Biosystems, Inc.) to score alleles and TANDEM v1.07 (Matschiner &

11 Salzburger 2009) to assign integer numbers to DNA fragment sizes. We estimated allelic richness (number of alleles at each locus; NA) and observed heterozygosities (HO), and tested whether allele frequencies met Hardy–Weinberg expectation (HWE) using GENALEX v6.3 (Peakall & Smouse 2006). We also estimated the number of shared alleles (those found in more than one population) at each locus. We used MICRO-CHECKER (van Oosterhout et al. 2004) to detect null alleles and estimate their frequencies. We used GENEPOP v4.0.10 (Rousset 2008) to conduct exact tests of pairwise linkage disequilibrium.

RESULTS AND DISCUSSION Of the 29 microsatellite loci tested, we confirmed 17 loci as polymorphic when tested over eight individuals from four populations. Six loci were monomorphic and the remaining were unresolved (Table 1). Allelic richness ranged from three to 13 alleles, with an average of 7.53 alleles per locus. When we considered all eight individuals as a single population, the average observed heterozygosity over 17 loci was 0.792 (range 0.375–1.000); the average number of shared alleles was 2.88 per locus (range 1–5), with 38% of alleles found in more than one population. We found evidence of a potential null allele at one locus (CMD102), with a frequency of 0.2739. However, this result may be due to the small number of individuals sampled. One locus showed significant deviation from HWE when all individuals were pooled (CMD102; P < 0.05) and we did not detect any evidence of linkage disequilibrium. From this study, we developed and characterized a minimum of 17 polymorphic loci from C. monodonta. Such a large number of loci suggests that we will be able to provide high-level resolution of genetic structure within and among extant populations of C. monodonta. As a result, estimates of population parameters such as gene flow, effective population size, etc. are likely to be robust. Such information will be of great value to agencies charged with conservation of this imperiled species.

ACKNOWLEDGEMENTS We thank Curt Elderkin for collection of mussel samples, Zac Taylor for technical advice, Alyssa McQueen for laboratory assistance, and Genetic Identification Services (GIS) for construction and sequencing of enriched microsatellite libraries.

12 REFERENCES Don RH, Cox PT, Wainwright BJ, Baker K, Mattick JS (1991) 'Touchdown' PCR to circumvent spurious priming during gene amplification. Nucleic Acids Research, 19, 4008. Drummond AJ, Ashton B, Cheung M, Heled J, Learse M, Moir R, Stones-Havas S, Thierer T, Wilson A (2010) Geneious v4.8, Available from http://www.geneious.com/. Eackles MS, King TL (2002) Isolation and characterization of microsatellite loci in Lampsilis abrupta (Bivalvia: Unionidae) and cross-species amplification within the genus. Molecular Ecology Notes, 2, 559-562. Elderkin CL (2009) Intragenomic variation in the rDNA internal transcribed spacer (ITS1) in the freshwater mussel Cumberlandia monodonta (Say, 1828). Journal of Molluscan Studies, 75, 419-421. Geist J, Rottmann O, Schroder W, Kuhn R (2003) Development of microsatellite markers for the endangered freshwater pearl mussel Margaritifera margaritifera L. (Bivalvia: Unionoidea). Molecular Ecology Notes, 3, 444-446. Harris JL, Posey II WR, Davidson CL, Farris JL, Oetker SR, Stoeckel JN, Crump BG, Barnett MS, Martin HM, Matthews MW, Seagraves JH, Wentz NJ, Winterringer R, Osborne C, Christian AD (2009) Unionoida (Mollusca: Margaritiferidae, Unionidae) in Arkansas, third status review. Journal of the Arkansas Academy of Science, 63, 50-86. Jones JW, Neves RJ, Ahlstedt SA, Hallerman EM (2006) A holistic approach to taxonomic evaluation of two closely related endangered freshwater mussel species, the oyster mussel Epioblasma capsaeformis and tan riffleshell Epioblasma florentina walkeri (Bivalvia: Unionidae). Journal of Molluscan Studies, 72, 267-283. Lydeard C, Cowie RH, Ponder WF, Bogan AE, Bouchet P, Clark SA, Cummings KS, Frest TJ, Gargominy O, Herbert DG, Hershler R, Perez KE, Roth B, Seddon M, Strog EE, Thompson FG (2004) The global decline of nonmarine mollusks. Bioscience, 54, 321- 330. Matschiner M, Salzburger W (2009) TANDEM: integrating automated allele binning into genetics and genomics workflows. Bioinformatics, 25, 1982-1983. Monroe EM (2008) Population genetics and phylogeography of two large-river freshwater mussel species at large and small spatial scales Ph.D. dissertation, Miami University, Oxford, Ohio.

13 Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6, 288-295. Rousset F (2008) GENEPOP'007: a complete re-implementation of the GENEPOP software for Windows and Linux. Molecular Ecology Resources, 8, 103-106. USFWS (US Fish and Wildlife Service). 2009. Endangered and threatened wildlife and plants; review of native species that are condidates for listing as endangered or threatened; annual notice of fininfs on resubmitted petition; annual description of progress on listing actions; proposed rule. Federal Register 74, 57803-57878. van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) MICRO-CHECKER: Software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes, 4, 535-538. Watters GT, Hoggarth MA, Stansbery DH (2009) The freshwater mussels of Ohio Ohio State University, Columbus, OH. Williams JD, Bogan AE, Garner JT (2008) Freshwater mussels of Alabama and the Mobile Basin in Georgia, Mississippi & Tennessee The University of Alabama Press, Tuscaloosa, AL.

14 TABLE Table 1. Characterization of 29 microsatellite loci for Cumberlandia monodonta.

Repeat TA Size Range Shared Locus Primer Sequence (5'-3') Motif (˚C) (bp) N NA HO Alleles Polymorphic CMA1 F: CATTTGGGCGATGATGTG AAC 52 282-288 7 3 0.429 3 R: ATGGACAGGCGGTAGCAC CMA2 F: TCCCTCTGTTGTTGATGATTT AAC 52 211-217 8 3 0.625 3 R; CTGCACCCATTCCACTACTAG CMB3 F: CATTGGGAAGGGAGTACATG ATG 52 257-293 8 9 1.000 5 R: GGATGGAACAATAATAAGGGTG CMB6 F: GCAACAGATCGAGTTACCTCAG ATG 52 235-259 7 7 0.857 4 R: AGCTTTTAATCTTTGGGTCAGC CMC5 F: ACCGCTACCTCCTTCTGG CATC TD 273-301 7 6 1.000 3 R: AAATGGGCAGCACTCAAC CMC11 F: CGAAAATACAAACACCTTGCA CATC 52 233-313 8 13 1.000 3 R: CTTGCCGAATGACACACC CMC109 F: TAGCCAGAACAGCCAGTG CATC TD 209-249 7 6 0.714 3 R: CGAGTGTATTTGTGCTCTCAC CMC115 F: TGCAGCCAATAATGTATGAGAA CATC 52 171-207 8 7 1.000 2 R: AGTGGGAGACATCAGTTACTGG CMC126 F: GGATTGCCTTGTAACTCCAAC CATC 55 107-115 8 3 0.500 2 R: AACCCTAAGTGCTGATGCTG CMD1 F: TGGAACAGAACCACTCATGTA TAGA 52 91-143 8 9 0.875 2 R: TACGTTTACCTGCGTATCTAGC CMD9 F: AAGCAAATGACGGCAGTT TAGA 52 281-301 7 5 0.714 4 R: TGGTGGGTCTGGTAGTAGG CMD101 F: AAGCTCGCAAGTGACATTC TAGA TD 276-344 8 9 0.625 2 R: GCAGGCATACTTATTACAAAGG CMD102 F: TAGGTGGTTAGGTAGGAAGAT TAGA TD 312-396 8 9 0.375 1 R: GACTGGGTAACTGCTATGTG CMH7 F: TCAAGCCAACCTTTGCATA TAGA 52 225-313 8 12 1.000 2 R: GCAATATGGGGTCGTGTC CMH8 F: CCTTCTCAACGTCGATTGT TAGA 52 266-310 8 8 0.750 3 R: CCGATCATTCCATCACTG CMH9 F: TGCTGGTTGTTCTATGGC TAGA 52 241-297 7 7 1.000 3 R: ATTGTCCGTTCTTGCTACAG CMH12 F: GTTCCTCTTCTTCTTCTTGTTG TAGA 52 202-362 8 12 1.000 4 R: CTGAATATGCCTGTTGTGGTA Monomorphic CMA11 F: CAGAGCCACAAAACTTTCTAG AAC 52 174 8 1 0.000 n/a R: GGCAACAGCATATTACTTAGG CMB8 F: TCCTCCTCCTCATCATCAA ATG 52 146 8 1 0.000 n/a R: TGCATGTTTTTAGTGACCATG CMC105 F: CTTGACGATTTTGGAGTTCTG CATC 52 278 8 1 0.000 n/a R: GCTATTCGCAATTCTTCAGTG CMG1 F: TCATTCGTTCAATGCTACATTC AAAC 52 170 5 1 0.000 n/a R: CTATGGAGAGGTCGAAACAGTT CMG3 F: TTTTCAGTCAGTCAGAGAGTCA AAAC 52 183 4 1 0.000 n/a R: ATGCTATGTGCAGTTGTGTTAC CMG11 F: CAGAATGCTGGACTGACTACAG AAAC 52 216 7 1 0.000 n/a R: TAAAGCAAGGCTGAACATAATG (table continues)

Table 1. (continued) Unresolved CMC1 F: GTGTTTCACAACAGTTTTCTCA CATC TD n/a 8 n/a n/a n/a R: GCTGGAACCTTGCAGTTC CMD8 F: AACAATCCATTTGGGTACAAG TAGA 52 n/a 8 n/a n/a n/a R: TTGCTCCCACTATTTCACAG CMD10 F: TGCAGTTCAAGCACATCTT TAGA 52 n/a 8 n/a n/a n/a R: GACGGACATACAGACAGTCAG CMD115 F: ATCCATCCATCCATCCATC TAGA TD n/a 8 n/a n/a n/a R: CACGCGACAGCTACTTATATTC CMG6 F: AATCCGTAGGTAACACAACG AAAC 52 n/a 8 n/a n/a n/a R: TCATTCAGTCAGACAGTCAGTC CMG10 F: CCCACACACAAACACACTCTC AAAC TD n/a 8 n/a n/a n/a R: ATCATCACACACGCCAACATA Mean 5.83 0.59 2.88 (standard error) (0.82) (0.084) (0.24) Shown are locus names, primer sequences, annealing temperature (TA) where TD is a touchdown

PCR method, size range of alleles in base pairs (bp), sample size (N), number of alleles (NA), observed heterozygosity (HO), and number of shared alleles (those found in more than one population). F = forward sequence; R = reverse sequence; n/a = not available.

16 Chapter 2: Predicting the effects of climate change on population connectivity and genetic diversity in riverine systems

Co-author: David J. Berg1 1 Department of Biology, Miami University, Hamilton, Ohio 45011

ABSTRACT In the face of changes in global climate regimes, organisms often respond to temperature increases by shifting their ranges poleward or to higher altitudes. However, the direction of range shifts in riverine systems is less clear. Because rivers are dendritic networks, there is only one dispersal route from any given location to another. Thus, range shifts are only possible if branches are connected by suitable habitat, and stream-dwelling organisms can disperse through these branches. Cumberlandia monodonta is a useful species for investigating the effects of climate change on population connectivity because a majority of contemporary populations are panmictic. We combined ecological niche models (ENMs) with population genetic simulations to investigate the effects of climate change on population connectivity and genetic diversity of C. monodonta. The ENMs were constructed using bioclimatic and landscape data to project shifts in suitable habitat under future climate scenarios. We then used forward-time simulations to project potential changes in genetic diversity and population connectivity based on these range shifts. ENM results under current conditions indicated highly suitable habitat across rivers where C. monodonta is known to be abundant; populations in the upper Mississippi River remain connected by suitable habitat, likely supporting panmixia. Future climate scenarios projected northward and headwater-ward range contraction and drastic declines in habitat suitability for most extant populations throughout the Mississippi River Basin. Simulations indicated that climate change would greatly reduce genetic diversity and connectivity across populations. Results suggest that a single, large population of C. monodonta will become fragmented into several smaller populations, each of which will be isolated from remaining populations. Because C. monodonta is a widely distributed species, our results suggest that persistence and connectivity of stream-dwelling organisms will be significantly altered in response to future climate change.

17 Keywords: CDPOP, forward-time population genetic simulation, ecological niche modeling, dendritic network, Cumberlandia monodonta, spectaclecase, MAXENT

INTRODUCTION Over the last 100 years, mean global temperature has increased by 0.85˚C due to human- induced climate change, which is predicted to continue into the future (IPCC 2013). Such rapid changes in global climatic regimes impact ecological processes in aquatic and terrestrial ecosystems. When species encounter changes in local climatic conditions, they often respond by adjusting their phenology and physiology to match new climatic optima, as most species tolerate short-term variability in climate through phenotypic plasticity (Walther et al. 2002). However, if a species’ distributional range reflects its adaptive capability, it will need to shift its distributional range poleward or to higher altitude when climate warms (Hampe & Petit 2005; Jump & Penuelas 2005). Range shifts associated with climate change have been observed and predicted in plants and animal species worldwide (Parmesan & Yohe 2003; Chen et al. 2011); such range shifts are varied among individual species depending on habitat specificity and dispersal capability (Chen et al. 2011). In terrestrial systems, spatially heterogeneous landscapes often have a lattice-like architecture, where patches of suitable habitat are connected by multiple dispersal paths (Urban & Keitt 2001). Range shifts in response to climate change, therefore, depend on species’ dispersal abilities and the connectivity of suitable habitat across the landscape. Riverine systems, on the other hand, have a hierarchical dendritic structure, where physical flows often dictate distance and directionality of dispersal (Grant et al. 2007; Peterson et al. 2013). In a dendritic network, ecological processes are carried out within the network; for instance, the dispersal route is a single path from any given location to another. In general, connectivity of riverine systems is predicted to be greater through mainstems relative to headwaters because mainstems allow movement of species among branches (Fagan 2002; Peterson et al. 2013). Thus, range shifts are expected only if all branches are well connected with suitable habitats and stream-dwelling organisms are capable of dispersing through branches and mainstems. Despite this theoretical work, few empirical studies have demonstrated range shifts in response to climate change in dendritic networks (e.g., Booth et al. 2011).

18 Given current rates of anthropogenically driven environmental and climate change, many researches have focused on projecting the species-level impacts of such changes. Recently, evidence has emerged linking genetic variation and patterns of environmental and climatic change (Geffen et al. 2004; Habel et al. 2011; Row et al. 2014; Sexton et al. 2014). These studies indicated that changes in climatic patterns would potentially alter population genetic structure. For example, contemporary population genetic structure of Canada lynx (Lynx canadensis) is strongly correlated with a winter climate gradient across the Pacific-North American and North Atlantic oscillations (Row et al. 2014). Snow conditions between the climate systems are predicted to diverge in future-climate scenarios, which will reduce connectivity between eastern and western lynx populations and lead to a threefold increase in genetic differentiation. In riverine systems, a few studies have focused on integrating dendritic landscape structure into spatial patterns of genetic variation and gene flow of aquatic taxa (Hughes et al. 2009; Alp et al. 2012). However, while studies have demonstrated correlations between population genetic structure and dendritic landscape structure, principal mechanisms of genetic-climate interactions with range shifts in riverine systems have not been fully articulated (Heino et al. 2009). Riverine ecosystems comprise a wide variety of habitat types and disproportionately high species richness, both of which are vulnerable to anthropogenic disturbances (Dudgeon et al. 2006; Vörösmarty et al. 2010). It has become increasingly important to understand how such changes affect ecological processes in riverine systems. Here, we investigated how changes in future climate might alter distributions and population connectivity of stream-dwelling organisms. We used a wide-ranging freshwater mussel, Cumberlandia monodonta, as a model to examine such changes. This species was distributed throughout the Mississippi River system in historic times. Until the early 20th Century, its range included the mainstems of the Mississippi, Ohio, and Tennessee rivers and 41 tributaries; the species currently occurs in the Mississippi River Basin in a few locally abundant populations scattered across approximately 20 streams within the historic range, covering a great range in latitude (Figure 1; USFWS 2012). The only relatively strong populations remaining are in the Meramec and Gasconade rivers, Missouri; the St. Croix River, Minnesota/Wisconsin; the Ouachita River, Arkansas; and the Green River, Kentucky (USFWS 2014). As with most freshwater mussels, C. monodonta possesses a complex life cycle and its larvae (glochidia) are likely obligate parasite of vertebrate hosts. Although numerous host-suitability trials have been

19 conducted, the host species is still unknown (USFWS 2012). A previous study reconstructing the phylogeography of C. monodonta revealed that the species occupied at least two glacial refugia during the mid-Pleistocene, and that populations from these refugia became admixed and spread across the current species range during the Holocene (Inoue et al. 2014). Based on population genetic analyses, populations in tributaries of the upper Mississippi River (from the headwaters to the confluence with the Ohio River) form a panmictic population, while the Ouachita River populations in the lower Mississippi basin are genetically distinct and isolated from the panmictic population. Because poleward range shift is expected in response to climate warming, current populations of C. monodonta will likely be fragmented if climate change reduces habitat connectivity and prevents poleward shifts in distribution. In this study, we assessed fine-scale population structure of C. monodonta using data from a previous study (Inoue et al. 2014) supplemented with additional populations, and then investigated the effects of future climate change on distributional shifts in suitable habitats and population genetic connectivity. We performed both ecological and genetic simulations to accomplish these aims. We created an ecological niche model (ENM) based on current climatic and landscape conditions to predict suitable habitat across the range where C. monodonta persists, and then projected this model based on near-future climate scenarios to examine future changes in suitable habitat and therefore, species distribution. We then used a forward-time genetic simulation to project potential changes in genetic diversity and population connectivity under these future-climatic scenarios. If future-climate scenarios show poleward shifts in distribution of suitable habitats, we predicted that habitat fragmentation due to the presence of unsuitable habitats would restrict gene flow, and thus create genetically isolated populations from the panmictic population.

METHODS Sampling, genetic data collection, and analyses of population genetic diversity We used a total of 416 individuals of C. monodonta, including samples from a previous study (Inoue et al. 2014), from seven rivers in the Mississippi River Basin (Table 1 and Figure 1). The new samples were collected from 12 localities in the seven rivers using snorkeling, SCUBA, and tactile methods. We collected between 20 and 37 individuals from at least two locations in each river, except only six individuals were collected from one of the sites in the

20 Ouachita River (OR2); and only one location from the Green River was sampled (Table 1). These rivers represent four biogeographic provinces identified by Haag (2009): Tennessee- Cumberland (Clinch River sites), Upper Mississippi (Gasconade, Meramec, Osage, and St. Croix rivers), Ohioan (Green River), and Mississippi Embayment (Ouachita River). All rivers, except the Ouachita, are tributaries of the upper Mississippi River. Nondestructive samples were collected from foot and mantle tissues by rubbing mucus and epidermal cells using buccal swabs (Epicentre Biotechnologies, Madison, WI). Mussels were returned to the river bottom; samples were preserved in 95% ethanol and stored at -20˚C. Total genomic DNA was extracted from swab samples using the ArchivePure DNA Cell/Tissue Kit (5 Prime, Gaithersburg, MD). Extracted DNA was diluted to 10 ng/µL and used as a template in polymerase chain reactions (PCR) that amplified the mtDNA cytochrome oxidase I (COI) and 16 microsatellite loci (Inoue et al. 2011; Inoue et al. 2014). Procedures and conditions for PCR, sequencing, fragment analyses, and post-sequencing/post-fragment-analyses followed those in Inoue et al. (2014).

We estimated population genetic indices from mtDNA sequences using DNASP v5.10 (Librado & Rozas 2009). We calculated number of haplotypes (H), mean number of basepair differences (K), and mean nucleotide diversity (π) for each locality. Because sample sizes at each locality differed, we used rarefaction to estimate the number of haplotypes (HR) after correcting for sample-size bias. We built a haplotype network using HAPLOVIEWER (available at http://www.cibiv.at/~greg/haploviewer). Because HAPLOVIEWER requires a parsimony tree, we constructed a consensus parsimony tree using PHYLIP v3.695 (Felsenstein 2005). For microsatellite loci, we conducted exact tests for pairwise linkage disequilibrium and deviation from Hardy-Weinberg expectation (HWE) using GENEPOP v4.0.10 (Rousset 2008) for each locality. We estimated population genetic indices (mean number of alleles per locus, NA; and observed and expected heterozygosities, HO and HE) for each locality using GENALEX v6.3 (Peakall & Smouse 2006). Additionally, we used rarefaction to correct mean allelic richness

(rarefied number of alleles per locus; AR) and mean number of private alleles per locus (NRP) using ADZE v1.0 (Szpiech et al. 2008). We set standardized sample size to 20, and excluded one Ouachita River population (OR2) due to small sample size. Population genetic structure We examined partitioning of genetic variation among localities by performing an analysis of molecular variation (AMOVA) using HIERFSTAT v0.04-10 (Goudet 2005) in R v3.0.2 (R

21 Development Core Team 2011). We partitioned genetic variation into four hierarchical levels:

(1) total genetic variation among provinces (FPT), (2) genetic variation among rivers within province (FRP), (3) genetic variation among localities within river (FSR), and (4) total genetic variation among localities (FST). Statistical significance of the deviation of each F from 0 was estimated using 1000 permutations for each partition. We then estimated two indices of genetic differentiation among localities: pairwise FST in GENALEX and Jost’s DEST (Jost 2008) in

DEMETICS v0.8-7 (Gerlach et al. 2010) in R. We tested for statistical significance of deviation from 0 using 9999 permutations for pairwise FST comparisons and using 100 bootstraps for pairwise DEST comparisons. We used only the microsatellite dataset for AMOVA and genetic differentiation indices.

Using STRUCTURE v2.3.4 (Pritchard et al. 2000), we evaluated population genetic structure of C. monodonta without a priori assignment of individuals to populations. We used the admixture model and allowed for correlated allele frequencies to account for ancestral admixture in the dataset. We ran STRUCTURE with a burn-in period of 500,000 Markov chain Monte Carlo (MCMC) generations followed by 200,000 iterations for k = 1 through 10 with 10 replicates for each k. We evaluated the log-likelihood [lnP(k)] for each k and estimated ∆k using

STRUCTURE HARVESTER (Earl & vonHoldt 2012) to determine the most likely number of distinct clusters. We averaged each individual's admixture proportions over the 10 replicates for the best k using CLUMPP (Jakobsson & Rosenberg 2007), then produced graphical display results using

DISTRUCT (Rosenberg 2004). Projecting the effects of climate change on suitable habitat To predict suitable habitat for C. monodonta in the present and future throughout the Mississippi River Basin, we developed ENMs employing the maximum entropy algorithm implemented in MAXENT v.3.3.3 (Phillips et al. 2006). We used geo-referenced occurrence points generated in this study and from museum records within the last 50 years. These points represent the present-day distribution of C. monodonta in the Mississippi River Basin. Given that occurrence data often show strong spatial bias in sampling efforts, we used SDMTOOLBOX V1.0b (Brown 2014) to reduce spatial autocorrelation in the occurrence data by selecting one record within a 5-km radius (Phillips et al. 2009; Kramer-Schadt et al. 2013). We then created a layer of the Gaussian kernel density of sampling locations (i.e., a bias layer) with a bandwidth of 50 km to control for background sampling efforts. We set our modeling area to include a 5-km buffer

22 around ≥3rd order streams in the Mississippi River Basin. We obtained current and near-future estimates for 19 bioclimatic parameters from WorldClim (http://www.worldclim.org; Hijmans et al. 2005); eight landscape layers were obtained from WorldClim and the NASA Socioeconomic Data and Applications Center (SEDAC; available at http://sedac.ciesin.columbia.edu; Table S1). Values for bioclimatic variables were based on “current conditions” (interpolations of observed data from 1950 to 2000) and four future climate scenarios (low and high greenhouse gas concentrations in 2050, average from 2041 to 2060; and in 2070, average from 2051 to 2080; IPCC 5th Assessment). We used two representative concentration pathways (RCPs) for low and high greenhouse gas concentration scenarios (RCP2.6 and RCP8.5, respectively). Because we were unable to obtain layers that projected future changes in the landscape, we assumed that landscape factors remained constant for the next 70 years. We used ARCGIS v10.2 (ESRI, Inc.) and SDMTOOLBOX to produce a base-map (i.e., a 5-km buffer around streams of the Mississippi River Basin), and bioclimatic and environmental layers with the same map projection and resolution (1 km2). Using SDMTOOLBOX, we identified eight uncorrelated layers (<0.6 Pearson correlation coefficient; see Table S1) to predict suitable habitat for C. monodonta.

Using MAXENT, we first built ENMs based on current bioclimatic and landscape layers, and then projected current ENMs to four future climate scenarios to predict future suitable habitats where C. monodonta might persist. Because future landscape variables were unavailable, we repeated the processes with only bioclimatic variables to check the effect of landscape variables. We checked the current ENM accuracy by using the 10-fold cross-validation method to calculate the area under the curve (AUC) of the receiver operating characteristics curve (ROC). The AUC ranges from 0.5 (random accuracy) to 1.0 (perfect discrimination). We estimated proportional changes in suitability scores for each of the occurrence points over time. Furthermore, we estimated distributional changes under the two greenhouse gas concentration scenarios. We first created binary predictions of suitable and unsuitable by classifying as “suitable” any cell with suitability values greater than or equal to the lowest value associated with any one of the occurrence points (Pearson et al. 2007). We used SDMTOOLBOX to estimate the distributional changes between two consecutive ENMs (i.e., ENMs of current and 2050, and ENMs of 2050 and 2070) and then calculated areas of range expansion, contraction, and no change in the species’ distribution under future climate scenarios.

23 Projection of the effects of climate change on genetic variation and connectivity

We used the spatially explicit, individual-based landscape genetic program CDPOP v1.2.19 (Landguth & Cushman 2010) to assess how future changes in stream connectivity might influence population genetic structure and genetic diversity of the extant populations of C. monodonta. CDPOP implements stochastic processes to simulate individual genotype frequencies through time as a function of individual-based reproduction, mortality, and dispersal on a continuous resistance surface. We considered individual dispersal in the stream network to be a function of stream connectivity through suitable habitats. Dispersal resistance matrices typically have required expert knowledge on the weighting of habitat heterogeneity to represent relevant friction values for calculation of least-cost paths (Brown 2014). Recently, inverted ENMs have been used to create friction landscapes, which is a more objective alternative to the use of expert knowledge. Thus, we converted the ENM layers to friction layers in SDMTOOLBOX by simply inverting the suitability scores, and calculated a stream resistance matrix for each pair of populations through stream branches based on the friction layers using Python scripts for ARCGIS described in Etherington (2011). Using this method, a path through highly suitable habitat is converted to a path of low dispersal cost. Because individuals in the upper Mississippi River are currently panmictic and there was no signature of isolation-by-distance (Inoue et al. 2014), we considered that individuals were able to disperse freely within this region. Thus, we used the maximum stream resistance value within the panmictic region as a dispersal threshold in CDPOP simulations. We simulated three models associated with future climatic scenarios: (1) stream resistance is constant through time (no-climate-change model), (2) stream resistance changes under the low greenhouse gas concentration scenario (RCP2.6 model), and (3) stream resistance changes under the high greenhouse gas concentration scenario (RCP8.5 model). We applied the current stream resistance matrix for the no-climate-change model throughout the simulation. For the RCP2.6 and RCP8.5 models, we first applied the current stream resistance matrix for the first 49 years, and then applied future stream resistance matrices derived from the 2050 ENMs and the 2070 ENMs at the 50th and 70th years, respectively.

CDPOP requires a series of demographic parameters (e.g., reproductive modes and age- specific mortality rates). We assumed average sexual maturity at 10 years; equal sex ratio at birth; maximum age of reproduction as 56 years; and average fecundity of five million glochidia

24 per female (Baird 2000; USFWS 2012, 2014). We calculated age-specific mortality rates estimated from a static life table (Baird 2000). Information for early life stages (e.g., glochidia and juvenile stages) is often limited. In our case, we do not have information for a transformation rate from glochidia to settled juvenile for C. monodonta. Thus, we used the known rate for Quadrula fragosa (Kjos et al. 1998) to estimate a mortality rate during metamorphosis. Furthermore because it is not practical to use such a large number of individual glochidia in the simulation, we used the number of first-year juveniles per female as an estimate of fecundity, calculated from number of glochidia per female multiplied by a mortality rate during metamorphosis estimated from Q. fragosa (Kjos et al. 1998) and the mortality rate between year 0 and year 1 estimated from a static life table (Baird 2000). Thus, we set fecundity (i.e., average number of first-year juveniles per female) at 0.453, with individual fecundities drawn from a Poisson distribution. We assumed that mating occurs completely within the local populations because adult mussels are incapable of dispersing long distances (USFWS 2012). We used empirical microsatellite data for the simulations, after excluding individuals with missing genotypes and the OR2 population due to small initial population size. Because varied population sizes can influence the outcome (Hall & Beissinger 2014; Landguth & Schwartz 2014), we added hypothetical individuals and made up each population to 50 individuals. Prior to the simulations, we ran short burn-in simulations (30 years) to assign genotypes for the hypothetical individuals based on empirical allele frequencies. After the burn- in period, we had a total of 800 individuals from 16 populations that contained a total of 374 alleles over 16 loci. We aimed to simulate long-term evolutionary potential rather than short- term conservation management. Thus, we ran each simulation for 1000 years, or approximately 40 generations, with 10 Monte Carlo replications. We estimated changes in population genetic structure by comparing simulated DEST at year 1000 with the empirical estimate of DEST. For each model at every year, we estimated two genetic diversity indices within the populations over

10 replicates: total number of alleles over 16 loci and mean HE.

RESULTS Population genetic diversity Estimates of within-population variation at the COI and microsatellite loci were obtained for 17 populations from seven rivers. We recovered a total of 62 haplotypes from 412 COI

25 sequences (Table 1). Overall mean number of basepair differences was 4.05-bp and nucleotide diversity was 0.0065. We observed similar genetic diversity among all populations, except those from the Ouachita River, which had consistently low numbers of haplotypes, numbers of base pair differences, and nucleotide diversity. Consistent with previous results (Inoue et al. 2014), we observed two haplotype lineages in our samples (Figure 2). A majority of individuals (66.7 – 99.6% of individuals within a population; Table 1) were from Lineage 1 in all populations, except those from the Ouachita River; all individuals from the Ouachita populations possessed Lineage 2 haplotypes (Figure 2), which were not shared with populations from other rivers. Microsatellite analyses showed no evidence of linkage disequilibrium, and few deviations from HWE (15% of all population-by-locus pairs) after sequential Bonferroni correction. However, these did not show any pattern across populations or loci and thus, we included them in further analyses. The number of alleles per locus ranged from five to 66 (for a total of 382 different alleles over 16 loci). The mean allelic richness across the 16 loci ranged from 6.366 in OR1 to 9.342 in CR1 (Table 1). The mean observed heterozygosity ranged from 0.632 in GR12 to 0.827 in GR1, and the mean expected heterozygosity ranged from 0.679 in OR2 to 0.854 in

MR1. After rarefaction, there were no alleles unique to the populations (NRP < 0; Table 1), indicating that all alleles are shared by two or more populations. Population genetic structure

Results of AMOVA showed that the greatest genetic variation was among localities (FST

= 0.038, P = 0.001; Table 2), followed by genetic variation among provinces (FPT = 0.033, P =

0.001). Genetic variation was much smaller among rivers within province (FRP = 0.001, P =

0.012) and among localities within river (FSR = 0.003, P = 0.008), indicating that rivers and provinces contributed a small amount of genetic variation to population structure of C. monodonta. A majority of pairwise comparisons of FST and DEST were significantly greater than

0; however, ranges of these indices were small (FST = 0 – 0.097; DEST = 0.010 – 0.518; Table S2, supplementary information). Furthermore, all high values of FST and DEST were found among the

Ouachita populations (FST = 0.057 – 0.097, DEST = 0.346 – 0.518) while all other populations were lower. Cumberlandia monodonta showed evidence of significant range-wide population genetic structure. The STRUCTURE analysis indicated differentiation between a cluster of the Ouachita populations and a cluster of all the other populations at k = 2 [∆k = 303.6, lnP(k) = -31453.0;

26 Figures 3 and S1, supplementary information]. We found no evidence of admixture between the two clusters and we did not recover further differentiation within clusters. Projection of the effects of climate change on suitable habitat All models had high AUC values (>0.85), indicating overall adequate performance. The climatic and landscape variables that most influenced prediction of suitable habitat were human population density (40.6%) and nitrogen fertilizer input (17.9%), followed by altitude (15.3%), seasonal precipitation (10.9%), and annual mean temperature (9.5%). Human population density, nitrogen fertilizer input, altitude, and temperature were negatively correlated with habitat suitability scores, while precipitation was positively correlated. The prediction of current habitat for C. monodonta showed high suitability scores across rivers with current and historical records of C. monodonta (Figures 4, 5, and S2, supplementary information; USFWS 2012, 2014). The average suitability score of the occurrence points was 0.614 (SE = 0.027); scores ranged from 0.253 to 0.844. We used the lowest suitability score (0.253) as the threshold for creating binary predictions of suitable and unsuitable habitats (Figures 4 and 5). Additionally, the ENMs detected potential suitable habitat in the upper White River (Arkansas) the Red River (the border of Arkansas and Oklahoma), and tributaries of the lower Mississippi River in Mississippi; no occurrences have been recorded from these. Populations in tributaries of the upper Mississippi River were connected by areas of high suitability scores, likely allowing high gene flow among populations and panmixia. Suitable habitat in the Ouachita River was disconnected from the upper Mississippi by unsuitable habitat in the lower Mississippi River, which likely resulted in the isolation of Ouachita River populations. The ENMs for both the RCP2.6 and RCP8.5 scenarios projected range contraction northward and upstream, with drastic declines in suitability scores of the occurrence points throughout the Mississippi River Basin (Figures 4, 5, and S2). In the RCP2.6 scenarios, suitable habitat declined by 62.7% from the present to 2050, with a slight increase in habitat from 2050- 2070 (Table 3; Figure 4). However, this range expansion was really just recovery from the range contraction in 2050; in 2070, amount of suitable habitat was still 62.1% of the present area of suitable habitat. The average suitability score of the occurrence points declined 53.6% from the present to 2050 (average suitability score = 0.285, SE = 0.018) and remained almost the same in 2070 (average suitability score = 0.284, SE = 0.019). In the RCP8.5 scenarios, on the other hand, suitable habitat declined 79.7% from the current conditions to 2050 and continued to decline

27 from 2050-2070 (Table 3; Figure 5). The average suitability score was 66.6% lower in 2050 (average suitability score = 0.205, SE = 0.016) than at the present, and continued to decline through 2070 (average suitability score = 0.126, SE = 0.014). The ENMs with both bioclimatic and landscape variables and bioclimatic-only variables were highly correlated (Pearson correlation coefficient = 0.80 – 0.91). Both ENMs showed similar distributional changes under future scenarios; the bioclimatic-only ENMs predicted a wider range of suitable habitats and higher suitability scores than those incorporating both bioclimatic and landscape variables (Figures S2 and S3). Thus, bioclimatic variables strongly affected the prediction of suitable habitats even though the contribution of each individual variable to the prediction was low. Landscape variables further narrowed down the distribution of suitable habitats. Projection of the effects of climate change on genetic variation and connectivity

The CDPOP simulations indicated that the magnitude of greenhouse gas concentrations and resultant decrease in stream connectivity due to loss of suitable habitat would alter population connectivity and genetic diversity of the extant populations in the future (Figures 6 and 7). In the no-climate-change model, populations in tributaries of the upper Mississippi River maintained low genetic differentiation and thus remained panmictic (Figure 6). As a result, genetic diversity remained relatively constant over the simulated period (Figure 7). The Ouachita populations, on the other hand, were predicted to slightly increase genetic differentiation while losing ~55% of their total alleles and ~25% of heterozygosity over this time period. Both future climate models predicted loss of panmixia in the upper Mississippi River. In the RCP2.6 model, the Clinch and St. Croix river populations increased their genetic differentiation versus other populations in the upper Mississippi basin (Figure 6). Combined with loss of suitable habitat among these populations, these results suggest that climate change under the RCP2.6 scenario would lead to isolation of these populations with significant loss of genetic diversity over time (~71% decrease in total alleles and 28% decrease in heterozygosity over 1000 years; Figure 7). Furthermore in the RCP8.5 model, the Green River population increased its genetic differentiation versus all the other populations (Figure 6), and accordingly, suffered a drastic reduction in genetic diversity over time (loss of 80% of total alleles and 46% of heterozygosity; Figure 7). In both RCP2.6 and RCP8.5 models, populations in the Gasconade, Meramec, and Osage rivers remained panmictic, and maintained relatively stable genetic diversity over time.

28 Regardless of the simulations, among-population variation within rivers remained low (Figure 6).

DISCUSSION Our extensive sampling of C. monodonta throughout its range suggests the existence of a large panmictic population in tributaries of the upper Mississippi River and a genetically isolated population in the Ouachita River, with little or no contemporary gene flow between Ouachita and the other populations. These results are congruent with a previous study that reconstructed phylogeographic patterns of C. monodonta (Inoue et al. 2014). Additionally, our study suggests that subtle, but significant, population genetic structure in C. monodonta is primarily due to unsuitable habitat in the lower Mississippi River that prevents dispersal between these regions. The presence of panmixia in the upper Mississippi River is made possible by current stream connectivity of highly suitable habitat. Furthermore, the ENMs under current conditions predicted potential suitable habitats in headwaters of the Ohio and Mississippi rivers. There are historic records of C. monodonta occurrence (USFWS 2012); however, the species is thought to be extirpated from these rivers. Subsequent climate change, however, would likely fragment the panmictic population. The ENMs under future-climate scenarios predict that increased greenhouse gas concentration is likely to be correlated with reduction of suitable habitat throughout the Mississippi River Basin. The ENMs of both future scenarios showed drastic range contraction northward and upstream, without range expansion, compared to the contemporary distributional range. In the RCP2.6 scenario, suitable habitat will remain in relatively large portions of the upper Mississippi River, whereas suitable habitat will be severely restricted to northern headwater tributaries of the Mississippi River in the RCP8.5 scenario. In either scenario, panmixia in the upper Mississippi River will not be sustained. Furthermore, most of the extant populations will suffer reductions of habitat suitability, especially in populations from the southern portion of the distributional range. Potential impacts of future climate change on freshwater biodiversity have been reviewed (e.g., Heino et al. 2009), and there is some evidence that freshwater species have exhibited range shifts in response to climate change (e.g., Hickling et al. 2006). However, these studies often fail to identify ecological processes (e.g., dispersal, population dynamics) that lead to these impacts, nor do they consider the genetic consequences of climate change (e.g., changes in gene flow and

29 genetic variability). While long-term ecological and evolutionary processes are often too complex to assess with manageable experiments or surveys, integration of ecological and genetic simulations allows us to create mechanistic explanations for understanding biological responses to climate change. In this study, when we assumed that the current extant populations of C. monodonta would remain at the same localities over time, increases in greenhouse gas concentrations will greatly decrease intra-population genetic diversity by reducing stream connectivity among populations. Our genetic simulations, which accounted for variation in stream resistance, showed that fragmentation due to long stretches of unsuitable habitat will lead to the loss of panmixia in the upper Mississippi region, resulting in severe loss of genetic diversity; such outcomes were very pronounced even in currently well-connected tributaries such as the St. Croix, Clinch, and Green rivers. Our genetic trajectories were based upon populations from seven rivers in the Mississippi River Basin where C. monodonta is currently known to occur. Because our ENMs showed potential suitable habitat in upper Ohio River and lower Mississippi River drainages, undiscovered populations in suitable habitat may serve as “bridges” that have the potential to maintain gene flow between extant populations. Given that the dynamics of range shifts involve two peripheral edges, the expanding “leading edge” and contracting “rear edge,” net shifts in terrestrial ecosystems associated with warming climates involve range movement toward the poles and/or to higher altitude (Hampe & Petit 2005; Moran & Alexander 2014). While leading edge populations are controlled by active colonization events and positive population growth, rear edge populations are often small and fragmented such that regional population dynamics cannot easily compensate for local extinction events (Hampe & Petit 2005). Such rear edge populations often suffer reduced genetic diversity within populations and high levels of genetic diversification among populations. In comparison with terrestrial ecosystems, ecological processes of stream-dwelling species occur along the network branches as a function of physical flow and network connectivity (Grant et al. 2007). Therefore, if species have limited dispersal capability and/or if suitable habitat for the species is fragmented, strong demographic and genetic isolation are expected to occur among localities along network branches, resulting in isolation by river distance (Fagan 2002; Primmer et al. 2006). In riverine systems flowing toward the equator, poleward movement of ranges also involves shifting from larger to smaller stream habitats. Our study indicates that range shifts of C. monodonta likely involve only the rear edge. Given their small size and prolonged isolation,

30 the current Ouachita River populations display characteristics of rear edge populations. Furthermore, our ENMs predict that populations in both southern and northern portions of the range will become rear edge populations as a consequence of climate change. As such, C. monodonta and other species with similar distributions will suffer constriction of ranges at the rear (southern or downstream) edge due to reduction of habitat quality and the loss of connectivity among suitable habitats. The net effect is that total range will decline with climate warming. Maintenance of rear edge populations is particularly important for the long-term conservation of genetic diversity and evolutionary adaptability (Lesica & Allendorf 1995). Furthermore, the outcomes of our study allow inferring of the consequences of climate warming for organisms inhabiting other rivers flowing equatorward, including the Colorado River and Rio Grande in the United States, and other great rivers such as the Mekong and Indus. Given the rate and magnitude with which climate patterns are expected to change in the near future, the link between climate and stream connectivity found in our study has important implications for conservation of C. monodonta and other aquatic species in the region. Reducing greenhouse gas emissions may serve to lessen the impacts of climate-change-driven losses in biodiversity and accordingly, prevent local extinctions. Applications of forward-time genetic simulations with ENMs allow prediction of future changes in stream connectivity and the consequences of such changes. We note, however, that there has been much debate about the accuracy and projections of ENMs in predicting range shifts in response to climate change (Jiménez-Valverde et al. 2008; Elith & Leathwick 2009). Much of the discussion points to the exclusion of biological interactions between species and of adaptive capacities of individual species from predicted models (Araújo & Guisan 2006; de Araújo et al. 2014). For example, warmer trends in climatic regimes can alter physiological and phenological traits to match new climatic conditions (Moran & Alexander 2014). Although many freshwater organisms, including freshwater mussels, may have broad thermal tolerances, warmer temperatures may alter seasonal life cycles and disrupt the timing of reproductive activities. Gametogenesis and glochidia production are triggered by seasonal temperature fluctuations in many mussel species (reviewed in Haag 2012), including C. monodonta (Gordon & Smith 1990). Because the use of host species varies among mussel species and is often limited to one or a few fishes, changes in the timing of reproductive events may affect biological interactions with hosts. Although hosts for C. monodonta have not been determined, given the panmixia evident among populations in the

31 upper Mississippi River, hosts are likely to be highly mobile and migratory. Migration and long- distance dispersal are often triggered by seasonal cues, such as water temperature (Ficke et al. 2007). These shifts in physiological and phenological traits in response to climate change may create a mismatch of host-parasite interactions (e.g., monophagous butterfly and its host plant; Schweiger et al. 2008). Although we solely examined future changes in genetic structure of the extant populations, each of these populations will undergo ecological and evolutionary changes in order to adapt to future climatic conditions. Most species tolerate short-term variability in climate through phenotypic plasticity; however, long-term and extreme variability require species to undergo evolutionary changes in order to survive such variability (Jump & Penuelas 2005). Studies of evolutionary responses to-date, which are limited to invasive species with generation times of 1 year or shorter, show that genetic responses generally develop in a few decades (~25 generations or greater; Moran & Alexander 2014). However, ongoing climate change is rather rapid relative to species with long generation times; for example, the mean generation time for C. monodonta is 26 years (Inoue et al. 2014) and thus, 25 generations is approximately 650 years. Our results suggest that short-term simulations (i.e., ~100 years projection, which are often used for conservation management) would likely fail to capture trajectories of genetic persistence for species with long generation times. High levels of overall genetic diversity and large effective population sizes increase chances of adapting to climate change within a few generations (Petit & Hampe 2006). Furthermore, high levels of heterozygosity may compensate for decreased individual fitness by counterbalancing deleterious mutations in populations (Chapman et al. 2009). The high mtDNA and microsatellite diversity of most C. monodonta populations would maintain the potential to adapt to rapid climate change for the panmictic populations, but not for the Ouachita populations. Global trends in climate change show evidence of poleward range shifts in plant and animal species worldwide. However, such range shifts are a function of species’ dispersal capability and landscape connectivity. While range shifts and changes in landscape connectivity in terrestrial systems are well documented, less is known about range shifts and habitat connectivity in dendritic systems such as streams. We suggest that riverine systems flowing equatorward may not exhibit poleward range shifts in biota due to limitations in dispersal ability of aquatic species, loss of connectivity among locations containing suitable habitat, and changes

32 in physical environments due to shifts from larger to smaller streams. Instead, aquatic species inhabiting such systems will be threatened with extirpation unless they are able to rapidly adapt to climate change.

ACKNOWLEDGEMENTS We would like to thank M. Davis, B. Sietman, S. McMurray, JS Faiman, D. Hubbs, J. Jones, S. Ahlstedt, W. Posey II, J. Harris, C. Lewis, L. Koch, and M. McGregor for help with sampling collections. Technical assistance was provided by numerous undergraduate technicians in the Berg Lab, and the Center for Bioinformatics and Functional Genomics at Miami University. We thank T. Smith, S. McMurray, JS Faiman, T. Crist, B. Cochrane, B. Keane, and R. Moore for comments on earlier drafts of this manuscript. Funding was provided by the Missouri Department of Conservation and the U.S. Fish and Wildlife Service.

REFERENCES Alp M, Keller I, Westram AM, Robinson CT (2012) How river structure and biological traits influence gene flow: a population genetic study of two stream invertebrates with differing dispersal abilities. Freshwater Biology, 57, 969-981. Araújo MB, Guisan A (2006) Five (or so) challenges for species distribution modelling. Journal of Biogeography, 33, 1677-1688. Baird MS (2000) Life history of the spectaclecase, Cumberlandia monodonta Say, 1829 (Bivalvia, Unionoidea, Margaritigeridae) M.S. thesis, Southwest Missouri State University, Springfield, Missouri. Booth DJ, Bond N, Macreadie P (2011) Detecting range shifts among Australian fishes in response to climate change. Marine and Freshwater Research, 62, 1027-1042. Brown JL (2014) SDMtoolbox: a python-based GIS toolkit for landscape genetic, biogeographic and species distribution model analyses. Methods in Ecology and Evolution, 5, 694-700. Chapman JR, Nakagawa S, Coltman DW, Slate J, Sheldon BC (2009) A quantitative review of heterozygosity-fitness correlations in animal populations. Molecular Ecology, 18, 2746- 2765. Chen I-C, Hill JK, Ohlemüller R, Roy DB, Thomas CD (2011) Rapid range shifts of species associated with high levels of climate warming. Science, 333, 1024-1026.

33 de Araújo CB, Marcondes-Machado LO, Costa GC, Silman M (2014) The importance of biotic interactions in species distribution models: a test of the Eltonian noise hypothesis using parrots. Journal of Biogeography, 41, 513-523. Dudgeon D, Arthington AH, Gessner MO, Kawabata Z-I, Knowler DJ, Leveque C, Naiman RJ, Prieur-Richard A-H, Soto D, Stiassny MLJ, Sullivan CA (2006) Freshwater biodiversity: importance, threats, status and conservation challenges. Biological Reviews, 81, 163-182. Earl DA, vonHoldt BM (2012) STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources, 4, 359-361. Elith J, Leathwick JR (2009) Species distribution models: ecological explanation and prediction across space and time. Annual Review of Ecology, Evolution, and Systematics, 40, 677- 697. Etherington TR (2011) Python based GIS tools for landscape genetics: visualising genetic relatedness and measuring landscape connectivity. Methods in Ecology and Evolution, 2, 52-55. Fagan WF (2002) Connectivity, fragmentation, and extinction risk in dendritic metapopulations. Ecology, 83, 3243-3249. Felsenstein J (2005) PHYLIP (phylogeny inference package) version 3.6. Department of Genome Sciences, University of Washington, Seatle, WA. Ficke AD, Myrick CA, Hansen LJ (2007) Potential impacts of global climate change on freshwater fisheries. Reviews in Fish Biology and Fisheries, 17, 581-613. Geffen E, Anderson MJ, Wayne RK (2004) Climate and habitat barriers to dispersal in the highly mobile grey wolf. Molecular Ecology, 13, 2481-2490. Gerlach G, Jueterbock A, Kraemer P, Deppermann J, Harmand P (2010) Calculations of

population differentiation based on GST and D: forget GST but not all of statistics! Molecular Ecology, 19, 3845-3852. Gordon ME, Smith DG (1990) Autumnal reproduction in Cumberlandia monodonta (Unionoidea: Margaritiferidae). Transactions of the American Microscopical Society, 109, 407-411. Goudet J (2005) HIERFSTAT, a package for R to compute and test hierarchical F-statistics. Molecular Ecology Notes, 5, 184-186.

34 Grant EHC, Lowe WH, Fagan WF (2007) Living in the branches: population dynamics and ecological processes in dendritic networks. Ecology Letters, 10, 165-175. Haag WR (2009) A hierarchical classification of freshwater mussel diversity in North America. Journal of Biogeography, 37, 12-26. Haag WR (2012) North American Freshwater Mussels: Natural History, Ecology, and Conservation Cambridge University Press, Cambridge, UK. Habel JC, Rödder D, Schmitt T, Nève G (2011) Global warming will affect the genetic diversity and uniqueness of Lycaena helle populations. Global Change Biology, 17, 194-205. Hall LA, Beissinger SR (2014) A practical toolbox for design and analysis of landscape genetics studies. Landscape Ecology, 29, 1487-1504. Hampe A, Petit RJ (2005) Conserving biodiversity under climate change: the rear edge matters. Ecology Letters, 8, 461-467. Heino J, Virkkala R, Toivonen H (2009) Climate change and freshwater biodiversity: detected patterns, future trends and adaptations in northern regions. Biological Reviews, 84, 39-54. Hickling R, Roy DB, Hill JK, Fox R, Thomas CD (2006) The distributions of a wide range of taxonomic groups are expanding polewards. Global Change Biology, 12, 450-455. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 1965- 1978. Hughes JM, Schmidt DJ, Finn DS (2009) Genes in streams: using DNA to uderstand the movement of freshwater fauna and their riverine habitat. BioScience, 59, 573-583. Inoue K, Monroe EM, Elderkin CL, Berg DJ (2014) Phylogeographic and population genetic analyses reveal Pleistocene isolation followed by high gene flow in a wide-ranging, but endangered, freshwater mussel. Heredity, 112, 282-290. Inoue K, Moyer GR, Williams A, Monroe EM, Berg DJ (2011) Isolation and characterization of 17 polymorphic microsatellite loci in the spectaclecase, Cumberlandia monodonta (Bivalvia: Margaritiferidae). Conservation Genetics Resources, 3, 57-60. IPCC (2013) Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

35 Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics, 23, 1801-1806. Jiménez-Valverde A, Lobo JM, Hortal J (2008) Not as good as they seem: the importance of concepts in species distribution modelling. Diversity and Distributions, 14, 885-890. Jost L (2008) Gst and its relatives do not measure differentiation. Molecular Ecology, 17, 4015- 4026. Jump AS, Penuelas J (2005) Running to stand still: adaptation and the response of plants to rapid climate change. Ecology Letters, 8, 1010-1020. Kjos C, Byers O, Miller P, Borovansky J, Seal US (1998) Population and habitat viability assessment workshop for the winged mapleleaf mussel (Quadrula fragosa): Final Report, p. 92, CBSG, Apple Valley, MN. Kramer-Schadt S, Niedballa J, Pilgrim JD, Schröder B, Lindenborn J, Reinfelder V, Stillfried M, Heckmann I, Scharf AK, Augeri DM, Cheyne SM, Hearn AJ, Ross J, Macdonald DW, Mathai J, Eaton J, Marshall AJ, Semiadi G, Rustam R, Bernard H, Alfred R, Samejima H, Duckworth JW, Breitenmoser-Wuersten C, Belant JL, Hofer H, Wilting A, Robertson M (2013) The importance of correcting for sampling bias in MaxEnt species distribution models. Diversity and Distributions, 19, 1366-1379. Landguth EL, Cushman SA (2010) CDPOP: A spatially explicit cost distance population genetics program. Molecular Ecology Resources, 10, 156-161. Landguth EL, Schwartz MK (2014) Evaluating sample allocation and effort in detecting population differentiation for discrete and continuously distributed individuals. Conservation Genetics, 15, 981-992. Lesica P, Allendorf FW (1995) When are peripheral population valuable for conservation? Conservation Biology, 9, 753-760. Librado P, Rozas J (2009) DnaSP v5: a software for comprehensive analysis of DNA polymorphism data. Bioinformatics, 25, 1451-1452. Moran EV, Alexander JM (2014) Evolutionary responses to global change: lessons from invasive species. Ecology Letters, 17, 637–649. Parmesan C, Yohe G (2003) A global coherent fingerprint of climate change impacts across natural systems. Nature, 421, 37-42.

36 Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6, 288-295. Pearson RG, Raxworthy CJ, Nakamura M, Peterson AT (2007) Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography, 34, 102-117. Peterson EE, Ver Hoef JM, Isaak DJ, Falke JA, Fortin MJ, Jordan CE, McNyset K, Monestiez P, Ruesch AS, Sengupta A, Som N, Steel EA, Theobald DM, Torgersen CE, Wenger SJ (2013) Modelling dendritic ecological networks in space: an integrated network perspective. Ecology Letters, 16, 707-719. Petit RJ, Hampe A (2006) Some evolutionary consequences of being a tree. Annual Review of Ecology, Evolution, and Systematics, 37, 187-214. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190, 231-259. Phillips SJ, Dudik M, Elith J, Graham CH, Lehmann A, Leathwick JR, Ferrier S (2009) Sample selection bias and presence-only distribution models: implications for background and pseudo-absence data. Ecological Applications, 19, 181-197. Primmer CR, Veselov AJ, Zubchenko A, Poututkin A, Bakhmet I, Koskinen MT (2006) Isolation by distance within a river system: genetic population structuring of Atlantic salmon, Salmo salar, in tributaries of the Varzuga River in northwest Russia. Molecular Ecology, 15, 653-666. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genetype data. Genetics, 155, 945-959. R Development Core Team (2011) R: A language and environment for statistical computing. R Fundation for Statistical Computing, Vienna, Austria. Rosenberg NA (2004) DISTRUCT: a program for the graphical display of population structure. Molecular Ecology Notes, 4, 137-138. Rousset F (2008) GENEPOP'007: a complete re-implementation of the GENEPOP software for Windows and Linux. Molecular Ecology Resources, 8, 103-106. Row JR, Wilson PJ, Gomez C, Koen EL, Bowman J, Thornton D, Murray DL (2014) The subtle role of climate change on population genetic structure in Canada lynx. Global Change Biology, 20, 2076-2086.

37 Schweiger O, Settele J, Kudrna O, Klotz S, Kühn I (2008) Climate change can cause spatial mismatch of trophically interacting species. Ecology, 89, 3472-3479. Sexton JP, Hangartner SB, Hoffmann AA (2014) Genetic isolation by environment or distance: which pattern of gene flow is most common? Evolution, 68, 1-15. Szpiech ZA, Jakobsson M, Rosenberg NA (2008) ADZE: a rarefaction approach for counting alleles private to combinations of populations. Bioinformatics, 24, 2498-2504. Urban D, Keitt T (2001) Landscape connectivity: a graph-theoretic perspective. Ecology, 82, 1205-1218. USFWS (US Fish and Wildlife Service). 2012. Endangered and threatened wildlife and plants; determination of endangered status for the sheepnose and spectaclecase mussels throughout their range, final rule. Federal Register, 77, 14914-14949. USFWS (U.S. Fish and Wildlife Service). 2014. Recovery outline for the spectaclecase mussel (Cumberlandia monodonta) - January 2014. US Fish and Wildlife Service, Fort Snelling, MN. Vörösmarty CJ, McIntyre PB, Gessner MO, Dudgeon D, Prusevich A, Green P, Glidden S, Bunn SE, Sullivan CA, Liermann CR, Davies PM (2010) Global threats to human water security and river biodiversity. Nature, 467, 555-561. Walther G-R, Post E, Convey P, Menzel A, Parmesan C, Beebee TJC, Fromentin J-M, Hoegh- Guldberg O, Bairlein F (2002) Ecological responses to recent climate change. Nature, 416, 389-395.

38 TABLES AND FIGURES Table 1. Descriptive statistics for COI sequences and 16 microsatellite loci for each collection site of Cumberlandia monodonta. Population IDs correspond to Figure 1. COI Microsatellites River ID n H HR K π %Lineage1 NA AR NRP HO HE Clinch River CR01 26 11 9.8 3.90 0.0063 76.0 13.9 9.342 0.326 0.771 0.850 Clinch River CR02 20 9 9.0 2.80 0.0045 90.0 12.1 8.880 0.235 0.694 0.839 Gasconade River GR01 24 6 5.7 3.57 0.0058 73.9 12.8 8.891 0.172 0.827 0.841 Gasconade River GR06 27 8 7.1 4.24 0.0069 66.7 12.5 8.528 0.222 0.662 0.831 Gasconade River GR09 23 7 6.5 3.25 0.0052 82.6 12.9 9.046 0.236 0.707 0.844 Gasconade River GR12 23 8 7.3 3.19 0.0052 82.6 11.3 8.281 0.095 0.632 0.824 Green River GRN 37 14 9.6 3.01 0.0049 86.5 14.7 9.278 0.200 0.676 0.851 Meramec River MR01 24 7 6.4 3.35 0.0054 75.0 13.6 9.308 0.324 0.815 0.854 Meramec River MR03 25 10 8.9 4.15 0.0067 72.0 12.9 8.839 0.078 0.724 0.837 Meramec River MR04 25 7 6.6 3.71 0.0060 76.0 12.4 8.560 0.161 0.717 0.827 Ouachita River OR1 20 3 3.0 0.48 0.0008 0.0 8.0 6.366 0.139 0.703 0.736 Ouachita River OR2 6 2 n.a. 0.40 0.0007 0.0 4.8 n.a. n.a. 0.648 0.679 Ouachita River OR3 30 3 2.9 0.50 0.0008 0.0 9.1 6.864 0.181 0.661 0.776 Osage River OSG1 24 5 4.7 2.89 0.0047 83.3 13.1 9.055 0.115 0.674 0.843 Osage River OSG2 32 9 6.7 3.39 0.0055 81.2 14.1 9.089 0.206 0.701 0.847 St. Croix River SC1 25 7 6.3 1.48 0.0024 96.0 13.2 9.046 0.127 0.804 0.843 St. Croix River SC2 25 4 3.7 1.59 0.0026 95.8 13.1 9.071 0.247 0.703 0.836 Total 416 62 9.8 4.05 0.0065 70.4 12.0 8.653 0.191 0.713 0.821 AR, rarefied allelic richness; COI, cytochrome oxidase subunit I; H, number of haplotypes; HE, mean expected heterozygosity; HO, mean observed heterozygosity; HR, rarefied number of haplotypes; K, mean number of base pair differences; n, number of samples;

NA, mean number of observed alleles; NPR, rarefied number of private alleles; π, nucleotide diversity; %Lineage 1, percentage of Lineage 1 haplotypes

39 Table 2. Results of the hierarchical analysis of molecular variance (AMOVA) in Cumberlandia monodonta. Seven rivers represent four biogeographic provinces: Tennessee-Cumberland (Clinch River), Upper Mississippi (Gasconade, Meramec, Osage, and St. Croix rivers), Ohioan (Green River), and Mississippi Embayment (Ouachita River). Variance component F-statistics P Among provinces FPT = 0.033 0.001 Among rivers within province FRP = 0.001 0.012 Among localities within river FSR = 0.003 0.008 Among all localities FST = 0.038 0.001

40 Table 3. Areas (km2) of binary predictions of suitable and unsuitable under current conditions and two greenhouse gas concentration scenarios in 2050 and 2070. RCP2.6 RCP8.5 Current 2050 2070 2050 2070 Range expansion -- 2,588.5 12,170.5 719.4 454.1 No change -- 64,124.5 55,653.1 35,688.7 11,328.2 Total occupancy 178,999.3 66,712.9 67,823.6 36,408.1 11,782.3 No occupancy 544,270.4 541,681.9 644,386.2 543,550.9 686,407.5 Range contraction -- 114,874.9 11,059.9 143,310.7 25,079.9 Total no occupancy 544,270.4 656,556.7 655,446.1 686,861.6 711,487.4

Figure 1. Map of the central United States indicating sites where Cumberlandia monodonta were sampled (black dots). See Table 1 for sampled rivers and sample size in each site. Colored watersheds represent historic (gray) and current (blue) distributions of C. monodonta obtained from NatureServe (http://explorer.natureserve.org; accessed June 25, 2015).

42 Lineage 1 CR 1 1 GR 3 GRN 1 1 MR 1 2 OR 1 1 OSG 1 1 SC 1 1 1 1 1 1 133 2 1 Lineage 2 1 1 1 2 1 1 1 1 1 1 1 1 1 1 5 1 46 59 28 1 1 1 1 58 1 3 1 3 1 1 1 1 1 1 1 1 19 1 1 1 1 1 Figure 2. A parsimony network of COI sequences for Cumberlandia monodonta. Each circle represent unique haplotype and lines between haplotypes represent one base pair; black dots are inferred missing haplotypes. Haplotype frequency is relative to the size and number in the circle. Colors represent drainages. Lineage designations were based on Inoue et al. (2014).

Figure 4. Prediction of suitable (blue) and unsuitable (light gray) habitats for Cumberlandia monodonta identified using ecological niche models (ENMs) under “current” bioclimatic conditions (interpolations of observed data from 1950 to 2000) and projections of near future climatic in 2050 (average for 2041-2060) and 2070 (average for 2051-2080) under low greenhouse gas concentration scenarios (IPCC 5th Assessment). Low greenhouse gas concentration was based on representative concentration pathways 2.6 (RCP2.6). Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs.

Figure 5. Prediction of suitable (blue) and unsuitable (light gray) habitats for Cumberlandia monodonta identified using ecological niche models (ENMs) under “current” bioclimatic condition (interpolations of observed data from 1950 to 2000) and projections of near future climatic in 2050 (average for 2041-2060) and 2070 (average for 2051-2080) under high greenhouse gas concentration scenarios (IPCC 5th Assessment). High greenhouse gas concentration was based on representative concentration pathway 8.5 (RCP8.5). Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs.

th Figure 6. Results of CDPOP to compare between observed (empirical) and projected DEST at 1000 year. Simulations were (A) no climate change model, (B) RCP2.6 model, and (C) RCP8.5 model. Dotted lines represent 1:1 relationship. Points above the line indicate that DEST increases in the future (population differentiation increases) and points below the line indicate that DEST decreases in the future. Red circle represents pairwise DEST between OR populations and all the other populations; blue circle represents pairwise

DEST between CR/SC populations and GR, GRN, MR, and OSG populations; and green clusters represent pairwise DEST between GRN population and CR, GR, MR, OSG, and SC populations.

Figure 7. Results of CDPOP simulations projecting trajectories of genetic diversity (i.e., total number of alleles and expected heterozygosity) over 1000 years under different climate change scenarios. Simulations were based on the no-climate-change model (A and D), the RCP2.6 model (i.e., low greenhouse gas concentration scenario; B and E), and RCP8.5 model (i.e., high greenhouse gas concentration scenario; C and F). Colored lines represent extant populations. Areas of light gray showed 95% confidence intervals of each population.

48 SUPPLEMENTARY INFORMATION Table S1. A list of bioclimatic, geographic, and landscape layers used in ecological niche models. We chose eight uncorrelated layers (<0.6 Pearson correlation coefficient; “X” in the Used column). Name Description Used Citations BIO01 Annual mean temperature X WorldClim; Hijmans et al. (2005) BIO02 Mean diurnal range WorldClim; Hijmans et al. (2005) BIO03 Isothermality X WorldClim; Hijmans et al. (2005) BIO04 Temperature seasonality WorldClim; Hijmans et al. (2005) BIO05 Maximum temperature of warmest month WorldClim; Hijmans et al. (2005) BIO06 Minimum temperature of coldest month WorldClim; Hijmans et al. (2005) BIO07 Temperature annual range WorldClim; Hijmans et al. (2005) BIO08 Mean temperature of wettest quarter X WorldClim; Hijmans et al. (2005) BIO09 Mean temperature of driest quarter WorldClim; Hijmans et al. (2005) BIO10 Mean temperature of warmest quarter WorldClim; Hijmans et al. (2005) BIO11 Mean temperature of coldest quarter WorldClim; Hijmans et al. (2005) BIO12 Annual precipitation WorldClim; Hijmans et al. (2005) BIO13 Precipitation of wettest month WorldClim; Hijmans et al. (2005) BIO14 Precipitation of driest month WorldClim; Hijmans et al. (2005) BIO15 Precipitation seasonality X WorldClim; Hijmans et al. (2005) BIO16 Precipitation of wettest quarter WorldClim; Hijmans et al. (2005) BIO17 Precipitation of driest quarter WorldClim; Hijmans et al. (2005) BIO18 Precipitation of warmest quarter WorldClim; Hijmans et al. (2005) BIO19 Precipitation of coldest quarter WorldClim; Hijmans et al. (2005) Altitude Altitude X WorldClim; Hijmans et al. (2005) Anthrome Anthropogenic biome (2001-2006) Ellis and Ramankutty (2008); obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) available at http://sedac.ciesin.columbia.edu Cropland Croplands (2000) Ramankutty et al. (2008); obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) available at http://sedac.ciesin.columbia.edu (table continues)

Table S1. (continued) HII Global human influence index (1995-2004) X Wildlife Conservation Society et al. (2005); obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) available at http://sedac.ciesin.columbia.edu HFP Global human footprint (1995-2004) Wildlife Conservation Society et al. (2005); obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) available at http://sedac.ciesin.columbia.edu Popdens Human population density (1990, 1995, 2000) X Balk et al. (2006); obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) available at http://sedac.ciesin.columbia.edu N firtilizer Nitrogen fertilizer application (1994-2001) X Potter et al. (2010); obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) available at http://sedac.ciesin.columbia.edu P firtilizer Phosphorus fertilizer application (1994-2001) Potter et al. (2010); obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) available at http://sedac.ciesin.columbia.edu REFERENCES Balk DL, Deichmann U, Yetman G, Pozzi F, Hay SI, Nelson A (2006) Determining global population distribution: methods, applications and data. Advances in Parasitology, 62, 119-156. Ellis EC, Ramankutty N (2008) Putting people in the map: anthropogenic biomes of the world. Frontiers in Ecology and the Environment, 6, 439-447. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 1965-1978. Potter P, Ramankutty N, Bennett EM, Donner SD (2010) Characterizing the Spatial Patterns of Global Fertilizer Application and Manure Production. Earth Interactions, 14, 1-22. Ramankutty N, Evan AT, Monfreda C, Foley JA (2008) Farming the planet: 1. Geographic distribution of global agricultural lands in the year 2000. Global Biogeochemical Cycles, 22, GB1003.

50 Table S2. Pairwise DEST (above diagonal) and FST (below diagonal) values for 16 microsatellite loci from 17 populations of Cumberlandia monodonta. CR01 CR02 GR01 GR06 GR09 GR12 GRN MR01 MR03 MR04 OR1 OR2 OR3 OSG1 OSG2 SC1 SC2 CR01 -- 0.063 0.089 0.106 0.056 0.089 0.049 0.058 0.083 0.116 0.459 0.430 0.382 0.076 0.068 0.025 0.057 CR02 0.007 -- 0.120 0.159 0.116 0.133 0.059 0.039 0.060 0.080 0.473 0.479 0.409 0.084 0.053 0.064 0.101 GR01 0.010 0.012 -- 0.096 0.032 0.095 0.127 0.116 0.098 0.106 0.483 0.481 0.413 0.059 0.024 0.024 0.087 GR06 0.012 0.015 0.012 -- 0.031 0.095 0.181 0.113 0.119 0.125 0.504 0.452 0.401 0.107 0.034 0.076 0.109 GR09 0.006 0.012 0.006 0.006 -- 0.082 0.133 0.051 0.083 0.069 0.471 0.435 0.384 0.036 0.010 0.047 0.043 GR12 0.010 0.016 0.012 0.014 0.007 -- 0.160 0.107 0.111 0.100 0.475 0.469 0.346 0.087 0.056 0.081 0.061 GRN 0.011 0.011 0.024 0.029 0.020 0.023 -- 0.058 0.068 0.116 0.487 0.452 0.418 0.113 0.097 0.067 0.087 MR01 0.006 0.002 0.009 0.010 0.003 0.011 0.012 -- 0.037 0.023 0.502 0.510 0.412 0.046 0.044 0.023 0.045 MR03 0.005 0.008 0.013 0.014 0.007 0.013 0.014 0.002 -- 0.017 0.506 0.510 0.434 0.050 0.025 0.062 0.090 MR04 0.011 0.011 0.018 0.015 0.009 0.013 0.020 0.004 0.003 -- 0.471 0.518 0.406 0.073 0.024 0.076 0.063 OR1 0.081 0.086 0.097 0.092 0.093 0.095 0.097 0.091 0.096 0.087 -- 0.091 0.051 0.506 0.494 0.432 0.472 OR2 0.071 0.081 0.091 0.086 0.082 0.086 0.081 0.084 0.091 0.090 0.020 -- 0.039 0.470 0.517 0.453 0.433 OR3 0.057 0.061 0.072 0.066 0.067 0.061 0.072 0.067 0.072 0.066 0.012 0.009 -- 0.419 0.419 0.355 0.376 OSG1 0.008 0.010 0.004 0.011 0.003 0.008 0.019 0.003 0.005 0.010 0.096 0.084 0.070 -- 0.030 0.055 0.070 OSG2 0.007 0.006 0.005 0.004 0.000 0.005 0.018 0.002 0.002 0.004 0.092 0.091 0.067 0.001 -- 0.018 0.049 SC1 0.003 0.005 0.004 0.008 0.006 0.012 0.015 0.001 0.006 0.012 0.088 0.083 0.064 0.004 0.003 -- 0.053 SC2 0.008 0.016 0.014 0.022 0.007 0.010 0.017 0.006 0.009 0.013 0.094 0.077 0.068 0.009 0.008 0.009 -- Values in bold indicate statistically significant differences from zero at P<0.05 with 9999 nonparametric permutations for FST and 100 bootstrap estimates for DEST.

Figure S1. (A) Log-likelihood [Ln P(k)] and (B) ∆k for each k over the 10 replicates in

STRUCTURE. Error bars are standard deviation.

Figure S2. Potential suitable habitat for Cumberlandia monodonta identified using ecological niche models (ENMs) using both “current” bioclimatic conditions (interpolations of observed data from 1950 to 2000) and landscape variables, and projections of four near-future climate scenarios (low and high greenhouse gas concentration in 2050, average for 2041-2060; and in 2070, average for 2051-2080; IPCC 5th Assessment). Low and high greenhouse gas concentrations were based on representative concentration pathways (RCP2.6 and RCP8.5, respectively). Gradient colors represent the probability of suitable habitat in percentage. Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs.

Figure S3. Potential suitable habitat for Cumberlandia monodonta identified using ecological niche models (ENMs) using “current” bioclimatic conditions (interpolations of observed data from 1950 to 2000), and projections of four near-future climate scenarios (low and high greenhouse gas concentration in 2050, average for 2041-2060; and in 2070, average for 2051-2080; IPCC 5th Assessment). Low and high greenhouse gas concentrations were based on representative concentration pathways (RCP2.6 and RCP8.5, respectively). Gradient colors represent the probability of suitable habitat in percentage. Models included ≥3rd order streams of the Mississippi River Basin; the maps show the eastern part of the Mississippi River Basin. Black dots on the Present map represent occurrence points included in the ENMs.

54 Chapter 3: Development and characterization of 20 polymorphic microsatellite markers for the Texas hornshell, Popenaias popeii (Bivalvia: Unionidae), through next-generation sequencing

Inoue K, Lang BK, Berg DJ (2013) Development and characterization of 20 polymorphic microsatellite markers for the Texas hornshell, Popenaias popeii (Bivalvia: Unionidae), through next-generation sequencing. Conservation Genetics Resources, 5, 195-198.

ABSTRACT We identified 28 microsatellite loci from Popenaias popeii, a freshwater mussel that has experienced population declines throughout its range. Twenty loci were polymorphic, with 4–10 alleles, observed heterozygosity values of 0.375–1.00, and 16 % of alleles shared by more than one population. These loci should be useful for describing population genetic diversity, which will facilitate ongoing conservation efforts for P. popeii.

Keywords: 454 shotgun pyrosequencing, de novo sequencing, freshwater mussels, microsatellite primers

INTRODUCTION Freshwater mussels (Unionoidea) are among the most endangered groups of animals in North America (Lydeard et al. 2004). Popenaias popeii (Lea 1857), the Texas hornshell, is endemic to the Rio Grande drainages in the southwest US and northern Mexico, along with coastal drainages in northeastern Mexico (Howells et al. 1996; Strenth et al. 2004). However, current US populations of P. popeii are limited to three rivers: lower Rio Grande and Devils River in Texas and a single 14 km reach of the Black River in New Mexico (Lang 2001); the status of Mexican populations is unknown (Strenth et al. 2004; Karatayev et al. 2012). The New Mexico population is isolated from Texas populations due to large dams and unsuitable water quality (Howells et al. 1996; Lang 2001; Strenth et al. 2004). Threats to this species are primarily anthropogenic activities including modification of physical conditions in streams, and reduction in water quality due to groundwater extraction and increased salinity levels resulting from irrigation water returns. Population reduction and current threats have made this species a candidate for listing under the US Endangered Species Act (USFWS 2005). Although life history

55 and population demography for this species have been studied (Smith et al. 2003; Levine et al. 2012), genetic assessment of extant populations is critical to inform conservation and management strategies. Development of highly variable microsatellite markers will allow such assessment of genetic structure within and among populations.

METHODS High throughput, next-generation sequencing (also known as 454 shotgun pyrosequencing) allows creation of massive genomic sequences that provide opportunities to develop microsatellite markers (Guichoux et al. 2011). We developed microsatellite markers for P. popeii using this de novo sequencing approach. We extracted genomic DNA from one individual using a DNeasy Kit (Qiagen, Inc.). The Plant–Microbe Genomics Facility at Ohio State University performed shotgun pyrosequencing using a Roche 454 FLX Titanium Genome Sequencer, which produced a total of 161,714 reads averaging 394 bp in length corresponding to ~63.8 Mbp generated from 1/4th of a picotiterplate. Putative microsatellite loci were identified from the dataset using MSATCOMMANDER v1.0.8 (Faircloth 2008) to design primers on flanking regions of microsatellite loci using PRIMER3 (Rozen & Skaletshky 2000) as its primer design engine and adding M13R or CAG tail sequence at the 5′-end of either forward or reverse primers. We identified 7,281 putative microsatellite motifs and were able to design primers for 699 microsatellite loci. Dinucleotides were the most abundant repeat type (45 % of all loci) followed by trinucleotides (38 %), tetranucleotides (13 %), hexanuleotides (3 %), and pentanucleotides (1 %). We genotyped four individuals each from the Black River, NM and Rio Grande, TX populations for 28 tetranucleotide loci. We conducted amplifications in a 10μL reaction volume using GoTaq Master Mix (Promega Corporation), 0.2 μM of universal fluorescently-labeled primer and non-tailed primer, 0.04 μM of tailed primer, and 10 ng of DNA template. PCR conditions were: initial denaturing at 95 °C for 2 min, followed by 35 cycles at 95 °C for 30 s, annealing at 58 °C for 45 s, extension at 72 °C for 45 s, and final extension at 72 °C for 30 min. For problematic primer pairs, we added a universal fluorescently-labeled primer to the PCR only in the final 10 cycles (de Arruda et al. 2010). We performed fragment analyses on an ABI Genetic Analyzer with LIZ600 size standard (Applied Biosystems, Inc.). We used PEAKSCANNER v1.0 (Applied Biosystems, Inc.) to score alleles and TANDEM v1.07 (Matschiner & Salzburger 2009) to assign integer numbers to DNA fragment sizes.

56 We estimated allelic richness (number of alleles per locus; NA) and observed heterozygosity (HO) using GENALEX v6.3 (Peakall & Smouse 2006). We also estimated the number of shared alleles at each locus. We used MICRO-CHECKER (van Oosterhout et al. 2004) to detect null alleles and GENEPOP v4.0.10 (Rousset 2008) to conduct exact tests of pairwise linkage disequilibrium.

RESULTS AND DISCUSSION Of the 28 microsatellite loci tested, eight showed unscoreable bands. The other 20 loci were polymorphic (Table 1). Allelic richness ranged from four to ten alleles, with an average of 7.15 alleles per locus. When we pooled all eight individuals as a single population, the average observed heterozygosity over 20 loci was 0.701 (range 0.375–1.000); the average number of shared alleles was 1.15 per locus (range 0–3), with 16 % of alleles found in both populations. We found homozygote excesses, suggesting the possibility of null alleles at two loci (Tetra08 and Tetra14). After correction with a sequential comparison Bonferroni technique (Lessios 1992), no locus pairs showed evidence of linkage disequilibrium. We were able to demonstrate use of next-generation sequencing techniques to easily identify and design primers for hundreds of microsatellite loci for a non-model organism. This approach was approximately one-fourth the cost of using a commercial laboratory to build an enriched library for development of microsatellites. A large number of the tested loci were polymorphic, suggesting that we will be able to provide high-level resolution of genetic structure within and among extant populations of P. popeii. As a result, estimates of population parameters such as effective population size and gene flow, and detection of demographic events such as historic bottlenecks, are likely to be robust. Such information will be of great value as agencies develop conservation strategies for this imperiled species.

ACKNOWLEDGEMENTS We thank the Plant–Microbe Genomics Facility at Ohio State University for next- generation sequencing, and staff in the Miami University Center for Bioinformatics and Functional Genomics for technical support. This study was funded by the New Mexico Department of Game and Fish.

57 REFERENCES de Arruda MP, Gonçalves EC, Schneider MPC, da Costa da Silva AL, Morielle-Versute E (2010) An alternative genotyping method using dye-labeled universal primer to reduce unspecific amplifications. Molecular Biology Reports, 37, 2031-2036. Faircloth BC (2008) MSATCOMMANDER: detection of microsatellite repeat arrays and automated, locus-specific primer design. Molecular Ecology Resources, 8, 92-94. Guichoux E, Lagache L, Wagner S, Chaumeil P, Léger P, Lepais O, Lepoittevin C, Malausa T, Revardel E, Salin F, Petit RJ (2011) Current trends in microsatellite genotyping. Molecular Ecology Resources, 11, 591-611. Howells RG, Neck RW, Murray HD (1996) Freshwater mussels of Texas Texas Park and Wildlife Department, Austin, TX. Karatayev AY, Miller TD, Burlakova LE (2012) Long-term changes in unionid assemblages in the Rio Grande, one of the world's top 10 rivers at risk. Aquatic Conservation: Marine and Freshwater Ecosystems, 22, 206-219. Lang BK (2001) Status of the Texas hornshell and native freshwater mussels (Unionoidea) in the Rio Grande and Pecos River of New Mexico and Texas. New Mexico Department of Game and Fish, Santa Fe, New Mexico. Lessios HA (1992) Testing electrophoretic data for agreement with Hardy-Weinberg expectations. Marine Biology, 112, 517-523. Levine TD, Lang BK, Berg DJ (2012) Physiological and ecological hosts of Popenaias popeii (Bivalvia: Unionidae): laboratory studies identify more hosts than field studies. Freshwater Biology, 57, 1854-1864. Lydeard C, Cowie RH, Ponder WF, Bogan AE, Bouchet P, Clark SA, Cummings KS, Frest TJ, Gargominy O, Herbert DG, Hershler R, Perez KE, Roth B, Seddon M, Strog EE, Thompson FG (2004) The global decline of nonmarine mollusks. Bioscience, 54, 321- 330. Matschiner M, Salzburger W (2009) TANDEM: integrating automated allele binning into genetics and genomics workflows. Bioinformatics, 25, 1982-1983. Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6, 288-295.

58 Rousset F (2008) GENEPOP'007: a complete re-implementation of the GENEPOP software for Windows and Linux. Molecular Ecology Resources, 8, 103-106. Rozen S, Skaletshky H (2000) Primer3 on the WWW for general users and for biologist programmers. In: Bioinformatics Methods and Protocols: Methods in Molecular Biology (eds. Krawetz S, Misener S), pp. 365-386. Humana Press, Totowa, NJ. Smith DG, Lang BK, Gordon ME (2003) Gametogenetic cycle, reproductive anatomy, and larval morphology of Popenaias popeii (Unionoida) from the Black River, New Mexico. Southwestern Naturalist, 48, 333-340. Strenth NE, Howells RG, Correa-Sandoval A (2004) New records of the Texas hornshell Popenaias popeii (Bivalvia: Unionidae) from Texas and northern Mexico. The Texas Journal of Science, 56, 223-230. USFWS (US Fish and Wildlife Service). 2005. Endangered and threatened wildlife and plants; review of native species that are candidates or proposed for listing as endangered or threatened; annual notice of findings on resubmitted petitions; annual description of progress on listing actions; proposed rule. Federal Register 70, 24870-24934. van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) MICRO-CHECKER: Software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes, 4, 535-538.

59 TABLE Table 1. Characterization of 20 polymorphic microsatellite loci for Popenaias popeii.

Repeat TA Size range Shared Locus Primer sequence (5'-3') motif (˚C) (bp) N NA HO alleles Tetra01 F: (GGAAACAGCTATGACCA)TACATCTACTCGCCGCTGG ACAT 58 235-275 8 7 0.625 0 R: GATGCGCATGAACAAACAGC Tetra02 F: CTCAGGTTTGTATGCCGAAATC ACAT 58 207- 267 8 10 0.750 0 R: (GGAAACAGCTATGACCA)TATGACCTCCTGCATCCGAG Tetra03 F: CTCATACAAGTCAGGTACGCC AAGT 58 331- 359 8 6 0.750 2 R: (GGAAACAGCTATGACCA)TGAAAGAATGCAGCCGGACG Tetra05 F: AGGGACCTAGTTAAAGCGGTC AGAT 58 243- 323 8 10 1.000 1 R: (GGAAACAGCTATGACCA)TGTGATAGGTGTTGTTGCTCG Tetra08 F: CTACATCCGACGTCATTGCC ACAT 58 356- 400 8 7 0.375 1 R: (CAGTCGGGCGTCA)TCATGTTACACACGCGAGATGG Tetra09 F: CGCCTAGTTCGCCTATTGTG ACAT 58* 416- 460 8 7 0.750 3 R: (GGAAACAGCTATGACCA)TGCCACGATATCTGCATTGGG Tetra14 F: GTCCGTGTGTGCAGGTTAAG AATG 58 165- 185 8 5 0.375 2 R: (GGAAACAGCTATGACCA)TGTACCGCGCGTTCCTTTAC Tetra15 F: TGGAACACTTGGCGAAACAG ACAT 58 150- 202 8 7 0.875 2 R: (CAGTCGGGCGTCA)TCATGGATTTGTGGATTGACTGGC Tetra17 F: GACTCGTAGCACTGTTTCGG AGAT 58 140- 184 8 10 1.000 1 R: (GGAAACAGCTATGACCA)TACATTTCACTCACTCACGCG Tetra19 F: TCTGCTTCTCTTGAGCGATG ACAT 58* 246- 286 8 6 0.500 1 R: (GGAAACAGCTATGACCA)TGCGCGCAATCCTAATCTTAC Tetra22 F: (GGAAACAGCTATGACCA)TGGGCACTATTCAACTCGGTG ACAT 58 288- 316 8 6 0.625 1 R: AACAGGAAGGTCGGTTAGCC Tetra23 F: (CAGTCGGGCGTCA)TCAAATCACATGTCTGCGGC ACAT 58 224- 304 8 10 0.875 0 R: GTGGGCGCAAATATGTCCTC Tetra24 F: AGATCTCTGCGTCCTTAGATTG AGAT 58 119- 163 8 6 0.625 1 R: (GGAAACAGCTATGACCA)TCTCACCCACTGAAAGATTCCC Tetra30 F: ACAGCTGCAACACCAATCTG ACAT 58* 262- 314 8 9 0.625 0 R: (GGAAACAGCTATGACCA)TAAACAGAATAGTGCTGATGTGC (table continues)

Table 1. (continued)

Tetra31 F: AGTTGATCTTGACCGTCGTG ACAT 58 308-368 8 10 0.875 0 R: (GGAAACAGCTATGACCA)TTCGCCGAATCCTAGACTCG Tetra33 F: (GGAAACAGCTATGACCA)TACATACACACACCGGTCTGC ACAT 58 210- 282 8 6 0.500 2 R: TGAGAGATGAACGGACGGAC Tetra36 F: (GGAAACAGCTATGACCA)TCTCTAAGTAAGACGCCGGC AGAT 58 196- 216 8 5 0.375 2 R: TCTTCCTGCTCCTGTCACAC Tetra37 F: GATATCCCAAACTGCCGCAC AAAT 58 179- 187 8 6 0.766 1 R: (CAGTCGGGCGTCA)TCATAGTATCAACGCAGGCACC Tetra40 F: TGCTTTGTTTCAAACGATGGG ACAT 58 145- 221 8 6 1.000 1 R: (CAGTCGGGCGTCA)TCAGCTTCTCGGGAATTTGTGC Tetra41 F: TTGTAGCGTTCACCCTCCC ACAT 58 203- 219 8 4 0.750 2 R: (GGAAACAGCTATGACCA)TCAAGTAGCTCGTGGAAACG Mean 7.15 0.701 1.15 (SE) (0.437) (0.046) (0.196) Shown are locus names, primer sequences (with a tail sequence in parentheses), repeat motif, annealing temperature (TA) size range of alleles in base pairs (bp), sample size (N), number of alleles (NA), observed heterozygosity (HO), number of shared alleles (those found in New Mexico and Texas populations). F = forward sequence, R = reverse sequence, SE = standard error and * = use of de Arruda et al. (2010) method.

61 Chapter 4: Past climate change drives current genetic structure of an endangered freshwater mussel species

Inoue K, Lang BK, Berg DJ (2015) Past climate change drives current genetic structure of an endangered freshwater mussel species. Molecular Ecology, 24, 1910-1926.

ABSTRACT Historical-to-recent climate change and anthropogenic disturbance affect species distributions and genetic structure. The Rio Grande watershed of the United States and Mexico encompasses ecosystems that are intensively exploited, resulting in substantial degradation of aquatic habitats. While significant anthropogenic disturbances in the Rio Grande are recent, inhospitable conditions for freshwater organisms likely existed prior to such disturbances. A combination of anthropogenic and past climate factors may contribute to current distributions of aquatic fauna in the Rio Grande basin. We used mitochondrial DNA and 18 microsatellite loci to infer evolutionary history and genetic structure of an endangered freshwater mussel, Popenaias popeii, throughout the Rio Grande drainage. We estimated spatial connectivity and gene flow across extant populations of P. popeii and used ecological niche models (ENMs) and approximate Bayesian computation (ABC) to infer its evolutionary history during the

Pleistocene. STRUCTURE results recovered regional and local population clusters in the Rio Grande. ENMs predicted drastic reductions in suitable habitat during the last glacial maximum. ABC analyses suggested that regional population structure likely arose in this species during the mid-to-late-Pleistocene, and was followed by a late-Pleistocene population bottleneck in New Mexico populations. The local population structure arose relatively recently, perhaps due to anthropogenic factors. Popenaias popeii, one of the few freshwater mussel species native to the Rio Grande basin, is a case study for understanding how both geological and anthropogenic factors shape current population genetic structure. Conservation strategies for this species should account for the fragmented nature of contemporary populations.

Keywords: Approximate Bayesian computation (ABC), aquatic connectivity, ecological niche modeling, Chihuahuan Desert, desert streams

62 INTRODUCTION A major interest in evolution and population genetics is the impact of past climate conditions and vicariant events on current species’ distributions and the genetic structure of natural populations (Avise 2000; Hewitt 2000). Climate change and geological processes often cause strong directional selection in natural populations (Hewitt 2000). For example, during the Pleistocene interglacial periods, times of drastic climate change altered freshwater systems by rerouting rivers and inundating large areas of land with meltwater (Hocutt et al. 1978; Mayden 1988; Strange & Burr 1997). Consequently in the Northern Hemisphere, organisms were extirpated over northern portions of their ranges, survived in isolated populations, and/or dispersed to new locations (Hewitt 2000). In fragmented populations, neutral forces such as genetic drift may have shaped genetic structure. Thus, current species distributions and genetic structure at both intra- and interspecific levels were formed through isolation of populations via vicariant events followed by recolonization with secondary contact during Pleistocene glacial- interglacial cycles (Mäkinen & Merilä 2008; Bossu et al. 2013; Inoue et al. 2014b). Studies of evolutionary responses to vicariant events are especially timely, given rapid changes in ecological conditions due to climate change and anthropogenic disturbance. Recently, the development of advanced statistical methods has led to improved ability to reconstruct the ways in which past climate change and geological events have shaped existing species’ distributions and phylogeographic patterns (Knowles 2009; Fordham et al. 2014). Coalescent-based population genetic approaches allow researchers to infer the evolutionary histories of species. For example, with the development of coalescent methods via approximate Bayesian computation (ABC), we can test hypothesized demographic and evolutionary scenarios by creating many simulated datasets, measuring similarity between those datasets and empirical data, and estimating the posterior probability that a given scenario is a true model (Beaumont 2010). By using approximations of likelihood inference, ABC can manage much more complex demographic and evolutionary models when full-likelihood methods are not efficient (Nielsen & Beaumont 2009). Ecological niche models (ENMs), which predict suitable habitats for a given species by estimating the relationship between species occurrences and environmental characteristics of locations, also provide powerful tools for evolutionary biologists, biogeographers, paleoecologists, and conservation biologists (Elith et al. 2006; Fordham et al. 2014). The combination of coalescent-based approaches and ENMs promises to shed light on a

63 variety of issues, including the evolutionary histories of species, effects of population fragmentation and connectivity, and future changes in distributional ranges and population connectivity (Carstens & Richards 2007; Alvarado-Serrano & Knowles 2014). Aquatic ecosystems in the Chihuahuan Desert of the southwestern United States (USA) and northern Mexico harbor many endemic species adapted to harsh environments (Kodric- Brown & Brown 2007; Seidel et al. 2009), which make the Chihuahuan Desert one of the world’s most biologically rich and diverse deserts (Olson & Dinerstein 1998). The region’s Rio Grande watershed (length: 3051 km; basin area: 471,900 km2) encompasses diverse aquatic ecosystems that are intensively exploited, making the Rio Grande one of the world’s top 10 rivers-at-risk (Wong et al. 2007). A doubling of the Rio Grande basin’s human population every 20 years (Schmandt 2002) has led to substantial modification and degradation of aquatic habitats. In the last 100 years, approximately half of the region’s fishes have been extirpated or have become threatened with extinction (Edwards et al. 2002). While anthropogenic disturbance seems to have affected the current distribution of freshwater fishes in the Rio Grande basin, some investigators question this hypothesis. A recent study of the widespread cyprinid Rhinichthys cataractae found that inhospitable habitat conditions likely existed prior to major human alteration of freshwater ecosystems (Kim & Conway 2014). Based on divergence time estimates between the lower and upper Rio Grande populations of R. cateractae, the authors hypothesized that inhospitable habitat conditions in the middle Rio Grande may have existed during the Pleistocene and created current population genetic structure of riffle-dwelling fishes. Thus, a combination of anthropogenic and geological factors may contribute to current distributions of aquatic fauna in Rio Grande drainages. We investigated the influence of past climate changes and contemporary anthropogenic disturbance on population genetic structure and connectivity of Popenaias popeii. This freshwater mussel is endemic to Rio Grande drainages in the USA and several Mexican tributaries of the Rio Grande, and coastal Gulf of Mexico drainages in northern Mexico (Strenth et al. 2004; Carman 2007). Freshwater mussels are among the most endangered groups of animals in North America (Lydeard et al. 2004), with ~70% of all mussel taxa having conservation status (Williams et al. 1993). These organisms possess a complex life cycle in which larvae (glochidia) are obligate parasites of vertebrate hosts. Because adult mussels tend to be sedentary, host movement is responsible for most mussel dispersal; this relationship has likely

64 led to similar patterns in local species composition and distribution of mussels and their hosts (Schwalb et al. 2013). In the case of P. popeii, three species of freshwater fishes have been identified as the primary hosts of its glochidia (Levine et al. 2012). Historically, P. popeii is known to have occurred in the lower Pecos River, from North Spring River (Chaves County, New Mexico) downstream to the Rio Grande confluence; in the Rio Grande from San Francisco Creek (Brewster County, Texas) downstream to the lower Rio Grande (Cameron County, Texas) near the Gulf of Mexico; and in major tributaries of the Rio Grande in the USA and Mexico. Currently, however, P. popeii populations in the USA are restricted to the lower Rio Grande and Devils River in Texas, and a 14-km stretch of the Black River (a Pecos River tributary) in New Mexico (Figure 1; Lang 2001; Karatayev et al. 2012) where P. popeii represents the last remaining native freshwater mussel in the state. Recent surveys of Mexican rivers did not discover live P. popeii specimens (Strenth et al. 2004). Current extant populations are likely isolated due to loss of connecting habitats and inability of hosts to move through stretches of river with intermittent flow, inhospitable water conditions, and a series of impoundments (Carman 2007). Reductions in its current range have resulted in P. popeii being declared “critically endangered” by the International Union for the Conservation of Nature (Bogan 1996) and a candidate for listing under the US Endangered Species Act (USFWS 2013). We used mitochondrial DNA (mtDNA) gene sequences and microsatellite loci to infer the evolutionary history and population genetic structure of P. popeii throughout the Rio Grande drainage. Because current population fragmentation is likely due to anthropogenic activities, we tested the hypothesis that among-drainage population divergence occurred with European settlement of the region. We used ENMs with current climatic conditions to predict suitable environments for P. popeii, and projected the model into the past based on mid-to-late- Pleistocene climatic scenarios. Employing an ABC framework, we then tested several alternative hypotheses regarding the demographic history of the species based on changes in population size, distributional range, and diversification among populations over evolutionary time. If recent (post-European arrival) anthropogenic activities are the major factor driving current distribution and genetic structure, we predict that populations would have experienced severe genetic bottleneck events and diverged as recently as several hundred years before present due to these disturbances. Our results identify both fine- and regional-scale genetic differentiation and

65 population structure within P. popeii—essential knowledge for conservation planning and management of imperiled species.

METHODS Sampling, genetic data collection, and analyses of population genetic diversity Using snorkeling, SCUBA, and tactile methods, we sampled 254 P. popeii from two tributaries and one main-stem region of the Rio Grande basin, including 193 individuals from eight locations in the Black River, New Mexico; 58 individuals from five locations in the mainstem Rio Grande, Texas; and three individuals from three locations in the Devils River, Texas (Table 1; Figure 1). Given the small sample size, we pooled samples from the Devils River into a single “population” prior to analyses. We sampled nondestructively using tissue swabs, collected from between foot and mantle tissues by rubbing mucus and epidermal cells using sample collection swabs (Epicentre Biotechnologies, Madison, WI), and then returned mussels to the riverbed. Samples were preserved in 95% ethanol and stored at 20˚C. Total genomic DNA was extracted using ArchivePure DNA Cell/Tissue Kits (5 Prime, Gaithersburg, MD), diluted to 10 ng/µL, and used as a template in polymerase chain reactions (PCR) for mtDNA and microsatellite analyses.

We used PRIMER3 (Untergasser et al. 2012) to design primers (forward: 5′- TGTGGGGTGAATCATTCCTT-3′ and reverse: 5′-TAAACCTCAGGATGCCCAAA-3′) from a complete mtDNA genome of the mussel Lampsilis ornata (GenBank accession number: NC_005335). These primers amplified about 810 basepairs of part of the cytochrome oxidase II gene and the cytochrome oxidase I gene (hereafter collectively abbreviated as COX). We followed procedures for conditions for PCR, sequencing, and post-sequencing analyses described in Inoue et al. (2014b).

Using DNASP v5.10 (Librado & Rozas 2009), we estimated population genetic indices from mtDNA sequences, including number of haplotypes (H), mean number of basepair differences (K), and mean nucleotide diversity (π) over the pooled dataset and within each locality and river because we did not have a priori population genetic information. To correct for sample-size bias, we estimated rarefied number of haplotypes (HR) with a standardized sample size of six individuals using the VEGAN package (Oksanen et al. 2015) in R v3.0.3 (R

66 Development Core Team 2011). We then used TCS v1.21 (Clement et al. 2000) to build a 95% confidence parsimony network from COX haplotypes. Assigning the shortest path from the most frequent haplotype simplified multiple connections between haplotypes (Fetzner & Crandall 2003). We genotyped populations of P. popeii at 20 tetra-nucleotide microsatellite loci (Inoue et al. 2013). Forward primers for each PCR were labeled with a 5′ fluorescent tag (6-FAM, NED, PET, or VIC) for visualization. We performed five sets of multiplex PCR (Plex1: Tetra17-19-41; Plex2: Tetra01-09-14-24-30-36; Plex3: Tetra02-03-22-23; Plex4: Tetra05-31-40; and Plex5:

Tetra08-15-33-37) designed by MULTIPLEXMANAGER (Holleley & Geerts 2009). Thermal cycling began with initial denaturing at 95˚C for 2 min; 35 cycles of 94˚C for 30 s, 60˚C for 1.5 min, and 72˚C for 1 min; and final extension at 72˚C for 30 min. We used previously published procedures for fragment analyses, allele scoring, and assignment of integer numbers to DNA fragment sizes (Inoue et al. 2014b).

We tested for the presence of null alleles and large allele dropout using MICRO-CHECKER (van Oosterhout et al. 2004). We checked for microsatellite loci under directional or balancing selection using LOSITAN (Antao et al. 2008). Using GENEPOP v4.0.10 (Rousset 2008), we conducted exact tests for pairwise linkage disequilibrium and deviation from Hardy-Weinberg expectation (HWE) for each locality, along with several population genetic indices (mean number of alleles per locus, NA; observed and expected heterozygosities, HO and HE; and number of private alleles, NP) using GENALEX v6.3 (Peakall & Smouse 2006). We used rarefaction to a standardized sample size of six individuals to correct mean allelic richness (rarefied number of alleles per locus; AR) in FSTAT v2.9.3 (Goudet 1995). To assess the effect of geographic distance on genetic structure, we examined the correlation of matrices of pairwise genetic differences (ST for mtDNA sequences and DEST for microsatellites) and geographic distances between pairs of localities using Mantel tests. We estimated the pairwise genetic differences using GENALEX. We measured total distance between pairs of localities along the rivers (river distance) using ArcGIS v10.2 (ESRI, Inc.). We excluded localities with small population size (<5 individuals) and performed Mantel tests in GENALEX with 9999 permutations.

67 Population genetic structure

We used STRUCTURE v2.3.4 (Pritchard et al. 2000) to evaluate population genetic structure without a priori assignment of individuals to populations. We ran STRUCTURE using the admixture model with correlated allele frequencies to account for possible ancestral admixture.

Each STRUCTURE run used a burn-in period of 500,000 Markov chain Monte Carlo (MCMC) generations followed by 100,000 iterations for k = 1 through 10 with 10 replicates for each k. To determine the most likely number of distinct clusters, we evaluated the log-likelihood [lnP(k)] for each k and estimated ∆k using STRUCTURE HARVESTER (Earl & vonHoldt 2012). We used CLUMPP (Jakobsson & Rosenberg 2007) to average each individual's admixture proportions over the 10 replicates for the best k, then produced graphical display results using DISTRUCT (Rosenberg

2004). We employed GENALEX to estimate pairwise FST and DEST among STRUCTURE-defined populations. For both indices, we tested for statistically significant differences from 0 using 9999 permutations. We used only the microsatellite dataset for these indices of genetic differentiation. Historic and contemporary migration among populations

Using MIGRATE-N v3.3.2 (Beerli 2006), we estimated migration rates among populations and mutation-scaled effective population size (θ; θ = 4Neµ, where Ne = effective population size and µ = mutation rate per generation) for each population. MIGRATE-N implements coalescent- based MCMC simulations to estimate a long-term average migration parameter (M; M = m/µ, where m = migration rate) representing migration between pairs of populations over approximately 4Ne generations in the past. We performed maximum likelihood analyses in

MIGRATE-N using ten short chains of 20,000 steps and two long chains of 200,000 steps, sampled every 20 steps following a burn-in of 1000 steps. To increase the efficiency of the MCMC, we used four-chain heating at approximately exponential increasing temperatures of 1.0, 1.5, 3.0, and

1,000,000.0. We randomly chose 20 individuals per population to use in our MIGRATE-N analyses because of computational demands and evidence that >20 individuals does not increase the accuracy of parameter estimates (Kuhner 2006). We repeated this procedure 10 times to ensure consistency of estimates, and reported average maximum likelihood estimates of θ and M along with 95% confidence intervals.

We calculated contemporary migration rates (over the last few generations; mc) between pairs of populations using BAYESASS v3.0.3 (Wilson & Rannala 2003). We performed a run with 10 million generations, sampling every 1000 generations after discarding the first 5 million

68 generations as burn-in. We set delta values (i.e., maximum parameter change per iteration) to 0.10, 0.30, and 0.30 for allele frequency, migration, and inbreeding parameter estimates, respectively. We estimated average mc and the 95% credible interval as mc ± 1.96 • standard deviation. For the MIGRATE-N and BAYESASS analyses, we used STRUCTURE-defined populations based on k = 3. Ecological niche modeling To predict suitable environments for P. popeii in the past and present throughout the Rio

Grande drainage, we developed ENMs employing the maximum entropy algorithm in MAXENT v3.3.3 (Phillips et al. 2006) using geo-referenced points from this study and collection records from Karatayev et al. (2012). These occurrence points represent the present-day distribution of P. popeii in the Rio Grande drainage. Given that occurrence data often show strong spatial bias in sampling efforts, we used SDMTOOLBOX v1.0b (Brown 2014) to reduce spatial autocorrelation in the occurrence data by selecting one record within a 5-km radius (Phillips et al. 2009; Kramer- Schadt et al. 2013). We then created a layer of the Gaussian kernel density of sampling locations (i.e., a bias layer) with a bandwidth of 50 km to control for background sampling effort. We set our modeling area to include a 10-km buffer around major streams where P. popeii potentially persists: the Rio Grande, Devils, and Pecos rivers in the USA; and Rio Conchos, Rio Salado, and Rio San Juan in Mexico (Figure 1). We obtained 19 bioclimatic layers from WorldClim

(http://www.worldclim.org; Hijmans et al. 2005), and used ARCGIS and SDMTOOLBOX to produce a base-map and bioclimatic layers with the same map projection and resolution (1 km2). Using

SDMTOOLBOX, we identified eight uncorrelated climatic layers (<0.7 Pearson correlation coefficient; Table S1, Supplementary Information) to identify suitable environments for P. popeii. Using MAXENT, we first built ENMs based on current bioclimatic data (interpolations of observed data from 1950 to 2000), then projected the current ENMs to paleoclimate scenarios from the last interglacial (LIG; 120 – 140 ka based on Otto-Bliesner et al. 2006) and the last glacial maximum (LGM; 21 ka based on the Model for Interdisciplinary Research on Climate [MIROC]) to predict suitable climatic areas during the Pleistocene. We checked model accuracy by using a random sample of 25% of the dataset as training data and the remaining 75% as testing data for model validation with 10 replicates. We evaluated the ENM results using the area under the curve (AUC) of the receiver operating characteristics (ROC) curve. The AUC ranges from 0.5 (random accuracy) to 1.0 (perfect discrimination).

69 Because freshwater mussels require vertebrate hosts to complete their life cycles, and mussel dispersal likely occurs during this parasitic glochidial stage, suitable environments for P. popeii may be similar to those of its host species. Thus, we also built ENMs for each of three primary host fishes for P. popeii (Carpiodes carpio, Cyprenella lutrensis, and Moxostoma congestum; Levine et al. 2012) using current bioclimatic data and projected the ENMs to paleoclimate scenarios described above. We chose primary hosts because these species represented 80% of all individual fishes infested and carried more than 99% of glochidia (Levine et al. 2012). The Global Biodiversity Information Facility (http://www.gbif.org/), Fishes of Texas Project (http://www.fishesoftexas.org), and survey records from New Mexico Department of Game and Fish, provided geo-referenced points for these species. We then reduced spatial autocorrelation in the occurrence data and created kernel density bias layers for each species as described above. Approximate Bayesian computation analysis of demographic history We investigated the demographic history of P. popeii by using the ABC framework implemented in DIYABC v2.0.4 (Cornuet et al. 2014) to estimate divergence times between

STRUCTURE-defined populations, and to compare demographic scenarios characterized by the existence of founder effects and population bottlenecks (Figure 2). In our DIYABC analyses, we only employed microsatellite markers due to the absence of variation in COX sequences from the

Black River populations (see Results). We tested four demographic scenarios in DIYABC. For each scenario, we placed current populations at time t0 (i.e., the present), divergence within the

Black River at t1, and divergence of the ancestral Black River and Rio Grande populations from the most recent common ancestral population (MRCAP) at t2 (Figure 2). Scenario 1 was a null model that held size constant for each population throughout the entire simulation. Because genetic diversity of Black River populations was consistently lower than Rio Grande populations (see Results), Scenario 2 assumed a small number of founders colonized the Black River followed by population expansion. Alternatively, Scenario 3 assumed a population bottleneck in the Black River populations. Scenario 4 assumed bottlenecks in both the Black River and Rio Grande populations.

Table 2 describes prior values for Ne at various points in time, which were chosen based on  estimated from MIGRATE-N and several preliminary ABC runs to confirm upper boundaries

70 of prior values. Based on the Ne priors, we calculated the following summary statistics: mean number of alleles, mean genic diversity, mean allele size variance, and mean Garza-Williamson's 2 M for each population. We also calculated FST, the classification index, and (δµ) distance for each pair of populations. Our DIYABC analyses employed a generalized stepwise mutation model with a mean mutation rate of 1 x 10-5 to 10-3. For each scenario, we ran one million simulations to estimate posterior probabilities using logistic regression (Cornuet et al. 2014). We created 500 pseudo-observed datasets (i.e., datasets simulated with known parameters; Cornuet et al. 2010; Inoue et al. 2014b) in order to evaluate Type I and Type II error rates. To assess precision of parameter estimation, we computed the relative median of the absolute error (RMAE; Cornuet et al. 2010) on 500 pseudo-observed datasets simulated under the best-fit scenario.

RESULTS Population genetic diversity We successfully sequenced 246 individuals and recovered 34 COX haplotypes (Table 1); sequences were submitted to GenBank (accession numbers KP779655 – KP779900). The mainstem Rio Grande (hereafter RG) showed the highest diversity with 33 haplotypes, whereas in the Black and Devils rivers (hereafter, BR and DR, respectively), only single haplotypes were recovered at all sites (Table 1; Figure 3). The BR haplotype was also the most common haplotype in the Rio Grande. The haplotype unique to the Devils River differed from the BR haplotype by a single base pair. Two microsatellites (Tetra41 and Tetra37) were excluded from analyses because Tetra41 amplified multiple fragments and we were unable to score alleles, and Tetra37 showed signs of positive selection. Over 18 loci, we found some evidence of null alleles (4.8% of all locality-by- locus pairs), no evidence of linkage disequilibrium, and some deviation from HWE (2% of all locality-by-locus pairs) after sequential Bonferroni correction. However, neither the null alleles nor the deviations from HWE showed any pattern across localities or loci; accordingly, we included these loci in all subsequent analyses. We scored a total of 349 different alleles over the 18 loci examined, ranging from five to 51 per locus. Mean rarefied allelic richness and heterozygosity across the 18 loci were similar among sites within rivers, but varied among rivers (Table 1). Overall mean AR was more than two times higher in the RG sites than the BR sites, and overall mean heterozygosities were more than

71 1.5 times higher in the RG sites than the BR sites. Overall, the BR sites exhibited the lowest genetic diversity (Table 1). There were significantly positive correlations between pairwise genetic differences and river distances (Mantel’s r = 0.655, P = 0.002 for ST; Mantel’s r = 0.991, P < 0.001 for DEST; Figure S1, Supplementary Information). However, these significant correlations were likely due to low genetic variance within rivers and high genetic variance among rivers. When we decomposed the relationships into pairs of localities within and among rivers, only DEST showed slight positive isolation-by-distance patterns within rivers (Figure S1E). Population genetic structure Popenaias popeii showed evidence of significant range-wide population genetic structure.

The STRUCTURE analysis recovered explicit boundaries between the Black River and Devils River/Rio Grande at k = 2, indicated by lnP(k) = −12576.0 and ∆k = 1305.5 (Figures 4 and S2,

Supplementary Information). At k = 3, STRUCTURE further split the BR sites into two clusters, and this assignment scheme received the highest log-likelihood estimate, lnP(k) = −12473.7, and the second best ∆k estimate, ∆k = 63.4 (Figures 4 and S2). The BR sites were split between three upstream sites (hereafter BR-u) and five downstream sites (hereafter BR-d) at k = 3, where BR-d had admixture of BR-u genotypes (Figure 4). We found no evidence of admixture between BR and RG populations when k = 2 or k = 3. Generally, ∆k is thought to capture the strongest patterns of population structure (Evanno et al. 2005; Coulon et al. 2008), and the highest lnP(k) can be useful to investigate finer population structure (e.g., Coulon et al. 2008; Fisher-Reid et al.

2013; but see Kalinowski 2011). The STRUCTURE analysis on the BR sites alone showed the split between BR-u and BR-d at k = 2 received the highest log-likelihood and ∆k estimates (lnP(k) = −6965.0, ∆k = 5.0); this assignment scheme was consistent with the overall dataset. Thus, our

STRUCTURE results showed that both the best likelihood and the second best ∆k reflect fine-scale structure in the Black River. Because the BR-u and BR-d groups of sites are separated by a low- water crossing which may inhibit fish movement (see Discussion), we concluded that k = 3 represents the most biologically relevant clustering scheme for P. popeii.

All pairwise comparisons of FST and DEST were significant (Table 3). Pairwise FST values between populations ranged from 0.033 between BR-u and BR-d to 0.255 between BR-d and RG.

72 Similar genetic differentiation was observed for DEST values, ranging from 0.015 between BR-u and BR-d to 0.751 between BR-u and RG. Historic and recent migration among populations

Estimates of θ from MIGRATE-N ranged from 0.928 (95% CI: 0.867-0.996) for BR-u to 35.120 (32.062-38.605) for RG (Table 4). Estimates of long-term average migration (M) indicated asymmetrical migration between pairs of populations. We inferred relatively high bidirectional migration between BR-u and BR-d. Migration between the two BR populations and RG was high and unidirectional, where RG was a net producer of migrants entering the BR populations.

In contrast, estimates of contemporary migration rates (mc) were consistently low among populations. Our BAYESASS analysis indicated no current migration among population pairs because all migration rates included 0 in the 95% credible interval, except migration was non- zero from BR-d into BR-u (0.320, 95% CI: 0.303-0.338; Table 3). Thus, RG was not a net producer of migrants to the two BR populations; instead, BR-d was a net producer of migrants entering BR-u. Ecological niche modeling All ENMs had high AUC values (>0.75; Table S1), indicating overall adequate model performance. In the ENMs for P. popeii, maximum temperature of warmest month (41.0%) and mean diurnal temperature range (24.2%) were the best predictors of suitable environments, followed by precipitation seasonality (15.4%) and precipitation of warmest quarter (13.9%; Table S1). The present-day ENM model predicted suitable environment that closely matched the known distribution of P. popeii in the USA (Figure 5). Additionally, the ENMs predicted potential suitable environment in the lower Pecos River and the Rio Salado. ENMs of the LGM model predicted drastic reduction of suitable environment with low suitability scores. However, the LIG predicted suitable environment similar to the present-day model, but with an even greater total area of predicted suitable environment (Figure 5). These results suggest that P. popeii populations were potentially continuously distributed and much more widespread throughout the Rio Grande drainage during the LIG, but the range was reduced and populations fragmented during the LGM. The ENMs for the host species predicted wider distributional ranges of suitable environments than those of P. popeii (Figure S3, Supplementary Information). Like ENMs for P.

73 popeii, maximum temperature of warmest month best predicted suitable environments for the host species; however, additional environmental variables varied among host species (Table S1). Consistent with the ENM results for P. popeii, the ENM models of the paleodistributions of the host species predicted a wide range of suitable environments in the Rio Grande drainage during the LIG, followed by drastic reduction of suitable environments during the LGM. Approximate Bayesian computation analysis of demographic history

Our ABC analysis using DIYABC identified Scenario 3, which specified a population bottleneck in the ancestral Black River, as the most highly supported scenario, with strong posterior probability (0.740; Table 5). All other scenarios received much lower support. Based on our ABC analyses of pseudo-observed datasets, the Type I error rate for this most-probable scenario was relatively high (Table 5), but the Type II error rate was low. Although a high Type I error indicated that this scenario was less frequently chosen when it was the true scenario, the high posterior probability and low Type II error suggest greater confidence that this is the best of the proposed scenarios.

The RMAEs were relatively high for NBR and te2, but low for the rest of the parameters, indicating most parameters estimated by ABC were reliable values (Table 2). The high posterior probability of Scenario 3 indicated that the divergence from MRCAP occurred approximately 9020 generations ago (95% CI: 2370 – 18,900; RMAE = 0.185). This was followed by a severe population bottleneck event in the ancestral BR population 3670 generations ago (95% CI: 614 – 11,000; RMAE = 0.316), and divergence between BR-u and BR-d about 125 generations ago (95% CI: 12 – 410; RMAE = 0.298). Although no species-specific generation time for P. popeii is known, when using an average generation time of 8.9 years (based on 8.1 - 9.6 years for Lampsilis radiata; Chagnon & de la Cheneliere 1998), we inferred that the time of the divergence from MRCAP, the population bottleneck, and the divergence between BR-u and BR-d are 80.3 ka (21.1 – 168.2 ka), 32.7 ka (5.5 – 97.9 ka), and 1.1 ka (0.1 – 3.6 ka), respectively. During the bottleneck event, the ancestral BR population fell by 96% (from Ne = 39,400 to 1630 individuals;

Table 2). Posterior distributions of current Ne varied among populations; the largest Ne was RG with 70,300 individuals (41,600 – 91,900; RMAE = 0.200), whereas the two BR populations were small (1190 individuals for BR-u and 4680 individuals for BR-d). Mean mutation rate over 18 microsatellite loci was estimated to be 1.53 × 10-4 per generation (7.48 × 10-5 – 3.39 × 10-4, RMAE = 0.214).

DISCUSSION Spatial and demographic history of Popenaias popeii during the Pleistocene We found that P. popeii populations were highly structured among Rio Grande drainages at a regional scale, and within the Black River at a more local scale. Based on our population genetic modeling and ENM analyses, we conclude that this pattern probably derives from historical demographic events associated with climate change. For example, the ENM results indicated considerable suitable environment for P. popeii during the LIG (ca. 120 – 140 ka; Figure 5), suggesting spatial connectivity among populations, followed by drastic reductions in suitable environment in the LGM (ca. 12 ka; Figure 5). These distributional shifts are nearly congruent with time estimates from coalescent simulations; ABC analyses suggested that population divergence from the MRCAP occurred during the mid-to-late Pleistocene (ca. 80 ka; Table 2) followed by a severe population bottleneck in the ancestral BR population (ca. 33 ka; Table 2), but not in the RG population. Even though P. popeii might not have had continuous suitable environments throughout the Pecos River during the LIG, the continuous distribution of suitable environments for all host species indicates that P. popeii was capable of dispersing throughout the drainage. In fact, estimates of historic gene flow were high but rather unidirectional from RG populations to BR populations, suggesting that RG populations were source populations for the Black River. Although MIGRATE-N and ABC analyses have different assumptions, where the former assumes populations in migration-drift equilibrium and the latter assumes divergence without gene flow, both analyses indicate that RG and BR populations diverged during the mid-to-late-Pleistocene and RG populations retained most of the ancestral polymorphisms. Furthermore, BR individuals share a single COX haplotype—the most common haplotype in the RG population—also consistent with historical migration between the rivers and recent divergence. Apparently, subsequent glaciation reduced the extent of suitable environment during the LGM (ca. 21 ka); such environments for P. popeii and its hosts were small and fragmented, presumably leading to greatly decreased levels of dispersal and gene flow among populations and a substantial bottleneck in the ancestral BR population. Although glaciers advanced only to southern New Mexico’s Sierra Blanca (Smith & Miller 1986), continental glaciation caused cooler summers and a significant increase in winter precipitation in much of the Chihuahuan Desert during the Pleistocene glacial-interglacial periods

75 (Metcalfe et al. 2000). Paleoecological studies have shown that the Pleistocene’s dynamic climate induced distributional shifts in unglaciated areas. For example, glacial-interglacial cycles prompted severe range contractions and population reductions in currently widespread species of fishes (Miller 1977) and more geographically restricted populations of land snails (Bequaert & Miller 1973; Metcalf 1997) in the Chihuahuan Desert. More recent studies using genetic and ENM approaches identified population isolation followed by species diversification in small mammals in North American deserts (Jezkova et al. 2009; Mantooth et al. 2013). Similarly, our results indicate that the LGM’s cooler climate reduced suitable environment for P. popeii, likely isolating the ancestral BR and RG populations. The subsequent bottleneck substantially reduced

Ne of the ancestral BR population; a lack of connectivity with the Rio Grande likely caused current relatively low levels of genetic variation and led to complete lineage sorting of COX haplotypes in the BR populations. In contrast, we did not detect bottlenecks in RG populations from ABC analyses, indicating that RG populations maintained higher Ne, and thus, higher genetic diversity through to the present. We estimated the timing of historical demographic events based on generalized mutation rates and average generation time for L. radiata, which might differ from species-specific values for P. popeii. Nevertheless, our results indicate that the BR and RG populations diverged during the mid-to-late-Pleistocene and the BR population experienced a severe population bottleneck during the late Pleistocene. Vicariant events associated with tectonic activity and climate change can also reroute rivers, shifting distributions of aquatic organisms. Such events frequently occurred in Rio Grande drainages during the Pleistocene (Thomas 1972; Galloway et al. 2011). Because we used current Rio Grande drainages to project historic ENMs, the ENMs might overestimate suitable environment for P. popeii and its hosts. For example, pluvial lakes formed along the mid-Rio Grande during the Pleistocene (Metcalfe et al. 2000; Galloway et al. 2011). Furthermore, the upper Rio Grande (above El Paso, Texas) was not connected with the lower Rio Grande. Instead, its headwaters, along with the upper Pecos River, were tributaries of the Canadian River (Mississippi River drainage) and headwaters of the Brazos River (Gulf of Mexico drainage) during the mid- and late-Pleistocene (Thomas 1972; Smith & Miller 1986). Thus, although the upper Pecos River and Rio Grande provided suitable environments for P. popeii and its host species, they likely did not occur there due to lack of stream connections. Additionally, our ENMs were based only on bioclimatic data. Although bioclimatic ENMs have been used for

76 freshwater fishes to predict their contemporary ranges and range shifts under future climate change scenarios (e.g., Bond et al. 2011), given the scale of resolution of such data, our bioclimatic ENMs might only give a rough indication of which parts of drainages are suitable for these species. Previous studies have found that P. popeii inhabits distinctive microhabitats, including undercut riverbanks and the bases of large boulders (Lang 2001; Karatayev et al. 2012; Inoue et al. 2014a), that are not distributed uniformly throughout these drainages (Inoue et al. 2014a). Additional geological and landscape information may increase accuracy of the ENMs. Currently, such information during the Pleistocene is not available for this region. Contemporary gene flow and genetic structure Although the BR and RG populations diverged during the Pleistocene, we detected finer spatial-scale genetic structure within segments of the Black River, perhaps driven by anthropogenic factors (Figure 4). ABC analyses indicated recent divergence between BR-u and

BR-d (0.1 – 3.6 ka; Table 2); however, because DIYABC does not allow migration among populations in demographic scenarios, the divergence time between the populations is likely overestimated. In fact, STRUCTURE and BAYESASS analyses indicated that these populations are not completely isolated. STRUCTURE analyses showed admixture of two clusters in BR-d (Figure 4), whereas estimates of contemporary gene flow indicated active but unidirectional (upstream) gene flow between BR populations (Table 4). A possible explanation for the inconsistent results between STRUCTURE and BAYESASS analyses is that BR-u was recently colonized by a few individuals from BR-d, with all individuals in BR-u being comprised of partial BR-d genetic characters. In fact, population genetic indices showed that the BR-u individuals had consistently lower genetic diversity and Ne than the BR-d individuals (Tables 1 and 2) and genetic divergence between the populations was statistically significant, but low (Table 3). The climatic history of the late Holocene (last 4000 years) in southeast New Mexico shows sequences of drier and wetter climates (Polyak & Asmerom 2001). While climate fluctuation might have caused BR populations to differentiate, it is unclear how this occurs with such small-spatial-scale isolation.

Alternatively, because estimates from BAYESASS reflect migration rates over the last few generations and there is genetic divergence between BR-u and BR-d, a few individual of P. popeii might have recently colonized BR-u and are currently isolated by a low-water culvert crossing (built 1932-1936; Figure 1), which can prevent fish dispersal (Warren & Pardew 1998).

77 Future studies of host fish movements may identify the means by which fine-scale population structure is maintained in BR populations of P. popeii. Because the lower Pecos River connects BR and RG populations, gene flow could occur in contemporary time. However, genetic analyses and ENMs under current bioclimatic conditions support discontinuity between these populations, most likely because impoundments and occasional intermittent conditions prevent dispersal of host fishes. Red Bluff and Amistad reservoirs (impounded in 1936 and 1969, respectively; Figure 1) are major impoundments between BR and RG populations, and a segment of river with intermittent flow exists downstream of Red Bluff Reservoir. Historic surveys in the lower Pecos River periodically reported the presence of a single or a few individuals of P. popeii; however, the most recent survey did not (Karatayev et al. 2012). Intermittent conditions are common in many of these drainages due to seasonal drought and use of river water for irrigation; these alter both water quality and quantity (Hubbs et al. 1977; Davis 1980). Studies of benthic macroinvertebrates in the lower Pecos River showed that species richness was significantly lower than in upper reaches due to high salinity and extreme physiochemical fluctuations (Davis 1980). These conditions continue until below Sheffield, Texas, where numerous spring-fed creeks contribute groundwater that dilutes releases from naturally saline Red Bluff Reservoir (Davis 1980). Conservation implications Population genetic analyses and ENMs indicate that the BR and RG populations are currently disconnected. Because the BR populations have been isolated from RG populations since the mid-Pleistocene and contain low genetic diversity, the BR and RG populations likely comprise separate evolutionarily significant units that must be considered when developing conservation strategies and species recovery plans. Furthermore, although current ENMs for P. popeii indicate stable suitable environments in the lower Rio Grande, landscape features (such as topographic complexity) and anthropogenic land- and water-uses can influence potential dispersal and availability of high-quality habitat. Over the past several decades, at least 30 springs have gone dry in the Rio Grande drainages (Contreras-Balderas & Lozano-Vilano 1994). Rising human demand for water will likely exacerbate intermittent conditions. A long-term demographic study revealed that reduced river discharge is associated with significantly decreased survival of P. popeii (Inoue et al. 2014a). Significant changes in hydrological regimes will likely cause significant declines in P. popeii populations. If the region’s human population

78 continues to grow, intensifying anthropogenic threats, P. popeii has a low probability of persistence even in the suitable environments predicted from ENMs. Given a previous study that estimated 48,006 individuals inhabiting the Black River

(Inoue et al. 2014a), we can estimate the ratio of genetically effective population size (Ne) and census population size (Nc) in the BR populations. With an estimated Ne of 5870 individuals in the Black River from ABC analyses, the Ne/Nc ratio is 0.12. Such estimates of the Ne/Nc ratio of a target population are critical to understanding whether a population can persist and maintain -4 -3 adaptive genetic variance. In general, the Ne/Nc ratio can be as low as 10 to 10 and fluctuate enormously throughout time in fish and invertebrates that have very high fecundity and juvenile mortality (i.e., Type III survivorship curve; Luikart et al. 2010; Hare et al. 2011). Although freshwater mussels fit these criteria, promiscuous reproduction through factors such as multiple paternity (Christian et al. 2007) may maintain higher Ne/Nc ratios (Balloux & Lehmann 2003;

Snook et al. 2009). The Ne/Nc ratio can help unravel the relative importance of environmental, ecological, and genetic factors that drive population persistence (Palstra & Fraser 2012). Conclusion The Chihuahuan Desert is one of the world’s most biologically rich and diverse deserts (Olson & Dinerstein 1998); its ecosystems harbor high numbers of endemics that are often evolutionarily unique (Metcalf & Smartt 1997; Kodric-Brown & Brown 2007). Unfortunately, desert ecosystems are fragile and recover slowly from perturbations, so they are susceptible to human activities. Our study integrated evolutionary history, estimates of distribution of suitable environment over time, and genetic connectivity among extant populations, to elucidate factors affecting current population genetic structure in a desert river. We found that climate change in the Pleistocene strongly influenced current genetic structure of P. popeii in the Rio Grande drainage. Future changes in climate and habitat due to anthropogenic activities mean that P. popeii populations will continue to be isolated from one another. Resource managers must consider environmental and genetic features when developing species-recovery strategies.

ACKNOWLEDGEMENTS We thank L. Burlakova, A. Karatayev, T. Miller, Y. Zhang, and T. Nobles for help with sample collections. Field and laboratory assistance was provided by the New Mexico Department of Game and Fish, the US Fish and Wildlife Service, numerous undergraduate technicians in the

79 Berg Lab, and the Center for Bioinformatics and Functional Genomics at Miami University. We thank A. Walters for help with ecological niche modeling and L. Burlakova for providing us with collection records for P. popeii. We also thank private landowners for access to the river. Comments from C. Franz Berg and four anonymous reviewers greatly improved the manuscript. Funding was provided by the New Mexico Department of Game and Fish, and the US Fish and Wildlife Service.

REFERENCES Alvarado-Serrano DF, Knowles LL (2014) Ecological niche models in phylogeographic studies: applications, advances and precautions. Molecular Ecology Resources, 14, 233-248. Antao T, Lopes A, Lopes RJ, Beja-Pereira A, Luikart G (2008) LOSITAN: a workbench to detect molecular adaptation based on a Fst-outlier method. BMC Bioinformatics, 9, 323. Avise JC (2000) Phylogeography: The History and Formation of Species Harvard University Press, Cambridge, Massachusetts. Balloux F, Lehmann L (2003) Random mating with a finite number of matings. Genetics, 165, 2313-2315. Beaumont MA (2010) Approximate Bayesian computation in evolution and ecology. Annual Review of Ecology, Evolution, and Systematics, 41, 379-406. Beerli P (2006) Comparison of Bayesian and maximum-likelihood inference of population genetic parameters. Bioinformatics, 22, 341-345. Bequaert JC, Miller WB (1973) The Mollusks of the Arid Southwest: With an Arizona Check List University of Arizona Press, Tucson, Arizona. Bogan AE (1996) Popenaias popeii. In: IUCN 2013. IUCN Red List of Threatened Species. Bond N, Thomson J, Reich P, Stein J (2011) Using species distribution models to infer potential climate change-induced range shifts of freshwater fish in south-eastern Australia. Marine and Freshwater Research, 62, 1043-1061. Bossu CM, Beaulieu JM, Ceas PA, Near TJ (2013) Explicit tests of palaeodrainage connections of southeastern North America and the historical biogeography of Orangethroat Darters (Percidae: Etheostoma: Ceasia). Molecular Ecology, 22, 5397-5417. Brown JL (2014) SDMtoolbox: a python-based GIS toolkit for landscape genetic, biogeographic and species distribution model analyses. Methods in Ecology and Evolution, 5, 694-700.

80 Carman SM (2007) Texas hornshell Popenaias popeii recovery plan, p. 57. New Mexico Department of Game and Fish, Santa Fe, NM. Carstens BC, Richards CL (2007) Integrating coalescent and ecological niche modeling in comparative phylogeography. Evolution, 61, 1439-1454. Chagnon G, de la Cheneliere V (1998) Status assessment and conservation action plan for Lampsilis radiata. McGill University, Montreal, Quebec, Canada. Christian AD, Monroe EM, Asher AM, M. LJ, Berg DJ (2007) Methods of DNA extraction and PCR amplification for individual freshwater mussel (Bivalvia: Unionidae) glochidia, with the first report of multiple paternity in these organisms. Molecular Ecology Resources, 7, 570–573. Clement M, Posada D, Crandall KA (2000) TCS: a computer program to estimate gene genealogies. Molecular Ecology, 9, 1657-1659. Contreras-Balderas S, Lozano-Vilano ML (1994) Water, endangered fishes, and development perspectives in arid lands of Mexico. Conservation Biology, 8, 379-387. Cornuet JM, Pudlo P, Veyssier J, Dehne-Garcia A, Gautier M, Leblois R, Marin JM, Estoup A (2014) DIYABC v2.0: a software to make approximate Bayesian computation inferences about population history using single nucleotide polymorphism, DNA sequence and microsatellite data. Bioinformatics, 30, 1187-1189. Cornuet JM, Ravigne V, Estoup A (2010) Inference on population history and model checking using DNA sequence and microsatellite data with the software DIYABC (v1.0). BMC Bioinformatics, 11, 401. Coulon A, Fitzpatrick JM, Bowman R, Stith BM, Makarewich CA, Stenzler LM, Lovette IJ (2008) Congruent population structure inferred from dispersal behaviour and intensive genetic surveys of the threathened Florida scrub-jay (Aphelocoma coerulenscens). Molecular Ecology, 17, 1685-1701. Davis JR (1980) Species composition and diversity of benthic macroinvertebrate population of the Pecos River, Texas. Southwestern Naturalist, 25, 241-256. Earl DA, vonHoldt BM (2012) STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources, 4, 359-361.

81 Edwards RJ, Garrett GP, Marsh-Matthews E (2002) Conservation and status of the fish communities inhabiting the Rio Conchos basin and middle Rio Grande, Mexico and U.S.A. Reviews in Fish Biology and Fisheries, 12, 119-132. Elith J, Graham CH, Anderson RP, Dudik M, Ferrier S, Guisan A, Hijmans RJ, Huettmann F, Leathwick JR, Lehmann A, Li J, Lohmann LG, Loiselle BA, Manion G, Moritz C, Nakamura M, Nakazawa Y, Overton JM, Peterson AT, Phillips SJ, Richardson K, Scachetti-Pereira R, Schapire RE, Soberon J, Williams S, Wisz MS, Zimmermann NE (2006) Novel methods improve prediction of species' distributions from occurrence data. Ecography, 29, 129-151. 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. Fetzner JW, Jr., Crandall KA (2003) Linear habitats and the nested clade analysis: an empirical evaluation of geographic versus river distances using an Ozark crayfish (Decapoda: Cambaridae). Evolution, 57, 2101-2118. 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. Fordham DA, Brook BW, Moritz C, Nogues-Bravo D (2014) Better forecasts of range dynamics using genetic data. Trends in Ecology & Evolution, 29, 436-443. Galloway WE, Whiteaker TL, Ganey-Curry P (2011) History of Cenozoic North American drainage basin evolution, sediment yield, and accumulation in the Gulf of Mexico basin. Geosphere, 7, 938-973. Goudet J (1995) FSTAT (version 1.2): a computer program to calculate F-statistics. Journal of Heredity, 86, 485-486. Hare MP, Nunney L, Schwartz MK, Ruzzante DE, Burford M, Waples RS, Ruegg K, Palstra F (2011) Understanding and estimating effective population size for practical application in marine species management. Conservation Biology, 25, 438–449. Hewitt G (2000) The genetic legacy of the Quanternary ice ages. Nature, 405, 907-913. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 1965- 1978.

82 Hocutt CH, Denoncourt RF, Stauffer JR JR (1978) Fishes of the Greenbrier River, West Virginia, with drainage history of the central Appalachians. Journal of Biogeography, 5, 59-80. Holleley CE, Geerts PG (2009) Multiplex Manger 1.0: a cross-platform computer program that plans and optimizes multiplex PCR. Biotechniques, 46, 511-517. Hubbs C, Miller RR, Edwards RJ, Thompson KW, Marsh E, Garrett GP, Powell GL, Norris DJ, Zerr RW (1977) Fishes inhabiting the Rio Grande, Texas and Mexico, between El Paso and the Pecos confluence. In: Importance, Preservation and Management of Riparian Habitat: A Symposium eds. Johnson RR, Jones DA). USDA Forest Service, Fort Collins, Colorado. Inoue K, Lang BK, Berg DJ (2013) Development and characterization of 20 polymorphic microsatellite markers for the Texas hornshell, Popenaias popeii (Bivalvia: Unionidae), through next-generation sequencing. Conservation Genetics Resources, 5, 195-198. Inoue K, Levine TD, Lang BK, Berg DJ (2014a) Long-term mark-and-recapture study of a freshwater mussel reveals patterns of habitat use and an association between survival and river discharge. Freshwater Biology, 59, 1872-1883. Inoue K, Monroe EM, Elderkin CL, Berg DJ (2014b) Phylogeographic and population genetic analyses reveal Pleistocene isolation followed by high gene flow in a wide-ranging, but endangered, freshwater mussel. Heredity, 112, 282-290. Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics, 23, 1801-1806. Jezkova T, Jaeger JR, Marshall ZL, Riddle BR (2009) Pleistocene impacts on the phylogeography of the desesrt pocket mouse (Chaetodipus penicillatus). Journal of Mammalogy, 90, 306-320. Kalinowski ST (2011) The computer program STRUCTURE does not reliably identify the main genetic clusters within species: simulations and implications for human population structure. Heredity, 106, 625-632. Karatayev AY, Miller TD, Burlakova LE (2012) Long-term changes in unionid assemblages in the Rio Grande, one of the world's top 10 rivers at risk. Aquatic Conservation: Marine and Freshwater Ecosystems, 22, 206-219.

83 Kim D, Conway KW (2014) Phylogeography of Rhinichthys cataractae (Teleostei: Cyprinidae): pre-glacial colonization across the Continental Divide and Pleistocene diversification within the Rio Grande drainage. Biological Journal of the Linnean Society, 111, 317-333. Knowles LL (2009) Statistical phylogeography. Annual Review of Ecology, Evolution and Systematics, 40, 593-612. Kodric-Brown A, Brown JH (2007) Native fishes, exotic mammals, and the conservation of desert springs. Frontiers in Ecology and the Environment, 5, 549-553. Kramer-Schadt S, Niedballa J, Pilgrim JD, Schröder B, Lindenborn J, Reinfelder V, Stillfried M, Heckmann I, Scharf AK, Augeri DM, Cheyne SM, Hearn AJ, Ross J, Macdonald DW, Mathai J, Eaton J, Marshall AJ, Semiadi G, Rustam R, Bernard H, Alfred R, Samejima H, Duckworth JW, Breitenmoser-Wuersten C, Belant JL, Hofer H, Wilting A, Robertson M (2013) The importance of correcting for sampling bias in MaxEnt species distribution models. Diversity and Distributions, 19, 1366-1379. Kuhner MK (2006) LAMARC 2.0: maximum likelihood and Bayesian estimation of population parameters. Bioinformatics, 22, 768-770. Lang BK (2001) Status of the Texas hornshell and native freshwater mussels (Unionoidea) in the Rio Grande and Pecos River of New Mexico and Texas. New Mexico Department of Game and Fish, Santa Fe, New Mexico. Levine TD, Lang BK, Berg DJ (2012) Physiological and ecological hosts of Popenaias popeii (Bivalvia: Unionidae): laboratory studies identify more hosts than field studies. Freshwater Biology, 57, 1854-1864. Librado P, Rozas J (2009) DnaSP v5: a software for comprehensive analysis of DNA polymorphism data. Bioinformatics, 25, 1451-1452. Luikart G, Ryman N, Tallmon DA, Schwartz MK, Allendorf FW (2010) Estimation of census and effective population sizes: the increasing usefulness of DNA-based approaches. Conservation Genetics, 11, 355-373. Lydeard C, Cowie RH, Ponder WF, Bogan AE, Bouchet P, Clark SA, Cummings KS, Frest TJ, Gargominy O, Herbert DG, Hershler R, Perez KE, Roth B, Seddon M, Strog EE, Thompson FG (2004) The global decline of nonmarine mollusks. Bioscience, 54, 321- 330.

84 Mäkinen HS, Merilä J (2008) Mitochondrial DNA phylogeography of the three-spined stickleback (Gasterosteus aculeatus) in Europe - evidence for multiple glacial refugia. Molecular Phylogenetics and Evolution, 46, 167-182. Mantooth SJ, Hafner DJ, Bryson RW, Jr., Riddle BR (2013) Phylogeographic diversification of antelope squirrels (Ammospermophilus) across North American deserts. Biological Journal of the Linnean Society, 109, 949-967. Mayden RL (1988) Vicariance biogeography, parsimony, and evolution in North America freshwater fishes. Systematic Zoology, 37, 329-355. Metcalf AL (1997) Land snail of New Mexico from an historical zoogeographic point of view. New mexico Museum of Natural History and Science, Bulletin, 10, 71-108. Metcalf AL, Smartt RA (1997) Land snails of New Mexico. New mexico Museum of Natural History and Science, Bulletin, 10, 1-145. Metcalfe SE, O'Hara SL, Caballero M, Davies SJ (2000) Records of Late Pleistocene–Holocene climatic change in Mexico — a review. Quanternary Science Reviews, 19, 699-721. Miller RR (1977) Composition and derivation of the native fish fauna of the Chihuahuan Desert region. In: Transactions of the Symposium on the Biological Resources of the Chihuahuan Desert Region, United States and Mexico (eds. Wauer RH, Riskind DH), pp. 365-381. Sul Ross State University, Alpine, Texas. Nielsen R, Beaumont MA (2009) Statistical inferences in phylogeography. Molecular Ecology, 18, 1034-1047. Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O'Hara RB, Simpson GL, Solymos P, Stevens MHH, Wagner H (2015) Vegan: community ecology package, R package version 2.2-1, available at http://cran.r-project.org/web/packages/vegan/. Olson DM, Dinerstein E (1998) The Global 200: a representation approach to conserving the Earth's most biologically valuable ecoregions. Conservation Biology, 12, 502-515. Otto-Bliesner BL, Marshall SJ, Overpeck JT, Miller GH, Hu A (2006) Simulating Arctic climate warmth and icefield retreat in the last interglaciation. Science, 311, 1751-1753. Palstra FP, Fraser DJ (2012) Effective/census population size ratio estimation: a compendium and appraisal. Ecology and Evolution, 2, 2357-2365. Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6, 288-295.

85 Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190, 231-259. Phillips SJ, Dudik M, Elith J, Graham CH, Lehmann A, Leathwick JR, Ferrier S (2009) Sample selection bias and presence-only distribution models: implications for background and pseudo-absence data. Ecological Applications, 19, 181-197. Polyak VJ, Asmerom Y (2001) Late Holocene climate and cultural changes in the southwestern United States. Science, 294, 148-151. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genetype data. Genetics, 155, 945-959. R Development Core Team (2011) R: A language and environment for statistical computing. R Fundation for Statistical Computing, Vienna, Austria. Rosenberg NA (2004) DISTRUCT: a program for the graphical display of population structure. Molecular Ecology Notes, 4, 137-138. Rousset F (2008) GENEPOP'007: a complete re-implementation of the GENEPOP software for Windows and Linux. Molecular Ecology Resources, 8, 103-106. Schmandt J (2002) Bi-national water issues in the Rio Grande/Río Bravo basin. Water Policy, 4, 137-155. Schwalb AN, Morris TJ, Mandrak NE, Cottenie K (2013) Distribution of unionid freshwater mussels depends on the distribution of host fishes on a regional scale. Diversity and Distributions, 19, 446-454. Seidel RA, Lang BK, Berg DJ (2009) Phylogeographic analysis reveals multiple cryptic species of amphipods (Crustacea: Amphipoda) in Chihuahuan Desert springs. Biological Conservation, 142, 2303-2313. Smith ML, Miller RR (1986) The evolution of the Rio Grande basin as inferred from its fish fauna. In: The Zoogeography of North American Freshwater Fishes (eds. Hocutt CH, Wiley EO), pp. 457-485. John Wiley & Sons, New York. Snook RR, Brustle L, Slate J (2009) A test and review of the role of effective population size on experimental sexual selection patterns. Evolution, 63, 1923-1933. Strange RM, Burr BM (1997) Intraspecific phylogeography of North American highland fishes: A test of the Pleistocene vicariance hypothesis. Evolution, 51, 885-897.

86 Strenth NE, Howells RG, Correa-Sandoval A (2004) New records of the Texas hornshell Popenaias popeii (Bivalvia: Unionidae) from Texas and northern Mexico. The Texas Journal of Science, 56, 223-230. Thomas RG (1972) The Geomorphic Evolution of the Pecos River System The Baylor University Press, Waco, TX. Untergasser A, Cutcutache I, Koressaar T, Ye J, Faircloth BC, Remm M, Rozen SG (2012) Primer3—new capabilities and interfaces. Nucleic Acids Research, 40, e115. USFWS (US Fish and Wildlife Service). 2013. Endangered and threatened wildlife and plants; review of native species that are candidates for listing as endangered or threatened; annual notice of findings on resubmitted petitions; annual description of progress on listing actions. Federal Register 78, 70104-70162. van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) MICRO-CHECKER: Software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes, 4, 535-538. Warren ML, Pardew MG (1998) Road crossings as barriers to small-stream fish movement. Transactions of the American Fisheries Society, 127, 637-644. Williams JD, Warren ML, Jr., Cummings KS, Harris JL, Neves RJ (1993) Conservation status of freshwater mussels of the United States and Canada. Fisheries, 18, 6-22. Wilson GA, Rannala B (2003) Bayesian inference of recent migration rates using multilocus genotypes. Genetics, 163, 1177-1191. Wong CM, Williams CE, Pittock J, Collier U, Schelle P (2007) World's top 10 rivers at risk. WWF International, Gland, Swizerland.

87 TABLES AND FIGURES Table 1. Descriptive statistics for COX sequences and 18 microsatellite loci for Popenaias popeii populations in the Black River, New Mexico and the Devils River and Rio Grande, Texas. COX Microsatellites River Site ID n H HR K π NA AR NP HO HE Black CC 3 1 -- 0 0 2.0 -- 0 0.380 0.314 Fall 24 1 1.0 0 0 3.3 2.8 0 0.469 0.483 Magby 19 1 1.0 0 0 3.6 2.9 1 0.469 0.477 DV 55 1 1.0 0 0 4.5 3.1 2 0.503 0.507 RF5 20 1 1.0 0 0 3.9 3.1 0 0.527 0.518 RF1 41 1 1.0 0 0 4.6 3.1 4 0.507 0.532 BS 10 1 1.0 0 0 3.4 3.0 0 0.450 0.480 RF3 21 1 1.0 0 0 3.9 3.1 0 0.513 0.518 All BR 193 1 1.0 0 0 6.0 3.0 8 0.495 0.528 Devils DR 3 1 -- 0 0 3.4 -- 2 0.556 0.608 Rio Grande RG1 29 21 5.9 3.56 0.0050 14.2 7.7 40 0.813 0.881 RG2 5 4 -- 2.00 0.0028 6.2 -- 5 0.800 0.793 RG3 2 2 -- 5.00 0.0070 3.4 -- 0 0.800 0.693 RG4 16 13 5.6 3.39 0.0047 11.4 7.6 11 0.805 0.872 RG5 6 5 5.0 4.60 0.0064 6.7 6.7 3 0.731 0.798 All RG 58 33 5.5 3.49 0.0049 16.7 7.3 167 0.806 0.902 Overall 254 34 2.2 1.41 0.0020 17.3 4.5 68 0.566 0.694 AR, rarefied allelic richness; COX, ~810 basepairs of part of the cytochrome oxidase II gene and the cytochrome oxidase I gene; H, number of haplotypes; HE, mean expected heterozygosity; HO, mean observed heterozygosity; HR, rarefied number of haplotypes; K, mean number of base pair differences between all pairs of individuals; n, number of mussels sampled; NA, mean number of observed alleles; NP, number of private alleles; π, nucleotide diversity

88 Table 2. Prior distributions of parameters for each scenario (Figure 2) and posterior parameter values for Scenario 3. The unit of time is generations, except that values in parentheses are 1000 years before present (ka) calculated with an average generation time of 8.9 years (based on 8.1 - 9.6 years for Lampsilis radiata; Chagnon & de la Cheneliere 1998). Prior distribution Posterior parameter estimates Population Parameter Distribution Median 95% credible interval RMAE Black River upstream population (BR-u) NBR-u ~U(10,5000) 1190 194 4230 0.212 Black River downstream population (BR-d) NBR-d ~U(10,50000 4680 3460 4980 0.222 Rio Grade population (RG) NRG ~U(10,100000) 70300 41600 91900 0.200 Ancestral Black River population NBR ~U(10,50000) 39400 5800 87200 0.422 Ancestral Rio Grande population NA-RG ~U(10,100000) ------Founder population into the Black River NF-BR ~U(10,50000) ------Bottlenecked population in the Black River NB-BR ~U(10,5000) 1630 344 4490 0.254 Bottlenecked population in the Rio Grande NB-RG ~U(10,100000) ------Most ancestral population NA ~LU(10,100000) 81000 45900 99100 0.170 Time of divergence within the Black River t1 ~U(1,500) 125 12 410 0.298 (1.113 ka) (0.107 ka) (3.649 ka) Time of expansion/bottleneck event te1 ~U(1,20000) ------Time of expansion/bottleneck event te2 ~U(1,20000) 3670 614 11000 0.316 (32.663 ka) (5.465 ka) (97.900 ka) Time of the ancestral divergence t2 ~U(1,20000) 9020 2370 18900 0.185 (80.270 ka) (21.093 ka) (168.210 ka) Mean mutation rate per generation µ ~U(1×10-5,1×10-3) 1.53×10-4 7.48×10-5 3.39×10-4 0.214 LU = log-uniform distribution; RMAE = relative median of absolute error; U = uniform distribution

89 Table 3. Pairwise FST (below diagonal) and DEST (above diagonal) for 18 microsatellite loci from

STRUCTURE-defined populations of Popenaias popeii. All values are statistically significantly different from 0 at P<0.05. BR-u BR-d RG BR -u -- 0.015 0.751 BR-d 0.033 -- 0.715 RG 0.246 0.255 --

90 Table 4. Asymmetric migration rates between pairs of STRUCTURE-defined populations estimated from (A) MIGRATE-N and (B) BAYESASS. In MIGRATE-N, values in italics are θ; other entries are migration rates (M) from column populations into row populations. In BAYESASS, all values refer to contemporary migration rates (mc). Values in parentheses are 95% confidence interval (or credible interval for BAYESASS) of estimates. BR-u BR-d RG (A) MIGRATE-N (θ in italics, M for other entries) BR-u 0.928 (0.677-0.996) 5.424 (5.036-5.833) 6.755 (6.322-7.218) BR-d 6.580 (6.135-7.043) 0.988 (0.919-1.064) 7.818 (7.338-8.321) RG 0.820 (0.767-0.875) 1.049 (0.990-1.111) 35.120 (32.062-38.605) (B) BAYESASS (mc) BR-u -- 0.320 (0.303-0.338) 0.006 (-0.006-0.018) BR-d 0.002 (-0.002-0.007) -- 0.002 (-0.002-0.007) RG 0.008 (-0.007-0.024) 0.005 (-0.005-0.015) -- θ = mutation-scaled effective population size for each population; M = m/µ, where m = migration rate and µ = mutation rate; mc = contemporary migration rate

91 Table 5 Type I and Type II error rates and Bayesian posterior probabilities for each scenario estimated using DIYABC. True scenario used Type II error rate Type I Posterior probability for simulation Scenario 1 Scenario 2 Scenario 3 Scenario 4 error rate (95% credible interval) Scenario 1 -- 0.310 0.084 0.048 0.442 0.045 (0.036-0.055) Scenario 2 0.328 -- 0.052 0.014 0.394 0.007 (0.002-0.012) Scenario 3 0.058 0.046 -- 0.400 0.504 0.740 (0.719-0.761) Scenario 4 0.008 0.016 0.310 -- 0.334 0.208 (0.189-0.227)

Figure 1. Map of the Rio Grande drainages in the southwest USA indicating major tributaries, reservoirs, and Popenaias popeii sampling sites (circles). Colors correspond to the parsimony network of COX sequences in Figure 3. The magnified inset map shows Black River sampling locations.

93 Scenario 1 Scenario 2 Scenario 3 Scenario 4

NA NA NA NA

NF-BR NBR NBR NA-RG N BR te2 NB-RG NB-BR NBR NB-BR te1

N N N N N N N N N NBRu NBRd NRG BR-u BR-d RG BR-u BR-d RG BRu BRd RG t 0 Figure 2. Four demographic scenarios tested using DIYABC: 1) all populations have constant size over time; 2) a small number of individuals founded the ancestral Black River population followed by population expansion at te2; 3) a large number of founders colonized the ancestral

Black River population followed by a population bottleneck at te2; and 4) Rio Grande populations experienced population bottleneck at te2 followed by population expansion at te1, in addition to

Scenario 3. All scenarios assume divergence between the ancestral BR and RG populations at t2, divergence between BR-u and BR-d at t1, and STRUCTURE-defined populations at the present time

(t0). Timing of events is shown at right. See Table 2 for detailed labels.

94 2

1 1

2 1 1 2 3 1 1 1 1 1 1 2 1 1 3 4 1

3 3 2 1 1 1 1 1 1 1 3 2 1

Figure 3. Parsimony network of COX sequences for Popenaias popeii. Each circle represents a unique haplotype; lines between haplotypes represent single-base-pair changes; black dots are inferred extinct or unsampled haplotypes. Haplotype frequency is relative to the size and number in the circle. The most common haplotype is shared between BR (n = 185) and RG (n = 10) individuals. Circle colors represent localities (Black River = dark gray; Devils River = light gray; Rio Grande = white).

Figure 5. Potential distribution of Popenaias popeii identified using ecological niche modeling under current bioclimatic conditions (1950 to 2000) and projections back to two paleoclimatic periods (last interglacial, 120 – 140 ka; last glacial maximum, 21 ka). Models included major rivers of the Rio Grande watershed in the USA and Mexico. Black dots on the Present map represent occurrence points included in the ENMs. Scale bars in the bottom-left corners represent 200 km.

97 SUPPLEMENTARY INFORMATION Table S1. Estimates of relative contributions (%) of bioclimatic variables ranked according to the

MAXENT models for Popenaias popeii and its host fish (Carpiodes carpio, Cyprinella lutrensis, and Moxostoma congestum) populations. AUCtest is the area under the curve in the receiver operating characteristic curve from a test dataset. Popenaias Carpiodes Cyprinella Moxostoma Variables Description popeii carpio lutrensis congestum Bio02 Mean diurnal temperature range 24.2 3.2 1.7 14.4 Bio03 Isothermality 4.2 0.6 0.3 12.4 Bio04 Temperature seasonality 1.4 10.3 14.3 9.2 Bio05 Maximum temperature of warmest month 41.0 61.0 64.4 37.2 Bio09 Mean temperature of driest quarter 0.0 0.1 0.1 1.8 Bio15 Precipitation seasonality 15.4 2.2 12.7 22.7 Bio18 Precipitation of warmest quarter 13.9 14.9 5.7 1.2 Bio19 Precipitation of coldest quarter 0.0 7.7 0.5 1.1

AUCtest 0.901 0.769 0.760 0.833

Figure S1. Relationships between pairwise genetic differences (ST for mtDNA sequences and

DEST for microsatellites) and geographic (river) distances among localities. Overall relationships are shown in A and D. Overall relationships were decomposed into within river (B and E) and among rivers (C and F). Colors show relationships within the Black River localities (black dots), within the Rio Grande localities (white dots), and among the Black River and Rio Grande localities (grey dots).

Figure S2. (A) Log-likelihood [Ln P(k)] and (B) ∆k for each k over the 10 replicates in

STRUCTURE. Error bars are standard deviation.

100

Figure S3. Potential distribution of three primary host fish species (Carpiodes carpio, Cyprinella lutrensis, and Moxostoma congestum) identified using ecological niche modeling under present climatic conditions (1950 to 2000) and projections of two paleoclimatic (last interglacial, 120 – 140 ka; last glacial maximum, 21 ka). Models included major rivers of the Rio Grande watershed in the United States and Mexico. Black dots on the Present map represent occurrence points included in the ENMs. Scale bars on the left-bottom corner showed 200 km.

101 Chapter 5: Long-term mark-and-recapture study of a freshwater mussel reveals patterns of habitat use and an association between survival and river discharge

Inoue K, Levine TD, Lang BK, Berg DJ (2014) Long-term mark-and-recapture study of a freshwater mussel reveals patterns of habitat use and an association between survival and river discharge. Freshwater Biology, 59, 1872-1883.

ABSTRACT Climate change and human population growth threaten the supply of fresh water for human use and freshwater biodiversity. Long-term studies are necessary to identify the effects of such temporal trends on biological and ecological phenomena; however, the collection of long- term data can be costly and time-consuming. We investigated the effect of hydrological variation over time on population dynamics in a perennial river of the northern Chihuahuan Desert, using an imperiled freshwater mussel (Popenaias popeii) as a model. We conducted a 15-year mark- and-recapture study, and distance sampling, to estimate demographic parameters while accounting for habitat heterogeneity and changes in river discharge. Recapture probability varied between microhabitats, and survival was positively correlated with river discharge. Survival and the finite rate of population growth were relatively stable over time. Over 60% of individuals were found at relatively high density in riffle habitats, which compose approximately16% of the total study area. Mean monthly temperature in the region increased over the past 100 years, and mean monthly discharge of the Black River declined over the past 65 years. With no significant trends in total monthly precipitation, declines in discharge suggest that reduction of streamflow is likely due to lowering of the water table and decreased groundwater recharge. Significant changes in climate and hydrological regimes, and increases in anthropogenic threats (increased water demand, degraded water quality) in the region may induce significant declines in population size of this imperiled mussel. We demonstrated the importance of considering habitat heterogeneity and hydrological cycles over time to examine population dynamics. Survival of benthic invertebrates in desert streams is sensitive to hydrological cycles, which are expected to be altered via climate change and extensive water use. Species recovery plans need to incorporate knowledge of spatial distributions when designing strategies for habitat assessment and making conservation decisions.

102 Keywords: endangered species, long-term data, Popenaias popeii, population dynamics, total population size

INTRODUCTION Global climate change, primarily due to anthropogenic activities, threatens both the supply of fresh water for human use and freshwater biodiversity by altering means and extremes of precipitation and evapotranspiration, rates of river discharge and thermal regimes (Vörösmarty et al. 2010; Malmqvist & Rundle 2002). In arid regions such as the southwest U.S.A, streams and springs have experienced significantly altered hydrological cycles over the last half century (Barnett et al. 2008). Additionally, growing human populations have significantly increased demands for water used for irrigation, domestic needs and industry in this region. Both the magnitude and duration of droughts and anthropogenic demands are predicted to increase in the future, as the intensity of anthropogenic activity remains high (MacDonald 2010; Cayan et al. 2010). Hydrological cycles in arid regions often include heavy storms that can dramatically change flow in streams and springs over short time-periods (Poff et al. 1997). Because many of these ecosystems are isolated from one-another by long distances and extremely dry environments, they harbor endemic species that have evolved in particularly harsh environments (Seidel et al. 2009; Witt et al. 2006). These species often have narrow environmental tolerances (Seidel et al. 2010) and are confronted with a high rate of imperilment related to habitat modification and surface/ground water withdrawal (Deacon et al. 2007; Bogan & Lytle 2011). Hydrological alterations associated with climate change and anthropogenic activities are predicted to have very large impacts on such ecosystems. Increasingly, species in these ecosystems are targeted for management by agencies charged with conservation of imperiled taxa (USFWS 2013). We used an imperiled freshwater mussel, Popenaias popeii (Bivalvia: Unionidae), as a model species to investigate the effects of hydrological variation over time on demography and population dynamics in a desert stream. Popenaias popeii is ideal for this task because it has a long life span (over 20 years; Carman 2007) and adults are relatively immobile, making them especially useful for long-term, mark-and-recapture studies. Popenaias popeii is a candidate for listing under the U.S. Endangered Species Act; it is considered “critically endangered” by the

103 International Union for the Conservation of Nature (Bogan 1996), with listing criteria of “a decline in area of occupancy, extent of occurrence, and/or quality of habitat.” It is currently restricted to a tributary of the Pecos River (New Mexico), and the lower Rio Grande and Devils River (Texas), U.S.A. (Howells et al. 1996). Although there are historic records of occurrence in Mexico (the Rio Salado and its tributaries), current occurrence of P. popeii in Mexican drainages is uncertain (Strenth et al. 2004; Smith et al. 2003). In 1997-1998, initial surveys were conducted to determine habitat occupancy and distribution of P. popeii in the Black River, Eddy County, New Mexico, a tributary of the Pecos River. Based on an initial 48-km survey, P. popeii was found to occur in an ~14-km stretch of the river between low-head dams (Figure 1; Lang 2001). Because the species does not occur outside of this stretch and the nearest populations in the Rio Grande and Devils River are hundreds of kilometers away, the Black River population is effectively closed. Individuals occupy distinctive microhabitats such as undercut riverbanks, crevices, shelves and the bases of large boulders with small-grained substrata, where shallow riffles flow over conglomerate bedrock. These microhabitats are believed to serve as flow refuges (sensu Strayer 1999) during large floods associated with annual precipitation events. In other extant populations in the Rio Grande, mussels are found in similar habitats, such as under large boulders or beneath limestone ledges where clay seems to provide a stable substratum (Karatayev et al. 2012). Understanding the demography and population dynamics of an endangered taxon is critical for developing effective plans for species conservation. Population surveys that do not track individuals may be informative to assess current age structure or abundance. However, better understanding of biological and ecological phenomena can be obtained by long-term monitoring programs that follow individuals, such as mark-and-recapture sampling, even when target organisms are sessile (Alexander et al. 1997). Long-term studies are able to identify the effects of less-frequent events (i.e. catastrophic events) and temporal trends on demographic parameters, including vital rates of populations. However, collection of long-term data is often limited by funding constraints, inconsistency of methods throughout the research period and changes in research direction over time (reviewed in Jackson & Füreder 2006). Nevertheless, such studies offer improved estimates of demographic parameters by accounting for changes in abundance as well as environmental and individual-animal covariates (Pollock 2002; Lindberg 2012). Moreover, because animal survey methods seldom detect all animals present in a surveyed

104 area, estimates of demographic parameters need to be corrected to account for the influence of detectability of the animal (Royle & Nichols 2003). This incomplete detectability can be corrected via estimates of detection probability, such as recapture probability from mark-and- recapture methods. Methods lacking such considerations may lead to underestimates of population size or insufficient descriptions of demographic parameters (Lindberg 2012). This is especially true with organisms that occur in heterogeneous environments and are sensitive to changes in the environment. We sought to (1) investigate differences in demographic parameters in various flow- refuge microhabitats (i.e. undercut riverbanks and the bases of large boulders in river channels; hereafter microhabitats), (2) evaluate effects of hydrological cycles on demographic parameters and (3) estimate total abundance of P. popeii in the Black River, taking into account heterogeneous micro- and macro-habitats (i.e. pool/run and riffle habitats; hereafter macrohabitats). We investigated heterogeneity in recapture probabilities and demographic parameters in two microhabitats, while also assessing responses to hydrological events over a 15- year period. We hypothesized that mussel detection would vary among microhabitats, with recapture probability predicted to be lower in undercut riverbanks than in river channels due to difficulty of sampling mussels in the former. Because P. popeii primarily inhabits flow refuges, we hypothesized that mussel survival would be affected more by maximum river discharge than by minimum river discharge. We estimated density and total population size of P. popeii in the Black River using adaptive distance-sampling methods (Strayer & Smith 2003). Because estimating demographic parameters and total population size are greatly influenced by detection efficiency of individuals, we estimated recapture probabilities using a “robust” mark-and- recapture design (Pollock 1982) to improve accuracy. Finally, we considered the implications of our results for the future survival of P. popeii and other aquatic species of conservation concern in desert streams given that river discharge is sensitive to climate change and anthropogenic water use.

METHODS Our study was conducted in the Black River in the northern Chihuahuan Desert (Figure 1). The river originates from several groundwater-fed springs in the foothills of the Guadalupe Mountains, and flows intermittently for 48-km before draining into the Pecos River upstream of

105 Malaga, Eddy County. Downstream of Blue Spring (the largest of these springs), the Black River becomes perennial until its confluence with the Pecos River. Median streamflow is low (0.16- 0.27 m3 s-1 in summer, 0.18-0.34 m3 s-1 in winter). The upstream-limit of perennial flow has fluctuated throughout the past century (Bjorklund & Motts 1959), likely due to groundwater pumping for agricultural and domestic uses. Additionally, a reduction in groundwater quality has been reported in the vicinity of the Black River due to increasing oil and gas development (Richard 1988; Richard & Boehm 1989). Nevertheless, the Black River sustains the only population of P. popeii in New Mexico, and also harbors a regionally rich ichthyofauna (Cowley & Sublette 1987; Sublette et al. 1990), including several state endangered/threatened species that serve as hosts for the parasitic glochidia larvae of P. popeii (Levine et al. 2012). All field sampling was conducted by tactile search with snorkels, SCUBA or surface air supplies. Demographic parameter estimates In 1997-1998, a mark-and-recapture study of river-channel habitats was initiated at three sites [Life History (LH) sites 1, 2 and 3], where large aggregations of P. popeii (>1 mussel m-2) occurred. Large floods in 2000 destroyed LH3, and thus, it was excluded from analyses. A fourth study site (LH4) upstream of the others was established in 2003. We sampled each LH site annually in the autumn (September or October) from 1997 to 2012, except in 1999 and 2001 because of a gap in funding and for logistical reasons. Beginning in July 2005, additional surveys at riverbank habitats were initiated at all LH sites. From 2005-2008, we carried out five short- interval samplings (1-2 week intervals; May 2005, May and July 2006, May 2007 and May 2008) which allowed us to assume that populations were closed (i.e. no migration, recruitment or death, and thus we could estimate recapture probability), meeting the assumptions of a robust design (Pollock 1982). We searched all LH sites thoroughly for mussels; captured individuals were marked with oval, 4 x 10 mm Floy laminated flex tags with unique numbers embedded in superglue gel along the valve hinge posterior to the umbo. Shell measurements (length, width and height) were recorded (± 0.1 mm) for all mussels (both new captures and recaptures), prior to replacing mussels in natural orientation at their points of capture. We implemented Pollock’s robust design in MARK (White & Burnham 1999) to estimate survival and recapture probabilities for three LH sites (LH1, LH2 and LH4) and two microhabitats within LH sites (river channels and riverbanks) during 2006 to 2008. We excluded 2005 data because we only sampled river-channel habitat at monthly sampling intervals. We

106 developed 432 models to evaluate whether or not survival and recapture probabilities varied among sites and microhabitats. We used the bias-corrected Akaike Information Criterion (AICc) to identify the most parsimonious model (Anderson & Burnham 1999). We compared parameter estimates [e.g. survival (), encounter probability (p), recapture probability (c) and abundance (N)] among models accounting for no-variance (.); variance of time (t), site (s) or microhabitat (h) and combinations of these. We also compared temporal migration estimates between open- population-surveys [e.g. the probability of emigration from populations (γ') and the probability of immigration to the populations (γ")] accounting for no temporal migration (i.e. γ' = 1, γ'' = 0) and no-variance of time, site or microhabitat. To ensure that patterns in the dataset were adequately represented, we estimated the variance inflation factor (ĉ) for the robust design model via goodness-of-fit tests. Because MARK is unable to conduct goodness-of-fit tests for the robust design, we fit the most general model to the dataset using RDSURVIV (Kendall et al. 1997); a cell-pooling algorithm was used to compute Pearson's χ2 test statistic, and we calculated the ĉ by hand. If a high ĉ was detected (indicating overdispersion), we converted AICc to quasi-AICc (QAICc) using ĉ in MARK. We calculated AICc for each candidate model and calculated ΔAICc (the difference in AICc between the candidate and the most parsimonious model) or ΔQAICc (if a correction was required). Models with ΔAICc or ΔQAICc < 2 were considered similarly parsimonious (Burnham & Anderson 2002). We implemented Cormack-Jolly-Seber (CJS) models (Lindberg 2012) for open- population capture-recapture in MARK to estimate survival for the three LH sites and two microhabitats from 1997 to 2012. These accounted for variation among sites, microhabitats and hydrological variables. The hydrological variables we used were minimum, maximum and mean monthly discharge during each interval between surveys; discharge data were obtained from a stream gauge just downstream of the study site (USGS gauge number 08405500). In addition, we determined 10th, 25th, 75th and 90th percentile daily discharge for the entire period of record (beginning 1 January 1947), and counted the number of days that fell below the 10th and 25th percentiles and above the 75th and 90th percentiles for each interval between surveys. Because we have spring samples (i.e. sampling in May) for 2005-2008, we also estimated seasonal survival (i.e. spring vs. autumn). We used recapture probabilities estimated from the robust design to condition the parameters. We developed 32 models and determined the most parsimonious

107 models for both annual and seasonal datasets using AICc (or QAICc). For CJS models, ĉ was estimated using bootstrapping methods in MARK and then used to correct AICc as necessary. We used the Pradel (1996) model from MARK to estimate finite rates of population growth (λ) over the 15-year study period. However, we excluded survey data in 2000 from the analysis due to low encounter rates; preliminary analysis showed unusually large fluctuations of λ when including these data. We only estimated annual λ because P. popeii is a short-term brooder, where spawning occurs from late April to mid-June (Smith et al. 2003). We used the most parsimonious CJS model as a base model, and used annual survival estimated from the CJS model and recapture probabilities estimated from the robust design to condition the parameters. Because each λ carries previous λ in a multiplicative fashion (Morris & Doak 2002), we calculated the geometric mean to estimate λ over the study period. Total population size estimate In 2011-2012, we conducted distance sampling to estimate population density and total population size of P. popeii within this 14-km stretch of the Black River. We used an adaptive distance-sampling method, implemented by traversing systematically located transects perpendicular to the riverbank and recording perpendicular distances between the transect line and any detected mussels (Strayer & Smith 2003). Initial surveys indicated that mussels tended to aggregate in flow refuge in shallow and narrow riffles (Lang 2001). Thus, we used satellite orthoimages verified by groundtruthing to identify two macrohabitat types in the river: shallow riffles, and deep pools and runs (hereafter riffle and pool habitats). We estimated total stream area of the 14-km stretch, areas of each macrohabitat, and areas between transect sites within macrohabitats using ArcGIS v10.1 (ESRI Inc., Redlands, CA, U.S.A.). Transect sites were systematically located in each of the two macrohabitat types, at 500 m intervals for pool habitats and 250 m intervals for riffle habitats. At each site, we located three line-transects at 10 m intervals. At each line-transect, we surveyed mussels one meter upstream and one meter downstream along the transect to estimate probability of mussel detection and density. We surveyed 27 transect sites in riffle habitats (total surveyed area = 1302.7 m2) and 21 transect sites in pool habitats (total surveyed area = 2484.0 m2; pools were wider than riffles and therefore line-transects were longer); thus, a total of 144 line-transects were surveyed. We recorded number of mussels found, distance from right-riverbank (facing downstream) on the line-transect and perpendicular distance between the line-transect and a detected mussel.

108 We used DISTANCE v6.0 (Thomas et al. 2010) to identify a detection function based on perpendicular distances from the transects. We fit our data to the most parsimonious model of the detection function of mussel distribution using AICc. Because the best model was a uniform distribution (see Results) and detection of mussels is a function of mussel availability along the line-transect and capture efficiency, we used the recapture probabilities estimated from the robust design to correct for the number of detected mussels at each line-transect. Furthermore, because recapture probabilities varied between the microhabitat types (river channels vs. riverbanks; see Results), we corrected the number of detected mussels in microhabitats separately. We considered mussels within 0.5 m from the riverbank as occupying riverbank habitat; all others were considered to be in the river channel. We estimated mussel density for each line-transect and then estimated mean density for each transect site. Using densities for transect sites, we estimated mean density with 95% confidence intervals for each of the macrohabitats using nonparametric bootstrap methods. We estimated total population sizes for each macrohabitat from estimated density multiplied by total area of each macrohabitat in the inhabited stretch of the Black River. Trends in climate and hydrology We obtained historical climate data (mean monthly temperature, total monthly precipitation and Palmer Hydrological Drought Index: PHDI) from 1895-2012 in southeast New Mexico (Climatic Division 7) from the National Climatic Data Center (http://www.ncdc.noaa.gov/cag/) and mean monthly discharge from 1948-2012 in the Black River from USGS gauge number 08405500. We examined long-term trends of climatic and hydrological changes by decomposing time series data into seasonal, trend and irregular components using moving averages. We then fit the output with a smooth spline (smoothing parameter = 0.5) and tested for monotonic trends in time series variables using a Mann-Kendall trend test (Hirsch et al. 1982). To examine effects of climate on river discharge, we regressed mean monthly discharge with climate data (i.e. discharge vs. temperature, discharge vs. precipitation and discharge vs. temperature + precipitation); we excluded PHDI from this analysis because the values are calculated from temperature and precipitation. We calculated AIC values for each model to identify the best model. Statistical analyses of climate data were conducted using R (R Development Core Team 2011).

109 RESULTS Demographic parameter estimates We marked 873 unique individuals over the 15-year period; 81.2% of them were recaptured at least once. Two individuals captured in 1997 were recaptured in 2012, and thus, these animals were a minimum of 15 years old at the end of the study. Mussels marked in this study averaged 93.6 mm in length (range = 33.9-123.2 mm). The most general model describing the robust design dataset from 2006-2008 indicated overdispersion, with a ĉ value of 9.056. Therefore, we corrected for the overdispersion using ΔQAICc (Table 1). The two most-parsimonious models had ΔQAICc = 1.9452; values < 2 indicate equally parsimonious models. The model with the lowest QAICc (-495.7067; QAICc weight = 0.48034) had constant survival among LH sites and microhabitats over the three-year period. The model with the second lowest QAICc (-493.7615; QAICc weight = 0.18162) contained variance in survival among microhabitats. The first model found that survival over the three-year period was 99.6% (95% CI: 86.73-99.99). The second model estimated that survival was constant over time, but varied by microhabitat (Table 1). However, this model provided no evidence for a significant difference in survival between microhabitats; survival was more variable in river channels (99.7%; 95% CI: 39.31-99.99) compared to riverbanks (99.5%; 95% CI: 88.30-99.98). Although the two models gave slightly different survival estimates, estimates of recapture probabilities were the same for both models: 0.802 for river channels (95% CI: 0.726-0.861) and 0.703 for riverbanks (95% CI: 0.647-0.753). Thus, we used these recapture probabilities for the CJS model. When estimating survival over the 15-year period, we detected overdispersion for annual and seasonal datasets described by the CJS models (ĉ value of 3.952 and 6.752, respectively), and therefore, we corrected using ΔQAICc (Table 1). For both datasets, models that predicted survival based on minimum monthly discharge and time were the most parsimonious. All other models had ΔQAICc > 2, indicating that minimum monthly discharge best-describes the survival of mussels over the 15-year period. Further, because models in which survival varied by site and/or microhabitat were not the most parsimonious, the best model suggests that survival did not vary by either site or microhabitat. Survival showed similar patterns between annual and seasonal datasets (Figure 2a). Overall, survival estimates ranged from 97.8% between 2011 and 2012 (95% CI: 96.8-98.5) to 99.7% between October 2004 and May 2005 (95% CI: 99.0-99.9),

110 with survivorship over time a function of minimum monthly discharge (Figure 3). The average survivorship was 98.6% over the 15-year period. The Pradel model revealed that annual λ from 1997-2011 ranged from 0.988 to 1.064, with significant positive growth observed in 1997-1998 and 2004-2005 (Figure 2b). While most annual estimates included 1 within the 95% confidence intervals, three data segments (1998- 2002, 2006-2007 and 2007-2008) showed significant negative growth. The geometric mean of λ was 1.005 over the 15-year period. However, this was driven by the high estimate from 1997- 1998 followed by a gap of four years; the geometric mean of λ was 0.999 from 2002-2012, indicating a slight decline in population size. Based on estimates of λ and current population size (see below), total population size was relatively stable over the 15-year period (Figure 2c). Total population size estimate We estimated that riffles comprise 15.9% and pools comprise 84.1% of the total area (total area of this 14-km stretch = 281 167.1 m2); thus, pools are over five times more extensive than riffles. In both macrohabitat types, the majority of mussels occurred along riverbanks (Figure 4). We estimated the detection function as a uniform distribution; in other words, detection of mussels was equally efficient anywhere along the 1-m lines perpendicular to the transects (data not shown). After microhabitat-specific correction for detection of mussels (using recapture probabilities estimated from the robust design analysis), average density was eight times higher in riffle habitats (0.65 mussels m-2; bootstrap 95% CI: 0.40-1.00) than in pool habitats (0.08 mussels m-2; bootstrap 95% CI: 0.05-0.12). Based on total area for each macrohabitat in this stretch of the Black River, we estimated 28 989 mussels (95% CI: 17 733-44 801) inhabiting riffle habitats and 19 017 mussels (95% CI: 11 116-29 326) inhabiting pool habitats. From these, our estimate of total population size is 48 006 mussels (95% CI: 28 849-74 127), with 60.3% of mussels occurring in riffle habitats and 39.7% of mussels occurring in pool habitats. Trends in climate and hydrology Smooth splines indicated a steep increasing trend in mean monthly temperature in southeast New Mexico since the mid-1970s, and a decreasing trend in mean monthly discharge in the Black River since the mid-1990s (Figure 5). Mann-Kendall trend tests indicated a significant increasing trend in mean monthly temperature over the past 100 years (0.0059˚C year-1, P < 0.001; Figure 5a) and a significant decreasing trend in mean monthly discharge over the past 65

111 years (-0.0015 m3 s-1 year-1, P = 0.032; Figure 5d). However, we did not observe significant trends in smooth splines or Mann-Kendall trend tests using total monthly precipitation or PHDI in the region during the same time period (P = 0.325 and 0.487, respectively; Figures 5b & c). Regression analyses showed that total monthly precipitation and a combination of mean annual temperature and precipitation are equally best predictors for mean monthly discharge (AIC = 687.5 and 688.5, respectively), whereas temperature alone was not (AIC = 794.7). Discharge was positively correlated with total monthly precipitation (R2 = 0.149, P < 0.001).

DISCUSSION Estimates of abundance and distribution are necessary for developing effective management strategies and are the most commonly used benchmarks when conservation agencies create recovery plans and delisting criteria (Neel et al. 2012). However, accurate estimation is often problematic because survey methods seldom detect all individuals present in a study area. Such estimation is further problematic for organisms occurring in patchy distributions in heterogeneous habitats. Our study revealed that detection of mussels was lower in riverbank habitats than river channels, but not among sites, which supported the hypothesis that detection of mussels varied in heterogeneous microhabitats. This is likely because surveys in undercut riverbanks are often challenging due to benthic debris and difficulty searching in deep undercut riverbanks, while surveys in river channels are relatively easy to accomplish. Estimates of abundance can be under- and/or over-estimated unless one accounts for heterogeneity in detection probability of animals. For instance, our estimates of total abundance of P. popeii would have been underestimated by 27.7% without incorporating detection probabilities at all or overestimated by 2.9% if detection probability was estimated only for riverbank habitats. While ad hoc estimates can be informative to develop adaptive sampling-designs for further population assessment, optimal sampling-design accounting for habitat heterogeneity and detection efficiency is necessary in order to achieve accurate estimation and make informed conservation decisions. In the Black River, P. popeii aggregates at high densities in narrow riffles, while a majority of the macrohabitat consists of wide and deep pools/runs. This is likely due to the availability of suitable microhabitats. In most instances, we observed that pool/run habitats had relatively few large boulders and were covered with thick silt, whereas riffle habitats consisted of

112 bedrock and large boulders. Regardless of macrohabitat type, the majority of individuals were found in riverbanks. Because pools comprise a large portion of the total area, 40% of mussels inhabit what was previously thought to be inhospitable habitat (Lang 2001). Species recovery plans need to design proper strategies for habitat assessment in order to obtain accurate knowledge of spatial distributions. Estimates of survival over the 15-year period were similar to those found for other freshwater mussels (Meador et al. 2011; Villella et al. 2004; Hart et al. 2001). In our case, however, survival was a function of minimum monthly discharge, with no significant differences among sites, or microhabitats. This rejects the hypothesis that mussel survival would be affected more by maximum discharge than by minimum discharge. Instead, survival declined with decreasing minimum monthly discharge. During the mark-and-recapture study, we observed that marked individuals typically stayed in the same place (i.e. the same undercut riverbank, or base of a large boulder in the river channel) over the study period. They buried the anterior half of the shell into clay to stabilize position during high streamflow, especially flash floods that often exceeded base flow by 3-4 orders of magnitude. However during low-stream-flow years, we observed accumulation of deposited silt on river bottoms of the flow refuges and a few dead individuals buried completely under silt. Although the association between fine-sediment deposition and mussel survival is disputed (reviewed in Haag 2012), lower survival during low stream-flow has been attributed to siltation (Smakhtin 2001; Wood & Armitage 1999). Low stream-flow during drought enhances fine-sediment accumulation in areas where sedimentation does not usually occur (Wood & Armitage 1999). We observed that P. popeii does not inhabit silt substrates in pool and run habitats. In addition, we noted that river bottoms were cleared of silt after flash flood events. We presume that periodically, extreme flows wash out deposited sediment from flow refuges leading to high survival of mussels, while low flows decrease survival by enhancing accumulation of fine-sediment. Siltation may further increase due to excessive grazing by cattle on riparian vegetation (Belsky et al. 1999), and vehicle traffic over low-water crossings. Swirling dust created by vehicles is a source of fine sediments and contaminants that enter aquatic ecosystems (Trombulak & Frissell 2000). Unbiased estimates of demographic parameters often require that studies be sufficiently long to account for temporal variation (Sandercock 2006). Our mark-and-recapture study was considerably longer than other such studies with freshwater mussels (3-9 years; Hart et al. 2001;

113 Meador et al. 2011; Villella et al. 2004) and is the first study that we know of to show that survival of benthic macroinvertebrates is associated with river discharge. Typically, life spans of freshwater mussels are longer than other benthic macroinvertebrates (4-5 years to more than 100 years; Haag & Rypel 2011). Such long life spans require long-term study to characterize the relationship between a species’ population dynamics and its environment because the interaction may reflect seasonal, annual or decadal events. We were able to estimate surprisingly stable survival and population growth rates over the 15-year period, while accounting for habitat heterogeneity and hydrological cycles when examining population dynamics. We should note, however, that our estimates of survival might only represent adult survival because the smallest mussels we could detect (approximately 30mm total length) during the study were likely sexually mature adults. Because juvenile mussels probably live buried in substratum for a few years before emerging to the benthic surface, detection of juvenile mussels is extremely difficult without appropriate sampling techniques (e.g. excavating and sieving substrata). Our tactile searches likely missed these juveniles. Future studies that are able to census and track juveniles should improve estimates of demographic parameters of mussel species. A declining trend in survival associated with low streamflow suggests that significant changes in precipitation, hydrological regimes and water demands in the region may induce significant declines in population size (Figure 6). Mean annual temperature and precipitation are predicted to be significantly altered due to human-induced climate change (USGCRP 2009), with current projections suggesting that hotter and drier climates will become the norm in the southwestern U.S.A. (Seager et al. 2007). Furthermore, water demands for domestic, agricultural and industrial uses are predicted to increase due to human population growth. Our trend analysis indicates significantly increasing temperature in the region and declining discharge in the Black River over a 65-year period. We should note that the significant trends we observed in temperature and discharge are monotonic averages over time; and thus, these trends did not necessarily start at the beginning of the data periods we investigated. In fact, smooth splines showed steeply increasing temperature in the last 40 years and decreasing discharge in the last 15 years. Furthermore, the lack of significant temporal trends in precipitation in the region but significant correlation of time with discharge of the Black River suggests that reduction of stream-flow is likely due to lowering of the water table due to decreased groundwater recharge. Because the Black River is primarily groundwater fed, studies of recharge in the region are

114 necessary to predict future hydrological trends in the face of climate- and human-induced changes in water availability. Additionally, Southeastern New Mexico’s location within the Permian Basin, which is a large oil- and natural gas-producing region, may further exacerbate trends in regional water quality and supply. In the past 20 years, the number of oil/gas wells has tripled in the Black River catchment (data obtained from New Mexico Energy, Minerals and Natural Resources Department; http://www.emnrd.state.nm.us/OCD/). The combination of hotter and drier climate, along with increase in water demands, all increase the risk of extinction for P. popeii and other endemic aquatic taxa of the northern Chihuahuan Desert. The overall status of the species is tenuous because P. popeii inhabits a substantially reduced range, and current extant populations are limited to a few localities. High survival and relatively stable λ over time suggest that protection of the P. popeii population of the Black River requires maintaining the status quo in the river, while taking steps to minimize anthropogenic catastrophes. For example, the volume of surface and groundwater withdrawal within the catchment should be restricted to minimize declines in the water table during drought periods. Accidental release of harmful substances in the catchment must be avoided at all cost; risk could be minimized by monitoring water quality and the effects of groundwater exploitation. At the same time, the reintroduction program into the Delaware River (Figure 1), part of the historic range of P. popeii and currently the target of significant restoration activities by the New Mexico Department of Game and Fish, the U.S. Bureau of Land Management and the U. S. Fish and Wildlife Service, should be continued. In 2012-2013, reintroduction of P. popeii and its host fishes to the Delaware River was initiated and survival is currently being monitored. This program would prevent extinction of the New Mexico population in the case of a natural or anthropogenic catastrophe in the Black River. Global biodiversity is facing a crisis of escalating species extinction and habitat loss due to climate change and anthropogenic activities. This is especially pronounced in desert aquatic ecosystems, which are often composed of large numbers of endemic species living in isolated environments. We demonstrated that survival of aquatic invertebrates in desert streams is sensitive to hydrological cycles, which are expected to be altered via climate change and extensive water use in the region. Long-lived organisms require long-term studies in order to identify threats such as these. Additionally, long-term study of freshwater ecosystems can inform understanding of many more-general topics (e.g. hydrologic regimes, natural and anthropogenic

115 disturbance, biological monitoring and climate change). Our results provide necessary information for making conservation decisions and developing recovery strategies. Furthermore, we hope that our results will encourage other researchers to conduct long-term studies to elucidate biological trends caused by climate change and other anthropogenic activities.

ACKNOWLEDGMENTS We thank the New Mexico Department of Game and Fish, the U.S. Fish and Wildlife Service, the Albuquerque BioPark and undergraduate and graduate students in the Berg Lab for field assistance, and the Statistical Consulting Center at Miami University for assistance with statistical analyses. We also thank private landowners for access to the river. We thank T. Crist, M. Youngquist, Berg Lab members and two anonymous reviewers for their helpful comments on earlier drafts of this manuscript. Funding was provided by the New Mexico Department of Game and Fish, and the U.S. Fish and Wildlife Service.

REFERENCES Alexander HM, Slade NA, Kettle WD (1997) Application of mark-recapture models to estimation of the population size of plants. Ecology, 78, 1230-1237. Anderson DR, Burnham KP (1999) Understanding information criteria for selection among capture-recapture or ring recovery models. Bird Study, 46, S14-S21. Barnett TP, Pierce DW, Hidalgo HG, Bonfils C, Santer BD, Das T, Bala G, Wood AW, Nozawa T, Mirin AA, Cayan DR, Dettinger MD (2008) Human-induced changes in the hydrology of the western United States. Science, 319, 1080-1083. Belsky AJ, Matzke A, Uselman S (1999) Survey of livestock influences on stream and riparian ecosystems in the western United States. Journal of Soil and Water Conservation, 54, 419-431. Bjorklund LJ, Motts WS (1959) Geology and water resources of the Carlsbad area, Eddy County, New Mexico. U.S. Department of Interior, U.S. Geological Survey. Bogan AE (1996) Popenaias popeii. In: IUCN 2013. IUCN Red List of Threatened Species. Bogan MT, Lytle DA (2011) Severe drought drives novel community trajectories in desert stream pools. Freshwater Biology, 56, 2070-2081.

116 Burnham KP, Anderson DR (2002) Model Selection And Multimodel Inference: A Practical Information Theoretic Approach Springer, New York, USA. Carman SM (2007) Texas hornshell Popenaias popeii recovery plan, p. 57. New Mexico Department of Game and Fish, Santa Fe, NM. Cayan DR, Das T, Pierce DW, Barnett TP, Tyree M, Gershunov A (2010) Future dryness in the southwest US and the hydrology of the early 21st century drought. Proceedings of the National Academy of Sciences of the United States of America, 107, 21271-21276. Cowley DE, Sublette JE (1987) Distribution of fishes in the Black River drainage, Eddy County, New Mexico. Southwestern Naturalist, 32, 213-221. Deacon JE, Williams AE, Williams CD, Williams JE (2007) Fueling population growth in Las Vegas: how large-scale groundwater withdrawal could burn regional biodiversity. Bioscience, 57, 688-698. Haag WR (2012) North American Freshwater Mussels: Natural History, Ecology, and Conservation Cambridge University Press, Cambridge, UK. Haag WR, Rypel AL (2011) Growth and longevity in freshwater mussels: evolutionary and conservation implications. Biological Reviews, 86, 225-247. Hart RA, Grier JW, Miller AC, Davis M (2001) Empirically derived survival rates of a native mussel, Amblema plicata, in the Mississippi and Otter Tail Rivers, Minnesota. American Midland Naturalist, 146, 254-263. Hirsch RM, Slack JR, Smith RA (1982) Techniques of trend analysis for monthly water quality data. Water Resource Research, 18, 107-121. Howells RG, Neck RW, Murray HD (1996) Freshwater mussels of Texas Texas Park and Wildlife Department, Austin, TX. Jackson JK, Füreder L (2006) Long-term studies of freshwater macroinvertebrates: a review of the frequency, duration and ecological significance. Freshwater Biology, 51, 591-603. Karatayev AY, Miller TD, Burlakova LE (2012) Long-term changes in unionid assemblages in the Rio Grande, one of the world's top 10 rivers at risk. Aquatic Conservation: Marine and Freshwater Ecosystems, 22, 206-219. Kendall WL, Nichols JD, Hines JE (1997) Estimating temporary emigration using capture- recapture data with Pollock's robust design. Ecology, 78, 563-578.

117 Lang BK (2001) Status of the Texas hornshell and native freshwater mussels (Unionoidea) in the Rio Grande and Pecos River of New Mexico and Texas. New Mexico Department of Game and Fish, Santa Fe, New Mexico. Levine TD, Lang BK, Berg DJ (2012) Physiological and ecological hosts of Popenaias popeii (Bivalvia: Unionidae): laboratory studies identify more hosts than field studies. Freshwater Biology, 57, 1854-1864. Lindberg MS (2012) A review of designs for capture–mark–recapture studies in discrete time. Journal of Ornithology, 152, 355-370. MacDonald GM (2010) Water, climate change, and sustainability in the southwest. Proceedings of the National Academy of Sciences of the United States of America, 107, 21256-21262. Malmqvist B, Rundle S (2002) Threats to the running water ecosystems of the world. Environmental Conservation, 29, 134-153. Meador JR, Peterson JT, Wisniewski JM (2011) An evaluation of the factors influencing freshwater mussel capture probability, survival, and temporary emigration in a large lowland river. Journal of the North American Benthological Society, 30, 507-521. Morris WF, Doak DF (2002) Quantitative Conservation Biology: Theory and Practice of Population Viability Analysis Sinauer Associates Inc., Sunderland, MA. Neel MC, Leidner AK, Haines A, Goble DD, Scott JM (2012) By the numbers: how is recovery defined by the US Endangered Species Act? Bioscience, 62, 646-657. Poff NL, Allan JD, Bain MB, Karr JR, Prestegaard KL, Richter BD, Sparks RE, Stromberg JC (1997) The natural flow regime. Bioscience, 47, 769-784. Pollock KH (1982) A capture-recapture design robust to unequal probability of capture. Journal of Wildlife Management, 46, 752-757. Pollock KH (2002) The use of auxiliary variables in capture-recapture modelling: An overview. Journal of Applied Statistics, 29, 85-102. Pradel R (1996) Utilization of capture-mark-recapture for study of recruitment and population growth rate. Biometrics, 52, 703-709. R Development Core Team (2011) R: A language and environment for statistical computing. R Fundation for Statistical Computing, Vienna, Austria.

118 Richard M (1988) Natural gas contamination at Rattlesnake Springs, Carlsbad Caverns National Park: report of the first field investigation, August 1988. Report 2. National Park Service Contract RFQ 7029-8-0025. Richard M, Boehm A (1989) Natural gas contamination at Rattlesnake Springs, Carlsbad Caverns National Park: final summary of the investigation. Report 4. National Park Service Contract RFQ 7029-8-0025. Royle JA, Nichols JD (2003) Estimating abundance from repeated presence-absence data or point counts. Ecology, 84, 777-790. Sandercock BK (2006) Estimation of demographic parameters from live-encounter data: a summary review. Journal of Wildlife Management, 70, 1504-1520. Seager R, Ting M, Held I, Kushnir Y, Lu J, Vecchi G, Huang HP, Harnik N, Leetmaa A, Lau NC, Li C, Velez J, Naik N (2007) Model projections of an imminent transition to a more arid climate in southwestern North America. Science, 316, 1181-1184. Seidel RA, Lang BK, Berg DJ (2009) Phylogeographic analysis reveals multiple cryptic species of amphipods (Crustacea: Amphipoda) in Chihuahuan Desert springs. Biological Conservation, 142, 2303-2313. Seidel RA, Lang BK, Berg DJ (2010) Salinity tolerance as a potential driver of ecological speciation in amphipods (Gammarus spp.) from the northern Chihuahuan Desert. Journal of the North American Benthological Society, 29, 1161-1169. Smakhtin VU (2001) Low flow hydrology: a review. Journal of Hydrology, 240, 147-186. Smith DG, Lang BK, Gordon ME (2003) Gametogenetic cycle, reproductive anatomy, and larval morphology of Popenaias popeii (Unionoida) from the Black River, New Mexico. Southwestern Naturalist, 48, 333-340. Strayer DL (1999) Use of flow refuges by unionid mussels in rivers. Journal of the North American Benthological Society, 18, 468-476. Strayer DL, Smith DR (2003) A Guide to Sampling Freshwater Mussel Populations American Fisheries Society, Bethesda, MD. Strenth NE, Howells RG, Correa-Sandoval A (2004) New records of the Texas hornshell Popenaias popeii (Bivalvia: Unionidae) from Texas and northern Mexico. The Texas Journal of Science, 56, 223-230.

119 Sublette JE, Hatch MD, Sublette M (1990) The fishes of New Mexico University of New Mexico Press, Albuquerque, New Mexico. Thomas L, Buckland ST, Rexstad EA, Laake JL, Strindberg S, Hedley SL, Bishop JRB, Marques TA, Burnham KP (2010) Distance software: design and analysis of distance sampling surveys for estimating population size. Journal of Applied Ecology, 47, 5-14. Trombulak SC, Frissell CA (2000) Review of ecological effects of roads on terrestrial and aquatic communities. Conservation Biology, 14, 18-30. USFWS (US Fish and Wildlife Service). 2013. Endangered and threatened wildlife and plants; determination of endangered species status for six west Texas aquatic invertebrates. Federal Register 78, 41228-41258. USGCRP (2009) Global Climate Change Impacts in the United States Cambridge University Press, New York, NY. Villella RF, Smith DR, Lemarié DP (2004) Estimating survival and recruitment in a freshwater mussel population using mark-recapture techniques. American Midland Naturalist, 151, 114-133. Vörösmarty CJ, McIntyre PB, Gessner MO, Dudgeon D, Prusevich A, Green P, Glidden S, Bunn SE, Sullivan CA, Liermann CR, Davies PM (2010) Global threats to human water security and river biodiversity. Nature, 467, 555-561. White GC, Burnham KP (1999) Program MARK: survival estimation from populations of marked animals. Bird Study, 46, S120-S139. Witt JDS, Threloff DL, Hebert BL (2006) DNA barcoding reveals extraordinary cryptic diversity in an amphipod genus: implications for desert spring conservation. Molecular Ecology, 15, 3073-3082. Wood PJ, Armitage PD (1999) Sediment deposition in a small lowland stream –management implications. Regulated Rivers: Research & Management, 15, 199-210.

120 TABLES AND FIGURES Table 1. Results for the five most-parsimonious models for the robust design in 2006-2008 and the Cormack-Jolly-Seber (CJS) models for annual and seasonal datasets in 1997-2012. Model parameters include survival (), migration factors (γ' = emigration, γ'' = immigration), encounter probability (p), recapture probability (c) and abundance (N). Parameters denoted in parentheses indicate no-variance (.) and variance by time (t), habitat (h) and site (s) for the robust design. Additionally for CJS models, parameters indicate variance by hydrological parameters (mean, minimum, and maximum discharge and numbers of days below the 10th and 25th percentile and above the 75th and 90th percentile of daily discharge). Asterisks (*) indicate combinations of variances. Encounter probability for CJS models is conditioned with recapture probabilities estimated from the robust design (RD). AICc No. Dataset Model QAICc a ΔQAICc b Weights parameters Robust (.) γ'(.) γ''(.) p(t) c(h) N(s*t) -495.7067 0 0.48034 19 (h) γ'(.) γ''(.) p(t) c(h) N(s*t) -493.7615 1.9452 0.18162 20 (.) γ'(.) γ''(.) p(t) c(.) N(s*t) -492.9747 2.7320 0.12255 18 (t) γ'(.) γ''(.) p(t) c(h) N(s*t) -492.2482 3.4585 0.08522 21 (h) γ'(.) γ''(.) p(t) c(.) N(s*t) -491.0303 4.6764 0.04635 19 Annual (minimum discharge*t) p(RD) 1378.9925 0 0.45912 2 (minimum discharge*h*t) p(RD) 1382.2403 3.2478 0.09051 4 (75th percentile*t) p(RD) 1382.3039 3.3114 0.08767 2 (90th percentile*t) p(RD) 1382.6006 3.6081 0.07559 2 (mean discharge*t) p(RD) 1382.8605 3.8680 0.06637 2 Seasonal (minimum discharge*t) p(RD) 1245.8264 0 0.54467 2 (mean discharge*t) p(RD) 1248.5866 2.7602 0.13701 2 (minimum discharge*s*t) p(RD) 1248.7543 2.9279 0.12599 6 (mean discharge*s*t) p(RD) 1251.0820 5.2556 0.03935 6 (25th percentile*t) p(RD) 1251.3506 5.5242 0.03440 2 a Corrected quasi-Akaike information criterion b Differences in QAICc units from the model with the lowest QAICc value.

121

Figure 1. Regional map (top-left corner) and the Pecos River and its perennial tributaries in Eddy County, New Mexico. Life history (LH) sites are shown as black dots. The study area for distance sampling is highlighted in grey.

122

Figure 2. (a) Annual and seasonal survival and (b) annual finite rate of population growth (λ) of Popenaias popeii over a 15-year period estimated from the most parsimonious mark-and- recapture model, and (c) total population size in the Black River estimated from λ and total abundance for 2012 estimated from distance sampling. In panel (a), the dotted line with circles represents the seasonal dataset and the solid line with diamonds represents the annual dataset. In panel (b), the dotted line indicates λ = 1 and asterisks (*) represent λ values significantly different than 1. Error bars are 95% confidence intervals.

123

Figure 3. Survival of Popenaias popeii plotted against minimum monthly discharge of the Black River. The line represents the best-fit regression line. Error bars are 95% confidence intervals.

124

Figure 4. Mussel distributions along transect lines for riffle (dark grey) and pool (light grey) habitats. Number of mussels found per distance category from the bank for riffle and pool habitats is shown in bar graphs. Cumulative frequencies for riffle (solid line) and pool (dashed line) habitats are shown in line graphs. The vertical dotted line indicates the boundary between the river channel and riverbank.

125

Figure 5. Monthly climatic and hydrological trends after removing seasonal effects (grey lines) in southeast New Mexico and the Black River. Climatic measures include mean monthly temperature (a), total monthly precipitation (b) and Palmer hydrological drought index (PHDI; c) from 1895-2012, while the hydrological measure is mean monthly discharge of the Black River (d) from 1948-2012. Black lines are smooth splines (smoothing parameter = 0.5). Asterisks (*) indicate statistical significance (P < 0.05) of Mann-Kendall trend tests.

126 Ultimate causes

Climate change + Human population growth - + Proximate causes Water uses • Domestic Precipitation • Agricultural • Industrial - - River discharge - - Water quality • Nutrient concentration • Water temperature • Sedimentation • Contamination -

Invertebrate survival

Figure 6. Conceptual diagram of the effects of climate change and increases in anthropogenic activities on aquatic invertebrate survival. Arrows indicate directional effects, and symbols along the arrows represent positive (+) and negative (-) effects.

127 Chapter 6: Integrating population viability analysis and genetic assessments into endangered species recovery planning for freshwater mussels

Co-authors: Brian K. Lang1, David J. Berg2 1 New Mexico Department of Game and Fish, Santa Fe, NM 87507 2 Department of Biology, Miami University, Hamilton, OH 45011

ABSTRACT An endangered species recovery plan establishes quantitative, measureable criteria for removal of the species from protection under the US Endangered Species Act (ESA). Population viability analysis (PVA) is commonly used as a conservation tool; however, its use is heavily weighted towards terrestrial vertebrates. Because freshwater mussels are among the most imperiled groups of animals, we reviewed 71 recovery plans for 90 listed mussel species to assess whether recovery plans have implemented demographic and genetic viability assessments. We then used empirical examples of two mussels, the recently listed Cumberlandia monodonta and the candidate species Popenaias popeii, to demonstrate that PVA integrated with genetic assessments can be implemented into recovery planning and downlisting/delisting criteria. While the concept of PVA and genetic assessments are pervasive in the recovery plans for listed mussel species, few plans explicitly implemented these assessments in their recovery strategies. Among the 90 listed species, five and 13 species have developed demographic PVA models and conducted genetic assessments, respectively. Our genetic PVA models revealed that demographic parameters during early life stages highly interact to influence the probability of extinction and finite rates of population growth for both species. In addition to population size, initial allele frequencies and population growth rates highly influence the trajectories of population genetic persistence for the foreseeable future. Failure to implement quantitative strategies and criteria in the recovery plans leads to uncertainty in achieving these goals. Our genetic PVAs effectively demonstrate that, when sufficient data exist, quantitative recovery criteria based on demographic and genetic data can create objective, measureable goals for recovery plans. We hope that this study will encourage others to utilize ecological and genetic modeling in developing recovery plans for species from this highly imperiled group.

128 Keywords: extinction risk assessment, genetic diversity, recovery plans, recovery criteria, Unionoida, Vortex, PVA, review

INTRODUCTION Endangered species conservation is an enormously complex task. Since the US Endangered Species Act (ESA) of 1973 was enacted, 1577 animal and plant species in the USA have been listed as threatened or endangered as of March 2015. Because the primary goals of the ESA are to prevent extinction and to recover species from the necessity of protection, the ESA prescribes development and implementation of recovery plans that include scientifically objective, measureable recovery criteria and management actions that mitigate threats (16 U.S.C. §1533 [f]). To establish such recovery criteria, population viability analysis (PVA) has been widely used as a tool for assessing extinction risks, exploring the consequences of management actions, and developing recovery strategies (Megens 2000; Reed et al. 2002). Population viability analysis is a quantitative demographic modeling process that uses empirical and hypothetical estimates of demographic parameters to evaluate demographic trajectories and assess extinction risks for a given species (Zeigler et al. 2013). These analyses have included spatially explicit models to assess the movements and fates of individual organisms across a complex landscape (Reed et al. 2002; Naujokaitis-Lewis et al. 2009); sensitivity and elasticity analyses to predict factors that most affect population growth and extinction probability (Cross & Beissinger 2001; Naujokaitis-Lewis et al. 2009); and more recently, genetic analyses to predict the maintenance of genetic variation and structure (Allendorf et al. 2010; Frankham et al. 2014). Because extinction is fundamentally a demographic process influenced by genetic and environmental factors, demography may be of more immediate importance for developing recovery criteria (Lande 1988). However, there is extensive evidence that inbreeding depression substantially reduces population viability (Frankham et al. 2014). Furthermore, along a long- term demographic trajectory, the degree of genetic variability within species dictates evolutionary adaptability to natural and anthropogenic environmental changes in the future. The maintenance of genetic diversity is a function of population size, where random genetic drift tends to reduce genetic variation in small, isolated populations, and thus, causes the loss of fitness and evolutionary adaptability (Vrijenhoek 1994). Population viability analysis models often incorporate inbreeding depression (loss of heterozygosity) to simulate the inbreeding

129 effects on vital rates of populations. Furthermore, the use of forward-time genetic simulations implemented in PVA models allows estimation of the persistence of genetic variability in the future under various demographic and environmental conditions (Hoban et al. 2011). The persistence of genetic variability in isolated populations is especially significant, given rapid changes in ecological conditions due to anthropogenic disturbance and climate change. The use of PVA integrated with genetic assessments to-date is, however, heavily weighted toward terrestrial vertebrate species and is rarely implemented into recovery planning of aquatic macroinvertebrate species (Naujokaitis-Lewis et al. 2009), even though freshwater mollusks are one of the most imperiled groups of organisms in the world (Lydeard et al. 2004). A recent review showed that only 4% of demographic PVA models were developed in invertebrate taxa (Naujokaitis-Lewis et al. 2009). This is because almost all PVA models require detailed estimates of demographic parameters (Reed et al. 2002) that are often difficult to acquire for many taxa due to uncertainties in life cycles (for instance, in early life stages), host interactions, recruitment and the effects of environmental variation. Particularly, freshwater mussels (order Unionoida) possess a complex life history in which larvae (glochidia) are obligate parasites of vertebrate hosts. An adult female produces from a few thousands to more than 10 million glochidia per reproductive season and uses various strategies to transmit glochidia to hosts (Williams et al. 2008). While annual adult survival is as high as 98% (Inoue et al. 2014a), survival during early life stages (e.g., glochidial and juvenile stages) is exceptionally low (reviewed in Haag 2012). Gathering such information is, however, difficult in short-term studies, given their secretive life history and long generation time (more than 20 years; Inoue et al. 2014b). In this study, we first reviewed recovery plans for endangered freshwater mussel species (order Unionoida) that have been approved by the US Fish and Wildlife Service (USFWS) in order to determine how demographic and genetic viability assessments have been used in the recovery plans. We examined 71 recovery plans for 90 mussels species/subspecies listed under the ESA as of April 2015. We then used empirical examples to develop PVAs that integrated genetic assessments (hereafter, genetic PVAs) in order to explore potential use of genetic PVAs in recovery planning. We focused on two imperiled mussel species (spectaclecase [Cumberlandia monodonta] and Texas hornshell [Popenaias popeii]) from North America as examples. Cumberlandia monodonta (family Margaritiferidae) was listed as endangered under

130 the ESA in 2012 (USFWS 2012) due to severe reduction in number of occurrences; its recovery plan has not yet been developed. Historically, this species occurred in a large portion of the Mississippi River basin; currently, however, its occurrence is limited to a few localities within the historic range. Popenaias popeii (family Unionidae) is currently a candidate for listing under the ESA (Priority 8; USFWS 2014). This species is endemic to the Rio Grande drainage in the USA and Mexico, and coastal Gulf of Mexico drainages in northern Mexico; however, currently P. popeii populations in the USA are restricted to a few localities in the mainstem of the Rio Grande and two tributaries (Carman 2007). The biology of the two species have been relatively well studied, including their reproductive biology (Gordon & Smith 1990; Smith et al. 2003; Levine et al. 2012), demographic characteristics and population dynamics (Baird 2000; Inoue et al. 2014a), and population genetics (Inoue et al. 2014b; Inoue et al. 2015). The goals of the genetic PVAs were to develop quantitative recovery criteria and strategies for these species prior to the development of their recovery plans. Through genetic PVA simulations, we incorporated empirical demographic and population genetic data of the extant populations into PVA models to determine the interaction between demographic and genetic persistence over time. The use of empirical data allows us to realistically assess the effects of changes in demographic parameters on demographic and genetic viabilities over time. We set specific objectives for the genetic PVA to (1) examine the effects of uncertainties in demographic parameters (i.e., early life stages) on the probability of extinction and population growth; (2) perform sensitivity analyses of the influence of demographic parameters on extinction probability; and (3) examine the effects of varying demographic parameters on the trajectories of genetic diversity over time. This study has broad-scale implications not only for evaluating current recovery plans and criteria for freshwater mussel conservation, but also to aid in the development of additional conservation management practices (e.g., captive propagation, augmentation, reintroduction) for these and other imperiled species.

METHODS The use of PVA and genetic assessment in recovery planning We modified the methods used in Zeigler et al. (2013) to assess use of PVA and genetic assessment in endangered species recovery planning. We reviewed 71 recovery plans (70 final and 1 draft plans) for 90 endangered and threatened mussel species/subspecies that were

131 approved by the USFWS prior to 1st April 2015. For each recovery plan, we searched for population viability keywords: population viability analysis, minimum viable population, matrix, model, and viability (Zeigler et al. 2013); and genetic assessments keywords: population genetics, genetic diversity/variability/variation, and effective population size. We recorded how PVA and genetic assessments were discussed in each of the recovery plans, including whether recovery criteria were written in the language of PVA (e.g., the plan determines number of individuals required to maintain a viable population) and genetic assessment (e.g., recovered population should have sufficient genetic variation to enable evolution and response to natural habitat changes), and whether PVA and/or genetic assessments were recommended as a part of the recovery strategies to determine, evaluate, or refine recovery criteria. In addition, we recorded recovery goals and presence/absence of quantitative statements related to PVA and genetic assessments in recovery criteria if downlisting and/or delisting were considered in the plan. In addition, we used ISI’s Web of Science database and the search engine Google Scholar to search for peer-reviewed literature and technical reports (e.g., 5-year reviews) related to PVA and genetic assessment for the listed mussel species. We searched for the keywords described above and found all publications prior to March 2015. We recorded how PVA and/or genetic assessment were conducted and used to inform development of conservation strategies. Case Studies Data collection We used two mussel species (C. monodonta and P. popeii) as examples to demonstrate the use of genetic PVA in recovery strategies. We employed Vortex v10.0.8.0 for each species to build PVA models to project population and genetic trajectories. Vortex simulates population trajectories through a series of demographic parameters and individual genotypes by tracking the sex, age, and parentage of each individual in the population through time, while incorporating demographic, environmental, and genetic stochasticity. Using a user-specified starting population and genetic relationships among individuals, Vortex allows tracking of both demographic metrics (population size, population growth rate, and time to extinction) and genetic metrics (heterozygosity and allelic richness) over the simulated time period. For each species, we parameterized Vortex with best-available information from published studies, field observations, and technical reports (Table 1). If species-specific

132 parameters were unavailable, we used parameters known for closely related species. Table 1 describes demographic parameters as input to Vortex for a base-model. Freshwater mussels produce a large number of glochidia: average fecundity of five million glochidia per female for C. monodonta (Baird 2000) and 200,000 glochidia for P. popeii (personal observation). Because of computational demands and the fact that it is not practical to simulate a large number of glochidia, we used the number of transformed juveniles from glochidia per female (hereafter, the number of juveniles-per-female) as an input for the number of offspring per female in Vortex (Table 1). Furthermore, we did not incorporate density dependence in reproduction (i.e., Allee effects) because high fertilization efficiency at low density is pervasive among freshwater mussels (Mosley et al. 2014). PVA models with uncertain parameters Although adult survival rate and reproductive modes for many mussel species are relatively well understood, information for early life stages (i.e., glochidial and juvenile stages) is often limited. In our case, we have no species-specific information for transformation rates from glochidia to juvenile nor mortality rates of glochidia and juveniles. Thus, for transformation rates we used an estimate from Margaritifera margaritifera (Young & Williams 1984) for C. monodonta, and the mean transformation rate estimated over five unionid species (Haag 2002) for P. popeii. Furthermore, although mortality rate of newly settled juveniles is predicted to be high, it can be highly variable between species and between natural and artificial conditions. We have no species-specific information for mortality rates of newly settled juveniles immediately after settlement; thus, we used mortality rates for the first year (i.e., mortality from settled juvenile to Age 1; hereafter, juvenile mortality rate) set at 93% for C. monodonta obtained from Young and Williams (1984) and 95% for P. popeii obtained from Hart et al. (2004). Because mortality rates in freshwater mussels are predicted to quickly decrease with their age, we set mortality rates for second and third years (hereafter, young mortality rate) at 15% for both species (Haag 2012). To examine the effects of uncertainty in these parameters on the probability of extinction and population growth, we built PVA models by varying the number of juveniles per female and juvenile mortality rates to determine tendencies of the probability of extinction and finite rate of population growth () over time. We used the lowest number of juveniles per female at 0.1 juveniles per female based on the lowest estimate for unionids in Haag (2012); we used the

133 highest estimates at 7.2 juveniles per female estimated from hatchery trials of M. margaritifera (Preston et al. 2007) for C. monodonta, and 12 juveniles per female from the maximum estimate for unionids in Haag (2002) for P. popeii. Furthermore, because detailed mortality rates for juveniles were unavailable, we varied the juvenile mortality rate from 90% to 99% based on the range used in Hart et al. (2004). We set the initial population size at 3000 individuals for C. monodonta and 1000 individuals for P. popeii with a stable age distribution, based on mean population sizes from field estimates (Baird 2000; personal observation). We performed Vortex runs by generating 1000 sets of parameters in which values for juveniles-per-female and juvenile mortality rate were drawn across the range of each parameter. We aimed for simulating demographic trajectories over at least 10 generations; thus, each of the 1000 parameter sets was evaluated based on 100 replicate simulations of 300 years each (approximately 10 generations for C. monodonta and 20 generations for P. popeii). We determined the association of the two parameters with the probability of extinction and mean  over 300 years. Sensitivity analysis Because freshwater mussels and other benthic invertebrates are often difficult to sample, detailed life histories are lacking and as a result, substantial uncertainty in demographic parameters may decrease the reliability of PVA. We performed a relative sensitivity analysis by generating 1000 parameter-sets in which values for eight demographic parameters were drawn randomly from a uniform distribution within the ranges described in Table 2. The sensitivity analysis also included the pairs of simulated number of juveniles per female and juvenile mortality rate described above (Table 2). Relative sensitivity analysis has been previously used when either the empirical distribution of demographic parameters was unknown or demographic parameters were difficult to derive from the literature (Cross & Beissinger 2001; Carroll et al. 2014). Each of the 1000 parameter-sets was evaluated based on 100 replicate simulations of 300 years each in extinction and two quasi-extinction scenarios. Extinction was defined as only one individual of either sex remaining in the population and two quasi-extinction scenarios as the mean population size falling below 250 or 500 individuals. We used standardized coefficients from logistic regression of parameters against the probability of extinction or quasi-extinction to rank the effects of demographic parameters on the extinction probability (Cross & Beissinger 2001). Dividing a regression coefficient by its standard error results in a standardized regression coefficient (z-value), which expresses the

134 unique contribution of that parameter scaled by the variability of the parameter (Cross & Beissinger 2001). The absolute z-values are unitless and interpretable only in comparison with other z-values in the same model. The effect of demography on population genetics We assessed the effects of changes in demographic parameters on population genetic trajectories for the two species over time. In the simulations, we assessed the effects of two demographic parameters, mean  and initial population size, on the maintenance of genetic diversity over time. However, Vortex does not allow us to directly control mean ; thus, we varied the number of juveniles per female with constant juvenile mortality rate (see Results) to vary mean , targeting the range of  from 0.95 to 1.03. We used empirical allele frequencies from 16 microsatellite loci and one mitochondrial DNA gene (cytochrome oxidase I; COI) obtained from Inoue et al. (2014b) for C. monodonta and 18 microsatellite loci and the COI gene obtained from Inoue et al. (2015) for P. popeii. Previous studies revealed that both species are currently composed of two genetically diverged populations throughout their ranges (Inoue et al. 2014b; Inoue et al. 2015), where one population has significantly less genetic diversity (e.g., allelic richness and level of heterozygosity) than the other. Thus, we simulated population genetic trajectories of two genetically diverged populations for each of the species by varying the demographic parameters. We used the demographic parameters for a base-model in Vortex. We performed Vortex runs by generating sets of two varying parameters in which values for the number of juveniles per female were drawn between 1 and 7 at intervals of 0.5, and initial population sizes were drawn between 50 and 2000 at intervals of 50. We conditioned the juvenile mortality rates at 93% for C. monodonta and 95% for P. popeii. Each of the paired parameter sets was evaluated based on 100 replicate simulations of 300 years each. At the end of each simulation, we measured two genetic diversity indices: mean allelic richness (AR) and expected heterozygosity (HE), and calculated the probability of genetic diversity loss from the original diversity after 300 years. We did not incorporate mutation rates because of short simulation times.

135 RESULTS Use of PVA and genetic assessment in recovery planning The use of PVA and genetic assessment in recovery planning was generally limited; only three of the 71 recovery plans explicitly mentioned both PVA and genetic assessment, and recommended implementing these in recovery strategies (six recovery plans recommended PVA only and 31 plans recommended genetic assessment only). However, the concept of PVA and genetic assessment in recovery planning was nearly pervasive; 70% of recovery plans mentioned and/or adopted “viable population” and 79% mentioned and/or adopted “genetic variability/viability,” in their recovery goals and downlisting/delisting criteria (Table S1, supplemental information). For instance, 44 recovery plans stated their recovery goals as “to maintain, enhance, and restore viable populations of the species to a significant portion of its historic range” and the definition of a viable population as “a naturally reproducing population that is large enough to maintain sufficient genetic variation to enable it to evolve and respond to natural environmental changes without further intervention” (Table S2, supplemental information). Interestingly, only four recovery plans defined a viable population as a stable or increasing population in addition to the above definition. Furthermore, 55 recovery plans had quantitative downlisting/delisting criteria of which 80% suggested downlisting/delisting when the species reached a target number of viable populations and/or reestablished new populations in its historic range (Table S3, supplemental information). Although most of the recovery plans considered a viable population to be one consisting of “sufficient” genetic variation for the “foreseeable” future, none of the recovery plans established a quantitative measure of genetic variation or a time frame for their recovery criteria. None of the peer-reviewed literature or technical reports contained PVAs or genetic assessments conducted prior to the approval of recovery plans. Studies of phylogenetic reconstruction and genetic-based species delimitation were relatively common across the 90 listed species. However, these studies solely examined systematics and taxonomic positions of the listed species and often did not assess intraspecific genetic variability; thus, we excluded these from our review. Among the listed species, only five species have been assessed with some sort of demographic PVA and 13 species have had intraspecific genetic diversity and population genetic structure assessed (Table S4, Supplemental information). Overall, all PVA studies consistently incorporated demographic stochasticity into the models and reported stochastic ,

136 the results of sensitivity analysis, and estimates of the probability of extinction. Among the 13 species with genetic assessments, mitochondrial DNA gene (mtDNA) sequences were the most used genetic markers (11 species) followed by microsatellite loci (eight species) and nuclear DNA gene sequences (five species). These studies assessed population genetic diversity within populations (13 species), population genetic structure among populations (10 species), demographic history (e.g., bottleneck events; six species), and effective population size (five species). We found that only two species have been assessed for both demographic and genetic trajectories in the same study (Jones et al. 2012); however, the study simulated demographic and genetic models independently, and thus, the genetic trajectory was not integrated into PVA models. Case studies PVA models with uncertain parameters The number of juveniles per female and juvenile mortality rate interacted to influence the probability of extinction and mean  over the simulated period in both species (Figures 1 and 2). Both parameters strongly affected the probability of extinction. Regardless of the species, the simulations showed acute transition from complete extinction to nonextinction within narrow ranges of the parameters (Figure 1). Although there were slight differences between the species, populations would likely be maintained when the number of juveniles per female was greater than 2 and juvenile mortality rates were less than 97% over the next 300 years. The mean  was relatively uniformly distributed across the range of parameters (Figure 2); the mean  ranged from 0.871 to 1.037 for C. monodonta and from 0.815 to 1.141 for P. popeii across all simulations. Stable populations ( = 1) were achieved when the number of juveniles per female was greater than 3 and juvenile mortality rate was less than 95% for C. monodonta, while the number of juveniles per female was greater than 2 and juvenile mortality rate was less than 97% for P. popeii (Figure 2). While P. popeii populations were only maintained (probability of extinction < 0.1) by increasing populations, C. monodonta populations could be maintained by slightly declining populations ( > 0.98) over the 300 years (Figures 1 and 2). Sensitivity analyses Logistic regressions of demographic parameters showed similar standardized coefficient values among extinction and two quasi-extinction scenarios (Table 2). Relative sensitivity analyses showed that the most important demographic parameters (absolute z-values > 10) were

137 the number of juveniles per female and juvenile mortality rate for both species (Table 2). In addition, adult mortality rate was also an important parameter for C. monodonta, but it was of intermediate importance for P. popeii (absolute z-values 4 – 10). Young mortality rate and percentage of females in breeding pool were of intermediate importance for C. monodonta, but not for P. popeii. The percentage of males in the breeding pool, the effect of inbreeding depression, and the initial population size were the least important demographic parameters for both species. The effect of demography on population genetics The mean  and initial population sizes interacted to influence the persistence of genetic diversity over the simulated periods in both species (Figures 3 and 4). Overall, larger population size with a stable or increasing population ( ≥ 1) would retain most of the initial genetic diversity over 300 years. Negative population growth caused high fluctuations in the probability of genetic persistence regardless of population size, indicating that declining populations were more susceptible to stochasticity events. The patterns and degrees of genetic diversity persistence were different between the genetic diversity indices. Mean expected heterozygosity could be achieved greater than 90% of its original level even in slightly declining population with moderate population size ( > 0.99; population size > 500). Mean allelic richness, on the other hand, was more susceptible to the changes in the mean  and initial population sizes (Figures 3 and 4). The persistence of mean allelic richness was varied between populations; one population of each of the species could not achieve greater than 90% of their original levels of allelic richness during our simulations (Figures 3a and 4c). However, these patterns were not associated with the original levels of genetic diversity in either population. For example, the upper Mississippi population of C. monodonta (Figure 3a and 3b) had higher original levels of genetic diversity (AR = 17.38; HE = 0.863) than the lower Mississippi population (Figure 3c and 3d; AR =

10.79, HE = 0.771), while the Black River population of P. popeii (Figure 4c and 4d) had smaller original levels of genetic diversity (AR = 4.81, HE = 0.528) than the Rio Grande population

(Figure 4a and 4b; AR = 16.83, HE = 0.715).

138 DISCUSSION Current trends in recovery planning for freshwater mussels The primary goal of an endangered species recovery plan is identification of scientifically objective, measurable recovery criteria and management actions for the species that will decrease the threat of extinction and ultimately, result in removal of the species from protection under the ESA. We found that a majority of recovery plans for listed mussel species included the concept of genetic PVA by stating that goals included maintenance or restoration of viable populations with sufficient genetic variation for the foreseeable future. These recovery plans often established recovery criteria, such as reestablishment of a target number of viable populations and active dispersal between multiple drainages, in order to downlist/delist from the ESA. However, such recovery criteria were rather arbitrary; they rarely defined “viable population,” “sufficient genetic variation,” and “foreseeable future” in quantitative recovery goals. Our review also revealed that the recovery plans rarely implemented PVA and genetic assessments into recovery strategies; the implementation of PVA is at a much lower rate than that for listed plant species (Zeigler et al. 2013). Failure to implement quantitative strategies leads to uncertainty in achieving recovery goals. The use of PVA and genetic assessment in recovery planning for endangered mussel species has been recommended in order to prioritize recovery strategies (Jones et al. 2006; Berg et al. 2008). The major obstacle to development of genetic PVAs for freshwater mussels is a lack of knowledge of demographic and genetic parameters. Because of the complex life cycle and long life span of these organisms, the acquisition of demographic parameters often requires long- term monitoring and experiments (e.g., Inoue et al. 2014a). In fact, most of the listed species lack such information; thus, recovery strategies often prioritize life history research to include such factors as reproduction, age and growth, and population dynamics (e.g., USFWS 2010). To-date, five listed species (excluding C. monodonta) had PVA models developed for extinction risk assessments, whereas only a single study implemented both demographic and genetic assessments in the context of PVA (Jones et al. 2012). Interestingly, genetic assessment studies were more pervasive; these studies focused on characterizing current population genetic structure and reconstructions of historic demography. Datasets from such studies might be utilized in genetic PVAs to investigate population genetic persistence and changes in genetic structure into the future. Based on our review, we recognized the deficiency in demographic and genetic

139 models in the recovery strategies. Below, we use the results of the present modeling to make recommendations to implement genetic PVAs in recovery strategies and in the establishment of quantitative recovery criteria. Uncertainty and sensitivity in PVA Our PVA models and sensitivity analyses for C. monodonta and P. popeii revealed that demographic parameters for the early life stages highly influenced population persistence and population growth rates of the species (Table 2; Figures 1 and 2). We found that narrow ranges of demographic parameters caused acute transition from complete extinction to nonextinction, and the probability of extinction corresponded to  in both species. Given that P. popeii requires increasing populations ( ≥ 1), while C. monodonta requires at least slightly declining populations to maintain population persistence over 300 years, duration of the simulations likely caused these differences. The mean generation times simulated from Vortex are 27 years for C. monodonta and 17 years for P. popeii. Given the low adult mortality and long lifespan for C. monodonta, remaining adults likely maintain population persistence in declining populations. Furthermore, these early-life-stage parameters happened to be the most uncertain parameters within the mussel life cycle. Although research on the transition from glochidia to metamorphosis is actively conducted due to the need for captive propagation programs, data on survivorship from metamorphosis to settlement are lacking (Berg et al. 2008). For example, even the best-studied endangered species, M. margaritifera, has limited information for the natural survival rate during these early life stages (Schmidt & Vandré 2010). Sources of uncertainty, such as poor quality of data and difficulties in demographic parameter estimations, affect reliability of PVA predictions (Beissinger & Westphal 1998). To avoid using fixed values for uncertain parameter estimates, we created distributions of the uncertain parameters in the simulations. We were able to observe the interactive effects of varied uncertain parameters on the probability of extinction and population growth rates. In addition to these parameters, population persistence was highly sensitive to adult mortality rates for C. monodonta, but not for P. popeii. As a result, P. popeii is likely to be more immune to low number of juveniles-per-female and high juvenile mortality rates (Figures 1 and 2). Previous PVA studies with sensitivity analyses indicate that the importance of demographic parameters can vary between species. For example, population persistence of Quadrula fragosa was highly sensitive to changes in female breeding success and juvenile mortality rates (Kjos et

140 al. 1998), whereas adult mortality rates were primary demographic factors for Amblema plicata and A. neislerii (Hart et al. 2001; Miller 2011). These results suggest that species-specific PVA models are required to evaluate population persistence and population growth rates. Quantitative recovery criteria for C. monodonta and P. popeii Although most recovery plans consider the maintenance of viable populations as their recovery goal, the definition of a viable population is rather vague. Thus, we modified the definition and considered a viable population to be a stable or increasing population ( ≥ 1) that retains >90% of genetic variation for ten generations. Based on the results of our PVA models, we delineate quantitative, measureable criteria for these species. Cumberlandia monodonta requires a minimum of five juveniles per female produced annually, with juvenile mortality rate of less than 93% to achieve a stable or increasing population (Figure 2). For P. popeii, a minimum of four juveniles per female are required, with a juvenile mortality rate of less than 95% (Figure 2). To maintain >90% of current genetic variation (measured as heterozygosity) over the next 300 years, both C. monodonta and P. popeii require at least 500 individuals with  ≥ 1 (Figures 3 and 4). However, defining quantitative criteria for the persistence of allelic richness is more difficult due to high inter-population variability and complex interaction between population size, , and initial allelic richness. PVA with genetic factors Genetic persistence is not often considered in PVAs because demographic and environmental stochasticities are thought to drive small populations to extinction before genetic factors become important (Lande 1988). Theoretical and empirical evidence, however, has found that genetic factors can contribute to extinction via inbreeding depression, loss of genetic diversity due to genetic drift and fixation of deleterious mutations that reduce fitness, and reduced evolutionary potential (Frankham et al. 2014). Inbreeding and loss of genetic diversity are especially inevitable in small, closed populations through genetic drift; endangered and threatened species are often already experiencing these conditions (Spielman et al. 2004). In addition to population size, our genetic PVA models suggested that initial allele frequencies and population growth rates highly influence the trajectories of population genetic persistence (Figures 3 and 4). We also found that stochasticity is greater in declining populations, causing high variability in the probability of genetic persistence (Figures 3 and 4). Population growth rates and population sizes were positively associated with population genetic persistence;

141 however, population growth rates had a strong effect on future genetic persistence. Traditionally, genetic management for endangered species recommends an effective population size (Ne) of at least 50 to avoid short-term inbreeding depression and Ne of 500 to retain evolutionary potential for the foreseeable future (50/500 rule; Franklin 1980; Soulé 1980). The 50/500 rule is often referred as “a rule of thumb” for genetic management in recovery plans of freshwater mussels and in recent PVA models (Jones et al. 2012). However, recent debates question the application of the 50/500 rule as a threshold of genetic management (e.g., Jamieson & Allendorf 2012; Frankham et al. 2013). Our genetic PVA models avoided this generalization. Instead, we simulated persistence of genetic diversity over time by simulating interactive effects of population growth rates and population sizes on genetic trajectories. Importance of PVA and genetic assessment in recovery plans for freshwater mussels Captive propagation followed by reintroduction and augmentation has been the recovery strategy of choice for mussel conservation (Haag & Williams 2013); nearly all of the recovery plans recommend establishment of propagation techniques and reestablishment of new populations with propagated juvenile mussels. While propagation programs provide individuals for establishment of new populations and enhance existing populations to facilitate recovery, the recent increased use of propagation raises serious concerns about genetic and ecological risks. Genetic issues including reduction of genetic variation, mixture of genetic stocks, and exposure to novel adaptation under propagating environments have been addressed (USFWS & NMFS 2000; Jones et al. 2006); however, awareness of these issues is rarely evident in recovery planning. We recommend using genetic PVAs to predict the consequences of introduction of propagated individuals into natural environments. Furthermore, we suggest that recovery planning should include detailed genetic management efforts and assessment of risk associated with introduction of propagated individuals. The ultimate goal of an endangered species recovery plan is to recover the species to the point at which protection under the ESA is no longer necessary. Population viability analysis integrated with genetic assessment becomes an important tool for identifying recovery strategies and criteria, for establishing quantitative goals of recovery, and for evaluating management strategies. Because PVAs are probabilistic models and uncertainty is an inherent feature of them, PVA predictions must be carefully evaluated in recovery planning and decision making. Our genetic PVAs effectively demonstrate that, when sufficient data exist, quantitative recovery

142 criteria based on demographic and genetic data can create objective, measureable goals for recovery plans. The importance for the long-term conservation of genetic diversity is evident in order to maintain evolutionary adaptability (Lesica & Allendorf 1995). Given current rates of environmental changes due to anthropogenic activities, high levels of overall genetic diversity and large effective population sizes increase chances of adapting to such change within a few generations (Petit & Hampe 2006). We hope that this study will encourage other researchers to implement ecological and genetic modeling to develop this highly imperiled group of animals.

ACKNOWLEDGMENTS We thank M. Youngquist, T. Crist, B. Cochrane, B. Keane, and R. Moore for their helpful comments on earlier drafts of this manuscript. Funding was provided by the New Mexico Department of Game and Fish, the Missouri Department of Conservation, and the US Fish and Wildlife Service.

REFERENCES Allendorf FW, Hohenlohe PA, Luikart G (2010) Genomics and the future of conservation genetics. Nature Reviews Genetics, 11, 697-709. Baird MS (2000) Life history of the spectaclecase, Cumberlandia monodonta Say, 1829 (Bivalvia, Unionoidea, Margaritigeridae) M.S. thesis, Southwest Missouri State University, Springfield, Missouri. Beissinger SR, Westphal MI (1998) On the use of demographic models of population viability in endangered species management. Journal of Wildlife Management, 62, 821-841. Berg DJ, Levine TD, Stoeckel JA, Lang BK (2008) A conceptual model linking demography and population genetics of freshwater mussels. Journal of the North American Benthological Society, 27, 395-408. Carman SM (2007) Texas hornshell Popenaias popeii recovery plan, p. 57. New Mexico Department of Game and Fish, Santa Fe, NM. Carroll C, Fredrickson RJ, Lacy RC (2014) Developing metapopulation connectivity criteria from genetic and habitat data to recover the endangered Mexican wolf. Conservation Biology, 28, 76-86.

143 Cross PC, Beissinger SR (2001) Using logistic regression to analyze the sensitivity of PVA models: a comparison of methods based on African wild dog models. Conservation Biology, 15, 1335-1346. Frankham R, Bradshaw CJA, Brook BW (2014) Genetics in conservation management: Revised recommendations for the 50/500 rules, Red List criteria and population viability analyses. Biological Conservation, 170, 56-63. Frankham R, Brook BW, Bradshaw CJ, Traill LW, Spielman D (2013) 50/500 rule and minimum viable populations: response to Jamieson and Allendorf. Trends in Ecology and Evolution, 28, 187–188. Franklin IR (1980) Evolutionary change in small populations. In: Conservation Biology: An Evolutionary-ecological Perspective (eds. Soulé ME, Wilcox BA), pp. 135-149. Sinauer Associates Inc., Sunderland, MA. Gordon ME, Smith DG (1990) Autumnal reproduction in Cumberlandia monodonta (Unionoidea: Margaritiferidae). Transactions of the American Microscopical Society, 109, 407-411. Haag WR (2002) Spatial, temporal, and taxonomic variation in population dynamics and community structure of freshwater mussels, University of Mississippi. Haag WR (2012) North American Freshwater Mussels: Natural History, Ecology, and Conservation Cambridge University Press, Cambridge, UK. Haag WR, Williams JD (2013) Biodiversity on the brink: an assessment of conservation strategies for North American freshwater mussels. Hydrobiologia, 735, 45-60. Hart RA, Grier JW, Miller AC (2004) Simulation models of harvested and zebra mussel colonized threeridhe mussel populations in the upper Mississippi River. American Midland Naturalist, 151, 301-317. Hart RA, Grier JW, Miller AC, Davis M (2001) Empirically derived survival rates of a native mussel, Amblema plicata, in the Mississippi and Otter Tail Rivers, Minnesota. American Midland Naturalist, 146, 254-263. Hoban S, Bertorelle G, Gaggiotti OE (2011) Computer simulations: tools for population and evolutionary genetics. Nature Reviews Genetics, 13, 110-122. Inoue K, Lang BK, Berg DJ (2015) Past climate change drives current genetic structure of an endangered freshwater mussel species. Molecular Ecology, 24, 1910-1926.

144 Inoue K, Levine TD, Lang BK, Berg DJ (2014a) Long-term mark-and-recapture study of a freshwater mussel reveals patterns of habitat use and an association between survival and river discharge. Freshwater Biology, 59, 1872-1883. Inoue K, Monroe EM, Elderkin CL, Berg DJ (2014b) Phylogeographic and population genetic analyses reveal Pleistocene isolation followed by high gene flow in a wide-ranging, but endangered, freshwater mussel. Heredity, 112, 282-290. Jamieson IG, Allendorf FW (2012) How does the 50/500 rule apply to MVPs? Trends in Ecology and Evolution, 27, 578-584. Jones JW, Hallerman EM, Neves RJ (2006) Genetic management guidelines for captive propagation of freshwater mussels (Unionoidea). Journal of Shellfish Research, 25, 527- 535. Jones JW, Neves RJ, Hallerman EM (2012) Population performance criteria to evaluate reintroduction and recovery of two endangered mussel species, Epioblasma brevidens and Epioblasma capsaeformis (Bivalvia: Unionidae). Walkerana, 15, 27-44. Kjos C, Byers O, Miller P, Borovansky J, Seal US (1998) Population and habitat viability assessment workshop for the winged mapleleaf mussel (Quadrula fragosa): Final Report, p. 92, CBSG, Apple Valley, MN. Lande R (1988) Genetic and demography in biological conservation. Science, 241, 1455-1460. Lesica P, Allendorf FW (1995) When are peripheral population valuable for conservation? Conservation Biology, 9, 753-760. Levine TD, Lang BK, Berg DJ (2012) Physiological and ecological hosts of Popenaias popeii (Bivalvia: Unionidae): laboratory studies identify more hosts than field studies. Freshwater Biology, 57, 1854-1864. Lydeard C, Cowie RH, Ponder WF, Bogan AE, Bouchet P, Clark SA, Cummings KS, Frest TJ, Gargominy O, Herbert DG, Hershler R, Perez KE, Roth B, Seddon M, Strog EE, Thompson FG (2004) The global decline of nonmarine mollusks. BioScience, 54, 321- 330. Megens ES (2000) Population viability analyses in plants: challenges and opportunities. Trends in Ecology and Evolution, 15, 51-56.

145 Miller PS (2011) Revised population viability analysis for the fat threeridge mussel (Amblema neislerii). In: Final Report, p. 23. IUCN/SSC Conservation Breeding Specialist Group, Apple Valley, MN. Mosley TL, Haag WR, Stoeckel JA (2014) Egg fertilisation in a freshwater mussel: effects of distance, flow and male density. Freshwater Biology, 59, 2137-2149. Naujokaitis-Lewis IR, Curtis JMR, Arcese P, Rosenfeld J (2009) Sensitivity analyses of spatial population viability analysis models for species at risk and habitat conservation planning. Conservation Biology, 23, 225-229. Petit RJ, Hampe A (2006) Some evolutionary consequences of being a tree. Annual Review of Ecology, Evolution, and Systematics, 37, 187-214. Preston SJ, Keys A, Roberts D (2007) Culturing freshwater pearl musselMargaritifera margaritifera: a breakthrough in the conservation of an endangered species. Aquatic Conservation: Marine and Freshwater Ecosystems, 17, 539-549. Reed JM, Mills LS, Dunning Jr. JB, Menges ES, McKelvey KS, Frye R, Beissinger SR, Anstett M-C, Miller P (2002) Emerging issues in population viability analysis. Conservation Biology, 16, 7-19. Schmidt C, Vandré R (2010) Ten years of experience in the rearing of young freshwater pearl mussels (Margaritifera margaritifera). Aquatic Conservation: Marine and Freshwater Ecosystems, 20, 735-747. Smith DG, Lang BK, Gordon ME (2003) Gametogenetic cycle, reproductive anatomy, and larval morphology of Popenaias popeii (Unionoida) from the Black River, New Mexico. Southwestern Naturalist, 48, 333-340. Soulé ME (1980) Thresholds for survival: maintaining fitness and evolutionary potential. In: Conservation Biology: An Evolutionary-ecological Perspective (eds. Soulé ME, Wilcox BA), pp. 151-170. Sinauer Associates Inc., Sunderland, MA. Spielman D, Brook BW, Frankham R (2004) Most species are not driven to extinction before genetic factors impact them. Proceedings of the National Academy of Sciences of the United States of America, 101, 15261-15264. USFWS (US Fish and Wildlife Service). 2010. Scaleshell mussel recovery plan (Leptodea leptodon). US Fish and Wildlife Service, Fort Snelling, MN. 118 pp.

146 USFWS (US Fish and Wildlife Service). 2012. Endangered and threatened wildlife and plants; determination of endangered status for the sheepnose and spectaclecase mussels throughout their range, final rule. Federal Register 77, 14914-14949. USFWS (US Fish and Wildlife Service). 2014. Endangered and threatened wildlife and plants; review of native species that are candidates for listing as endangered or threatened; annual notice of findings on resubmitted petitions; annual description of progress on listing actions; proposed rule. Federal Register 79, 72450-72497. USFWS & NMFS (US Fish and Wildlife Service and National Marine Fisheries Service). 2000. Policy regarding controlled propagation of species listed under the Endangered Species Act. Federal Register 65, 56916-56922. Vrijenhoek RC (1994) Genetic diversity and fitness in small populations. In: Conservation Genetics (eds. Tomiuk J, Jain SK), pp. 37-53. Birkhäuser Verlag, Basel, Switzerland. Williams JD, Bogan AE, Garner JT (2008) Freshwater mussels of Alabama and the Mobile Basin in Georgia, Mississippi & Tennessee The University of Alabama Press, Tuscaloosa, AL. Young M, Williams J (1984) The reproductive biology of the freshwater pearl mussel Margaritifera margaritifera (Linn.) in Scotland I. Field studies. Archiv für Hydrobiologie, 99, 405-422. Zeigler SL, Che-Castaldo JP, Neel MC (2013) Actual and potential use of population viability analyses in recovery of plant species listed under the U.S. Endangered Species Act. Conservation Biology, 27, 1265-1278.

147 TABLES AND FIGURES Table 1. Lists of base-line demographic parameters for Cumberlandia monodonta and Popenaias popeii used in Vortex simulations. Standard deviations are in parentheses. Cumberlandia Popenaias Demographic parameters monodonta popeii Citations Age of first offspring for female 10 5 Baird 2000; Personal obs Age of first offspring for male 10 5 Baird 2000; Personal obs Maximum age of female reproduction 56 30 Baird 2000; Personal obs Maximum age of male reproduction 56 30 Baird 2000; Personal obs Maximum lifespan 56 30 Baird 2000; Personal obs Sex ratio at birth 50:50 50:50 Baird 2000; Personal obs % Adult female breeding 70 (25) 80 (25) Baird 2000; Personal obs Number of juveniles per female 4 (25) 4.6 (25) Haag 2002; Haag 2012 Juvenile mortality rate (settled juvenile to Age 1) 93 (5) 95 (5) Young & Williams 1984; Hart et al. 2004 Annual young mortality rate (Age 2 to Age 3) 15 (5) 15 (5) Haag 2012 Annual adult mortality rate (>Age 3) 5 (2) 2 (2) Baird 2000; Inoue et al. 2014 % Males in the breeding pool 80 (25) 80 (25) Initial population size 3000 1000 Baird 2000; Personal obs Carrying capacity 5000 4000

148 Table 2. Results of relative sensitivity analyses using standardized coefficients (z-values) from logistic regression of demographic parameter sets against probability of extinction and quasi-extinction. QE (250) = quasi-extinction at N = 250 individuals; QE (500) = quasi-extinction at N = 500 individuals. Z-value Parameter Minimum Estimate Maximum Extinction QE (250) QE (500) Cumberlandia monodonta Juvenile mortality 90 93 99 12.52 10.74 10.53 Annual young mortality 10 15 20 4.07 4.51 3.34 Annual adult mortality 2 5 8 11.31 10.45 10.10 Percentage of female in breeding pool 60 70 100 -4.75 -6.09 -5.82 Percentage of male in breeding pool 60 80 100 -0.22 -0.25 -0.10 Number of juveniles per female 0.1 4 7.2 -12.81 -10.90 -10.70 Effect of inbreeding depression 4.72 6.29 7.86 0.00 2.07 0.34 Initial population size 2000 3000 4000 -0.74 -1.50 0.59 Popenaias popeii Juvenile mortality 90 95 99 13.65 13.26 13.16 Annual young mortality 10 15 20 1.68 1.09 2.63 Annual adult mortality 1 2 5 6.37 4.48 4.85 Percentage of female in breeding pool 60 70 100 -2.87 -3.69 -3.21 Percentage of male in breeding pool 60 80 100 -0.75 1.49 -1.58 Number of juveniles per female 0.1 4.6 12 -13.70 -13.43 -13.43 Effect of inbreeding depression 4.72 6.29 7.86 0.41 -0.57 0.53 Initial population size 500 1000 1500 -1.12 0.21 -1.09

149

Figure 1. Heat map showing the probability of extinction over 300 years relative to the number of juveniles-per-female and juvenile mortality rate based on Vortex population simulations of Cumberlandia monodonta (a) and Popenaias popeii (b). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles-per-female and juvenile mortality rate, with 100 replicate simulations of 300 years each. Note that the scale on x-axis varies between species.

150

Figure 2. Heat map showing finite rate of population growth () relative to number of juveniles- per-female and juvenile mortality rate based on Vortex population simulation of Cumberlandia monodonta (a) and Popenaias popeii (b). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles-per-female and juvenile mortality rate, with 100 replicate simulations of 300 years each. Note that the scale on x-axis varies between species.

151

Figure 3. Heat map showing the probability of genetic diversity persistence (a and c: mean allelic richness, AR; b and d: mean expected heterozygosity, HE) relative to initial population size and mean finite rate of population growth () based on Vortex population simulation of two populations of Cumberlandia monodonta: upper Mississippi population (a and b) and lower Mississippi population (c and d). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles per female and initial population size, with 100 replicate runs per scenario.

152

Figure 4. Heat map showing the probability of genetic diversity persistence (a and c: mean allelic richness, AR; b and d: mean expected heterozygosity, HE) relative to initial population size and mean finite rate of population growth () based on Vortex population simulation of two populations of Popenaias popeii: Rio Grande population (a and b) and Black River population (c and d). Simulations are based on 1000 scenarios derived from randomized combinations of number of juveniles per female and initial population size, with 100 replicate runs per scenario.

153 SUPPLEMENTARY INFORMATION Table S1. Recovery goals in recovery plans that were approved by the US Fish and Wildlife Service for 71 mussel species listed under the US Endangered Species Act. Recovery goal Number of speciesa To remove the species from the Federal list of endangered and threatened 54 species To maintain, enhance, and restore viable populations of the species to a 42 significant portion of its historic range To prevent the extinction of the species by protecting its remaining range 13 To locate, maintain, and enhance any known populations of the species 3 aNumbers do not sum to 71 recovery plans because some species have more than one criterion

154 Table S2. Definitions of “a viable population” mentioned in 45 recovery plans that were approved by the US Fish and Wildlife Service. Definition of "a viable population" Number of speciesa A naturally reproducing population 45 A population large enough to maintain sufficient genetic variation 40 A stable or increasing population 6 A population sustained without immigration from other populations 3 aNumbers do not sum to 45 recovery plans because some species have more than one criterion

155 Table S3. Quantitative downlisting/delisting criteria in 55 recovery plans that were approved by the US Fish and Wildlife Service. Quantitative viability criterion Number of speciesa Recent recruitment 7 A given number of populations that ensure viability 41 Reestablishment/discoveries of new populations 44 Protected from present and foreseeable anthropogenic and natural events 21 With characteristics that ensure viability over a given period 3 Population connectivity 13 aNumbers do not sum to 55 recovery plans because some species have more than one criterion

156 Table S4. Use of population viability analysis (PVA) and genetic assessments (Genetic) for mussel species listed under the US Endangered Species Act. Assessments Species PVA reference Genetic reference PVA Amblema neislerii Miller (2001) n/a Genetic Arcidens wheeleri n/a Inoue et al. (2014a) PVA/Genetic Cumberlandia monodonta Present study Inoue et al. (2014b) Present study Genetic Cyprogenia stegaria n/a Grobler et al. (2011) PVA/Genetic Epioblasma brevidens Jones et al. (2012) Jones et al. (2012) Jones et al. (2015) PVA/Genetic Epioblasma capsaeformis Jones et al. (2012) Jones et al. (2006) Jones et al. (2012) Jones et al. (2015) Genetic Epioblasma florentina walkeri n/a Jones et al. (2006) Genetic Epioblasma torulosa rangiana n/a Jones et al. (2006) Zanatta & Murphy (2007) Genetic Epioblasma triquetra n/a Zanatta & Murphy (2008) Genetic Lampsilis higginsii USACE (2002) Bowen (2004) Genetic Margaritifera hembeli n/a Curolé et al. (2004) PVA/Genetic Quadrula fragosa Kjos et al. (1998) Roe & Boyer (2015) Genetic Pleurobema collina n/a Petty (2005) Genetic Pleuronaia dolabelloides n/a Grobler et al. (2006) n/a = not available

REFERENCES Bowen, B. S. 2004. Genetic variability and geographic structure of Lampsilis higginsii mussels in upper Mississippi River and tributaries. Iowa State University, Ames, Iowa. Curolé, J. P., D. W. Foltz, and K. M. Brown. 2004. Extensive allozyme monomorphism in a threatened species of freshwater mussel, Margaritifera hembeli Conrad (Bivalvia: Margaritiferidae). Conservation Genetics 5:271-278. Grobler, P. J., J. W. Jones, N. A. Johnson, B. Beaty, J. Struthers, R. J. Neves, and E. M. Hallerman. 2006. Patterns of genetic differentiation and conservation of the slabside pearlymussel, Lexingtonia dolabelloides (Lea, 1840) in the Tennessee River drainage. Journal of Molluscan Studies 72:65-75. Grobler, J. P., J. W. Jones, N. A. Johnson, R. J. Neves, and E. M. Hallerman. 2011. Homogeneity at nuclear microsatellite loci masks mitochondrial haplotype diversity in the endangered fanshell pearlymussel (Cyprogenia stegaria). Journal of Heredity 102:196-206.

157 Inoue, K., A. L. McQueen, J. L. Harris, and D. J. Berg. 2014a. Molecular phylogenetics and morphological variation reveal recent speciation in freshwater mussels of the genera Arcidens and Arkansia (Bivalvia: Unionidae). Biological Journal of the Linnean Society 112:535-545. Inoue, K., E. M. Monroe, C. L. Elderkin, and D. J. Berg. 2014b. Phylogeographic and population genetic analyses reveal Pleistocene isolation followed by high gene flow in a wide- ranging, but endangered, freshwater mussel. Heredity 112:282-290. Jones, J. W., R. J. Neves, S. A. Ahlstedt, and E. M. Hallerman. 2006. A holistic approach to taxonomic evaluation of two closely related endangered freshwater mussel species, the oyster mussel Epioblasma capsaeformis and tan riffleshell Epioblasma florentina walkeri (Bivalvia: Unionidae). Journal of Molluscan Studies 72:267-283. Jones, J. W., R. J. Neves, and E. M. Hallerman. 2012. Population performance criteria to evaluate reintroduction and recovery of two endangered mussel species, Epioblasma brevidens and Epioblasma capsaeformis (Bivalvia: Unionidae). Walkerana 15:27-44. Jones, J. W., R. J. Neves, and E. M. Hallerman. 2015. Historical demography of freshwater mussels (Bivalvia: Unionidae): genetic evidence for population expansion and contraction during the late Pleistocene and Holocene. Biological Journal of the Linnean Society 114:376–397. Kjos C, Byers O, Miller P, Borovansky J, Seal US (1998) Population and habitat viability assessment workshop for the winged mapleleaf mussel (Quadrula fragosa): Final Report, CBSG, Apple Valley, MN. Miller, P. S. 2011. Revised population viability analysis for the fat threeridge mussel (Amblema neislerii). Final Report. IUCN/SSC Conservation Breeding Specialist Group, Apple Valley, MN. 23 pp. Petty, M. A. 2005. Distribution, genetic characterization, and life history of the James spinymussel, Pleurobema collina (Bivalvia: Unionidae), in Virginia and North Carolina. Fisheries and Wildlife Science. Virginia Polytechnic Institute and State University, Blacksburg, Virginia. Roe, K. J., and S. L. Boyer. 2015. A comparison of genetic diversity between symatric populations of the endangered winged-mapleleaf (Quadrula fragosa) and the pimpleback (Amphinaias pustulosa) in the St. Croix, USA. American Malacological Bulletin 33:1-9.

158 USACE (US Army Corps of Engineers). 2002. Definite project report and environmental assessment for relocation plan for the endangered Higgins' eye pearlymussel (Lampsilis higginsii), upper Mississippi River and tributaries Minnesota, Wisconsin, Iowa, and Illinois. US Army Corps of Engineers, St. Paul, MN. 122 pp. Zanatta, D. T., and R. W. Murphy. 2007. Range-wide population genetic analysis of the endangered northern riffleshell mussel, Epioblasma torulosa rangiana (Bivalvia: Unionoida). Conservation Genetics 8:1393-1404. Zanatta, D. T., and R. W. Murphy. 2008. The phylogeographical and management implications of genetic population structure in the imperiled snuffbox mussel, Epioblasma triquetra (Bivalvia: Unionidae). Biological Journal of the Linnean Society 94:371-384.

159 Summary and General Conclusion

Conservation of endangered species requires the available information on biology, status, and threats for a species to develop recovery strategies that include scientifically objective, measureable recovery criteria and management actions that mitigate threats (Neel et al. 2012). My dissertation used a comprehensive approach from population genetics to extinction risk assessments to elucidate 1) evolutionary history as a consequence of past environmental changes; 2) contemporary population dynamics and demography; and 3) trajectories of species’ distribution, genetic variation, and demography in the foreseeable future. Furthermore, by comparing two endangered freshwater mussel species, I was able to characterize features unique to each species. The assessment of population genetic structure often requires highly polymorphic genetic markers, such as microsatellites (Guichoux et al. 2011). I used two approaches to develop microsatellite markers: a conventional enriched library approach (Chapter 1) and a next- generation sequencing approach (Chapter 3). Both approaches provided a large number of polymorphic microsatellite markers for non-model organisms (17 loci for Cumberlandia monodonta and 20 loci for Popenaias popeii). However, using the next-generation sequencing approach, I was able to identify and design primers for hundreds of microsatellite loci; this approach promises to be cost- and time-effective relative to one using a commercial laboratory to build an enriched library for development of microsatellites. Species diversification and current population structure reflect historical and contemporary ecological and evolutionary forces, both of which are often associated with past climate and environmental changes (Hewitt 2000). Given current rates of human-induced environmental change, studies of evolutionary responses to past ecological events are especially timely and allow us to infer demographic and evolutionary trajectories in the foreseeable future. Recent development of advanced statistical methods has led to improved ability to reconstruct the ways in which past climate change and geological processes shaped existing species’ distributions and phylogeographic patterns (Knowles 2009). Furthermore, the use of forward- time genetic simulations and predicted future climate scenarios allows estimation of the persistence of genetic and demographic variability in the future under various environmental conditions (Hoban et al. 2011; Fordham et al. 2014). Using microsatellite loci developed as part of my dissertation, I revealed that climate change during the Pleistocene likely shaped current

160 distribution and population genetic structure in both species (Chapter 4; Inoue et al. 2014). A previous study suggested that the Pleistocene interglacial periods created two glacial refugia; post-glacial admixture of C. monodonta populations was then followed by simultaneous dispersal throughout the current species range (Inoue et al. 2014). Current genetic structure includes a panmictic population in the upper Mississippi River and a divergent population in the lower Mississippi River, which is likely a consequence of connectivity of suitable habitat (Chapter 2). However, future climate change is predicted to reduce connectivity across populations, leading to fragmentation of the panmictic population into several isolated populations. These results suggest that persistence and connectivity of stream-dwelling organisms will be significantly altered in response to future climate change. Although the Rio Grande region was not directly affected by the Pleistocene glaciation, climate change associated with this event still impacted current species distribution and population genetic structure of P. popeii (Chapter 4). Previously, extant populations were thought to be isolated due to loss of connecting habitats and inability of hosts to move through stretches of river with intermittent flow, inhospitable water conditions, and a series of impoundments (Carman 2007). However, my study revealed that regional population structure likely arose in this species during the mid-to-late Pleistocene and was followed by a late Pleistocene population bottleneck in the northern populations. My preliminary attempt to predict species distribution under future climate change indicated stable trends in suitable habitat over the next 50 years (Figure 1); however, current anthropogenic land- and water-uses threaten the availability of high-quality habitat for P. popeii. Over the past several decades, at least 30 springs have gone dry in the Rio Grande drainages (Contreras-Balderas & Lozano-Vilano 1994). Rising human demand for water will likely exacerbate intermittent river conditions. A long-term demographic study revealed that reduced river discharge is associated with significantly decreased survival of P. popeii (Chapter 5). Significant changes in climate and hydrological regimes, and increases in anthropogenic threats (increased water demand, degraded water quality) in the region, will induce significant declines in P. popeii populations. If the region’s human population continues to grow, intensifying anthropogenic threats, P. popeii will not persist even in suitable habitat. Population viability analysis (PVA) that integrates information from both population biology and conservation genetics has been recommended for recovery planning for freshwater

161 mussels (Jones et al. 2006; Berg et al. 2008). In Chapter 6, PVAs integrated with genetic assessments (genetic PVAs) revealed that demographic parameters during early life stages interact to influence the probability of extinction and vital rates of population growth for both species. The trajectories of population genetic persistence were influenced by initial allele frequency within populations, population size, and population growth rates. My genetic PVAs effectively demonstrate that, when sufficient data are available, quantitative recovery criteria based on demographic and genetic data can create objective, measureable goals for recovery plans. While freshwater ecosystems represent a small proportion of total ecosystems, they harbor a disproportionally high level of biodiversity (Dudgeon et al. 2006). Because these ecosystems are often closed systems in which ecological processes are carried out within the system (e.g., lakes, rivers), stream-dwelling organisms are often evolutionarily unique (Vörösmarty et al. 2010). However, freshwater ecosystems are among the most threatened ecosystems due to huge impact from anthropogenic activities. Because these ecosystems are fragile and recover slowly from perturbations, conservation of freshwater biodiversity is urgently needed. In my dissertation, I used freshwater mussels, the most endangered group of animals in the world (Lydeard et al. 2004; Strayer et al. 2004), to demonstrate the importance of considering evolution and ecology of organisms when developing species recovery strategies. Recovery plans need to incorporate knowledge of multiple aspects of the biology of target species, from demography to evolution when designing strategies for habitat assessment and making conservation decisions. I hope that this comprehensive approach to conservation biology of highly endangered species encourages others to utilize ecological and genetic studies in developing recovery plans for species from this and other highly imperiled groups.

REFERENCES Berg DJ, Levine TD, Stoeckel JA, Lang BK (2008) A conceptual model linking demography and population genetics of freshwater mussels. Journal of the North American Benthological Society, 27, 395-408. Carman SM (2007) Texas hornshell Popenaias popeii recovery plan, p. 57. New Mexico Department of Game and Fish, Santa Fe, NM.

162 Contreras-Balderas S, Lozano-Vilano ML (1994) Water, endangered fishes, and development perspectives in arid lands of Mexico. Conservation Biology, 8, 379-387. Dudgeon D, Arthington AH, Gessner MO, Kawabata Z-I, Knowler DJ, Leveque C, Naiman RJ, Prieur-Richard A-H, Soto D, Stiassny MLJ, Sullivan CA (2006) Freshwater biodiversity: importance, threats, status and conservation challenges. Biological Reviews, 81, 163-182. Fordham DA, Brook BW, Moritz C, Nogues-Bravo D (2014) Better forecasts of range dynamics using genetic data. Trends in Ecology & Evolution, 29, 436-443. Guichoux E, Lagache L, Wagner S, Chaumeil P, Léger P, Lepais O, Lepoittevin C, Malausa T, Revardel E, Salin F, Petit RJ (2011) Current trends in microsatellite genotyping. Molecular Ecology Resources, 11, 591-611. Hewitt G (2000) The genetic legacy of the Quanternary ice ages. Nature, 405, 907-913. Hoban S, Bertorelle G, Gaggiotti OE (2011) Computer simulations: tools for population and evolutionary genetics. Nature Reviews Genetics, 13, 110-122. Inoue K, Monroe EM, Elderkin CL, Berg DJ (2014) Phylogeographic and population genetic analyses reveal Pleistocene isolation followed by high gene flow in a wide-ranging, but endangered, freshwater mussel. Heredity, 112, 282-290. Jones JW, Hallerman EM, Neves RJ (2006) Genetic management guidelines for captive propagation of freshwater mussels (Unionoidea). Journal of Shellfish Research, 25, 527- 535. Knowles LL (2009) Statistical phylogeography. Annual Review of Ecology, Evolution and Systematics, 40, 593-612. Lydeard C, Cowie RH, Ponder WF, Bogan AE, Bouchet P, Clark SA, Cummings KS, Frest TJ, Gargominy O, Herbert DG, Hershler R, Perez KE, Roth B, Seddon M, Strog EE, Thompson FG (2004) The global decline of nonmarine mollusks. BioScience, 54, 321- 330. Neel MC, Leidner AK, Haines A, Goble DD, Scott JM (2012) By the numbers: how is recovery defined by the US Endangered Species Act? BioScience, 62, 646-657. Strayer DL, Downing JA, Haag WR, King TL, Layzer JB, Newton TJ, Nichols SJ (2004) Changing perspectives on pearly mussels, North America's most imperiled animals. BioScience, 54, 429-439.

163 Vörösmarty CJ, McIntyre PB, Gessner MO, Dudgeon D, Prusevich A, Green P, Glidden S, Bunn SE, Sullivan CA, Liermann CR, Davies PM (2010) Global threats to human water security and river biodiversity. Nature, 467, 555-561.

164

Figure 1. Potential distribution of Popenaias popeii identified using ecological niche modeling under current bioclimatic conditions (1950-2000) and projections of two future climate scenarios (low and high greenhouse gas concentrations in 2050). Models included major streams of the Rio Grande watershed in USA and Mexico. Black dots on the Present map represent occurrence points included in the ENMs. Scale bars in the bottom-left corners represent 200 km.

165