W&M ScholarWorks
Dissertations, Theses, and Masters Projects Theses, Dissertations, & Master Projects
Summer 2018
Evaluation of Population Structure and the Interspecific Relationship of Striped Marlin (Kajikia Audax) and White Marlin (K. Albida) Based on Traditional or Genome-Wide Molecular Markers
Nadya Mamoozadeh College of William and Mary - Virginia Institute of Marine Science, [email protected]
Follow this and additional works at: https://scholarworks.wm.edu/etd
Part of the Fresh Water Studies Commons, and the Oceanography Commons
Recommended Citation Mamoozadeh, Nadya, "Evaluation of Population Structure and the Interspecific Relationship of Striped Marlin (Kajikia Audax) and White Marlin (K. Albida) Based on Traditional or Genome-Wide Molecular Markers" (2018). Dissertations, Theses, and Masters Projects. Paper 1530192501. http://dx.doi.org/10.25773/v5-dj58-qc74
This Dissertation is brought to you for free and open access by the Theses, Dissertations, & Master Projects at W&M ScholarWorks. It has been accepted for inclusion in Dissertations, Theses, and Masters Projects by an authorized administrator of W&M ScholarWorks. For more information, please contact [email protected]. Evaluation of Population Structure and the Interspecific Relationship of Striped Marlin (Kajikia audax) and White Marlin (K. albida) Based on Traditional or Genome-wide Molecular Markers
A Dissertation
Presented to
The Faculty of the School of Marine Science
The College of William and Mary in Virginia
In Partial Fulfillment
of the Requirements for the Degree of
Doctor of Philosophy
by
Nadya R. Mamoozadeh
May 2018 APPROVAL PAGE
This dissertation is submitted in partial fulfillment of
the requirements for the degree of
Doctor of Philosophy
Nadya R. Mamoozadeh
Approved by the Committee, April 2018
John E. Graves, Ph.D. Committee Chair / Advisor
Jan R. McDowell, Ph.D. Committee Co-chair / Co-advisor
Bongkeun Song, Ph.D.
Michael A. Banks, Ph.D. Hatfield Marine Science Center, Oregon State University Newport, OR, USA
Bruce B. Collette, Ph.D. National Marine Fisheries Service, Smithsonian Institution Washington D.C., USA
TABLE OF CONTENTS
ACKNOWLEDGMENTS ...... viii
LIST OF TABLES ...... x
LIST OF FIGURES ...... xii
AUTHOR’S NOTE ...... xv
ABSTRACT ...... xvi
CHAPTER I—Introduction: Challenges Related to the Conservation and Management of
Istiophorid Billfishes ...... 2
Motivation ...... 3
Research approach ...... 6
Background information ...... 8
Population structuring in white marlin ...... 8
Population structuring in striped marlin ...... 11
Relationship of striped marlin and white marlin...... 13
Dissertation objectives and structure ...... 14
References ...... 17
CHAPTER II—Genetic Evaluation of Population Structure in White Marlin (Kajikia albida): The Importance of Statistical Power ...... 25
Abstract ...... 26
Introduction ...... 27
Materials and methods ...... 30
Sample collection and generation of nuclear genotype data ...... 30
Generation of mtDNA sequence data ...... 32
iii Analyses of microsatellite genotype data ...... 32
Analyses of mtDNA sequence data ...... 35
Results ...... 36
Summary statistics and genetic diversity ...... 36
Population structure inferred from genetic data...... 37
Statistical power ...... 38
Temporal replicates ...... 39
Discussion ...... 39
Acknowledgments...... 49
References ...... 50
Tables ...... 63
Figures...... 65
Supplementary materials ...... 69
CHAPTER III—Population Structure for Striped Marlin (Kajikia audax) in the Pacific and Indian Oceans Based on Genome-wide Single Nucleotide Polymorphisms ...... 76
Abstract ...... 77
Introduction ...... 78
Methods and materials ...... 82
Sample collection and DNA isolation...... 82
DArTseqTM 1.0 library preparation and sequencing ...... 83
DArTseqTM genotype calling ...... 84
SNP filtering ...... 85
Detection of outlier loci ...... 86
iv Temporal replicates ...... 87
Evaluation of diversity and differentiation by sampling location ...... 88
Organization of sample collections into populations ...... 89
Evaluation of diversity and differentiation by population ...... 91
Population connectivity ...... 91
Results ...... 92
SNP filtering ...... 92
Detection of outlier loci ...... 93
Temporal replicates ...... 93
Evaluation of diversity and differentiation by sampling location ...... 94
Organization of sample collections into populations ...... 95
Evaluation of diversity and differentiation by population ...... 101
Population connectivity ...... 102
Discussion ...... 103
Biological significance of statistically significant comparisons ...... 104
Biological support for genetically distinct populations ...... 105
Temporal stability of allele frequencies ...... 113
Maintenance of population structure ...... 114
Movements inferred via putative migrants ...... 115
Management implications and concluding remarks ...... 116
Acknowledgments...... 118
References ...... 120
Tables ...... 135
v Figures...... 143
Supplementary materials ...... 151
Appendix ...... 158
CHAPTER IV—Use of Genome-wide Molecular Markers for Evaluating the
Evolutionary Relationship of Two Highly Migratory Sister Species Inhabiting Distinct
Ocean Basins ...... 166
Abstract ...... 167
Introduction ...... 168
Materials and methods ...... 172
Sample collection and DNA preparation ...... 172
DArTseqTM 1.0 genotyping ...... 173
Bioinformatic analyses...... 174
Lineage delimitation ...... 175
Lineage characterization ...... 178
Phylogenetic inference ...... 179
Results ...... 179
Bioinformatic analyses...... 179
Lineage delimitation ...... 180
Lineage characterization ...... 183
Phylogenetic inference ...... 185
Discussion ...... 185
Evolutionary relationship of striped marlin and white marlin ...... 186
Significance of individuals with shared ancestry...... 192
vi Concluding remarks ...... 194
Acknowledgments...... 195
References ...... 196
Tables ...... 207
Figures...... 212
Supplementary materials ...... 218
CONCLUSIONS...... 224
Contributions...... 224
Significance...... 227
Future research directions ...... 228
References ...... 232
VITA ...... 236
vii ACKNOWLEDGMENTS
This dissertation represents a culmination of efforts by a number of individuals who have contributed to the success of this work over the past five years. To my co- advisors, Drs. John Graves and Jan McDowell, thank you for bringing me into the wonderful world of fisheries genetics. You’ve both allowed me to take the time to develop skills that will help me succeed in my future career. John, thank you for keeping me focused on the big picture, sharing valuable career advice, and spending countless hours guiding the development of my written work. Jan, thank you for the many discussions on data analysis, including the opportunity to learn beyond my own projects, and overall support throughout my program. I am also thankful for the support of my academic committee over the years. To Dr. Michael Banks, thank you for taking an interest in my work. In particular, your efforts to help me prepare for my qualifying exam and interact with your own graduate students are greatly appreciated. To Dr. Bruce Collette, thank you for sharing your enthusiasm for the istiophorid billfishes with me–your keen interest in the research questions addressed in my dissertation has been wonderful. To Dr. Bongkeun ‘BK’ Song, thank you for continuing to check in with me as I progressed through my labwork, and taking the time for detailed discussions of methodology early on. This dissertation really would not have been possible without help from many collaborators around the world. The personal efforts of Roy Bealey and the African Billfish Foundation were critical for sample collection. Sampling and outreach efforts by Jason Schratwieser and the International Game Fish Association, including a myriad of enthusiastic international representatives (Fouad Sahiaoui, Jose Luis Martinez Jenssen, Iain Richardson, and others), were also essential. Drs. Jerome Bourjea and Evgeny Romanov of the Indian Ocean Tuna Commission have been helpful from the start, as has Dr. Sofia Ortega-Garcia. Captains Trevor Hansen and Tim Richardson couldn’t have been more eager to support this work. Dr. Ken Neill, you’ve supported several projects in our lab, including mine, thank you for your help. There is a long list of others that contributed to this work (including Morgan O’Kennedy, Johan Smal, John Holdsworth, Sam Williams, Bob Kurz, Xavier Perez, Dr. Alberto Amorim, Dr. Jay Rooker, Dr. Catherine Purcell, Rick Weber and the Mid-Atlantic 500 Tournament, Estrella Malca and the SEFSC Early Life History Lab, Dr. Bob Humphreys and others at NMFS PIFSC)– know that your efforts are greatly appreciated. This work also would not have been possible without an array of both private and public support. The Virginia Sea Grant Graduate Research Fellowship program provided financial support and several outreach opportunities for my work. Financial support was also provided by the International Women’s Fishing Association, the Virginia Institute of Marine Science, and the Virginia Institute of Marine Science Foundation. Thank you all for valuing me and my work.
viii The VIMS community is a special asset of this program–thank you for your support over the past five years. To Heidi Brightman, you have provided unwavering guidance in the molecular lab, and I am thankful we also became friends along the way. To the VIMS IT and William & Mary HPC teams (especially Jay Kanukurthy and Adam Miller), thank you for the numerous program installs and troubleshooting help. Jen Hay, Cathy Cake, Cindy Forrester, Grace Tisdale, Heather Longest, Maxine Butler, thank you for your help with countless administrative questions and forms that I brought your way. Finally, to the VIMS student community–you guys are awesome. Making so many new friends has been a really enjoyable part of my time in this program. Lastly, to my family and friends that are like family, thank you for your persistent love and support over the years. To my mom Nancy and stepdad David, your belief in my success has been a motivating factor on many occasions. To my dad Abbas, thank you for teaching me to work hard and never settle. To my husband Steve, I can’t imagine a more supportive partner, thank you for helping me accomplish my goals, even though this particular goal meant four years of travelling on I-95.
ix LIST OF TABLES
Table Page
CHAPTER II—Genetic Evaluation of Population Structure in White Marlin (Kajikia albida): The Importance of Statistical Power
1. Summary information for mitochondrial DNA control region sequence data ...... 63
2. Matrix of pairwise FST values based on microsatellite data and pairwise фST values based on mitochondrial DNA sequence data ...... 64
S1. Summary statistics for the 24 microsatellite loci surveyed in this study ...... 69
S2. Collection information for temporal replicate sample collections ...... 71
S3. Observed and expected heterozygosities across sample collections ...... 72
S4. Pairwise Fst values for United States mid-Atlantic temporal replicates ...... 73
CHAPTER III—Population Structure for Striped Marlin (Kajikia audax) in the Pacific and Indian Oceans Based on Genome-wide Single Nucleotide Polymorphisms
1. Details for striped marlin sample collections...... 135
2. Number of single nucleotide polymorphism loci retained after each filtering step .....136
3. Results from analysis of molecular variance for scenarios with striped marlin sample collections organized by sampling location or by region ...... 137
4. Diversity metrics calculated for striped marlin sample collections organized by sampling location ...... 139
5. Pairwise FST values calculated between striped marlin sample collections organized by sampling location ...... 140
6. Diversity metrics calculated for striped marlin populations ...... 141
7. Pairwise FST values calculated between striped marlin populations ...... 142
S1. Diversity metrics calculated for striped marlin populations with the western Australia sample collection grouped separately ...... 151
S2. Pairwise FST values calculated between striped marlin populations with the western Australia sample collection grouped separately ...... 152
x CHAPTER IV—Use of Genome-wide Molecular Markers for Evaluating the Evolutionary Relationship of Two Highly Migratory Sister Species Inhabiting Distinct Ocean Basins
1. Sampling details for striped marlin and white marlin sample collections ...... 207
2. Number of single nucleotide polymorphism loci retained after each filtering step .....208
3. Results from the comparison of species delimitation scenarios using Bayes factor delimitation ...... 209
4. Pairwise FST and DS between populations of white marlin and striped marlin ...... 210
5. Diversity metrics for populations of white marlin and striped marlin ...... 211
S1. Pairwise distances between populations of white marlin and striped marlin ...... 218
S2. Percentage of loci exhibiting fixed differences between populations of white marlin and striped marlin ...... 219
xi LIST OF FIGURES
Figure Page
CHAPTER II—Genetic Evaluation of Population Structure in White Marlin (Kajikia albida): The Importance of Statistical Power
1. Geographic sampling locations for sample collections ...... 65
2. STRUCTURE results based on microsatellite genotype data ...... 66
3. Median joining haplotype network generated from mitochondrial DNA control region sequence data ...... 67
4. Results from power analyses based on empirical allele frequency data generated in the present study and in Graves & McDowell (2006) ...... 68
S1. Two-dimensional plot of axes one and two from principal component analysis performed using microsatellite genotype data ...... 74
S2. Locus-by-locus FST values associated with the pairwise comparison of Gulf of Mexico adult and larval sample collections ...... 75
CHAPTER III—Population Structure for Striped Marlin (Kajikia audax) in the Pacific and Indian Oceans Based on Genome-wide Single Nucleotide Polymorphisms
1. Map displaying geographic sampling locations and numbers of samples for striped marlin sample collections ...... 143
2. Two-dimensional plots of results from principal coordinate analysis and from discriminant analysis of principal components for temporal replicates sampled off Ecuador and eastern Australia ...... 144
3. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using dataset with all striped marlin sample collections ...... 145
4. Two-dimensional plot of axes one and two from principal coordinate analysis of striped marlin genotype data ...... 146
5. Two-dimensional plot of results for K = 4 scenario assessed with discriminant analysis of principal components for striped marlin genotype data ...... 147
6. Two-dimensional plot of results for K = 5 scenario assessed with discriminant analysis of principal components for striped marlin genotype data ...... 148
xii 7. Parameter estimates for mutation-scaled effective population sizes and effective migration rates estimated using MIGRATE-N ...... 149
8. World map displaying spatial distribution of striped marlin overlaid with regional fisheries management organization jurisdictions and genetically determined population structure...... 150
S1. Schematic displaying the processing of raw Illumina sequencing reads ...... 153
S2. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using dataset limited to striped marlin sample collections limited to the Pacific Ocean and western Australia ...... 154
S3. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using dataset limited to striped marlin sample collections limited to the Indian Ocean, eastern Australia, and New Zealand ...... 155
S4. Two-dimensional plot of axes one and three from principal coordinate analysis of striped marlin genotype data ...... 156
S5. Two-dimensional plot of axes two and three from principal coordinate analysis of striped marlin genotype data ...... 157
A1. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using neutral markers and markers putatively under selection and all striped marlin sample collections...... 158
A2. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using neutral markers and markers putatively under selection, and striped marlin sample collections limited to the Pacific Ocean and western Australia ...159
A3. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using neutral markers and markers putatively under selection, and striped marlin sample collections limited to the Indian Ocean, eastern Australia, and New Zealand ...... 160
A4. Two-dimensional plot of results for K = 4 scenario assessed with discriminant analysis of principal components for striped marlin genotype data including neutral markers and markers putatively under selection ...... 161
A5. Two-dimensional plot of results for K = 5 scenario assessed with discriminant analysis of principal components for striped marlin genotype data including neutral markers and markers putatively under selection ...... 162
A6. Two-dimensional plot of axes one and two from principal coordinate analysis of striped marlin genotype data including neutral markers and markers putatively under selection ...... 163
xiii A7. Two-dimensional plot of axes one and three from principal coordinate analysis of striped marlin genotype data including neutral markers and markers putatively under selection ...... 164
A8. Two-dimensional plot of axes one and three from principal coordinate analysis of striped marlin genotype data including neutral markers and markers putatively under selection ...... 165
CHAPTER IV—Use of Genome-wide Molecular Markers for Evaluating the Evolutionary Relationship of Two Highly Migratory Sister Species Inhabiting Distinct Ocean Basins
1. Two-dimensional plot of axes one and two from principal coordinate analysis of striped marlin and white marlin genotype data ...... 212
2. Two-dimensional plot of results for K = 2 scenario assessed with discriminant analysis of principal components for striped marlin and white marlin genotype data ...... 213
3. Barplot displaying results from fastSTRUCTURE simulations with K = 2...... 214
4. NetView network topology based on pairwise differences between individuals and k- nearest neighbor = 65 ...... 215
5. Unrooted neighbor-joining tree based on pairwise differences between individuals ..216
6. Species tree and maximum likelihood gene tree for striped marlin and white marlin 217
S1. Two-dimensional plot of axes one and two from principal coordinate analysis of striped marlin and white marlin genotype data ...... 220
S2. Two-dimensional plot of axes one and three from principal coordinate analysis of striped marlin and white marlin genotype data ...... 221
S3. Two-dimensional plot of axes two and three from principal coordinate analysis of striped marlin and white marlin genotype data ...... 222
S4. Unrooted maximum likelihood phylogeny inferred using RAxML ...... 223
xiv AUTHOR’S NOTE
Chapters II and III of this dissertation have been prepared as manuscripts for publication in peer-reviewed journals. As a result, these chapters were written in the third person to reflect the contributions of co-authors, and use the citation and formatting styles required by each journal. Chapter IV will not be submitted for publication in its current form, but will be prepared for publication in the near future.
The citations for Chapters II and III are:
Chapter II
Mamoozadeh, N. R., McDowell, J. R., Rooker, J. R., and Graves, J. E. 2018. Genetic evaluation of population structure in white marlin (Kajikia albida): the importance of statistical power. ICES Journal of Marine Science. doi: 10.1093/icesjms/fsx047
Chapter III
Mamoozadeh, N. R., McDowell, J. R., and Graves, J. E. In preparation. Population structure for striped marlin (Kajikia audax) in the Pacific and Indian oceans based on genome-wide single nucleotide polymorphisms.
xv ABSTRACT
The istiophorid billfishes (marlins, spearfishes, and sailfish) are highly migratory pelagic fishes exhibiting broad and continuous spatial distributions in the Atlantic, Pacific, and Indian oceans. These species are targeted by a number of recreational, commercial, artisanal, and subsistence fisheries worldwide, and are also caught as bycatch in pelagic longline fisheries targeting tunas and swordfish. Though stock assessments have not been conducted for all istiophorids, assessments available for some species indicate that many istiophorid stocks are overfished and/or experiencing overfishing. However, the development of stock-specific recovery efforts is often impeded by a lack of information on basic species biology, including stock structure. The species status of some istiophorids is also uncertain, further complicating management efforts as well as strategies to conserve genetic diversity characteristic of distinct evolutionary lineages. In this dissertation, a molecular approach is used to address questions currently contributing uncertainty to the conservation and management of two istiophorid billfishes, white marlin (Kajikia albida) and striped marlin (K. audax). These closely related sister species are distributed in the Atlantic and Indo-Pacific oceans, respectively. Previous assessment of genetic population structure for white marlin based on mitochondrial (mt) DNA and five nuclear microsatellite markers suggested the possibility of population structuring for this species; however, results from the evaluation of mtDNA and 24 microsatellites across a larger number of samples, including a collection of larvae, are consistent with the presence of a single genetic stock (Chapter II). This result highlights the importance of analyses based on large numbers of molecular markers and samples, as well as a biologically informed sampling design, for studies of population structure in highly migratory pelagic species. Compared to the apparent lack of genetic population structure for white marlin, analysis of nearly 4,000 single nucleotide polymorphism (SNP) molecular markers across collections of striped marlin from the Pacific and, for the first time, Indian oceans resolved multiple genetically distinct populations (Chapter III). These populations correspond with striped marlin sampled from the western Indian Ocean, Oceania, North Pacific Ocean, and eastern central Pacific Ocean. Results from individual-based cluster analyses also suggest the presence of a second genetically distinct population in the North Pacific Ocean. Comparisons of replicate sample collections for some regions demonstrate the stability of allele frequencies across multiple generations. Finally, the uncertain species status of striped marlin and white marlin was evaluated using over 12,000 genome-wide SNPs surveyed across large numbers of exemplars per species (white marlin: n = 75, striped marlin: n = 250; Chapter IV). Results from individual-based cluster and maximum likelihood phylogenetic analyses suggest the presence of distinct evolutionary lineages for striped marlin and white marlin. This result is consistent with levels of genetic differentiation between striped marlin and white marlin which are an order of magnitude higher than those calculated between populations of striped marlin. Collectively, results of this dissertation
xvi provide practical insights for improving the conservation and management of white marlin and striped marlin, including revised stock structures which should be recognized in assessment and management plans for striped marlin. Future genomic studies should focus on addressing uncertainties regarding rangewide stock structure and species relationships for other istiophorids. Additionally, studies which continue to improve the genomic resources available for istiophorid billfishes and other large pelagic fishes may ultimately facilitate the evaluation of questions previously unexplored for the pelagic marine environment, such as localized adaptation and speciation.
xvii
Evaluation of Population Structure and the Interspecific Relationship of Striped Marlin (Kajikia audax) and White Marlin (K. albida) Based on Traditional or Genome-wide Molecular Markers
CHAPTER I
Introduction: Challenges Related to the Conservation and Management of Istiophorid
Billfishes
2 MOTIVATION
The istiophorid billfishes (marlins, spearfishes, and sailfish) are large, predatory fishes with broad and continuous spatial distributions spanning temperate, sub-tropical, and tropical pelagic waters of the Atlantic, Pacific, and Indian oceans (Nakamura 1985).
Many of these species undertake long-distance seasonal migrations, presumably to utilize geographically distant spawning and feeding grounds. Fisheries targeting istiophorid billfishes include recreational fisheries, in which these species are prized for their large size and spectacular fight displays, and a number of commercial fisheries. Istiophorids are also targeted in small-scale artisanal and subsistence fisheries; however, reporting from these fisheries is incomplete and their contribution to overall billfish fishing mortality is unknown. The primary source of fishing mortality for istiophorid billfishes is attributed to bycatch in commercial pelagic longline fisheries targeting tunas and swordfish (Peel et al. 2003). Understanding the sources and magnitude of fishing mortality for istiophorid billfishes is critical because a number of istiophorid stocks are considered overfished and/or experiencing overfishing. In addition, istiophorids assessed according to International Union for Conservation of Nature Red List standards include some species with declining population trends that are also vulnerable or near threatened, while other species lack sufficient data to determine conservation status (Collette et al.
2011). Conservation and management measures have been implemented for istiophorids in some fisheries, and limited degrees of recovery have been observed in certain regions
(Graves & Horodysky 2015); however, additional efforts are necessary for rebuilding overfished stocks and/or reducing fishing effort to sustainable levels.
3 In many instances, the development of targeted management measures for fisheries that interact with istiophorid billfishes is limited by a lack of information on the basic biology of these species. The pelagic habitat and rare-event nature of istiophorid billfish encounters makes field observations of these species challenging, especially for smaller size classes that do not frequently interact with local fisheries. Patterns of habitat utilization, including seasonal migrations and spatial regions important for feeding or spawning, are poorly understood across istiophorid species. Additionally, the extent to which stocks seasonally mix on shared spawning and feeding grounds is not well known.
This lack of information limits the ability of fisheries managers to develop management measures which promote the recovery of stocks characterized by distinct biological attributes, including varying propensities for sustaining fishing pressure.
Effective management of istiophorid billfishes is also frequently challenged by a lack of information regarding population structure (i.e. the number and geographic extent of distinct populations) for many of these species. Presumably, a continuous marine environment that lacks obvious physical barriers to dispersal would not be conducive to the development of population structure for highly migratory species such as billfishes.
However, the presence of distinct populations within and/or between ocean basins has been reported for a number of istiophorids (reviewed by Graves & McDowell 2003,
2015). For example, previous genetic studies have confirmed that blue marlin (Makaira nigricans) comprises genetically distinct populations between the Atlantic and Pacific oceans (Finnerty & Block 1992, Graves & McDowell 1995, Buonaccorsi et al. 1999,
2001, Sorenson et al. 2013); however, population structuring for this species within either ocean basin has not been detected (McDowell et al. 2007). In comparison, the presence of
4 multiple genetically distinct populations has been reported for sailfish (Isitophorus platypterus) within both the Pacific and Indian oceans (McDowell 2002, Hoolihan et al.
2004, Lu et al. 2015). Though information on population structuring is available for some istiophorid billfishes, most studies have focused on only a portion of the species’ range.
In addition, the biological interpretation of results from studies of population structure for some istiophorids has been equivocal, while for other species information on population structure is lacking altogether.
In addition to a lack of clarity regarding population-level relationships for istiophorid billfishes, a number of species-level relationships within this family are uncertain. The discrimination of istiophorid species is primarily dependent upon differences in morphological and meristic characters, and in reported spatial distributions.
However, similar morphological features, particularly among some istiophorids, makes species-level identifications in the field challenging, especially in regions where spatial distributions among species overlap. Difficulties associated with species identifications have resulted in complicated taxonomic histories for a number of istiophorids. For example, despite an original species description in the year 1840, the validity of roundscale spearfish (Tetrapturus georgii) remained uncertain for over a century due to morphological similarity and an overlapping spatial distribution with white marlin
(Kajikia albida; Shivji et al. 2006). In addition, distinct Indo-Pacific and Atlantic species were previously recognized for both blue marlin and sailfish based on slight morphological differences (Nakamura 1985); however, genetic results are consistent with single, globally distributed species for blue marlin and sailfish (Collette et al. 2006).
5 Though a number of issues regarding the specific status of istiophorid billfishes have been resolved, uncertainties persist for some species.
An informed understanding of population- and species-level relationships is necessary for identifying and conserving evolutionarily significant units (e.g. Moritz
1994), and for estimating levels of genetic diversity. Consequences of reduced genetic diversity include populations and/or species less capable of withstanding disease and fluctuations in environmental conditions (Gaggiotti & Vetter 1999, Allendorf et al. 2008).
In the context of fisheries management, a mismatch between biologically distinct units and units recognized for management can result in the unintentional overfishing of some populations relative to others (Allendorf et al. 2008, Reiss et al. 2009, Allendorf et al.
2014, Pinksy & Palumbi 2014, Ovenden et al. 2015).
RESEARCH APPROACH
The primary goals of the research encompassed by this dissertation are to clarify population- and species-level relationships for two closely related istiophorid billfishes, striped marlin (K. audax) and white marlin. These fishes represent a unique study system comprising closely related sister species that inhabit distinct ocean basins: striped marlin is distributed in waters of the Pacific and Indian oceans, and white marlin is distributed in the Atlantic Ocean (Nakamura 1985). Currently, information on population structuring within these species is equivocal or incomplete, and the status of striped marlin and white marlin as distinct species has been questioned. To address these issues, a genetic approach is employed for evaluating the presence of population structuring within striped
6 marlin and white marlin, and for determining the evolutionary relationship of these species.
The application of a genetic approach for assessing population- and species-level relationships in highly migratory pelagic fishes is especially timely given recent advances in molecular technology. These advances are based on the development and subsequent widespread availability of next-generation sequencing (NGS) technology (e.g. Mardis
2008). Compared with previous technologies, NGS enables the rapid and economical characterization of large portions of the genome across large numbers of samples, including for species with limited or no prior genetic information. The unprecedented level of statistical power facilitated by NGS-based methodologies is especially useful for resolving intra- and inter-specific relationships for species characterized by high levels of gene flow, such as istiophorid billfishes. In Chapters III and IV of this dissertation, NGS- based methodology is employed for the discovery of large numbers of single nucleotide polymorphism (SNP) molecular markers for both striped marlin and white marlin. Newly developed SNPs are then used to evaluate the presence of genetically distinct populations for striped marlin in the Pacific and Indian oceans, and to assess the interspecific relationship of striped marlin and white marlin.
A large number of striped marlin and white marlin tissue samples were necessary for the genetic analyses comprising this dissertation. In particular, assessments of genetic population structure require sample collections from locations representative of the full species range, and that comprise enough individuals to accurately characterize allele frequencies for each geographic region. Collecting large numbers of samples for istiophorid billfishes is challenging given the pelagic habitat and comparative rarity of
7 these species. Experimental designs for population genetic studies of istiophorids are typically characterized by samples opportunistically collected from diverse geographic locations in seasons where recreational and/or commercial fisheries interact with these species. However, if population structuring is present, populations are most likely to be separated at the time of spawning, and genetic studies that focus sampling efforts on larvae and reproductively active adults on spawning grounds will improve the ability to resolve population structure (Graves et al. 1996, Waples 1998, Bowen et al. 2005,
Carlsson et al. 2007, Graves & McDowell 2015). In practice, such an informed sampling design is challenging to implement for istiophorid billfishes due to a number of factors, including limited information on timing and location of spawning, a lack of larval sampling efforts, and difficulty determining reproductive status using non-lethal methods.
For this dissertation, tissue samples of striped marlin and white marlin were primarily obtained through the development of a global sampling network comprising numerous local and international recreational anglers, scientists, and conservation agencies. Tissues obtained through this network were analyzed in conjunction with previously collected tissues to facilitate insightful comparisons of sample collections spanning multiple generations.
BACKGROUND INFORMATION
Population structuring in white marlin
The presence of stock structure for white marlin in the Atlantic Ocean is unclear, and this uncertainty is reflected in the management history of this species. The International
Commission for the Conservation of Atlantic Tunas (ICCAT), the regional fisheries
8 management organization (RFMO) for white marlin and a number of other highly migratory pelagic fishes in the Atlantic Ocean, currently recognizes a single Atlantic- wide stock of white marlin for assessment and management purposes. However, this single-stock model was preceded by a two-stock model that recognized distinct stocks of white marlin in the North Atlantic and South Atlantic oceans. Biological support for a two-stock model included the presence of discrete spawning grounds north and south of the equator, a perceived lack of catches for white marlin in equatorial waters, and a lack of trans-equatorial movements reported for tagged fish. ICCAT switched to the single- stock model in 2001 primarily in response to the results of a genetic study that was unable to reject the null hypothesis of a single Atlantic-wide population for white marlin
(Graves & McDowell 2001). In addition, the expansion of pelagic longline fisheries into equatorial waters revealed a continuous distribution of white marlin catches across this region, and a few tag recaptures representing trans-equatorial movements had been reported.
Since 2001, additional information from genetic as well as conventional and electronic tagging studies suggest the possibility of population structuring for white marlin. A 2006 genetic study based on the analysis of additional molecular markers and sample also failed to reject the null hypothesis of a single ocean-wide population for white marlin (Graves & McDowell 2006), but reported a statistically significant level of genetic differentiation between sample collections from the western North Atlantic and western South Atlantic oceans. Analysis of conventional tag recapture data also suggest the possibility of white marlin movements occurring on annual time scales (Ortiz et al.
2003, Snodgrass et al. 2011). Finally, valuable year-long tag deployment periods for two
9 individuals carrying pop-up satellite archival tags demonstrate cyclical migration patterns which include time spent on known spawning grounds during the spawning season
(Rooker et al. 2013, Loose 2014). Though observations from available tagging data do not provide direct evidence of spawning-site fidelity in white marlin, they suggest the possibility that white marlin are not panmictic.
Given the results of these more recent genetic and tagging studies, it is clear that the possibility of population structuring for white marlin requires further examination.
Previous genetic studies of population structure for white marlin have analyzed relatively small numbers of molecular markers. Graves & McDowell (2001) surveyed genetic variation at four nuclear microsatellite markers and whole mitochondrial (mt) DNA, and
Graves & McDowell (2006) evaluated five microsatellite markers and the mtDNA control region. It is possible that statistical power associated with the analysis of a small number of molecular markers is insufficient for detecting low but biologically meaningful levels of genetic differentiation between populations of white marlin.
Populations of marine fishes are expected to display levels of genetic differentiation that are at least an order of magnitude lower than populations of terrestrial or freshwater species (Ward et al. 1994, Waples 1998). In instances where genetic differentiation between populations is low, a high level of statistical power facilitated by the analysis of large numbers of molecular markers and individuals is necessary for resolving population structure. Relative to previous genetic studies of population structure for white marlin, an assessment based on larger numbers of molecular markers would provide a more powerful evaluation of population structuring in this species. Furthermore, incorporating
10 sample collections of larvae and/or reproductively active adults will also improve the ability to identify genetically distinct populations, if present.
Population structuring in striped marlin
In contrast to the purported lack of population structuring for white marlin, significant population structuring has been reported for striped marlin in the Pacific Ocean. Early genetic studies of population structure for striped marlin surveyed variation at allozyme loci (Morgan 1992) and in whole mtDNA (Graves & McDowell 1994), and observed low but statistically significant heterogeneity among sample collections obtained from geographically distant regions of the Pacific Ocean. More recently, McDowell & Graves
(2008) resolved four genetically distinct populations of striped marlin in the Pacific
Ocean based on the analysis of five microsatellite markers and the mtDNA control region. Those populations correspond with striped marlin sampled off Baja California,
Ecuador, eastern Australia, and across the North Pacific Ocean (including fish sampled off Japan, Taiwan, Hawaii, and California). These populations were also resolved by
Purcell & Edmands (2011) based on the comparison of a larger number of microsatellite markers (n = 12) and the mtDNA control region across a larger number of samples; however, a lack of samples from waters south of Baja California prohibited the evaluation of population structure in the eastern central Pacific Ocean. In addition,
Purcell & Edmands (2011) resolved mature striped marlin sampled off Hawaii as belonging to a genetically distinct population relative to juvenile fish from this region.
Results from genetic studies of population structure for striped marlin are generally consistent with information derived from conventional and pop-up satellite archival tags
11 deployed on fish from locations throughout the Pacific Ocean, although some exceptions exist. In contrast to white marlin, conventional tag recaptures for striped marlin do not correspond with patterns suggestive of movements occurring on annual time scales (Ortiz et al. 2003). Studies based on the use of electronic tags demonstrate localized movements restricted to specific regions of the Pacific Ocean; these regions generally correspond with the spatial extent of striped marlin populations resolved in genetic studies (Domeier
2006, Holdsworth et al. 2009, Sippel et al. 2011, Holdsworth & Saul 2014).
Despite multiple assessments of genetic population structure for striped marlin in the Pacific Ocean, population structuring for this species in the Indian Ocean remains unexplored. Given this lack of information, the Indian Ocean Tuna Commission, the
RFMO for striped marlin and other highly migratory pelagic fishes in the Indian Ocean, utilizes a single-stock model for striped marlin management and assessment. Possible incongruence between the number of units recognized for management and the number of biologically distinct populations may be especially problematic in the Indian Ocean because striped marlin in this region was recently identified as overfished and experiencing overfishing (IOTC 2017). It is clear that efforts to assist the recovery of striped marlin to sustainable levels of biomass in the Indian Ocean would benefit from an improved understanding of population structure. Results from such a study would also enable the prioritization of stocks for management intervention, and would provide information on gene flow between striped marlin in the Pacific and Indian oceans.
12 Relationship of striped marlin and white marlin
Though currently recognized as distinct species, morphological and molecular similarity of white marlin and striped marlin have resulted in a questionable taxonomic status for these species. Morphological comparisons suggest that slight differences in fin morphology are useful for discriminating white marlin and striped marlin (Nakamura
1985); however, in practice, the subtlety and variation of these characters has resulted in species identifications primarily based on geographic location of capture (Atlantic vs.
Indo-Pacific). Previous genetic comparisons of white marlin and striped marlin include surveys of nuclear markers and mtDNA using a variety of methodologies (Chow 1994,
Graves & McDowell 1995, Graves 1998, Collette et al. 2006, Hanner et al. 2011, Santini
& Sorenson 2013), including analyses based on cytochrome oxidase subunit I mtDNA
‘barcode’ region sequences (e.g. Hebert et al. 2003; Hanner et al. 2011). These studies have been unsuccessful in resolving white marlin and striped marlin as reciprocally monophyletic lineages. McDowell et al. (unpublished data) recovered white marlin and striped marlin as distinct lineages based on a segment of the mtDNA control region, but net nucleotide sequence divergence between lineages was 2.25%, and is less than the level of divergence reported for geographically distant populations of striped marlin
(McDowell & Graves 2008).
The distinct species status of striped marlin and white marlin contrasts with the taxonomic status of blue marlin and sailfish, which are also distributed across the
Atlantic, Pacific, and Indian oceans. Blue marlin and sailfish were historically recognized as each comprising separate Atlantic and Indo-Pacific species (Nakamura 1985).
However, analyses of mtDNA revealed a lack of fixed differences between Atlantic and
13 Indo-Pacific sample collections of blue marlin and of sailfish, and levels of inter-oceanic differentiation were consistent with those expected for populations of the same species
(Graves & McDowell 1995, Buonaccorsi et al. 2001, McDowell 2002). In comparison, the level of divergence observed between white marlin and striped marlin is less than half of that observed between Atlantic and Indo-Pacific populations of blue marlin and sailfish (Graves & McDowell 1995).
White marlin and striped marlin have maintained distinct species status despite similar morphology and an inability of previous genetic studies to resolve these species as distinct evolutionary lineages. To improve our understanding of the interspecific relationship of white marlin and striped marlin, a powerful assessment based on the analysis of genome-wide molecular markers is necessary. A number of empirical studies demonstrate the utility of NGS-based phylogenomic approaches for resolving species relationships, particularly in closely related taxa (Rubin et al. 2012, Wagner et al. 2013,
Cruaud et al. 2014). Such a study would also facilitate subsequent analyses focused on evaluating demographic history for white marlin and striped marlin, providing novel insights into speciation in the pelagic marine environment.
DISSERTATION OBJECTIVES AND STRUCTURE
This dissertation comprises three studies focused on the use of genetic methods for evaluating population- and species-level relationships of striped marlin and white marlin.
Primary objectives of this dissertation include:
• Evaluate the ability of an informed sampling design and a large number of
microsatellite markers to clarify population structure for white marlin
14 • Employ NGS-based methodology to discover genome-wide SNPs for striped
marlin and white marlin
• Assess genetic population structure for striped marlin in the Pacific and Indian
oceans using genome-wide SNPs
• Delimit species boundaries for white marlin and striped marlin using genome-
wide SNPs
The general structure of this dissertation is as follows:
Chapter II details an assessment of genetic population structure for white marlin in the Atlantic Ocean. Relative to previous studies, this assessment is based on the comparison of a larger number of microsatellite markers as well as mtDNA across a larger number of samples to facilitate greater statistical power for resolving genetically distinct populations. Additionally, this study incorporates a biologically informed experimental design that includes the sampling of larvae. Results from this study have direct implications for the management of white marlin, and provide key insights into statistical power associated with population genetic studies of highly migratory pelagic fishes.
Chapter III describes a genomic evaluation of population structure for striped marlin in the Pacific and Indian oceans. This study utilizes NGS-based methodology to discover and characterize thousands of SNPs across striped marlin sample collections representative of the full species range. SNP genotype data are analyzed to discriminate genetically distinct populations, evaluate multi-generational temporal stability of allelic frequencies, and to estimate genetic connectivity among populations. Results from this
15 study have direct implications for the management of striped marlin in the Pacific and
Indian oceans, and provide information on the long-term stability of low but statistically significant levels of genetic differentiation between populations.
Chapter IV comprises a study focused on the use of genome-wide SNPs for assessing the evolutionary relationship of striped marlin and white marlin. This study employs a range of population genomic and species delimitation methods to evaluate scenarios for which white marlin and striped marlin are regarded as distinct species or as sub-populations of a single species. Results from this study provide clarification regarding the taxonomic status of white marlin and striped marlin.
16 REFERENCES
Allendorf, F., Berry, O., and Ryman, N. 2014. So long to genetic diversity, and thanks for
all the fish. Molecular Ecology, 23: 23–25.
Allendorf, F., England, P., Luikart, G., Ritchie, P., and Ryman, N. 2008. Genetic effects
of harvest on wild animal populations. Trends in Ecology and Evolution, 23: 327–
337.
Bowen, B., Bass, A., Soares, L., and Toonen, R. 2005. Conservation implications of
complex population structure: lessons from the loggerhead turtle (Caretta
caretta). Molecular Ecology, 14: 2389–2402.
Buonaccorsi, V., McDowell, J., and Graves, J. 2001. Reconciling patterns of inter-ocean
molecular variance from four classes of molecular markers in blue marlin
(Makaira nigricans). Molecular Ecology, 10: 1179–1196.
Buonaccorsi, V., Reece, K., Morgan, L., and Graves, J. 1999. Geographic distribution of
molecular variance within the blue marlin (Makaira nigricans): a hierarchical
analysis of allozyme, single-copy nuclear DNA, and mitochondrial DNA markers.
Evolution, 53: 568–579.
Carlsson, J., McDowell, J., Carlsson, J., and Graves, J. 2007. Genetic identity of YOY
bluefin tuna from the eastern and western Atlantic spawning areas. The Journal of
Heredity, 98: 23–28.
Chow, S. 1994. Identification of billfish species using mitochondrial cytochrome b gene
fragment amplified by polymerase chain reaction. ICCAT Collective Volume of
Scientific Papers, 41: 549–556.
17 Collette, B., Carpenter, K., Polidoro, B., Juan-Jordá, M., Boustany, A., Die, D., Elfes, C.,
Fox, W., Graves, J., Harrison, L., McManus, R., Minte-Vera, C., Nelson, R.,
Restrepo, V., Schratwieser, J., Sun, C., Amorim, A., Brick Peres, M., Canales, C.,
Cardenas, G., Chang, S., Chiang, W., de Oliveira Leite, Jr., N., Harwell, H.,
Lessa, R., Fredou, F., Oxenford, H., Serra, R., Shao, K., Sumaila, R., Wang, S.,
Watson, R., and Yáñez, E. 2011. High value and long live–double jeopardy for
tunas and billfishes. Science, 333: 291–292.
Collette, B., McDowell, J., and Graves, J. 2006. Phylogeny of recent billfishes
(Xiphioidei). Bulletin of Marine Science, 79: 455–468.
Cruaud, A., Gautier, M., Galan, M., Foucaud, J., Sauné, L., Genson, G., Dubois, E.,
Nidelet, S., Deuve, T., and Rasplus, J. 2014. Empirical assessment of RAD
sequencing for interspecific phylogeny. Molecular Biology and Evolution, 31:
1272–1274.
Domeier, M. 2006. An analysis of Pacific striped marlin (Tetrapturus audax) horizontal
movement patterns using pop–up satellite archival tags. Bulletin of Marine
Science, 79: 811–825.
Finnerty, J., and Block, B. 1992. Direct sequencing of mitochondrial DNA detects highly
divergent haplotypes in blue marlin (Makaira nigricans). Molecular Marine
Biology and Biotechnology, 1: 206–214.
Gaggiotti, O., and Vetter, R. 1999. Effect of life history strategy, environmental
variability, and overexploitation on the genetic diversity of pelagic fish
populations. Canadian Journal of Fisheries and Aquatic Sciences, 56: 1376–1388.
18 Graves, J. 1998. Molecular insights into the population structures of cosmopolitan marine
fishes. Journal of Heredity, 89: 427–437.
Graves, J., and Horodysky, A. 2015. Challenges of estimating post-release mortality of
istiophorid billfishes caught in the recreational fishery: a review. Fisheries
Research, 166: 163–168.
Graves, J., Gold, J., Ely, B., Quattro, J., Woodley, C., and Dean, J. 1996. Population
genetic structure of bluefin tuna in the North Atlantic Ocean. Identification of
variable genetic markers. ICCAT Collective Volume of Scientific Papers, 45:
155–157.
Graves, J., and McDowell, J. 1994. Genetic analysis of striped marlin (Tetrapturus
audax) population structure in the Pacific Ocean. Canadian Journal of Fisheries
and Aquatic Sciences, 51: 1762–1768.
Graves, J., and McDowell, J. 1995. Inter-ocean genetic divergence of istiophorid
billfishes. Marine Biology, 122: 193–203.
Graves, J., and McDowell, J. 2001. A genetic perspective on the stock structures of blue
marlin and white marlin in the Atlantic Ocean. ICCAT Collective Volume of
Scientific Papers, 53: 180–187.
Graves, J., and McDowell, J. 2003. Stock structure of the world’s istiophorid billfishes: a
genetic perspective. Marine and Freshwater Research, 54: 287–298.
Graves, J., and McDowell, J. 2006. Genetic analysis of white marlin (Tetrapturus
albidus) stock structure. Bulletin of Marine Science, 79: 469–482.
Graves, J., and McDowell, J. 2015. Population structure of istiophorid billfishes.
Fisheries Research, 166: 21–28.
19 Hanner, R., Floyd, R., Bernard, A., Collette, B., and Shivji, M. 2011. DNA barcoding of
billfishes. Mitochondrial DNA, 22: 27–36.
Hebert, P., Cywinska, A., Ball, S., and deWaard, J. 2003. Biological identifications
through DNA barcodes. Proceedings of the Royal Society of London B, 270:
313–321.
Holdsworth, J., and Saul, P. 2014. New Zealand billfish and gamefish tagging, 2012–13.
Wellington, New Zealand. Ministry for Primary Industries.
Holdsworth, J., Sippel, T., and Block, B. 2009. Near real time satellite tracking of striped
marlin (Kajikia audax) movements in the Pacific Ocean. Marine Biology, 156:
505–514.
Hoolihan, J., Premanandh, J. D’Aloia–Palmieri, M., and Benzie, J. 2004. Intraspecific
phylogeographic isolation of Arabian Gulf sailfish Istiophorus platypterus
inferred from mitochondrial DNA. Marine Biology 145: 465–475.
IOTC. 2017. Report of the 15th session of the IOTC working party on billfish. Victoria
Mahé, Seychelles. Indian Ocean Tuna Commission.
Loose, E. 2014. Seasonal movements, habitat utilization, and comparative scale
morphology of white marlin (Kajikia albida) and roundscale spearfish
(Tetrapturus georgii). Master’s thesis. Virginia Institute of Marine Science,
College of William & Mary, Gloucester Point, VA, USA.
Lu, C. –P., Alvarado Bremer, J., McKenzie, J., and Chiang, W. –C. 2015. Analysis of
sailfish (Istiophorus platypterus) population structure in the North Pacific Ocean.
Fisheries Research, 166: 33–38.
20 Mardis, E. 2008. The impact of next-generation sequencing technology on genetics.
Trends in Genetics 24: 133–141.
McDowell, J. 2002. Genetic stock structure of the sailfish, Istiophorus platypterus, based
on nuclear and mitochondrial DNA. Doctoral dissertation. Virginia Institute of
Marine Science, College of William & Mary, Gloucester Point, VA, USA.
McDowell, J., and Graves, J. 2008. Population structure of striped marlin (Kajikia audax)
in the Pacific Ocean based on analysis of microsatellite and mitochondrial DNA.
Canadian Journal of Fisheries and Aquatic Sciences, 65: 1307–1320.
McDowell, J., Carlsson, J., and Graves, J. 2007. Genetic analysis of blue marlin (Makaira
nigricans) stock structure in the Atlantic Ocean. Gulf and Caribbean Research,
19: 75–82.
Morgan, L. 1992. Allozyme analysis of billfish population structure. Master’s thesis.
Virginia Institute of Marine Science, College of William & Mary, Gloucester
Point, VA, USA.
Moritz, C. 1994. Defining ‘Evolutionarily Significant Units’ for conservation. Trends in
Ecology and Evolution, 9: 373–375.
Nakamura, I. 1985. Billfishes of the world: an annotated and illustrated catalogue of
marlins, sailfishes, spearfishes, and swordfishes known to date. FAO Fisheries
Synopsis No. 125, vol. 5. 65 pp. Rome, Italy: Food and Agricultural Organization.
Ortiz, M., Prince, E., Serafy, J., Holts, D., Davy, K., Pepperell, J., Lowry, M., and
Holdsworth, J. C. 2003. Global overview of the major constituent-based billfish
tagging programs and their results since 1954. Marine and Freshwater Research,
54: 489–507.
21 Ovenden, J., Berry, O., Welch, D., Buckworth, R., and Dichmont, C. 2015. Ocean’s
eleven: a critical evaluation of the role of population, evolutionary and molecular
genetics in the management of wild fisheries. Fish and Fisheries, 16: 125–159.
Peel, E., Nelson, R., and Goodyear, C. 2003. Managing Atlantic marlin as bycatch under
ICCAT. The fork in the road: recovery or collapse. Marine and Freshwater
Research, 54: 575–584.
Pinsky, M., and Palumbi, S. 2014. Meta-analysis reveals lower genetic diversity in
overfished populations. Molecular Ecology, 23: 29–39.
Purcell, C., and Edmands, S. 2011. Resolving the genetic structure of striped marlin,
Kajikia audax, in the Pacific Ocean through spatial and temporal sampling of
adult and immature fish. Canadian Journal of Fisheries and Aquatic Sciences, 68:
1861–1875.
Reiss, H., Hoarau, G., Dickey–Collas, M., and Wolff, W. 2009. Genetic population
structure of marine fish: mismatch between biological and fisheries management
units. Fish and Fisheries, 10: 361–395.
Rooker, J., Kitchens, L., Dance, M., Wells, R., Falterman, B., and Cornic, M. 2013.
Spatial, temporal, and habitat-related variation in abundance of pelagic fishes in
the Gulf of Mexico: potential implications of the Deepwater Horizon oil spill.
PLoS ONE, 8: e76080.
Rubin, B., Ree, R., and Moreau, C. 2012. Inferring phylogenies from RAD sequence
data. PLoS ONE, 7: e33394.
22 Santini, F., and Sorenson, L. 2013. First molecular timetree of billfishes
(Istiophoriformes: Acanthomorpha) shows a Late Miocene radiation of marlins
and allies. Italian Journal of Zoology, 80: 481–489.
Shivji, M., Magnussen, J., Beerkircher, L., Hinteregger, G., Lee, D., Serafy, J., and
Prince, E. 2006. Validity, identification, and distribution of the roundscale
spearfish, Tetrapturus georgii (Teleostii: Istiophoridae): morphological and
molecular evidence. Bulletin of Marine Science, 79: 483–491.
Sippel, T., Holdsworth, J., Dennis, T., and Montgomery, J. 2011. Investigating behaviour
and populations dynamics of striped marlin (Kajikia audax) from the Southwest
Pacific Ocean with satellite tags. PLoS ONE, 6: e21087.
Snodgrass, D., Orbesen, E., Hoolihan, J., and Prince, E. 2011. The U.S. conventional
tagging database updates for Atlantic white marlin (1954-2008). ICCAT
Collective Volume of Scientific Papers, 66: 1760–1766.
Sorenson, L., McDowell, J., Knott, T., and Graves, J. 2013. Assignment test method
using hypervariable markers for blue marlin (Makaira nigricans) stock
identification. Conservation Genetics Resources, 5: 293–297.
Wagner, C., Keller, I., Wittwer, S., Selz, O., Mwaiko, S., Greuter, L., Sivasundar, A., and
Seehausen, O. 2013. Genome-wide RAD sequence data provide unprecedented
resolution of species boundaries and relationships in the Lake Victoria cichlid
adaptive radiation. Molecular Ecology, 22: 787–798.
Waples, R. 1998. Separating the wheat from the chaff: patterns of genetic differentiation
in high gene flow species. The Journal of Heredity, 89: 438–450.
23 Ward, R., Woodwark, M., and Skibinski, D. 1994. A comparison of genetic diversity
levels in marine, freshwater, and anadromous fishes. Journal of Fish Biology, 44:
213–232.
24
CHAPTER II
Genetic Evaluation of Population Structure in White Marlin (Kajikia albida): The
Importance of Statistical Power
25 ABSTRACT
The genetic basis of population structure in white marlin (Kajikia albida) is not well understood. Previous evaluation of genetic population structure in this species utilized a small number of molecular markers to survey genetic variation across opportunistically collected samples of adults, resulting in statistically significant levels of genetic differentiation for some pairwise comparisons and global levels of genetic differentiation that approached statistical significance. The present study increased statistical power to improve resolution of genetic population structure in white marlin by surveying a larger number of molecular markers across sample collections of increased size, including collections from additional geographic locations and a robust collection of larvae.
Increased statistical power resulted in lower levels of genetic heterogeneity compared with the previous study, and results were consistent with the presence of a single genetic stock of white marlin in the Atlantic Ocean. These results indicate that when statistical power is low, the ability to distinguish noise from a true signal of population structure is compromised. This relationship is especially important for population genetic assessments of marine fishes where genetic differentiation, if it exists, is expected to be low.
26 INTRODUCTION
Population genetic studies of marine fishes provide information to fisheries managers useful for the identification of biological units relevant for assessment and management, and for maintaining unique genetic variation (Allendorf et al., 1987; Ward, 2000;
Ovenden et al., 2013). Despite their utility, these studies are particularly challenging for many marine fishes because the large effective population sizes of these species limit genetic drift, resulting in low levels of genetic differentiation among populations (Ward et al., 1994; Waples, 1998). A high level of statistical power is required to detect genetic differentiation in population genetic studies of marine fishes, and can be achieved by evaluating larger numbers of molecular markers and/or samples per putative population
(Waples, 1998; Ryman et al., 2006). While the former strategy is facilitated by currently available laboratory methodologies, obtaining appropriate sample sizes for population genetic studies of relatively rare-event species can be difficult.
Additional considerations for population genetic studies of highly migratory marine fishes are associated with sampling designs appropriate for these species. Highly migratory marine fishes are capable of long distance movements, and opportunistic experimental designs based on the sampling of geographically distant locations may not be informative because individuals of the same stock could be sampled from multiple locations depending on the time of year (Graves et al., 1996; Carlsson et al., 2007;
Graves & McDowell, 2015). Many highly migratory marine fishes also display seasonal mixed assemblages, and opportunistic sampling may lead to sample collections representative of more than one stock, resulting in a noisy genetic signal that may mask genetic differentiation (Waples, 1998; Bowen et al., 2005). For highly migratory marine
27 fishes that display population structure, populations are most likely to separate at the time of spawning (Graves et al., 1996; Carlsson et al., 2007; Graves & McDowell, 2015).
Replacing opportunistic sampling with a biologically informed design that targets larvae and/or reproductively active adults improves the ability to detect population subdivision in these species, if it exists. This concept is illustrated in Atlantic bluefin tuna (Thunnus thynnus) for which eastern and western Atlantic stocks were initially recognized based on the presence of distinct spawning grounds in the Gulf of Mexico and Mediterranean Sea as well as differences in biological characteristics between fish from these regions
(Fromentin & Powers, 2005; Rooker et al., 2007). Early genetic studies based on opportunistic sample collections were unable to detect population subdivision between putative stocks; however, subsequent genetic analyses that incorporated a biologically informed sampling design targeting larvae and mature adults on spawning grounds during the spawning season detected statistically significant population subdivision consistent with inferences from non-genetic data (Carlsson et al., 2007; Boustany et al., 2008).
Population genetic studies that incorporate sampling designs also inclusive of temporal replicates provide the ability to determine if an observation of statistically significant differentiation is stable over time and unlikely to result from artifacts such as the non-random sampling of populations, stochastic fluctuation in allele frequencies, or variation in reproductive success (Allendorf & Phelps, 1981; Waples & Teel, 1990;
Hedgecock, 1994; Waples, 1998; Tessier & Bernatchez, 1999). Temporal replicates are particularly important for marine fishes because the level of genetic divergence among populations, if present, is expected to be low. In such cases, it becomes increasingly challenging to distinguish between noise and a weak but meaningful level of
28 heterogeneity (Waples, 1998). Obtaining temporal replicates for some highly migratory marine fishes is especially challenging due to the relatively rare event nature of a number of these species. Frequently, these challenges result in very low sample sizes per sampling location in any given year, and necessitate the use of sample collections that are pooled across years for each location.
White marlin (Kajikia albida) is a highly migratory marine fish distributed throughout the Atlantic Ocean in temperate, sub-tropical, and tropical waters (Nakamura,
1985). This species is managed along with a number of other pelagic fishes in the
Atlantic Ocean by the member nations of the International Commission for the
Conservation of Atlantic Tunas (ICCAT). Prior to the year 2000, ICCAT utilized a two stock management model for white marlin that recognized distinct populations in the
North Atlantic and South Atlantic oceans. This model was based on the presence of seasonally displaced spawning grounds north and south of the equator, fishery dependent data that reflected limited catches in equatorial waters, and a lack of trans-equatorial movements reported for tagged fish (ICCAT, 1994). ICCAT adopted a single stock management model in 2000 in response to new fisheries data that indicated continuous catches across the equator, trans-oceanic and trans-equatorial tag returns for a few individuals, and a population genetic study that was unable to detect statistically significant intra-oceanic heterogeneity (Graves & McDowell, 2001).
Since the adoption of the single stock management model for white marlin, additional genetic study has suggested the possibility of stock structure in this species.
Analysis of variation at five nuclear microsatellite markers revealed statistically significant genetic differentiation between collections of white marlin from the western
29 North Atlantic and western South Atlantic oceans, and global levels of genetic differentiation that approached statistical significance based on the microsatellite data and on mitochondrial (mt) DNA control region sequence data (Graves & McDowell, 2006).
The authors concluded that while the null hypothesis of genetic homogeneity could not be rejected, genetic population structure could exist but may not have been detected due to low statistical power and/or the opportunistic sampling design employed in the study
(Graves & McDowell, 2006). Thus, there is considerable uncertainty regarding the suitability of a single stock management model for white marlin, and management of this species would benefit from a more thorough understanding of population structure.
In the present study, the null hypothesis of a single Atlantic-wide stock of white marlin was evaluated by surveying genetic variation at 24 microsatellite loci and at the mtDNA control region in multiple collections of white marlin sampled from locations throughout the Atlantic Ocean. Relative to Graves & McDowell (2006), the ability to detect genetic population structure was improved by analyzing a larger number of microsatellite markers, increasing the number of sampling locations and sample sizes for some locations, and including a robust collection of larvae from the Gulf of Mexico.
MATERIALS & METHODS
Sample Collection & Generation of Nuclear Genotype Data
Samples consisting of muscle tissue from landed adult white marlin or fin clips from live- released adult white marlin were opportunistically collected between 1992 and 2014 from the following locations: United States mid-Atlantic coast (USM, n = 263) off Cape May,
NJ and Ocean City, MD; Caribbean Sea (CAR, n = 40) off the Dominican Republic and
30 Cumaná, Venezuela; Gulf of Mexico (GOM Adults, n = 49) off Veracruz, Mexico and from National Marine Fisheries Service pelagic longline survey stations throughout the
Gulf of Mexico; western Central Atlantic (WCA, n = 55) off the northern and central coasts of Brazil; western South Atlantic (WSA, n = 39) off Santos, Brazil; and eastern
North Atlantic (ENA, n = 33) off Morocco (Figure 1). Samples were stored at room temperature in 95% ethanol or a 10% DMSO solution. Biological information including sex, length, and weight was available for a limited number of samples; spawning condition was unknown for most specimens. White marlin larvae sampled from the northern Gulf of Mexico (GOM Larvae, n = 75) between 2006 and 2013 were also evaluated; ichthyoplankton sampling methods are described in Rooker et al. (2012).
Morphological species identification of larval white marlin was confirmed using the microsatellite genotype and mtDNA control region sequence data generated in this study, wherein larvae of species other than white marlin appeared as extreme outliers when compared with genotype and sequence data for adult white marlin.
Isolation of total genomic DNA was performed using standardized kits including the DNeasy Blood and Tissue Kit (Qiagen) and the ZR Genomic DNA Tissue MiniPrep
Kit (Zymo Research). A subset of microsatellite loci previously developed for other istiophorid billfishes were optimized for use in white marlin (Buonaccorsi & Graves,
2000; Purcell et al., 2009; Sorenson et al., 2011; Bernard et al., 2012; Williams et al.,
2015a). Twenty-four loci with consistent PCR amplification in white marlin
(Supplementary Table S1) were developed into multiplex amplification reactions using forward primers modified at the 5’ end with locus-specific tag sequences and fluorescent dye labels. Amplifications were performed using the Type-It Microsatellite PCR Kit
31 (Qiagen) following the manufacturer’s protocol. Amplification products were sized on an ABI 3130XL Genetic Analyzer (Applied Biosystems Inc.), and visualized using
GeneMarker v2.6.0 software (SoftGenetics, LLC). Allele size calls were manually inspected for quality and a subset of samples (20%) were genotyped a second time to verify call consistency.
Generation of mtDNA Sequence Data
A segment of the mtDNA control region (approximately 875 bp) was sequenced for a random subset of individuals from each sample collection and represented ≥ 29% of the total number of individuals in each collection (Table 1). PCR amplifications were performed using Pro-5 forward primer (Palumbi, 1996), an internal reverse primer
(Graves & McDowell, 2006), and a PCR Core Reagents Kit (Qiagen). Amplification products were purified using the QIAquick PCR Purification Kit (Qiagen). Sequencing and consensus sequence construction were performed according to Portnoy et al. (2010), except with use of the MUSCLE multiple sequence alignment algorithm (Edgar, 2004).
Control region sequence data previously generated for white marlin and reported in
Graves & McDowell (2006) were downloaded from GenBank (accession numbers
DQ835191 - DQ835281) and included in the final alignment.
Analyses of Microsatellite Genotype Data
Microsatellite genotype data were evaluated for the presence of scoring errors, null alleles, and large allele dropout using Micro-Checker v2.2.3 (Van Oosterhout et al., 2004,
2006). Loci were tested for selective neutrality using the Fst outlier detection method
32 implemented in Lositan v1.0 (Antao et al., 2008) using 500,000 simulations, a stepwise mutation model, and neutral and forced mean Fst. Conformance to the expectations of
Hardy-Weinberg and linkage equilibriums were evaluated for each locus using Arlequin v3.5 (10,000 iterations; Excoffier & Lischer, 2010) and Genepop v4.5 (10,000 iterations;
Raymond & Rousset, 1995; Rousset, 2008), respectively. Allelic richness and the presence of private alleles were determined using the rarefaction methodology implemented in HP-RARE (version issued February 2009; Kalinowski 2004, 2005); a sample size of 62 genes was utilized to reflect the minimum number of genes observed in a sample collection. To account for the large difference in sample sizes across sample collections, mean expected heterozygosity corrected for uneven sample sizes among sample collections was calculated using GenClone v2.0 (Arnaud-Haond & Belkhir,
2007). Resampling was performed based on a corrected sample size of 33 individuals to reflect the smallest sample collection (ENA), and using 1,000 permutations. Mean expected heterozygosity was also calculated as a ratio based on the comparison of the smallest sample collection (ENA) to every other sample collection using the R package diveRsity and 1,000 bootstrap replicates (Keenan et al., 2013). The standard error for each mean expected heterozygosity ratio was also calculated in diveRsity.
Genetic differentiation across all samples and pairwise differentiation between sample collections was assessed by calculating Fst values in Arlequin v3.5 (Excoffier &
Lischer, 2010). Significance of Fst values was assessed based on 10,000 permutations of the data and a critical value corrected for multiple pairwise comparisons using a modified false discovery rate (Benjamini & Yekutieli, 2001; Narum, 2006). Bayesian model-based cluster analyses were performed in STRUCTURE v2.3.4 (Pritchard et al., 2000) using
33 500,000 MCMC repetitions, 25 iterations of each K, and an admixture ancestry model
(Falush et al., 2007; Hubisz et al., 2009). STRUCTURE simulations were performed with and without the use of sampling location as a prior, and results were compiled in
STRUCTURE Harvester v0.6.94 (Evanno et al., 2005; Earl & VonHoldt, 2012). The scaleGen function in the R package adegenet v1.4-2 was used to perform principal component analysis (PCA) with centered allele frequencies and missing data replaced with mean allele frequencies (Jombart, 2008; Jombart et al., 2009, 2010).
Power analyses were executed in POWSIM v4.1 (Ryman & Palm, 2006) using empirical allele frequency data generated in the present study and in Graves & McDowell
(2006). Two scenarios based on the ability to detect the presence of two populations with a low (Fst = 0.005) or high (Fst = 0.05) level of genetic differentiation were evaluated; these levels were chosen based on the range of genetic differentiation reported in previous studies of genetic population structure in istiophorid billfishes. POWSIM simulates genetic drift in populations of effective size Ne over a specified number of generations t to produce a level of genetic differentiation described by Fst = 1 - (1 -
t 1/2Ne) (Nei, 1987). Many combinations of Ne and t will produce a particular Fst (Ryman
& Palm, 2006). In this study, t was altered between scenarios of low and high genetic differentiation to produce the desired Fst. Both simulation scenarios were completed with
1,000 replications, effective population sizes of 1,500, and 15 (low scenario) or 155 (high scenario) generations of drift. The proportion of significant results (defined as p < 0.05) was used to represent statistical power and was reported using Fisher’s exact test (Ryman
& Palm, 2006). The probability of committing a Type I statistical error was also evaluated in each simulation.
34 Temporal replicates consisting of ≥ 10 individuals sampled from the same geographic location in more than one year were available for some sampling locations
(Supplementary Table S2). These replicates were used to evaluate the consistency of a statistically significant population structure signal over time within a geographic region.
Pairwise Fst values were generated for temporal replicates within a sampling location using Arlequin. Significance of Fst values was assessed based on 10,000 permutations of the data and a critical value corrected for multiple pairwise comparisons (Benjamini &
Yekutieli, 2001; Narum, 2006). The minimum number of samples required for a temporal replicate to be considered in this analysis (n = 10) was chosen based on the minimum number of samples associated with an acceptable (≥ 95% probability of detection) level of statistical power in power analyses performed using the high Fst scenario.
Analyses of mtDNA Sequence Data
Arlequin was used to determine the number of haplotypes and associated measures of haplotype and nucleotide diversity. Arlequin was also used to evaluate genetic differentiation globally across all samples, and pairwise between sample collections. The statistical significance of global and pairwise levels of genetic differentiation was assessed with 10,000 permutations of the data, and by correcting for multiple pairwise comparisons using a modified false discovery rate method (Benjamini & Yekutieli, 2001;
Narum, 2006). A median joining haplotype network was generated in PopArt v1.7 (Leigh
& Bryant, 2015).
35 RESULTS
Summary Statistics & Genetic Diversity
A total of 554 individuals representative of six geographic regions and two demographic groups were genotyped at 24 microsatellite loci (Supplementary Table S1). There was no indication of null alleles or large allele dropout in the genotype data. Genotype calls for the subset of samples genotyped twice were consistent between runs. One microsatellite locus (Isin29) demonstrated statistically significant deviation from the expectations of
Hardy-Weinberg equilibrium in the WSA sample collection (p = 0.004). This deviation was associated with a deficiency of heterozygotes. As there was no pattern of deviation across multiple sample collections, Isin29 was retained for subsequent analyses.
Observed and expected heterozygosities, including mean expected heterozygosity corrected for uneven sample sizes, were similar among sample collections
(Supplementary Tables S1 and S3). No loci were identified as experiencing selection.
The total number of alleles per locus ranged from 3 at Ta149 to 35 at Isin1 (mean a =
16.33); mean allelic richness within sample collections ranged from 9.96 to 10.39 and the mean number of private alleles per sample collection ranged from 0.23 to 0.46. Allelic richness for GOM Larvae (aR = 10.37; Supplementary Table S1) did not differ from allelic richness in any other sample collection (aR = 9.96 - 10.39), nor from the pooled collection of all adult samples (aR = 10.27).
An 858 bp alignment of the mtDNA control region was analyzed for a total of 276 individuals, including 185 newly generated sequences (GenBank accession numbers
K595230 - K595414) and 91 sequences downloaded from GenBank (Table 1).
Nucleotide composition consisted of 31.09% A, 29.37% T, 22.70% C, and 16.84% G.
36 There were 354 polymorphic sites, 311 transitions, and 18 transversions. A total of 242 haplotypes were identified. Haplotype and nucleotide diversities across all samples were
0.99 and 0.031, respectively. Within sample collections, nucleotide diversity ranged from
0.014 to 0.036, and was lowest for USM (0.014) and CAR (0.016). Haplotype and nucleotide diversities in the larval collection (h = 1.00, π = 0.032; Table 1) were similar to that in all other sample collections (h = 0.99 - 1.00, π = 0.014 - 0.036) and to the pooled collection of all adult samples (h = 0.99, π = 0.031).
Population Structure Inferred from Genetic Data
Overall genetic differentiation based on the microsatellite genotype data was not statistically significant (global Fst = -0.00009, p = 0.525). Fst values associated with the pairwise comparison of sample collections ranged from 0 to 0.003 and were not statistically significant, with the exception of the comparison between the larval and adult sample collections from the Gulf of Mexico (Fst = 0.003, p = 0.010; Table 2). Results from STRUCTURE did not indicate the presence of more than one genetic group, and iterations with K = 1 were associated with the highest log likelihood values (Figure 2).
STRUCTURE results were similar between admixture models that did or did not utilize sampling information as a prior to inform clustering. PCA also demonstrated a single grouping of individuals (Supplementary Figure S1). The eigenvalues associated with this analysis were similar across principal components.
Overall genetic differentiation based on the mtDNA sequence data was not statistically significant (global фst = -0.00190, p = 0.614). Levels of genetic differentiation associated with pairwise comparisons of sample collections were not
37 statistically significant (Table 2). The high haplotype diversity of the mtDNA control region was reflected in a midpoint spanning haplotype network (Figure 3). The three clusters of haplotypes apparent in this network were separated by 12 and 17 mutational differences.
Statistical Power
Results from power simulations based on empirical allele frequency data generated in this study and in Graves & McDowell (2006) indicate that when the level of genetic differentiation between populations is relatively high (Fst = 0.05), an acceptable ( 95% probability of detection) level of statistical power is possible with low sample sizes (n
10) per population and either a small (n = 5) or large (n = 24) number of microsatellite markers (Figure 4). When the level of genetic differentiation between populations is low
(Fst = 0.005), considerably larger numbers of samples and microsatellite markers are necessary to provide a suitable level of statistical power. In simulations where Fst = 0.05, both the present study and Graves & McDowell (2006) displayed similarly high levels of statistical power; however, the five microsatellite loci surveyed in the 2006 study did not facilitate a 95% probability of detection at any of the sample sizes explored with simulations when Fst = 0.005. In comparison, the 24 microsatellite loci surveyed in the present study provided an acceptable level of statistical power with sample sizes 40 individuals per putative population in the low differentiation scenario. Two sample collections in this study included < 40 individuals: WSA (n = 39) and ENA (n = 33). The
Type I error rate remained low for all simulation scenarios based on both empirical datasets.
38 Temporal Replicates
Pairwise Fst values were not statistically significant for the two temporal replicates available for the WCA (Fst = 0.0034, p = 0.067) and for the GOM Adults sample collections (Fst = 0.0006, p = 0.419), and were also not significant for the three temporal replicates available for the GOM Larvae sample collection (Fst = -0.0074 to -0.0045, p =
0.911 - 0.983). Fourteen temporal replicates with sample sizes ≥ 10 were available for the
USM sample collection; sample sizes for these replicates ranged from 10 to 42 individuals per year. Statistically significant levels of genetic differentiation were observed for five pairwise comparisons between USM temporal replicates and were consistently associated with replicates from the year 1998 or 2012 (Supplementary Table
S4). Sample sizes for the 1998 and 2012 temporal replicates were 11 and 14, respectively. All microsatellite loci conformed to the assumptions of Hardy-Weinberg equilibrium within all USM temporal replicates, with the exception of MnE in the 1995 replicate (p = 0.008) and MnI in the 1998 replicate (p = 0.011).
DISCUSSION
The objective of this study was to evaluate genetic population structure in white marlin using a higher level of statistical power compared to previous studies. Increased statistical power was achieved by surveying a larger number of molecular markers across greater numbers of samples, including sample collections from additional geographic locations and a collection of larvae. With the increase in statistical power, levels of genetic heterogeneity in the present study decreased compared to those observed in previous studies with lower power. Graves & McDowell (2006) reported global levels of
39 genetic differentiation that approached statistical significance (Fst = 0.0022, p = 0.057;
фst = 0.0163, p = 0.069); however, in the present study global levels of differentiation were considerably reduced (Fst = -0.00009, p = 0.525; фst = -0.00190, p = 0.614). The null hypothesis of a single genetic stock of white marlin could not be rejected based on the lack of genetic heterogeneity observed in this study.
This study represents the first population genetic analysis of an istiophorid billfish species to include a collection of larvae. Ideally, the inclusion of two or more collections of tissue from spawning adults or larvae sampled from geographically distinct spawning grounds would enable the direct evaluation of genetic differentiation between putative source populations of white marlin. However, confirmation of spawning status in adult fish requires direct observation of the release of gametes or the histological examination of gonadal tissue. While the former approach may be possible for actively spawning females that may release hydrated oocytes when external pressure is applied to the ovaries, the latter approach requires the sacrifice of individual fish in order to inspect gonadal tissue. White marlin is an overfished species and the collection of fin clips as in this study makes it unnecessary to sacrifice fish for the collection of tissues for genetic analysis. The collection of larvae from istiophorid billfishes is also challenging due to limited information on spawning in these species, and due to the logistics of obtaining robust sample sizes. In this study, sample collections of adult fish were not known to include actively spawning individuals, and only a single larval collection was available for genetic analysis.
Nevertheless, evaluation of a single collection of larvae can be informative of population structure when compared with collections of adults from throughout the
40 species’ range. In the case of genetically discrete populations, a larval collection would be expected to display a statistically significant level of genetic differentiation when compared to collections of adults representative of another source population. In the present study, there were no statistically significant levels of genetic differentiation between sample collections based on either the microsatellite or the mtDNA sequence data, with the exception of the microsatellite-based comparison of GOM Larvae and
GOM Adults. The statistically significant heterogeneity observed between the Gulf of
Mexico sample collections was driven by two loci (Ta149 and Isin29; Supplementary
Figure S2) and resulted in an Fst for which the statistical significance was just below the critical value (p = 0.010, pcrit = 0.014). This result was likely due to random sampling error and is not likely to be biologically meaningful (Waples, 1998). The lack of support for more than one genetic cluster in the STRUCTURE results is also consistent with the observed lack of genetic differentiation among sample collections, although the ability of
STRUCTURE to estimate the true number of populations is reduced when genetic differentiation between populations is low (Waples & Gaggiotti, 2006). In addition, observed levels of genetic differentiation between sample collections based on the mtDNA control region sequence data may be influenced by homoplasy at this gene region (Reeb et al., 2010; Bradman et al., 2011); however, support for population subdivision based on the nuclear genotype data is lacking.
If genetically discrete populations exist, a larval collection may also display lower genetic diversity relative to that of a pooled collection of adult samples. Additionally, genetic diversity in the larval collection may also be lower than non-pooled geographically distant sample collections if those collections comprised mixed stocks. In
41 this study, levels of genetic diversity based on the microsatellite and the mtDNA sequence data were similar between the Gulf of Mexico larval collection and the pooled collection of all adult samples. Levels of genetic diversity based on both marker types were also similar between the larval collection and all individual (non-pooled) sample collections. Collectively, the absence of genetic differentiation and similar levels of genetic diversity among sample collections, particularly between the larval collection and the pooled collection of all adult samples, do not provide sufficient evidence to reject the null hypothesis of a single genetic stock of white marlin in the Atlantic Ocean.
The lack of statistically significant genetic heterogeneity observed for white marlin in this study differs from results reported by Graves & McDowell (2006). The increase in statistical power associated with the present study resulted in decreased levels of global genetic differentiation compared to the earlier study, and reduced genetic differentiation between sample collections from the western South Atlantic and the
United States mid-Atlantic coast. Power analysis simulations based on empirical allele frequency data indicate that the number of loci and samples analyzed in the present study facilitated relatively high statistical power even at a low level of genetic differentiation between simulated populations. For example, with sample sizes of 35 individuals and an
Fst of 0.005, the current study was associated with a nearly 90% probability of detecting genetic differentiation. In comparison, power analysis simulations based on empirical allele frequency data from the five microsatellite markers surveyed by Graves &
McDowell (2006) indicate a less than 60% probability of detecting genetic differentiation with sample sizes of 35 individuals and an Fst of 0.005. For istiophorid billfishes in which the analysis of microsatellite markers has previously revealed population subdivision,
42 statistically significant Fst values ranged from 0.007 to 0.047 (McDowell & Graves,
2008; Purcell & Edmands, 2011; Lu et al., 2015; Williams et al., 2015b). This suggests that statistical power associated with the present study was high enough to facilitate the detection of genetic population subdivision in white marlin, if it exists. Three of the loci surveyed by Graves & McDowell (2006) were not included in the present study, and additional analysis of the 2006 data indicates that the genetic heterogeneity observed in that study was not driven by one locus in particular. The lack of genetic heterogeneity observed in the present study compared with Graves & McDowell (2006) is likely due to the large difference in statistical power resulting from the limited number of microsatellite markers surveyed in the earlier study.
The temporal replicates evaluated in this study also provide useful insights into the importance of statistical power in interpreting the results of population genetic studies. Collectively, results of this study lack any evidence to suggest the presence of more than one genetic stock of white marlin; however, a number of statistically significant pairwise comparisons were observed between temporal replicates from the
USM sample collection. These statistically significant comparisons are presumed to reflect random noise rather than a true population structuring signal. The USM temporal replicates consisted of sample sizes ranging from 10 to 42 individuals per replicate; however, only replicates of very small sample sizes (1998, n = 11; 2012, n = 14) were associated with statistically significant pairwise comparisons. These results suggest that when statistical power is low due to small sample sizes and/or limited numbers of molecular markers, the ability to distinguish between noise and a true population structuring signal is diminished, resulting in a higher probability of detecting a false
43 signal and committing a Type I statistical error (Waples, 1998). Conversely, low statistical power may also result in failure to detect a low but biologically meaningful population structuring signal and lead to a Type II statistical error.
The apparent lack of genetic population structure in white marlin has also been observed for other istiophorid billfishes. Previous evaluations of genetic population structure for blue marlin (Makaira nigricans) in the Atlantic Ocean have failed to detect population subdivision despite the analysis of multiple types of nuclear and mitochondrial markers (Buonaccorsi et al., 1999, 2001; McDowell et al., 2007). A lack of population subdivision was also reported for sailfish (Istiophorus platypterus) in the
Atlantic Ocean based on the analysis of mtDNA control region sequence data and a small number (n = 5) of microsatellite loci (McDowell, 2002; McDowell & Graves, 2002).
Analysis of mtDNA control region sequence data for blue marlin and sailfish has demonstrated the presence of two distinct mitochondrial lineages for both species in the
Atlantic Ocean (McDowell & Graves, 2002; McDowell et al., 2007). The existence of these lineages is attributed to historical isolation between Atlantic and Indo-Pacific populations of these species, followed by subsequent re-introduction of Indo-Pacific fish to the Atlantic Ocean (Graves & McDowell, 1995). In the current study, haplotype diversity at the mtDNA control region is high, but distinct lineages are not present in white marlin. This result is consistent with a lack of historical isolation in white marlin compared to blue marlin and sailfish, perhaps due to differences in thermal tolerance and/or the spatial distributions of these species.
Despite the apparent lack of genetic population structure for some istiophorid billfishes in the Atlantic Ocean, genetic heterogeneity has been demonstrated for other
44 highly migratory marine fishes in this ocean. Analyses of mtDNA control region sequence data and a number of nuclear loci, including microsatellite markers, revealed distinct populations of swordfish (Xiphias gladius) in the North Atlantic and South
Atlantic (Chow & Takeyama, 2000; Alvarado-Bremer et al., 2005; Kasapidis et al.,
2006). These populations correspond with the presence of distinct spawning grounds in the northern and southern hemispheres (Alvarado-Bremer et al., 2005). In the Pacific
Ocean, genetic population structure has been reported for sailfish and black marlin
(Istiompax indica; McDowell, 2002; Lu et al., 2015; Williams et al., 2015b). In addition, the presence of at least four stocks has been genetically determined for striped marlin (K. audax), the closely related sister species of white marlin, in the Pacific Ocean (McDowell
& Graves, 2008; Purcell & Edmands, 2011). STRUCTURE analysis of genotype data previously generated for striped marlin (McDowell & Graves, 2008) and based on only five microsatellite loci clearly demonstrates the presence of four genetic clusters (Figure
2). While Figure 2 provides an interesting comparison of the level of genetic subdivision observed for congeneric species that inhabit different ocean basins, the lack of populations resolved by STRUCTURE for white marlin may only reflect the limitations of this analysis in resolving subtle population structure in this species.
The genetic homogeneity observed for white marlin in this study may be consistent with what we currently know about the biology of this species. The distribution of catches based on fisheries-dependent data suggests a continuous species distribution in the Atlantic Ocean (ICCAT, 2015). Movements of white marlin inferred from the tagging of individual fish include a small number of conventional tag recaptures representing trans-oceanic (n = 7) or trans-equatorial (n = 3) dispersal (Snodgrass et al.,
45 2013). Information from electronic tags reflects highly inconsistent directionality of movements by individuals tagged from seasonal assemblages of white marlin (Loose,
2014; Schlenker, 2014). In addition, seasonally displaced spawning grounds have been confirmed in the northern and southern hemispheres, including locations in the western
North Atlantic, Gulf of Mexico, and Caribbean Sea for which spawning appears to primarily occur in the second quarter of the year (de Sylva & Breder, 1997; Luthy et al.,
2005; Prince et al., 2005; Arocha & Barrios, 2009; Richardson et al., 2010; Rooker et al.,
2012), and in the western South Atlantic where spawning primarily occurs in the fourth quarter (Ueyanagi et al., 1970; Arfelli et al., 1986; Amorim & Arfelli, 2003; Schmidt et al., 2015). Because the spawning season in these regions is not concurrent, it is possible that white marlin may spawn in multiple locations within a year, facilitating gene flow among geographically distant regions and resulting in the lack of genetic subdivision observed in this study.
The inability to reject the null hypothesis of a single genetic stock of white marlin in this study does not necessarily mean that ecological stocks of this species do not exist.
Results of this study suggest that ocean-wide connectivity in white marlin is sufficient to prohibit the detection of genetic heterogeneity with the molecular markers surveyed in this study, but the magnitude of intraspecific connectivity beyond this threshold is unknown and could reflect dispersal in a small to large proportion of individuals. A 365- day geolocation track from Loose (2014) provides evidence that some white marlin complete cyclical migrations to seasonal foraging and spawning grounds on an annual time scale. Rooker et al. (2013, Supplementary Figure S5) also report cyclical movement for white marlin in the Gulf of Mexico based on a year-long geolocation track.
46 Conventional tag recapture data also suggest annual periodicity associated with feeding and spawning assemblages of white marlin (Ortiz et al., 2003). If annual migrations are associated with some level of fidelity to discrete spawning grounds, this could result in the presence of ecological stocks; however, whether white marlin display spawning site fidelity is currently unknown. Atlantic-wide electronic tagging efforts that incorporate multi-year deployment periods would provide an informative approach for inferring ocean-wide connectivity in white marlin; however, tag deployment periods of this duration have not been previously successful.
Recently developed next-generation sequencing technology now facilitates the discovery of thousands of molecular markers representative of the whole genome, providing an unprecedented ability to detect genetic population structure (Davey &
Blaxter, 2010; Narum et al., 2013). For example, application of this methodology to yellowfin tuna (Thunnus albacares; Grewe et al., 2015; Pecoraro et al., 2016) revealed inter- and intra-oceanic genetic heterogeneity previously unresolved by molecular markers such as allozymes, mtDNA, and microsatellites (Ward et al., 1994; Diaz-James
& Uribe-Alcocer, 2006). Elucidation of unresolved or cryptic genetic heterogeneity can have significant consequences for fisheries management, particularly for stocks currently threatened by unsustainable fishing practices. The evaluation of genome-wide molecular markers may significantly improve the ability to detect genetic population structure, especially for weakly differentiated populations of marine species. In addition, genome- wide methodologies facilitate the identification of putative adaptive loci, which may provide insights into localized adaptation and regional genetic diversity, even if the level
47 of intraspecific connectivity is enough to obscure heterogeneity at neutral markers
(Allendorf et al., 2008; Andrews et al., 2016).
The ability to detect genetic population structure in white marlin was considerably improved in this study by increasing the number of molecular markers surveyed across a larger number of samples, including collections from additional geographic locations and larvae from the Gulf of Mexico. This increase in statistical power resulted in lower levels of genetic heterogeneity relative to those reported in a previous study that utilized fewer molecular markers and lower numbers of opportunistically collected samples. These results highlight the importance of statistical power in population genetic assessments of species for which population subdivision, if present, is expected to be shallow. Results from this study are consistent with the presence of a single genetic stock of white marlin in the Atlantic Ocean, and with the single-stock assessment and management model currently utilized by ICCAT. However, these inferences are based on molecular markers that may be limited in their ability to resolve low levels of genetic differentiation, and the degree of Atlantic-wide connectivity beyond that required to mask heterogeneity using the molecular markers surveyed in this study is unknown. It is especially important that
Atlantic-wide connectivity in white marlin is understood considering the most recent stock assessment identified this species as overfished, with biomass in 2010 at half of that necessary to produce maximum sustainable yield (ICCAT, 2012). Additional investigation using genome-wide molecular markers and collections of larvae and/or reproductively active adults from additional spawning locations to inform inferences on ocean-wide genetic connectivity, as well as studies to determine the biological significance of genetic results, are warranted.
48 ACKNOWLEDGMENTS
We thank a number of recreational anglers throughout the Atlantic region for providing samples for this study. We also thank the Mid-Atlantic Tournament for facilitating the opportunity to collect samples over a number of years. This study was supported by a grant from NOAA-NMFS and is Virginia Institute of Marine Science, College of William and Mary Contribution No. 3627.
49 REFERENCES
Allendorf, F., England, P., Luikart, G., Ritchie, P., and Ryman, N. 2008. Genetic effects
of harvest on wild animal populations. Trends in Ecology and Evolution, 23: 327-
337.
Allendorf, F., and Phelps, S. 1981. Use of allelic frequencies to describe population
structure. Canadian Journal of Fisheries and Aquatic Sciences, 38: 1507-1514.
Allendorf, F., Ryman, N., and Utter, F. 1987. Genetics and fishery management: past,
present, and future. In Population Genetics and Fishery Management, pp. 1-19. Ed.
by N. Ryman, and F. Utter, University of Washington Press. 440 pp.
Alvarado Bremer, J., Vinas, J., Mejuto, J., Ely, B., and Pla, C. 2005. Comparative
phylogeography of Atlantic bluefin tuna and swordfish: the combined effects of
vicariance, secondary contact, introgression, and population expansion on the
regional phylogenies of two highly migratory pelagic fishes. Molecular
Phylogenetics and Evolution, 36: 169-187.
Amorim, A., and Arfelli, C. 2003. Review of white marlin (Tetrapturus albidus) fishery
biology off the southern Brazilian coast (1971-2001). ICCAT Collective Volume
of Scientific Papers, 55: 467-474.
Andrews, K., Good, J., Miller, M., Luikart, G., and Hohenlohe, P. 2016. Harnessing the
power of RADseq for ecological and evolutionary genetics. Nature Reviews
Genetics, 17: 81-92.
Antao, T., Lopes, A., Lopes, R., Beja-Pereira, A., and Luikart, G. 2008. LOSITAN: a
workbench to detect molecular adaptation based on a Fst-outlier method. BMC
Bioinformatics, 9: 323.
50 Arfelli, C., Amorim, A., and Galhardo-Amado, J. 1986. Analysis on Tetrapturus albidus
Poey (1861) caught off south and southeast of Brazil (1971-1984). ICCAT
Collective Volume of Scientific Papers, 25: 202-217.
Arnaud-Haond, S., and Belkhir, K. 2007. Genclone: a computer program to analyse
genotypic data, test for clonality and describe spatial clonal organization.
Molecular Ecology Notes, 7: 15-17.
Arocha, F., and Barrios, A. 2009. Sex ratios, spawning seasonality, sexual maturity, and
fecundity of white marlin (Tetrapturus albidus) from the western central Atlantic.
Fisheries Research, 95: 98-111.
Benjamini, Y., and Yekutieli, D. 2001. The control of false discovery rate under
dependency. Annals of Statistics, 29: 1165-1188.
Bernard, A., Feldheim, K., and Shivji, M. 2012. Development and characterization of 11
novel microsatellite loci for the roundscale spearfish Tetrapturus georgii and their
cross-species amplification among other istiophorid species. Journal of Fish
Biology, 81: 1781-1786.
Boustany, A., Reeb, C., and Block, B. 2008. Mitochondrial DNA and electronic tracking
reveal population structure of Atlantic bluefin tuna (Thunnus thynnus). Marine
Biology, 156: 13-24.
Bowen, B., Bass, A., Soares, L., and Toonen, R. 2005. Conservation implications of
complex population structure: lessons from the loggerhead turtle (Caretta caretta).
Molecular Ecology, 14: 2389-2402.
51 Bradman, H., Grewe, P., and Appleton, B. 2011. Direct comparison of mitochondrial
markers for the analysis of swordfish population structure. Fisheries Research,
109: 95-99.
Buonaccorsi, V., and Graves, J. 2000. Isolation and characterization of novel
polymorphic tetra-nucleotide microsatellite markers from the blue marlin, Makaira
nigricans. Molecular Ecology, 9: 820-821.
Buonaccorsi, V., McDowell, J., and Graves, J. 2001. Reconciling patterns of inter-ocean
molecular variance from four classes of molecular markers in blue marlin
(Makaira nigricans). Molecular Ecology, 10: 1179-1196.
Buonaccorsi, V., Reece, K., Morgan, L., and Graves, J. 1999. Geographic distribution of
molecular variance within the blue marlin (Makaira nigricans): a hierarchical
analysis of allozyme, single-copy nuclear DNA, and mitochondrial DNA markers.
Evolution, 53: 568-579.
Carlsson, J., McDowell, J., Carlsson, J., and Graves, J. 2007. Genetic identity of YOY
bluefin tuna from the eastern and western Atlantic spawning areas. The Journal of
Heredity, 98: 23-28.
Chow, S., and Takeyama, H. 2000. Nuclear and mitochondrial DNA analyses reveal four
genetically separated breeding units of swordfish. Journal of Fish Biology, 56:
1087-1098.
Davey, J., and Blaxter, M. 2010. RADSeq: next-generation population genetics. Briefings
in Functional Genomics, 9: 416-423.
52 de Sylva, D., and Breder, P. 1997. Reproduction, gonad histology, and spawning cycles
of North Atlantic billfishes (Istiophoridae). Bulletin of Marine Science, 60: 668-
697.
Diaz-Jaimes, P., and Uribe-Alcocer, M. 2006. Spatial differentiation in the eastern Pacific
yellowfin tuna revealed by microsatellite variation. Fisheries Science, 72: 590-596.
Earl, D., and VonHoldt, B. 2012. STRUCTURE HARVESTER: a website and program
for visualizing STRUCTURE output and implementing the Evanno method.
Conservation Genetics Resources, 4: 359-361.
Edgar, R. 2004. MUSCLE: multiple sequence alignment with high accuracy and high
throughput. Nucleic Acids Research, 32: 1792-1797.
Evanno, G., Regnaut, S., and Goudet, J. 2005. Detecting the number of clusters of
individuals using the software STRUCTURE: a simulation study. Molecular
Ecology, 14: 2611-2620.
Excoffier, L., and Lischer, H. 2010. Arlequin suite ver 3.5: a new series of programs to
perform population genetics analyses under Linux and Windows. Molecular
Ecology Resources, 10: 564-567.
Falush, D., Stephens, M., and Pritchard, J. 2007. Inference of population structure using
multilocus genotype data: dominant markers and null alleles. Molecular Ecology
Notes, 7: 574-578.
Fromentin, J. –M., and Powers, J. E. 2005. Atlantic bluefin tuna: population dynamics,
ecology, fisheries and management. Fish and Fisheries, 6: 281-306.
Graves, J., Gold, J., Ely, B., Quattro, J., Woodley, C., and Dean, J. 1996. Population
genetic structure of bluefin tuna in the North Atlantic Ocean. Identification of
53 variable genetic markers. ICCAT Collective Volume of Scientific Papers, 45: 155-
157.
Graves, J., and McDowell, J. 1995. Inter-ocean genetic divergence of istiophorid
billfishes. Marine Biology, 122: 193-203.
Graves, J., and McDowell, J. 2001. A genetic perspective on the stock structures of blue
marlin and white marlin in the Atlantic Ocean. ICCAT Collective Volume of
Scientific Papers, 53: 180-187.
Graves, J., and McDowell, J. 2006. Genetic analysis of white marlin (Tetrapturus
albidus) stock structure. Bulletin of Marine Science, 79: 469-482.
Graves, J., and McDowell, J. 2015. Population structure of istiophorid billfishes.
Fisheries Research, 166: 21-28.
Grewe, P., Feutry, P., Hill, P., Gunasekera, R., Schaefer, K., Itano, D., Fuller, D., Foster,
S., and Davies, C. 2015. Evidence of discrete yellowfin tuna (Thunnus albacares)
populations demands rethink of management for this globally important resource.
Scientific Reports, 5: 16916.
Hedgecock, D. 1994. Does variance in reproductive success limit effective population
sizes of marine organisms? In Genetics and Evolution of Aquatic Organisms, pp.
122-134. Ed. By A. Beaumont. Chapman and Hall, London.
Hubisz, M., Falush, D., Stephens, M., and Pritchard, J. 2009. Inferring weak population
structure with the assistance of sample group information. Molecular Ecology
Resources, 9: 1322-1332.
ICCAT. 1994. Report of the second ICCAT billfish workshop. ICCAT Collective
Volume of Scientific Papers, 41: 13-166.
54 ICCAT. 2012. Report of the 2012 white marlin stock assessment meeting. ICCAT
Collective Volume of Scientific Papers, 55: 1-69.
ICCAT. 2015. Statistical bulletin: volume 42(2) (1960-2013). Madrid, Spain.
Jombart, T. 2008. Adegenet: a R package for the multivariate analysis of genetic markers.
Bioinformatics, 24: 1403-1405.
Jombart, T., Devillard, S., and Balloux, F. 2010. Discriminant analysis of principal
components: a new method for the analysis of genetically structure populations.
BMC Genetics, 11: 1471-2156.
Jombart, T., Pontier, D., and Dufour, A. –B. 2009. Genetic markers in the playground of
multivariate analysis. Heredity, 102: 330-341.
Kalinowski, S. 2004. Counting alleles with rarefaction: private alleles and hierarchical
sampling designs. Conservation Genetics, 5: 539-543.
Kalinowski, S. 2005. HP-RARE 1.0: a computer program for performing rarefaction on
measures of allelic richness. Molecular Ecology Notes, 5: 187-189.
Kasapidis, P., Mejuto, J., Tserpes, G., Antoniou, A., Garcia-Cortes, B., Peristeraki, P.,
Oikonomaki, K., Kotoulas, G., and Magoulas, A. 2006. Genetic structure of the
swordfish (Xiphias gladius) stocks in the Atlantic using microsatellite DNA
analysis. ICCAT Collective Volume of Scientific Papers, 61: 89-98.
Keenan, K., McGinnity, P., Cross, T., Crozier, W., and Prodohl, P. 2013. diveRsity: an R
package for the estimation and exploration of population genetics parameters and
their associated errors. Methods in Ecology and Evolution, 4: 782-788.
Leigh, J., and Bryant, D. 2015. POPART: full-feature software for haplotype network
construction. Methods in Ecology and Evolution, 6: 1110-1116.
55 Loose, E. 2014. Seasonal movements, habitat utilization, and comparative scale
morphology of white marlin (Kajikia albida) and roundscale spearfish
(Tetrapturus georgii). Unpublished master’s thesis. Virginia Institute of Marine
Science, Gloucester Point, VA.
Lu, C. –P., Alvarado Bremer, J., McKenzie, J., and Chiang, W. –C. 2015. Analysis of
sailfish (Istiophorus platypterus) population structure in the North Pacific Ocean.
Fisheries Research, 166: 33-38.
Luthy, S. Cowen, R., Serafy, J., and McDowell, J. 2005. Toward identification of larval
sailfish (Istiophorus platypterus), white marlin (Tetrapturus albidus), and blue
marlin (Makaira nigricans) in the western North Atlantic Ocean. Fisheries
Bulletin, 103: 588-600.
McDowell, J. 2002. Genetic stock structure of the sailfish, Istiophorus platypterus, based
on nuclear and mitochondrial DNA. Doctoral dissertation. Virginia Institute of
Marine Science, College of William & Mary.
McDowell, J., Carlsson, J., and Graves, J. 2007. Genetic analysis of blue marlin (Makaira
nigricans) stock structure in the Atlantic Ocean. Gulf and Caribbean Research, 19:
75-82.
McDowell, J., and Graves, J. 2002. A genetic perspective on Atlantic sailfish stock
structure. ICCAT Collective Volume of Scientific Papers, 54: 805-810.
McDowell, J., and Graves, J. 2008. Population structure of striped marlin (Kajikia audax)
in the Pacific Ocean based on analysis of microsatellite and mitochondrial DNA.
Canadian Journal of Fisheries and Aquatic Sciences, 65: 1307-1320.
56 Nakamura, I. 1985. Billfishes of the world: an annotated and illustrated catalogue of
marlins, sailfishes, spearfishes, and swordfishes known to date. FAO Fisheries
Synopsis No. 125, vol. 5. 65 pp. Rome, Italy: Food and Agricultural Organization.
Narum, S. 2006. Beyond Bonferroni: less conservative analyses for conservation
genetics. Conservation Genetics, 7: 783-787.
Narum, S., Buerkle, C., Davey, J., Miller, M., and Hohenlohe, P. 2013. Genotyping-by-
sequencing in ecological and conservation genomics. Molecular Ecology, 22,
2841-2847.
Nei, M. 1987. Molecular evolutionary genetics. New York: Columbia University Press.
Ortiz, M., Prince, E., Serafy, J., Holts, D., Davy, K., Pepperell, J., Lowry, M., and
Holdsworth, J. C. 2003. Global overview of the major constituent-based billfish
tagging programs and their results since 1954. Marine and Freshwater Research,
54: 489-507.
Ovenden, J., Berry, O., Welch, D., Buckworth, R., and Dichmont, C. 2015. Ocean’s
eleven: a critical evaluation of the role of population, evolutionary and molecular
genetics in the management of wild fisheries. Fish and Fisheries, 16: 125-159.
Palumbi, S. 1996. Nucleic acids II. the polymerase chain reaction. In Molecular
systematics, 2nd edn, pp. 205-247. Ed. by D. Hillis, C. Moritz, and B. Mable.
Sinauer Associates.
Pecoraro, C., Babbucci, M., Villamor, A., Franch, R., Papetti, C., Leroy, B. Ortega-
Garcia, S., et al. 2016. Methodological assessment of 2b-RAD genotyping
technique for population structure inferences in yellowfin tuna (Thunnus
albacares). Marine Genomics, 25: 43-48.
57 Portnoy, D., McDowell, J., Heist, E., Musick, J., and Graves, J. 2010. World
phylogeography and male-mediated gene flow in the sandbar shark, Carcharhinus
plumbeus. Molecular Ecology, 19: 1994-2010.
Prince, E., Cowen, R., Orbesen, E., Luthy, S., Llopiz, J., Richardson, D., and Serafy, J.
2005. Movements and spawning of white marlin (Tetrapturus albidus) and blue
marlin (Makaira nigricans) off Punta Cana, Dominican Republic. Fisheries
Bulletin, 103: 659-669.
Pritchard, J., Stephens, M., and Donnelly, P. 2000. Inference of population structure
using multilocus genotype data. Genetics, 155: 945-959.
Purcell, C., and Edmands, S. 2011. Resolving the genetic structure of striped marlin,
Kajikia audax, in the Pacific Ocean through spatial and temporal sampling of adult
and immature fish. Canadian Journal of Fisheries and Aquatic Sciences, 68: 1861-
1875.
Purcell, C., Harrison, J., and Edmands, S. 2009. Isolation and characterization of 10
polymorphic microsatellite markers from striped marlin, Tetrapturus audax.
Molecular Ecology Resources, 9: 1556-1559.
Raymond, M., and Rousset, F. 1995. GENEPOP (Version 1.2): population genetics
software for exact tests and ecumenicism. Journal of Heredity, 86: 248-249.
Reeb, C., Arcangeli, L., and Block, B. 2000. Structure and migration corridors in Pacific
populations of the swordfish Xiphius gladius, as inferred through analyses of
mitochondrial DNA. Marine Biology, 136: 1123-1131.
58 Richardson, D., Llopiz, J., Guigand, C., and Cowen, R. 2010. Larval assemblages of
large and medium-sized pelagic species in the Straits of Florida. Progress in
Oceanography, 86: 8-20.
Rooker, J., Alvarado-Bremer, J., Block, B., Dewar, H., De Metrio, G., Corriero, A.
Kraus, R., et al. 2007. Life history and stock structure of Atlantic bluefin tuna
(Thunnus thynnus). Reviews in Fisheries Science, 15: 265-310.
Rooker, J., Kitchens, L., Dance, M., Wells, R., Falterman, B., and Cornic, M. 2013.
Spatial, temporal, and habitat-related variation in abundance of pelagic fishes in
the Gulf of Mexico: potential implications of the Deepwater Horizon oil spill.
PLoS ONE, 8: e76080.
Rooker, J., Simms, J., Wells, R., Holt, S., Holt, G., Graves, J., and Furey, N. 2012.
Distribution and habitat associations of billfish and swordfish larvae across
mesoscale features in the Gulf of Mexico. PLoS ONE, 7: e34180.
Rousset, F. 2008. GENEPOP’007: a complete re-implementation of the GENEPOP
software for Windows and Linux. Molecular Ecology Resources, 8: 103-106.
Ryman, N., and Palm, S. 2006. POWSIM: a computer program for assessing statistical
power when testing for genetic differentiation. Molecular Ecology Notes, 6: 600-
602.
Ryman, N., Palm, S., Andre, C., Carvalho, G., Dahlgren, T., Jorde, P., Laikre, L.,
Larsson, L., Palme, A., and Ruzzante, D. 2006. Power for detecting genetic
divergence: differences between statistical methods and marker loci. Molecular
Ecology, 15: 2031-2045.
59 Schlenker, L. 2014. Physiological stress and post-release mortality of white marlin
(Kajikia albida) caught in the U.S. recreational fishery. Master’s thesis. Virginia
Institute of Marine Science, College of William & Mary.
Schmidt, R., Rodrigues, T., Pimenta, E., Hilsdorf, A., and Amorim, A. 2015. Preliminary
occurrence of Istiophoridae larvae (Perciformes, Xiphioidei) in southern Brazil.
ICCAT Collective Volume of Scientific Papers, 71: 2256-2261.
Snodgrass, D., Orbesen, E., Hoolihan, J., and Prince, E. 2011. The U.S. conventional
tagging database updates for Atlantic white marlin (1954-2008). ICCAT Collective
Volume of Scientific Papers, 66: 1760-1766.
Sorenson, L., McDowell, J., and Graves, J. 2011. Isolation and characterization of
microsatellite markers for blue marlin, Makaira nigricans. Conservation Genetics
Resources, 3: 721-723.
Tessier, N., and Bernatchez, L. 1999. Stability of population structure and genetic
diversity across generations assessed by microsatellites among sympatric
populations of landlocked Atlantic salmon (Salmo salar L.). Molecular Ecology, 8:
169-179.
Ueyanagi, S., Kikawa, S., Uto, M., and Nishikawa, Y. 1970. Distribution, spawning, and
relative abundance of billfishes in the Atlantic Ocean. Bulletin of the Far Seas
Fisheries Research Laboratory, 3: 15-25.
Van Oosterhout, C., Hutchinson, W., Wills, D., and Shipley, P. 2004. MICRO-
CHECKER: software for identifying and correcting genotyping errors in
microsatellite data. Molecular Ecology Notes, 4: 535-538.
60 Van Ooserhout, C., Weetman, D., and Hutchinson, W. 2006. Estimation and adjustment
of microsatellite null alleles in nonequilbrium populations. Molecular Ecology
Notes, 6: 255-256.
Waples, R. 1998. Separating the wheat from the chaff: patterns of genetic differentiation
in high gene flow species. The Journal of Heredity, 89: 438-450.
Waples, R., and Gaggiotti, O. 2006. What is a population? An empirical evaluation of
some genetic methods for identifying the number of gene pools and their degree of
connectivity. Molecular Ecology, 15: 1419-1439.
Waples, R., and Teel, D. 1990. Conservation genetics of Pacific salmon I. Temporal
changes in allele frequency. Conservation Biology, 4: 144-156.
Ward, R. 2000. Genetics in fisheries management. Hydrobiologia, 420: 191-201.
Ward, R., Elliott, N., Grewe, P., and Smolenski, A. 1994. Allozyme and mitochondrial
DNA variation in yellowfin tuna (Thunnus albacares) from the Pacific Ocean.
Marine Biology, 118: 531-539.
Ward, R., Woodwark, M., and Skibinski, D. 1994. A comparison of genetic diversity
levels in marine, freshwater, and anadromous fishes. Journal of Fish Biology, 44:
213-232.
Williams, S., Bennett, M., Pepperell, J., Morgan, J., and Ovenden, J. 2015b. Spatial
genetic subdivision among populations of the highly migratory black marlin
Istiompax indica within the central Indo-Pacific. Marine and Freshwater Research,
166: 29-32.
61 Williams, S., Pepperell, J., Corley, S., and Ovenden, J. 2015a. Isolation and
characterisation of 18 polymorphic microsatellite loci for black marlin (Istiompax
indica) and their utility for Pacific billfish species. Fisheries Research, 166: 29-32.
62
Table 1. Summary information for mitochondrial DNA control region sequence data. Sample collections are labeled as in Figure 1. h = haplotype diversity, π = nucleotide diversity.
No. Sample No. No. Polymorphic h π Collection Sequences Haplotypes Sites WSA 17 17 134 1.00 0.036 WCA 40 39 166 0.99 0.031 USM 75 73 225 0.99 0.014 CAR 37 34 178 0.99 0.016 ENA 32 32 165 1.00 0.032 GOM Adults 44 41 185 0.99 0.032 GOM Larvae 31 31 176 1.00 0.032 Pooled Adults 245 216 347 0.99 0.031 All Samples 276 242 354 0.99 0.031
63
Table 2. Matrix of pairwise Fst values based on microsatellite data (below diagonal) and pairwise фst values based on mitochondrial DNA sequence data (above diagonal). Significance values are shown in parentheses, and statistically significant comparisons are italicized. A critical value corrected for multiple pairwise comparisons and equal to 0.014 was used. Sample collections are labeled as in Figure 1.
GOM GOM WSA WCA USM CAR ENA Adults Larvae
0.01796 0.01175 -0.00206 -0.00170 -0.00139 -0.01356 WSA -- (0.126) (0.167) (0.411) (0.407) (0.398) (0.760)
-0.00015 -0.00142 0.00561 -0.00386 -0.00445 -0.00912 WCA -- (0.447) (0.471) (0.221) (0.544) (0.578) (0.800)
0.00006 -0.00125 0.01025 -0.00676 0.00285 -0.00421 USM -- (0.444) (0.964) (0.109) (0.769) (0.253) (0.613)
-0.00118 -0.00066 -0.00053 -0.01313 -0.00283 -0.00704 CAR -- (0.733) (0.607) (0.690) (0.923) (0.474) (0.651)
0.00008 -0.00160 -0.00049 0.00018 -0.00542 -0.01261 ENA -- (0.440) (0.783) (0.643) (0.448) (0.596) (0.904)
GOM 0.00253 -0.00025 0.00128 0.00283 0.00134 -0.01496 -- Adults (0.059) (0.478) (0.069) (0.041) (0.209) (0.987)
GOM -0.00092 -0.00166 -0.00035 -0.00054 0.00102 0.00308 -- Larvae (0.688) (0.915) (0.698) (0.601) (0.222) (0.010)
64
USM ENA GOM
CAR
WCA
WSA
Figure 1. Geographic sampling locations for sample collections. Points reflect representative sampling location for each region. USM = United States mid-Atlantic, GOM = Gulf of Mexico, CAR = Caribbean, WCA = western Central Atlantic, WSA = western South Atlantic, ENA = eastern North Atlantic.
65
Figure 2. STRUCTURE results based on microsatellite genotype data generated for 24 loci in white marlin in this study (Panel A; plot shown for K = 3) and for 5 loci in striped marlin in McDowell & Graves (2008; Panel B; plot shown for K = 4). STRUCTURE results were similar with or without the use of sampling location as a prior to inform clustering. Individuals are organized by sample collection for both species. White marlin sample collections are labeled as in Figure 1. Sample collection labels for striped marlin are as follows: AUS = Port Stephens, Australia, CAL = San Diego, California, ECU = Manta, Ecuador, HAW = Kona, Hawaii, MEX = Cabo San Lucas, Mexico, TAW = Taiwan, and JPN = Japan.
66
WSA WCA USM CAR GOM Adults 10 samples ENA 1 sample GOM Larvae
Figure 3. Median joining haplotype network generated from mitochondrial DNA control region sequence data. Ancestral nodes inferred from observed data are shown in black. The number of mutational differences between nodes are represented by hatch marks along edges. Nodes are colored according to sample collection, and node size corresponds to the number of occurrences for a particular haplotype. Sample collections are labeled as in Figure 1.
67
Figure 4. Results from power analyses based on empirical allele frequency data generated in the present study (black lines) and in Graves & McDowell (2006; red lines). Solid lines reflect scenario with high level of genetic differentiation between simulated populations (Fst = 0.05), dashed lines reflect low level of genetic differentiation between simulated populations (Fst = 0.005). Dotted lines reflect alpha level of statistical error, and blue line represents 0.95 probability of detection.
68
SUPPLEMENTARY MATERIALS
Table S1. Summary statistics for the 24 microsatellite loci surveyed in this study. SC = sample collection, a = observed number of alleles, aR = rarefaction allelic richness, aP = rarefaction private allelic richness, HO = observed heterozygosity, HE = expected heterozygosity, P = statistical significance associated with comparison of HO and HE. A critical value corrected for multiple pairwise comparisons and equal to 0.013 was used. Statistically significant comparisons are italicized. Locus SC Mn08 Mn90 MnE MnI MnAA MnKK MnK MnY Ta149 Ta105 Ta155 Ta157 Ta162 Western South Atlantic (WSA; n = 39) a 22 14 13 10 15 12 8 17 1 13 9 3 11 aR 21.31 12.89 12.13 9.17 14.05 11.33 7.23 16.48 1.00 12.57 8.17 2.79 10.26 aP 0.39 0.02 0.32 0.10 0.02 0.20 0.67 0.30 0.00 0.00 0.04 0.79 0.00 HO 1.000 0.974 0.897 0.897 0.846 0.923 0.368 0.949 0.000 0.897 0.487 0.128 0.769 HE 0.951 0.880 0.839 0.820 0.853 0.871 0.423 0.932 0.000 0.893 0.539 0.123 0.742 P 0.376 0.460 0.089 0.119 0.201 0.699 0.152 0.797 -- 0.049 0.234 1.000 0.663 Western Central Atlantic (WCA; n = 55) a 24 15 16 10 19 15 7 19 2 19 9 3 11 aR 20.78 13.36 14.13 9.31 16.33 12.39 6.29 16.79 1.81 16.45 8.33 2.91 9.70 aP 1.33 0.06 0.33 0.04 1.06 0.09 0.00 0.09 0.00 0.28 0.06 0.00 0.00 HO 0.909 0.964 0.868 0.945 0.887 0.891 0.600 0.907 0.036 0.889 0.642 0.145 0.709 HE 0.941 0.903 0.878 0.857 0.874 0.878 0.522 0.927 0.036 0.925 0.605 0.139 0.705 P 0.016 0.103 0.909 0.316 0.738 0.524 0.528 0.709 1.000 0.336 0.656 1.000 0.554 United States Mid-Atlantic (USM, n = 263) a 29 20 22 12 23 17 10 27 3 22 10 5 12 aR 19.25 13.08 14.91 8.77 16.04 11.77 6.89 18.51 1.87 15.46 8.04 2.72 8.72 aP 0.63 0.44 0.54 0.21 0.11 0.27 0.33 0.61 0.01 0.36 0.03 0.29 0.22 HO 0.928 0.867 0.824 0.840 0.863 0.897 0.502 0.954 0.046 0.934 0.554 0.152 0.764 HE 0.942 0.888 0.862 0.834 0.864 0.877 0.500 0.939 0.045 0.913 0.567 0.143 0.750 P 0.530 0.415 0.620 0.331 0.058 0.486 0.980 0.952 1.000 0.231 0.830 0.764 0.770 Caribbean Sea (CAR, n = 40) a 20 12 15 8 18 12 6 17 2 18 9 2 10 aR 19.36 11.68 14.27 7.95 16.35 11.49 5.50 16.34 1.78 16.48 8.38 2.00 9.26 aP 0.70 0.01 0.80 0.00 0.81 0.18 0.01 0.00 0.00 1.21 0.04 0.00 0.00 HO 0.897 0.925 0.795 0.900 0.872 0.825 0.375 0.974 0.025 0.795 0.538 0.125 0.725 HE 0.947 0.897 0.858 0.844 0.839 0.865 0.463 0.928 0.025 0.905 0.631 0.119 0.735 P 0.245 0.680 0.427 0.690 0.997 0.701 0.112 0.648 1.000 0.250 0.261 1.000 0.273 Eastern North Atlantic (ENA; n = 33) a 19 13 15 6 17 12 7 18 2 15 7 3 10 aR 18.63 12.75 14.63 6.00 16.68 11.87 6.87 17.69 2.00 14.69 6.93 2.94 9.81 aP 0.94 0.16 1.67 0.00 0.03 0.60 0.00 0.36 0.00 0.13 0.00 0.00 0.00 HO 0.848 0.879 0.818 0.727 0.788 0.727 0.333 1.000 0.125 0.879 0.455 0.121 0.636 HE 0.938 0.874 0.850 0.794 0.864 0.860 0.396 0.937 0.119 0.904 0.501 0.117 0.756 P 0.167 0.297 0.590 0.564 0.265 0.097 0.163 0.115 1.000 0.569 0.124 1.000 0.133 Gulf of Mexico Adults (GOM Adults; n = 49) a 20 16 15 8 20 14 8 22 3 19 9 3 11 aR 18.33 14.79 14.13 7.52 17.67 12.39 6.98 19.83 2.90 16.32 8.10 2.87 9.96 aP 0.20 0.07 0.25 0.36 0.41 0.02 0.01 1.52 0.80 1.22 0.02 0.00 0.00 HO 0.857 0.896 0.886 0.813 0.896 0.918 0.449 0.915 0.156 0.854 0.543 0.163 0.755 HE 0.939 0.905 0.852 0.831 0.872 0.893 0.460 0.942 0.148 0.911 0.477 0.171 0.758 P 0.061 0.102 0.686 0.828 0.960 0.736 0.219 0.260 1.000 0.014 0.948 0.161 0.534 Gulf of Mexico Larvae (GOM Larvae; n = 75) a 24 20 18 14 21 13 9 25 2 17 9 4 10 aR 19.85 15.49 14.91 10.07 16.36 11.87 6.83 19.52 1.41 13.76 8.06 3.20 8.60 aP 1.47 1.19 1.05 1.87 0.04 0.49 0.44 1.25 0.00 0.57 0.01 0.36 0.00 HO 0.917 0.930 0.812 0.803 0.819 0.833 0.486 0.959 0.013 0.907 0.573 0.133 0.732 HE 0.943 0.901 0.842 0.838 0.857 0.891 0.538 0.937 0.013 0.898 0.649 0.151 0.774 P 0.019 0.948 0.685 0.147 0.689 0.048 0.691 0.600 1.000 0.293 0.030 0.372 0.158 Pooled Adults (n = 479) a 31 21 24 13 25 19 10 28 3 25 10 6 12 aR 19.39 13.07 14.59 8.62 16.15 11.99 6.74 18.38 1.98 15.41 8.07 2.72 9.05 aP 3.30 1.56 3.38 1.34 3.29 1.94 1.35 2.68 0.65 3.66 1.10 0.34 1.71 HO 0.916 0.895 0.838 0.851 0.863 0.883 0.475 0.949 0.055 0.902 0.549 0.146 0.745 HE 0.942 0.890 0.861 0.833 0.862 0.878 0.481 0.937 0.054 0.911 0.562 0.139 0.744 P 0.076 0.090 0.718 0.288 0.491 0.636 0.539 0.733 1.000 0.662 0.675 0.501 0.872 Total (n = 554) a 33 22 25 16 25 19 11 30 3 26 10 6 12
69
Table S1. (continued)
Locus Average SC Across Ta24 Tge54 Tge139 Tge151 Isin11 Isin40 Tge121 Tge144 Tge105 Isin1 Isin29 Loci Western South Atlantic (WSA; n = 39) a 14 20 5 4 4 5 13 6 6 21 6 10.50 aR 13.66 18.62 4.99 3.79 3.99 4.79 12.12 5.95 5.79 20.01 5.96 9.96 aP 0.85 0.93 0.00 0.00 0.23 0.00 0.03 0.00 0.00 0.75 0.00 0.24 HO 0.872 0.897 0.538 0.590 0.487 0.718 0.846 0.385 0.615 0.923 0.641 0.694 HE 0.917 0.897 0.621 0.536 0.519 0.626 0.791 0.420 0.682 0.946 0.708 0.689 P 0.209 0.527 0.389 0.275 0.603 0.090 0.931 0.326 0.530 0.207 0.004 Western Central Atlantic (WCA; n = 55) a 17 19 5 6 4 5 14 7 9 19 6 11.67 aR 15.05 16.01 4.91 5.12 3.48 4.81 12.12 6.59 7.13 16.98 5.92 10.28 aP 0.60 0.59 0.00 0.06 0.01 0.00 0.04 0.03 0.83 0.02 0.00 0.23 HO 0.909 0.873 0.691 0.564 0.545 0.673 0.764 0.473 0.764 1.000 0.745 0.725 HE 0.918 0.896 0.632 0.562 0.534 0.617 0.818 0.483 0.662 0.924 0.701 0.706 P 0.141 0.219 0.780 0.801 0.040 0.917 0.242 0.606 0.271 0.783 0.491 United States Mid-Atlantic (USM; n = 263) a 20 23 5 7 6 7 22 9 9 32 7 14.96 aR 14.51 15.75 4.67 4.68 3.49 5.60 13.36 6.67 6.08 20.01 6.06 10.29 aP 0.31 0.12 0.00 0.26 0.32 0.12 1.25 0.01 0.08 0.91 0.00 0.31 HO 0.924 0.897 0.597 0.555 0.498 0.677 0.798 0.506 0.563 0.958 0.693 0.700 HE 0.916 0.885 0.621 0.566 0.515 0.639 0.835 0.516 0.652 0.935 0.686 0.704 P 0.295 0.849 0.866 0.161 0.750 0.582 0.014 0.535 0.176 0.889 0.113 Caribbean Sea (CAR; n = 40) a 16 18 5 5 4 6 13 7 7 21 6 10.71 aR 15.21 16.87 4.77 4.77 3.77 5.75 12.30 6.68 6.28 19.49 5.95 10.11 aP 1.09 0.01 0.00 0.07 0.02 0.01 0.00 0.02 0.00 1.62 0.00 0.27 HO 0.875 0.925 0.650 0.550 0.475 0.590 0.838 0.500 0.750 0.872 0.725 0.688 HE 0.917 0.914 0.597 0.604 0.516 0.642 0.838 0.493 0.642 0.939 0.712 0.703 P 0.914 0.544 0.924 0.201 0.475 0.659 0.228 0.658 0.400 0.044 0.816 Eastern North Atlantic (ENA; n = 33) a 14 19 5 5 3 4 13 8 6 21 7 10.38 aR 13.88 18.57 4.94 4.94 3.00 4.00 13.00 7.87 5.94 20.45 7.00 10.21 aP 0.00 0.11 0.00 0.41 0.00 0.00 0.21 0.40 0.00 0.83 0.77 0.28 HO 1.000 0.879 0.636 0.606 0.667 0.545 0.871 0.469 0.656 0.970 0.839 0.686 HE 0.920 0.922 0.587 0.566 0.550 0.526 0.834 0.490 0.644 0.927 0.751 0.693 P 0.930 0.391 0.286 0.966 0.113 0.803 0.317 0.446 0.567 0.751 0.450 Gulf of Mexico Adults (GOM Adults; n = 49) a 16 16 5 4 5 6 15 7 7 24 6 11.63 aR 14.11 13.88 4.94 3.99 4.40 5.13 13.24 6.45 5.77 20.05 5.52 10.39 aP 0.24 0.04 0.00 0.00 0.26 0.02 0.69 0.01 0.11 1.32 0.00 0.32 HO 0.816 0.898 0.633 0.551 0.521 0.571 0.872 0.490 0.592 0.959 0.625 0.693 HE 0.894 0.883 0.637 0.495 0.543 0.623 0.845 0.489 0.634 0.934 0.548 0.695 P 0.026 0.365 0.569 0.270 0.969 0.513 0.854 0.686 0.168 0.586 0.651 Gulf of Mexico Larvae (GOM Larvae; n = 75) a 16 16 5 6 5 6 15 8 8 26 6 12.63 aR 13.69 13.92 4.80 5.00 3.73 5.60 11.64 6.90 6.71 21.02 5.96 10.37 aP 0.33 0.12 0.00 0.36 0.24 0.01 0.02 0.05 0.38 0.72 0.00 0.46 HO 0.931 0.847 0.587 0.560 0.587 0.653 0.775 0.479 0.667 0.959 0.721 0.695 HE 0.901 0.890 0.603 0.571 0.524 0.653 0.799 0.457 0.670 0.939 0.718 0.707 P 0.302 0.278 0.293 0.657 0.043 0.739 0.618 0.489 0.054 0.539 0.737 Pooled Adults (n = 479) a 21 23 5 8 7 7 22 9 10 35 7 15.88 aR 14.48 15.97 4.73 4.59 3.74 5.31 12.92 6.58 6.07 19.91 6.03 10.27 aP 2.05 4.01 0.15 0.44 1.32 0.31 2.91 0.95 0.76 3.42 0.22 1.78 HO 0.908 0.896 0.614 0.562 0.515 0.653 0.813 0.487 0.616 0.954 0.700 0.699 HE 0.916 0.892 0.619 0.558 0.524 0.625 0.831 0.497 0.652 0.934 0.683 0.701 P 0.486 0.784 0.707 0.453 0.306 0.751 0.056 0.822 0.458 0.636 0.779 Total (n = 554) a 21 23 5 8 7 7 22 9 10 35 7 16.33
70
Table S2. Temporal replicate sample collections comprising ≥10 individuals sampled from the same geographic location in more than one year. Sample collections are labeled as in Figure 1.
No. Year Collected Individuals WCA 2005 29
2006 26
USM 1994 20 1995 11
1996 17 1998 11 2004 16 2006 11 2007 13 2008 42
2009 23
2010 11 2011 10 2012 14 2013 13 2014 22
GOM Adults
2006 31
2008 16 GOM Larvae 2007 17 2008 27 2009 22
71
Table S3. Observed and expected heterozygosities across sample collections, including heterozygosities calculated with a correction for uneven sample sizes. Sample collections are labeled as in Figure 1. n = number of samples, HO = mean observed heterozygosity, HE = mean expected heterozygosity, HEC = mean expected heterozygosity standardized by the smallest sample collection (ENA); HER = ratio of expected heterozygosity standardized by the smallest sample collection (ENA); HER SE = standard error for HER.
GOM GOM WSA WCA USM CAR ENA Adults Larvae n 39 55 263 40 33 49 75
HO 0.694 0.725 0.700 0.688 0.686 0.693 0.695
HE 0.689 0.706 0.704 0.703 0.693 0.695 0.707
HEC 0.697 0.741 0.726 0.748 0.709 0.769 0.771
HER 1.014 1.041 1.044 1.034 1.000 1.031 1.047 H ER 0.107 0.104 0.104 0.106 0.103 0.105 0.106 SE
72
--
2014
0.0026 0.0038
0.0045 0.0029 0.0011 0.0142 0.0027 0.0020 0.0042 0.0015 0.0046 0.0002 0.0100
- -
--
2013
0.813
0.0025
0.0006 0.0060 0.0017 0.0096 0.0007 0.0092 0.0035 0.0000 0.0062 0.0004 0.0112
-
A critical value value Acritical
--
Atlantic temporal temporal Atlantic
2012
-
0.031 0.008
0.0026 0.0120 0.0032 0.0055 0.0055 0.0127 0.0065 0.0090 0.0095 0.0078 0.0173
--
italicized.
2011
0.003 0.398 0.307
0.0029
0.0003 0.0067 0.0001 0.0181 0.0057 0.0058 0.0058 0.0033 0.0033
-
--
2010
0.127 0.067 0.125 0.576
0.0060 0.0025 0.0036 0.0190 0.0028 0.0052 0.0076 0.0056 0.0057
--
2009
0.098 0.203 0.024 0.519 0.069
0.0018
0.0011 0.0025 0.0067 0.0196 0.0007 0.0012 0.0010
-
--
2008
0.300 0.065 0.063 0.008 0.148 0.202
0.0012 0.0050
0.0001 0.0036 0.0100 0.0009 0.0001
- -
--
2007
0.446 0.383 0.087 0.137 0.117 0.080 0.123
0.0064 0.0095 0.0073 0.0225 0.0016 0.0098
--
2006
0.077 0.351 0.426 0.161 0.543 0.017 0.666 0.247
0.0001
0.0041 0.0016 0.0011 0.0196
-
--
2004
0.572 0.448 0.966 0.720 0.304 0.174 0.164 0.515 0.226
0.0031
0.0021 0.0005 0.0021
-
--
1998
0.466 0.036 0.010 0.022 0.003 0.035 0.040 0.252 0.168 0.016
0.0023 0.0132 0.0055
values (above diagonal) based on microsatellite data for United for data States diagonal) mid microsatellite on valuesbased (above
st
F
--
1996
0.220 0.482 0.368 0.061 0.086 0.044 0.172 0.353 0.202 0.358 0.286
0.0007 0.0038
--
1995
0.221 0.118 0.413 0.404 0.077 0.560 0.297 0.278 0.089 0.029 0.194 0.203
0.0014
-
Matrix of pairwise pairwise of Matrix
--
1994
0.607 0.413 0.816 0.215 0.114 0.445 0.368 0.121 0.405 0.276 0.479 0.079
0.2382
1994 1995 1996 1998 2004 2006 2007 2008 2009 2010 2011 2012 2013 2014
corrected for multiple pairwise comparisons and equal to 0.010 was used. 0.010 to equal and comparisons pairwise multiple for corrected replicates. Significance values are shown below diagonal. Statistically significant comparisons are significant comparisons are Statistically shown diagonal. values Significance below are replicates. S4. Table
73
Figure S1. Two-dimensional plot of axes one and two from principal component analysis performed using microsatellite genotype data. Individuals are represented by points colored by sample collection according to legend. Percentage of total genetic variation explained by each axis is shown. Top right: Eigenvalues for principal components. Components shaded in black reflect components one and two displayed in plot.
74
Figure S2. Locus-by-locus Fst values associated with the pairwise comparison of the Gulf of Mexico adult and larval sample collections.
75
CHAPTER III
Population Structure for Striped Marlin (Kajikia audax) in the Pacific and Indian Oceans
Based on Genome-wide Single Nucleotide Polymorphisms
76
ABSTRACT
Striped marlin (Kajikia audax) is a highly migratory pelagic species of significant commercial and recreational importance throughout its range in the Pacific and Indian oceans. Previous studies of genetic population structure for striped marlin have been limited to the analysis of small numbers of molecular markers across sample collections restricted to the Pacific Ocean, and the presence of population structuring for striped marlin in the Indian Ocean remains unexplored. In this study, we assessed nearly 4,000 single nucleotide polymorphisms across sample collections from both the Pacific (n =
199) and Indian oceans (n = 46) to evaluate population structure for striped marlin throughout its range. Our results demonstrate the presence of genetically distinct populations in the western Indian Ocean, Oceania, and eastern central Pacific Ocean, as well as two populations in the North Pacific Ocean (FST = 0.0169–0.0836). Comparisons of replicate collections for some regions demonstrate stability of allele frequencies across multiple generations. Collectively, our results provide novel insights into population structuring for striped marlin, and highlight inconsistencies between biologically distinct units and units currently recognized for fisheries management.
77
INTRODUCTION
Highly migratory pelagic fishes are of significant commercial and recreational value worldwide, yet for many of these species population structure is poorly understood.
Historically, it has been assumed that a continuous marine environment which lacks obvious physical barriers to dispersal would not be conducive to the development of population structure in highly migratory pelagic fishes. However, a number of genetic studies demonstrate the presence of population structuring for several of these species
(e.g. Reeb et al. 2000; Carlsson et al. 2007; Davies et al. 2011; Grewe et al. 2015;
Williams et al. 2015). Most of these studies were based on the assessment of small numbers of molecular markers across sample collections which represent only a portion of the species range, and report low but statistically significant levels of genetic differentiation among populations. In addition, few studies have evaluated the persistence of such shallow structuring over multiple generations, making it difficult to distinguish biologically meaningful levels of differentiation from stochastic noise. Improved resolution of previously identified populations of highly migratory pelagic fishes, as well as the detection of additional population structuring, may be possible with genetic studies based on genome-wide molecular markers and sample collections representative of the full species range. Assessments that also include temporally spaced sample collections which span multiple generations enable evaluation of the temporal stability of observed structure.
The striped marlin (Kajikia audax) is an istiophorid billfish (marlins, spearfishes, sailfish) distributed throughout temperate, sub-tropical, and tropical waters of the Pacific and Indian oceans (Nakamura 1985). This highly migratory pelagic species is targeted in
78
recreational fisheries throughout its range, many of which are primarily catch-and-release and contribute limited fishing mortality (Bromhead et al. 2003; Peel et al. 2003). Striped marlin is also targeted in a number of commercial and small-scale artisanal and subsistence fisheries; however, reporting from these fisheries is often limited by a lack of species-level information or, in many cases, lacking altogether (Bromhead et al. 2003).
The primary source of fishing mortality for striped marlin is attributed to bycatch in commercial pelagic longline fisheries targeting tunas and swordfish. Fisheries which interact with striped marlin are managed by three regional fisheries management organizations (RFMOs) with jurisdictions spanning continuous portions of the species range. In the eastern Pacific Ocean, the Inter-American Tropical Tuna Commission
(IATTC) recognizes a single stock of striped marlin for assessment and management purposes. The Western and Central Pacific Fisheries Commission (WCPFC) recognizes two distinct stocks of striped marlin corresponding with the western and central North
Pacific Ocean and the western South Pacific Ocean. Finally, the Indian Ocean Tuna
Commission (IOTC) recognizes a single ocean-wide stock of striped marlin in the Indian
Ocean. Recent stock assessments indicate that the eastern Pacific Ocean stock of striped marlin is neither overfished nor experiencing overfishing (IATTC 2017); however, stocks in the Indian Ocean and the western and central North Pacific Ocean are considered overfished and experiencing overfishing (WCPFC 2015; IOTC 2017). The stock in the western South Pacific is estimated to be approaching an overfished state (WCPFC 2012).
The management units currently recognized for striped marlin are primarily based on pragmatic RFMO boundaries that do not necessarily correspond with genetically- determined populations of this species. McDowell and Graves (2008) surveyed genetic
79
variation at five nuclear microsatellite markers and the mitochondrial (mt) DNA control region across collections of striped marlin from geographically distant regions of the
Pacific Ocean. That study resolved four genetically distinct populations: 1) a population spanning the North Pacific and comprising striped marlin sampled off Japan, Taiwan,
Hawaii (United States), and southern California (United States), 2) a population in the eastern North Pacific corresponding with striped marlin sampled off Baja California
(Mexico), 3) a population in the eastern central Pacific comprising striped marlin sampled off Ecuador, and 4) a population in the western South Pacific corresponding with striped marlin sampled off eastern Australia. These populations were also resolved in a subsequent genetic study based on the analysis of larger numbers of microsatellites
(n = 12) and samples (Purcell and Edmands 2011); however, a number of discrepancies between studies were observed. Purcell and Edmands (2011) reported the presence of a genetically distinct population corresponding with mature fish sampled off Hawaii. In addition, a lack of sampling effort for striped marlin off Ecuador or surrounding waters prohibited verification of a population in this region. In both studies, levels of genetic differentiation observed between striped marlin populations were low (FST ≤ 0.0377) relative to those reported for populations of terrestrial or freshwater species (e.g. Ward et al. 1994), but high compared with studies of other pelagic fishes (Carlsson et al. 2004,
2007). These results suggest that a more powerful assessment of population structure based on the analysis of larger numbers of molecular markers and samples may result in the detection additional population structuring for striped marlin.
Population structure for striped marlin in the Indian Ocean has not been investigated, and the relationship between striped marlin from the Pacific and Indian
80
oceans is also unknown. This lack of information is problematic given that striped marlin is heavily overfished (biomass in 2015 estimated to be 0.24–0.62 of that required for maximum sustainable yield) and likely experiencing heavy overfishing (fishing effort in
2015 estimated to be 1.32–3.04 of that required for maximum sustainable yield) in the
Indian Ocean (IOTC 2017). Uncertainty regarding stock structure for striped marlin in the Indian Ocean has made it impossible to determine if the single ocean-wide stock currently recognized by the IOTC reflects the biological organization of this species in this region. An understanding of population structure for striped marlin in the Indian
Ocean is necessary for the development of management measures that promote the recovery of this stock to sustainable levels, and to conserve genetic diversity important for short- and long-term population persistence (Gaggiotti and Vetter 1999; Allendorf et al. 2008; Reiss et al. 2009; Allendorf et al. 2014; Pinksy and Palumbi 2014; Ovenden et al. 2015; Spies et al. 2015).
In this study, we use next-generation sequencing (NGS; e.g. Mardis 2008) to discover genome-wide single nucleotide polymorphism (SNP) molecular markers for evaluating the genetic population structure of striped marlin throughout its range. The primary objectives of this study include: 1) determine the number and geographic extent of striped marlin populations in both the Pacific and Indian oceans, 2) evaluate the temporal stability of observed population structure, and 3) assess the degree of genetic connectivity among populations of striped marlin.
81
METHODS AND MATERIALS
Sample collection and DNA isolation
Samples analyzed in this study consisted of fin tissue from striped marlin released alive following capture by recreational anglers or from striped marlin caught as bycatch on commercial pelagic longline vessels. Additional samples consisting of muscle tissue were obtained from striped marlin available in local markets. Samples were opportunistically collected during the period 1992 through 2017 from locations throughout the species range, including waters off South Africa (SAF), Kenya (KEN), and northwestern
Australia (WAUS) in the Indian Ocean, and eastern Australia (EAUS), New Zealand
(NZ), Japan (JAP), Taiwan (TAI), Hawaii (HAW), southern California (CAL), Baja
California (BAJA), Ecuador (ECU), and Peru (PERU) in the Pacific Ocean (Table 1; Fig.
1). All samples were preserved in 95% ethanol or a 10% dimethyl sulfoxide solution
(Seutin et al. 1991) and maintained at room temperature until DNA isolation. Total genomic DNA was isolated from tissues using a DNeasy Blood & Tissue Kit (Qiagen) or a ZR-96 Genomic DNA Kit (Zymo Research). DNA from each sample was visualized on
5% agarose gels that included a lane containing 1Kb Plus DNA Ladder (Thermo Fisher
Scientific). DNA isolations that recovered high molecular weight DNA were quantified using a Qubit 2 fluorometer and dsDNA BR assays (Thermo Fisher Scientific). Isolations with sufficient DNA for NGS analysis were normalized to 700 ng total DNA at 50 ng per uL and stabilized in GenTegra-DNA (GenTegra LLC). Stabilized high quality DNA isolations were submitted to Diversity Arrays Technology Pty. Ltd. (DArT PL; Canberra,
Australia) for DArTseqTM 1.0 genotyping.
82
DArTseqTM 1.0 library preparation and sequencing
DArTseqTM genotyping (e.g. Sansaloni et al. 2011) involves genomic complexity reduction followed by NGS, and is similar to other commonly utilized approaches for
NGS of reduced genomic representations (e.g. Peterson et al. 2012). Genomic complexity reduction was principally performed as described in Kilian et al. (2012), but with a double restriction enzyme (RE) digestion and ligation with RE-specific adapters. Four RE combinations were tested at the DArT PL facility (data not shown) and digestion with
PstI and SphI was selected based on the size of the representation and the fraction of the genome selected. Custom proprietary adapters used in ligation reactions were similar to those described by Elshire et al. (2011) and Kilian et al. (2012). A PstI-compatible forward adapter included an Illumina flowcell attachment sequence, a sequencing primer sequence, and a variable length barcode. A SphI-compatible reverse adapter included an
Illumina flowcell attachment region.
Following double RE digestion and adapter ligation, fragments with PstI-SphI overhangs were preferentially amplified in PCR reactions using the following conditions: initial denaturation at 94 ˚C for 1 min, 30 cycles of 94 ˚C for 20 sec, 58 ˚C for 30 sec, and
72 ˚C for 45 sec, and a final extension at 72 ˚C for 7 min. PCR amplification products were subsequently cleaned using a GenElute PCR Clean-Up Kit (Sigma-Aldrich), visualized on 0.8% agarose gels, and quantified using ImageJ software (Abràmoff et al.
2004). Samples for which RE digestion appeared to be incomplete and/or PCR amplification was unsuccessful were excluded from further library preparation. Samples were normalized and pooled at equimolar ratios into multiplex libraries each comprising
94 samples and two controls, and sequenced on single lanes of an Illumina HiSeq 2500
83
platform (Illumina, Inc.) at the DArT PL facility. Cluster generation and amplification were performed using a HiSeq SR Cluster Kit V4 cBot (Illumina, Inc.), and was followed by 77 bp single-end sequencing.
DArTseqTM genotype calling
Raw Illumina reads were processed in CASAVA v1.8.2 (Illumina, Inc.) for initial assessment of read quality and sequence representation, and to produce FASTQ output files. Resulting FASTQ files were analyzed in the proprietary DArTseq analytical software pipeline DArTtoolbox, wherein quality filtering, variant calling, and generation of final genotypes were performed in sequential primary and secondary workflows (Fig.
S1). In the primary workflow, reads with Q < 25 for at least 50% of bases were removed, followed by the removal of reads with Q < 30 in the barcode region; this latter filtering step was performed to ensure accurate assignment of reads to individual samples. Reads were de-multiplexed according to sample-specific barcodes, then aligned and queried against catalogued sequences in the National Center for Biotechnology Information
(NCBI) GenBank and proprietary DArTdb databases to identify and remove reads associated with viral or bacterial contamination.
In the secondary workflow, a catalog of reduced representation loci (RRL) was created de novo by first aligning identical reads within and among sequenced individuals to form read clusters. Read clusters were catalogued in DArTdb then matched against each other based on degree of similarity and size to form RRL. Polymorphic positions within RRLs were distinguished as SNP variants, and major and alternate alleles for each variant were identified. A matrix of SNP genotypes based on the following DArT scores
84
was then generated: “0” = major allele homozygote, “1” = alternate allele homozygote,
“2” = heterozygote. Robust variant calling was ensured by removing SNP loci that met any of the following conditions: monomorphic clusters, clusters containing tri-allelic or aberrant SNPs, clusters with overrepresented sequences, and/or loci lacking both allelic states (homozygote and heterozygote). A proportion of loci were produced a second time to assess technical replication error. Each remaining SNP locus was then characterized by calculating major and alternate allele frequency, heterozygote and homozygote frequency, polymorphism information content (PIC), call rate, and average reproducibility. DArT PL supplied a final genotype matrix containing 61,908 SNP loci and metadata associated with each locus.
SNP filtering
Additional quality filtering of SNP data received from DArT PL was performed in R version 3.3.1 (R Core Team 2017) using the dartR v0.93 package (Gruber et al. 2018;
Fig. S1). Loci missing ≥ 10% of genotype calls followed by individuals missing ≥ 20% of genotype calls were excluded from the dataset. To ensure high quality loci with reliable genotype calls, loci with average reproducibility < 95% were removed. All monomorphic loci were also removed. In instances where more than one SNP originated from a RRL, a single SNP was retained by selecting the locus associated with the highest reproducibility and PIC, in that order. This step was performed to reduce the probability of linked loci in the final dataset. Sequencing and/or PCR error as well as ascertainment bias resulting from non-random sampling of a gene pool have been shown to bias estimates of genetic connectivity and genetic assignment of individuals to source populations (Bradbury et al.
85
2011; Roesti et al. 2012). To address these issues, we removed loci with a minor allele frequency < 0.05 across all samples. Finally, loci that did not conform to the expectations of Hardy-Weinberg equilibrium (HWE) were identified and excluded from subsequent analyses using the exact methodological approach described by Wigginton et al. (2005).
HWE was evaluated within sample collections organized by sampling location. Statistical significance of HWE comparisons was assessed using a critical value corrected by a modified false discovery rate based on the formulation originally described by Benjamini and Yekutieli (2001) and tested by Narum (2006). Loci that did not conform to the expectations of HWE in more than one sample collection were removed.
Detection of outlier loci
In order to reduce bias from non-neutral processes in estimates of genetic connectivity
(Beaumont and Nichols 1996; Luikart et al. 2003), SNPs potentially under the influence of natural selection were removed prior to subsequent analyses. To reduce the probability of committing type I or type II statistical errors, we employed two approaches for the identification of outlier loci that may be under the influence of natural selection (Narum and Hess 2011; Lotterhos and Whitlock 2014). The program BayeScan v2.1 (Foll and
Gaggiotti 2008) implements a Bayesian-based methodology that compares allele frequencies among populations to directly estimate the probability that each locus is exposed to natural selection (Beaumont and Balding 2004; Foll and Gaggiotti 2008). We performed BayeScan analyses using default settings, except conservative prior odds for the neutral model (10,000:1) were used to reduce the probability of false positives.
Outlier loci were identified from BayeScan output using a false discovery rate of 0.05.
86
We also employed the outlier detection method of Excoffier et al. (2009) implemented in the program Arlequin v3.5 (Excoffier and Lischer 2010). This method assumes a finite island model of migration to obtain a distribution of FST values across loci as a function of average heterozygosity within populations. Arlequin outlier detection analyses were performed with 500,000 simulations, and a p-value of 0.05 was used to identify outliers from simulation output. A final list of outlier loci was generated by including only those loci identified as outliers in both the BayeScan and Arlequin analyses; these loci were excluded from a final neutral dataset that was used for all subsequent analyses.
Temporal replicates
We first assessed the temporal stability of allelic frequencies for sampling locations with replicate collections spanning multiple generations of striped marlin. We also used this information to determine whether temporal replicates could be combined into single collections for subsequent analyses. Temporal replicates were defined as sample collections obtained from a similar geographic region in more than one year. We evaluated only those temporal replicates with sample sizes ≥ 15 individuals per replicate, and for which replicates spanned at least one generation (average generation time estimated at 4.4 years for striped marlin; Collette et al. 2011). Temporal stability was assessed for samples collected off Ecuador in the years 1992 (ECU 1992, n = 15; Table
1) and 2016 (ECU 2016, n = 22). We also evaluated temporal stability for samples collected off eastern Australia in the year 1994 (EAUS 1994, n = 16) and in the years
2010, 2011, 2012, and 2015 (EAUS 2010–2015, n = 19), the latter of which were pooled to facilitate the valuable comparison of replicates spanning 16 to 21 years.
87
To assess temporal stability of allelic frequencies for the ECU and EAUS replicate collections, we performed individual-based cluster analyses, hierarchical analysis of covariance components (i.e. analysis of molecular variance, AMOVA;
Excoffier et al. 1992; Weir 1996; Rousset 2000), and calculated pairwise levels of genetic differentiation. Individual-based cluster analyses consisted of principal coordinate analysis (PCoA) and discriminant analysis of principal components (DAPC; Jombart et al. 2009, 2010). We used dartR to perform PCoA based on a Euclidean distance matrix
(Gower 1966) generated from SNP genotype data for ECU or EAUS temporal replicates.
DAPC performed using the R package adegenet v2.0.1 (Jombart 2008) was used to assess the distinctiveness of clusters corresponding with temporal replicates for ECU or EAUS.
We used Arlequin to perform AMOVA and calculate associated ΦST values (Weir and
Cockerham 1984) for scenarios with temporal replicates for ECU or for EAUS grouped separately or together. Statistical significance of AMOVA results was determined using
10,000 permutations of the data. Finally, genetic differentiation between temporal replicates was assessed within ECU or EAUS by calculating pairwise measures of FST in
Arlequin. Statistical significance of FST values was determined based on 10,000 permutations of the data.
Evaluation of diversity and differentiation by sampling location
To genetically characterize collections of striped marlin sampled from geographically distant regions, we calculated diversity metrics and pairwise levels of genetic differentiation for samples organized by sampling location. Observed and expected heterozygosities were calculated for each sample collection using the R packages poppR
88
v2.5.0 (Kamvar et al. 2014) and dartR, respectively. We used the R package
PopGenReport v3.0.0 (Adamack and Gruber 2014) to perform rarefaction allelic richness analyses based on the smallest number of alleles observed for a sample collection.
Genetic differentiation among sample collections was determined by calculating pairwise
FST values as described above for temporal replicates. Lastly, we evaluated sample collections for the presence of SNPs exhibiting fixed differences (i.e. private alleles) using dartR.
Organization of sample collections into populations
To determine the number and geographic extent of striped marlin populations represented by the sample collections evaluated in this study, we employed a variety of individual- based clustering approaches consisting of Bayesian-based simulations and multivariate analyses. We also used AMOVA to evaluate a range of population structuring scenarios.
The clustering of genetically similar individuals was evaluated using the Bayesian simulation algorithm implemented in the program STRUCTURE v2.3.4 (Pritchard et al.
2000; Falush et al. 2003; Hubisz et al. 2009). All STRUCTURE simulations were performed using an admixture model of ancestry (Falush et al. 2003), a burn-in of 50,000 followed by 500,000 Markov Chain Monte Carlo simulations, and five iterations of each
K. Previous evaluations of STRUCTURE performance demonstrate that the presence of strongly differentiated genetic clusters may obfuscate resolution of weakly differentiated clusters (Vähä et al. 2006; Waples and Gaggiotti 2006; Janes et al. 2017). In preliminary analyses of our dataset, the highest levels of genetic differentiation were observed between sample collections from the Pacific Ocean and western Indian Ocean. Thus, to
89
improve the resolution of more weakly differentiated clusters, we performed
STRUCTURE analyses on three datasets: 1) all sample collections, 2) all Pacific Ocean sample collections and the western Australia collection, 3) all Indian Ocean sample collections and the eastern Australia and New Zealand collections. Scenarios with K equal to two through eight were evaluated for each dataset. Results from each K scenario for each dataset were summarized in CLUMPP v1.1.2 (Jakobsson and Rosenberg 2007), and barplots displaying individual admixture proportions were visualized using
DISTRUCT v1.1 (Rosenberg 2004). The most likely K for each dataset was identified using the method described by Evanno et al. (2005) and implemented in the program
Structure Harvester v0.6.94 (Earl and vonHoldt 2012). We also performed multivariate analyses including PCoA and DAPC as described above for temporal replicates, except we evaluated a range of values for K to represent the number of clusters described by
DAPC. The most likely K for our dataset was determined by generating a Bayesian information criterion (BIC) score for each K scenario, and selecting those scenarios with the lowest BIC scores to assess with DAPC. Finally, we assessed various population structuring scenarios for striped marlin using AMOVA performed as above for temporal replicates. Results from individual-based cluster analyses were used to inform scenarios evaluated in AMOVA analyses.
Results from individual-based cluster analyses and AMOVA were compared to each other and to available biological information to determine the most likely number of striped marlin populations represented by our data. Based on this information, sample collections were combined into groups comprising genetically similar individuals representative of populations. Populations were then characterized in subsequent analyses
90
to calculate population-level genetic diversity, evaluate genetic differentiation between populations, and estimate genetic connectivity among populations.
Evaluation of diversity and differentiation by population
We calculated metrics of genetic diversity for each population, including observed and expected heterozygosities and rarefaction allelic richness, and surveyed populations for the presence of SNPs exhibiting fixed allelic differences between populations. Finally, we calculated pairwise FST values to assess the level of genetic differentiation between populations. Estimates of genetic diversity, fixed differences, and pairwise FST were performed as above for individual sample collections. The minimum number of alleles used to calculate rarefaction allelic richness was based on the smallest number of alleles observed for a population.
Population connectivity
To assess the level of genetic connectivity among populations resolved using individual- based cluster analyses and AMOVA, we estimated mutation-scaled effective population sizes (횹) and mutation-scaled effective migration rates (M) between populations in the program MIGRATE-N v3.6.11 (Beerli and Palczewski 2010). Briefly, MIGRATE-N uses
Bayesian inference to generate posterior probability distributions for parameters of interest based on coalescence theory. Within a simulation, MIGRATE-N estimates the genealogical history of each molecular marker comprising a dataset, therefore limiting the number of markers that can be included for analysis. To account for this limitation, we performed MIGRATE-N simulations using a reduced dataset comprising 700 SNPs
91
randomly selected from the final neutral dataset. To eliminate bias from the exclusion of invariable sites, all MIGRATE-N analyses were performed using the full RRL sequence for each SNP. We evaluated a single scenario where bidirectional gene flow was possible among all of the populations resolved using individual-based cluster analyses and
AMOVA. All MIGRATE-N simulations were performed using a burn-in of 40,000 followed by 400,000 Markov Chain Monte Carlo steps. Six independent simulations of our model were completed to assess the consistency of parameter estimates among model runs.
RESULTS
SNP filtering
The original DArT PL dataset consisted of 61,908 SNP loci (Table 2). A total of 4,016
SNPs remained after quality filtering of loci based on percent missing genotype calls, average reproducibility, monomorphic RRL reads, the presence of > 1 SNP per RRL, and minor allele frequency < 0.05 across all samples. Four individuals were missing genotype calls at ≥ 20% of loci and were excluded from the dataset. Results from the testing of quality filtered SNPs for conformance to the expectations of HWE indicated that 41 loci violated these expectations in more than one sample collection; these loci were removed from the dataset. Collectively, these filtering steps resulted in a dataset comprising 245 individuals (Table 1) genotyped across 3,975 SNP loci (Table 2).
92
Detection of outlier loci
From the 3,975 SNPs remaining after quality filtering and HWE testing, a genome scan performed using BayeScan identified 59 loci (1.48%) as outliers potentially under the influence of natural selection. FST values associated with these outlier loci ranged from
0.091 to 0.552, and all outliers were candidates for divergent selection (alpha = 0.728–
3.657). Results from outlier detection analyses performed using Arlequin included the identification of 341 loci (8.6%) as putative outliers. Of those outliers, 159 were candidates for exposure to balancing selection (per locus FST no different from 0.000) and
182 were candidates for directional selection (per locus FST = 0.081–0.680). All 59 of the loci identified as outliers by BayeScan were identified as outliers likely under the influence of directional selection by Arlequin; these loci comprised a list of candidate
SNPs under selection and were excluded from a final dataset composed of the remaining
3,916 putatively neutral loci (Table 2). This final neutral dataset was used for all subsequent analyses.
Temporal replicates
We assessed the temporal stability of allelic frequencies for replicate collections obtained off Ecuador (ECU 1992, 2016) and off eastern Australia (EAUS 1994, 2010–2015).
Results from multivariate analyses did not provide any evidence suggesting biologically meaningful differences in allele frequencies between replicate collections for either location. PCoA performed using datasets comprising temporal replicates for either ECU or EAUS resulted in single clusters of individuals for each dataset (Figs. 2A, 2B). DAPC performed for scenarios with K equal to two for either the ECU or EAUS temporal
93
replicates produced two distinct clusters of individuals for both datasets (Figs. 2C, 2D); however, these clusters did not correspond with replicate collections for either location and likely represent noise rather than biologically distinct groups of individuals. Results from AMOVA scenarios with ECU or EAUS temporal replicates grouped together included among sample ΦST values that were an order of magnitude lower than among region ΦST values associated with population-level comparisons (Table 3). FST values between temporal replicates were low but statistically significant for both locations
(ECU: FST = 0.0030, p = 0.002; EAUS: FST = 0.0041, p = 0.001). These levels of genetic differentiation are < 25% of those observed for population-level comparisons (see below), and in some instances are a full order of magnitude lower. Collectively, these results are consistent with the temporal stability of allele frequencies for striped marlin off Ecuador and off eastern Australia for time periods spanning multiple generations.
Replicate collections for ECU and EAUS were combined into single collections for all subsequent analyses.
Evaluation of diversity and differentiation by sampling location
To genetically characterize sample collections evaluated in this study, we calculated metrics of genetic diversity for each collection, evaluated collections for the presence of fixed differences, and calculated pairwise levels of genetic differentiation. Rarefaction allelic richness was calculated using the smallest number of alleles observed for a sample collection (n = 4), and was lowest for sample collections from the Indian Ocean, including SAF, KEN, WAUS (aR = 1.259–1.269; Table 4). These sample collections also displayed the lowest expected heterozygosities (HE = 0.145–0.148). The highest levels of
94
genetic diversity were associated with striped marlin sampled off Hawaii (aR = 1.318, HE
= 0.177). There were no fixed differences observed among sample collections. Pairwise levels of genetic differentiation ranged from FST = -0.0003 between JAP and HAW to FST
= 0.0653 between KEN and PERU (Table 5). The highest levels of genetic differentiation corresponded with comparisons between sample collections from the western Indian
Ocean (SAF, KEN) and collections from the North Pacific and eastern central Pacific oceans (JAP, TAI, HAW, CAL, BAJA, ECU, PERU; FST = 0.0410–0.0653). With the exception of eight instances (SAF vs. KEN, EAUS vs. NZ, JAP vs. TAI, JAP vs. HAW,
TAI vs. CAL, BAJA vs. ECU, BAJA vs. PERU, ECU vs. PERU), all pairwise levels of genetic differentiation between sample collections were highly statistically significant (p
= 0.000–0.009, pcrit = 0.010). Non-statistically significant FST values all corresponded with comparisons between sample collections comprising the same population (see population-level results below); however, five statistically significant comparisons between sample collections representative of the same population were observed (WAUS vs. EAUS, WAUS vs. NZ, JAP vs. CAL, TAI vs. HAW, HAW vs. CAL).
Organization of sample collections into populations
We employed multiple approaches, including Bayesian-based simulation and multivariate analyses, for the delineation of striped marlin populations from our final SNP dataset.
Individual admixture proportions inferred from STRUCTURE analyses performed using the dataset inclusive of all sample collections are shown in Fig. 3. A number of clusters were consistently resolved across the K scenarios evaluated with this dataset. These included a cluster comprising striped marlin from the western Indian Ocean (SAF, KEN);
95
this cluster was characterized by a low degree of shared ancestry with striped marlin from the Pacific Ocean. A second cluster comprising individuals from Oceania (WAUS,
EAUS, NZ) was also consistently identified. This cluster corresponded with high degrees of shared ancestry with striped marlin from elsewhere in both the Pacific and Indian oceans. A third cluster comprising striped marlin sampled from locations in the eastern central Pacific and North Pacific oceans was also resolved across all K scenarios. This cluster was characterized by a low degree of shared ancestry with Indian Ocean striped marlin. In scenarios with K equal to three through five (Fig. 3B–D), the cluster comprising sample collections from the eastern central Pacific and North Pacific oceans was subdivided into two clusters corresponding with these geographic regions. Results from Structure Harvester indicated that the most likely K for this dataset was five.
Examination of admixture proportions estimated for individuals in the K equal five scenario revealed that in addition to distinct clusters corresponding with sample collections from the western Indian Ocean, Oceania, North Pacific Ocean, and eastern central Pacific Ocean, a fifth cluster corresponded with a subset of striped marlin sampled from the North Pacific Ocean off Japan (n = 5; hereafter referred to as JAP2) and Hawaii
(n = 6; hereafter referred to as HAW2). Across K scenarios, admixture proportions for a small number of striped marlin sampled off Hawaii (n = 4) and Ecuador (n = 3) were consistent with striped marlin from Oceania.
Results from STRUCTURE analyses performed using the dataset limited to Pacific
Ocean and WAUS sample collections also reflected the presence of distinct genetic clusters corresponding with Oceania, the eastern central Pacific Ocean, and two genetically distinct clusters in the North Pacific Ocean (Fig. S2). For the Oceania cluster,
96
individual admixture proportions were similar across the WAUS, EAUS, and NZ sample collections for all K scenarios. Structure Harvester identified the most likely K for this dataset as four. Results from STRUCTURE analyses performed using the dataset limited to
Indian Ocean, EAUS, and NZ sample collections resolved at least two distinct clusters across K scenarios (Fig. S3). One of these clusters corresponded with sample collections from the western Indian Ocean (SAF, KEN), and a second cluster comprised EAUS+NZ.
Admixture proportions for the WAUS sample collection were intermediate to SAF+KEN and EAUS+NZ, and Structure Harvester identified the most likely K for this dataset as three.
We also employed multivariate analyses for the delineation of striped marlin populations. Results from PCoA of SNP genotype data are shown in Fig. 4. PCoA axes one and two collectively explained 5.1% of total genetic variation and resolved multiple distinct clusters of individuals. One of these clusters corresponded with sample collections from the eastern central Pacific Ocean (BAJA, ECU, PERU). A second cluster was composed of sample collections spanning the North Pacific Ocean (JAP, TAI,
HAW, CAL). Striped marlin associated with the JAP2 (n = 5) and HAW2 (n = 6) sample collections comprised a third cluster distinct from other North Pacific Ocean sample collections; this cluster was also distinct on PCoA axes one and three (Fig. S4), and on axes two and three (Fig. S5). Striped marlin sampled off eastern Australia (EAUS) and
New Zealand (NZ) clustered separately from all other Pacific Ocean sample collections, and were positioned adjacent to sample collections from the Indian Ocean. The sample collection from western Australia (WAUS) was placed intermediate to the cluster comprising EAUS+NZ and a cluster comprising remaining Indian Ocean collections
97
(KEN, SAF). In the PCoA results described here, eight striped marlin grouped with sample collections geographically distant from the sampling locations of these individuals. Four fish sampled off Hawaii and three fish sampled off Ecuador grouped with EAUS+NZ; in results from STRUCTURE analyses, admixture proportions for these fish were similar to striped marlin sampled from Oceania. Similarly, one fish sampled off
California grouped with the cluster comprising sample collections from the eastern central Pacific Ocean (BAJA, ECU, PERU). These eight individuals are hereafter referred to as putative migrants, and were retained with their original sample collection so that realistic assemblages of striped marlin could be characterized in subsequent analyses.
We employed DAPC to evaluate the distinctiveness of clusters comprising groups of genetically similar individuals. BIC scores were generated for each of eleven clustering scenarios based on sequentially increasing values of K ranging from two through twelve. The lowest observed BIC score corresponded with the scenario for K equal to two (BIC = 1228.53), but similar BIC scores were produced for scenarios with K equal to three through five (BIC = 1230.09, 1231.74, and 1234.36, sequentially). In the scenario with K equal to two, one cluster comprised all Pacific Ocean sample collections except EAUS and NZ, and a second cluster comprised EAUS, NZ, and all Indian Ocean sample collections. The scenario with K equal to three also resolved a cluster comprising
EAUS, NZ, and Indian Ocean sample collections, as well as clusters corresponding with fish sampled from the North Pacific Ocean (JAP, JAP2, TAI, HAW, HAW2, CAL) or from the eastern central Pacific Ocean (BAJA, ECU, PERU). Clusters comprising North
Pacific Ocean and eastern central Pacific Ocean sample collections were also resolved in the scenario with K equal to four (Fig. 5), but sample collections from Oceania (WAUS,
98
EAUS, NZ) were assigned to a cluster separate from western Indian Ocean sample collections (KEN, SAF). These four clusters were also resolved in the scenario with K equal to five, except fish comprising JAP2 and HAW2 assigned to a fifth cluster (Fig. 6).
For each of the clustering scenarios described here, DAPC analysis resulted in clusters that were clearly resolved (i.e. non-overlapping) in two-dimensional plots, except for the scenario with K equal to five (Fig. 6). In the plot for that scenario, the cluster comprising
JAP2 and HAW2 overlapped considerably with the cluster comprising remaining collections from the North Pacific Ocean. Across all K scenarios, the mean posterior probability of assignment to the cluster to which an individual assigned was 99.72
(minimum = 0.56, maximum = 1.00, standard deviation = 0.03). The four putative migrants sampled off Hawaii and three putative migrants sampled off Ecuador consistently assigned to the same cluster as WAUS+EAUS+NZ. The single putative migrant sampled off California consistently assigned to the same cluster as sample collections from the eastern central Pacific Ocean (BAJA, ECU, PERU). In DAPC scenarios with K equal to two and five, an additional individual from the Hawaii sample collection grouped with WAUS+EAUS+NZ.
Finally, we used AMOVA to evaluate various population structuring scenarios for striped marlin. To limit the range of possible scenarios to test with AMOVA, we combined individual sample collections into larger regional groups reflecting those groups consistently resolved in individual-based cluster analyses. Those groups corresponded with the western Indian Ocean (SAF, KEN), eastern central Pacific Ocean
(BAJA, ECU, PERU), and the larger grouping of individuals from the North Pacific
Ocean (JAP, TAI, HAW, CAL). Results from AMOVA (Table 3) indicated that
99
differences among regions based on percent variation and ΦST were maximized (percent variation = 3.36, ΦST = 0.0336) in the scenario with sample collections grouped as follows: western Indian Ocean, Oceania (WAUS, EAUS, NZ), eastern central Pacific
Ocean, North Pacific Ocean, and a second population in the North Pacific Ocean comprising JAP2 and HAW2. However, nearly identical results were obtained for the scenario where the same groups were recognized, except Oceania was subdivided into
WAUS and EAUS+NZ (percent variation = 3.35, ΦST = 0.0335). This latter scenario also corresponded with the lowest observed levels of variation among populations within regions (percent variation = 0.22, ΦST = 0.0023).
Results from individual-based cluster analyses and AMOVA were used to inform the grouping of sample collections into larger regional groups representative of genetically distinct populations. Genetic clusters corresponding with the western Indian
Ocean (SAF, KEN) and eastern central Pacific Ocean (BAJA, ECU, PERU) were consistently resolved across cluster analyses and were therefore recognized as distinct populations in subsequent analyses. Cluster analyses also consistently resolved sample collections from the North Pacific Ocean (JAP, TAI, HAW, CAL) as distinct, as well as a second group in the North Pacific Ocean corresponding with a subset of striped marlin sampled off Japan (JAP2) and Hawaii (HAW2). Finally, a group comprising sample collections from Oceania (WAUS, EAUS, NZ) was consistently resolved in cluster analyses; however, some results suggest the possibility that WAUS comprises a separate cluster. FST values associated with the pairwise comparison of WAUS and EAUS, as well as WAUS and NZ, were low but statistically significant (Table 5). Results from PCoA included the placement of WAUS as adjacent to, rather than overlapping with,
100
EAUS+NZ (Fig. 4), but assignment of individuals to genetically distinct clusters prior to
DAPC consistently placed WAUS in the same cluster as EAUS+NZ. Results from
STRUCTURE analyses performed using the dataset limited to Indian Ocean, EAUS, and
NZ sample collections included admixture proportions for WAUS that were intermediate to those characterizing EAUS+NZ and SAF+KEN (Fig. S3), and Structure Harvester identified the most likely K for that dataset as three. To account for uncertainty regarding the relationship of WAUS and EAUS+NZ, metrics of genetic diversity and differentiation were calculated twice: once with WAUS grouped with EAUS+NZ (five total populations), and a second time with WAUS grouped separately (six total populations).
Evaluation of diversity and differentiation by population
Genetic diversity metrics, the presence of fixed differences, and pairwise levels of genetic differentiation were assessed for striped marlin populations identified in previous analyses. Rarefaction allelic richness was calculated using the smallest number of alleles observed for a population (n = 16), and was highest for the population in the North
Pacific Ocean comprising JAP2 and HAW2 (aR = 1.501; Table 6); expected heterozygosity was also highest for this population (HE = 0.204). The lowest observed value for rarefaction allelic richness corresponded with the western Indian Ocean population (aR = 1.463); this population also displayed the lowest expected heterozygosity (HE = 0.147). For diversity calculations performed with WAUS grouped separately from EAUS+NZ (Table S1; minimum number of alleles for rarefaction allelic richness = 6), the North Pacific Ocean population corresponding with JAP2 and HAW2 again displayed the highest levels of diversity (aR = 1.418, HE = 0.204), but the lowest
101
levels of diversity were associated with WAUS (aR = 1.317, HE = 0.145). This result for
WAUS was likely due to the small number of samples in this collection (n = 8). Results from the evaluation of fixed differences revealed a lack of private alleles between any populations, including for the dataset with WAUS grouped separately.
We assessed the level of genetic differentiation between populations by calculating pairwise FST values. These values ranged from 0.0169 between the larger
North Pacific Ocean population (JAP, TAI, HAW, CAL; Table 7) and the population in the eastern central Pacific Ocean (BAJA, ECU, PERU), to 0.0836 between the western
Indian Ocean population (SAF, KEN) and the smaller population in the North Pacific
Ocean (JAP2, HAW2). All FST values calculated between populations were statistically significant at p = 0.000. For calculations performed with WAUS grouped separately from
EAUS+NZ (Table S2), FST between WAUS and EAUS+NZ was equal to 0.0069 and was statistically significant (p = 0.007, pcrit = 0.015). All other pairwise levels of genetic differentiation associated with this dataset were also statistically significant.
Population connectivity
To assess genetic connectivity among striped marlin populations, coalescent-based simulations were performed using MIGRATE-N. Populations in the North Pacific Ocean were combined into a single population for these analyses due to the small sample size of the population corresponding with JAP2 and HAW2 (n = 11). We first evaluated a scenario which included the division of Oceania into sub-populations corresponding with
WAUS and EAUS+NZ. Initial optimization of this model revealed high levels of migration and an inability to achieve convergence for M estimated between WAUS and
102
EAUS+NZ. Subsequent model optimization and final parameter estimates were therefore performed using a model with WAUS+EAUZ+NZ (i.e. Oceania). Results from simulations based on this model revealed low levels of genetic connectivity (M < 25) between the western Indian Ocean and all other populations, including Oceania (Fig. 7).
The highest estimates for M corresponded with migration from the North Pacific Ocean to Oceania (M = 242.93, standard error = 14.01) and to the eastern central Pacific Ocean
(M = 235.68, standard error = 14.28). Migration from Oceania to the North Pacific Ocean
(M = 167.29, standard error = 13.83) and to the eastern central Pacific Ocean (M =
184.67, standard error = 14.87) was also high. Estimates for mutation-scaled effective population sizes (횹) were low, and ranged from 0.001 (standard error = 0.004) for the population in the western Indian Ocean to 0.389 (standard error = 0.001) for the eastern central Pacific Ocean population.
DISCUSSION
The primary goal of this study was to evaluate the population structure of striped marlin throughout its range using genome-wide molecular markers. To accomplish this goal, we characterized nearly 4,000 SNPs across collections of striped marlin from locations throughout the Pacific and Indian oceans. We report the presence of multiple genetically distinct populations corresponding with striped marlin in the western Indian Ocean,
Oceania, and eastern central Pacific Ocean. We also observed the presence of two genetically distinct populations in the North Pacific Ocean. Allele frequencies for replicate collections spanning multiple generations of striped marlin in the eastern central
Pacific Ocean and western South Pacific Ocean were found to be stable for both regions.
103
Finally, estimates of migration rates among populations reveal varying degrees of genetic connectivity between geographically distant regions, including between oceans.
Collectively, the results of this study provide practical information for improving the management of striped marlin.
Biological significance of statistically significant comparisons
The populations of striped marlin resolved in this study were separated by large degrees of genetic differentiation (FST = 0.0169–0.0836) that were also highly statistically significant (p = 0.000), but in a number of instances, comparatively low levels of genetic differentiation were also identified as statistically significant. Low but statistically significant FST values (FST = 0.0067–0.0091, p = 0.001–0.009) were observed for five comparisons between sample collections which, based on results from individual-based clustering, comprise the same population. Low but statistically significant FST values were also observed for comparisons of temporal replicate collections (EAUS: FST =
0.0041, p = 0.001; ECU: FST = 0.0030, p = 0.002). Collectively, FST values observed for these comparisons were considerably smaller than those calculated among populations of striped marlin. This observation suggests that such low levels of genetic differentiation may not be biologically meaningful, and may instead represent statistical error due to the high level of statistical power associated with surveying a large number of genome-wide molecular markers, but comparatively low power associated with the small sample sizes
(n = 8–37 per sample collection; n = 15–22 per temporal replicate) evaluated in this study. However, further exploration of the relationship between FST, sample size, and
104
number of molecular markers is necessary for interpreting the biological significance of low levels of genetic differentiation based on genome-wide SNPs.
Biological support for genetically distinct populations
This study represents the first genetic assessment of population structure for striped marlin in the Indian Ocean. A genetically distinct population of striped marlin in the western Indian Ocean was unambiguously resolved in all individual-based cluster analyses. This population exhibits limited gene flow with striped marlin from the Pacific
Ocean, as evidenced by low degrees of shared ancestry and comparatively small migration rates between these regions. Similarly, the highest levels of genetic differentiation observed in this study corresponded with pairwise comparisons between the western Indian Ocean and Pacific Ocean populations of striped marlin. Migration rates estimated between the western Indian Ocean and Oceania were also low; however, this could be due to the small size of the WAUS sample collection relative to other collections comprising Oceania (EAUS, NZ). The presence of a genetically distinct population of striped marlin in the western Indian Ocean is consistent with biological information suggesting spawning in this region; striped marlin larvae have been collected from waters off the island nations of Réunion and Mauritius as well as from waters extending from Somalia to Tanzania (Nishikawa 1978; Pillai and Ueyanagi 1978).
Information on seasonal movements for striped marlin in the Indian Ocean are limited, but catch per unit effort data from pelagic longline fisheries operating in the western
Indian Ocean suggest north-south migrations corresponding with seasonal aggregations off Kenya and off South Africa (Bromhead et al. 2003). Similarly, satellite tags deployed
105
on striped marlin in waters off Kenya demonstrate movements restricted to the western
Indian Ocean basin (Roy Bealey, African Billfish Foundation, personal communication).
Conventional tagging efforts in this region also include a number of tag recaptures within the western Indian Ocean, except for a single striped marlin which was recaptured off
Perth, Australia (Roy Bealey, African Billfish Foundation, personal communication).
Additional tagging efforts spanning the Indian Ocean are necessary to enable a better understanding of movement patterns for striped marlin in this region.
A genetically distinct population of striped marlin in Oceania was also consistently resolved in individual-based cluster analyses. Results from STRUCTURE and pairwise FST values suggest that the Oceania population is associated with high levels of gene flow with populations from elsewhere in both the Pacific and Indian oceans.
Migration rates estimated using MIGRATE-N also reflect a high level of genetic connectivity between Oceania and populations in the Pacific Ocean; however, migration rates with the western Indian Ocean are comparatively low. Given these results, it is possible that Oceania is a zone of admixture which may facilitate gene flow between the
Pacific and Indian oceans, but a larger number of samples from WAUS is necessary to determine if the apparently high levels of connectivity between Oceania and elsewhere in the Pacific are influenced by comparatively larger sample sizes for EAUS and NZ. Our results provide some evidence that the Oceania population could be divided to reflect distinct sub-populations in the eastern Indian Ocean (off western Australia) and the western South Pacific Ocean (off eastern Australia and New Zealand). This possibility is suggested by results from PCoA, some STRUCTURE simulations, and AMOVA. FST values associated with the pairwise comparison of WAUS to the EAUS and NZ sample
106
collections (FST = 0.0078 and 0.0079 for comparisons with NZ and EAUS, respectively), as well as the level of genetic differentiation between WAUS and EAUS+NZ (FST =
0.0069), were statistically significant; however, these values are less than half of those observed for other population-level comparisons (FST = 0.0169–0.0836). It is possible that the statistical significance of these comparatively low levels of genetic differentiation are the result of type I statistical error at least partially due to the comparatively smaller size (n = 8) of the WAUS sample collection (Waples 1998; discussed in more detail above).
The presence of a genetically distinct population of striped marlin in Oceania is supported by a number of biological observations. Striped marlin spawning has been confirmed for locations off both the eastern and western coasts of Australia (Jones and
Kumaran 1964; Kume and Joseph 1969; Ueyanagi 1974; Hanamoto 1977a; Nishikawa et al. 1978; Nakamura 1983; Kopf et al. 2012). Striped marlin larvae have also been collected off northern Australia in the Banda and Timor seas (Ueyanagi and Wares 1975).
Tagging efforts in Oceania have largely been limited to waters off eastern Australia and
New Zealand, and are consistent with relatively localized movements in this region, although a number of long distance migrations as far east as French Polynesia have been observed (Ortiz et al. 2003; Domeier 2006; Holdsworth et al. 2009; Sippel et al. 2011;
Holdsworth and Saul 2014). No inter-oceanic movements have been reported for striped marlin; however, tagging and reporting efforts in the Indian Ocean have been limited.
Possible mechanisms facilitating genetic connectivity between striped marlin off eastern and western Australia include passive inter-oceanic drift of eggs and/or larvae, as well as directed inter-oceanic movements and subsequent spawning of adult fish. A
107
number of studies demonstrate the significant impact of larval exchange among geographically distant regions on observed patterns of genetic connectivity in marine species (White et al. 2010; Selkoe and Toonen 2011). Given the location of striped marlin spawning grounds off Australia, inter-oceanic drift of striped marlin eggs and/or larvae may be more likely to occur around the northern coast of Australia; however, information on pelagic larval duration for striped marlin as well as additional ichthyoplankton sampling efforts are required to test this hypothesis. In comparison, genetic connectivity facilitated by inter-oceanic movements of mature fish may be more likely to occur around southern Australia, as the Torres Strait connecting the Coral Sea to the Gulf of
Carpentaria off northern Australia is characterized by extensive shallow (15–20 m deep) stretches with limited flow (Wolanski et al. 1988). Relative to other istiophorid billfishes, striped marlin typically inhabit more temperate waters (20–25 ºC sea surface temperature;
Howard and Ueyanagi 1965; Sippel et al. 2007) and are seasonally abundant in waters as far south as Tasmania (Bromhead et al. 2003); these characteristics suggest the possibility of inter-oceanic movements for striped marlin in some years.
A population of striped marlin that spans the North Pacific Ocean was also consistently resolved in individual-based cluster analyses. Multiple lines of evidence suggest that this population displays high levels of gene flow with the eastern Central
Pacific Ocean and Oceania, but limited gene flow with the western Indian Ocean.
Biological information consistent with the presence of a genetically distinct population of striped marlin in the North Pacific Ocean includes spawning activity reported for waters off Taiwan and Japan (Nakamura 1949; Nishikawa et al. 1978; Sun et al. 2011; Chang et al. 2017). The identification of a small number of striped marlin larvae off Hawaii (Hyde
108
et al. 2006), as well as seasonally abundant juveniles (approximately 5–30 kg; Bromhead et al. 2003), suggests that this region of the North Pacific may also be used for spawning.
A large number of tag recaptures are consistent with frequent movements of striped marlin between Hawaii and southern California (Ortiz et al. 2003); however, movements between the western North Pacific Ocean and elsewhere in the North Pacific have not been observed, presumably due to comparatively limited tagging and reporting efforts in that region.
Our results are consistent with the presence of a second population in the North
Pacific Ocean, corresponding with a subset of striped marlin sampled off Japan and
Hawaii. This population was resolved across individual-based cluster analyses, and
AMOVA scenarios with this population grouped separately were favorable to scenarios with all striped marlin from North Pacific Ocean grouped together. Previous studies have suggested the presence of more than one genetically and/or biologically distinct group of striped marlin off Hawaii. Purcell and Edmands (2011) report a statistically significant level of genetic differentiation between reproductively immature and mature striped marlin sampled off Hawaii. However, that result was based on genotype data corrected for the presence of null alleles, and comparisons with non-corrected data were not significant. In addition, maturity was indirectly determined using values for length or weight at first maturity previously published for striped marlin from the Coral Sea
(Hanamoto 1977b), and may not be accurate predictors of maturity for striped marlin from elsewhere in the Indo-Pacific. Bromhead et al. (2003) report a bimodal size distribution corresponding with the presence of very young (mode 110 cm EFL) and older (mode 160 cm EFL) striped marlin caught on pelagic longline gear in the central
109
North Pacific Ocean. In the present study, biological data is incomplete for the subset of striped marlin sampled off Japan and Hawaii identified as genetically distinct from other striped marlin sampled from the North Pacific Ocean. It is possible that the presence of two genetically distinct populations in the North Pacific Ocean is facilitated by spawning in the western North Pacific off Japan and Taiwan, and in the central North Pacific off
Hawaii (discussed above). It is also possible that one of the North Pacific Ocean populations corresponds with a spawning ground previously confirmed in the central
South Pacific Ocean: Kume and Joseph (1969) report the presence of mature female striped marlin in the region spanning 125˚W–130˚W and 20˚S–25˚S, and the presence of larvae between 140˚W–145˚W and 15˚S–20˚S was reported by Nishikawa et al. (1978).
To improve our understanding of population structure for striped marlin in the North
Pacific Ocean, sampling efforts that target the range of striped marlin demographic groups seasonally abundant off Japan and Hawaii are necessary.
A population of striped marlin in the eastern central Pacific Ocean was also unambiguously resolved in this study, and the presence of this population is consistent with available biological information. Striped marlin in the eastern central Pacific Ocean are characterized by high degrees of gene flow with populations in the North Pacific
Ocean and Oceania, but gene flow with the western Indian Ocean population is limited.
Spawning of striped marlin in the eastern central Pacific Ocean has been confirmed for waters extending from the Gulf of California to the central coast of Mexico (Kume and
Joseph 1969; Eldridge and Wares 1974; González-Armas et al. 1999, 2006). In contrast to the genetic results presented here and in previous studies (McDowell and Graves 2008;
Purcell and Edmands 2011), tagging studies demonstrate frequent movements of striped
110
marlin between southern California and Baja California (Ortiz et al. 2003; Bromhead et al. 2003; Domeier 2006). Conventional tags deployed on striped marlin off southern
California have been recovered off Hawaii and off Baja California (Ortiz et al. 2003;
Bromhead et al. 2003). Similarly, movements inferred from satellite tags reveal significant spatial overlap between striped marlin tagged off southern California and off
Baja California (Domeier 2006). The discrepancy between results from genetic and tagging studies is likely due to the seasonal mixing of populations from the North Pacific and eastern central Pacific on feeding grounds spanning waters off southern California and Baja California (Bromhead et al. 2003). A resident eastern central Pacific Ocean population purportedly remains in the Baja California region for spawning (Ortega-
García et al. 2003), while individuals from the North Pacific population spawn elsewhere.
The Pacific Ocean populations of striped marlin resolved in this study are similar to those identified in previous studies using limited numbers of molecular markers
(Graves and McDowell 1994; McDowell and Graves 2008; Purcell and Edmands 2011).
Graves and McDowell (1994) observed statistically significant genetic heterogeneity among striped marlin sampled from waters off Baja California, Ecuador, eastern
Australia, and Hawaii based on restriction fragment length polymorphism analysis of mtDNA. Genetically distinct populations corresponding with these locations were resolved in subsequent studies based on the analysis of microsatellite markers and mtDNA, except additional sample collections from the North Pacific Ocean revealed the wide geographic extent of this population (McDowell and Graves 2008; Purcell and
Edmands 2011). In the present study, we also resolved striped marlin sampled from the
North Pacific Ocean, western South Pacific Ocean, and eastern central Pacific Ocean as
111
genetically distinct; however, in contrast to previous studies, we identified striped marlin sampled off Ecuador as belonging to the same population as striped marlin sampled off
Baja California. This discrepancy between studies is likely due to differences in statistical power; sample sizes for these regions are similar between studies, but the significantly larger number of molecular markers (and total number of alleles) evaluated in the present study corresponds with an improved ability to resolve genetic population structure
(Chapter II).
A range of population structuring scenarios has been reported for other istiophorid billfishes with spatial distributions spanning the Pacific and Indian oceans. Based on the analysis of microsatellite markers and mtDNA, Williams et al. (2015) resolved three genetically distinct populations of black marlin (Istiompax indica) corresponding with fish sampled off eastern Australia, western and northern Australia, and in the South
China Sea. The more temperate thermal preference of striped marlin and potential movement of this species around southern Australia may account for differences in population structure observed between striped marlin and black marlin in waters off
Australia. Assessments of microsatellite markers and mtDNA have also revealed the presence of population structuring for sailfish (Istiophorus platypterus) in the Pacific
Ocean, including genetically distinct populations in the eastern and western Pacific
(McDowell 2002; Lu et al. 2015). In the Indian Ocean, a distinct population of sailfish has been identified in the Arabian Gulf (Hoolihan et al. 2004); however, comparisons between fish sampled off Kenya and off western Australia suggest sailfish from these regions comprise a single population (McDowell 2002). In comparison, genetic assessments of population structure for blue marlin (Makaira nigricans) in the Pacific
112
Ocean based on mtDNA and a variety of nuclear markers are consistent with a single ocean-wide population for this species (Buonaccorsi et al. 1999, 2001); genetic population structure for blue marlin in the Indian Ocean has not been evaluated.
Differences in population structuring among istiophorids in the Indo-Pacific likely correspond with variation in biological characteristics, including thermal preferences, dispersal capabilities, and spawning site fidelity.
Temporal stability of allele frequencies
The temporal replicate sample collections evaluated in this study provide a rare opportunity to evaluate the stability of allelic frequencies across multiple generations of striped marlin. Based on an average generation time of 4.4 years (Collette et al. 2011), temporal replicates collected off Ecuador and off eastern Australia span approximately five generations and four generations of striped marlin, respectively. Results from individual-based cluster analyses and AMOVA provided no evidence of a difference in allele frequencies between temporal replicates for both regions. In contrast, FST values calculated between temporal replicates were statistically significant. The lack of consistency in these results could be due to multiple factors, including differing capabilities of cluster-based analyses and FST to represent low levels of genetic differentiation. However, FST values associated with the temporal replicates are less than
25% of those associated with population-level comparisons, and suggest that the level of genetic differentiation observed between temporal replicates likely corresponds with sampling error rather than a meaningful shift in allele frequencies over time (discussed in more detail above). Statistical significance of FST values for comparisons of temporal
113
replicates may also be attributed to intergenerational noise (e.g. due to differential reproductive success [Hedgecock 1994] or environmental variation [Selkoe et al. 2008]).
Collectively, results from the comparison of temporal replicate collections in this study are consistent with the stability of observed spatial structuring for at least the past 16 years.
Maintenance of population structure
The presence of genetically distinct populations for some highly migratory pelagic fishes suggests a mechanism that limits intraspecific gene flow among some geographic regions. This mechanism may be multi-faceted, but must at least include a high degree of fidelity to natal spawning location, as even a relatively small number of migrants per generation could mask genetic differentiation between populations (Waples 1998). While a comparatively large number of satellite tags have been deployed on striped marlin
(Domeier 2006; Sippel et al. 2007; Holdsworth et al. 2009), tag deployment periods have been less than one year and are not useful for evaluating annual movement patterns or spawning site fidelity. A small number of conventional tag recaptures with times at liberty in excess of one year have also been reported for striped marlin, but these data only provide information on locations of tag deployment and recapture, and do not correspond with any patterns suggestive of annual site fidelity (Ortiz et al. 2003). Recent studies that have used satellite tags to study seasonal movements in other istiophorid billfish species include a few reports of deployment periods spanning approximately one year (Rooker et al. 2013; Loose 2014; Lam et al. 2016). Movements for these individuals show a cyclical pattern that includes time spent on known spawning grounds during the
114
spawning season. Though such observations do not provide information on the occurrence of spawning site fidelity in istiophorids, they do provide evidence that at least some individuals make regional movements on annual time scales, and such movements correspond with what is known regarding spatiotemporal spawning in these species.
Movements inferred via putative migrants
Individual-based cluster analyses not only revealed the presence of four genetically distinct populations of striped marlin, but also identified eight putative migrants sampled from locations geographically distant from their presumed source population. Seven of these putative migrants were genetically consistent with striped marlin from Oceania, but were sampled off Hawaii (n = 4; eye fork length = 118–170 cm) and off Ecuador (n = 3; biological information not available). An eighth putative migrant was genetically consistent with striped marlin from the eastern central Pacific Ocean, but was sampled off southern California (biological information not available). Migration rates corresponding with the movements represented by these putative migrants are high (M = 75.21–184.67), including for movements in the opposite direction (M = 99.15–242.93), and suggest some regularity to the straying of individuals between these populations. Though movements between striped marlin from the eastern central Pacific Ocean and southern California have been frequently observed in tagging studies, movements between Oceania and
Hawaii or Ecuador have not been previously reported. The occurrence of these putative migrants provides information essential for understanding the degree of connectivity among geographically distant populations of striped marlin. Of particular concern to fisheries managers is the identification of regions utilized by multiple stocks, where it is
115
possible for fisheries to interact with striped marlin characterized by different propensities for sustaining fishing pressure.
Management implications and concluding remarks
The results of this study highlight inconsistencies between stocks currently recognized for the assessment and management of striped marlin, and what is known regarding the life history and genetic population structure of this species (Fig. 8). Accounting for spatial genetic structure is essential for conserving population-level genetic diversity necessary for short- and long-term population persistence (Gaggiotti and Vetter 1999;
Allendorf et al. 2008, Allendorf 2014; Pinksy and Palumbi 2014; Spies et al. 2015). This is especially important given that striped marlin is overfished and/or experiencing overfishing in a number of regions throughout the species range. Based on our results, we recommend that the single ocean-wide stock of striped marlin currently recognized by the
IOTC be subdivided to reflect a distinct population of striped marlin in the western Indian
Ocean. This population is of particular management concern given its overfished status and comparatively limited genetic connectivity with other populations of striped marlin.
In addition, the western Indian Ocean population is associated with the lowest levels of genetic diversity and the smallest effective population size observed in this study. The presence of a distinct population of striped marlin in Oceania demonstrates the need for joint management efforts by the IOTC and WCPFC. The distinct management units currently recognized by the IATTC and WCPFC in the western and central North Pacific
Ocean and in the eastern Pacific Ocean should be combined to reflect a single stock spanning the North Pacific Ocean. Additional study is required to determine the most
116
appropriate management strategy for a second population in the North Pacific Ocean corresponding with a proportion of striped marlin off Japan and Hawaii. Finally, a separate management unit for striped marlin in the eastern central Pacific Ocean should be reflected in IATTC management plans for this species. Collectively, these changes will reduce uncertainties currently associated with the management of striped marlin and improve the effectiveness of management and conservation efforts for this species.
It is possible that additional fine-scale population structure for striped marlin exists but was not observed despite the large number of genome-wide molecular markers evaluated in this study. A relatively small number of migrants per generation could facilitate enough gene flow to mask biologically distinct populations of striped marlin
(Waples 1998), precluding the detection of some populations with genetic methods.
Based on migration rates estimated between the populations resolved in this study, rates higher than approximately 250 migrants per generation would be necessary to mask additional population structure. In such instances, it would be important to evaluate population structure using a range of non-genetic approaches (e.g. Cadrin et al. 2005). It is also possible that the opportunistic sampling design employed in this study prohibited the detection of additional population structure. Ideally, population genetic studies of highly migratory pelagic fishes would employ experimental designs which focus sampling efforts on source populations by targeting larvae and/or reproductively active adults on spawning grounds (Chapter II; Graves et al. 1996; Bowen et al. 2005; Carlsson et al. 2007; Graves and McDowell 2015). However, larval sampling efforts for striped marlin and other highly migratory pelagic fishes are limited, and determining the reproductive status of adult fish is challenging without the use of lethal methods.
117
Regardless of these factors, it is clear from our results that at least four genetically distinct populations of striped marlin are present in the Indo-Pacific.
Further study is required to improve our understanding of factors contributing to the development and maintenance of population structure in highly migratory pelagic fishes. For example, satellite tag technology (e.g. internal archival tags; Block et al. 2005) that enables deployment periods of greater than one year could provide information on the occurrence of spawning site fidelity in these species. In addition, continuing efforts to improve genomic resources available for highly migratory pelagic fishes will allow researchers to address evolutionary questions that have been unexplored for the pelagic environment, including the role of natural selection in facilitating localized adaptation and population structuring. Working toward a better scientific understanding of istiophorids and other highly migratory pelagic fishes will assist the development of informed management and conservation efforts which promote the availability of these resources for future generations.
ACKNOWLEDGMENTS
We thank a number of recreational anglers throughout the Indo-Pacific region for providing samples for this study, especially captain Trevor Hansen. We also thank multiple scientists and conservation groups for facilitating sample collection, including
Sofia Ortega-Garcia, Sam Williams, John Holdsworth, and the NOAA Pacific Islands
Fisheries Science Center longline fishery observer program. In particular, efforts from the
African Billfish Foundation (Roy Bealey), International Game Fish Association (Rob
Kramer, Jason Schratwieser, and several international representatives), and Japan Game
118
Fish association (Tomo Higashi) were essential to the success of this project.
119
REFERENCES
Abràmoff, M., Magalhães, P., and Ram, S. 2004. Image processing with ImageJ.
Biophotonics Intern. 11: 36–42.
Adamack, A., and Gruber, B. 2014. POPGENREPORT: simplifying basic population
genetic analyses in R. Methods Ecol. Evol. 5: 384–387.
Allendorf, F., Berry, O., and Ryman, N. 2014. So long to genetic diversity, and thanks for
all the fish. Mol. Ecol. 23: 23–25.
Allendorf, F., England, P., Luikart, G., Ritchie, P., and Ryman, N. 2008. Genetic effects
of harvest on wild animal populations. Trends Ecol. Evol. 23: 327–337.
Beaumont, M., and Balding, D. 2004. Identifying adaptive genetic divergence among
populations from genome scans. Mol. Ecol. 13: 969–980.
Beaumont, M., and Nichols, R. 1996. Evaluating loci for use in the genetic analysis of
population structure. Proc. R. Soc. Lond. B. 263: 1619–1626.
Beerli, P. and Palczewski, M. 2010. Unified framework to evaluate panmixia and
migration direction among multiple sampling locations. Genetics, 185: 313–326.
Benjamini, Y., and Yekutieli, D. 2001. The control of false discovery rate under
dependency. Ann. Stat. 29: 1165–1188.
Block, B., Teo, S., Walli, A., Boustany, A., Stokesbury, M., Farwell, C., Weng, K.,
Dewar, H., and Williams, T. 2005. Electornic tagging and population structure of
Atlantic bluefin tuna. Nature, 434: 1121–1127.
Bowen, B., Bass, A., Soares, L., and Toonen, R. 2005. Conservation implications of
complex population structure: lessons from the loggerhead turtle (Caretta
caretta). Mol. Ecol. 14: 2389–2402.
120
Bradbury, I., Hubert, S., Higgins, B., Bowman, S., Paterson, I., Snelgrove, P., Morris, C.,
Gregory, R., Hardie, D., Borza, T., and Bentzen, P. 2011. Evaluation SNP
ascertainment bias and its impact on population assignment in Atlantic cod,
Gadus morhua. Mol. Ecol. Resour. 11: 218–225.
Bromhead, D., Pepperell, J., Wise, B., and Findlay, J. 2003. Striped marlin: biology and
fisheries. Canberra, Australia. Bureau of Rural Sciences.
Buonaccorsi, V., McDowell, J., and Graves, J. 2001. Reconciling patterns of inter–ocean
molecular variance from four classes of molecular markers in blue marlin
(Makaira nigricans). Mol. Ecol. 10: 1179–1196.
Buonaccorsi, V., Reece, K., Morgan, L., and Graves, J. 1999. Geographic distribution of
molecular variance within the blue marlin (Makaira nigricans): a hierarchical
analysis of allozyme, single–copy nuclear DNA, and mitochondrial DNA
markers. Evolution, 53: 568–579.
Cadrin, S., Friedland, K., and Waldman, J. 2005. Stock Identification Methods:
Applications in Fishery Science. Elsevier Academic Press, Burlington,
Massachusetts, USA.
Carlsson, J., McDowell, J., Carlsson, J., and Graves, J. 2007. Genetic identity of YOY
bluefin tuna from the eastern and western Atlantic spawning areas. J. Hered. 98:
23–28.
Carlsson, J., McDowell, J., Díaz-Jaimes, P., Carlsson, J., Boles, S., Gold, J., and Graves,
J. 2004. Microsatellite and mitochondrial DNA analyses of Atlantic bluefin tuna
(Thunnus thynnus thynnus) population structure in the Mediterranean Sea. Mol.
Ecol. 13: 3345–3356.
121
Chang, H., Sun, C., Yeh, S., Chang, Y., Su, N., and DiNardo, G. 2017. Reproductive
biology of female striped marlin Kajikia audax in the western Pacific Ocean. J.
Fish Biol. 92: 105–130.
Collette, B., Carpenter, K., Polidoro, B., Juan-Jordá, M., Boustany, A., Die, D., Elfes, C.,
Fox, W., Graves, J., Harrison, L., McManus, R., Minte-Vera, C., Nelson, R.,
Restrepo, V., Schratwieser, J., Sun, C., Amorim, A., Brick Peres, M., Canales, C.,
Cardenas, G., Chang, S., Chiang, W., de Oliveira Leite, Jr., N., Harwell, H.,
Lessa, R., Fredou, F., Oxenford, H., Serra, R., Shao, K., Sumaila, R., Wang, S.,
Watson, R., and Yáñez, E. 2011. High value and long live–double jeopardy for
tunas and billfishes. Science, 333: 291–292.
Davies, C., Gosling, E., Was, A., Brophy, D., and Tysklind, N. 2011. Microsatellite
analysis of albacore tuna (Thunnus alalunga): population genetic structure in the
North–East Atlantic Ocean and Mediterranean Sea. Mar. Biol. 158: 2727–2740.
Domeier, M. 2006. An analysis of Pacific striped marlin (Tetrapturus audax) horizontal
movement patterns using pop–up satellite archival tags. Bull. Mar. Sci. 79: 811–
825.
Earl, D., and vonHoldt, B. 2012. STRUCTURE HARVESTER: a website and program for
visualizing STRUCTURE output and implementing the Evanno method. Conserv.
Genet. Resour. 4: 359–361.
Eldridge, M., and Wares, P. 1974. Some biological observations of billfishes taken in the
Eastern Pacific Ocean. In Proceedings of the International Billfish Symposium,
Kailua-Kona, Hawaii, 9-12 August 1972. Part 2. Review and Contributed Papers.
122
Edited by R. Shomura, and F. Williams. National Marine Fisheries Service,
Seattle, Washington, USA. NOAA Technical Report NMFS SSRF-675.
Elshire R., Glaubitz, J., Sun, Q., Poland, J., Kawamoto, K., Buckler, E., and Mitchell, S.
2011. A robust, simple genotyping–by–sequencing (GBS) approach for high
diversity species. PLoS ONE, 6: e19379.
Evanno, G., Regnaut, S., and Goudet, J. 2005. Detecting the number of clusters of
individuals using the software STRUCTURE: a simulation study. Mol.Ecol. 14:
2611–2620.
Excoffier, L., Hofer, T., and Foll, M. 2009. Detecting loci under selection in a
hierarchically structured population. Heredity, 103: 285–298.
Excoffier, L., and Lischer, H. 2010. Arlequin suite ver 3.5: a new series of programs to
perform population genetics analyses under Linux and Windows. Mol. Ecol.
Resour. 10: 564–567.
Excoffier, L., Smouse, P., and Quattro, J. 1992. Analysis of molecular variance inferred
from metric distances among DNA haplotypes: application to human
mitochondrial DNA restriction data. Genetics, 131: 479–491.
Falush, D., Stephens, M., and Pritchard, J. 2003. Inference of population structure using
multilocus genotype data: linked loci and correlated allele frequencies. Genetics,
164: 1567–1587.
Foll, M., and Gaggiotti, O. 2008. A genome–scan method to identify selected loci
appropriate for both dominant and codominant markers: a Bayesian perspective.
Genetics, 180: 977–993.
123
Gaggiotti, O., and Vetter, R. 1999. Effect of life history strategy, environmental
variability, and overexploitation on the genetic diversity of pelagic fish
populations. Can. J. Fish. Aquat. Sci. 56: 1376–1388.
González–Armas, R., Klett–Traulsen, A., and Hernández–Herrera, A. 2006. Evidence of
billfish reproduction in the southern Gulf of Califormia, Mexico. Bull. Mar. Sci.
79: 705–717.
González–Armas, R., Sosa–Nishizaki, O., Funes–Rodríguez R., and Levy–Pérez, V.
1999. Confirmation of the spawning area of the striped marlin, Tetrapturus audax,
in the so–called core area of the eastern tropical Pacific off Mexico. Fish.
Oceanogr. 8: 238–242.
Gower, J. 1966. Some distance properties of latent root and vector methods used in
multivariate analysis. Biometrika, 53: 325–338.
Graves, J., Gold, J., Ely, B., Quattro, J., Woodley, C., and Dean, J. 1996. Population
genetic structure of bluefin tuna in the North Atlantic Ocean. Identification of
variable genetic markers. Coll. Vol. Sci. Papers ICCAT, 45: 155–157.
Graves, J., and McDowell, J. 1994. Genetic analysis of striped marlin (Tetrapturus
audax) population structure in the Pacific Ocean. Can. J. Fish. Aquat. Sci. 51:
1762-1768.
Graves, J., and McDowell, J. 2015. Population structure of istiophorid billfishes. Fish.
Res. 166: 21–28.
Grewe, P., Feutry, P., Hill, P., Gunasekera, R., Schaefer, K., Itano, D., Fuller, D., Foster,
S., and Davies, C. 2015. Evidence of discrete yellowfin tuna (Thunnus albacares)
124
populations demands rethink of management for this globally important resource.
Sci. Rep. 5: 16916.
Gruber, B., Unmack, P., Berry, O., and Georges, A. 2018. DARTR: and R package to
facilitate analysis of SNP data generated from reduced representation genome
sequencing. Mol. Ecol. Resour. doi: 10.1111/1755-0998.12745.
Hanamoto, E. 1977a. Fishery oceanography of striped marlin. I. Fishing season, fishing
ground and movement pattern of the fish in the southern Coral Sea. Bull. Jap. Soc.
Sci. Fish. 43: 649–657.
Hanamoto, E. 1977b. Fishery oceanography of striped marlin. II. Spawning activity of
the fish in the Southern Coral Sea. Bull. Jap. Soc. Sci. Fish. 43: 1279–1286.
Hedgecock, D. 1994. Does variance in reproductive success limit effective population
sizes of marine organisms? In Genetics and Evolution of Aquatic Organisms.
Edited by A. Beaumont. Chapman and Hall, London.
Holdsworth, J., and Saul, P. 2014. New Zealand billfish and gamefish tagging, 2012–13.
Wellington, New Zealand. Ministry for Primary Industries.
Holdsworth, J., Sippel, T., and Block, B. 2009. Near real time satellite tracking of striped
marlin (Kajikia audax) movements in the Pacific Ocean. Mar. Biol. 156: 505–
514.
Hoolihan, J., Premanandh, J. D’Aloia–Palmieri, M., and Benzie, J. 2004. Intraspecific
phylogeographic isolation of Arabian Gulf sailfish Istiophorus platypterus
inferred from mitochondrial DNA. Mar. Biol. 145: 465–475.
125
Howard, J., and Ueyanagi, S. 1965. Distribution and relative abundance of billfishes
(Istiophoridae) of the Pacific Ocean. University of Miami Press, Miami, Florida,
USA.
Hubisz, M., Falush, D., Stephens, M., and Pritchard, J. 2009. Inferring weak population
structure with the assistance of sample group information. Mol. Ecol. Resour. 9:
1322–1332.
Hyde, J., Humpherys, Jr., R., Musyl, M., Lynn, E., and Vetter, R. 2006. A central North
Pacific spawning ground for striped marlin, Tetrapturus audax. Bull. Mar. Sci.
79: 683–690.
IATTC. 2017. Tunas, billfishes and other pelagic species in the eastern Pacific Ocean in
2016. La Jolla, California, USA. Inter-American Tropical Tuna Commission.
IOTC. 2017. Report of the 15th session of the IOTC working party on billfish. Victoria
Mahé, Seychelles. Indian Ocean Tuna Commission.
Jakobsson, M., and Rosenberg, N. 2007. CLUMPP: a cluster matching and permutation
program for dealing with label switching and multimodality in analysis of
population structure. Bioinformatics, 23: 1801–1806.
Janes, J., Miller, J., Dupuis, J., Malenfant, R., Gorrell, J., Cullingham, C., and Andrew, R.
2017. The K = 2 conundrum. Mol. Ecol. 26: 3594–3602.
Jombart, T. 2008. Adegenet: a R package for the multivariate analysis of genetic markers.
Bioinformatics, 24: 1403–1405.
Jombart, T., Devillard, S., and Balloux, F. 2010. Discriminant analysis of principal
components: a new method for the analysis of genetically structured populations.
BMC Genet. 11: 1471–2156.
126
Jombart, T., Pontier, D., and Dufour, A. 2009. Genetic markers in the playground of
multivariate analysis. Heredity, 102: 330–341.
Jones, S., and Kumaran, M. 1964. Distribution of Larval Billfishes (Xiphiidae and
Istiophoridae) in the Indo-Pacific with Special Reference to the Collections Made
by the Danish Dana Expedition. In Proceedings of the Symposium on Scombroid
Fishes, Part I. 12-15 January 1962. Mandapam, India.
Kamvar, Z., Tabima, J., and Grünwald, N. 2014. Poppr: an R package for genetic
analysis of populations with clonal, partially clonal, and/or sexual reproduction.
PeerJ, 2: e281.
Kilian, A., Wenzl, P., Huttner, E., Carling, J., Xia, L., Blois, H., Caig, V., Heller-
Uszynska, K., Jaccoud, D., Hopper, C., Aschenbrenner-Kilian, M., Evers, M.,
Peng, K., Cayla, C., Hok, P., and Uszynski, G. 2012. Diversity arrays technology:
a generic genome profiling technology on open platforms. In Data Production and
Analysis in Population Genomics: Methods and Protocols. Edited by F.
Pompanon, and A. Bonin. Humana Press, Totowa, New Jersey, USA.
Kopf, R., Davie, P., Bromhead, D., and Young, J. 2012. Reproductive biology and
spatiotemporal patterns of spawning in striped marlin Kajikia audax. J. Fish. Biol.
81: 1834–1858.
Kume, S., and Joseph, J. 1969. Size composition and sexual maturity of billfish caught by
the Japanese longline fishery in the Pacific Ocean east of 130W. Bull. Far Seas
Fish. Res. Lab. 2: 115–162.
127
Lam, C., Galuardi, B., Mendillo, A., Chandler, E., and Lutcavage, M. 2016. Sailfish
migrations connect productive coastal areas in the West Atlantic Ocean. Sci. Rep.
6: 38163.
Loose, E. 2014. Seasonal movements, habitat utilization, and comparative scale
morphology of white marlin (Kajikia albida) and roundscale spearfish
(Tetrapturus georgii). Master’s thesis. Virginia Institute of Marine Science,
College of William & Mary, Gloucester Point, Virgina, USA.
Lotterhos, K., and Whitlock, M. 2014. Evaluation of demographic history and neutral
parameterization on the performance of FST outlier tests. Mol. Ecol. 23: 2178–
2192.
Lu, C., Alvarado–Bremer, J., McKenzie, J., and Chiang, W. 2015. Analysis of sailfish
(Istiophorus platypterus) population structure in the North Pacific Ocean. Fish.
Res. 166: 33–38.
Luikart, G., England, P., Tallmon, D., Jordan, S., and Taberlet, P. 2003. The power and
promise of population genomics: from genotyping to genome typing. Nature Rev.
Genet. 4: 981–994.
Mamoozadeh, N., McDowell, J., Rooker, J., and Graves, J. In press. Genetic evaluation
of population structure in white marlin (Kajikia albida): the importance of
statistical power. ICES J. Mar. Sci. doi: 10.1093/icesjms/fsx047.
Mardis, E. 2008. The impact of next-generation sequencing technology on genetics.
Trends Genet. 24: 133–141.
128
McDowell, J. 2002. Genetic stock structure of the sailfish, Istiophorus platypterus, based
on nuclear and mitochondrial DNA. Doctoral dissertation. Virginia Institute of
Marine Science, College of William & Mary, Gloucester Point, Virgina, USA.
McDowell, J., and Graves, J. 2008. Population structure of striped marlin (Kajikia audax)
in the Pacific Ocean based on analysis of microsatellite and mitochondrial DNA.
Can. J. Fish. Aquat. Sci. 65: 1307–1320.
Nakamura, H. 1949. The tunas and their fisheries. Takeuchi Shobo, Tokyo, Japan.
Nakamura, I. 1983. Systematics of billfishes (Xiphiidae and Istiophoridae). Publications
of the Seto Marine Biological Laboratory 28: 255–396.
Nakamura, I. 1985. Billfishes of the world: an annotated and illustrated catalogue of
marlins, sailfishes, spearfishes, and swordfishes known to date. FAO Fisheries
Synopsis No. 125, vol. 5. 65 pp. Food and Agricultural Organization, Rome, Italy.
Narum, S. 2006. Beyond Bonferroni: less conservative analyses for conservation
genetics. Conserv. Genet. 7: 783–787.
Narum, S., and Hess, J. 2011. Comparison of FST outlier tests for SNP loci under
selection. Mol. Ecol. Resour. 11: 184–194.
Nishikawa, Y., Kikawa, S., Honma, M., and Ueyanagi, S. 1978. Distribution atlas of
larval tunas, billfishes and related species. Results of larval surveys by R/V
Shunyo Maru and Shoyo Maru, 1956-1975. Far Seas Fisheries Research
Laboratory, Shimizu, Japan.
Ortega-García, S., Klett-Traulsen, A., and Ponce-Díaz, G. 2003. Analysis of sportfishing
catch rates of striped marlin (Tetrapturus audax) at Cabo San Lucas, Baja
129
California Sur, Mexico, and their relation to sea surface temperature. Mar.
Freshwater Res. 54: 483–488.
Ortiz, M., Prince, E., Serafy, J., Holts, D., Davy, K., Pepperell, J., Lowry, M., and
Holdsworth, J. C. 2003. Global overview of the major constituent–based billfish
tagging programs and their results since 1954. Mar. Freshwater Res. 54: 489–507.
Ovenden, J., Berry, O., Welch, D., Buckworth, R., and Dichmont, C. 2015. Ocean’s
eleven: a critical evaluation of the role of population, evolutionary and molecular
genetics in the management of wild fisheries. Fish Fish. 16: 125–159.
Peel, E., Nelson, R., and Goodyear, C. 2003. Managing Atlantic marlin as bycatch under
ICCAT. The fork in the road: recovery or collapse. Mar. Freshwater Res. 54:
575–584.
Peterson, B., Weber, J., Kay, E., Fisher, H., and Hoekstra, H. 2012. Double digest
RADseq: an inexpensive method for de novo SNP discovery and genotyping in
model and non–model species. PLoS ONE, 7: e37135.
Pillai, P., and Ueyanagi, S. 1978. Distribution and biology of the striped marlin
Tetrapturus audax (Philippi) taken by the longline fishery in the Indian Ocean.
Bull. Natl. Res. Inst. Far Seas Fish. 16: 9–32.
Pinsky, M., and Palumbi, S. 2014. Meta–analysis reveals lower genetic diversity in
overfished populations. Mol. Ecol. 23: 29–39.
Pritchard, J., Stephens, M., and Donnelly, P. 2000. Inference of population structure
using multilocus genotype data. Genetics, 155: 945–959.
130
Purcell, C., and Edmands, S. 2011. Resolving the genetic structure of striped marlin,
Kajikia audax, in the Pacific Ocean through spatial and temporal sampling of
adult and immature fish. Can. J. Fish. Aquat. Sci. 68: 1861–1875.
R Core Team. 2017. R: a language and environment for statistical computing.
Reeb, C., Arcangeli, L., and Block, B. 2000. Structure and migration corridors in Pacific
populations of the swordfish Xiphius gladius, as inferred through analyses of
mitochondrial DNA. Mar. Biol. 136: 1123–1131.
Reiss, H., Hoarau, G., Dickey–Collas, M., and Wolff, W. 2009. Genetic population
structure of marine fish: mismatch between biological and fisheries management
units. Fish Fish. 10: 361–395.
Roesti, M., Salzburger, W., and Berner, D. 2012. Uninforomative polymorphisms bias
genome scans for signatures of selection. BMC Evol. Biol. 12: 94.
Rooker, J., Kitchens, L., Dance, M., Wells, R., Falterman, B., and Cornic, M. 2013.
Spatial, temporal, and habitat–related variation in abundance of pelagic fishes in
the Gulf of Mexico: potential implications of the Deepwater Horizon oil spill.
PLoS ONE, 8: e76080.
Rosenberg, N. 2004. DISTRUCT: a program for the graphical display of population
structure. Mol. Ecol. Notes. 4: 137–138.
Rousset, F. 2000. Inferences from spatial population genetics. In Handbook of Statistical
Genetics. Edited by D. Balding, M. Bishop, and C. Cannings. Wiley & Sons, Inc.,
Hoboken, New Jersey, USA.
Sansaloni, C., Petroli, C., Jaccoud, D., Carling, J., Detering, F., Grattapaglia, D., and
Kilian, A. 2011. Diversity Arrays Technology (DArT) and next–generation
131
sequencing combined: genome–wide, high throughput, highly informative
genotyping for molecular breeding of Eucalyptus. BMC Proceedings, 5: P54.
Selkoe, K., Henzler, C., and Gaines, S. 2008. Seascape genetics and the spatial ecology
of marine populations. Fish Fish. 9: 363–377.
Selkoe, K., and Toonen, R. 2011. Marine connectivity: a new look at pelagic larval
duration and genetic metrics of dispersal. Mar. Ecol. Prog. Ser. 436: 291–305.
Seutin, G., White, B., and Boag, P. 1991. Preservation of avian blood and tissue samples
for DNA analyses. Can. J. Zool. 69: 82–90.
Sippel, T., Davie, P., Holdsworth, J., and Block, B. 2007. Striped marlin (Tetrapturus
audax) movements and habitat utilization during a summer and autumn in the
Southwest Pacific Ocean. Fish. Oceanogr. 16: 459–472.
Sippel, T., Holdsworth, J., Dennis, T., and Montgomery, J. 2011. Investigating behaviour
and populations dynamics of striped marlin (Kajikia audax) from the Southwest
Pacific Ocean with satellite tags. PLoS ONE, 6: e21087.
Spies, I., Spencer, P., and Punt, A. 2015. Where do we draw the line? A simulation
approach for evaluating management of marine fish stocks with isolation–by–
distance stock structure. Can. J. Fish. Aquat. Sci. 72: 968–982.
Sun, C., Chang, S., Chang, Y., Yeh, S., Chiang, W., and Su, N. 2011. Reproductive
biology of female striped marlin (Kajikia audax) in the waters off Taiwan.
International Scientific Committee for Tuna and Tuna-like Species in the North
Pacific Ocean.
Ueyanagi, S. 1974. On an additional diagnostic character for the identification of billfish
larvae with some notes on the variations in pigmentation. In Proceedings of the
132
International Billfish Symposium, Kailua-Kona, Hawaii, 9-12 August 1972. Part
2. Review and Contributed Papers. Edited by R. Shomura, and F. Williams.
National Marine Fisheries Service, Seattle, Washington, USA. NOAA Technical
Report NMFS SSRF-675.
Ueyanagi, S., and Wares, P. 1975. Synopsis of biological data on striped marlin,
Tetrapturus audax (Philippi, 1887). In Proceedings of the International Billfish
Symposium, Kailua-Kona, Hawaii, 9-12 August 1972. Part 3. Species Synopses.
Edited by R. Shomura, and F. Williams. NOAA Technical Report NMFS SSRF-
675.
Vähä, J., and Primmer, C. 2006. Efficiency of model–based Bayesian methods for
detecting hybrid individuals under different hybridization scenarios and with
different numbers of loci. Mol. Ecol. 15: 63–72.
Waples, R. 1998. Separating the wheat from the chaff: patterns of genetic differentiation
in high gene flow species. J. Hered. 89: 438–450.
Waples, R., and Gaggiotti, O. 2006. What is a population? An empirical evaluation of
some genetic methods for identifying the number of gene pools and their degree
of connectivity. Mol. Ecol. 15: 1419–1439.
Ward, R., Woodwark, M., and Skibinski, D. 1994. A comparison of genetic diversity
levels in marine, freshwater, and anadromous fishes. J. Fish Biol. 44: 213–232.
WCPFC. 2012. Stock assessment of striped marlin (Kajikia audax) in the southwest
Pacific Ocean. Kolonia, Pohnpei State, Federated States of Micronesia. Western
and Central Pacific Fisheries Commission.
133
WCPFC. 2015. Stock assessment update for striped marlin (Kajikia audax) in the western
and central North Pacific Ocean through 2013. Kolonia, Pohnpei State, Federated
States of Micronesia. Western and Central Pacific Fisheries Commission.
Weir, B. 1996. Genetic Data Analysis II: Methods for Discrete Population Genetic Data.
Sinauer Assoc., Inc., Sunderland, Massachusetts, USA.
Weir, B., and Cockerham, C. 1984. Estimating F–statistics for the analysis of population
structure. Evolution, 38: 1358–1370.
White, C., Selkoe, K., Watson, J., Siegel, D., Zacherl, D., and Toonen, R. 2010. Ocean
currents help explain population genetic structure. Proc. R. Soc. B. 277: 1685–
1694.
Wigginton, J., Cutler, D., and Abecasis, G. 2005. A note on exact tests of Hardy–
Weinberg equilibrium. Am. J. Hum. Genet. 7: 887–883.
Williams, S., Bennett, M., Pepperell, J., Morgan, J., and Ovenden, J. 2015. Spatial
genetic subdivision among populations of the highly migratory black marlin
Istiompax indica within the central Indo–Pacific. Mar. Freshwater Res. 166: 29–
32.
Wolanski, E., Ridd, P., and Inoue, M. 1988. Currents through Torres Strait. J. Phys.
Oceanogr. 18: 1535–1545.
134
Table 1. Details for striped marlin sample collections analyzed in this study.
No. Sampling Region Code Year Total Individuals
Indian Ocean South Africa SAF 2017 1 11 2016 3 2015 7
Kenya KEN 2016 13 27 2015 14
Western Australia WAUS 2016 8 8
Total: 46 Pacific Ocean Eastern Australia EAUS 2015 3 35 2012 3 2011 7 2010 6 1994 16
New Zealand NZ 2017 22 22
Japan JAP 2015 18 18
Taiwan TAI 2016 4 11 2015 5 2014 2
Hawaii HAW 2015 21 21
California CAL 2016 2 15 2000 13
Baja California BAJA 2015 21 22
Ecuador ECU 2016 22 37 1992 15
Peru PERU 2016 19 19 Total: 199 Grand Total: 245
135
Table 2. Number of single nucleotide polymorphism (SNP) loci retained after each filtering step.
Filter No. Retained Loci
Loci received from Diversity Arrays 61,908 Technology Pty. Ltd.
Quality Filter
Missing ≥ 10% genotypes 41,613
Average reproducibility < 95% 41,540
Monomorphic 11,831
More than one SNP per reduced 10,220 representation locus
Minor allele frequency < 0.05 4,016
Hardy-Weinberg Equilibrium
P < 0.006 in > 1 sample collection 3,975
Outlier Identification
Putatively neutral 3,916
Putatively under selection 59
136
Table 3. Results from analysis of molecular variance (AMOVA) for scenarios with striped marlin sample collections organized by sampling location or by region. Sample collections are labeled as in Table 1. Comparisons with sample collections grouped by region are identified as follows: WIO = western Indian Ocean sample collections SAF and KEN; Oceania = sample collections WAUS, EAUS, and NZ; NPO = North Pacific Ocean sample collections JAP, TAI, HAW, and CAL; NPO2 = North Pacific Ocean sample collections JAP2 and HAW2; ECPO = eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
Percent Grouping Source of Variation Φ p-value ST Variation Temporal Replicates EAUS 1994, EAUS Among samples 0.0041 0.41 0.000 2010-2015 Within samples 99.59
ECU 1992, ECU 2016 Among samples 0.0030 0.30 0.002 Within samples 99.70 Regions WIO, Oceania, NPO, Among regions 0.0300 3.00 0.000 ECPO Among populations within regions 0.0037 0.36 0.000 Within populations 0.0336 96.64 0.000
WIO, Oceania, NPO, Among regions 0.0336 3.36 0.000 NPO2, ECPO Among populations within regions 0.0029 0.28 0.000 Within populations 0.0363 96.37 0.000
WIO, WAUS, Among regions 0.0299 2.99 0.000 EAUS+NZ, NPO, ECPO Among populations within regions 0.0033 0.32 0.000 Within populations 0.0330 96.69 0.000
WIO, WAUS, Among regions 0.0335 3.35 0.000 EAUS+NZ, NPO, NPO2, Among populations within groups 0.0023 0.22 0.002 ECPO Within populations 0.0358 96.42 0.000
WIO+WAUS, Among regions 0.0288 2.88 0.000 EAUS+NZ, NPO, ECPO Among populations within regions 0.0046 0.45 0.000 Within populations 0.0332 96.68 0.000
WIO+Oceania, NPO, Among regions 0.0248 2.48 0.000 ECPO Among populations within regions 0.0103 1.00 0.000 Within populations 0.0349 96.51 0.000
137
Table 3. (continued)
Percent Grouping Source of Variation Φ p-value ST Variation WIO, Among regions 0.0298 2.98 0.017 Oceania+NPO+ECPO Among populations within regions 0.0196 1.90 0.000 Within populations 0.0488 95.12 0.000
WIO+WAUS, Among regions 0.0269 2.69 0.004 EAUS+NZ+NPO+ECPO Among populations within regions 0.0192 1.86 0.000 Within populations 0.0456 95.44 0.000
WIO+Oceania, Among regions 0.0236 2.36 0.001 NPO+ECPO Among populations within regions 0.0156 1.52 0.000 Within populations 0.0388 96.12 0.000
138
Table 4. Diversity metrics calculated for striped marlin sample collections organized by sampling location. Sample collections are labeled as in Table 1. Cells are colored as a heat map ranging from green (low diversity values) to red (high diversity values).
Sample Collection N aR HE HO SAF 11 1.260 0.145 0.138 KEN 27 1.269 0.148 0.142 WAUS 8 1.259 0.145 0.136 EAUS 35 1.290 0.160 0.172 NZ 22 1.273 0.151 0.143 JAP 18 1.294 0.163 0.181 TAI 11 1.272 0.152 0.143 HAW 21 1.318 0.177 0.214 CAL 15 1.280 0.154 0.147 BAJA 21 1.271 0.150 0.146 ECU 37 1.285 0.157 0.165 PERU 19 1.271 0.150 0.142 N = number of individuals comprising sample collection aR = rarefaction allelic richness HE = expected heterozygosity HO = observed heterozygosity
139
--
.000 .000 .000 .000 .000 .000 .000 .000 .000
0 0 0 0 0 0 0 0 0
PERU 0.033* 0.171*
to values)
ST
--
.000 .000 .000 .000 .000 .000 .000 .000 .000
ECU
0 0 0 0 0 0 0 0
0
0.299* 0.0019
0
--
.000 .000 .000 .000 .000 .000 .000 .000 .000
. Statistical significance significance Statistical . 0 0 0 0 0 0 0 0 0
BAJA
0.0006 0.003
2 0 0
--
.000 .000 .000 .000 .000
CAL
0 0 0 0 0 0.004 0.001
0.122* 0.021 0.018 0.020
1 1 2
--
.000 .000 .000 .000 .000
0 0 0 0 0 0.004
HAW
0.087* 0.009 0.024 0.0176 0.024
shown above diagonal shownabove
--
.000 .000 .000 .000 .000
TAI
0 0 0 0 0
0.036* 0.0067 0.0078 0.0207 0.0179 0.0229
statistically significant comparisons are denoted with with significant comparisons an are denoted statistically
-
4
--
.000 .000 .000 .000 .000
JAP
0 0 0 0 0
0.0003
0.004 0.0077 0.0225 0.0186 0.0229
-
4 3 4 0 1 4
Cells are colored as a heat ranging green from map heat are a as colored F (low Cells
. .
--
.000 .000
NZ
0 0 0.009
0.931* 0.026 0.027 0.0185 0.026 0.036 0.031 0.039
3 0 4 2 2 4
) calculated between striped marlin sample collections organized by sampling sampling by organized collections sample striped between marlin calculated )
Yekutieli Yekutieli Non (2001).
--
.000 .000
0 0 0.006
EAUS
0.001 0.024 0.025 0.0149 0.023 0.035 0.0296 0.037
= 0.010) corrected for multiple pairwise comparisons (n = 66) using the modified false modified pairwise for comparisons the using 66) = (n corrected 0.010) = multiple
and
crit
3 1
(p
--
.000 .000
0 0
0.0079 0.0078 0.0356 0.034 0.021 0.0378 0.0425 0.0355 0.0449
WAUS
4 2 0 3 2 3
values associated with each pairwise comparison are each are comparison with associated pairwise values
--
-
values (below diagonal (below values
KEN
critical value critical
0.432* 0.0199 0.0285 0.029 0.0565 0.0556 0.045 0.057 0.061 0.055 0.065
ST
cribed by Benjamini Benjamini by cribed
1 1 0 0 0 0
lues). P lues).
--
va
SAF
0.0027 0.027 0.027 0.0288 0.0537 0.0557 0.041 0.0577 0.062 0.054 0.065
ST
Pairwise F Pairwise
Table 5. Table 1 Table labeled in as are collections Sample location. F (high red wasassessed a using des rate discovery asterisk. SAF KEN WAUS EAUS NZ JAP TAI HAW CAL BAJA ECU PERU
140
Table 6. Diversity metrics calculated for striped marlin populations defined according to results from individual-based cluster analyses and analysis of molecular variance (AMOVA). Cells are colored as a heat map ranging from green (low diversity values) to red (high diversity values). WIO = western Indian Ocean, NPO = North Pacific Ocean, ECPO = eastern central Pacific Ocean.
Population Sample Collections N aR HE HO WIO SAF, KEN 38 1.463 0.147 0.144 Oceania WAUS, EAUS, NZ 65 1.488 0.156 0.162 NPO JAP, TAI, HAW, CAL 54 1.489 0.155 0.156 NPO2 JAP2, HAW2 11 1.501 0.204 0.304 ECPO BAJA, ECU, PERU 77 1.472 0.154 0.160 N = number of individuals comprising population
aR = rarefaction allelic richness HE = expected heterozygosity HO = observed heterozygosity
141
Table 7. Pairwise FST values (below diagonal) calculated between striped marlin populations defined according to results from individual-based cluster analyses and analysis of molecular variance (AMOVA). Cells are colored as a heat map ranging from green (low FST values) to red (high FST values). P-values associated with each pairwise comparison are also shown (above diagonal). Sample collections are grouped as follows: WIO = western Indian Ocean sample collections SAF and KEN; Oceania = sample collections WAUS, EAUS, and NZ; NPO = North Pacific Ocean sample collections JAP, TAI, HAW, and CAL; NPO2 = North Pacific Ocean sample collections JAP2 and HAW2; ECPO = eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
WIO Oceania NPO NPO2 ECPO WIO -- 0.000 0.000 0.000 0.000 Oceania 0.0261 -- 0.000 0.000 0.000 NPO 0.0497 0.0198 -- 0.000 0.000 NPO2 0.0836 0.0555 0.0394 -- 0.000 ECPO 0.0580 0.0330 0.0169 0.0556 --
142
Japan (JAP) California (CAL) n = 18 n = 15 Taiwan (TAI) Baja California (BAJA) n = 11 Hawaii (HAW) n = 21 n = 21
Ecuador (ECU) Kenya (KEN) n = 37 n = 27 Eastern Peru (PERU) n = 19 Western Australia (EAUS) Australia (WAUS) n = 35 n = 8 South Africa (SAF) n = 11 New Zealand (NZ) n = 22
Fig. 1. Map displaying geographic sampling locations and sample sizes for collections of striped marlin evaluated in this study. Points indicate representative sampling region.
143
temporal temporal
dimensional plots of results of dimensionalplots
-
. Percentage of total genetic variation explained by each axis is shown. Insets at top right top shown. is of explained axis at each by Insets genetic variation total Percentage .
dimensional plots of axes one and two resulting from principal coordinate from axes of analysis dimensional(PCoA) plots principal one of resulting two and
-
Two
scriminant analysis of principal components (DAPC) for EAUS (C) or ECU (D) temporal replicates. Insets at top left left top at Insets replicates. principal of (DAPC) components scriminant (D) ECU or (C) analysis temporal EAUS for
A and B: B: Aand
Fig. 2. Fig. EAUS for (A) (B) ECU or replicates Two D: PCoA, with and C bars). (black including associated the axes plotted showeigenvalues di from bars). (darkly with retained shaded axes DAPC, (PCA) component eigenvalues analysis including associated showprincipal
144
Fig. 3. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using the dataset with all striped marlin sample collections. Panels A through D correspond with scenarios for K equal to two through five, sequentially. Individuals are ordered identically in each panel. Open arrowheads indicate subset of samples corresponding with JAP2 (n = 5) and HAW2 (n = 6) which comprise a distinct population in the North Pacific Ocean. Filled arrowheads indicate putative migrants sampled off Hawaii (n = 4) and off Ecuador (n = 3). Individual sample collections are denoted at bottom of figure and regional populations are denoted at top: WIO = western Indian Ocean sample collections SAF and KEN; Oceania = sample collections WAUS, EAUS, and NZ; NPO = North Pacific Ocean sample collections JAP, TAI, HAW, and CAL; ECPO = eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
145
Eigenvalues 6 SAF KEN WAUS EAUS NZ
4 JAP TAI HAW CAL
) JAP2
%
7 2
. HAW2
1
(
2 BAJA C
P ECU
PERU
0
2 −
−4 −2 0 2
PC1 (3.4%)
Fig. 4. Two-dimensional plot of axes one and two resulting from principal coordinate analysis (PCoA) of striped marlin genotype data. Percentage of total genetic variation explained by each axis is shown. Sample collections are labeled as in Table 1. Each collection is represented by a unique color according to the legend at top left; similar colors are used to represent larger geographic regions. Inset at top left shows eigenvalues associated with the PCoA, including plotted axes (black bars).
146
Fig. 5. Two-dimensional density plot of results for K = 4 scenario assessed with discriminant analysis of principal components (DAPC). Inset at top left shows principal component analysis (PCA) eigenvalues associated with the DAPC (darkly shaded bars denote retained axes). Genetic clusters are numbered one through four and, with the exception of individuals identified as putative migrants, correspond with the following sample collections: 1) Oceania sample collections WAUS, EAUS, and NZ, 2) eastern central Pacific Ocean sample collections BAJA, ECU, and PERU, 3) North Pacific Ocean sample collections JAP, TAI, HAW, CAL as well as JAP2 and HAW2, 4) western Indian Ocean sample collections SAF and KEN.
147
Fig. 6. Two-dimensional density plot of results for K = 5 scenario assessed with discriminant analysis of principal components (DAPC). Inset at top left shows principal component analysis (PCA) eigenvalues associated with the DAPC (darkly shaded bars denote retained axes). Genetic clusters are numbered one through four and, with the exception of individuals identified as putative migrants, correspond with the following sample collections: 1) western Indian Ocean sample collections SAF and KEN, 2) North Pacific Ocean sample collections JAP2 and HAW2, 3) Oceania sample collections WAUS, EAUS, and NZ, 4) North Pacific Ocean sample collections JAP, TAI, HAW, and CAL, 5) eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
148
Fig. 7. Parameter estimates for mutation-scaled effective population sizes (횹; inside circles) and mutation-scaled effective migration rates (M; along arrows) derived from Bayesian-based coalescent simulations performed using MIGRATE-N. Standard errors for each parameter estimate are shown in parentheses. Weight of arrows is scaled according to migration rate. Circles represent populations and are labeled as follows: WIO = western Indian Ocean, NPO = North Pacific Ocean, and ECPO = eastern central Pacific Ocean.
149
Fig. 8. World map displaying spatial distribution of striped marlin (Kajikia audax; dark blue) overlaid with jurisdictional regions for the Indian Ocean Tuna Commission (IOTC; green), Western and Central Pacific Fisheries Commission (WCPFC; pink), and Inter- American Tropical Tuna Commission (IATTC; light blue). Points correspond with sampling locations for collections of striped marlin evaluated in the present study, and are colored according to genetically distinct population. Currently, the IOTC recognizes a single stock of striped marlin in the Indian Ocean, the WCPFC recognizes distinct stocks in the western and central North Pacific and in the western South Pacific oceans, and the IATTC recognizes a single stock in the eastern Pacific Ocean. WIO = western Indian Ocean, NPO = North Pacific Ocean, ECPO = eastern central Pacific Ocean.
150
SUPPLEMENTARY MATERIALS
Table S1. Diversity metrics calculated for striped marlin populations defined according to results from individual-based cluster analyses and analysis of molecular variance (AMOVA), except WAUS is grouped separately. Cells are colored as a heat map ranging from green (low diversity values) to red (high diversity values). WIO = western Indian Ocean, NPO = North Pacific Ocean, ECPO = eastern central Pacific Ocean.
Population Sample Collections N aR HE HO WIO SAF, KEN 38 1.332 0.147 0.144 WAUS 8 1.317 0.145 0.136 EAUS, NZ 57 1.351 0.156 0.162 NPO JAP, TAI, HAW, CAL 54 1.350 0.155 0.156 NPO2 JAP2, HAW2 11 1.418 0.204 0.304 ECPO BAJA, ECU, PERU 77 1.345 0.154 0.160 N = number of individuals comprising grouped collections aR = rarefaction allelic richness HE = expected heterozygosity HO = observed heterozygosity
151
Table S2. Pairwise FST values (below diagonal) calculated between striped marlin populations defined according to results from individual-based cluster analyses and analysis of molecular variance (AMOVA), except WAUS is grouped separately. Cells are colored as a heat map ranging from green (low FST values) to red (high FST values). P- values associated with each pairwise comparison are shown above diagonal. Statistical significance was assessed using a critical value (pcrit = 0.015) corrected for multiple pairwise comparisons (n = 15) using the modified false discovery rate described by Benjamini and Yekutieli (2001). Sample collections are grouped as follows: WIO = western Indian Ocean sample collections SAF and KEN; NPO = North Pacific Ocean sample collections JAP, TAI, HAW, and CAL; NPO2 = North Pacific Ocean sample collections JAP2 and HAW2; ECPO = eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
WIO WAUS EAUS+NZ NPO NPO2 ECPO WIO -- 0.000 0.000 0.000 0.000 0.000 WAUS 0.0208 -- 0.006 0.000 0.000 0.000 EAUS+NZ 0.0279 0.0069 -- 0.000 0.000 0.000 NPO 0.0497 0.0291 0.0195 -- 0.000 0.000 NPO2 0.0836 0.0512 0.0541 0.0394 -- 0.000 ECPO 0.0580 0.0380 0.0331 0.0169 0.0556 --
152
Fig. S1. Schematic displaying the processing of raw Illumina sequencing reads to identify the single nucleotide polymorphisms (SNPs) evaluated in this study. DArTtoolbox refers to the proprietary bioinformatic pipeline employed by Diversity Arrays Technology Pty. Ltd. (DArT PL). Genotypes received from DArT PL were further processed using the dartR package (Gruber et al. 2018) for the R statistical computing environment.
153
Fig. S2. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using the dataset limited to striped marlin sample collections from the Pacific Ocean and western Australia. Panels A through C correspond with scenarios for K equal to two through four, sequentially. Individuals are ordered identically in each panel. Open arrowheads indicate subset of samples corresponding with JAP2 (n = 5) and HAW2 (n = 6) which comprise a distinct population in the North Pacific Ocean. Filled arrowheads indicate putative migrants sampled off Hawaii (n = 4) and off Ecuador (n = 3). Individual sample collections are denoted at bottom of figure and regional populations are denoted at top: Oceania = sample collections WAUS, EAUS, and NZ; NPO = North Pacific Ocean sample collections JAP, TAI, HAW, and CAL; ECPO = eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
154
Fig. S3. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using the dataset comprising striped marlin sample collections limited to the Indian Ocean, eastern Australia, and New Zealand. Panels A through C correspond with scenarios for K equal to two through four, sequentially. Individuals are ordered identically in each panel. Individual sample collections are denoted at bottom of figure and regional populations are denoted at top: WIO = western Indian Ocean sample collections SAF and KEN; Oceania = sample collections WAUS, EAUS, and NZ.
155
Eigenvalues SAF KEN WAUS EAUS NZ
0 JAP 1 TAI HAW CAL JAP2
) HAW2
% 3
. BAJA
1
5 (
ECU 3
C PERU
P 0
−4 −2 0 2
PC1 (3.4%)
Fig. S4. Two-dimensional plot of axes one and three resulting from principal coordinate analysis (PCoA) of striped marlin genotype data. Percentage of total genetic variation explained by each axis is shown. Sample collections are labeled as in Table 1. Each collection is represented by a unique color according to the legend at top right; similar colors are used to represent larger geographic regions. Inset at top right shows eigenvalues associated with the PCoA, including plotted axes (black bars).
156
SAF Eigenvalues KEN WAUS EAUS NZ
0 JAP 1 TAI HAW CAL JAP2
) HAW2
% 3
. BAJA
1
5 (
ECU 3
C PERU
P 0
−2 0 2 4 6
PC2 (1.7%)
Fig. S5. Two-dimensional plot of axes two and three resulting from principal coordinate analysis (PCoA) of striped marlin genotype data. Percentage of total genetic variation explained by each axis is shown. Sample collections are labeled as in Table 1. Each collection is represented by a unique color according to the legend at top right; similar colors are used to represent larger geographic regions. Inset at top right shows eigenvalues associated with the PCoA, including plotted axes (black bars).
157
APPENDIX
Fig. A1. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using the dataset with neutral markers and markers putatively under the influence of natural selection, and including all striped marlin sample collections. Panels A through D correspond with scenarios for K equal to two through five, sequentially. Individuals are ordered identically in each panel. Open arrowheads indicate subset of samples corresponding with JAP2 (n = 5) and HAW2 (n = 6) which comprise a distinct population in the North Pacific Ocean. Filled arrowheads indicate putative migrants sampled off Hawaii (n = 4) and off Ecuador (n = 3). Individual sample collections are denoted at bottom of figure and regional populations are denoted at top: WIO = western Indian Ocean sample collections SAF and KEN; Oceania = sample collections WAUS, EAUS, and NZ; NPO = North Pacific Ocean sample collections JAP, TAI, HAW, and CAL; ECPO = eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
158
Fig. A2. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using the dataset with neutral markers and markers putatively under the influence of natural selection, and limited to striped marlin sample collections from the Pacific Ocean and western Australia. Panels A through C correspond with scenarios for K equal to two through four, sequentially. Individuals are ordered identically in each panel. Open arrowheads indicate subset of samples corresponding with JAP2 (n = 5) and HAW2 (n = 6) which comprise a distinct population in the North Pacific Ocean. Filled arrowheads indicate putative migrants sampled off Hawaii (n = 4) and off Ecuador (n = 3). Individual sample collections are denoted at bottom of figure and regional populations are denoted at top: Oceania = sample collections WAUS, EAUS, and NZ; NPO = North Pacific Ocean sample collections JAP, TAI, HAW, and CAL; ECPO = eastern central Pacific Ocean sample collections BAJA, ECU, and PERU.
159
Fig. A3. Barplots displaying individual admixture proportions inferred from STRUCTURE analyses performed using the dataset with neutral markers and markers putatively under the influence of natural selection, and limited to striped marlin sample collections limited to the Indian Ocean, eastern Australia, and New Zealand. Panels A through C correspond with scenarios for K equal to two through four, sequentially. Individuals are ordered identically in each panel. Individual sample collections are denoted at bottom of figure and regional populations are denoted at top: WIO = western Indian Ocean sample collections SAF and KEN; Oceania = sample collections WAUS, EAUS, and NZ.
160
Fig. A4. Two-dimensional density plot of results for K = 4 scenario assessed with discriminant analysis of principal components (DAPC) and dataset inclusive of neutral markers as well as markers putatively under the influence of natural selection. Inset at top left shows principal component analysis (PCA) eigenvalues associated with the DAPC (darkly shaded bars denote retained axes). Genetic clusters are numbered one through four and, with the exception of individuals identified as putative migrants, correspond with the following sample collections: 1) Oceania sample collections WAUS, EAUS, and NZ, 2) eastern central Pacific Ocean sample collections BAJA, ECU, and PERU, 3) western Indian Ocean sample collections SAF and KEN, 4) North Pacific Ocean sample collections JAP, TAI, HAW, CAL as well as JAP2 and HAW2.
161
Fig. A5. Two-dimensional density plot of results for K = 5 scenario assessed with discriminant analysis of principal components (DAPC) and dataset inclusive of neutral markers as well as markers putatively under the influence of natural selection. Inset at top left shows principal component analysis (PCA) eigenvalues associated with the DAPC (darkly shaded bars denote retained axes). Genetic clusters are numbered one through four and, with the exception of individuals identified as putative migrants, correspond with the following sample collections: 1) Oceania sample collections WAUS, EAUS, and NZ, 2) western Indian Ocean sample collections SAF and KEN, 3) eastern central Pacific Ocean sample collections BAJA, ECU, and PERU, 4) North Pacific Ocean sample collections JAP, TAI, HAW, and CAL, 5) North Pacific Ocean sample collections JAP2 and HAW2.
162
Eigenvalues SAF
KEN 4 WAUS EAUS NZ JAP TAI
2 HAW CAL
) JAP2
% 0
. HAW2
2
(
2 BAJA C
P ECU 0
PERU
2 −
−4 −2 0 2
PC1 (4.6%)
Fig. A6. Two-dimensional plot of axes one and two resulting from principal coordinate analysis (PCoA) using dataset inclusive of neutral markers as well as markers putatively under the influence of natural selection. Percentage of total genetic variation explained by each axis is shown. Sample collections are labeled as in Table 1. Each collection is represented by a unique color according to the legend at top left; similar colors are used to represent larger geographic regions. Inset at top left shows eigenvalues associated with the PCoA, including plotted axes (black bars).
163
SAF Eigenvalues KEN WAUS EAUS NZ
0 JAP 1 TAI HAW CAL JAP2
) HAW2
% 3
. BAJA
1
5 (
ECU 3
C PERU
P 0
−4 −2 0 2
PC1 (4.6%) Fig. A7. Two-dimensional plot of axes one and three resulting from principal coordinate analysis (PCoA) using dataset inclusive of neutral markers as well as markers putatively under the influence of natural selection. Percentage of total genetic variation explained by each axis is shown. Sample collections are labeled as in Table 1. Each collection is represented by a unique color according to the legend at top left; similar colors are used to represent larger geographic regions. Inset at top left shows eigenvalues associated with the PCoA, including plotted axes (black bars).
164
SAF Eigenvalues KEN WAUS EAUS NZ
0 JAP 1 TAI HAW CAL JAP2
) HAW2
% 3
. BAJA
1
5 (
ECU 3
C PERU
P 0
−2 0 2 4
PC2 (2.0%)
Fig. A8. Two-dimensional plot of axes two and three resulting from principal coordinate analysis (PCoA) using dataset inclusive of neutral markers as well as markers putatively under the influence of natural selection. Percentage of total genetic variation explained by each axis is shown. Sample collections are labeled as in Table 1. Each collection is represented by a unique color according to the legend at top left; similar colors are used to represent larger geographic regions. Inset at top left shows eigenvalues associated with the PCoA, including plotted axes (black bars).
165
CHAPTER IV
Use of Genome-wide Molecular Markers for Evaluating the Evolutionary Relationship of
Two Highly Migratory Sister Species Inhabiting Distinct Ocean Basins
166
ABSTRACT
Species delimitation is necessary for identifying and protecting genetic diversity characteristic of separately evolving lineages, but is challenging for closely related species where a high level of statistical power is required to detect low levels of divergence. In this study, we employ next-generation sequencing and a range of population genomic and species delimitation methods for evaluating the evolutionary relationship of two sister species of highly migratory pelagic fishes, striped marlin
(Kajikia audax) and white marlin (K. albida). Previous genetic assessments based on nuclear and mitochondrial DNA have been unsuccessful in resolving striped marlin and white marlin as distinct evolutionary lineages. We surveyed over 12,000 single nucleotide polymorphisms across samples of striped marlin (n = 250) and white marlin (n
= 75) collected from locations throughout the distributional ranges of both species.
Individual-based cluster analyses identified a small number of individuals (n = 4) that displayed high levels of shared ancestry, but consistently resolved striped marlin and white marlin as separate clusters. Bayes factor delimitation identified a species tree comprising distinct lineages for striped marlin and white marlin as highly favorable
(Bayes factor = 23,161) relative to a tree with a single lineage for these species. Genetic differentiation between striped marlin and white marlin (DS = 0.0431, FST = 0.4802) was considerably higher, and in many instances an order of magnitude greater, than intraspecific comparisons among striped marlin populations (DS = 0.0013–0.0037, FST =
0.0217–0.0703). We identified 20 SNPs exhibiting fixed differences between species.
Collectively, our results suggest that striped marlin and white marlin represent distinct evolutionary lineages for which the current taxonomic status of these species is valid.
167
INTRODUCTION
Evaluating evolutionary relationships among species not only improves our understanding of global biodiversity, but also enables the identification of distinct evolutionary lineages (e.g. separately evolving ancestor-descendant series; de Queiroz
2007). Such information is a necessary first step for identifying and protecting genetic diversity characteristic of species, and for facilitating the study of processes relevant to the divergence of populations and species. However, delimiting evolutionary lineages for closely related species is challenging because the level of divergence between lineages is expected to be low and therefore more difficult to detect. This challenge is especially relevant to many pelagic marine species, where large effective population sizes and high dispersal capabilities across broad geographic regions presumably result in the slow accumulation of genetic differences between lineages (Mayr 1954, Martin et al. 1992; but see Palumbi 1992). Traditionally, the ability to resolve low levels of genetic divergence has been limited by technological and economic constraints on the number of genomic regions that can practically be compared between species. Recently, technological advancements corresponding with the development of next-generation sequencing (NGS; e.g. Mardis 2008) have made possible the rapid and cost-effective comparison of extensive genomic regions across large numbers of samples. This advancement facilitates new opportunities for the delimitation of evolutionary lineages (Bickford et al. 2006,
Lemmon & Lemmon 2013, Narum et al. 2013), particularly for highly migratory pelagic species exhibiting low degrees of divergence.
White marlin (Kajikia albida) and striped marlin (K. audax) are istiophorid billfishes (marlins, spearfishes, and sailfish) which comprise sister species distributed in
168
the Atlantic and Indo-Pacific oceans, respectively. These pelagic fishes display broad and continuous spatial distributions in temperate, sub-tropical, and tropical waters (Nakamura
1985), and are capable of long distance movements thought to correspond with spawning and feeding behaviors (Ortiz et al. 2003). White marlin and striped marlin are the most temperate of the istiophorid billfishes, and regularly occur in waters with temperatures of
20–26 ˚C (striped marlin; Ortega-Garcia et al. 2003, Sippel et al. 2007) or 24–28 ˚C
(white marlin; Schlenker 2014, Hoolihan et al. 2015). Though spawning activity has been confirmed in multiple geographically distant regions for both species (reviewed in
Bromhead et al. 2003, White Marlin Biological Review Team 2007), population structuring has been observed only for striped marlin (Chapter III; Graves & McDowell
1994, McDowell & Graves 2008, Purcell & Edmands 2011), and molecular results are consistent with the presence of a single ocean-wide population for white marlin (Chapter
II).
The ability of istiophorid billfishes to travel long distances suggests that these species may be capable of inter-oceanic movements. Such movements are more likely to occur around the Cape of Good Hope (South Africa) given the lower latitude and warmer water temperatures characterizing this region relative to Cape Horn (South America).
Movement between the Atlantic and Indian oceans has been reported for other istiophorid billfishes, including blue marlin (Makaira nigricans; Ortiz et al. 2003) and shortbill spearfish (Tetrapturus angustirostris; McDowell et al. 2018). Relative to other istiophorids, inter-oceanic movements may be more likely for white marlin and striped marlin given the more temperate thermal preferences of these species. Accordingly,
Talbot & Penrith (1962) and Penrith & Cram (1974) collectively report the capture of 13
169
striped marlin from the eastern South Atlantic Ocean off Cape Town, South Africa based on the morphological examination of vouchered museum specimens. A number of white marlin have also been reported from pelagic longline fisheries operating in the Indian
Ocean off the southern and southeastern coasts of Africa (Wendy West, Department of
Agriculture, Forestry and Fisheries, South Africa, unpublished data); however, these reports are based on field identifications of unvouchered specimens. Regardless, reports such as these suggest that overlapping spatial distributions for striped marlin and white marlin may occur off the coast of South Africa in some years. Determining whether such spatial overlap corresponds with gene flow is important for understanding the evolutionary relationship of striped marlin and white marlin, and may provide key insights to the evolution of species pairs distributed across the Atlantic and Indo-Pacific.
Though currently recognized as distinct species, morphological similarities between white marlin and striped marlin have contributed to complicated taxonomic histories for these species. White marlin was originally described as Tetrapturus albidus (Poey 1860), but a number of synonyms have been described in the years since. Several synonyms have also been described for striped marlin, which was originally described as Histiophorus audax (Philippi 1887) but eventually placed in the genus Tetrapturus (Robins & de Sylva 1961). More recently, Collette et al. (2006) identified striped marlin and white marlin as genetically distant enough from other species comprising Tetrapturus to warrant reclassification to a previously described genus, Kajikia Hirasaka & Nakamura (1947). Morphological comparisons of striped marlin and white marlin note differences in fin morphology corresponding with the anterior lobe of the first dorsal fin and tips of the pectoral and first anal fins as
170
distinguishing features of these species (Nakamura 1985). However, in practice, the subtlety of these characters has resulted in species identifications primarily based on geographic location of capture, wherein Atlantic fish are typically identified as white marlin and Indo-Pacific fish as striped marlin. Though geographic location of capture may be suitable for distinguishing striped marlin and white marlin in some regions, this character is unreliable for species identifications in regions where the spatial distributions of these species overlap, such as off South Africa.
Results from genetic studies of striped marlin and white marlin have also contributed uncertainty to the taxonomic relationship of these species. Phylogenetic assessments of the istiophorid billfishes based on a variety of mitochondrial (mt) and nuclear markers report a lack of fixed differences between striped marlin and white marlin (Finnerty & Block 1995, Graves & McDowell 1995, Collette et al. 2006, Hanner et al. 2011). These studies have also been unsuccessful in unambiguously resolving striped marlin and white marlin as reciprocally monophyletic lineages. Such results are consistent with low levels of divergence between striped marlin and white marlin, which are also less than those reported for other istiophorids with spatial distributions spanning the Atlantic and Indo-Pacific. Based on restriction fragment length polymorphism analysis of whole mtDNA, Graves & McDowell (1995) observed corrected mean nucleotide sequence divergences of 0.15% and 0.45% between Atlantic and Indo-Pacific collections of blue marlin and sailfish (Istiophorus platypterus), respectively. In comparison, the level of divergence observed between white marlin and striped marlin was 0.06% (Graves & McDowell 1995). Despite molecular observations such as these, the distinct taxonomic status of white marlin and striped marlin has persisted, while
171
separate Atlantic and Indo-Pacific species previously recognized for blue marlin and sailfish have been reclassified as single, globally distributed species (Collette et al. 2006).
Collectively, the molecular and morphological factors discussed here question the validity of the taxonomic status of striped marlin and white marlin, and demonstrate the need for further studies characterized by high degrees of statistical power.
In this study, we use NGS to facilitate a statistically powerful evaluation of the evolutionary relationship of white marlin and striped marlin based on genome-wide molecular markers. We employ a range of population genomic and species delimitation methods to evaluate scenarios for which white marlin and striped marlin are regarded as distinct species or as sub-populations of a single species. Our objectives were to: 1) apply
NGS to identify novel genome-wide single nucleotide polymorphism (SNP) molecular markers for white marlin and for striped marlin; 2) investigate the presence of distinct evolutionary lineages corresponding with white marlin and striped marlin; and 3) genetically characterize resolved lineages.
MATERIALS AND METHODS
Sample collection and DNA preparation
Samples analyzed in this study were opportunistically collected from a variety of sources throughout the ranges of white marlin and striped marlin during the period 1992 to 2017
(Table 1). White marlin and striped marlin fin tissues were sampled prior to the live release of fish captured by recreational anglers, and from fish caught as bycatch on commercial pelagic longline vessels. Additional tissue samples were obtained from striped marlin and white marlin available in local markets. Two collections of white
172
marlin larvae acquired through fisheries-independent surveys were also analyzed.
Tentative species identifications for larvae were performed by evaluating diagnostic morphological characters and by sequencing a diagnostic segment of the mitochondrial
DNA control region following the methodology described in Chapter II. Tissue samples and whole larvae were preserved in 95% ethanol or a 10% dimethyl sulfoxide solution
(Seutin et al. 1991) and maintained at room temperature until DNA isolation.
Total genomic DNA was isolated from tissues and larvae using a DNeasy Blood
& Tissue Kit (Qiagen) or a ZR-96 Genomic DNA Kit (Zymo Research). DNA isolations were visualized on 5% agarose gels that included a standard DNA ladder. Isolations that recovered high molecular weight DNA were quantified using a Qubit 2 fluorometer and dsDNA BR assays (Thermo Fisher Scientific). Isolations with sufficient DNA for NGS analysis were normalized to 700 ng total DNA at 50 ng per uL and stabilized in
GenTegra-DNA (GenTegra LLC). Stabilized high quality DNA isolations were submitted to Diversity Arrays Technology Pty. Ltd. (DArT PL; Canberra, Australia) for DArTseqTM
1.0 genotyping.
DArTseqTM 1.0 genotyping
DArTseqTM genotyping (e.g. Sansaloni et al. 2011) consists of a genomic complexity reduction step followed by NGS; this methodology is similar to other methodologies commonly utilized for NGS of genomic complexity reductions (e.g. Peterson et al. 2012).
DArTseqTM library preparation was performed as in Chapter III. Briefly, this involved double restriction enzyme (RE) digestion of DNA isolations followed by ligation of fragments with RE-compatible adapters containing sample-specific barcodes. Fragments
173
containing both RE overhangs were preferentially amplified in PCR reactions, and samples displaying successful PCR amplification were normalized and pooled at equimolar ratios into multiplex libraries. Libraries were then used to perform single-end sequencing for 77 cycles on an Illumina HiSeq 2500 platform (Illumina, Inc.) at the
DArT PL facility.
Bioinformatic analyses
DArTseqTM genotype calling was performed as in Chapter III. Briefly, FASTQ output files generated from raw Illumina reads were analyzed using a proprietary DArTseqTM analytical software pipeline for quality filtering, variant calling, and generation of final genotypes. Quality filtering consisted of removing reads with Q < 25 for at least 50% of bases. Reads containing barcodes with Q < 30 were also removed. Remaining reads were demultiplexed according to sample-specific barcodes, then queried against NCBI
GenBank and a proprietary DArT PL database to identify and remove reads associated with viral or bacterial contamination. Variant calling consisted of creating a catalog of reduced representation loci (RRL) by first aligning identical reads within and among sequenced individuals to form read clusters, then matching read clusters based on degree of similarity to produce RRL. Polymorphic positions within RRL were identified as SNP variants, and major and alternate alleles for each variant were identified. DArT PL supplied a final matrix of genotypes comprising 61,908 SNP loci and metadata for each locus.
We performed additional filtering of SNP loci supplied by DArT PL using R v3.3.1 (R Core Team 2017) and the dartR v0.93 package (Gruber et al. 2018). Loci
174
missing ≥ 25% of genotype calls followed by samples missing ≥ 20% of genotype calls were excluded from the dataset. Loci with average reproducibility < 0.95 were removed in order to retain only high quality SNP variants. To reduce the probability of linked loci in the final dataset, we retained only a single SNP locus from RRL where more than one
SNP variant was identified. In such instances, the locus corresponding with the highest reproducibility and polymorphism information content, in that order, was retained.
Finally, all monomorphic loci were excluded from the final dataset.
Lineage delimitation
Because any given approach for species delimitation encompasses only a portion of the parameter space potentially relevant to the identification of distinct evolutionary lineages
(Carstens et al. 2013), we employed a variety of approaches for delimiting species from our final SNP dataset. These approaches included methods for determining the organization of samples according to lineage, followed by methods for validating these lineages. Confirmed lineages were then characterized by evaluating their level of genetic differentiation, genetic diversity, and phylogenetic relationship. Genetic differentiation and diversity calculations were also performed at the population level in order to facilitate comparisons between white marlin and previously confirmed populations of striped marlin (Chapter III), except populations resolved in the North Pacific Ocean were combined into a single population to increase the sample size for this region.
To organize samples according to lineage, we used individual-based cluster analyses consisting of multivariate methods and Bayesian inference. We also used network analysis and tree estimation to visualize pairwise differences between
175
individuals. Multivariate methods included principal coordinate analysis (PCoA) and discriminant analysis of principal components (DAPC; Jombart et al. 2009, 2010). We used dartR to perform PCoA based on a Euclidean distance matrix (Gower 1966) generated from SNP genotype data. The R package adegenet v2.0.1 (Jombart 2008) was used to identify clusters of genetically similar individuals based on successive K-means clustering, and to describe clusters using DAPC. We evaluated multiple sequential values for K and selected the most likely K by comparing Bayesian information criterion (BIC) scores associated with each K scenario. DAPC was then used to estimate posterior probabilities of assignment of individuals to each cluster, and to assess the genetic distinctiveness of clusters. Bayesian model-based simulations were performed using the variational algorithm implemented in the program fastSTRUCTURE v1.0 (Raj et al. 2014).
The simple allele frequencies prior was used for all fastSTRUCTURE simulations.
Individual admixture proportions estimated by fastSTRUCTURE were visualized using
DISTRUCT v1.1 (Rosenberg 2004). Because fastSTRUCTURE assumes loci conform to the expectations of Hardy-Weinberg equilibrium (HWE), SNPs that failed to meet these expectations were excluded from fastSTRUCTURE analyses. Conformance to HWE was evaluated using the exact methodological approach described by Wigginton et al. (2005), and statistical significance was assessed using a p-value corrected for multiple pairwise comparisons with a modified false discovery rate (Benjamini & Yekutieli 2001, Narum
2006). HWE testing was performed within groups of samples organized by species, except striped marlin samples were organized to account for previously determined population structure (Chapter III). Finally, a matrix of pairwise differences between individuals was generated using the R package ape v5.0 (Paradis et al. 2004). Pairwise
176
differences were visualized using the mutual k-nearest neighbor graph algorithm implemented in the R package NetView v1.0 (Neuditschko et al. 2012, Steinig et al.
2016). We used the cover_tree algorithm for performing the mutual nearest neighbor search, and the Infomap algorithm for community detection in NetView. Pairwise differences were also visualized by calculating an unrooted neighbor-joining tree (Saitou
& Nei 1987) in ape. Results from the visualization of pairwise differences were considered together with results from multivariate and fastSTRUCTURE analyses to organize samples into groups representative of putative lineages.
To confirm the lineages inferred using individual-based cluster analyses and calculation of pairwise differences, we employed the Bayes factor species delimitation method implemented in BFD* (Grummer et al. 2014, Leaché et al. 2014). Briefly, BFD* facilitates the comparison of marginal likelihoods calculated for competing species delimitation scenarios to determine the most likely species tree. We estimated species trees directly from SNP data using the program SNAPP v1.3.0 (Bryant et al. 2012) implemented in BEAST2 v2.4.7 (Bouckaert et al. 2014). Path sampling for marginal likelihood calculation was performed using 24 steps and a burn-in of 10,000 followed by
100,000 Markov Chain Monte Carlo generations. Tracer v1.7 (Rambaut et al. 2018) was used to assess model convergence. We compared marginal likelihoods for two species trees based on scenarios with white marlin and striped marlin comprising one or two distinct lineages. For computational tractability, species tree estimation was performed using a reduced dataset composed of 15 striped marlin and 15 white marlin. To create this reduced dataset, one to three individuals were randomly selected from each sampling
177
location for each species. This reduced dataset also included SNP genotype data for a single blue marlin, which served as an outgroup for species tree estimation.
Lineage characterization
Genetic characterization of the lineages confirmed through Bayes factor delimitation was performed by computing genetic distances and evaluating fixed differences between lineages, calculating genetic diversity within lineages, and by performing lineage-based assignment testing. These calculations were also performed at the population level to assess the relationship of white marlin to previously confirmed populations of striped marlin (Chapter III). For population-level calculations, samples of white marlin were retained as a single population given an apparent lack of genetic population structure for this species (Chapter II). Samples of striped marlin were subdivided to reflect populations in the western Indian Ocean, Oceania, North Pacific Ocean, and eastern central Pacific
Ocean (Table 1). The R package hierfstat v0.04-22 (Goudet 2005) was used to calculate standard genetic distance (DS, Nei 1972), minimum genetic distance (DM, Nei 1973), and chord distance (DC, Cavalli-Sforza & Edwards 1967) between lineages and among populations. FST (Wright 1951, Weir & Cockerham 1984) was also calculated between lineages and among populations. The presence of fixed differences between lineages or among populations was evaluated using dartR. Observed and expected levels of heterozygosity were calculated for each lineage or population using the R packages dartR and poppr v2.6.1 (Kamvar et al. 2014), respectively. Mean allelic richness was also calculated for each lineage or population using the R package PopGenReport v3.0.0
(Adamack & Gruber 2014). Finally, we used the Bayesian model-based clustering framework implemented in newhybrids v1.1b3 (Anderson & Thompson 2002) to
178
calculate the posterior probability that each individual belongs to one of three genotype categories: pure striped marlin, pure white marlin, or F1 striped marlin x white marlin hybrid. These genotype categories were selected to allow for the identification of individuals displaying genomic signatures representative of both parental species.
Simulations in newhybrids were executed using a burn-in of 100,000 followed by
1,000,000 Markov Chain Monte Carlo generations.
Phylogenetic inference
Phylogenetic relationships among the individuals analyzed in this study were estimated using maximum likelihood inference as implemented in RAxML v8 (Stamatakis 2014).
To achieve more accurate estimates of branch lengths (Leaché et al. 2015), RAxML analyses were performed using a dataset comprising concatenated RRL sequences of
SNP loci. Heterozygous genotypes were represented by standard IUPAC codes. We used the GTR+gamma model of evolution for RAxML analyses, as recommended by the authors, and assessed support for each node with 1,000 rapid bootstrap replicates
(Stamatakis et al. 2008). Concatenated RRL sequences for a single blue marlin were used as an outgroup to root the maximum likelihood phylogeny.
RESULTS
Bioinformatic analyses
Of the 61,908 SNPs supplied by DArT PL, 12,762 SNPs remained after filtering to remove loci with ≥ 25% missing genotype calls, average reproducibility < 0.95, monomorphic loci, and after retaining only a single locus for RRL associated with more
179
than one SNP (Table 2). In addition, eight samples missing ≥ 20% of genotype calls were removed. Collectively, these filtering steps resulted in a final dataset comprising 250 striped marlin and 75 white marlin (Table 1) genotyped across 12,762 SNPs (Table 2).
Lineage delimitation
We employed a range of approaches for evaluating the number of distinct lineages represented by our SNP genotype data, and to genetically characterize and infer the phylogenetic relationship of confirmed lineages. We first used individual-based cluster analyses as well as visualization of pairwise differences to determine the organization of samples according to lineage. Results from PCoA indicated that the majority of variation in our genotype data was explained by principal component axis one (34.8%), with each subsequent axis explaining ≤ 1.9% of variation. A two-dimensional plot of PCoA axes one and two (Fig. 1) resolved individuals identified as white marlin and individuals identified as striped marlin as belonging to discrete clusters. Four white marlin (one sampled off Brazil and three sampled off the United States mid-Atlantic) were positioned as outliers adjacent to the cluster comprising all other white marlin. Within striped marlin, multiple clusters of individuals consistent with previously resolved populations were apparent (Fig. S1). White marlin and striped marlin were also resolved as distinct clusters on PCoA axes one and three, and axes two and three (Figs. S2, S3). On those axes, a single striped marlin (sampled off New Zealand) was positioned as an outlier adjacent to clusters comprising all other striped marlin. Additional white marlin outliers were also resolved on PCoA axes one and three, as well as axes two and three; these consisted of two individuals sampled off Brazil and one individual sampled off Angola.
180
K-means clustering performed prior to DAPC produced BIC scores that were similar across a range of values for K (BIC = 1556.16–1561.26 for K = 2–6); however, scenarios with K > 2 corresponded with varying degrees of individual admixture within striped marlin individuals. For the scenario with K = 2, all individuals identified as striped marlin or as white marlin assigned to distinct genetic clusters; these clusters were clearly differentiated in a two-dimensional plot of DAPC results (Fig. 2). Posterior probabilities of assignment to the cluster to which an individual was assigned were equal to 1.00
(standard deviation = 0.00) for all individuals in the K = 2 scenario.
Individual-based clustering assessed through fastSTRUCTURE simulations also confirmed the presence of two distinct groups corresponding with individuals identified as striped marlin or as white marlin. We performed fastSTRUCTURE simulations using a dataset for which 195 SNPs that did not conform to the expectations of HWE were removed. In results for K = 2 (Fig. 3), all striped marlin were resolved as belonging to a cluster distinct from individuals identified as white marlin. Individuals comprising these clusters were characterized by little to no admixture, except for a small number of individuals (white marlin: n = 4; striped marlin: n = 1) which displayed levels of admixture ranging from 13.79% to 35.73%. For white marlin, these individuals corresponded with the within-species outliers resolved across all PCoA scenarios discussed above (one white marlin sampled off Brazil and three white marlin sampled off the United States mid-Atlantic). The striped marlin displaying a high level of admixture also corresponded with the single within-species outlier resolved in PCoA analyses
(sampled off New Zealand). Results from fastSTRUCTURE simulations for scenarios with
K > 2 resolved only two clusters of individuals corresponding with striped marlin and
181
white marlin. Individuals of both species were characterized by little to no admixture, except for the four white marlin and one striped marlin described above.
Pairwise distances calculated between individuals were visualized using network analysis and by generating an unrooted neighbor-joining tree. To visualize topologies representative of broad-scale (i.e. species-level) community structures, we examined
NetView results corresponding with k ≥ 65. Networks associated with these values for k displayed two clusters of individuals corresponding with striped marlin and white marlin; individuals were highly interconnected within both clusters. In the network for k = 65
(Fig. 4) a small number of connections (n = 5) existed between the striped marlin and white marlin communities. A single striped marlin was positioned intermediate to the two clusters comprising the network; this individual corresponded with the striped marlin sampled off New Zealand that was previously identified as a within-species outlier in
PCoA analyses and as a highly admixed individual in fastSTRUCTURE results. In addition, two white marlin were positioned within the striped marlin cluster. These individuals were sampled off Brazil and off Angola, and were identified as within-species outliers on PCoA axes one and three as well as on axes two and three. Finally, an unrooted neighbor-joining tree inferred using pairwise distances (Fig. 5) resolved white marlin as a monophyletic group genetically distant from striped marlin. White marlin outliers identified through PCoA, fastSTRUCTURE, and/or NetView analyses partially comprised a group of individuals displaying greater pairwise differences relative to the majority of white marlin samples. Collectively, results from the visualization of pairwise differences were consistent with those from multivariate and fastSTRUCTURE analyses in resolving two distinct groups corresponding with individuals identified as striped marlin
182
or as white marlin. Samples were organized to reflect these groups for all subsequent analyses. With the exception of results from NetView network analysis, striped marlin or white marlin individuals identified as within-species outliers consistently grouped with their respective species, and were therefore retained with these species for subsequent analyses.
Putative lineages resolved by individual-based clustering and the evaluation of pairwise differences were confirmed using Bayes factor delimitation. Results from BFD* included strong support for the scenario with white marlin and striped marlin comprising separate lineages representative of distinct species (Table 3). The Bayes factor corresponding with the comparison of the two scenarios we evaluated was 23,161. The species tree associated with the favored species delimitation model (Fig. 6A) also reflects a high degree of consistency among model runs in resolving lineages corresponding with striped marlin and white marlin.
Lineage characterization
To genetically characterize lineages confirmed through Bayes factor delimitation, we evaluated genetic differentiation and the presence of fixed differences between lineages corresponding with striped marlin and white marlin. We also calculated genetic diversity within lineages. Pairwise FST between lineages was 0.4802. Distance metrics more appropriate for comparisons at the species level included DS, DM, and DC; values for these metrics were equal to 0.0431, 0.0403, and 0.0520, respectively. We identified 20 loci (0.16% of SNPs in the final dataset) exhibiting fixed differences between lineages.
Observed and expected heterozygosities were similar between striped marlin and white
183
marlin (striped marlin: HE = 0.0433; white marlin: HE = 0.0442); mean allelic richness was also similar between species (striped marlin: AR = 1.044; white marlin: AR = 1.047).
Distance and diversity metrics, as well as the presence of fixed differences, were also assessed at the population level to determine the relationship of white marlin to previously resolved populations of striped marlin. Pairwise FST values ranged from
0.4865 to 0.4989 for comparisons of white marlin to populations of striped marlin (Table
4). A similar pattern was observed for pairwise distances based on other metrics. DS, DM, and DC calculated between white marlin and populations of striped marlin ranged from
0.0433–0.0458, 0.0404–0.0428, and 0.0533–0.0561 (Tables 4, S1), respectively. Across metrics, the greatest levels of differentiation corresponded with comparisons of white marlin to the western Indian Ocean population of striped marlin. Genetic diversity calculated at the population level was similar between white marlin (HE = 0.0442, AR =
1.047; Table 5) and populations of striped marlin (HE = 0.0402–0.0448, AR = 1.040–
1.045). Finally, 0.18–0.19% of SNPs (23–24 loci) corresponded with fixed differences between white marlin and populations of striped marlin (Table S2).
Results from lineage-based assignment testing performed using newhybrids included the assignment of all individuals identified as striped marlin to the pure striped marlin genotype category with 100% probability, except for a single individual which assigned to the F1 hybrid category with a probability of 99.4%. That individual corresponded with the striped marlin sampled off New Zealand which displayed a high degree of admixture in fastSTRUCTURE results (Fig. 3), and which appeared as a within- species outlier in results from PCoA (Figs. S2, S3). For individuals identified as white marlin, results from newhybrids included assignment of all individuals to the pure white
184
marlin genotype category with 100% probability, except for three individuals. Two of those individuals assigned to the F1 hybrid category with probabilities equal to 97.8% and
95.6%. A third individual was associated with assignment probabilities of 41.6% and
58.5% to the F1 hybrid and pure white marlin genotype categories, respectively. These three individuals corresponded with the white marlin sampled of the United States mid-
Atlantic which displayed high degrees of admixture in fastSTRUCTURE results (Fig. 3), and comprised three of the four within-species outliers identified across all PCoA scenarios (Figs. 1, S2, S3).
Phylogenetic inference
Phylogenetic relationships of white marlin and striped marlin were inferred using maximum likelihood estimation. Our RAxML phylogeny recovered two major clades corresponding with individuals identified as striped marlin or as white marlin (Figs. 6B,
S4). Both of these clades were monophyletic and displayed high levels of bootstrap support.
DISCUSSION
The primary goal of this study was to employ genome-wide molecular markers and a range of population genomic and species delimitation methods to evaluate the evolutionary relationship of white marlin and striped marlin. Results from this study are consistent with the presence of distinct evolutionary lineages for white marlin and striped marlin, although a small number of individuals displayed mixed ancestry and suggest historical and/or contemporary gene flow between these species.
185
Evolutionary relationship of striped marlin and white marlin
Multiple lines of evidence suggest the presence of distinct evolutionary lineages for striped marlin and white marlin, which is in agreement with the current taxonomy of these two species. Individual-based cluster analyses consistently resolved fish identified as striped marlin or white marlin as comprising two distinct clusters. These clusters were well-differentiated in results from multivariate and fastSTRUCTURE analyses, and in a neighbor-joining tree based on pairwise differences. Results from network visualization of the relationship of striped marlin and white marlin also included the presence of two distinct clusters corresponding with these species; however, two white marlin were placed within the striped marlin cluster, and a single striped marlin was positioned adjacent to the white marlin cluster. These results were likely due to the admixed genomic composition of these individuals (discussed below), and the limited ability of graph-based approaches to resolve low levels of divergence compared with model-based approaches grounded in population genetic theory (Steinig et al. 2016). Finally, strong support for a species tree composed of separate lineages for striped marlin and white marlin corroborates the presence of distinct evolutionary lineages for these species.
Comparisons of genetic distances within and between species are also consistent with the presence of distinct evolutionary lineages for striped marlin and white marlin, and provide novel insights on population- and species-level differentiation expected for genomic comparisons of pelagic fishes. The FST value between striped marlin and white marlin was an order of magnitude higher than most FST values between regional populations of striped marlin. This same pattern was also observed for FST values between white marlin and individual populations of striped marlin. An even greater
186
difference between inter- and intra-specific levels of differentiation was observed for additional distance metrics calculated in this study. In some instances, DS, DM, or DC calculated among populations of striped marlin were < 5% of values calculated between striped marlin and white marlin. Collectively, these results demonstrate considerably greater genetic divergence between striped marlin and white marlin than between populations of striped marlin, and suggest that striped marlin and white marlin correspond with species- rather than population-level comparisons.
Genetic distances reported in genomic studies of other marine species also suggest that the level of genetic differentiation between striped marlin and white marlin corresponds with that expected for species. Based on the assessment of over 6,000 genome-wide SNPs, Pecoraro et al. (2016) reported FST values ranging from 0.0171–
0.0474 for inter-oceanic comparisons of yellowfin tuna (Thunnus albacares), a globally distributed highly migratory pelagic fish. These values are an order of magnitude lower than FST calculated between striped marlin and white marlin in this study (= 0.4802), and are similar to intraspecific comparisons of striped marlin (= 0.0217–0.0703). Though additional examples from genomic studies of other highly migratory pelagic fishes are currently unavailable, comparisons with genomic studies focused on other marine systems are still informative. Gaither et al. (2015) surveyed nearly 4,000 SNPs across
Pacific reef fishes of the genus Acanthurus, and reported FST values ranging from 0.000 to 0.027 and 0.093 to 0.114 for comparisons among populations and species, respectively. Species-level divergences reported in that study are more similar to the population-level divergences reported in the present study. FST values based on nearly
19,000 SNPs surveyed across cryptic species of the Australian golden perch (Macquaria
187
ambigua), a highly dispersive freshwater fish, ranged from 0.546–0.589 between species
(Beheregaray et al. 2017), and are similar to the FST observed between striped marlin and white marlin in this study. Finally, Lal et al. (2018) reported values of DM ranging from
0.252 to 0.289 between species of the black-lip pearl oyster genus Pinctada. These values are considerably higher than DM calculated between striped marlin and white marlin in this study (= 0.0403). Though genetic distances expected for species-level comparisons vary according to species’ life history and additional biological factors, results from other genomic studies are consistent with the identification of striped marlin and white marlin as distinct species rather than sub-populations comprising a single species.
The levels of inter- and intra-specific differentiation calculated in this study also provide insights into the relationship of white marlin to the four regional populations recognized for striped marlin. Across distance metrics, the range of values associated with comparisons of white marlin to individual populations of striped marlin is small, but the greatest genetic distances were consistently associated with comparisons of white marlin to the western Indian Ocean population of striped marlin. In comparison, the lowest genetic distances were associated with comparisons of white marlin to the North
Pacific Ocean population of striped marlin, except for DC, where the smallest value corresponded with striped marlin from Oceania but is similar to that for the North Pacific
Ocean. Maximum likelihood inference of phylogenetic relationships also resolved a monophyletic lineage comprising white marlin as most closely related to striped marlin from the Pacific Ocean. Additionally, a slightly higher proportion of loci corresponded with fixed differences between white marlin and the western Indian Ocean population of striped marlin (0.19% compared to 0.18% for all other comparisons). A closer genetic
188
relationship of white marlin to Pacific Ocean striped marlin, rather than to Indian Ocean striped marlin, is counter-intuitive to what might be expected for these species. These results suggest that the limited range of variation observed across genetic distances associated with comparisons of white marlin to regional populations of striped marlin is not informative for inferring the evolutionary history of these species. Additional analyses (such as those discussed below) and larger sample sizes per geographic region may be required to address this question.
Though the genomic results of the present study support the validity of striped marlin and white marlin as distinct species, distinguishing these species based on non- genomic characters is challenging. Striped marlin and white marlin are morphologically similar, though striped marlin generally attain larger body sizes (in excess of 350 cm total length and 200 kg total weight, but variable among regions; Namakura 1985) relative to white marlin (in excess of 280 cm total length and 82 kg total weight; Nakamura 1985).
Reports from morphological examinations of vouchered specimens, including type material for striped marlin (Ueyanagi & Wares 1975), suggest striped marlin can be distinguished by pointed tips on the first dorsal, first anal, and pectoral fins, whereas these fin tips are rounded in white marlin (Nakamura 1983, 1985). However, tips of the pectoral fins and of the first dorsal fin are reported as rounded in most, but not all, specimens of white marlin examined by Nakamura (1983). This observation suggests that fin morphology, at least for the pectoral and first dorsal fins, is not a reliable character for distinguishing striped marlin and white marlin.
Several molecular studies representing a progression of methodological advancements over a twenty-year time period have attempted to resolve the genetic
189
relationship of striped marlin and white marlin. Early investigation of 42 presumptive gene loci across 20 white marlin and 20 striped marlin revealed a lack of fixed allelic differences between these species (Morgan 1992). Restriction fragment length polymorphism (RFLP) analysis of a 350 bp segment of the mitochondrial (mt) DNA cytochrome b gene region using 13 restriction endonucleases revealed identical restriction profiles across 21 striped marlin and 14 white marlin (Chow 1994). Graves &
McDowell (1995) performed RFLP analysis of whole mtDNA for 35 white marlin and
166 striped marlin using ten restriction endonucleases. Results from that study included a corrected mean nucleotide sequence divergence of only 0.06%, as well as a lack of fixed differences, between species. Additionally, Graves & McDowell (1995) identified two mtDNA haplotypes that were common to both species, and the most common striped marlin haplotype differed from the most common white marlin haplotype by only a single restriction site. Finnerty & Block (1995) analyzed sequence data for a 590 bp segment of the mtDNA cytochrome b gene region across three striped marlin and two white marlin.
Phylogenetic analyses performed in that study resolved a monophyletic clade comprising striped marlin and white marlin; however, sequences of these species differed by < 0.5% and no fixed differences between species were observed. McDowell et al. (unpublished data) observed 2.25% net nucleotide sequence divergence between striped marlin and white marlin based on sequences for an 854 bp segment of the mtDNA control region; however, this result was based on a small number of individuals (striped marlin: n = 7; white marlin: n = 9). McDowell et al. (unpublished data) also evaluated four nuclear microsatellite loci across samples of striped marlin and white marlin, but considerable overlap in allele frequencies between species was observed.
190
More recent phylogenetic comparisons based on a variety of nuclear and mt markers have also failed to resolve striped marlin and white marlin as reciprocally monophyletic lineages. Collette et al. (2006) resolved white marlin and striped marlin as comprising a well-supported clade based on maximum parsimony and Bayesian analyses of three mt gene regions and an anonymous nuclear locus, but within this clade species were not resolved as distinct lineages. A similar result was reported by Hanner et al.
(2011) based on maximum parsimony and neighbor-joining analyses of sequence data for the nuclear rhodopsin gene and the mtDNA ‘barcode’ region (COI; e.g. Hebert et al.
2003), as well as previously published mtDNA control region sequences (Graves &
McDowell 2006, McDowell & Graves 2008). Hanner et al. (2011) also identified a COI haplotype that was shared between striped marlin and white marlin. Recently, Santini &
Sorenson (2013) analyzed previously published sequence data for three nuclear genes and seven mtDNA gene regions. Maximum likelihood and Bayesian phylogenies based on these sequences resolved two well-supported clades comprising single exemplars for striped marlin or for white marlin. A time-calibrated Bayesian phylogeny also reported by Santini & Sorenson (2013) estimated a divergence time of 0.4 to 2.6 million years ago
(Ma; 95% highest posterior density intervals) for these species. Collectively, difficulty distinguishing striped marlin and white marlin based on morphological characteristics and traditional molecular comparisons such as those discussed here demonstrates the utility of NGS-based genomic approaches for resolving the evolutionary relationship of these species.
191
Significance of individuals with shared ancestry
Results from multivariate analyses and comparisons of pairwise differences identified a number of white marlin and a single striped marlin as within-species outliers positioned adjacent to or slightly distant from larger groups of individuals comprising the same species. This striped marlin (sampled off New Zealand) and a subset of these white marlin (three individuals sampled off the United States’ mid-Atlantic) also exhibited high levels of shared ancestry with the alternate species, and displayed high probabilities of assignment to an F1 striped marlin x white marlin hybrid genotype category. Possible explanations for shared genetic variation among these individuals include the incomplete sorting of evolutionary lineages due to recent divergence, and/or divergence followed by secondary contact and introgression (Coyne & Orr 2004, Carstens & Knowles 2007,
Harrison & Larson 2014, Zhou et al. 2017). Hudson & Coyne (2002) estimate that nine to twelve N generations, where N is the historically effective population size of each descendant, are required for diverging allopatric populations to achieve reciprocal monophyly of lineages at > 95% of sampled nuclear loci. Though average generation lengths for striped marlin and white marlin are short (estimated at 4.4 years for striped marlin and 5.5 years for white marlin; Collette et al. 2011), these species are estimated to be recently diverged (0.4 to 2.6 Ma; Santini & Sorenson 2013) and presumably display large effective population sizes. Given these factors, it is possible that lineage sorting for striped marlin and white marlin is incomplete; however, this mechanism would be expected to correspond with a range of admixture across a large proportion of individuals for both species, rather than a handful of individuals displaying high degrees of shared genetic variation.
192
Other factors suggest the possibility of isolation followed by secondary contact and introgression as a mechanism facilitating shared genetic variation in striped marlin and white marlin. Given the divergence time estimated for these species, isolation between Atlantic and Indo-Pacific populations could have been facilitated by glaciation during the Pleistocene geological epoch (initiated approximately 2.58 Ma; Gibbard et al.
2010). During this time, colder water temperatures likely resulted in a contracted North-
South spatial distribution for the most recent common ancestor of striped marlin and white marlin, thereby limiting inter-oceanic connectivity that may have occurred around present-day South Africa. However, because white marlin and striped marlin display lower levels of genetic divergence relative to other istiophorid billfishes with spatial distributions spanning the Atlantic and Indo-Pacific (e.g. sailfish and blue marlin; Graves
& McDowell 1995), inter-oceanic isolation of striped marlin and white marlin may have been comparatively incomplete and/or persisted for a shorter period of time. This scenario is plausible given the more temperate waters inhabited by striped marlin and white marlin compared with other istiophorids. Secondary contact of striped marlin and white marlin may have been facilitated by warming waters after the most recent glaciation event, resulting in the shared genetic variation observed in this study.
Incomplete lineage sorting and secondary contact following a period of isolation produce similar genetic signals that are challenging to disentangle, but a number of statistical methods facilitate evaluation of the relative contribution of these mechanisms to observed patterns of genetic variation. Several of these methods implement a simulation approach for inferring evolutionary history based on a mathematical model describing the random evolution of genetic variation over time (e.g. the coalescent,
193
Kingman 1982). For example, the IM program suite (Nielsen & Wakeley 2001, Hey &
Nielsen 2007, Sethuraman & Hey 2016) can be used to compare competing models of demographic history by estimating the relative effects of migration and isolation on genetic diversity among species. Other coalescent-based methods use information derived from the allele frequency spectrum to infer the demographic history of species
(Gutenkunst et al. 2009, Excoffier et al. 2013). Similarly, a variety of test statistics are useful for identifying evolutionary mechanisms relevant to a study system. These include the D-statistic (e.g. ABBA-BABA test; Green et al. 2010, Durand et al. 2011, Patterson et al. 2012), which quantifies the frequency of allelic patterns expected for incomplete lineage sorting or introgression. Collectively, methods focused on inferring demographic history are useful for identifying possible evolutionary mechanisms operating on a species complex, but these methods first require an understanding of distinct lineages
(e.g. species) represented in the system.
Concluding remarks
We demonstrate the presence of distinct evolutionary lineages for striped marlin and white marlin, and support the current taxonomy of these species, the status of which has been in question for the past few decades. Information resulting from the comparison of intra- and inter-specific levels of genetic divergence in this study also provides a valuable reference for future genomic assessments of population- and species-level relationships in istiophorid billfishes and other pelagic fishes. Results from these comparisons also suggest that genomic evaluations of Atlantic and Indo-Pacific populations of blue marlin and sailfish may be warranted, given that inter-oceanic divergence in these species has
194
previously been shown to be greater than that of striped marlin and white marlin. Finally, results from this study are also important for facilitating subsequent analyses focused on evaluating the evolutionary history of striped marlin and white marlin. Such an assessment will enable the identification of evolutionary processes responsible for the shared genetic variation observed among a small number of individuals in this study.
More broadly, an understanding of evolutionary processes relevant to the demographic history of striped marlin and white marlin will provide key insights into the poorly understood process of speciation in the pelagic marine environment.
ACKNOWLEDGMENTS
The authors are indebted to a large number of recreational anglers worldwide for the contribution of samples to this project, particularly captains Trevor Hansen, Tim
Richardson, and Ken Neill III. We also thank several scientists who provided samples, including Sofia Ortega-Garcia, Sam Williams, John Holdsworth, and Wei-Chuan Chiang.
Finally, a number of conservation groups were essential to the success of this project, especially the African Billfish Foundation (Roy Bealey), International Game Fish
Association (Rob Kramer, Jason Schratwieser, and several international representatives), and Japan Game Fish Association (Tomo Higashi).
195
REFERENCES
Adamack, A., & Gruber, B. (2014). POPGENREPORT: simplifying basic population
genetic analyses in R. Methods in Ecology and Evolution, 5, 384–387.
Anderson, E., & Thompson, E. (2002). A model-based method for identifying species
hybrids using multilocus genetic data. Genetics, 160, 1217–1229.
Beheregaray, L., Pfeiffer, L., Attard, C., Sandoval-Castillo, J., Domingos, F., Faulks, L.,
Gilligan, D., & Unmack, P. (2017). Genome-wide data delimits multiple climate-
determined species ranges in a widespread Australian fish, the golden perch
(Macquaria ambigua). Molecular Phylogenetics and Evolution, 111, 65–75.
Benjamini, Y., & Yekutieli, D. 2001. The control of false discovery rate under
dependency. Annals of Statistics, 29, 1165–1188.
Bickford, D., Lohman, D., Sodhi, N., Ng, P., Meier, R., Winker, K., Ingram, K., & Das, I.
(2006). Cryptic species as a window on diversity and conservation. Trends in
Ecology and Evolution, 22, 148–155.
Bouckaert, R., Heled, J., Künert, D., Vaughan, T., Wu, C-H., Xie, D., Suchard, M.,
Rambaut, A., & Drummond, A. (2014). BEAST 2: a software platform for
Bayesian evolutionary analysis. PLoS Computational Biology, 10, e1003537.
Bromhead, D., Pepperell, J., Wise, B., and Findlay, J. (2003). Striped marlin: biology and
fisheries. Canberra, Australia: Bureau of Rural Sciences.
Bryant, D., Bouckaert, R., Felsenstein, J., Rosenberg, N., & RoyChoudhury, A. (2012).
Inferring species trees directly from biallelic genetic markers: bypassing gene
trees in a full coalescent analysis. Molecular Biology and Evolution, 29, 1917–
1932.
196
Carstens, B., & Knowles, L. (2007). Estimating species phylogeny from gene-tree
probabilities despite incomplete lineage sorting: an example from Melanoplus
grasshoppers. Systematic Biology, 56, 400–411.
Carstens, B., Pelletier, T., Reid, N., & Satler, J. (2013). How to fail at species
delimitation. Molecular Ecology, 22, 4369–4383.
Cavalli-Sforza, L., & Edwards, A. (1967). Phylogenetic analysis: models and estimation
procedues. Evolution, 21, 550–570.
Chow, S. (1994). Identification of billfish species using mitochondrial cytochrome b gene
fragment amplified by polymerase chain reaction. ICCAT Collective Volume of
Scientific Papers, 41, 549–556.
Collette, B., Carpenter, K., Polidoro, B., Juan-Jordá, M., Boustany, A., Die, D., Elfes, C.,
Fox, W., Graves, J., Harrison, L., McManus, R., Minte-Vera, C., Nelson, R.,
Restrepo, V., Schratwieser, J., Sun, C., Amorim, A., Brick Peres, M., Canales, C.,
Cardenas, G., Chang, S., Chiang, W., de Oliveira Leite, Jr., N., Harwell, H.,
Lessa, R., Fredou, F., Oxenford, H., Serra, R., Shao, K., Sumaila, R., Wang, S.,
Watson, R., & Yáñez, E. (2011). High value and long live–double jeopardy for
tunas and billfishes. Science, 333, 291–292.
Collette, B., McDowell, J., & Graves, J. (2006). Phylogeny of Recent billfishes
(Xiphioidei). Bulletin of Marine Science, 79, 455–468.
Coyne, J., & Orr, H. (2004). Speciation. Sinauer Associates Inc., Sunderland, Mass. de Queiroz, K. (2007). Species concepts and species delimitation. Systematic Biology, 56,
879–886.
197
Durand, E., Patterson, N., Reich, D., & Slatkin, M. (2011). Testing for ancient admixture
between closely related populations. Molecular Biology and Evolution, 28, 2239–
2252.
Excoffier, L., Dupanloup, I., Huerta-Sánchez, E., Sousa, V., & Foll, M. (2013). Robust
demographic inference from genomic and SNP data. PLoS Genetics, 9, e1003905.
Finnerty, J., & Block, B. (1995). Evolution of cytochrome b in the Scombroidei
(Teleostei): molecular insights into billfish (Istiophoridae and Xiphiidae)
relationships. Fishery Bulletin, 93, 78–96.
Gaither, M., Bernal, M., Coleman, R., Bowen, B., Jones, S., Simison, W., & Rocha, L.
(2015). Genomic signatures of geographic isolation and natural selection in coral
reef fishes. Molecular Ecology, 24, 1543–1557.
Gibbard, P., Head, M., Walker, M., & The Subcommission on Quaternary Stratigraphy.
(2010). Formal ratification of the Quaternary System/Period and the Pleistocene
Series/Epoch with a base at 2.58 Ma. Journal of Quaternary Science, 25, 96–102.
Goudet, J. (2005). HIERFSTAT, a package for R to compute and test hierarchical F-
statistics. Molecular Ecology Resources, 5, 184–186.
Gower, J. (1966). Some distance properties of latent root and vector methods used in
multivariate analysis. Biometrika, 53, 325–338.
Graves, J., & McDowell, J. (1994). Genetic analysis of striped marlin (Tetrapturus
audax) population structure in the Pacific Ocean. Canadian Journal of Fisheries
and Aquatic Sciences, 51, 1762–1768.
Graves, J., & McDowell, J. (1995). Inter-ocean genetic divergence of Istiophorid
billfishes. Marine Biology, 122, 193–203.
198
Graves, J., & McDowell, J. (2006). Genetic analysis of white marlin (Tetrapturus
albidus) stock structure. Bulletin of Marin Science, 79, 469–482.
Green, R., Krause, J., Briggs, A., Maricic, T., Stenzel, U., Kircher, M., Patterson, N., et
al. (2010). A draft sequence of the Neandertal genome. Science, 328, 710–722.
Gruber, B., Unmack, P., Berry, O., & Georges, A. 2018. DARTR: and R package to
facilitate analysis of SNP data generated from reduced representation genome
sequencing. Molecular Ecology Resources. doi: 10.1111/1755-0998.12745.
Grummer, J., Bryson Jr., R., & Reeder, T. (2014). Species delimitation using Bayes
factors: simulations and application to the Sceloporus scalaris species group
(Squamata: Phrynosomatidae). Systematic Biology, 63, 119–133.
Gutenkunst, R., Hernandez, R., Williamson, S., & Bustamante, C. (2009). Inferring the
joint demographic history of multiple populations from multidimensional SNP
frequency data. PLoS Genetics, 5, e1000695.
Hanner, R., Floyd, R., Bernard, A., Collette, B., & Shivji, M. (2011). DNA barcoding of
billfishes. Mitochondrial DNA, 22, 27–36.
Harrison, R., & Larson, E. (2014). Hybridization, introgression, and the nature of species
boundaries. Journal of Heredity, 105, 795–809.
Hebert, P., Cywinska, A., Ball, S., & deWaard, J. (2003). Biological identifications
through DNA barcodes. Proceedings of the Royal Society of London B, 270, 313–
321.
Hey, J., & Nielsen, R. (2007). Integration within the Felsenstein equation for improved
Markov chain Monte Carlo methods in population genetics. Proceedings of the
National Academy of Sciences, 104, 2785–2790.
199
Hirasaka, K., & Nakamura, H. (1947). On the Formosan spear-fishes. Bulletin of the
Oceanographic Institute, Taiwan, 3, 9–24.
Hoolihan, J., Luo, J., Snodgrass, D., Orbesen, E., Barse, A., & Prince, E. (2015). Vertical
and horizontal habitat use by white marlin Kajikia albida (Poey, 1860) in the
western North Atlantic Ocean. ICES Journal of Marine Science, 72, 2364–2373.
Hudson, R., & Coyne, J. (2002). Mathematical consequences of the genealogical species
concept. Evolution, 56, 1557–1565.
Jombart, T. (2008). Adegenet: a R package for the multivariate analysis of genetic
markers. Bioinformatics, 24, 1403–1405.
Jombart, T., Devillard, S., & Balloux, F. (2010). Discriminant analysis of principal
components: a new method for the analysis of genetically structured populations.
BMC Genetics, 11, 1471–2156.
Jombart, T., Pontier, D., & Dufour, A. (2009). Genetic markers in the playground of
multivariate analysis. Heredity, 102, 330–341.
Kamvar, Z., Tabima, J., & Grünwald, N. (2014). Poppr: an R package for genetic
analysis of populations with clonal, partially clonal, and/or sexual reproduction.
PeerJ, 2, e281.
Kingman, J. (1982). The coalescent. Stochastic Processes and their Applications, 13,
235–248.
Lal, M., Southgate, P., Jerry, D., & Zenger, K. (2018). Genome-wide comparisons reveal
evidence for a species complex in the black-lip pearl oyster Pinctada
margaritifera (Bivalvia: Pteriidae). Scientific Reports, 8, 191.
200
Leaché, A., Banbury, B., Felsenstein, J., Nieto-Montes de Oca, A., & Stamatakis, A.
(2015). Short tree, long tree, right tree, wrong tree: new acquisition bias
corrections for inferring SNP phylogenies. Systematic Biology, 64, 1032–1047.
Leaché, A., Fujita, M., Minin, V., & Bouckaert, R. (2014). Species delimitation using
genome-wide SNP data. Systematic Biology, 63, 534–542.
Lemmon, E., & Lemmon, A. (2013). High-throughput genomic data in systematics and
phylogenetics. Annual Review of Ecology, Evolution, and Systematics, 44, 99–
121.
Mardis, E. (2008). The impact of next-generation sequencing technology on genetics.
Trends in Genetics, 24, 133–141.
Martin, A., Naylor, G., & Palumbi, S. (1992). Rates of mitochondrial DNA evolution in
sharks are slow compared with mammals. Nature, 357, 153–155.
Mayr, E. (1954). Geographic speciation in tropical echinoids. Evolution, 8, 1–18.
McDowell, J., & Graves, J. (2008). Population structure of striped marlin (Kajikia audax)
in the Pacific Ocean based on analysis of microsatellite and mitochondrial DNA.
Canadian Journal of Fisheries and Aquatic Sciences, 65, 1307–1320.
McDowell, J., Mamoozadeh, N., Brightman, H., & Graves, J. (2018). Use of rapidly
evolving molecular markers to distinguish species and clarify range uncertainties
in the spearfishes (Istiophoridae, Tetrapturus). Bulletin of Marine Science, 94.
doi: https://doi.org/10.5343/bms.2017.1130.
Morgan, L. 1992. Allozyme analysis of billfish population structure. Master’s thesis.
Virginia Institute of Marine Science, College of William & Mary, Gloucester
Point, VA, USA.
201
Nakamura, I. 1983. Systematics of billfishes (Xiphiidae and Istiophoridae). Publications
of the Seto Marine Biological Laboratory, 28, 255–396.
Nakamura, I. (1985). Billfishes of the world: an annotated and illustrated catalogue of
marlins, sailfishes, spearfishes, and swordfishes known to date. FAO Fisheries
Synopsis No. 125, vol. 5. 65 pp. Rome, Italy: Food and Agricultural Organization.
Narum, S. 2006. Beyond Bonferroni: less conservative analyses for conservation
genetics. Conservation Genetics, 7, 783–787.
Narum, S., Campbell, N., Meyer, K., Miller, M., & Hardy, R. (2013). Thermal adaptation
and acclimation of ectotherms from differing aquatic climates. Molecular
Ecology, 22, 3090–3097.
Nei, M. (1972). Genetic distance between populations. The American Naturalist, 106,
283–291.
Nei, M. (1973). The theory and estimation of genetic distance. In N. Morton (Ed.),
Genetic Structure of Populations (pp. 45-54). Honolulu, HI: University Press of
Hawaii.
Neuditschko, M., Khatkar, M., & Raadsma, H. (2012). NETVIEW: a high-definition
network-visualization approach to detect fine-scale population structures from
genome-wide patterns of variation. PLoS ONE, 7, e48375.
Nielsen, R., & Wakeley, J. (2001). Distinguishing migration from isolation: a Markov
chain Monte Carlo approach. Genetics, 158, 885–896.
Ortega-García, S., Klett-Traulsen, A., & Ponce-Díaz, G. (2003). Analysis of sportfishing
catch rates of striped marlin (Tetrapturus audax) at Cabo San Lucas, Baja
202
California Sur, Mexico, and their relation to sea surface temperature. Marine and
Freshwater Research, 54, 483–488.
Ortiz, M., Prince, E., Serafy, J., Holts, D., Davy, K., Pepperell, J., Lowry, M., &
Holdsworth, J. C. (2003). Global overview of the major constituent–based billfish
tagging programs and their results since 1954. Marine and Freshwater Research,
54, 489–507.
Palumbi, S. (1992). Marine speciation on a small planet. Trends in Ecology and
Evolution, 7, 114–118.
Paradis, E., Claude, J., & Strimmer, K. (2004). APE: analyses of phylogenetics and
evolution in R language. Bioinformatics, 20, 289–290.
Patterson, N., Moorjani, P., Luo, Y. Mallick, S., Rohland, N., Zhan, Y., Genschoreck, T.,
Webster, T., & Reich, D. (2012). Ancient admixture in human history. Genetics,
192, 1065–1093.
Pecoraro, C., Babbucci, M., Villamor, A., Franch, R., Papetti, C., Leroy, B., Ortega-
Garcia, S., Muir, J., Rooker, J., Arocha, F., Murua, H., Zudaire, I., Chassot, E.,
Bodin, N., Tinti, F., Bargelloni, & Cariani, A. (2016). Methodological assessment
of 2b-RAD genotyping technique for population structure inferences in yellowfin
tuna (Thunnus albacares). Marine Genomics, 25, 43–48.
Penrith, M., & Cram, D. (1974). The Cape of Good Hope: a hidden barrier to billfishes.
In R. Shomura and F. Williams (Eds.), Proceedings of the International Billfish
Symposium, Kailua-Kona, Hawaii, 9-12 August 1972. Part 2. Review and
contributed papers (pp. 175–187). NOAA Technical Report NMFS SSRF-675.
203
Peterson, B., Weber, J., Kay, E., Fisher, H., & Hoekstra, H. (2012). Double digest
RADseq: an inexpensive method for de novo SNP discovery and genotyping in
model and non–model species. PLoS ONE, 7, e37135.
Purcell, C., & Edmands, S. (2011). Resolving the genetic structure of striped marlin,
Kajikia audax, in the Pacific Ocean through spatial and temporal sampling of
adult and immature fish. Canadian Journal of Fisheries and Aquatic Sciences, 68,
1861–1875.
R Core Team. (2017). R: a language and environment for statistical computing.
Raj, A., Stephens, M., & Pritchard, J. (2014). fastSTRUCTURE: variational inference of
population structure in large SNP data sets. Genetics, 197, 573–589.
Rambaut, A., Drummond, A., Xie, D., Baele, G., & Suchard, M. (2018). Tracer v1.7.
Rosenberg, N. (2004). DISTRUCT: a program for the graphical display of population
structure. Molecular Ecology Notes, 4, 137–138.
Saitou, N., & Nei, M. (1987). The neighbor-joining method: a new method for
reconstructing phylogenetic trees. Molecular Biology and Evolution, 4, 406–425.
Sansaloni, C., Petroli, C., Jaccoud, D., Carling, J., Detering, F., Grattapaglia, D., &
Kilian, A. (2011). Diversity Arrays Technology (DArT) and next–generation
sequencing combined: genome–wide, high throughput, highly informative
genotyping for molecular breeding of Eucalyptus. BMC Proceedings, 5, P54.
Santini, F., & Sorenson, L. (2013). First molecular timetree of billfishes
(Istiophoriformes: Acanthomorpha) shows a Late Miocene radiation of marlins
and allies. Italian Journal of Zoology, 80, 481–489.
204
Schlenker, L. (2014). Physiological stress and post-release mortality of white marlin
(Kajikia albida) caught in the U.S. recreational fishery (Master’s thesis). Virginia
Institute of Marine Science, College of William & Mary, Gloucester Point, VA,
USA.
Sethuraman, A., & Hey, J. (2016). IMa2p - parallel MCMC and inference of ancient
demography under the Isolation with Migration (IM) model. Molecular Ecology
Resources, 16, 206–215.
Seutin, G., White, B., & Boag, P. (1991). Preservation of avian blood and tissue samples
for DNA analyses. Canadian Journal of Zoology, 69, 82–90.
Sippel, T., Davie, P., Holdsworth, J., & Block, B. (2007). Striped marlin (Tetrapturus
audax) movements and habitat utilization during a summer and autumn in the
Southwest Pacific Ocean. Fisheries Oceanography, 16, 459–472.
Stamatakis, A. (2014). RAxML version 8: a tool for phylogenetic analysis and post-
analysis of large phylogenies. Bioinformatics, 30, 1–2.
Stamatakis, A., Hoover, P., & Rougemont, J. (2008). A rapid bootstrap algorithm for the
RAxML web servers. Systematic Biology, 57, 758–771.
Steinig, E., Neuditschko, M., Khatkar, M., Raadsma, H., & Zenger, K. (2016). NETVIEW
P: a network visualization tool to unravel complex population structure using
genome-wide SNPs. Molecular Ecology Resources, 16, 216–227.
Talbot, F., & Penrith, M. (1962). Tunnies and marlins of South Africa. Nature, 193, 558–
559.
Ueyanagi, S., & Wares, P. (1975). Synopsis of biological data on striped marlin,
Tetrapturus audax (Philippi, 1887). In R. Shomura, and F. Williams (Eds.),
205
Proceedings of the International Billfish Symposium, Kailua-Kona, Hawaii, 9-12
August 1972. Part 3. Species synopses (pp. 132–159). NOAA Technical Report
NMFS SSRF-675.
Weir, B., & Cockerham, C. (1984). Estimating F–statistics for the analysis of population
structure. Evolution, 38, 1358–1370.
White Marlin Biological Review Team. (2007). Atlantic white marlin status review.
Report to National Marine Fisheries Service, Southeast Regional office (88 pp.).
Wigginton, J., Cutler, D., & Abecasis, G. (2005). A note on exact tests of Hardy–
Weinberg equilibrium. American Journal of Human Genetics, 7, 887–883.
Wright, S. (1951). The genetical structure of populations. Annals of Eugenics, 15, 323–
354.
Zhou, Y., Duvaux, L., Ren, G., Zhang, L., Savolainen, O., & Liu, J. (2017). Importance
of incomplete lineage sorting and introgression in the origin of shared genetic
variation between two closely related pines with overlapping distributions.
Heredity, 118, 211–220.
206
Table 1. Sampling details for striped marlin and white marlin sample collections analyzed in this study. Striped marlin sample collections are organized according to the populations resolved in Chapter III.
Sampling Location Years Collected Number Striped marlin Western Indian Ocean South Africa 2015 - 2017 11 Kenya 2015 - 2016 26
Oceania Western Australia 2016 8 Eastern Australia 1994, 2010 - 2012, 2015 37 New Zealand 2017 23
North Pacific Ocean Japan 2015 18 Taiwan 2014 - 2016 12 Hawaii 2015 21 California 2000, 2016 16
Eastern Central Pacific Ocean Baja California 2015 21 Ecuador 1992, 2016 37 Peru 2016 20
Total: 250 White marlin Angola 2014, 2015 2 Azores 2012 1 Brazil 2006, 2015 19 Caribbean Sea* 2016 9 Gulf of Mexico* 2007, 2008 8 Morocco 1995, 2016 18 United States mid-Atlantic 2015 18
Total: 75 *Larval sample collection
207
Table 2. Number of single nucleotide polymorphism (SNP) loci retained after each filtering step.
Filter No. Loci Retained Loci received from Diversity 61,908 Arrays Technology Pty. Ltd. Missing ≥ 25% genotypes 49,046
Average reproducibility < 95% 48,894 More than one SNP per 28,005 reduced representation locus Monomorphic 12,762
208
Table 3. Results from the comparison of species delimitation scenarios using the Bayes factor delimitation method implemented in BFD*. MLE = marginal likelihood estimate.
Model Species MLE Rank Bayes factor Current taxonomy 2 -48,501 1 -- Alternative 1 -60,082 2 23,161
209
Table 4. Pairwise distances between populations of white marlin and striped marlin, where white marlin comprises a single population (WHM) and striped marlin comprises four previously resolved populations (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). FST (Weir & Cockerham 1984) is shown in lower left of table and DS (Nei 1972) is shown in upper right. Cells are colored as a heat map ranging from green (low distance values) to red (high distance values).
WIO Oceania NPO ECPO WHM WIO -- 0.0018 0.0034 0.0037 0.0458 Oceania 0.0289 -- 0.0016 0.0022 0.0438 NPO 0.0590 0.0253 -- 0.0013 0.0433 ECPO 0.0703 0.0406 0.0217 -- 0.0437 WHM 0.4989 0.4865 0.4764 0.4928 --
210
Table 5. Diversity metrics for populations of white marlin and striped marlin, where white marlin comprises a single population (WHM) and striped marlin comprises four previously resolved populations (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). Cells are colored as a heat map ranging from green (low diversity values) to red (high diversity values). N = number of individuals, HO = observed heterozygosity, HE = expected heterozygosity, AR = mean allelic richness.
N HO HE AR WIO 37 0.0400 0.0402 1.040 Oceania 68 0.0448 0.0423 1.043 NPO 67 0.0517 0.0448 1.045 ECPO 78 0.0416 0.0404 1.041 WHM 75 0.0445 0.0442 1.047
211
Fig. 1. Two-dimensional plot of axes one and two resulting from principal coordinate analysis (PCoA) of striped marlin and white marlin genotype data. Percentage of total genetic variation explained by each axis is shown. Points are colored by species according to legend. Inset at lower right shows eigenvalues associated with the PCoA, including plotted axes (black bars). STM = striped marlin, WHM = white marlin.
212
Fig. 2. Two-dimensional plot of results for K = 2 scenario assessed with discriminant analysis of principal components (DAPC) for striped marlin and white marlin genotype data. Red density peak corresponds with the genetic cluster comprising striped marlin, and blue peak corresponds with cluster comprising white marlin. Inset at top right shows principal component analysis (PCA) eigenvalues associated with DAPC (darkly shaded bars denote retained axes).
213
WHM
mulations with K = 2. STM = striped marlin, WHM = white marlin. K WHM STM 2. = with striped = marlin, marlin. white = mulations
si
STM
TRUCTURE
Barplot displaying results from fastS from results displaying Barplot
3. Fig.
214
Fig. 4. NetView network topology based on pairwise differences between individuals and k-nearest neighbor = 65. Nodes are colored by species according to legend, edges depicting connections among individuals are shown in gray. STM = striped marlin, WHM = white marlin.
215
Fig. 5. Unrooted neighbor-joining tree based on pairwise differences between individuals. Branch tips are colored according to the legend, and include the identification of populations previously resolved for striped marlin (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). STM = striped marlin, WHM = white marlin. Individuals identified as within- species outliers by principal coordinate analysis are denoted with single arrowhead, and individuals also displaying high levels of shared ancestry and high assignment probabilities to F1 striped marlin x white marlin hybrids are denoted with double arrowhead.
216
) )
M. nigricans M.
del are shown. del
bluemarlin (
Species tree for favored species delimitation model based on analyses performed in on in based model for analysestree performed species delimitation Species favored
Panel A: Panel
Support values based on 1,000 bootstrap replicates are shown. 1,000 on values are Support replicates based bootstrap
Maximum likelihood phylogeny inferred using RAxML. Monophyletic clades comprising striped marlin or white or striped using Maximum phylogeny comprising clades marlin inferred RAxML. Monophyletic likelihood
Species tree and gene tree for striped marlin (STM) and white marlin (WHM). (STM) A striped for tree marlin single marlin gene white Speciesand tree and
Fig. 6. Fig. trees. both root wasto used trees of this for inferred set complete mo the density which for tree a as depicted is SNAPPBFD*. and Tree B: Panel collapsed. are marlin
217
SUPPLEMENTARY MATERIALS
Table S1. Pairwise distances between populations of white marlin and striped marlin, where white marlin comprises a single population (WHM) and striped marlin comprises four previously resolved populations (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). DC (Cavalli-Sforza & Edwards 1967) is shown in lower left of table and DM (Nei 1973) is shown in upper right. Cells are colored as a heat map ranging from green (low distance values) to red (high distance values).
WIO Oceania NPO ECPO WHM WIO -- 0.0018 0.0033 0.0036 0.0428 Oceania 0.0049 -- 0.0015 0.0021 0.0409 NPO 0.0066 0.0041 -- 0.0013 0.0404 ECPO 0.007 0.0047 0.0036 -- 0.0409 WHM 0.0561 0.0533 0.0534 0.0541 --
218
Table S2. Percentage of loci exhibiting fixed differences between populations of white marlin and striped marlin, where white marlin comprises a single population (WHM) and striped marlin comprises four previously resolved populations (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean).
WIO Oceania NPO ECPO WHM WIO -- 0.00 0.00 0.00 0.19 Oceania 0.18 -- 0.00 0.00 0.18 NPO 0.18 0.00 -- 0.00 0.18 ECPO 0.18 0.00 0.00 -- 0.18 WHM 0.19 0.00 0.00 0.00 --
219
Fig. S1. Two-dimensional plot of axes one and two resulting from principal coordinate analysis (PCoA) of striped marlin and white marlin genotype data. Percentage of total genetic variation explained by each axis is shown. Points are colored by species according to legend, and include the identification of populations previously resolved for striped marlin (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). Inset at lower right shows eigenvalues associated with the PCoA, including plotted axes (black bars). STM = striped marlin, WHM = white marlin.
220
Fig. S2. Two-dimensional plot of axes one and three resulting from principal coordinate analysis (PCoA) of striped marlin and white marlin genotype data. Percentage of total genetic variation explained by each axis is shown. Points are colored by species according to legend, and include the identification of populations previously resolved for striped marlin (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). Inset at upper left shows eigenvalues associated with the PCoA, including plotted axes (black bars). STM = striped marlin, WHM = white marlin.
221
Fig. S3. Two-dimensional plot of axes two and three resulting from principal coordinate analysis (PCoA) of striped marlin and white marlin genotype data. Percentage of total genetic variation explained by each axis is shown. Points are colored by species according to legend, and include the identification of populations previously resolved for striped marlin (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). Inset at upper right shows eigenvalues associated with the PCoA, including plotted axes (black bars). STM = striped marlin, WHM = white marlin.
222
Fig. S4. Unrooted maximum likelihood phylogeny inferred using RAxML. Branch tips are colored according to the legend, and include the identification of populations previously resolved for striped marlin (Chapter III; WIO = Western Indian Ocean, NPO = North Pacific Ocean, ECPO = Eastern Central Pacific Ocean). STM = striped marlin, WHM = white marlin. Support values (> 50) based on 1,000 bootstrap replicates are shown.
223
CONCLUSIONS
The goal of this dissertation was to employ a genetic approach, including the use of traditional and newly developed methodologies, for addressing questions regarding the stock structure and species status of striped marlin and white marlin. The following discussion highlights original contributions resulting from this dissertation, their significance to the broader scientific community, and possible future directions for this work.
Contributions
For studies focused on identifying the number and geographic extent of genetically distinct populations of a species in a given region, perhaps the most perplexing result is an inability to reject the null hypothesis of a single genetic population. Such a negative result could reflect a true lack of population structuring; however, it is also possible that genetically distinct populations exist, but statistical power was too low to enable detection (Waples 1998). For pelagic marine fishes, a negative result is especially challenging because genetic differentiation between populations of these species, if present, is expected to be low and difficult to detect relative to populations of terrestrial and freshwater species (Ward et al. 1994, Waples 1998). To address this challenge, evaluations of genetic population structure for pelagic marine fishes require large numbers of molecular markers and sampled individuals to achieve levels of statistical power suitable for resolving populations separated by low but biologically meaningful levels of genetic differentiation. Special consideration must also be given to the experimental design associated with population genetic studies of pelagic marine fishes,
224
wherein sampling efforts focused on source populations (e.g. larvae and/or reproductively active adults on spawning grounds) may be necessary to resolve population structure
(Graves et al. 1996, Waples 1998, Bowen et al. 2005, Carlsson et al. 2007, Graves &
McDowell 2015). In Chapter II of this dissertation, these principles (increased numbers of molecular markers and samples, as well as the inclusion of larval sample collections) are incorporated into an assessment of genetic population structure for white marlin
(Kajikia albida). Previous genetic assessments of this highly migratory pelagic species have included statistically significant comparisons of geographically distant sample collections, but have been unable to reject the null hypothesis of a single ocean-wide population. These studies have been based on small numbers of molecular markers and lack sample collections from known source populations, therefore the biological significance of genetic results has been uncertain. Relative to the results of previous studies, results from Chapter II include lower levels of genetic differentiation between geographically distant sample collections of white marlin. These results empirically demonstrate the importance of statistical power in facilitating accurate assessments of genetic population structure for pelagic marine fishes. Such insights are necessary for informing the biological interpretation of genetic results, which may have direct implications for the conservation and management of wild populations.
An additional challenge to the conservation and management of istiophorid billfishes is a lack of information on population structure for many of these species. A limited understanding of population structure prohibits the identification of biologically meaningful stocks for assessment and management. In addition, an understanding of population structure is necessary for conserving genetic diversity essential for short- and
225
long-term population persistence (Gaggiotti & Vetter 1999, Allendorf et al. 2008).
Factors such as these are especially important for istiophorid billfishes because stock assessments available for some of these species indicate that many istiophorid stocks are overfished and/or experiencing overfishing. In Chapter III of this dissertation, the genomic assessment of population structure for striped marlin (K. audax) throughout its range provides information on the presence of genetically distinct populations of this species within both the Pacific and Indian oceans. This information is especially important for the Indian Ocean because striped marlin is heavily overfished and experiencing overfishing in this region, and recent management actions have been impeded by a lack of information on stock structure.
Effective conservation and management of istiophorid billfishes is also challenged by uncertainty regarding species-level relationships for a number of istiophorids. An understanding of species relationships is necessary for identifying distinct evolutionary lineages characterized by unique genetic properties essential for species persistence. Such knowledge is not only important for improving our understanding of global biodiversity, but is a necessary first step for protecting genetic variation characteristic of species, and for studying evolutionary processes (e.g. speciation) responsible for observed genetic variation. In Chapter IV, long-standing uncertainty regarding the evolutionary relationship and taxonomic status of striped marlin and white marlin is clarified using genome-wide molecular markers. Results from
Chapter IV also include evidence for historical and/or contemporary gene flow between striped marlin and white marlin, corresponding with reports of inter-oceanic movements for istiophorids typically regarded as confined to distinct ocean basins.
226
Significance
Results from this dissertation are significant in providing practical information that can be incorporated into assessment and management plans for white marlin and striped marlin. Specifically, findings from the assessment of genetic population structure for white marlin are consistent with the single ocean-wide stock currently recognized by the
International Commission for the Conservation of Atlantic Tunas, though further study focused on ecologically (rather than genetically) distinct management units is warranted.
More broadly, results from this study provide practical insights on experimental designs, including statistical power and sampling strategy, appropriate for population genetic studies of pelagic marine fishes. Results from the genomic evaluation of population structure for striped marlin demonstrate inconsistencies between genetically distinct populations and stocks currently recognized for fisheries management, particularly in the
Indian and North Pacific oceans. In a broader context, findings from this study contribute to a growing body of literature on the use of newly developed genomic methods for resolving population structure in highly dispersive species. Results from the assessment of the evolutionary relationship of striped marlin and white marlin include verification of the distinct taxonomic status of these species after decades of uncertainty. These findings also highlight the utility of genomic methods for resolving relationships of closely related species which may not be readily distinguishable based on non-genomic characters. As a whole, results from this dissertation are significant in providing information necessary for conserving genetic diversity characteristic of striped marlin and white marlin, and for maintaining the evolutionary potential of these species.
227
The importance of collaborative scientific research is also a common theme to the work encompassed by this dissertation, and highlights the significance of public involvement in science. A large number of recreational anglers, scientists, and conservation and management agencies were fundamental in providing samples necessary for this research. Such collaborations are important for fostering a positive relationship between scientists and stakeholders, and enhances the outreach potential of this work.
Future Research Directions
While the results of this dissertation provide clarification regarding population- and species-level relationships of striped marlin and white marlin, these results also highlight several research areas to be addressed in future work. One particular question of interest is the exploration of mechanisms facilitating the presence of population structuring for striped marlin in both the Pacific and Indian oceans, but an apparent lack of population structuring for white marlin in the Atlantic Ocean. Differences in ocean-wide genetic connectivity for striped marlin and white marlin likely result from a number of factors, including the younger age and smaller size of the Atlantic Ocean relative to the Pacific
Ocean, and variation in biological characteristics such as thermal preferences, pelagic larval duration, and dispersal capabilities. To evaluate the role of environment in shaping observed patterns of genetic connectivity for striped marlin and white marlin, recent statistical advances in the emerging field of seascape genomics (Selkoe et al. 2008,
Benestan et al. 2016, Cooke et al. 2016, Liggins et al. 2016, Riginos et al. 2016) may be useful. Insights into environmental variables most conducive to gene flow in highly
228
migratory pelagic species may also enable the prediction of responses to a changing global climate. Similarly, the presence of population structuring for striped marlin implies some degree of spawning site fidelity, but this behavior has not been demonstrated for this or other closely related species. Electronic tagging studies which employ technology that enables multi-year tag deployment periods, such as internal archival tags (e.g. Block et al. 2005), are necessary to provide information on spawning site fidelity in istiophorid billfishes. Collectively, addressing questions related to the presence of genetic population structuring for striped marlin, and lack of such structuring for white marlin, will not only improve our understanding of barriers to gene flow in the marine environment, but may also provide practical insights into population structuring scenarios expected for other highly migratory pelagic fishes of commercial and/or recreational importance.
Another future research direction involves the exploration of evolutionary history for striped marlin and white marlin. Genomic data generated for these species in Chapter
IV can be evaluated in a context which enables the identification of evolutionary processes most relevant to shaping observed patterns of genetic variation. Evaluations of evolutionary history in other species have been useful for identifying modes of divergence in allopatric populations, including patterns of historical and/or contemporary gene flow (Sousa & Hey 2013, Filatov et al. 2016, Oswald et al. 2017). Similarly, given the distinct evolutionary lineages resolved for striped marlin and white marlin in Chapter
IV, evaluation of the relationship between Atlantic and Indo-Pacific populations of blue marlin and of sailfish based on genome-wide markers appears warranted. The generation of genomic data for these studies would also facilitate the assessment of evolutionary
229
history for blue marlin and sailfish. Comparison of evolutionary history across striped marlin, white marlin, blue marlin, and sailfish would provide valuable insights into the evolution of species pairs between the Atlantic and Indo-Pacific, and improve our understanding of the poorly understood process of speciation (for example, the relative contribution of genetic drift vs. natural selection) in highly dispersive pelagic species.
Finally, genome-wide single nucleotide polymorphisms (SNPs) identified as part of this dissertation represent the first application of genomic methodology to striped marlin and white marlin. The continued development of genomic resources for these and other highly migratory pelagic fishes will facilitate the assessment of questions previously unexplored for these species. Specifically, identifying genomic regions associated with phenotypic variation will provide information on the adaptation of highly migratory pelagic fishes to regional environmental conditions. Such information is also necessary for the conservation of adaptive genetic variation important for population persistence. Though the identification of putative adaptive loci is possible without detailed information on phenotype (for example, through FST outlier detection tests such as those used in Chapter III; Storz 2005, Stapley et al. 2010), insights possible with such approaches are relatively limited. Ideally, controlled tank experiments could be used to compare molecular responses to altered environmental conditions among individuals sampled from a range of environments (Narum et al. 2013, Defaveri & Merilä 2014, de
Villemereuil et al. 2016). Although this approach is not practical for many highly migratory pelagic species, it may be possible for some species (such as small tunas and/or juveniles of larger species; Estess et al. 2017), and insights derived from such studies
230
could eventually be used to inform targeted molecular comparisons among other species
(Bradbury et al. 2010, Pespeni & Palumbi 2013).
231
REFERENCES
Allendorf, F., England, P., Luikart, G., Ritchie, P., and Ryman, N. 2008. Genetic effects
of harvest on wild animal populations. Trends in Ecology and Evolution, 23: 327–
337.
Benestan, L., Quinn, B., Maaroufi, H., Laporte, M., Clark, F., Greenwood, S., Rochette,
R., and Bernatchez, L. 2016. Seascape genomics provides evidence for thermal
adaptation and current-mediated population structure in American lobster
(Homarus americanus). Molecular Ecology, 25: 5073–5092.
Block, B., Teo, S., Walli, A., Boustany, A., Stokesbury, M., Farwell, C., Weng, K.,
Dewar, H., and Williams, T. 2005. Electronic tagging and population structure of
Atlantic bluefin tuna. Nature, 434: 1121–1127.
Bowen, B., Bass, A., Soares, L., and Toonen, R. 2005. Conservation implications of
complex population structure: lessons from the loggerhead turtle (Caretta
caretta). Molecular Ecology, 14: 2389–2402.
Bradbury, I., Hubert, S., Higgins, B., Borza, T., Bowman, S., Paterson, I., Snelgrove, P.,
Morris, C., Gregory, R., Hardie, D., Hutchings, J., Ruzzante, D., Taggart, C., and
Bentzen, P. 2010. Parallel adaptive evolution of Atlantic cod on both sides of the
Atlantic Ocean in response to temperature. Proceedings of the Royal Society B,
277: 3725–3734.
Carlsson, J., McDowell, J., Carlsson, J., and Graves, J. 2007. Genetic identity of YOY
bluefin tuna from the eastern and western Atlantic spawning areas. The Journal of
Heredity, 98: 23–28.
232
Cooke, G., Schlub, T., Sherwin, W., and Ord, T. 2016. Understanding the spatial scale of
genetic connectivity at sea: unique insights from a land fish and a meta-analysis.
PLoS ONE, 11: e0150991. de Villemereuil, P., Gaggiotti, O., Mouterde, M., and Till-Bottraud, I. 2016. Common
garden experiments in the genomic era: new perspective and opportunities.
Heredity, 116: 249–254.
Defaveri, J., and Merilä, J. 2014. Local adaptation to salinity in the three-spined
stickleback? Journal of Evolutionary Biology, 27: 290–302.
Estess, E., Klinger, D., Coffey, D., Gleiss, A., Rowbotham, I., Seitz, A., Rodriguez, L.,
Norton, A., Block, B., and Farwell, C. 2017. Bioenergetics of captive yellowfin
tuna (Thunnus albacares). Aquaculture, 468: 71–79.
Filatov, D., Osborne, O., and Papadopulos, A. 2016. Demographic history of speciation
in a Senecio altitudinal hybrid zone on Mt. Etna. Molecular Ecology, 25: 2467–
2481.
Gaggiotti, O., and Vetter, R. 1999. Effect of life history strategy, environmental
variability, and overexploitation on the genetic diversity of pelagic fish
populations. Canadian Journal of Fisheries and Aquatic Sciences, 56: 1376–1388.
Graves, J., Gold, J., Ely, B., Quattro, J., Woodley, C., and Dean, J. 1996. Population
genetic structure of bluefin tuna in the North Atlantic Ocean. Identification of
variable genetic markers. ICCAT Collective Volume of Scientific Papers, 45:
155–157.
Graves, J., and McDowell, J. 2015. Population structure of istiophorid billfishes.
Fisheries Research, 166: 21–28.
233
Liggins, L., Treml, E., Possingham, H., and Riginos, C. 2016. Seascape features, rather
than dispersal traits, predict spatial genetic patterns in co-distributed reef fishes.
Journal of Biogeography, 43: 256–267.
Narum, S., Campbell, N., Meyer, K., Miller, M., and Hardy, R. 2013. Thermal adaptation
and acclimation of ectotherms from differing aquatic climates. Molecular
Ecology, 22: 3090–3097.
Oswald, J., Overcast, I., Mauck III, W., Andersen, M., and Smith, B. 2017. Isolation with
asymmetric gene flow during the nonsynchronous divergence of dry forest birds.
Molecular Ecology, 26: 1386–1400.
Pespeni, M., and Palumbi, S. 2013. Signals of selection in outlier loci in a widely
dispersing species across an environmental mosaic. Molecular Ecology, 22:
3580–3597.
Riginos, C., Crandall, E., Liggins, L., Bongaerts, P., and Treml, E. 2016. Navigating the
currents of seascape genomics: how spatial analyses can augment population
genomic studies. Current Zoology, 62: 581–601.
Selkoe, K., Henzler, C., and Gaines, S. 2008. Seascape genetics and the spatial ecology
of marine populations. Fish and Fisheries, 9: 363–377.
Sousa, V., and Hey, J. 2013. Understanding the origin of species with genome-scale data:
the role of gene flow. Nature Reviews Genetics, 14: 404–414.
Stapley, J., Reger, J., Feulner, P., Smadja, C., Galindo, J., Ekblom, R., Bennison, C., Ball,
A., Beckerman, A., and Slate, J. 2010. Adaptation genomics: the next generation.
Trends in Ecology and Evolution, 25: 705–712.
234
Storz, J. 2005. Using genome scans of DNA polymorphism to infer adaptive population
divergence. Molecular Ecology, 14: 671–688.
Waples, R. 1998. Separating the wheat from the chaff: patterns of genetic differentiation
in high gene flow species. The Journal of Heredity, 89: 438–450.
Ward, R., Woodwark, M., and Skibinski, D. 1994. A comparison of genetic diversity
levels in marine, freshwater, and anadromous fishes. Journal of Fish Biology, 44:
213–232.
235
VITA
Nadya R. Mamoozadeh
Born in Slippery Rock, Pennsylvania on April 13, 1986. Graduated from Slippery Rock
High School, Slippery Rock, PA in 2004. Earned a B.Sc. in Biology, with minors in
Chemistry and Marine Science, from Slippery Rock University of Pennsylvania in 2008.
Earned a M.Sc. in Marine Science from the University of North Carolina Wilmington in
2010. Entered the doctoral program at the Virginia Institute of Marine Science, College of William & Mary, School of Marine Science in 2013.
236