This item is only available as the following downloads:
1 POPULATION GENETICS, SPECIATION, AND HYBRIDIZATION IN DICERANDRA (LAMIACEAE), AN ENDEMIC OF THE SOUTHEASTERN USA By ADAM PAYTON A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DE GRE E OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2012
2 2012 Adam Payton
3 To my Parents
4 ACKNOWLEDGMENTS I thank my advisor Douglas Soltis f or giving me the opportunity to further my education in a field I am passionate about. His guidance and support have been critical for the completion of this project. I also thank the members of my advisory committee: Pam Soltis, Matt Gitzendanner, and Walter Judd who have significantly contributed to my ability to advan c e this study. I thank Robin Huck for her involvement in initially shaping the direction of this project and her lifelong dedication to the study of the diversity and evolution of this group of unique plants. I am grateful to Tom Patrick of the Georgia Di vision of Natural Resources and Alison McGee of the Nature Conservancy for their assistance with Dicerandra radfordiana I thank the staff of the University of Florida Herbarium, particularly Kent Perkins, and the members of the U.F. Biology Department fo r their knowledge, encouragement, and support. Finally, I thank my family and friends who have been patient and endlessly supporting throughout my graduate career, without which, this would not have been possible.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ............................. 9 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 Stud y System ................................ ................................ ................................ .......... 14 Study Organisms ................................ ................................ ................................ .... 15 Distribution ................................ ................................ ................................ ....... 15 Ecology ................................ ................................ ................................ ............. 17 Past Research on Dicerandra ................................ ................................ ................. 19 Taxonomic History ................................ ................................ ............................ 19 Biological Researc h ................................ ................................ .......................... 20 Motivation for this Study ................................ ................................ .......................... 20 Study 1: Dicerandra linearifolia Complex ................................ ................................ 22 Study 2: Dicerandra radfordiana and D. odoratissima ................................ ............ 25 2 MATERIALS AND METHODS ................................ ................................ ................ 34 Sampling Locations ................................ ................................ ................................ 34 454 Sequencing ................................ ................................ ................................ ...... 35 Microsatellite Discovery and Primer Design ................................ ............................ 36 Primer Testi ng ................................ ................................ ................................ ........ 37 Genotyping ................................ ................................ ................................ ............. 38 Chromosome Counts ................................ ................................ .............................. 39 Flow Cytometry ................................ ................................ ................................ ....... 40 Population Genetic Analyses ................................ ................................ .................. 41 Linkage Disequilibrium, Hardy Weinberg Equilibrium, Genetic Diversity .......... 41 Population Structure ................................ ................................ ......................... 43 Isolation by Distance ................................ ................................ ........................ 44 3 RESULTS ................................ ................................ ................................ ............... 49 Next Generation Sequencing ................................ ................................ .................. 49 Microsatellite Discovery, Primer Design, and Testing ................................ ............. 49 Genotyping ................................ ................................ ................................ ............. 50
6 Chromosome Counts ................................ ................................ .............................. 5 1 Flow Cytometry ................................ ................................ ................................ ....... 51 Locus Characteristics and Fragment Variation ................................ ....................... 52 Linkage Disequilibrium, Hardy Weinberg Equilibrium, and Heterozygosity ............ 54 Genetic Diversity Among Popu lations ................................ ................................ ..... 56 Population Structure ................................ ................................ ............................... 60 Hybrid Detection ................................ ................................ ................................ ..... 63 4 DISCUSSIO N ................................ ................................ ................................ ......... 77 Next Generation Sequencing for Microsatellite Discovery ................................ ...... 77 Genetic Diversity ................................ ................................ ................................ ..... 78 Population Structure and Species Delimitation ................................ ....................... 81 Conservation Implications ................................ ................................ ....................... 88 Future Research ................................ ................................ ................................ ..... 91 Conclusions ................................ ................................ ................................ ............ 92 APPENDIX: PAIRWISE POPULATION COMPARISONS ................................ ......... 96 LIST OF REFERENCES ................................ ................................ ............................. 103 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 108
7 LIST OF TABLES Table page 1 1 Character differences between D. linearifolia var. linearifolia D. fumella and D. linearifolia var. robustior ................................ ................................ ................. 29 1 2 Character differences between D. odoratissima and D. radfordiana .................. 30 2 1 Sampling locations ................................ ................................ ............................. 46 3 1 R eads with microsatellites containing 8 or more repeats ................................ .... 67 3 2 R eads with mi crosatellites containing fewer than 8 repeats ............................... 67 3 3 Description of the primer sequences, loci characteristics, and allelic complement ................................ ................................ ................................ ........ 68 3 4 Hardy Weinberg equilibrium p values, inbreeding coefficient ( F IS ), and observed and expected heterozygosity ( H o H e ) ................................ ................. 69 3 5 Genetic diversity metrics D EST and F ST ................................ ............................... 70 A 1 Pairwise F ST values ................................ ................................ ............................. 97 A 2 D ................................ ................................ ..... 100
8 LIST OF FIGURES Figure page 1 1 Approximate distribution of Dicerandra species ................................ .................. 31 1 2 Geographic distribution of Dicerandra populations used in current phylogeny ... 32 1 3 Current phylogeny ................................ ................................ ............................. 33 2 1 Locations sampled in Florida panhandle. ................................ ........................... 47 2 2 Locations sampled in Georgia ................................ ................................ ............ 48 3 1 Examples of microsatellite chromatograms ................................ ........................ 71 3 2 Chromosome images ................................ ................................ ......................... 72 3 3 Genetic distance based neighbor joining trees ................................ ................... 73 3 4 Mantel test results evaluating isolation by distance ................................ ............ 74 3 5 Structure analysis of all individuals ................................ ................................ ..... 75 3 6 Structure analysis of D. linearifolia var. linearifolia and D. odoratissima populations with a focus on the hyb rid ................................ ................................ 76
9 LIST OF ABBREVIATION S C Degrees Celsius 2 n Diploid chromosome number AMOVA Analysis of Molecular Variance BLAST Basic local alignment search tool bp Basepairs ca Circa CHR Dicerandra christmanii cm Centimeter COR Dicerandra cornutissima CTAB Hexadecyltrimethylammonium D D D. Dicerandra DAPI 4',6 diamidino 2 phenylindole DEN Dicerandra densiflora D EST D diH2O Deionized water dm Decimeter DNA Deoxyribonucleic acid DNR Division of Natural Res ources dNTP Deoxyribonucleotide triphosphate EDTA Ethylenediaminetetraacetic acid Eff. Effective et al. Et alia F IS Inbreeding coefficient
10 FLAS University of Florida Herbarium FNAI Florida Natural Areas Inventory FRU Dicerandra frutescens F ST F ixation index FUM Dicerandra fumella GC Guanine and cytosine GPB General purpose buffer H e Expected heterozygosity H o Observed heterozygoisty HWE Hardy Weinberg equilibrium HYB Hybrid IMM Dicerandra immaculata variety immaculata ITS Internal transcribed sp acer K Clusters km Kilometer LD Linkage disequilibrium LIN Dicerandra linearifolia variety linearifolia m Meter M Molar matK Maturase K gene MID Multiplex identifier mL Milliliter mm Millimeter mM Millimolar MOD Dicerandra modesta
11 NAD83 North American dat um 1983 ng Nanogram Num. Number ODO Dicerandra odoratissima PCR Polymerase chain reaction POP Population psi Pounds per square inch RAD Dicerandra radfordiana ROB Dicerandra linearifolia variety robustior s.l. Sensu lato SAV Dicerandra immaculata variety s avannarum sp. Species Std. Standard THI Dicerandra thinicola trnL Transfer ribonucleic acid L gene trnT Transfer ribonucleic acid T gene USFWS United States Fish and Wildlife Service UTM Universal Transverse Mercator v. Version var. Variety g Microgram L Microliter M Micromolar
12 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science POPULATION GENETICS, SPECIATION, AND HYBRIDIZATION IN DICERANDRA (LAMIACEAE), AN ENDEMIC OF THE SOUTHEASTERN USA By Adam Payton August 2012 Chair: Douglas Soltis Major: Botany Endemic organisms are valuable biologic and intrinsic components of ecosystems and the study of endemics can address a variety of evolutionary questions. The southeastern United States harbors a high number of endemics including the eleven species of the genus Dicerandra Recent phylogenetic analysis revealed accessions of the same species of D. linearifolia var. linearifolia a nd D. odoratissima from different geographic regions were not sister to each other. These phylogenetic discontinuities raised questions about the evolutionary history, relationships, and taxonomic circumscription within these species. This study used mic rosatellite based population level analyses of 42 populations from four of the annual species ( D. linearifolia var. linearifolia D. linearifolia var. robustior D. fumella D. odoratissima and D. radfordiana ), to address the questions of population struc ture, gene flow, and hybridization. Strong support was found for the species level recognition of the recently described D. fumella from the Florida panhandle with no evidence for hybridization with the proximate populations of D. linearifolia var. robust ior. Dicerandra linearifolia var. linearifolia showed some regional cohesion of populations, seen in STRUCTURE analyses and neighbor joining trees based on genetic distance, but there was no consistent
13 geographic pattern to the clustering of populations f or the recognition of regional populations at the species level. Dicerandra radfordiana showed consistent clustering in STRUCTURE analyses and distance trees with proximate populations of D. odoratissima Since D. radfordiana is found at the southeastern extreme of the D. odoratissima range these populations may represent the beginning of speciation by isolation. While there are morphological distinctions between D. odoratissima and D. radfordiana there is no molecular support for a distinct D. radfordia na entity. This may result from incomplete lineage sorting of alleles early in the speciation process. As a result the taxonomic circumscription of D. radfordiana at the species rank is in question. Overall, there is significant genetic diversity found a t the populations level for all Dicerandra annuals, 11.39% of the total genetic diversity is found among populations within species. This value is similar to the proportion of genetic variation found between species, 12.49%. The isolated nature of many p opulations, limited dispersal capability, and frequent visitation from generalist pollinators produce conditions where drift can cause significant differentiation between populations, even within the same geographic area. However the obligate outcrossing nature of these plants and the potential for multi generational populations resulting from recruitment from the seed bank are likely stabilizing factors maintaining heterozygosity within populations.
14 CHAPTER 1 INTRODUCTION Study System Endemic organisms are valuable biologic and intrinsic components of ecosystems. The study of endemics can address a variety of evolutionary questions. For example, analyses of endemics can provide insights into the recent evolutionary history of a region, as many endem ic s are the result of recent speciation and thus represent the leading edge of the evolutionary process. In contrast, endemics can be among the last remaining members of ancient communities and thus provide a glimpse into the past, showing what forms of lif e previously occupied a region. Endemics, especially those that are rare, threatened, or endangered, receive a relatively high degree of public interest and scientific study. Such studies guide conservation efforts, which must begin with a s trong understa nding of the life history of an organism. Ecology and demography are often the first aspects studied as they can reveal how an organism fits into and functions within a community. Much can be learned about the niche that a given endemic fills by utilizin g these approaches. Equally as important, but typically less studied, is the pattern and role of genetic diversity and population structure. Such studies provide insights into overall genetic differentiation, the degree of migration or admixture between populations, and the evolutionary history of an organism. Understanding genetic diversity and how that diversity is distributed across a landscape provides fundamentally different insights than ecological studies, but is equally important. The southeaster n Unite d States contains a large number of endemic plant species (Estill and Cruzan 2001) Many of these species occur in specific habitats, which are
15 also rare, such as scrub, sandhill, and coastal dune communities (Myers and Ewel 1990). Often studies a ddressing endemics focus on individual species that are members of genera/clades containing widely distributed species along with a few narrow endemics. This approach can yield insights into the underlying genetic factors such as hybridization, polyploidy and introgression, which can be major factors driving the evolution of rare species. We have chosen to investigate a clade composed solely of southeastern endemic plants. This approach facilitates the comparison of multiple closely related species to i nvestigate patterns of genetic diversity and structure, which in turn provides new insights into common factors or mechanisms that lead to the formation and maintenance of clades containing rare species. Understanding these speciation patterns and process es forms the foundation for understanding the origin of organismal diversity, and ultimately the biotic diversity of a region. We chose for investigation the annual clade of the southeastern endemic mint genus Dicerandra Benth. (Huck 1987, Oliveria 2007). We used a microsatellite based population genetic approach to investigate genetic variation, population structure, evidence for hybridization, and degree of differentiation between members of this annual clade Study Organisms Distribution Dicerandra (Lam iaceae) is endemic to the southeastern United States. First describe d by Bentham (1848 ), Dicerandra is characterized by spurs on the anther sacs. character remains a significan t distinguishing synapomorphy of all species in the genus.
16 Dicerandra currently comprises 11 species, of which two have varietal designations (Huck 1987, 2010). There are five annual species [ D. densiflora D. fumella D. linearifolia (composed of var. li nearifolia and var. robustior ), D. odoratissima and D. radfordiana ] and six perennial species [ D. christmanii D. cornutissima D. frutescens D. immaculata (composed of var. immaculata and var. savannarum ) D. modesta and D. thinicola ]. The perennial sp ecies are restricted to ancient dune ridges in central peninsular Florida and inland Atlantic coastal dunes. Many perennials have very narrow ranges ( Figure 1 1) small population sizes, and have experienced decrease and fragmentation of critical habitat. As a result, all of the perennials except D. thinicola are listed as federally endangered (US FWS 2007 ). The combined range of the annual species of Dicerandra is considerably wider than that of the perennials. They occur from the Florida panhandle through southeastern Alabama and southern Georgia. Their northern limit is at the Fall Line of Georgia, a geologic barrier separating the hard crystalline rock of the Piedmont from the sedimentary deposits of the Coastal Plains. All annual species, with the exc eption of D. radfordiana have wide distributions occupying hundreds of square miles; however, populations are often isolated from one another creating a sparse local distribution. Dicerandra radfordiana has only two known populations in McIntosh Co., Geo rgia. In eastern Georgia the ranges of D. linearifolia var. linearifolia and D. odoratissima overlap, and rare hybrids have been described from this contact zone where populations of the two species are adjacent or intermixed. In the Florida panhandle D. fumella D. linearifolia var. robustior and D. densiflora have parapatric distributions with
17 a proposed hybridization zone between D. fumella and D. linearifolia var. robustior in the central panhandle ( Figure 1 1) (Huck 2010). Ecology All Dicerandra speci es are found on deep, well draining, sandy soils, typical of the coastal plain of the southeastern U.S. (Huck 1987). The soil, quartzipsamments, consists of deep, unstructured, acidic sand deposits with little to no organic horizon and is low in available nutrients especially phosphorous (Huck 1987, Myers and Ewel 1990). Disturbance is a critical component for persistence of populations. A significant increase of individuals is often seen following perturbation of a site, mechanical or pyrogenic (Huck 19 87, Menges 1992, personal observation). Annual species are almost exclusively found in disturbed areas near road cuts, power line rights of way, firebreaks, or trails. Sites are typically characterized by having a sparse forb and graminoid community, min imal to no overstory canopy cover, and low water availability. Plants can be found within natural openings in former sandhill or high pine communities but are more common in disturbed areas adjacent to dense, closed canopy, later successional variants of sandhill or high pine (xeric hammock or turkey oak barrens) communities or adjacent to pine plantations presumably occupying former sandhill or other high pine community types (Myers and Ewel 1990, FNAI 2010). Dicerandra species are considered obligate out crossers. Self compatibility experiments showed 1.9% of crosses resulted in mature flowering plants, while selfing studies showed 0.7% of selfed offspring reached maturity (Huck 1987). Artificial hybridization experiments involving all Dicerandra species revealed a low frequency of viable interspecific offspring, approximately 12%, with the vast majority resulting from crosses between annual species (Huck 1987). In most cases the
18 allopatric distribution of species prohibits reproductive contact; however, several contact zones have been considered to contain natural hybrid individuals. In eastern Georgia D. linearifolia var. linearifolia has been observed to hybridize with D. odoratissima (Huck 1987, personal observation), and in the Florida panhandle hybr ids between D. fumella and D. linearifolia var. robustior have been proposed (Huck 2010). Dispersal capability of Dicerandra is thought to be limited. The pericarp wall is coved with hydrophobic mucilaginous cells, which provide buoyancy to seeds, suggest ing that water is an important dispersal mechanism (Huck 1987). This is corroborated by many populations of Dicerandra being located near waterways or on sand ridges deposited by ancient waterways and may provide a mechanism to explain the overall distrib ution of species from a historical standpoint ( Figures 2 1 & 2 2). Limited dispersal is also supported by phylogenetic data, which show a strong correlation between clades of annual species and the watershed in which they occur (Oliveira et al. 2007). Yet localized dispersal by water is likely limited to precipitation events of sufficient volume to create sheet flows extending beyond the population perimeter. While precipitation events of this scale do occur, the isolated nature of many populations and the ir relatively small sizes suggest long distance seed dispersal is limited. These observations suggest that the primary mechanism for genetic exchange between populations is pollinator mediated. However, relatively little is known about the pollinators fo r the annual species. Various insects have been observed systematically visiting flowers, with the majority being generalists, such as the cloudless sulphur butterfly, Gulf fritillary, and various bees (personal observation, Huck 1987), but detailed
19 study would be required to determine the precise composition and frequency of visits by these species and their efficiency as pollinators. Past Research on Dicerandra Taxonomic History The first description of Dicerandra was a specimen of D. linearifolia var. l inearifolia, from eastern Georgia, collected by Stephen Elliot in 1821. Originally Ceranthera linearifolia the n ame was changed by Bentham (1848 ) to Dicerandra The description of additional species continued over the next century with the genus receivi ng taxonomic treatment by Shinners (1962) and then Kral (1982). Huck (1987) provided the first thorough investigation of the taxonomy, phylogenetics, reproductive biology, and extent of hybridization. This work resulted in the recognition of two sections within the genus. Section Dicerandra contained all of the perennial species as well as D. linearifolia var. linearifolia D. linearifolia var. robustior and D. densiflora Section Lecontea contained the remaining two species, D. odoratissima and D. rad fordiana These sectional designations were based primarily on a suite of floral characters and their primary type of secondary compound; either menthol based, which produced a minty odor, or cineol based, which produced a cinnamon like odor. A molecular phylogenetic analysis (Oliveira et al. 2007) showed the genus was in fact composed of two sister clades; one contained the perennials, redesignated as subgenus Kralia and the other the annuals, designated subgenus Dicerandra Huck (2010) elevated the Flor ida panhandle populations that had been referred to as D. linearifolia var. linearifolia to the rank of species, naming them D. fumella based on phylogenetic results in which the panhandle populations of D. linearifolia var. linearifolia
20 were in a separat e clade from the populations found in Georgia (Oliveira et al. 2007) and morphological characteristics (Huck 2010). Biological Research Due to conservation concerns, the perennial species have received considerable research attention with studies investiga ting reproductive and pollination biology (Evans et al. 2004), production of secondary leaf compounds (McCormick 1993, Huck 1989), microhabitat preference and response to disturbance (Menges et al. 1999, Menges 1992), and genetic diversity across scrub end emics in past and present landscapes (McDonald and Hamrick 1996, Menges et al. 2001, Menges 2010). The annuals have received less attention, although D. linearifolia was the focus of a series of studies investigating development and plasticity of leaf var iation (Winn 1996a, 1996b, 1999) and functional trait comparisons ( Merchant 2006 ). Both perennials and annuals were included in a chromosome analysis showing 2 n =32 ( D. cornutissima D. christmanii D. frutescens D. linearifolia var. robustior D. odorati ssima D. radfordiana and D. thinicola ) and 2 n =48 in ( D. densiflora D. linearifolia var. linearifolia ) with the 2 n =48 species considered to be the result of recent genome duplication events within the genus (Huck and Chambers 1997). Motivation for this S tudy The phylogenetic analysis performed by Oliveira et al. (2007) clearly separated annuals and perennials into distinct clades; however, it provided limited resolution in the annual clade especially within geographic regions ( Figure 1 2). Within the annua ls, three clades were recovered which correspond to major geographic regions; river basins draining to either the Atlantic Ocean or Gulf of Mexico and the highlands of west
21 central Georgia. While regional geographic structure was evident, polytomies invol ving multiple species were common within the clades ( Figure 1 2). The analysis of Oliveira et al. (2007) also revealed several instances where accessions of the same species from different geographic areas did not form a clade ( Figure 1 2). While these result s are not particularly uncommon in species level phylogenies when multiple populations of a species are sampled (Comes and Abbott 2001, Jakob and Blattn er 2006, Syring et al. 2007, Ramdhani et al. 2011), they raise questions regarding the mechanisms respon sible for such phylogenetic discontinuities and the validity of current taxonomic circumscriptions. Factors such as incomplete lineage sorting or gene flow from ancient or recent hybridizations can result in intraspecific non monophyly (Edwards et al. 200 8). Distinguishing between incomplete lineage sorting and hybridization is often difficult if not impossible (Rieseberg and Soltis 1991, Wendel and Doyle 1998) yet progress is being made with the development of new analytical methods to address these ques tions (see Joly et al. 2009, Yu 2011). However, many currently used loci for phylogenetic analysis cannot determine the degree of gene flow between species or populations because they often lack sufficient variability to discriminate between recently dive rged species or clades. Assessing the extent of gene flow or degree of isolation is a critical component of many species concepts and thus is necessary for addressing species boundaries (Wheeler and Meier 2000). Population level studies with highly varia ble markers are the best way to determine admixture, migration, and population structure, all of which may ultimately aid in determining species limits.
22 This study aims to address the discontinuities observed in the phylogeny of Oliveira et al. (2007) util izing microsatellite markers and population level analyses. With this approach, we will use analyses of population structure, divergence, and gene flow to address species boundaries. Study 1: D icerandra linearifolia C omplex The most widespread and morphol ogically diverse annuals are members of the Dicerandra linearifolia complex. This complex is composed of D. linearifolia var. linearifolia D. linearifolia var. robustior, and the recently named D. fumella (Huck 2010). Dicerandra linearifolia var. linear ifolia occurs in west central Georgia and eastern Georgia, D. linearifolia var. robustior is primarily found in the eastern panhandle of Florida between the Apalachicola and Suwannee Rivers, extending into southern Georgia, and D. fumella occurs primarily in the Florida panhandle west of the Apalachicola River and into southern Alabama ( Figure 1 1) (Kral 1982, Huck 1987, 2010). To frame the questions for this study, consider the phylogenetic analysis of Oliveira et al. (2007) with the Florida panhandle popula tions of D. linearifolia var. linearifolia relabeled to reflect the recent taxonomic revision to D. fumella ( Figure 1 3). The phylogenetic tree shows the Florida panhandle taxa, D. fumella and D. linearifolia var. robustior forming a clade ( Figure 1 3). Int erestingly, population 2, representing D. fumella has identical DNA sequences for the markers used as population 8, identified as D. linearifolia var. robustior The voucher specimens for these populations were consulted to verify proper identification o f these species, and despite identical sequences, they are properly identified based on the morphological characters used to delimit each species. Additionally, the voucher specimen for population 3 labeled as D.
23 linearifolia var. linearifolia in the Oliv eira et al. (2007) phylogeny, has subsequently been determined to be D. linearifolia var. robustior and is thus relabeled to reflect this change in identification. However, even with these changes, D. linearifolia var. linearifolia remains unresolved beca use of population 4, from western Georgia ( Figure 1 3 ). We were interested in investigating at a population level the degree of molecular support for species designation of D. fumella and the relationship of the Georgia populations of D. linearifolia var l inearifolia Recognition of D. fumella as a species was based on its phylogenetic placement in a separate clade from the Georgia populations and on morphological differentiation. However, the morphological similarity between D. fumella and D. linearifolia var linearifolia is rather striking upon first examination. The differences within the following characters are subtle, but have been cited by Huck (2010) as distinct for each species: leaf width, inflorescence architecture, number of flowers per cyme, corolla upper lobe shape, corolla tube length, and anther color (Table 1 1). A greater morphological distinction exists between D. fumella and D. linearifolia var. robustior despite their parapatric distribution in the Florida panhandle and identical seq uences at the surveyed loci (Table 1 1). However, sufficient morphological intergradation occurred near the geographic contact point between these two species for Huck (2010) to propose the existence of a relatively wide hybrid zone in the central panhand le of Florida. Within D. linearifolia var linearifolia the west central and eastern populations did not form a clade, and hybrids with D. odoratissima have been observed at the eastern contact zone where populations of the two species are found in proxim ity or are intermixed (Huck 1987). These rare hybrids display either an intermediate or chimeric
24 morphology between the two otherwise distinct parental species and show 90% pollen viability (Huck 1987). To address more thoroughly the phylogenetic incongru ence within the D. linearifolia complex and assess the molecular signature of hybrids, sampling was increased to represent the full ranges of D. fumella D. linearifolia var. linearifolia and D. linearifolia var. robustior and D. odoratissima Sampling included the proposed hybrid zone in the Florida panhandle and the two hybrid populations in eastern Georgia. Twenty populations totaling 265 individuals were genotyped at eight highly variable microsatellite loci to determine population structure, extent of gene flow, and hybridization. Polyploidy has played a role in the evolution of Dicerandra (Huck and Chambers 1997). Chromosome analysis of meiotic pollen cells indicated differing chromosome numbers within the annuals: D. fumella 2 n = 48, D. linearifo lia var. robustior 2 n = 32, with no available data for D. linearifolia var. linearifolia (Huck and Chambers 1997, Huck 2010). These high chromosome numbers for diploids (2 n = 32) indicate the genus is likely the descendent of an ancient genome duplication event in an ancestral lineage having 2 n = 16. While Huck and Chambers (1997) included all species recognized at the time, their counts were based on a limited geographic range and a small number of samples per species, in many cases only one. Thus, the actual extent of polyploidy within the annuals remains unknown. To address the ploidy of the annual species of Dicerandra in greater detail, live plants were collected and chromosome counts were performed from mitotically dividing root tip cells of D. lin earifolia var. robustior (2 n = 32) and D. densiflora (2 n = 48). Flow cytometry of silica dried leaf tissue was then used to
25 screen a wider sample of individuals within the D. linearifolia complex. Verification of chromosome counts and denser sampling all ow for a better understanding of the distribution of cytotypes and can provide increased insight into the role of genome duplication in the evolu tionary history of the annuals. The non monophyly of species in the phylogeny of Oliveira et al. (2007) may ref lect species divergence within the D. linearifolia complex, with some regions potentially representing independently evolving lineages with significant geographic D. fumella ). T h e disparate populations contained in these geographic regions may merit taxonomic recognition as distinct entities. To investigate these hypotheses, a fine scale molecular analysis was necessary to: 1) assess the genetic structure within populations o f D. linearifolia var linearifolia D. fumella, and D. linearifolia var. robustior 2) determine if population genetic structure corresponds to geographic distribution and phylogenetic relationships, 3) evaluate the hypothesized hybrid zone between D. fum ella and D. linearifolia var. robustior as well as between D. linearifolia var linearifolia and D. odoratissima, 4) determine the extent of polyploidy within the complex, 5) assess species delimitations and potential conservation implications, and 6) give further insights into the evolutionary patterns of this clade of regional endemics. Study 2: D icerandra radfordiana and D odoratissima Dicerandra radfordiana has an extremely narrow distribution with only two recorded populations in McIntosh Co., Georgia Its range could extend farther north within coastal plain communities but the definitive extent of its range remains unknown (Robin Huck and Tom Patrick Georgia DNR lead botanist, pers. comm.). Dicerandra odoratissima is widespread in eastern Georgia a nd extends north to the southernmost
26 county of South Carolina, Jasper. Dicerandra radfordiana and D. odoratissima are morphologically divergent from other members of the genus, sharing most notably the distinct floral synapomorphies of tubular corollas, c ucullate superior corolla lobe, slightly protruding anthers, and distinct cineol based secondary compounds that producing the scent of cinnamon. All other annual species of Dicerandra display sharply geniculate, funnel shaped corollas, an erect superior c orolla lobe, strongly protruding anthers, and menthol based secondary compounds that produce the scent of mint (Huck 1987). The morphological differences between D. odoratissima and D. radfordiana are primarily based on scale, with D. radfordiana having v ery similar floral and leaf architecture to D. odoratissima, but with larger dimensions (Table 1 2). Different pollinators have been suggested for D. radfordiana and D. odoratissima hawkmoth and bee respectively (Huck 1987) but studies specifically addre ssing pollinators are needed. The chromosome analysis by Huck and Chambers (1997) showed that D. odoratissima and D. radfordiana have chromosome counts of 2 n = 32. Currently, D. radfordiana is recognized as an endangered species by the state of Georgia a nd is listed federally as a category 2 species of concern with an unknown trend (USFWS 1993), with the need for additional work to establish range limits, assess population diversity, and verify species rank. As with D. linearifolia var linearifolia the a nalyses of Oliveira et al. (2007) revealed that D. odoratissima is non monophyletic ( Figure 1 2). The D. radfordiana sample used had identical sequences (ITS, matK and trnT trnL ) to the most geographically proximate population of D. odoratissima and these samples shared one chloroplast and two nuclear synapomorphies. The remaining two populations of D. odoratissima had
27 unique, yet identical sequences. Geographically the three populations of D. odoratissima spanned the range of the species with considerabl e distance between the three sampled populations: 74, 81, and 110 km (Oliveira et al. 2007). However, the distance between D. radfordiana and the closest sampled population of D. odoratissima was only 750 m, and these two populations had identical sequenc es and are found at the southeastern extreme of range. Multiple hypotheses could account for the observed non monophyly and identical nucleotide sequences noted above. The shared sequences of the population of D. odoratissima and D. radf ordiana could result from recent divergence of D. radfordiana from D. odoratissima with incomplete lineage sorting producing the observed similarity in DNA sequences. Alternatively, introgression of D. radfordiana genes into neighboring D. odoratissima th rough hybridization and backcrossing could also produce these phylogenetic results. Both of these mechanisms could produce DNA sequence similarities, yet maintain the morphological differences used to define these two entities as distinct species. Anothe r hypothesis for the sequence similarities is clinal variation or isolation by distance. D icerandra radfordiana occurs at the very southeastern portion of D. odoratissima range and may appear as a distinct entity when compared morphologically or molecul arly to individuals from the no rthern or western portion of the range. But when compared to geographically proximate populations, those distinguishing characteristics diminish. Given the small sampling from both species in the phylogeny of Oliveira et al (2007), addressing these hypotheses requires broader geographic sampling and finer scale molecular analyses.
28 We sampled 175 individuals from 10 populations of D. odoratissima spanning 150 km between the northern and southernmost samples and both known D. radfordiana populations Seven highly variable microsatellite loci were developed to investigat e the relationship between D. radfordiana and D. odoratissima allowing us to assess: 1) if there is unique and significant genetic structure between D. radfor diana and D. odoratiss ima 2) the degree of gene flow between geographically proximate populations of D. radfordiana and D. odoratissima 3) the extent of isolation by distance, and 4) the impli cations of population structure or hybridization on specie s de limitation and conservation.
29 Table 1 1. Character differences between D. linearifolia var. linearifolia D. fumella and D. linearifolia var. robustior (modified from Huck ( 2010 ) Dicerandra fumella (Lamiaceae), a new species in the Florida panhandle and adjacent Alabama, with comments on the D. linearifolia complex. Rhodora, 112, 215 227 ). Character D. linearifolia var. linearifolia D. fumella D. linearifolia var. robustior Leaf shape Linear Linear Obovate, rhombic Leaf width 0.5 1 mm 1 5 mm 2 10 mm L eaf vestiture Rugose, hispid Rugose, hispid Smooth Inflorescence Simple Simple or compound Compound Inflorescence aspect Open verticils Very open verticils Compact or loose verticils Flowers per cyme 1 2 1 7, usually 3 5 3 7 Corolla upper lobe shape, size ratio in upright position Ovate, appearing wider than tall Cordate, appearing taller than wide Cordate, appearing taller than wide Corolla tube length 4 5 mm 6 7 mm 6 7 mm Corolla color White, pale pink, rarely deep pink White, cream, pale pink, sal mon pink, pale lavender, very rarely magenta Vivid reddish purple, vivid purplish red, magenta, deep pink, rarely pale pink Anther color Yellow Yellow, orangish yellow, streaked with reddish purple Reddish brown, maroon
30 Table 1 2. Character differenc es between D. odoratissima and D. radfordiana Measurements of leaf size, corolla length, tube length, and orifice width are means with ranges in parentheses (modified from Huck ( 1987 ) Systematics and evolution of Dicerandra (Labiatae). Phanerogamarum mon ographiae, 19, 1 343 ). Character D. odoratissima D. radfordiana Plant height Up to 5 dm Up to 6.5 dm Habit Often bushy Virgate Leaf shape Linear Narrowly obovate Leaf size 25(15 45) mm x 1(1 3) mm 40(19 55) mm x 4(2 6) mm Corolla length 14(14 21) mm 32(27 44) mm Tube length 11(11 18) mm 25(20 32) mm Orifice width 2 mm 4 mm Pistil length ca. 35 mm ca. 68 mm
31 Figure 1 1. Approximate distribution of Dicerandra species. Species with very limited ranges are represented by points.
32 Fig ure 1 2 Geographic distribution of Dicerandra populations sa mpled in Oliveira et al. (2007). This cladogram is from a maximum likelihood analysis of the combined nuclear and chloroplast data sets. Numbers on branches correspond to sampling locations on the ma p; where several samples of the same species had identical sequences, they are represented as multiple numbers on a single branch. Bootstrap values are in parentheses. Two major clades correspond to the northern annuals and the southern perennials, respe ctively, with further geographic subdivision between the Atlantic and Gulf of Mexico drainages within the annual clade. Note the non monophyly of both D. linearifolia var. linearifolia (LIN) and D. odoratissima (ODO) (Figure modified from Oliveira et al. ( 2007 ) Molecular phylogeny, biogeography, and systematics of Dicerandra (Lamiaceae), a genus endemic to the southeastern United States. American Journal of Botany, 94, 1017 1027 ).
33 Figure 1 3. Oliveira et al. (2007) phylogeny with D. linearifolia var. linearifolia populations from the Florida panhandle relabeled to reflect the taxonomic revisions of Huck (2010) ( Dicerandra fumella (Lamiaceae), a new species in the Florida panhandle and adjacent Alabama, with comments on the D. linearifolia complex. Rhod ora, 112, 215 227) Taxa differing from original phylogeny are indicated in red. Population 3, formerly labeled as D. linearifolia var. linearifolia in the Oliveira et al. phylogeny, was relabeled as D. linearifolia var. robustior to reflect subsequent t axonomic revision to the original voucher specimen. Note the D. fumella (FUM) specimen, population 2, has an identical sequence as a specimen of D. linearifolia var. robustior population 8. (Figure modified from Oliveira et al. (2007) Molecular phylogeny, biogeography, and systematics of Dicerandra (Lamiaceae), a genus endemic to the southeastern United States. American Journal of Botany, 94, 1017 1027).
34 CHAPTER 2 MATERIALS AND METHODS Sampling Locations Forty seven populations were collected from Florid a, Georgia, and southern Alabama ( Fig ure s 2 1 & 2 2). The number of individuals collected at each population varied based on overall size and collecting conditions; 38 populations had 16 individuals sampled, and 9 populations had 8 individuals sampled (T able 2 1). Due to the annual habit, lower dispersal capability, and habitat specificity of these Dicerandra species, populations are almost always confined to areas of several hundred square meters or less with easily demarcated boundaries. Thus, for thi s study a population is defined as a sampling area where plants were determined to be consistently distributed. Sampling locations were initially identified through herbarium collections with additional locations discovered through exploration of similar h abitat in areas surrounding known samples. Collection took place from August through November 2010 and 2011, with the majority of sampling taking place in 2010. Upon arrival at a site, an outer perimeter was identified, general overview photographs were taken, and leaf material was collected from up to 16 individuals throughout the site. Leaf tissue was immediately stored in silica desiccant. Voucher specimens were collected at all sites and typically included both flowering and vegetative individuals. If flowering individuals were absent at the initial time of collection, sites were revisited to ensure sampling of identifiable individuals. Voucher specimens were deposited at the University of Florida herbarium (FLAS). Site locations were recorded util izing GPS derived Universal Transverse Mercator (UTM) coordinates in the North American Datum 1983 (NAD 83).
35 454 Sequencing Genomic DNA was isolated from a single individual of D. linearifolia var. robustior using a modified CTAB extraction procedure (Doyle and Doyle 1987 ). In order to produce the desired quantity of 4 ug of high molecular weight DNA, multiple extractions were performed, the DNA was cleaned using a Geneclean Turbo kit, and individual extractions were combined and conce ntrated by sodium Hind III ladder showed DNA fragments were between 8,000 11,500 base pairs in length. Nanodrop (Thermo Fisher Scientific, Inc. Waltham, MA, USA) analysis indicated sample conce ntrations of 360 ng/ L with ratios of DNA to proteins and organic compounds above the desired threshold of 2. Based on prior studies using next generation sequencing for microsatellite discovery, it was determined that a full lane (1/8 th of a sequencing pl ate) was not necessary to generate the desired quantity of data (Castoe et al. 2010). Thus, the Dicerandra sample was barcoded during library preparation and combined with a sample of the monocot Nolina bri t toniana (Nolinaceae), and pooled to run on 1/8 th of a sequencing plate. Unique MID barcodes were ligated onto the genomic DNA during library construction, prior to pooling of the two samples, allowing for accurate identification and separation of data for the two species. Sequencing was performed on a Roche 454 GS FLX utilizing Titanium chemistry (Roche Diagnostics Corp., Indianapolis, IN, USA). Following sequencing, the data were separated by barcode using FASTX Barcode splitter (Gordon 2008), and the barcodes were removed using FASTX Trimmer (Gordon 2008).
36 Microsatellite Discovery and Primer Design Microsatellite discovery and initial primer design were accomplished by utilizing a modified Perl scrip originally published by Castoe et al. (2010), and modified by M. Gitzendanner (University of Florida). This script searched the sequence data and identified repeated units of nucleotides, indicating the presence of a microsatellite region. Primer3 (Rozen and Skaletsky 2000) was then run, and primers flanking the repeat were generated. For this study di, tri, and tetra nucleotide repeats were targeted, with a minimum repeat number of 8. Primers were excluded from the 35 base pair region flanking each microsatellite. Primer design specifications, as allowed by Primer3, were as follows: optimum primer siz e: 20 (range: 18 30), product size range: 60 400, primer GC content 40 60%, GC clamp: 2 bp, optimal melting temp: 62 (range: 58 65), maximum self complementarity: 4, maximum pair complementarity: 4, maximum end complementarity: 3, maximum homopolymers: 3. The selected microsatellite repeats and their primers were further screened manually to ensure the microsatellites were perfect repeats, contained only short homopolymer runs (<8), and had high quality base calling scores from the 454 software. Primers co ntaining greater than 3 base pairs with a low quality score were discarded. This screening produced 101 loci, 46 of which were chosen for PCR amplification testing across all species. To ensure microsatellite loci were not located in known genic regions o r in the chloroplast genome the 101 candidate loci were subjected to a BLAST search in GenBank. Any loci showing greater than 80% similarity with a known coding region or to a region within the chloroplast genome would be discarded. This approach ensured
37 that loci would be as neutrally evolving as possible, avoiding the potential for selection driving false population structure for a given locus. Primer Testing Primers selected for amplification testing had an M13 label, genotyping utilizing fluorescently labeled probes. Primers were tested against a standardized panel of DNA composed of samples from D. linearifolia var. linearifolia, D. linearifolia var. robustior D. odoratiss ima and D. radfordiana Initial testing took place with M13 primers not labeled with the florescent probe. All primers were tested for amplification twice against the same panel under the same PCR conditions to assess reliability of primer pairs. PCR re actions were carried out in 10 L volumes consisting of 1.5 L H 2 O, 2 uL 5 M betaine, 2 L 5xGoTaq buffer, 0.8 L 25 mM MgCl 2 0.6 L 0.8 uM forward primer, 0.9 L 5 M reverse primer, 0.9 L 5 mM fluorescently labeled M13 primer (FAM, VIC, NED, or PET), 0 .2 L 2.5 mM dNTP, 0.1 L Taq polymerase, and 1 L 15 ng/ L DNA sample. Reactions were carried out under the following conditions: 1) 3 min at 94C, 2) 30 sec at 94C, 3) 30 sec at 52C, 4) 45 sec at 72C, 5) 35 cycles of steps 2 4, and 6) 20 min at 72C Initial reactions were performed in 96 well plates, and subsequent reactions used 384 well plates, once successful loci were identified. Amplification of microsatellites developed in one species which are then used across several species may experience some degree of allele dropout, where amplification consistently fails in a particular species but is successful in others. To maximize the benefit of successfully amplifying loci, even those not amplifying in every species, loci were divided into three ca tegories based on the questions they were
38 capable of addressing. The first category was loci that amplified in all species of interest, the second group was those that amplified in at least the species of the D. linearifolia complex, and the third group c onsisted of those that amplified in at least D. odoratissima and D. radfordiana This approach ensured there was a set of loci that could be used for addressing similarity across all species, as well as a potentially larger subset that could be used to ad dress similarity on a finer scale, such as between two taxa or between geographic regions of a single taxon. Genotyping All samples were amplified via PCR with fluorescently labeled primers that allow for detection of PCR product length by sequencing instr uments (Schuelke 2000). Samples were genotype on an ABI 3730XL (Life Technologies Corporation, Carlsbad, CA, USA), which can detect five fluorescent dyes, one for an internal size standard and four for the sample, allowing multiple loci of a single individ ual to be amplified and then pooled together for analysis. Further pooling of loci can take place by using the same dye if there is a significant length difference in the PCR products (Levitt et al. 1994). When the same dye is used for loci with differin g PCR fragment sizes, the resulting chromatogram will display the allele size peaks of each locus with sufficient space separating the two to allow for confident assignment of alleles to their proper locus. Using this procedure, PCR products for 8 loci of a single individual were pooled and then genotyped. Due to the phenomenon of dye shift, where the observed allele size changes solely as a result of the fluorophore used, a single dye was chosen for each locus and used consistently throughout the study ( Sutton et al. 2011). PCR products were diluted prior to genotyping based on the strength of amplification as observed from agarose gel electrophoresis and inherent dye intensity;
39 VIC and FAM are stronger fluorophores than PET and NED. In general, loci amp lified with VIC and FAM were diluted 1 L PCR product in 39 L diH 2 O to a final volume of 40 L and loci amplified with PET and NED were diluted 4 L PCR product in 36 L diH 2 O to a final volume of 40 L Genotyping was performed by the Interdisciplinary Center for Biotechnology Research at the University of Florida. GeneMarker v .1.6 (SoftGenetics LLC, State College, PA, USA) was used to visualize chromatograms and design automated allele calling panels. Additionally, all allele calls were verified manua lly. Chromosome Counts Huck and Chambers (1997) performed meiotic chromosome counts from dividing microsporocytes and described multiple ploidal levels within the annual species of Dicerandra To verify ploidy across species, as well as to calibrate flow cytometry analyses, chromosomes from dividing root tip cells from D. densiflora and D. linearifolia var. robustior two species with different published ploidal levels (Huck and Chambers 1997), were counted. Live specimens of D. densiflora (populations 46 and 47) and D. linearifolia var. robustior site 7 reared in the Department of Biology greenhouse, with vo ucher specimens deposited at FLAS. Root tips were collected from vigorously growing individ uals and pretreated by exposure to nitrous oxide at 15 0 psi for 1.5 hours in a pressurized chamber. Pretreatment disrupts the development of microtubules, resulting in an accumulation of cells arrested at metaphase with condensed chromosomes capable of be ing visualized. Root tips were then fixed in 90% glacial acetic acid for 10 minutes, rinsed with 70% ice cold ethanol, and stored in 70% ethanol at 20C. To prepare the root tips for enzyme
40 digestion, the samples were washed for 20 minutes in 10 mL of 1 x citric buffer, repeated three times. The terminal 1 2 mm of the root tip was removed and placed in a 20 L solution of 1% pectinase and 2% cellulase. The digestion incubated for 20 min at 37C. Root tips were then rinsed with 70% ice cold ethanol, 20 L of glacial acetic acid were added, and the tips were broken up using a blunt dissecting probe. Three microscope slides were placed in a high humidity chamber, and 3.5 L of the glacial acetic acid root tip slurry were dropped onto each slide. Slides w ere allowed to dry over the course of 10 minutes. Chromosomes were stained with DAPI and visualized using a Zeiss Axio Imager M2 microscope (Carl Zeiss Microscopy LLC, Thornwood, NY, USA). To ensure accuracy, counts were only made from ruptured cells that were isolated and had few to no overlapping chromosomes. Each slide produced a sufficient quantity of isolated cells with highly condensed chromosomes so that a minimum of 3 cell counts could be made per slide. Flow Cytometry To assess the ploidy of a la rge number of populations, without requiring the rearing of live material for root tip counts, flow cytometry was used to assess the genome size of individuals from populations spanning each species distribution. Tissue was prepared following the procedur es outlined in Suda et al. (2006). A critical aspect of successfully extracting intact nuclei is pairing the species of interest with an appropriate buffer. Four buffers were tried with varying degrees of success; extraction Gruihuber et al. 2007 ) were either inconsistent or consistently failed. LB01, however, provided consistent, high quality nuclei extractions if the buffer was prepared fresh or if it was
41 stored frozen. LB01 use d in this experiment consisted of: 15 mM tris base, 2 mM Na2EDTA, 80 mM KCl, 20 mM NaCl, 0.1% (v/v) Triton X mercaptoethanol. Silica dried Dicerandra tissue was chopped in 1 mL LB01 buffer using a fresh disposable razor blade with fresh tissu e from a single individual of Zea mays which served as an internal standard. Approximately 1 cm 2 of intervenous Zea mays tissue was used. The quantity of Dicerandra leaf tissue varied, partly as a function of the amount available and partly due to the s ize of the leaves, with 5 9 leaves used for D. linearifolia var. linearifolia D. odoratissima, and D. fumella and 1 3 leaves used for D. linearifolia var. robustior D. densiflora, and D. radfordiana Following chopping, the buffer/nuclei slurry was filt ered through a 30 m screen into a 10 mL test tube, 40 L of 50 g/ mL propidium iodide solution was added, and the samples were incubated on ice for 60 min. Analyses were performed on an Accuri C6 flow cytometer (BD Biosciences San Juan, CA, USA) recordi ng a minimum of 100 events within the gated region per sample. Population Genetic Analyses Linkage Disequilibrium, Hardy Weinberg Equilibrium, Genetic D iversity Loci were tested for linkage disequilibrium (LD) and Hardy Weinberg equilibrium (HWE) using GEN EPOP v.4.1 (Rousset 2008). LD was evaluated to test the null hypothesis that genotypes at one locus are uncorrelated with genotypes at another locus. Significant departure from HWE was assessed for each population. Markov chain sampling was used to eval Markov chain search parameters set to run 300 batches with 5,000 dememorization steps and 100,000 iterations per batch, and P exact test. Arlequin v.126.96.36.199 (Excoffi er & Lischer 2010) was used to calculate observed
42 and expected heterozygosity. GenAlEx v.6.4 (Peakall & Smouse 2006) was used to generate the descriptive statistics of alleles per locus, allele frequency by population and locus, allelic richness, and pres ence of private alleles. F statistics were calculated with GENEPOP v.4.1. The genetic divergence metric of D (Jost 2008) was calculated in GENODIVE v.2.0 (Meirmans and Van Tienderen 2004). D was chosen over the traditional F ST because the latter is less suited for highly polymorphic microsatellite loci. The upper value for F ST of 1, denoting complete population differentiation, is mathematically related to the expected heterozygosity ( H e ) of the population. As heterozygosity increases, the maximum value of F ST decreases proportionally; this relationship can be described mathematically as F ST (max)=1 H e (Meirmans and Hedrick 2011). For example, if a H e value of 0.8 is found then the maximum value of F ST is 0.2, which may appear to describe a relatively low degree of population structure, yet this value actually D does not rely on H e but rather on the effective number of alleles. Jost recognized that increases in H e do not scale linearly with increases in genetic divergence and thus metrics relying on H e such as F ST do not intuitively reflect population divergence (Jost 2008). For example, a two fold increase in H e does not correspond to a two fold increase in F ST and when comp aring multiple populations with differing H e values, the magnitude of differentiation becomes difficult to interpret because the metric does not scale linearly, plus it is affected by a reduced upper limit ( F ST (max)=1 H e ). Because of these properties of F ST a metric not reliant on H e D relies on the effective number of alleles, which does scale linearly with diversity and is not affected by a
43 diminishing upper boundary as the number of alleles increases. Thus, D is a more appropriat e metric than F ST for describing microsatellite based variation. As there is a long history of including F ST values in population genetic studies even when highly polymorphic microsatellite loci are used, we present F ST as well as D metrics for our analys is. To visually assess the magnitude of genetic differentiation among populations, pairwise genetic distance matrices using F ST and D were used to construct unrooted neighbor joining trees in PAUP* v.4.0 (Swafford 2003). To assess the distribution of genet ic variance an analysis of molecular variance ( AMOVA ) was performed using Arlequin v188.8.131.52 (Excoffier and Lischer 2010). The 32 populations were assigned to g roups based on current species delimitations to quantify how the total observed genetic variati on was distributed: among groups, among populations within groups, among individuals within populations, and within individuals. Significance was tested using 10,100 permutations of the data. Population Structure To determine population structure the Baye sian clustering program STRUCTURE v.2.3.1 (Pritchard et al. 2000) was used to determine the likelihood for the placement of individuals into a series of increasing numbers of clusters ( K ). This method is used to help determine the maximum amount of organi zation, or distinct units, within a sample. The data were evaluated at varying scales, depending on the questions being addressed, for a total of 5 nested analyses. The most inclusive analysis tested all individuals across all species using 8 loci. This initial analysis was used to evaluate large scale genetic structure primarily at the species level. The next analysis consisted of samples from the D. linearifolia complex using the same 8 loci, and was run with and
44 without the D. linearifolia x D. odora tissima hybrid populations from eastern Georgia (pop 23 and 25). The final nested analysis contained samples of D. odoratissima and D. radfordiana using data from 7 loci, with the analysis run with and without the hybrid populations (pop 23 and 25). For e ach analysis STRUCTURE was run for 1,000,000 generations with a 200,000 generation burn in for values of K ranging from 1 8. Each analysis was performed for 10 iterations per K value using the admixture model, correlated allele frequencies, with no a prio ri grouping of individuals by population ID. Estimation of the highest level of population structure ( K ) was determined using the Delta K method of Evanno et al. (2005) as implemented in Structure Harvester (Earl and VonHoldt 2011). The program CLUMP v.1 .1.2 (Jakobsson and Rosenberg 2007) was used to average the probability of cluster assignment for each individual across the 10 independent iterations, and DISTRUCT v.1.1 (Rosenberg 2004) was used for plot visualization. Pairwise comparisons of F ST for all populations were calculated using Arlequin v184.108.40.206 (Excoffier and Lischer 2010). A permutation test of 10,000 replicates followed by a Bonferroni correction for multiple comparisons was used to determine significance differences between p 0.05) (Rice 1989 ). Currently, no pairwise D Isolation by Distance To test the effect of geography on genetic differentiation, a Mantel test was used to evaluate the effects of isolation by distan ce. GenAlEx V.6.4 (Peakall and Smouse D est statistic and log transformed linear geographic distances between site locations. Log geographic distance was used over linear distance because log distance best describes spatial
45 relationships in a two dimensional space, such as a landscape. Additionally, to accurately evaluate the correlation between a metric that varies between 0 and 1, such D EST comparisons must be made against log tra nsformed distances. These data were randomly permuted 9,999 times to construct a null distribution to evaluate the hypothesis that increasing geographic distance correlates with increasing genetic differentiation. This analysis was performed on the compl ete data set as well as on a data set containing species within the D. linearifolia complex and a data set of D. odoratissima and D. robustior with the latter two excluding the hybrid populations (pop 23 and 25).
46 Table 2 1. Sampling locations, number o f individuals per populations, and whether the population was utilized for this study. State Population Number Species Num. of Individuals Sampled Used in Molecular Analysis GA 1 LIN 2 N GA 2 LIN 11 Y GA 3 LIN 16 Y GA 4 RAD 16 Y GA 5 ODO 16 N GA 6 OD O 16 N GA 7 ODO 16 N GA 8 ODO 10 N GA 9 ODO 16 Y GA 10 LIN 16 Y GA 11 ODO 16 N GA 12 ODO 16 N GA 13 ODO 16 Y GA 14 ODO 16 Y GA 15 ODO 16 Y GA 16 ODO 16 Y GA 17 ODO 16 Y FL 18 FUM 16 Y FL 19 FUM 16 N GA 20 ODO 16 Y GA 21 ODO 16 Y GA 22 LIN 1 6 Y GA 23 HYB,LIN,ODO 22 Y GA 24 LIN 16 Y GA 25 HYB, ODO 16 Y GA 26 LIN 16 N GA 27 LIN 16 Y GA 28 LIN 16 Y GA 29 LIN 16 Y GA 30 ODO 8 Y GA 31 ODO 8 Y GA 32 LIN 8 Y GA 33 LIN 8 Y FL 34 FUM 16 Y AL 35 FUM 14 Y FL 36 FUM 16 Y FL 37 FUM 8 Y FL 38 FUM 8 Y FL 39 FUM 8 N FL 40 ROB 5 Y FL 41 ROB 5 Y GA 42 RAD 15 Y
47 Figure 2 1. Locations sampled in Florida panhandle. Populations designated by orange circles were sampled but not included in the population level analysis. Site 7 is the locat ion from which the sample for the 454 sequencing run was collected; this population was not used for further analyses.
48 Figure 2 2. Locations sampled in Georgia. All populations shown were used in the analyses. The insert displays the populations along the Altamaha River, including the two populations of D. radfordiana (pop 4 and 42) and the two populations containing putative hybrid individuals (pop 23 and 25). Population 10 and population 25 are approximately 30 meters from one another and thus are represented by a single point.
49 CHAPTER 3 RESULTS Next Generation Sequencing The 454 titration run sequenced a total of 51,540,789 base pairs in 137,609 reads with an average length of 374 bp and an average quality score of 33.3. Of the total reads; 63,196 were from the Dicerandra sample, 72,761 were from Nolina and 1,652 reads could not be matched during the sorting of reads by barcode. Of the total reads assignable to species, 46.48% were Dicerandra and 53.51% were Nolina close to the ideal 1:1 r atio for samples pooled in the same sequencing lane. Microsatellite Discovery, Primer Design, and Testing Of the 63,196 Dicerandra reads generated during the sequencing run; 3,349 contained microsatellite loci, 1,065 dinucleotide loci of 6 repeats or great er, 1,247 trinucleotide loci of 4 repeats or greater, and 1,037 tetranucleotide loci of 3 repeats or greater. Despite the high number of microsatellites detected, primers could only be designed for 1,668 loci, and of these, 1,567 contained microsatellites with fewer than the desired 8 repeats. This resulted in 101 reads with 8 or more repeats with primers meeting out specifications (Tables 3 1 & 3 2). The BLAST search did not reveal any of the 101 loci to be within known coding regions, thus lowering the probability of them being under selection, nor were any within the chloroplast genome. Each locus and its associated primer sites were visually inspected, with 46 loci being selected for amplification testing. These 46 loci were divided into two categori es: 1) 25 loci with high quality base calls in the primer sites, unbroken microsatellite repeats, and a low frequency of homopolymer runs; and 2) 21
50 loci with poorer quality base calls or longer homopolymer runs, but all contained unbroken repeats. Within the first category, 6 of the 25 loci did not amplify in any of the 4 species in the test panel. Eight of the loci showed amplification across the entire test panel, and 11 amplified in 2 or 3 of the species. Of the 21 loci in the second category, 14 fail ed to amplify in any of the species in the test panel, but the remaining 8 loci amplified in all of the species. The 16 loci that amplified in all species of the test panel were then genotyped. Of the species within the test panel, D. linearifolia var. ro bustior had the highest proportion of amplifying loci, consistent with the fact that this was the species sequenced. However, when comparing the number of loci which amplified in some but not all of the species, there was no pattern between phylogenetic r elatedness and probability of amplification. Taxa more closely related to D. linearifolia var. robustior did not have a greater proportion of loci amplify than more distantly related taxa. Genotyping The 16 loci that amplified across the test panel were g enotyped in a subset of individuals from populations across the range of each species. This enabled the early detection of monomorphic loci as well as loci producing chromatograms that could not be scored easily and consistently. Of the 16 loci genotyped 3 appeared monomorphic across all species and were excluded from subsequent tests. Five loci produced chromatograms that could not be accurately scored. The morphology of these chromatograms either consisted of several scattered peaks across a wide ran ge of base pairs, often with considerable background noise, or consisted of a smooth mound of peaks, each one base pair apart, with no distinguishable features to identify the true
51 fragment length ( Figure 3 1). These 5 loci were also excluded from further a nalysis. The remaining 8 loci (locus 2, 5, 8, 10B, 16B, 18, 20, 23) produced clear chromatograms that were scored easily and consistently ( Figure 3 1). Despite showing amplification across all species in the test panel, not all of these 8 loci consistently amplified across all the species when the number of populations was increased. Once all individuals were genotyped, 5 loci amplified in greater than 95% of individuals. The remaining 3 amplified in either the D. linearifolia complex or in D. odoratissima and D. radfordiana (Table 3 3). Chromosome Counts Huck (1997) reported D. densiflora and D. linearifolia var. robustior as 2 n= 48 and 2 n= 32, respectively. To verify these observations we sampled 4 individuals of D. densiflora from populations 46 and 47 a nd 6 individuals of D. linearifolia var. robustior number between these species; all 10 individuals had 2 n= 32 ( Figure 3 2). Each least three metaphase or prometaphase cells, and often more, with minimal physical overlap. Flow Cytometry Analyses of genomic DNA concentrations in relation to a standard of known size were performed to assess variation in ploidal level. The log differen ce between the standard and a sample of D. densiflora with a confirmed chromosome count of 2 n = 32, was calculated and used to compare against samples to determine if there was an increase in DNA content. The difference in chromosome numbers between 2 n = 32 and 48 represents a 33.3% increase in DNA content. Screening for specimens with 2 n = 48
52 standard, with the difference of the known 32 chromosome sample and the standard. Any samples with an approximately 33% increase in DNA content would have been considered a polyploid. Flow cytometry was performed on 2 individuals each from 15 populations: Populations 46 and 47 of D. densiflora population s 40 and 41 of D. linearifolia var. robustior populations 34, 18, and 36 of D. fumella populations 27, 29, 22, and 33 of D. linearifolia var. linearifolia populations 15, 13, and 21 of D. odoratissima and population 4 of D. radfordiana Across all samples no significant deviation in th e observed absorbance compared to the known 32 chromosome sample was found. This lack of observed polyploids from the flow cytometry analysis is consistent with the observations from the chromosome counts, indicating all annual species of Dicerandra have a chromosome number of 2 n = 32. The lack of chromosome number variation was corroborated by the micro satellite data, where all loci evaluated had a maximum of two alleles. If polyploids existed it is likely there would be loci that contained more than two alleles, since the genome would consist o f more than two copies of a locus resulting from the duplication event; however this was not observed in any of the loci examined. Locus Characteristics and Fragment Variation The number of alleles observed at each locus varied considerably, ranging from a maximum of 39 alleles at locus 20 to a minimum of 5 alleles at locus 23 (Table 3 3). The nature of microsatellite expansion, increasing in discrete intervals, predicts that a locus with a high number of alleles wi ll also show high variation in fragment length, since differences in fragment length are assumed to reflect differences at the microsatellite repeat. Locus 20 displayed the largest variance in fragment size, 48bp, and also had the greatest number of allel es. Similarly, locus 23, with the fewest
53 observed alleles, exhibited the smallest variance in fragment size, 13bp. To verify if the large variation in fragment length seen at certain loci was the result of increases in tandem repeats at the microsatellit e, instead of indels in the flanking region, a limited number of homozygous individuals exhibiting large differences in allele length were sequenced. Loci 2, 5, and 10B, all of which showed a high number of alleles (Table 3 3), were sequenced in 12 indivi duals. Within this relatively small sample, multiple sources contributed to the variation in fragment length other than expansion of the microsatellite repeat. Indels in the flanking region and point mutations in the repeat or the flanking sequences were found in all loci sequenced. Locus 2 contained 20 alleles across all species, with some being significantly smaller fragments than expected: for example, 173 bp and 174 bp in length compared to the expected 234 266 bp. Sequencing revealed the small alle le size was the result of a 72 bp deletion, which included the microsatellite repeat. Locus 5 contained 31 alleles spanning a fragment size range of 45 bp. The vast majority of variation at this locus was due to expansion of tandem repeats; however, a 3 bp deletion in the flanking region was present in several sequenced individuals. Locus 10B displayed 21 alleles with a fragment size range of 26 bp. Of the individuals sequenced, variation in the repeat region was low, with only point mutations being det common and ranged from 1 14 bp. Overall, sequencing of these microsatellite loci showed the variation in fragment length was attributed to both expansion of the microsatellite as well as indel s in the flanking sequences. While not strictly attributed to expansion of the microsatellite repeat, mutations in the flanking regions are heritable and thus still remain valuable in
54 assessing population structure and genetic similarity. For this reason no attempt to correct allele size for variation in the flanking region was made. Additionally, because the source of fragment length variation was not exclusively from the microsatellite repeat, a stepwise mutation model is not valid for analyzing this data set and thus R ST was not used as a divergence metric. Instead, F ST D which both rely on an infinite alleles model, were used. Linkage Disequilibrium, Hardy Weinberg Equilibrium and Heterozygosity None of the loci analyzed showed evidence of significant linkage disequilibrium Weinberg equilibrium (HWE) (Table 3 4). The low percentage of populations in HWE is not of great surprise given the HWE assumptions and the life history traits of these organisms. Populations not in HWE may have resulted from founders events or have undergone bottlenecks. Current factors such as selection or small population size can also result in deviation from HWE; however, the size of the population did not predict whether a population was individuals at the time of sampling were in HWE. In contrast populations 30, 37, 38, and 40 contained fewer than 50 individuals and we re also in HWE. The same is true for those populations not in HWE; size at time of sampling does not appear to predict if a population is in HWE or not. However, given the annual nature of these plants, the isolation of most populations, and relatively s mall size of many populations, genetic drift could be a considerable factor contributing to the observed levels of diversity within each population. Mean observed heterozygosity ( H o ) among populations was 0.584 with a range of 0.275 0.8 I n 28 of the 32 populations H o was lower than expected heterozygosity ( H e )
55 (Table 3 4). Within these 28 populations 23 showed a statistically significant deficiency of heterozygotes, including both populations of D. radfordiana The f our populations showing h igher H o than H e were all D. fumella with two showing a statistically signif icant excess of heterozygotes. The populations observed to be deficient in heterozygotes and thus not in HWE, may be deficient due to biological factors mentioned previously (fou nders events, bottlenecks, small population sizes). However, an alternative explanation for the observed deficiency of heterozygotes is population substructure producing a Wahlund effect. If there was strong differences in allele frequencies between two portions of a population which were analyzed as a single population, this could result in a significant deviation from HWE (Wahlund 1928). This would be perceived as a deficiency in heterozygotes even though each subpopulation may individually be in HWE. In order for a significant deviation in HWE to be solely from a Wahlund effect there must be a high frequency of individuals homozygous for different alleles within the population. Close evaluation of the 8 locus data set for all populations with a signi ficant deviation from HWE revealed that at each locus the vast majority of homozygous individuals were homozygous for the same allele. This indicates the deviation from HWE and the observed deficiency of heterozygotes is not due to unrecognized population substructure but more likely the result of biological factors acting on the population as a whole. The overwhelming presence of a single allele for all homozygotes is unusual, especially since there was little to no representation of homozygotes for the alternative alleles seen in the heterozygotes. The low frequency of homozygotes with the
56 alternative alleles and fewer than expected heterozygotes is likely the cause for the significant deviation from HWE and the deficiency in heterozygotes. Genetic D ive rsity Among Populations Genetic differentiation was measured using two metrics, F ST and D EST both with a range of 0 1. D EST D that takes into account small population size and the uncertainty of not knowing the true allele frequ encies. These values were calculated using the 5 locus data set, which had the greatest coverage across all species (>95% of individuals could be genotyped at these 5 loci), and allows for a direct comparison across species. In general, F ST and D EST valu es showed a similar pattern (Table 3 5) despite having different values. This is to be expected since F ST calculations are likely low er due to the highly polymorphic nature of the loci used and the inability of F ST to scale with increasing levels of heter ozygosity. D EST values for D. odoratissima and D. linearifolia var. linearifolia showed the highest degree of interpopulational variation, D EST = 0.560 and 0.437 respectively. Both species had an equal number of populations, n=10, and spanned the widest g eographic area of all sampled species (Table 3 5). Dicerandra radfordiana and D. linearifolia var. robustior had lower D EST values, 0.304 and 0.205, respectively. However, both species were represented by only two populations, by necessity for D. radford iana given that only two populations are known to exist. But D. linearifolia var. robustior is more widespread, and its corresponding D EST value could change if additional populations were included. There is high uncertainty associated with the D EST valu e for D. linearifolia var. robustior as demonstrated by the standard error, 0.223, which is higher than the D EST value itself, 0.205; thus this value should be interpreted with significant caution. Dicerandra fumella had the lowest D EST 0.263, indicating less interpopulational
57 differentiation than the other species, despite the samples representing a relatively wide geographic area. Mean F ST values varied among species ranging from a minimum of 0.149 in D. fumella to a maximum of 0.254 in D. odoratissima (Table 3 5). Overall, F ST values were highest in species with the greatest geographic range as seen in Dicerandra linearifolia var. linearifolia and D. odoratissima Reciprocally, species with smaller geographic ranges, D. fumella, D. linearifolia var. robustior and D. radfordiana showed less genetic differentiation among populations (Table 3 5). The AMOVA showed that of the total genetic variance 12.49% was among species, 11.39% was among populations within species, 9.36% was among individuals within p opulations, and 66.76 % was within individuals. Permutations tests indicated all values were significantly different than zero D using permutation based tests for statistical significance is not available. However, Bonferroni corrected pairwise comparisons of F ST values, using all individuals for the 5 locus data set, revealed mostly statistically significant differences between populations. Of the 496 pairwise comparisons only 15 were not statistically different (Table 3 6). Of these 15 pairwise comparisons, one was between two populations of D. odoratissima seven w ere between populations of D. linearifolia s.l. four were between populations of D. fumella two were between D. linearifolia var. linearifolia pop 33 and both hybrid populations (pop 23 and 25), and the final non significant comparison was between two po pulations of different species, D. linearifolia var. linearifolia and D. odoratissima Most of the populations displaying non significant pairwise comparisons are found in the same
58 geographic region, but are not necessarily proximate. The shortest distan ce between two of these populations is 9 km, but many are separated by over 30 km. In most cases there are populations geographically closer, yet these have differentiated sufficiently to be considered statistically different. Given the distances involve d, these non significant associations are probably not due to current gene flow, especially given that the insects observed visiting flowering plants are generalists, and thus not specifically seeking Dicerandra individuals. Additionally, the habitat is o ften discontinuous, further reducing the probability of pollinator mediated gene flow. Visualization of the magnitude of genetic differentiation was achieved using neighbor joining trees constructed from pairwise distance matrices of F ST and D Ove rall, the trees are similar in t opology and branch length and both place individuals into two main groups, one containing the D. linearifolia complex and the other containing D. odoratissima and D. radfordiana ( Figure 3 3). Yet within these two groups popul ations of D. linearifolia var. linearifolia do not cluster together, nor do all populations of D. odoratissima The D. linearifolia var. linearifolia cluster also contains the populations of D. linearifolia var. robustior nested within it ( Figure 3 3). Wit hin the D. linearifolia s.l. cluster there are geographic trends where populations in the same region are typically more genetically similar to each other than to populations in a different region; however, most geographic regions do not exclusively contai n populations from that area. For example, t here is general cohesion among the western populations of D. linearifolia var. linearifolia ( Figure 3 3, blue); however, population 22 from eastern Georgia is nested among this predominantly western cluster. Simi larly, the central Georgia populations ( Figure 3 3, orange) also do not form an exclusive
59 cluster. Thus, there is not support, in the form of exclusive geographic clusters, for a clear genetic distinction between geographic regions of D. linearifolia var. l inearifolia. While there is support for the exclusive clustering of D. linearifolia var. robustior ( Figure 3 3, light blue), these populations a re nested closely among the central Georgia D. linearifolia var. linearifolia and do not show an affinity to the proximate D. fumella populations from the Florida panhandle. Both the F ST and D E ST trees show all populations of D. fumella forming an exclusive cluster Interestingly the length of the branch leading to the D. fumella clade i s as great or greater than b ranch es leading to other named species. This supports species. There is also considerable genetic distinction between D. fumella and D. linearifolia var. robustior de spite their geographic proximity The genetic distinction of D. fumella in these trees is also well supported in the STRUCTURE analysis ( Figure 3 5), giving further credence to the recognition of D. fumella as a species. Similar to the situation with the D. linearifolia complex, D. odoratissima populations also do not from an exclusive cluster. While t he two populations of D. radfordiana c onsistently are sister, they are found on the same branch as two geographically proximate populations of D. odoratissima (pop s 20 and 21) Additionally, t hese four populations are the same ones recognized as a distinct cluster in the STRUCTURE analyses ( Figure 3 5). There is little geographic cohesion among the populations of D. odoratissima : the northeastern ( Figure 3 3, gree n), the central ( Figure 3 3, orange), and one southeastern population ( Figure 3 3, yellow) are all nested together and do not form geographic
60 based clusters. The possible exceptions are the two D. radfordiana populations and their proximate D. odoratissima po pulations (pop 20 and 21). These populations of a relatively long branch, suggesting their location at the edge of the range is also associated with increased genetic dif ferentiation. However, there is a 50 km gap between D. odoratissima population 21 and the next closest D. odoratissima population, 9. This gap was not sampled but is know to contain populations of D. odoratissima thus the length of the branch may be mis leadingly long due to the absence of these populations. Isolation by distance, evaluated with a Mantel test, showed a significant correlation between geographic distance and genetic distance for D. odoratissima and a combined analysis of D. linearifolia va r. linearifolia and D. linearifolia var. robustior ( Figure 3 4). No correlation was seen with D. fumella even though the sampled populations Population Structure STRUCTURE analysis of all individuals using both t he 5 and 8 locus data sets showed significant genetic structure which generally followed currently delimited species boundaries ( Figure 3 5). The Delta K method for determining the maximum number of genetically significant clusters showed the highest suppo rt for K =4 (Evanno et al. 2005). However, determining the true number of clusters can be contentious as different methods for determining the maximum number of clusters often produce differing results (Ball et al. 2010). It is important to consider the b iological aspects of the organism when evaluating the clustering pattern at a given K As a result, it is often prudent to examine the clustering patterns for values of K surrounding what was
61 determined by the given method, especially if there could be bi ologically pertinent differences captured by an alternative clustering pattern. In this analysis, K =4 was consistently chosen by the Delta K method. However, the Delta K method relies heavily on the observed variance of the log likelihood value of multip le independent STRUCTURE runs. The K value with the smallest variance is often mathematically favored when compared to K values with even slightly larger variances. In this analysis, K =4 showed very low variance. When compared to K =5, which also had a v ery low variance but was larger than that of K =4, K =4 was selected as having the highest K Since K =5 also exhibited a very low variance the clustering pattern at this value was also evaluated. Within the combined STRUCT URE analysis the first cluster, red, includes all populations of D. fumella 18 and 34 38 from the Florida panhandle, which form a distinct genetic cluster with low to no signs of genotypic admixture with other clusters ( Figure 3 5). These populations were collected along a 115 km stretch of the Florida panhandle from the Escambia River in the west to the center of the Marianna lowland in the east. This includes samples from the type locality of D. fumella population 34, as well as two populations, 37 and 38, within the zone of hybridization proposed by Huck (2010) between D. fumella and D. linearifolia var. robustior However, our analyses found no evidence of admixture in any of the individuals of D. fumella with individuals of D. linearifolia var. robus tior or D. linearifolia var. linearifolia Individuals of D. linearifolia var. linearifolia and D. linearifolia var. robustior formed the second cluster, blue ( Figure 3 5, K =4). This cluster spans the greatest geographic area, ranging from the Florida panha ndle to west central Georgia and east to the
62 Altamaha River. This cluster displays a high degree of cohesion even though the most extreme populations are separated by as much as 300 km, with many separated by 100 km or more. The final two clusters, purple and yellow, contain populations of D. odoratissima and D. radfordiana These clusters, while well supported, do not follow current species delimitations. The yellow cluster includes the two D. radfordiana populations as well as two nearby populations of D. odoratissima The remaining eight D. odoratissima populations constitute the purple cluster. The placement of D. radfordiana and D. odoratissima populations in the same cluster correlates strongly with geography. The distance between the two D. radfo rdiana populations is 3 km, with the two closest D. odoratissima populations (pop 20 and 21) being 1.8 km and 7.3 km from the closest D. radfordiana population. All four of these populations (4, 42, 20, and 21) occur across 11.7 km. The next closest D. o doratissima population sampled (pop 9) is 50 km away and is a member of a different cluster (purple), which does not show any signs of admixture with any of the individuals from the yellow D. radfordiana cluster. The two D. odoratissima populations within the same cluster as D. radfordiana are also supported in the neighbor joining trees based D and F ST distance matrices ( Figure 3 3) The clusters observed in the analysis of all individuals were identical to those observed in the two less inclusive analyses focusing on the D. linearifolia complex and D. odoratissima and D. radfordiana Additionally, there was no change in cluster assignment of populations between the 5 and 8 locus data sets.
63 In general, the cluster assignments of most populations do not change from K =4 to K =5 except for certain populations of the D. linearifolia var. linearifolia and D. linearifolia var. robustior cluster ( Figure 3 5). At K =5, the two populations of D. linearifolia var. robustior are pulled out in a cluster, green, along with two populations of D. linearifolia var. linearifolia from eastern Georgia, as well as the D. linearifolia var. linearifolia individuals of the putative hybrid populations, pop 23 and 25, also from eastern Georgia. The western populations of D. linearifolia var. linearifolia remain in a well defined cluster, blue. Populations 32 and 33 are D. linearifolia var. linearifolia from central Georgia, and both populations show a high degree of admixture with the eastern (green) and western (blue) clust ers. Based on this pattern, it could be hypothesized that there is clinal variation in D. linearifolia var. linearifolia with the east and west being genetically distinct regions and central Georgia being an area of intergradation. This pattern would be supported ecologically as well, because dispersal capability is limited and pollinator mediated gene transfer is likely low, allowing for genetic drift, or selection, to reinforce particular genotypes at either extreme of the range. However, population 2 2, which is the eastern most population of D. linearifolia var. linearifolia, is made up of individuals primarily of the western genotype. The placement of population 22 in the same cluster as the western populations is also seen in both the F ST s D trees ( Figure 3 3). Hybrid Detection To evaluate the molecular support for a hybridization zone between D. fumella and D. linearifolia var. robustior STRUCTURE analyses containing all populations of D. fumella, D. linearifolia var. linearifolia, and D. linearifolia var. robustior were performed ( Figure 3 6). In all circumstances the highest support was for K =2, with one cluster
64 containing populations of D. fumella and the second cluster containing all the D. linearifolia populations. However, regardless of the K value there was no support for any individuals of D. fumella or D. linearifolia var. robustior being of hybrid origin. Additional STRUCTURE analyses containing individuals from mixed populations of D. odoratissima and D. linearifolia var. linearif olia in eastern Georgia were performed t o determine if there was a detectable molecular signature for putative hybrid individuals, identified in the field based on intermediate or chimeric morphology. The two putative hybrid populations, pop 23 and 25, co ntained individuals that were morphologically identified as D. linearifolia var. linearifolia, D. odoratissima and as putative hybrids. The putative hybrids had corolla architecture similar to D. linearifolia var. linearifolia except the stamens were pos itioned along the upper lobe or were upwardly curving with minimal stamen exertion, a characteristic of D. odoratissima Additionally, the putative hybrid individuals had a distinct odor more reminiscent of the cineol compounds of D. odoratissima than the menthol compounds of D. linearifolia var. linearifolia despite showing floral morphology more similar to D. linearifolia var. linearifolia The STRUCTURE analysis of the populations containing the putative hybrids showed many individuals with mixed geno types with portions from both of the suspected parental species, as well as individuals with genotypes composed primarily of D. linearifolia var. linearifolia (blue and green depending on the K value employed) or D. odoratissima (purple) ( Figure 3 6). The De lta K method identified K =2 as having the greatest support. The clustering pattern for K =3 was also investigated, but did not reveal additional meaning ful patterns. When the individuals in populations 23 and 25 are segregated based on identification in t he field, we see there is not a clear molecular distinction
65 between the two parental species. Individuals identified as D. linearifolia var. linearifolia and D. odoratissima have large portions of their genotype that are assigned to the other species (Fig 3 6). In population 23 at K =2, the genotypes for two of the eight samples of D. linearifolia var. linearifolia are almost entirely assigned to the D. odoratissima cluster. Similarly, 4 of the 8 samples of D. odoratissima have genotypes almost exclusivel y assigned to the D. linearifolia var. linearifolia cluster. Within the putative hybrids only two individuals show a genotype that is significantly mixed, each with approximately 65% assigned to D. linearifolia var. linearifolia and 35% assigned to D. odo ratissima This pattern suggests there has been backcrossing of hybrids to the parental lines, creating individuals that are phenotypically more similar to one parent but genetically more similar to another. The observation of individuals that are not a 50:50 mixture of each parental species suggests the observed hybrids are likely not F 1 hybrids and may represent F 2 or later generation progeny and/or backcrosses. Population 25 consists of individuals of D. odoratissima and putative hybrids, but populatio n 10 which is approximately 30 m from population 25 is exclusively D. linearifolia var. linearifolia The STRUCTURE analysis strongly supports population 10 being exclusively D. linearifolia var. linearifolia with no evidence of admixture within this pop ulation ( Figure 3 6). This suggests there may be a directional component to hybridization, where hybrids are produced with D. linearifolia var. linearifolia exclusively being the pollen donor and D. odoratissima exclusively being the pollen recipient. Yet within population 25, of the eight individuals identified as D. odoratissima three a re genotypically identical to D. linearifolia var. linearifolia two display a near 50:50 genotypic mixture, and the remaining three display approximately 80:20 genotypic
66 mixture between the two parental species. A similar situation is observed in the individuals identified as hybrids with five of the eight being exclusively D. linearifolia var. linearifolia genotypes, one displaying almost an exclusively D. odoratissima g enotype, and two show a 35:65 genotypic mix. In both of these putative hybrid populations there is a clear molecular signature showing certain individuals are of hybrid origin. However, the individuals identified as hybrids based on possessing what appear ed to be intermediate or mosaic morphology did not necessarily carry a molecular signature for a hybrid origin.
67 Table 3 1. 101 reads with microsatellites containing 8 or more repeats for which primers could be constructed Monomer Size, bp Num. of Repe ats Occurrences 2 8 27 9 24 10 13 11 1 12 4 Total 69 3 8 17 9 6 10 3 12 1 14 1 16 1 17 2 19 1 Total 32 Table 3 2. 1 567 reads with microsatellites containing fewer than 8 repeats for which primers could be constructed Monomer Size, bp Num. of Repeats Occurrences 2 6 143 7 64 Total 207 3 4 528 5 159 6 54 7 38 Total 779 4 3 520 4 40 5 19 7 2 Total 581
68 Table 3 3. Description of the primer sequences, loci characteristics, and allelic complement for the 8 loci utilized across all the annual species. Single asterisk denotes loci that amplified consistently in all species, double asterisk only amplified in species of the D. linearifolia complex, triple asterisk loci amplified across spec ies but was inconsistent between populations. Locus Num. Locus Name Repeat Motif Dye Used Num. of Alleles Fragm ent Size (bp) 2* H2BD8 F: AAAGGAAGACAAGGGTCAAGG (TC)8 VIC 20 173, 174, R: AGAAGCCAGCAACAAGAACG 234 266 5* H41X1 F: TCGCAGAACAAGTAGTCAAAGC (TC)10 PET 31 211 256 R: GGAAGCTCAATCCATAGAAAGG 8* H61XY F: GAAGCACCATATCAAGATTCACC (AC)8 FAM 23 198 245 R: GGACATTCAGCTCTCCATCG 10B*** IAXR6 F: TGGTGCTACTGCTACTGCTGG (AC)9 NED 21 211 237 R: CGACATTGGTCTGAATCATGG 16B*** IGK5Z F: TTGTGGAAAGTTATTGACACCC (TC)9 PET 11 184 209 R: CCTCACGGACTCGACTGG 18* IMEMS F: CTCACCTATGCGGAGATGG (TC)9 NED 19 237 265 R: AACTCGACTCAAAGCAATCG 20* IAPG4 F: AGGCGT CCCATTCTGAAGC (TC)12 FAM 39 257 305 R: AGGTTGGTGAGGGTGTTGG 23** IJQ5D F: TTTCGTAGGAGGTAGATGTAGAAG (AT)9 VIC 5 302 315 R: GAGCAAACTTGGGAAGGAGG
69 Table 3 4. Hardy Weinberg equilibrium p values, inbreeding coefficient ( F IS ), heteroz ygosity deficiency and excess tests, and observed and expected heterozygosity ( H o H e ) Asterisk denotes significant values. POP Num. Species HWE p value F IS H e def. p value H e excess p value H o H e 2 LIN 0 0.361 0 1.0 0.475 0.764 3 LIN 0 0.198 0 1.0 0.520 0.643 4 RAD 0 0.270 0 0.999 0.417 0.619 9 ODO 0 0.243 0 1.0 0.489 0.644 10 LIN 0.208 0.059 0.069 0.931 0.582 0.618 13 ODO 0 0.229 0 1.0 0.530 0.703 14 ODO 0.001 0.203 0 0.999 0.518 0.616 15 ODO 0.043 0.120 0.032 0.969 0.553 0.685 16 O DO 0 0.308 0 1.0 0.512 0.735 17 ODO 0 0.255 0 1.0 0.574 0.779 18 FUM 0.014 0.070 0.222 0.779 0.743 0.698 20 ODO 0 0.294 0 1.0 0.423 0.628 21 ODO 0 0.201 0 0.999 0.602 0.765 22 LIN 0.020 8 0.111 0.015 0.985 0.613 0.701 23 HYB 0 0.245 0 1.0 0.623 0. 821 24 LIN 0 0.190 0 0.999 0.645 0.793 25 HYB 0 0.242 0 1.0 0.635 0.830 27 LIN 0 0.313 0 1.0 0.455 0.699 28 LIN 0.062 0.076 0.041 0.956 0.643 0.711 29 LIN 0 0.269 0 1.0 0.566 0.771 30 ODO 0.101 0.074 0.152 0.845 0.535 0.588 31 ODO 0.019 0.156 0.0 31 0.967 0.589 0.704 32 LIN 0.955 0.031 0.295 0.707 0.716 0.739 33 LIN 0.325 0.103 0.059 0.935 0.692 0.783 34 FUM 0.008 0.086 0.591 0.413 0.739 0.684 35 FUM 0.026 0.216 0.995 0.003 0.757 0.627 36 FUM 0.037 0.124 0.878 0.129 0.781 0.694 37 FUM 0.371 0.207 0.993 0.007 0.800 0.671 38 FUM 0.590 0.114 0.046 0.956 0.675 0.756 40 ROB 0.143 0.586 0.001 0.999 0.275 0.597 41 ROB 0.376 0.200 0.037 0.966 0.600 0.733 42 RAD 0.035 0.263 0.002 0.998 0.393 0.526
70 Table 3 5. Genetic diversity met rics D EST and F ST Num. of pops. Mean Num. of alleles Eff. Num. of alleles D EST D EST Std. error Mean F ST F ST Std. error D. fumella 6 12.667 3.412 0.263 0.137 0.149 0.036 D. linearifolia var. linearifolia 10 16.857 3.453 0.437 0.046 0.221 0 .067 D. linearifolia var. robustior 2 4.167 2.096 0.205 0.223 0.151 0.07 D. odoratissima 10 17.000 2.824 0.560 0.059 0.254 0.033 D. radfordiana 2 4.429 2.134 0.304 0.13 0.158 0.05
71 Figure 3 1. Examples of microsatellite chromatograms showing peaks from loci that were unscorable (top and middle) as well as those easily and consistently scorable (bottom)
72 A B C D Figure 3 2. Chromosome images. A) 32 prometaphase chromosomes from D. linearifolia var. robustior B) 32 metaphase chromosomes from D. linearifolia var. robustior C&D) 32 metaphase chromosomes from D. densiflora
73 B Figure 3 3. Genetic distance based neighbor joining trees of all annual populations of Dicerandra D distance matrix. B) F ST distance matrix. Both trees two main clades containing 1) the D. linearifolia complex 2) D. odoratissima and D. radfordiana Colors represent geographic regions: Pink: Florida panhandle, Light Blue: central Florida, Yellow: southeast Geo rgia, Red: eastern Georgia, Orange: central Georgia, Blue: western Georgia, Green: northeastern Georgia.
74 Figure 3 4. Mantel test results evaluating isolation by distance. Asterisk indicates significance of the correlation: *p= < 0.05, **p= < 0.0 05, ***p= < 0.0005. A significant correlation between increasing geographic distance and genetic distance was found in D. linearifolia var. linearifolia and D. linearifolia var robustior as well as D. odoratissima and D. radfordiana
75 Figure 3 5. St ructure analysis of all 448 individuals at 8 loci, D elta K analysis indicates maximum population structure occurs at K =4. Population numbers are listed below, corresponding species names are listed above. Populations are arranged geographically.
76 Figur e 3 6. Structure analysis of D. linearifolia var. linearifolia and D. odoratissima populations with a focus on the hybrid populations 23 and 25 which contain individuals of suspected hybrid origin as well as putative parental species, D. linearifolia var. linearifolia and D. odoratissima Blue and green represents the clusters containing individuals of D. linearifolia var. linearifolia purple represents the cluster containing individuals of D. odoratissima
77 CHAPTER 4 DISCUSSION Next Generation Sequencing for Microsatellite Discovery The use of next generation sequencing for the discovery and development of population level markers has seen a remarkable increase in use since its start only a few years ago (e.g. Adbelkrim et al. 2009, Allentoft et al. 2009, Castoe et al. 2010). Our recovery of 3,349 microsatellite loci from one half of 1/8 of a sequencing plate further shows the utility of this method for marker discovery. However, for a variety of reasons, the detection of a large quantity of loci does not directly result in a high number of usable loci. The lack of sufficient flanking sequence for primer design and flanking sequences not meeting primer design specifications reduced the number of potential loci to 1,668. When loci with under eight repeats were then removed, this significantly reduced the number of loci to 101, yet this is still more than sufficient for population genetic analyses. Reducing the threshold for the minimum number of repeats would have provided a larger pool of candidate loci if additional loci were desired. This was especially true for tri and tetranucleotide repeats. After removing loci with fewer than eight repeats, two thirds of the remaining loci were dinucleotide repeats. Relaxing the number of repeats to seven would have produced an additional 104 loci (64 di 38 tri and two tetranucleotide loci), all of which had primers meeting the desired specifications (Tables 3 1 & 3 2). The quality of the base calls in the sequence from which the primers were designed was str ongly associated with the success of PCR amplification of the locus. The 46 loci for which primers were tested were divided into two categories. The first category containing primers designed from sequences with few to no poor quality base
78 calls had 25% t hat failed to amplify in every species. However, the second category containing loci with primers designed from sequences with some lower quality base calls had 66% that did not amplify in any species. While several factors could contribute to the differ ential success in amplification between the two categories, the primary justification for assigning loci into their respective category was the quality of the sequence from which the primers were designed. Based on this observation, we suggest only testin g primers designed from sequences with the highest quality score to maximize successful amplification of candidate loci. Overall, amplification of loci across species boundaries was successful. Of the 26 loci that amplified, only three were exclusively fou nd in D. linearifolia var. robustior, the species used for the sequencing effort, and 11 amplified in one or two additional species Consistent cross species amplification was required for this study, and thus only loci showing amplification in all member s of the test panel were considered for further testing. Because loci that amplified in all species were specifically selected, this has likely introduced some bias into our estimates of genetic diversity. These estimates are likely conservative since mo re rapidly evolving sites have a greater probability of accumulating mutations in the primer binding sites and thus cease to amplify within those species or populations. Those loci, that did not show amplification in all species were removed from the anal ysis. Thus, it is unlikely that our estimates of genetic diversity are elevated as a result of marker selection. Genetic Diversity The AMOVA results showed there was substantial genetic differentiation found among populations within a species, 11.39% and this quantity is similar to the amount of variation found between species, 12.49%. This further illustrates the importance of
79 populations for maintaining overall genetic diversity within a species. Population level variation is further illustrated with comparison s of F ST and D EST where g enetic differentiation between populations was found to be high within most species. This is especially true in the widely distributed species D. odoratissima and D. linearifolia var. linearifolia where D EST values were 0.56 and 0.437, respectively, and F ST values were 0.254 and 0.221 (Table 3 5). This level of differentiation was also reflected in the branch lengths of the genetic distance trees ( Figure 3 5). Long branches often separate populations within the same spec ies, indicating the magnitude of genetic differentiation among clusters of the same species. Furthermore, the Mantel test revealed a positive correlation between geographic distance and genetic distance in D. odoratissima and D. linearifolia var. linearif olia indicating proximate populations are more genetically similar to one another than they are to populations farther away. This correlation is consistent with high D EST values and the long branch lengths observed in the distance trees. The patterns se en across these three measures of genetic divergence consistently show there is considerable genetic diversity contained at the population level. In the case of D. odoratissima and D. linearifolia var. linearifolia, 56% and 43.7%, respectively, of the tot al variation is contained within populations, as indicated in the D EST values, the remainder is found among individuals (Table 3 5). In D. fumella and D. radfordiana there was less variation at the population level: 26.3% and 30.4%, respectively, based on D EST values. The magnitude of this reduced variation between populations in D. fumella is represented in the distance tree where populations are found on considerably shorter branches than those of D. odoratissima and D. linearifolia var. linearifolia ( Figure 3 5). Additionally the lack of a correlation
80 between geographic distance and genetic distance in the Mantel test further illustrates D. fumella has less genetic divergences among the populations ( Figure 3 4). This is further corroborated by pairwise F ST comparisons, where three of the six populations were not significantly different from one another. This is noteworthy because the vast majority of pairwise comparisons, 481 out of 496, between all the populations were statistically different from one ano ther. Interestingly, even though population level variation is lower within D. fumella than other species, D. fumella exhibits high divergence from all other Dicerandra specie. This is exemplified by the length of the branch in the distance trees leading to the D. fumella group ( Figure 3 5). While the degree of genetic differentiation between populations varies by species, low dispersal capability, isolated populations and visitation by generalist pollinators all contribute to reduce gene flow between populations. These life history characteristics would readily allow individual populations to differentiate from one another. Additionally, this tendency for population s to be genetically divergent is accompanied by a reduction in heterozygosity. In 23 out of 32 populations, there was a significant deficiency in heterozygotes (Table 3 4). This pattern of population differentiation, along with a reduction in heterozygos ity, is consistent with what would be expected from isolation and drift, potentially accelerated by fluctuations in population size, non random mating, or consistently low population numbers. These factors eventually could lead to the observed reduction o f heterozygosity. However, the obligate outcrossing nature of these plants has probably aided in maintaining heterozygosity despite drift and limited gene flow. Recruitment from the seed bank may also aid in maintaining heterozygosity
81 within these popula tions. Within the perennial species D. frutescens reestablishment of populations solely from the seed bank occurs following fire s which eliminate all individuals from the population (Menges 1992). This indicates that at least some Dicerandra species pos sess seeds capable of persisting for several seasons within the seed bank. This type of recruitment from the seed bank would produce populations with individuals from multi generations living at a given time. The presences of a multi generational populat ion would aid in maintaining heterozygosity in an annual outcrossing plant because individuals with previous years allele combinations would be present and capable of reproducing with more recent individuals. Population Structure and Species Delimitation D elimiting species boundaries requires adherence to a theoretical framework that provides both a sound description of what constitutes a species and evolutionarily meaningful evidence used to circumscribe populations as a single species. The unified specie s concept recognizes that most competing species concepts share the common feature of describing separately evolving metapopulation lineages, but they differ in the criteria employed to recognize those lineages or delimit metapopulations (De Queiroz 2007). While the difference among distinguishing criteria may seem to simply shift the name of the debate from species concepts to criteria, recognizing that the collective goal is to name populations of organisms that represent units moving through time provid es a critical ontological foundation for recognizing and delimiting species. Under this species concept, many lines of evidence can be used to recognize metapopulations, such as reciprocal monophyly, diagnosability, reproductive isolation, apomorphic char acters, or niche specialization (De Queiroz 2007). These are biologically and evolutionar il y important characteristics that can delimit particular groups
82 of organisms as species, but they often are in conflict with one another for one reason: the speciatio n process occurs over a long time scale. These characteristics do not necessarily evolve in a consistent order, over a consistent time period, nor do they all manifest themselves in every taxon. They all capture important elements of the evolutionary proc ess, which can be used as evidence for delimiting species, the difference being that each criterion recognizes a different aspect of the speciation process. Evolution is often a gradual process, and the stage of diversification at which an organism is obs erved will dictate the magnitude of divergence that has occurred. Under this approach, there is no single defining criterion, but the totality of the evidence is appreciated, recognizing the validity and biological significance of differing sources of dat Dicerandra four criteria best enable the evaluation of the degree of divergence and thus species boundaries: monophyly of populations in phylogenetic and distance based tr ees, clustering patterns based on allele frequencies, gene flow, and mo rphological differentiation. The STRUCTURE analysis and genetic distance trees that depict the magnitude of differentiation tell a congruent story regarding the relationships among spec ies and among populations within species. Overall, most current taxonomic designations are well supported at their respective rank. There is strong support for the recognition of D. fumella at the species level. The STRUCTURE analyses at both K =4 and K =5 are consistent in placing all populations of D. fumella within a single cluster. There was no evidence of admixture even with D. geographic neighbor, D. linearifolia var. robustior ( Figure 3 5). Clustering of
83 all populations of Dicerandra fumell a in the genetic distance trees further supports recognition at the species level. Additionally D. fumella has experienced substantial genetic differentiation from other Dicerandra species; this is depicted in the length of the branch leading to the D. fu mella clade. The branch length is as great as or greater than the branches separating other Dicerandra species, i.e., D. odoratissima and D. linearifolia s.l. ( Figure 3 3), which have long been recognized by systematists (Huck 1987). The genetic evidence p resented here in conjunction with the morphological characteristics presented by Huck (2010) provide substantial support for the maintenance of D. fumella at the species rank. Populations of D. linearifolia s.l. are supported at the species rank in both th e STRUCTURE analysis and the genetic distance trees ( Figures 3 3 & 3 5). In the distance trees, all populations of D. linearifolia s.l. form an exclusive cluster. Strong support is also seen from the clustering of these populations together in STRUCTURE at K =4, the value indicated by the Delta K method as the maximum number of statistically significant genetic clusters ( Figure 3 5). There is also support at K =5 where the additional cluster results purely from the subdivision of the D. linearifolia cluster of K =4. This new cluster contains populations of D. linearifolia var. robustior and several populations of D. linearifolia var. linearifolia from central Georgia. Taxonomic designations within D. linearifolia s.l. are complex. The varietal designation of D. linearifolia var. robustior is supported by morphology and in the distance trees where populations of this variety are consistently sister to each other ( Figure 3 3). However, the recognition of var. robustior results in a paraphyletic D. linearifolia var. linearifolia This is not necessarily of great concern, as the function of
84 varietal designations is to describe meaningful variation that is seen within a species, and monophyly in such cases should neither be assumed nor required. For D. linearifolia v ar. robustior the taxonomic designation as a variety, and not a species, best characterizes this group of populations for the following reasons: m orphological characteristics are capable of differentiating var. robustior from other taxa of Dicerandra, incl uding var. linearifolia populations of var. robustior are more genetically similar to each other than they are to var. linearifolia but var. robustior is nested well within var. linearifolia and populations of var. robustior are geographically disjunct f rom those of var. linearifolia. Similarly, D. linearifolia var. linearifolia is best kept as a variety. Elevation of regional groups of var. linearifolia to species rank is not supported because there is not evidence for consistently cohesive geographic c lusters within var. linearifolia ( Figures 3 3 & 3 5 ). Nor is there a meaningful way to formally classify the exclusive groups that are observed in the distance trees without introducing an additional two taxonomic units, to designate the eastern, western, a nd central populations, which are not morphologically or genetically distinct. Populations of D. odoratissima also did not for exclusive clusters in the distance trees and in the STRUCTURE analyses ( Figures 3 3 & 3 5). Populations of D. radfordiana were nes ted among populations of D. odoratissima in the distance trees and D. radfordiana and proximate populations of D. odoratissima formed a distinct cluster in STRUCTURE analyses. This cluster ( Figure 3 5, yellow) comprises populations at the southeastern extre me of the range of D. odoratissima and shows little admixture with more northern or western populations of D. odoratissima except in
85 population 21 where there is an individual with a large contribution of its genotype coming from the other D. odoratissima cluster ( Figure 3 5, purple). Population 21 marks one edge of the 50 km sampling gap separating these two clusters. This gap may contain further admixed populations suggesting clinal variation or the beginning of speciation by isolation at the extreme sou theastern portion of the range, yet additional data are needed to address this hypothesis. Regardless of the sampling gap, the recognition of D. radfordiana at the species rank causes D. odoratissima to contain populations which are more genetically simil ar to a different species, D. radfordiana which is in conflict with our species concept. This is further complicated because some D. odoratissima populations are more genetically similar to D. radfordiana yet they are morphologically more similar to all other populations of D. odoratissima There are two potential solutions to this problem, which are each considered below. 1) Recognize the genetically distinct populations, represented by the yellow cluster of the STRUCTURE analysis ( Figure 3 5), as a sing le species, D. radfordiana If this change were made, certain populations of D. radfordiana s.l. (placed in that species because of their genetic similarity) would be morphologically indistinguishable from populations of D. odoratissima To a degree this defeats the purpose of a classification, because it makes some populations impossible to identify accurately without genetic analysis. Therefore, the recognition of D. radfordiana in an expanded circumscription, based on this genetic information, would o nly introduce confusion and erroneous identification of populations because the species would lack distinguishing morphological characteristics Finally, t he state of Georgia recognizes D. radfordiana as an endangered species and continuing to recognize it as a species by widening the
86 circumscription to include genetically similar, but morphologically different populations would only negatively impact its management, as one c ould no longer definitively recognize an individual of that species in the field 2) Recognize the morphologically distinct populations of radfordiana as a variety of D. odoratissima This approach recognizes the monophyly of D. odoratissima maintains the morphological diagnosability of the radfordiana entity, allowing for consiste nt and positive identification of individuals, and continues support for the nomenclatural recognition of radfordiana populations as different from D. odoratissima Maintaining the nomenclatural recognition at an infraspecific rank not only acknowledges i ts distinct morphology it also alludes to its differing evolutionary trajectory. This is significant b ecause the genetic data suggest that these populations are at the beginning stages of speciation by isolation, and may evolve into separate species, if s ufficient time elapses and the processes driving the divergence persist. However, currently the genetic differentiation is insufficient for re cognition at the species level. Within D. linearifolia var. linearifolia a surprising outcome from the population structure analyses was the placement of the far eastern population 22 within populations found exclusively in western Georgia ( Figures 3 3 & 3 5). There are several biogeographic hypotheses that could have given rise to this pattern. The first is long dis tance dispersal of individuals from the western populations to the far eastern edge of the range. However, this is implausible due to the dispersal limitations discussed earlier, the sheer geographic distance (260 km separate population 22 and the western populations), as well as geography (the two regions are located in different watersheds one draining to the Atlantic the other to the Gulf of Mexico). Even though
87 seed transport by water has been proposed as a significant influence in current species dis tributions, the probability of a long distance dispersal event capable of establishing a viable population of annual plants appears small. A second scenario is the reverse of the first, i.e., one in which the western populations were derived from an easte rn population, and while the western populations thrived and spread, their eastern relatives remained rare and local. But, this scenario is also improbable for the same reasons as the first. The third scenario is shaped by large scale local extinction. Prior to the practice of widespread fire suppression and land conversion for anthropogenic activities, it is likely there was a broader distribution of suitable habitat with a greater frequency of Dicerandra populations. Clinal variation may still have ex isted with a predominantly western genotype and a predominantly eastern genotype with an area of admixture in the intermediate geographic region. Under this scenario, populations in the intermediate region could consist of populations dominated by one gen otype or the other, or contain truly admixed individuals. Since dispersal is low, populations of a single genotype could easily be maintained genetically due to the plants annual habit; removal of alleles from migrants can be swift, after four generations of backcrossing an individual contains only 6.25% of the introduced alleles. If local extinction increased, eliminating the mosaic of eastern and western genotypes in the intermediate region, we would be left with a genotypic distribution that would look like clinal variation with an occasional outlier, which is precisely what is seen today (Fig 3 5, K =5). These populations would seem genetically out of place given the current distribution of genotypes, but they may represent relict populations from an e ra where intergradation of the two genotypes was more common. However, all three of these hypotheses
88 require additional testing in a phylogeographic framework to assess them more effectively. Conservation Implications Dicerandra as an endemic genus to th e southeast, is a valuable biologic and intrinsic component of this region. Given the narrow distribution and the isolated nat it is important to consider the conservation implication of any research f ocusing on this genus. This study has addressed the p opulation genetic aspect of these organisms biology and has shown there is considerable diversity found at the population level for the species investigated. As with all conservation decisions it is i mportant to consider multiple biologic aspects of an organism when deciding how we can best aid in the maintenance of these species. All Dicerandra species rely on perturbation to maintain their habitat. Field work conducted for this study found many popul ations along road cuts, power line rights of way, and in recently disturbed natural areas with almost all populations adjacent to later su ccessional closed canopy forest; testament to these species requirement for disturbance. Even wi thin the widely distr ibuted spe c i es of D. linearifolia var. linearifolia D. fumella and D. odoratissima the greatest threat to the long term persistence of populations is habitat loss. This loss most often appears to be the result of habitat succession resulting from fire s uppression or wholesale land conversion for anthropogenic purposes (typically silviculture or agriculture). Research by Menges (1992) has shown the rapid response of populations following disturbance and allowing natural processes, such as periodic fire, to remain within ec osystems is a critical component for the long term persistence of these species.
89 Within the annuals examined for this study, D. linearifolia var. linearifolia D. fumella and D. odoratissima have the largest distribution with many popul ations These populations are likely stable and will persist if suitable habitat remains available While each species has a wide distribution certain areas have a greater presence of populations. For D. linearifo lia var. linearifolia the greatest density of populations occur in western Georgia near the upper Flint River, centered in Talbot and Taylor counties. Dicerandra linearifolia var. robustior is common in the eastern portion of the Florida panhandle betwee n the Ochlockonee and Suwannee Rivers. Dicerandra fumella is widespread in the western portion of the Florida panhandle with populations commonly found in Okaloosa, Walton, Holm es, and Washington counties east of the Escambia Rive and surrounding the Choc tawhatchee River. Dicerandra odoratissima populations are common in eastern Georgia along the Altamaha River north to Emanuel county. These regions represent the core po pulations for these species where efforts to maintain habitat would provide maximum b enefit in terms of conserving the greatest quantity of populations and thus gene tic diversity. Alternatively D. radfordiana has a very narrow distribution only occurring in McIntosh county, Georgia and is currently listed by the state of Georgia as an enda ngered species. This study found the populations of D. radfordiana are not genetically distinct from proximate populations of D. odoratissima despite showing morphological differentiation. We presented t w o approaches for addressing the taxonomy of radfor diana Each approach has distinct conservation implications; either radfordiana and the proximate populations of D. odoratissima are recognized as D. radfordiana (based on their genetic similarity ( Figure 3 5) ) but would lack morphological
90 characters to dis tinguish populations from one another making identif ication in the field impossible or D. radfordiana is recognized as a variety of D. odoratissima thus removing its species rank and any conservation protections this rank carries, but would continu e to re cognize it s unique morphology and evolutionary trajectory. It can often be argued when addressing designation and conservation of threatened or endangered species that the potential damage associated with a type II error (accepting the null of no differen ce where a difference actually exists) is significantly greater than a type I error (rejecting the null when there really is no difference) (Dayton 1998; McGarvey 2007). However, there also exists a delicate balance between available resources for conserv ation and the degree of need of each species. As funding for conservation is typically limited, it is prudent to apply those resources to species where the benefit will be greatest. Part of determining which species and projects will be most beneficial i s identifying previously unrecognized or under studied species that truly merit species status or concerted efforts to aid their survival. Alternatively, understanding which species do not require conservation effort is equally as important to ensure fund ing is only directed toward organisms that truly need it. For example, understanding the biological requirements of these species, determining conservation actions, and establishing priority among species represent signifi cant investments of resources. Di cerandra radfordiana has been the focus of recent conservation efforts by the Georgian Division of Natural Resources to open habitat surrounding certain populations and has benefited from the corporation of the Nature Conservancy and the Georgia Rare Plant Alliance in assessing population trends and identifying the extent of it's
91 range. These efforts have provided invaluable benefits to aid in the persistence of D. radfordiana Regardless of the taxonomic designation of radfordiana these kinds of investme nts are critical to maintaining the unique variation that resulted in the initial recognition of radfordiana popu l ations as being unique from other populations of Dicerandra and any taxonomic revision should not be interpreted as diminishing the evolutiona ry significance of these populations. Future Research This study has made significant advances by increasing the quantity of genetic resources available for this genus and furthering the understanding of genetic distribution and relatedness of species with in the annual clade of Dicerandra Yet, additional questions still remain unanswered. The population level sampling for this study sets the stage for further more detailed work toward understanding the mechanisms responsible for the distributions of thes e species. With additional sequence data, the current geographic sampling could easily be used in a phylogeographic study to address the pattern of expansion within D. linearifolia var. linearifolia in Georgia. Similar questions could be addressed in con nection with the relationship and pattern of expansion between D. fumella and D. linearifolia var. robustior in Florida and D. linearifolia var. linearifolia Georgia. Additional sampling in a few key geographic areas such as central Georgia would help eva luate the strength of the eastern D. linearifolia var. linearifolia populations as a cohesive unit and provide additional insight into the potential role local extinction, land use changes, or habitat reduction have played in the current geographic distrib ution of this taxon These types of analyses could be extended to include the perennial species of peninsular Florida to provide a more complete understanding of how these species have radiated and
92 evolved from their hypothesized glacial refugium in peni nsular Florida following the end of the last glacial maximum. With the addition of a relatively small number of samples, the genus could provide a unique opportunity to investigate the dynamics associated with the climate shift following glacial retreat and the subsequent radiation of a genus of plants endemic to the southeastern United States. While there is substantial genetic evidence for the recognition of D. fumella as a species, presented in this study as well as in the phylogenetic analysis of Oliv eira et al. (2007), the morphological basis for species identification and circumscription would significantly benefit from a more detailed analysis. Variation among the populations of D. fumella is substantial, as noted by Huck (2010), so much so that sh e proposed a hybrid zone with D. linearifolia var. robustior yet hybridization was not supported by genetic data in this study. Several factors could drive this morphological variation: drift, phenotypic plasticity, environmental variation in resource av ailability, or local adaptation. Further studies with ordination analyses based on detailed morphological comparisons, reciprocal transplant or common garden experiments, and microsite comparisons could help advance our understanding of the genetic and ec ological roles associated with the observed phenotypic variation. These types of studies could also be used to assess the rather minimal pattern of morphological differentiation between the western and central Georgia genotypes of D. linearifolia var. lin earifolia Conclusions This st udy aimed to investigate phylogenetic incongruences observed by Oliveira et al. (2007) between multiple accessions of the same species of Dicerandra using a population level approach. Each of the following questions were prop o sed at the onset of the study.
93 Within Dicerandra linearifolia s.l. : 1) Did population structure mirror the east west structure observed in the phylogeny? We found there was general geographic cohesions of Georgia populations of D. linearifolia var. linea rifolia However, populations did not for m exclusive clusters, as illustrated by STRUCTURE analyses and distance based trees. Central Georgia populations showed admixture of both eastern and western genotypes and population 22, found at the eastern edge of the range was composed exclusively of individuals with western genotypes. 2) Is ther e molecular evidence for hybrids between D. linearifiolia var. linearifolia and D. odoratissima ? Putative hybrid individuals identified in the field did carry a mixed genotypic signature however not all putative hybrids showed this. Some hybrids displayed genotypes that were almost exclusively D. linearifolia var. linearifolia or D. odoratissima despite showing intermediate or chimeric m orphology in the field. Additi onally hybridization may be direction al as seen by the presences of individuals with mixed genotypes within a D. odoratissima population but the absence of genotypically mixed i ndividuals within the D. linearifolia var. linearifolia population separated b y only 30 m 3) What is the extent of po lyploidy within the genus? All annual species of Dicerandra have a ploidal level of 2 n =32. While this number indicates this genus is the result of historic genome duplication in an ancestral lineage there was no e vidence of recent polyploidy within the extant members of this genus. 4) Is there molecular support for the recently described species D. fumella ? There was strong support for the recognition of D. fumella at the species rank. Populations of D. fumella formed exclusive clusters in STRUCTURE analyses and within the distance based trees. Additionally D. fumella had considerable genetic differentiation from all other Dicerandra spe c ies investigated. 5) Is there
94 support of a hybridization zone between D. fumella and D. linearifolia var. robustior ? While there is considerable morphological variation at the eastern edge of range giving the appearance of a potential hybridization zone with neighboring D. linearifolia var. linearifolia there was no genetic support for this hypothesis Within D. odoratissima and D. radfordiana : 1) Is there significant genetic structure separating populations of D. odoratissima and D. radfordiana ? The two D. radfordiana populations were found to be genetically dist inct from most but not all populations of D. odoratissima Two populations of D. odoratissima that occur within 11 km of D. radfordiana consistently clustered with both D. radfordiana popu lations in the STRUCTURE analyse s and the distance trees. While th e two D. radfordiana populations and the two proximate D. odoratissima populations are genetically distinct from the remaining D. odoratissima populations morphologically these two D. odoratissima populations are indistinguishable from the remaining popul ations of D. odoratissima It is likely these are in the process of diverging from the remaining D. odoratissima populations ; however insufficient time has elapsed for them to be both genetically and morphological distinct. 2) Does geography play a sign ificant role in explaining the observed non monophyly of D. odoratissima ? Mantel tests showed there was a significant correlation between geographic distance and genetic distance within D. odoratissima and D. radfordiana The D. radfordiana populations a nd their proximate D. odoratissima populations are found at the far southeastern edge of range and may represent the beginning of speciation by isolation. Overall, there is significant genetic diversity found at the populations level for all Dicerandra annuals, 11.39% of the total genetic diversity is found among populations
95 within species. This value is similar to the proportion of genetic variation found between species, 12.49%. The isolated nature of many populations, limited dispersa l capability, and frequent visitation from generalist pollinators produce conditions where drift can cause significant differentiation between populations, even within the same geographic area. However the obligate outcrossing nature of these plants and t he potential for multi generational populations resulting from recruitment from the seed bank are likely stabilizing factors maintaining heterozygosity within populations. With the exception of D. radfordiana all the annual species are widely distributed w ithin their respective ranges and often populations containing individuals exist. Disturbance is require d for the maintenance of their habitat and allowing for continued natural disturbance such as periodic fire it is likely many of the se populations will persist well into the future.
96 APPENDIX PAIRWISE POPULATION COMPARISON S
97 Table A 1. Pairwise F ST values. Asterisks indicate non significant Bonferroni corrected p values (p>0.05). Column and r ow headers are color coded by species: Blue: D. linearifolia var. linearifolia Yellow: D. radfordiana Purple: D. odoratissima Red: D. fumella Green: D. linearifolia var. robustior White: hybrid populations. LIN_2 LIN_3 RAD_4 ODO_9 LIN_10 ODO_1 3 ODO_1 4 ODO_1 5 ODO_1 6 ODO_1 7 FUM_18 ODO_2 0 L IN_3 0.0475 RAD_4 0.2475 0.3164 ODO_9 0.2569 0.3131 0.3345 LIN_10 0.2307 0.2765 0.3195 0.3445 ODO_1 3 0.2283 0.2965 0.3113 0.2074 0.3301 ODO_1 4 0.1799 0.1931 0.1997 0.2737 0.3076 0.1768 ODO_1 5 0.1654 0.2322 0.3039 0.1843 0.2754 0.1788 0.1918 ODO_1 6 0.1359 0.2098 0.2192 0.1460 0.2497 0.1672 0.1854 0.0561 ODO_1 7 0.1500 0.2144 0.2370 0.1755 0.2428 0.1553 0.2057 0.0542 0.0481* FUM_18 0.1991 0.2548 0.2932 0.2573 0.2831 0.2477 0.2699 0.2221 0. 1881 0.2049 ODO_2 0 0.3066 0.3587 0.1815 0.3140 0.3804 0.3311 0.3114 0.2908 0.2290 0.2444 0.3086 ODO_2 1 0.2062 0.2628 0.1268 0.2180 0.2591 0.1967 0.2214 0.1775 0.1695 0.1670 0.2277 0.0797 LIN_22 0.0896 0.0834 0.2498 0.2227 0.1930 0.2136 0.1410 0.1658 0.1263 0.1318 0.2019 0.2774 HYB_23 0.0997 0.1475 0.1997 0.1629 0.1417 0.1176 0.1671 0.1084 0.1114 0.1350 0.1653 0.2215 LIN_24 0.1310 0.1774 0.2056 0.1521 0.1310 0.1648 0.1806 0.1535 0.1515 0.1550 0.1681 0.2389 HYB_25 0.0772 0.1137 0.2106 0.1764 0.1331 0 .1652 0.1553 0.0968 0.0916 0.1264 0.1685 0.2691 LIN_27 *0.0664 0.0685 0.3149 0.2719 0.2430 0.2530 0.1848 0.2564 0.1548 0.1738 0.2161 0.3235 LIN_28 0.1316 0.1150 0.2785 0.2885 0.2687 0.2662 0.1992 0.1476 0.2081 0.1988 0.2595 0.3259 LIN_29 0.0673 0.1209 0 .1922 0.2173 0.1597 0.1891 0.1361 0.1417 0.1314 0.1428 0.1721 0.2436 ODO_3 0 0.2610 0.3374 0.3839 0.2164 0.3759 0.2236 0.2934 0.2266 0.1983 0.1904 0.3076 0.3815 ODO_3 1 0.1775 0.2422 0.2692 0.1830 0.2625 0.2002 0.2201 0.1371 0.1136 0.1155 0.2348 0.2710 LI N_32 0.0836 0.1518 0.2150 0.1854 0.2088 0.1817 0.1612 0.1097 0.0964 0.0948 0.1473 0.2594 LIN_33 *0.0846 0.1439 0.2650 0.1565 0.2139 0.1621 0.1553 0.1715 *0.0688 0.1155 0.1311 0.2614 FUM_34 0.1863 0.2331 0.2791 0.2708 0.2610 0.2548 0.2375 0.2152 0.2007 0. 2113 0.0507 0.3176 FUM_35 0.2899 0.3110 0.3897 0.3943 0.3876 0.3766 0.3134 0.3421 0.3271 0.3238 0.1604 0.4283 FUM_36 0.1818 0.2194 0.2823 0.2871 0.2734 0.2603 0.2555 0.2159 0.2202 0.2111 0.0952 0.3079 FUM_37 0.2981 0.3392 0.3714 0.3756 0.3442 0.3872 0.3 617 0.3416 0.3098 0.3046 0.1804 0.3894 FUM_38 0.2054 0.2617 0.3024 0.2629 0.2912 0.2531 0.2625 0.2221 0.2021 0.2066 *0.0514 0.3059 ROB_40 0.2676 0.3229 0.4280 0.3482 0.3782 0.4187 0.3606 0.3288 0.2426 0.2803 0.3091 0.4509 ROB_41 0.1168 0.1497 0.3046 0.2 632 0.2033 0.3036 0.2317 0.2260 0.1621 0.1865 0.1918 0.3455 RAD_42 0.3805 0.4333 0.1834 0.4410 0.4696 0.4311 0.3512 0.3798 0.3499 0.3299 0.4042 0.2258
98 Table A 1. Continued ODO_21 LIN_22 HYB_23 LIN_24 HYB_25 LIN_27 LIN_28 LIN_29 ODO_30 OD O_31 LIN_32 LIN_33 LIN_2 LIN_3 RAD_4 ODO_9 LIN_10 ODO_13 ODO_14 ODO_15 ODO_16 ODO_17 FUM_18 ODO_20 ODO_21 LIN_22 0.1779 HYB_23 0.1158 0.0914 LIN_24 0.1385 0.0921 0.0531 HYB_25 0.1707 0.0668 0.0435 0.0821 LIN_27 0.2196 0.0883 0.0919 0.1163 0.1089 LIN_28 0.2244 0.1228 0.1233 0.1688 0.1410 0.08 36 LIN_29 0.1593 0.0802 0.1007 0.0905 0.0899 0.1010 0.1272 ODO_30 0.2413 0.2558 0.1769 0.2149 0.2204 0.2917 0.2930 0.2309 ODO_31 0.1511 0.1654 0.1014 0.1359 0.1171 0.2109 0.2323 0.1484 0.1700 LIN_32 0.1610 0.0602 0.0860 0.0792 0.0717 0.0974 0.1455 0.0807 0.2399 0.1322 LIN_33 0.1568 0.0995 *0.0571 0.0709 *0.0616 0.0933 0.1528 0.0888 0.2302 0.1603 0.0668 FUM_34 0.2171 0.1885 0.1456 0.1507 0.1417 0.2129 0.2267 0.1641 0.3167 0.2327 0.1445 0.1628 FUM_35 0.3296 0.3047 0.2726 0.2891 0.2 742 0.3187 0.3011 0.2565 0.4518 0.3599 0.2802 0.2764 FUM_36 0.2253 0.1869 0.1723 0.1754 0.1682 0.2063 0.2247 0.1658 0.3119 0.2286 0.1543 0.1660 FUM_37 0.2979 0.2856 0.2304 0.2356 0.2469 0.3067 0.3273 0.2595 0.4230 0.3245 0.2460 0.2666 FUM_38 0.2184 0.19 93 0.1688 0.1591 0.1799 0.2048 0.2346 0.1637 0.3145 0.2312 0.1400 0.1380 ROB_40 0.3283 0.2497 0.2355 0.2442 0.2228 0.2901 0.3138 0.2441 0.4446 0.3408 0.2329 0.2293 ROB_41 0.2155 0.0946 0.0958 *0.0677 0.0948 0.1121 0.1806 0.1153 0.3155 0.1908 *0.0843 *0.0 899 RAD_42 0.2453 0.3417 0.3267 0.3454 0.3612 0.4078 0.3823 0.3080 0.5031 0.4085 0.3533 0.3866
99 Table A 1. Continued FUM_34 FUM_35 FUM_36 FUM_37 FUM_38 ROB_40 ROB_41 RAD_42 LIN_2 LIN_3 RAD_4 ODO_9 LIN_10 OD O_13 ODO_14 ODO_15 ODO_16 ODO_17 FUM_18 ODO_20 ODO_21 LIN_22 HYB_23 LIN_24 HYB_25 LIN_27 LIN_28 LIN_29 ODO_30 OD O_31 LIN_32 LIN_33 FUM_34 FUM_35 0.1007 FUM_36 0.0221 0.1077 FUM_37 0.0976 0.1759 0.0913 FUM_38 0.0294 0.1215 *0.0186 0.1043 ROB_40 0.3294 0.4474 0.3179 0.4602 0.3431 ROB_41 0.1756 0.3125 0. 1763 0.2872 0.1979 *0.1922 RAD_42 0.3975 0.4914 0.4020 0.5085 0.4261 0.5477 0.4482
100 Table A 2 D Column and r ow headers are color coded by species: Blue: D. linearifolia var. linearifolia Yellow: D. radfordiana Pur ple: D. odoratissima Red: D. fumella Green: D. linearifolia var. robustior White: hybrid populations. LIN_2 LIN_3 RAD_4 ODO_9 LIN_10 ODO_13 ODO_14 ODO_15 ODO_16 ODO_17 FUM_18 ODO_20 LIN_3 0.1167 RAD_4 0.8440 0.8966 ODO_9 0.8236 0.8546 0.9264 LIN_10 0.6879 0.6980 0.8439 0.9261 ODO_13 0.8012 0.8585 0.8739 0.4827 0.9232 ODO_14 0.5613 0.4848 0.5036 0.7254 0.8399 0.4800 ODO_15 0.6591 0.6490 0.8214 0.6192 0.7597 0.6150 0.6679 ODO_16 0.5276 0.6918 0.7637 0.4240 0.8136 0.5771 0.5991 0.4189 ODO_17 0.5970 0.6922 0.7977 0.5078 0.7622 0.5248 0.6706 0.3088 0.1670 FUM_18 0.7600 0.8031 0.9801 0.7889 0.8711 0.8577 0.8906 0.8915 0.7318 0.8008 ODO_20 0.9188 0.9311 0.3578 0.7054 0.9508 0.7811 0.762 2 0.7948 0.6478 0.6757 0.8850 ODO_21 0.8330 0.8586 0.4099 0.6352 0.7946 0.6421 0.6956 0.6762 0.6698 0.6266 0.8681 0.1673 LIN_22 0.3268 0.2177 0.7440 0.7070 0.5744 0.7749 0.4278 0.5837 0.5290 0.5242 0.7972 0.8030 HYB_23 0.4513 0.5241 0.8250 0.5658 0.483 7 0.5284 0.6140 0.5804 0.5264 0.6327 0.7404 0.7222 LIN_24 0.5984 0.6320 0.8019 0.5191 0.4240 0.6906 0.6488 0.7551 0.7252 0.7205 0.7237 0.7596 HYB_25 0.3278 0.3668 0.8474 0.5973 0.4318 0.6937 0.5400 0.5131 0.4142 0.5868 0.7488 0.9121 LIN_27 0.2321 0.1770 0.9148 0.8102 0.6797 0.7951 0.5545 0.7823 0.6104 0.6140 0.7963 0.8595 LIN_28 0.4303 0.2853 0.8600 0.8589 0.7616 0.8617 0.5563 0.5086 0.7712 0.7088 0.9267 0.9116 LIN_29 0.2889 0.4255 0.7728 0.8365 0.5624 0.8602 0.4846 0.7800 0.6525 0.7114 0.7966 0.8400 ODO_30 0.7327 0.8234 0.9072 0.3932 0.8734 0.4122 0.6984 0.5549 0.5291 0.4832 0.8706 0.7799 ODO_31 0.6555 0.7147 0.7902 0.4616 0.7256 0.5543 0.6573 0.4915 0.3705 0.3810 0.8624 0.6607 LIN_32 0.4160 0.5584 0.8534 0.6371 0.7498 0.7453 0.6032 0.6637 0.4793 0. 4699 0.6907 0.8729 LIN_33 0.3887 0.4824 0.9117 0.4589 0.7187 0.5268 0.5330 0.7215 0.3239 0.5032 0.5577 0.7186 FUM_34 0.7094 0.7076 0.8532 0.8307 0.7638 0.8447 0.7517 0.7450 0.8025 0.8328 0.1558 0.8912 FUM_35 0.7277 0.6554 0.8937 0.9102 0.8541 0.8574 0.6 863 0.8106 0.9140 0.8839 0.3194 0.9284 FUM_36 0.6697 0.6446 0.9169 0.8932 0.8161 0.8872 0.8055 0.7677 0.8704 0.8061 0.2785 0.8745 FUM_37 0.8537 0.8134 0.8695 0.9088 0.7541 0.9554 0.9349 0.9395 0.9448 0.9034 0.4126 0.8393 FUM_38 0.8342 0.8388 0.9754 0.80 29 0.8891 0.8260 0.8688 0.8745 0.8366 0.8367 0.1403 0.8419 ROB_40 0.6658 0.6806 0.9694 0.6817 0.7778 0.8963 0.8497 0.7206 0.6105 0.7404 0.8025 0.9653 ROB_41 0.3710 0.3675 0.9027 0.7383 0.4915 0.9364 0.6825 0.6533 0.5896 0.6867 0.6494 0.9315 RAD_42 0.884 7 0.9274 0.2307 0.9170 0.9841 0.8573 0.6746 0.8552 0.8356 0.7561 0.9734 0.3001
101 Table A 2 Continued ODO_21 LIN_22 HYB_23 LIN_24 HYB_25 LIN_27 LIN_28 LIN_29 ODO_30 ODO_31 LIN_32 LIN_33 LIN_2 LIN_3 RAD_4 ODO_9 LIN_10 ODO_13 ODO_14 ODO_15 ODO_16 ODO_17 FUM_18 ODO_20 ODO_21 LIN_22 0.6981 HYB_23 0.4938 0.4263 LIN_24 0.5888 0.4230 0.2496 HYB_25 0.7840 0.2934 0.2081 0.4164 LIN_27 0.7808 0.2838 0.3965 0.5103 0.4638 LIN_28 0.7891 0.4023 0.4767 0.6643 0.5465 0.2473 LIN_29 0.7567 0.3755 0.5742 0.4813 0.5011 0.4803 0.5016 OD O_30 0.6223 0.7242 0.5534 0.6751 0.7032 0.7417 0.7765 0.7966 ODO_31 0.5060 0.6055 0.4169 0.5774 0.4809 0.6967 0.7909 0.6867 0.3397 LIN_32 0.7962 0.3274 0.5318 0.4800 0.4291 0.4744 0.6179 0.5300 0.7730 0.6111 LIN_33 0.6435 0.4456 0.3021 0.3853 0 .3597 0.2864 0.6174 0.5576 0.6206 0.6396 0.4947 FUM_34 0.8127 0.7196 0.6640 0.6641 0.6350 0.7449 0.7770 0.7815 0.8635 0.8236 0.7012 0.6997 FUM_35 0.8669 0.7987 0.8233 0.8516 0.7959 0.7685 0.7039 0.7602 0.9539 0.8822 0.8065 0.7376 FUM_36 0.8310 0.7006 0 .7460 0.7344 0.7150 0.7175 0.7460 0.7311 0.8756 0.8140 0.7030 0.7169 FUM_37 0.8362 0.8258 0.7396 0.7365 0.7823 0.8003 0.8798 0.8872 0.8945 0.8462 0.7935 0.8144 FUM_38 0.8431 0.8152 0.7973 0.7173 0.8534 0.7388 0.8324 0.8011 0.8683 0.8648 0.7037 0.6165 RO B_40 0.8854 0.6058 0.7053 0.7272 0.6265 0.6654 0.7735 0.7753 0.8006 0.7796 0.6355 0.5962 ROB_41 0.7792 0.2867 0.3638 0.2373 0.3588 0.3094 0.5561 0.4877 0.7772 0.6123 0.3667 0.3655 RAD_42 0.4759 0.7696 0.8965 0.9128 0.9789 0.9401 0.8423 0.8075 0.9299 0.86 15 0.8854 0.9545
102 Table A 2. Continued FUM_34 FUM_35 FUM_36 FUM_37 FUM_38 ROB_40 ROB_41 RAD_42 LIN_2 LIN_3 RAD_4 ODO_9 LIN_10 ODO_13 ODO_14 ODO_15 ODO_16 ODO_17 F UM_18 ODO_20 ODO_21 LIN_22 HYB_23 LIN_24 HYB_25 LIN_27 LIN_28 LIN_29 ODO_30 ODO_31 LIN_32 LIN_33 FUM_34 FUM_35 0.1727 FUM_36 0.0654 0.2071 FUM_37 0.1897 0.2513 0.1973 FUM_38 0.0873 0.2253 0.0467 0.2202 ROB_40 0.8237 0.8335 0.8254 0.9163 0.9086 ROB_41 0.5751 0.6865 0.5738 0.6907 0.7135 0.3159 RAD_42 0.9050 0.8804 0.9514 0.9567 0.9877 0.9593 0. 9525
103 LIST OF REFERENCES A bdelkrim J, Robertson BC, Stanton JL, Gemmel NJ (2009) Fast, cost effective development of species specific microsatellite markers by genomic sequencing. BioTechniques 46 185 192. Allentoft ME, Schuster SC, Holdaway RN et al. (2009) Identifica tion of microsatellites from an ext inct moa species using high throughput (454) sequence data. BioTechniques 46 195 200. Bentham G (1848) Dicerandra DC Proceedings 12 242 243. Castoe T, Poole A, GU W, et al. (2010) Rapid identificatio n of thousands of copperhead snake ( Agkistrodon contortrix ) microsatellite loci from modest amounts of 454 shotgun genome sequence. Molecular Ecology Resources 10 341 347. Comes H, Abbott R (2001) Molecular phylogeography, reticulation, and lineage sorti ng in Mediterranean Senecio sect. Senecio (Asteraceae). Evolution 55 1943 1962. Dayton P (1998) Reversal of the Burden of Proof in Fisheries Management. Science 279 821 822. De Queiroz K (2007) Species concepts and species delimitation. Systematic Biol ogy 56 879 886. Doyle JJ, Doyle JL (1987) A rapid isolation procedure for small quantities of fresh tissue. Phytochemical Bulletin 19 11 15. Earl D, V onHoldt B (2012) STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and imple menting the Evanno method. Conservation Genetics Resources 4 359 361. Edwards C, Solits D, Soltis P (2008) Using patterns of genetic structure based on microsatellite loci to test hypotheses of current hybridization, ancient hybridization and incomplete lineage sorting in Conradina (Lamiaceae). Molecular Ecology 17 5157 5174. Estill J, Cruzan M (2001) Phytogeography of rare plant species endemic to the southeastern United States. Castanea 66 3 23. Evanno G, Regnaut S, Goudet J (2005) Detecting the num ber of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology 14 2611 2620. Evans M, Menges E, Gordon D (2004) Mating systems and limits to seed production in two Dicerandra mints endemic to Florida scrub. Biodiversit y and Conservation 13 1819 1832.
104 Excoffier L, Lischer H (2010) Arlequin suite ver 3.5: a new series of programs to perform population genetics ana lyses under Linux and Windows. Molecular Ecology Resources 10 564 567. Florida Natural Areas Inventory (FNA I) (2010) Guide to the natural communities of Florida. 2010 edition, pp. 32 44. Florida Natural Areas In ventory, Tallahassee, Florida. Gordon A (2008) Fastx toolkit http://hannonlab.cshl.ed u/fastx_toolkit/index.html Last accessed July 2012. Greilhuber J, Temsch E, Loureiro J (2007) Nuclear DNA content measurements. In: Flow cytometry with plant cells: analysis of genes, chromosomes and genomes (ed. Dolezel and Suda), pp. 67 99. Wiley VCH, Weinheim, Germany. Huck R (1987) Systematics and evolution of Dicerandra (Labiatae). Phanerogamarum monographiae 19 1 343. Huck R, Chambers H (1997) Polyploidy: a factor in the evolution of Dicerandra Benth.(Labiatae). Edinburgh Journal of Botany 54 21 7 229. Huck R, Judd W, Whitten W, et al. (1989) A new Dicerandra (Labiatae) from the Lake Wales Ridge of Florida, with a cladistic analysis and discussion of endemism. Systematic Botany 14 197 212. Huck, R (2010) Dicerandra fumella (Lamiaceae), a new spe cies in the Florida panhandle and adjacent Alabama, with comments on the D. linearifolia complex. Rhodora 112 215 227. Jakob S, Blattner F (2006) A chloroplast genealogy of Hordeum (Poaceae): Long term persisting haplotypes, incomplete lineage sorting, r egional extinction, and the consequences for phylogenetic inference. Molecular Biology and Evolution 23 1602 1612. Jakobsson M, Rosenberg N (2007) CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in an a lysis of population structure. Bioinformatics 23 1801 1806. Joly S, McLenachan P, Lockhart P (2009) A statistical approach for distinguishing hybridization and incomplete lineage. American Naturalist 174 54 70. Jost L (2008) GST and its relatives do n ot measure differentiation. Molecular Ecology 17 4015 4026. Kral R (1982) Some notes on Dicerandra (Lamiaceae). SIDA 9 238 262.
105 Levitt R, Kiser M, Dragwa C, et al. (1994) Fluorescence based resource for semiautomated genomic analyses using microsatelli te m arkers. Genomics 24 361 365. McCormick K, Deyrup M, Menges E et al. (1993) Relevance of chemistry to conservation of isolated populations: the case of volatile leaf components of Dicerandra mints. Proceedings of the National Academy of Science 90 7 701 7705 McDonald D, Hamrick J (1996) Genetic variation in some plants of Florida scrub. American Journal of Botany 83 21 27. McGarvey D (2007) Merging precaution with sound s cience under the Endangered Species Act. BioScience 57 65 70. Meirmans P, Va n Tiederen P (2004) GENOTYPE and GENODIVE: two programs for the analysis of genetic d iversity of asexual organisms. Molecular Ecology Notes 4 792 794. Meirmans P Hedrick P (2011) Assessing population structure: FST and related measures. Molecular Ecolog y Resources 11 5 18. Menges E (1992) Habitat preferences and response to disturbance for Dicerandra frutescens a Lake Wales Ridge (Florida) endemic plant. Bulletin of the Torrey Botanical Club 119 308 313. Menges E, Dolan R, Yahr R, Gordon D (2001) Co mparative genetics of seven plants endemic to Florida's Lake Wales Ridge. Castanea 66 98 114. Menges E, McIntyre P, Finer M, et al. (1999) Microhabitat of the narrow Florida scrub endemic Dicerandra christmanii with comparisons to its congener D. frutes cens Journal of the Torrey Botanical Society 126 24 31. Menges ES, Dolan RW, Pickert R, et al. (2010) Genetic Variation in Past and Current Landscapes: Conservation Implications Based on Six Endemic Florida Scrub Plants. International Journal of Ecology 2010 1 13. Merchant A (2006) Variation in functional morphology and physiological responses of Dicerandra (Lamiaceae) congeners native to sandhill habita ts and Florida scrub. Doctoral d issertation, University of F lorida, Gainesville, Florida. Myers R, E wel J (1990) In: Ecosystems of Florida (ed. Myers and Ewel), pp. 35 69, 150 193. University Presses of F lorida, Gainesville, Florida. Oliveira L, Huck R, Gitzendanner M, et al. (2007) Molecular phylogeny, biogeography, and systematics of Dicerandra (Lamiac eae), a genus endemic to the southeastern United States. American Journal of Botany 94 1017 1027.
106 Peakal R, Smouse P (2006) GENALE X 6: genetic analysis in Excel. Population genetic soft ware for teaching and research. Molecular Ecology Notes 6 288 295. Ramdhani S, Barker N, Cowling R (2011) Revisiting monophyly in Haworthia Duval (Asphodelaceae): Incongruence, hybridization and contemporary speciation. Taxon 60 1001 1014. Pritchard J, Stephens M, Donnelly P (2000) Inference of population structure usin g multilocus genotype data. Genetics 155 945 959. Rice W (1989) The sequential Bonferroni test. Evolution 43 223 225. Rieseberg L Soltis D (1991) Phylogenetic c onse quences of cytoplasmic gene flow in p lants. Evolutionary Trends in Plants 5 65 84. Ro senberg N (2004) Distruct: a program for the graphical display of population structure Molecular Ecology Notes 4 137 138. implementation of the GENEPOP software for Windows and Linux. Molecular Ecology Resource s 8 103 106. Rozen S, Skaletsky H (2000) Primer3 on the WWW for general users and for biologist programmers. In: Bioinformatics Methods and Protocols: Methods in Molecular Biology. (ed. Krawetz S, Misener S), pp. 365 386. Hum ana Press, Totowa, New Jersey Schuelke M (2000) An economic method for the fluorescent labeling of PCR fragments. Nature biotechnology 18 223 224. Shinners L (1962) Synopsis of Dicerandra (Labiatae). SIDA 1 89 91. Suda J, Trvncek P (2006) Reliable DNA ploidy determination in de hydrated tissues of vascular plants by DAPI flow cytometry -new prospects for plant research. Cytometry. T he journal of the International Society for Analytical Cytology 69 273 280. Sutton J, Robertson B, Jamieson I (2011) Dye shift: a neglected source o f genotyping error in molecular ecology. Molecular Ecology Resources 11 514 520. Swofford D (2003) PAUP* Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4. Sinauer Associat es, Sunderland, Massachusetts. Syring J, Farrell K, Businsky R et al. (2007) Widespread genealogical nonmonophyly in species of Pinus subgenus Strobus Systematic Biology 56 163 181. USFWS (1993) Endangered and threatened wildlife and plants; review of plant taxa for listing as endangered or threatened species. Fe deral Register 58 51144 51190.
107 USFWS (2007) Endangered and threatened wildlife and plants; 5 year review of 22 southeastern species. Federal Register 72 20866 20868. Wahlund S (1928) Composition of populations and correlation appearances view in relati on to the studies of inheritance. Hereditas 11 65 106. Wendel J, Doyle J (1998) Phylogenetic incongruence: window into genome history and molecular evolution. In: Molecular Systematics of Plants II (ed. Soltis, Soltis, and Doyle), pp. 265 296. Kluwer Academic Publi shers, Norwell, Massachusetts. Wheeler Q, Meier R (2000) In: Species Concepts and Phylogenetic Theory (ed. Wheeler and Meier), pp. 1 208. Columbia University Press, West Sussex, New York. Winn A (1996a) Adaptation to fine grained environmental variation: an analysis of within individual leaf variation in an annual p lant. Evolution 50 1111 1118. Winn A (1996b) The contributions of programmed developmental change and phenotypic plasticity to within individual variation in leaf traits in Dicerandra linearifolia Journal of Evolutionary Biology 9 737 752. Winn A (199 9) Is seasonal variation in leaf traits adaptive for the annual plant Dicerandra linearifolia ? Journal of Evolutionary Biology 12 306 313. Yu Y, Than C, Degnan J H, Nakhleh L (2011) Coalescent histories on phylogenetic networks and detection of hybridizat ion despite incomplete lineage s orting. Systematic Biology 60 138 149.
10 8 BIOGRAPHICAL SKETCH Adam Payton completed high school in Sandpoint, Idaho and attended Western State College of Colorado in Gunnison, for his undergraduate education. It was his exp eriences living and recreating in the western US that sparked his interest in biology. This personal interest depended into a professional interested and as he pursued his bot anist during the summers and analyzed data during the school semesters. His work with the Bureau of Land Management was pivotal in shaping his interests and desi re to further his education in g raduate school. In 2008 he graduated with a degree in ecology and evolutionary biology. In 2009 he began his graduate career at the University of Florida under the supervision of Douglas Soltis, where he studied population genetics and the evolution of the southeastern endemic genus Dicerandra Adam grad uated with a Master of Science in 2012 and will continue his work with population genetics and evolution in the McDaniel lab at the University of Florida.