<%BANNER%>

Phylogeny Reconstruction and Conservation Genetics of Conradina and Related Southeastern U.S. Endemic Mints (Lamiaceae)

Permanent Link: http://ufdc.ufl.edu/UFE0021647/00001

Material Information

Title: Phylogeny Reconstruction and Conservation Genetics of Conradina and Related Southeastern U.S. Endemic Mints (Lamiaceae)
Physical Description: 1 online resource (168 p.)
Language: english
Creator: Edwards, Christine Elizabeth
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: clinopodium, conradina, conservation, dicerandra, florida, genetic, heterozygosity, hybridization, lamiaceae, microsatellites, phylogeny, piloblephis, polymorphism, southeastern, stachydeoma
Botany -- Dissertations, Academic -- UF
Genre: Botany thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Conradina, comprising six described species, is a morphologically homogeneous group of narrow-leaved, aromatic mints endemic to the southeastern U.S. Despite the fact that several Conradina species are endangered and are targets of federal recovery programs, species limits in Conradina are unclear, several federally endangered species are of questionable validity, and hybridization has been proposed. The goals of this study were to reconstruct the evolutionary relationships among Conradina species, to clarify the relationship of Conradina to related mint genera such as Dicerandra, Clinopodium, Piloblephis, and Stachydeoma, to clarify species debates in Conradina, to understand patterns of genetic structure within and among each species of Conradina, and to use results of this study to aid in conservation efforts. A molecular phylogeny was inferred by sequencing four plastid regions and ITS and from each species of Conradina and related genera. ITS sequence data strongly supported the monophyly of Conradina, in agreement with evidence from morphology, but plastid sequence data did not support a monophyletic Conradina and placed the genus as paraphyletic to Clinopodium, Stachydeoma, and Piloblephis. Similar plastid haplotypes were shared by different genera, perhaps due to incomplete lineage sorting or hybridization. Next, three members of the nuclear GapC gene family were isolated, two of which were used to reconstruct the phylogeny of Conradina and related mints. Separate phylogenetic analyses of the two GapC loci did not resolve species relationships. Two approaches were used to concatenate and carry out combined analyses of the two heterozygous GapC loci with the ITS and plastid data sets. Trees resulting from the two concatenation approaches were similar in the resolution and support of generic relationships, but differed substantially in resolution and support for relationships within Conradina, probably due to differing treatment of conflict in the data set. Next, methods of concatenation for combined analyses of multiple heterozygous nuclear genes were investigated. All available methods of concatenating data from more than one heterozygous nuclear gene were evaluated to assess their utility, and six approaches were used to combine the ITS, plastid and two heterozygous GapC data sets. Two of these methods allow for multiple placements of individuals in a single tree and allow relationships to collapse in the presence of conflict among data sets, so these methods may be the best ways to illustrate the conflict resulting from heterozygosity. Next, microsatellite loci were isolated, characterized, and screened for amplification success in Conradina. These microsatellite loci were used to genotype individuals in populations of each species of Conradina in order to clarify species limits, investigate genetic structure in Conradina, and test hypotheses of interspecific hybridization vs. incomplete lineage sorting suggested by phylogenetic analyses. Microsatellite data differentiated the six described species, and little hybridization was apparent, supporting the hypothesis of incomplete lineage sorting as the primary source of conflict in phylogenetic analyses. All described species appear to be valid biological entities, so the current listing status of the endangered species of Conradina is appropriate; however, two populations that were thought to be C. etonia are highly genetically differentiated, suggesting that they may possibly merit description as a new species.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Christine Elizabeth Edwards.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Soltis, Pamela S.
Local: Co-adviser: Soltis, Douglas E.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021647:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021647/00001

Material Information

Title: Phylogeny Reconstruction and Conservation Genetics of Conradina and Related Southeastern U.S. Endemic Mints (Lamiaceae)
Physical Description: 1 online resource (168 p.)
Language: english
Creator: Edwards, Christine Elizabeth
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: clinopodium, conradina, conservation, dicerandra, florida, genetic, heterozygosity, hybridization, lamiaceae, microsatellites, phylogeny, piloblephis, polymorphism, southeastern, stachydeoma
Botany -- Dissertations, Academic -- UF
Genre: Botany thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Conradina, comprising six described species, is a morphologically homogeneous group of narrow-leaved, aromatic mints endemic to the southeastern U.S. Despite the fact that several Conradina species are endangered and are targets of federal recovery programs, species limits in Conradina are unclear, several federally endangered species are of questionable validity, and hybridization has been proposed. The goals of this study were to reconstruct the evolutionary relationships among Conradina species, to clarify the relationship of Conradina to related mint genera such as Dicerandra, Clinopodium, Piloblephis, and Stachydeoma, to clarify species debates in Conradina, to understand patterns of genetic structure within and among each species of Conradina, and to use results of this study to aid in conservation efforts. A molecular phylogeny was inferred by sequencing four plastid regions and ITS and from each species of Conradina and related genera. ITS sequence data strongly supported the monophyly of Conradina, in agreement with evidence from morphology, but plastid sequence data did not support a monophyletic Conradina and placed the genus as paraphyletic to Clinopodium, Stachydeoma, and Piloblephis. Similar plastid haplotypes were shared by different genera, perhaps due to incomplete lineage sorting or hybridization. Next, three members of the nuclear GapC gene family were isolated, two of which were used to reconstruct the phylogeny of Conradina and related mints. Separate phylogenetic analyses of the two GapC loci did not resolve species relationships. Two approaches were used to concatenate and carry out combined analyses of the two heterozygous GapC loci with the ITS and plastid data sets. Trees resulting from the two concatenation approaches were similar in the resolution and support of generic relationships, but differed substantially in resolution and support for relationships within Conradina, probably due to differing treatment of conflict in the data set. Next, methods of concatenation for combined analyses of multiple heterozygous nuclear genes were investigated. All available methods of concatenating data from more than one heterozygous nuclear gene were evaluated to assess their utility, and six approaches were used to combine the ITS, plastid and two heterozygous GapC data sets. Two of these methods allow for multiple placements of individuals in a single tree and allow relationships to collapse in the presence of conflict among data sets, so these methods may be the best ways to illustrate the conflict resulting from heterozygosity. Next, microsatellite loci were isolated, characterized, and screened for amplification success in Conradina. These microsatellite loci were used to genotype individuals in populations of each species of Conradina in order to clarify species limits, investigate genetic structure in Conradina, and test hypotheses of interspecific hybridization vs. incomplete lineage sorting suggested by phylogenetic analyses. Microsatellite data differentiated the six described species, and little hybridization was apparent, supporting the hypothesis of incomplete lineage sorting as the primary source of conflict in phylogenetic analyses. All described species appear to be valid biological entities, so the current listing status of the endangered species of Conradina is appropriate; however, two populations that were thought to be C. etonia are highly genetically differentiated, suggesting that they may possibly merit description as a new species.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Christine Elizabeth Edwards.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Soltis, Pamela S.
Local: Co-adviser: Soltis, Douglas E.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021647:00001


This item has the following downloads:


Full Text

PAGE 1

1 PHYLOGENY RECONSTRUCTION AND CONSERVATION GENETICS OF Conradina AND RELATED SOUTHEASTERN U.S. ENDEMIC MINTS (LAMIACEAE) By CHRISTINE E. EDWARDS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

2 2007 Christine E. Edwards

PAGE 3

3 ACKNOWLEDGMENTS I thank my advisors, Pamela Soltis and Dougl as Soltis, and my committee members, David Reed and Walter Judd, for support and guidance throughout my PhD program. I also thank former committee members Rebecca Kimball and Doria Gordon; current and former Soltis lab members, including Chuck Bell, Luiz Oliviera, Ma tt Gitzendanner, Andrew Doust, Josh Clayton, Michael Moore, Ashley Morris, Monica Arak aki, Sam Brockington and Vaughn Symonds for their interesting discussions, help with experi mental design, lab procedures, and analytical methods; Kent Perkins, Norris Williams, and the FLAS herbarium staff and students for assistance with collections and loans; Phil Ca ntino, Jay Walker, Kurt Neubig, Richard Abbott and Luiz Oliviera for plant material; David Lefkowitz, Amber Pouncey, Zera Damji, Suneel Modani, and Aisha Goodman for assistance with la b work; Alan Prather and Rachel Williams for access to unpublished data; and Claude Bailey, Br ian Wender, Eric Tillman, Gretchen Ionta, Sam Brockington, Chuck Bell, Jason Ulev, Ll oyd and Mary Edwards, Ann Cox, David Lefkowitz, and Pam and Doug Soltis for assistance with field work. I th ank my family, friends, and lab members for moral support and encour agement throughout my dissertation. Funding for this project was provided by the Florida Native Plant Society, Fl orida Division of Forestry’s Florida Statewide Endangered and Threatened Plant Conservation Program, the Garden Club of America’s Catherine Beattie Fellowship, Sigma XI Grants-in-Aid of Research, a Graduate Student Research Award from the American Society of Plant Taxonomists, and a Botanical Society of America Genetics Section Award.

PAGE 4

4 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................3 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES................................................................................................................ .........8 ABSTRACT....................................................................................................................... ............10 CHAPTER 1 GENERAL INTRODUCTION..............................................................................................12 2 MOLECULAR PHYLOGENY OF CONRADINA AND OTHER SCRUB MINTS (LAMIACEAE) FROM THE SOUTHEASTERN USA: EVIDENCE FOR HYBRIDIZATION IN PLEI STOCENE REFUGIA?............................................................15 Introduction................................................................................................................... ..........15 Materials and Methods.......................................................................................................... .20 Sampling and Outgroups.................................................................................................20 DNA Extraction, PCR Amplification, and Sequencing..................................................20 DNA Sequence Alignment and Phylogenetic Analysis..................................................22 Congruence Tests and Hypothesis Testing......................................................................24 Analysis of Spatial Genetic Structure..............................................................................25 Results........................................................................................................................ .............26 Monophyly of the SE Scrub Mint Clade.........................................................................26 ITS Sequence Characteristics and Phylogeny Reconstruction........................................29 Congruence Tests............................................................................................................30 Analyses of the Relationship betwee n Geographic and Genetic Distance......................31 Discussion..................................................................................................................... ..........32 The Monophyly of Conradina and Causes of Discordan ce between ITS and Plastid Sequence Data..............................................................................................................32 Other Generic-Level Relationships Re garding the SE Scrub Mint Clade......................35 Relationships within Conradina ......................................................................................37 Taxonomic Questions within Conradina : Status of C. brevifolia and the Santa Rosa Populations...................................................................................................................3 8 3 PHYLOGENY OF CONRADINA AND RELATED SOUTHEASTERN SCRUB MINTS (LAMIACEAE) BASED ON GAPC GENE SEQUENCES.....................................50 Introduction................................................................................................................... ..........50 Materials and Methods.......................................................................................................... .53 Initial Isolation of GapC in the SE Scrub Mint Clade.....................................................53 Determination of GapC Copy Number...........................................................................54

PAGE 5

5 Taxon Sampling for Phylogeny Reconstruction Using GapC-1 and GapC-2 ................57 Analyses of Recombination.............................................................................................58 Separate Analyses of GapC-1 and GapC-2 Congruence Analyses, and Combined Phylogenetic Analyses.................................................................................................59 Results........................................................................................................................ .............62 GapC Copy Number in Ne w World Mentheae...............................................................62 Analyses of Recombination.............................................................................................63 Phylogenetic Analysis of GapC-1 and GapC-2 ..............................................................63 Congruence Analyses......................................................................................................65 Combined Phylogenetic Analyses...................................................................................66 Removal of Single Partitions...........................................................................................67 Discussion..................................................................................................................... ..........68 Copy Number and Evolution of GapC in New World Mentheae...................................68 Separate Phylogenetic Analysis of GapC-1 and GapC-2 ................................................69 Combined Analyses—Congruence and the Effect of Concatenation Approach.............70 Combined Phylogenetic Analyses—Relations hips in the SE Scrub Mint Clade............73 Combined Phylogenetic Anal yses–Relationships among Conradina Species................75 4 COMBINED PHYLOGENETIC ANALYS IS OF HETEROZYGOUS NUCLEAR LOCI: AN EXAMPLE USING CONRADINA AND RELATED MINTS (LAMIACEAE).................................................................................................................... ..84 Introduction................................................................................................................... ..........84 Materials and Methods.......................................................................................................... .86 Characteristics of Data Matrices.....................................................................................86 Methods of Concatenating Multiple Heterozygous Nuclear Genes................................87 Methods of Phylogenetic Analysis..................................................................................92 Results........................................................................................................................ .............93 Discussion..................................................................................................................... ..........96 5 ISOLATION, CHARACTERIZATION, AN D CROSS-SPECIES AMPLIFICATIONS OF MICROSATELLITE LOCI FROM CONRADINA (LAMIACEAE).............................108 Introduction................................................................................................................... ........108 Materials and Methods, Re sults, and Conclusions...............................................................108 6 PATTERNS OF GENETIC STRUCTUR E BASED ON MICROSATELLITE LOCI REVEAL SPECIES COHESION AND S HARED ANCESTRAL POLYMORPHISM RATHER THAN HYBRIDIZATION IN CONRADINA .....................................................113 Introduction................................................................................................................... ........113 Materials and Methods.........................................................................................................1 17 Population sampling, DNA Ex traction, and Genotyping..............................................117 Statistical Analysis of Microsatellite Data....................................................................118 Results........................................................................................................................ ...........120 Summary Statistics........................................................................................................120 Analyses of Genetic Structure among Conradina Species............................................121

PAGE 6

6 Patterns of Genetic Structure within Each Conradina Species.....................................123 Discussion..................................................................................................................... ........127 Deviations from Hardy-Weinberg Equilibrium.............................................................127 Species Limits and Distinguishing Between Hybridization vs. Incomplete Lineage Sorting in Conradina .................................................................................................128 Patterns of Genetic Structure among Conradina Species.............................................130 Patterns of Genetic Structure in Endangered Conradina Species and Conservation Implications................................................................................................................134 Conclusions...................................................................................................................1 38 7 GENERAL CONCLUSIONS...............................................................................................150 LIST OF REFERENCES............................................................................................................. 154 BIOGRAPHICAL SKETCH.......................................................................................................168

PAGE 7

7 LIST OF TABLES Table page 2-1 Taxon name, distribution, and conservation st atus of the SE scrub mint taxa included in this study.................................................................................................................. ......40 2-2 Accessions used for phylogeny reconstructi on (with all necessary landowner or state permits obtained before collection), vouche r information (all vouchers are deposited in FLAS), locality, and genbank accession numbe rs for regions that were sequenced in the following order: ITS, trnT-trnL spacer (AB), trnL intron (CD), the trnK 5 intron, psbA-trnH spacer....................................................................................................41 2-3 Sequence characteristics of regions used in phylogeny reconstruction.............................43 2-4 Results of SH tests........................................................................................................ .....44 3-1 Characteristics of DNA matrices and results of phylogenetic analyses............................77 4-1 Characteristics of indivi dual DNA regions and analyses................................................101 4-2 Results of combined analyses of the GapC-1 GapC-2 ITS and plastid data partitions using the concatenation methods. The PO FAD method is not shown here because it was not analyzed using parsimony..................................................................................102 5-1 Results of initial primer sc reening in all six species of Conradina (plus the Santa Rosa populations), Clinopodium ashei, Clinopodium georganium and Piloblephis rigida ............................................................................................................................... .111 5-2 Characteristics of 11 micros atellite loci developed for Conradina .................................112 6-1 Populations of Conradina species sampled in this study. ..............................................140 6-2 Diversity indices for each species of Conradina ..........................................................142 6-3 The range in pairwise FST values between all 56 p opulations, grouped into the described species of Conradina, including the Santa Rosa populations.........................144 6-4 Results of hierarchical AMOVA analyses.......................................................................145

PAGE 8

8 LIST OF FIGURES Figure page 2-1 Geographic distributions of all Conradina species, including the questionable Santa Rosa county populations....................................................................................................45 2-2 Collection locations of accessions included in this study..................................................46 2-3 One of 96,300 most parsimonious tr ees (length = 111, CI = 0.946, RI = 0.921, RC = 0.872) based on combined plastid sequence da ta (excluding a 22-bp inversion in the psbA-trnH data set)............................................................................................................47 2-4 One of 99,400 most parsimonious trees based on ITS sequence data (length = 164, CI = 0.841, RI = 0.864, RC = 0.727).................................................................................48 2-5 A plot of geographic distance (km) vs. genetic distance (HKY85) using the plastid and ITS data sets within Conradina and within the SE scrub mint clade (excluding Dicerandra )........................................................................................................................49 3-1 Distributions of Conradina species...................................................................................78 3-2 Collection locations of accessions included in this study (figure reproduced from Edwards et al. (2006))........................................................................................................7 9 3-3 Relationships among GapC loci in the SE scrub mint clade.............................................80 3-4 Trees based on separate analyses of the GapC data sets...................................................81 3-5 Phylogram of one of the most parsimonious trees resulting from combined analysis of GapC-1 GapC-2 ITS and plastid data sets, w ith heterozygotes concatenated using the IUB code method................................................................................................82 3-6 Phylogram of one of 220 most parsimonious trees result ing from analysis based on combined analysis of GapC 1 GapC 2 ITS, and plastid data sets, with heterozygotes concatenated in all possibl e combinations (maximum of four per individual).................................................................................................................... ......83 4-1 Phylogram of one of the most parsimonious trees resulting from combined analysis of GapC 1 GapC 2 ITS and plastid data sets, with heterozygotes concatenated by randomly selecting a single allele to repres ent an individual in each data set. ..............103 4-2 Phylogram of one of the most parsimonious trees resulting from combined analysis of GapC 1 GapC 2 ITS and plastid data sets, w ith heterozygotes concatenated using the IUB/N method..................................................................................................104

PAGE 9

9 4-3 Phylogram of one of 220 most parsimonious trees result ing from combined analysis of GapC 1 GapC 2 ITS, and plastid data sets, with heterozygotes concatenated in all possible combinations (maxim um of four per individual)..........................................105 4-4 Phylogram of one of the most parsimonious trees resulting from combined analysis of GapC 1 GapC 2 ITS, and plastid data sets, w ith heterozygotes concatenated using the CI method. ......................................................................................................106 4-5 NeighborNet trees resulting from an alyses of the POFAD data sets...............................107 6-1 Collection locations and population designa tions of populations used in this study.......146 6-2 Neighbor joining phylogram of pa irwise genetic distances of all Conradina populations included in this study. Distances calculated using Nei’s standard genetic distances...................................................................................................................... .....147 6-3 Results of STRUCTURE analyses to investigate species boundaries in Conradina ....148 6-4 Results of STRUCTURE analyses in each species of Conradina ...................................149

PAGE 10

10 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PHYLOGENY RECONSTRUCTION AND CONSERVATION GENETICS OF CONRADINA AND RELATED SOUTHEASTERN U.S. ENDEMIC MINTS (LAMIACEAE) By Christine E. Edwards December 2007 Chair: Pamela S. Soltis Cochair: Douglas E. Soltis Major: Botany Conradina comprising six described species, is a morphologically homogeneous group of narrow-leaved, aromatic mints ende mic to the southeastern U.S. Despite the fact that several Conradina species are endangered and are targets of fe deral recovery programs, species limits in Conradina are unclear, several federally endangered species are of questionable validity, and hybridization has been proposed. The goals of th is study were to reconstruct the evolutionary relationships among Conradina species, to clarify the relationship of Conradina to related mint genera such as Dicerandra Clinopodium, Piloblephis and Stachydeoma, to clarify species debates in Conradina to understand patterns of genetic structure within and among each species of Conradina and to use results of this study to ai d in conservation efforts. A molecular phylogeny was inferred by sequenci ng four plastid regions and ITS and from each species of Conradina and related genera. ITS sequence da ta strongly suppor ted the monophyly of Conradina, in agreement with evidence from mor phology, but plastid sequence data did not support a monophyletic Conradina and placed the genus as paraphyletic to Clinopodium Stachydeoma, and Piloblephis Similar plastid haplotypes were shared by different genera, perhaps due to incomplete lineage sorting or hyb ridization. Next, three members of the nuclear

PAGE 11

11 GapC gene family were isolated, two of which were used to recons truct the phylogeny of Conradina and related mints. Separate phylogenetic analyses of the two GapC loci did not resolve species relationshi ps. Two approaches were used to concatenate and carry out combined analyses of the two heterozygous GapC loci with the ITS and plasti d data sets. Trees resulting from the two concatenation approaches were si milar in the resolution and support of generic relationships, but differed substantially in re solution and support for relationships within Conradina probably due to differing treatment of conf lict in the data set. Next, methods of concatenation for combined analyses of multiple heterozygous nuclear genes were investigated. All available methods of concatenating data from more than one heteroz ygous nuclear gene were evaluated to assess their utility, and six approaches were used to combine the ITS, plastid and two heterozygous GapC data sets. Two of these methods allow for multiple placements of individuals in a single tree and al low relationships to collapse in the presence of conflict among data sets, so these methods may be the best wa ys to illustrate the conflict resulting from heterozygosity. Next, microsatellite loci were isolated, characterized, and screened for amplification success in Conradina These microsatellite loci we re used to genotype individuals in populations of each species of Conradina in order to clarify species limits, investigate genetic structure in Conradina and test hypotheses of interspecific hybridization vs. incomplete lineage sorting suggested by phylogenetic an alyses. Microsatellite data differentiated the six described species, and little hybridization wa s apparent, supporting the hypot hesis of incomplete lineage sorting as the primary source of conflict in phylogenetic analyses. All described species appear to be valid biological entities so the current listing status of the endangered species of Conradina is appropriate; however, two popula tions that were thought to be C. etonia are highly genetically differentiated, sugges ting that they may possibly mer it description as a new species.

PAGE 12

12 CHAPTER 1 GENERAL INTRODUCTION Conradina (Lamiaceae), comprising six describe d species, is a morphologically homogeneous group of narrow-leaved, aromatic, shrubs from the southeastern U.S. and is readily distinguished from other mint s by a distinctively geniculate corolla (Gray 1965). Four Conradina species are federally endange red or threatened, and four are endemic to Florida. Conradina etonia C. glabra and C. brevifolia are federally endangered Florida endemics, Conradina verticillata is federally threatened and found only in the drainage basins of certain rivers in Tennessee and Kentucky, and Conradina grandiflora is endemic to Florida and is threatened in the state (Chafin 2000) The most widespread species is C. canescens which is locally abundant along the Gulf Coast of th e Florida panhandle a nd into Alabama and Mississippi. Conradina is related to other endemic genera from the southeastern U.S., including Dicerandra (nine species), Piloblephis (monotypic) Stachydeoma (monotypic), and the woody, southeastern U.S. species of Clinopodium : C. ashei, C. georgianum, C. coccineum, and C. dentatum (Crook 1998; Edwards et al. 2006; Edwards et al. in press; Trusty et al. 2004). The relationship of Conradina to these related genera is unclear and intergeneric hybridization has been proposed; thus, some goals of this study were to clarify relationships among Conradina and related genera, and to evaluate the hy potheses of intergen eric hybridization. Despite the fact that several species are fe derally endangered or threatened and have federal recovery programs in place, species limits in Conradina are unclear; hybridization has been proposed among Conradina species, and the taxonomic status of several protected species is questionable. Conradina brevifolia is listed as federally endangered yet taxonomically questionable because it is morphologically similar to the relatively widespread, disjunct species, C. canescens (USFWS 1996), and populations of Conradina found in Santa Rosa County, FL,

PAGE 13

13 are problematic because they are morphological intermediates between the federally endangered C. glabra and the widespread C. canescens Another goal of this study was to clarify species limits and the status of narrow endemics in Conradina Little is known about th e population genetics of Conradina Because some Conradina species are extremely rare, it has been suggested that these species have low genetic diversity; for example, Gray (1965) hypothesized th at the federally endangered species C. glabra would show high levels of homozygosity because of morphological homogeneit y and high levels of male sterility, and C. verticillata was hypothesized to have low ge netic diversity because of low reproductive success and high levels of vegeta tive reproduction (USFWS 1996). However, allozyme studies of Conradina revealed that for species with limited distributions, some had unexpectedly high levels of gene tic variation (Crook 1998). Howeve r, these earlier studies did not measure genetic diversity in all members of the genus or samp le more than th ree populations of any species. Because these studies have been incomplete, it is difficult to determine what forces have most strongly influenced genetic diversity of each species of Conradina, or to use this information to conserve the e ndangered and threatened species. Thus, the main goals of this study were: to clarify the relationship of Conradina to other genera of southeastern U.S. endemic mints and investigate hypotheses of intergeneric hybridization; to clarify phylogeneti c relationships in Conradina ; to clarify species boundaries in Conradina and investigate the status of several taxonomically questionable narrow endemics; to investigate the hypothesis of hybridization among Conradina species; to investigate patterns of genetic diversity and genetic structure in Conradina species; to use this information to he lp conserve these endangered Conradina species.

PAGE 14

14 These goals are addressed throughout the fo llowing six chapters: In chapter 2, I reconstruct the phylogeny of Conradina and related southeastern U.S. mints (the SE scrub mint clade) and investigate the limits and status of several narro w endemics using DNA sequences from four plastid regions and the Internal Transcribed Spacer (ITS) region of nuclear rDNA. In chapter 3, I present the resu lts of the isolation of three copies of the glyceraldehyde-3phosphate dehydrogenase ( GapC ) gene family from Conradina and related mints and the use two of these loci to reconstruct the evolutio nary history of the SE scrub mint clade. In chapter 4, I research met hods of combining heterozygous nuc lear data and used several of these methods to concatenate th e ITS, plastid, and two heterozygous GapC data sets from Conradina and related mints. I evaluate how eac h of these methods performs and make recommendations about the optimal method for c oncatenating multiple heterozygous nuclear loci for combined phylogenetic analysis. In chapter 5, I present the results of the isol ation and characterizati on of 11 microsatellite loci from three Conradina species, to be used to investig ate hypotheses of hybridization, to clarify species boundaries, and to elucidate the patterns of genetic diversity within and among populations of each species of Conradina. In chapter 6, I investigate patterns of gene tic structure within a nd among each species of Conradina using 10 microsatellite loci to help clarify species bounda ries, understand the partitioning of genetic variation in Conradina, and to use this genetic information to help inform conservation decisions in the rare species of Conradina Chapter 7 is a general conclu sion in which I summarize what has been learned about the evolutionary dynamics of Conradina and to propose future directions that will help us gain a more complete understanding of these endangered species.

PAGE 15

15 CHAPTER 2 MOLECULAR PHYLOGENY OF CONRADINA AND OTHER SCRUB MINTS (LAMIACEAE) FROM THE SOUTHEASTERN USA: EVIDENCE FOR HYBRIDIZATION IN PLEISTOCENE REFUGIA? Introduction †Conradina (Lamiaceae), a genus of six species endemic to the southeastern United States, is a morphologically homogeneous group of narrow-leaved, aromatic shrubs. Conradina is readily distinguished from other mints by a unique corolla morphology. From the stem, the calyx and corolla point at a downward angle; bey ond the calyx (in the throat region) the corolla tube reflexes upwards at an angl e of almost 90 and opens to fo rm the two lobes of the corolla (Gray 1965). The strongly bilabiate flowers of Conradina are grouped in axillary clusters and are white, soft pink, or lavender. The lower lip of the corolla is strongly th ree-lobed, and the mouth of the lower corolla tube is lined with dark purple spots. Most Conradina species inhabit only the imperiled sand pine scrub or sandhill habitats of Florida. These xeric, pyrogenic habitats occ upy well-drained, sandy soils in the southeastern United States and support a large number of ende mic species due to their roles as Pleistocene refugia (Christman and Judd 1990; Huck 1989; Myer s 1990). It is hypothesized that scrub once dominated most of peninsular Fl orida, but it has become restri cted geographically in the last 12,500 years due to climate change (Delcourt and Delcourt 1981). Today, these habitats exist only on a fragmented network of xeric ridges and dunes occurring thr oughout Florida and the southeastern United States (M yers 1990), and they are quic kly diminishing due to human † Reprinted with permission from the Amer ican Society of Plant Taxonomists. Original publication: Edwards, C. E., D.E. Soltis, and P. S. Soltis, 20 06. Molecular phylogeny of Conradina and other scrub mints (Lamiaceae) from the southeastern U. S.: evidence for hybridization in Pl eistocene refugia? Systematic Botany 31:193-207.1

PAGE 16

16 development. As scrub and sandhill endemics, Conradina species occupy a mostly allopatric distributional pattern that corresponds to ar eas of favorable habitat (Fig. 2-1). Because Conradina species are endemic to a dwindling habitat, four of the six described species are listed as federally e ndangered or threatened (U. S. Fish and Wildlife Service; Table 2-1). The rarest of all Conradina species is the recently described Conradina etonia, which inhabits a small, isolated patch of inland scr ub in Putnam County in northeastern Florida (Kral and McCartney 1991). Conradina glabra is found exclusively in the sandhill habitat of the ancient bluffs of Torreya State Park in the Florida panha ndle (USFWS 1994). Conradina brevifolia is found on the Lake Wales Ridge of pe ninsular Florida in Polk, Osceola, and Highlands counties (USFWS 1996). Conradina verticillata the only species that does not occur in Florida, is unique in habita t preference, found only in sandy ar eas of rocky cobble bars on the banks of rivers in the Cumberland River draina ge basin in Tennessee and Kentucky (Gray 1965). Conradina canescens the most widespread and morphologi cally variable species (Gray 1965), occupies coastal scrub along the Gulf coast from the Florida panhandle to Mississippi. Conradina grandiflora is also found in coastal scrub and occu pies the eastern co ast of peninsular Florida from Palm Beach County to Volusia County (Gray 1965). Conradina is a member of the Nepetoideae (Lamiaceae) which is divided into four tribes: Escholtzieae, Lavanduleae, Ocimeae, and Mentheae (Cantino et al. 1992). Conradina belongs to a clade of New World Mentheae (Crook 1998; Trusty et al. 2004), and in recent analyses, Conradina has been shown to be nested within a clade that includes other endemics of the southeastern United States that share similar morphology and habitat preference. These closely related genera include Dicerandra (nine species), Piloblephis (monotypic) Stachydeoma (monotypic), and the woody, southeastern U. S. species of Clinopodium such as C. ashei, C.

PAGE 17

17 georgianum, C. coccineum, and C. dentatum (Crook 1998; Prather et al 2002; Trusty et al. 2004). Hereafter, these taxa will be referred to as the southeastern (SE) scrub mint clade. There are several nomenclatural or taxonom ic issues that involve members of the SE scrub mint clade, including: 1) Stachydeoma graveolens has been considered to be Hedeoma by some authors (e. g.,Wunderlin 1998), although A. Prather (pers. comm.) has found that it is rela ted to the SE scrub mint clade (not to Hedeoma species) and we thus retain Stachydeoma 2) the species of Clinopodium in this study are often referred to as Calamintha (e. g.,Wunderlin 1998), although all Calamintha were transferred to Clinopodium based on the finding that Calamintha is nested within Clinopodium (Cantino and Wagstaff 1998; Doroszenko 1985), and 3) the Clinopodium species of the SE scrub mint clade do not appear to be related to ei ther the Old World or herbaceous New World species of Clinopodium (Crook 1998; Trusty et al. 2004; A. Prather, pers. comm.). Regardless of these taxonomic issu es, in recent analyses th e SE scrub mint clade (as defined above) has been found to be m onophyletic (Crook 1998; Trus ty et al. 2004; A. Prather pers. comm.), although the sister group to the SE scrub mint clade is unresolved. Some analyses place eastern U. S. genera such as Monarda, Pycnanthemum or Blephilia as sister, while other analyses place the eastern South American Hesperozygis as sister (Crook 1998; Trusty et al. 2004; A. Prather pers. comm.). Other hypothesized western U. S. and European taxa do not appear to be cl osely related (Trusty et al. 2004; A. Prather pers. comm). Furthermore, although the SE scrub mint clade (as defined above) appears to be monophyletic, the relationship of Conradina to other members of the SE scrub mint clade is still unclear; some analyses place various SE Clinopodium species as a paraphyletic sister to Conradina (Crook 1998), whereas others have found a close relationship between Conradina and Piloblephis rigida (Trusty et al. 2004). The only fine-s cale study that atte mpted to clarify

PAGE 18

18 relationships among Conradina species and their relationship to the SE scrub mint clade is that of Crook (1998), who analyzed the group us ing morphology, plastid DNA markers, and allozyme loci. The morphological analysis of all Conradina taxa and various other members of the SE scrub mint clade resulted in a monophyletic Conradina which forms a polytomy with an unresolved assemblage of southeastern Clinopodium species. In Crook’s combined analysis of rbcL, ndhF and trnL-F, taxon sampling was very limited (incl uding only four species of the SE scrub mint clade), and the resolution was poor, but the results conflict with morphology, showing Conradina as a polyphyletic assemblage with Piloblephis rigida and Clinopodium ashei Because of this result, Crook (1998) questioned the relationships among Conradina Clinopodium species from the southeastern United States, and Piloblephis Furthermore, within Conradina, species readily cross with one another experimentally, leading Gray (1965) to hypothesize that species boundaries are probably maintained sole ly by geographic isolation. Because several Conradina species have proximal geographic ra nges (Fig. 2-1), this raises the possibility of hybridization between species. Although six species are currently recogni zed, the circumscriptions of several Conradina species have long been debated. Disa greements have centered primarily on C. brevifolia and a population of Conradina found in Santa Rosa County, FL Conradina brevifolia was first described by Shinners (1962) and was dist inguished from the morphologically similar C. canescens by the presence of shorter leaves, a shor ter lower lip of the corolla, and a disjunct geographic range. This classi fication was recognized by severa l authors (Gray 1965; Kral and McCartney 1991; Wunderli n et al. 1980), but later treatments lumped C. brevifolia into C. canescens, often without listing C. brevifolia as a synonym (Chafi n 2000; Delaney and

PAGE 19

19 Wunderlin 1989; Wunderlin 1982, 1 998). The USFWS (1996) lists C. brevifolia as federally endangered yet taxonomically questionable. The other taxonomic debate concerns the st atus of populations f ound along the Blackwater River in Santa Rosa County, Florida, referred to hereafter as the Sant a Rosa populations. The Santa Rosa populations have glabro us leaves, leading some to cons ider them to be the federally endangered Conradina glabra (Godfrey 1998; Martin 1992; Shinners 1962). However, the USFWS (1994) currently extends protection only to the glabrous populations that are restricted to the area around Torreya State Park in Libert y County, Florida, while the questionable Santa Rosa populations are 200 km away and para patric to populations of the widespread Conradina canescens Furthermore, despite resembling C. glabra in pubescence characters, the Santa Rosa populations resemble C. canescens in floral morphology and habit. One of the goals of this study is to determine if the Santa Rosa populatio ns represent 1) disjun ct populations of the endangered C. glabra 2) anomalous populations of the widespread C. canescens or 3) a new species. If the Santa Rosa popul ations indeed repres ent a distinct species, federal protection may be necessary, given that ther e are only approximately 6 populati ons that inhabit only a small portion of a county (L. Anderson pers. comm.). The purposes of this study were 1) to investigate the monophyly of Conradina and its relationship to other proposed re latives in the SE scrub mint clade, including assessment of possible hybridization between genera, 2) to elucidate relationships among Conradina species using an autapomorphic species concept (e. g., Donoghue 1985; Mishler 1985) and 3) to evaluate the validity of species status for C. brevifolia and the Santa Rosa populations to determine if they merit federal protection.

PAGE 20

20 Materials and Methods Sampling and Outgroups Most tissue samples were collected in the fi eld and placed in silica gel for preservation prior to DNA extraction. When field collecti on was not possible, herbarium specimens or greenhouse plants of a documented origin were us ed. All voucher specimens were deposited in the FLAS herbarium, and the accession details for the specimens used in this study are listed in Table 2-2. Collection locations ar e shown in Fig. 2-2. At least two individuals of every named species of Conradina and the Santa Rosa popula tions were included, with the exception of the highly disjunct and locally distributed Conradina verticillata of which we included only one individual. To assess the monophyly of Conradina and determine its relationship to other SE scrub mints, one individual each of Clinopodium coccineum Clinopodium dentatum Clinopodium georgianum, Dicerandra cornutissima, Dicerandra thinicola and Piloblephis rigida and two individuals of Stachydeoma graveolens were also included. We sampled Clinopodium ashei more intensively to investig ate the possibility of hybridization between C. ashei and Conradina brevifolia as hypothesized by Huck (1994); four populations were sampled, with Conradina brevifolia co-occurring in two sampling loca tions. As outgroups, we included Monarda fistulosa and Pycnanthemum muticum because these genera were often placed as sisters to the SE scrub mint clade in pr evious analyses (Prath er et al. 2002; Trusty et al 2004, A. Prather pers. comm.), and we included an unidentified specimen of Mentha as a more distantly related outgroup. DNA Extraction, PCR Amplific ation, and Sequencing Total DNA was extracted from leaf tissue us ing a CTAB extraction protocol (Doyle and Doyle 1987). Many samples apparently containe d impurities that inhibi ted PCR amplification, and these were cleaned using a QiaQuick PCR pur ification kit (Qiagen Corp, Valencia, CA.).

PAGE 21

21 Amplification of DNA was conducted with a Bi ometra T3 Thermocycl er or an Eppendorf Mastercycler using PCR r eactions in 50-l volumes containing 2.5 units of Taq polymerase, 0.5 M of each primer, 0.2 mM of each dNTP in an equimolar ratio, 10X buffer containing 1.5 mM MgCl2, 20-30 ng of genomic DNA, and dH20. 10% DMSO or 1M Betaine was included in all ITS reactions to relax secondary structure. The ITS-1, 5.8S, and ITS-2 regions of the nuc lear ribosomal DNA were amplified using the primers N-nc18S10 and C26A (Wen and Zi mmer 1996) and the following PCR conditions: 1) an initial heating step at 95C for 5 min, 2) 94C for 1 min, 3) annealing at 53C for 1 min, and 4) elongation at 72C for 2.5 min. Steps 2-4 were repeated for 5 cycles, except for dropping the annealing temperature by 1C in each of the five cycles until 48C, and then annealing temperature was maintained at 48C and steps 2-4 repeated for a total of 35 cycles. The trnT-trnL spacer region and the trnL intron were amplified using the primer pairs A-B and C-D (Taberlet et al. 1991), respectively, and the psbA-trnH spacer region was amplified using primers developed by Sang et al. (1997). The entire 5 and 3 trnK intron region (including the matK gene) was amplified using the primers tr nK1F (Manos and Steele 1997) and trnK2R (Johnson and Soltis 1994). In an initia l screening of several species of Conradina the only variable region was the 5 trnK intron. Thus, the 5 trnK intron plus a small portion of the 5 end of the matK gene (referred to hereafter as the 5 trnK intron) was amplified using the primers trnK1F and DIC1100R (Oliveira et al. 2007). PCR am plification conditions fo r these four plastid primer pairs were similar to those of ITS, excep t the initial annealing temperature was 58C, and was then dropped one degree during each of the fi rst 6 cycles to 52C, where it was held constant for the remaining 34 cycles.

PAGE 22

22 PCR products were visualized by agarose ge l electrophoresis and cl eaned using either Sephadex columns or treatment with shrimp al kaline phosphatase and exonuclease I (ExoSAP; USB Corporation, Cleveland, OH). Cleaned PCR products were quantified using gel electrophoresis prior to cycle sequencing. Cycle sequencing reactions were performed using the CEQ Dye Terminator Cycle Sequencing Quick Star t kits (Beckman Coulter, Fullerton, CA) and the PCR primers. PCR products were sequenced in both directions using a Beckman-Coulter CEQ 2000 or CEQ 8000 Automated Sequencer accordi ng to the manufacturer’s specifications. Because they were placed in surprising positions in the plastid data set (see Results), accessions of Conradina grandiflora 171-1, Stachydeoma graveolens 162-24 and 162-15, Clinopodium dentatum Clinopodium georgianum and Piloblephis rigida were extracted and sequenced an additional time to verify that ou r results were not due to error. DNA Sequence Alignment and Phylogenetic Analysis Sequences were assembled and edited us ing Sequencher 4.0.5 (Gene Codes Corp., Ann Arbor, MI). ITS sequences were inspect ed for additional copy types by checking for polymorphisms in their chromatograms, and we did not find evidence for additional copy types in this analysis. Because of high sequence simila rity, sequences were easily aligned by eye using the program Se-Al (Rambaut 2003). Sequences th at were redundant thro ughout all data sets were excluded from analyses. Conradina glabra 166-1 and Conradina glabra 109-2 were identical, as were Conradina brevifolia 198-1 and Conradina brevifolia 197-1; thus, Conradina glabra 166-1 and Conradina brevifolia 198-1 were removed from all phylogenetic analyses. Aligned datasets and trees were deposited in TreeBASE (study accession number S1262, matrix accession numbers M2204 and M2205). All phylogenetic analyses we re conducted using PAUP* 4.0b10 (Swofford 2002). The following data partitions were analyzed separately using both parsimony and maximum

PAGE 23

23 likelihood optimality criteria: 1) only the ITS sequences collected in this study, 2) the combined plastid data sets, and 3) an expanded ITS data set that in cluded downloaded sequences of hypothesized relatives in the Mentheae. The latt er was carried out to conduct another assessment of the monophyly of the SE scrub mint clade in relation to other Menthe ae (Crook 1998; Prather et al. 2002; Trusty et al. 2004). We downloaded ITS sequences of Blephilia ciliata, Blephilia hirsuta, Pycnanthemum incanum, Hesperozygis spathulata, Monarda punctata, Monardella linoides, Clinopodium vulgare, Satureja hortensis, and Thymus mastichina, with Lavandula as the outgroup. We then analyzed these data in co mbination with the ITS sequence data from the SE scrub mint clade. In all analyses, gaps were coded as missing da ta. However, in the plastid data set, we conducted analyses with or wit hout including indels coded as pr esence/absence characters using a simple gap coding strategy (Simmons and Ochoterena 2000). In al l parsimony analyses, heuristic searches were conducted using 1000 ra ndom addition replicates and TBR branch swapping, saving 100 trees per replicate. Bootstra p analyses with 1000 replicates (Felsenstein 1985) were used to assess branch support using a heuristic search with TBR branch swapping, with 10 random additions per replicate, saving no more than 100 trees per replicate. For maximum likelihood (ML) analyses a likelihood ratio test (Gol dman 1993; Huelsenbeck and Crandall 1997) as implemented in ModelTest 3.06 (Posada and Cranda ll 1998) was used to determine the maximum likelihood model that best fit each data set. Model parameter values generated by ModelTest were then used in the ML analyses. The model selected for the smaller ITS data set was the TrN + model (Tamura and Nei 1993; Yang 1994); parameter values were freqA = 0.2099, freqC = 0.3240, freqG = 0.2911, freqT = 0.175, and gamma shape = 0.3239. The model for the expanded ITS data set was also the TrN + model; parameter values were

PAGE 24

24 freqA = 0.2130, freqC = 0.3085, freqG = 0.2754, freqT = 0.2031, and gamma shape = 0.3087. The model selected for the plastid data set was HKY + (Hasegawa et al. 1985; Yang 1994); parameter values were freqA = 0.3568, freq C = 0.1481, freqG = 0.1512, freqT = 0.3439, Ti/Tv ratio = 0.8066 and gamma shape = 0.0172. Heuristi c searches were co nducted using 100 random addition replicates and TBR branch swapping. We measured support for branches using the same likelihood model with 100 bootstrap replicates, 10 random additions per replicate, and TBR branch swapping, saving all trees. Congruence Tests and Hypothesis Testing To test for congruence between ITS and pl astid data partitions, we conducted the Incongruence Length Difference (ILD) Test of Farri s et al. (1994) as implemented in PAUP* as the Partition Homogeneity Test (Swofford 2002). Although some studies ha ve shown that this test is sometimes overly conservative or unable to detect incongruence in certain conditions (Darlu and Lecointre 2002; Yoder et al. 2001) recent authors suggest that it may be a good, conservative first approximation of congruence betw een data partitions (Hipp et al. 2004). For all ILD analyses, we compared the ITS data set with the combined plastid da ta set, including only parsimony-informative characters. For each test we performed 1000 replicates. Because the significance value of the ILD test is derived using the number of replicates, in some cases it was necessary to increase the number of replicates (up to 10,000) to determine the value of highly significant P values. We used heuristic searches, TBR branch swapping, gaps coded as missing data, 10 random additions per replicate, saving no more than 100 trees per replicate. ILD tests were conducted to test the null hypo thesis of congruence with 1) all accessions included, 2) one conflicting accession removed at a time to identify specific accessions that may be causing the incongruence, and 3) blocks of conflicting accessions removed.

PAGE 25

25 To test alternative hypotheses of phylogeny in our data sets, we used the ShimodairaHasegawa (SH) test to compare the most likely trees of each data set to the most likely trees when a constraint tree is enforced (Shimodair a and Hasegawa 1999; Shimodaira 2002), with the null hypothesis being that there is no difference. Trees constrai ned to a specific relationship (Table 2-4) were generated using MacClade 4.1 (Maddison and Maddison 2001). The data sets were analyzed with an enforced constraint tr ee using the ML search strategy outlined above and ML parameter values estimated for that data se t using ModelTest (see above). SH tests were conducted as implemented in PAUP* 4.0b10 (Swofford 2002) with RELL optimization and 1000 bootstrap replicates. Analysis of Spatial Genetic Structure. Under a scenario of isolation by distan ce (Wright 1943, 1946), restricted gene flow between regions may cause novel mutations to be geographically isolated (Neigel et al. 1991). Gene flow among genetically different lineages in a region will lead to introgression and will cause geographically localized ge notypes to be shared by more than one lineage in a region (Schaal and Olsen 2000). If horiz ontal transfer occurs betwee n lineages, then the regional distribution of shared haplotypes may cause a po sitive correlation between genetic distance and geographic distance when examining this relationship across different species or genera. To test this possibility, we used a Mantel test to evaluate the relation ship between geographic distance and genetic distance. This type of test was used because the non-independence of the entries violates the assumptions of a simple correlation (M antel 1967). However, th e significance of the normalized Mantel coefficient cannot be found from standard statistical tests because the observations within the matri ces are not independent (Rosen berg 2001). Thus, significance is tested through a randomization test by s huffling the entries in one matrix.

PAGE 26

26 To conduct this analysis, we calculated pairwi se genetic distances be tween taxa using the HKY85 model of evolution (Hasegaw a et al. 1985) and constructed a geographic distance matrix by finding the pairwise distance between GPS poin ts. A Mantel test was used to find the correlation between the two distance matrices wi th 5,000 permutations as implemented in the computer program PASSAGE (Rosenberg 2001). To test for spatial genetic structure in each of the data sets, we performed this analysis using genetic distances derived separately from the ITS and plastid data sets. We also performed th is test for both data partitions using only Conradina to test for migration between geographically clos e species, as well as using all taxa in the SE scrub mint clade excluding Dicerandra (which was sister to the remainder of the SE scrub mint clade in all analyses) to inves tigate the possibility of gene flow occurring between genera. Results Monophyly of the SE Scrub Mint Clade Sequence characteristics for all DNA regions ar e listed in Table 2-3. For our expanded ITS data set, the aligned lengt h was 643 characters, of which 113 were parsimony-informative, 102 were variable but parsimony-uninformative, and 428 characters were constant. Parsimony analyses of the expanded ITS data set result ed in 99,100 trees of length = 396, CI = 0.705, RI = 0.766, and RC = 0.540. Maximum likelihood analysis recovered 49 trees of -ln likelihood 2959.64841. The topologies derived from the strict consensus of the parsimony trees and the ML trees were almost identical. These topolog ies (not shown) show a monophyletic SE scrub mint clade, which forms a polytomy with Pycnanthemum, Blephilia, Hesperozygis and Monarda The western U. S. taxon ( Monardella ) is sister to all of the above, with Old World taxa ( Mentha, Clinopodium vulgare, Thymus and Satureja hortensis ) more distantly related. Plastid Sequence Characteristics and Phylogeny Reconstruction

PAGE 27

27 A large part of the variation in the plastid data set was in the form of indels. Ten out of the 19 indels occurred in only one accession, and thus do not influence relationships in the trees. However, all but one of the remaining nine inde ls fell into two classes: mutations occurring in monoor di-nucleotide repeats (f our instances) that likely aros e from slipped-strand mispairing (Kelchner 2000), and tandem repeat s (four instances) that are co mmonly associated with highly homoplasious loop regions (Kelchner 2000). Wh en we included these indels coded as presence/absence characters in th e analysis of the plastid data set, resolutio n and bootstrap support for clades decreased, perhaps indicative of a homoplastic patt ern of length-mutation evolution. Because these types of indels may often occur independently multiple times (Kelchner 2000), homology assessment is questiona ble, so we excluded the presence/absence indel characters in all analyses shown here. A 22-bp inversion fl anked by an inverted repeat was found in all taxa in the psbA-trnH spacer region. Inversions in hairpin regions of non-coding plastid spacers have been shown to occur fre quently and homoplastical ly (Kelchner and Wendel 1996; Kelchner 2000), especially in the psbA-trnH spacer region (Sang et al. 1997; Tate and Simpson 2003). Indeed, when ma pped onto a tree constructed fr om the plastid data set using only base substitutions, the inversion occurred at least four times. Because homology assessment was again questionable, the inversion and flanking inverted repeats were also omitted from all analyses. The aligned sequences of the four plastid re gions comprised a total of 2034 characters, of which 29 were parsimony-informative and 74 were variable but parsimony-uninformative (Table 2-3). In general, base substitutions were common in the plastid data set, but very few of these mutations were parsimony-informative. Instead, th e mutations were frequently autapomorphies

PAGE 28

28 for individual accessions, which re sulted in long terminal branch lengths. In fact, we found very few shared plastid haplotypes, even for di fferent populations of the same species. Parsimony analysis of the combined plastid regions resulted in 96,300 most parsimonious trees of length 111, consistency index (CI) of 0.946, retention index (RI) of 0.921, and rescaled consistency index (RC) of 0.827. One of the most parsimonious trees (chosen at random) resulting from this analysis is shown in Fig. 2-3. ML analysis resulted in two most likely trees of -ln likelihood = 3505.7778. Because parsimony and ML reconstructions are extremely similar, only the parsimony reconstruction is shown here (Fig. 2-3). Plastid sequence data weakly support (55% bootstrap (BS)-parsimony, 60% BS -ML) the monophyly of the SE scrub mint clade excluding Dicerandra and place Conradina as paraphyletic in relation to Piloblephis Stachydeoma and most Clinopodium species. The strict consensus (not shown) of the parsimony trees and both maximum likelihood tr ees shows a trichotomy made up of 1) a clade consisting of Clinopodium georgianum and several Conradina accessions, 2) a clade consisting of Clinopodium ashei Clinopodium dentatum Piloblephis Stachydeoma, and the remainder of Conradina, including a well-supported clade consisting of Conradina grandiflora 215-1, Conradina canescens 130-1 and 142-1, and both accessions of Conradina glabra and 3) Clinopodium coccineum However, none of the relationships received >50% BS support in parsimony or ML analyses. Plastid sequen ce data do not suppor t the placement of Dicerandra in the SE scrub mint clade, and instead place Dicerandra in a clade with Monarda and Pycnanthemum although this relationship is not strongly supported. Plastid sequence data do not support th e monophyly of any of the species of Conradina One extreme example is that of C. grandiflora which is placed in thr ee separate clades (Fig. 23): one accession is placed in a clade with two accessions of both C. canescens and C. glabra

PAGE 29

29 one accession is placed in a clade with Clinopodium ashei 147/147-1 and Stachydeoma graveolens 162-24, and two accessions are placed in a clade with the remaining Conradina accessions. Another example is that of C. brevifolia, in which all accessions group together except C. brevifolia 168-2, which is in a completely differe nt clade. The only species which may be monophyletic are Conradina etonia, C. glabra and the Santa Rosa populations, as both accessions of each are placed in the same clade (although the two accessions from the Santa Rosa population were taken from the sa me population and differ by many steps). ITS Sequence Characteristics a nd Phylogeny Reconstruction The aligned length of the ITS region was 624 characters, of which 55 were parsimonyinformative and 68 were variable but parsimonyuninformative (Table 2-3). Parsimony analyses of the ITS data set resulted in 99,400 trees of length = 164, CI = 0.841, RI = 0.864, and RC = 0.727. One of the most parsimonious trees (chosen at random) resulting from analyses of the ITS data set is shown in Fig. 2-4. Maximum lik elihood analysis recovered 14 trees of -ln likelihood 1790.81347. Because the topologies deri ved from the strict consensus of the parsimony trees and the ML trees were almost id entical, differing only in rearrangements within Conradina only a parsimony tree is shown he re (Fig. 2-4). In contrast to the plastid trees, both parsimony and ML reconstructions using ITS sequence data support th e monophyly of not only the SE scrub mint clade, including Dicerandra (89% BS-parsimony, 100% BS-ML), but also Conradina (94% BS-parsimony, 96% BS-ML). Parsimony and ML trees place Piloblephis as sister to Conradina (with <50% BS support), with Clinopodium georgianum as sister to Conradina + Piloblephis Both analyses support a clade composed of Conradina, Clinopodium, Piloblephis and Stachydeoma (64% BS-parsimony, 71% BS-ML) and place Dicerandra as sister to the rest of the SE scrub mint clade.

PAGE 30

30 Unlike plastid sequence data, ITS sequence data do not resolve many relationships within Conradina. In fact, rearrangements within Conradina cause the large number of trees resulting from analyses of ITS. The only branch within Conradina with >50% BS support in both parsimony and ML analyses of ITS is that which unites Conradina etonia 111-1 with the northernmost population of Conradina grandiflora 215-1. A clade consisting of both accessions of the Santa Rosa populations and Conradina canescens 128-2 is found in the strict consensus of the parsimony trees and is weakly supported (53% BS support) in ML analyses. Congruence Tests Partition homogeneity tests revealed that there was a highly significant level of incongruence ( P = 0.0005) between the ITS and plastid da ta sets. Visual examination of the topologies from the two data matrices reveals major differences (see Discussion), which we examined by carrying out ILD tests with various accessions or blocks of accessions removed. In all cases, removal of single accessions did not notably impact the significance of the ILD P values (results not shown). In f act, we found large increases in ILD P values only when blocks of conflicting accessions were removed. For exam ple, we found increases in ILD values after removal of all accessions of Clinopodium, Stachydeoma, P iloblephis, and Dicerandra ( P = 0.25), after removal of Dicerandra, Clinopodium georgianum and Stachydeoma ( P = 0.051), and after removal of all Clinopodium, Stachydeoma and Piloblephis ( P = 0.051). Thus, the incongruence does not appear to be due to confli cting signal from a singl e accession, but rather is more widespread throughout the data set. Interestingly, remova l of all conflicting Conradina accessions does not reduce the significance of the ILD P value ( P = 0.006), suggesting that a large portion of the inco ngruence may be due to conflicting data from the remainder of the SE scrub mint clade.

PAGE 31

31 Furthermore, SH tests were also used to exam ine the source of incongr uence in the data set and investigate the possibili ty of alternative hypotheses of evolution in the SE scrub mint clade. Results of the SH tests (Table 2-4) generally sup port the results of the IL D tests. For example, when the constraints were the most likely tree or strict consensus of the most parsimonious trees of the rival data set, we were able to reject the null in every case. Furtherm ore, in contrast to ITS data, plastid data reject both the monophyly of Conradina and Clinopodium (Table 2-4). Because the obvious and significant differences between the ITS and plastid partitions may represent different evolutionary histories, we did not combine the two matrices and perform a single analysis. Analyses of the Relationship betw een Geographic and Genetic Distance A Mantel test was used to test the null hypot hesis that genetic dist ance is unrelated to geographic distance either 1) within the SE scru b mint clade (excluding Dicerandra because it was found as sister to the remai nder of the clade in ITS analyses and outside of the SE scrub mint clade in plastid analyses) or 2) within Conradina as opposed to the a lternative hypothesis, which is that geogra phic distance and geneti c distance are related Plots of all analyses of geographic distance vs. genetic dist ance are shown in Fig. 2-5. Within the SE scrub mint clade, Mantel tests of the plastid data set were significant ( R =0.1644, P =0.034), while we could not reject the null hypothesis usi ng genetic distances derived from the ITS data set ( R =-0.1463, P =0.06539). Within Conradina analyses using the plastid da ta set were also significant ( R =0.234, P =0.024). We could not reject the null hypothesis within Conradina using the ITS data ( R =0.1772, P =0.07731).

PAGE 32

32 Discussion The Monophyly of Conradina and Causes of Discordance between ITS and Plastid Sequence Data ITS sequence data strong ly support the monophyly of Conradina relative to the other genera included in this st udy (Fig. 2-4). The monophyly of Conradina is further supported by evidence from morphology; species of Conradina are readily distinguishable from other species in the SE scrub mint clade, united by the morphological syna pomorphy of a distinctively geniculate corolla (Crook 1998). Species of Conradina are morphologically extremely similar to each other and are generally distinguishable by only small differences in leaf, flower size, or pubescence characters (Gra y 1965; Wunderlin 1998). In contrast to the ITS sequence data, plas tid sequence data do not support the monophyly of Conradina and both parsimony (Fig. 2-3) and maxi mum likelihood analyses of the plastid data place Conradina as paraphyletic to Piloblephis Stachydeoma and some Clinopodium species, albeit with <50% BS support (Fig. 2-3) The obvious incongruence between the ITS and plastid data sets is further supported by ILD and SH tests (Tab le 2-4). Furthermore, the incongruence does not appear to be limited to a single taxon, but instead appears to be caused by widespread incongruence involvi ng many accessions in our study. Because the plastid data set is incongruent wi th both ITS and morphological evidence, this suggests that the patterns of variation in the plastid data may be due to processes such as introgression or shared ancestr al polymorphism and lineage sorting (Schaal and Olsen 2000; Wendel and Doyle 1998). It is often impossibl e to distinguish betw een shared ancestral polymorphism and introgression in a data set such as this (Agu ilar and Feliner 2003; Larena et al. 2002; Neigel and Avise 1986). However, in general there are very few well-documented cases of ancestral polymorphism that are sh ared across species boundaries (but see Mason-

PAGE 33

33 Gamer et al. 1999), let alone across generic bounda ries (Wendel and Doyle 1998). Furthermore, if ancestral polymorphism and lineage sorting ar e indeed the cause of the incongruence between the ITS and plastid trees, then polymorphism would have been ma intained in the plastid genome during the divergence of Conradina and throughout the speciation events of each species of Conradina. This would also mean that coalescence did not occur within Conradina in the plastid genome, whereas it did occur in IT S. However, if mutation rates are equal, coalescence is much more rapid in organellar genomes in comparison to nuclear genes because haploid genomes have a smaller effective population size (Moore 1995). Thus, this suggest s that coalescence may have already occurred in the plastid genome in Conradina and that ancestral polymorphism is probably not responsible for the obser ved patterns of plastid variation. Introgression appears to be the more likely explanation for the incongruence between the ITS and plastid data sets (e. g., Rieseberg and So ltis 1991; Rieseberg et al. 1996). Within the SE scrub mint clade, Huck (1994) reporte d possible reticulate evolution between Conradina and Clinopodium as evidenced by putative hybrids between Clinopodium ashei (as Calamintha ashei ) and Conradina brevifolia occurring in their areas of sympat ry in the Lake Wales Ridge of peninsular Florida. Although we did not find ev idence of recent introgr ession between these two species, we did find examples of putative introgression between ge nera of the SE scrub mint clade (all of which were re-extra cted and sequenced to verify that they were not due to sampling or PCR error). The first involves Clinopodium georgianum which is firmly nested within a clade composed strictly of Conradina accessions. Clinopodium georgianum differs from many accessions of Conradina by only one base substitution, most notably Santa Rosa population 1331, which occurs within 100 km of the southernmost populations of Clinopodium georgianum Another putative example of intr ogression involves the sympatric Clinopodium dentatum and

PAGE 34

34 Stachydeoma graveolens 162-15 which possess identical haplotypes, although another accession of Stachydeoma graveolens from the same population has a diffe rent plastid haplotype. The last instance may be that of Conradina grandiflora 171-1, which is separate from all other accessions of Conradina in a clade with Clinopodium ashei 147/147-1 and Stachydeoma graveolens 161-24. Although these species are not curre ntly sympatric, the ranges of Clinopodium ashei and Conradina grandiflora are less than 100 km apart. In each of these cases, the fact that identical or similar plastid haplotypes are shared in tw o different, geographically close species suggests that introgression may have occurred (L arena et al. 2002; Sc haal et al. 1998). The fact that several clades of Conradina appear to be more closel y related to other genera of the SE scrub mint clade than to other Conradina accessions, based on the plastid data set but not ITS, may also be due to introgression that occurred during the Pleistocene. At that time, scrub habitat was favored by climatic conditions and thus was widespread throughout Florida (Delcourt and Delcourt 1981). Howe ver, glacial cycles during the Pleistocene caused drastic sea level fluctuations in Florida; at high sea levels only small areas of high elevation were above water, and at low sea levels large areas of land that are currently covere d by water were exposed, opening up habitat in newly deposited dunes (Del court and Delcourt 1981; Myers 1990). During each rise in sea level, all members of the SE sc rub mint clade may have been forced to occupy the same refugia, at which time extensive hybrid ization and introgression may have occurred among lineages. This would explain why some accessions of Conradina have a haplotype that is similar to that of Clinopodium, Stachydeoma, and Piloblephis ; if this hybridization scenario is true, either Conradina was polymorphic and introgression caused the other members of the SE scrub mint clade to capture one plastid type of Conradina or alternatively, some of the polymorphism currently observed in Conradina could have resulted from introgression of the

PAGE 35

35 plastids of other genera into Conradina Range expansions following drops in sea level may have allowed the dispersal of divergent haplotype s, resulting in even more complicated patterns of plastid variation in Conradina and related genera. Shared re fugia during the Pleistocene and subsequent range expansion have been shown to account for shared plastid haplotypes across species boundaries in multiple European Quercus species (Petit et al. 1997) and in Armeria (Larena et al. 2002). The significant positive correlation between geographic distance and genetic distance in the plastid data set of the SE scrub mint clade (Fig. 2-5) further supports the hybridization hypothesis. Because no relations hip was found between geographic distance and genetic distance in analyses of the ITS data set (F ig. 2-5), we believe that this data set is not influenced by introgression and may reflect a clearer picture of evolutionary relationships. Other Generic-Level Relationships Re garding the SE Scrub Mint Clade Although a comprehensive phylogeny of the Me ntheae will likely be the only way to verify that we sampled all taxa in the SE scr ub mint clade, the monophyly of the SE scrub mint clade with respect to other hypothesized relative s in the expanded ITS analysis leads us to believe that our sampling is a reasonable approxi mation of the taxa that belong in the SE scrub mint clade. However, while analyses of IT S sequence data support the monophyly of the SE scrub mint clade, including Dicerandra (Fig. 2-4; expanded analysis not shown), Dicerandra is not placed within the SE scrub mint clade base d on plastid sequence data (Fig. 2-3). Despite lack of support in the plastid data set, all members of the SE scrub mint clade, including Dicerandra are also united by morphology. All members have a very distinc tive habit; they all have a stem that is profusely branched from th e base, with branches that grow erectly. Most species of this clade inhabit dr y, frequently disturbed, or pyrogen ic habitats. ITS sequence data place Dicerandra as sister to the remainder of the SE sc rub mint clade with high support, and the two species of Dicerandra included in this study form a str ongly supported clade in both ITS and

PAGE 36

36 plastid analyses (Figs. 2-3, 2-4). Dicerandra consists of nine sp ecies of herbaceous to suffrutescent shrubs which are united by spur-lik e anther appendages that serve as a trigger mechanism for pollen di spersal (Huck 1987). The remainder of the SE scrub mint clade, without Dicerandra, is monophyletic in both ITS and plastid analyses (Figs. 2-3, 2-4). A ll of these remaining members are exclusively woody. Woodiness may be a synapomorphy for this clade, although this hypothesis remains to be tested by morphological anal yses. Relationships within the clade are unclear. The monophyly of Clinopodium is not supported in our expande d ITS analysis; the woody, New World Clinopodium species (the SE scrub mint species) we re found to be more closely related to the SE scrub mint clade (and other Eastern U. S. species such as Monarda Pycnanthemum and Blephilia ) than to Clinopodium vulgare (the type of the genus). However, because generic relationships within the Mentheae are still unclear, a comprehensive Mentheae phylogeny will likely be the only way to clarify the limits of Clinopodium The monophyly of the SE Clinopodium is also questionable. ITS data group all Clinopodium species together except C. georgianum and SH tests of ITS data could not reject the hypothesis of monophyly of SE Clinopodium However, all analyses of the plasti d data set place the SE scrub mint Clinopodium species in various clades with Conradina, Stachydeoma and Piloblephis There is considerable variation in floral morphol ogy among the SE species of Clinopodium and no apparent morphological synapomorphy that unites them (Cr ook 1998). Further analyses of more rapidly evolving regions of the nuclear genome will be necessary to resolve the relationships among Conradina, Clinopodium, Piloblephis, and Stachydeoma New circumscriptions of generic limits may be necessary.

PAGE 37

37 Relationships within Conradina While ITS sequence data were useful for reve aling relationships among the members of the SE scrub mint clade, levels of ITS sequence divergence in Conradina are extremely low and there is very little species-level resolution in the trees resulting fr om this data set (Fig. 2-4). There is more resolution within Conradina in the plastid sequence data, although there is not strong support for species-le vel relationships. Of all Conradina species, the only species in which all accessions are placed in the same clad e are the two most endangered narrow endemics, Conradina glabra and Conradina etonia (and the Santa Rosa populations-see below for discussion). Coalescence occurs more rapidly in populations w ith small effective population sizes (Hudson 1990), and because both of thes e species have very small population sizes, coalescence may have occurred more rapidly than in other, more widespread species of Conradina The same may also be said for the other narrow endemic, Conradina brevifolia which all accessions are grouped together except C. brevifolia 168-2. The latter accession is placed with the two most geograp hically proximal populations of Conradina grandiflora and may be another example of introgression. Besides the few exceptions noted above, plastid sequence data do not support the monophyly of most trad itionally defined Conradina species, and in fact, clades consisting of different species often receive moderate s upport (Fig. 2-3), even though each species of Conradina has distinguishing morphological characteristi cs that are present throughout its range. The non-monophyly of Conradina species found in the plastid data set may again reflect an evolutionary history with ances tral polymorphism and/or intr ogression. For example, one wellsupported clade consists of both accessions of Conradina glabra and two accessions of Conradina canescens all of which are from the Florida pa nhandle. Also included in this clade is Conradina grandiflora 215-1, which is from the east coas t. Although it is possible that

PAGE 38

38 Conradina glabra and Conradina canescens have hybridized given their close geographic proximity, the likelihood that either could hybridize with a population of Conradina grandiflora located ~600 km away on the Atla ntic coast is unlik ely (although specific information about the identity and dispersal ranges of their pollinators is unknown). This suggests that either introgression occurred in the pa st and geographic ranges have cha nged, or that this particular clade may have originated before speciation and thus may be a case of shared ancestral polymorphism. There are also examples of what appears to be hybridi zation and chloroplast capture in Conradina As described previously, geogra phically proximal populations of Conradina grandiflora and Conradina brevifolia share plastid haplotypes, which is likely the result of introgression of the plastid genome of C. grandiflora into C. brevifolia The significant relationship within Conradina between geographic distance and genetic distance found from the Mantel test of the plastid da ta (Fig. 2-5) further supports a hypothesis of interspecific hybridization and ch loroplast capture. Taxonomic Questions Within Conradina : Status of C. brevifolia and the Santa Rosa Populations. Plastid sequence data place most accessions of C. brevifolia as separate from C. canescens (the exception being C. brevifolia 168-2, which is likely pl aced in a clade with C. canescens 1282 because of hybridization with the geographically proximal C. grandiflora ), although their relationships are unres olved in the ITS trees Thus, we tentatively recommend that C. brevifolia be retained as a distinct, protec ted species until more evidence is available. The Santa Rosa populations are not place d in a clade with C. glabra in either ITS or plastid analyses, so we recommend that they not be considered a disjunct population of the narrow endemic, C. glabra However, the close relationship found be tween the Santa Rosa populations and C. canescens 128-2 in both ITS and plastid analyses (Figs. 2-3, 2-4) may be the result of several possible

PAGE 39

39 scenarios: 1) it is a recent derivative of C. canescens 2) it has hybridized with C. canescens 1282, or 3) it is merely another population of C. canescens and does not deserve species status. However, further work is needed to elucidate the evolutionary processe s that are occurring in Conradina so that we may make stronger recommendations about the delimitation of these species. To that end, we are currently working with rapidly evolving low-copy nuclear regions to resolve relationships within Conradina and developing microsatellite loci to examine patterns of genetic similarity within and among populations of all species of Conradina Both sources of data will help us to understand more fully the fi ne-scale patterns of evol ution in this group so that we may make more concrete r ecommendations on the management of Conradina species.

PAGE 40

40 Table 2-1. Taxon name, distribution, and conservation status of the SE scrub mint taxa included in this study. FL-E/T = Florida state endange red/threatened species (Florida Fish and Wildlife Conservation Commission), GA-T = Georgia state threatened species (Georgia Department of Natural Heritage), KT-E = Kentucky state threatened species (Kentucky State Nature Preserves Commi ssion), TN-T = Tennessee state threatened species (Tennessee Department of Natu ral Heritage), US-E/T = Federally endangered/threatened species (U. S. Fish and Wildlife Service). Q denotes a taxonomic question either the taxon is not ge nerally recognized as valid, or there is reasonable concern about its validity or identity globally or at the state level. Taxon Distribution Federal a nd state protection status Conradina brevifolia FL endemic US-EQ / FL-E Conradina canescens FL, AL, MS Not protected Conradina etonia FL endemic US-E / FL-E Conradina glabra FL endemic US-E / FL-E Conradina grandiflora FL endemic FL-T Conradina sp. (Santa Rosa population) FL endemic Not named, of questionable validity Conradina verticillata TN, KT US-T / TN-T, KT-T Clinopodium ashei FL, GA FL-T, GA-T Clinopodium coccineum SE U. S. Not protected Clinopodium dentatum FL endemic FL-T Clinopodium georgianum AL, GA, FL, MS, NC, SC FL-E Dicerandra densiflora FL endemic Not protected Dicerandra thinicola FL endemic Not protected Piloblephis rigida FL endemic Not protected Stachydeoma graveolens FL endemic FL-T

PAGE 41

41 Table 2-2. Accessions used for phylogeny reconstr uction (with all necessary landowner or state permits obtained before collection), vouche r information (all vouchers are deposited in FLAS), locality, and genbank accession numbe rs for regions that were sequenced in the following order: ITS, trnT-trnL spacer (AB), trnL intron (CD), the trnK 5 intron, psbA-trnH spacer. Clinopodium ashei (Weath.) Small Edwards and Tancig 147 (from voucher) Ocala National Forest, Marion County, FL. AY9434 81, AY943411, AY943446, AY943516 AY943551 Clinopodium ashei (Weath.) Small Edwards and Tancig 147 accession 1 Ocala National Forest, Marion County, FL. AY943482, AY943412, AY943447, AY943517, AY943552. Clinopodium ashei (Weath.) Small Edwards and Bell 160 Near Avon Park, Highlands County, FL. AY943483, AY943413, AY 943448, AY943518, AY943553. Clinopodium ashei (Weath.) Small Edwards and Bell 196 Near Lake Placid, Highlands County, FL. AY943484, AY943414, AY 943449, AY943519, AY943554. Clinopodium coccineum (Nutt.ex Hook.) Kuntze. Edwards and Ionta 102 Near Ochlocknee bay, Liberty County, FL. AY943485, AY943415, AY943450, AY943520, AY943555. Clinopodium dentatum (Chapm.) Kuntze Edwards 146 Near Torreya State park, Liberty County, FL. AY943486 AY943416, AY 943451, AY943521, AY943556. Clinopodium georgianum R. M. Harper Edwards 178 Biophilia nature center, Elberta, AL. AY943487, AY943417, AY943452, AY943522, AY943557. Conradina brevifolia Shinners Edwards and Brockington 159 Near Avon Park, Highlands County, FL. AY943462, AY943392, AY943427, AY943497, AY943532. Conradina brevifolia Shinners Edwards and Cox 168 Lake Wales Ridge State Forest, Polk County, FL. AY943463, AY943393, AY943428, AY943498, AY943533. Conradina brevifolia Shinners Edwards and Bell 197 Near Sebring, Highlands County, FL. AY943464, AY943394, AY943429, AY943499, AY943534. Conradina brevifolia Shinners Edwards and Bell 198 Near the city of Sun Ray, Polk County, FL. AY943465, AY943395, AY943430, AY943500, AY943535. Conradina brevifolia Shinners Bok Tower Garden Collection. AY943461, AY943391, AY943426, AY943496, AY943531. Conradina canescens Gray Edwards et al. 128 Carrabelle, Franklin County, FL. AY943466, AY943396, AY943431, AY943501, AY943536. Conradina canescens Gray Edwards et al. 130 San Destin, Walton County, FL. AY943467, AY943397, AY943432, AY943502, AY943537 Conradina canescens Gray Edwards and Tillman 142 Near Ft. Morgan, Baldwin County, AL. AY943468, AY943398, AY943433, AY943503, AY943538. Conradina etonia Kral & McCartney Edwards et al. 111 Etoniah Creek State Forest, Putnam County, FL. AY943470, AY943400, AY943435, AY943505, AY943540. Conradina etonia Kral & McCartney Oliviera and Miller S. N. Dunn’s Creek State Park, Putnam County, FL. AY943469, AY 943399, AY943434, AY943504, AY943539. Conradina glabra Shinners Edwards and Ionta 109 accession 1 Torreya state park, Liberty County, FL. AY943471, AY943401, AY943436, AY943506, AY943541. Conradina glabra Shinners Edwards and Ionta 109 accession 2 Torreya state park, Liberty County, FL. AY943472, AY 943402, AY943437, AY943507, AY943542. Conradina glabra Shinners Edwards and Ionta 166 Torreya state park, Liberty County, FL. AY943473, AY943403, AY943438, AY943508, AY943543.

PAGE 42

42 Table 2-2. Continued Conradina grandiflora Small Edwards and Katz 134 Seacrest Scrub Natural Area, Palm Beach County, FL. AY943476, AY 943406, AY943441, AY943511, AY943546. Conradina grandiflora Small Edwards and Katz 135 Juno Dunes Natural Area, Palm Beach County, FL. AY943475, AY 943405, AY943440, AY943510, AY943545. Conradina grandiflora Small Edwards and Scanlon 171 St. Sebastian River State Preserve, Indian River County, FL. AY 943477, AY943407, AY943442, AY943512, AY943547. Conradina grandiflora Small Edwards 215 Titusville, Brevard County, FL. AY943474, AY943404, AY943439, AY943509, AY943544. Conradina sp. (Santa Rosa population) Edwards et al. 133 accession 1 Near Milton, Santa Rosa County, FL. AY943478, AY943408, AY943443, AY943513, AY943548. Conradina sp. (Santa Rosa population) Edwards et al. 133 accession 30 Near Milton, Santa Rosa County, FL. AY943479, AY943409, AY943444, AY943514, AY943549. Conradina verticillata Jennison Edwards 176 From Tripple Brook Farm, Southhampton MA. AY943480, AY943410, AY943445, AY943515, AY943550. Dicerandra densiflora Benth. Huck S. N. AY943488, AY943418, AY943453, AY943523, AY943558. Dicerandra thinicola H. A. Mill. Huck S. N. Titusville, Brevard County, FL. AY943489, AY943419, AY943454, AY943524, AY943559. Piloblephis rigida (W. Bartram ex Benth) Raf. Edwards and Tancig 149 Ocala National Forest, Marion County, FL. AY943490, AY943420, AY943455, AY943525, AY943560 Stachydeoma graveolens Edwards and Ionta 162 accession 15 Apalachicola National Forest, Liberty County, FL. AY943491, AY943421, AY943456, AY943526, AY943561, Stachydeoma graveolens Edwards and Ionta 162 accession 24 Apalachicola National Forest, Liberty County, FL. AY943492, AY943422, AY943457, AY943527, AY943562. Mentha sp. Edwards 177 Cultivated plant, Gainesville, FL. AY943495, AY943425, AY943460, AY943530, AY943565 Monarda fistulosa L. Edwards 174 Obtained from Biophilia nature center, Elberta, AL. AY943493, AY943423, AY943458, AY943528, AY943563 Pycnanthemum muticum (Michx.) Pers Edwards 175 From Biophilia Nature Center, Elberta, AL. AY943494, AY943424, AY943459, AY943529, AY943564

PAGE 43

43 Table 2-3. Sequence characte ristics of regions used in phylogeny reconstruction. Length (bp) Parsimony informative characters Aligned length Parsimony informative characters Variable, parsimony uninformative characters Gaps 2 bp ITS region 617-619 55 624 55 68 1 ITS with expanded sampling 643 113 102 6 trnTtrnL (AB) 517-518 5 518 5 16 1 trnL intron (CD) 488-522 11 521 71 15 5 5 trnK intron 644-668 7 671 7 14 3 psbA-trnH (without inversion) 285-300 6 318 6 29 7 Combined plastid regions n/a 29 2034 29 74 16

PAGE 44

44 Table 2-4. Results of SH tests. Numbers indicate P value / difference in -ln likelihood score. *Indicates significance at the P = 0.05 level. Data Set Constraint ITS Combined Plastid Most likely tree of riva l data set 0.002* / 296.0078 0.006* / 156.0934 Strict consensus of most parsimonious trees of rival data set 0.001* / 214.566 0.003* / 89.203 Conradina monophyly --0.034* / 21.15561 SE Clinopodium monophyly 0.163 / 4.955 0.025* / 32.975

PAGE 45

45 Figure 2-1. Geographic di stributions of all Conradina species, including the questionable Santa Rosa county populations.

PAGE 46

46 Figure 2-2. Collection locations of accessions included in this study. Numbers following taxon names indicate the accession number of the population and the number of the individual in the population. Accessions obtai ned from native nurseries or botanical gardens are not shown (these include Clinopodium georgianum Conradina brevifolia (from Bok Tower Garden), and outgroups Mentha sp ., Pycnanthemum muticum and Monarda fistulosa ).

PAGE 47

47 Figure 2-3. One of 96,300 most parsimonious trees (length = 111, CI = 0.946, RI = 0.921, RC = 0.872) based on combined plastid sequence da ta (excluding a 22-bp inversion in the psbA-trnH data set). Arrows indicate branches that collapse in the strict consensus, numbers above branches indicate branch le ngths, numbers below branches indicate parsimony/maximum likelihood bootstrap support >50%, and numbers following taxon names indicate the accession number of the population and the number of the individual in the population.

PAGE 48

48 Figure 2-4. One of 99,400 most parsimonious tr ees based on ITS sequence data (length = 164, CI = 0.841, RI = 0.864, RC = 0.727). Arrows indicate branches that collapse in the strict consensus, numbers above branches indicate branch lengths, numbers below branches indicate parsimony/maximum likelihood bootstrap support >50%, and numbers following taxon names indicate the accession number of the population and the number of the indivi dual in the population.

PAGE 49

49 0 0.002 0.004 0.006 0.008 0.01 0.012 050010001500 Pairwise geographic distance (km) between Conradina accessionsITS Pairwise genetic distance (HKY) between Conradina accessions. 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 050010001500 Pairwise geographic distance (km) between Conradina accessionsPlastid pairwise genetic distance (HKY85 ) between Conradina accessions 0 0.01 0.02 0.03 0.04 0.05 0.06 050010001500 Pairwise geographic distance (km) between SE scrub mint taxa (excluding Dicerandra )ITS Pairwise Genetic distance (HKY85) between SE scrub mint taxa (excluding Dicerandra ) 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 050010001500 Pairwise geographic distance (km) between SE scrub mint taxa (excluding Dicerandra)Plastid Pairwise Genetic distance (HKY85 ) between SE scrub mint taxa (excluding Dicerandra) Figure 2-5. A plot of geographic distance (km) vs. ge netic distance (HKY85) using the plastid and ITS data sets within Conradina and within the SE scrub mint clade (excluding Dicerandra ).

PAGE 50

50 CHAPTER 3 PHYLOGENY OF CONRADINA AND RELATED SOUTHEASTERN SCRUB MINTS (LAMIACEAE) BASED ON GAPC GENE SEQUENCES Introduction §The use of single-copy nuclear genes to reso lve phylogenetic relationships among closely related species has become increasingly common b ecause the introns of these genes often have a much more rapid evolutionary rate than othe r commonly used regions, such as ribosomal or organellar markers (Sang 2002; 1998; Small et al. 2004). Introns can therefore yield a high proportion of variable characte rs that may lead to increa sed resolution and support in phylogenetic analyses. However, because of their short length or lack of coalescence, nuclear gene introns may still not have sufficient info rmation to resolve evol utionary relationships, especially at shallow phylogenetic levels (Zha ng and Hewitt 2003). In these cases, phylogenetic analyses that combine intron data with sequen ce data from other genes may improve resolution and support. The purpose of this study was to use separate and combined analyses of singlecopy nuclear genes to reconstruct th e phylogeny of species in the genus Conradina and related mints (Lamiaceae). Conradina comprising six described species (Fi g. 3-1), is a morphologically homogeneous group of diploid (all species n =12; Gray 1965), aromatic shrubs that are easily distinguished from other mints by a distinctivel y geniculate corolla (Gray 1965). Conradina is related to four other genera of Lamiaceae from the southeastern U.S.: Dicerandra (nine species), four woody species of Clinopodium ( C. ashei, C. georgianum, C. coccineum, and C. dentatum), and the monotypic genera Stachydeoma and Piloblephis (Crook 1998; Edwards et al. 2006; Trusty et al. §Accepted for publication by th e International Journal of Plant Sciences on 10/08/07. www.journals.uchicago.edu/IJPS/home.html

PAGE 51

51 2004). Because most species in this clade are restricted to the sand pine scrub or sandhill habitats of Florida or the southeastern U.S., we will hereafter refer to them as the southeastern (SE) scrub mint clade (Edwards et al. 2006). The SE scrub mint clade (Mentheae; Nepetoideae) is part of a larger group cont aining only New World species; molecular phylogenetic analyses indicate that these species diversified after a single colonization event from the Old World (Crook 1998; Trusty et al. 2004). Despite the fact that several Conradina species ( C. brevifolia, C. etonia C. glabra and C. verticillata ) are federally endangered or threatened, species limits are unclear. Hybridization may occur among species (Edwards et al. 2006), a nd the taxonomic status of several species is questionable. For example, C. brevifolia is listed as federally endangered yet taxonomically questionable because it is morphologically similar to the relatively widespread, disjunct species, C. canescens (USFWS 1996). Likewise, populations of Conradina found in Santa Rosa County, FL (referred to hereafter as the Santa Rosa populations) are problematic because they are morphological intermediates betw een the federally endangered C. glabra and the widespread C. canescens The relationship of Conradina to other members of the SE scrub mint clade is also unclear. Hybridization was reported between Clinopodium ashei (as Calamintha ashei ) and Conradina brevifolia based on morphological eviden ce (Huck 1994), and a phylogenetic analysis of the SE scrub mint clade (with limited taxon sampli ng) using plastid regions found that Conradina was non-monophyletic with regard to Clinopodium and Piloblephis (Crook 1998). In a subsequent phylogenetic study of Conradina and the SE scrub mint cl ade (Edwards et al. 2006), Conradina was found to be monophyletic based on ITS sequence data, in agreement with morphology (Crook 1998); however, trees based on plastid data were weakly s upported, but largely failed to

PAGE 52

52 support the monophyly of each Conradina species, the genus Conradina or that of most other genera (Edwards et al. 2006). The non-monophyly of species and ge nera in the plastid tree may be the result of low levels of informative nucle otide variation, introg ression, or incomplete sorting of ancestral po lymorphism (Edwards et al. 2006). The combined ITS and plastid tree largely agreed with relationships found in the ITS data set, perhap s due to the larger number of informative ITS characters or more informative signal. Because ITS and plastid data yielded trees wi th low levels of resolution and support, and because hybridization or incomplete lineage sorting (or both) have been hypothesized, more data are needed to clarify phylogenetic relationships in Conradina and the SE scrub mint clade. The quickly evolving introns of bipa rentally inherited, single-copy nuc lear genes are ideal in cases such as these. To this end, we focused on the glyceraldehyde 3-phosphate dehydrogenase gene family ( GapC or g3pdh ), a nuclear-encoded gene family i nvolved in sugar phosphate regulation in the cytosol (Figge et al. 1999; Martin et al. 1993a). Previous studies using partial sequences of GapC for phylogeny reconstruction (Howarth a nd Baum 2005; Martin et al. 1993b; Wall 2002) or analyses of within-species diversity (Ingvarsson 2005; Morrell et al. 2005; Olsen and Schaal 1999; Perusse and Schoe n 2004) have revealed that GapC introns have high substitution rates and are phylogenetically informative at low taxonomic levels. In this study, we isolated and sequenced multiple GapC loci from Conradina and related mints, identified GapC gene copy number, quantified levels of sequence diverg ence and heterozygosity in the SE scrub mint clade, and carried out tests to detect recombination. We then focused on two GapC loci, both of which were heterozygous in many individuals, fo r phylogeny reconstruction of the SE scrub mint clade, and compared these results to previous studies using ITS and plastid sequence data. Because combined analyses often increase support and resolution of relationships, we then used

PAGE 53

53 two methods to concatenat e the two heterozygous GapC data sets with the ITS and plastid data partitions of Edwards et al. (2006) and carried out combined analyses of the four data partitions. Finally, we conducted combined analyses with each data partition removed, one at a time, to understand the contribution of each da ta partition to the combined analyses, and to understand if nuclear genes increase resolution and support of phylogenetic relationships. Materials and Methods Initial Isolation of GapC in the SE Scrub Mint Clade Total DNA was isolated from leaf tissue as de scribed in Edwards et al. (2006). Initially, the GapC gene was amplified from the genomic DNA of several representatives of the Mentheae using the PCR primers gpdx7F and gpd9r (Strand et al. 1997). These primers amplify exons 7 to 9, including introns 7 and 8, of the GapC gene. Amplifications were carried out in 25 l volumes with one unit of Amplitaq gold DNA polymerase (Applied Biosystems, Foster City, CA) or Invitrogen Platinum Taq High Fidelity DNA polymerase (Invitrogen Corp., Carlsbad, CA), 1X buffer (provided by the manufacturer s), 1M betaine, and dNTP and primer concentrations following the manufacturers’ specifications. PCR conditions were as follows: 1) an initial denaturation step at 95C for 5 min, 2) 94C for 1 min, 3) annea ling at 53C for 1 min, 4) elongation at 72C for 2.5 min, and 5) a final elongation step of 72C for 2.5 min. Steps 2-4 were repeated for 5 cycles, except for dropping th e annealing temperature by 1C in each of the five cycles until 48C, and then annealing temperature was maintained at 48C and steps 2-4 repeated for a total of 35 cycles. PCR pr oducts were visualized on 1.5% agarose gels. Amplifications produced several PCR bands rangi ng from 700 to 1,000 base pairs (bp) in length. PCR products were cloned into pCR 4.0 TOPO vector for sequencing (Invitrogen Corp., Carlsbad, CA). Eight to 16 colonies per PCR r eaction were amplified using the plasmid primers M13F and M13R and the PCR reaction mixtures specified above. P CR conditions were as

PAGE 54

54 follows: 1) an initial denaturation step at 95C for 3 min, 2) 95C for 45 sec, 3) annealing at 50C for 45 sec, 4) elongation at 72C for 45 s ec. Steps 2-4 were repeated for 30 cycles, followed by a final step of 72C for 10 min. PC R products were cleaned and sequenced using the M13 primers following Edwards et al. (2006). Determination of GapC Copy Number Our initial screening re covered two copies of GapC, which were then used to design specific primers for the Mentheae. For GapC 1 the primers are ConAmF (5' CAGCCTCGTTCAACATCATC)/ ConAmR (5 CTTGAGCTTCGTCTCCGATG), and for GapC-2 they are CONGPD-F (5' GATGGT CCGTCGAGCAAGGATT)/ CONGPD-R (5' CCTGCTGTCACGAAGTCTGT). To verify GapC copy number, we selected individuals from 11 species: Pycnanthemum muticum, Monarda fistulosa Stachydeoma graveolens Clinopodium dentatum, C. ashei 147-1, Dicerandra thinicola, D. densiflora Conradina brevifolia 198-2, C. etonia (Dunn’s Creek), and C. grandiflora 135-1 Accession information for these individuals is given in Edwards et al. (2006). To increase ta xonomic range of the copy number analyses, we also added accessions of Blephilia hirsuta (Cantino 1421, BHO) and Hedeoma pulegioides (Cantino 1424, BHO). We used the two specific primer pair s to amplify and sequence all possible GapC alleles from these individuals. We used the same PCR reagents as in the in itial screening, and the temperature cycling was as follows: 1) an initial heating step at 95C for 5 min, 2) 94C for 1 min, 3) annealing for 1 min at a gradient between 48.2C and 53.1C for ConGPDF/R and between 55C and 58C for ConAmF /R, 4) elongation at 72C fo r 2.5 min, 5) steps 2-4 were repeated for 30 cycles, and 6) a final elongation step of 72C fo r 10 min. To avoid preferential PCR of only some copies (Wagner et al. 1994) and to increase the possibility that all alleles were amplified, at least four reactions per individual per primer pair were amplified simultaneously

PAGE 55

55 following Emshwiller and Doyle (2002). The f our reactions were pooled, cloned, amplified using the plasmid primers, and sequenced as sp ecified above. We sequenced 6-15 colonies per cloning reaction, each to at least 2X coverage in an attemp t to recover all possible alleles, although there is a possibility that some alleles were not recove red via PCR. Sequences were assembled and edited using Sequencher 4.0.5 (Gene Codes Corp., Ann Arbor, MI). All resulting sequences were first subjected to BLAST s earches (Altschul et al. 1997) of GenBank ( http://www.ncbi.nlm.nih.gov/ ) to verify that they are indeed most similar to cytosolic GapC Next, to ensure that the multiple sequence t ypes recovered were not a result PCR-mediated recombination (Cronn et al. 2002), we inspected and discarded any divergent sequences that were a chimera of two copy types. In many cases, we repeated PCR, cloning, and sequencing using independent genomic DNA extr actions to verify our original sequences. We also inspected sequences for Taq error; especially when using Amplitaq gold, we found that some colonies had 1-bp substitutions that were not repeatable after sequencing ma ny colonies or re-sequencing colonies from a new DNA extraction. Because Taq polymerases, even those with proofreading ability, have a small rate of error (Cline et al 1996), we assumed that these nonrepeatable 1-bp substitutions were due to Taq error and did not include these in our analyses. All other alleles were retained and treated as terminals to allow for the possibility of multiple GapC loci. All alignments were initially produced using Clustal X (Thompson et al. 1997) and then adjusted “by eye” using the progr am Se-Al (Rambaut 2003). When all alleles recovered from these 13 individu als were aligned, the alignmen t was ambiguous in the intron regions due to high levels of sequence divergen ce and length variation among allele types, so introns were excluded from the in itial copy number analysis.

PAGE 56

56 All phylogenetic analyses were carried out using the Phyloinformatics Cluster for High Performance Computing at the Florida Museum of Natural History (http://c luster.flmnh.ufl.edu). The exon-only data set was analyzed usi ng parsimony as implemented in PAUP* 4.0b10 (Swofford 2002). We conducted parsimony analyses with heuristic searches using 1000 random addition replicates and TBR branch swapping, savi ng 100 trees per replicate. Bootstrap analyses (Felsenstein 1985) with 1000 replicates were used to asse ss branch support using a heuristic search with TBR branch swapping, one random addition per replicate, saving no more than 100 trees per replicate. We also analyzed the data set using Bayesian phylogenetic analysis (Huelsenbeck et al. 2001; Larget and Simon 1999; Yang and Ranna la 1997) as implemented in MrBayes 3.1.1 (Ronquist and Huelsenbeck 2003). We ran two anal yses using four chains each, three hot and one cold, with the temperature set to the default of 0.2. All analys es were run using flat priors, using the optimal model of evolution chosen fo r each data set by the AIC as implemented in Modeltest (Posada and Crandall 1998). The optim al model chosen for the exon-only data set was GTR + We ran analyses for 10 million generations, sampling a tree every 1000 generations. To determine the burn-in value, we checked the log like lihood scores of the resulting trees and the split frequencies between the two independent runs, and discarded any trees that were saved prior to the stabilization of the scores or that had a split frequency of greater than 0.1. For the two independent runs, posterior probabilities for each branch were found by constructing a majority-ru le consensus of trees in th e posterior distribution using PAUP* 4.0b10 (Swofford 2002). The two identical, independent analyses were checked for convergence in topology a nd branch lengths.

PAGE 57

57 Unrooted trees placed alleles into two clades separated by an extremely long branch (see Results), with alleles of every species found in each clade. We inferred that the high level of divergence between the exons of the two clusters of alleles was due to fact that there are at least two putative gene copies, and we rooted the tree between these two clades. We will hereafter refer to one of these putative loci as GapC 1 (see Results). The other clade placed alleles into two sister clades, each containing alleles from most individuals, possibly corresponding to two additional putative loci. To determine if intron differences s upported these two latter clades/putative loci separate from GapC-1 we conducted an additional analysis of the introns+exons of these alleles using both parsimony and Bayesian phylogenetic analyses, with the settings described above, but with Bayesian analyses instead performed using the HKY + model as selected by Modeltest. The intron+exon data strongly supported the two sister clades recovered using only exon data (see Results), an d we will hereafter refer to these clades as GapC-2 and GapC-3 GapC 3 was not recovered in all indivi duals due to poor amplification (the locus was amplified by the prim ers that were designed to amplify GapC 2 ); thus, we did not pursue GapC-3 for phylogeny reconstruction and focused only on GapC 1 and GapC 2 Taxon Sampling for Phylogeny Reconstruction Using GapC-1 and GapC-2 For phylogenetic analysis of the SE scrub mint clade, we sampled all species of Conradina including multiple accessions of most species. We also sampled individuals from all other genera of the SE scrub mint clade, i.e., Clinopodium, Dicerandra, Piloblephis, Stachydeoma and the outgroups Monarda fistulosa, Pycnanthemum muticum, and Mentha sp. The sampling strategy, collection lo cations, and voucher information are the same as in Edwards et al. (2006; Fig. 3-2). Because evidence suppor ts the radiation of the New World Mentheae following a single colonization event from the Old World (Trusty et al. 2004), and because relationships among New World mints have ye t to be resolved, we used Old World Mentha sp.

PAGE 58

58 as our designated outgroup in all analyses to ensure that possible New World ingroup taxa were not designated as an outgroup. Using this taxon sampling strategy, we attempted to amplify all possible alleles of GapC 1 and GapC 2 screened all resulting sequences for DNA polymerase error or PCR recombination (see below), and ali gned all alleles recovere d from all individuals using the protocols specified above. During alignment, the introns of all GapC loci were easily distinguishable from one another due to large differences in length and sequence motif, and we designated the loci accordingly. Analyses of Recombination Because recombination may have a significan t impact on the phylogenetic accuracy of a topology (Posada and Crandall 2002), it is therefor e important to understand if and how alleles of nuclear genes have recombined pr ior to their use in phylogeny reconstruction. We thus tested the two GapC loci for the presence of recombination pr ior to phylogenetic analysis. Simulations and empirical studies have revealed a wide ra nge in the power of different methods, depending on levels of genetic diversity, r ecombination rate, and rate varia tion among sites; therefore, the use of multiple methods of detecting recombinat ion may help minimize fa lse positive results and maximize the power of the methods (Posada 2002). We tested for recombination in GapC 1 and GapC 2 separately using the following methods: the Phi test for recombination (Bruen et al. 2006) in Splitstree version 4.0 (Huson and Bryant 2006) and RDP, MaxChi, Chimera, and Geneconv in the program RDP2 (Martin et al. 200 5). For all analyses using RDP2, we used settings for linear sequences, a P -value of 0.05 as the highest acceptable P -value, Bonferroni multiple comparison correction, polished breakpoi nts, with consensus daughter sequences and breakpoints. For RDP, we used “internal re ferences only” and a window of 10, 20, 30, 40, and 50 bp. For Geneconv, we scanned sequence trip lets, with indel blocks treated as one polymorphism, and a G-scale value of 0. For Ma xChi, we scanned sequence triplets with a

PAGE 59

59 variable window size, the fraction of variab le sites per window set to 0.1, 0.15, and 0.2, and 1000 permutations with a P -value of 0.05. For Chimera, we scanne d with a variable window size, the fraction of variable sites pe r window set to 0.1, 0.15, and 0.2, and 1000 permutations with a P value of 0.05. Separate Analyses of GapC-1 and GapC-2 Congruence Analyses, and Combined Phylogenetic Analyses The GapC 1 and GapC-2 data sets were aligned separate ly using Se-Al (Rambaut 2003). The data sets were analyzed separately usi ng parsimony and Bayesian methods, with the same settings used in the analyses of copy number (see above), but with th e approporiate model as chosen by Modeltest; for GapC 1 the model was GTR + and for GapC 2 the model was HKY + (Hasegawa et al. 1985; Yang 1994). To concatenate the data matrices, we chos e two methods by which to code heterozygous sites. The first method combines two alleles of a heterozygote into a single sequence, by substituting any heterozygous site s with the appropriate Internat ional Union of Biochemistry (IUB) ambiguity code following Howarth and Baum (2005). During parsimony analysis using PAUP*, the program can treat the IUB codes as e ither “uncertainties” or “polymorphisms.” If the “uncertainties” option is selected, this assume s that the state is not known but is one of the options designated by the IUB code, and the termin al is assigned the most parsimonious state (Swofford and Begle 1993; Wiens 1999). The “polymorphism” option is very similar in that PAUP* assigns the most parsimonious of the present states to the node; however, it differs in that it treats all polymorphic si tes as present in the terminal, and thus always places the less parsimonious character states as extra steps al ong the branch (Swofford and Begle 1993). The trees resulting from the “polymorphism” option will therefore be longer than those using the “uncertainty” option, but will be topologically id entical. Thus, for simplicity, we only carried

PAGE 60

60 out analyses treating the IUB codes as “uncerta inties.” MrBayes also treats IUB codes as uncertainties (Ronquist et al. 2005). The second concatenation method includes all possible combina tions of alleles from an individual as terminals in the data matrix (Sota and Vogler 2003). For example, if an individual is homozygous for GapC 1 and heterozygous for GapC 2 the individual would be represented twice in the matrix; the two concatenated sequ ences would be identical for ITS, plastid, and GapC-1 and would differ for GapC 2 If an individual is heterozygous for both GapC loci, all four combinations of alleles would be concatenat ed with the ITS and plastid data partitions and included in the analysis. We chose these two co ncatenation approaches because they represent two extremes; the IUB code method removes all variation resulting from heterozygosity, while the “all combinations” approach includes all possible information. The difference in the results therefore should reflect the influence of in cluding heterozygosity on the resulting topology. We then used the two concatenated data ma trices to test for incongruence among data partitions ( GapC-1 GapC-2 ITS, and plastid). We conduc ted Incongruence Length Difference (ILD) tests (Farris et al. 1994) as implemented in PAUP* as the Partition Homogeneity Test (Swofford, 2002). We conducted an ILD analysis to compare all possible pairs of the four data partitions, as well as one that compared all four data partitions to one another simultaneously. We conducted ILD analyses with 1000 replicat es, heuristic searches TBR branch swapping, gaps coded as missing data, and 10 random additions per replicate, saving no more than 100 trees per replicate. However, although the ILD test is a good first approxi mation of incongruence among data partitions (Hipp et al. 2004), the reje ction of the null using the ILD test does not necessarily predict whether the tr ees resulting from the combined data sets will be accurate (Cunningham 1997). It is a conservativ e test that often rejects congr uence due to factors such as

PAGE 61

61 differences in evolutionary rates, among-site rate variation among par titions, high amounts of noise, or differences in partition length (Bar ker and Lutzoni 2002; Darlu and Lecointre 2002; Dolphin et al. 2000; Dowton and Austin 2002). Even if data sets have different evolutionary histories, combined phylogenetic analysis can improve resolution and support (Wiens 1998). Thus, we generally followed the methodology of Wiens (1998) for combining data sets; we compared trees resulting from analysis of separa te data sets to identify conflicting topological differences that received >80% bootstrap support. Because there were almost no cases in which topological congruence was strongl y supported in the different da ta sets (see Results), we analyzed the combined data sets. However, any groupings that were found to be conflicting in the various individual data sets were treated as tentative in all combined analyses. For each concatenation method employed, we conducted a simultaneous analysis of all four data partitions ( GapC 1 GapC 2 ITS, and plastid) using both parsimony and Bayesian phylogenetic methods. For parsimony searches, we us ed the same search strategy as specified in the copy number analysis above. For Bayesian an alysis, we used the same search settings as specified above, but we used mixed models an alysis (Nylander et al. 2004) to assign the appropriate evolutionary mode l to analyze each data pa rtition. The models for GapC-1 and GapC-2 were those that were chosen in the analyses of separate GapC loci, the model used for the ITS data set was GTR + and the model chosen for th e plastid data set was HKY + For each concatenation method, we performed four addi tional analyses, each with one data partition removed at a time, to explore the topological in fluence that each data partition exerts on the total-evidence topology. Each combination of data partiti ons was analyzed using the parsimony and Bayesian methods as described above.

PAGE 62

62 Results GapC Copy Number in New World Mentheae All sequences of GapC-1 GapC-2 and GapC-3 were deposited in GenBank (GapC1: EU179328-EU179374, GapC2 : GapC3 :). Phylogenetic analysis of the exon-only data set to infer copy number contained 49 GapC alleles from 13 individuals, representing 13 species in 9 genera. Across all genera and all copy types, the aligned length of the coding regions was 301 characters, of which 216 were invariant, 65 were parsimony-informative, and 20 were variable but parsimony-uninformative. Parsimony analys es resulted in 59,873 trees of length = 113, CI = 0.876, RI = 0.973, and RC = 0.852. One of the most parsimonious trees resulting from the exononly data set, chosen at ra ndom, is presented in Fig.3-3 A The unrooted exon-only tree placed alleles into two clades separated by a long branch with greater than 20 substitutions; th e clades received 100% bootstra p support (BS) and 1.0 Bayesian Posterior Probability (PP). Alle les of all individuals were found in both clades, with alleles of Old World Mentha placed as sister to the remainder of the alleles in each clade. We used midpoint rooting to root the tree between the two divergent clades (Fig. 3-3 A ), which we interpret to correspond putatively to GapC 1 and GapC 2+3 (Fig. 3-3 A ). Within the GapC 2+3 clade, Mentha (100% BS, 1.0 PP) was sister to a clade of New World mints (80% BS, 0.86 PP). The New World mints were divided into two closel y related clades, with a lleles of all individual placed in both clades. In the subsequent anal ysis of alleles using both intron and exon sequence data, the aligned length of this data set was 528 characters, of which 402 were invariant, 95 were parsimony-informative, and 31 were variable but parsimony-uninformative. Parsimony analyses resulted in 1 most parsimonious tree of length = 169, CI = 0.846, RI = 0.930, and RC = 0.787 (Fig. 3-3 B ). In the most parsimonious tree, the two clades recovered in the exon-only trees were very strongly supported (100% and 99% BS, respectively, each with PP=1.0; Fig. 3-3 B ),. We

PAGE 63

63 interpret these clades to corres pond to two additiona l putative loci; GapC-2 and GapC-3 However, while we identified three putative loci, the primers did not consistently amplify GapC3 ; therefore, for phylogeny reconstruction, we used only GapC 1 and GapC 2. Analyses of Recombination For GapC 1 the phi test ( P =0.18), RDP, and Geneconv did not detect significant evidence for recombination. However, Chimera and MaxChi found 2 and 5 recombination events, respectively, in GapC 1 All of these recombination events were detected in one allele of Mentha ; parental sequences for the recombination even ts were identified to be distantly related ingroup accessions which have had no physic al opportunity to interbreed with Mentha Furthermore, the two types of recombination anal yses did not identify the same recombination locations, visual inspection did not reveal any obvious recombination, and phylogenetic analyses with and without the Mentha allele (results not sh own) did not influence phylogenetic analyses. Thus, we believe that the recombination events detected by these programs are most likely false positives resulting from oversensitivity of the anal ytical methods. None of the methods detected any significant evidence of recombination in GapC 2 Phylogenetic Analysis of GapC-1 and GapC-2 GapC 1 and GapC 2 each have a higher percentage of variable, parsimony-informative characters than either ITS or a combined matrix (~2000 bp) of five plastid regions; characteristics of the data matrices and tree sear ches for all DNA regions are presented in Table 3-1. For both GapC loci, the Bayesian 50% majority-ru le consensus trees and the parsimony strict consensus topologies were nearly identical, vary ing only in the exte nt to which they collapse different short branches. Characteristics of the GapC-1 data matrix and analyses are presented in Table 3-1. GapC 1 contained 45 alleles from 35 i ndividuals; 12 individuals were heterozygous, and 23 were

PAGE 64

64 homozygous. Parsimony strict consensus and Ba yesian majority rule consensus trees were almost identical; we thus only present one of th e most parsimonious trees selected at random and designate the branches that collapse in the strict consensus tree (Fig. 3-4 A ). Resolution among genera was poor in trees based on the GapC-1 data set (Fig. 3-4 A ); Monarda fistulosa Pycnanthemum, and the SE scrub mint clade formed a large polytomy, and the SE scrub mint clade was subdivided into many subclades, severa l of which collapsed either in the Bayesian trees or the parsimony strict consensus (Fig. 3-4 A ). One clade found in the parsimony consensus tree (BS <50%), which collapsed in the Ba yesian topology, did not correspond to generic boundaries; it contained one allele of Conradina verticillata two alleles of C. glabra 109, three alleles of Stachydeoma graveolens and two alleles of Clinopodium dentatum ; sister to this clade was a clade containing all Clinopodium ashei alleles and a single Clinopodium coccineum allele. The remaining clades corresponded to generic bo undaries; one contained a ll alleles of the two Dicerandra species, one contained both alleles of Clinopodium georgianum one contained 13 alleles from five Conradina species, and the last contained ei ght alleles from all six described Conradina species. The clades containing Conradina alleles had no appare nt connection with geography. Characteristics of the GapC-2 data matrix and analyses are presented in Table 3-1. GapC-2 contained 59 alleles from 35 individu als; the amount of heterozygosity at GapC 2 was double that of GapC-1 with heterozygotes tota ling 24 of 35 individuals heterozygous, and 11 were homozygous for GapC-2 There was a 96-bp insertion in exon 9 of the Conradina verticillata GapC-2 allele; this insertion was identical in sequence to a portion of GapC-2 intron 8, and was not found in any other taxon. Because it was not in formative, this insertion was removed from all analyses. Parsimony strict consensus and the Bayesian majority rule consensus trees were

PAGE 65

65 almost identical; we present one of the most pa rsimonious trees selected at random and designate the branches that collapse in the strict consensus tree (Fig. 3-4 B ). Using Mentha as the sole designated outgroup, a clade of Monarda fistulosa and Pycnanthemum muticum was sister to the SE scrub mint clade, although the latter receiv ed <50% BS and 0.52 PP. The SE scrub mint clade formed a large polytomy consisting of many clades of varying size, some of which had high bootstrap support and posterior probability va lues, but relationships among these clades were poorly resolved and collapsed in consensus trees (Fig. 3-4 B ). Many small clades consisted of multiple accessions or multiple al leles of a single species. In Conradina several large clades contained alleles from multiple species of Conradina two of which loosely correspond to geographic location. For example, except for one allele from the Santa Rosa populations from the Florida panhandle, one large clade (13 alleles) was composed entirely of alleles from species from peninsular Florida ( Conradina grandiflora C. brevifolia, and C. etonia) Another geographically coherent clade contained alleles fr om species distributed in the Florida panhandle and northern regions, includi ng an allele each from C. verticillata C. glabra Santa Rosa population 133-30, and C. etonia The final Conradina clade does not correspond to geographic boundaries and contained two alleles from Conradina brevifolia two alleles from the Santa Rosa populations, and one allele from C. canescens. Interestingly, C. canescens and the taxa of uncertain status, C. brevifolia and the Santa Rosa popul ations, were involved in all instances in which alleles do not correspond to geographic locations. Congruence Analyses All pairwise ILD analyses rejected the null hypo thesis of partition homogeneity. However, visual inspection of the four gene trees identified only one species, Conradina canescens 142-1, that was placed with >80% BS support in conf licting locations; it was placed in a clade (85% BS) with Conradina glabra, C. grandiflora and C. canescens 130-2 in the plastid data set, as

PAGE 66

66 sister (83% BS) to Conradina canescens 128-2 in the GapC-2 data set, and in a clade with Conradina etonia C. brevifolia 198-2, C. verticillata and the Santa Rosa population 133-1 in the GapC-1 data set. Because there was only one instance of strongly su pported incongruence, we carried out combined analyses using all f our data sets. However, given the conflicting placements of C. canescens 142-1, we regarded it with cauti on in all combined analyses. Combined Phylogenetic Analyses Characteristics of the data matrices and tree searches for the combined analyses are presented in Table 3-1. When all four data sets were conc atenated by combining heterozygotes into one sequence by coding hete rozygous sites using an IUB c ode, the Bayesian and parsimony trees were almost identical and almost completely resolved; one of the most parsimonious trees selected at random, with branches indicated that collapse in the stri ct consensus tree, is presented in Fig. 3-5. Both parsimony and Baye sian topologies placed alleles from Monarda fistulosa and Pycnanthemum muticum as successive sisters to a clade composed of alleles from all SE scrub mint clade taxa. A clade of Dicerandra alleles was sister to the remainder of the SE scrub mint clade. The remaining SE scrub mint taxa were divided into two clades The first clade was composed of all alleles of Clinopodium ashei with Clinopodium coccineum Clinopodium dentatum, and Stachydeoma as successive sisters. The othe r large clade was composed of a monophyletic Conradina with Piloblephis rigida and Clinopodium georgianum placed as successive sisters to Conradina Within Conradina very few branches received >50% BS support, but most received PP va lues of greater than 0.95, and re lationships were well resolved; all taxa were grouped into two geographically structured clad es. One clade contained only alleles from species that inhabit the Florida panhandle and northern habitats: C. glabra, C. canescens, C. verticillata and one accession from the Sant a Rosa populations. The second Conradina clade, with the exception of the second accession from the Santa Rosa populations,

PAGE 67

67 was composed of alleles from peninsular Conradina species: C. brevifolia C. grandiflora and C etonia No species was monophylet ic in either clade. All four data partitions were also concaten ated using the “all combinations” method and analyzed simultaneously; one of the most parsimoni ous trees selected at random is presented in Fig. 3-6. Relationships among gene ra recovered from analyses of the “all combinations” method were almost identical to those found in trees based on the IUB code method, with the only difference being that instead of alleles of Clinopodium georgianum and Piloblephis rigida being placed as successive sisters to Conradina a clade of C. georgianum alleles was sister to Conradina and P. rigida collapsed into a polytomy. Within Conradina parsimony and Bayesian trees exhibited a large amount of topological variation that collapsed in consensus trees. In comparison to the high level of re solution and geographical structuring found in Conradina in analyses using the IUB code concaten ation method (Fig. 3-5), trees based on the “all combinations” method (F ig. 3-6) did not group Conradina into two geographically structured clades; Conradina alleles formed a large polytomy composed of several clades, which generally included a lleles or accessions of a single species. Removal of Single Partitions Analyses with one data partition removed usi ng the data matrices resulting from the two concatenation methods produced similar results (results no t shown). When GapC 1 was removed, resolution among SE scrub mint genera d ecreased in comparison to the total evidence tree, and relationships among many Conradina species collapsed. When GapC-2 or the plastid data set was removed, relationships within Conradina were much less reso lved and lost their geographic structure, and the sister group to Conradina became unresolved. When ITS was removed, relationships among SE scrub mint genera were drastically altered compared to the total evidence tree, indicating that the signal in the ITS data se t strongly influenced relationships

PAGE 68

68 in the combined analyses. In these trees overa ll resolution decreased markedly, and in the IUB code tree, Conradina became nonmonophyletic, with some Conradina alleles forming a clade with Stachydeoma, Clinopodium ashei, C. dentatum, and C. coccineum. Discussion Copy Number and Evolution of GapC in New World Mentheae At least three putative GapC loci are present in the New World Mentheae sampled in this study, concordant with a study of GapC evolution in another asterid, Amsinckia (Boraginaceae; Perusse and Schoen, 2004). However, while three copies were found in both the SE scrub mint clade and Amsinckia spectabilis we aligned and carried out phylogenetic analysis of the sequences from both studies and found that they are not orthologous. All putative copy types found in Amsinckia are very homogeneous in sequence and structure and clustered with GapC-1 whereas the three copies in New World Mentheae ar e much more divergent from one another. In the New World Mentheae, an initial duplication event resulted in GapC 1 and the ancestor of GapC 2 and GapC-3 Because these loci are present in a ll mint taxa sampled in this study, this initial gene duplication must predat e the origin of these genera a nd may be ancient, as evidenced by the large amount of sequence divergence betw een the two copy types. Indeed, the intron regions are so divergent that sequence ali gnment and homology assessment between the two copy types were impossible, and in 301 bp of protein-coding regi ons, the two loci differ by >20 substitutions (Fig. 3-3 A ). A wide survey of GapC copy number in the asterids, or perhaps the eudicots, would likely be necessary to understand the phylogenetic origin of this duplication. A second duplication event yielded GapC-2 and GapC-3 This is a much more recent duplication as evidenced by the fact that the exons of the two loci are almost identical (Fig. 3-3 A ) and their introns are similar enough to be aligned. Because all New World mint taxa have both GapC 2 and GapC 3 while only one copy is found in the Old World Mentha sp., the duplication

PAGE 69

69 event may have occurred since the divergen ce of the New World ta xa from Old World Mentha However, the single copy type in Mentha may also be due to other processes such as gene loss, so a more thorough sampling of GapC-3 in the Mentheae would be necessary to pinpoint the exact origin of this duplication. One possible cause of the multiple loci in the New World Mentheae, given the relatively high base chromo some number of many members of this clade (for example, in Conradina, n =12; Gray, 1965), is an ancient pol yploidization event. However, regardless of whether the multiple loci are paralogs or homeologs, the multiple GapC loci found in this study also high light the importance of ensuring that only orthologous genes are included in comparative evolutionary studies. Separate Phylogenetic Analysis of GapC-1 and GapC-2 Although both GapC loci have a much higher percentage of parsimony-informative characters than ITS or plastid sequence data (Table 3-1), neither contained enough information to resolve phylogenetic relationships in the New World Mentheae, even between morphologically divergent genera. For example, both GapC loci only weakly differen tiated the SE scrub mint clade from other morphologically di stinct New World mints such as Monarda and Pycnanthemum and GapC loci did not contain enough info rmation to clarify phylogenetic relationships among SE scrub mint taxa (Figs. 4 A B ). The monophyly of most SE scrub mint genera is unresolved; some gene ra are placed as non-monophyletic in the strict consensuses of both GapC trees, but none of these clad es received high BS and/or PP values. Instead, all clades with high BS and/or PP values were composed of alleles from a singl e genus; however, none of these strongly supported clades contained all alleles of the entire genus. Thus, because the monophyly and relationships among the genera rema in unclear, this precludes most assessments of intergeneric hybridization (E dwards et al. 2006) based on GapC data.

PAGE 70

70 One clade in GapC 1 does suggest that hybridization among genera may have occurred; alleles from Conradina glabra Conradina verticillata Stachydeoma graveolens and Clinopodium dentatum were placed in the same clade. Although this clade did not receive bootstrap support >50%, it was recovered in bo th parsimony and Bayesian trees. Plastid haplotypes for C. dentatum and one accession of S. graveolens are identical, this clade corresponds to geography (Fig. 3-1), and the species flower during the same time period, allowing for the possibility of hybridization. Cros sing studies may be the best way to evaluate this hypothesis. Within Conradina, GapC 1 and GapC 2 individually provided very little resolution. Reconstructions based on GapC 1 did not place Conradina alleles into clades corresponding to species limits or geographic location. Based on GapC-2 several multi-species Conradina clades corresponded to geographic location, but no clade contained all alle les that would be expected on the basis of geographic location. One noteworthy clade in the GapC-2 trees contained some alleles of C. canescens and some alleles from the two taxa of uncertain status, the Santa Rosa populations and C. brevifolia. This clade may indicate that C. canescens, C. brevifolia, and the Santa Rosa populations may share a common ancestor or alternativel y, experienced gene flow at some time in their evolutionary history. Howeve r, in general, there was little resolution of relationships among the various Conradina clades; many of these clades were poorly supported, and thus we are unable to make strong conclusions regarding sp ecies relationships or taxonomic circumscriptions within Conradina using either GapC data set alone. Combined Analyses—Congruence and th e Effect of Concatenation Approach Although many studies have hail ed singleor lowcopy nuclear genes as a solution for resolving relationships among closely related taxa such as these (Sang 2002), the alleles of individuals in the GapC data sets either do not coalesce or are of insufficient length to resolve

PAGE 71

71 relationships among closely related Conradina species. Depending on substitution rate, a much longer gene region would likely be necessary to resolve relationships. In many cases, the most reasonable solution to this problem is to combin e nuclear genes with sequence data from other genomes or gene regions for a total eviden ce analysis. However, congruence among data partitions should first be assessed, and heterozygos ity of the nuclear genes must be taken into account prior to phylogenetic analysis. Significant levels of incongruen ce exist between ITS and plastid data partitions based on ILD tests (Edwards et al., 2006) and in the present study we al so found significant levels of incongruence among all four data partitions. However, because many different factors could cause a significant ILD result, we visually insp ected the topologies and found only one instance of “hard incongruence” (Seelanan et al. 1997): Conradina canescens 142-1 was placed with >80% bootstrap support in conflicting positions in di fferent trees. Because there was very little well-supported conflict, we carried out co mbined analysis of our data sets. Trees based on the two concatenation methods were generally very similar in their resolution and support along the backbone of the trees (i.e., relationships among genera). Within Conradina however, the two methods produced very di fferent topologies; the topology derived using the IUB code method (Fig. 3-5) is well resolved and strongly supported, but the trees resulting from the “all combinations” (Fig. 3-6) ar e markedly less resolved. The variation in the topologies resulting from the concatenation met hods may be the result of differing treatment of conflict and heterozygosity in the Conradina sequence data. If the data sets conflict because of hybridization or incomplete lineage sor ting, both of which may be present in Conradina, the “all combinations,” method may support multiple diffe ring placements of a taxon due to conflicting signal across the length of a seque nce, causing branches to collapse in consensus trees. The “all

PAGE 72

72 combinations” (Fig. 3-6) method also allows an individual to occupy multiple terminals, which permits alternative placements of an individual due to heteroz ygosity. Conversely, the IUB code approach ignores both heterozygosity and conflict Using this method, the most parsimonious of the states contained in the ambigu ity code is assigned to a bran ch after analysis (Swofford and Begle 1993). Thus, even if the two alleles of an individual should be placed in differing positions in a tree, using this method the individu al can only occupy one branch and the variation will be ignored. This reduces conflict and result s in a more strongly supported and well-resolved tree, but if conflict truly exists in a data set, using this me thod may result in a well-resolved and strongly supported tree that is perh aps not an accurate portrayal of relationships. Furthermore, it is assumed that all alleles represented in a terminal are monophyletic (Swofford and Begle 1993), and if the two alleles of a heterozygote are not monophyletic because of processes such as hybridization or incomplete so rting, the IUB code method vi olates the assumptions of phylogenetic analysis. Thus, if e ither hybridization or incomplete lineage sorting is suspected, it may be more valid to use the “all combinations” to portray the uncertainty involved in the phylogenetic hypothesis. Because the true evolutionary tree of the SE scrub mint clade is unknown, and we are unable to compare the results of the concatenation methods with a known tree, it is impossible to ascertain which, if either, of th e concatenation approaches is be tter for combining data. One of the most interesting prospects for further resear ch on combined analysis of heterozygous loci would be a simulation study to explore how vary ing methods of concat enating heterozygotes would affect phylogenetic accuracy of results under bifurcating evolution, after hybridization, or after incomplete sorting. Simulations would enable us to compare the resu lts of analyses based on the concatenation approaches against a true tree and would provide valuable information on

PAGE 73

73 how concatenation methods perform with heteroz ygosity caused by different processes, so that we may understand whether any of these methods can recover accurate phylogenetic relationships when heterozygosity is present. Furthermore, a simula tion study could test our hypotheses of whether disagreement among the c oncatenation methods results from conflicting signal among data sets, and whether the IUB code produces a misleading, well-resolved tree because it ignores conflict in the data set, or conversely, if it is actually an accurate representation of phylogene tic relationships. Combined Phylogenetic Analyses—Relatio nships in the SE Scrub Mint Clade In general, combined analyses of the four data partitions fully resolved relationships among genera of the SE scrub mint clade, with much higher BS and PP values than obtained with individual data sets. The resolution of thes e higher-level relationships in the topologies is largely due to the signal found in the ITS and GapC 1 data sets; when either data partition was removed, particularly ITS, relationships among SE scrub mint taxa collapsed. Topologies derived from all concatenation methods are co ngruent and support a monophyletic SE scrub mint clade, with Dicerandra as sister to the remainder of the clade. One morphol ogical feature that may unite this clade is habit: most New World mint species are herbaceous, Dicerandra species are herbaceous and suffrutescent, and the rema inder of the SE scrub mint clade is woody (Edwards et al. 2006). The remaining SE scrub mint taxa were divided into two clades. The first clade consisted of Stachydeoma graveolens Clinopodium dentatum and C. coccineum as successive sisters to Clinopodium ashei Interestingly, Principal Com ponents Analysis (PCA) of quantitative morphological char acters found a high degree of mo rphological similarity among Clinopodium dentatum C. coccineum and C. ashei while other Clinopodium species, notably C. georgianum were morphologically distinct (Crook 1998). Stachydeoma graveolens was not included in Crook’s (1998) anal ysis; however, it is morphologi cally distinct. The three

PAGE 74

74 Clinopodium species are upright shrubs to 0.5 m tall a nd have flowers in loose axillary clusters, whereas Stachydeoma is a sprawling shrub to 0.2 m tall, with a much more compressed floral axis. The composition of the second cl ade of SE scrub mints vari es depending on combination method, indicating that conflict in pla cement may exist in the sequences of Clinopodium georgianum and Piloblephis rigida The IUB code (Fig. 3-5) placed C. georgianum as sister to Piloblephis + Conradina, whereas the “all combinations” (Fig. 3-6) method placed Clinopodium georgianum as sister to Conradina. Clinopodium georgianum was not placed with the remaining three Clinopodium species in any of our trees, in ag reement with Crook (1998), who found that C. georgianum was morphologically distin ct from the other three Clinopodium species; however, Crook (1998) found no morphological similarity between C. georgianum and Conradina or Piloblephis Regardless, the woody Clinopodium species from the southeastern U.S. do not appear to be monophyletic. Th e close relationship between Conradina and Piloblephis has also been found in other studies; for example, the two taxa have identical plas tid DNA restriction site profiles (Wagstaff et al. 1995). However, Crook et al. (1998) found them to be very distinct morphologically; for example, instead of flower s grouped in loose axillary clusters, as in Conradina the terminal flowering branches of Piloblephis rigida are compressed to form a tight head, and the species is unique in having rigid, tria ngular hairs covering all surfaces of the plant. The only obvious characters that may unite Conradina and Piloblephis are similar habit and the fact that both have reduced, needle-like leaves with tightly rolled margins. Both concatenation methods strongly support the monophyly of Conradina. This relationship is also suppor ted by morphology; species of Conradina are readily distinguishable from other species in the SE scrub mint cl ade, united by the morphological synapomorphy of a

PAGE 75

75 distinctively geniculate corolla (Crook 1998). Species of Conradina are morphologically very similar to each other and are generally distingu ishable by only small differences in leaf, flower size, or pubescence char acters (Gray 1965). Combined Phylogenetic Analyses–Relationships among Conradina Species Signal in the GapC-2 and plastid sequence data cont ributed most to resolution of phylogenetic relationships among Conradina species. However, the two concatenation methods differ in their amount of resolution within Conradina The “all combinations” method did not resolve relationships among Conradina species (Fig. 3-6), while trees based on the IUB code method (Fig. 3-5) are almost completely re solved. Although the IUB code method ignores polymorphic sites and violates the assumptions of analysis because th e two alleles of many Conradina individuals included in a terminal ar e not monophyletic, the resulting trees are difficult to ignore because they present a hypothesi s of evolutionary relationships which is in agreement with geography. With only one exception, species of Conradina are placed into two clades that correspond to geographi c location; one clade contains the peninsular species, and the other contains the northern/panhandle specie s. This is a common biogeographical split, described as “the Apalachicola River Basin disc ontinuity” (Soltis et al. 2006), and hypothesized to have arisen because the panhandle and penins ular areas became separate islands during the high sea levels of Pleistocene interglacial period s (Soltis et al. 2006) The panhandle and peninsular clades of Conradina may have diverged on these isla nds and subsequently dispersed to their current distributions after sea levels dropped. However, because this geographic structure was only supported using one of the concatenation methods, this hypothesis should be tested further using different types of data and more rigorous statistical methods. Although there may be some geogra phical signal within the genus Conradina none of the individual data sets or the combined analyses recovere d the monophyly of any Conradina

PAGE 76

76 species. The apparent lack of species m onophyly reported here based on sequence data correlates with the high degree of morphological similarity in Conradina as well as the ability of Conradina species to hybridize in crossing experi ments (Gray 1965). One explanation for the nonmonophyly of the described Conradina species recovered in this study is that the taxonomic groupings are not representative of true biological entities. Howeve r, another explanation is that Conradina species are closely related and probably very recently derived, so the alleles of each species may not yet have coalesced. It may therefore be unreasonabl e to expect species monophyly using DNA sequence data. Other type s of data, such as rapidly evolving microsatellite data, will likely be more useful for determining if interspecific hybridization has occurred, and for clarifying species boundaries and taxonomic groupings in Conradina

PAGE 77

77Table 3-1. Characteristics of DNA matri ces and results of phylogenetic analyses Data set Number of Terminals Percent HeterozygosityLength Percent Parsimony Informative Characters Number of MostParsimonious Trees Tree length CI RI RC Locus1 45 34.3% 472 13.7% 441 171 0.848 0.884 0.750 Locus2 59 68.5% 512 18.3% 92,800 176 0.795 0.877 0.697 ITS (from Edwards et al. 2006) 35 0% 624 8.8% 99,400 164 0.841 0.864 0.727 Plastid (from Edwards et al. 2006) 35 0% 2034 1.4% 96,300 111 0.946 0.921 0.821 ITS+Plastid combined 35 0% 2658 3.1% 2001 284 0.856 0.846 0.724 IUB code method 35 n/a 3642 4.4% 46 589 0.793 0.751 0.595 “All combinations” method 80 n/a 3642 11.4% 220 770 0.684 0.880 0.602

PAGE 78

78 Figure 3-1. Distributions of Conradina species

PAGE 79

79 Figure 3-2. Collection locations of accessions in cluded in this study (figure reproduced from Edwards et al. (2006)). Numbers follo wing taxon names indicate the accession number of the population and the number of the individual in the population. Accessions obtained from native nurseries or botanical gardens are not shown (these include Clinopodium georgianum Conradina brevifolia from Bok Tower Garden, and outgroups Mentha sp ., Pycnanthemum muticum and Monarda fistulosa ).

PAGE 80

80 Figure 3-3. Relationships among GapC loci in the SE scrub mint clade. Parsimony bootstrap values >50%/Bayesian posterior probabil ities are indicated above the branches, arrows indicate branches that collapse in the strict consensu s, numbers following taxon names indicate the accession number of the population and number of the individual in the population, and numbers fo llowing locus names represent the allele number for that locus. An asterisk denot es individuals that are heterozygous for a particular locus. ( A ) One of the most parsimonious tr ees resulting from analysis of the exon-only data set, which includes all al leles recovered from a reduced data set of 13 taxa. The two hypothesized duplication ev ents and putative lo ci are indicated. ( B ) The most parsimonious tree resulting from analysis of intron and exon data of GapC2 and GapC-3 to test if intron differences also support clades rec overed in the exononly trees

PAGE 81

81 Figure 3-4. Trees based on se parate analyses of the GapC data sets. Parsimony bootstrap values >50%/Bayesian posterior probabilities are indicated above the branch es, arrows indicate branches that coll apse in the strict consensus, numbers following taxon names indicate the accession number of the popul ation and the number of the individual in the population, and asterisks denote heterozygotes. A) P hylogram of one of the most parsimonious trees based on analysis of GapC-1, and B) phylogram of one of the most parsimoni ous trees based on analysis of GapC-2.

PAGE 82

82 Figure 3-5. Phylogram of one of the most parsimonious trees resu lting from combined analysis of GapC-1 GapC-2 ITS and plastid data sets, with heterozygotes concatenated using the IUB code method. Parsimony bootstra p values >50%/Bayesian posterior probabilities are indicated above the branches, arrows indicate branches that collapse in the strict consensus, and numbers following taxon names indicate the accession number of the population and the number of the individual in the population.

PAGE 83

83 Figure 3-6. Phylogram of one of 220 most parsimonious trees re sulting from analysis based on combined analysis of GapC 1 GapC 2 ITS, and plastid data sets, with heterozygotes concatenated in all possible combinations (maximum of four per individual). Each unique combination of alleles of heteroz ygotes is denoted at the end of the taxon name as A, B, C, or D. Parsimony b ootstrap values >90%/Bayesian posterior probabilities are indicated above the branch es. Posterior probability values below 0.90 are not included due to variation among Baye sian runs in that branches that have lower than 0.90 PP. Arrows indicate branches that collapse in the strict consensus and numbers following taxon names indicate the accession number of the population and the number of the indivi dual in the population

PAGE 84

84 CHAPTER 4 COMBINED PHYLOGENETIC ANALYSIS OF HETEROZYGOUS NUCLEAR LOCI: AN EXAMPLE USING CONRADINA AND RELATED MINTS (LAMIACEAE) Introduction The use of single-copy nuclear genes for resolving phylogenetic relationships among closely related species has become increasingly common because the introns of these genes often have a much more rapid evoluti onary rate than other regions su ch as ribosomal or organellar markers (Sang 2002; Small et al. 1998; Small et al. 2004). However, although individual nuclear gene introns often have higher s ubstitution rates than ot her types of sequence data, they may still lack sufficient information to resolve evolut ionary relationships, especially at shallow phylogenetic levels (Zhang and Hewitt 2003), perh aps because of short length or lack of coalescence. Combined analysis of multiple genes often increases resolution and support for nodes (Bardeleben et al. 2005; Ga ines et al. 2005; Pereira et al. 2002; Slade et al. 1994). However, nuclear genes are bipare ntally inherited and may be he terozygous, which can lead to phylogenetic inconsistencies. In phylogenetic an alyses of a single heterozygous nuclear gene, heterozygosity is usually accounted for by treating each allele as a terminal (Wiens 1999). This approach is problematic for combined analyses because it is possible fo r the two alleles of a heterozygote to occupy different positions in the resulting tree (Doyle 1995; 1997). The question then arises as to which allele will “represent” an individual when concatenating gene sequences of each individual into a single, multigene “s equence” for combined phylogenetic analysis. Several methods have been used to deal with heterozygosity of a nucle ar gene in combined analysis. Authors have randomly chosen one a llele per individual (Ari as and Sheppard 2005; Larkin et al. 2006; Malcomber 2002), coded the heterozygous site as missing or by using an ambiguity code (Calcagnotto et al. 2005; Emsh willer and Doyle 2002; Howarth and Baum 2005; Iglesias et al. 2005; Reeder et al. 2006; Sota and Vogler 2003 ; Strand et al. 1997; Won et al.

PAGE 85

85 2006), or represented heterozygotes by using more than one terminal per individual (Howarth and Baum 2005; Sota and Vogler 2003). However, if a heterozygote is represented by multiple terminals, inclusion of more than one heteroz ygous locus further comp licates analysis. If multiple nuclear genes are closely linked and heterozygous, several computer programs can determine the allelic phase of th e linked genes (Niu et al. 2002; Niu 2004; Scheet and Stephens 2006; Stephens et al. 2001; Zhang et al. 2006), a nd thus an individual can be represented by two terminals in a data matrix, each corresponding to a linked set of genes inheri ted from one parent. However, if genes are unlinked and more than one te rminal is used to represent an individual, the process of concatenating the alleles is not straightforward. Unlinked genes will assort independently and are not associated in any part icular way; thus, concatenation of the alleles should be random. Although many studies have proposed concatenation methods to deal with heterozygosity in combined phylogenetic analysis, no study has reviewed or evaluated the validity of these methods. The purpose of this study is to review previously published methods of concatenating two or more nuclear loci for combined phyloge netic analyses and expl ore the use of these methods using four previously published data sets (ITS, plastid, and two heterozygous nuclear data sets from the GapC gene family) from the southeastern (SE) scrub mint clade (Edwards et al. 2006; Edwards et al. in press). The SE scrub mint clade is a group which includes Conradina and other related mints (Lamiaceae), i.e., Dicerandra (nine species), Piloblephis (monotypic) Stachydeoma (monotypic), and the woody, southeastern U. S. species of Clinopodium : C. ashei, C. georgianum, C. coccineum, and C. dentatum (Crook 1998; Edwards et al. 2006; Edwards et al. in press; Trus ty et al. 2004).

PAGE 86

86 After surveying the literature for methods th at account for heterozygosity in combined phylogenetic analyses (i.e. different concatenation methods), we evaluated the validity and utility of these methods. Because ignoring polymorphi sm can reduce the accuracy of phylogenetic inference (Wiens 1999), we also evaluated how each of the concatenatio n methods accounts for polymorphism in a data set. We then used six methods to concatenate the two heterozygous GapC data sets with the ITS and pl astid data: 1) random selection of an allele to represent an individual, 2) all possible combin ations of alleles, following Sota and Vogler (2003), 3) the CI was used to reduce the number of terminals to 2 per individual following Howarth and Baum (2005), 4) two consensus approaches, replacing he terozygous site s with an N (Sota and Vogler 2003) or the appropriate ambiguity code follo wing Howarth and Baum (2005), and 5) the distance-based POFAD method of Joly and Bruneau (2006). We then conducted phylogenetic analyses and compared the trees resulting from these concatenation methods to examine their effects on tree topology, resoluti on, and support. Finally, we ma de recommendations as to the methods that best account for heterozygosity in combined analyses of nuclear genes. Materials and Methods Characteristics of Data Matrices We used the ITS, plastid, and two GapC data sets from the SE scrub mint clade to investigate concatenation methods (Edwards et al 2006; Edwards et al. in press); characteristics of these data matrices are presented in Table 4-1. The ITS and plastid data sets were both singlecopy, while GapC 1 contained 45 alleles from 35 individuals; 12 indi viduals were heterozygous, and 23 were homozygous. GapC-2 contained 59 alleles from 35 i ndividuals, and the amount of heterozygosity at GapC 2 was double that of GapC-1 ; of the 35 individuals, 24 were heterozygous, and 11 were homozygous for GapC-2 Tests for recombination in the GapC loci

PAGE 87

87 and pairwise congruence tests were carried out in previous analyses of these data sets (Edwards et al. in press). Methods of Concatenating Multiple Heterozygous Nuclear Genes. Seven methods for conducting combined analyses of heterozygous data sets were evaluated and assessed for suitability with the SE scrub mint clade data. Method 1. Reduce the data set to e xclude all heterozygotes This method is only reasonable if only a few individu als are heterozygous; otherwise, most individuals, and therefore most taxa, will be excluded. We did not pursue this method because only 7 of the 32 SE scrub mint individuals remained after removal of all individuals heterozygous for at least one GapC locus. Method 2. Select one allele of a heterozygote To select an allele, authors have used a priori selection criteria, elimina ting any heterozygosity prior to phylogen etic analysis by sequencing only one colony of a cloning reaction (e.g., Larkin et al. 2006) or using only the most common allele recovered from sequencing multipl e colonies of a cloni ng reaction (e.g., Arias and Sheppard 2005). Although these are valid met hods for randomly selecting an allele to represent an individual in a data matrix, they do not take al lelic variation into account. An a posteriori allele selection crit erion was used by Malcomber (2002) in a study of PepC sequence data in Gaertnera ; after phylogenetic analysis, all PepC alleles of a single individual clustered together, so the sequence with the shortest patristic distance to the base of the species cluster was selected to be concatenated with other sequenc e data. However, this method is only suited to data sets in which heterozygosity is not informative; we did not use the a posteriori method because alleles of a single individual did not alwa ys cluster together (for five individuals in GapC 1 and ten individuals in GapC 2 ).

PAGE 88

88 Method 3. Code any heterozygous site as missing/unknown. In this method all sites that differ between the two alleles are coded as an N. During analysis using PAUP*, these sites are used neither as a synapomorphy nor homoplasy dur ing tree reconstruction. Instead, the state for the missing/unknown character is assi gned after the tree is recons tructed; any of the possible character states can be assigned, even ones that are not present in an individual, depending on which is the most parsimonious optimization (Swofford and Begle 1993; Wiens 1999). For a detailed description of the stre ngths and weakness of this met hod, see method 4 below, which is very similar to the N method. Method 4. Combine two alleles of a heterozygote into a single sequen ce, substituting any heterozygous sites with the appropr iate International Union of Biochemistry (IUB) ambiguity code This method is by far the most commonly us ed to code heterozygotes for phylogenetic analysis (Buckley et al. 2006; Calcagnotto et al. 2005; Emshwiller and Doyle 2002; Howarth and Baum 2005; Iglesias et al. 2005; Reeder et al. 2006; Sota and Vogler 2003; Strand et al. 1997; Won et al. 2006). We will hereafter refer to this method as the IUB code method. During parsimony analysis using PAUP*, the program can tr eat the IUB codes as either “uncertainties” or “polymorphisms.” While most authors do not specify which setting is used, treating the character as “uncertainties” is th e default setting in PAUP* and is thus likely the most widely used method. If the “uncertainties” option is se lected, this assumes that the state is not known but is one of the options designated by th e IUB code (Swofford and Begle 1993). During analysis, the terminal is assigned the most pars imonious of the states in cluded in the IUB code (Swofford and Begle 1993; Wiens 1999). The “polymorphism” option is very similar in that, depending on the tree topology, PAUP* assigns the mo st parsimonious of the present states to the node; however, it differs in that it treats all polymorphic sites as pres ent in the terminal, and

PAGE 89

89 thus places the less parsimonious character states as extra steps along th e branch (Swofford and Begle 1993). The trees resulting from the “polym orphism” option will therefore be longer than those using the “uncertainty” option. If a data set is coded usi ng the IUB method and analyzed using parsimony in PAUP*, PAUP* assumes that the consensus sequence for an individual is composed of a monophyletic group of sequences, and that th e common ancestor of the sequen ces possessed only one of the observed states (Swofford and Begle 1993). The program does not allow internal nodes to be polymorphic. Thus, if alleles of a heterozygote are not monophyletic and derived from different sources, as in cases of hybridization, this violates the assumptions of analysis. In this case, if alleles with disparate evolutionary histories occupy a single terminal in the tree and are treated as a monophyletic group, this may influence the topology that is recovered. Method 5. Include all possible combinations of alle les from an individual as terminals in the data matrix (Sota and Vogler 2003). Usi ng our data set as an example, if an individual is homozygous for GapC 1 and heterozygous for GapC 2 the individual would be represented twice in the matrix; the two concatenated sequ ences would be identical for ITS, plastid, and GapC-1 and would differ for GapC 2 If an individual is heterozygous for both GapC loci, all four combinations of alleles would be concatenat ed and included in the analysis. The weakness of this method is that if some data partitions are homozygous or haploid, as in the case of organellar genes, concatenated sequences for a single individual will be identical for the homozygous or haploid partitions. Any autapomorphi es in the haploid data sets will become synapomorphies in the concatenated sequences of a single individual and may force the concatenated alleles of an indivi dual together in the phylogeny.

PAGE 90

90 A similar method to that of Sota and Vogler (2003) is one in whic h length mutations and polymorphic sites are treated differently (Creer et al. 2006); polymorphic sites resulting from nucleotide substitution were coded as an N and gaps were combined in all possible combinations based on length variation. We did not use this method because it was developed to incorporate length variation into phylogenetic analyses, which is outside the scope of this study. Method 6. Include only two terminals per individual by choosing a particular combination of alleles If an individual is heterozygous for two genes, A and B, each with two alleles designated as 1 and 2, only two configurations of terminals are possible that contain all of the alleles present in an individual, wi thout using the same allele twice. These are: 1) terminal A1B1 and terminal A2B2 and 2) terminal A1B2 and te rminal A2B1. In this approach, the consistency indices of the possible configur ations are evaluated, and the configuration with the higher Consistency Index (CI) is selected for inclusion in the data matrix (Howarth and Baum, 2005). This approach results in only two terminals per i ndividual, and these terminals contain all alleles present in an individual, without repetition. However, when selecting an optimal configuration of alleles, the only way to truly maximize the CI w ould be to evaluate the CI of the configuration of alleles for each individual in relation to the configurati on of alleles for every other heterozygous individual, which would result in a large number of matrices to be evaluated. For example, if two heterozygous nuclear gene s are going to be concatenated, with n individuals heterozygous at both loci, then 2n matrices must be evaluated. If there are three heterozygous genes, and n is again the number of individuals heterozygous at all loci, there are 8n possible matrices to be evaluated. Another weakness of this method is its use of the CI to select the optimal configuration of alleles; although the CI may be highest for a certain configuration of

PAGE 91

91 alleles, this metric is not m eaningful biologically because a lleles of unlinked genes assort independently. Another approach to choosing a pa rticular configuration of allele s is that of Petersen et al. (2006), who examined individual gene trees to identify the position of the alleles of a heterozygote. Alleles that were placed in congruent positions in different gene trees were concatenated. However, Petersen et al. (2006) expected to have trees with heterozygotes placed in congruent positions because the heterozygosity was due to allopolyploidization; congruent placement of alleles of different loci was expect ed because they were inherited from the same diploid progenitor. This method is not relevant for independently asso rting loci in diploid individuals. Method 7. Use the Phylogeny for Organisms From Allelic Data (POFAD) algorithm (Joly and Bruneau 2006). This algorithm was developed specifically to account for heterozygosity in combined analyses of nuclear genes. The progr am uses pairwise distances among all alleles of all individuals as its input. The output is a distance matr ix in which each individual is represented once; this matrix can be input into anothe r program, such as PAUP* or NeighborNet, to produce a graphical representa tion of distance among individuals in the form of a network, UPGMA, or neighbor joining tree. To comput e pairwise distance betw een individuals (not alleles), the program uses three different algorith ms to account for the following situations: 1) if two individuals are homozygotes, the distance valu e will simply be the distance between that pair of alleles, 2) if one individual is homozygous and anot her heterozygous, the distance between the individuals will be the average of the distances between the one allele of the homozygote and the two alleles of the heterozygo te, and 3) if both indi viduals are heterozygous, the algorithm calculates the distance between two i ndividuals as the mean of the shortest pair of

PAGE 92

92 distances between their alleles (Joly and Bruneau 2006). Mul tiple data sets are combined by averaging the pairwise distances be tween taxa for all data sets. We used methods 2, 3, 4, 5, and 6 to concatenat e our data sets, and method 7 was also used to combine our data (see below). For method 3, a ny sites that varied between the two alleles of an individual were coded as an N, and lengt h between the two alleles was coded using the sequence without a gap. For method 4, the IU B code method, we followed Howarth and Baum (2005). For method 5, the “all combinations” appr oach, we followed Sota and Vogler (2003). For method 2, we used the data matrix genera ted using method 5 and randomly selected one allele per data set by deleting all but terminal A for each indi vidual. For method 6, we followed Howarth and Baum (2005). Methods of Phylogenetic Analysis For all concatenation methods employed, we c onducted a simultaneous analysis of all four data partitions ( GapC 1 GapC 2 ITS, and plastid) using both parsimony and Bayesian phylogenetic methods. All phylogenetic analyses were carried out using the Phyloinformatics Cluster for High Performance Computing at the Florida Museum of Natural History (http://cluster.flmnh.ufl.edu). We conducted pars imony analyses with heuristic searches using 1000 random addition replicates and TBR branch sw apping, saving 100 trees per replicate. When heterozygous sites were coded us ing the IUB ambiguity code, we conducted two analyses, one in which polymorphisms were treated as “polymor phisms” and another in which polymorphisms were treated as “uncertainties.” Bootstrap analyses (Felsenstein 1985) with 1000 replicates were used to assess branch support us ing a heuristic search with TBR branch swapping, one random addition per replicate, saving no more than 100 trees per replicate. We also analyzed the data using Bayesian phylogenetic analys is (Huelsenbeck et al. 2001; Larget and Simon 1999; Yang and Rannala 1997) as implemented in MrBayes 3.11 (Ronquist and Huelsenbeck 2003). We ran

PAGE 93

93 two analyses using four chains each, three hot and one cold, with the temperature set to the default of 0.2. All analyses were run using flat priors, and we used mixed models analysis (Nylander et al. 2004) to assign the appropriate evolutionary m odel chosen for each data set by the AIC as implemented in Modeltest (Posada an d Crandall 1998). The model used for the ITS and GapC-1 data sets was GTR + and the model chosen for the plastid data set and GapC-2 was HKY + (Hasegawa et al. 1985; Yang 1994). We ran analyses for 10 million generations, sampling a tree every 1000 generations. To dete rmine the burn-in value, we checked the log likelihood scores of the resulti ng trees and the split frequencie s between the two independent runs, and discarded any trees that were saved prior to the stabiliza tion of the scores or that had a split frequency of greater than 0.1. For the two in dependent runs, posterior probabilities for each branch were found by constructing a majority-rule consensus of trees in the posterior distribution using PAUP* 4.0b10 (Swofford 2002). The two iden tical, independent analyses were checked for convergence in topology and branch lengths. Results Characteristics of the data matrices and tree searches for the combined analyses for all concatenation methods are presented in Table 4-2. Method 2 — When one allele from each data partition was chosen randomly to represent an individual in a concatenated matrix (Fig. 4-1), both parsimony and Bayesi an topologies placed alleles from Monarda fistulosa and Pycnanthemum muticum as successive sisters to a clade composed of alleles from all SE scrub mint clade taxa. A clade composed of the alleles from the two accessions of Dicerandra was sister to the remainder of th e SE scrub mint clade, which was divided into three clades. The first clade was composed of both alleles of Stachydeoma The second clade containe d all alleles of Clinopodium ashei with Clinopodium coccineum and Clinopodium dentatum as successive sisters. The other large clade was composed of a

PAGE 94

94 monophyletic Conradina with Piloblephis rigida and Clinopodium georgianum placed as unresolved sisters to Conradina Both parsimony and Bayesian searches recovered trees that exhibited a large amount of topological variation within Conradina, which collapsed into a large polytomy in consensus trees. Very few rela tionships received >50% BS or >0.95 PP. Method 3 — When any heterozygous sites were code d as an N, results were almost completely identical to the IUB code method. Because of the extremely high degree of similarity, we hereafter refer to them as the IUB code/N method (see method 4 below for results). Method 4 — When all four data sets were concat enated by combining heterozygotes into one sequence by coding heterozygous sites using an IUB code or N, we treated the ambiguity codes as both “uncertainties” and “polymorphism s” during parsimony searches in PAUP*. In comparison to treating data as “uncertainties,” treating IUB codes or N as “polymorphisms” increased the number of parsimonyinformative sites, tree lengt h, CI, and RC. However, both the “polymorphisms” and “uncertainties” op tions recovered comparable numbers of topologically identical, most parsimonious trees, with highly similar bootstrap values. Because the strict consensus trees using both methods are topologically identical, for simplicity we present only searches that treated ambiguities as “uncertainties.” When all four data partitions were con catenated using the IUB code/N method and analyzed simultaneously, results of the Bayesian an alyses were almost identical to that of the parsimony strict consensus (Fig. 4-2), and both were almost completely re solved. Relationships among genera of the SE scrub mint clade were similar to the “random selection” method (see method 2 above), with a slight increase in re solution in two locations : 1) instead of being unresolved, the two accessions of Stachydeoma graveolens were placed as sister to the clade

PAGE 95

95 composed of Clinopodium ashei, C. coccineum, and C. dentatum and 2) instead of being unresolved sisters to Conradina Piloblephis rigida and Clinopodium georgianum were placed as successive sisters to Conradina Within Conradina although very few branches received >50% BS support, most received high PP values, and rela tionships were much more resolved than with any other concatenation method. All taxa were grouped into two geographically structured clades: one clade contained only alleles from sp ecies that inhabit the Florida panhandle and northern habitats: C. glabra, C. canescens, C. verticillata and one accession from the Santa Rosa populations. The second Conradina clade, with the exception of the second accession from the Santa Rosa populations, was completely co mposed of alleles from peninsular Conradina species: C. brevifolia C. grandiflora and C etonia No species was monophyletic in either clade. Method 5.— When all four data partitions were co ncatenated using the “all combinations” method (Fig. 4-3) and analyzed simultaneously, re lationships among genera of the SE scrub mint clade were almost identical to those found in trees based on the IUB code/N method. The only difference was that instead of alleles of Clinopodium georgianum and Piloblephis rigida being placed as successive sisters to Conradina a clade of C. georgianum alleles was sister to Conradina and the P. rigida alleles collapsed into a polytomy with the two large Conradina and Clinopodium/Stachydeoma clades. Like the random method (see method 2 above), both parsimony and Bayesian searches recovered trees that exhibited a larg e amount of topological variation within Conradina, which collapsed in consensus trees. In comparison to the high level of resolution and geogra phical structuring found in Conradina in analyses in which heterozygous sites were coded using the IUB code/N met hod (see method 4 above), trees based on the “all combinations” method (Fig. 4-3) did not group Conradina alleles into two geographically

PAGE 96

96 structured clades; Conradina alleles formed a large polytomy composed of several clades, each of which generally included alleles or accessions of a single species (compare Figs. 4-2 and 4-3). Method 6.— When the number of terminals was reduced to two per each heterozygous individual by selecting the pair of terminals that contained all al leles and had the highest CI (Fig. 4-4); hereafter referred to as the CI method), relationships among genera were identical to relationships recovered based on method 5, the “all combinations” method (compare Figs. 4-3 and 4-4). Within Conradina the CI method had fewer terminals and is slightly more resolved than the “all combinations” method, but gene rally, both methods recover very similar relationships among Conradina taxa. Method 7.— The equally weighted and proportionally weighted distance matrices created by POFAD reconstructed the same general relatio nships among SE scrub mint genera, and are concordant with relationships recovered using both of the other concatenation methods. However, like the “all combinations” method, th e POFAD method recovere d variable topologies within Conradina species (Fig. 4-5 a, b). Based on th e equally weighted matrix (Fig. 4-5a), clusters generally corresponded to the panhandle/ peninsular geographic splits found in the IUB code/N topologies. When the matrices were proportionally weighted (Fig. 4-5b), geographic structure was not apparent; clusters cont ained both panhandle a nd peninsular taxa. Discussion Within genera of the SE scrub mint clade, trees based on the six concatenation methods differed in the placement of either Clinopodium georgianum or Piloblephis rigida as the sister taxon to Conradina, and the “random selection” method (F ig. 4-1) yielded trees that were slightly less resolved than those from other methods However, trees based on the six concatenation methods were generally very si milar in their resolution and support along the backbone of the trees (i.e., relationships among ge nera), suggesting that concatenation approach

PAGE 97

97 has little influence on the topology in the backbone of the trees. Within Conradina however, the six methods produced variable t opologies; the IUB code/N met hod topology (Fig. 4-2) is well resolved and strongly supported, the trees resul ting from the “all combinations” (Fig. 4-3), “random method” (Fig. 4-1) and “CI method” (Fig. 4-2) are mark edly less resolved than trees resulting from the IUB code/N approach, and the POFAD method resulted in only two trees (Figs. 4-5 A B ), but relationships among species of Conradina differ between the two. The variation in the topologies resulting from the concatenat ion methods may be the result of differing treatment of conf lict and heterozygosity in the Conradina sequence data. If the data sets conflict because of hybrid ization or incomplete lineage sorting, both of which may be the present in the case of Conradina, the “all combinations,” random selection, and CI methods may support multiple differing placements of a taxon due to conflicting signal across the length of a sequence, causing branches to collapse in cons ensus trees. However, although they may all portray conflict in data sets in a similar manner, these three methods do not represent heterozygosity equally. The “all combinations” (Fig. 4-3) and CI (Fig. 4-4) methods allow an individual to occupy multiple terminals, which al lows for alternative placements of an individual due to heterozygosity, while the random sel ection method (Fig. 4-1) completely ignores heterozygosity and cannot show alternative placeme nts of an individual in a single tree. If alternative placements of individu als are of interest in a study, we do not recommend the random selection method. Of the two methods that do allo w for alternative placements of an individual, both the “all combinations” and CI methods r ecover similar topologies, suggesting an equal accommodation of heterozygosity. Thus, they appear to be equivalent approaches to concatenating sequences when he terozygosity and conflict among da ta partitions are present.

PAGE 98

98 Conversely, the IUB code/N approaches ignore both heterozygosity a nd conflict. Using this method, polymorphic sites are ignored during the search for the optimal trees. Next, the most parsimonious of the states contained in the ambiguity code is assigned to a branch after analysis. Thus, even if the two alleles of an individual should be placed in differing positions in a tree due to many differences between its two alleles (which may also cause conflict across a concatenated matrix), only the most parsimonious of the sites will be optimized onto tree. This reduces conflict and results in a more strongly supported and wellresolved tree, which may not be an accurate portrayal of relationships. Because it is assumed that all alleles represented in a terminal are monophyletic, if the two alleles of a heterozygote are not monophyletic because of processes such as hybridization or incomplete sorting, the IU B code/N method violates our assumptions of phylogenetic analysis. However, if the conflict resulting from heterozygosity is the result of noise in phylogenetic signal, then the IUB code will eliminate the noise and may be the best method to concatenate heterozygous nucl ear genes; the problem is determining whether noise, hybridization, or incomplete lineage sorting is the cause of the heterozygosity in the data set. Using the POFAD method, the equally weighted (Fig. 4-5 A ) and proportionally weighted matrices (Fig. 4-5 B ) reconstructed different relationships among Conradina taxa. Interestingly, analyses of the equally weighted matrix recovere d a cluster that was almost identical to the IUB code/N method consensus tree. This is unexpected because the cluster based on the proportionally weighted PO FAD matrix is different than th e IUB code/N method trees, even though both are proportionally weight ed. Regardless, because the POFAD matrix is represented by a reticulate network, this may be a more accura te representation of ambiguity or reticulation in the data sets than the bifurcating trees th at result from the IUB code/N method. However,

PAGE 99

99 because disparate alleles of an individual, which may be placed in different positions in a gene tree, are forced to occupy a single terminal, this method still does not portray the alternative placements of two alleles of an individual in a cluster, as would the “all combinations” and CI methods. Because allelic information is taken into account in the POFAD method, the individual will be placed in an intermediate position; however, this does not provide insight into the relative placements of the two alleles. In parts of the data matrix that do not have c onflicting signal, the six methods appear to be equally good approaches to concatenating multipl e heterozygous data part itions. Thus, if all methods recover similar topologies for all taxa in a data matrix, then any of the concatenation methods presented here may be a valid approach. However, as in our study, if parts of the tree have conflicting signal among data sets due to incomplete lineage sor ting, hybridization, or noise, then the six methods will differ in their topo logies due to differences in their treatment of heterozygosity. If the topologies resulting from th e concatenation methods differ, then it may be possible to compare the amount of resolution and placement of taxa in these differing topologies to identify areas of conflict in the data set. After using se veral concatenation methods, if relationships among certain taxa are less resolv ed when using the “all combinations” or CI methods in comparison to the IUB code/N method, th is may indicate that multiple data partitions conflict for those taxa. If this is the case, then the “all combinations” or CI method may provide a more valid portrayal of the uncertainty of th e placement of those taxa resulting from allelic variation and conflicting signal. However, the IUB code method may be the best in eliminating noise in phylogenetic signal. Although comparing the results of concatenation methods can pr ovide an insight into their utility, it is difficult to make any absolute conclu sions because the true evolutionary tree of the

PAGE 100

100 SE scrub mint clade is unknown, and we are unabl e to compare the results of the concatenation methods with a known tree. One of the most interesting prospects for further research on combined analysis of heterozygous loci woul d be a simulation study to explore how varying methods of concatenating heterozygotes woul d affect phylogenetic accuracy of results under bifurcating evolution, afte r hybridization, or after incomplete sorting. Simulations would enable us to compare the results of an alyses based on the concatenation approaches against a true tree and would provide valuable information on how concatenation me thods perform with heterozygosity caused by different processes, so that we may understand whether any of these methods can recover accurate phylo genetic relationships when heterozygosity is present. Furthermore, a simulation study could test our hypotheses of whether disagreement among the concatenation methods results from conflicti ng signal among data sets, and whether the IUB code/N method produces a misleading, well-resolved tree because it ignores c onflict in the data set, or conversely, if it is act ually an accurate representation of phylogenetic relationships.

PAGE 101

101Table 4-1. Characteristics of indi vidual DNA regions and analyses Number of terminals Percent heterozygosity Length Percent parsimony informative Characters Number of mostparsimonious trees Tree length CI RI RC Locus1 45 34.3% 472 13.7% 441 171 0.848 0.884 0.750 Locus2 59 68.5% 512 18.3% 92,800 176 0.795 0.877 0.697 ITS (from Edwards et al. 2006) 35 0% 624 8.8% 99,400 164 0.841 0.864 0.727 Plastid (from Edwards et al. 2006) 35 0% 2034 1.4% 96,300 111 0.946 0.921 0.821 ITS+Plastid combined 35 0% 2658 2001 284 0.856 0.846 0.724

PAGE 102

102Table 4-2. Results of combined analyses of the GapC-1 GapC-2 ITS and plastid data partitions using the concatenation methods. The POFAD method is not shown here because it was not analyzed using parsimony. Number of terminals Length Percent parsimony informative characters Number of mostparsimonious trees Tree length CI RI RC Method 2—random selection method 35 3187 4.9% 34 653 0.761 0.7080.539 Methods 3 and 4—IUB code/N method 35 3642 4.4% 46 589 0.793 0.7510.595 Method 5—all combinations method 80 3642 11.4% 220 770 0.684 0.8800.602 Method 6—CI method 62 3642 12.4% 1856 742 0.710 0.8380.595

PAGE 103

103 Figure 4-1. Phylogram of one of the most parsimonious trees resu lting from combined analysis of GapC 1 GapC 2 ITS and plastid data sets, with heterozygotes concatenated by randomly selecting a single allele to repr esent an individual in each data set. Parsimony bootstrap values >50%/Bayesia n posterior probabili ties are indicated above the branches, but posterior probabil ity values below 0.90 are not included due to variation among Bayesian runs in branch es that have lower than 0.90 PP. Arrows indicate branches that collapse in the strict consensus and numbers following taxon names indicate the accession number of the population and the number of the individual in the population.

PAGE 104

104 Figure 4-2. Phylogram of one of the most parsimonious trees resu lting from combined analysis of GapC 1 GapC 2 ITS and plastid data sets, with heterozygotes concatenated using the IUB/N method. Parsimony bootstrap values >50%/Bayesian posterior probabilities are indicated above the branches, arrows indicate branches that collapse in the strict consensus, and numbers following taxon names indicate the accession number of the population and the number of the individual in the population.

PAGE 105

105 Figure 4-3. Phylogram of one of 220 most parsimonious trees resu lting from combined analysis of GapC 1 GapC 2 ITS, and plastid data sets, with heterozygotes concatenated in all possible combinations (maximum of four per individual). Each unique combination of alleles of heterozygotes ar e denoted as A, B, C, or D. Parsimony bootstrap values >90%/Bayesian posterior probabilities are indicated above the br anches. Posterior probability values below 0.90 are not include d due to variation among Bayesian runs in branches that have lower than 0.90 PP. A rrows indicate branches that collapse in the strict consensus and numbers foll owing taxon names i ndicate the accession number of the population and the number of the individual in the population.

PAGE 106

106 Figure 4-4. Phylogram of one of the most parsimonious trees resu lting from combined analysis of GapC 1 GapC 2 ITS, and plastid data sets, w ith heterozygotes concatenated using the CI method. Each unique combin ation of alleles of heterozygotes are denoted as A or B. Parsimony bootst rap values >90%/Bayesian posterior probabilities are indicated above the branch es. Posterior probability values below 0.90 are not included due to variation among Ba yesian runs for branches that have lower than 0.90 PP. Arrows indicate branches that collapse in the strict consensus and numbers following taxon names indicate the accession number of the population and the number of the indivi dual in the population.

PAGE 107

107 Figure 4-5. NeighborNet trees resu lting from analyses of the PO FAD data sets. (a) Diagram resulting from the matrix in which the four data sets were equally weighted. (b) Diagram resulting from the matrix in whic h the four data sets were weighted proportionally according to number of most parsimonious sites.

PAGE 108

108 CHAPTER 5 ISOLATION, CHARACTERIZATION, AND CROSS-SPECIES AMPLIFICATIONS OF MICROSATELLITE LOCI FROM CONRADINA (LAMIACEAE) Introduction ‡Conradina comprising six described species, is a morphologically homogeneous group of shrubby, aromatic mints (Lamiaceae) endemic to the southeastern U.S. Several Conradina species are federally endangered or threatened (including C. brevifolia, C. etonia C. glabra and C. verticillata ), but species limits are unclear and th e taxonomic status of several protected species is questionable (Edwards et al 2006). Although phylogenetic analyses of Conradina were not sufficiently resolved to clarify speci es relationships, all an alyses revealed the nonmonophyly of most Conradina species; because each species is distinguishable morphologically, incomplete lineage sorting or hybridization was suggested as possible causes of these patterns (Edwards et al. 2006; Edwards et al. in press). The purpose of this study was to develop microsatellite loci for the genus Conradina to: 1) investigate species limits, 2) investigate the possibility of interspecific hybr idization, 3) investigate the population structure of each Conradina species, and 4) use these da ta to facilitate the manageme nt of critica lly endangered Conradina species. Materials and Methods, Results, and Conclusions Genomic DNA was extracted from one individual each of Conradina brevifolia C. glabra and C. canescens using a modified CTAB DNA extr action protocol (Doyle and Doyle 1987). We used a microsatelli te enrichment procedure base d on (Kandpal et al. 1994). DNA ‡ This article was reproduced with permission from Molecular Ecology Notes, Blackwell Publishing, www.blackwellpublishing.com/men Edwards CE, DE Soltis, PS Soltis 2007b Isolation, characteriza tion, and cross-species amplifications of microsatellite loci in Conradina Mol Ecol Notes in press.

PAGE 109

109 was digested with Sau 3AI, purified/fractionated using Chroma Spin columns (Clontech Laboratories) to remove fragments < 400 base pairs (bp), and ligated with Sau 3AI linkers; fragments were amplified by polymerase chain reaction (PCR). DNA was enriched for (CA/GT) repeats by hybridization to a (CA)15 biotinylated nucleic acid pr obe. Probe-target fragments were isolated using VECTREX Avidin D (Vector Laboratories), amplified by PCR, ligated into a TOPO TA pCR 4.0 vector, transformed into One Shot E. coli competent cells (Invitrogen), and grown overnight on Luria Broth (LB)/Kanamycin /agar plates. Colonies containing the repeat were detected using the Quick-Light hybridizatio n kit (FMC BioProducts), amplified using M13 primers, and sequenced in both directions on a CEQ 2000 or CEQ 8000 Automated Sequencer (Beckman-Coulter) using M13 primers and CEQ Dy e Terminator Cycle Sequencing Quick Start chemistry (Beckman-Coulter). Sequences were assembled using Sequencher 4.2 (Gene Codes Corp.). Primers were designed manually or using PRIMER3 (Rozen & Skaletsky 2000). An M13 tail (CACGACGTTGTAAAAC) was added to the 5 end of each forward primer for amplification using a universal dye-labeling method (Boutin-Gan ache et al. 2001). Of 59 colonies sequenced, 23 contained microsatellites, and primers were desi gned for 20 loci. To ensure amplification and variability in all Conradina and related species, we screened th ese loci in two individuals from every species of Conradina and one individual each of Clinopodium ashei, Piloblephis rigida and Clinopodium georgianum all part of the SE scrub mint clade, with Conradina (Edwards et al. 2006) Prior to genotyping, DNA was cleaned using GENECLEAN Turbo Kits (Qbiogene). PCR was carried out in 10 L reactions c ontaining 0.5 unit of GoTaq Flexi DNA polymerase (Promega), 1X Promega Colorless GoTaq Flexi Buffer, 2.0 mM MgCl2, 0.45 M each of the reverse primer and 6FAM labe led M13 primer (Applied Biosystems), 0.012 M of the extended

PAGE 110

110 forward primer, and 50 M of each dNTP usi ng temperature cycling conditions following Edwards et al. (Edwards et al. 2007a). Labeled PCR products were diluted 1:20 and run on an ABI 3730xl DNA Analyzer (Applied Bi osystems) at the ICBR facility, University of Florida, by loading 1 l of the diluted PCR product, 9.9 l of formamide, and 0.1 l LIZ 600 size standard (Applied Biosystems). Fragments were analyzed using GENEMARKER 1.5 (Soft Genetics LLC) and scored manually. Of 20 primer pairs, five demonstrated poor amplification, two were monomorphic in all taxa, and two demonstrated nonspecific am plification; these were not pursued further. The remaining 11 loci amp lified consistently and varied in most Conradina species and closely relate d taxa (Table 5-1). These 11 loci were characterized in 20 individuals from one population of C. brevifolia (Table 5-2) PCR conditions were as specified above, ex cept that four different fluorescent dyes were used to label the M13 PCR primer to combin e up to four loci of an individual in a single run on the ABI 3730xl DNA Analyzer. The loci that were run together, the fluorescent dye used to label each locus, and the volume of each PCR r eaction added (with other loci) to 20 l of H2O, are detailed in Table 5-2; the pooled, diluted P CR products were loaded as specified above. Data were analyzed to assess linkage disequilibrium among loci and deviations from Hardy-Weinberg equilibrium using Fisher’s exact tests in FSTAT version 2.9.3.2 (Goudet 20 02); significance was assessed using permutation procedures. Descriptiv e statistics (Table 5-2) were measured using GENEPOP version 3.4 (Raymond and Rousset 1995). After Bonferroni correction for multiple comparisons, no linkage disequilibrium or signi ficant deviations from Hardy-Weinberg equilibrium were detected. We plan to use these markers to assess species boundaries, detect interspecific hybridization, and unde rstand population structure in each Conradina species.

PAGE 111

111Table 5-1. Results of initial primer screening in all six species of Conradina (plus the Santa Rosa populations), Clinopodium ashei, Clinopodium georganium and Piloblephis rigida For each species/locus combination, th e number of individuals in which amplification was observed, the number of individuals for which amplification wa s attempted (in parentheses), the number of observed alleles, and size range of the obser ved alleles (following a colon) is given. Locus C. brevifolia C. canescens C. etonia C. glabra C. grandiflora Santa Rosa populations C. verticillata Clinopodium ashei Clinopodium georgianum Piloblephi s rigida CSRC1 2 (2) 2:159-161 2 (2) 3:143-169 2 (2) 4:153-175 2 (2) 3:145-157 2 (2) 2:153-157 2 (2) 3:141-155 2 (2) 3:144-156 1 (1) 2:157-165 1 (1) 2:145-147 1 (1) 2:149-155 CSRD 2 (2) 3:186-200 2 (2) 4:186-204 1 (2) 1:194 0 (2) no amp 1 (2) 1:186 2 (2) 4:186-204 2 (2) 1:202 1 (1) 2:200-202 1 (1) 1:186 1(1) 1:224 CSRE 2 (2) 2:274-282 2 (2) 3:254-260 1 (2) 1:292 2 (2) 2:258-280 2 (2) 3: 278-284 2 (2) 2:260-262 2 (2) 2:278-288 1 (1) 1:282 1 (1) 2:268-278 1(1) 1:278 CSRF 2 (2) 4:162-202 2 (2) 3:184-194 0 (2) no amp 1 (2) 2:188-190 2 (2) 3:164-174 2(2) 2:190-192 2 (2) 3:188-192 1 (1) 1:160 1 (1) 1:188 1 (1) 2:156-164 CSR 304 2 (2) 3:166-180 2 (2) 180-186 1 (2) 2:190 2 (2) 3:164-190 2 (2) 2:166-168 2 (2) 3:188-204 2 (2) 3:188-198 1 (1) 2:166-200 1 (1) 2:170-178 0 (1) no amp CSR 507 2 (2) 2:179-185 2 (2) 2:179-189 1 (2) 1:179 2 (2) 2:183-191 2 (2) 2:187-203 2 (2) 3:185-201 2 (2) 1:181 1 (1) 1:169 0 (1) no amp 0 (1) no amp BR-E 2 (2) 3:323-347 1 (2) 1:343 1 (2) 1:333 0 (2) no amp 2 (2) 4:321-329 1 (2) 1:327 2 (2) 2:324-332 1 (1) 1:341 1 (1) 2:323-327 1 (1) 2:319-322 B50-2 2 (2) 3:204-238 2 (2) 1:212 1 (2) 2:208-214 0 (2) no amp 2 (2) 3:210-218 0 (2) no amp 2 (2) 1:210 1 (1) 2:210-218 1 (1) 2:204-206 0 (1) no amp GLA-G 2 (2) 3:254-262 2 (2) 2:254-256 1 (2) 2:258-260 0 (2) no amp 1 (2) 2:268-272 2 (2) 3:248-258 1 (2) 2:250-254 1 (1) 2:254-264 1 (1) 2:244-246 1(1) 258-260 GLAH 2 (2) 3:264-288 2 (2) 274-290 1 (2) 1:268 1 (2) 2:256-284 1 (2) 2:264-266 2 (2) 3:272-296 2 (2) 1:266 1 (1) 2:254-270 1 (1) 2:258-272 1 (1) 2:260-270 GLA 304 2 (2) 2:269-273 2 (2) 3:237-263 1 (2) 1:275 0 (2) no amp 1 (2) 2:269-271 2 (2) 3:237-263 1 (2) 2:261-271 1 (1) 2:269-273 0 (1) no amp 1(1) 2:273-277

PAGE 112

112Table 5-2. Characteristics of 11 mi crosatellite loci developed for Conradina including locus name (names beginning with CSR=isolated from C. canescens BR=isolated from C. brevifolia GLA=isolated from C. glabra ), primer sequences (5 3 ), GenBank accession number, repeat motif, th e conditions used to pool the loci, incl uding the loci that were run together, the fluorescent dye used during PCR to label the M13 primer, and the volume of each PCR (in l) that was pooled in 20 l of H20 and loaded (with formamide and si ze standard) into the sequencer. Primers were used to genotype a sample of 20 individuals from one population of Conrad ina brevifolia (all individuals were succ essfully genotyped at all loci), and we report a, the number of alleles found in all individuals; the size range of the obs erved alleles; HE, expected heterozygosity; and HO, observed heterozygosity. Conditions used to pool multiple loci Characterization in a population of Conradina brevifolia (n=20) Locus name Primer sequence (5 3 ) GenBank accession number Repeat motif in clone run Fluorescent dye Volume (l) a Size (bp) HE HO CRC1 F:*tgaacgcacttcgcctagta R:tcatgctggcaaattagctg EF635393 (tg)24 1 VIC 1 8 155-171 0.795 0.70 CSRD F:*accagctcaagcaagcctat R:gcttgggatcacgtgctatt EF635394 (ca)20 2 PET 3 13186-226 0.876 0.80 CSRE F:*cactgcatcgaccatcaatc R:accactcgaacattggttcc EF635395 (gt)15 2 NED 2 7 258-284 0.626 0.70 CSRF F:*cactttttcccggtacctca R:ccccatagagaattggcaaG EF635396 (tg)17 2 VIC 1 7 159-201 0.809 0.80 CSR30-4 F:*gcttggtcgttgctaccttc R: gctgtttcaggcgcaatataa EF635397 (gt)15(ga)15 3 NED 2 7 166-214 0.756 0.70 CSR50-7 F:*ttctgataaacctgtctcaactt R:aacctgaaaataaatcatacggg EF635398 (tg)14 2 6FAM 1 5 177-201 0.496 0.30 BR-E F:*cgcgtcggactgaattatag R:cttcgccttggtaaagttgg EF635399 (ca)25 1 NED 3 13321-347 0.845 0.85 BR50-2 F:*cgatggccatttttgttctt R:tgcctaacccattgccttta EF635400 (gt)11 3 VIC 1 11205-237 0.891 0.90 GLA-G F:*ggtgaaaagaggctgagctg R:ggaataaggattgaaggggaag EF635401 (gt)8 3 PET 3 11251-267 0.813 0.75 GLA-H F:*accagtgtatcaacatctgttcg R:aagcccatagaatgaagagca EF635402 (ca)3(tatg)2(tg)4t(t g)5(ga)7a(ag)4 1 6FAM 1 6 256-268 0.764 0.75 GLA30-4 F:*aagctcaataattaacattggcact R:ggaaagcatcagatgtagccagag EF635403 (ca)12(ta)3(ca) 2 3 6FAM 3 9 259-281 0.810 0.55 *M13 tag (cacgacgttgtaaaacgac) added to 5 end of primer for amplifica tion with fluorescently labeled M13

PAGE 113

113 CHAPTER 6 PATTERNS OF GENETIC STRUCTURE BASE D ON MICROSATELLITE LOCI REVEAL SPECIES COHESION AND SHARED ANCEST RAL POLYMORPHISM RATHER THAN HYBRIDIZATION IN CONRADINA Introduction Species-level phylogenetic studies that include multiple individuals or populations of each sampled species commonly and unexpectedly support the non-monophyly of species. Such results are unexpected when morphology suggest s otherwise, and possible causes are shared ancestral polymorphism across species boundaries (e.g. incomplete lineage sorting), past or present interspecific hybridizati on, or both. It is often difficu lt or impossible to distinguish whether hybridization or incomp lete lineage sorting is respon sible for the apparent nonmonophyly of species using phylogenetic data al one, simply because these processes leave identical phylogenetic signatures (Wendel and Doyle 1998). However, it is important to determine which of these phenomena is res ponsible for the non-monophyly of species for practical reasons; because ongoing gene flow is an important component of many species concepts, discovering if interspecific hybridizatio n has occurred may result in changes to species classifications. The best method to determine whic h of the two processes ha s occurred is the use of more rapidly evolving data and more exte nsive taxon sampling to i nvestigate the genetic structure of the clade. One group in which hybridization and incomplete lineage sorting have been suggested is Conradina Conradina comprising six described species is a group of morphologically distinctive, narrow-leaved, aromatic shr ubs in the mint family (Gray 1965). Conradina species are endemic to the southeastern United States, an d four of the six descri bed species are federally endangered or threatened ( Conradina brevifolia, C. glabra, C. etonia and C. verticillata ). The species occupy an allopatric distri bution pattern centered primarily in Florida (Fig. 6-1). Despite

PAGE 114

114 the fact that each Conradina species occupies a unique ge ographical region and most are distinguished by distinctive morphological char acters (Gray 1965), phylogenetic analyses of Conradina indicate that most species are non-monophylet ic (Edwards et al. 2006; Edwards et al. in press). Using DNA sequence data from four data partitions (plas tid DNA, ITS, and two members of the GapC gene family), separate phylogenetic analyses demonstrated low levels of sequence divergence and were not sufficiently re solved to determine species relationships. However, in cases in which relationships within Conradina were resolved, most genes revealed the non-monophyly of most Conradina species (Edwards et al. 2006; Edwards et al. in press). Combined phylogenetic analyses of the four data partitions in Conradina including the two heterozygous GapC loci, incorporated two methods for concatenating the heterozygous nuclear genes with the single-copy plastid and in variant ITS genes. Tr ees resulting from both concatenation methods supported the monophyly of Conradina but resolution of relationships within Conradina varied depending on concatenation met hod (Edwards et al. in press); one method did not resolve relationships in Conradina while the other method recovered two clades corresponding to geographic locations (peninsu lar species vs. Florida panhandle/northern species). Although this suggests a geogra phic structure to relationships within Conradina neither of the combined analys es supported the monophyly of any Conradina species (Edwards et al. in press). Because the species are morphologically distinguishable, incomplete lineage sorting, rapid speciation rates, and hybridization were suggested as possible causes of the nonmonophyly of species (Edwards et al. 2006; Edwards et al. in press). Distinguishing between hybr idization and incomplete lineage sorting in Conradina is important because of the endangered status of most Conradina species. If widespread gene flow is occurring among allopatric populations of multiple Conradina species, then the described

PAGE 115

115 species may simply represent disjunct subpopulations of a large, widespread, panmictic species. If this is the case, species status may not be appropriate, and the fede ral listing and recovery programs for four Conradina species may not be necessary. Conversely, if hybridization has not recently occurred among the Conradina species, and shared ancestr al polymorphism is more likely to be responsible for the patterns of non-monophyly revealed in phylogenetic analyses, then the current species classifi cation and listing may be valid. B ecause valuable time, effort and resources are used to conserve the endangered Conradina species, it is imperative that we clarify whether hybridization, incomplete lineage sorting, or both have occurred in this group. Another motivation for understanding patterns of geneti c structure in Conradina is to clarify the status of problem atic species. Although most Conradina species are morphologically distinguishable, two debates about species status continue based on morphological grounds: C. brevifolia is listed as federally e ndangered yet taxonomically questionable because it is morphologically similar to the relati vely widespread, disjunct species, C. canescens (USFWS 1996). Likewise, populations of Conradina found in Santa Rosa County, FL, have been problematic because they have morphological char acteristics in common with both the federally endangered C. glabra and the widespread C. canescens The Santa Rosa populations and C. canescens have peripatric geographic ranges, and Edwards et al. ( 2006) found that they may be more closely related to C. canescens ; the only morphological characteristics that these populations share with C. glabra is the lack of pubescence. Ho wever, it is important to confirm the identity of C. brevifolia and the Santa Rosa populations in order to clarify whether they merit federal protection. To clarify species debates and determine if hybridizat ion and/or incomplete lineage sorting have occurred in Conradina, more variable genetic data are ne cessary to reveal the patterns of

PAGE 116

116 genetic structure in Conradina To this end, we used 10 polym orphic microsatellite loci to genotype 1098 individuals fr om 56 populations of all six described species of Conradina, including populations of the questionable species, to investigat e the patterns of genetic structure in Conradina Our first goal was to test whethe r patterns of genetic variation in Conradina correspond to species boundaries to determine whet her incomplete lineage sorting is responsible for the non-monophyly of species rev ealed in phylogenetic analyses, or instead, if gene flow is occurring among the described Conradina species. Next, we evalua ted whether current species boundaries adequately represent patterns of variation in Conradina and evaluated the species status of C. brevifolia and the Santa Rosa populations. Further goals were to investigate the partitioning of genetic variation within and among populations of each Conradina species to understand how rarity influences levels of genetic diversity in Conradina to clarify patterns of genetic di versity and migration within and among populations of each species, and to evaluate the genetic effects of previous reintroduction efforts in endangered species. We evaluated th e levels of genetic diversity in each Conradina species to determine whether the rarest species of Conradina ( C. glabra and C. etonia) have reduced levels of genetic diversity in comparison with more wide spread species because of genetic drift or other factors commonly associated with rarity. We then estimated patterns of ge netic structure within and among populations in each species to eval uate whether population fragmentation is influencing gene flow in endangered populations of Conradina Furthermore, prior to public acquisition of the lands that cont ained all natural populations of C glabra cuttings from one population were experimentally translocated to a neighboring park; we evaluated how this translocation has influenced the genetic structure of C. glabra We then made recommendations

PAGE 117

117 about how manage populations of endangered species of Conradina to conserve the highest possible levels of genetic diversity Materials and Methods Population sampling, DNA Extraction, and Genotyping Populations were sampled throughout the geographic range of each Conradina species. We used species-specific mor phological attributes and the al lopatric geogra phic ranges of Conradina species (Gray 1965; Kral a nd McCartney 1991) to classify populations into species. We sampled 7-10 populations from each of the six described species of Conradina, plus four populations of the taxonomically questionable Santa Rosa popul ations; sampling locations are presented in Fig. 6-1. In each population, one vouch er specimen was collected and deposited in FLAS; collection information is presented in Table 6-1. In ea ch population, we collected leaf material from multiple individuals. DNA was extracted for up to 24 individuals in each population using a modified CTAB protocol (Doyle and Doyle 1987) and cleaned with GENECLEAN Turbo Kits (Qbiogene). Each individual was genotyped at 10 microsatellite loci using the PCR primers, M13-tailed PCR reaction conditions, and electrophoresis conditions previously developed by Edwards et al. (2007b). In PCR reactions, four different fluore scent dyes were used to label the M13 PCR primer to combine up to four loci in a single run; the four loci were pooled, diluted 1:20, and run on an ABI 3730xl DNA Analyzer (Applied Biosyste ms) at the ICBR facility, University of Florida, by loading 1 l of the diluted PCR prod uct, 9.9 l of formamide, and 0.1 l LIZ 600 size standard (Applied Biosystems). The loci that were run together, the fluorescent dye used to label each locus, and the volume of each PCR reactio n added (with other loci) to 20 l of H2O, are detailed in Edwards et al. (2007b). Fragment analysis and scoring was carried out using automated fragment scoring panels developed fo r each locus in Genemarker version 1.6 (Soft

PAGE 118

118 Genetics LLC). Failed PCR reactions were repeated once; non-successful amplifications were treated as missing data. Any indivi duals that did not amplify at f our or more loci were excluded from the analysis. Statistical Analysis of Microsatellite Data After scoring the data, 8 of the 10 loci e xhibited 1-bp size differe nces between alleles instead of the 2-bp differences expected from loci with dinucleotide repeats. These 1-bp differences are often due to indels in the fl anking regions or the repeat region of the microsatellite. Although RST has been proposed as a more accurate model for analyses of microsatellites than FST (Slatkin 1995), we did not use any analyses based on a stepwise mutation model as it does not appear that these microsatellite alleles in Conradina are evolving in a stepwise manner. We instead conducted analyses based on FST (Wright 1965), which uses allele identity rather than allele size, as this appear s to be a more appropriate model given our data. The weakness of using FST for microsatellite data is that if our microsatellite loci are evolving in a stepwise manner, FST may underestimate population s ubdivision (Slatkin 1995). For each species, we tested for linkage disequ ilibrium (LD) between pairs of loci in each population and for deviations from Hardy-Weinbe rg equilibrium (HWE) at each locus in each population using Fisher’s exact te sts in GenePop version 4.0 (Rou sset 2007). Diversity indices and general statistics were estimated using MSA (Dieringer and Schlotte rer 2003). To investigate population subdivision, we used MSA to calculate pairwise FST between all pairs of 56 populations of Conradina included in this study, with 100,000 permutations to assess significance. We used Bonferroni corrections for all tests involving multiple comparisons. For a graphical representation of popula tion structure, we computed a pa irwise genetic distance matrix for all 56 populations using Nei’s standard gene tic distance (Nei 1972) in MSA and used PAUP*

PAGE 119

119 (Swofford 2002) to construct an unrooted Nei ghbor Joining (NJ) phyl ogram of the genetic distances. To understand the hierarchical partitioning of genetic structure within and among species of Conradina we carried out an analysis of molecu lar variance, or AMOVA (Excoffier et al. 1992), in Arlequin ver. 3.11 (Excoffier et al. 200 5). We grouped indivi duals into populations based on sampling location, and grouped populations into species according to a priori species identifications based on morphology and geography. We carried out a locus-by-locus AMOVA, investigating the amount of variance attributab le to the “among speci es,” “among populations within species,” and the “among indi viduals within populatio ns” levels using all data from all six species, with the Santa Rosa Populations included in Conradina canescens Significance for AMOVA analyses was ascertained using 10,000 permutations. To investigate gene flow and genetic structure in Conradina without a priori grouping of individuals into species or popul ations, we used the Bayesian program STRUCTURE (Pritchard et al. 2000) to estimate the number of groups present in the data set ( K ) and assign admixture proportions of each individual to th ese groupings. We analyzed patt erns of genetic structure in the entire data set of 1098 i ndividuals representing all Conradina species, allowing K to vary from 1 to 7, equal to the numb er of described species in Conradina plus the Santa Rosa populations. We used an admixture model, assu med correlated allele frequencies, and carried out five independent replicates at each K. After preliminary anal yses to determine the adequate burn-in and number of generations we used a burn-in of 100,000 ge nerations and a run length of 500,000 generations. To understand how the patterns of genetic structure change as the number of groupings increases, we examined all groupings from K =2 to K =8. Distruct ver. 1.1 (Rosenberg 2004) was used to create figures of the STRUCTURE results.

PAGE 120

120 We next carried out STRUCTURE analyses to investigate the pa tterns of genetic structure within each of the cluste rs revealed in the initial structure an alysis. Analyses of the smaller data sets were carried out usi ng the same settings as used in the initial analysis of all individuals. However, because STRUCTURE finds the uppermost hi erarchical levels of structure in the data instead of the true number of clusters, we us ed a statistic based on the change in the log probabilities of the data in successive K levels ( K ) to find the true K (Evanno et al. 2005); the weakness of this method is that it is only able to evaluate K =2 and higher (Evanno et al. 2005). We also plotted the log proba bilities of data at each K to identify the K at which the log probability of the data began to plateau; if this K differed from that selected by the K statistic, we examined the patterns of genetic struct ure proposed at the alternative values of K We then examined pairwise FST values to obtain another estimate of the relationships between populations in each species (see above). Fina lly, we used AMOVA analyses to understand the partitioning of genetic variation within each species at the “among populations” and “among individuals within populations” levels, using our a priori population groupings based on collection location and the same settings as in the initial AMOVA analyses of all populations. Results Summary Statistics Expected heterozygosity and number of allele s at each locus varied among populations and species of Conradina ; expected heterozygosity at each lo cus ranged from 0-0.92, and alleles at each locus in a species ranged from 0-30 (Table 6-2). Conradina etonia had the overall lowest number of alleles per locus a nd expected heterozygosity, while C. brevifolia and C. canescens had the highest. After Bonferr oni correction, nine instances of significant LD were detected between pairs of loci within populations. Five of these significant results were between CSRE and CSR30-4; all other results did not reveal a ny consistent patterns of linkage between loci.

PAGE 121

121 Because linkage between CSRE and CSR30-4 was only found in 5 of 56 populations, this result indicates weak linkage between the two loci, if any at all, and we consider these to be independent loci in all subsequent analyses. After Bonferroni correction for multiple comparis ons, probability tests revealed that 47 of the 560 population/locus combina tions showed significant depart ures from Hardy-Weinberg equilibrium (HWE), involving nine of the ten loci, all due to hete rozygote deficiencies. Of the population/locus combinations that showed signi ficant departures from HWE, five were in Conradina brevifolia ; 13 were in C. canescens and the Santa Rosa popul ations; three were in C. etonia ; seven were in C. grandiflora ; six were in C. glabra ; and eight were in C. verticillata. Most of the significant results were found in the CSRF, CG L30-4, and CS50-7 loci, which demonstrated 10, 9, and 10 significant departur es from HWE, respectively. Five of the departures from HWE at the CSRF locus were in C. brevifolia eight of the departures at the CS50-7 locus were in C. canescens, and at the GLAG locus, three departures were found in C. glabra All of the remaining departures appear to be distributed randomly among loci and species. Analyses of Genetic Structure among Conradina Species We calculated pairwise FST between all 56 sampled populati ons from all six described species of Conradina and the Santa Rosa populations; pairwise FST values varied widely between species and populations pairs. Al l comparisons between populations of different species yielded highly significant P -values (all tests P <0.001 after Bonferroni corrections for multiple comparisons); non-significant pairwise FST values were found only between populations of the same species. Interspecific comparisions were lower within th e panhandle/northern and peninsular geographic regions indicated in Fig. 1 (Table 6-3). Within the panhandle/northern region, the lowest pairwise FST values were between C. canescens and the Santa Rosa

PAGE 122

122 populations (0.058-0.180), and sl ightly higher pairwise FST values were found between populations of C. glabra and C. canescens (0.120-0.233) or the Sant a Rosa populations (0.1340.207; Table 6-3). Pairwise FST involving C. verticillata populations were the highest of all comparisons in the panhandle/northern region, the lowest of which involved populations of C. glabra (0.195-0.358; Table 6-4). In the peninsular region, pair wise comparisons involving populations of C. etonia were highest, while pairwise FST values were lower between C. brevifolia and C. grandiflora (0.148-0.305; Table 6-4). To obtain a graphical interpretation of simila rity between populati ons, we calculated a pairwise genetic distance matrix between all of the 56 populations of Conradina in this study and used PAUP* to create a Neighbor Joining (NJ) tree based on the ge netic distances. The resulting unrooted NJ phylogram (Fig. 6-2) grouped popula tions into six distinct clusters, each corresponding to one of the six described species of Conradina with all clusters containing only populations of a single species. The questionable Santa Rosa populations clustered with C. canescens The clusters were also gr ouped geographically according to the regions designated in Fig. 6-1; the clusters of the panhandle/northern species, i.e., C. canescens the Santa Rosa populations, C. glabra and C. verticillata, were grouped together, wh ile the clusters of the peninsular Conradina species, i.e., C. brevifolia, C. etonia and C. glabra were grouped together. AMOVA analyses were conducted to investigate hierarchical patterns of genetic structure at the “among species,” “among populations with in species,” and “among individuals within a population” level using all 1098 individuals, with species a nd populations defined using a priori species designations based on morphology and geography (with the Santa Rosa populations included in Conradina canescens ; Table 6-4). These analyses found that 22% percent of the

PAGE 123

123 variation was attributable to the “among speci es” level (P<0.0001), wh ereas only 8.5% of the variation was partitioned between populations within species (which was still highly significant, P <0.0001; Table 6-4). The largest percentage of va riation (69%) was attri butable to differences among individuals within populations ( P <0.0001). In analyses using the comput er program STRUCTURE, we an alyzed the genetic structure of the entire data set of 1098 individu als, allowing the number of groupings ( K ) to vary from 1 to 6. For each value of K we present the results of the run wi th the best probability (Fig. 6-3). When K =2, the peninsular and panhandle/northern species were grouped into two separate clusters; when K =3, Conradina verticillata was separated from the panhandle/northern group into its own group; at K =4, C. etonia was placed into its own group; at K =5, C. glabra was separated into its own group; and at K =6, C. brevifolia and C. grandiflora were divided into separate groups (Fig. 6-3). At K =7 (results not shown), the Santa Rosa populations and C. canescens did not split into mutually exclusive gr oups, but instead grouped according to the intraspecific groupings of C. canescens and the Santa Rosa populations shown in Fig. 6-4B; several populations of C. canescens from the western panhandle grouped with the Santa Rosa populations, and the remaining populations of C. canescens were grouped separately. In all cases, each individual was assign ed membership in the appropriate species cluster, and the percentage of admixture from other cl usters was very small (Fig. 6-3) Patterns of Genetic Structure within Each Conradina Species In STRUCTURE analyses of Conradina brevifolia two populations were selected as the true K using the method of Ev anno et al (2005). At K =2, the genetic structure loosely corresponds to population boundaries ; most individuals in popul ations 161, 197, and 198 were placed in one group, while most individuals in populations from the Lake Wales Ridge State Forest (168, 169, and 170) were placed into the second group (Fig. 6-4A), although some

PAGE 124

124 individuals in each population were assigned memb ership in the other group (Fig. 6-4A). Individuals in populations 158 a nd 159 were assigned in more or less equal numbers to both groupings (Fig. 6-4A). Pairwise FST analyses were low between populations of C. brevifolia (0.028-0.115; Table 6-3), with an average of 0.040. Population 197 had the highest FST values in comparison to the other populations ( FST =0.077-0.115). In AMOVA anal yses, only 6.3% of the variation was partitioned among populations in C. brevifolia while 93.7% of the variation was partitioned among individuals within populations (Table 6-4). In analyses of Conradina canescens including the Santa Rosa populations, K =2 was selected as the optimal K ; one group contained populati ons 131, 138, 141, and 142, all from Alabama and the western panhandle of Florida, and the other contained the Santa Rosa populations and populations of C. canescens from the eastern panhandl e of Florida (populations 127, 128, 129, 130, 179, 180; Fig. 6-4B). When K =3, STRUCTURE divided the Santa Rosa populations and the eastern panhandle populations of C. canescens into two separate groups (Fig. 6-4B). Pairwise FST analyses between populations of C. canescens, not including the Santa Rosa populations, ranged from 0.034-0.185, comparisons be tween pairs of Sant a Rosa populations ranged from 0.058-0.094, and comparisons between populations of C. canescens and the Santa Rosa populations ranged from 0.0760.194 (Table 6-3). Pairwise FST was lowest between pairs of populations of 131, 138, 141, and 142 and between pairs of the Santa Rosa populations. AMOVA analyses of C. canescens and the Santa Rosa populations revealed that 12.8% percent of the genetic variation was attributable to the “among popu lations” level, while 87.2% was attributable to variation among i ndividuals within populat ions (Table 6-4). To test alternative groupings of the Santa Rosa populations, we first carried out an AMOVA with C. canescens and the Santa Rosa populations assigne d to separate groups; these anal yses revealed that only 3.4%

PAGE 125

125 of the variation was partitioned among-groups (P=0.001), while 11.1% was attributable to among-population variation within each group, and 85.5% was among individuals. When we used the groupings revealed in the STRUCTUR E analysis (i.e., the Santa Rosa populations grouped with C. canescens from the eastern panhandle), 6.0% of the variation was among groups, 9.7% of the variation was among popul ations within groups, and 84.3% was among individuals. In STRUCTURE analyses of Conradina etonia the most likely true K was found to be two. Individuals in the Dunn’ s Creek populations were placed in one population, while individuals in the remaining populations were as signed to the second cluster, with almost no admixture (Fig 6-4 C ). Pairwise FST values within C. etonia ranged from -0.016-0.551 (Table 63); the highest values were found in comparis ons between the Dunn’s Creek populations and the remainder of C. etonia ( FST =0.320-0.550). The pairwise FST between the two Dunn’s Creek populations was extremely small ( FST =-0.016), and it was larger between populations in the remainder of the species ( FST=0.031-0.292). In AMOVA analys es, when populations from Dunn’s Creek and the remaining populations we re treated as a single group, 23.9% of the variation was partitioned among populations and 76.1% was partitioned among individuals within populations. When the Dunn’s Creek and Etonia Creek populations were considered as two separate groups, 31.3% of the variation was found betwee n the two groups, 6.8% of the variation was partitioned among populations within groups, a nd 61.2% of the variation was partitioned among individuals within populations. In STRUCTURE analyses of Conradina glabra two was again selected as the true K Most individuals in population 109 and the three translocated populations, 143, 144, and 145, were assigned to the first group (Fig. 6-4 D ). Most individuals in populations 163, 165, 166, and

PAGE 126

126 167 were assigned to the second gr oup, and individuals in populat ion 164 were assigned nearly equally to both groups (Fig. 6-4 D ). Pairwise FST was generally low between C. glabra populations averaging 0.065 (0.020-0.159; Table 6-3). P opulation 167 had the highest pairwise FST values, especially in comparisons involving populations 109, 143, 144, 145, and 163 ( FST =0.120-0.159). In AMOVA analyses of C. glabra 6.1% of the variation was partitioned among populations, and 93.9% was partitioned am ong individuals within populations. In Conradina grandiflora the true K was estimated to be two; the northern populations were placed in one group, the s outhern populations were placed in the other, and populations in between were distributed between the two clusters (Fig. 6-4 E ). In AMOVA analyses, 10.7% of the genetic variation was partitioned among populations, and 89.2% was partitioned among individuals. Pairwise FST between populations of C. grandiflora ranged from 0.048-0.224 (Table 6-3). Comparisons involving the two norther nmost populations (193 and 215) and all other populations yielded th e highest pairwise FST values (0.10-0.22). In Conradina verticillata the best value of K was selected to be K =2; populations 184, 186, 190, 191, and 192 were grouped into one cluste r, populations 187, 188, and 189 were placed into the second cluster, and individuals from population 181 were distri buted equally between the two clusters (Fig. 6-4 E ). Because the plot of the l og probability of the data for C. verticillata began to plateau at K =3, we examined these groupings; populations 181, 184, and 186 were placed into one cluster, 187, 188, and 189 into another cluster, and 190, 191, and 192 into the third cluster (Fig. 6-4 F ), corresponding to the three river dr ainages from which the populations were sampled (Table 1). Pairwise FST values between populations of C. verticillata ranged from 0.028-0.226 and were lowest between pairs of populations 187, 188, and 189 (0.028-0.063), between pairs of 190, 191, 192 (0.076-0.106), between 181 and 188 (0.067), and between 181

PAGE 127

127 and 189 (0.074); pairwise FST values were highest between po pulations from these groups and between population186 and most other populations (0.138-0.226) In hierarchical AMOVA analyses, 12.5% of the variation was pa rtitioned among populations, while 87.5% of the variation was partitioned among indivi duals within populations(Table 6-4). Discussion Deviations from Hardy-Weinberg Equilibrium In total, 57 population/locus combinations of 560 showed significant departures from HWE, in nine out of the ten loci and in all species. These deviations are scattered among all species and loci; because they are not consiste ntly found across all loci in any population, or across all populations for a locus, these devi ations from HWE likel y do not reflect true deviations for either a populat ion or a locus. All of the de viations from HWE are due to heterozygote deficiency, which may be caused by scoring errors due to large allele dropout or stuttering, a Wahlund effect non-random mating, or null alleles. We did not find an excess of homozygotes in any particular size class, and the loci did not exhibit a hi gh amount of stuttering, ruling out scoring errors. A Wahl und effect, where two populations are sampled and analyzed as one, is unlikely because this pattern would be evident in all loci in a population, and there is also no reason to suspect that crypt ic population boundaries exis t within populations of any Conradina sampled in this study Another possibility is non-random mating, and this may be possible in Conradina given the fact that seeds are gr avity-dispersed and probably do not germinate very far from the parent plant. However, if non-random mating were the cause of deviations from HWE in Conradina then we would expect that more than one locus in a population would show departures from HWE, and no population exhi bited deviations at more than one locus. Another possibility is the presence of null alleles, which are non-amplifying alleles due to mutations in the priming site. Indeed, higher levels of missing data (i.e. non-

PAGE 128

128 amplification of a locus in an individual) we re present at the CS507, CSRF, GLAG, and B50-2 loci in the populations with deviations fr om HWE, even though these PCR reactions were attempted twice. The greater prevalence of miss ing data in populations of certain species is therefore likely due to a null allele that is pres ent in these popul ations. However, amplification was successful for the remainder of the loci s howing deviations from HWE, so we are unsure why these loci show deviations from HWE. Future analyses will focus on excluding locus/population combinations, and using altered gene frequencies in analyses of genetic structure to determine if the deviations from HWE influence the results of analyses presented here. Species Limits and Distinguishing Between Hy bridization vs. Incomplete Lineage Sorting in Conradina In AMOVA analyses of the entire Conradina data set, using groups based on the traditional species boundaries, around 21.5% of the genetic variation in Conradina was partitioned among species, indicating si gnificant genetic differentiation among Conradina species (Table 6-4). Neighbor jo ining trees based on genetic distan ce clustered all populations of each species together (Fig. 6-2), indicating that p opulations of each species are most similar to one another. When indi viduals were not grouped a priori into populations or species, analyses of genetic structure clearly par titioned the 1098 individuals accord ing to species boundaries. In runs using the program STRUCTURE, when the number of groupings ( K ) was allowed to vary from 1-8; at K =6 individuals were assigned to groups in concordance with species assignments based on morphological and geographical information (Fig. 6-3). At K =7 and K =8, populations of each species began to be divi ded into the intraspecific groups shown in Fig 6-4. Very little admixture was evident among the species groupings at K =6, and in all cases, each individual was assigned to the appropriate specie s cluster, with only a small percentage of membership in any

PAGE 129

129 other species cluster. Pairwise FST between all populations of Conradina also support differences among species; the only nonsignificant pairwise FST values were found between pairs of populations from the same species. Pairwise FST values between populations of different species ranged from 0.120-0.597, demonstrating moderate to extremely high levels of genetic differentiation (Wright 1965). Although patterns of genetic stru cture indicate a high degree of genetic subdivision in Conradina that is partitioned according to species boundaries, the question remains as to whether the species are sufficiently differentiated genetically to be considered as separate species, or whether interspecific hybr idization is occurring at sufficient levels to blur species boundaries. STRUCTURE analyses grouped individuals acc ording to species boundaries, but small percentages of individuals were as signed to other species clusters that disagreed with the species assignment based on morphology and geography. Many of these assignments may be due to noise from shared ancestral polymorphism, and given the small amount of admixture apparent within individuals, none of the individuals sa mpled in this study a ppeared to be recent interspecific hybrids. However, it is possible that interspecific hybr idization has occurred in past generations, especially within specific geographic regions. Se veral interspecific pairwise FST values are lower than expected; the lowest pairwise FST between populations of C. glabra and C. canescens was 0.120, and between C. brevifolia and C. grandiflora was 0.148, indicative of only moderate genetic differentiation. These values do not indicate recent hybridization, but several interspecific pairwise FST values were higher than the highest intraspecific pairwise Fst value, which may support the hypothesis of past interspecific hybridi zation. Indeed, hybridization between C. brevifolia and C. grandiflora was also suggested by Edwa rds et al. (2006) based on phylogenetic analyses of DNA seque nce data. Alternatively, these results could be due to recent

PAGE 130

130 divergence; sufficient time may not have passed for some species to have become more highly differentiated. Further analyses to quantify the amount of migration among species and to date divergences between the species will help to clar ify which of these processes have occurred in Conradina However, because these microsatellite loci indicate genetic subdivision among the species, and because our results do not support rece nt or widespread interspecific hybridization, there is no reason to change the cu rrently defined species limits in Conradina. Because recent interspecific hybr idization is not evident in Conradina and because there is only weak evidence th at hybridization may have occurred in the past, there is very little support for the hypothesis that interspecific hybr idization is responsible for the non-monophyly of species found in previous phylogenetic analyses of Conradina (Edwards et al. 2006). Instead, these patterns are more likely the result of incomp lete sorting of ancestral variation and the slow rate of DNA substitution relative to the rapid rate of speciation. If a large, polymorphic ancestral species undergoes rapid speciation, or at least more rapid than the evolutio n of the genes, then the resulting daughter species may share polymorphisms across species boundaries. Furthermore, because base substitutions in DNA sequence data occur at a slow rate, there may not have been enough mutations to track the evolution of the species of Conradina Both of these processes have probably c ontributed to the apparent nonmonophyly of species revealed in phylogenetic analyses. Micr osatellite loci evolve at a much more rapid rate (Jarne and Lagoda 1996), which increases the probability that mu tational events will track speciation events, therefore recovering patterns of genetic structure that corre spond to the boundaries of the described species. Patterns of Genetic Structure among Conradina Species The microsatellite loci in this study demonstr ated clear patterns of genetic structuring among Conradina species. Analyses group species according to the geographic regions

PAGE 131

131 indicated in Fig. 6-1 with species in the panhandle/northern region ( C. canescens, C. glabra, C. verticillata, and the Santa Rosa populations) and the peninsular region ( C. etonia, C. grandiflora and C. glabra ) most closely related to one anothe r. These two groupings are supported by multiple lines of evidence; unrooted NJ trees base d on genetic distance clustered species into the two geographically structured groups (Fig. 6-2), pairwise FST were lower in populations in geographic regions (Table 6-3), STRUCTURE analyses at K =2 grouped species into the two geographically structured groups (Fig. 6-3), and previous phylogene tic analyses of DNA sequence data of Edwards et al. {, in press #172} recovered the same geographic structuring as found in this study. However, Edwards et al. (in press) questioned the tree that recovered these geographically structured relationships because of questions about the methodology used to concatenate heterozygotes in combined phyloge netic analyses. Because similar patterns of genetic structure were also found in this study using microsatell ite data, perhaps the methodology to concatenate the heterozygotes is valid. In the STRUCTURE analyses at each successive K the two geographically structured clusters were further subdivided according to species boundaries. At K =3, Conradina verticillata was placed in its own cluster, suggesting that within the panhandle/northern group, C. verticillata probably has the most divergent patterns of allele frequencies. This result is supported by large pairwise FST values (0.195-0.597), indicating that C. verticillata is genetically well differentiated from other Conradina species. Conradina verticillata is geographically isolated from the remainder of the Conradina species, occupying river drainages in Tennessee and Kentucky, while all other species are found in Florida. This geographic isolation leaves virtually no possibility for migration between species. Conradina verticillata is also divergent morphologically; it is a small subshrub no more th an 0.3 m tall, while the other species reach 1

PAGE 132

132 m. At K =4 in STRUCTURE analyses, C. etonia was grouped into its own cluster, which again was supported by large pairwise FST values (0.21-0.597). This high degree of genetic differentiation from the remainder of Conradina species may be a consequence of small population size ; C. etonia is the most endangered Conradina species and had the lowest number of alleles per locus and lowest expected heterozy gosity (Table 6-2). This suggests that factors associated with small numbers of individuals, su ch as genetic drift, a genetic bottleneck, or a founder effect, may have caused a reduction in the genetic diversit y of the species, leading to high levels of genetic differentiation between C. etonia and the remainder of the species. At K =5, Conradina glabra was separated from the group containing C. canescens and the Santa Rosa populations. Pairwise FST between C. glabra and C. canescens revealed moderate to great genetic differentiation betw een the two species (0.120-0.223), but values were smaller than those in all other popula tion pairs. The species have proxi mal geographic distributions, and the similarity of the species may be caused by recent divergence and/or past in terspecific gene flow, although more analyses are necessary to test th is hypothesis. Because STRUCTURE analyses separated C. glabra from the group containing C. canescens and the Santa Rosa populations, and because the Santa Rosa populations are much more differentiated genetically from C. glabra than C. canescens (lowest Fst= 0.134 vs. 0.070, respectively), we conc lude that the Santa Rosa populations are not populations of C. glabra. Instead, the Santa Rosa populations are probably part of C. canescens based on several lines of evidence. STRUCTURE analyses from K =1-7 never separated the Santa Rosa populations from C. canescens Instead, at K =7, STRUCTURE subdivided the cluster of C. canescens/ Santa Rosa populations, but not as expected; instead, four populations of C. canescens from the western panhandle of Florida and Alabama were separated into one group, while the other populations of C. canescens and the Santa Rosa populations were

PAGE 133

133 placed into another. This split did not corres pond to any morphological features or patterns expected given geographic location, because instead of being grouped with neighboring populations of C. canescens the Santa Rosa populations, which are from the western panhandle of Florida, were grouped w ith disjunct populations of C. canescens from the eastern panhandle of Florida. AMOVA analyses to test these gr oupings further supported th ese relationships; when the Santa Rosa populations and C. canescens were treated as separate groups, only 3% of the variation was partitioned among groups, but wh en populations were grouped according to the STRUCTURE results, 6% of the variation wa s partitioned among groups, indicating that the patterns of genetic structur e revealed using STRUCTURE better represent the genetic subdivision in C. canescens There is no meaningful way to divide C. canescens and the Santa Rosa populations; it thus appear s that the Santa Rosa populations are simply glabrous populations of C. canescens and we will hereafter consider them to be C. canescens The fact that both C. glabra and the Santa Rosa populations are bot h glabrous may be due to phenotypic plasticity or parallel ad aptation to a similar habita t; the pubescent populations of C. canescens were all collected in exposed coastal areas, wh ile the glabrous Santa Rosa populations were collected in more protected, shady, in land areas, similar to the habitat of C. glabra It appears that amount of pubescence may be dependent on exposure to sun and wind, and if these differences are fixed in individuals, amount of pubescence is an adaptive trait, but if it is variable in individuals depending on habitat, it is a plastic trait. Common garden experiments would be the best way to determine which of these hypotheses is supported in C. canescens Finally, at K =6, Conradina brevifolia and C. grandiflora were split into separate clusters. Pairwise FST values also revealed a high leve l of genetic similarity between C. brevifolia and C. grandiflora This is an interesting result because C. brevifolia has been considered to be

PAGE 134

134 conspecific with C. canescens by many authors (e.g., Wunderlin 1998) due to morphological similarities between the two. However, C. brevifolia and C. canescens are not genetically similar; lumping C. brevifolia into C. canescens does not adequately represent patterns of variation in these species. Instead, thes e results support the c ontinued recognition of C. brevifolia as a separate species that is endemic to th e Lake Wales Ridge in peninsular Florida. Patterns of Genetic Structure in Endangered Conradina Species and Conservation Implications AMOVA analyses revealed th at among-population variation was responsible for only a small percentage of the genetic variation in most species of Conradina while the majority of the genetic variation was partitione d among individuals with in species (Table 6-4). Populations demonstrated low levels of differentiation acco rding to a priori population boundaries. This may be due to how populations were defined; it is often difficult to determine the population boundaries of Conradina species in the field because some species have a somewhat continuous distribution over a large geographi c area; our population designati ons based on sampling location were not adequate for describing the true population boundaries. However, STRUCTURE analyses indicate that there are definite patter ns of population structure in several species of Conradina although they are grouped into fewer groups then originally defined. In STRUCTURE analyses of each species, the optimal number of groupings was K =2. Three species demonstrated high levels of genetic structure corresponding to geographic regions ( C. canescens C. etonia and C. verticillata ), and efforts should be made to conserve populations of each group to adequately conserve the extant ge netic diversity in each spcies. However, the method that we used to select K is not able to evaluate K =1, and for three species ( C. brevifolia, C. glabra and C. grandiflora ) there may indeed be very lit tle population subdivision, given the high level of admixture in populations revealed at K =2.

PAGE 135

135 In Conradina brevifolia at K =2, populations were divided into one group that contained most of the individuals from Lake Wales Ridge State Forest (populations 168, 169, and 170), and one group that contained most of the indi viduals from populations 161, 197, and 198; some individuals in populations 158 and 159 were assigned to one group and the rest to another. Most populations have individuals that were assigned membership in both clusters, indicating a large amount of admixture between populations, and AMOVA analyses found only a small percentage of variation partitioned among populations. This suggests that populations of C. brevifolia are not genetically differentiated, contrary to expectations, because C. brevifolia occupies isolated patches of scrub interrupted by development on the Lake Wales Ridge in Florida. This population fragmentation does not appear to have reduced gene flow; thus, the network of preserves that has been establis hed on the Lake Wales ridge may be adequate to maintain gene flow and prevent inbreeding depression in C. brevifolia Conversely, because these are longlived shrubs, and fragmentation is recent, the levels of genetic similarity may be due to gene flow that occurred before fragmenta tion. Recent restrictions might only be evident by sampling seeds or seedlings. In Conradina canescens at K =2, populations were grouped according to the Western panhandle and Santa Rosa populations/eastern panhandle groups (see above; Fig. 6-4B). The Santa Rosa populations and eastern panhandle populati ons were divided when K =3, with populations 130 and 131 admixed be tween the three groupings. C. canescens thus demonstrates high levels of genetic differentiation accordi ng to geography, and as the Florida panhandle becomes more developed in future years, conservation efforts should focus on preserving populations in all three geographic areas to conserve the highest le vels of genetic diversity in C. canescens

PAGE 136

136 In Conradina etonia extremely high levels of subdivision were found between populations from Etoniah Creek State Forest an d Dunn’s Creek State Park populations based on pairwise FST, STRUCTURE, and AMOVA analyses. In AMOVA analyses, when populations from the two locations were treated as a single group, an extremely high level of amongpopulation variation was found, but wh en populations from the two re gions were placed into two separate groups the amount of among-population va riation decreased drama tically and most of the variation was found between the groups. STRUCTURE placed the Etoniah Creek and Dunn’s Creek populations into two groups, and pairwise FST values are higher than most interspecific pairs of populations (~0.5). W ithin each of these regi ons, populations are not differentiated, probably due to the narrow range that each occupies The Dunn’s Creek populations were discovered recently and were tentatively placed into C. etonia based on overall morphological similarity. However, the extremel y high levels of genetic differentiation between the two sampling locations indicate that the Dunn ’s creek populations may merit species status. More in-depth morphological analyses are necessa ry to determine if a ny morphological features can be used to differentiate the Etoniah Creek and Dunn’s Creek populations. If no morphological differences are appare nt and it appears that the tw o sampling locations are both C. etonia, then one approach to in crease levels of genetic diversity in the species may be to translocate individuals between the two populations. However, mo re analyses are necessary to determine which is the most appropriate classi fication for the Etoniah Creek and Dunn’s Creek populations. In Conradina grandiflora STRUCTURE clustered indivi duals according to geographic locations; northern and southern populations were placed into two clusters, with admixture occurring between the two regions, perhaps indi cating a pattern of isolation by distance.

PAGE 137

137 AMOVA analyses revealed higher amounts of variation partitioned among populations in C. grandiflora than most species, probably because C. grandiflora occupies a much wider geographic range than in most of the other species, causing large differences between the northernmost and southernmost populations. Because STRUCTURE does not perform well in cases of isolation by distan ce, future analyses in C. grandiflora will focus on determining the amount of isolation by distance within the specie s. The lack of population differentiation in C. grandiflora is encouraging; despite the fact that C. grandiflora occupies the east coast of Florida and the distribution is extremely fragmented due to development, gene flow may be occurring among populations, indicating that fragmen tation does not hinder dispersal in C. grandiflora. However, this gene flow again may be due to the long lifespan of C. grandiflora ; populations may currently be restricted genetically but differentiation may not be evident because we sampled long-lived plants. In Conradina glabra analyses revealed very limite d population subdivision or genetic structure. STRUCTURE analyses place the three translocated populations of C. glabra (143, 144, and 145) (Gordon 1996) into a group with population 109, from which the three populations were derived, and the remaining populations were placed into the other group, but with much more admixture in these populations (163. 164, 165, 166, and 167). Gene flow may be occurring between the translocated populati ons and the other populations of C. glabra ; however, more indepth analysis is necessary to understand patterns of migration. Despite being one of the most endangered species of Conradina C. glabra still maintains comparable levels of genetic diversity and low levels of popul ation subdivision relative to widespread congeners; however these patterns may be due to th e fact that all populations of C. glabra were sampled within a very

PAGE 138

138 small geographic range (i.e., in two neighboring parks in a si ngle county), while widespread species were sampled over much wider distances. In Conradina verticillata AMOVA results found more am ong-population variation than in most other species of Conradina STRUCTURE analyses grouped individuals corresponding closely with the three river dr ainages from which populations we re collected, with very little admixture. The three river drainages sampled in this study are separated by a larger geographical distance than are populations of the other federally endangere d or threatened species of Conradina and the regions are further isolated by m ountainous or hilly terrain. These barriers have apparently restricted gene flow among rive r drainages, causing high er levels of population differentiation in this species. Because of the high level of population differentiation among the river drainages, efforts should be taken to protect populations in all three drainages to conserve the maximum amount of genetic diversity in C. verticillata Because Conradina species are easily established from cuttings it has been suggested that C. verticillata may reproduce vegetatively if a branch breaks off and root s farther downriver (USFWS). The genetic structuring among river drainages do es not discount this hypothesis but more in-depth parentage analysis will be necessary to de termine if this is occurring. Conclusions Although previous phylogenetic analyses were no t able to resolve sp ecies relationships, patterns of genetic structure in Conradina clearly group individuals according to species boundaries defined based on morphol ogical data. Because genetic structure corresponds to species boundaries, and because there is only we ak support for hybridization, it appears that incomplete lineage sorting is probably the mo st reasonable hypothesis for the non-monophyly of species revealed in phy logenetic analyses of Conradina Morphological evolution in Conradina appears to have occurred as ra pidly as evolution in microsate llite regions, both of which are

PAGE 139

139 much more rapid than the rate of substitutions in the DNA sequence regions that we sampled; perhaps morphological phylogenetic analyses wo uld be more appropriate for resolving the phylogeny of Conradina if enough characters are present. W ithin species, very little population differentiation is present, possibl y indicating a high level of gene flow. However, because these plants are long-lived, the lack of genetic structure may correspond to past gene flow; analyses to understand the genetics of recent generations (i.e seeds) would distinguish between these two hypotheses. In the future, we plan to use MIGRATE (Beer li and Felsenstein 2001) to estimate rates and patterns of long-term migration between species and among populations in each species. Within each species, we will carry out Mantel tests to test for isolation by distance, as STRUCTURE does not perform well in such cases. We will also investigate the divergence time between species pairs using IM (Hey a nd Nielsen 2007). Most importantl y, we would like to carry out correlations between geographic ra nge size and levels of geneti c diversity to understand if phylogeny is a determinant of levels of genetic divers ity, or if rarity is a be tter predictor of levels of genetic diversity in Conradina

PAGE 140

140 Table 6-1. Populations of Conradina species sampled in this study. All necessary permits were obtained prior to sampling, and voucher specim ens collected were deposited in FLAS Species/ population N Voucher Locality County C. brevifolia 170 brev158 22 Edwards 158 Saddleblanket Lakes Preserve Polk Co., FL brev159 23 Edwards 159 Carter Creek Preserve, Avon Park Highlands Co., FL brev161 22 Edwards 160 Sun and Lakes Subdivision Highlands Co., FL brev168 18 Edwards 168 Lake Wales Ridge State Forest Polk Co., FL brev169 24 Edwards 169 Lake Wales Ridge State Forest Polk Co., FL brev170 23 Edwards 170 Lake Wales Ridge State Forest Polk Co., FL brev197 17 Edwards 197 Silver Lakes Preserve, Near Sebring Highlands Co. FL brev198 21 Edwards 198 Sun Ray FL Polk Co., FL C. canescens 213 can127 16 Edwards 127 Carrabelle Franklin Co, FL can128 24 Edwards 128 Southeast of Port St. Joe Gulf Co., FL can129 22 Edwards 129 West of Panama City Bay Co. FL can130 22 Edwards 130 East of Sandestin Walton Co., FL can131 22 Edwards 131 Gulf Breeze Santa Rosa Co. FL can138 24 Edwards 138 Mary Ester Okaloosa Co. FL can141 23 Edwards 141 Gulf Shores Baldwin Co., AL can142 19 Edwards 142 Near Fort Morgan Baldwin Co., AL can179 22 Edwards 179 Northwest of Mexico Beach Bay Co., FL can180 19 Edwards 180 West of Port St. Joe Gulf Co. FL C. etonia 126 ce111 24 Edwards 111 Etonia Creek State Park Putnam Co. FL ce153 14 Edwards 153 Etonia Creek State Park Putnam Co. FL ce152 2 Edwards 152 Etonia Creek State Park Putnam Co. FL ce154 24 Edwards 154 Etonia Creek State Park Putnam Co. FL ce155 11 Edwards 155 Etonia Creek State Park Putnam Co. FL ce156 24 Edwards 156 Etonia Creek State Park Putnam Co. FL ce-dc1 3 Oliveira and Miller S.N Dunn’s Creek State Pa rk Putnam Co. FL cd-dc2 24 Oliveira and Miller S.N Dunn’s Creek State Pa rk Putnam Co. FL C. glabra 171 cgl109 10 Edwards 109 Torreya State Park Liberty Co. FL cgl143 22 Edwards 143 Apalachicola Bluffs and Ravines Park Liberty Co. FL cgl144 18 Edwards 144 Apalachicola Bluffs and Ravines Park Liberty Co. FL

PAGE 141

141 Table 6-1. Continued cgl145 16 Edwards 145 Apalachicola Bluffs and Ravines Park Liberty Co. FL cgl163 23 Edwards 163 Torreya State Park Liberty Co. FL cgl164 24 Edwards 164 Torreya State Park Liberty Co. FL cgl165 24 Edwards 165 Torreya State Park Liberty Co. FL cgl166 24 Edwards 166 Torreya State Park Liberty Co. FL cgl167 10 Edwards 167 Torreya State Park Liberty Co. FL C. grandiflora 176 cgr134 19 Edwards 134 Seacrest Scrub Natural Area Palm Beach Co. FL cgr137 24 Edwards 137 Johnathan Dickinson State Park Martin Co. FL cgr135 24 Edwards 135 Juno Dunes Natural Area Palm Beach Co. FL cgr171 20 Edwards 171 St. Sebastian River State Preserve Indian River Co., FL cgr172 20 Edwards 172 Near Vero Beach St. Lucie Co. FL cgr173 23 Edwards 173 By St. Lucie River Martin Co., FL cgr193 23 Edwards 193 Volusia Co., FL cgr215 23 Edwards 215 Titusville Well Fields Brevard Co. FL C. verticillata 160 vert181 17 Edwards 181 Along Daddy’s creek, Obed Wild and Scenic River Morgan Co. TN vert184 23 Edwards 184 Lilly Bridge on the Clear Creek, Obed Wild and Scenic River Morgan Co. TN vert186 7 Edwards 186 Obed Junction, Obed Wild and Scenic River Morgan Co. TN vert187 24 Edwards 187 Big Island, Big South Fork Nation River and Recreation Area Scott Co, TN vert188 21 Edwards 188 Leatherwood Ford, Big South Fork Nation River and Recreation Area Scott Co, TN vert189 23 Edwards 189 Burnt Mill Bridge, Big South Fork Nation River and Recreation Area Scott Co, TN vert190 15 Edwards 190 Caney Fork River White Co., TN vert191 20 Edwards 191 Caney Fork River White Co., TN vert192 10 Edwards 192 Caney Fork River White Co., TN Santa Rosa Populations 93 csr133 22 Edwards 133 NW of Milton Airport Santa Rosa Co. FL csr209 24 Edwards 209 Near Milton Airport Santa Rosa Co. FL csr211 24 Edwards 211 Along Blackwate r river Santa Rosa Co. FL csr212 23 Edwards 212 Along Blackwate r river Santa Rosa Co. FL

PAGE 142

142Table 6-2. Diversity indi ces for each species of Conradina HE =expected heterozygosity, a =number of alleles GLAH CSC1 CS507 CSRE CSRD CGL30-4 GLAG B502 CSR30-4 CSRF HE a HE a HE a HE a HE a HE a HE a HE a HE a HE a Br158 0.83 9 0.86 11 0.5880.5540.91140.806 0.8170.86110.76140.789 Br159 0.81 8 0.90 13 0.6360.7460.90100. 737 0.88100.88120.75100.8310 Br161 0.87 9 0.66 10 0.7470.6650.90140.716 0.7860.8780.7790.8710 Br168 0.79 7 0.63 6 0.6260.7860.85100.716 0.8680.89110.7060.8810 Br169 0.71 6 0.82 9 0.6060.6070.87140.827 0.8490.88110.7870.727 Br170 0.80 8 0.85 10 0.5750.5060.86100.715 0.78100.7390.6350.819 Br197 0.89 9 0.73 8 0.5830.5550.88110.614 0.6950.90110.7870.898 Br198 0.75 9 0.88 10 0.4240.4860.8170.765 0.86100.87110.7470.817 can127 0.74 9 0.87 9 0.7960.4340.5970.868 0.8380.1830.92180.8710 can128 0.89 11 0.42 5 0.8260.7350.7790.8512 0.6960.0010.6670.8911 can129 0.88 12 0.75 10 0.8180.7870.7590.9113 0.4950.0010.92140.8612 can130 0.88 12 0.82 11 0.8480.7650.6190.8511 0.5770.2120.89140.7610 can131 0.92 15 0.90 14 0.8780.7360.6270.869 0.6790.0010.92160.7310 can138 0.90 13 0.85 11 0.7880.7050.8280.8510 0.5580.2640.84110.768 can141 0.87 10 0.85 10 0.8690.7360.8290.8010 0.6360.0010.87110.789 can142 0.86 11 0.79 8 0.5240.8060.8680.8410 0.7670.0010.81100.827 can179 0.86 9 0.77 6 0.8270.6960.5580.836 0.6770.0010.8180.689 can180 0.82 10 0.72 8 0.8690.7970.5850.768 0.4240.0010.6390.806 ce111 0.29 2 0.66 4 0.0010.5330.1220.483 0.0520.1620.5130.312 ce153 0.00 1 0.56 3 0.0010.6330.1420.663 0.5220.3630.8170.142 ce152 0.50 2 0.50 2 0.0010.6720.0010.001 0.0010.5020.6720.001 ce154 0.34 3 0.76 5 0.0010.4750.0010.584 0.5430.5330.7770.282 ce155 0.17 2 0.69 3 0.0010.6930.0010.393 0.5220.2520.7260.512 ce156 0.47 3 0.75 4 0.0010.5940.0820.583 0.3120.5030.7860.512 ce-dc1 0.33 2 0.33 2 0.0010.3320.5320.532 0.6030.3320.3320.001 cd-dc2 0.51 2 0.38 3 0.0010.6460.5650.614 0.4440.5020.5540.342 cgl109 0.76 5 0.77 7 0.7550.8150.6640.856 0.6730.3530.8780.725 cgl143 0.63 6 0.85 9 0.8260.7770.7770.724 0.4730.0920.85130.736 cgl144 0.76 8 0.86 8 0.7040.7870.6750.828 0.6430.2920.89130.675 cgl145 0.57 6 0.85 8 0.6860.8060.6850.665 0.5930.0010.92130.847 cgl163 0.81 7 0.89 14 0.7860.7580.7460.808 0.6040.4530.83110.001 cgl164 0.75 8 0.89 11 0.5050.84100.6370. 819 0.6730.3170.9013n.d. n.d cgl165 0.88 9 0.83 9 0.6340.4670.7770.8 27 0.6840.1730.8514n.d. n.d cgl166 0.80 7 0.82 10 0.6070.6280.7580.809 0.7440.1620.90120.757 cgl167 0.72 6 0.83 7 0.8770.5150.3930.734 0.5930.3920.8470.523 cgr134 0.42 2 0.76 7 0.8270.8060.0010.724 0.8680.7490.4240.133

PAGE 143

143Table 6-2. Continued. cgr137 0.69 5 0.83 9 0.7450.7750.0830.537 0.8590.8490.6550.325 cgr135 0.53 3 0.77 7 0.6980.8170.1340.625 0.8490.79100.5830.672 cgr171 0.46 3 0.82 8 0.7180.6460.1530.727 0.7870.7050.563n.d. n.d cgr172 0.76 7 0.78 6 0.8580.7550.3420.858 0.7670.7150.503n.d. n.d cgr173 0.67 4 0.70 6 0.6860.5750.0010.839 0.7560.7150.5360.733 cgr193 0.66 5 0.52 5 0.7250.7560.1730.555 0.6460.5940.6960.414 cgr215 0.17 4 0.55 4 0.6040.4140.0010.755 0.6560.7340.6130.695 vert181 0.33 3 0.74 5 0.00 1 0.7850.6570.252 0.7550.6740.83100.672 vert184 0.63 7 0.74 6 0.00 1 0.8390.4160.764 0.6050.3420.87100.725 vert186 0.62 3 0.66 4 0.00 0.8460.4940.804 0.6530.0010.8970.613 vert187 0.48 5 0.71 9 0.00 1 0.6040.8490.342 0.6850.6340.8080.532 vert188 0.60 4 0.71 8 0.00 1 0.7050.7270.372 0.6750.7250.8370.503 vert189 0.36 5 0.78 9 0.04 2 0.7160.6980.492 0.5840.4520.87120.502 vert190 0.75 4 0.80 8 0.13 3 0.7570.6860.583 0.5840.5320.6550.765 vert191 0.67 6 0.79 8 0.05 2 0.8180.7860.734 0.5730.4420.7590.833 vert192 0.56 3 0.73 7 0.00 2 0.7960.8160.493 0.6930.4730.8480.896 csr133 0.88 11 0.88 12 0.7880.5340.6570.798 0.6150.1740.6891.004 csr209 0.86 10 0.74 9 0.7180.8260.7550.8 07 0.5660.1540.8713n.d. n.d csr211 0.81 7 0.83 7 0.6850.7660.8180.815 0.2940.2230.93130.001 csr212 0.80 9 0.88 9 0.5930.6570.8380.859 0.6070.4040.8290.001

PAGE 144

144 Table 6-3. The range in pairwise FST values between all 56 p opulations, grouped into th e described species of Conradina, including the Santa Rosa populations. All comparisons between populations of different species are significantly different than zero at the P<0.01 level. Species brevifolia canescens etonia glabra grandiflora Santa Rosa verticillata brevifolia 0.028-0.115 canescens 0.186-0.271 0.030-0.180 etonia 0.213-0.395 0.334-0.491 -0.016-0.551 glabra 0.175-0265 0.120-0.223 0.393-0.555 0.020-0.159 grandiflora 0.148-0.305 0.195-0.350 0.244-0.560 0.204-0.365 0.048-0.224 Santa Rosa 0.198-0.265 0.070-0.195 0.361-0.537 0.134-0.207 0.216-0.385 0.058-0.091 verticillata 0.219-0.324 0.233-0.358 0.411-0.597 0.195-0.358 0.291-0.415 0.241-0.371 0.028-0.199

PAGE 145

145 Table 6-4. Results of hier archical AMOVA analyses. Data set Number of species Number of populations Percent variation among species Percent variation among populations within species Percent variation among individuals All individuals 6 56 22.1 8.6 69.3 C. brevifolia 1 8 n/a 6.3 93.7 C. canescens and Santa Rosa Pops. 1 14 n/a 12.8 87.2 C. etonia 1 8 n/a 23.9 76.1 C. glabra 1 9 n/a 6.1 93.9 C. grandiflora 1 8 n/a 10.7 89.3 C. verticillata 1 10 n/a 12.5 87.5

PAGE 146

146 Figure 6-1. Collection locations and population designations of popul ations used in this study.

PAGE 147

147 Figure 6-2. Neighbor joining phylogram of pairwise genetic distances of all Conradina populations included in this study. Distances calculated using Nei’s standard genetic distances.

PAGE 148

148 Figure 6-3. Results of STRUCTURE analyses to investigate species boundaries in Conradina

PAGE 149

149 Figure 6-4. Results of STRUCTUR E analyses in each species of Conradina

PAGE 150

150 CHAPTER 7 GENERAL CONCLUSIONS The purpose of this study was to unde rstand the patterns of evolution in Conradina. The first objective was to r econstruct the phylogeny of Conradina and related mints using ITS and plastid sequence data. ITS sequence da ta strongly supported the monophyly of Conradina, in agreement with evidence from morphology, but di d not resolve species relationships within Conradina Plastid sequence data did not support a monophyletic Conradina or the monophyly of most traditionally defined species of Conradina that are distinguishable morphologically, perhaps due to shared ancestr al polymorphisms or introgre ssion. More rapidly evolving sequence data from nuclear markers were needed to clarify relationships in Conradina and related mints from the southeastern United States. To clarify the phylogeny of Conradina and related mints further, three nuclear genes in the GapC gene family were isolated and two of thes e genes were used to carry out phylogenetic analyses. Because separate analyses of the two GapC loci did not resolve species relationships, two approaches were used to c oncatenate the two heterozygous GapC loci with the ITS and plastid data sets and combined analyses were carried out. Trees resulting from the two concatenation approaches were similar in the re solution and support of ge neric relationships, but differed drastically in resolution a nd support for relationships within Conradina Conradina species are probably very recently derived, and it may be unreasonable to reconstruct species relationships in Conradina using DNA sequence data due to wide spread hybridization or lack of coalescence. Next, we investigated the utility of six c oncatenation methods for carrying out combined analyses of two or more heterozygous nuclear gene us ing the ITS, plastid, and GapC data sets. Relationships among the southeastern scrub mint genera were similar in topologies resulting

PAGE 151

151 from the six concatenation methods; Dicerandra was sister to the remainder of the SE scrub mint clade, Conradina was monophyletic with Piloblephis or Clinopodium georgianum as sister, and Stachydeoma was sister to a clade of Clinopodium species. Within Conradina trees resulting from two concatenation methods in particular we re much more resolved than others, possibly because these two methods ignore conflict due to heterozygosity in the data sets. If trees resulting from differing concatenatio n methods agree, then little conf lict may be present in a data set, and all methods yield equa lly adequate representations of phylogenetic relationships. However, if less resolution is present in tr ees resulting from some methods, methods that collapse in the presence of heterozygosity may be the most conservative way to display the conflict resulting from heteroz ygosity in combined analyses. Although phylogeny reconstruction wa s able to resolve relationships among genera of the SE scrub mint clade, these analyses di d not resolve species relationships in Conradina The next goal was to understand patterns of genetic structure in Conradina using genotype data from quickly evolving microsatellite loci and to investigate hypotheses of hybridization and incomplete lineage sorting. All analyses indi cated a clear differentiation of individuals according to species boundaries, with a large percentage of variation partitioned among species. There were no recent hybrids uncovered in our an alyses. Populations of some of the species were genetically similar, but se veral explanations, such as rece nt species divergences or past hybridization, may be responsible for these patterns. The hypot hesis of incomplete lineage sorting, raised via phylogenetic analyses of DNA sequence data, is the more probable explanation for non-monophyly of sp ecies recovered in phy logenetic analyses, and differences in coalescence time due to variation in mutational rate in DNA sequen ces vs. microsatellite data are probably responsible for the differe nces in resolution of relationsh ips between the two data sets.

PAGE 152

152 Results of within-species variati on indicate that very l ittle population different iation is present in each species of Conradina suggesting that high levels of outcrossing have occurred among populations of each species, even in populations that are fragme nted geographically. This may indicate that fragmentation is not affecting gene flow, and current management efforts are adequate. However, because these plants are l ong-lived, the lack of ge netic structure in each species may reflect past gene flow that occurred prior to fragmenta tion; investigating the patterns of genetic structure in seeds and seedlings may be the best way to determine if current gene flow among populations is occurring. Because all describe d species appear to be valid entities, the current listing status of the endangered species of Conradina is appropriate gi ven the patterns of genetic data revealed in this study. In the future, morphological phylogenetic anal yses will be conducted and combined with molecular data sets to help resolve phylogene tic trees, identify morphological synapomorphies, and create a revised classification and key to the species in the SE scrub mint clade. These trees will be used to understand the evolution of leve ls of genetic diversity in the clade, and to evaluate if rarity necessarily corresponds to low levels of genetic diversity. Further analyses of microsatellite data in Conradina will focus on understanding longterm patterns of migration among populations in each species and among species of Conradina and how migration patterns relate to conservation of the endangered species. Microsatellite data will also be used to date divergences between species pairs in Conradina using the program IM (Hey and Nielsen 2007) Overall, I have learned that the evolutionary dynamics of closely re lated species complexes can be very complicated. Speciation is complex and recent divergences or incipient species may not be detectable using the genetic markers commonly used to reconstruct species-level relationships. Occurrences such as hybridization or incomple te lineage sorting, which would

PAGE 153

153 probably be found more frequently if more studi es included multiple populations of each species, can complicate the patterns recovered in phylogene tic analysis, but contin ued investigations can help to clarify conflicting patte rns that may arise through these processes. Although evolutionary patterns can be difficult to detect in closely related species complexes using DNA sequence data, the complicated patterns that emer ge, and the questions that arise based on these results are often more interesting than the original questions. Furthermore, by using more in-depth data to resolve relationships, we can learn much more about the evolutionary dynamics of these organisms than originally anticipated.

PAGE 154

154 LIST OF REFERENCES Aguilar JF, GN Feliner 2003 Additive polymorphi sms and reticulation in an ITS phylogeny of thrifts ( Armeria Plumbaginaceae). Mol P hylogenet Evol 28:430-447. Altschul SF, TL Madden, AA Schaffer, JH Zha ng, Z Zhang, W Miller, DJ Lipman 1997 Gapped BLAST and PSI-BLAST: a new generation of pr otein database search programs. Nucleic Acids Res 25:3389-3402. Arias MC, WS Sheppard 2005 Phylogenetic relati onships of honey bees (Hymenoptera: Apinae: Apini) inferred from nuclear and mito chondrial DNA sequence data. Mol Phylogenet Evol 37:25-35. Bardeleben C, RL Moore, RK Wayne 2005 A mo lecular phylogeny of the Canidae based on six nuclear loci. Mol Phylogenet Evol 37:815-831. Barker FK, FM Lutzoni 2002 The utility of the incongruence length difference test. Syst Biol 51:625-637. Beerli P, J Felsenstein 2001 Maximum likelihood es timation of a migration matrix and effective population sizes in n subpopulati ons by using a coalescent ap proach. Proc Natl Acad Sci U S A 98:4563-4568. Boutin-Ganache I, M Raposo, M Raymond, CF Deschepper 2001 M13-tailed primers improve the readability and usability of microsatelli te analyses performed with two different allele-sizing methods BioTechniques 31:24-25. Bruen TC, H Philippe, D Bryant 2006 A simple a nd robust statistical test for detecting the presence of recombination. Genetics 172:2665-2681. Buckley TR, M Cordeiro, DC Marshall, C Si mon 2006 Differentiating between hypotheses of lineage sorting and introgression in New Zealand alpine cicadas ( Maoricicada Dugdale). Syst Biol 55:411-425. Calcagnotto D, SA Schaefer, R DeSalle 2005 Re lationships among charac iform fishes inferred from analysis of nuclear and mitochondr ial gene sequences. Mol Phylogenet Evol 36:135-153.

PAGE 155

155 Cantino PD, RM Harley, SJ Wagstaff 1992 Genera of the Labiatae: status and classification, Pages 511-522 in R. M. Harley, and T. Reynolds, eds. Advances in Labiatae science. Kew, Royal Botanic Gardens Press. Cantino PD, SJ Wagstaff 1998 A reexamination of north American Satureja s.l. (Lamiaceae) in light of molecular evidence. Brittonia 50:63-70. Chafin L 2000 Field guide to the rare plants of Florida. Tallahassee, FL Florida Natural Areas Inventory. Christman SP, WS Judd 1990 Notes on plants ende mic to the Florida scrub. Fla Sci 53:52-73. Cline J, J Braman, H Hogrefe 1996 PCR fidelity of pfu DNA polymerase and other thermostable DNA polymerases. Nucleic Acids Res 24:3546-3551. Creer S, CE Pook, A Malhotra, RS Thor pe 2006 Optimal intron analyses in the Trimeresurus radiation of Asian pitvip ers. Syst Biol 55:57-72. Cronn R, M Cedroni, T Haselkor n, C Grover, JF Wendel 2002 PCRmediated recombination in amplification products derived from pol yploid cotton. Theor Appl Genet 104:482-489. Crook R 1998 Systematics of Conradina (Lamiaceae). Ph.D. Diss. University of Georgia, Athens, GA. Cunningham CW 1997 Is congruence between data pa rtitions a reliable pr edictor of phylogenetic accuracy? Empirically testi ng an iterative procedure for choosing among phylogenetic methods. Syst Biol 46:464-478. Darlu P, G Lecointre 2002 When does the incong ruence length difference test fail? Mol Biol Evol 19:432-437. Delaney RK, RP Wunderlin 1989 A new species of Crotalaria (Fabaceae) from the Florida Central Ridge. SIDA Contrib Bot 13:315-324. Delcourt PA, HR Delcourt 1981 Vegetation maps fo r eastern North America: 40,000 YR BP to the present, Pages 123-165 in R. C. Romans, ed. Geobotany II. New York, Plenum Publishing Co.

PAGE 156

156 Dieringer D, C Schlotterer 2003 MICROSA TELLITE ANALYSER (MSA): a platform independent analysis tool for large microsat ellite data sets. Mo l Ecol Notes 3:167-169. Dolphin K, R Belshaw, CDL Orme, DLJ Quic ke 2000 Noise and incongruence: Interpreting results of the incongruence length differe nce test. Mol Phylogenet Evol 17:401-406. Donoghue MJ 1985 A critique of the biological species concept and recommendations for a phylogenetic alternative. Bryologist 88:172-181. Doroszenko A 1985 Taxonomic studies on the Satureja complex (Labiatae). PhD dissertation Edinburgh University. Dowton M, AD Austin 2002 Increased congruence does not necessarily indicate increased phylogenetic accuracy: The behavior of th e incongruence length difference test in mixed-model analyses. Syst Biol 51:19-31. Doyle JJ, JL Doyle 1987 A rapid isolation pro cedure for small quantities of fresh tissue. Phytochem Bull 19:11-15. Doyle JJ 1995 The irrelevance of allele tree topologies for species delimitation, and a nontopological alternative. Syst Bot 20:574-588. ——— 1997 Trees within trees: Genes and specie s, molecules and morphology. Syst Biol 46:537-553. Edwards CE, DE Soltis, PS Soltis 2006 Molecular phylogeny of Conradina and other scrub mints (Lamiaceae) from the southeastern USA: evidence for hybridization in pleistocene refugia? Syst Bot 31:193-207. Edwards CE, M Arakaki, P Quintana-Ascencio, DE Soltis, PS Soltis 2007a Isolation and characterization of microsate llite loci from the endangered highlands scrub hypericum ( Hypericum cumulicola ). Mol Ecol Notes in press. Edwards CE, DE Soltis, PS Soltis 2007b Isola tion, characterization, and cross-species amplifications of microsatellite loci in Conradina Mol Ecol Notes in press. Edwards CE, D Lefkowitz, DE Soltis, PS Soltis in press Phylogeny of Conradina and related southeastern scrub mints (Lamiaceae) based on GapC gene sequences. Int J Plant Sci.

PAGE 157

157 Emshwiller E, JJ Doyle 2002 Origins of do mestication and polyploidy in oca ( Oxalis tuberosa : Oxalidaceae). 2. Chloroplast-expressed glut amine synthetase data. Am J Bot 89:10421056. Evanno G, S Regnaut, J Goudet 2005 Detecting the num ber of clusters of individuals using the software STRUCTURE: a simu lation study. Mol Ecol 14:2611-2620. Excoffier L, PE Smouse, JM Quattro 1992 Analysis of Molecular Variance Inferred from Metric Distances among DNA Haplotypes App lication to Human Mitochondrial-DNA Restriction Data. Genetics 131:479-491. Excoffier L, G Laval, S Schneider 2005 Arlequin ver. 3.0: An integrated software package for population genetics data analysis. Evol utionary Bioinformatics Online 1:47-50. Farris JS, M Kallersjo, AG Kluge, C Bult 1994 Testing significance of incongruence. Cladistics 10:315-319. Felsenstein J 1985 Confidence limits on phylogenies-an approach using the bootstrap. Evolution 39:783-791. Figge RM, M Schubert, H Brinkmann, R Cerff 1999 Glyceraldehyde-3-phosphate dehydrogenase gene diversity in eubacteria a nd eukaryotes: evidence for intraand interkingdom gene transfer. Mo l Biol Evol 16:429-440. Gaines CA, MP Hare, SE Beck, HC Rosenbaum 2005 Nuclear markers confirm taxonomic status and relationships among highly endangered and closely related right whale species. Proc R Soc Biol Sci Ser B 272:533-542. Godfrey RK 1998 Trees, shrubs, and woody vines of Northern Florida and adjacent Georgia and Alabama. University of Ge orgia Press, Athens. 322 pp. Goldman N 1993 Simple diagnostic statistical tests of models for DNA substitution. J Mol Evol 37:650-661. Gordon DR 1996 Experimental translocation of the endangered shrub Apalachicola rosemary Conradina glabra to the Apalachicola Bluffs and Ravines Preserve, Florida. Biol Conserv 77:19-26.

PAGE 158

158 Goudet J 2002 FSTAT (version 2.9.3.2): A computer program to calculate F-statistics. Gray TC 1965 A monograph of the genus Conradina. PhD diss. Vanderbilt University, Nashville, TN. Hasegawa M, H Kishino, TA Yano 1985 Dating of th e human-ape split by a molecular clock of mitochondrial DNA. J Mol Evol 22:160-174. Hey J, R Nielsen 2007 Integration within the Felsenstein equation for improved Markov chain Monte Carlo methods in popul ation genetics. Proc Natl Acad Sci U S A 104:2785-2790. Hipp AL, JC Hall, KJ Sytsma 2004 Congruence versus phylogenetic accuracy: Revisiting the incongruence length difference test. Syst Biol 53:81-89. Howarth DG, DA Baum 2005 Gen ealogical evidence of homoploi d hybrid speciation in an adaptive radiation of Scaevola (Goodeniaceae) in the Ha waiian Islands. Evolution 59:948-961. Huck RB 1987 Systematics and evolution of Dicerandra (Labiatae). Phanerogamarum Monogr 19:1-343. ——— 1994 Complex patterns of evolution in perenni al labiates in Florida. Am J Bot Suppl 81:168-169. Huck RB, W. S. Judd, W. M. Whitten, J. D. Skean, R. P. Wunderlin, and K. R. Delaney. : 1989 A new Dicerandra from the Lake Wales Ridge of Florida, with a cladistic analysis and discussion of endemism. Syst Bot:197-213. Hudson RR 1990 Gene genealogies and the coales cent process. Oxf Surv Evol Biol 7:1-44. Huelsenbeck JP, KA Crandall 1997 Phylogeny estimation and hypothesis testing using maximum likelihood. Annu Rev Ecol Syst 28:437-466. Huelsenbeck JP, F Ronquist, R Nielsen, JP Boll back 2001 Bayesian inference of phylogeny and its impact on evolutionary biology. Science 294:2310-2314. Huson DH, D Bryant 2006 Application of phylogene tic networks in evolutionary studies. Mol Biol Evol 23:254-267.

PAGE 159

159 Iglesias SP, G Lecointre, DY Sellos 2005 Extens ive paraphylies within sharks of the order Carcharhiniformes inferred from nuclear and mitochondrial DNA. Mol Phylogenet Evol 34:569-583. Ingvarsson PK 2005 Nucleotide polymorphism a nd linkage disequilbrium within and among natural populations of European Aspen ( Populus tremula L., Salicaceae). Genetics 169:945-953. Jarne P, PJL Lagoda 1996 Microsatellites, from mo lecules to populations and back. Trends Ecol Evol 11:424-429. Johnson LA, DE Soltis 1994 matK DNA-Sequences and Phylogenetic Reconstruction in Saxifragaceae S-Str. Syst Bot 19:143-156. Joly S, A Bruneau 2006 Incorporating allelic varia tion for reconstructing th e evolutionary history of organisms from multiple genes: An example from Rosa in North America. Syst Biol 55:623-636. Kandpal RP, G Kandpal, SM Weissman 1994 Constr uction of Libraries Enriched for Sequence Repeats and Jumping Clones, and Hybridizat ion Selection for Regi on-Specific Markers. Proc Natl Acad Sci U S A 91:88-92. Kelchner SA, JF Wendel 1996 Hairpins create minute inversions in non-coding regions of chloroplast DNA. Curr Genet 30:259-262. Kelchner SA 2000 The evolution of non-coding ch loroplast DNA and its application in plant systematics. Ann Mo Bot Gard 87:482-498. Kral R, RB McCartney 1991 A new species of Conradina (Lamiaceae) from northeastern peninsular Florida. SIDA Contrib Bot 14:391-398. Larena BG, JF Aguilar, GN Feliner 2002 Gl acial-induced altitudinal migrations in Armeria (Plumbaginaceae) inferred from patterns of chloroplast DNA haplotype sharing. Mol Ecol 11:1965-1974. Larget B, DL Simon 1999 Markov chain Monte Carl o algorithms for the Bayesian analysis of phylogenetic trees. Mol Biol Evol 16:750-759.

PAGE 160

160 Larkin LL, JL Neff, BB Simpson 2006 Phylogeny of the Callandrena subgenus of Andrena (Hymenoptera: Andrenidae) based on mitochondrial and nuclear DNA data: Polyphyly and convergent evolution. Mo l Phylogenet Evol 38:330-343. Maddison DR, WP Maddison 2001 MacClade 4: An alysis of phylogeny and character evolution, version 4.05. Sinauer Asso ciates, Inc., Sunderland. Malcomber ST 2002 Phylogeny of Gaertnera Lam. (Rubiaceae) based on multiple DNA markers: Evidence of a rapid radiation in a widespread, morphologically diverse genus. Evolution 56:42-57. Mantel N 1967 Detection of dis ease clustering and a generalize d regression approach. Cancer Res 27:209-&. Martin DP, C Williamson, D Posada 2005 RDP2: R ecombination detection and analysis from sequence alignments. Bioinformatics 21:260-262. Martin EL 1992 Patterns of genetic diversity in the rare mint Conradina glabra and its nearest relative, Conradina canescens M. S. Thesis University of West Florida. Martin W, H Brinkmann, C Savonna, R Cerff 1993a Evidence for a chimeric nature of nuclear genomes: Eubacterial origin of eukaryo tic glyceraldehyde-3-phosphate dehydrogenase genes. Proc Natl Acad Sci U S A 90:8692-8696. Martin W, D Lydiate, H Brinkmann, G Fo rkmann, H Saedler, R Cerff 1993b Molecular phylogenies in angiosperm evol ution. Mol Biol Evol 10:140-162. Mason-Gamer RJ, KE Holsinger, RK Jansen 1999 Chloroplast DNA variation in Coreopsis nuecensoides and C-nuecensis (Asteraceae), a presumed progenitor-derivative species pair. Plant Syst Evol 218:5-12. Mishler BD 1985 The morphological, developmental, and phylogenetic basis of species concepts in Bryophytes. Bryologist 88:207-214. Moore WS 1995 Inferring phylogenies from mtDNA Variation mito chondrial gene trees versus nuclear gene trees. Evolution 49:718-726.

PAGE 161

161 Morrell PL, DM Toleno, KE Lundy, MT Clegg 2005 Low levels of linkage disequilibrium in wild barley ( Hordeum vulgare ssp spontaneum ) despite high rates of self-fertilization. Proc Natl Acad Sci U S A 102:2442-2447. Myers RL 1990 Scrub and high pine, Pages 150-193 in R. L. Myers, and J. Ewell, eds. Ecosystems of Florida. Orlando, Univ ersity of Centra l Florida Press. Nei M 1972 Genetic distance between populations. Am Nat 106:283–291. Neigel JE, JC Avise 1986 Phylogenetic rela tionships of mitochondrial DNA under various demographic models of speciation, Pages 515-534 in E. Nevo, and S. Karlin, eds. Evolutionary processes and theo ry. New York, Academic Press. Neigel JE, RM Ball, JC Avise 1991 Estimation of single generation migration distances from geographic-variation in animal m itochondrial-DNA. Evolution 45:423-432. Niu TH, ZHS Qin, XP Xu, JS Liu 2002 Bayesian haplotype inference for multiple linked singlenucleotide polymorphisms. Am J Hum Genet 70:157-169. Niu TH 2004 Algorithms for inferring haplotypes. Genet Epidemiol 27:334-347. Nylander JAA, F Ronquist, JP Huelsenbeck, JL Nieves-Aldrey 2004 Bayesian phylogenetic analysis of combined data. Systematic Biology 53:47-67. Oliveira LO, RB Huck, MA Gitzendanner, WS Judd, DE Soltis, PS Soltis 2007 Molecular phylogeny, biogeography, and systematics of Dicerandra (Lamiaceae), a genus endemic to the southeastern United States. Am J Bot 94:1017-1027. Olsen KM, BA Schaal 1999 Evidence on the origin of cassava: Phylogeography of Manihot esculenta Proc Natl Acad Sci USA 96:5586-5591. Pereira SL, AJ Baker, A Wajntal 2002 Combin ed nuclear and mitochondrial DNA sequences resolve generic relationships within the Cracidae (Galliformes, Aves). Systematic Biology 51:946-958. Perusse JR, DJ Schoen 2004 Mo lecular evolution of the GapC gene family in Amsinckia spectabilis populations that differ in outcr ossing rate. J Mol Evol 59:427-436.

PAGE 162

162 Petersen G, O Seberg, M Yde, K Berthe lsen 2006 Phylogenetic relationships of Triticum and Aegilops and evidence for the origin of the A, B, and D genomes of common wheat ( Triticum aestivum ). Mol Phylogenet Evol 39:70-82. Petit RJ, E Pineau, B Demesure, R Bacilieri, A Ducousso, A Kremer 1997 Chloroplast DNA footprints of postglacial recolonizati on by oaks. Proc Natl Acad Sci U S A 94:999610001. Posada D, KA Crandall 1998 MODELTEST: te sting the model of DNA substitution. Bioinformatics 14:817-818. Posada D 2002 Evaluation of methods for de tecting recombination from DNA sequences: Empirical data. Mol Biol Evol 19:708-717. Posada D, KA Crandall 2002 The effect of recombination on the accuracy of phylogeny estimation. J Mol Evol 54:396-402. Prather LA, AK Monfils, AL Posto, RA Williams 2002 Monophyly and phylogeny of Monarda (Lamiaceae): Evidence from the internal tr anscribed spacer (ITS) region of nuclear ribosomal DNA. Syst Bot 27:127-137. Pritchard JK, M Stephens, P Donnelly 2000 Infere nce of population structure using multilocus genotype data. Genetics 155:945-959. Rambaut A 2003 Se Al: A manual sequen ce alignment editor, version 2.0 a11. Raymond M, F Rousset 1995 Genepop (Version-1.2) Population-genetics software for exact tests and ecumenicism. J Hered 86:248-249. Reeder SA, DS Carroll, CW Edwards, CW Kilpatrick, RD Bradley 2006 Neotomineperomyscine rodent systematics based on combined analyses of nuclear and mitochondrial DNA sequences. Mol Phylogenet Evol 40:251-258. Rieseberg LH, DE Soltis 1991 Phylogenetic conseque nces of cytoplasmic gene flow in plants. Evol Trend Plant 5:65-84. Rieseberg LH, J Whitton, CR Linder 1996 Molecular ma rker incongruence in plant hybrid zones and phylogenetic trees. Ac ta Bot Neerl 45:243-262.

PAGE 163

163 Ronquist F, JP Huelsenbeck 2003 MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19:1572-1574. Ronquist F, HP Huelsenbeck, P van der Mark 2005 MrBayes 3.1 manual. Rosenberg MS 2001 PASSAGE. Pattern analysis, sp atial statistics, and geographic exegesis. Version 1.0. Department of Biology, Arizona State University, Tempe. Rosenberg NA 2004 DISTRUCT: a program for the graphical display of population structure. Mol Ecol Notes 4:137-138. Rousset F 2007 Genepop'007: a complete reimpl ementation of the Genepop software for Windows and Linux. Mol Ecol Notes in press. Sang T, DJ Crawford, TF Stuessy 1997 Chlo roplast DNA phylogeny, reticulate evolution, and biogeography of Paeonia (Paeoniaceae). Am J Bot 84:1120-1136. Sang T 2002 Utility of low-copy nuclear gene sequences in plant phylogenetics. Crit Rev Biochem Mol Biol 37:121-147. Schaal BA, DA Hayworth, KM Olsen, JT Rausch er, WA Smith 1998 Phylogeographic studies in plants: problems and pros pects. Mol Ecol 7:465-474. Schaal BA, KM Olsen 2000 Gene genealogies and population variation in pl ants. Proc Natl Acad Sci U S A 97:7024-7029. Scheet P, M Stephens 2006 A fast and flexib le statistical model fo r large-scale population genotype data: Applications to inferring missing genotypes and haplotypic phase. Am J Hum Genet 78:629-644. Seelanan T, A Schnabel, JF Wendel 1997 Congr uence and consensus in the cotton tribe (Malvaceae). Syst Bot 22:259-290. Shimodaira H, M Hasegawa 1999 Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol Biol Evol 16:1114-1116. Shimodaira H 2002 An approximately unbiased test of phylogenetic tree selection. Syst Biol 51:492-508.

PAGE 164

164 Shinners LH 1962 Synopsis of Conradina (Labiatae). SIDA Contrib Bot 1:84-88. Simmons MP, H Ochoterena 2000 Gaps as characte rs in sequence-based phylogenetic analyses. Syst Biol 49:369-381. Slade RW, C Moritz, A Heideman 1994 Multipl e nuclear gene phylogenies: Application to pinnipeds and comparison with a mitoc hondrial DNA gene phylogeny. Mol Biol Evol 11:341-356. Slatkin M 1995 A Measure of Population S ubdivision Based on Micr osatellite Allele Frequencies. Genetics 139:457-462. Small RL, JA Ryburn, RC Cronn, T Seelanan, JF Wendel 1998 The tortoise and the hare: Choosing between noncoding plastome and nuclear Adh sequences for phylogeny reconstruction in a recently diverg ed plant group. Am J Bot 85:1301-1315. Small RL, RC Cronn, JF Wendel 2004 Use of nuc lear genes for phylogeny reconstruction in plants. Aust Syst Bot 17:145-170. Soltis DE, AB Morris, JS McLachlan, PS Manos, PS Soltis 2006 Comparative phylogeography of unglaciated eastern North America. Mol Ecol 15:4261-4293. Sota T, AP Vogler 2003 Reconstructing species phylogeny of th e carabid beetles Ohomopterus using multiple nuclear DNA sequences: Heterogeneous information content and the performance of simultaneous analys es. Mol Phylogenet Evol 26:139-154. Stephens M, NJ Smith, P Donnelly 2001 A new st atistical method for ha plotype reconstruction from population data. Am J Hum Genet 68:978-989. Strand AE, J LeebensMack, BG Milligan 1997 Nuclear DNA-based markers for plant evolutionary biology. Mol Ecol 6:113-118. Swofford DL, DP Begle 1993 Phylogenetic An alysis Using Parsimony Version 3.1. User's Manual. Illinois Natural History Survey, Champaign, Illinois. 257 pp. Swofford DL 2002 PAUP*. Phylogenetic Analys is Using Parsimony (* and other methods), version 4.10b10. Sinauer Asso ciates, Sunderland, MA.

PAGE 165

165 Taberlet P, L Gielly, G Pa utou, J Bouvet 1991 Universal Prim ers for Amplification of 3 Noncoding Regions of Chloroplast DNA. Plant Mol Biol 17:1105-1109. Tamura K, M Nei 1993 Estimation of the Number of Nucleotide Substitutions in the Control Region of Mitochondrial-DNA in Humans and Chimpanzees. Mol Biol Evol 10:512-526. Tate JA, BB Simpson 2003 Paraphyly of Tarasa (Malvaceae) and diverse origins of the polyploid species. Syst Bot 28:723-737. Thompson JD, TJ Gibson, F Plewniak, F Jean mougin, DG Higgins 1997 The ClustalX windows interface: Flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res:4876-4882. Trusty JL, RG Olmstead, DJ Bogler, A Santos -Guerra, J Francisco-Ortega 2004 Using molecular data to test a biogeographic conne ction of the Macaronesian genus Bystropogon (Lamiaceae) to the New World: A case of conflicting phylogenies. Systematic Botany 29:702-715. USFWS 1994 Recovery plan fo r Apalachicola rosemary ( Conradina glabra ). Atlanta, U. S. Fish and Wildlife Service. ——— 1996 Recovery Plan for Nineteen Centra l Florida Scrub and High Pineland Plants (Revised). Atlanta, GA, U. S. Fish and Wildlife Service. Wagner A, N Blackstone, P Cartwright, M Dick, B Misof, P Snow, GP Wagner, J Bartels, M Murtha, J Pendleton 1994 Surveys of gene families using polymerase chain-reaction: PCR Selection and PCR Drif t. Syst Biol 43:250-261. Wagstaff SJ, RG Olmstead, PD Cantino 1995 Pa rsimony analysis of cpDNA restriction site variation in subfamily Nepetoid eae (Labiatae). Am J Bot 82:886-892. Wall DP 2002 Use of the nuclear gene glyc eraldehyde 3-phosphate dehydrogenase for phylogeny reconstruction of recently diverged lineages in Mitthyridium (Musci: Calymperaceae). Mol Phylogenet Evol 25:10-26. Wen J, EA Zimmer 1996 Phylogeny and biogeography of Panax L (the ginseng genus, Araliaceae): Inferences from ITS sequences of nuclear ribosomal DNA. Mol Phylogenet Evol 6:167-177.

PAGE 166

166 Wendel JF, JJ Doyle 1998 Phylogenetic incongrue nce: window into genome history and molecular evolution, Pages 265-296 in D. E. Soltis, P. S. Soltis, and J. J. Doyle, eds. Molecular systematics of plants II: DNA sequencing. Boston, Kluwer Academic Publishers. Wiens JJ 1998 Combining data sets with differe nt phylogenetic histories. Syst Biol 47:568-581. ——— 1999 Polymorphism in systematics and comparative biology. Annu Rev Ecol Syst 30:327-362. Won YJ, Y Wang, A Sivasundar, J Raincrow, J He y 2006 Nuclear gene vari ation and molecular dating of the cichlid species flock of Lake Malawi. Mol Biol Evol 23:828-837. Wright S 1943 Isolation by distance. Genetics 28:114-138. ——— 1946 Isolation by distance under dive rse systems of mating. Genetics 31:39-59. ——— 1965 The Interpretation of Popul ation-Structure by F-Statistics with Special Regard to Systems of Mating. Evolution 19:395-420. Wunderlin R, D Richardson, B Ha nsen 1980 Status report on Conradina brevifolia Jacksonville, U. S. Fish and Wildlife Service. Wunderlin R 1982 Guide to the vascul ar plants of central Florida. University Presses of Florida, Gainesvillepp. ——— 1998 Guide to the vascular plants of Florida. Gainesville, University Presses of Florida. Yang ZH 1994 Maximum likelihood phylogenetic esti mation from DNA sequences with variable rates over sites: approximate methods. J Mol Evol 39:306-314. Yang ZH, B Rannala 1997 Bayesian phylogenetic inference using DNA sequences: A Markov Chain Monte Carlo method. Mol Biol Evol 14:717-724. Yoder AD, JA Irwin, BA Payseur 2001 Failure of the ILD to determine data combinability for slow loris phylogeny. Syst Biol 50:408-424.

PAGE 167

167 Zhang DX, GM Hewitt 2003 Nuclear DNA analyses in genetic studies of populations: practice, problems and prospects. Mol Ecol 12:1687-1687. Zhang Y, TH Niu, JS Liu 2006 A coalescenceguided hierarchical Bayesian method for haplotype inference. Am J Hum Genet 79:313-322.

PAGE 168

168 BIOGRAPHICAL SKETCH Christine E. Edwards was born in Denver, CO and lived there with her family until 1988, when they moved to Vancouver, WA. Christin e then moved to Houston, TX in 1991 with her family and attended I.H. Kempner High School in Sugarland, TX, wher e she graduated cum laude in 1995. She attended the Universi ty of Colorado at Boulder from 1995-1999 and graduated with a Bachelor of Arts in 1999 w ith majors in environmental, population, and organismic biology and Spanish literature. Afte r holding seasonal field te chnician positions in Massachusetts and Florida, Chri stine began her PhD at the Univ ersity of Florida in 2001 under the supervision of Drs. Pamela and Douglas Soltis, where she studi ed the evolutionary relationships in a group of endange red southeastern U.S. mints. Christine graduated with a PhD in Botany in December 2007, and began a postd octoral research position in January 2008 studying the genetic architecture of flow ering time and circadian genes in Brassica under the supervision of Cynthia Weinig at the University of Wyoming.