A MOLECULAR PHYLOGENY OF THE LYTHRACEAE AND INFERENCE OF THE EVOLUTION OF HETEROSTYLY
A dissertation submitted to Kent State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy
By
Julie A. Morris
August 2007 Dissertation written by Julie A. Morris B.S., Mesa State College, 1995 M.S., Kent State University, 2002 Ph.D., Kent State University, 2007
Approved by
______, Chair, Doctoral Dissertation Committee Andrea L. Case ______, Members, Doctoral Dissertation Committee Andrea E. Schwarzbach ______W. Randolf Hoeh ______L. Gwenn Volkert ______Alison J. Smith
Accepted by
______, Chair, Department of Biological Sciences James L. Blank ______, Dean, College of Arts and Sciences John R. D. Stalvey
ii TABLE OF CONTENTS
LIST OF FIGURES ……………………………………………………………………v
LIST OF TABLES ……………………………………………………………………vi
ACKNOWLEDGMENTS …………………………………………………………...vii
CHAPTER
I. A Molecular Phylogeny of the Lythraceae (Myrtales) Based on Combined Analysis of Five Chloroplast Regions and ITS
Abstract……………………………………………………………1 Introduction………………………………………………………..3 Methods……………………………………………………………9 Taxon Sampling and DNA Sequencing…………………...9 Data Analysis…………………………………………….15 Results…..………………………………………………………..17 Individual Data Sets……………………………………...17 Combined Data Sets……………………………………...21 Discussion………………………………………………………..29 Data Sets and Analysis………………………………...…29 Relationships Among Taxa………………………………30 Literature Cited…………………………………………………..36
II. A Molecular Phylogeny of Lythrum (Lythraceae): Preliminary Analyses Based on the atpB-rbcL Intergenic Spacer and ITS
Abstract…………………………………………………………..39 Introduction………………………………………………………41 Methods…………………………………………………………..48 Taxon Sampling and DNA Sequencing………………….48 Data Analysis…………………………………………….52 Results……………………………………………………………53 Discussion………………………………………………………..60 Peplis…………………………………………………...... 60 iii Relationships Within Lythrum…………………………...62 North American Clade…………………………...63 L. junceum –L. hyssopifolia Clade……………….65 L. virgatum – L. salicaria Clade…………………66 Conclusions………………………………………………………67 Literature Cited…………………………………………………..69
III. Evolution of Heterostyly in the Lythraceae (Myrtales): Approaches to Ancestral Character State Reconstruction Using a New Molecular Phylogeny
Abstract.………………………………………………………….72 Introduction………………………………………………………74 Methods…………………………………………………………..80 Taxon Sampling and Molecular Data……………………80 Ancestral Character State Reconstruction……………….83 Results……………………………………………………………86 Discussion………………………………………………………..96 Limitations of Current Models…………………………...96 Repeated Loss of Heterostyly in Lythraceae………….…99 Literature Cited…………………………………………………104
iv LIST OF FIGURES
1-1. Distribution of the Lythraceae…………………………………………………….5
1-2. Results of Analyses of Individual Data Sets……………………………………..19
1-3. Results of Combined Analysis of Data Sets with No Missing Data……………..23
1-4. Results of Combined Analysis of All Available Data…………………………...25
1-5. Selected Morphological Characters Mapped onto the Phylogeny……………….34
2-1. Botanical Illustration of Lythrum salicaria………………………………………44
2-2. Results of Combined Analysis…………………………………………………...55
2-3. Results of Combined Analysis with Branch Lengths …………………………...59
3-1. Schematic representation of Heterostyly………………………………………...76
3-2. Maximum Likelihood Character Reconstruction………………………………..88
3-3. Maximum Parsimony Optimization; Equally Weighted…………………………91
3-4. Maximum Parsimony Optimization; Weighting Gains to Losses 2:1…………...93
3-5. Maximum Parsimony Optimization; Weighting Gains to Losses 3.14:1………..95
v LIST OF TABLES
1-1. Sample List with Voucher Information and GenBank Accession Numbers…….10
1-2. PCR and Sequencing Primers……………………………………………………14
1-3. Comparative Statistics for Maximum Parsimony Analyses……………………..20
1-4. Summary Statistics for Pair-wise Topology Tests……………………………….26
2-1. Current Species List for Lythrum………………………………………………..46
2-2. Sample List with Voucher Information and GenBank Accession Numbers…….49
2-3. PCR and Sequencing Primers……………………………………………………51
2-4. Comparative Statistics for Maximum Parsimony Analyses……………………..56
3-1. Sample List and Character Coding Information…………………………………81
vi ACKNOWLEDGEMENTS
During my work on this dissertation, a substantial number of people have helped me in a great number of ways. In fact, properly acknowledging all of them would likely take almost as much time as the dissertation itself. However (as anyone who knows me will not be surprised to hear), I am preparing these comments with very (very) little time remaining prior to the submission deadline, and so am forced to be brief☺.
I am extremely grateful to Shirley Graham, Andrea Schwarzbach, and Andrea
Case, without whom this work would not have been possible. Most students would consider themselves fortunate to have had one great advisor during the course of their
graduate career, and I was privileged enough to have three!! I should also acknowledge
Barbara Andreas and Oscar Rocha who regularly and willingly provided advice and
support. All of these people contributed a great deal to my professional development at
Kent State University, and I am honored to count them as mentors and friends.
Many thanks are also due: to Eric C. for his gifted (and patient) technical
assistance in all of my analyses; to Holger and the Vals for making our lab a fun place to
work; to Andrea, Diana, Esther, Rajlakshmi, Maria, Jenn C., and Eric F. (and his magic
coffee) for holding me together during “the final days”; to Diana, Esther, and Raj for
providing defense refreshments that have already become legendary; to Chris, Sharon,
Pat, Robin, Gail, and Linda -without whom the biology department would not function;
to Chris R. for keeping the green things growing; to Joe, Pam, Min, Ken, Michelle, John,
vii Darrel, and Dale for keeping the “good old days” alive; to Jenn C., Justin M., Joel, Eric
F., Justin R., Julie P., Steph, Dan, Steve, Doug, and my entire PCC crew for reminding
me how to stay young; to Oscar and all of the Costa Rica Crew for an experience I will
never forget; to Diana and Brian for being there from start to finish; to Holger, Ana, Jen,
Esther, Philip, Laura, Raj, Amiya, Oscar, Maria, Andrea, Pat, and Mason for great
friendships, great conversations, and a lot of great food!!; to DW and Jeannie (among
“what I will miss most about Kent”) for their friendship and support (including, but not
limited to: dog-walking, orchid-watering, door-opening, car-starting, movie-going, and postcard-sending); to Al for all he has done and continues to do, and to Tanya, Porter,
Polly, Mark and Evan for all of the laughs and all of the hugs. I have been extraordinarily blessed with the friendships I have made during my time at Kent State,
and I will never forget the lab meetings, field days, journal clubs, floral hikes, MEEC
meetings, Thanksgiving meals, Darwin parties, kitty rescues, beer fests, Spanish lunches,
knitting nerds, Blossom nights, mental breakdowns, afternoon tea times, Ska concerts,
pot-lucks, Bronco games, backyard barbecues, bird-watching trips, Wednesday dinners,
walks in the woods, AND countless “crawls” down Franklin Ave. Thank you all!!!!!!!!
Finally, special thanks are due to Towner’s, Beetle, Rana, and the Boys -the
source of my sanity; and to my family –the source of all of my good qualities: to my Dad
from whom I gained focus and strength; to my Mom from whom I gained patience and curiosity; and to my brothers from whom I gained my sense of humor. There is nothing
that I have accomplished, or will achieve, that would be possible without them. viii
CHAPTER 1
A Molecular Phylogeny of Lythraceae (Myrtales) Based on Combined Analysis of Five Chloroplast Regions and ITS
Abstract
The Lythraceae (Myrtales) comprises approximately 32 genera and 600 species with a worldwide distribution. Members of the family show extreme variation in habit, ranging from tall trees and woody shrubs to small aquatic herbs. This vegetative variation, combined with a fairly generalized floral morphology, has complicated inferences of relationships among these genera based on traditional morphological characters. Previous molecular analyses have been unable to distinguish among various hypotheses of basal relationships within the family, making it difficult to correctly polarize characters for the study of morphological evolution. In this study, molecular data from two newly sequenced chloroplast regions (atpB-rbcL intergenic spacer and the trnK-matK region) are added to previous sequence data (rbcL, trnL-trnF, psaA-ycf3, and
ITS) and analyzed using maximum parsimony, maximum likelihood, and Bayesian approaches. Topologies resulting from these analyses are congruent and generally well resolved. The relationships of terminal clades are more stable than in previous analyses, however, statistical support for basal relationships remains relatively low, possibly suggesting an early rapid radiation among these taxa. It is clear that multiple
1 2 convergences in morphological characters and life history traits have occurred during the evolution of this family.
3
Introduction
As currently defined, the Lythraceae includes approximately 32 genera and 600 species (Graham et al. 2005). The family has been traditionally classified in the order
Myrtales and closely allied with the Onagraceae based on morphological, anatomical, and embryological evidence (Dahlgren and Thorne, 1984; Johnson and Briggs, 1984).
Subsequent molecular analyses have confirmed these hypotheses (Conti et al., 1996,
1997; Sytsma et al., 2004). Earliest fossil evidence of the family dates from the late
Cretaceous to early Paleocene (see review in Graham et al., 2005), however divergence dates estimated from molecular analyses of the Myrtales suggest a considerably older origin, closer to the Early Cretaceous (Sytsma et al., 2004). Both fossil and molecular evidence support a relatively early establishment of the family, followed by a rapid expansion, and then widespread isolation and extensive extinction (Graham et al., 2005), resulting in a worldwide distribution (Figure 1-1) of generally small and very distinctive genera ranging from trees and woody shrubs to small aquatic herbs. A majority of the genera in the family are either monotypic or ditypic, and some of them have been treated as distinct subfamilies within the Lythraceae, or even as separate monogeneric families
(Koehne, 1903; Cronquist 1981; Takhtajan, 1987; Dahlgren and Thorne, 1984; Johnson and Briggs, 1984). The extreme variation in habit found within the family combined with a fairly generalized floral morphology has made determining the relationships among these genera very difficult.
The only monograph of the family (Koehne 1903) includes 22 genera distinguished by opposite entire leaves, bell-shaped to cylindrical flowers with stamens,
4
Figure 1-1. Distribution of the Lythraceae, with the number of genera/number of endemic genera indicated for the following areas: United States and Canada; Central America, and the Antilles; South America; western Europe; eastern Europe and northern Asia, including Japan; Africa; Madagascar and Mauritius; southeastern Asia and the Pacific Islands; Australia (reprinted with permission from S. Graham (Graham et al., 2005)).
5
6 petals, and sepals fused to a floral tube surrounding a superior ovary, two whorls of deeply set stamens, and a many-seeded dry capsular fruit. This early classification divided the family into two tribes (Lythreae and Nesaeeae), distinguished on the basis of complete versus incomplete septation of the ovary, and four subtribes (Lythrinae,
Diplusodontinae, Nesaeinae, and Lagerstroemiinae), based primarily on the presence or absence of wings on the seeds. Subsequent morphological studies have shown these divisions to be largely artificial and based on nonexclusive or erroneous character descriptions (Tobe et al., 1998, Graham et al., 1993).
Three very distinctive genera (Sonneratia, Duabanga, and Punica), were excluded from previous classifications primarily on the basis of their semi-inferior or inferior ovaries. In Koehne’s classification (Koehne 1881), he excluded these genera, recognizing instead the families Sonneratiaceae (Sonneratia and Duabanga) and
Punicaceae (Punica), and he suggested that they were closely allied to the Lythraceae and
Onagraceae. Due to its extremely divergent morphology, the genus Trapa was also placed in its own monogeneric family, Trapaceae. Until recently, this remained the most common treatment of these taxa (Melchior, 1964; Hutchinson,1973; and Cronquist,
1981). Based on increasing anatomical, palynological, and embryological data, other authors variously included Sonneratia, Duabanga, and Punica as subfamilies within the
Lythraceae (Dahlgren, 1975; Dahlgren and Thorne, 1984; Thorne, 1981 and 1992), an idea that was supported by the first comprehensive cladistic analyses of morphological characters in the Myrtales (Johnson and Briggs, 1984), and in the Lythraceae (Graham et al., 1993).
7
In the first broad molecular study including many of these taxa, Conti et al.,
(1997) compared sequences of the chloroplast gene rbcL for several genera in the
Myrtales including Trapa, ten members of the Onagraceae, and seven genera in the
Lythraceae. Their analyses found strong support for the sister relationship between
Lythraceae and Onagraceae (92% bootstrap), and surprising evidence that a
monophyletic Lythraceae should include Trapa, in addition to Duabanga and Punica
(Sonneratia was not included in this study). This unexpected result has since been
confirmed by further molecular work in the family (Shi et al., 2000; Huang and Shi,
2002; Graham et al., 2005).
Huang and Shi (2002) sequenced two chloroplast regions (rbcL and psaA-ycf3)
and nuclear ITS for 16 Lythraceae genera including Trapa, Sonneratia, Duabanga, and
Punica. Their analyses found strong support for a monophyletic, more broadly defined
Lythraceae (98% bootstrap), and suggested that these genera are actually relatively derived members of the family. Well-supported sister-group relationships (>70%
bootstrap) were found for Punica and Duabanga with the “classical” Lythraceae genera
Pemphis and Lagerstroemia, respectively; Sonneratia and Trapa were very unexpectedly
found to be sister to one another (89% bootstrap). Various other terminal clades were
also well supported, however deeper nodes in the phylogeny received mainly weak
support.
In an effort to confirm these findings and gain further understanding of the basal
relationships and character evolution in the family, Graham et al., (2005) expanded their
analysis to include all but four Lythraceae genera, an additional chloroplast region (trnL-
8
trnF), and morphological characters. This study provided unambiguous molecular support for the inclusion of Duabanga, Sonneratia, Punica, and Trapa in the family,
increased support for a majority of the terminal clades, well-supported placement of many of the additional taxa, and a significantly improved understanding of how
morphological characters may be distributed throughout the family. However, these
analyses were unable to distinguish between three conflicting hypotheses of basal
relationships in the family: 1) Decodon is basal, and sister to the remainder of the family;
2) the base of the family diverges as two superclades; or 3) more than two lineages
emerge at the base of the family. This lack of resolution makes it difficult to assign
intrafamilial classification and correctly polarize characters in order to study their
evolution.
In this study, molecular data from two newly sequenced chloroplast regions, atpB-rbcL intergenic spacer and the trnK-matK region, are added to previous sequence
data (rbcL, trnL-trnF, and psaA-ycf3, and ITS) and analyzed using maximum parsimony
and Bayesian approaches. The main objectives are: 1) to distinguish among competing
hypotheses of basal relationship in the family, 2) to produce a consistent, well-supported
phylogeny that can be used as a basis for an intrafamilial classification of the Lythraceae
reflecting phylogenetic relationships; and 3) to evaluate current hypotheses of character
evolution and biogeography in light of these new phylogenetic hypotheses.
9
Methods
Taxon Sampling and DNA Sequencing
All but three of the 32 proposed genera of Lythraceae were represented in these analyses, including Didiplis that had not been previously sampled; missing taxa include
Crenea, Haitia, and Hionanthera. The data set is complete for 18 genera. Select members of the Onagraceae, previously shown to be closely allied to the Lythraceae
(Johnson and Briggs 1984, Conti et al., 1997, Sytsma et al., 2004), were chosen as outgroups.
In order to maximize taxon sampling and consider as much of the available data as possible, some sequences (rbcL, trnL-trnF, and psaA-ycf3, and ITS) were retrieved from GenBank (Huang and Shi, 2002; Graham et al., 2005). Some GenBank sequences were incomplete, making it necessary to resequence some taxa for rbcL and trnL-trnF.
Additional taxa were also resequenced in an effort to reduce the possibility of sequencing error as a source of noise in the data set, especially given that several previously published sequences were produced with manual sequencing techniques (Huang and Shi,
2002; Graham et al., 2005). All new sequences, and the earlier sequences from
GenBank, for all taxa and gene regions are provided in Table 1-1.
Total DNA used to generate new sequence data for this study was extracted from fresh tissue, dried material preserved in silica gel, or dried herbarium specimens (voucher information is provided in Table 1-1). DNA extractions were conducted using a modified
CTAB protocol (as in Graham et al., 2005), or the QIAGEN DNeasy Plant Mini Kit
(QIAGEN Inc., Chatsworth, CA). Some difficulties exist with DNA extraction in this
10
Table 1-1. Information for all plant samples included in this study, including taxon name, GenBank accession numbers, and voucher location. * = sequence from another study, downloaded from
used in Combined used in Combined Taxon DNA Region GenBank Accession # Analysis #1 Analysis #2 Voucher Adenaria floribunda atpB-rbcL waiting submission X X Breedlove 38133 (CAS) Adenaria floribunda rbcL waiting submission X X Breedlove 38133 (CAS) Adenaria floribunda trnLF waiting submission X X Breedlove 38133 (CAS) Adenaria floribunda MatK waiting submission X X Breedlove 38133 (CAS) Ammannia latifolia atpB-rbcL waiting submission X X Liogier 10314 (MO) Ammannia latifolia rbcL waiting submission X X Liogier 10314 (MO) Ammannia latifolia trnLF waiting submission X X Liogier 10314 (MO) Ammannia latifolia MatK waiting submission X X Liogier 10314 (MO) Ammannia baccifera * psaA-ycf3 AY035733 X Tang, S. Q. 99010301 (SYS) Ammannia baccifera * ITS AY905419 X Tang, S. Q. 99010301 (SYS) Capuronia madagascariensis atpB-rbcL waiting submission X D’Arcy 15439 (MO) Capuronia madagascariensis * rbcL AY905405 X D'Arcy 15439 (MO) Capuronia madagascariensis trnLF waiting submission X D’Arcy 15439 (MO) Capuronia madagascariensis MatK waiting submission X D’Arcy 15439 (MO) Capuronia madagascariensis * psaA-ycf3 AY905435 X D’Arcy 15439 (MO) Capuronia madagascariensis * ITS AY905420 X D’Arcy 15439 (MO) Cuphea lanceolata * rbcL AY036137 X Shi 99090201 (SYS) Cuphea lanceolata * psaA-ycf3 AY035723 X Shi 99090201 (SYS) Cuphea lanceolata * ITS AY035763 X Shi 99090201 (SYS) Decodon verticillatus atpB-rbcL waiting submission X X Graham 917 (MO) Decodon verticillatus rbcL waiting submission X X Graham 917 (MO) Decodon verticillatus trnLF waiting submission X X Graham 917 (MO) Decodon verticillatus MatK waiting submission X X Graham 917 (MO) Decodon verticillatus * psaA-ycf3 AY035728 X Hill, Steven R. 18871 (A) Decodon verticillatus * ITS AY035752 X Hill, Steven R. 18871 (A) Didiplis diandra atpB-rbcL waiting submission X X my collection 1/28/2004 -preparing voucher Didiplis diandra rbcL waiting submission X X my collection 1/28/2004 -preparing voucher Didiplis diandra trnLF waiting submission X X my collection 1/28/2004 -preparing voucher Didiplis diandra MatK waiting submission X X my collection 1/28/2004 -preparing voucher Duabanga molucanna atpB-rbcL waiting submission X X Chai s.n. in 1990, (MO) Duabanga molucanna rbcL waiting submission X X Chai s.n. in 1990, (MO) Duabanga molucanna trnLF waiting submission X X Chai s.n. in 1990, (MO) Duabanga molucanna MatK waiting submission X X Chai s.n. in 1990, (MO) Duabanga grandiflora * psaA-ycf3 AY035738 X Huang, S.D. 990401 (SYS) Duabanga grandiflora * ITS AF163695 X Huang, S.D. 990401 (SYS) Galpinia transvaalica atpB-rbcL waiting submission X X Balsinhas 3263 (MO) Galpinia transvaalica rbcL waiting submission X X Balsinhas 3263 (MO) Galpinia transvaalica trnLF waiting submission X X Balsinhas 3263 (MO) Galpinia transvaalica MatK waiting submission X X Balsinhas 3263 (MO) Galpinia transvaalica * psaA-ycf3 AY905443 X Balsinhas 3263 (MO) Galpinia transvaalica * ITS AY905423 X Balsinhas 3263 (MO) Ginoria americana atpB-rbcL waiting submission X S. Graham 1137 (MO) Ginoria americana rbcL waiting submission X S. Graham 1137 (MO) Ginoria americana trnLF waiting submission X S. Graham 1137 (MO) Ginoria americana * psaA-ycf3 AY905437 X S. Graham 1127 (MO) Ginoria americana * ITS AY078421 X S. Graham 1127 (MO) Heimia salicifolia atpB-rbcL waiting submission X X S. Graham 1062 (MO) Heimia salicifolia rbcL waiting submission X X S. Graham 1062 (MO) Heimia salicifolia trnLF waiting submission X X S. Graham 1062 (MO) Heimia salicifolia MatK waiting submission X X S. Graham 1062 (MO) Heimia myrtifolia * psaA-ycf3 AY035735 X Tang, S. Q. 99070502 (SYS) 11
Table 1-1 (cont).
used in Combined used in Combined Taxon DNA Region GenBank Accession # Analysis #1 Analysis #2 Voucher Heimia myrtifolia * ITS AF201693 X Tang, S. Q. 99070502 (SYS) Koehneria madagascariensis atpB-rbcL waiting submission X X D’Arcy & Rakotozafy 15317 (MO) Koehneria madagascariensis rbcL waiting submission X X D’Arcy & Rakotozafy 15317 (MO) Koehneria madagascariensis trnLF waiting submission X X D’Arcy & Rakotozafy 15317 (MO) Koehneria madagascariensis MatK waiting submission X X D’Arcy & Rakotozafy 15317 (MO) Koehneria madagascariensis * psaA-ycf3 AY605444 X D’Arcy & Rakotozafy 15317 (MO) Koehneria madagascariensis * ITS AY905424 X D’Arcy & Rakotozafy 15317 (MO) Lafoensia acuminata atpB-rbcL waiting submission X X Neil 8930 (MO) Lafoensia acuminata rbcL waiting submission X X Neil 8930 (MO) Lafoensia acuminata trnLF waiting submission X X Neil 8930 (MO) Lafoensia acuminata MatK waiting submission X X Neil 8930 (MO) Lafoensia acuminata * psaA-ycf3 AY905438 X Neil 8930 (MO) Lafoensia acuminata * ITS AY905425 X Neil 8930 (MO) Lagerstroemia indica atpB-rbcL waiting submission X X USA: Texas, cultivated, no voucher Lagerstroemia indica rbcL waiting submission X X USA: Texas, cultivated, no voucher Lagerstroemia indica trnLF waiting submission X X USA: Texas, cultivated, no voucher Lagerstroemia indica MatK waiting submission X X USA: Texas, cultivated, no voucher Lagerstroemia speciosa * psaA-ycf3 AY035737 X Shi, S. H. 99060103 (SYS) Lagerstroemia indica * ITS AF201689 X Shi, S. H. 9963001 (SYS) Lawsonia inermis atpB-rbcL waiting submission X X Correll 45915 (TEX) Lawsonia inermis rbcL waiting submission X X Correll 45915 (TEX) Lawsonia inermis trnLF waiting submission X X Correll 45915 (TEX) Lawsonia inermis MatK waiting submission X X Correll 45915 (TEX) Lawsonia inermis * psaA-ycf3 AY035734 X Qiu, H. X. 99070201 (SYS) Lawsonia inermis * ITS AY078424 X Correll 45915 (TEX) Lourtella resinosa * psaA-ycf3 AY905445 X S. Graham 1116 (MO) Lourtella resinosa * ITS AY905427 X S. Graham 1116 (MO) Lythrum ovalifolium atpB-rbcL waiting submission X X M. C. Johnston s.n. in Oct. 1975 (MO) Lythrum ovalifolium rbcL waiting submission X X M. C. Johnston s.n. in Oct. 1975 (MO) Lythrum ovalifolium trnLF waiting submission X X M. C. Johnston s.n. in Oct. 1975 (MO) Lythrum ovalifolium MatK waiting submission X X M. C. Johnston s.n. in Oct. 1975 (MO) Lythrum ovalifolium ITS waiting submission X M. C. Johnston s.n. in Oct. 1975 (MO) Lythrum salicaria * psaA-ycf3 AY035727 X Lei, Y.B. 005 (SYS) Lythrum salicaria * psaA-ycf3 AF421495 X Tsang, W.T. 27844 (IBSC) Lythrum virgatum atpB-rbcL waiting submission X X cultivated, Gottingen Bot. Garden, no voucher Lythrum virgatum rbcL waiting submission X X cultivated, Gottingen Bot. Garden, no voucher Lythrum virgatum trnLF waiting submission X X cultivated, Gottingen Bot. Garden, no voucher Lythrum virgatum MatK waiting submission X X cultivated, Gottingen Bot. Garden, no voucher Lythrum virgatum ITS waiting submission X cultivated, Gottingen Bot. Garden, no voucher Nesaea aspera atpB-rbcL waiting submission X X Drummond 11446 (SRGH) Nesaea aspera rbcL waiting submission X X Drummond 11446 (SRGH) Nesaea aspera trnLF waiting submission X X Drummond 11446 (SRGH) Nesaea aspera MatK waiting submission X X Drummond 11446 (SRGH) Nesaea aspera * psaA-ycf3 AY905441 X Drummond 11446 (SRGH) Nesaea aspera * ITS AY905429 X Drummond 11446 (SRGH) Pehria compacta atpB-rbcL waiting submission X X P. Berry s.n. in 1979 (MO) Pehria compacta rbcL waiting submission X X P. Berry s.n. in 1979 (MO) Pehria compacta trnLF waiting submission X X P. Berry s.n. in 1979 (MO) Pehria compacta MatK waiting submission X X P. Berry s.n. in 1979 (MO) Pehria compacta * psaA-ycf3 AY905446 X P. Berry s.n. in 1979 (MO) Pehria compacta * ITS AY905430 X P. Berry s.n. in 1979 (MO) Pemphisp acidula * rbcL AY036138 X Liao 1150 (A) 12
Table 1-1 (cont).
used in Combined used in Combined Taxon DNA Region GenBank Accession # Analysis #1 Analysis #2 Voucher Pemphis acidula * psaA-ycf3 AY035725 X Liao 1150 (A) Pemphis acidula * ITS AY035762 X Liao 1150 (A) Peplis portula atpB-rbcL waiting submission X X Montezuma s.n. (MO) Peplis portula rbcL waiting submission X X Montezuma s.n. (MO) Peplis portula trnLF waiting submission X X Montezuma s.n. (MO) Peplis portula MatK waiting submission X X Montezuma s.n. (MO) Peplis portula * psaA-ycf3 AY035726 X Montezuma s.n. (KE) Peplis portula ITS waiting submission X Montezuma s.n. (MO) Physocalymma scaberrimum * psaA-ycf3 AY905448 X Shi, S. H. 861 (SYS) Physocalymma scaberrimum * ITS AY905432 X Shi, S. H. 861 (SYS) Pleurophora anomala atpB-rbcL waiting submission X X Cavalcanti et al. 368 (MO) Pleurophora anomala rbcL waiting submission X X Cavalcanti et al. 368 (MO) Pleurophora anomala trnLF waiting submission X X Cavalcanti et al. 368 (MO) Pleurophora anomala MatK waiting submission X X Cavalcanti et al. 368 (MO) Punica granatum atpB-rbcL waiting submission X X cultivated, no voucher Punica granatum rbcL waiting submission X X cultivated, no voucher Punica granatum trnLF waiting submission X X cultivated, no voucher Punica granatum MatK waiting submission X X cultivated, no voucher Punica granatum * psaA-ycf3 AY035742 X Wang, J.B. 00-041602 Punica granatum * ITS AY035761 X Wang, J.B. 00-041602 Rotala ramosior atpB-rbcL waiting submission X S. Graham 1028 (MO) Rotala ramosior * rbcL AY905417 X S. Graham 1028 (MO) Rotala ramosior trnLF waiting submission X S. Graham 1028 (MO) Rotala ramosior MatK waiting submission X S. Graham 1028 (MO) Rotala indica * psaA-ycf3 AY035736 X Tang, S. Q. 99070503 (SYS) Rotala indica * ITS AF420220 X Jian, S.G. 200508 (SYS) Sonneratia caseolaris * rbcL AY036143 X Huang 990435 (SYS) Sonneratia caseolaris * psaA-ycf3 AY035731 X Huang 990435 (SYS) Sonneratia caseolaris * ITS AF208696 X Huang 990435 (SYS) Sonneratia ovata * trnLF AY905487 X no voucher -see Graham et al. 2005 Tetrataxis salicifolia * psaA-ycf3 AY905450 X Lorence 1231 (MO) Tetrataxis salicifolia * ITS AY078423 X Lorence 1231 (MO) Trapa maximowiczii * psaA-ycf3 AY035729 X Wang, J.B. 2000-041601 (SYS) Trapa maximowiczii * ITS AY035757 X Wang, J.B. 2000-041601 (SYS) Trapa natans * rbcL L10226 X no voucher -see Conti et al. 1993 Woodfordia fruticosa atpB-rbcL waiting submission X X cultivated, USDA Agricultural Station, Homestead, Fl. (MO) Woodfordia fruticosa rbcL waiting submission X X cultivated, USDA Agricultural Station, Homestead, Fl. (MO) Woodfordia fruticosa trnLF waiting submission X X cultivated, USDA Agricultural Station, Homestead, Fl. (MO) Woodfordia fruticosa MatK waiting submission X X cultivated, USDA Agricultural Station, Homestead, Fl. (MO) Woodfordia fruticosa * psaA-ycf3 AY035722 X Tang, S. Q. 99070504 (SYS) Woodfordia fruticosa * ITS AF201692 X Tang, S. Q. 99070504 (SYS) 13
family, and extended storage in the freezer (> 1 year) and/or substantial dilution is often
required prior to successful PCR (pers. obs.).
Information concerning all PCR and sequencing primers can be found in Table 1-
2. No additional sequencing was carried out for psaA-ycf3. PCR amplification was
carried out using standard PCR techniques and a MJ Research Thermocycler. Reactions
were performed in a volume of 25 µl containing 6 µl of a Tricine buffer mix (final
concentration: 30 mM Tricine pH 8.4, 2.0 mM MgCl2, 50 mM KCl, and 0.25 mM each
dNTP), 0.05 µg of each primer, 1 U Taq polymerase, and approximately 0.5 ng genomic
DNA. In a few cases, genomic DNA needed to be substantially diluted (0.005-0.05 ng/
25 µl reaction) before PCR was successful. The thermocycler program included 1 minute
for initial strand separation at 94˚C; followed by 37 cycles of 1 minute at 94˚C, 45
seconds at 50˚, 4 minutes at 72˚C; and a final 10 minute step at 72˚C. PCR products were
isolated using 1.5% agarose gels and purified using the QIAGEN QIAquick Gel
Extraction Kit (QIAGEN Inc., Chatsworth, CA). Cycle sequencing reactions were
conducted in a 10µl volume, using 2 µl ABI BigDye cycle sequencing terminator
chemistry (Perkin-Elmer Corp.), 2 µl Half Term Dye (Bioline Inc.), 2.0 ng each primer,
and approximately 50 ng of template DNA. Reactions were purified using sephadex in
Cetrisep columns according to manufacturer’s instructions, and then were completely
lyophilized using a ThermoSavant DNA120 SpeedVac. Sequencing was conducted on an
Applied Biosystems 3730 DNA Analyzer at the Sequencing Core Facility, Brigham
Young University, Provo, Utah. Completed sequences were edited, assembled, compiled
14
Table 1-2. Base composition and citation information for all PCR and sequencing primers used and/or developed for these analyses.
Primer Sequence 5'-3' Citation
*rbcL forward ATG TCA CCA CAA ACA GAA ACT AAA GC Fay et al. (1998)
*rbcL reverse CTT TTA GTA AAA GAT TGG GCC GAG Fay et al. (1998)
^rbcL INT1forward AGC CTG TTG CTG GAG AAG AA this study
^rbcL INT2reverse CCA AAG ATC TCG GTC AGA GC this study
*trnLF A forward CAT TAC AAA TGC GAT GCT CT Taberlet et al. (1991)
*trnLF F reverse ATT TGA ACT GGT GAC ACG AG Taberlet et al. (1991)
^trnLF B TCT ACC GAT TTC GCC ATA TC Taberlet et al. (1991)
^trnLF C CGA AAT CGG TAG ACG CTA CG Taberlet et al. (1991)
^trnLF D GGG GAT AGA GGG ACT TGA AC Taberlet et al. (1991)
^trnLF E GGT TCA AGT CCC TCT ATC C Taberlet et al. (1991)
*atpB-rbcL forward GAA GTA GTA GGA TTG ATT CTC Schwarzbach and Ricklefs (2000)
*atpB-rbcL reverse AGT TTC TGT TTG TGG TGA CAT Schwarzbach and Ricklefs (2000)
^atpB-rbcL INT1forward TGG TTC TTT ATT AGA CCA TGG TAT TTG this study
*trnK-matK forward GGG GGT GCT AAC TCA ACG G Mano and Steele (1997)
*trnK-matK reverse AAC TAG TCG GAT GGA GTA G Mano and Steele (1997)
^trnK-matK INT1reverse CGC ACA AAT CGG TCG ATA AT this study
^trnK-matK INT2reverse TTC TTC CAA TAA TTC CGA ACC this study
^trnK-matK INT3forward GGG AAA GAA AAA GCA ACG AG this study
^trnK-matK INT4forward TCA AAA GAG CGA TTG GGT TG this study
^trnK-matK INT5reverse TTG CCA TAA ATT GAC AAG GTA A this study
^trnK-matK INT6forward ATC TCG ACA CCA CGA CTT CC this study
*ITS A GGA AGG AGA AGT CGT AAC AAG Blattner (1999)
*ITS B CTT TTC CTC CGC TTA TTG ATA TG Blattner (1999) * For amplification and sequencing. ^ For sequencing only. 15
with GenBank sequences, and initially aligned using Sequencher 4.1 (GenesCode Co.
Ann Arbor, MI). Manual adjustments were then made to ensure proper alignment.
Data analysis
Phylogenetic analyses were conducted using maximum parsimony, maximum
likelihood and Bayesian approaches with various combinations of the molecular dataset.
Each of the data sets for which new sequences were produced were analyzed
individually, then combined (including psaA-ycf3 and ITS) in two ways: 1) including
only the 18 taxa for which all data sets were complete, resulting in no missing data
(‘CNM’); 2) including all sampled, but accepting some missing data (CM –including 28
taxa). Individual and combined non-partitioned data sets were analyzed using maximum
parsimony (MP) implemented in PAUP*4.0b10 (Swofford, 2001). Analysis methods
include: heuristic searches using random stepwise addition with 1000 replicates, tree
bisection reconnection (TBR) branch-swapping, accelerated transformation (ACCTRAN) character optimization, MulTrees, and the ‘steepest descent’ option not selected.
Characters were unordered and equally weighted, and gaps were coded as missing.
Bootstrap analyses (Felsenstein, 1985) were conducted using a heuristic search and 1000 replicates. Tree length, consistency index (CI) excluding autapomorphic characters, and retention index (RI) were calculated.
Bayesian analyses were conducted for each of the combined non-partitioned data
sets using MrBayes (v.3.1.2; Huelsenbeck and Ronquist, 2003; Ronquist and
Huelsenbeck, 2005). Modeltest (v.3.5; Posada and Crandall, 1998) was employed to
determine which model best fit the data. Analyses were run using the general time
16 reversible + invariant gamma model (GTR+I+Γ) on eight chains for 5 million generations, saving trees every 500 generations. The first 1000 trees were discarded
(500,000 generations), and a 50% majority rule consensus tree was constructed from the remaining 9000 saved trees using PAUP*. Maximum likelihood analyses, including
1000 bootstrap replicates, were conducted using the default settings in GARLI (Genetic
Algorithm for Rapid Likelihood Inference,
(ww.bio.utexas.edu/faculty/antisense/garli/Garli.html; Zwickl, 2006).
To further distinguish among hypotheses of basal relationships in the family
(Graham et al., 2005), various pair-wise tests for significant differences between topologies were used to compare the consensus tree from unconstrained Bayesian analysis of the most-inclusive taxon sample (CM), with the consensus trees produced by constraining: 1) all Lythraceae taxa other than Decodon to the same clade, and 2) all
Lythraceae taxa other than Decodon, Lythrum and Peplis to the same clade. The parsimony-based Kishino-Hasegawa test (Kishino & Hasegawa, 1989), Templeton test
(Wilcoxon signed-ranks test; Templeton, 1983) and winning sites (sign) test (Prager &
Wilson, 1988), as well as the likelihood-based Kishino-Hasegawa and Shimodaira-
Hasegawa tests (Shimodaira & Hasegawa, 1999) were conducted using PAUP*.
17
Results
Individual Data Sets
The overall topologies obtained from each of the individual MP analyses are generally congruent with one another, however the degree of resolution and relationships among subclades varies (Figure 1-2). All of the data sets are congruent with respect to the following groupings: A) Adenaria, Pehria, Woodfordia, Koehneria, and Pleurophora
(including Cuphea in data sets where it was present); B) Capuronia, Galpinia, Lafoensia, and Punica (including Pemphis when present); C) Ammannia, Nesaea, and Lawsonia
(including Ginoria when present); D) Duabanga, Lagerstroemia, (and Sonneratia sister to Trapa when present); E) Peplis and Lythrum; and F) Rotala and Didiplis (and Heimia when present). The only incongruence supported by a bootstrap value greater than 70% occurs in the atpB-rbcL tree, where Koehneria and Pehria are clustered together. In other trees where these relationships are resolved, Koehneria is sister to Woodfordia, and
Pehria is sister to Adenaria. The relative position of Decodon varies in the analysis of each individual data set, but is never well supported.
Of the two newly sequenced data sets (trnK-matK and atpB-rbcL), trnK-matK has the greater number of parsimony-informative characters. However, when overall length is taken into account, the percent of informative characters was very similar to that found for trnL-trnF (Table 1-3). Additionally, the topology resulting from the individual analysis of trnK-matK better resolves basal relationships, and the topology produced is more similar to that found in the overall combined analyses than any of the other individual data sets (Figure 1-2D). In comparison, trnL-trnF (used in previous studies
18
Figure 1-2. 50% majority-rule consensus trees produced in maximum parsimony analyses of individual data sets. Bootstrap values > 50% shown above branches (red) (A) Consensus of 44 most parsimonious trees based on trnL-trnF data. (B) Consensus of 6 most parsimonious trees based on rbcL data. (C) Consensus of 292,026 most parsimonious trees based on atpB-rbcL data. (D) Consensus of 21 most parsimonious trees based on trnK-matK data. Tree lengths and summary statistics are shown in Table 1-3. 19
20
Table 1-3. Comparative statistics for maximum parsimony analyses of individual DNA regions, and combined analyses (presenting the number of Lythraceae genera included in the analysis, total aligned length, number and % of parsimony informative characters, number and length of most parsimonious trees, consistency index (CI), and retention index (RI)).
rbc L trn K-mat K trn L-trn F atp B-rbc L CNM* CM^
# of Lythraceae Genera 25 20 23 21 18 28
Aligned Length 1381 2640 1058 824 5903 6737 # of Parsimony Informative Characters 101 267 116 47 453 642
% Informative 7.31 10.11 10.96 5.7 7.67 9.53
# of Most Parsimonious Trees 6 21 44 292,026 14 5
Length of Most Parsimonious Tree 433 941 452 200 1727 2623
Consistency Index 0.468 0.645 0.713 0.624 0.631 0.560
Retention Index 0.59 0.731 0.795 0.693 0.713 0.645
* Combined analysis with no missing data. ^ Combined analysis accepting some missing data (also includes ITS and psa A-yc f3). 21 and reanalyzed here) has a slightly higher percentage of parsimony-informative characters (Table 1-3), but results in less overall resolution, especially of basal relationships (Figure 1-2). The atpB-rbcL region yielded the least amount of basal resolution (Figure 1-2), few parsimony-informative characters, and an exceptionally large number of equally parsimonious trees (Table 1-3). Variable characters in this data set primarily supported well-established terminal clades.
Combined Data Sets
As has been found in previous molecular phylogenetic studies of the Lythraceae
(Huang and Shi 2002, Graham et al., 2005), analyses of the combined data sets in this study resulted in fewer, more resolved trees with improved branch support compared to analyses of individual data sets. The MP and Bayesian analyses of each of the combined data sets were generally congruent (Figures 1-3 and 1-4). In both cases, the earliest branching event results in two major clades containing nearly equal numbers of taxa: I)
Adenaria-Pemphis (11 genera in CM) with 77% bootstrap support; and II) Ammannia-
Heimia (14 genera in CM) with < 50% bootstrap support. According to all pair-wise tests, this topology was statistically better than the topology obtained by constraining these major clades to exclude Decodon+Lythrum and Peplis (P < 0.0005), but not significantly better than the topology obtained by constraining the major clades to exclude only Decodon (P < 0.1547). Excluding Decodon required 10 extra steps, while excluding Decodon +Lythrum and Peplis required 18 extra steps (Table 1-4).
The combined analysis including taxa with some missing data (CM) differed from the combined analysis with no missing data (CNM) only in the placement of Decodon.
22
Figure 1-3. Majority-rule Bayesian consensus of 9000 trees based on the combined analysis of all Lythraceae taxa with data sets having no missing data (CNM). Posterior probabilities are shown above the branches and maximum parsimony bootstrap values > 50% are shown below the branches. Tree length and summary statistics are provided in Table 1-3.
23
24
Figure 1-4. Majority-rule Bayesian consensus of 9000 trees based on the combined analysis of all available molecular data (CM). Posterior probabilities are shown above the branches and maximum parsimony and (maximum likelihood) bootstrap values > 50% are show below the branches. Tree length and summary statistics are provided in Table 1-3. The composition of superclades I and II, and subclades A-F are indicated to the right of the tree. 25
26
Table 1-4. Summary statistics for all pair-wise topology tests used to evaluate alternative hypotheses of basal relationship in the Lythraceae. The consensus tree produced by an unconstrained Bayesian analysis of the most inclusive data set (CM) is compared to consensus trees produced by constraining all Lythraceae taxa other than Decodon to the same clade (Decodon basal), and all Lythraceae taxa other than Decodon , Lythrum , and Peplis to the same clade (Decodon +Lythrum /Peplis basal).
Tree Test
Parsimony-based tests Length Difference Kishino-Hasegawa Templeton Winning sites
CM Unconstrained 2627
Constrained 2637 10 P = .1139 P = .1138 P = .1547 (Decodon basal)
Constrained 2645 18 P = .0004* P = .0004* P = .0005* (Decodon + Lythrum /Peplis basal)
Liklihood-based tests -Ln L Difference Kishino-Hasegawa Shimodaira-Hasegawa
CM Unconstrained 24246.99
Constrained 24268.08 21.08 P = 0.004* P = 0.003* (Decodon basal)
Constrained 24268.10 21.1 P = 0.004* P = 0.003* (Decodon + Lythrum /Peplis basal) * P < 0.05 27
This taxon was placed with very weak support (PP=0.62 and BS<50%) sister to a clade
containing Duabanga and Lagerstroemia in CNM, and more strongly supported
(PP=0.78 and BS=76%) with Lythrum and Peplis in CM (Figures 1-3 and 1-4). The
Decodon/Lythrum/Peplis clade is better supported by morphology (Graham et al., 1993),
and is the more common placement of Decodon in other molecular analyses that have
included these genera (Huang and Shi, 2002; Graham et al., 2005). The placement of
Decodon with Duabanga and Lagerstroemia in CNM is very likely caused by signal
from the trnK-matK region, considering that this data set contributes 44.7% of the
characters in this analysis, and shows the same relationship when analyzed individually.
The relative placement of Decodon varies more than any other genus in the family,
depending on the gene region and type of analysis. The reason for this is unclear and will
need to be addressed using a better population-level sampling of the genus, as well as
more molecular data, which might help to better discriminate among alternative
hypotheses.
CNM resulted in a smaller number of most parsimonious trees, shorter trees, and
a slightly higher CI and RI (Table 1-3). Bootstraps and posterior probabilities (PP) were sometimes slightly higher when using the complete data set (CNM). For example, the node joining Adenaria, Pehria, Koehneria, and Woodfordia has a bootstrap of 98% and a
PP of 1.00 in CNM compared to a bootstrap of 78% and a PP of 0.91 when Cuphea is added to CM. In some cases, branch support is relatively unaffected by taxon and data sampling. The node joining Ammannia, Nesaea, and Lawsonia has 100% bootstrap support and a PP of 1.00 in CNM, compared to a bootstrap of 99% and a PP of 1.00 when
28
Ginoria is added to CM. In no case was there a dramatic decrease in the value for either measure of support (Figures 1-3 and 1-4).
29
Discussion
Data Sets and Analyses
CNM resulted in a slightly more resolved topology, and all taxa missing from CM were placed in clades in CNM that would be predicted from morphology or previous molecular analyses (Graham et al., 1993; Huang and Shi, 2002; Graham et al., 2005).
This supports the assertion by Wiens (2006) that incomplete taxa may be added to phylogenetic analyses without negative effect as long as there is enough data to place those taxa with good branch support. Using simulated data sets, they found that when the overall number of characters is high (>2000), the entire tree could be reconstructed correctly even when half of the taxa were 90% incomplete, except in cases where branches were extremely long. This could be potentially problematic in an analysis of the Lythraceae since terminal branches are often relatively long. However, Wiens (2006) also found that adding taxa in this situation could actually improve results in some cases, by subdividing potentially problematic long branches. Wiens (2006) also found that, for almost all methods and conditions, taxa that were 50% complete were just as beneficial as taxa that were 100% complete. In model-based methods such as Bayesian and likelihood approaches, taxa that were as much as 75-90% incomplete were found to cause
“dramatic increases in accuracy,” however highly incomplete taxa were less successful in rescuing parsimony-based analyses. In our analysis, incomplete taxa ranged from 12-
61% complete, and there were no differences in topology between parsimony and
Bayesian approaches. While Wiens (2006) found no significant difference in the relationship between levels of support and the levels of completeness, in our comparison
30
there were slight decreases in branch support for several clades when new taxa were
added. For example, support for the clade containing Ammannia, Nesea, and Lawsonia
dropped from 100% BS to 74% BS when Ginoria and Tetrataxis were added (in this case
the PP remained the same at 100%). Additionally, taxa with the most missing data
(Lourtella, Physocalymma, and Tetrataxis –all 12% complete) were placed in clades with
relatively low support (58-65% after addition of these taxa).
Relationships Among Taxa
Despite the addition of two new chloroplast regions (> 3400 bp) and increased
taxon sampling, statistical support for basal relationships in the family remains relatively low. However, the relationships of terminal clades are more stable (with combined MP and Bayesian analyses resulting in congruent topologies) compared to the analyses of
Graham et al. (2005), where basal relationships varied considerably depending on the type of analysis (including MP, ML, and Bayesian approaches). This indicates that the additional parsimony informative characters added in this analysis may be resulting in an improved phylogenetic signal.
Bootstrap values for the two basal clades are 88% and <50%. As has been generally observed in other studies (Suzuki et al., 2002; Cummings et al., 2003),
Bayesian support for the same branches is stronger (100% and 55%, respectively). The sampled genera are divided almost equally between the two major clades, with clade I containing 13 of the 28 sampled genera and clade II containing the remaining 15 genera
(see Figure 1-4). The combined analysis of Huang and Shi (2002) recovered a completely congruent topology with almost identical bootstrap support (83% and 54%
31
respectively), although their sampling included only 4 taxa from clade I, and 12 taxa from
clade II. The topology is also similar to those recovered by the maximum likelihood
analyses of trnL-trnF and combined data sets in Graham et al., (2005), with the only
incongruence being the relative placement of the clade containing Rotala and Heimia
(including Didiplis in my analysis). However, the placement of this clade is not well supported in any case, and may best be treated as part of a 3-way polytomy at the base of the family.
The six “crown” clades recovered by all analyses of Graham et al. (2005), were congruent to terminal subclades found by Huang and Shi (2002), with the only differences caused by variation in the taxa sampled in each analysis. These clades continue to be well supported in the current analysis (A-F in Figure 1-4). Crown clades labeled I, II, and III by Graham et al. (2005), correspond exactly to C, D, and B, respectively, in the current combined analyses.
Completed data sets for Adenaria, Pehria, Koehneria, and Pleurophora are included here in combined analyses for the first time. There is now substantial molecular evidence to support their close relationship to Cuphea and Woodfordia (crown clade IV
from Graham et al., 2005), as has been previously suggested by morphological analyses
(they all share resin-secreting trichomes; Graham et al., 1993), and the individual analysis
of trnL-trnF (Graham et al., 2005). There is some molecular evidence suggesting that
Lourtella, Physocalymma, and Diplusodon are also included in this group. However, they are currently represented by data sets that are <20% complete, and additional sequence data will be required to further address this possibility. New molecular evidence
32
presented here supports Webb’s (1967) recommendation (based on morphology) that
Lythrum and Peplis should be merged, and further analyses including a more detailed sampling of these taxa are ongoing in an effort to better address this question (Chapter 2).
Additionally, there is now substantial molecular evidence (BS> 99% and PP >0.99 in all individual and combined analyses), implying that Didiplis is sister to Rotala and closely related to Heimia, a relationship never previously suggested. A more detailed discussion of this proposed relationship will be presented elsewhere (Morris et al., in prep).
As found in previous analyses (Graham et al., 2005), all morphological characters
appear to be homoplastic when mapped onto the tree of the most inclusive combined
analysis. For almost all traits, multiple evolutionary events (gains and/or losses) occur across, and often within, each of the major clades. For example, most of the family has superior ovaries, with inferior or semi-inferior ovaries arising at least twice (Figure 1-5
A). Inflorescences terminating in a flower were apparently derived from inflorescences
terminating in a vegetative meristem at least four times (Figure 1-5 B). Both major
clades contain taxa with three-pseudocolpate and six-pseudocolpate pollen interspersed among taxa without pseudocolpi (Figure 1-5 C). Even considering that the current analysis has a slightly different basal topology than that used for character optimization by Graham et al. (2005), my interpretation of character evolution differs from earlier analyses only in the case of seed coat trichomes. In Graham et al. (2005), taxa having straight seed trichomes were distributed in both major clades, while in my analysis, taxa with straight seed trichomes are confined to clade I, and taxa with spiral seed trichomes
33
Figure 1-5. Selected morphological characters mapped on the majority-rule Bayesian consensus tree of 9000 trees based on the combined analysis of all available molecular data (for comparison to Figure 6 in Graham et al., 2005). (A) Ovary position (B) Inflorescence type (C) Pollen pseudocolpi (D) Seed coat trichomes.
34
I I
II II
I I
II II 35
are confined to clade II. However, multiple independent gains and/or losses of both
trichome states are still inferred (Figure 1-5 D).
Relationships within the Lythraceae have puzzled researchers for decades.
Molecular data now provide the first new independent means of addressing these
questions. The current combined analysis based on a large sequence dataset (> 6500 bp)
has greatly improved our understanding the relationships within the family. However, a
complete resolution requires a complete taxon sampling and more data. It is clear that
multiple convergences in morphological characters and life history traits have occurred
during the evolution of this family, and it is possible that the poor resolution at the base
of the family might reflect an early rapid radiation resulting in a number of lineages evolving within a short period of time. If this is the case, it is possible that no amount of additional data will ever be able to completely resolve these relationships.
36
Literature Cited
Conti E, Litt A, Wilson PG, Graham SA, Briggs BG, Johnson LAS, and Sytsma KJ. 1997. Interfamilial relationships in Myrtales: molecular phylogeny and patterns of morphological evolution. Systematic Botany 22(4): 629-647.
Cronquist A. 1981. An integrated system of classification of flowering plants. Columbia University Press, New York.
Cummings MP, Handley SA, Myers DS, Reed DL, Rokas A, and Winka K. 2003. Comparing bootstrap and posterior probability values in the four-taxon case. Systematic Biology 52: 477-487.
Dahlgren R. 1975. A system of classification of the angiosperms to be used to demonstrate the distribution of characters. Botaniska Notiser 128: 119-147.
Dahlgren R and Thorne RF. 1984. The order Myrtales: circumscription, variation, and relationships. Annals of the Missouri Botanical Garden 81: 419-450.
Fay MF, Bayer C, Alverson WS, de Bruijin AY, and Chase MW. 1998. Plastid rbcL sequence data indicate a close affinity between Diegodendron and Bixa. Taxon 47: 43- 50.
Felsenstein J. 1985. Confidence limits on phylogenies: An approach using the bootstrap. Evolution, 39: 783-791.
Graham SA, Crisci JV, and Hoch PC. 1993b. Cladistic analysis of the Lythraceae sensu lato base on morphological characters. Botanical Journal of the Linnaean Society. 113: 1-33.
Graham SA, Hall J, Sytsma K, and Shi S. 2005. Phylogenetic analysis of the Lythraceae based on four gene regions and morphology. International Journal of Plant Sciences 166(6): 995-1017.
Huang Y-L and Shi S-H. 2002. Phylogenetics of Lythraceae sensu lato: a preliminary analysis based on chloroplast rbcL gene, psaA-ycf3 spacer, and nuclear rDNA internal transcribed spacer (ITS) sequences. International Journal of Plant Sciences. 163: 215- 225.
Hutchinson J. 1973. The families of flowering plants. 3d ed. Clarendon, Oxford.
Johnson LAS, and Briggs BG. 1984. Myrtales and Myrtaceae –a phylogenetic analysis. Annals of the Missouri Botanical Garden. 71:700-756.
37
Kishino H and Hasegawa M. 1989. Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order of the Hominoidea. Journal of Molecular Evolution. 29: 170-179.
Koehne E. 1881. Lythraceae monographice describuntur. Botanische Jahrbücher Systematik 1: 142-157.
Koehne E. 1903. Lythraceae. In A Engler, [ed]. Das Pflanzenreich 4. 216. (Heft#17). 1-326. W Engelmann, Weinheim, Germany.
Maddison WP and Maddison DR. 2003. Mesquite: a modular system for evolutionary analysis. Version 1.05. http://mesquite-project.org.
Manon PS and Steele KP. 1997. Phylogenetic analyses of “higher” Hamamelididae based on plastid sequence data. American Journal of Botany 84: 1407-1419.
Melchior H. 1964. Myrtiflorae. Pages 345-366 in H Melchior, ed. Englander’s Syllabus der Pflanzenfamilien. 12th ed. Vol 2. Bornträger, Berlin.
Posada D and Crandall KA. 1998. Modeltest: Testing the model of DNA substitution. Bioinformatics. 14: 817-818.
Prager EM and Wilson AC. 1988. Ancient origin of lactabumin from lysosome : analysis of DNA and amino acid sequences. Journal of Molecular Evolution. 27: 326- 335.
Schwarzbach AE and Ricklefs RE. 2000. Systematic affinities of Rhizophoraceae and Anisophylleaceae, and intergeneric relationships within Rhizophoraceae, based on chloroplast DNA, nuclear ribosomal DNA, and morphology. American Journal of Botany 87(4): 547-564.
Shi S, Huang Y, Tan F, He X and Boufford DE. 2000. Phylogenetic Analysis of the Sonneratiaceae to Lythraceae based on ITS Sequences of nrDNA. Journal of Plant Research. 113: 253-258.
Shimodaira H and Hasegawa M. 1999. Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Molecular Biology and Evolution. 16: 1114-1116.
Susuki Y, Glazko GV, Nei M. 2002. Over credibility of molecular phylogenetics obtained by Bayesian phylogenetics. Proc Natl Acad Sci USA. 99: 16138-16143.
Swofford DL. 2001. PAUP*. Phylogenetic Analysis Using parsimony (*and other Methods). Sinauer Associates, Sunderland MA.
38
Sytsma KJ, Litt A, Zihra, ML, Pires JC, Nepokroeff M, Conti E, Walker J, and Wilson PG. 2004. Clades, clocks, and continents: Historical and biogeographical analysis of Mrytaceae, Vochysiaceae, and relatives in the Southern Hemisphere. International Journal of Plant Sciences. 165(4 suppl.): S85-S105.
Tablerlet PL, Gielly GP, and Bouvet J. 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Molecular Biology. 17: 1105-1109.
Takhtajan, A. 1987. System of Magnoliophyta. Academy of Sciences USSR, Leningrad.
Templeton AR. 1983. Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to humans and apes. Evolution.37: 221-244.
Thorne RF. 1981. Phytochemistry and angiosperm phylogeny, a summary statement. Pages 233-295 in DA Young, DS Seigler, eds. Phytochemistry and angiosperm phylogeny. Praeger, New York.
Thorne RF. 1992. An updated phylogenetic classification of the flowering plants. Aliso 13: 365-389.
Tobe H, Graham SA, and Raven PH. 1998. Floral morphology and evolution in Lythraceae sensu lato. Pages 329-344 in SJ Owens, PJ Rudall, eds. Reproductive biology. Royal Botanic Gardens, Kew.
Webb DA. 1967. Generic limits in European Lythraceae. Feddes Repertorium Specierum Novarum Regni Vegetables. 74: 10-13.
Wiens JJ. 2006. Missing data and the design of phylogenetic analyses. Journal of Biomedical Informatics. 39: 34-42.
CHAPTER 2
A Molecular Phylogeny of Lythrum (Lythraceae): Preliminary Analyses Based on the atpB-rbcL Intergenic Spacer and ITS
Abstract
The genus Lythrum (Lythraceae) is comprised of 25-36 species of herbaceous
annuals and perennials with a distribution almost entirely in the Northern Hemisphere. It
is probably best known for its introduced and invasive members, including the notorious
invasive, L. salicaria (purple loosestrife); however it also includes many rare and narrow endemics. Recent analyses at the family level suggest that a monophyletic Lythrum also includes the genus Peplis, and a more comprehensive sampling of these taxa is required to confirm these results. Classification within Lythrum based on traditional
morphological characters is challenging because of considerable vegetative variability
within species and few readily apparent differences among species. Clusters of species
form complexes across Eurasia and North America and Mexico, and to date, relationships
among the various groups remain unclear. This study analyzes sequence data from the
atpB-rbcL intergenic spacer and ITS for ten species of Lythrum, one species of Peplis,
and appropriate outgroups using maximum parsimony, maximum likelihood, and
Bayesian approaches. There is now substantial molecular evidence to support the
assertion that Peplis should be included as species of Lythrum. The current subgeneric
classification of Lythrum is not supported, however strong biogeographical patterns exist.
39 40
Population level sampling, more variable genetic markers, and an improved understanding of the distribution of morphological variation across the phylogeny, will be required before useful recommendations concerning revisions to taxonomy can be made.
41
Introduction
The genus Lythrum (Lythraceae) is comprised of approximately 25-36 species of
herbaceous annuals and perennials with a distribution almost entirely in the Northern
Hemisphere. It is probably best known for its introduced and invasive members,
including the notorious invasive, L. salicaria (purple loosestrife). This pervasive species
is native to Europe, Asia and Japan, and has been introduced to Africa and Australia. It
has become extremely invasive in wetland areas across North America, where economic
losses due to decreased forage and control costs have been estimated at $45 million per
year (ATTRA, 1996). Evidence suggests that negative effects on native wetland biota
and ecosystem function are also high (reviewed in Blossey et al., 2001). Another
widespread species, L. hyssopifolia, native to southern Europe, Asia and the Middle East, also has become invasive in many parts of the world, including Australia and North
America. Conversely, some Lythrum species are considered rare, narrow endemics, including the North American species L. curtissi (endemic to southern Georgia and northern Florida) and L. flagellare (endemic to central Florida), which are both state- listed as threatened or endangered (USDA, NRCS, 2007). Clusters of the remaining
Lythrum species form complexes across Eurasia, and North America into Mexico.
Members of the genus are most commonly herbs, or occasionally small shrubs
found primarily in wet or seasonally wet habitats such as freshwater and coastal marshes, pond and lake margins, and along rivers, streams, and ditches. Most Lythrum species are relatively easily distinguished from other genera in the Lythraceae by having 6-merous, conspicuously ribbed, elongate, tubular, axillary flowers; rosy purple to white petals
42 inserted just within the margin of the floral tube; twelve stamens arranged in two whorls; and a two-carpellate superior ovary (Koehne, 1903; Graham, 1964; Webb, 1967; illustrated in Figure 2-1). However, exceptions to this general pattern occur, causing taxonomic confusion with other wetland herbaceous genera in the family, including
Ammannia, Didiplis, and especially Peplis (Graham, 1964; Webb, 1967).
Peplis was historically distinguished from Lythrum by minute or absent petals, 4-
6 stamens, and a hemispherical floral tube. Pollen morphology also supports recognition of two genera (Graham et al., 1990). In his treatment of the European Lythraceae, Webb
(1967) merged Peplis with Lythrum, citing apparently intermediate taxa such as L. thesiodes and L. borysthenicum, species with flowers like Peplis (0-4 petals, 4-6 stamens, and short, wide floral tubes). He distinguished Ammannia from Lythrum based on differences in their inflorescence (Ammannia generally having small clusters of flowers in their leaf axils, and Lythrum having one, or at most two flowers per leaf axil), and maintained the North American monotypic genus Didiplis (often included within Peplis on the basis of its similar habit and apetalous 4-merous flowers). These four genera were considered separately in the first comprehensive cladistic analyses of morphological characters in the family (Graham et al., 1993), and clustered together with the other small, herbaceous marsh genera in the Lythraceae, including Nesaea, Rotala, and
Hionanthera.
Subsequent molecular evidence did not support the relationships predicted based on morphology (Huang and Shi, 2002; Graham et al., 2005; Morris et al., in prep.). In
43
Figure 2-1. Botanical illustration of Lythrum salicaria by Otto Wilhelm Thomé. Flora von Deutschland, Österreich und der Schweiz 1885, Gera, Germany. (A) - Upper part of plant stem showing two whorls of leaves and an inflorescense. (1) - A single flower bud clearly showing the ribbed floral tube and epicalyx appendages. (2) - A single fully developed flower. (3) - Longitudinally dissected floral tube opened flat to illustrate the relative attachment location of the single ovary, twelve anthers, and six petals. (4) and (8) – Stigma. (5) and (7) - A single flower longitudinally dissected along two different planes; illustrating the two-carpellate superior ovary. (6) - Ovary, style and stigma. (9) - Two types of anthers found in flowers of L. salicaria. (10) - Fully developed fruit with seeds. 44
45
molecular analyses, Lythrum and Peplis form a well-supported clade (100% bootstrap in all analyses) only distantly related to a clade containing Ammannia and Nesaea, and a third clade contains Didiplis and Rotala (Hionanthera has not yet been sampled). This suggests that many of the characters traditionally used to group these taxa have likely arisen multiple times within the Lythraceae, perhaps in response to similar environmental pressures. Each of these studies included only one species of Lythrum and of Peplis, so results were unable to address their relative distinction. However, more recent analyses of Lythraceae containing 2 species of Lythrum, and P. portula (Chapter 1) suggest that a monophyletic Lythrum includes Peplis, however, a more comprehensive sampling of these taxa is required to confirm these results.
Determination of relationships within Lythrum are “notoriously difficult”
(Shinner, 1953) because of considerable vegetative variability within species, few readily
apparent differences among species, and frequent introgression in areas where ranges
overlap (Shinner, 1953; Graham, 1975). The most recent complete treatment of the
genus (Koehne, 1903) includes two subgenera, five sections, two subsections, and 24
species. Modern North American, European, and Asian floras currently recognize up to
36 species (Table 2-2), as well as a considerable number of subspecies, varieties, and
forms. It is likely that some of these taxa are based on minor phenotypic differences in
more widespread, variable species (S. Graham, pers com.).
This study analyzes sequence data from the atpB-rbcL intergenic spacer and ITS
sequences for ten species of Lythrum, one species of Peplis, and appropriate outgroups
using maximum parsimony and Bayesian approaches. The main objectives are: 1) to
46
Table 2-2: List of all species of Lythrum currently included in modern North American, European, and Asian floras; including species designation, relative distribution, and current classification.
According to Koehne, 1903 Species Distribution Subgenus Section Subsection 1 Lythrum acutangulum Spain, France, N. Africa H Sal na 2 Lythrum alatum var. alatum E. US H Euh Pyt 3 Lythrum alatum var. lanceolatum SE US H Euh Pyt 4 Lythrum album Mexico H Euh Pyt 5 Lythrum anatolicum Turkey H Euh Pen 6 Lythrum anceps Japan S na na 7 Lythrum baeticum Spain, N. Africa, Middle East H Sal na 8 Lythrum bryanti Baja, Mexico H Euh Pyt 9 Lythrum californicum SW US H Euh Pyt 10 Lythrum castellanum Spain H Sal na 11 Lythrum curtissii Georgia, Florida H Euh Pyt 12 Lythrum flagellare Florida H Euh Pyt 13 Lythrum flexuosum Spain H Euh Pen 14 Lythrum gracile Mexico H Euh Pyt 15 Lythrum hispidulum S. Europe, N. Africa H Mid na 16 Lythrum hyssopifolia Eurasia (invasive worldwide) H Euh Pen 17 Lythrum intermedium Eurasia, Far East S na na 18 Lythrum junceum Spain H Euh Pen 19 Lythrum komarovii Former USSR H Sal na 20 Lythrum lineare E and SE US H Euh Pyt 21 Lythrum linifolium Middle East, Asia H Euh Pen 22 Lythrum maritimum C. and S. America, intr. Hawaii H Euh Pyt 23 Lythrum nanum Central Asia H Sal na 24 Lythrum ovalifolium Texas, N. Mexico H Euh Pyt 25 Lythrum paradoxum Australia H Mes na 26 Lythrum rotundifolium Africa H Hoc na 27 Lythrum salicaria Eurasia, Middle East S na na 28 Lythrum schelkovnikovii Caucasus H Euh Pen 29 Lythrum silenoides C. Asia, Middle East H Euh Pen 30 Lythrum theodori Caucasus H Euh Pen 31 Lythrum thesoides SE Europe, Caucasus H Euh Pen 32 Lythrum thymifolia S. Europe, N. Africa, C. Asia H Euh Pen 33 Lythrum tribracteatum Mediteranean, C. Asia, S. Russia H Sal na 34 Lythrum virgatum C. Europe to N. Mongolia S na na 35 Lythrum vulneraria Mexico H Euh Pyt 36 Lythrum wilsonii Australia H Euh Pen H = Hyssopiflolia; S = Salicaria Sal = Salzmannia; Euh = Euhyssopifolia; Mid = Middendorfia; Mes = Mesolythrum; Hoc = Hochstetteria Pyt = Pythagorea; Pen = Pentaglossum; na = current classification does not include this level of division 47 provide increased support for the hypothesis that a monophyletic Lythrum should include
Peplis; 2) to compare a molecular phylogeny of Lythrum to its current subgeneric classification; and 3) to evaluate the utility of these molecular markers for ongoing work on more comprehensive molecular analysis of the genus.
48
Methods
Taxon Sampling and DNA Sequencing
Ten species of Lythrum and one species of Peplis were sampled. In molecular analyses of the Lythraceae (Huang and Shi, 2002; Graham et al., 2005; Morris et al., in prep {see chapter one}), Decodon is the taxon most often found in a sister-relationship with the clade containing Lythrum and Peplis (usually lacking high bootstrap support).
Based on this, Decodon was used as an outgroup in these analyses, along with the more distantly related genera Ammannia and Heimia. ITS sequences for Ammannia and
Heimia were obtained from GenBank, and all remaining sequences were generated for this study (Table 2-3).
Total DNA was extracted from fresh tissue, dried material preserved in silica gel, or dried herbarium specimens. All voucher information and GenBank accession numbers are provided in Table 2-3. DNA extractions were conducted using a modified CTAB protocol (as in Graham et al., 2005), or the QIAGEN DNeasy Plant Mini Kit (QIAGEN
Inc., Chatsworth, CA). Some difficulties exist with DNA extraction in this family, and extended storage in the freezer and/or substantial dilution is often required prior to successful PCR.
Because of difficulties in sequencing atpB-rbcL in Peplis, it was necessary to design one new internal sequencing primer for this genus, which was done using sequence data from a consensus of available Lythrum species and the primer design program Primer3 (Rozen and Skaletsky, 1998). Sequence information for this primer, in addition to all other PCR and sequencing primers used in this study can be found in Table
49
Table 2-3. Information for all plant samples included in this study; including species designation, GenBank accession numbers, and location of voucher.
Taxon DNA Region GenBank Accession # Voucher Ammannia latifolia atpB-rbcL waiting submission Liogier 10314 (MO) Ammannia baccifera * ITS AY905419 Tang, S. Q. 99010301 (SYS) Decodon verticillatus atpB-rbcL waiting submission Graham 917 (MO) Decodon verticillatus ITS waiting submission Graham 917 (MO) Heimia salicifolia atpB-rbcL waiting submission S. Graham 1062 (MO) Heimia myrtifolia * ITS AF201693 Tang, S. Q. 99070502 (SYS) Lythrum alatum var. alatum atpB-rbcL waiting submission S. Graham 1129 (MO) Lythrum alatum var. alatum ITS waiting submission S. Graham 1129 (MO) Lythrum album atpB-rbcL waiting submission J. Rzedowski s.n. in Nov. 1963 (MO) Lythrum album ITS waiting submission J. Rzedowski s.n. in Nov. 1963 (MO) Lythrum flagellare atpB-rbcL waiting submission S. Graham 1103 (MO) Lythrum flagellare ITS waiting submission S. Graham 1103 (MO) Lythrum hyssopifolia atpB-rbcL waiting submission cultivated by S. Graham (Park Seed Company) Lythrum hyssopifolia ITS waiting submission cultivated by S. Graham (Park Seed Company) Lythrum junceum atpB-rbcL waiting submission cultivated by S. Graham (Gottingen Botanical Garden -Index Seminum 2000, Catalog #2017) Lythrum junceum ITS waiting submission cultivated by S. Graham (Gottingen Botanical Garden -Index Seminum 2000, Catalog #2017)
Lythrum lineare atpB-rbcL waiting submission B. Brown s.n. (MO) Lythrum lineare ITS waiting submission B. Brown s.n. (MO) Lythrum ovalifolium atpB-rbcL waiting submission M. C. Johnston s.n. in Oct. 1975 (MO) Lythrum ovalifolium ITS waiting submission M. C. Johnston s.n. in Oct. 1975 (MO) Lythrum salicaria var. anceps atpB-rbcL waiting submission cultivated by S. Graham (Gottingen Botanical Garden -Index Seminum 2000, Catalog #2015) Lythrum salicaria var. anceps ITS waiting submission cultivated by S. Graham (Gottingen Botanical Garden -Index Seminum 2000, Catalog #2015)
Lythrum salicaria var. salicaria atpB-rbcL waiting submission B. Brown s.n. (MO) Lythrum salicaria var. salicaria ITS waiting submission B. Brown s.n. (MO) Lythrum virgatum atpB-rbcL waiting submission cultivated by S. Graham (Gottingen Botanical Garden -Index Seminum 2000, Catalog #2021) Lythrum virgatum ITS waiting submission cultivated by S. Graham (Gottingen Botanical Garden -Index Seminum 2000, Catalog #2021)
Peplis portula atpB-rbcL waiting submission Montezuma s.n. (MO) Peplis portula ITS waiting submission Montezuma s.n. (MO) * = sequence from another study, downloaded from genbank 50
2-4. PCR amplification was carried out using standard PCR techniques and a MJ
Research Thermocycler. These reactions were performed in a volume of 25 µl containing
6 µl of a Tricine buffer mix (final concentration: 30 mM Tricine pH 8.4, 2.0 mM MgCl2,
50 mM KCl, and 0.25 mM each dNTP), 0.05 µg each primer, 1 U of Taq polymerase, and approximately 0.5 ng of genomic DNA. In a few cases, genomic DNA needed to be substantially diluted (0.005-0.05 ng/25 µl reaction) before PCR was successful. The thermocycler program included 1 minute for initial strand separation at 94˚C; followed by
37 cycles of 1 minute at 94˚C, 45 seconds at 50˚, 4 minutes at 72˚C; and a final 10 minute step at 72˚C. PCR products were isolated using 1.5% agarose gels and purified using the
QIAGEN QIAquick Gel Extraction Kit (QIAGEN Inc., Chatsworth, CA). Cycle sequencing reactions were conducted in a 10 µl volume, using 2 µl ABI BigDye cycle sequencing terminator chemistry (Perkin-Elmer Corp.), 2 µl Half Term Dye (Bioline
Inc.), 2.0 ng each primer, and approximately 50 ng of template DNA. Reactions were purified using sephadex in Cetrisep columns according to manufacturer’s instructions, and then were completely lyophilized using a ThermoSavant DNA120 SpeedVac.
Sequencing was conducted on an Applied Biosystems 3730 DNA Analyzer at the
Sequencing Core Facility, Brigham Young University, Provo, Utah. Completed sequences were edited, assembled, compiled with GenBank sequences, and initially aligned using Sequencher 4.1 (GenesCode Co. Ann Arbor, MI). Manual adjustments were then made to ensure proper alignment.
51
Table 2-4. Base composition and citation information for all PCR and sequenceing primers used and/or developed for these analyses.
Primer Sequence 5'-3' Citation
*atpB-rbcL forward GAA GTA GTA GGA TTG ATT CTC Schwarzbach and Ricklefs (2000)
*atpB-rbcL reverse AGT TTC TGT TTG TGG TGA CAT Schwarzbach and Ricklefs (2000)
^atpB-rbcL INT1forward TGG TTC TTT ATT AGA CCA TGG TAT TTG this study
*ITS A GGA AGG AGA AGT CGT AAC AAG Blattner (1999)
*ITS B CTT TTC CTC CGC TTA TTG ATA TG Blattner (1999) * For amplification and sequencing. ^ For sequencing only. 52
Data analysis
Each of the data sets was analyzed individually and examined for overall
congruence before being combined. Individual and combined non-partitioned data sets
were analyzed using a maximum parsimony (MP) approach implemented in
PAUP*4.0b10 (Swofford, 2001). Analysis methods include heuristic searches using
random stepwise addition with 1000 replicates, tree bisection reconnection (TBR) branch
swapping, accelerated transformation (ACCTRAN) character optimization, MulTrees,
and the ‘steepest descent’ option not selected. Characters were unordered and equally
weighted, and gaps were coded as missing. Bootstrap analyses (Felsenstein, 1985) were
conducted using a heuristic search with the same settings as above and 1000 replicates.
Tree length, consistency index (CI) excluding autapomorphic characters, and the
retention index (RI) were calculated.
Bayesian analyses were conducted for the combined non-partitioned data set
using MrBayes (v.3.1.2; Huelsenbeck and Ronquist, 2003; Ronquist and Huelsenbeck,
2005). ModelTest (v.3.5; Posada and Crandall, 1998) was employed to determine which
model best fit the data. Analyses were run using the general time reversible + invariant
gamma model (GTR+I+Γ) on eight chains for 5 million generations, saving trees every
500 generations. The first 1000 trees were discarded (500,000 generations), and a 50%
majority rule consensus tree was constructed from the remaining 9000 saved trees using
PAUP*. Maximum likelihood analyses, including 1000 bootstrap replicates, were
conducted using the default settings in GARLI (Genetic Algorithm for Rapid Likelihood
Inference, www.bio.utexas.edu/faculty/antisense/garli/Garli.html; Zwickl, 2006).
53
Results
The atpB-rbcL intergenic spacer is over 100 bp longer than ITS within the taxa sampled for these analyses, but is considerably less variable, having approximately 1/4 the number of parsimony informative characters (Table 2-5). Separate analyses of these data sets produce congruent topologies, however they differ considerably in their degree of resolution. The atpB-rbcL data set contributes very little resolution overall, but it does provide increased support for the relative monophyly of the genus, in addition to recognizing the split between the two major subclades (I and II in Figure 2-2). Most of the resolution in the combined analysis comes from ITS, which provides support for the relative distinctiveness of Peplis and three strongly supported clusters of Lythrum taxa
(A, B, and C in Figure 2-2).
The maximum parsimony (MP), maximum likelihood (ML), and Bayesian analyses of the combined dataset are completely congruent (Figure 2-2), with each of the analyses showing strong support for a monophyletic Lythrum including Peplis. The molecular evidence clearly does not support the currently existing sub-classification of
Lythrum species (Koehne, 1903; Table 2-1). This classification designates two subgenera, separated on the basis of inflorescence type: subgenus Salicaria (including only L. salicaria and L. virgatum), having compound inflorescences with multiple trimorphic flowers in axils, or terminal spikes or racemes; and subgenus Hyssopifolia
(including all remaining Lythrum species), generally with only 1-3 flowers per axil.
Based on molecular data, Lythrum hyssopifolia and L. junceum (subgenus Hyssopifolia)
54
Figure 2-2. Majority-rule Bayesian consensus of 9000 trees based on the combined analysis of atpB-rbcL and ITS datasets. Posterior probabilities are shown above the branches and maximum parsimony and (maximum likelihood) bootstrap values > 50% are show below the branches. The composition of superclades I and II, and subclades A, B, and C are indicated to the right of the tree. (H) = Subgenus Hyssopifolia; (S) = Subgenus Salicaria (Koehne, 1903). NA = native North American distribution, EA = native Eurasian distribution. 55
56
Table 2-5. Comparative statistics for the maximum parsimony analyses for each of the individual data sets (atpB-rbcL and ITS), and for the combined analysis; including total aligned length of the DNA region, number and percentage of parsimony informative characters, number and length of most parsimonious tree produced, consistency index (CI) and retention index (RI).
atp B-rbc L ITS Combined
Aligned Length 726 595 1321
# of Parsimony Informative Characters 24 83 107
% Informative 3.3 13.9 7.4
# of Most Parsimonious Trees 1 1 1
Length of Most Parsimonious Tree 79 149 328
Consistency Index 0.962 0.772 0.778
Retention Index 0.955 0.842 0.858 57
are grouped in subclade II with Peplis and species from subgenus Salicaria, rather than
with the other included species from subgenus Hyssopifolia, all found in subclade I.
There is a strong biogeographical pattern evident, with the North American taxa
of subgenus Hyssopifolia forming clade I, while the Eurasian taxa and Peplis form clade
II. The Eurasian taxa form two subclades, one exclusively consisting of species from
subgenus Hyssopifolia, and the others with species of subgenus Salicaria (B and C in
Figure 2- 2). Relationships within the terminal clusters (especially A and C) remain poorly resolved and are supported by very few molecular characters (Figure 2-3).
58
Figure 2-3. Majority-rule Bayesian consensus of 9000 trees based on the combined analysis of atpB-rbcL and ITS datasets showing branch lengths to scale.
59
60
Discussion
Peplis
As generally recognized, the genus Peplis includes two species of small aquatic
annuals with a Eurasian distribution. Peplis portula occurs in wet habitats across Europe
(except the extreme north), and is described as having creeping stems, opposite obovate
leaves with distinct petioles, and 6-merous flowers (rarely with reduced purple petals).
Peplis alternifolia has a more northern and eastern distribution, shorter erect stems,
alternate linear leaves with shorter petioles, and generally 5-merous flowers lacking
petals (Koehne, 1903; Webb, 1967). These species were recognized as being closely
allied with Lythrum due to many shared morphological characters, including small
tubular to nearly campanulate flowers found in leaf axils, and a prominent epicalyx
(Koehne, 1903). However, they were retained as separate genera based on reduced or
absent petals, generally fewer stamens and more hemispherical (wider than long) floral
tubes (Koehne, 1903, Webb, 1967).
Webb (1967) argued that both species of Peplis should be merged with Lythrum,
pointing out specific taxa with intermediate morphologies, including L. thesioides
(having minute or missing petals, 4-6 stamens, and a bell-shaped floral tube), L.
thymifolia and L. tribracteatum (both often found with reductions in the numbers of
petals and stamens), and L. borysthenicum (sometimes classified as a species of Peplis
due to its similar habit and general appearance). Despite this recommendation, Peplis
and Lythrum are still generally addressed as separate genera in most treatments of the family (S. Graham, pers. com.). Graham et al. (1993) discussed their possible congeneric
61
status, but treated them separately in the first comprehensive cladistic analysis of
morphological characters in the Lythraceae, citing the absence of floral nectaries found to
occur in all other Lythrum species (Tobe and Raven, 1983), and distinctive pollen
morphology (Graham et al., 1987). Not surprisingly, based on this morphological
analysis, Peplis clustered with Rotala, Hionanthera, and Didiplis; all of which are small
herbaceous annuals lacking secondary tissue development, and found in marshy or
aquatic habitats. These four genera were also closely allied with other genera having
annual herbaceous species (Nesaea, Ammannia, and Lythrum). It is now clear from
molecular phylogenetic studies of the family (Huang and Shi, 2002; Graham et al., 2005,
see also Chapter 1) that this habit has likely evolved multiple times in the family and does
not necessarily indicate a close relationship.
Recent molecular phylogenetic analyses of the Lythraceae (Huang and Shi, 2002;
Graham et al., 2005) including various species of Lythrum and Peplis, also treated these
taxa separately. Huang and Shi (2002) included P. portula and two samples of L. salicaria, while Graham et al. (2005) included two samples of P. portula, and two different species of Lythrum (L. hyssopifolia and L. salicaria). These studies provided strong support for the close relationship of Lythrum and Peplis, but were unable to
adequately address their possible congeneric status. It is interesting to note that these
analyses show the same relative relationships of Peplis, L. salicaria, and L. hyssopifolia
that are found in the current analyses, but were unable to demonstrate that Peplis is nested
within Lythrum because of insufficient sampling.
62
Results of the current analysis show P. portula to be well nested within a monophyletic Lythrum, at the base of subclade II (also including the rest of the Eurasian
Lythrum species included in this analysis; Figure 2-2). Unfortunately, due to the difficulties involved in getting fresh material from the Middle East and former Soviet
Republics, and problems associated with extracting quality DNA from herbarium specimens (especially problematic in this family), we were unable to include P. alternifolia or any of the other diminutive Lythrum species with intermediate morphologies in the current analysis. As we are able to add these species to the analysis, it will be interesting to see if Peplis remains in a basal position with respect to the rest of the Eurasian Lythrum species, and how morphological features previously relied on to separate the genera have evolved across the phylogeny of Lythrum.
Relationships Within Lythrum
All native North American taxa included in the analysis form one major clade (all from subsection Pythagorea), while all sampled Eurasian taxa (from both subgenera and including Peplis) form a second major clade. Current evidence suggesting an Old World origin for the family (see review in Graham et al., 2005), and the relative distribution of genetic diversity found in the current analysis, supports a Eurasian origin for the genus.
The very low level of genetic diversity found within North American clade is suggestive of a bottleneck event, and is the likely result of a single introduction into North America.
Additionally, all of the North American species share a chromosome number of n=10 (including other North American and Mexican taxa not included in this analysis).
However, this number is also found in the Eurasian L. hyssopifolia. Chromosome
63 numbers in the Eurasian clade are considerably more variable, with n=5 being reported for Peplis and L. junceum (in addition to all other Eurasian species not included in this analysis), and n=15, 25, 30 being reported for the L. virgatum/ L. salicaria clade (see review in Graham and Cavalcanti, 2001). Evidence from this analysis of chromosome numbers in the entire family supports a base number of n=5 for Lythrum, suggesting that the formation of polyploids has likely occurred multiple times in the history of the genus, and may have played some role in the diversification of species.
North American clade
The low genetic diversity among the native North American taxa in this analysis was not unsuspected. A lack of readily distinguishable morphological characters among taxa and the occurrence of apparent hybrids in areas where ranges overlap have been previously reported (Shinners, 1953; Graham 1964, Graham 1975). The North American species are generally distinguished on the basis of geographic distribution, habitat, overall size and growth habit, relative degree of branching, color of flowers (ranging from dark pink to white), and the size and shape of leaves. After a close examination of
Lythrum specimens from all herbaria in the southeastern United States, Graham (1975) recognized 5 distinct species (and two varieties), distributed across the southern and eastern states. An additional species occurs across the southwestern states, a narrow endemic is found in south and central Texas, and two or three distinct species occur in
Mexico (L. album, L. gracile, and L. vulneraria) (Shinners, 1953; S. Graham pers com.)
Lythrum alatum is the most widespread North American species, having two recognized varieties divided primarily by stature, leaf shape, and geographical range.
64
Lythrum alatum var. alatum (included in the current analysis) ranges from the North East
to the Midwest, and has more rounded leaf bases, multiple slender stems growing from an enlarged root stock, and larger flowers. Comparatively, L. alatum var. lanceolatum has
leaves tapering at the base, taller and more robust stems, and is primarily restricted to the
South East. The morphological differences persist even when grown in the greenhouse
under the same conditions, suggesting they are at least partly genetically based and not
simply ecological variants (Graham, 1975).
In the current analysis, there was almost no sequence variation found between L.
alatum and three other more narrowly distributed taxa, including L. flagellare (a rare and
narrow endemic of central Florida, having short creeping stems and strictly opposite
leaves), L. ovalifolium (a narrow endemic of southern and central Texas, having a much reduced size and short oval leaves), and L. album (a shrubby white flowered species of
south central Mexico). In turn, these four taxa were only slightly differentiated in this
analysis from L. lineare, a species morphologically and ecologically more distinct.
Lythrum lineare is restricted to coastal brackish marshes from Georgia around to
Louisiana, and has tall, slender stems and a more open growth habit. The flowers also lack a nectariferous ring at the base of the ovary that is found in all other native North
American species (Shinners, 1953; Graham 1975).
Missing from this analysis are L. californicum (found from California, across the
Southwest and northern Mexico through Texas, with smaller and more slender stature
and narrow, gray-green leaves that are firm and waxy), L. curtissii (a narrow endemic of
southern Georgia and northern Florida, with a slender much-branched habit and narrow
65 yellow-green leaves) (pers.obs.), and L. gracile (another species reported from central
Mexico; S. Graham pers.com.). Materials have been collected from each of these taxa, but so far, we have been unable to extract usable DNA. These species are also reported to hybridize with L. alatum and L. lineare wherever their ranges overlap (Shinners, 1953;
Graham 1975). At this point, based on morphological evidence alone, we expect them to tightly cluster with the rest of the North American clade. Population-level sampling and molecular analyses (especially in areas where ranges overlap) will be required to further clarify relationships and to assess the relative distinctiveness of these taxa.
L. junceum - L. hyssopifolia clade
In this analysis, L. junceum and L. hyssopifolia appear as sister to one another in a strongly supported subclade. Terminal branch lengths in this clade are longer than those found in either of the other two clades of Lythrum species, suggesting a greater degree of differentiation between L. junceum and L. hyssopifolia than is seen among the North
American species. Morphological and cytological evidence are in concordance with these molecular data. These two species were placed in the same subsection by Koehne
(1903) with other Eurasian Lythrum species sharing 1-2 axillary flowers, a thickened nectary ring never circling the ovary, and comparatively long styles. Lythrum hyssopifolia (n=10) is generally described as being a weedy annual having erect ascending branches and homomorphic flowers with 4-6 stamens (Koehne, 1903; Webb,
1968). It is found in disturbed and seasonally wet places in southern Eurasia and has become invasive in many other parts of the world. Lythrum junceum (n=5) is distinguished as a tristylous perennial having decumbent straggling branches and 12
66 stamens. It occurs across southern Europe and the Middle East. Given their easily distinguishable morphology, differences in style morphology and chromosome counts, and the relative genetic distance between them, they are clearly distinct from one another.
However, many of the newly described Lythrum species across Europe and Asia are described as being similar to one or the other of these species, suggesting that an increased sampling of taxa from those areas would likely reveal species complexes of their own.
L. virgatum – L. salicaria clade
Koehne (1903) recognized the close relationship of these species, placing them in their own subgenus (Salicaria) based on shared complex inflorescences and the tristylous condition. Later cytological studies found that they also share higher ploidy levels than the rest of the genus (n=5 or 10), with n=15 reported for L. virgatum, and n=15, 25, and
30 reported for L. salicaria (Graham and Cavalcanti, 2001). Lythrum virgatum is described as glabrous, having epicalyx appendages barely equaling the length of the lobes of the calyx, and leaves acute at the base; as opposed to the rarely glabrous L. salicaria, with epicalyx appendages longer than the lobes of the calyx, and round or cordate leaf bases. Some taxonomists have questioned this distinction in both their native and introduced ranges (Rendall, 1989; S. Graham pers.com). These questions have become particularly important to ecosystem managers and horticulturists in the northeastern US and Canada, where plants of both descriptions are popular cultivars, and pervasive damaging invaders in natural habitats (Lindgren and Clay, 1993; Anderson and Ascher,
1993, Strefeler et al., 1996a, 1996b). The sale or cultivation of L. salicaria and any of its
67
cultivars has been outlawed in several northern states, however these laws often do not include L. virgatum and its cultivars (Strefeler et al., 1996b). Adding to the confusion,
most horticulturists and law enforcement officials do not clearly understand the
distinction (Anderson and Ascher, 1993). In an effort to clarify this situation, studies
were conducted comparing isozyme variation among cultivars from both groups and wild
populations (Strefeler et al., 1996 a, 1996b). These studies concluded that it was not
possible to distinguish between naturalized wild populations of L. salicaria and L.
virgatum, or their cultivars. However, they did find that the L. salicaria-L. virgatum-
cultivar group was genetically distinct from wild populations of L. alatum that co-occur
with L. salicaria and L. virgatum.
Conclusions
The current analysis includes sequence information from chloroplast and nuclear
markers from ten Lythrum taxa in addition to Peplis and three appropriate outgroup taxa.
Seven of the ten Lythrum species have never previously been sampled, and their addition
to these analyses has significantly improved our understanding of molecular variation and
phylogenetic relationships within the genus. There is now substantial molecular evidence
to support Webb’s (1967) assertion that Peplis should be included as species of Lythrum.
Additional sampling of the remainder of Eurasian species (especially those with “Peplis-
like” morphology) will be required to see if the reduced “Peplis” morphology has arisen
multiple times within the genus, or if the former species of Peplis form a sister clade to
the other European Lythrum. It is also clear that the current subgeneric classification
does not reflect evolutionary relationships, and will need to be revised. Some critical
68 species remain to be sampled (especially L. rotundifolium, L. maritimum, L. tribracteatum, and L. paradoxum, each representing other important morphological variation or range extensions missing in the current analysis), before useful recommendations concerning revisions to taxonomy can be made. Evidence suggests that Lythrum is composed of several species complexes and/or wide-ranging, highly variable species. Population level sampling (especially in areas of symparty), more variable genetic markers, and an improved understanding of the distribution of morphological variation across the phylogeny, will be required to make these distinctions.
69
Literature Cited
Anderson NO, Ascher PD. 1993. Male and female fertility of loosestrife (Lythrum) cultivars. Journal of the American Society of Horticultural Science. 118(6): 851-858.
ATTRA. 1996. Purple loosestrife: Public enemy #1 on federal lands. ATTRA Interior Helper 1(2):2.
Blossey B, Skinner LC, and Taylor J. 2001. Impact and management of purple loosestrife (Lythrum salicaria) in North America. Biodiversity and Conservation 10: 1787-1807.
Felsenstein J. 1985. Confidence limits on phylogenies: An approach using the bootstrap. Evolution, 39: 783-791.
Graham A, Nowicke J, Skvarla J, Graham S, Patel V, Lee S. 1987. Palynology and systematics of the Lythraceae. III. Genera Physocalymma through Woodfordia, addenda, and conclusions. American Journal of Botany 77: 159-177.
Graham A, Graham S, Nowicke J, Patel V, Lee S. 1990. Palynology and systematics of the Lythraceae III. Genera Physocalymma through Woodfordia, addenda, and conclusions. American Journal of Botany 77: 159-17.
Graham SA. 1964. The genera of Lythraceae in the Southeastern United States. Journal of the Arnold Arboretum. 45: 235-250.
Graham SA. 1975. Taxonomy of the Lythraceae in the Southeastern United States. SIDA 6(2): 80-103.
Graham SA, Cavalcanti TB. 2001. New Chromosome counts in the Lythraceae and a review of chromosome numbers in the family. Systematic Botany 26: 445-458.
Graham SA, Crisci JV, Hoch PC. 1993. Cladistic analysis of the Lythraceae sensu lato based on morphological characters. Botanical Journal of the Linnean Society 113: 1-33.
Graham SA, Hall J, Sytsma K, and Shi S. 2005. Phylogenetic analysis of the Lytraceae based on four gene regions and morphology. International Journal of Plant Sciences 166(6): 995-1017.
Hickey LJ. 1981. Leaf architecture of Myrtales. XIII International Botanical Congress, Sydney, Australia, Abstracts, p. 131.
70
Huang Y-L and Shi S-H. 2002. Phylogenetics of Lythraceae sensu lato: a preliminary analysis based on chloroplast rbcL gene, psaA-ycf3 spacer, and nuclear rDNA internal transcribed spacer (ITS) sequences. International Journal of Plant Sciences. 163: 215- 225.
Kishino H and Hasegawa M. 1989. Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order of the Hominoidea. Journal of Molecular Evolution. 29: 170-179.
Koehne E. 1903. Lythraceae. In A Engler, [eds]. Das Pflanzenreich 4. 216. (Heft17). 1-326. W Engelmann, Weinheim, Germany.
Lindgren CJ, and Clay RT. 1993. Fertility of “Morden Pink” Lythrum virgatum L. transplanted into wild stands of L. salicaria in Manitoba. HortScience. 28 (9): 954.
Maddison WP and Maddison DR. 2003. Mesquite: a modular system foe evolutionary analysis. Version 1.05. http://mesquite-project.org.
Posada D and Crandall KA. 1998. Modeltest: Testing the model of DNA substitution. Bioinformatics. 14: 817-818.
Prager EM and Wilson AC. 1988. Ancient origin of lactabumin from lysosome: analysis of DNA and amino acid sequences. Journal of Molecular Evolution. 27: 326- 335.
Rendall J. 1989. The Lythrum story: A new chapter. Minnesota Horticulture. 117(2): 22-24.
Rozen S and Skaletsky HJ. 1998. Primer3. Code available at http://www- genome.wi.mit.edu/genome_software/other/primer3.html.
Schwarzbach AE and Ricklefs RE. 2000. Systematic affinities of Rhizophoraceae and Anisophylleaceae, and intergeneric relationships within Rhizophoraceae, based on chloroplast DNA, nuclear ribosomal DNA, and morphology. American Journal of Botany 87(4): 547-564.
Shimodaira H and Hasegawa M. 1999. Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Molecular Biology and Evolution. 16: 1114-1116.
Shinners LH. 1953. Synopsis of the United States species Lythrum (Lythraceae). Field and Lab. 21: 80-89.
71
Strefeler MS, Darmo E, Becker RL, and Katovich EJ. 1996a. Isozyme characterization of genetic diversity in Minnesota populations of Purple Loosestrife, Lythrum salicaria (Lythraceae). American Journal of Botany 83(3): 265-273.
Strefeler MS, Darmo E, Becker RL, and Katovich EJ. 1996b. Isozyme variation in cultivars of Purple Loosestrife (Lythrum sp.). HortScience 31(2): 279-282.
Swofford DL. 2001. PAUP*. Phylogenetic Analysis Using parsimony (*and other Methods). Sinauer Associates, Sunderland MA.
Templeton AR. 1983. Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to humans and apes. Evolution.37: 221-244.
Tobe H and Raven PH. 1983. An embryological analysis of the Myrtales: its definition and characteristics. Annals of the Missouri Botanical Garden 70: 71-94.
Tobe H, Raven PH and Graham S. 1986. Chromosomes counts for some Lythraceae sens. str. (Myrtales) and the base number of the family. Taxon. 35:13-20.
USDA, 2007. Plants Database. http://plants.usda.gov.
Webb DA. 1967. Genetic limits in European Lythraceae. Feddes Repert Spec Nov Regni Veg. 74: 10-13.
Webb DA. 1968. Lythraceae in Flora Europaea 2: 300-303.
Zwickl, D. J., 2006. Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. Ph.D. dissertation, The University of Texas at Austin.
CHAPTER 3
Evolution of Heterostyly in the Lythraceae (Myrtales): Approaches to Ancestral Character State Reconstruction Using a New Molecular Phylogeny
Abstract
Heterostyly is a rare polymorphism in plants in which flower morphs differ
reciprocally in the height of their anthers and stigmas, usually accompanied by various
incompatibility systems and other ancillary characters. The mode of inheritance, function, and adaptive significance of heterostyly have been the focus of extensive study, however its evolutionary history remains poorly understood. Well-resolved phylogenies of groups having both heterostylous and homostylous members could provide a source of evidence to compare current evolutionary hypotheses. Lythraceae is one of very few families containing heterostylous species, and a phylogeny of the family could significantly contribute to our understanding of the evolutionary history of this interesting specialized breeding system. In this study, we perform ancestral character state reconstruction of heterostyly using the current phylogeny of the Lythraceae using parsimony and maximum likelihood optimization methods. Various reconstruction methods and character weighting schemes provide ambiguous or conflicting results, with estimates ranging from 5 independent origins of heterostyly, to a single origin for the
entire family followed by repeated parallel reversions to homostyly. While the direction
of change is somewhat ambiguous based on these analyses, multiple changes have
72 73 undoubtedly occurred, and there is some evidence suggesting that a single evolutionary gain followed by multiple parallel losses may be likely.
74
Introduction
Heterostyly is a genetically controlled floral polymorphism in which a plant species
may have either two (distyly) or three (tristyly) flower morphs differing reciprocally in
the height of their anthers and stigmas. This condition is thought to promote outcrossing
and increase pollination precision. Having spatial separation of the anthers and stigmas
within a flower (herkogamy) is relatively common among flowering plants. However
heterostyly is a special and rare case in which this morphological variation is reciprocal,
and usually accompanied by various incompatibility systems and other ancillary
characters (Barrett, 1992) (Figure 3-1). Evidence suggests that distyly may have evolved
independently at least twenty-three different times among flowering plant families (Lloyd
and Webb, 1992), while tristyly is found much less frequently (occurring in only five
families: the Lythraceae, Amaryllidaceae, Connaraceae, Oxalidaceae, and
Pontederiaceae). The complex and unusual aspects of this breeding system, in addition to
its simple genetic basis, and the ease with which it can be identified under field
conditions, have made it a “model system” for addressing a variety of questions in evolutionary biology (Barrett, 1992). The mode of inheritance, function, and adaptive significance of heterostyly have been the focus of extensive study (see review in Barrett,
1992), beginning with publications by Darwin (1865, 1877). In spite of this attention, little is known about its evolutionary history (Graham and Barrett, 2004).
There are two competing models for the evolution of heterostyly (see reviews in
Barrett, 1992). These primarily differ in the order in which various characters evolved, and in what type of breeding system is hypothesized to be ancestral. Darwin (1877)
75
Figure 3-1. Schematic representation of stamen and style configuration in distylous and tristylous flowers. Arrows indicate compatible pollinations, while other pollen transfer combinations usually result in reduced or no seed set (redrawn from Barrett, 1992). 76
stigma style anther
floral tube
ovary Distyly
Tristyly 77 suggested that reciprocal differences in stigma and anther height (reciprocal herkogamy) may have evolved first due to selection for increased accuracy of pollen transfer, and that self-incompatibility mechanisms and other ancillary features developed later (an idea supported by Lloyd and Webb, 1992). Other researchers have challenged Darwin’s hypothesis, suggesting that the features of heterostyly, including self-incompatibility, evolved simultaneously (Mather and de Winton, 1941), or that self-incompatibility evolved before the flowers became herkogamous (Baker, 1966; Ganders, 1979).
Charlesworth and Charlesworth (1979) developed the first mathematical model concerning the evolution of distyly. They hypothesized that distyly arose from a homostylous ancestor, and that self-incompatibility mechanisms evolved first to prevent selfing, followed later by reciprocal herkogamy in order to increase the efficiency of cross pollination. More recently, Lloyd and Webb (1992) argued that the ancestors of heterostylous species were approach herkogamous (having a style longer than the stamens) rather than homostylous. They suggested that reciprocal herkogamy developed first (reviving Darwin’s idea), and that incompatibility mechanisms and ancillary characters developed secondarily as pleiotropic effects of genes involved in the competitive ability of pollen.
Phylogenetic analyses of groups with heterostylous members could provide a source of evidence to assess these models (Barrett, 1992). However, very few of these studies have been conducted, and in most cases they were unable to discern ancestral states or evolutionary pathways (Schoen et al., 1997; Kohn et al., 1996). Some recent progress has been made in the genus Primula (Primulaceae; Mast et al., 2006; Guggisberg et al.,
78
2006), which is well-known for having many distylous members, and in the monocot
genus Narcissus (Amaryllidaceae) which also exhibits a range of stylar conditions
(Graham and Barrett, 2004; Perez-Barrales et al., 2006). The ancestral state in Primula
appears to have been distyly, with all stylar monomorphism in the genus being derived.
However, it is still unclear how many times distyly has arisen in the family (Mast et al.,
2006). In Narcissus, distyly appears to have arisen from simple stigma height dimorphism, but problems with resolution make exact interpretations difficult (Graham and Barrett, 2004). A well-resolved phylogeny of the Lythraceae could significantly contribute to this question because of its homostylous, distylous, and tristylous members.
The Lythraceae is historically well known for having heterostylous members, with six
genera containing heterostylous species: Decodon, Pemphis, Adenaria, Nesaea, Rotala
and Lythrum. A large number of studies have been conducted on the inheritance,
breeding system, and population variation of the tristylous Lythrum salicaria (Darwin,
1865, 1867; Haldane, 1936; Fisher and Mather, 1943; Heuch 1980; Oneil, 1992;
Andrsson, 1994; Oneil and Schmitt, 1993; Anderson and Ascher, 1995, 2000; Agren and
Ericson, 1996; Agren, 1996; Eckert et al., 1996; Mal, 1998; Mal et al., 1999; Hermann et
al., 1999; Mal and Hermann, 2000). Similarly detailed work has also been conducted for
L. junceum (Dulberger, 1970; Ornduff, 1975), Pemphis acidula (Lewis and Rao, 1971),
and Decodon verticillatus (Eckert and Barrett 1992, 1993a, 1993b, 1994, 1995; Eckert
and Mavraganis, 1996).
In his discussion of the modifications of tristylous breeding systems, Weller (1992)
states that “without a better understanding of the phylogeny of Lythraceae, it is difficult
79
to determine whether tristyly in Lythrum, Nesaea, and Decodon represents a primitive breeding system retained in these genera, or a breeding system that has evolved on several occasions in the family.” Subsequent molecular phylogenetic studies of the family (Huang and Shi, 2002; Graham et al., 2005) could not fully resolve basal relationships, and so have been unable to address this question. In this study, we perform a preliminary phylogenetic reconstruction of stylar polymorphisms on the most recent phylogenetic analyses of the Lythraceae (Chapter 1) in an effort to address this question.
80
Methods
Taxon Sampling and Molecular Data
For these analyses, 28 out of the 32 genera of Lythraceae were sampled for
generation of the molecular phylogeny, including all genera having any heterostylous
members. Select members of the Onagraceae, previously shown to be closely allied to
the Lythraceae (Johnson and Briggs 1984, Conti et al., 1997, Sytsma et al., 2004), were
used as outgroups. In order to maximize taxon sampling and consider as much of the available data as possible, some sequences (rbcL, trnL-trnF, and psaA-ycf3, and ITS)
were retrieved from GenBank (Huang and Shi, 2002; Graham et al., 2005). Some of the
GenBank sequences were incomplete, making it necessary to resequence some taxa for
rbcL and trnL-trnF. Additional taxa were also resequenced in an effort to reduce the
possibility of sequencing error as a source of noise in the data set, especially given that several of the previously published sequences were produced with manual sequencing techniques (Huang and Shi, 2002; Graham et al., 2005). No additional sequencing was carried out for psaA-ycf3. Voucher information and GenBank numbers for all sampled taxa and gene regions are provided in Table 3-1.
Total DNA used to generate new sequence data for this study was extracted from
fresh tissue, dried material preserved in silica gel, or dried herbarium specimens. DNA
extractions were conducted using a modified CTAB protocol (as in Graham et al., 2005),
or the QIAGEN DNeasy Plant Mini Kit (QIAGEN Inc., Chatsworth, CA). Some
difficulties exist with DNA extraction in this family, and extended storage in the freezer
(> 1 year) and/or substantial dilution is often required prior to successful PCR (pers.
81
Table 3-1. Voucher information for all taxa sampled for this study, and the character coding that was used for ancestral character state reconstruction of heterostyly in the Lythraceae. Information on the molecular data collected for each of these taxa (including GenBank accession numbers) can be found in Table 1-1.
Taxon Voucher Character State Adenaria floribunda Breedlove 38133 (CAS) 1 Ammannia latifolia Liogier 10314 (MO) 0 Capuronia madagascariensis D’Arcy 15439 (MO) 0 Cuphea lanceolata Shi 99090201 (SYS) 0 Decodon verticillatus Graham 917 (MO) 1 Didiplis diandra my collection 1/28/2004 -preparing voucher 0 Duabanga molucanna Chai s.n. in 1990, (MO) 0 Galpinia transvaalica Balsinhas 3263 (MO) 0 Ginoria americana S. Graham 1137 (MO) 0 Heimia salicifolia S. Graham 1062 (MO) 0 Koehneria madagascariensis D’Arcy & Rakotozafy 15317 (MO) 0 Lafoensia acuminata Neil 8930 (MO) 0 Lagerstroemia indica USA: Texas, cultivated, collected by S. Graham, no voucher 0 Lawsonia inermis Correll 45915 (TEX) 0 Lourtella resinosa S. Graham 1116 (MO) 0 Lythrum ovalifolium M. C. Johnston s.n. in Oct. 1975 (MO) 1 Lythrum virgatum cultivated, Gottingen Bot. Garden, collected by S. Graham; no voucher 1 Nesaea aspera Drummond 11446 (SRGH) 1 Pehria compacta P. Berry s.n. in 1979 (MO) 0 Pemphis acidula Liao 1150 (A) 1 Peplis portula Montezuma s.n. (MO) 0 Physocalymma scaberrimum Shi, S. H. 861 (SYS) 0 Pleurophora anomala Cavalcanti et al. 368 (MO) 0 Punica granatum cultivated, collected by S. Graham; no voucher 0 Rotala ramosior S. Graham 1028 (MO) 1 Sonneratia caseolaris Huang 990435 (SYS) 0 Tetrataxis salicifolia Lorence 1231 (MO) 0 Trapa natans no voucher -see Conti et al. 1993 0 Woodfordia fruticosa cultivated, USDA Agricultural Station, Homestead, Fl. (MO) 0 0 = homostylous; 1 = heterostylous 82
obs.). PCR amplification was carried out using standard PCR techniques and a MJ
Research Thermocycler. These reactions were performed in a volume of 25 µl containing
6 µl of a Tricine buffer mix (final concentration: 30 mM Tricine pH 8.4, 2.0 mM MgCl2,
50 mM KCl, and 0.25 mM each dNTP), 0.05 µg of each primer, 1 U Taq polymerase, and approximately 0.5 ng genomic DNA. In a few cases, genomic DNA needed to be substantially diluted (0.005-0.05 ng /25 µl reaction) before PCR was successful. The thermocycler program included 1 minute for initial strand separation at 94˚C; followed by
37 cycles of 1 minute at 94˚C, 45 seconds at 50˚, 4 minutes at 72˚C; and a final 10 minute step at 72˚C. PCR products were isolated using 1.5% agarose gels and purified using the
QIAGEN QIAquick Gel Extraction Kit (QIAGEN Inc., Chatsworth, CA). Cycle sequencing reactions were conducted in a 10 µl volume, using 2 µl ABI BigDye cycle sequencing terminator chemistry (Perkin-Elmer Corp.), 2 µl Half Term Dye (Bioline
Inc.), 2.0 ng each primer, and approximately 50 ng of template DNA. Reactions were purified using sephadex in Centrisep columns according to manufacturer’s instructions, and then were completely lyophilized using a ThermoSavant DNA120 SpeedVac.
Sequencing was conducted on an Applied Biosystems 3730 DNA Analyzer at the
Sequencing Core Facility, Brigham Young University, Provo, Utah. Completed sequences were edited, assembled, compiled with GenBank sequences, and initially aligned using Sequencher 4.1 (GenesCode Co. Ann Arbor, MI). Manual adjustments were then made to ensure proper alignment.
83
Ancestral Character State Reconstruction
Phylogenetic analyses were conducted as described in Chapter 1, and subsequent
character state reconstructions were conducted using the fully resolved Bayesian
consensus tree (identical to the best ML tree). To investigate possible evolutionary
scenarios of heterostyly in the Lythraceae, the breeding system was coded as present
(heterostylous = 1) or absent (homostylous = 0; Table 3-1), and mapped onto the
phylogeny of the family using maximum parsimony and maximum likelihood
optimization methods in Mesquite (v. 1.05, Maddison and Maddison 2003). The overall likelihood values of the data under the alternative 1-rate (Mk1) and 2-rate (asymmetric)
models were calculated and compared using likelihood ratio tests in Mesquite (Pagel
1999). In all cases, the 2-rate model was found to have a significantly better fit to the
data (a difference in log-likelihood scores > than 2.0; Edwards, 1972), and so was used in
all ML character state reconstructions reported here.
A likelihood bias could have been caused by including more than one
heterostylous species of Lythrum in character state reconstructions at the family level when all other genera were sampled only once. This possibility was explored by conducting multiple analyses: 1) including only one heterostylous species of Lythrum; 2) including both Lythrum species, but coding one as heterostylous and one as homostylous; and 3) including both species of Lythrum and coding them both as heterostylous. No significant difference was found in the likelihood rate model chosen, or in character state reconstructions at any of the nodes among these three, so results are presented here for reconstructions including both heterostylous species.
84
Because maximum likelihood methods rely on relative branch lengths to infer
ancestral states, errors or inconsistencies in branch lengths may have a significant effect
on reconstructions (Cunningham et al., 1998). Missing data in the current analyses does
not appear to affect placement of taxa (see discussion in Chapter 1), but it is likely
affecting relative branch lengths (Wiens, 2006). In an attempt to address this issue,
reconstructions were conducted using a tree with branch lengths inferred from the
Bayesian analysis, and then repeated using a tree with all branch lengths set to equal
length using the ‘alter/transform branch lengths’ function in Mesquite.
For parsimony-based character state reconstructions, the most conservative
assumption is to treat any character as unordered, with transitions in either direction
being equally weighted. However, in the case of heterostyly, there is strong evidence to
suggest that more than one independent origin of this complex character is much less
likely than its loss. Previous parsimony reconstructions of heterostyly in other taxa
(Kohn et al., 1996; Schoen et al., 1997; Guggisberg. 2006; Mast et al., 2006) addressed
this problem using various weighting schemes in which the loss of heterostyly is favored
over its gain (ranging from 2:1 up to 20:1). In the current study, the result of equal weighting, and of differential weighting using a 2:1 ratio (loss favored over gain), is compared to a weighting scheme in which the loss to gain ratio is set equal to the ratio of forward and backward rates calculated from the maximum likelihood ancestral state reconstructions (=3.14 : 1). Using this weighting scheme and the ‘state changes and stasis’ tool in MacClade (v. 4.03 Maddison and Maddison, 2003), the minimum,
85
maximum, and total number of transitions between monomorphism and heterostyly in either direction were estimated over all 9000 trees produced from the Bayesian analyses.
86
Results
Using the data currently available and the most commonly used maximum
likelihood and parsimony based approaches, attempts at reconstructing the evolutionary
history of heterostyly in the Lythraceae results in a substantial amount of uncertainty.
Different reconstruction methods and character weighting schemes provide ambiguous or
conflicting results, with estimates ranging from 5 independent origins, to a single origin
for the entire family followed by repeated parallel reversions to homostyly.
Likelihood estimates clearly support a two-rate model, and the rates of character- loss were consistently estimated to be over 3 times the rates of character gain. The likelihood of the data using inferred branch lengths and the asymmetrical two-rate model
(-lnL = 16.03; rate of gain = 2392.72; rate of loss = 7519.98) was significantly better than that of the equal rates model (-lnL = 20.03; rate of change = 23.08) based on the
likelihood ratio test (difference in lnL > 2.0). Even when using the tree having
transformed branch lengths, the likelihood of the data given the two-rate model (-lnL =
16.03; rate of gain = 3.74; rate of loss = 11.77) is still significantly better than that given
the equal rates model (-lnL = 18.37; rate of change = 0.18), and the ratio of forward and
backward rates remains approximately the same.
Unfortunately, no internal nodes on the tree were found to have ancestral
character state reconstructions considered to be statistically significant. The proportional
likelihoods of each of the internal nodes are relatively equal, and equivalent to the ratio of
backward and forward rates calculated during the analysis (3.14:1; Figure 3-2). These
results do not change when calculated using a tree with transformed branch lengths (all
87
Figure 3-2. Likelihood reconstruction of the evolutionary history of heterostyly using the asymmetrical 2-parameter model (forward rate = 2392.72 and backward rate = 7519.98), and a majority-rule Bayesian consensus of 9000 trees based on the combined analysis of all available molecular data. Terminal taxa represented with a black circle are heterostylous, while taxa represented with a white circle are monomorphic. Pie charts at each node indicate proportional likelihood calculated for each state, none of which are statistically significant in these analyses. 88
89
equal to one). This suggests that differences in the relative branch lengths inferred from
the phylogenetic analyses are not responsible for the lack of statistically significant
character reconstructions for the internal nodes. It seems likely that the lack of
significance could be primarily due to the scattered distribution of the heterostylous
condition among the various major clades in the phylogeny. Given the distribution of
this character within the family, a large number of changes in one or both directions have
undoubtedly occurred, with maximum parsimony estimates of the total number of
transitions from across all 9000 Bayesian trees ranging from 9-14.
Equally-weighted parsimony analyses suggest 5 independent origins of
heterostyly in the family (Figure 3-3), however, this weighting scheme seems
inappropriate given the complexity of this character, and the rate bias towards losses in
the likelihood analyses. Despite the lack of nodal significance in the maximum likelihood reconstructions, evidence suggests that independent losses have occurred more
frequently than independent gains. Applying the maximum likelihood rate bias as a
weighting scheme (3.14:1), parsimony estimates of character state gains (transitions from
0 to 1) across all 9000 Bayesian trees range from 0-2, while estimates of character state
losses (transitions from 1 to 0) range from 7-14. If these estimates are accurate, it
appears that heterostyly may have been the ancestral condition in the family, despite the
fact that there are currently more homostylous than heterostylous taxa. Even when
weighting losses 2:1 over gains, parsimony reconstruction predicts heterostyly to be the
basal condition in the family (Figure 3-4), and use of the 3.14:1 weighting scheme
suggested by likelihood analyses further strengthens this argument (Figure 3-5).
90
Figure 3-3. Equal weight parsimony optimization of the evolutionary history of heterostyly using a majority-rule Bayesian consensus of 9000 trees based on the combined analysis of all available molecular data. Terminal taxa represented with a black circle are heterostylous, while taxa represented with a white circle are monomorphic. Pie charts at each node indicate proportional likelihood calculated for each state, none of which are statistically significant in these analyses. 91
92
Figure 3-4. Parsimony reconstruction of the evolutionary history of heterostyly using a 2:1 weighting scheme (favoring the loss of heterostyly over its gain), and a majority-rule Bayesian consensus of 9000 trees based on the combined analysis of all available molecular data. Terminal taxa represented with a black circle are heterostylous, while taxa represented with a white circle are monomorphic. Pie charts at each node indicate proportional likelihood calculated for each state. 93
94
Figure 3-5. Parsimony reconstruction of the evolutionary history of heterostyly using a weighting scheme in which the loss to gain ratio was set equal to the ratio of forward and backward rates calculated from the maximum likelihood analyses (3.14:1), and a majority-rule Bayesian consensus of 9000 trees based on the combined analysis of all available molecular data. Terminal taxa represented with a black circle are heterostylous, while taxa represented with a white circle are monomorphic. Pie charts at each node indicate proportional likelihood calculated for each state. 95
96
Discussion
Limitations of Current Models
Using phylogenies to reconstruct ancestral character states can provide a powerful
mechanism for understanding historical patterns in biological evolution (Pagel, 1999,
Swofford and Maddison, 1998). However, results of these analyses depend very strongly
on the quality of available phylogenetic information that you start with, and the approach
(model) used for reconstruction (Cunningham et al., 1998). Although our understanding
of phylogenetic relationships in the Lythraceae continues to improve, uncertainty still
limits our ability to perform ancestral character state reconstructions of heterostyly for
these taxa in a statistically rigorous way. It now seems likely that the available models
may never be able to conclusively address all of the difficulties faced in phylogenetic
analyses, including those in the current analyses of the Lythraceae.
Historically, maximum parsimony was the most widely used method for reconstructing ancestral character states (Cunningham et al., 1998). This method maps characters on to a given topology by minimizing the number of character transitions.
Studies have shown that under some conditions this type of analysis can be misleading, including cases when there are considerable differences in relative branch lengths, when rates of evolution are rapid, and when the probabilities of gains to losses are not equal
(Felsenstein, 1973; Frumhoff and Reeve, 1994; Collins et al., 1994; Maddison, 1994,
Cunningham et al., 1998). Each of these conditions could present potential problems in reconstructing evolutionary history of heterostyly in the Lythraceae. Evidence is increasing that the major lineages in this family may have diversified as a result of an
97
early and relatively rapid radiation followed by long periods of isolation and more
gradual accumulation of changes within the major lineages (see Chapter 1). This could
explain the short, weakly supported internal branches and long terminal branches in the
current molecular phylogenies, and if true, it is likely that no amount of additional data
will be able to resolve this issue. As previously mentioned, there is strong evidence
suggesting that multiple independent origins of heterostyly are probably much less likely
than repeated loss, and that this might be particularly true in the Lythraceae (see
discussion below). Parsimony reconstructions of heterostyly in other taxa have addressed
this problem using various weighting schemes in which the loss of heterostyly is favored
over its gain (ranging from 2:1 up to 20:1; Kohn et al., 1996; Schoen et al., 1997; Mast et
al., 2006). However, different weighting schemes can result in very different character
state reconstructions, and the choice of one scheme over the other is difficult to justify.
Additionally, there are no available comparative statistics for evaluating one parsimony
reconstruction with respect to another.
More recently, maximum likelihood approaches of character reconstruction have
been used. These methods use a given model of character evolution to estimate the probabilities of all possible character state reconstructions at every node on a given
topology (Schluter et al., 1997; Pagel, 1999). These probabilities are determined by the
distribution of characters in terminal taxa, the rate of evolution assigned to the character,
and by the length of the branches between nodes (Cunningham et al., 1998; Cunningham,
1999). Maximum likelihood methods are often favored because they can estimate the
relative probability of each character state at every node, resulting in a measure of
98
statistical support for a particular character reconstruction (Maddison, 1995). Various
models differ in their assumptions about the rates of change between character states,
which can be assumed to be equal or unequal (Mooers and Schluter, 1999). One
limitation to maximum likelihood approaches is that they assume that every character
evolves at a constant rate across the entire tree, which may underestimate changes on
short internal branches that result from an adaptive radiation (Schluter et al. 1997), as is
suspected for the Lythraceae.
Taxon sampling in both outgroups and ingroups is also known to affect the
accuracy of ancestral state estimation using any method (Maddison, 1984; Salisbury and
Kim, 2001, Mast et al. 2006). Choice of outgroup can affect the ability of reconstruction
methods to resolve the ancestral character state at the base of the ingroup (Maddison,
1984). Because of the lack of heterostylous breeding systems in any other taxon closely
related to the Lythraceae, the particular choice of outgroup taxa will not likely influence
the reconstruction of heterostyly within the family, however it will affect any attempt to determine what type of breeding system was likely to have preceded the evolution of heterostyly within the Lythraceae. Sampling of ingroup taxa can have a substantial effect, increasingly a problem in small clades or sampling regimes that are biased toward the presence or absence of the character in question (Salisbury and Kim, 2001). This is
not likely to be a problem in a reconstruction at the family level for Lythraceae
considering that only three genera remain to be sampled and all genera with heterostylous members are represented in the analysis. However, future attempts to reconstruct the
99
evolution of heterostyly focusing on Lythrum would obviously be inconclusive without a
more thorough sampling, especially of Eurasian taxa.
It is evident that further statistical support for reconstructions of heterostyly in the
Lythraceae will require an improved confidence in the phylogeny (especially among the
basal branches) and/or novel methods of analyses that can more appropriately account for
the remaining phylogenetic uncertainties. New evidence suggests that a more recently
proposed Bayesian Markov Chain Monte Carlo procedure could provide a more rigorous
way to address uncertainties that can be obscured by other methods focusing on a single
set of model parameters and associated ancestral states (Vanderpoorten and Goffinet,
2006). This method derives the posterior probability distribution of rate coefficients and
ancestral character states, allowing a range of model parameters and ancestral states to be
sampled according to their posterior probabilities (Pagel and Lutzoni, 2002; Pagel et al.,
2002).
Repeated Loss of Heterostyly in Lythraceae
Despite current difficulties in obtaining statistical support for any particular ancestral state reconstruction, evidence suggests that losses of heterostyly (or reductions from tristyly to distyly) within the Lythraceae may have occurred multiple times, and that heterostyly may in fact be the primitive breeding system in the family. This is in contrast to conventional assumptions of gains on separate occasions, as this data may have traditionally been interpreted, and provides increased support for the practice of weighting gains more heavily than losses.
100
Evidence suggests that many complex biological characters have a low
probability of origin, but can be lost or reduced by the action of a few genes. Once lost,
these characters are generally thought to be difficult to regain, especially after a long
period of time (Maddison, 1994; Hart et al, 1997, Strathmann and Eernisse, 1994).
Heterostyly is probably not an exception, and it has been postulated that the breakdown
of distyly to monomorphism has a simple genetic basis, being brought about by
recombination in the distyly locus, or by changes in modifier loci (Lewis and Jones
1992). It is now clear that heterostyly has frequently been modified, giving rise to other
sexual systems (see reviews in Ganders, 1979; Barrett, 1988 and 1989; and Weller, 1992) including alternate outcrossing systems (Baker, 1966; Weller, 1976; Lloyd 1979;
Ganders, 1979), self-fertilization (Ornduff, 1972; Barrett and Shore, 1987; Barrett et al.,
1989), or loss of sexuality (Baker, 1966; Berry, Tobe, and Gomez, 1991; Barrett, Eckert, and Husband, 1993); and that these modifications can occur for random and/or deterministic reasons (Ornduff, 1972; Weller, 1976; Barrett et al., 1989; Eckert and
Barrett 1992, 1995).
Loss of floral morphs from populations in heterostylous species has been well
documented (see review in Barrett, 1992). Within the Lythraceae, detailed population-
level studies have been conducted on both Lythrum salicaria (Schoch-Bodmer, 1938;
Heuch, 1980; Eckert and Barrett, 1992) and Decodon verticillatus (Eckert and Barrett,
1992, 1993a, 1993b,1994, 1995; Eckert and Mavraganis, 1996) in a effort to understand
the relative importance of drift and selection in explaining variation in style morph
frequencies in these species. In a comparison of populations from both species, Eckert
101
and Barrett (1992) found that a loss of morph was more likely in small populations, with
the short morph being more commonly lost, followed by the medium morph, and rarely in the long morph. Comparisons of populations of both L. salicaria and D. verticillatus from their native and introduced ranges suggested the rate of morph loss increased during colonization and migration. This tendency towards stochastic morph loss might be evolutionarily significant when associated with long-distance dispersal events or other forms of geographic isolation that would prevent reestablishment of normal ratios of floral polymorphism (Eckert and Barrett, 1992).
Comparisons of floral morphology among distylous and tristylous species of
Lythrum and Nesaea (the only genera in the Lythraceae having both distylous and
tristylous species) seem to provide evidence that these population-level processes could
have played a role in speciation in these taxa. In a comparison of the floral morphology
of tristylous L. junceum and some distylous North American taxa including L.
californicum, Ornduff (1979) noted that these distylous species have a single set of six
stamens alternating with the petals in both short and long morphs, while L. junceum has
twelve stamens arranged in two whorls of six. In L. junceum, stamen insertion of the
short set of stamen is opposite the petals, while stamens of mid- and long-stamen sets
alternate with the petals. Consequently, the stamens and styles of the distylous species
are located in the same relative positions as stamens and styles in the mid and long morphs of L. junceum. Conducting pair-wise comparisons of style length, anther length, and pollen size between morphs, Ornduff (1979) also found that the long:mid ratios of these measurements in L. junceum were very similar to long:short ratios in the North
102
American species. He interpreted this as evidence that distyly in the North American
taxa was the result of the loss of the short-styled morph of a tristylous ancestor (through
the loss of the lowest position stamens inserted opposite the petals). The new molecular
phylogeny of Lythrum presented here(chapter 2) suggests that L. junceum is not ancestral
to the North American taxa, as was previously suggested by the placement of these taxa
in the section (Euhyssopifolia) by Koehne (1903). This hypothesis will need to be
reevaluated in light of these new data. However, the low level of genetic diversity found
within the North American clade is suggestive of a bottleneck event (Chapter 2), and
could be further evidence that stochastic processes were involved in this case. Ornduff
(1978, 1979) speculated that morph loss has likely occurred more than once within
Lythrum, pointing out that L. rotundifolium (the only species of Lythrum found in Africa)
is also distylous, but in this case has twelve stamens and may be missing the mid morph.
He also predicted at least two origins of distyly in the genus Nesaea, observing that in
one section (Salicariastrum) trimorphic species have two sets of stamens and distylous
species have one (as in the Lythrum comparison); while in section Heimiastrum, both
trimorphic and dimorphic species have two sets of stamens. No morphological or
molecular data are available with which to evaluate these hypotheses in Nesaea, but should be an important component of future studies of the evolution of heterostyly in this family.
There is also evidence to suggest that some uniformly distylous or homostylous
genera in the Lythraceae may not have been so in the past. After an examination of the floral morphology of Pemphis acidula, Lewis and Rao (1971) concluded that the species
103
was trimorphic in the past and is still “adjusting” to being distylous. They observed that this species has two sets of stamens, and that the whorl of stamens that would be equivalent to the mid whorl in a fully trimorphic flower converges towards being long (as found in short morph) in some populations; and being shorter (as found in short morph) in other populations. They also found stamen length and pollen size in this mid-like whorl varies considerably among populations, while no variation is seen in these characters in the short or long morphs. Similarly, as part of his classic work on the evolution of plant breeding systems, Darwin (1877) suggested that the uniformly monomorphic genus Lagerstroemia may have been formerly heterostylous, citing considerable population variation in the length and structure of the two sets of stamens, and the fact that many flowers produce pollen of two different colors; a characteristic found in other tristylous members of the Lythraceae.
While it is not currently possible to unequivocally discern the evolution of
heterostyly in the Lythraceae based on ancestral character state reconstructions, this study
has resulted in an improved understanding of some possible evolutionary relationships
within these taxa. Future progress in this area will require fewer phylogenetic
uncertainties (requiring improved sampling, and more data to increase our confidence in
basal relationships and relative branch lengths), and a model that can address the unique
challenges that are posed by this unusual family of plants and their unusual breeding
systems.
104
Literature Cited
Anderson NO, and Ascher PD. 1995. Style morph frequencies in Minnesota populations of Lythrum (Lythraceae). II. Tristylous L. salicaria. Sexual Plant Reproduction. 8: 105-112.
Baker HG. 1966. The evolution, functioning and breakdown of heteromorphic incompatibility systems I. The Plumbaginaceae. Evolution. 18: 507-512.
Barrett SCH. [ED.]. 1992. Evolution and function of heterostyly. Springer-Verlag, Berlin, Germany.
Charlesworth D and Charlesworth B. 1979. A model for the evolution of heterostyly. American Naturalist. 114: 467-498.
Conti E, Litt A, Wilson PG, Graham SA, Briggs BG, Johnson LAS, and Sytsma KJ. 1997. Interfamilial relationships in Myrtales: molecular phylogeny and patterns of morphological evolution. Systematic Botany 22(4): 629-647.
Cunningham CW, Omland KE, and Oakley TH. 1998. Reconstructing ancestral character states: a critical reappraisal. Trends in Ecology and Evolution. 13(9): 361-366.
Darwin C. 1865. On the sexual relations of the three forms of Lythrum salicaria. Journal of the Linnean Society London. 8:169-196.
Darwin C. 1877. The different forms of flowers on plants of the same species. John Murray, London, UK.
Dulberger R. 1970. Tristyly in Lythrum salicaria. New Phytologist 69: 751-759.
Eckret CG and Barrett SCH. 1992. Stochastic loss of style morphs from populations of tristylous Lythrum salicaria and Decodon verticillatus (Lythraceae). Evolution 46(4): 1014-1029.
Eckret CG and Barrett SCH. 1993a. Clonal Reproduction and patterns of genotypic diversity in Decodon verticillatus (Lythraceae). American Journal of Botany. 80(10): 1175-1182.
Eckret CG and Barrett SCH. 1993b. The inheritence of tristyly in Decodon verticillatus (Lythraceae). Heredity. 71: 473-480.
Eckret CG and Barrett SCH. 1994. Tristyly, self-compatibility and floral variation in Decodon verticillatus (Lythraceae). Biological Journal of the Linnean Society. 53: 1-30.
105
Eckret CG and Barrett SCH. 1995. Style morph ratios in tristylous Decodon verticillatus (Lythraceae) Selection vs Historical contigency. Ecology. 76(4): 1051-1066.
Eckret CG and Mavraganis K. 1996. Evolutionary consequences of extensive morph loss in tristylous Decodon verticillatus (Lythraceae): a shift from tristyly to distyly? American Journal of Botany 83(8): 1024-1032.
Felsenstein J. 1985. Confidence limits on phylogenies: An approach using the bootstrap. Evolution, 39: 783-791.
Fisher RA, Mather K. 1943. The inheritance of style length in Lythrum salicaria. Ann Eugenics 12: 1-23.
Ganders FR. 1979. The biology of heterostyly. New Zealand Journal of Botany. 17:607-635.
Graham SW and Barrett SCH. 2004. Phylogenetic reconstruction of the evolution of stylar polymorphims in Narcissus (Amaryllidaceae). American Journal of Botany. 91(7): 1007-1021.
Graham SA, Hall J, Sytsma K, and Shi S. 2005. Phylogenetic analysis of the Lytraceae based on four gene regions and morphology. International Journal of Plant Sciences 166(6): 995-1017.
Guggisberg A, Mansion G, Kelso S and Conti E. 2006. Evolution of biogeographic patterns, ploidy levels, and breeding systems in a diploid-polyploid species complex of Primula. New Phytologist. 171: 617-632.
Haldane JBS. 1936. Some natural populations of Lythrum salicaria. Journal of Genetics 32: 393-397.
Heuch I. 1979. Loss of incompatibility types in finite populations of the heterostylous plant Lythrum salicaria. Hereditas 92: 53-57.
Heuch I. 1980. Equilibrium populations of heterostylous plants. Theoretical Population Biology 15: 43-57.
Huang Y-L and Shi S-H. 2002. Phylogenetics of Lythraceae sensu lato: a preliminary analysis based on chloroplast rbcL gene, psaA-ycf3 spacer, and nuclear rDNA internal transcribed spacer (ITS) sequences. International Journal of Plant Sciences. 163: 215- 225.
Johnson LAS, and Briggs BG. 1984. Myrtales and Myrtaceae –a phylogenetic analysis. Annals of the Missouri Botanical Garden. 71:700-756.
106
Koehne E. 1903. Lythraceae. In A Engler, [eds]. Das Pflanzenreich 4. 216. (Heft17). 1-326. W Engelmann, Weinheim, Germany.
Kohn JR, Graham SW, Morton B, Doyle JJ, Barrett SCH. 1996. Reconstruction of the evolution of reproductive characters in Pontederiaceae using phylogenetic evidence from chloroplast DNA restriction-site variation. Evolution. 50:1454-1469.
Lewis D, Rao AN. 1971. Evolution of dimorphism and population polymorphism in Pemphis acidula. Forst. Proceedings of the Royal Society of London Ser B. 178: 79-94.
Lloyd DG and Webb CJ. 1992. The evolution of heterostyly. In SCH Barrett [eds.]. Evolution and function of heterostyly. 179-207. Springer-Verlag, Berlin, Germany.
Maddison WP and Maddison DR. 2003. Mesquite: a modular system foe evolutionary analysis. Version 1.05. http://mesquite-project.org
Mast AR, Kelso S and Conti E. 2006. Are any primroses (Primula) primitively monomorphic? New Phytologist. 171: 605-616.
Mather K, and DeWinton D. 1941. Adaptation and counter-adaptation of the breeding system in Primula. Annals of Botany II. 5: 297-311.
O’Neil P, and Schmitt J. 1993. Genetic constrains on the independent evolution of male and female reproductive characters in the tristylous plant Lythrum salicaria. Evolution. 47: 1457-1471.
Ornduff R. 1975. Pollen flow in Lythrum junceum, a tristylous species. New Phytologist 75:161-166.
Pagel M. 1999. Inferring the historical patterns of biological evolution. Nature. 401: 877- 884.
Schoen DJ, Johnston MO, Heureux A-M, and Marsolais JV. 1997. Evolutionary history of the mating system in Amsinckia (Boraginaceae). Evolution 51. 1090-1099.
Swofford DL. 2001. PAUP*. Phylogenetic Analysis Using parsimony (*and other Methods). Sinauer Associates, Sunderland MA.
Sytsma KJ, Litt A, Zihra, ML, Pires JC, Nepokroeff M, Conti E, Walker J, and Wilson PG. 2004. Clades, clocks, and continents: Historical and biogeographical analysis of Mrytaceae, Vochysiaceae, and relatives in the Southern Hemisphere. International Journal of Plant Sciences. 165(4 suppl.): S85-S105.
107
Wiens JJ. 2006. Missing data and the design of phylogenetic analyses. Journal of Biomedical Informatics. 39: 34-42.
Weller SG. 1992. Evolutionary Modifications of Tristylous Breeding Systems. In: Barrett SCH (ed) Evolution and function of heterostyly. Springer-Verlag, Berlin, Germany.
Zwickl, D. J., 2006. Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. Ph.D. dissertation, The University of Texas at Austin.