Relationships among fragrance, phylogeny and pollination in southwestern Nyctaginaceae
Item Type text; Dissertation-Reproduction (electronic)
Authors Levin, Rachel Ann
Publisher The University of Arizona.
Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
Download date 30/09/2021 20:23:11
Link to Item http://hdl.handle.net/10150/280128 INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter fece, while others may be from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.
In the unlikely event that the author dkl not send UMI a completp mar^>.iscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overiaps.
Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6' x 9" black and white photographic prints are available for any photographs or illustrations appearing In this copy for an additional charge. Contact UMI directly to order.
Bell & Howell Information and Learning 300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA 800-521-0600
RELATIONSHIPS AMONG FRAGRANCE, PHYLOGENY AND POLLINATION
IN SOUTHWESTERN NYCTAGINACEAE
by
Rachel Ann Levin
Copyright © Rachel Ann Levin 2001
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF ECOLOGY AND EVOLUTIONARY BIOLOGY
In Partial Fulfillment of the Requirements For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
200 1 UMI Number: 3002512
Copyright 2001 by Levin, Rachel Ann
All rights reserved.
__ (ft UMI
UMI Microform 3002512 Copyright 2001 by Bell & Howell Information and Leaming Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.
Bell & Howell Information and Leaming Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 2
THE UNIVERSITY OF ARIZONA « GRADUATE COLLEGE
As members of the Final Examination Committee, we certify that ve have read the dissertation prepared by Levin
entitled ilelationships among Fragrance. Phvlnppnv anri Pnii inatinn
in Southwestern Nyctaginaceae
and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Pliilosophy
/! yi4c 00 LUtlinda A. ..IcUadi D^e U. I^JkcLuxj^ /'^ Cd Rob^^t if.) Robichaux Date
Robert A./Raguso Date
Q/diOp, /. -BKmrnur^ IS her en .^^ith L. ISronstein Date
Date
Final approval and acceptance of this dissertation is contingent upon the candidate's submission of the final copy of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.
Dissertation Director Date Lucinda A. :tcDade 3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the copyright holder.
SIGNED: 4
ACKNOWLEDGMENTS
This dissertation would not have been possible without the collaboration and assistance of many people. I would especially like to thank my advisor, Lucinda McDade, for encouraging me to join her and Rob Raguso in collaborating on a project studying relationships among fragrance, phylogeny, and pollinators. Further, I greatly appreciate Lucinda's amazing support, her love of systematics, and her considerable help in preparing manuscripts for publication. I cannot thank Rob enough for his enthusiasm and for teaching me how to collect and analyze fragrance and injecting numerous samples into the GC-MS for me. Thanks also to the other members of my committee, Drs. Judie Bronstein and Rob Robichaux, for their helpful comments and criticisms. The fragrance collections from field locations that were often at least a day's drive from Tucson (and often much farther) depended on the kind assistance of Amy Boyd. Thomas Tully, Mariette Manktelow, and David Heam; I am very grateful for their help, photographic expertise, and companionship on these long field trips. I would also like to thank Alice Le Due and Mark Fishbein, who provided much needed seeds of species that I was unable to visit in the field. A hearty thank you to both present and past graduate students of the University of Arizona Department of Ecology and Evolutionary Biology for discussions, support, relaxation, and friendship. I want to especially thank Jill Miller, Juliann Aukema, Henar Alonso-Pimentel, Maria Clauss, Margrit Mcintosh. Mark Fishbein, Betsy Arnold. Margaret Evans, Amy Boyd, Ken Moriuchi, Wendy Goodfriend, Francisco Omelas, Sandy Adondakis, Susan Masta, and Amy Faivre. A number of undergraduates associated with the McDade lab were also very helpful: these include Katy Riley. Alexandra Chetochine, Amanda Fox, and Alissa Martin. I want to thank the staff at Big Bend National Park, TX and the Jornada LTER. NM for permission to collect fragrance and plant materials, and their assistance in locating species within these areas. For help in locating Mirabilis macfarlanei. I thank the staff at the Bureau of Land Management in Idaho, and Joe Duft for allowing fragrance collection of this species from a plant growing in his personal garden. Many people have helped me locate plants in the field; I would especially like to thank Richard Spellenberg and Jackie Poole. I am grateful to Mats Thulin for providing me with leaf material of Selinocarpus sontalensis; I was able to extract DNA from this material and, thus, include this species in my phylogenetic analysis of the genus. I am very thankful for the support of my parents, Ellen and Joe Levin, my sister. Sara Levin, and my invaluable friends, Jill Miller, Henar Alonso-Pimentel. and Juliann Aukema; all of these people were indispensable in the completion of this dissertation. A research assistantship from a National Science Foundation grant to Lucinda McDade and Rob Raguso provided me with the flexibility of time and sufficient resources to conduct my dissertation research. I was also supported by the American Society of Plant Taxonomists, Sigma Xi, the Department of Ecology and Evolutionary Biology at the University of Arizona, and the University of Arizona Research Training Group in the Analysis of Biological Diversification. 5
TABLE OF CONTENTS
ABSTRACT 6
CHAPTER 1: INTRODUCTION 8
Frcifirance 9
Hawkmoth Pollinators 10
Phylogenetic History 11
Evaluating Ecological and Evolutionary Effects on Fragrance 13
This Study 13
Explanation of Dissertation Format 13
CHAPTER 2: PRESENT STUDY 16
APPENDIX A: PHYLOGENETIC RELATIONSHIPS WITHIN NYCTAGINACEAE TRIBE NYCTAGINEAE: EVIDENCE FROM NUCLEAR AND CHLOROPLAST GENOMES 20
APPENDIX B: TAXONOMIC STATUS OF ACLEISANTHES, SELINOCARPUS, AND AMMOCODON (NYCTAGINACEAE) 36
APPENDIX C: FRAGRANCE CHEMISTRY AND POLLINATOR AFRNITIES IN NYCTAGINACEAE 51
APPENDIX D: RELATIONSHIPS AMONG FRAGRANCE, PHYLOGENY, AND POLLINATION IN THREE GENERA OF NYCTAGINACEAE 87
REFERENCES 151 6
ABSTRACT
Fragrances appear to act in conjunction with visual cues to attract specific pollinators. Besides the ecological influence of pollinator type on fragrance, as with many other attributes of organisms, phylogenetic history may also affect fragrance composition. In this dissertation I examine the quality and quantity of floral and vegetative fragrance, and explore the relationships among fragrance, pollinators, and phylogenetic history in the plant family Nyctaginaceae.
Using DNA sequence data, I inferred phylogenetic relationships among and within the Nyctaginaceae genera. Adeisanthes, Selinocarpus, and Mirabilis. There is a high incidence of hawkmoth pollination within these genera, in addition to multiple pollinator transitions. Results suggest that neither Acleisanthes nor Selinocarpus are monophyletic, but that together they comprise a monophyletic lineage. Because of this finding, I have taxonomically combined these two genera into a single genus.
Analyses of floral and vegetative fragrance from Acleisanthes. Selinocarpus, and
Mirabilis species included in the phylogenetic study show that each species has a unique fragrance profile. Further, although there is substantial variation among individuals within species, intraspecific variation is significantly lower than interspecific variation in fragrance profiles. Fragrances are composed of 5-108 different compounds from at least seven different biosynthetic classes. Some species produce most of their fragrance vegetatively, while floral emissions are the sole source of volatiles in other taxa.
Results show that neither total amount of volatiles nor the amount of floral volatiles per (ig floral tissue is correlated with pollinator type. However, the emission of nitrogen-bearing compounds appears to have been lost in those lineages that have also 7 lost moth pollination, suggesting that the presence of nitrogen-bearing compounds may be important for moth attraction.
Although the phylogenetic signal in the fragrance data is not entirely congruent with the signal in the DNA sequence data, certain compounds and biosynthetic pathways do support the independent phylogeny inferred using the DNA data. However, it is also clear that many compounds are highly homoplastic, yielding limited phylogenetic information. Overall results suggest that phylogenetic relationships rather than pollinator affinities are better predictors of fragrance composition among these Nyctaginaceae species. g
CHAPTER 1
INTRODUCTION
Floral fragrances have long been thought to attract specific pollinators. Faegri and van der Pijl (1979) included fragrance, along with other floral traits, as characterizing specific pollination syndromes. However, the syndrome concept is clearly a gross over- generalization with many exceptions (e.g., Fishbein and Venable 1996; Herrera 1996;
Waser et al. 1996; Wilson and Thomson 1996). It is also a concept that has been more useful in ecology than in evolutionary biology because virtually all pollination syndromes have evolved multiple times. Species may share certain traits because of shared ancestry rather than ecological factors. Therefore, knowledge of phylogenetic relationships enables us to decouple historical and ecological forces in studying the evolution of character traits. Specifically, a phylogenetic context can elucidate the number of times a character has evolved and whether ecological factors, including pollinator relationships, may be correlated with character evolution.
This dissertation investigates the evolution of a particular suite of characteristics within a lineage of plants. Specifically, I examine the quality and quantity of floral and vegetative fragrance, and explore the relationships among fragrance, types of pollinators, and phylogenetic history. Previous studies have associated fragrance with floral visitors
(Knudsen andTollsten 1993; Borg-Karlson et al. 1994; Kaiser and Tollsten 1995;
Knudsen and Tollsten 1995; Bestmann et al. 1997; Dobson et al. 1997; Miyake et al.
1998) but not in an explicitly phylogenetic context. The unique approach presented here
allows the possible effects on fragrance composition of both pollinators and phylogenetic
history to be distinguished. 9
Fragrance
Fragrances emitted by plants are composed of a mixture of compounds that are of
low molecular weight and, thus, volatile (100-250 Da; Dudareva et al. 2000). These compounds are produced by several different biosynthetic pathways (Croteau and Karp
1991; Azuma et al. 1997; Dudareva et al. 2000), and include terpenoids, aromatics,
aliphatics, and nitrogen-bearing compounds. Together these compounds create a
fragrance profile. Although specific studies are lacking, it is likely that compounds
combine and interact with animals' sensory systems in a synergistic fashion.
Evolutionarily, floral fragrance may have originated from chemical compounds that
served as herbivore deterrents in the ancestors of insect-pollinated plants (Peilmyr and
Thien 1986; Peilmyr et al. 1991). There are many reasons why pollinators are attracted
to floral odors; they may act as chemical cues for mating sites (Gottsberger 1988, 1989)
and/or food (Peilmyr and Thien 1986). Fragrance may also be attractive through
deception via similarities to insect sexual pheromones (Borg-Karlson 1990; Peilmyr et al.
1990).
The location from which fragrance is emitted from a flower varies among plant
taxa. The majority of scented flowers studied to date produce a diffuse odor from
epidermal cells of the petals (Vogel 1963), although other floral parts may also emit
fragrance (Pichersky et al. 1994; Dobson et al. 1999). However, occasionally scent
emitting cells are localized into floral scent secretory structures called osmophores
(Vogel 1963). It appears that these cells store starch, which is then used to provide
energy for the biosynthesis of fragrance compounds (Vogel 1963). In addition to
emission of fragrance from floral structures, fragrances may also be produced by
vegetative tissues and by fruits. 10
Although there is a diversity of possible fragrance profiles (as a result of an essentially unlimited number of chemical combinations), it has been suggested that pollinators may cue on specific compounds as attractants to flowering plants (Knudsen and Tollsten 1993; Borg-Karlson et al. 1994; Kaiser and Tollsten 1995; Knudsen and
Tollsten 1995; Bestmann et al. 1997; Dobson et al. 1997; Miyake et al. 1998; Grison et al. 1999; Helversen 2000). Historically, floral fragrance was described qualitatively by
Faegri and van der Pijl (1979) in connection with different pollination syndromes. For example, beetle pollinated flowers tend to have sweet/fruity odors, bat pollinated flowers are "musty," and hummingbird flowers are scentless. Only recently, however, have biologists had the analytical tools to test these generalizations and to understand their chemical basis. Thus, we now know that visually oriented floral visitors, including many bees, butterflies and nectar-feeding flies, visit flowers with highly variable scent chemistry and emission rates (Dobson 1994). In contrast, animals that use fragrance as a primary orientation cue such as bats and carrion flies pollinate guilds of flowers showing strong evidence for convergent evolution in scent chemistry, with floral fragrances dominated by sulfur and nitrogen-bearing compounds, respectively (Vogel 1963; Smith and Meeuse 1966; Kaiser 1993; Knudsen and Tollsten 1995; Bestmann et al. 1997).
Hawkmoth Pollinators
Hawkmoths and the flowers they visit are a good example of plant-pollinator co-
adaptation. Throughout warm regions worldwide, hawkmoths visit often pale, tubular,
night-blooming flowers that produce abundant nectar and emit perfumey floral fragrances
(Grant 1983; Haber and Frankie 1989). The fragrances appear to function as long
distance attractants (Tinbergen 1958) and to elicit feeding in naive, captive hawkmoths
(Brantjes 1978; Raguso and Willis 2000). Surveys of hawkmoth-pollinated flowers 11
suggest that the scent chemistry of these flowers is only loosely convergent (Knudsen and
Tollsten 1993; Raguso 1997; Miyake et al. 1998). Their floral fragrances are variable blends of mono- and sesquiterpenes, aromatic esters and alcohols, and nitrogen-bearing compounds, few of which are unique to hawkmoth-pollinated flowers (Kaiser 1993;
Knudsen and Tollsten 1993; Miyake et al. 1998).
Electro-physiological assays indicate that hawkmoths are olfactorily sensitive to plant volatiles of all major biochemical classes (Raguso et al. 1996; Raguso and Light
1998). Although naive hawkmoths appear to use floral scent as a distance orientation cue, wind tunnel assays suggest that nearly any floral blend (but not single compounds) is sufficient to guide upwind flight (Raguso and Willis 2000). Thus, the observed variance in floral scent chemistry among species that share hawkmoth pollination indicates that different classes of scent compounds may be functionally equivalent in hawkmoth attraction, suggesting that convergence in this system may involve emission of any blend of plant volatiles. However, Knudsen and Tollsten (1993) and Kaiser (1993) proposed
that the presence of particular compounds was more related to moth pollination than were
entire fragrance blends. Specifically, it has been suggested that the presence of nitrogen-
bearing compounds in fragrance is characteristic of moth-pollinated species (Nilsson et
al. 1985; Kaiser 1991; Knudsen and Tollsten 1993; Dobson et al. 1997), and that
oxygenated sesquiterpenes are more characteristic of the fragrances of flowers pollinated
by hawkmoths than those pollinated by noctuid moths (Knudsen and Tollsten 1993).
Phylogenetic History
Like many other characteristics of organisms, aspects of fragrance may provide
information about patterns of phylogenetic relationships. Secondary compounds have
been found to be phylogenetically informative at higher taxonomic levels (e.g.. 12
Waterman 1998; Grayer et al. 1999) and to some extent at lower levels (e.g., Azuma et al.
1997; Adams 1999; Azuma et al. 1999; Williams and Whitten 1999; Barkman 2001).
However, there have been difficulties in determining the most appropriate manner for coding such data for use in phylogenetic inference. Specifically, the fragrance profile of a species is composed of both qualitative data, i.e., which compounds are present, and quantitative information, i.e., how much of each compound is present. Further, each of these compounds is a product of a specific biosynthetic pathway; fragrance profiles often include multiple compounds from the same biosynthetic pathway. Barkman (2001) suggests that the use of individual compounds violates statistical assumptions of non- independence and, thus, fragrance components should only be coded according to biosynthetic pathway. In addition to questions regarding how to code this qualitative information (i.e., which compounds/pathways are present), it is also unclear how to approach the use of quantitative fragrance data (i.e., how much of each compound/pathway is present). There is evidence that environmental conditions can influence the quantity of compounds sampled (Jakobsen and Olsen 1994) and. thus, many researchers have avoided using these data and have simply coded for the presence or absence of a compound or biosynthetic pathway (Barkman et al. 1997; Dobson et al.
1997; Adams 1999; but see Azuma et al. 1997; Williams and Whitten 1999).
In addition to problems in coding these data, proper evaluation of which coding method yields the best estimate of phylogenetic relationships requires the ability to compare the phylogeny inferred from fragrance data with a phylogenetic hypothesis
based on an independent set of data. The absence of a robust phylogeny inferred using characters independent of fragrance (e.g., morphological or DNA sequence characters) to
which a phylogeny inferred using fragrance data can be compared may be a limitation for
examining the phylogenetic utility of fragrance data. 13
Evaluating Ecological and Evolutionary Effects on Fragrance
A phylogenetic context facilitates explicit tests of the relationship between pollinator class and fragrance composition. Further, inclusion of information on phylogenetic relationships, pollinator affinities, and fragrance profiles permits evaluation of the relative importance of evolutionary history versus current ecological relationships in predicting fragrance composition.
This Study
My dissertation research has involved the examination of relationships among fragrance, type of pollinators, and phylogenetic history. This research has been done with species of two closely related lineages in the plant family Nyctaginaceae.
Nyctaginaceae are woody or herbaceous perennials (rarely annuals) distributed mainly in subtropical and tropical regions of the New World. In southwestern North America the flowers of many Nyctaginaceae species have nocturnal anthesis and are visited by hawkmoths (Grant 1983). I have focused my studies on three genera of Nyctaginaceae:
Acleisanthes (7 spp.), Selinocarpiis (9 spp.), and Mirabilis section Qiiamoclidion (6 spp.).
Explanation of Dissertation Format
The chapters in this dissertation are included as four appendices. Appendices C and D represent work that I did as part of a larger project in collaboration with Drs.
Lucinda A. McDade and Robert A. Raguso; thus, they are both co-authors on these papers.
Appendix A reconstructs the phylogenetic history of species within
Nyctaginaceae tribe Nyctagineae, with special emphasis on relationships within
Acleisanthes, Selinocarpus, and Mirabilis section Quamoclidion (6 spp.). Here I present 14
results of phylogenetic analyses based on molecular sequence data from the internal transcribed spacer regions of nuclear ribosomal DNA (ITS) and the intergenic region between the rbcL and accD genes plus a 300 bp region of accD on the chloroplast genome. These analyses were conducted to elucidate phylogenetic relationships within
Acleisanthes and Selinocarpus and test the monophyly of these two genera. In addition, this study endeavored to test the monophyly of Mirabilis and to clarify relationships within Mirabilis section Quamoclidion and among species in this section and a sampling of species in other sections of the genus.
Results from this phylogenetic study (Appendix A) prompted a reevaluation of current taxonomic circumscriptions of Acleisanthes and Selinocarpus. Thus, Appendix B presents new taxonomic combinations and a key to all of the species in these two genera.
Appendix C presents data from analyses of floral and vegetative fragrance in a twenty species subset of the taxa included in Appendix A. To determine whether there are species-specific fragrance profiles, we compared the amount of intraspecific variation in fragrance composition to the amount of interspecific variation in fragrance composition. Fragrance differences across species and possible relationships to pollination were also examined. Further, we explored the relative importance of floral versus vegetative compounds to overall species' fragrance composition.
Appendix D examines these fragrance data in the context of the evolutionary
relationships inferred in Appendix A. The ecological relationship between floral and
vegetative fragrance and type of pollinators was explored by testing the hypothesis that
the fragrance profiles of plants pollinated by hawkmoths are more similar to each other
than to closely related species that are pollinated by other organisms. We also examined
whether fragrance is absent in taxa pollinated by organisms lacking the ability to smell. 15
as might be expected if fragrance production is costly in terms of metabolic resources and apparency to herbivores.
In addition to allowing phylogenetically independent contrasts between fragrance and pollinators, the molecular sequence data (Appendix A) provided a phylogeny to which we could compare the phylogenetic information in the fragrance data. These comparisons involved coding the fragrance data in a variety of ways to determine whether there was phylogenetic signal, and how this signal varied depending on the coding method used. We also explored whether the phylogenetic signal in the fragrance data was congruent to that in the molecular DNA sequence data. Finally, given the molecular sequence phylogeny, we determined which compounds and pathways were most phylogenetically informative. 16
CHAPTER 2
PRESENT STUDY
The methods, results, and conclusions of this study are presented in the papers appended to this dissertation. The following is a summary of the major findings in these papers.
In Appendi.x A, I present a hypothesis of phylogenetic relationships within
Nyctaginaceae tribe Nyctagineae, with an emphasis on Acleiscmthes, Selinocarpus, and
Mirahilis section Qiiamoclidion. Thirty-two species that are mainly from southwestern
North America were included in this study. Phylogenetic analyses based on molecular sequence data from the nuclear and chloroplast genomes suggest that neither Acleisanthes nor Selinocarpus are monophyletic as currently circumscribed, but that together, along with Ammocodon chenopodioides, they form a monophyletic lineage. Interestingly, the only Selinocarpus species that occurs outside of southwestern North America. 5. somalensis, was supported as being a true member of this lineage, and its phylogenetic placement suggests that dispersal is the most likely explanation for its disjunct occurrence in Somalia. The genus Mirabilis is strongly supported as monophyletic, but the monophyly of section Quamoclidion is equivocal due to the uncertain relationship of M. triflora to the other tlve species in this section. Further sampling of species closely related to those in section Quamoclidion is needed to help resolve this issue. In addition, increased sampling of other species in the tribe Nyctagineae is necessary to determine whether the genus Mirabilis is in fact sister to Acleisanthes plus Selinocarpus.
Appendix B updates the taxonomy of Acleisanthes and Selinocarpus to reflect the
phylogenetic results presented in Appendix A. As the generic names Selinocarpus and
Acleisanthes have the same nomenclatural priority, I combined all Selinocarpus species 17
into Acleisanthes to form a monophyletic Acleisanthes. Ammocodon chenopodioides was also combined into Acleisanthes. Further, three species of Selinocarpiis are synonymized because any distinctions observed appear insufficient to warrant specific status of these taxa. Thus, Acleisanthes is now composed of 16 species. In this paper I also present a key to all 16 species.
Presented in Appendix C are the floral and vegetative fragrance data for the 20 species of Acleisanthes, Selinocarpiis, and Mirabilis from which fragrance was collected.
We show that intraspecific variation in fragrance profiles is significantly less than interspecific variation. This finding supports the premise that each species has a unique fragrance profile. Fragrance profiles included 5-108 different compounds, and were composed of compounds from at least seven different biosynthetic classes:
monoterpenoids, sesquiterpenoids, fatty acid and amino acid-derived aliphatic compounds and lactones, benzenoids, phenylpropanoids, and nitrogen-bearing compounds. From these data it is clear that amino-acid derived aliphatics, lactones,
phenylpropanoids, benzenoids, and nitrogen-bearing compounds are almost solely emitted from fioral tissue, whereas many monoterpenoids, sesquiterpenoids, and fatty- acid derived aliphatic compounds may be emitted by vegetation. Further, some species
produce much more of their overall fragrance vegetatively, while floral emissions are the
sole source of volatiles in other taxa. Despite the remarkable variation in fragrance
profiles among closely related species, the fragrances of the moth-pollinated taxa
included in this study are generally consistent with previous descriptions of moth-
pollinated fragrances from species in a variety of plant families (Kaiser 1993; Knudsen
and Tollsten 1993; Miyake et al. 1998).
Appendix D examines the ecological relationship between floral and vegetative
fragrance and pollinators, and the evolutionary relationship between fragrance and 18
phylogenetic history. We present evidence from field observations and the literature that the majority of species in Acleisanthes, Selinocarpus, and Mirabilis section
Qiiamoclidion are likely pollinated by hawkmoths or smaller, noctuid moths. Using a phylogeny based on molecular sequence data (Appendix A) for the 20 species from which we collected fragrance (a subset of those included in Appendix A), we were able to factor out phylogenetic relationships to determine whether specific aspects of fragrance were correlated with type of pollinators. We found that neither total amount of volatiles nor the amount of floral volatiles per ^g floral tissue was correlated with pollinator class.
Rather, it appears that these fragrance measures are strongly related to phylogenetic relationships, with species in Selinocarpus and Acleisanthes emitting much higher amounts of volatiles than Mirabilis species. However, when the presence of nitrogen- bearing compounds was examined in relation to type of pollination, we did find that emission of nitrogen-bearing compounds has been lost in those lineages that have also lost moth pollination. This finding is consistent with previous studies that have suggested a link between moth pollination and nitrogen-bearing volatiles (Nilsson et al. 1985;
Kaiser 1993; Knudsen and Tollsten 1993). For the most part, our results suggest that pollinators are not a major factor in predicting fioral scent composition. However, the number of transitions to different pollinators is quite low among Acleisanthes.
Selinocarpus, and Mirabilis section Quamoclidion, thus limiting the power to test for such associations.
In Appendix D we also explored different methods for coding fragrance data as phylogenetic characters, and compared the phylogenetic signal in the fragrance data with that in the molecular sequence data (Appendix A). Although the phylogenetic signal in the fragrance data was never entirely congruent with the signal in the molecular sequence data, certain compounds and biosynthetic pathways did support the phylogeny inferred 19
using the sequence data. Overall results from this study suggest that fragrance data can be phylogenetically informative at the level of compound, biosynthetic pathway, amount, and presence/absence; thus, it may be useful to consider fragrance data in all of these ways. However, it is also clear that many compounds are highly homoplastic, yielding limited phylogenetic information. As with many morphological characters (e.g., leaf size), certain fragrance characters may be too homoplastic for use in phylogenetic
Inference. On the other hand, we propose that certain fragrance characters may be useful in providing support for phylogenies inferred using other types of data. In general, our results suggest that phylogenetic relationships rather than pollinator affinities are better predictors of fragrance composition among these Nyctaginaceae species. 20
APPENDIX A
PHYLOGENETIC RELATIONSHIPS WITHIN NYCTAGINACEAE TRIBE NYCTAGINEAE. EVIDENCE FROM NUCLEAR AND CHLOROPLAST GENOMES 21
SYSTEMATIC BOIANY Volume 25, Number 4 October-December 2000
Cover Illustration: Holmes and Pniski describe a new spedes of Mikania (Compositae) from Peru, whidi they name after J. Wurdak who collected the only known spedmen of this new spedes.
TABLE OF CONTENTS New Species of Mikania (Compositae: Eupalorieae) from Ecuador and Ftru Walter C. Holmes and John F. Pniski 571 Morphdmetric Analysis of Eseobaria stuedii var. sneedii, E. sneedii var. leei, and E. guad- alupensis (Cactaceae) Marc A. Baker and Robert A. Johnson 577 Taxonomic Revision of South American Spedes of the Genus Acacia Subgenus Acacia (Fa- baceae: Mimqsoideae) ]ohn E. Ebinger, David S. Seigler, and H. David Clarke 588 Implications of Chloroplast DNA Restriction Site Variation for Systematics of Acacia (Fa- baceae: Mimosoideae) H. David Clarke, Stephen R. Downie, and David S. Seigler 618 Evolution of Endemic Sideritis (Lamiaceae) in Macaronesia: Insights from a Chloroplast DNA Restriction Site Analysis faaet C. Barber, Javier Francisco Ortega, Amoldo Santos'Guerra, Aguedo Marrero, and Robert K. Jansen 633 Expansion and Contraction of the Chloroplast Inverted Repeat in Apiaceae Subfamily Apioideae Gregory M. Plunkett and Stephen R. Downie 648 Phylogenetic Relationships in the Commelinaceae: I. A Qadistic Analysis of Morphological Data Timothy M. Evans, Robert B. Faden, Michael G. Simpson, and Kenneth J. Sytsma 668 Congruence and Complementarity of Morphological and tmL-tmF Sequence Data and the Phytogeny of the African Restionaceae Pia K. Eldenas and H. Peter Under 692 Phytogeny of the American Amaryllidaceae based on nrONA ITS sequences Alan W Steerauf Charles L Guy and Qin-Bao Li, and Si-Lin Yang 708 Phylogenetic Relationships of Corylus Species (Betulaceae) Based on Nuclear Ribosomal DNA ITS Region and Chloroplast matK Gene Sequences Mli Erdogan and Shawn A. Mehlenbacher 727 Phylogenetic Relationships Within Nyctaginaceae Tribe Nyctagineae: Evidence from Nu clear and Chloioplast Genomes Radiel A. Levin 738 Book Reviews Legume (Fabaceae) Fruits and Seeds: Interactive Identification and Infijrmation Retrieval by Kirkbride, J.H. Jr, Gunn, C.R, Weitzman, A.L tc Daillwitz, M.J. Reviewed by M. Lock 751 Erratum 753 Reviewers of Manuscripts 755 Index 756
i 22
Date: Wed, 20 Dec 2000 13:53:14 -0700 To: [email protected] From: Melissa Luckow
Dear Rachel: You have permission from the ASPT to reprint your paper in your thesis, and for University Microfilms. The way we work is that you continue to own your work and we hold the copyright for you. We appreciate knowing when and where you republish your work, just so we know that any reprinting is done with your permission. Best of luck! Melissa Luckow
At 04:49 PM 12/12/00 -0700, you wrote; Dear Dr. Luckow, Patrick Herendeen suggested that I should contact you regarding a request for permission for University Microfilms, Inc. to publish and distribute my dissertation that includes a paper to be published in Systematic Botany. The paper is titled Phylogenetic relationships within Nyctaginaceae tribe Nyctagineae: Evidence from nuclear and chloroplast genomes. It is currently in press and should appear in volume 25(4), October-December 2000 issue.
Please let me know If you have any questions. I appreciate your help in this matter. Thank you.
Sincerely,
Rachel Levin
Rachel A. Levin Department of Ecology & Evolutionary Biology University of Arizona Tucson, AZ 85721 phone: 520-621-8220 fax: 520-621-9190 23
SytfrnoTK SiMOfty (2000i 2S<4^ pfi 738-750 C Copyright 2000 by (he Anvncan Sooecy of PUnt Taunamuts
Phylogenetic Relationships Within Nyctaginaceae Tribe Nyctagineae: Evidence from Nuclear and Chloroplast Genomes
RACHEL A. LEVIN Department of Ecology ic Evolutionary Biology, University of Arizona, Tucsoa Arizona 85721
Communicating Editor Aaron Uston
ABSTRACT. Nycuginaceae are A small family of mainly Sew World tropical and subtropical trees, shrubs, and herbs. To dale phylogenetic relationships within the family have not been examined. This study provides the first phylogenetic hypothesis of relatiotiships within Nyctaginaceae tribe Nyctagineae based on sequence data from both nuclear (ITS) and chloroplast (occO 5'coding region and intergenic region between the rhL and mrcO genes). Morphological characters are also discussed as they relate to the phytogeny inferred using molecular data. Results suggest that neither Adeisanths nor Sttimxarpus is monophyletic but tlut together they compnse a monophyletic lineage The genus Minbilis is strongly supported as monophyletic but the monophyly of two of its sections is suspect Morphology generally agrees with the molecular data and in some instances reinforces clades weakly supported by nuclear and chloroplast data. Further sampling will help clarify relationships of these genera witliin Nyctaginaceae
.Vyctaginaceae are a small funily of ca. 30 genera Rwler and Turner 1977; Fosberg 1978; Pilz 1978; and -too spedes distributed primarily in the tropical Spellenberg 1993; Turner 1994; Le Due 1995; Mahrt and subtropical regions of the New World (Mab- and Spellenberg 1995), although no researchers berley 1997). Monophyly of the family is suggested have yet taken a phylogenetic approach to the study by tvk-o synapomorphies: the absence of a corolla of Nyctaginaceae at any level (flov»rers possess an often petaloid, showy calyx) It is this paudty of knowledge of phylogenetic and a specialized anthocarp fruit-type; an achene relationships within Nyctaginaceae that 1 have be or utricle enclosed In perianth tissue (Bogle 1974, gun to address. 1 am studying phylogenetic rela but see discussion therein of conflicting anthocarp tionships both within and among genera in Nyc definitions). tagineae; with a focus on the subtribes Nyctagini Relationships within the family ate less clear de nae and Boerhaviinae sensu Bittrich and Kiihn spite considerable morphological diversity. For in (1993). As part of a larger study on the relationship stance; it appears that similar morphologies have between the evolution of floral fragrance and hawk- evolved multiple times, making within-fomily tax moth-pollination (R. Levia L. McDade; and R. Ra- onomy challenging. Perhaps as a result of these dif guso, in mss.), I have reconstructed phylogenetic ficulties, the latest family-level revision was almost relationships within and among three genera: Mir- sevent>- years ago (Heimerl 1934). Mote recently an Mis, Selinocarpus, and AcUisanthes. Based on mor updated treatment (Bittrich and Ktihn 1993) includ phology, it has been suggested that Selinocarpus is ed the current literature on various genera within the dosest relative of Acleisanthts (Smith 1976). Nyctaginaceae; but was still based on Heimerl's Within Mirabilis there are six sections, with most (1934) classificatioa According to this treatment, spedes in sections Mirabilis, Oxybaphus, and Quum- the majority of species occur in tribes Nyctagiiwae octidion (Heimerl 1934; Le Due 1995). In section (ca. 140 spp.) and Pisoitieae (ca. 200 spp.). Quamoclidion (6 spp.) there has been question about In Nyctagineae the most spedes-cidi subtribes the placement of Mirabilis triflora. a vine with red, are Boerhaviinae and Nyctagininae (Table 1). These tubular hummingbird-pollinated flowers that is two subtribes represent a segregation of Heimerl's unique in the section (Pilz 1978). Pilz (1978) placed (1934) Boerhaviinae apparently based on the pres M. triflora withiri this section because it has united ence in Nyctagininae of conspicuous, sometimes ca involucral bracts subtending the inflorescence that lyx-like mvolucral bracts subtending the inflores appear to be homologous to those that characterize cence. However, the validity of such a distinction is the rest of section Quamoclidion. questionable as floral bracts are present in both These genera are mostly found in southwestern subtribes. VVithin these subtribes there have been North America, with Acleisanthts and Selinocarpus ta.xonomic studies of many genera (Smith 1976; mainly restricted to the Chihuahuan Desert al 738 24
2000] LEVIN: PHYLOCENETICS OF MYCTACINACEAE 739
TABU I. Tribe Nyctagineae sensu Bittrich and Kuhn (1993). Asterisks indicate those genera included in this study. 'Note that Boerhma is diverse and often di^-ided into multiple genera (eg., Bittrich and Kuhn 1993: Spellenberg 1993; Mahrt and Spellenberg 1995).
Ctfnua • sppL Ccof^rjphic itinge Suurce Subtribe .Nyctagininae 'Minbila L ca. 60 North and South America. 1 Hi- Le Due 1993 malavan sp. 'Allionia L 1-2 SW USA to Chile Turner 1994 Cuscatknui Standi I El Salvador Bittrich and Kuhn 1993 Sycta^inia Chuisy 1 Tnas and .Mexico Bittrich and Kuhn 1993 Subtribe Boerhaviinae 'Burfuciu L' ca. 50 pan tropical Mabberley 1997 'Selinocarpus A. Gray • 9 SW USA to northern Mexico, 1 Fowler and Turner 1977 Somalian sp. '.^i:tfisanlha A. Cray 7 SW USA to Mexico Smith 1976 Okenia Schdl. it Cham. 1-2 Florida to Mexico and Nicaragua Bogle 1974; Bittrich and Kuhn 1993 Canlw Alain 1 Cuba Bittrich and Kiihn 1993 Subtribe Colignoniinae Coligmnia Endl. 6 Colombia to Argenhna Bohlin 1988 Piscnidk (Heimerl) Standi. 1 .Mexico to Argentina Bittrich and Kiihn 1993 Subtribe Phacoptilinae Phaniptilum Radlk. 1 South Africa Bittrich and Kiihn 1993
though one spedes of Selinocarpus occurs in Soma subtribes Boerhaviinae and Nyctagininae (Bittrich lia (Thulin 1993)! Most species have long, salver- and Kiihn 1993). I will also discuss morphological form or funnelform nocturnal flowers that are pri characters as they relate to the phytogeny inferred marily visited by hawkmoths (Cruden 1970; Fowler using molecular data. and Turner 1977; Spellenberg and Delson 1977; Pilz 1978; Grant and Grant 1983; Martinez del Rio and MATERIALS A.\D METHOIJS Burquez 1986: R. Levin, unpubl. data), although many are also capable of self-pollination (Cruden Taxon Sampling. Sampling was mainly focused 1973; Spellenberg and Delson 1977; Martinez del within Mirabilis, Selinocarpus, and Acleisanthes (Table Rio and Burquez 1986; Hernandez 1990). Plants of 2). Within Mirabilis all six spedes traditionally Selitwcarpus spp. and Acleismlhts spp. are morpho placed into section Quamoclidion were included. logically very similar, although Selinocarpus spp. are Seven taxa representing other sections within Mir usually woody perermials, whereas Acleisanthes abilis, including sections Mirabilis and Oxybaphus spp. are herbaceous perennials. Further, Selinocar (sensu Le Due 1995) were also used to test mono pus spp. have characteristic five-winged fruits, un phyly of Quamoclidion and to permit polarization of like the ribbed but wingless anthocarps of Acleis anthes spp: character states within that section. Also induded Here I present a phylogenetic hypothesis for re in this study were all seven spedes of Acleisanthes lationships within and among Mimbilis. Selinocar and eight of the nine spedes of Selinocarpus. I fol pus, and Acleisanthes using DMA sec{uence data lowed Bittrich and Kiihn (1993) in considering Am- from both the dtloroplast and nuclear genomes. mocodon chenopodioides (Cray) Standley as a Selino This study endeavors to 1) enhance our under carpus. Selinocarpus palmeri HemsL was not induded standing of relationships among spedes of Selino because plant material was unavailable; it is otUy carpus and Acleisanthes, 2) eluddate relationships known from one locality, was last collected in 1978, within Mirabilis section Quamoclidion and clarify and could not be found in 1998. Allionia incamata some sectional relationships within Mirabilis, 3) test was induded as an outgroup within subtribe Nyc- the monophyly of these three genera, and 4) pro taginiiuie; and Boerhaoia coccinea was induded as an vide a preliminary test of the validity of the divi outgroup within Boerhaviinae The more distant sion of subtribe Boerhaviinae (H^erl 1934) into outgroups Abronia fragrans (monogeneric tribe 740 SYSTEMATIC BOTANY [Volume 25
TABLE 1 Toxa, voucher specimens, and CenBank accession raunbers for both nuclear ITS and chloroplast (cp) se quences. Samples from spedes in section Mirabilis were made from fresh material grown from seeds of the voucher collections. BBNP is Big Bend National Park. Brewster Co, Texas. All vouchers not already labeled with location are deposited in the herbarium at the University of Arizona.
Toicn Vlxichrf Uxality ITS cp
Conus Str/rmKdrpui A. Cray S mgusltfblius Torrcj' RL97-I BBNP.TX AF211997 NA S dKnofwdiaiiies A. Gray ITS: Worthington Hudspeth Co., TX AF2U998 NA 6694. AZ cp:RL99-2 Dona Ana Ca, MM NA AF214694 S Jiffusus A. Gray RL98-1 CIaikCo,,MV AR11991 AF2I4688 S. lum'eolatui VVooton RL97-3 Culbetson Co., TX AF2U995 AF214691 S parvtfbliui (Torrey) Standi. RL97-2 B8NP,TX AF2I1993 AF214689 S. purpusutnus Heimerl RL98-6 Coihuila, .MX AF211994 AF214690 S. somaUnyi:) Chiov. Thulin Ic Dahir Somalia AF2n992 AF214692 6517, UPS S. miiulatus Fowler & Turner RL98.7 Coahuila, MX AF2n9% AF214693 Genus AcUtsantka A. Cray .•<. acutifolia Standi. RL98-9 Coahuila, MX AF211988 AF214685 .4. amsopMla A. Gray RL 98-13 Coahuila, MX AF211986 AF214683 A. crasit/blia A. Cray RL 98-12 Coahuila, MX AF211987 AF2146W •A. Itinjfiflora A. Gray RL97-3 BBNP,TX Anil985 AF214682 A. nana 1. M. fohnst. RL 98-15 San Luis Potosi, MX AF2U990 AF214687 .4. oblusa (Choisy) Standi. RL97^ U Salle Co, TX AF2U984 AF214681 A. wri^hlii (A. Gray) Bcnth. & RL 98-11 Coahuila, M.X AF2n989 AF2I4686 Hook. Genus Mirabilis L. Sod. Ouanwclitiion .Vt. alipfs (S. Watson) Pib RL98-3 Churchill Co, NV AF112DOO AF214696 .\l. ifwni'i S. WaLson RL98-4 Siskiyou Ca, CA AF212001 AF2I4697 .M. macfiirlanci Constance & RAR98-101C Ada Co, ID AF211999 AF214695 Rollins .Vf. multijhim (Torrey) A. Gray RL98-5 Pima Ca, A2 AF212002 AF214698 M. puJica Bameby RL98-2 Lincoln Co, NV AF212003 AF214699 .\r. tri/bra Benth. RL 98-23 Baja Calif. Sur, M.X AF212005 AF2I4701 Soct. Ot^ijoplms •Vf, bi^ltirii A. Cray RL 98-24 Mineral Co, NV AF212011 .MA •Vl. lenuiltiba S. Watson RL 98-25 Baja Calif. MX AF2I2004 AF214700 Sect. Mirabilis M. iaiapa L. U Due 228. TEX Puebla.MX AF212009 NA .\f, tangiflara L U Due 234. TEX MDrelos,MX AF212006 AF214702 .Vf. polanti Le Due LeDuc259,TE.X Nuevo Leda M.X AF212007 NA iVI. Srinjumoi var. brei'tfbm Le U Due 254. TEX Jalisco, MX AF212008 NA Due Sect. Mirabilapsis M. axcinea (Torrey) Benth. k RL9S-20 Pima Co, AZ AF2I2010 NA Hook. Outgroups Atlionia incarnata L RL 98-21 Pima Co, AZ AF212012 NA Bixrharia axcinea .Miller None Pima Co. AZ Af2l20l3 NA Abrvnia fragrans Nutt RAR98-I01B Fremont Co, ID AF212014 AF2U703 Pisjnia capitata (S. Watson: RL99-1 UA campus AF2I2015 NA Standi. 26
2000] LEVLV: PHYUXENEnCS OF NKCTACINACEAE 741
TABLE 3. Primers used for the amplification and sequencing of ca. 90 bp of rbcL, the intergenic region between rbcL and the accD gene, and the accD 3' coding legioa [n the aligned sequences used in this study, Nya-338 and Nyct-1061 are positions from the begiiuiing of the intergenic region.
N4me Strand Sn^iencr Source 1378 sense 5'-GA.nGTATGGAnGGAAATCA-3* V^sui Sc Ohnishi 1998 .\'yct-558 antisense 5'-ACAT.=ITAGATCCAATTACTC-3' This paper \yct-1061 antisense S'-CTCTTAGATCGAGTAGTC-J' This paper 2+42 antisense 5'-GGCAATGAAGATAACTGTC-3' Yasui Abrorueae sensu Bittridi and Kiihn 1993) and Pi- between the 3' rbcL gene and the accD gene; and sonia capitata (tribe Pisptueae) were also included. the accD 5' coding region (encodes one of the sub- For the chloroplast locus sampling differed units of acetyl-CoA carboxylase; Yasui and Ohnishi slightly from that of the nuclear locus (Table 2). The 1998). Most taxa were amplified using primer 1378 chloroplast region was challenging to amplify be at the 3' end of rbcL (Yasui and Ohnishi 1998) and cause the primers on the 3' end did not always an primer Nyct-1061 at the 5' end of accD (instead of neal or yield a single product Because of this am Nyct-1061 a few were amplified with primer 2442 plification problem, as well as the strong support of Yasui and Ohnishi 1998) using an aiuiealing tem for Mirabilis section Mirabilis from the nuclear locus perature of 52°C and 40 amplification cycles. I was (see below), the taxa sampled were a subset of unable to use these primer pairs for amplification those studied for the nuclear locus. All taxa within in some of the taxa. Consequently, for these taxa a Mirabilis section Quamoclidion were included as well shorter region was amplified using primer 1378 as the seven spedes of Acleisanthes and all Selino- (Yasui and Ohnishi 1998) and Nyct-558, a primer carpus except S. angustifblius and S palmeri. Two positioned at the 3' end of the intergenic regio(\. taxa representing two different sections of Mirabilis SeifueHcing. Cleaned PCR products (QlAquick® were also included. Abronia fragrans was used as the PCR purification kit, QIACEN) were sequenced in outgroup; the other outgrcxips used for the nuclear both directions. Sequencing used the same primers locus were not included because they did not am as for amplification and the Big-D^-e Ready Reac plify for the chloroplast locus. tion Kit (Perkin-Elmer). Sequencing reactions con DNA Extraction and Amplifkation. Fresh or sil sisted of 25 cycles and were conducted on an ABl ica gel-dried material was available for all but two automated sequencer at the University of Arizona taxa; for these herbarium material was used (Table DNA sequencing focility. 2). Total genomic DMA was extracted using the Alignmtnt and Analtfsis. Sequences were edit modified CTAB method of Doyle and Doyle (1987). ed and aligned using AutoAssembler® DNA Se NUCLEAR LOCUS. PCR amplification of the in quence Assembly Software version 1.4.0 (Applied ternal transcribed spacer (ITS) region of nuclear ri- Biosystems 1989-95) and SeqApp (Gilbert 1993). bosomal DNA, composed of ITS-l, the 5.8S gen^ Character by taxon matrices were prepared in and rTS-2 (Baldwin 1992; Baldwin et aL I'.^S), used MacQade (Addison and Maddison 1999). the primers C26A and N-ncl8slO (Vfen and Zim- PAUP* (Swoffotd 1998) was used to reconstruct mer 1996). PCR conditions varied but mainly con phylogenetic relationships among all 32 taxa for sisted of a "touchdown" procedure; with an initial which ITS sequence data were available Sequences aiuiealing temperature of 58°C, then a decrease to in the data matrix had an average of 0.0094% tniss- 56°C, then to M°C with two cycles at each temper ing data, and an average of 433% gaps (indels). ature. Subsequently the temperature was decreased Parsimony analyses were conducted using the heu by rC every two cycles until 49°C At 49'C thirty ristic search option and TOO random addition se amplification cycles were conducted. This proce quence replicates with tree-bisection reconnection dure pelded a single product of ca. 700 nucleotides (TBR) branch-swapping. Caps were treated as in length. missing data. Analyses were also done including CHLOROP1.AST Locrs. PCR amplification of an extra characters for the presence/absence of >1 bp approximately 1100 base pair (bp) sequence used indels. Multiple equally parsimonious trees were the primers listed in Table 3. This sequence includes combined as a strict consensus tree ca. 90 bp at the 3' end of rbcL the intergeiuc region Phylogenetic recorutruction of the 23 taxa for 27 742 SYSTEMATIC BOTANY [Volume 25 which chioroplast data were collected was similar branch and bound searches were conducted to find to above. For four taxa 1 obtained partial sequences the shortest trees consistent with each constraint. such that there was an average of 39.4% missing The additional steps required for a particular con data and 4.63% gaps. For the remaining 19 taxa straint are given by the difference in length be there was an average of 0.021% missing data and tween the shortest trees consistent with that con 6.57% gaps. As the number of parsimony infor straint and the globally shortest trees. mative characters was very low in comparison to Ma.-umum likelihood (ML) analyses were also the entire data set, I used only the parsimony in conducted for each data set and the combined data formative characters and conducted branch and set using PAUP* (Swofford 1998). Nucleotide fre bound searches. Gaps were treated as missing data. quencies were empirically determined. ML analyses Analyses were also done including extra characters were begun to estimate the transitio[\:transversion for the presence/absence of >1 bp indels. Multiple ratio and rates of evolution. These analyses were most parsimonious trees were combined in a strict stopped before completion and the estimated tran- consensus tree. sition:tTansversion ratio and gamma distribution A partition homogeneit\* test (Farris et al. 1995) shape parameter were then specified in analyses was conducted (PAUP*; Swofford 1998) to deter that were completed. mine the congruence of the chioroplast and nuclear data sets. For this test lOO replicates were per RESULTS formed, using the heuristic search option with sim ple addition, TBR branch-swapping, and gaps treat Nuclear. Aligned length for the entire ITS re ed as missing data. Because of memory constraints gion was 690 bp. There were four >1 bp indel char only variable characters were included. acters that were also added to the data set. The in Data sets were combined for the 23 taxa for clusion of these indels had no effect on the overall which both nuclear and chioroplast data were a\-ail- tree topology, and this analysis will not be dis able. Parsimony analyses were conducted using the cussed further. The ITS data included 158 parsi branch and bound search optioa with gaps treated mony informatiw characters and yielded 260 most- as missing data. Analyses were also done including parsimonious trees. The strict consensus of these extra characters for the presence/absence of >1 bp trees (Fig. 1) shows a highly supported clade com indels. Multiple equally parsimonious trees were posed of Achrisanthes and Selinocarpus {Acleisanthes combined as a strict consensus tree. lineage; BS=100, DI=15) but does not support the For the ITS and combined data sets, the strength monophyly of each genus individually. However, of support for individual tree branches was esti there Is strong support for the clade composed of mated using bootstrap values (BS) (Felsenstein A. abtusa. A. longi^ra, and A. miisophyda (Arleisan- 1985) and decay indices (DI) (Bremer 1988; Dono- ttes A; BS=94, Dl=3), as well as for a clade of the ghue et al. 1992). Bootstrap values were from 100 remaining four species of A'leisanthes and all Stli- full heuristic replicates (150 for ITS) with 100 ran mcarpus {Ackisantlies B; BS=88, Dl=3). Addition dom addition sequences (10 for ITS), TBR branch ally, within this last clade there is strong support swapping, and the MulTrees option that saves all for a clade composed of A. acutijblia, A. nana, A. minimal trees. Decay values for each branch were U-'rightii. and S. diffusus (BS=98, DI=5). determined by first using the decay index PAUP file The monophyly of Minbilis is well supported command in MacGade (Maddison and Maddison (BS=100, Dl=16), and Mirai>ilis section Quamocli- 1999) to prepare a set of trees each with a single dion is supported as a monophyletic group exclud branch resolved. This file was then executed in ing Mirabtlis trifbra (BS=90, DI=3). Further, Afira- PAUP* (Swofford 1998) using the heuristic search bilis section Mimbilii is confidently placed as a option to find the shortest trees consistent with monophyletic lineage (BS=99, DI=7). Boerhavia coc- each constraint. The decay index for each branch in cinea and AUiania incamata are supported as each question is the difference in length behveen the other's closest relative in this analysis (BS=81, shortest trees consistent with a particular con DI=3). straint and the globally shortest trees. Chioroplast The aligned length for 90 bp at the Constraint trees were constructed in MacQade 3" end of rbcL the intergenic region between rbcL (Maddison and Maddiscn 1999) to test various hy and accD. and ca. 300 bp at the 5' end of the accD potheses of phylogenetic relationships. These trees gene was 1101 bp. These data on their own yielded were then loaded into PAUP* (Swofford 1998) and very little resolution (topology not shovvn), because 28 20001 LEVIN: PHYLOCENEnCSOF NVCTAaNACEAE 743 A. anisophyUa A. obtusa A. langiflara A. erassifolia A. acutifoUa A. nana A. wrightii Ia S. difftuus & S. somalensis B S. parvi/olius s S. purpusianus S. undulatus S. angiutifolius S. laneeotatus S. ehenopodioidts M. eoccinea sect Mirabitopsis M. sanguinea — M. langiflara sect Mirabilis lU.poloHii M. jalapa M.tenuUoba I , ^ . , sect. Oxytophus M.bigelom | Af. triflora M. multiflan M. maefartanei sect. Quamoclidian M. greenei M.alipes M. pudiea Baerharia coccinea Altiania incantata Abronia Jragnmi Pisania eapUata FIG. 1. The slnct consensu^ of 260 most-parsimonious trees using the ITS data (TL=544. 0=0.73. RI=0.86). The numbers above the branches are bootstrap values and the numbers bdow are deciv indices. A and B indicate the A and ALietsauHws B clades within the AcUtAtnthts lineage. 29 744 SYSTEMATl : BOTANY [Volume 25 there were only 28 parsimony informative diarac- is dear that in Nyctaginaceae ITS evolves much ters, not including indels. However, they did yield faster than the intergenic region between rbcL and better resolution within Mirabilis sectior. Quamocli- accD (K^R) and the 5' accD coding region (Table 4). dion than ITS, suggesting that M. macfarlanei is sis Although comparable in length, the ITS region has ter to a dade composed of M. greenei and M. mul- five times the number of parsimony informative tiflora. characters as the ICR, and over ei^t times that of Combined. Almost all the random partitions of the much shorter 5' accD region. Despite the large the data had summed lengths greater than the number of changes, the 113 region is not much summed length of the original partitioa However, more homoplastic than the chloroplast regions, as the lengths of the random partitions exceeded the indicated by the high consistency indices for all summed length of the original partition by orJy three regions (Table 4). It therefore appears that the 0.2-1.2% (i.e, maximally 5 steps longer). Therefore; chloroplast ICR and 5' accD coding region are not the nuclear and chloropl^t data were combined. In as phylogenetically informative for d^udng pat cluding indels there were 1811 characters, 134 of terns of relationships among spedes and closely re whicfi were parsimony informative. The strict con lated genera as is the nudear region. This result sensus of 15 trees is shown in Fig. 1 might be expected as most of the ITS region is non- There is strong support for monophyly of ArJris- coding. However, it should be noted that the anthes plus Selinocarpus {AcUisanthes lineage; amount of the 5' accD region that is completely cod BS=100, DI=30) and for monophyly of Mirabilis ing (i.e., does not contain internal stop codons) has (BS=100. DI=32). Within the Acl^nths lineage been reduced in many of the taxa included in this there is moderate support for Acleisanthes A com study, suggesting that this region is no longer fully posed of three spedes of Acleisanthes (BS=87, DI=2) functional. and strong support for the Acleisanthes B dade that The results for Nyctaginaceae are in contrast to indudes the rest of Acleisanthes and all Selinocarpus the phylogenetic utility of these chloroplast regions (BS= 100, Dl=8). Within Acleisanthes B there is mod among spedes of Fagopyrum (Polygonaceae)(Yasui erate support for a dade composed of the other and Ohnishi 1998). What nay explain this discor four spedes of Acleisanthes and Selinocarpus diffusus dance is that the 5' accD region sequenced by Yasui (BS=87, Dl=3), vrith very strong support for Aclits- and Ohnishi (1998) is twice as long as that for Nyc anthes 8 abo\-e Acleisanthes crassifilia (BS=100, taginaceae (due at least in part to diffierert prim DI=5). Relationships are less dear among other ers). The accD gene, which is 1539 bp long in Ni- members of Acleisanthes B, although there is mod cotiana (Shinozaki et al. 1986), is thought to have a erate support for sister taxon relationships between very divergent 5' region and a more conserved 3' Selinocarpus parvifblius and S. lanceolatus (BS=83, region (Doyle et al. 1995). This suggests that the 5' DI=2) and S. somalensis and S. chenopodioides region should be phylogenetically informative in (BS=70; DI=2). Nyctaginaceae and other plant fomilies; the chal In Mirabilis there is strong support for a dade lenge is to find conserved primer sites in such a composed of taxa traditionally placed in section variable area. Quamoclidion plus M. tenuiloba (BS=100, DI=6). Ad The number of scored indels is also notable ditionally, there is moderate support for a dade among the nuclear ITS and chloroplast regions (Ta composed of section Quamoclidion as circumscribed ble 4). The dUoroplast ICR has three times more by Pilz (1978) but excluding Mirabilis triflora scored indels than does the nudear ITS regioa Fur (BS=71, DI=2). ther, the KiR has four times more scored indels ML versus Parsimony. Maximum likelihood than the 5' accD regioa although the 5' accD region analyses yielded tree topologies (not shown) that has more variable sites, parsimony informative were a subset of the resultant topologies assuming sites, and distance between taxa. These differences parsimony. The main difference was that ML help explain the relati%'e ease in locating primes in seordies found fewer optimal tree topologies than the KJR versus the 5' accD region. The number of parsimoity. indels in both the ICR and 5' accD regions is similar to that observed in Fagopyrum (Yasui and CDhnishi DTBCUSSION 1998). Acleisanthes LiTieage. The combined molecular Molecular Evolution. By comparing the nudear evidence strongly supports the monophyly of ribosomal ITS region with the chloroplast locus, it Acleisanthes plus ^linocarpus, but not the mom^y- 30 20001 I.E\1N: PHYLOCENETICS OF NYCTACINACEAE 745 A. anisophyUa A. obtusa A. hngiflora A. crassifolia A. wrightii 100 Ia 30 S. diffusus S A. acuHfolia A. nana 100 S. undulatus S. somalensis 70 c S. chenopodioides S. lanceolatus 67 S. parvi/olius cE S. purpmianus M. hngiflora M. tenuiloba 100 32 M. triflora — too M. multiflora M. macfarlanei sect. QuamocUdion M. greenei M. alipes M. pudica , Abroniajragrans Fic. 2. The strict consensus of 13 most-paisimonious trees using the combined data set of nuclear and diloroplast data and >I bp indels {TL=406, 0=0.87. RI=0.94). The numbers above the branches are bootstrap values and the numbers below are decay indices. A and B indicate the Adaisanthes A and Acleisantha B clades within the Acleisanthes lineage 31 746 SYSTEMATIC BCJTANV [Volume 25 TABLE 4. Comparison of the nuclear ribosomal ITS region and the chloroplast locus differentiated into the intergenic region bet^veen the rbcL and accD genes (ICR) and the 5' accD coding region. Riirwise distances were calculated using an HKY85 distance metric (PAUP*; Swofford 1998). Only the 19 tajia for which sequences were available for all three regions were included. "nr-LTS region includes 25 bp of the 18S and 65 bp of the 26S genes in addition to ITS-I. ITS-2, and the 5.3S gene. ICR 5' «cO nr-FTS Range of raw length 614-640 297-331 655-665 Aligned length 670 333 690 Variable sites (proportion) 4-1 (0 066) 38(0.114) 171 (0J48) Parsimony informative 17 (0.025) 10 (0.030) 88 (0.128) sites (proportion) Pairwise distances 0-3^13% 0-7.I59"(, 0.152-17.41°', Scored indels • 12 3 4 Consistency index 0.96 0.97 0.87 ly of the two genera as currently drcumscribed. for does not have these glands). Further, plants of these two genera to each be monophyletic, it would .^fkisant/us A produce cleistogamous flowers that require 23 more steps (-6% longer). These molec have five stamens rather than the usual two sta ular results are consistent with the similar floral mens found in cleistogamous flowers in the rest of morphology of these two genera. Further, members the genus as well as in many Selinocarpus including of both genera often produce both cleistogamous S. diffusus (Smith 1976; R. Levin, unpubl. data). Ad and chasmogamous flowers on the same individual. ditional morphological support for these relation VVithin this lineage there is strong support for S ships comes from the strong vegetative and floral dtffusus nested within a clade of four species of similarities between ,4. wrightii and .4. acuti/blia, the Acltismlhes (although not included in the analyses main distinction being a slight difference in the lo presented here, another individual of S. diffusus was cation of the resinous glands on the fruit and a also sequenced and yielded the identical result). more robust growth habit m A acuti/blia. Acleisan- This was a surprising finding as S. diffusus has lltes crassijblia has traditionally been considered in winged anthocarp fruits similar to those found in termediate betv«-een the resinous gland clade and all other members of Selinocarpus. However, winged the glandless clade (Smith 1976); although it lacks anthocarp fruits have clearly e\'olved multiple resinous glands, its cleistogamous flowers possess times within the family; they are knovm in Stlino- only two stamens. Thus, morphology is concordant carpiis and Boahavia (Nyctagineae: Boerhaviinae), wi^ the position of A crassijblia as basal to the res Coligmnia (Nyctagineae: Colignotuinae), Phaeoptil- inous gland clade um (Nyctagineae; Phaeoptilinae), Abronia Juss. The traditionally circumscribed Selinocarpus ex (tribe Abronieae), and Grajalesia Miranda (tribe Pt- cluding S. diffusus 1=Selinocarpus sensu stricto) may sonieae) (VVillson and Spellenberg 1977; Bittrich be monophyletic and nested within a polyphyletic and Kiihn 1993). Further morphological evidence Acleisanthes; constraining Selinocarpus s.s. as mono for the phylogenetic placement of S diffusus is its phyletic requires only one additional step. Alter herbaceous habit, similar to Acleisanthes, rather than natively, for monophyly of Selinocarpus sa and the woody habit of most Selinocarpus (Fig. 3). In monophyly of Acleisanthes plus 5. diffusus there is a additioa it has been previously noted that vegeta- cost of eight additional steps (2% longer). Within tively A., wri^tii and S. diffusus look identical Selinocarpus ss. there is limited molecular and mor (Smith 1976). Fifteen more steps (-4% longer) are phological evidence for relationships among taxa. required for Selinocarpus to fbrm a monophyletic Hovwi-er, the ITS data (Fig. 1) do suggest a sister group containing 5. diffusus. ta.xon relationship between 5. angustijblius (cp data The phylogenetic relationships within Acleisanthes were not available for this spedes) and S. undulatus, are concordant with morphology (Fig. 3). The basal which is consistent with morphology. Plants of the .^L-leisanthes A clade including A. obtusa, A langifbn, two spedes are only distinguishable by the undu and A anisophylla all lack resinous glands on their lating leaf margins of S undulatus, and this spedes anthocarps, whereas the other Acleisanthes except previously has been considered a subspecies of S. for /\. CTassijblia haw resinous glands (S. diffusus angustijblius (Fowler and Turner 1977). Further, al- 32 2000] LEVIN; PMXCCENETLCS OF WCTACINACEAE 747 A. anisophylla A. obtusa Five stamens A A. longiflora A. crassifolia A. wrightii S. difjusus PiGlands A. acutifolia Glands I A. nana B 5. undulatus S. somalensis C S. chenopodioides S. lanceolatus S. parvifolius I I Yellow Divar S. purpusianus perianth branch Fic. 3. AL-hrisantlKS linCiige from Fig. 2 ivith morphological characlcrs that support clades within this lineage: The woody habit is indicated b>' gray sh.-.i!ii;g; all other taxa are herbaceous (unclear for S. somlmsis). Gosed bars indicate charactcr gain and open bars indicate loss. The ^leisanllurs A clade shares five stamens in their cleistogamous flowers. Glands are resinous glands on the anthocarp; Di\-ar branch is a divuncating branching pattern of plant form. though support for a clade composed of S. parvifol data (R. Le^in, R. Raguso, and L McOadev in mss.) ius, S. lanceolatits. and S. purpusianus is weak (Rg. 2; further corroborate this clade as well "as the sister BS=67, DI=l), there is strong morphological evi ta.ton relationship between S. paroifblius and S. lan dence for this dadei including a yellow perianth ceolatus. and divaricating brandting pattern unique to this Because of its unique floral and inflonscence three-membered clade (Fig. 3). Floral fragrance morphology. Selinocarpus chenopodioides A. Gray has 748 SYSTEMATIC BOTANY (Volume 25 often been considered as the sole member of its lish the phylogenetic status of Oxybaphus and place own genus Ammocodon (Standley 1916; fbwler and ment of Quamoclidion with respect to members of Turner 1977). However, I agree with its original this large section. placement as a member of Selinocarpus. Molecular Within section Quamoclidion ss., there is no mo data support its phylogenetic position within Sdi- lecular support, but good morphological evidence nocarpus s.s. Selinocarpus somalensis is also well sup for a sister relationship between M. alipes and M. ported as a member of this genus, laying to rest pudica. Plants of these spedes have almost identical doubts about the proper placement of this Somalian campanulate floivers. The shared presence of lin- endemic in an otherwise North American clade. alool oxides in their floral fragrance further sup Further, this taxon does not appear to be sister to ports this relationship (R. Levin. R. Ragusa and L the North American Adeisanthes lineage (requires McDade, in mss.). Similarly, although the clade of 10 more steps, 2.5% longer), suggesting that long M. multifbra, M. macfarlanei, and M. greenei is weak distance dispersal is the most reasonable explana ly supported with molecular data (BS=67, Dl=l), tion for its disjunct occurrence in Somalia. This is all three share magenta, funnelform flowers and are in contrast to Thulin (1994), who explains this dis very similar vegetatively. junct distribution by vicariance combined with ex Higher Relationships. The results presented tinction events. here suggest that the division of subtribe Boethav- Mirabilis. Molecular evidence strongly sup iinae (Heimerl 1934) into subtribes Boerhaviinae ports monophyly of the genus Mirabilis. This is con and Nyctagininae (Bittrich and Kiihn 1993) is not sistent with the presence of often calyx-like invo- warranted. Allionia (Nyctagininae) and Bcerhaoia lucral bracts subtending one to many flowered in (Boerhaviinae) are mote dosely related to each oth florescences. Although only four of ten taxa from er than to either Mirabilis or the Acleisanlhes lineage section Mirabilis were included, this is a relatively (Fig. 1). To support the subtribal drcumscriptionof homogeneous section. Therefore, I would suggest Bittridi and Kiihn (1993), Bcerhavia should be more that there is strong molecular support for the closely related to the Acleisanlhes lineage than it is monophyly of section Mirabilis. This is consistent to either Allionia or Mirabilis. When Boerhaoia was with the shared presence of one flower per invo- constrained to be a member of the Acleisanthes lin luae; creating the appearance of a cal>'x subtending eage and Allionia more dosely related to Mirtbilis, a corolla (Le Due 1995). there were an additional 161 steps required (-40% The monophyly of section Quanwclidion sensu longer). However, before strong condusions can be stricto (i.e., excluding M. triflora) is also well sup made at the subtribal level, more taxa must be in ported by both molecular and morphological data. cluded, with increased sampling within Boerhmna All Quanwclidion SA share an inflorescence com and the indusion of Nyctaginia, Okenia, Caribea, and posed of Ave partially uiuted bracts surrounding Cuscatlania. Further, spedes in the genera Colignon- six or more flowers; most often there is one whorl ia, Pisoniella, and Phaeoptilum should also be indud- of six flowers composed of a central solitary flower ed from the other subtribes within Nyctagineae: We surrounded by five others attached at the bases of now have a phylogenetic understanding of major the five involucral bracts. In M. triflora, inflores genera within Nyctagineae Increased sampling cences have only three flowers arranged as a soli from all tribes and genera of Nyctaginaceae should tary central flower and two borne at the bases of further darify relationships in the entire family. the two largest involucral bracts. This morphology is in foct more similar to that of some species tra ACK.NOVVLEDCE.VE\-TS. The author thanks A. Le Due, ditionally placed in section Oxybaphus [e.g., M. pum- M. Thulin, R. A. Raguso. and M. Fishbein for providing ila (Standley) Standley). It is undear from my re plant Ruteriats and P. jetUuns for idenlifytng the Psonia sults whether M. trifora is basal to Quamoclidion s.s. capitaa. t am also grateful to Big Bend Mitionat Puk and or if it is actually more closely related to spedes in the k>mada LTER for permission to collect plant nuteriaK section Oxybaphus. Section Oxybaphus may contain K. Riley and S. Mosta tor help in the molecular lob, and L A. StcDade, J. S. Miller. H. Ballard and R. SpeOesibog ca. 40 species and is diverse; with both single-flow for critically reading earlier versions of the nunuscript ered and multiple-flowered inflorescences found in This researdi was partially supported tiy NSF grant the section. This diversity in ii\florescence structure DEB9707693 to L. A. McDade and NSF grant DEB9806840 suggests that section Quamcclidum may be nested to L A. McOade and R. A. Raguso. Further support was within Oxybaphus. Howevta; increased sampling of provided by small grants to R. Levin from the Americm spedes from sectiorv 0.xybaphus is needed to estab Sodety of Plant To-xonomists. Sigma Xi. the Deportment of 2000] LEVIN: PHYLOCENEnCS OF NYCTAaXACEAE 749 Ecology tc Evolutionary Biology at the University of Ari GILBERT, D. G 1993. S^qApp: .4 biostquence tditor and anal- zona, and the University of Arizona Research Training pis applicatum. Shareware a\'ailable by author from Croup in the Analysis of Biological Diversification (NSF ftp: 11 iubio.bio.indiana.edu I molbio I seqapp I. DtR-9n336Z BIR-9602246). This paper is part of a disser GR.\.NT, V. and K. A. GRA.vr. 1983. Hawkmoth pollination tation 10 be submitted to the University of Arizona in par of Mirabilii longiflora (Nyctaginaceae). Pnxreedings of tial fulfillment of the re<;uiremenls for the degree of doctor the National Academy of Sciences SO; 1298-1299. of philosophy. HENERL. A. 1934. Nyctaginaceae Pp 56-134 in Die Na- liirlicken Pfhnzenfamilien. 2"^ edition, I6c, eds. A. En- LITERATURE CITED gler and K. Prantl. Leipzig: Engelmann. HER.NA.NDEZ, H. M. 1990. Autopoliruzacion en Mirabilis BALDWIN, B. C 1992. Phylogenetic utility of the internal bnpfhra L (Nyctaginaceae). Acta Botanica Mexicana transcribed spacers of nuclear tibosomal DNA in 11 25-30. plants: an example hmi the Compositae Molecular LE DCC. A. 1995. A re\'ision of Mirabilis sechon Mirabilis Phylogenetics and Evolution 1: 3-16. (Nyctaginaceae). Sida 16; 613-648. , M. J. SA-MDERSON, ). M. PORTER, M. F. WOICIE- MABBERLEY, D. J. 1997. The plant-bock: A portable Jictuinary CHOWSKL, S. S. CA.MPBEA, and .M. J. DONOCHL-E. 1995. of the vascular plants, 2~ edition. Cambridge: Cam The ITS region of nuclear ribosomal DNA; a valuable bridge University Press. source of evidence on angiosperm phylogeny. AnnaU MADDBO.N. D. R. and VV. R MADDISON. 1999. MacClade: of the Missouri Botanical Garden 81 247-277. Analysis of Phylogeny and Character Eailutwn. Test ver BrmucH. V. and U. KCH.\. 1993. Nyctaginaceae Pp. 473- sion 4.0all. To be distributed by Sinauer Associates. 486 in 77K fimilia and genera of vascular plants. Vol. 2. Sunderland, .VIA. eds. K. Kubitzki, J. G Rohwer, and V. Bittrich. Berlin; .VIAHRT, M. and R. SPELLE-NBERC 1995. Ta.tonomy of Cy- Springer-Vbrlag. pkameris (Nyctaginaceae) based on multivariate anal BOCLE, A. L 1974. The genera of Nyctaginaceae in the yses of geographic variation. Sida 16; 679-697. southeastern United States. Journal of the Arnold Ar .VIARTLNEZ DEL RIO, C. and A. BLRQLEZ. 1986. Nectar pro boretum 55; 1-37. duction and temperature dependent pollination in BOHLIN, )-E. 1988. A monograph of the genus Catignonia Mirabilis jalapa L. Biotropica IS; 28-31. (Nyctaginaceae). Nordic )oumal of Botany 8:231-252. PiLi G E. 1978. Sj'stematics of Mirabilis subgenus Quam- BKE.MER, K. 1988. The limits of amino add sequence data ocluiion (Nyctaginaceae). Madroiio 25; 113-132. in angiosperm phylogenetic reconstructioa Eralution SHLNOZAKI. K., .\1. OHME. M. TANAKA. I WAKASLCI, .V. 42; 795-803. H,\YASHROA. T. MATSLBAYASHL N. ZAITA. J. CHLN- CRLOE.N, W. 1970. R. Hawkmoth pollination of Mirabilis woNCSE, |, OBOKATA, K. YASUCLCHI-SHLNOZAKL C (Nyctaginaceae). Bulletin of the Torrey Botanical Qub OHTO, K. TORAZAWA, B. Y. ME.\C. .VI. SccrrA, H. 97; 89-91. DENG. T. K.\.MOCASHIRA, K. YASIADA. J. KL-SLDA, F . 1973. Reproductive biology of weedy and culti TAKAWA. A. KATO, N. TOHDOH. H. SHRIADA, and vated Mirabtlis (Nyctagiiuceae). American (oumal of M. SLCILRA. I9S6. The complete nucleotide sequence Botany 60; 802-809. of tobacco chloroplast genome: its gene orgaiuzation DONCCHUE. M. I.. R. G O-MSTEAD, J. F S.Mna and J. D. and expression. Embo Journal 5; 20I3-2049. PAL.VIER. 1991 Phylogenetic relationships of Dipsa- S.MITH, J. VL 1976. A txxonomic study of Acleisanthes (Nyc cales based on rbcL sequences. Annals of the Vfissouri taginaceae). Wrightia 5; 261-276. Botanical Garden 79; 333-345. SPELLENBERC, R. 1993. Taxonomy of Anulocaulis (Nyctagi DOYLE, J. J. and J. L DOYLE. 1987. A rapid DNA isolation naceae). Sida 15; 373-389, procedure for small quantities of fresh leaf tissue and R. K. DEISON. 1977. Aspects of reproduction Phytodiemical Bulletin 19: 11-15. in Chihuahuan Desert Nyctaginaceae Pp 273-287 in and J. D. PALMER. 19RE. .Multiple independent loss es of two genes and one intron from legume chloro- Transactions of the symposium an the biological resources plast genomes. Systematic Botany 20; 272-294. of the Chihuahuan Desert regain United States and Mezica. eds. R. H. Wauer and D. H. Riskind. U. S. Department FARRB, J. S., KALLERSIO, A. G KLUCE, and C BLLT. 1995. Testing significance of incongruence. Cadistics of the Interioc 10: 315-319. STA.NT3LEY, P. C 1916. Ammocodon, a new genus of Allion- FELSENSTEIN, J. 1985. Confidence limits on phylogenies: An iaceae from the southwestern United States, foumal approach using the bootstrap. Evolution 39:783-791. of the Washington Academy of Sciences 6: 629-631. FOSBERC, F R. 1978. Studies in the genus Boerharm L (Nyc SWOFTORD, D. L 1998. PAUP'. Phylogenetic analysis using taginaceae), 1-5. Smithsonian Contributians to Bota- parsimony (' and other methods). Version 4. Sunderland; m- 39: 1-20. Sinauer .Associates. FOWXER, B. A and B. L TLTLNER. 1977. Taxonomy of Seli- THLTJN. M. 1993. Nyctaginaceae Pp 168-175 in Flora of mxarpus and Amrnxodm (Nyctaginaceac). Phvtologia Somalia. VoLl, ed. M. ThulirL Kew: Royal Botanic Gar 37; 177-208. dens. 750 SYSTEMATIC BOTANY [Volume 25 . 1994. Aspects of disjunct distributions and ende- DNA. .VIolecular Phylogenetics and Evolution 6: 167- mism in the arid parts of the horn of Africa, partic 177. ularly Somalia. Pp. 1105-1119 in Proceedings of the WILLSON. J. and R. SPELLENBERC. 1977. Observations on XIW Plenary iVtetmy AEIF. APPENDIX B TAXONOMIC STATUS 0¥ ACLEISANTHES, SEUNOCARPUS, AND AMMOCODON (NYCT AGIN ACE AE) 37 Taxonomic Status of Acleisanthes, Selinocarpus, and Ammocodon (Nyctaginaceae) Rachel A. Levin Department of Ecology & Evolutionary Biology University of Arizona, Tucson, AZ 85721, USA Tel. 520-621-8220 FAX 520-621-9190 E-mail; [email protected] 38 ABSTRACT. Recent molecular and morphological studies suggest that neither Acleisanthes nor Selinocarpus (Nyctaginaceae) are monophyletic as currently circumscribed. Further, these data provide no basis for continued recognition of the monotypic genus Ammocodon as separate from these other two genera. Therefore, I transfer Selinocarpus and Ammocodon to Acleisanthes. I also reduce three Selinocarpus species to synonymy. The genus Acleisanthes is now composed of sixteen species, distributed primarily in the Chihuahuan Desert of North America. 39 The genus Adeisanthes A. Gray consists of seven species distributed from southern Arizona east to the Texas coast and south to San Luis Potosf, Mexico, with its center of distribution in the Chihuahuan Desert (Smith, 1976). Selinocarpus A. Gray is another small genus of ca. nine species found in the southwestern United States and northern Mexico, with one species occurring in Somalia (Fowler & Turner, 1977). Like Adeisanthes, Selinocarpus is most species-rich in the Chihuahuan Desert of Texas and Mexico. Within this region, many Selinocarpus species appear to prefer gypseous soils (Fowler & Turner, 1977), whereas Adeisanthes inhabit a variety of soils (Smith, 1976: pers. obs.). Although the genera are morphologically similar, Selinocarpus species have been characterized by the presence of winged fruits similar to those found in its namesake genus Selinum (Apiaceae). Further, the majority of Selinocarpus species are perennial shrubs, whereas Adeisanthes species are all herbaceous perennials. However, the genera are similar in most other aspects, including the production of cleistogamous flowers in addition to the often tubular, hawkmoth-pollinated chasmogamous flowers. Since the original descriptions of these genera, Standley (1916) segregated Selinocarpus chenopodioides A. Gray into the monotypic genus Ammocodon Standley as A. chenopodioides (A. Gray) Standley based on characters now understood to be autapomorphic, including a reduction in stamen number and its many-flowered umbellate inflorescences versus the usually solitary flowers found in other species of Selinocarpus. Recent phylogenetic study (Appendix A) of these genera based on molecular sequence data and confirmed by morphology suggests that their current circumscriptions need to be reevaluated. For example, there is very strong support for a sister taxon 40 relationship between A. wrightii and S. dijfiisiis. Morphologically this result is not surprising, as these species are very similar vegetatively, including an herbaceous habit in 5. dijfusus. Further, phylogenetic results clearly split Acleisanthes into two groups. One group is composed of three Acleisanthes species: A. anisophylla, A. longiflora, and A. obtusa. The other four Acleisanthes species including A. wrightii are more closely related to Selinocarpus species and Ammococlon chenopodioides than to the other three Acleisanthes species. Therefore, as neither genus is monophyletic as currently circumscribed. I propo.se that a single genus composed of both Acleisanthes and Selinocarpus species be recognized. Further, it is apparent from both molecular and morphological data that the genus Ammocodon does not warrant recognition at the generic level. Asa Gray (1853) first described both Acleisanthes and Selinocarpus at the same time; thus, neither name has priority. Therefore, I propose that all Selinocarpus species including Ammocodon chenopodioides be transferred into Acleisanthes. As such Acleisanthes is now composed of 16 species, characterized by a perennial habit and the presence of both cleistogamous flowers as well as chasmogamous flowers that are often present simultaneously. All species produce tubular flowers that are frequently white or yellow, less often pink or purple, and each flower is subtended by 1—4 subulate bracts. Flowers are generally axillary, most are hawkmoth-pollinated and fragrant, and fruits are either ribbed or winged. Many of these species can be locally common, but are rare at the landscape level and tend to be limited to certain substrates. The following key to Acleisanthes is based upon my study of both fresh and herbarium material, and relevant literature (Smith, 1976; 41 Fowler & Turner, 1977; Thulin, 1993). The fruit characters included in the key to Acleisanthes are not difficult to observe, because plants are rarely found without fruits from chasmogamous flowers or cleistogamous flowers in various stages of maturity. Thus, most plants are readily assessed using fruit characteristics. KEY TO ACLEISANTHES: 1 a. Plants herbaceous perennials 2a. Fruits winged 3a. Flowers in an umbellate inflorescence of ca. 10 or more flowers; perianth magenta-purple, 4—5 mm long; stamens usually 2 A. chenopodioides (A. Gray) R. A. Levin 3b. Flowers usually solitary; perianth white, 20—50 mm long; stamens usually 5 4a. Perianths 30—50 mm long, plants found on various substrates in the southwestern United States A dijfusiis (A. Gray) R. A. Levin 4b. Perianths 20—35 mm long, plants found on gypsum in Somalia A. somalensis (Chiovenda) R. A. Levin 2b. Fruits ribbed, not winged 5a. Fruits with resinous glands at or near proximal end 6a. Plants small, stems 2—8 cm long; perianths purple, 8—10 mm long A. nana L M. Johnston 6b. Plants larger, with stems 1—4 dm long; white perianths, 25-45 mm long 7a. Plants upright, resinous glands below proximal end of fruit, plants occurring from 750-2500 m elevation A. aciitifolia Standley 7b. Plants more prostrate, resinous glands at proximal end of fruit, plants occurring from 250—1000 m elevation A. wrightii (A. Gray) Bentham & Hooker 5b. Fruits without resinous glands 8a. Cleistogamous flowers with 2 stamens, veins of upper leaf surfaces with white conspicuous trichomes A. crassifolia A. Gray 8b. Cleistogamous flowers with 5 stamens, veins of upper leaf surfaces glabrous 9a. Plants prostrate, stems to 15 cm long; leaves at a node strongly anisophyllous; veins deeply incising upper leaf surface; perianth pink; fruits with 10 ribs A. anisophylla A. Gray 9b. Plants clambering, stems to 1 m long; leaf pairs equal to subequal; veins of upper leaf surface inconspicuous; perianth white or pinkish white; fruits with 5 ribs 10a. Flowers solitary, perianths to 170 mm long A. longiflora A. Gray 42 10b. Flowers solitary or in cymes of 2--5 flowers, perianths 25— 50 mm long A. obtiisa (Choisy) Standley lb. Plants woody perennial shrubs 1 la. Plants with divaricating branching, having profusely-branched stems; perianths usually yellow, 30--50 mm long 12a. Shrubs small, to 3 dm in height; leaves up to 40 mm long; perianths yellow to white A lanceolatus (Wooton) R. A. Levin 12b. Shrubs to 6 dm tall, leaves to 20 mm in length, perianths usually yellow 13a. Leaves linear to lanceolate: stems grayish-purple; perianths yellow, white, or pink; plants occurring on gypsum 14a. Perianths pink, mature fruits 5—7 mm long A. palmed (Hemsley) R. A. Levin I4b. Perianths yellow or white, mature fruits 7.5—10 mm long A. piirpiisiamis (Heimerl) R. A. Levin 13b. Leaves ovate, stems not gray; perianths yellow, found more often on sandstone or limestone A. pcirvifolius (Torrey) R. A. Levin I lb. Plants virgate, stems having few branches; perianths greenish-yellow to red- orange, 6~ 12 mm long 15a. Leaf margins sinuous, perianths ochreous A. iindulatiis (B. A. Fowler & B. L. Turner) R. A. Levin 15b. Leaf margins straight, perianths greenish-yellow A. angustifoliiis (Torrey) R. A. Levin Acleisanthes angustifolius (Torrey) R. A. Levin, comb. nov. Basionym: Selinocarpus angiistifolius Torrey, Bot. Mex. Bound. 170. 1859. TYPE: Mexico. Chihuahua: Presidio del Norte (Ojinaga), July 1852, Parry s. n. [lectotype, NY not seen (Fowler and Turner 1977); isotype, US]. Specimens examined. USA. Texas, Brewster County: ca. 75 mi. S of Alpine, Powell 2180 (NMSU); Tomillo Creek in Big Bend National Park, 29° 11.447'N 103°00.389'W, Levin 97-1 (ARIZ). Texas, Presidio County: 2 mi. N of Presidio. Spellenberg & Moore 2623 (NMSU). MEXICO. Chihuahua: base of north end of Sierra del Cuchillo Parado where crossed by Ojinaga-Cd. Chihuahua Hwy. 60 km W of Ojinaga, 29°35'30"N 104°54'W, Chiang. Wendt & Johnston 9772 (NMSU). Coahuila: ca. 6.9 mi. SW of Cuatro Cienegas on hwy. to San Pedro, then 0.7 mi E of hwy. in first 43 canyon on N side of limestone Sierra San Marcos del Pino, near 26°54'N 102°6'W, Henrickson 20402 (ARIZ). Durango: ca. 0.5 mi SW of Lerdo on first limestone road cut, Spellenberg & Syvertsen 3763 (NMSU). Acleisanthes chenopodioides (A. Gray) R. A. Levin, comb. nov. Basionym: Selinocarpiis chenopodioides A. Gray, Am. J. Sci. II 15: 262. 1853. Ammocodon chenopodioides (A. Gray) Standley, J. Wash. Acad. Sci. 6: 629. 1916. TYPE: USA. Texas or New Mexico, 1851 or 1852, C. Wright 1707, left side of sheet [lectotype. GH (Fowler & Turner 1977)]. Specimens examined. USA. Arizona, Cochise County: Portal-Rodeo Rd.. 4 mi. NW of Rodeo, near cattle guard, McConnick & Assoc. 215 (ARIZ). New Mexico, Dona Ana County: The Jornada LTER, G-BASN NPP site, northeast comer. Lower Summerford Mtn. Bajada slope, 32.31°46.64'N 106.47° 11.79'W, Levin 99-1 (ARIZ). Texas, Hudspeth County: N end of the Quitman Mtns. at a roadside park along I-10. 31°12'50"N 105°29'30"W, Worthington 6694 {ARIZ). Acleisanthes diffusus (A. Gray) R. A. Levin, comb. nov. Basionym: Selinocarpus diffiisiis A. Gray, Am. J. Sci. U 15: 262. 1853. TYPE: USA. Texas: Pecos Co., 1851 or 1852, C. Wright 1708 [lectotype, GH (Fowler & Turner 1977)]. Selinocarpus nevadensis (Standley) B. A. Fowler & B. L. Turner, Phytologia 37:201. 1977. Syn. nov. Selinocarpus diffusus A. Gray subsp. nevadensis Standley, Contr. 44 U.S. Nat. Herb. 12:388. 1909. TYPE: USA. Nevada: Lincoln Co., Overton, May 1891, Bailey 1932 (holotype, US). Fowler and Turner (1977) defined 5. nevadensis as distinct from 5. dijfusiis on the basis of having wider, more rounded leaves and a more narrow, delicate perianth. However, these differences are likely due to environmental effi'fts on very plastic morphological traits, especially as these two taxa do not occur in sympatry. Therefore, I do not believe there is sufficient evidence that these species are separate, and I am reducing 5. nevadensis to synonymy with A. dijfusus. Specimens examined. USA. Arizona, Mohave County: along US Hwy. 93 ca. 30 mi. NW of Chloride. Barr 68-283 (ARIZ). Nevada, Clark County: Mojave Desert, ca. 30 mi. N of Las Vegas, 19 mi. S of Indian Springs, E side of US Hwy. 95, Spellenberg 9482 (NMSU); ca. 2 mi. N of Jean, off of frontage road by 1-15, 35°47.652'N 115°18.16rW, Levin 98-1 (ARIZ); Mt. Springs Summit, 2.1 miles below Blue Diamond cutoff. Niles & Plum 972 (ARIZ). New Mexico, De Baca County: 48 air km S of Fort Sumner, E side Pecos River, summit of Espia Peak, SW 1/4 S24 R25E T3S, 104°16.57'W 34°01.67'N, Spellenberg & Zucker 12404 (NMSU). Oklahoma, Harmon County: on gypsum 6 mi. south of Hollis, Waterfall 9001 (ARIZ). Acleisanthes lanceolatus (Wooton) R. A. Levin, comb. nov. Basionym: Selinocarpus lanceolatiis Wooton, Bull. Torr. Bot. Club 25: 304. 1898. TYPE: USA. New Mexico: Dona Ana Co., just S of the White Sands, 26 August 1897, Wooton 389 (holotype, US; isotypes, NMSU, US). 45 Selinocarpiis maloneanus B. L. Turner, Phytologia 75:239. 1993. Syn. nov. TYPE: USA. Texas: Hudspeth Co., N end of Malone Mtns., near Finley, 29 July 1958, D. S. Correll & I. M. Johnston 20358 (holotype, LL not seen). Selinocarpiis megaphyllus (B. A. Fowler & B. L. Turner) B. L. Turner, Phytologia 75:242. 1993. Syn. nov. Selinocarpus lanceolatus Wooton var. megaphyllus B. A. Fowler & B. L. Turner, Phytologia 37:181. 1977. TYPE: Mexico. Chihuahua: ca. 15 mi. SW of Estacion Moreon on Rfo Conchos Lake Road, Sierra de las Monillas, 25 May 1971, Powell 2105 (holotype, LL not seen). The main distinctions between S. maloneanus, S. megaphyllus, and ^4. lanceolatus are in leaf shape and width, and 5. megaphyllus is reportedly distinguished from the other two species by a larger, white perianth (Fowler & Turner, 1977; Turner. 1993). However, Fowler & Turner (1977) note that there is morphological overlap between S. megaphyllus and A. lanceolatus, and the characters that have been used to separate these three species are known to be quite plastic in plants. An excellent example of the wide variation in leaf shape within a population is the collection of Spellenberg & Syvertsen 3758. listed below, that includes specimens from multiple individuals at the same locality. These individuals were clearly conspecific, with gradations in leaf width as the only distinction among the individuals. I have examined specimens from the type localities of both S. maloneanus (Spellenberg & Syvertsen 3753) and S. megaphyllus (Henrickson 6799) as well as a wide range of other specimens of A. lanceolatus, and I do not believe that the distinctions outlined by Turner (1993) warrant specific recognition. 46 Specimens examined. USA. New Mexico, Otero County: about 14 mi. W of Alamogordo on US hwy., at Point of Sands just west of White Sands National Monument, Spellenberg 3910, (NMSU). Texas, Culberson County: State Hwy. 54 N of Van Horn, 15 mi. S of junction with US Hwy. 62-180 in NW comer of county, Spellenberg & Ward 97H (NMSU); 36 mi. N of Van Horn on Hwy. 54, 31°31.82rN I04°50.745'W, Levin 97-5 (ARIZ). MEXICO. Chihuahua: 12 mi. NE of Cuchillo Parado along old road in a gypsum outcropping N of Rfo Conchos, 29°33'N 104°4rW, Henrickson 6799 (NMSU); 13.6 km W of Camargo on road to Lago Boquilla, on gypsum river terraces, Spellenberg NE of railroad crossing. Chihuahua Hwy. 16, Spellenberg Acleisanthes palmeri (Hemsley) R. A. Levin, comb. nov. Basionym: Selinocarpus palmeri Hemsley, Biol. Centr. Am. Bot. 3: 6. 1882. TYPE: Mexico. Coahuila: San Lorenzo de Laguna, May 1880, Palmer II18 (holotype, K not seen; isotype, US). The characters that appear to distinguish this species from A. piirpusianiis are a smaller fruit and the presence of a pink perianth, rather than the yellow or white perianth found in A. piirptisianus. In addition to these slight morphological differences, A. palmeri is known only from two proximate localities. Therefore, it may be more appropriate to synonymize A. palmeri with A. piirpusianus, although nomenclatural priority would be given to A. palmeri. However, as I have not observed A. palmeri in the field and have only viewed two herbarium specimens of this species, I do not believe it is appropriate at the present time to reduce these two species to synonymy. A1 Specimens examined. MEXICO. Coahuila: ca. 15 road mi. E of Torreon and about 2.6 mi. by winding road E of El Coyote, at the NW end of a small mtn. range, the Sierra de Solis, mostly on the tops of knolls, Spellenberg & Syvertsen 3768 (NMSU). Acleisanthes parvifolius (Torrey) R. A. Levin, comb. nov. Basionym: Selinocarpus dijfiisiis A. Gray var. parvifolius Torrey, Bot. Mex. Bound. 168. 1859. Selinocarpus parvifolius (Torrey) Standley, Contr. U. S. Nat. Herb. 12: 388. 1909. TYPE: USA. Texas: Canons of the Rio Grande or Presidio del Norte, 1852, Me.x. Bound. Sur. (Bigelow, Parry et al.) s. n. [lectotype, GH (Fowler & Turner 1977); isotype, US|. Specimens examined. USA. Texas, Brewster County: Tomillo Creek, Big Bend National Park, 29°1.447'N !03°00.389'W, Levin 97-2 (ARIZ). Texas, Hudspeth County: 33 mi. SE from freeway on Texas Rd. 192, on county portion of road, between Quitman Mts. and the Rio Grande, Spellenberg & Ward 9688, (NMSU). MEXICO. Chihuahua: 10.7 road miles south of Ojinaga along Hwy. 18 (Ojinaga-C. Camargo), open sandstone hill along hwy., 29°24'N 104°20'W. Hendrickson 7704 (NMSU). Acleisanthes purpusianus (Heimerl) R. A. Levin, comb. nov. Basionym: Selinocarpus purpusianus Heimerl, Oesterr. Bot. Zeits. 63: 353. 1913. TYPE: Mexico. Coahuila: Sierra del Rey, June 1910, Purpus 4505 (holotype, US). Specimens examined. MEXICO. Coahuila: Km 134 on Hwy. 30 towards Torreon, 26°37.316'N 102° 15.701'W, Levin 98-6 (ARIZ); Hwy. 30, 68 mi. N of San Pedro, on gypseous flat, about 7 mi. S of Puente Arroyo Seco, Spellenberg, Willson & Feather 48 4050 (NMSU); ca. 12 (air) mi. SW of Cuatro Cienegas (13.5 road mi.) along hwy. to San Pedro and 4.4 mi. W on trail, 26°53'N 102°12'W, Henrickson 12552 (NMSU); Hwy 57, 1.2 mi. S of Hermanas, Spellenberg & Syvertsen 3779 (NMSU). Acleisanthes somalensis (Chiovenda) R. A. Levin, comb. nov. Basionym: Selinocarpiis sonialensis Chiovenda, Flora Somala 284. 1929. TYPE: Somalia. Costa dei Migiurtini. dintomi di Biaddo, June 1924, Puccioni and Stefanini 814 (holotype, FT not seen). Specimens examined. SOMALIA. Buran (Sorl): Lat 10-13N, Long 48-47E. Collenette 125 (K). Galguduud: 4 km S of Ceelbuur (El Bur) on road to Gal Hareeri, gypsum plain, 4:39N 46;37E. Thulin & Dahir 6517 (K). Acleisanthes undulatus (B. A. Fowler & B. L. Turner) R. A. Levin, comb. nov. Basionym: Selinocarpiis undulatus B. A. Fowler & B. L. Turner. Phytologia 37: 194. TYPE: Mexico. Coahuila: 4 mi. W of Cuatro Cienegas, mouth of canyon, 24— 26 August 1938,1. M. Johnston 7159 (holotype, GH). Specimens examined. MEXICO. Coahuila: outside of Cuatro Cienegas at Km 94 on Hwy. 30 towards Torreon, 26°54.062'N 102°08.430'W, Levin 98-7 (ARIZ); 8 km W of Cuatro Cienegas on S slopes and foothills of Sierra de la Madera, 26°58'30"N [02°08'W, Johnston, Chiang & Morofka 12076 (NMSU); Mpio. Parras. S side of Sierra Parras along road toward Menchaca, ca. 14 air km S of Parras, just N of Sierra Prieta. 25°18'N 102°13'W, 7651 (NMSU). 49 Acknowledgements. I thank Big Bend National Park and the Jornada LTER for permission to collect plants, R. Spellenberg for help in locating plant populations, and the staff at ARIZ for their assistance. The following herbaria allowed me to borrow specimens: GH, K, NMSU, US. I thank L. A. McDade for critically reading an earlier version of this manuscript. Financial support was provided by the National Science Foundation DEB9806840 to L. A. McDade and R. A. Raguso and by small grants from The American Society of Plant Taxonomists, Sigma Xi, the Department of Ecology & Evolutionary Biology at the University of Arizona, and the University of Arizona Research Training Group in the Analysis of Biological Diversification (NSF DIR- 9113362. BIR-9602246). This paper is part of a dissertation submitted to the University of Arizona in partial fulfillment of the requirements for doctor of philosophy. Literature Cited Chiovenda, E. 1929. Flora Somala. Sindicato Italiano Arti Grafiche, Rome. Fowler, B. A. & B. L. Turner. 1977. Taxonomy of Selinocarpiis and Ammocodon (Nyctaginaceae). Phytologia 37: 177-208. Gray, A. 1853. Brief characters of some new genera and species of Nyctaginaceae, principally collected in Texas and New Mexico. American Journal of Science and Arts, second series 15: 259-263. Heimerl, A. von. 1913. Eine neue Art der Gattung Selinocarpiis. Osterreichische Botanische Zeitschrift 63: 353-356 50 Hemsley, W. B. 1882. Biologia Centraii-Americana; or, contributions to the knowledge of the fauna and flora of Mexico and Central America, Botany, Vol. 3. R. H. Porter & Dulau & Co., London. Smith, J. M. 1976. A taxonomic study of Acleisanthes (Nyctaginaceae). Wrightia 5: 261- 276. Standley. P. C. 1909. The Allioniaceae of the United States with notes on the Mexican species. Contributions from the United States National Herbarium 12(8): 303-389. Standley, P. C. 1916. Ammocodon, a new genus of Allioniaceae, from the southwestern United States. Journal of the Washington Academy of Sciences 6: 629-631. Thulin, M. 1993. Nyctaginaceae. Pp. 168—175 in M. Thulin (editor). Flora of Somalia. Vol.1. Royal Botanic Gardens, Kew. Torrey, J. 1859. Report on the United States and Mexican Boundary Survey, Botany. Vol. 2. A O. P. Nicholson, Washington DC. Turner, B. L. 1993. New species and combinations in Selinocarpus (Nyctaginaceae). Phytologia 75: 239-242. Wooton, E. O. 1898. New plants from New Mexico.—II. Bulletin of the Torrey Botanical Club 25: 304-310. 51 APPENDIX C FRAGRANCE CHEMISTRY AND POLLINATOR AFRNITIES IN NYCTAGINACEAE 52 Fragrance chemistry and pollinator affinities in Nyctaginaceae Rachel A. Levin" *, Robert A. Raguso'', and Lucinda A. McDade" 'Department of Ecology & Evolutionary Biology University of Arizona. Tucson, AZ 85721, USA Tel. 520-621-8220 FAX 520-621-9190 ''Department of Biology Coker Life Sciences Building University of South Carolina Columbia SC 29208. USA E-mail: [email protected] 53 Abstract We present results of dynamic head-space collections and GC-MS analyses of floral and vegetative fragrances for twenty species in three genera of Nyctaginaceae: Acleisanthes, Mirabilis, and Selinocarpus. Most of the species included in this study are either hawkmoth- or noctuid moth-pollinated. A wide variety of compounds were observed, including mono- and sesquiterpenoids, aromatics (both benzenoids and phenylpropanoids), aliphatic compounds, lactones, and nitrogen-bearing compounds. Intraspecific variation in fragrance profiles was significantly lower than interspecific variation. Each species had a unique blend of volatiles, and the fragrance of many species contained species-specific compounds. The fragrance profiles presented here are generally consistent with previous studies of fragrance in a variety of moth-pollinated angiosperms. Keywords: Acleisanthes', Mirabilis: Selinocarpus-, Nyctaginaceae; Four o'clock; floral fragrance; variation; pollination; hawkmoths 54 1. Introduction In many pollination interactions, floral fragrance appears to act in concert with visual cues to attract floral visitors (see review by Dobson, 1994). Consequently, a number of studies have examined the relationship of fragrance composition to type of floral visitor (Knudsen andTollsten, 1993; Borg-Karlson et al., 1994; Kaiser and Tollsten, 1995; Knudsen and Tollsten, 1995; Bestmann et al., 1997; Dobson et al., 1997; Miyake et al., 1998; Grison et al., 1999). The fragrances of hawkmoth-pollinated flowers in a number of plant families have been of particular interest (Knudsen and Tollsten, 1993; Kaiser and Tollsten, 1995; Raguso and Pichersky. 1995; Barthlott et al.. 1997; Miyake et al., 1998). Such flowers often have similar morphologies, with long, tubular perianths that are usually white, occasionally yellow or pink (Grant, 1983; Haber and Frankie, 1989). Because these flowers generally emit fragrance at night, when most hawkmoths are active, it has been suggested that convergent evolution for hawkmoth attraction to plants of various angiosperm families may involve scent chemistry as well as floral morphology (Nilsson et al., 1985; Knudsen and Tollsten, 1993; Miyake et al., 1998). To date there has been some support for this idea (Kaiser, 1993; Knudsen and Tollsten, 1993; Miyake et al., 1998); oxygenated terpenoids and nitrogen-bearing compounds are frequently present in the fragrance of hawkmoth-pollinated flowers. However, the high degree of among species variation in fragrance chemistry identified in these studies suggests that convergence is but one of several factors affecting scent chemistry. 55 Nyctaginaceae are a small family of trees, shrubs, and herbs, distributed mainly in tropical and subtropical regions of the New World. This family is commonly called the Four o'clock family, as most species have flowers that open in the late afternoon to early evening. Further, the flowers of many species are pollinated by hawkmoths (Grant, 1983; Grant and Grant, 1983; Martinez del Rfo and Biirquez, 1986; Hodges, 1995). Thus, it is an ideal group in which to examine volatiles produced from hawkmoth-pollinated flowers. Here we present results of head-space volatile collections from flowers and vegetation of species in three genera of Nyctaginaceae: Acleisanthes, Mirabilis, and Selinocarpiis (Table 1). This is the first report of fragrance compounds for any Nyctaginaceae except Mirabilis jalapa (Heath and Manukian, 1994). In particular, we investigated within versus among species variation in fragrance compounds for 20 species of Nyctaginaceae. The prediction was that intraspecific variation would be less than interspecific variation, supporting the premise that species have distinct fragrance profiles. Further, we asked how fragrances differed within and among these three genera of Nyctaginaceae, and how these differences are related to pollinator affinities. 2. Results and discussion Our analyses identified a wide variety of volatile compounds across all 20 species examined (Table 2), including representatives of at least seven different biosynthetic classes: monoterpenoids, sesquiterpenoids, fatty acid and amino acid-derived aliphatic compounds and lactones, benzenoids, phenylpropanoids, and nitrogen-bearing 56 compounds [the nitrogen-bearing compounds are known to result from three different biosynthetic pathways (Radwanski and Last, 1995; Wink, 1997)]. The most commonly occurring fragrance compounds were 4,8-dimethyl-nona-l,3,7-triene (present in all species), and trans-p-ocimene, cis-3-hexenyl acetate, and methyl salicylate, found in 18 of 20 species (Table 2). There was also a preponderance of cis-3-hexenyl esters across all three genera, including both aliphatic and benzenoid compounds; this may be an example of a single enzyme or a group of similar enzymes reacting with multiple substrates (see Wang and Pichersky, 1999; Dudareva and Pichersky, 2000). 2.1. Within species variation Intraspecific variation in floral and vegetative fragrance quality (types of compounds) and quantity (amounts of compounds) was significantly lower than interspecific variation [P<0.001,Wilcoxon Signed Ranks Test (SPSS Inc., 1999); the median % dissimilarity of relative amounts of volatiles within species = 10.8 vs. median % dissimilarity among species = 17.5]. These results are consistent with those of Barkman et al. (1997), who found that whereas intraspecific variation was substantial in Cypripediiim (Orchidaceae). it was not as great as interspecific variation. Thus, it is reasonable to conclude that each species has its own characteristic fragrance profile, which can then be compared to the fragrance profiles of other taxa. It should be noted that the among species comparisons spanned a wide range of phylogenetic relationships (i.e., some species were more closely related than others), and it would be ideal to control for degree of phylogenetic relatedness (Appendix D). Further, the amount of intraspecific variation did vary among taxa. In particular, in species that produce many different volatiles, there are necessarily 57 more sources of variation and, thus, higher levels of intraspecific variation relative to other species. For example, the fragrance profile of A. wrightii contains the most compounds (Table 2), and this species also has one of the highest levels of intraspecific variation. By contrast, Mirabilis species released fewer volatiles than Acleisanthes and Selinocarpus (Table 2) and had fairly low levels of intraspecific variation. Further, the Selinocarpus species that released the fewest compounds (i.e., Selinocarpus angustifolius and S. undiilatiis\ Table 2) had levels of intraspecific variation similar to Mirabilis. Our results are generally consistent with a previous report of fioral fragrance in Mirabilis jalapa (Heath and Manukian, 1994). Despite methodological differences (we sampled fragrance from one individual whereas they sampled from 12) and selective reporting of fragrance compounds by Heath and Manukian (1994), many similarities are apparent. Our study confirms the presence of trans-P-ocimene as the major component of the fragrance of M. jalapa (38% of total volatiles emitted; Table 2). In further agreement with the previous study, we found cis-3-hexenyl acetate and myrcene. Our study did not detect benzaldehyde and indole; this may not be surprising, as Heath and Manukian (1994) report that neither of these compounds was found in all of their M. jalapa fragrance samples. We did find other compounds not reported by Heath and Manukian (1994), including a-famesene (37%) and benzyl acetate (5%). 2.2. Volatiles in taxa with diurnal pollination The two species in this study known to be pollinated by organisms other than moths (Table 1), Mirabilis triflora and M. macfarlanei, had low total volatile production (Table 2). Hummingbirds have been shown to ignore fragrance while nectar-foraging (Bene, 58 1945; van Riper, 1960), and hummingbird-pollinated flowers are generally scentless (Faegri and van der Pijl, 1979). Nevertheless, flowers of the hummingbird-pollinated M. triflora do release some fragrance (Table 2). It is possible that the volatile compounds that are no longer necessary for pollinator attraction may be retained for herbivore deterrence (Mullin et al., 1991; Berenbaum and Seigler 1992; Dobson, 1994). However, as its closest relatives also produce fragrance (Appendi.x A), the emission of volatiles by flowers of M. triflora may suggest more about its phylogenetic heritage than an absence of strong selection against fragrance production due to its mode of pollination (Appendix D). Mirabilis macfarlanei has been shown to be mainly bee-pollinated (Barnes. 1996). In addition to low total volatile production, its flowers also release few compounds, with the monoterpene limonene as the main compound (43%; Table 2). This compound is known to be perceived by bees (reviewed by Dobson, 1994) and may be sufficient for attraction in combination with visual cues. 2.3. Volatiles in moth-pollinated taxa The majority of species in this study are moth-pollinated and have fragrance, but there are many quantitative and qualitative differences among them. One notable result is that the Mirabilis species generally have lower total amounts of volatiles than do Acleisanthes and Selinocarpus species (Table 2). There is also a substantial difference in the number of fragrance compounds emitted, with Acleisanthes and Selinocarpus species averaging twice as many compounds as Mirabilis species (mean number of compounds 59 emitted by Mirabilis spp. = 24; mean number of compounds emitted by Acleisanthes and Selinocarpiis spp. = 62; Table 2). In terms of biosynthetic classes, Mirabilis fragrances are less diverse, with no species producing phenylpropanoids or lactones (Table 2). The presence of compounds from these two biosynthetic classes also appears to be phylogenetically informative, being characteristic of closely related species (Appendix A) within both Acleisanthes (phenylpropanoids) and Selinocarpiis (lactones). Interestingly, most of the lactones observed do not appear to have been reported previously from floral fragrances (Knudsen etal., 1993). In addition to the identity of individual fragrance compounds, the source of the volatiles should also be considered when comparing fragrance profiles among species, as both vegetative and floral volatiles are known to attract floral visitors (Raguso. 2001). For example, crepuscular hawkmoths respond to fragrance from close range (0-10 m) by approaching its source and then probing at bright objects (Brantjes. 1978; Raguso and Willis. 2000); they do not functionally distinguish between floral and vegetatively- derived odors. Acleisanthes acutifolia and A. wrightii have very fragrant vegetation, with most of the total volatile production occurring from vegetation rather than from flowers {ca. 82%; Table 2). Specifically, sesquiterpenoids are released in large amounts from the vegetation of both species. A similar situation is found in Selinocarpiis angiistifolius and S. iindiilatiis: plants of these species produce terpenoids vegetatively. In addition to the terpenoids, the lipoxygenase-derived "green leaf volatiles" (mainly compounds with a cis-3-hexenyl moiety) are often released from vegetation (Hatanaka et al., 1986; Croft et 60 al., 1993). Thus, it is not surprising that most of these compounds are also emitted vegetatively in these four species. Interestingly, S. angustifolius and S. iindulatiis are strongly supported as sister species, and A. wrightii and A. acutifolia are also very closely related within the genus (Appendix A). This further suggests that phylogenetic history explains some of the variation in fragrance profiles (Appendix D). By contrast, across all taxa some compounds are mainly produced by the flowers (Table 2). These include the lactones and phenylpropanoids, the isoleucine-derived tiglates (Hill et al., 1980), cis-jasmone, and the majority of the benzenoid compounds. Further, most of the nitrogen-bearing compounds are floral in origin. Thus, whether floral or vegetative volatiles dominate the fragrance profile depends on the species; across all twenty species the volatiles emitted from flowers varied from 5-l(X)% of the total amount of fragrance emitted (Table 2). The amount of floral compounds produced per |ig of floral mass varies substantially among species (Table 2). Although compounds such as trans-P-ocimene, nerolidol, cis- jasmone, and methyl benzoate are produced by flowers of many species, the amount produced can differ greatly, such that the relative importance of a particular compound in moth attraction may depend not simply on its presence but also the amount of the compound that is released. In general, our results are consistent with other studies that have detected floral fragrance in moth-pollinated taxa; however, the one exception is the absence of fragrance in Mirabilis greenei. In addition to its low volatile production, of the five volatile compounds collected from this species, only the homoterpene 4,8-dimethyl-nona-1.3,7- 61 triene was likely to be floral, with the others emitted from the vegetation (Table 2). This lack of fragrance is surprising if this species is indeed moth-pollinated. However, any interpretation of the fragrance of M. greenei should be considered preliminary, as the fragrance samples from this species were collected in the field when temperatures were unusually low (these conditions were unique to this species' fragrance collection); low temperatures have been reported to inhibit volatile production (Jakobsen and Olsen. 1994). 2.4. Fragrance and pollination Several authors have suggested a link between the presence of particular fragrance compounds and moth pollination (Nilsson et al.. 1985; Kaiser, 1993; Knudsen and Tollsten, 1993; Miyake et al., 1998). Kaiser (1993) described moth-pollinated flowers as having awhite-floral'" odor, containing acyclic terpene alcohols such as linalool and nerolidol and simple aromatic alcohols like benzyl alcohol, 2-phenylethanol, and their esters. Jasmonates, tiglates, lactones, and nitrogen-bearing compounds are also prevalent in a variety of moth-pollinated flowers (Kaiser, 1993). Knudsen and Tollsten (1993) surveyed hawkmoth- and noctuid moth-pollinated flowers among 15 species in 12 genera and 9 plant families with results consistent with Kaiser's (1993) description of moth- pollinated floral fragrances. However, they suggested that the presence of geraniolic compounds and oxygenated sesquiterpenes distinguishes hawkmoth-pollinated flowers from noctuid moth-pollinated flowers. The results presented here are consistent with Kaiser (1993) and support the suggestion of Knudsen and Tollsten (1993) that oxygenated sesquiterpenes are more 62 characteristic of hawkmoth- than noctuid moth-pollinated species. Of the 12 likely hawkmoth-pollinated taxa in the present study, the fragrances of 8 of them contained at least one oxygenated sesquiterpene, although the fragrances of 2 of the likely noctuid moth-pollinated species also have oxygenated sesquiterpenes (Tables 1, 2). We also observed that Selinoccirpus angustifolius and S. undulatiis, the two Selinocarpus species that are likely pollinated by noctuid moths (Table 1), emitted the lowest total amount of volatiles and had the least diversity of compounds compared to other Selinocarpus. Based on floral morphology, Selinocarpus chenopodioides would appear to be noctuid moth-pollinated, but the white-lined sphinx moth {Hyles lineata) has also been observed visiting its flowers (Table 1; Appendix D). This is not surprising, as flowers of 5. chenopodioides are similar to flowers of hawkmoth-pollinated Selinocarpus species in that they produce a large quantity of a variety of volatiles, including oxygenated sesquiterpenes (Table 2). Thus, while pollination by smaller insects may be occurring, the flowers are also attractive to hawkmoths. By contrast, the flowers of Mirabilis alipes, which are likely noctuid moth-pollinated, emit a low diversity of volatiles and lack oxygenated sesquiterpenoids. However, flowers of M. alipes also attract Hyles lineata, although the hawkmoths collected while visiting this species did not carry any pollen (Table 1; Appendix D). 2.5. Conclusions The fragrances of all of the moth-pollinated species included in this study (except A/. greenei) are similar in containing mono- and sesquiterpenoids, aliphatic compounds, benzenoids, and nitrogen-bearing compounds. These results of fragrance analyses are 63 consistent with the '"white-floral"' odor described by Kaiser (1993) and the moth- pollinated fragrances reported by Knudsen and Tollsten (1993) and Miyake et al. (1998). However, the floral volatiles reported here and in previous studies are not unique to fragrances of hawkmoth-pollinated flowers; for example, many of these volatiles have also been reported from beetle-pollinated flowers (Azuma et al., 1997). In addition to the chemical similarities of fragrance across moth-pollinated taxa, there are many differences that give flowers of each species their distinctive odor. These include species-specific compounds (e.g., vanillin in A. wrightii and methyl nicotinate in 5. chenopodioides), compounds shared by closely related species (e.g., 5-octalactone and 5-nonalactone in S. lanceolatiis and S. piirvifolius), and species-specific blends of more ubiquitous compounds. Fragrance plays diverse roles in hawkmoth pollination, attracting moths from a distance (Tinbergen, 1958; Haber, 1984), combining with visual cues to elicit floral approaches and feeding (Brantjes, 1978: Raguso and Willis, 2000), and facilitating associative learning and discrimination (Daly and Smith, 2000). Among diverse hawkmoth-pollinated flowers, the presence of common classes of compounds, such as oxygenated terpenoids and aromatic esters, combined with the species-specific nature of each blend documented in our study and its antecedents, provides a complex ecological signal with the potential to support all of these functions. 3. Experimental 3.1. Study taxa 64 Species in three genera of Nyctaginaceae were included in this study: Acleisanthes, Mirabilis, and Selinocarpiis. Fragrance was collected from multiple individuals (sample sizes are given in Table 2) of a total of twenty species (Table I), the majority of which are found in the Chihuahuan Desert in southwestern North America. These taxa were chosen because of the prevalence of hawkmoth pollination and multiple evolutionary losses of this pollination system (Appendix D). Vouchers of all species from which volatiles were collected are housed in the herbarium at the University of Arizona (ARIZ; see Table 2 in Appendix A). 3.2. Volatile collection Fragrance was collected in the field using the dynamic head-space collection method [see Raguso and Pellmyr (1998) and references therein]. This method adapted for the field involves enclosing flowers of a living plant within a polyacetate bag (Reynolds*' oven bags) where volatile compounds emitted from the plant accumulate and are trapped in adsorbent cartridges through the use of a battery operated diaphragm pump (KNF Neuberger, Inc.). Glass cartridges were packed with 100 mg of the adsorbent Porapak®Q (80-100 mesh), and the pump pulled air over the flowers and into the adsorbent trap at a flow rate of ca. 250 ml/min. Simultaneous collections of ambient and vegetative volatiles were used to distinguish between truly floral compounds, compounds emitted from the vegetation, and ambient contaminants. Scent collections typically commenced at floral anthesis and continued for 12 hrs. Most of the species included in this study have nocturnal floral anthesis such that fragrance collection usually began at dusk and ended early the following morning. As all these species have flowers that last only one day, this 65 collection period generally encompassed the entire period during which a flower was open. For a few species, we were unable to collect floral fragrance in the field, and instead did so in the greenhouse using the same protocol as outlined above. In general one fragrance collection per individual plant was done, and the number of flowers included for each fragrance collection was noted. After fragrance collection, flowers were removed from the plants and weighed to obtain an average fr. wt per flower for each species. To distinguish floral from vegetative volatiles, fragrance collections of only vegetative material were necessary. The architecture of these plants makes it impossible to avoid trapping vegetative volatiles while also collecting volatiles from a number of flowers. Ideally the mass of vegetation from which fragrance was collected would be quantifled; however, this would have required destructive sampling of plants (in many cases fragrance was collected from an entire plant), which was not feasible. 3.3. Chemical and data analysis After the 12 hr collection period, cartridges were wrapped in aluminum foil and kept chilled. Cartridges were then eluted with 3 ml of hexane, and the eluate was stored frozen in glass vials. Before GC-MS analysis, samples were concentrated to 75 |il with N;. One |il (occasionally 3 |il) aliquots of each sample were injected into a Shimadzu GC- 17A equipped with a Shimadzu QP5000 quadrupole electron impact MS as a detector. All analyses were done using splitless injections on a polar GC column [diameter 0.25 mm, length 30 m. Film thickness 0.25 Jim (EC WAX); Alltech Associates, Inc.]. The carrier gas was helium with a flow rate of I ml/min and a split ratio of 12. 66 The injector temp, was 240°C and detector temp, was 260°C. The oven program for ail analyses began with injection at 60°C and a constant temp, for 3 min. The temp, then increased by 10°C per min until 260°C, where it was held for 7 min. The column pressure at injection was 60.6 kilo Pascals (kPa). The pressure then increased to 400 kPa for 48 s followed by a decrease to 60.6 kPa for 2 min (this pressure pulse was included for peak sharpening at the suggestion of L. Evanicke, Shimadzu Scientific Instruments, Inc.). Pressure was then increased at 3.9 kPa per min until 132.9 kPa, where it was held for 7 min. Compounds were tentatively identified using computerized mass spectral libraries [Wiley and NIST libraries (>120,000 mass spectra)]. The identity of many compounds was also verified using retention times of known standards (Table 2). To ensure that even small quantities of compounds were detected, we also looked for the presence of specific mass fragments on the chromatograms. For example, the nitrogen-bearing compounds indole and phenylacetonitrile were often present in very small amounts; by searching for the molecular ion (117 Da) at the appropriate retention times, we were able to detect trace amounts of these compounds. Quantification of compound amounts was achieved by integrating individual GC peak areas using Class-5000 software (Shimadzu Co.. 1993-1996). For each species these compound areas were averaged across individuals and divided by the total average amount of volatiles produced to yield the relative amount of each compound present in the floral and vegetative fragrance per 12 hr collection period (Table 2). For those volatiles mainly or solely released from flowers. 67 compound areas were standardized by adjusting for the number and mass of flowers sampled to yield an amount of each compound present per |i.g of floral mass (Table 2). 3.4. Comparison of intra versus interspecific variation We compared intraspecific variation in fragrance profiles to interspecific variation. To do this we used the relative compound amounts of both floral and vegetative volatiles for each individual (rarely multiple samples from the same individual were included, see Table 2) to calculate a dissimilarity matrix based on Euclidean distance (SPSS Inc., 1999). Variables were standardized to have a mean of zero and a standard deviation of one. We then compared the pair-wise % dissimilarities between individuals within a species to those between individuals among species using a Wilcoxon Signed Rank Test (SPSS Inc., 1999). A lower amount of dissimilarity between individuals within species compared to between individuals among species would suggest that intraspecific variation in fragrance profiles is less than among species variation. Acknowledgements We are grateful to T. Tully, A. Boyd, M. Manktelow, D. Heam, and A. Chetochine, without whose assistance all of the fragrance collections could not have been made. We thank Big Bend National Park, the Jornada LTER, and J. Duft for access to plants, and A. Le Due and M. Fishbein for providing seeds. For use of a GC-MS and injection of many of the samples we thankfully acknowledge E. Pichersky and J. D'Auria, Univ. of Michigan. This manuscript was greatly improved by J. S. Miller, who read an earlier version of the manuscript. The main source of funding for this research was provided by 68 the National Science Foundation DEB9806840 to L. A. McDade and R. A. Raguso. Further support was provided by small grants to R. Levin from The American Society of Plant Taxonomists, Sigma Xi, the Department of Ecology & Evolutionary Biology at the University of Arizona, and the University of Arizona Research Training Group in the Analysis of Biological Diversification (NSF DIR-9113362, BIR-9602246). This paper is part of a dissertation by R. Levin in partial fulfillment of the requirements for the degree of doctor of philosophy. University of Arizona. References Azuma, H., Toyota, M., Asakawa, Y.. Yamaoka, R., Garcia-Franco, J. G., Dieringer, G.. Thien, L. B., Kawano, S., 1997. Chemical divergence in floral scents of Magnolia and allied genera (Magnoliaceae). Plant Species Biol. 12, 69-83. Barkman, T. J., Beaman, J. H., Gage, D. A., 1997. Floral fragrance variation in Cypripedium: Implications for evolutionary and ecological studies. Phytochemistry 44, 875-882. Barnes, J. L., 1996. Reproductive ecology, population genetics, and clonal distribution of the narrow endemic: Mirabilis macfarlanei (Nyctaginaceae). Master's thesis, Utah State University, Logan. Barthlott, W., Porembski, S., Kluge, M., Hopke, J., Schmidt, L., 1997. Selenicereiis wittii {CactaceaeY an epiphyte adapted to Amazonian Igapo inundation forests. PI. Syst. Evol. 206, 175-185. 69 Bene, F., 1945. The role of learning in the feeding behavior of Black-chinned Hummingbirds. The Condor 47, 3-22. Berenbaum, M., Seigler, D., 1992. Biochemicals: engineering problems for natural selection. In: Roitberg, B. D., Isman, M. B. (Eds.), Insect Chemical Ecology: An Evolutionary Approach. Chapman & Hall, New York, pp. 89-121. Bestmann, H. J., Winkler, L., von Helversen, O., 1997. Headspace analysis of volatile flower scent constituents of bat-pollinated plants. Phytochemistry 46, 1169-1172. Borg-Karlson, A.-K., Valterova, I., Nilsson, L. A., 1994. Volatile compounds from flowers of six species in the family Apiaceae: bouquets for different pollinators? Phytochemistry 35, 111-119. Brantjes, N. B. M., 1978. Sensory responses to flowers in night-flying moths. In: Richards, A. J. (Ed.), The Pollination of Flowers by Insects. Academic Press. London, pp. 13-19. Croft, K. P. C., Jiittner, F., Slusarenko, A. J., 1993. Volatile products of the lipoxygenase pathway evolved from Phaseolus vulgaris (L.) leaves inoculated with Pseudomonas syringae pv phaseolicola. Plant Physiol. 101, 13-24. Cruden, R. W., 1970. Hawkmoth pollination of Mirabilis (Nyctaginaceae). Bull. Torr. Bot. Club 97, 89-91. Daly, K., Smith, B. H., 2000. Associative olfactory learning in the moth Manduca sexta. J. Exper. Biol. 203, 2025-2038. Dobson, H. E. M., 1994. Floral volatiles in insect biology. In: Bemays, E. A. (Ed.), Insect-Plant Interactions, Vol. V. CRC Press, Boca Raton, pp. 47-81. 70 Dobson, H. E. M., Arroyo, J., Bergstrom, G., Groth, I., 1997. Interspecific variation in floral fragrances within the genus Narcissus (Amaryllidaceae). Biochem. Syst. Ecol. 25, 685-706. Dudareva, N., Pichersky, E., 2000. Biochemical and molecular genetic aspects of floral scents. Plant Physiol. 122, 627-633. Faegri, K., van der Pijl, L., 1979. The Principles of Pollination Ecology, 3"* ed. Pergamon Press, Oxford. Fowler, B. A., Turner, B. L., 1977. Taxonomy of Selinocarpiis vtnd Ammocodon (Nyctaginaceae). Phytologia 37, 177-208. Grant, V., 1983. The systematic and geographical distribution of hawkmoth flowers in the temperate North American flora. Bot. Gaz. 144,439-449. Grant, V., Grant, K. A., 1983. Hawkmoth pollination of Mirabilis longiflorci (Nyctaginaceae). Proc. Nat. Acad. Sci. 80, 1298-1299. Grison, L., Edwards, A. A., Hossaert-McKey, M., 1999. Interspecies variation in floral fragrances emitted by tropical Ficus species. Phytochemistry 52, 1293-1299. Haber. W. A., 1984. Pollination by deceit in a mass-flowering tropical tree Phimeria rubra L. (Apocynaceae). Biotropica 16, 269-275. Haber, W. A., Frankie, G. W., 1989. A tropical hawkmoth community: Costa Rican dry forest Sphingidae. Biotropica 21, 155-172. Hatanaka, A., Kajiwara, T., Sekiya, J., 1986. Fatty acid hydroperoxidase lyase in plant tissues: volatile aldehyde formation from linoleic and linolenic acid. In: Parliament, 71 T. H., Croteau, R. (Eds.), Biogeneration of Aromas. American Chemical Society, Washington, DC, pp.167-175. Heath, R. R., Manukian, A., 1994. An automated system for use in collecting volatile chemicals released from plants. J. Chem. Ecol. 20, 593-607. Hill, R. K.. Rhee, S.-W., Leete, E., McGaw, B. A., 1980. Stereospecificity of enzymatic dehydrogenation during tiglate biosynthesis. J. Am. Chem. Soc. 102, 7344-7348. Hodges, S. A., 1995. The influence of nectar production on hawkmoth behavior, self- pollination, and seed production in Mirabilis multiflora (Nyctaginaceae). Am. J. Bot. 82, 197-204. Jakobsen, H. B., Olsen, C. E., 1994. Influence of climatic factors on emission of flower volatiles in situ. Planta 192, 365-371. Kaiser, R., 1993. The Scent of Orchids: Olfactory and chemical investigations. Elsevier. Amsterdam. Kaiser, R., Tollsten, L., 1995. An introduction to the scent of cacti. Flav. Fragr. J. 10, 153-164. Knudsen, J. T., Tollsten, L., 1993. Trends in floral scent chemistry in pollination syndromes: floral scent composition in moth-pollinated taxa. Bot. J. Linn. Soc. 113, 263-284. Knudsen, J. T., Tollsten, L., 1995. Floral scent in bat-pollinated plants: a case of convergent evolution. Bot. J. Linn. Soc. 119, 45-57. Knudsen, J. T., Tollsten, L., Bergstrom, L. G., 1993. Floral scents—a checklist of volatile compounds isolated by head-space techniques. Phytochemistry 33, 253-280. Martinez Del Rfo, C., Biirquez, A., 1986. Nectar production and temperature dependent pollination in Mirabilis jalapa L. Biotropica 18, 28-31. Miyake, T., Yamaoka, R., Yahara, T., 1998. Floral scents of hawkmoth-pollinated flowers in Japan. J. Plant Res. Ill, 199-205. Mullin, C. A., Alfatafta, A. A., Harman, J. L., Serino, A. A., Everett, S. L., 1991. Com rootworm feeding on sunflower and other Compositae: influence of floral terpenoid and phenolic factors. In: Hedin, P. A. (Ed.), Naturally Occurring Pest Bioregulators. American Chemical Society, Washington, DC, pp. 278-292. Nilsson, L. A., Jonsson, L., Rason, L., Randrianjohany, E., 1985. Monophily and pollination mechanisms in Angraecum arachnites Schltr. (Orchidaceae) in a guild of long-tongued hawk-moths (Sphingidae) in Madagascar. Biol. J. Linn. Soc. 26. 1-19. Pilz, G. E., 1978. Systematics of Mirabilis subgenus Quamoclidion (Nyctaginaceae). Madrono 25, 113-132. Radwanski, E. R., Last, R. L., 1995. Tryptophan biosynthesis and metabolism: biochemical and molecular genetics. The Plant Cell 7, 921-934. Raguso, R. A., 2001. Floral scent, olfaction and scent-driven foraging behavior. In: Chittka, L., Thomson, J. (Eds.), Pollinator Behaviour and its Implications for Plant Evolution. Cambridge Univ. Press, in press. Raguso, R. A., Pellmyr, O., 1998. Dynamic headspace analysis of floral volatiles; a comparison of methods. Oikos 81, 238-254. 73 Raguso, R. A., Pichersky, E., 1995. Floral volatiles from Clarkia brewerii and C concinmi {Onagraceae): recent evolution of floral scent and moth pollination. PI. Syst. Evol. 194, 55-67. Raguso, R. A., Willis, M. W., 2000. The importance of olfactory and visual cues in nectar foraging by nocturnal hawkmoths. In; Boggs, C. L., Watt, W. B., Ehrlich, P. R. (Eds.), Proceedings of the Third Symposium on Butterfly Ecology and Evolution. University of Chicago Press, in press. Shimadzu Co., 1993-1996. Class-5000, version 2.0.0. Shimadzu chemical laboratory analysis system & software. Shimadzu Co. Spellenberg, R., Delson, R. K., 1977. Aspects of reproduction in Chihuahuan Desert Nyctaginaceae. In: Wauer, R. H., Riskind, D. H. (Eds.), Transactions of the symposium on the biological resources of the Chihuahuan Desert region United States and Mexico. U. S. Department of the Interior, pp. 273-287. SPSS Inc., 1999. SPSS for Windows, version 10.0.5. SPSS Inc., Chicago. Tinbergen, N., 1958. Curious naturalists. Doubleday & Co., Garden City, NY. Van Riper, W., 1960. Does a hummingbird find its way to nectar through its sense of smell? Sci. Amer. 202, 157-166. Wang, J., Pichersky, E., 1999. Identification of specific residues involved in substrate discrimination in two plant O-Methyltransferases. Arch. Biochem. Biophys. 368. 172- 180. Wink, M., 1997. Special nitrogen metabolism. In: Dey, P. M., Harbome, J. B. (Eds.), Plant Biochemistry. Academic Press, San Diego, pp. 439-486. Table I. Species included in this study, and what is known about pollinators. Except where indicated, all data are from personal observations of R. Levin, including observations of floral visitation, examination of visitors for pollen on their bodies, and micro.scopic study of stigmas for the presence of moth scales. In addition to nocturnal lloral anthesis, lloral morphological traits taken to indicate hawkmoth pollination included a perianth > 2 cm long; a perianth < 2 cm long suggested nociuid moth pollination. Species Likely pollinators Evidcni-c Genus Acleisunthes A. Grii> A. aculifolia Slundl. Hawkinoths Floral morphology A. crassifolia A. Gray Hawkmolhs Pers. observations A. Umgiflora A. Gray Huwkinolhs Pers. observations; Spellenberg and Delson, 1977 A. ohtusa (Choisy) Standi. Hawkmolhs Pers. observations A. wrighlii (A. Gray) Bcnth. & Hook. Hawkmolhs I'ers. observations Genus Mirahilis L. M. aUpes (S. Wal.st)n) Pi!/. Nociuid molhs, hawkmolhs Pers. observations M. higelovii A. Gray Nociuid molhs Floral morphology M. greenei S, Walson Hawkmolhs Floral morphology M. jalapa L. Hawkmolhs Marline/del Rio and Biirquc/, 1986 M. longijloru L. Hawkmolhs Grant and Grant, I9K3 M. macfiirlanei Ct)nstanL-e & Rollins Bees Barnes, 1996 Spccius Likely pollinators Kvidence M. multijhmi (Torrcy) A. Griiy Hawknu)lhs Pers. observations; Crudcn, 1970; Hodges, 1995 M. pudica Burneby Noctuid molhs I'crs. observations M. trijhira Bciuh. Hummingbirds Pilz, 1978 Genus Selinocarpus A. Gray 5. mgustifolius Torrcy NoLiuid moths Floral morphology S. chenopodioides A. Gray Noctuid moths, hawkmoths, long-tongued flics Pers. observations S. lanceoltiiHX Woolon Hawkmoths Floral morphology S. parvifoliiis (Torrcy) Slandl. Hawkmoths Pers. observations S. piirpusiunus Hcimcrl Hawkmoths Pers. observations; Fowler and Turner, 1977 S. unduUitus Fowler & Turner Noctuid moths Floral morphology -J Ui Table 2. Average relative amount ol each compound as a percentage ol the total amount of volatiles found in the lloral and vegetative fragrance of each of the twenty species. Total amount is an average of the total unit area (as integrateil peak area; see lixperimental) of all lloral and vegetative volatiles produced per 12 hrs, divided by 10'' for ease of pre.sentation. '7t Floral is the jiercentage of the total volatiles released by llowcrs. 'Ilie total number of both lloral and vegetative compounds and the total number of compounds produced solely or mainly from flowers are presented. 'Ilie amount of volatiles released from llowers (an average of the total unit area) has been adjusted by lloral mass to give a total amount of lloral volatiles relea.sed per of lloral tissue. Compounds within classcs are li.sted according to retention time, and for each compound class its percentage of the total is pre.sented. Tracc = <0.01 % of the total .sample. The sample si/e (n) is the number of individuals from which fragrance was collected; unless otherwise noted, only one fragrancc collcclion per individual was included here. Entries in bold idcniify those compounds that were .solely or mainly lloral. Mass fragments for unknowns are listed with the putative molecular ion first (if known or inferred), followed by the base peak and then other mass fragments in order of decreasing abundance. Abbreviations; Acleisamhes aailifolia (Aac), A. crassifoUu (Acr), A. Umfiijlora (Alo), A. obtiim (A()b), A. wrightii (Awr), Mirtihilis aUpes (Mai), M. hif^elovii (Mbi), M. greenei (Mgr), M.jaUipa (Mja), M. lonnijlorn (Mlo), M. mucfurlanei (Mmf), M. imiliijloiv (Mmu), M. pudica (Mpu), M. triflora (Mtr), Selinocurpus angustifoUus (San), S. chenopodioides (Sch), S. Umceolatus (Sla), S. ptmnfolius (Spa), S. purpusianus (Spu), S. undnlatus (Sun). Asterisks indicate (hose compounds for which we have retention times from standards in addition (o agreement with (he muss spectral library. C(>inpuunil KT Aac Acr Alo Aoti Awr Mai Mtii Mgr Mja Mlu Mint Mniu Mpu Mir San .Sch Sla Spa Spu Sun 11=1' ii=.S n=4 n=5 11=5 n=7 n=l' n=6 n=l n=2 n=l' 11=5 n=5 11=2- 11=3 11=9 n=4 n=4 n=4 n=4 Total uniiHinl 611.1 .i6.77 •S.'i.i.'; 99 H7 1.15.« 9.84 63.73 0 32 666 7.13 0 19 4.87 13 92 0.96 3.30 174.3 61 44 415.4 333.5 1.85 % I'loral IK.IK W5^ '>2 2.1 69.7.1 17.98 94 61 95.92 1.3 71 92.05 94 22 54.95 99.47 100.0 73.99 523 82.17 83.87 64.14 96.65 40.04 Total no. cmpds 60 62 47 7.1 lOK 29 41 5 21 27 6 .10 32 24 36 63 51 63 81 39 # Floral cmpds 1.1 2.5 55 3.1 23 34 1 15 22 2 29 .12 17 8 50 47 43 77 17 Amount floral/ 515.5 264'J 2.1.1.7 102.6 10 4 8.43 75.78 0.(H 3.84 5.59 0 13 2.84 9.75 3.52 0.90 2.16.5 43.59 352.8 3.14.6 3.42 MK floral mass Monotcrpcnoids/lrrcKular tcrpcncs u-Plnene* 2.5 0.32 Trace Tracc 0..12 0.06 — — Trace 0.29 — 0.40 Trace Tracc 0.57 Campticne* 2,9 0.67 — Tracc I.IS OJI — — — — 0.25 — 4J9 — — 0.54 (J-I'inene* 4.0 — Tracc Tracc 1.1.1 0.02 0.0.3 0.45 0.S6 0.06 Tracc 0 14 — — — Coiiipoiinil RT Aac Aa Alii ADII Awr Mai Mhi Mgr Mja Mlii Mini Minu Mpii Mii .S;ui Sch Sla Spa Spu Sun 11=1' 11=5 n=4 ti=.S 11=5 11=7 11=1' ii=ft 11=1 11=2 n=l* ii=5 n=5 11=2" ii=.t ii='> n=-l ii=4 n=4 11=4 SabiiiL-iiL'* 4.1 0 15 Tract 6 50 Tracc 0.29 OOK Trace 2-Carcne* 44 15.10 I'hcllandrcni;* 5.1 065 Myrccnc' 5 1 109 0.59 Tract- 1.26 0.11 OJI 0.08 05H 0.30 2 KO 3.09 0 6.1 0.73 1.27 u-Tcrpincnc* 5.5 Trace 0.04 0.23 Liinnncnc* 5.H 0.54 0.32 0.24 15.26 2 29 0 14 3 K9 43JS 1.94 0.88 4.76 .114 0 25 6.87 0 5« 0.07 9.85 (l-Tcrpinene 5.9 9.91 I 19 cis-p-Ocinicni:* 6.5 0.16 U.I8 Trui'c 0.13 U05 I.OS 0.13 IJI 0.83 0.02 4.6H 0.10 1.52 0..1.1 1.05 1.41 lrans-|i- 6.9 1.5U 10.08 U4.1 SJ9 1)82 60.79 7.21 38.10 3.44 1.72 2.24 IJO 12 14 0 76 5.87 0.76 57.81 3.45 Ociiiiene* a-Tciplnnleiie* 7..1 O.Wi Trace 0.28 0.06 0.01 0.24 Trace 0.08 0.94 0.15 Trace 6-Mclhyl-5- «..1 6.16 4.99 1J4 4.16 052 0.76 0.58 3.67 072 0.10 553 heplen-2-one a-Pinene 8.6, 0.14 epoxides 9.2 2.7-MelhyI 8.6 003 Ocladlene Alli)-Ocinicnu* H.H Trace 0.02 Trace 0.02 Trace 0.52 0.16 Cincrone/ 9.3. __ . 0.15 Pipcrilcnone 9.9 Furanoid linalool 9.8, 003 7.18 _ _ 4.16 2.28 oxides* 10.3 Menlhalricne 9.9 0.61 Trace 0.17 Sunlenc* 10.7 Trace (Xiadiene III 0.47 0.07 Compound RT A;ic Acr Alo Aob Avvr Mai Mbi Mgr Mja MIo Minf Miiui Mpu Mir San Sch Sla Spa Spu Sun 11=1'' n^.*) n=4 11=5 n=.') n=7 n^l' n=(< n-l n=l' n=5 n=5 n=.^ n=-l n--l n=-4 n=4 Linalool* 113 0 19 (U2 0 51 O.HO 2 52 0.66 0.38 _ _ , 0 53 2.0U «55 1 K5 0.V2 3.93 1.21 OJO 3 72 Isocilroncllcn 11,7 OJl 0.76 •J-Tcrpincol* 11.9 _ 7.9.3 cis-Cilral* 12 8 Tracc 0.03 Trace 0.08 Liinonenc 13- 0.08 Tracc 0.80 0.91 0.07 diepuxidcs 13.9 Eutarvime* 13.4 O.ll (O-Vcrlicnone'" 13.5 0.73 Irons-Cilral* 13.6 0.06 Pyranoid linalDol 13.7, 10.60 0.68 20.04 1.70 oxides* 13.9 Nerol* 14.3 2.72 0.03 0..I9 0J5 3.59 8 27 7.02 2.81 a-lononc* 14.5 0.08 Carvcol* 14.6 0.03 0.05 0.05 0.09 0.04 Thujol 164 0.15 . 0.05 _ Unkmtwns with 150 as mol. iim: 150.135,43. 7.8 . Trace Trace 0.02 67, 55 l.50,.39,95,77, 11.1 0.08 91 150,39,41,91, 17.2 0 16 _ 0J6 79,107,1.35 1.50,41,107, 19.2 0.20 5.3, 79,1.3.5,91 Other unknowns 1..37 0.10 0.09 1.34 0.38 1.41 0.02 3 12 0.46 0.05 Trace 0 72 Trace 1.65 OJl 1.43 10.83 Totals 8.78 11.76 7.44 .18.65 7.42 84.41 K.07 4..16 43 11 .35.25 43.87 5 23 .10.56 1903 .17 71 9.73 33.21 5.57 63.83 52.02 >1 C'onipouiul RT Aac Acr Alo Aoh Aur Mai Mhi Mgr Mja MIo Mini Miiui Mpu Mtr San Sch Sla Spa Spii Sun n=l' n=.S n=4 n=5 n=.'> n=7 n=l* n=6 n=l n=2 n=l' n=^ ii=.^ n=2'' n=.1 n='> n--4 n=-l n=-l n=-l Si'sqiiilcrpvnoid.s -».«-iliinclhyl- 7 8 0 84 0.47 0.20 0.68 Trarc 5.0.1 23.84 1.1.71 0 81 4.41 11.60 5.18 12.25 2 91 15 21 I 2y 10.37 0 51 0.02 2.01 nona-1,1,7- iricnc* Olher likely 105, 0.14 0.02 0.06 0.21 0.74 — 0.21 1.04 091 Tracc 0.14 hi)nu)ler|H.'nc!i 14.1, 14.4 a-Cubcbcne* o.y 0.02 Trace 6-Elenicne' 10.2 0.05 — 0.04 Trace 0.12 — 0.15 0.44 0.08 — («-Copacne' 10.5 0.16 0.0.1 Tract- 0 12 OJI — — P-Bourhtmene* 10.8, 0.74 0 51 242 Tracc — — I0'> in-Curjunenc III — 0.02 — Trace — |l-Cubfbciie* 112 0.42 O.UI 0.01 0.01 — — — — u-Bergonii)lene* 116 — 0.04 1.42 — 0.29 — — 0.65 — 0.20 — (l-Elcniene* 11,7 6.86 0.04 12 74 0.10 0.04 — __ — — — Caryophyllenc* 12.0 7..16 3.52 1.52 4.15 24.85 0.93 0.40 9.84 071 2.19 0.41 •— Choinigrene 12.1 Trace — ,— — — Tracc — 1.07 — Trace 0.08 0.47 P'Selinene 12.5 0.61 0.17 Tracc — P-Farncaenc* 12.7, I.Ofi Tnii'K 0.26 IJ6 0.05 0.09 — — — Tracc 0.12 — — 11 1 u-Huinulene* 12.y 4.11 0.22 l.ll 11.71 4J8 Tracc — 2.17 0.14 0.14 — — P-CoUinene* M.I 0.07 Gennacrene I)* 114, 12.27 Tract! 2J4 I..57 Tracc 4 11 0.01 1.26 0.18 — — — — 1.1.7 -J v£) Coinpoiinil RT Aac Acr Alo Aob Awr Mai Mhi Mgt Mja Mlii Minf Mmu Mpu Mir San Sell Sla Spa Spu Sun n=r 11=^ n=4 n=5 n=.^ n=7 ii=l' n=() n=l n=2 n=l' n=5 n=fi n=2" n=.^ n='j n=-4 n=4 n=4 n=-4 (iFarnL'senc' 1.1 S 2.00 0.52 18.95 ()()7 0.42 0.02 .17.03 4J7 0.97 0.5.1 0.09 2.6.1 0.52 1.1 H 8-Giiaicnc* 1.1.5 7.72 Juniponc 1.1.5 0.70 2'W 0.49 Y-Cailincne* 14.0 «.()! Trace 6.K3 0.01 0.06 u-Muurolene* 14,1 0.5« . 0.22 Caluincncnu* 14.H 0.78 0.16 Gcruiiyl accloni:* 14.9 0.4.5 0 10 _ 0.45 0J5 1 2« 1 .18 0.6.1 1.02 0 .16 0.08 2.46 Caryophyllcne 16.2, 6.18 0.12 Tracc 3.75 2J9 0.23 oxides* I6.,5 Nerolidol* 16.5- 2.55 0.14 80.46 1.06 028 0.05 0.21 0.14 Trace 0.01 J6 9 1,5-Cytlo- 16.7 0.06 undecudienu' 12-Oxabicycli> 17.0 057 0.01 0.98 0.58 dodecadienc'' Geniiacrone 17.3 005 Carotol 17.6 0.07 SpalhulcnnI 17.8 Oil Aromadeiidrene 18.0 0.22 epoxide 6-Cudini)l 18.3 9.14 Gualol 18.3 0.22 _ Arislolenc 18.4 0..15 00 o Compound KT Aac Acr Alo Aoh Awr Mai Mhi Mgr Ntju MIo Miiif Mnui Mpii Mir San Sch Sla Spa Spu Sun n=l' n=.'i n=4 n=5 n=5 n=7 n=l' n-(i n=l n=2 n=r n=5 n=5 n=2'' ii=9 n=4 n--4 n=-l n=j (i,|M:'(i Furncsol* IH.'J- OW 0.19 0.95 .1.02 3.55 19,8 Neroliitol 21.6 0 12 cpoxyacclule Unkniiwn sesquilerpeiumh 204,41,107. 13.9 0.05 Trace 0.09 0.03 2 34 0_W 81.91,67.121 222,43,161, 15.4, 1.27 0.08 0.08 0.15 Tracc 0 70 O.-M 0.03 .55,81,91,207, 15.8 105,119 222,81,41,55, 15.9 Trace OJI 1.02 _ 1.55 0.12 0.04 93.121,69, 1.36. 109 222,41,81,.5.5, 15.9 0.66 _ 93.69,109, 161, 121,207 220,43,79,93, 16.5 2.55 .5.5,69,109.121 222,81,43, 17.1 0 41 0.44 _ 161.105,71, 91,55,119,207 222,41,119. 17.2 0.33 _ I61..55.179. 105.81.95.204 Other unknowns 5.10 Trace 2.60 Trace Total % 71.14 7.08 82.90 31.92 79.13 658 5655 13.71 38.75 1069 ll.W) 6.03 13.78 6 80 40 13 4 66 1542 2 04 2.86 5.77 Ci)iiipi)uiul RT Aac Acr Alo Aoh Avvr Mai Mhi Mgr Mja MIo Mini Mnm Mpii Mir San Sell Sla Spa Spu Sun n=l' n=4 n=5 n=5 n=7 n=l' n=h n=l n=J n=l' ii=5 n^J" n=.^ n=9 n=-l n=-l n=-> n=4 Aliphatic ('ompiiunds 2-lli:xenal* ft 2 0.05 0.24 0.4-1 0 70 L-is-.l-lkxenyl 1}) 2.lh 9.52 .1.27 2 (X) 5.7.1 .1.1.1 29.00 52 41 144 6.49 .11.22 5.51 ll .fth '>.28 7 71 15 ()4 I OK ll .iy ucclulc* lraiis-2-Hcxeiiyl 8 2 0.01 0.04 Tract- 0.1)1 0.4.1 0 7ft acclulc* 1-Hexnnol* 8.4 0.23 0..5I 1.06 0.05 Tracc 0.1.1 11.15 2165 1.6.1 0.34 0.93 3.26 0.74 0.08 1.91 l-Oclcn-.1i)l «.«> _ 0.25 0.37 acclalc* cis-3-Hexcn- 8.9 022 5.08 103 0.86 091 168 321 18 15 0.79 094 19.19 12.89 12.95 17.83 1 44 1 4.1 7.49 1.1.96 0.14 2.90 l-ol* i:is-.1-Hcxcnyl 9.0 0.01 0.29 0.46 prDpiunale* Bulanoic aciil, 9.1. 0.13 0.20 0 02 0.04 US 1.55 12.64 0.05 3.20 4.85 0.16 3-hL'Xfnyl L'slcr 10.1 lrans-2-Hexcn- 9 4 _ 0.03 Trace 0 19 0.26 0.76 0.21 0 17 1-ol» 7-(Xicn-4-ol 9.9 OJI 0.13 Trace 6.64 016 030 025 7.74 1.68 0.62 055 cis-3-Hexenyl 2- 10.3. 0.09 0.10 Trace 0()4 0.11 0.43 1J4 2.74 11.02 0.12 2.19 5.16 O il methyl huianoaie 10.5 2-Nonenal I I.I Trace 0.53 __ _ 177 Tracc 0 01 168 fis-3-Hexenyl 3- 11.4 Trace 0.20 0.17 methyl hutartoate Bulanoic acid, 11.8 2.07 . . . 2-hexcnyl ester 2,6-NonaUienal 11.8 _ 5 48 00 I-J C'oinpDuiul RT A;ic Acr Ali> Aob Awr Mai Mhi Mpr Mja Mli> Miiif Mimi Mpii Mu .S:iii Sell Sla Spa Spu Sun 11=1' 11=5 n=-l 11=5 11=7 11=1' 11=6 11=1 11=1' =5 11=5 11=2" 11=1 n=y 11=4 11=4 11=4 11=4 llexaimic acid. 12 1. (M)2 Tnicc 0.92 2.26 Vhexenyl csler 12 7 iHolnilyl llghile 12 2 0.26 ci.s-.l-llcxcnyl 128 0.02 Tracc 0.05 0..32 0.29 5.22 .M.ll 0.13 liglulc lleptanoic acid. 1.1.4 Tracc 0.25 0.38 •l-hcxenyl cslcr Mcxyl huluniialu 15..3 0.14 0.07 4.04 cis-Jasnuine* 16.1 0.14 16.83 Tniic Tracv 0.84 IJ2 4.09 0.45 0J7 24.71 Benzyl Tigluli 17,6 0.49 0.09 Melhyl cuprule 18 6 0.18 Unkmiwns Trace 0.14 0 16 0.29 ()(H 0.22 0.53 0 54 Tnicc 0 49 1.20 0 39 TdINI % 2.K4 .32.40 7.42 11.88 8.25 .5.19 .34 11 81.93 .5.97 5 28 41 (H 25.31 49.02 .56.50 13.31 18.37 31.60 7604 27.80 28.15 Bcnzenuidii Anisole' 8 4 0.13 0.17 0 13 2J8 340 0.78 Benzyl melhyl 9.1 0.48 elher* Olher benzyl 10.8, 0.47 elhers 13 3 Ben/aldehyde* III 0.18 1.67 0.54 0.91 0J7 2J2 0.30 5.25 Melhyl 12.3 0.18 44.44 0.62 8.05 Tracc 0.44 44.53 0.75 I (M) 35.90 1.93 0_36 Trace 1.16 bcnzoale* Phenyl 12 5 0.04 U.05 1.74 1.67 0.09 0.24 0.05 acclaldehyde Elhyl benziiaie* 12.9 0.16 0.09 0.07 0.05 Ci)ii)pi)imd RT Aac Acr Alo Aob Awr Mai Mhi Myr Mja Mlo Minf Miiui Mpii Mir .San .Sch .Sla •Spa .Spii .Sun n=l' n=5 n=4 n=5 n=5 n=7 11=1' n=(i n=l n=2 11=1* n=5 n=5 n=2' n=3 n=y n=4 n=4 n=4 n=4 Salicylalilehyilc 13 0 0.01 Trace — 0 17 0.65 Trace Trace — — — — — — — 13.6 0 10 1 20 Hen/yl acelato* — — 0.07 0.92 0.09 5-31 — - 1.27 0.33 0.63 0.45 — 0.03 0.03 Propyl iK-n/oalc* 13.9 — — — — — — 0.03 -- — — Methyl M 2 2,54 0.19 0.18 0.8.3 0.06 o.so 0.86 0.57 003 2.81 0.21 8.14 1 59 2.80 0.92 0.23 0.05 1 20 — salicylulc* Isohuiyl 14.3 0.12 Trace 0.16 Trace 0.02 0.11 — — — bun/oale* Hhcnclhyl 146 Trace 0.16 — 0.18 Trace Trace — — — — acelale* Btilyl hcnzoalC 15.1 0.09 —. 0.03 0.04 — — — 0J3 0.45 0.08 0.05 — Benzyl 15 2 Trace 0.02 0.27 0.06 Trace — — — — — — bulanuule* Benzyl alcohiil* 15.3 0.15 1.02 0.03 — 1.67 3.48 1.S4 0.17 4.14 2.22 1.28 0.72 0.10 0.09 4.13 Benzyl 3- 15.4 0.01 0.20 — 0.21 0.03 — — — — — — — nielhylbulanoate Isoprnpyl 15 5 0.05 salicylule 2-Hhenyl- 15.3 Trace 0.07 0.12 Trace 0.43 0.93 0.26 0.U 0.27 0.07 0.47 — — — elhanol* Heniyl benzoale* 15.7 U.42 0.08 — 0.02 0.16 — — — 0.25 — 0.10 0.12 Hexyl benzoale* 163 — 0.01 — — — — Trace — — 0.62 — 0.04 0.01 Tolyl/Allyl 16.9 0.21 Trace Trace 0.15 Trace 0.21 0.12 — — — — — — — benzoale Cyclohexyl 17.2 0J6 0.04 Trace 0.13 IJI Trace 0.70 0.07 0.25 — — — — — benzoale cis-3-Hexenyl 17.6- 1.72 2.21 0.0M 0.23 0.2S 0.14 0.27 7.87 0.28 2 14 3.09 1.13 3.73 1.88 — — benzoale* 18.0 Coiiiptniiul KT A;ic Acr Alo Aiih Awr Mul Mhi Mjir Mja Ml» Mini Miiui Mpu Mir Sun Sch Sla Spa Spu Sun n=l'' n=5 n=-4 ii=5 n=> n=7 n=l' n=h ii=l n=3 n-l' n=5 n=5 n=2'' ii=.^ ii==9 ii=-> n=4 n=-i n=-> I'hcneihyl 17 K _ _ _ _ 0.83 _ _ . IK-II/OUIC Benzyl 22 1 0.21 Tracc 4.03 0 21 3.63 _ ^ 1.43 0.14 I..Vt 7.41 0J2 bon/Dalu* Unknown: lO.S, I.S.2 0.13 0.12 77, 41,.SI, 67 Olher unknowns Trace 0.05 Tracc Total % 4.'jy 4».26 1.41 17.07 l.(W 3 .il 1.22 12.17 2 40 .V48 6! 7.S 5 12 17.6K K.K5 .'>7.67 64.'i .'i.71 .162 12.22 Phcnylprupanuids Eugcnol* 18.2 2.53 _ _ 0.05 Eugcnol uccialc/ I'J.l 0.08 Isucugcnol* Isoeugenol* I')}) 6.55 2.91 Trace Vanillin* 21.7 __ 0.20 Vanillyl 22.5 0.19 inclhylkeiune Total % y.07 3..18 0.05 Ijiclones 4,4-Uiim:lhyl 122 0.05 hul-2-cnoli(lc Buiyroloclone 12.4 0.09 Y-Vinyl-Y- 12.9 0.41 1.13 valerolaclunu Y-Caprolaclone I.1..1 Trace 0.56 0_VI ft-Oclalnclone 16.2 0.62 0.26 oo C'oinpounii RT Aac Acr Alo Aob A»r Mai Mhi Mgr Mja Mlo Mini Mmu Mpu Mir San Sell Sla Spa Spu Sun 11=1' 11=5 n=4 11=5 n=5 n=7 ii=l' 11=6 n=l n=2 ii=l' n=5 ii=5 n=2" n=.1 ii=y ii=-l n=4 n=4 n=4 S-Nimalacionu 18-J U.IU 0.26 _ _ 7-l)ctcn-5-oliilc I'/O 2.95 9.87 9.44 Total It 2 % 0.41 Trace 11 15 10 .v; 0 05 1.1.1 Nilrugcn-hcaring C'onipuunds Melhyl 14.2 8.89 nicotinate* Phcnylaccio- 15.9 0.12 Trace 0 ()6 0.05 0.04 1.52 0.54 2.IK 0.22 0.06 0.42 nilrile* Melhyl 18.9 0.10 .1IJI onlhranilalu* Indole* 20.8 0.48 0.26 0.45 OJW 0.02 15.07 1.68 0.09 0.03 Trace TuIHI % 0 12 0 49 0.06 0.36 0.49 0.«) 0 ()ft 46..18 I 68 1..52 9.52 2.18 0 25 0.06 0.42 Uni(nuwn!> 0.10 0.36 0.12 0.24 1.77 0.30 "Only one plant was avuiluhic, but multiple collections were made of thi.s plant. ''Multiple collections from two plants were made during both the day and night for this diurnal species. "1,5-Cyclo-undecadiene, 8,K-dimcthyl-9-mclhylenc; ** 12-Oxabicyclo 9.1.0 dodeca-3,7-dicne, 1,5,5,8-tclramclhyI-, 1R- ([email protected],7E,IIR@). 00 ON 87 APPENDIX D RELATIONSHIPS AMONG FRAGRANCE, PHYLOGENY. AND POLLINATION IN THREE GENERA OF NYCTAGINACEAE 88 Relationships among fragrance, phylogeny, and pollination in three genera of Nyctaginaceae Rachel A. Levin\ Lucinda A. McDade\ and Robert A. Raguso'' ^Department of Ecology & Evolutionary Biology Biosciences West, Rm. 310 University of Arizona Tucson, AZ 85721. USA Tel. 520-621-8220 FAX 520-621-9190 ''Department of Biology Coker Life Sciences Building University of South Carolina Columbia SC 29208, USA E-mail: [email protected] 89 ABSTRACT. Floral fragrances have long been thought to attract specific classes of pollinators. We tested the relationship between hawkmoth pollination and floral and vegetative fragrances within three closely related Nyctaginaceae genera. A robust phylogeny previously inferred from molecular sequence data allowed us to conduct phylogenetically independent contrasts to evaluate the effect of fragrance composition on type of pollinator. Further, we examined relationships between fragrance and phylogeny by rigorously exploring methods for coding and comparing the phylogenetic information available in fragrance data with the phylogenetic signal in the molecular sequence data. The main relationship between fragrance and class of pollinators was the significant correlation of presence of nitrogen-bearing compounds and moth pollination. However, for the most part we found that the ecological effect of type of pollinator was less important than the effect of phylogenetic relationships on determining fragrance composition in this group of plants. Experimentation with various fragrance coding methods did not yield a definitive answer as to the method that preserves the most phylogenetic signal. Rather, both quantitative and qualitative coding were informative for certain compounds and biosynthetic pathways, especially those that were relatively rare. When the phylogenetic signal in the molecular sequence data was compared to that in various fragrance data sets, we found them to be highly incongruent. However, close examination of consistency and retention indices of various fragrance compounds and pathways revealed that, whereas the majority of compounds were homoplastic, a number of compounds and pathways were phylogenetically informative. Of these, many provided support for sister taxa relationships. Therefore, we suggest that although fragrance data as a whole may not be useful in phylogeny reconstruction, these data can certainly provide additional support for clades reconstructed with other types of characters. 90 Fragrances are emitted by both vegetative and reproductive structures in plants belonging to diverse families. These odors are composed of volatile compounds that are products of secondary metabolism. Specifically, these volatiles are products of several biosynthetic pathways, and can be produced from vegetation, flowers, and fruits (Shreier 1984; Croteau and Karp 1991; Dudareva and Pichersky 2000). Fragrance has many functions, including pollinator attraction and associative learning, and the deterrence of herbivores (Berenbaum and Seigler 1992; Dobson 1994; Raguso 2001). Specialist herbivores and parasitoids can also use these fragrances as cues for feeding and/or oviposition (Metcalf and Metcalf 1992; Dobson 1994; Birkett et al. 2000). As elements of plant reproductive strategies, floral fragrance profiles are analogous to animal courtship songs (e.g., frogs and birds) in that they have many compounds or notes, respectively, with the potential for synergistic interactions or "harmonics" among blend components (Ryan and Rand 1990, 1999). Important in the analysis of fragrance profiles is both the quality, i.e., which compounds are present, and the quantity, i.e.. how much of each compound is present. The complexity (i.e., potential information content) of fragrance profiles suggests that plant odors not only attract food-foraging pollinators, but also could function as learned species-specific cues that enhance floral constancy (Gerber et al. 1996; Raguso and Roy 1998; Domhaus and Chittka 1999). Thus, character displacement of fragrance chemistry might be expected to evolve among members of pollinator-limited plant guilds (Brown and Kodric-Brown 1979; Feinsinger 1983) or among closely related, sympatric plant species (Hills et al. 1972; Whitten and Williams 1992; Knudsen 1999). Finally, from the perspective of both ecological relationships and evolution, it is useful to consider whether fragrance is being produced by flowers, vegetation, or both. Selection pressure by pollinators and herbivores on fragrance composition may occur regardless of the site of fragrance emission. Further, the location 91 on a plant from which fragrance is emitted could be informative about evolutionary relationships. Although there is a diversity of possible fragrance profiles (as a result of an essentially unlimited number of chemical combinations), it has been suggested that pollinators may cue on specific odors as attractants to flowering plants (Knudsen and Tollsten 1993; Borg-Karlson et al. 1994; Kaiser and Tollsten 1995; Knudsen and Tollsten 1995; Bestmann et al. 1997; Dobson et al. 1997; Miyake et al. 1998; Grison et al. 1999). Indeed, behavioral assays have identified specific compounds and subsets of fragrance profiles that attract free flying bees (Williams and Whitten 1983; Whitten et al. 1986; Whitten et al. 1988; Roy and Raguso 1997; Dobson et al. 1999; Knudsen et al. 1999), moths (Cantelo and Jacobsen 1979; Naumann et al. 1991), beetles (Ervik, Tollsten and Knudsen 1999), and bats (von Helversen et al. 2000) to artificial flowers in field settings. Hawkmoths and the flowers they visit are a classic example of plant-pollinator co- adaptation (Thompson 1994). Throughout warm regions worldwide, hawkmoths often visit and pollinate pale, tubular, night-blooming flowers that produce abundant sucrose- rich nectar and emit sweet, perfumed floral fragrances (Grant 1983; Haber and Frankie 1989). The fragrances appear to function as long distance attractants (Tinbergen 1958) and, in combination with visual cues (Knoll 1925, Kugler 1971). are required for nectar- feeding by captive and wild nocturnal hawkmoths (Brantjes 1978; Raguso and Willis 2000). Surveys of hawkmoth-pollinated flowers suggest that the scent chemistry of these flowers is only loosely convergent (Knudsen and Tollsten 1993; Raguso 1997; Miyake et al. 1998). Floral scents are variable blends of mono- and sesquiterpenes, aromatic esters and alcohols, and nitrogen-bearing compounds, few of which are unique to hawkmoth- pollinated flowers (Kaiser 1993; Knudsen and Tollsten 1993; Appendix C). Electro physiological assays indicate that hawkmoths are olfactorily sensitive to plant volatiles of 92 all major biochemical classes (Raguso et al. 1996; Raguso and Light 1998). Although naive hawkmoths appear to use floral scent as a distance orientation cue, wind tunnel assays suggest that nearly any floral blend (but not single compounds) is sufficient to guide upwind flight (Raguso and Willis 2000). Thus, the observed variation in floral scent chemistry among species that share hawkmoth pollination indicates that different classes of scent compounds may be functionally equivalent in hawkmoth attraction, suggesting that convergence in this system may simply require the emission of any blend of plant volatiles. Whereas selection by pollinators and herbivores no doubt contributes to the overall fragrance composition, phylogenetic history is also likely to be important in explaining the remarkable variation in odors among plants. Like other characteristics of organisms, aspects of fragrance may provide information about patterns of phylogenetic relationships. To date there has been limited study of the extent to which phylogenetic history has contributed to the diversity of observed fragrances (but see Azuma et al. 1997; Barkman et al. 1997; Dobson et al. 1997; Azuma et al. 1999; Williams and Whitten 1999; Barkman 2001). One difficulty in asking this question is that it requires a robu.st phytogeny inferred from independent data for the group of interest in addition to fragrance data from a majority of the taxa in this group. Before data can be used directly for phylogeny reconstruction or mapped onto a pre-existing phylogeny to examine character evolution, these data need to be coded as characters. Whereas the coding of molecular sequence data is quite straightforward, coding fragrance data presents a number of difficulties. Analysis of fragrance composition yields a matrix, the cells of which represent the amount of each volatile compound present in the floral and vegetative fragrance of each species. These quantitative data could then be used directly as characters; however, individual 93 compounds are not independent from each other (Seaman and Funk 1983; Dobson et al. 1997; Barkman 2001). Instead, fragrances can contain compounds from more than ten different biosynthetic pathways, and the presence of one compound from a biosynthetic pathway is often correlated with the presence of another (Barkman 2001). Therefore, it is an open question whether it is best (i.e., most phylogenetically informative) to code individual compounds as characters or each biosynthetic pathway as a character. An elegant method for dealing with the non-independence of characters is the use of step- matrices (Barkman 2001). However, this method is not very effective when multiple compounds from a single biosynthetic pathway are present together in the fragrance of a single species (pers. obs.; Barkman 2001), a frequent occurrence in our study species. Another issue in coding fragrance characters is whether it is more phylogenetically informative to include information on the amount of a particular compound or pathway in the coding scheme, compared to coding these data as binary characters, reflecting simply the presence or absence of a particular compound or biosynthetic pathway. Many previous studies of volatiles have ignored quantitative differences and simply scored presence/absence of compounds or biosynthetic pathways (Barkman et al. 1997; Dobson et al. 1997; Adams 1999). However, a few studies have experimented to a limited extent with retaining quantitative information when coding fragrance data (Azuma et al. 1997; Williams and Whitten 1999). One frequent problem with evaluating the efficacy of these various coding methods is the lack of a robust phylogeny reconstructed using other sources of data upon which fragrance characters could be traced or to which a phylogenetic tree inferred using fragrance characters could be compared. Thus, to date there is no consensus as to the best methods of coding fragrance data, nor the extent to which useful phylogenetic information can be found in these data. 94 In this study we used novel approaches to address two main issues: 1) the ecological relationship between floral and vegetative fragrance and type of pollinator, and 2) the evolutionary relationship between fragrance and phylogenetic history. First, we examined the nature of the relationship between fragrance and pollination in species of three genera of Nyctaginaceae. We tested the hypothesis that flowers visited by hawkmoths have the same fragrance profile. If selection by pollinators is the main factor determining fragrance composition, then we should find support for this hypothesis. We also hypothesized that species pollinated by organisms whose attraction to flowers is not olfactory should have reduced fragrance emission or should lack fragrance entirely. Further, using phylogenetically independent contrasts, we tested for specific relationships between aspects of pollination and fragrance, including relationships previously suggested in the literature between the presence of nitrogen-bearing compounds and moth pollination, and the presence of oxygenated sesquiterpenes and hawkmoth pollination. The analysis of relationships between fragrance and type of pollinators within a phylogenetic context will provide fresh insight into the extent of convergence in fragrance among taxa that share a pollinator class. To determine if, and at what level, fragrance data are phylogenetically informative, we asked a number of questions. Previously we have shown (Appendix C) that while there is intraspecific variation in fragrance profiles, there is a much higher level of interspecific variation, suggesting that each species has its own unique fragrance profile and, thus, this information may be phylogenetically informative. Here we specifically addressed how coding influences the distinctiveness of each species' fragrance profile. However, we were primarily interested in determining the best way to handle fragrance data. Therefore, we asked which coding methods retained the most phylogenetic signal, by comparing the phylogenetic signal of the fragrance data with that 95 of an independently derived molecular (DNA sequence) data set. These varied approaches will provide the first rigorous examination of the relationship between fragrance and phytogeny and should offer insight into the utility of fragrance data in phylogenetic inference. MATERIALS AND METHODS Study System and Sampling The plant family Nyctaginaceae is composed of woody or herbaceous perennials (rarely annuals) distributed mainly in subtropical and tropical regions of the New World. In southwestern North America, the flowers of many Nyctaginaceae species have nocturnal anthesis and are visited by hawkmoths (Grant 1983). We focused on three genera of Nyctaginaceae: Adeisanthes (7 spp.), Selinocarpus (9 spp.), and Mirabilis section Quanioclidion (6 spp.). With a well-supported phylogenetic hypothesis inferred from molecular sequence data (Appendix A), we were able to examine character evolution and interactions between pollination and fragrance among these species. Fragrance was collected from multiple individuals of twenty species (Appendix C), the majority of which occur in the Chihuahuan Desert of southwestern North America. We were able to collect fragrance from 5 spp. of Adeisanthes, 6 spp. of Selinocarpus, and all 6 species in Mirabilis sect. Quamodidion. Fragrance profiles of 3 additional species of Mirabilis {M. bigelovii, M. jalapa, and M. longiflora) were also analyzed. Vouchers of all species from which volatiles were collected are housed in the herbarium at the University of Arizona (ARIZ; see Table 2 in Appendix A). Phylogenies were inferred using sequence data from the internal transcribed spacer (ITS) region of nuclear ribosomal DNA and the intergenic region between the rbcL and accD genes plus a 300 bp region of accD on the chloroplast genome (Appendix 96 A). Because we collected fragrance from a 20 species subset of the 32 species included in (Appendix C), we inferred phylogenetic relationships of this reduced sample of taxa. Parsimony analysis using the branch and bound search algorithm was done on a combined data set of the nuclear and chloroplast regions (PAUP*; Swofford 2000) with all species of Mirabilis designated as the monophyletic outgroup for rooting. The strength of support for individual tree branches was estimated using bootstrap values (Felsenstein 1985) and decay indices (Bremer 1988; Donoghue et al. 1992). Bootstrap values were from 200 full heuristic replicates with 10 random addition sequences and TBR branch swapping. Decay values for each branch were determined by using the decay index PAUP file command in MacClade (Maddison and Maddison 2000) to prepare a set of trees each with a single branch resolved. This file was then executed in PAUP* (Swofford 2000) using the heuristic search option to find the shortest trees consistent with each constraint. The decay index for each branch in question was the difference in length between the shortest trees consistent with a particular constraint and a single resolved globally shortest tree. We made observations of floral visitation for all species that were flowering in the field. Due to time constraints, observation periods were generally limited to a few hours around dusk for each species. Floral visitors were observed, and in many instances captured in order to determine if they were carrying pollen. Senesced flowers were also collected and the stigmas were examined under dissecting and compound microscopes for the presence of moth scales. When a moth (noctuid, hawkmoth, or other type of moth) has touched the stigmatic surface, it invariably leaves behind some of its scales (Nilsson and Rabakonandrianina 1988). Pollinators were inferred from perianth tube length and time of floral anthesis when no floral visitors were observed using the above methods and previous reports of floral visitation from literature were unavailable. 97 Fragrance Collection and Analysis Fragrance was collected in the field using the dynamic head-space collection method [see Raguso and Pellmyr (1998) and references therein]. This method adapted for the field involves enclosing flowers of a living plant within a polyacetate bag (Reynolds*^ oven bags) where volatile compounds emitted from the plant accumulate and are trapped in adsorbent cartridges through the use of a battery-operated diaphragm pump (KNF Neuberger, Inc.). Glass cartridges were packed with 100 mg of the adsorbent Porapak®Q (80-100 mesh), and the pump pulled air over the flowers and into the adsorbent trap at a flow rate of ca. 250 ml/min. Simultaneous collections of ambient and vegetative volatiles were used to distinguish between truly floral compounds, compounds emitted from the vegetation, and ambient contaminants. Scent collections typically commenced at floral anthesis and continued for 12 hrs. Most of the species included in this study have nocturnal floral anthesis such that fragrance collection usually began at dusk and ended early the following morning. Each of these species have flowers that last only one day; thus, the volatile collection generally encompassed the entire period during which a flower was open. For a few species, we were unable to collect floral fragrance in the field, and instead did so in the greenhouse using the same protocol as outlined above. In general, one fragrance collection per individual plant was done, and the number of flowers included for each fragrance collection was noted. After fragrance collection, flowers were removed from the plants and weighed to obtain an average fresh weight per flower for each species. To distinguish floral from vegetative volatiles, fragrance collections of only vegetative material were necessary. The architecture of these plants makes it impossible to avoid trapping vegetative volatiles while also collecting volatiles from a number of flowers. Ideally the mass of vegetation from which fragrance was collected would be 98 quantified; however, this would have required destructive sampling of plants (in many cases fragrance was collected from an entire plant), which was not feasible. Therefore, we were able to determine which compounds were emitted vegetatively, but we could not quantify the amount of these compounds that were produced. Analyses of fragrance samples were done on a gas chromatograph (GC) equipped with a mass spectrometer (Shimadzu Scientific Instruments, Inc.) (see Appendix C for complete details). Compounds were tentatively identified using computerized mass spectral libraries [Wiley and NIST libraries (>120,000 mass spectra)]. The identity of many compounds was also verified using retention times of known standards (see Table 2 in Appendix C). Quantification of compound amounts was achieved by integrating individual GC peak areas using Shimadzu Class-5000 software (Version 2.0). For each species these compound amounts were averaged across individuals and divided by the total average amount of volatiles produced to yield the relative amount of each compound present in the floral and vegetative fragrance per 12 hr collection period (Table 2 in Appendix C). For those volatiles principally or solely released from flowers, compound amounts were standardized by adjusting for the number and mass of flowers sampled to yield an amount of each compound present per jig of floral mass. Pollination and Fragrance Phylogenetically independent contrasts (PICs) were done using a fully resolved tree topology that was one of the most-parsimonious trees (MPTs) inferred for the 20 species from analysis of molecular sequence data. The Comparative Analysis by Independent Contrasts (CAIC) (Purvis and Rambaut 1995) program was used to examine correlations between a discrete variable (type of pollinator) and continuous variables including the total amount of fragrance produced and the amount of floral fragrance produced per |ig of floral tissue. The prediction was that if a certain class of pollinators 99 requires more fragrance to be attracted to flowers, then there should be a phylogenetically independent association between pollinator and amount of fragrance. Type of pollinator was coded in three different ways: as a multistate character with 3 states, (hawkmoth, noctuid moth and other); as a binary character with 0 = absence of moth pollination and 1 = presence of moth pollination (both noctuid and hawkmoth); as a binary character with 0 = not hawkmoth-pollinated and I = hawkmoth-pollinated. CAIC was also used to examine relationships between pairs of continuous characters. To test the prediction that the quantity of fragrance emitted is directly related to tlower size, we calculated PICs for the amount of floral fragrance produced per flower vs. perianth length and floral mass. A PIC was also done to test for a trade-off between the production of floral and vegetative fragrance; this PIC compared the total amount of floral volatiles produced per |i,g floral tissue and the percentage of the total volatiles that were released from the vegetation. Linear contrasts between one discrete and one continuous variable were done using the more conservative "brunch" algorithm in CAIC (Purvis and Rambaut 1995). For linear contrasts between two continuous variables, the "crunch" algorithm was used (Purvis and Rambaut 1995). For discrete variables, CAIC assumes that characters evolved parsimoniously, with ancestral states optimized according to the states of its descendents. Change in continuous characters is modeled by CAIC as a random walk, the assumptions of which are better met by transformed variables. In our analyses, all continuous variables were first transformed [natural log or arcsine-square root (proportions)] before being contrasted. CAIC cannot be used when comparing two discrete characters; for such contrasts, the Concentrated Changes Test (Maddison 1990) as implemented in MacClade (Maddison and Maddison 2000) was used. This allowed us to test the predictions put forth by Nilsson et al. (1985), Kaiser (1993) and Knudsen and Tollsten (1993) that 100 nitrogen-bearing compounds are characteristic of fragrances of moth-pollinated taxa (i.e., pollinated by either hawkmoths or noctuid moths), and that oxygenated sesquiterpenoids are characteristic of hawkmoth-pollinated taxa. With the Concentrated Changes Test, we examined whether the loss of two classes of fragrance compounds (nitrogen-bearing compounds and oxygenated sesquiterpenoids) was concentrated on branches of the phylogeny that had lost moth pollination (nitrogen-bearing compounds) or hawkmoth pollination (oxygenated sesquiterpenoids). The analyses outlined above incorporate the knowledge that phylogenetically related species are not independent. It is useful to compare these results to "phylogeny- free" analyses. One such analysis used linear regression (JMPIN 3.2.1, SAS Institute, Inc., 1997) to test whether the amount of floral fragrance produced (dependent variable) scales with flower size (independent variable). Second, to further examine the relationship between pollinators and fragrance profiles, we calculated a Euclidean distance matrix based on dissimilarities (SPSS, Inc. 1999) using the relative amount of all fragrance compounds (both floral and vegetative) for each individual of each species (these variables were first standardized to mean = 0, standard deviation = 1). We then used this matrix to ask whether overall fragrance profiles among hawkmoth-pollinated taxa are more similar to each other than they are to fragrance profiles for species with fiowers visited by other organisms. A Wilcoxon Signed Rank Test (SPSS, Inc. 1999) was used to compare the mean percent dissimilarities between paired comparisons of all hawkmoth-pollinated taxa and the mean percent dissimilarities of paired comparisons of hawkmoth-pollinated to non-hawkmoth-pollinated taxa. Phylogeny and Fragrance We examined the phylogenetic information content of fragrance characters using a variety of methods. First, we examined fragrance profiles as a whole by comparing the 10! impact of various coding methods on estimates of intraspecific versus interspecific dissimilarities in fragrance profiles and by mapping continuous characters onto the phylogeny inferred from molecular sequence data. We then explored the components of the fragrance profiles, the individual compounds or biosynthetic pathways, and how different coding methods for these components influence the phylogenetic information content of the fragrance data. To accomplish this, we inferred phylogenies from fragrance data coded in four ways, and examined congruence between the trees inferred from fragrance data and a phylogeny inferred from molecular sequence data. Finally, we mapped fragrance characters onto the phylogeny inferred from sequence data and compared the phylogenetic signal of characters across data sets. One method of comparing entire fragrance profiles used distance measures, as discussed in the previous section. With fragrance data for each individual of each species, we calculated dissimilarity matrices based on Euclidean distance (SPSS, Inc. 1999) for the data coded using 4 different methods: relative amounts of all compounds, presence/absence of all compounds, relative amounts of biosynthetic pathways, and presence/absence of biosynthetic pathways (as previously mentioned, relative amount variables were first standardized to have a mean = 0 and standard deviation = 1). Wllcoxon Signed Rank Tests (SPSS, Inc. 1999) were used to test whether the percent dissimilarity in fragrance composition of pairs of individuals within a species was equal to the percent dissimilarity of interspecific pairs. The results of these tests were compared to determine if coding affected whether there were clearly distinctive fragrance profiles for each species. In other words, was intraspecific variation in fragrance chemistry always less than interspecific variation, or did certain coding methods mask this difference? 102 To examine the evolution of fragrance profiles among these three Nyctaginaceae genera, we used squared change parsimony in MacClade (Maddison and Maddison 2000) to map a number of continuous characters onto the phylogeny inferred from molecular sequence data. These characters included: total amount of floral and vegetative volatiles produced, total number of volatile compounds produced, total amount of floral volatiles released per fig of floral mass, percentage of the total amount of volatiles produced that is released from vegetation, and percentage of the total amount of volatiles produced that is emitted from flowers. In addition to contrasting fragrance profiles as units, we explored a variety of phylogenetic methods for comparing the individual components of fragrance profiles. Because it is unclear what coding method best preserves the information content of fragrance data, we experimented with a number of coding techniques. Character-by- taxon matrices were assembled in four different ways: 1) including all compounds produced fiorally (not vegetatively) according to the amount of each compound produced per fig of floral tissue; 2) including all compounds produced from both flowers and vegetation coded as binary, presence/absence characters; 3) recoding compounds according to biosynthetic pathway, with characters defined as the presence/absence of any products of each biosynthetic pathway in flowers and vegetation; 4) including the amount of fioral compounds from each biosynthetic pathway per |i,g of floral tissue (coded as discrete character states rather than continuous). The discrete characters were multistate, with 12 different states, from absent to greater than 100 parts per |ig of floral tissue. Phylogenies were inferred from the fragrance data using PAUP* (Swofford 2000). Because PAUP* cannot analyze continuous characters, we reconstructed phylogenetic relationships using data sets 2-4 above. For data set 4, the effect of two different weighting methods was also explored; the muhi-state characters were treated as 103 unordered as well as ordered. In the unordered analysis transitions between character states (i.e., discrete amounts of compounds synthesized by each pathway) were all weighted equally (i.e., cost the same). By contrast, ordering multistate characters weights some transitions more than others; here, transitions from small to large quantities were weighted more (i.e., cost more) than transitions from small to intermediate quantity. For all data sets we conducted parsimony analyses using the heuristic search option and 100 random addition sequence replicates. As Acleisanthes plus Selinocarpiis and Mirabilis are well supported as two distinct clades by both molecular and morphological data (Appendix A), we rooted all phylogenies between these two groups. Fifty-percent majority rule consensus trees were constructed. To determine if there was phylogenetic signal in the fragrance data, we calculated the g, statistic (Hillis and Huelsenbeck 1992) as implemented in PAUP* (Swofford 2000). As the amount of structure in the data increases, the distribution of tree length frequences becomes more left skewed, resulting in a more negative value for g, (Hillis and Huelsenbeck 1992). If the fragrance data have phylogenetic signal that is congruent with evidence from DNA sequences, then the trees constructed from the fragrance data should be congruent with the tree inferred from molecular sequence data. To examine congruence, we conducted partition homogeneity tests (Farris et al. 1995) and calculated tree-to-tree distances using PAUP* (Swofford 2000). Partition homogeneity tests were done in a pairwise fashion, for each fragrance data set in combination with the molecular sequence data set. For these tests, 100 replicates were conducted, using the heuristic search option with 10 random addition sequence replicates, TBR branch-swapping, and gaps treated as missing data. Because of memory constraints only variable characters were included, and Mulpars was turned off except for data set 2, presence/absence of compounds. We calculated tree-to-tree distances (PAUP*; Swofford 2000) from each of 104 the four consensus trees inferred from fragrance data to the 50% majority rule consensus tree topology inferred using the molecular sequence data. The tree-to-tree distance metric used was symmetric-difference, which counts the number of groups present on only one of the two tree topologies. Therefore, tree topologies that are more similar will have a lower value of symmetric-difference. If the phylogenetic signal across the molecular and fragrance data sets is similar, then the partition homogeneity tests will be insignificant and all of the tree-to-tree distances will be low. Further, the fragrance coding method that retains phylogenetic information most congruent to that of the molecular sequence data will have the lowest tree-to-tree distance. If there is no difference in the phylogenetic information content across the four fragrance coding methods, then the tree-to-tree distances from all trees inferred from fragrance data to the phylogeny inferred from DNA sequence data should be the same. The last approaches for evaluating the phylogenetic information content of the fragrance data involved mapping fragrance characters to examine their evolution in the context of the hypothesis of phylogenetic relationships based on molecular sequence data. Using MacClade (Maddison and Maddison 2000), all characters from data sets 2-4. including both the unordered and ordered weighting method for matrix 4, were mapped onto the MPT topology inferred using molecular sequence data, which was also used for the phylogenetically independent contrasts discussed above. MacClade (Maddison and Maddison 2000) was used to calculate the consistency (CI) and retention indices (RI) for each parsimony-informative character from data sets 2-4 as mapped onto this tree. CIs and RIs were also calculated for each parsimony-informative character from the molecular sequence data set when mapped onto the molecular tree topology. Nonparametric Kruskal-Wallis Tests (JMPIN 3.2.1, SAS Institute, Inc., 1997) were used to compare the values for these two indices to determine whether the amount of 105 homoplasy differed among the four fragrance coding methods and the molecular sequence data set. In particular, we might expect that coding as presence/absence of a biosynthetic pathway would be less subject to homoplasy than coding as presence/absence of individual compounds, as pathways may combine highly labile compounds into a single phylogenetically informative pathway. By the same reasoning, the alternative could be just as likely, with the phylogenetic information of individual compounds becoming diluted in a larger, ubiquitous biosynthetic pathway. We also hypothesized that coding as presence/absence of individual compounds would be less homoplastic than coding as amount of compound. However, again it seems as likely that amounts would be more phylogenetically informative, with compounds appearing more homoplastic when coded as presence/absence. In addition, we used Templeton's test (Templeton 1983) as implemented in PAUP* (Swofford 2000) to test for differences in lengths of characters when a data set is optimized on one tree versus another. Four pair-wise tests were done; each of the three fragrance data sets and both the unordered and ordered weighting for data set 4 were optimized onto both the 50% majority-rule consensus tree inferred using that fragrance data set and the 50% majority-rule consensus tree inferred from molecular sequence data. The larger the differences in the lengths of the characters from a given data set on the two trees, the greater the differences in phylogenetic signal. Significant differences in lengths of characters were determined in PAUP* (Swofford 2000) by 2-tailed Wilcoxon Signed Ranks Tests. Across all four fragrance data sets, we determined the compounds and pathways with the highest CI plus RI (> I, with neither CI nor RI = 0) and the lowest CI plus RI (< 0.39). Although these cut-off points are admittedly arbitrary, examining characters in this way permitted us to determine whether certain sorts of characters and coding 106 methods are likely to be more or less phylogenetically informative. We were also able to examine differences in phylogenetic content among biosynthetic pathways and, particularly, whether pathways that have many compounds present in these species' fragrance profiles are more likely to be homoplastic than those with fewer compounds. In addition, we examined whether floral or vegetative compounds tended to be more homoplastic. The compounds and pathways with the highest summed CIs and RIs guided further mapping of characters to illustrate the phylogenetic information content of fragrance characters. RESULTS Pollination and Fragrance Figure 1 shows one of the most-parsimonious trees (MFTs) from the phylogenetic analysis of molecular sequence data; this phylogeny was used for the phylogenetically independent contrasts and character mapping. The phylogeny is one of 36 MPTs, and all but 2 clades (of these, one was present in 33% of MPTs, the other in 50% of MPTs) were present in the strict consensus of all most-parsimonious trees. Where support for branches is weak, there are often morphological data that also support the branch (e.g.. sister taxa M. piidica and M. alipes are morphologically very similar). Pollinators (Table 1) have been indicated on the phylogeny (Fig. I). There were no significant correlations between the total amount of volatiles released and pollinator type (coded in three different ways) [phylogenetically independent contrasts (PICs); Table 2; Fig. 1]. Note also that M. triflora, the only species pollinated by an organism for which scent is irrelevant (i.e.. hummingbirds), still emits fragrance at an amount similar to that of the closely related M. greenei and M. macfarlanei. However, these three species clearly have reduced fragrance emissions in 107 comparison to the other Mirabilis species (Fig. 1). There were also no significant correlations between the amount of floral volatiles produced per |ig of floral tissue and type of pollinator (coded in three different ways) (PICs, Table 2). Further, non significant correlations were observed between the amount of floral volatiles emitted per flower and perianth length, and between the amount of floral volatiles per flower and floral mass (PICs, Table 2). However, the linear regression between amount of floral volatiles produced per flower and perianth length was significant (P<0.05) when the outlier M. greenei (see Appendix C) was excluded (analysis not shown). There is clearly a stronger relationship between these two variables within Acleisanthes and Selinoccirpiis, as the regression was highly significant when all Mirabilis species were excluded. The PIC between the amount of floral volatiles produced per ^g of floral tissue and the percentage of the total volatiles that were produced vegetatively showed a significantly negative relationship (P<0.01; Table 2). The Concentrated Changes Test (Maddison and Maddison 2000) showed that the loss of nitrogen-bearing compounds in fragrance was significantly concentrated on branches of the phylogeny that had also lost moth pollination (P<0.05; Fig. 2). However, the loss of oxygenated sesquiterpenes was not concentrated on branches that had lost hawkmoth pollination. The result of comparing dissimilarities in fragrance profiles according to type of pollination showed that the mean percent differences in fragrance between pairs of hawkmoth-pollinated taxa was higher than the percent differences between hawkmoth- pollinated and non-hawkmoth-pollinated taxa (Wilcoxon Signed Ranks Test; f<0.001). Phylogeny and Fragrance The comparison of dissimilarities among fragrance profiles showed that for all methods of coding, the amount of variation in fragrance profiles within species was lower 108 than among species, as reflected in their lower mean percentage dissimilarity (Wilcoxon Signed Ranks Test; P<0.001). This trend was enhanced when we used presence/absence of compounds rather than relative amounts; any overlap between the two groups disappeared, such that in all rank comparisons the dissimilarity was greater for interspecific rather than intraspecific comparisons. Many of the measures for entire fragrance profiles appear phylogenetically informative when mapped onto the molecular tree topology, although there is also considerable convergence (Figs. 1, 3). The total amount of volatiles produced (both fioral and vegetative) is much lower in the Mirabilis clade than in the Acleisanthes plus Selinocarpiis lineage (Fig. 1); a similar pattern was observed by mapping the total amount of floral compounds produced per ^g of floral tissue (tree not shown). Not only is the total amount of volatiles emitted higher in the Acleisanthes plus Selinocarpiis lineage, but species in this group also emit a greater number of floral and vegetative compounds (Fig. 3A). The percentage of the total amount of volatiles that was released vegetatively also appears to be related to phylogeny (Fig. 3B), with sister taxon pairs often being most similar (the high percentage of volatiles released vegetatively by M. greenei is due to almost negligible volatile production). Note that the sister species of A. wriglitii and A. acutifolia, and 5. angiistifolius and S. undulatus share very high percentages of volatile production from vegetation. Mapping the percentage of the total volatiles that were released from flowers gives similar results (Fig. 3C). Character coding for data set 2 (i.e., with all compounds coded as presence/absence) yielded 204 characters, 128 of which were parsimony-informative. Data set 3 yielded 17 presence/absence biosynthetic pathway characters, 11 of which were parsimony-informative. Data set 4 included the amount of floral compounds from each biosynthetic pathway per ng of floral tissue (coded as discrete character states rather 109 than continuous); this coding method yielded 17 multi-state characters, 15 of which were parsimony-informative. All data sets appeared to have phylogenetic signal, as all had negative values for gi (Table 3). However, due to a stronger negative skew, the phylogenetic structure appears greater in the molecular sequence data and fragrance data set 2, presence/absence of all compounds (Table 3). The partition homogeneity tests allowed the rejection of the null hypothesis of homogenous data partitions for all pair-wise comparisons. All three fragrance data sets and both the unordered and ordered weighting of data set 4 were incongruent with the molecular sequence data set (^<0.010). Further, the tree-to-tree distances suggested that the tree topology from the phylogenetic reconstruction using the molecular sequence data was distinct from the tree topologies from phylogenetic analyses using fragrance data whether these were coded as the presence/absence of all volatile compounds (data set 2), the presence/absence of biosynthetic pathways (data set 3), or unordered and ordered discrete amounts of floral compounds per [ig of floral tissue according to biosynthetic pathway (data set 4) (Table 3). The trees inferred from the presence/absence data (data sets 2 and 3) were more similar to the phylogeny reconstructed using the molecular sequence data than were the trees inferred from the unordered and ordered discrete amounts of floral compounds grouped by biosynthetic pathway, with the topology from the unordered coding being the least similar (Table 3). When fragrance characters were mapped onto the molecular DNA sequence tree topology, there were diverse results in terms of consistency and retention indices for each character (Table 4). Not surprisingly, the molecular sequence characters were mapped onto the molecular tree topology with the highest CI and RIs. The fragrance data set with the highest median CI was the unordered discrete amounts of floral compounds per |ig of floral tissue from each biosynthetic pathway (data set 4), whereas the other three coding 110 methods had low median consistency indices. The fragrance data set that mapped onto the DNA sequence phylogeny with the highest RI was data set 3 (i.e., fragrance was coded as presence/absence of pathways); the lowest mean RI was from the unordered discrete amounts of floral compounds per ^ig of floral tissue from each biosynthetic pathway. Results of Templeton's tests (Table 5) suggest that the phylogenetic signal in fragrance data set 2 (presence/absence of compounds) is the most different from that in the molecular sequence data (P<0.(X)01; 2-taiIed Wilcoxon Signed Ranks Test). However, the lengths of the fragrance characters in data set 4 when both unordered and ordered were also significantly longer when optimized onto the molecular sequence topology rather than the 50% majority rule phylogeny inferred using that fragrance data set (unordered P<0.01, ordered P<0.05-, 2-tailed Wilcoxon Signed Ranks Test). By contrast, in fragrance data set 3 (presence/absence of biosynthetic pathways) the character lengths were actually shorter when optimized onto the molecular sequence tree rather than the phylogeny inferred using data set 3, although not significantly so (P>0.1; 2- tailed Wilcoxon Signed Ranks Test). This is possibly due to the much shorter total tree length for the phylogeny inferred using data set 3 in comparison to the much longer phylogenies inferred from data sets 2 and 4 (Table 5). When we closely examined the CIs and RIs for each character from data sets 2-4. interestingly, no one data set had a higher percentage of phylogenetically informative characters (sum of CI and RI > I; Table 6). However, it is notable that there appears to be a pattern across data sets as to which compounds/pathways have the highest CIs and RIs (Table 6). Whereas specific compounds from the sesquiterpenoid pathway are common among the phylogenetically informative fragrance characters (i.e., with CI + RI > I; Table 6), none of these compounds is among the more homoplastic characters (CI + Ill RJ < 0.39). By contrast, benzoates and cyclic and acyclic monoterpenoids more often are highly homoplastic (Table 6). There may also be a homoplasy pattern with data sets 3 and 4. which coded compounds according to biosynthetic pathway. For example, the three pathways that had summed CIs and RIs greater than or equal to one (phenylpropanoids. jasmonates, and tiglates) represented a small number of compounds. This is in contrast to the pathways with low summed CIs and RIs; half of these included a large number of compounds (benzoates, salicylates, cyclic monoterpenes). and half represented a small number of compounds (anthranilate. linoleic acid derivatives, benzyl ethers). In terms of whether floral or vegetative compounds are more homoplastic, there does not appear to be a clear pattern, although many of the purely vegetative sesquiterpenoids were phylogenetically informative (Table 6). A number of the characters with high CIs and RIs were also parsimoniously mapped (MacClade; Maddison and Maddison 2000) onto the DNA sequence phytogeny to illustrate the phylogenetic level at which each character was informative (Fig. 4). It appears that the presence of cis-jasmone is basal to Acleisantlies, Selinoccirpus. and Mirabilis but has been lost three times: on the branch leading to the six species in Mirabilis sect. Quamoclidion (A/, alipes. M. greenei, M. macfarlanei, M. multijlora, M. pudica, and M. triflora), on the branch leading to the sister taxa S. angustifolius and S. imdulatus, and in the S. lanceolatus lineage (Fig. 4). The oxygenated sesquiterpene farnesol has likely been gained twice within the clade of Acleisantlies and Selinocarpus: on the branch leading to A. obtiisa and A. longiflora, and on the branch leading to the other three Acleisanthes species (A. crassifolia, A. wrightii, and A. acutifolia) (Fig. 4). The vegetatively-produced sesquiterpenoids a,P-eudesmol, a-muurolene, calamenene. and a-cubebene are only emitted by the sister taxa A. acutifolia and A. wrightii, whereas the emission of phenylpropanoids is a synapomorphy for the clade containing S. 112 chenopociioides and A. acutifolia, with a loss along the A. crassifolia lineage (Fig. 4). Similarly, the presence of 8-octalactone and 6-nonalactone is a synapomorphy for the sister species 5. lanceolatus and 5. parvifoliiis (Fig. 4). DISCUSSION Ecological Influences on Fragrance Given the established pattern of association between types of fragrances and different animal pollinators (Faegri and van der PijI 1979; Kaiser 1993; Knudsen and Tollsten 1995), we might expect to find that pollinators are one of the main factors determining the presence and composition of floral fragrance in plants. Whereas most studies simply focus on floral fragrances when considering interactions with pollinators, pollinators can also use fragrances produced by vegetation as long distance odor cues, with visual cues used at close range to locate flowers (Dobson 1994; Raguso and Willis 2000). There has been interest in whether there is a fragrance profile that is characteristic of hawkmoth-pollinated taxa worldwide, as they smell distinctive to most pollination biologists (Knudsen and Tollsten 1993; Miyake et al. 1998; Appendix C). Our data set for 20 species of Nyctaginaceae, including numerous hawkmoth-pollinated taxa. allowed us to test explicitly whether fragrance profiles of hawkmoth-pollinated species were more similar to each other than to fragrance profiles of species pollinated by other animals, especially noctuid moths. Our data show the opposite, that fragrance profiles of hawkmoth-pollinated species were actually more different from each other than they were to the fragrance profiles of non-hawkmoth-pollinated species. These results agree with previous studies that suggested that there is a wide variety of fragrances to which hawkmoths could be attracted (Knudsen and Tollsten 1993; Raguso 1997; Miyake et al. 113 1998), and support the idea that species-specific odors could be learned by hawkmoths as conditioning stimuli associated with ample nectar rewards (Daly and Smith 2000). However, Knudsen and Tollsten (1993) and Kaiser (1993) suggested that the presence of particular compounds was more related to hawkmoth pollination than were entire fragrance profiles. Specifically, it was suggested that the presence of nitrogen- bearing compounds in fragrance was characteristic of moth-pollinated species (i.e., noctuid- and hawkmoth-pollinated) (Nilsson et al. 1985; Kaiser 1991; Knudsen and Tollsten 1993; Dobson et al. 1997), although this idea has never before been examined in a phylogenetic context. Further, Knudsen and Tollsten (1993) found higher amounts of nitrogenous compounds in fragrances of hawkmoth-pollinated taxa than in noctuid moth- pollinated species. Our results show that the loss of nitrogen-bearing compounds is concentrated onto branches of the phylogeny that have also lost moth pollination (Fig. 2). However, in our study, hawkmoth-pollinated taxa did not produce greater quantities of nitrogen-bearing compounds than did species pollinated by other types of moths (Appendix C). At present, the reason for the association between moth pollination and nitrogen-bearing compounds remains unclear. Results of electroantennograms showed that hawkmoths do respond to nitrogen-bearing compounds, but not more so than to a variety of other plant volatiles (Raguso et al. 1996; Raguso and Light 1998). Bioassays are definitely needed to determine whether a fragrance blend that attracts moths when a nitrogen-bearing compound is present will still attract moths in the absence of such a compound. Knudsen and Tollsten (1993) also suggested that oxygenated sesquiterpenes were more characteristic of hawkmoth- than noctuid moth-pollinated fragrances. Our results for Nyctaginaceae support this idea to some extent, although not all of the hawkmoth- pollinated taxa emitted these compounds in their fragrance, and the fragrance of other 114 taxa that are likely noctuid moth-pollinated also contained these compounds (Appendix C). Moreover, results of the Concentrated Changes Test did not show that the loss of oxygenated sesquiterpenes was concentrated on branches that had also lost hawkmoth pollination. Instead, there appears to be an association of the presence of oxygenated sesquiterpenes with degree of phylogenetic relatedness. Specifically, oxygenated sesquiterpenes are lacking in all species in Mirabilis sect. Quamoclidion (i.e., excluding M. bigelovii, M. jalapa, and M. longiflord) and in the sister taxa S. angustifolius and 5. undulatus. Odor not only functions for pollinator attraction but also to repel herbivores/florivores (Galen 1985; Mullin et al. 1991; Berenbaum and Seigler 1992; Dobson 1994; Euler and Baldwin 1996; Galen 1999). In fact, fragrance blends are often composed of both attractive and repellent compounds (Euler and Baldwin 1996. Omura et al. 2000), with certain compounds being at once attractive to some insects and repellent to others (Berenbaum and Seigler 1992; Birkett et al. 2000; Dobson and Bergstrom 2000). Further, floral odor has been shown to act as a feeding and/or oviposition cue for herbivores (Metcalf and Metcalf 1992; Dobson 1994), thus suggesting a strong cost to the olfactory apparency provided by fragrance production (Baldwin et al. 1997). Therefore, just as there may be selection by pollinators acting on fragrance, there may also be stabilizing selection by herbivores on fragrance, and one could imagine that these two factors could exert conflicting selection pressures on fragrance production (Berenbaum and Seigler 1992; Raguso 2001). One way to test this idea is to ask whether species that are pollinated by organisms that ignore floral odors also lack fragrance. The Nyctaginaceae data set has very limited ability to test this, as only one of the species included in this study, Mirabilis triflora, is pollinated by such an organism, in this case hummingbirds (Table I). Further, while flowers of this species produce a small amount 115 of fragrance, two of the other species release even smaller quantities of fragrance (Appendix C). One possible reason for the maintenance of fragrance in M. triflora is that the importance of fragrance as an herbivore deterrent may outweigh the cost of increased olfactory apparency to herbivores. Alternatively, as its closest relatives emit fragrance, M. triflora may simply retain some fragrance emission due to a developmental association between flower and fragrance production (i.e., fragrance produced concomitantly with flowers) and a lack of strong selection against fragrance emission. Whereas volatiles may be simultaneously released from a plant's floral and vegetative structures, all species emitted some compounds exclusively from flowers (Appendix C). This allows floral visitors to discern, even from a distance, whether a plant is flowering. However, because of the cost of fragrance production, both in terms of resources and apparency (Berenbaum and Seigler 1992; Euler and Baldwin 1996; Baldwin et al. 1997), and the ability of floral visitors to respond to both floral and vegetative compounds (Dobson 1994; Raguso and Willis 2000), there may be a tradeoff between the production of floral and vegetative volatiles. Our results support the idea of a tradeoff, with a significant negative phylogenetically independent correlation between the amount of floral volatiles produced per fig of floral mass and the percentage of the total volatiles produced vegetatively (Table 2). Although a negative correlation could also be observed if the amount of volatiles emitted from vegetation remained constant and there was simply an increase in the amount of floral volatiles produced per |ig of floral mass, it is clear that the amount of volatiles emitted from vegetation does differ across species, with some totally lacking vegetative fragrance (Figs. 3B-C; Appendix C). We also investigated whether floral fragrance production scales proportionately with flower size. Neither the PIC of the amount of floral volatiles produced per flower and perianth length nor the PIC of the amount of floral volatiles produced per flower and 116 floral mass showed significant correlations between the two characters (Table 2). This result suggests that phylogeny may instead play a role. Indeed, there was a significant linear regression, with the amount of floral volatiles increasing with perianth length. However, this relationship is mainly due lo Acleisanthes and Selinocarpus species and not to Mirabilis species. Further evidence of an influence of phylogenetic relationships on fragrance comes from the insignificant correlations between the total amount of volatiles released and type of pollinator, and between the amount of floral volatiles produced per |J,g of floral mass and pollinator (Table 2). Although there may be a decrease in fragrance with the loss of hawkmoth pollination (Fig. 1), there may not be a sufficient number of pollinator transitions among these three Nyctaginaceae genera to yield a significant PIC. Further complicating the situation is the clear relationship between the amount of fragrance and phylogeny, with all Mirabilis species (including the three hawkmoth-pollinated taxa) producing a much lower quantity of fragrance than Acleisanthes and Selinocarpus species (Fig. 1). Pfiylogenetics of Fragrance Consistent with findings of Barkman et al. (1997) for Cypripedium (Orchidaceae). comparisons of dissimilarities in fragrance profiles within and among species strongly showed that although there is intraspecific variation in fragrance profiles, this variation is much lower than interspecific variation in fragrance composition. These differences between intra- and interspecific comparisons are further magnified when data are coded as presence/absence rather than as the relative amount of each compound, suggesting that within species there are fewer qualitative than quantitative differences. Further, these results support the conclusion that while the fragrances of different species may share many compounds, each species has its own unique fragrance profile. 117 From the results of the PICs, it is apparent that there is some effect of phylogeny on fragrance. Further, mapping continuous characters, including total amounts and number of volatiles, strongly suggests that there are phylogenetic relationships among these characters. It is clear that members of the genus Mirabilis emit a lower total amount of volatiles composed of fewer compounds than do Selinocarpiis and Acleisanthes (Figs. 1, 3A). By contrast to these higher level phylogenetic patterns, the percentages of the total amount of volatiles emitted from flowers versus vegetation is mainly phylogenetically informative among closely related species (Figs. 3B, 3C). When we examine fragrance compounds and biosynthetic pathways rather than large-scale patterns in odor production, it is much more difficult to discern phylogenetic information. The highly negative g, statistics (Table 3) suggest that there is hierarchical structure in the data and, thus, phylogenetic signal. However, clearly this structure is not congruent between molecular sequence and fragrance data sets, as shown by the significant partition homogeneity tests and large tree-to-tree distances (Table 3). Results of the tree-to-tree distances (Table 3) and the comparison of CIs and RIs of fragrance characters mapped onto the DNA sequence phylogeny (Table 4) showed that fragrance characters tend to be much more homoplastic than molecular characters. It is also not clear how to best code these fragrance data. Neither presence/absence versus amount, nor all compounds versus biosynthetic pathways appears to have a clear advantage in terms of the amount of phylogenetic information retained, although results of Templeton's tests suggest that coding as presence/absence of pathways may yield characters that are most compatible with the hypothesis of phylogenetic relationships inferred from the molecular sequence data (Table 5). Generally, it appears that as with numerous types of characters, including morphological traits such as leaf length, many fragrance characters are just too IIS convergent to be useful in phylogenetic inference (Williams and Whitten 1999, Barkman 2001). However, as with morphology, that does not mean we should exclude all fragrance characters; it simply suggests that there are certain characters that are phylogenetically informative. It is these characters that we should either include in phylogenetic analyses along with other types of data or map onto a preexisting phylogenetic hypothesis for a given group. Table 6 can act as a guide to which of these many compounds and biosynthetic pathways are the least convergent and, thus, more useful for phylogenetic inference. For example, sesquiterpenoids, lactones, and phenylpropanoids have much higher CIs and RIs than compounds from most other pathways (Table 6). This is not surprising, as these compounds are less ubiquitous (Knudsen et al. 1993), and, therefore, more likely to be homologous rather than convergent in their occurrence. Conversely, the more common compounds such as the benzoic acid esters (i.e., benzoates; Table 6) are liable to be convergent and, thus, may be more likely involved in adaptation for the attraction of hawkmoths. We are currently designing behavioral bioassays to test this hypothesis. It becomes clear from examining Table 6 that fragrance compounds, biosynthetic pathways, and amounts may ail be phylogenetically informative, although the Information varies depending on the particular compound or pathway. For example, the sesquiterpenoid compounds are not phylogenetically informative when grouped into one biosynthetic pathway as a presence/absence character, because the presence of sesquiterpenoids is fairly common (Appendix C). Further, while the amount of these compounds produced florally is somewhat phylogenetically informative (Table 6; ordered CI=0.36, RI=0.33), what is most informative is the presence of specific sesquiterpenoid compounds in the flowers or vegetation, with many of these vegetative sesquiterpenoids (a,P-eudesmol, a-muurolene, calamenene, and a-cubebene) acting as 119 synapomorphies for A. acutifolia plus A. wrightii (Fig. 4). The oxygenated sesquiterpene famesol, which is released from flowers, is also phylogenetically informative (Fig. 4, Table 6; CI=0.50, RI=0.75). Interestingly, famesol is a synapomorphy for both clades of Acleisanthes\ the sister taxa A. obtusa and A. longiflora share this compound, as do A. acutifolia, A. crassifoUa, and A. wrightii (Fig. 4). Lactone compounds are another example of when coding as a biosynthetic pathway is not as representative of phylogenetic relationships as coding separate compounds. Lactones are not very common in floral fragrances (Knudsen et al. 1993), being more associated with fruit odors (Tressl and Albrecht 1986). In Nyctaginaceae the presence of lactones in flowers is phylogenetically informative deep in the phylogeny; they occur in species of the closely related genera Acleisanthes and Selinocarpiis, but are absent from Mirabilis (Table 6; Appendix C). However, two specific lactones (6-octalactone and 6- nonalactonei are also synapomorphies for the sister species S. lanceolatiis and S. parvifoliiis (Fig. 4). A good example of when both quality and quantity are phylogenetically informative is the occurrence of the monoterpenoid linaiool oxides. These compounds are only found in four species, with high amounts released from flowers of the sister taxa M. piidica and M. alipes (Appendix C). Jasmonate compounds have been thought characteristic of moth-pollinated flowers (Kaiser 1993). One of these compounds, cis-jasmone, is released from the flowers of a number of Nyctaginaceae species (Fig. 4). This compound is also somewhat phylogenetically informative (Fig. 4; Table 6); for example, within Mirabilis cis-jasmone is only present in the sister taxa M. longiflora and M. jalapa. Jasmonic acid and its metabolites are important signal transductants in herbivore-induced plant defenses (see reviews by Dicke and van Loon 2(X)0; Ryan 2000), and are derived from linolenic acid via a special branch of the lipoxygenase pathway (Vick and Zimmerman 1984; Farmer 120 and Ryan 1992; Croft et al. 1993; Koch et al. 1997). It is possible that interspecific differences in the demands for plant defense limit the distribution of cis-jasmone in floral fragrance and the availability of its precursors in our study plants. Another group of compounds that is phylogenetically informative at the level of biosynthetic pathway are the phenylpropanoid products related to eugenol and vanillin. High amounts of these compounds only occur in the sister species A. aciitifolia and A. wriglitii, with a small quantity of eugenol compounds also produced by S. clienopodioides (Fig. 4; Table 6). Lastly, a group of compounds that are mainly phylogenetically informative at the level of amount are the isoleucine-derived esters of tiglic acid (Hill et al., 1980) (Table 6). In particular, high amounts of cis-3-hexenyl tiglate are released by flowers of the sister taxa 5. parvifolius and 5. lanceolatiis. Conclusions Comparisons of fragrance and phylogeny in these three Nyctaginaceae genera suggest that, whereas many compounds or pathways appear phylogenetically informative, most are either highly homoplastic or autapomorphic (Appendix C). Further, our results suggest that valuable information would be ignored if fragrance data were coded only by pathway (e.g., Azuma et al. 1999). Additionally, although the quantitative data are subject to environmental conditions (Jakobsen and Olsen 1994), it seems misguided to disregard quantitative information, especially as it is clearly phylogenetically informative in many cases. As with morphological characters, it appears that it may be most appropriate to be selective in including only certain fragrance characters in phylogenetic inference. Data from overall fragrance profiles, including total amount of volatiles, appear to distinguish the Acleisanthes plus Selinocarpus clade from the Mirabilis clade. Further, individual fragrance compounds and biosynthetic pathways appear to be especially informative for helping to define sister taxa relationships, while generally not 121 being as informative for higher level relationships. The utility of fragrance data in supporting relationships among sister taxa is consistent with a number of previous studies of volatiles (Azumaet al. 1997, 1999; Adams 1999; Williams and Whitten 1999). Therefore, it is clear that while it may not be that advantageous to use fragrance characters for phylogeny reconstruction, many of these characters can certainly provide support for a phylogeny inferred using other types of data. In these twenty Nyctaginaceae species, phylogeny appears to be more important than ecological factors in predicting fragrance composition. This is in contrast to Knudsen and Tollsten's (1993) conclusions that there has been marked selection on the fragrance chemistry of hawkmoth-pollinated taxa from nine different angiosperm families, so much so that it would greatly obscure any phylogenetic patterns in fragrance. However, the species included in this study were mainly hawkmoth-pollinated or noctuid moth-pollinated. We are currently comparing the results presented here to similar data from groups with more pollinator transitions in Onagraceae and Solanaceae. Behavioral bioassays are also in progress to determine the responses of nocturnal hawkmoths to specific compounds and blends of compounds. Acknowledgements. We are grateful to T. Tully, A. Boyd, M. Manktelow. D. Heam, and A. Chetochine, without whose assistance all of the fragrance collections could not have been made. We thank Big Bend National Park, the Jornada LTER, and J. Duft for access to plants, and A. Le Due and M. Fishbein for providing seeds. For use of a GC-MS and injection of many of the samples we thankfully acknowledge E. Pichersky and J. D'Auria, Univ. of Michigan. The main source of funding for this research was provided by the National Science Foundation DEB9806840 to L. A. McDade and R. A. Raguso. Further support was provided by small grants to R. Levin from The American Society of 122 Plant Taxonomists, Sigma Xi, the Department of Ecology & Evolutionary Biology at the University of Arizona, and the University of Arizona Research Training Group in the Analysis of Biological Diversification (NSF DIR-9113362, BIR-9602246). This paper is part of a dissertation by R. Levin in partial fulfillment of the requirements for the degree of doctor of philosophy, University of Arizona. LITERATURE CITED Adams, R. P. 1999. Systematics of multi-seeded eastern hemisphere Jiiniperiis based on leaf essential oils and RAPD DNA fingerprinting. Biochem. Syst. Ecol. 27:709-725. Azuma, H.. M. Toyota, Y. Asakawa, R. Yamaoka, J. G. Garcia-Franco, G. Dieringer, L. B. Thien, and S. Kawano. 1997. Chemical divergence in floral scents of Magnolia and allied genera (Magnoliaceae). Plant Species Biol. 12:69-83. Azuma, H.. L. B. Thien, and S. Kawano. 1999. Molecular phylogeny of Magnolia (Magnoliaceae) inferred from cpDNA sequences and evolutionary divergence of the fioral scents. J. Plant Res. 112:291-306. Baldwin, I. T., C. Preston, M. Euler, and D. Gorham. 1997. Patterns and consequences of benzyl acetone fioral emissions from Nicotiana atteniiata plants. J. Chem. Ecol. 23:2327-2343. Barkman, T. J. 2001. Character coding of secondary chemical variation for use in phylogenetic analyses. Biochem. Syst. Ecol. 29:1-20. Barkman, T. J., J. H. Beaman, and D. A. Gage. 1997. Floral fragrance variation in Cypripediiim: Implications for evolutionary and ecological studies. Phytochemistry 44:875-882. 123 Barnes, J. L. 1996. Reproductive ecology, population genetics, and clonal distribution of the narrow endemic: Mirabilis macfarlanei (Nyctaginaceae). Master's thesis, Utah State University, Logan. Berenbaum, M., and D. Seigler. 1992. Biochemicals: Engineering problems for natural selection. Pp. 89-121 in B. D. Roitberg and M. B. Isman, eds. Insect Chemical Ecology: An Evolutionary Approach. Chapman & Hall, New York. Bestmann, H. J., L. Winkler, and O. von Helversen. 1997. Headspace analysis of volatile flower scent constituents of bat-pollinated plants. Phytochemistry 46:1169-1172. Birkett. M. A., C. A. M. Campbell, K. Chamberlain, E. Guerrieri, A. J. Hick, J. L. Martin. M. Matthes, J. A. Napier, J. Pettersson, J. A. Pickett, G. M. Poppy, E. M. Pow, B. J. Pye, L. E. Smart, G. H. Wadhams, L. J. Wadhams, and C. M. Woodcock. 2000. New roles for ctj-jasmone as an insect semiochemical and in plant defense. Proc. Nat. Academy of Sciences 97:9329-9334. Borg-Karlson, A.-K., I. Valterova. and L. A. Nilsson. 1994. Volatile compounds from flowers of six species in the family Apiaceae: bouquets for different pollinators? Phytochemistry 35:111-119. Brantjes, N. B. M. 1978. Sensory responses to flowers in night-flying moths. Pp. 13-19 in A. J. Richards, ed. The Pollination of Flowers by Insects. Academic Press. London. Bremer, K. 1988. The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42:795-803. Brown, J. H., and A. Kodric-Brown. 1979. Convergence, competition, and mimicry in a temperate community of hummingbird-pollinated flowers. Ecology 60:1022-1035. 124 Cantelo, W., and M. Jacobson. 1979. Phenylacetaldehyde attracts moths to bladder flower and to blacklight traps. Env. Ent. 8:444-447. Croft, K. P. C., F. Juttner, and A. J. Slusarenko. 1993. Volatile products of the lipoxygenase pathway evolved from Phaseolus vulgaris (L.) leaves inoculated with Pseudomonas syringae pv phaseolicola. Plant Physiology 101:13-24. Croteau, R., and F. Karp. 1991. Origin of natural odorants. Pp. 101-126 in P. M. Miiller and D. Lamparsky, eds. Perfumes: Art, Science and Technology. Elsevier, London. Cruden, R. W. 1970. Hawkmoth pollination of Mirabilis (Nyctaginaceae). Bull. Torr. Bot. Club 97:89-91. Daly, K., and B. H. Smith. 2000. Associative olfactory learning in the moth Mandiica sexta. J. Exper. Biol. 203:2025-2038. Dicke, M., and J. J. A. van Loon. 2000. Multitrophic effects of herbivore-induced plant volatiles in an evolutionary context. Entomologia Experimentalis et Applicata 97:237-249. Dobson, H. E. M. 1994. Floral volatiles in insect biology. Pp. 47-81 in E. A. Bemays. ed. Insect-Plant Interactions, Vol. V. CRC Press, Boca Raton. Dobson, H. E. M., and G. Bergstrom. 2000. The ecology and evolution of pollen odors. Plant Syst. Evol. 222:63-87. Dobson, H. E. M., J. Arroyo, G. Bergstrom, and 1. Groth. 1997. Interspecific variation in floral fragrances within the genus Narcissus (Amaryllidaceae). Biochem. Syst. Ecol. 25:685-706. 125 Dobson, H. E. M., E. M. Danielson, and I. D. van Wesep. 1999. Pollen odor chemicals as modulators of bumble bee foraging on Rosa rugosa Thunb. (Rosaceae). Plant Species Biology 14:153-166. Donoghue. M. J., R. G. Olmstead, J. F. Smith, And J. D. Palmer. 1992. Phylogenetic relationships of Dipsacales based on rbcL sequences. Annals of the Missouri Botanical Garden 79:333-345. Domhaus, A., and L. Chittka. 1999. Insect behaviour - Evolutionary origins of bee dances. Nature 401:38. Dudareva, N., and E. Pichersky. 2000. Biochemical and molecular genetic aspects of floral scents. Plant Physiology 122:627-633. Ervik, F., L. Tollsten, and J. T. Knudsen. 1999. Floral scent chemistry and pollination ecology in phytelephantoid palms (Arecaceae). PI. Syst. Evol. 217:279-297. Euler, M., and I. T. Baldwin. 1996. The chemistry of defense and apparency in the corollas of Nicotiana attemiata. Oecologia 107:102-112. Faegri, K., and L. van der Pijl. 1979. The Principles of Pollination Ecology, 3"^ ed. Pergamon Press, Oxford. Farmer, E. E, and C. A. Ryan. 1992. Octadecanoid precursors of jasmonic acid activate the synthesis of wound-inducible proteinase inhibitors. Plant Cell 4:129-134. Farris, J. S., M. Kallersjo, A. G. Kluge, and C. Bult. 1995. Testing significance of incongruence. Cladistics 10:315-319. Feinsinger, P. 1983. Coevolution and pollination. Pp. 282-310 in D. J. Futuyma and M. Slatkin, eds. Coevolution. Sinauer, Sunderland, MA. 126 Felsenstein, J. 1985. Confidence limits on phylogenies: An approach using the bootstrap. Evolution 39:783-791. Fowler, B. A., and B. L. Turner. 1977. Taxonomy of Selinocarpus and Ammococlon (Nyctaginaceae). Phytologia 37:177-208. Galen, C. 1985. Regulation of seed-set in Polemoniiim viscosum: floral scents, pollination, and resources. Ecology 66:792-797. Galen, C. 1999. Flowers and enemies: predation by nectar-thieving ants in relation to variation in floral form of an alpine wildflower, Polemoniiim viscosum. Oikos 85:426-434. Gerber, B., N. Geberzahn, F. Hellstem, J. Klein, O. Kowalksy, D. Wustenberg, and R. Menzel. 1996. Honey bees transfer olfactory memories established during fiower visits to a proboscis extension paradigm in the laboratory. Animal Behaviour 52:1079-1085. Grant, V. 1983. The systematic and geographical distribution of hawkmoth flowers in the temperate North American flora. Bot. Gaz. 144:439-449. Grant, V., and K. A. Grant. 1983. Hawkmoth pollination of Mirabilis longiflora (Nyctaginaceae). Proc. Nat. Acad. Sci. 80:1298-1299. Grison, L.. A. A. Edwards, and M. Hossaert-McKey. 1999. Interspecies variation in floral fragrances emitted by tropical Ficiis species. Phytochemistry 52:1293-1299. Haber, W. A., and G. W. Frankie. 1989. A tropical hawkmoth community: Costa Rican dry forest Sphingidae. Biotropica 21:155-172. Helversen, O. von, L. Winkler, and H. J. Bestmann. 2000. Sulphur-containing "perfumes" attract flower-visiting bats. J. Comp. Physiol. A 186:143-153. Hill, R, K., S.-W. Rhee, E. Leete, and B. A. McGaw. 1980. Stereospecificity of enzymatic dehydrogenation during tiglate biosynthesis. J. Am. Chem. Soc. 102:7344- 7348. Hillis, D. M., and J. P. Huelsenbeck. 1992. Signal, noise, and reliability in molecular phylogenetic analyses. J. Heredity 83:189-195. Hills, H. G., N. H. Williams, and C. H. Dodson. 1972. Floral fragrances and isolating mechanisms in the genus Catasetum (Orchidaceae). Biotropica 4:61-76. Hodges, S. A. 1995. The influence of nectar production on hawkmoth behavior, self- pollination, and seed production in Mirabilis multiflora (Nyctaginaceae). Am. J. Bot. 82:197-204. Jakobsen, H. B., and C. E. Olsen. 1994. Influence of climatic factors on emission of flower volatiles in situ. Planta 192:365-371. Kaiser, R. 1991. Trapping, investigation and reconstitution of flower scents. Pp. 213-250 in P. M. Miiller and D. Lamparsky, eds. Perfumes: Art, Science and Technology. Elsevier, London. Kaiser, R. 1993. The Scent of Orchids: Olfactory and chemical investigations. Elsevier, Amsterdam. Kaiser, R., and L. Tollsten. 1995. An introduction to the scent of cacti. Flav. Fragr. J. 10:153-164. 128 Knoll, F. 1925. Lichtsinn und Bliitenbesuch des falters von Deilephila livomica. Zeitschrift fiir Vergleichende Physiologic 2:329-380. Knudsen, J. T. 1999. Floral scent chemistry in geonomoid palms (Palmae: Geonomeae) and its importance in maintaining reproductive isolation. Memoirs of the New York Botanical Garden 83:141-157. Knudsen, J. T., and L. Tollsten. 1993. Trends in floral scent chemistry in pollination syndromes: floral scent composition in moth-pollinated taxa. Bot. J. Linn. Soc. 113:263-284. Knudsen. J. T., and L. Tollsten. 1995. Floral scent in bat-pollinated plants: a case of convergent evolution. Bot. J. Linn. Soc. 119:45-57. Knudsen, J. T., L. Tollsten, and L. G, Bergstrdm. 1993. Floral scents—a checklist of volatile compounds isolated by head-space techniques. Phytochemistry 33:253-280. Knudsen. J. T., S. Andersson, and P. Bergman. 1999. Floral scent attraction in Geonoma macrostachys, an understorey palm of the Amazonian rain forest. Oikos 85:409-418. Koch, T., K. Bandemer, and W. Boland. 1997. 64. Biosynthesis of cw-jasmone: a pathway for the inactivation and the disposal of the plant stress hormone jasmonic acid to the gas phase? Helvetica Chimica Acta 80:838-850. Kugler, H. 1971. Zur Bestaubung grossblumiger Datura-Arten. Rora 160:511-517. Maddison, W. P. 1990. A method for testing the correlated evolution of two binary characters: Are gains or losses concentrated on certain branches of a phylogenetic tree? Evolution 44:539-557. Maddison, D. R., and W. P. Maddison. 2000. MacClade 4: Analysis of Phylogeny and Character Evolution. Sinauer Associates, Sunderland, MA. Martinez Del Ri'o, C., and A. Burquez. 1986. Nectar production and temperature dependent pollination in Mirabilis jalapa L. Biotropica 18:28-31. Metcalf, R. L., and E. R. Metcalf. 1992. Plant kairomones in insect ecology and control. Chapman & Hall, New York. Miyake, T., R. Yamaoka, andT. Yahara. 1998. Floral scents of hawkmoth-pollinated flowers in Japan. J. Plant Res. Ill;199-205. Mullin, C. A., A. A. Alfatafta, J. L. Harman. A. A. Serino, and S. L. Everett. 1991. Com rootworm feeding on sunflower and other Compositae: influence of floral terpenoid and phenolic factors. Pp. 278-292 in P. A. Hedin, ed. Naturally Occurring Pest Bioregulators. American Chemical Society, Washington, DC. Nilsson L. A, and E. Rabakonandrianina. 1988. Hawkmolh scale analysis and pollination specialization in the epilithic Malagasy endemic Aerangis ellisii (Reichenb. Fil.) Schltr. (Orchidaceae). Botanical Journal of the Linnean Society 97:49-61. Nilsson, L. A., L. Jonsson, L. Rason, and E. Randrianjohany. 1985. Monophily and pollination mechanisms in Angraeciim arachnites Schltr. (Orchidaceae) in a guild of long-tongued hawk-moths (Sphingidae) in Madagascar. Biol. J. Linn. Soc. 26:1-19. Naumann, C. M., P. Ockenfels, J. Schmitz, F. Schmidt, and W. Francke. 1991. Reactions of Zygaena moths to volatile compounds of Knautia arvensis (Lepidoptera, Zygaenidae). Entomologia Generalis 15:255-264. Omura, H., K. Honda, and N. Hayashi. 2000. Floral scent of Osmanthus frag runs discourages foraging behavior of cabbage butterfly, Pieris rapae. J. Chem. Ecol. 26:655-666. Pilz, G. E. 1978. Systematics of Mirabilis subgenus Quamoclidion (Nyctaginaceae). Madrono 25:113-132. Purvis, A., and A. Rambaut. 1995. Comparative analysis by independent contrasts (CAIC): an Apple Macintosh application for analysing comparative data. Computer Appl. Biosciences 11:247-251. Raguso, R. A. 1997. Convergent evolution in hawkmoth pollinated flowers? Chemical analysis of floral advertisements and rewards. Int. Chem. Ecol. Soc., Vancouver. BC. Canada. July 1997. Raguso, R. A. 2001. Floral scent, olfaction and scent-driven foraging behavior. In: Chittka, L., Thomson, J. (Eds), Pollinator Behaviour and its Implications for Plant Evolution. Cambridge Univ. Press, in press. Raguso, R. A., and D. M. Light. 1998. Electroantennogram responses of male Sphinx perelegans hawkmoths to floral and 'green-leaf volatiles'. Entomologia Experimentalis et Applicata 86:287-293. Raguso, R. A., and O. Pellmyr. 1998. Dynamic headspace analysis of floral volatiles: a comparison of methods. Oikos 81:238-254. Raguso, R. A., and B. A. Roy. 1998. 'Floral' scent production by Puccinia rust fungi that mimic flowers. Molecular Ecology 7:1127-1136. 131 Raguso, R. A., and M. W. Willis. 2000. The importance of olfactory and visual cues in nectar foraging by nocturnal hawkmoths. In press in C. L. Boggs, W. B. Watt, and P. R. Ehrlich, eds. Proceedings of the Third Symposium on Butterfly Ecology and Evolution. University of Chicago Press. Raguso, R. A., D. M. Light, and E. Pichersky. 1996. Electroantennogram responses of ffv/es lineata (Sphingidae: Lepidoptera) to volatile compounds from Clarkia breweri (Onagraceae) and other moth-pollinated flowers. J. Chem. Ecol. 22:1735-1766. Rohlf, F. J., and R. R. Sokal. 1995. Stati.stical Tables, 3"^ edition. W. H. Freeman and Co., New York. Roy, B. A., and R. A. Raguso. 1997. Olfactory versus visual cues in a flora! mimicry system. Oecologia 109:414-426. Ryan, C. A. 2000. The systemin signaling pathway: differential activation of plant defensive genes. Biochimicaet Biophysica acta-Protein Structure and Molecular Enzymology 1477:112-121. Ryan, M. J., and A. S. Rand. 1990. The sensory basis of sexual selection for complex calls in the Tungara frog, Physalaemiis piistiilosus (sexual selection for sensory exploitation). Evolution 44:305-314. Ryan, M. J., and A. S. Rand. 1999. Phylogenetic influence on mating call preferences in female tungara frogs, Physalaemiis pustidosus. Animal Behaviour 57:945-956. SAS Institute, Inc. 1997. JMPIN, 3.2.1. SAS Institute Inc., Cary, NC. Schreier, P. 1984. Chromatographic Studies of Biogenesis of Plant Volatiles. Dr. Alfred Hiithig Verlag, Heidelberg. 132 Seaman, F. C., and V. A. Funk. 1983. Cladistic analysis of complex natural products: Developing transformation series from sesquiterpene lactone data. Taxon 32:1-27. Shimadzu Co., 1993-1996. Class-5000, version 2.0.0. Shimadzu chemical laboratory analysis system & software. Shimadzu Co. Spellenberg, R., and R. K. Delson. 1977. Aspects of reproduction in Chihuahuan Desert Nyctaginaceae. Pp. 273-287 in R. H. Wauer and D. H. Riskind, eds. Transactions of the symposium on the biological resources of the Chihuahuan Desert region United States and Mexico. U. S. Department of the Interior. SPSS Inc., 1999. SPSS for Windows, version 10.0.5. SPSS Inc., Chicago. Swofford, D. L. 2000. PAUP*. Phylogenetic analysis using parsimony (* and other methods). Version 4.0b4a. Sinauer Associates, Sunderland, MA. Templeton, A. R. 1983. Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to the humans and apes. Evolution 37:221-244. Thompson, J. N. 1994. The Coevolutionary Process. University of Chicago Press, Chicago. Tinbergen, N. 1958. Curious naturalists. Doubleday & Co., Garden City, NY. Tressl, R., and W. Albrecht. 1986. Biogenesis of aroma compounds through acyl pathways. Pp. 114-133 in T. H. Parliament and R. Croteau, eds. Biogeneration of aromas. American Chemical Society, Washington, DC. Vick, B. A., and D. C. Zimmerman. 1984. Biosynthesis of jasmonic acid by several plant species. Plant Physiology 75:458-461 133 Whitten, W. M., and N. H. Williams. 1992. Floral fragrances of Stanhopea (Orchidaceae). Lindleyana 7:130-153. Whitten, W. M., N. H. Williams, W. S. Armbruster, M. A. Battiste, L. Strekowski, and N. Lindquist. 1986. Carvone oxide: An example of convergent evolution in euglossine pollinated plants. Systematic Botany 11:222-228. Whitten, W. M., H. G. Hills, and N. H. Williams. 1988. Occurrence of ipsdienol in floral fragrances. Phytochemistry 27:2759-2760. Williams, N. H., and W. M. Whitten. 1983. Orchid floral fragrances and male euglossine bees; methods and advances in the last sesquidecade. Biological Bulletin 164:355- 395. Williams, N. H., and W. M. Whitten. 1999. Molecular phylogeny and floral fragrances of male euglossine bee-pollinated orchids: A study of Stanhopea (Orchidaceae). Plant Species Biology 14:129-136. Table I. Species included in this study, and what is known about pollinator relationships. Except where indicated, ail data are from personal observations of R. Levin including observations of Horal visitation, examination of visitors for pollen on their bodies, and microscopic study of stigmas for the pre.sence of moth scales. In addition to nocturnal floral anthesis, floral morjihological traits taken to indicate hawkmoth pollination included a perianth > 2 cm long; a perianth < 2 cm long suggested noctuid moth pollination. NA in the second column indicates that neither personal ob.servations nor reports of floral visitors in the literature were available; NA in the third column means that stigmas were not examined for moth scale deposition. Species Observed floral visitors Moth scales Likely pollinators Evidence Genus Acleistmthes A. Gray A. acuiifolia Standi. NA NA Hawkmoths Floral morphology A. crassifolia A. Gray None yes Hawkmoths Pers. observations A. Umgiflora A. Gray Hawkmoth: Manduca NA Hawkmoths Pcrs. observations; quiiuiuemaculaia; carrying pollen Spellenberg and Delson 1977 A. ohliisa (Choisy) Standi. Hawkmoth, noctuid moths, none yes Hawkmoths Pers. observations were caught A. wrighlii (A. Gray) Bcnih. & Hook. None yes Hawkmoths Pers. observations Genus Mirabilis L. M. (ilipes (S. Walson) Pilz Hawkmoths: Hyles linetila; not yes Noctuid moths Pers. observations carrying pollen M. higelovii A. Gray Noctuid moths; none caught NA Noctuid moths Floral morphology M. greenei S. Walson None possibly Hawkmoths Floral morphology M. jalapa L. Hawkmoths: Erinnyis ello and NA Hawkmoths Martinez del Rio and Burquez Hyles lineala; carrying pollen 1986 M. longijlora L. Hawkmoths: Manduca NA Hawkmoths Grant and Grant 1983 quinquemaadauf, carrying pollen Other visitors included small moths and honeybees M. nmcfurlanei Conslancc & Rollins Bees, mainly in Apidae; rarely none Bees Barnes 1996 hummingbirds or moths Species Observed floral visitors Moth scales Likely pollinators Evidence M. iiuiliijioni (Torrey) A. Gray Mawkniolhs: Hyh-s liiu'tinr, yes Hawkmoths Pcrs. ob.servations; Cruden carrying pollen; other hawknuilh 1970; Hodges 1995 visitors previously reported: SpliiiiK cliersis and Pholus iichfnitm (pollen on body) M. pudicd Barneby Noctuid moths, no pollen found yes Noctuid moths Pers. observations M. triflom Bcnih. Hummingbirds: Hylocharis NA Hummingbirds Pil/ I97S xanthusii\ carried pollen Genus Selinocarpus A. Gray S. anfiustifolms Torrey None NA Noctuid moths Floral morphology S. chenopodioiili's A. Gray Hawkmoth, long-tongued tly, NA Noctuid moths Pers. observations; bees, Pieris sp. Spellenberg and Dclson 1977 S. lanceolatus Woolon None NA Hawkmoths Floral morphology S. parvifolius (Torrey) Standi. Hawkmoths; Manduca NA Hawkmoths Pers. observations {/uinquemaciddtd, Hyles Uneahr, pollen on Hyles linenui S. purpmimus Heimcrl Hawkmoths: Hyles lineuui yes Hawkmoths Pers. observations; Fowler and Turner 1977 S. undulatus Fowler & Turner None none Noctuid moths Floral morphology Table 2. The phylogenelically independent contrasts that were conducted in CAIC (Purvis and Rambaut 1995), and the significance of these contrasts. Note that for the first two contrasts in the table, type of pollinator was coded in three different ways, and all yielded insignificant contrasts; for these contrasts the range of significance values is shown. *This contrast indicated a significant negative correlation between the two characters. Phylogenelically Independent Contrast P value Total amount volatiles vs. type of pollinator 0.14-0.30 Amount of floral volatiles per (ig floral tissue vs. type of pollinator 0.31-0.48 Amount of floral volatiles per flower vs. perianth length 0.30 Amount of floral volatiles per flower vs. floral mass 0.19 Amount of floral volatiles per ^g flora! tissue vs. % total volatiles from vegetation 0.0018* Table 3. A comparison of the 50% majority rule consensus tree topology inferred using the molecular DNA sequence data set and the majority rule consensus tree topologies inferred from data sets of volatile compounds using coding methods 2-4; presence/absence of compounds, presence/absence of biosynthetic pathways, and the unordered and ordered amount of each floral compounds expressed in terms of pathways. The g, values indicate the amount of hierarchical structure in the data set; the more negative the number, the greater the structure. Tree-to-tree distances from the topology ba.sed on molecular sequence data to each of the other four topologies are symmetric-difference distances calculated in PAUP* (Swofford 2000). For comparison we have also included the tree-to-tree distance between MPTs based on DNA sequence data, which is an average of the pairwise distances between all 36 MPTs. DNA sequence +/- cmpds +/- pthwys Amt pthwys unordered Amt pthwys ordered g, skewness -0.49 -0.42 -0.30 -0.30 -0.36 # Nodes Resolved 14 13 9 15 12 Tree-to-tree distance 5 21 23 29 26 from DNA tree Table 4. Median consistency and retention indices for each parsimony-informative character from each data set when mapped onto the phylogeny inferred from molecular DNA sequence data. The mean consistency (CI) and retention indices (RI) across all data sets were significantly different (Kruskal-Wallis; P Data set CI median (range) Rl median (range) DNA sequence 1 (0.25-1) 1 (0-1) +/- cmpds 0.3.3(0.11-1) 0.33 (0-1) +/- pthwys 0.33 (0.13-0.5) 0.50 (0-0.78) Amt pthwys unordered 0.50 (0.36-1) 0.17 (0-0.60) Ami plhwys ordered 0.31 (0.18-0.79) 0.28 (0-0.58) Table 5. Results ofTempleton's tests for each of the fragrance data sets with character length compared between character optimization onto the tree inferred using DNA sequence data versus the tree inferred using that particular fragrance data set. For n < 25, the probability was determined from Rohlf and Sokal (1995). Abbreviations for data sets are as in Table 3. Data set Pairwise tree comparison Tree length z value P value 2) DNA sequence 534 -4.24 <0.0001 +/- Compounds 480 — — 3) DNA sequence 45 -0.087 >0.10 +/- Pathways 46 — — 4) DNA sequence 140 -2.82 <0.01 Ami pthwys unordered 114 — — 4) DNA sequence 343 -2.64 <0.05 Ami pthwys ordered 295 — — Table 6. The fragrance characters with the highest consistency index (CI) and retention index (RI) (Ci + RI > 1; any character with a CI or RI of I but with a 0 for the other value was excluded) and the lowest CI and RI (CI + RI < 0.39) across data sets 2-4. For the presence/absence compounds the percentage with a high CI and RI was 17%, with an almost equal percentage with a low CI plus RI (16%). Of the eleven parsimony-informative characters coded as presence/ab.sence of biosynthetic pathways, 18% have both a high CI + RI and a low CI + RI. Of the characters coded as unordered floral amounts of biosynthetic pathways, both 6% had a high CI -h RI and 6% had a low CI + RI. When floral amounts of pathways were coded as ordered, 20% had a high CI and RI and 33% had a low CI and RI. Emi.ssion indicates whether the compound or pathway products were mainly produced vegetatively, florally, or both. Data set abbreviations are as in Table 3. 12-Oxabicyclo 9.1.0 dodeca-3,7-diene,1,5,5,8-letramethyI-,IR-( IR@,3E,7E,11R@) Compound/Pathway Emission Patiiway Dataset CI RI CI + RI ^ 1 Farnesol Floral Sesquiterpenoid +/-compound 0.5 0.75 a,p-Eudesmol Veg Sesquiterpenoid -h/-compound 1 1 P-Bourbonene Veg Sesquiterpenoid +/-compound 0.5 0.67 P-cubebene Floral/Veg Sesquiterpenoid +/-compound 0.5 0.67 a-Muurolene Veg Sesquiterpenoid +/-compound 1 I Calamenene Veg Sesquiterpenoid +/-compound 1 1 a-Cubebene Veg Se.squiterpenoid +/-compound 1 1 Oxabicyclododecadiene" Floral/Veg Sesquiterpenoid -h/-compound 0.5 0.67 p-Selinene Veg Sesquiterpenoid +/-compound 0.5 0.5 Junipene Floral/Veg Sesquiterpenoid +/-compound 0.5 0.5 cis-p-Ocimene Floral/Veg Acyclic monoterpenoid +/-compound 0.5 0.67 Compound/Pathway Emission Pathway Dataset CI R1 a-Pinene Floral/Veg Cyclic monoierpenoid +/-conipound 0.33 0.75 cis-Jasnione Floral Jasmonate +/-compound 0.33 0.78 8-Octalactone Floral Lactone +/-compound 1 1 5-Nonalaclone Floral Lactone +/-compound 1 1 y-Caprolactone Floral Lactone +/-compound 0.5 0.5 7-Decen-5-o)ide FJoraJ Lactone +/-compound 0.5 0.5 2-Hexenal Floral/Veg Linolenic acid derivatives +/-compound 0.5 0.67 cis-3-Hexenyl propionate Floral/Veg Linolenic acid derivatives +/-compound 0.5 0.5 Hexanoic acid, 3-hexenyl ester Floral Linolenic ucid derivatives +/-compound 0.5 0.5 Isoeugenol Floral Phenylpropanoid +/-compound 0.5 0.5 Jasmonate Floral Jasmonate +/-pathway 0.33 0.78 Phenylpropanoids Floral Phenylpropanoid +/-pathway 0.5 0.5 Jasmonate Floral Jusmonate Amt pthwys unordered 0.71 0.6 Jasmonate Floral Jasmonate Amt pthwys ordered 0.48 0.58 Phenylpropanoids Floral Phenylpropanoid Amt pthwys ordered 0.79 0.57 Tiglates Floral Isoleucine Amt pthwys ordered 0.58 0.5 CI + RI^O.39 Ethyl benzoate Floral Benzoate +/-compound 0.25 0 Benzyl butanoate Floral Benzoate +/-compound 0.2 0 Compound/Pathway Emission Pathway Dataset CI R1 Benzyl 3-methyl butanoate Floral Benzoale +/-conipound 0.25 0 Hexyl benzoale Floral Benzoate +/-compound 0.2 0 Benzyl benzoate Floral Benzoate +/-compound 0.13 0.22 Cyclohexyl benzoate Floral Benzoate +/-compound 0.14 0.25 cis-Citral Floral Acyclic monoterpenoid +/-compound 0.25 0 Allo-Ocimene Floral Acyclic monoterpenoid +/-compound 0.14 0 Nero! Floral/Veg Acyclic monoterpenoid +/-compound 0.14 0.14 Limonene diepoxide Floral Cyclic monoterpenoid +/-compound 0.2 0 a-Terpinene Floral Cyclic monoterpenoid +/-compound 0.33 0 Limonene Floral/Veg Cyclic monoterpenoid +/-compound 0.33 0 Menthatriene Floral Cyclic monoterpenoid +/-compound 0.33 0 a-Terpinolene Floral/Veg Cyclic monoterpenoid +/-compound 0.14 0.25 a-Bergamotene Floral Sesquiterpenoid +/-compound 0.2 0 cis-3-Hexenyl 2-methyl butanoate Floral/Veg Linolenic acid derivatives +/-compound 0.17 0.17 cis-3-Hexenyl 3-methyl butanoate Floral Linolenic acid derivatives +/-compound 0.33 0 Hexyl butanoate Floral/Veg Linolenic acid derivatives +/-compound 0.33 0 Butanoic acid, 3-hexenyl ester Floral Linolenic acid derivatives +/-compound 0.14 0.25 Indole Floral Anthranilate +/-coiTipound 0.13 0.22 Benzoates Floral Benzoate +/-pathway 0.25 0 Anthranilate Floral Anthranilate +/-pathway 0.13 0.22 Compound/Pathway Emission I'athway Dataset CI R1 Linoleic acid derivatives Florai/Veg Linoleic acid derivatives Ann pthwys unordered 0.38 0 Benzyl ethers Floral/Veg Benzyl ethers Amt pthwys ordered 0.31 0 Salicylates Floral/Veg Salicylate Amt pthwys ordered 0.19 0.15 Linoleic acid derivatives Floral/Veg Linoleic acid derivatives Amt pthwys ordered 0.18 0.1 Cyclic monoterpenoids Floral/Veg Cyclic monoterpenoid Amt pthwys ordered 0.22 0.13 Anthranilate Floral Anthranilate Amt pthwys ordered 0.18 0.18 144 FIGURE LEGENDS Figure 1. One of the most parsimonious trees inferred using nuclear and chloroplast DNA sequence data, and the tree used for ail analyses in this paper. Bootstrap values >50 are indicated above the branches and decay indices are below. The likely pollinators (as shown in Table 1) are listed beside each species' name: Hmoth = hawkmoths. Moth = noctuid moths, Hbird = hummingbirds. Onto this tree has been mapped the average total amount (total unit area) of floral and vegetative volatiles produced per 12 hrs, divided by 10" for ease of presentation. Gray indicates the lowest volatile emission, striped is intermediate emission, and black indicates high volatile emission. The character state at the tree root is unclear. The values for average total amounts of volatiles (standard error) are listed next to each taxon name; as data were available for only one individual of M. jalapa, there is no standard error. Figure 2. The MPT shown in Fig. 1 with the loss of nitrogen-bearing compounds indicated in gray and pollinator abbreviations as in Fig. I. Figure 3. Continuous characters mapped on the MPT in Fig. I using squared change parsimony but with classes of numbers grouped together for ease of illustration. Characters have been mapped onto the phylogeny, and the exact values for each character are listed beside each taxon name along with pollinator abbreviations as in Fig. I. A. The total number of both floral and vegetative volatiles emitted by each species. B. The % of total volatiles produced that were emitted by the vegetation. C. The % of total volatiles produced that were emitted by the flowers. Figure 4. The MPT in Fig. I with losses and gains of particularly informative characters indicated. Pollinator abbreviations beside each taxon name are as in Fig. 1. Open bars indicate a loss, closed bars indicate a gain. F = famesol, J = cis-jasmone, L = 5- octalactone and 5-nonalactone, P = phenylpropanoids, VS = the vegetatively produced sesquiterpenoids a,p-eudesmol, a-muurolene, calamenene, and a-cubebene. 145 Figure 1. 99 y/jimx/. A. obtiisa Hmoth 100(37) A. longiflora Hmoth 554(131) IQQ A. acutifolia Hmoth 611 (185) 8 a. wrightii Hmoth 136 (57) ' A. crassifolia Hmoth 57 (8) S. chenopodioides Moth 174(26) 94 88 S. angustifolius Moth 3 (0.4) 5 2 S. undulatus Moth 2 (0.3) 81 S. lanceolatiis Hmoth 61 (17) 79 I 2 S. parvifolius Hmoth 415 (103) S. purpusianus Hmoth 333 (114) M. alipes Moth 10(2) M. piidica Moth 14(5) 72 • 0 ^ M. macfarlanei Bees 0.2 (0.1) I M. greenei Hmoth? 0.3 (0.1) 1 M. multiflora Hmoth 5 (1) 84 M. triflora Hbird 1 (0.4) 100 M. bigelovii Moth 64(52) 57 100 M. jalapa Hmoih 7 (NA) 6 M. longiflora Hmoth 7 (0.9) 146 A. obtusa Hmoth A. longiflora Hmoth A. acutifolia Hmoth A. wrightii Hmoth A. crassifolia Hmoth 5. angiistifoliiis Moth S. undulatus Moth 5. lanceolatus Hmoth S. parvifolius Hmoth S. purpusianus Hmoth M. alipes Moth M. pudica Moth M. macfarianei Bees M. greenei Hmoth? M. multiflora Hmoth M. triflora Hbird M. bigelovii Moth M jalapa Hmoth Af. longiflora Hmoth 147 Figure 3A. •y/y////A A. obtusa Hmoth 73 A. longiflora Hmoth 47 ifIf A. aciitifolia Hmoth 60 If if r'"" A. wrightii Hmoth 108 rii li if jpiii^iiu^tZ cadsnneManeM:!neann A. crassifolia Hmoth 62 if ri if K M if K HciKinieMMMMinK^^ S. chenopodioides Moth 63 » n K X ISmajj S. angustifolius Moth 36 S. undulatus Moth 39 nnearj gunnnp S. lanceolatiis Hmoth 51 M XM S. parvifoliits Hmoth 63 KXXK! S. piirpiisianus Hmoth 81 M. alipes Moth 29 M. pudica Moth 32 M. macfarlanei Bees 6 M. greenei Hmoth? 5 M. midtiflora Hmoth 30 M. triflora Hbird 24 M. bigelovii Moth 41 M. jalapa Hmoth 21 M. longiflora Hmoth 27 I I = 5-26 •= 26-46 BS = 46-67 67-87 1 = 87-108 148 Figure 3B. A. obtiisa Hmoth 30 A. longiflora Hmoth 8 A. aciitifolia Hmoth 82 A. wrightii Hmoth 82 A. crassifolia Hmoth 0 S. chenopodioides Moth 18 S. angustifoliiis Moth 95 S. iindulatiis Moth 60 S. lanceolatus Wmoih 16 S. parvifolius Hmoth 36 S. piirpusianus Hmoth 3 M. alipes Moth 5 M. pudica Moth 0 M. macfarlanei Bees 45 M. greenei Hmoth? 86 M. multiflora Hmoth I M. triflora Hbird 26 M. bigelovii Moth 4 M. jalapa Hmoth 8 M. longiflora Hmoth 6 •= 0-19 •= 19-38 88 = 38-57 ^=57-76 •= 76-95 149 Figure 3C. y/MM//. A. obtiisa Hmoth 70 A. longiflora Hmoth 92 A. acutifolia Hmoth 18 A. wrightii Hmoth 18 W//M A. crassifoUa Hmoth 100 S. chenopodioides Moth 82 S. angustifoliiis Moth 5 S. iindulatiis Moth 40 S. lanceolatus Hmoth 84 ^ iyyyyyM S. parvifoliiis Hmoth 64 S. purpusiamis Hmoth 97 M. alipes Moth 95 M. pudica Moth 100 M. macfarlanei Bees 55 M. greenei Hmoth? 14 Af. multiflora Hmoth 99 M. triflora Hbird 74 M. bigelovii Moth 96 M. jalapa Hmoth 92 M. longiflora Hmoth 94 ••=5-24 •= 24-43 BB = 43-62 ^=62-81 •= 81-100 150 Figure 4. A. obtiisa Hmoth A. longiflora Hmoth A. aciitifolia Hmoth A. wrightii Hmoth A. crassifolia Hmoth S. chenopodioides Moth S. angmtifoliiis Moth S. iindiilatiis Moth S. lanceolatus Hmoth S. parvifoliiis Hmoth S. purpusianus Hmoth M. alipes Moth M. piidica Moth M. macfarlanei Bees M. greenei Hmoth? M. multiflora Hmoth M. triflora Hbird M. bigelovii Moth M. jalapa Hmoth M. longiflora Hmoth 151 REFERENCES Adams, R. P. 1999. Systematics of multi-seeded eastern hemisphere Jiiniperiis based on leaf essential oils and RAPD DNA fingerprinting. Biochem. Syst. Ecol. 27:709-725. Azuma, H., M. Toyota, Y. Asakawa, R. Yamaoka, J. G. Garcia-Franco, G. Dieringer, L. B. Thien, and S. Kawano. 1997. Chemical divergence in floral scents of Magnolia and allied genera (Magnoliaceae). Plant Species Biol. 12:69-83. Azuma, H., L. B. Thien, and S. Kawano. 1999. Molecular phylogeny of Magnolia (Magnoliaceae) inferred from cpDNA sequences and evolutionary divergence of the floral scents. J. Plant Res. 112:291-306. Barkman, T. J. 2001. Character coding of secondary chemical variation for use in phylogenetic analyses. Biochem. Syst. Ecol. 29:1-20. Barkman, T. J., J. H. Beaman, and D. A. Gage. 1997. Floral fragrance variation in Cypripediuin: Implications for evolutionary and ecological studies. Phytochemistry 44:875-882. Bestmann, H. J., L. Winkler, and O. von Helversen. 1997. Headspace analysis of volatile flower scent constituents of bat-pollinated plants. Phytochemistry 46:1169-1172. Borg-Karlson, A.-K. 1990. Chemical and ethoiogical studies of pollination in the genus Ophrys (Orchidaceae). Phytochemistry 29:1359-1387. Borg-Karlson, A.-K., I. Valterova, and L. A. Nilsson. 1994. Volatile compounds from flowers of six species in the family Apiaceae: bouquets for different pollinators? Phytochemistry 35:11 l-l 19. Brantjes, N. B. M. 1978. Sensory responses to flowers in night-flying moths. Pp. 13-19 in A. J. Richards, ed. The Pollination of Flowers by Insects. Academic Press, London. Croteau, R., and F. Karp. 1991. Origin of natural odorants. Pp. 101-126 in P. M. Muller and D. Lamparsky, eds. Perfumes: Art, Science and Technology. Elsevier. London. Dobson, H. E. M. 1994. Roral volatiles in insect biology. Pp. 47-81 in E. A. Bemays, ed. Insect-Plant Interactions. Vol. V. CRC Press, Boca Raton. Dobson, H. E. M., E. M. Danielson, and I. D. van Wesep. 1999. Pollen odor chemicals as modulators of bumble bee foraging on Rosa rugosa Thunb. (Rosaceae). Plant Species Biol. 14:153-166. 152 Dobson, H. E. M., J. Arroyo, G. Bergstrom, and I. Groth. 1997. Interspecific variation in floral fragrances within the genus Narcissus (Amaryllidaceae). Biochem. Syst. Ecol. 25:685-706. Dudareva, N., B. Piechulla, and E. Pichersky. 2000. Biogenesis of floral scents. Pp. 31-54 in J. Janick, ed. Horticultural Reviews, Vol. 24. John Wiley & Sons, Inc., New York. Faegri, K., and L. van der Pijl. 1979. The Principles of Pollination Ecology, S"' ed. Pergamon Press, Oxford. Fishbein, M., and D. L. Venable. 1996. Diversity and temporal change in the effective pollinators of Asclepias tiiberosa. Ecology 77:1061-1073. Gottsberger, G. 1988. The reproductive biology of primitive angiosperms. Taxon 37:630- 643. Gottsberger, G. 1989. Beetle pollination and flowering rhythm of Annona spp. (Annonaceae) in Brazil. Plant Syst. Evol. 167:165-187. Grant, V. 1983. The systematic and geographical distribution of hawkmoth flowers in the temperate North American flora. Bot. Gaz. 144:439-449. Grayer, R. J., M. W. Chase, and M. S. J. Simmonds. 1999. A comparison between chemical and molecular characters for the determination of phylogenetic relationships among plant families: An appreciation of Hegnauer's "Chemotaxonomie der Pflanzen". Biochem. Syst. Ecol. 27:369-393. Grison, L., A. A. Edwards, and M. Hossaert-McKey. 1999. Interspecies variation in floral fragrances emitted by tropical Ficus species. Phytochemistry 52:1293-1299. Haber, W. A., and G. W. Frankie. 1989. A tropical hawkmoth community: Costa Rican dry forest Sphingidae. Biotropica 21:155-172. Helversen, O. von, L. Winkler, and H. J. Bestmann. 2000. Sulphur-containing "perfumes" attract flower-visiting bats. J. Comp. Physiol. A 186:143-153. Herrera, C. M. 1996. Floral traits and plant adaptation to insect pollination: A devil's advocate approach. Pp. 65-87 in D. G. Lloyd and S. C. H. Barrett, eds. Floral Biology: Studies on floral evolution in animal-pollinated plants. Chapman & Hall, New York. Jakobsen, H. B., and C. E. Olsen. 1994. Influence of climatic factors on emission of flower volatiles in situ. Planta 192:365-371. 153 Kaiser, R. 1993. The Scent of Orchids; Olfactory and chemical investigations. Elsevier, Amsterdam. Kaiser, R., and L. Tollsten. 1995. An introduction to the scent of cacti. Flav. Fragr. J. 10:153-164. Knudsen, J. T., and L. Tollsten. 1993. Trends in floral scent chemistry in pollination syndromes: floral scent composition in moth-pollinated taxa. Bot. J. Linn. Soc. 113:263-284. Knudsen, J. T., and L. Tollsten. 1995. Floral scent in bat-pollinated plants: a case of convergent evolution. Bot. J. Linn. Soc. 119:45-57. Kugler, H. 1971. Zur Bestaubung grossblumiger Datura-Arten. Flora 160:511-517. Miyake, T., R. Yamaoka, and T. Yahara. 1998. Floral scents of hawkmoth-pollinated flowers in Japan. J. Plant Res. 111:199-205. Nilsson, L. A., L. Jonsson, L. Rason, and E. Randrianjohany. 1985. Monophily and pollination mechanisms in Angraecum arachnites Schltr. (Orchidaceae) in a guild of long-tongued hawk-moths (Sphingidae) in Madagascar. Biol. J. Linn. Soc. 26:1- 19. Pellmyr, O., and L. B. Thien. 1986. Insect reproduction and floral fragrances: keys to the evolution of the Angiosperms? Taxon 35:76-85. Pellmyr, O., L. B. Thien, G. Bergstrom, and L Groth. 1990. Pollination of New Caledonian Winteraceae: opportunistic shifts or parallel radiation with their pollinators? Plant Syst. Evol. 173:143-157. Pellmyr, O., W. Tang, L Groth, G. Bergstrom, and L. B. Thien. 1991. Cycad cone and angiosperm floral volatiles: inferences for the evolution of insect pollination. Biochem. Syst. Ecol. 19:623-627. Pichersky, E., R. A. Raguso, E. Lewinsohn, and R. Croteau. 1994. Floral scent production in Clarkia (Onagraceae). Plant Physiol. 106:1533-1540. Raguso, R. A. 1997. Convergent evolution in hawkmoth pollinated flowers? Chemical analysis of floral advertisements and rewards. Int. Chem. Ecol. Soc., Vancouver, BC. Canada. July 1997. 154 Raguso, R. A., and D. M. Light. 1998. Electroantennogram responses of male Sphinx perelegans hawkmoths to floral and 'green-leaf volatiles'. Entomologia Experimentalis et Applicata 86:287-293. Raguso, R. A., and M. W. Willis. 2000. The importance of olfactory and visual cues in nectar foraging by nocturnal hawkmoths. In press in C. L. Boggs, W. B. Watt, and P. R. Ehrlich, eds. Proceedings of the Third Symposium on Butterfly Ecology and Evolution. University of Chicago Press. Raguso, R. A., D. M. Light, and E. Pichersky. 1996. Electroantennogram responses of Hyles lineata (Sphingidae: Lepidoptera) to volatile compounds from Clarkia hreweri (Onagraceae) and other moth-pollinated flowers. J. Chem. Ecol. 22:1735- 1766. Smith, B., and B. J. D. Meeuse. 1966. Production of volatile amines and skatole at anthesis in some arum lily species. Plant Physiol. 41:343-347. Tinbergen, N. 1958. Curious naturalists. Doubleday & Co., Garden City, NY. Vogel, S. 1963. The role of scent glands in pollination: on the structure and function of osmophores. A. A. Balkema, Rotterdam. Waser, N. M., L. Chittka, M. V. Price, N. M. Williams, and J. Ollerton. 1996. Generalization in pollination systems, and why it matters. Ecology 77:1043-1060. Waterman, P. G. 1998. Chemosystematics—Current status. Phytochemistry 49:1175- 1178. Williams, N. H., and W. M. Whitten. 1999. Molecular phytogeny and floral fragrances of male euglossine bee-pollinated orchids: A study of Stanhopea (Orchidaceae). Plant Species Biol. 14:129-136. Wilson, P., and J. D. Thomson. 1996. How do flowers diverge? Pp. 88-111 in D. G. Lloyd and S. C. H. Barrett, eds. Floral Biology: Studies on floral evolution in animal-pollinated plants. Chapman & Hall, New York.