i
“…neither the flower nor the insect will ever understand the significance of their
lovemaking. I mean, how could they know that because of their little dance the
world lives? But it does. By simply doing what they're designed to do, something
large and magnificent happens.”
- John Laroche, from the film Adaptation
ii
Comparison of Chiloglottis valida (left) and Chiloglottis aff. jeanesii (right)
iii
THE EVOLUTIONARY BIOLOGY OF POLLINATION: STUDIES IN A
GENUS OF AUSTRALIAN SEXUALLY DECEPTIVE ORCHIDS
MICHAEL ROBERT WHITEHEAD
JULY, 2012
A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY OF
THE AUSTRALIAN NATIONAL UNIVERSITY
iv DECLARATION
The research presented in this thesis is my own original work except where due reference is given in the text. All the chapters were the product of investigations carried out jointly with others but in all cases I am the principal contributor to the work. No part of this thesis has been submitted for any previous degree.
……………………………………………….. Michael Robert Whitehead July, 2012
v THESIS PLAN
This thesis is presented in six chapters. Figures appear at the end of each chapter. All photographs and images are my own. Chapters 1, 3-6 are either published, submitted for publication or presented as manuscripts intended for submission. As such, the pronoun
“we” is used to represent co–authors in material intended for publication. Below I outline the contribution of my co–authors:
Chapter 1: Integrating floral scent, pollination ecology and population genetics
I performed the literature review, conceptual development and writing. Rod Peakall (RP) contributed written material based on his own work, the conceptual development of the chapter and editorial comments.
Published as article:
Whitehead, M.R. and Peakall, R. (2009) Integrating floral scent, pollination ecology and population genetics. Functional Ecology 23, 863-874.
Chapter 2: Introduction to Chiloglottis: a model system for plant–pollinator studies.
I was sole author and contributor to this chapter.
Chapter 3: Pollinator specificity and strong pre–pollination reproductive isolation in sympatric sexually deceptive orchids.
I was responsible for: field sampling and design, laboratory work, symbiotic orchid culture, writing, statistical analyses.
Christine Hayes and RP contributed to additional chemical and genetic data collection and analyses.
vi Chapter 4: Multiple paternity and outcrossing in self-compatible clonal orchids.
I was responsible for field sampling, study design, symbiotic orchid culture, laboratory work, analysis and writing. RP contributed programming and development of the
GENALEX simulation and paternity exclusion routine.
Chapter 5: Microdot technology for individual marking of small arthropods.
I was responsible for all design, field work, writing and statistical analyses. RP offered editorial comments.
Published as article:
Whitehead, M.R. and Peakall, R. (2012) Microdot technology for individual marking of small arthropods. Agricultural and Forestry Entomology 14, 171-175.
Chapter 6: Short term but not long term patch avoidance in an orchid-pollinating solitary wasp.
I was responsible for design, field work, writing and statistical analyses. Murray Efford, author of SECR software, contributed direction to analyses. RP provided editorial comments.
Whitehead, M.R. and Peakall, R. Short term but not long term patch avoidance in an orchid-pollinating solitary wasp Behavioral Ecology, in revision.
vii ACKNOWLEDGMENTS
I am grateful for the many contributing forces, small and large, that have allowed me to carry out this project. These many influences have all in some way allowed me to learn, develop, teach, write, speak and travel all in the name of my peculiar study plants. I am grateful that I not only got the opportunity to indulge my interest in this field, but that I was able to make it my prime occupation for the last 4.5 years. Among them:
My supervisor Rod Peakall whose expert supervision, unfailing availability, unflagging support, strategic advice, deep scientific expertise and creative accounting have kept me in the game at every stage of the way. Utmost gratitude to you for being a pleasure to work with and generously sharing your first research love.
The Australian Pacific Science Foundation, for supplying primary funding and taking a punt on an honours student. Vic Stephens at DataDot for the positive support.
Christine Hayes for unparalleled lab support. I will be surprised if I encounter such experience and efficiency in the lab ever again, and I am supremely grateful I was able to benefit from it.
Celeste Linde, my mycology guru, tutor and occasional counsellor. Your advice is always refreshing, grounded and highly valued.
Friends and colleagues at UNSW. Bill Sherwin who mentored me. Lee Ann Rollins for always encouraging words, Clare Holleley for sharing her PhD and Emily Miller for the fun and belief in me. I wouldn’t have started without your support from all of you.
Maurizio Rossetto, a mentor and friend. I was more scarce during the project than I would have liked, but your perspective and energy was always valuable.
Emily Miller, Thomas Wallenius, Myles Menz, Yann Triponez, Rachel Slatyer, Kate Griffiths,
Samantha Vertucci, Peri Bolton, Tonya Haff, Hanna Kokko for sharing field work, whether fun or dull! viii Leon Smith for always being generous with his time and practical lab knowledge.
ANU friends; Renee Catullo, Brian Mautz, Jules Livingston, Mitzy Pepper, Sophia Callander,
Anja Skroblin, Andrew Kahn, Jussi Lehtonen, Hanna Kokko, Michael Jennions, Dave Rowell,
Rob Lanfear, Tonya Haff, Isobel Booksmythe, Richie and Melita Milner, Thomas Wallenius,
Renae Pratt, Richard Carter, Dave Moore, Dom Roche, Sandra Binning, Duncan Fitzpatrick.
Biologists are the best people and you’re proof.
Jules Livingston, Michael Jennions, Renee Catullo, Brian Garms, Emily Miller. At a critical point all of these people said something I needed to hear.
Samantha Vertucci, for her love, company and support, especially in these final moments.
She, more than any other, endured the grumpy mood and self–absorption that can be occasional by products of PhD research. Love you kid. You make things better.
My brothers Michael Farrell, Nick Evershed, Matt Faulkner, Ben Zemanek, James Barrett for continued friendship as solid as the Hume Highway is long.
My full-sibs for being my best friends and needing only to appear in my thoughts to make me laugh. You lighten my spirit and I love you both.
Mum and Dad, for encouraging my interests, imbuing me with optimism, lovingly supporting me and making me who I am. I love you.
Thank you finally to my study species. At no point did you bite me, defecate on me, run away from me, require me to stay up late, get up early or burden me in excessive bureaucracy and paperwork. I don’t know why every biologist doesn’t want to work on pollination!
ix PRÉCIS
There are few other structures in nature from which evolution has generated such wide diversity as the flower or inflorescence, and this diversity is commonly attributed to the influence of their animal visitors. By outsourcing their mate choice to pollinators, plants have left themselves—and especially their flowers—subject to the selective forces imposed by the behaviour, cognition and perception of the pollinators that serve them.
The orchids provide some of the most remarkable and extreme examples of adaptations to specific animal pollinators. Perhaps one of the most peculiar of these strategies is sexual deception, whereby male insects are lured to the flower by mimicry of the female sex pheromone. This seemingly unlikely strategy has evolved multiple times independently on different continents in different parts of the orchid phylogeny which raises the question of what adaptive advantages might underlie such a strategy.
This multidisciplinary thesis studies gene flow and pollinator behaviour in two sympatric sexually deceptive orchids in the genus Chiloglottis. The two species attract their specific wasp pollinators through emission of distinct species–specific semiochemicals. Since floral volatiles play a pre–eminent role in pollinator attraction, Chiloglottis provides an excellent case study for examining the interaction between floral volatile chemistry, pollinator behaviour and the evolutionary dynamics of populations.
The thesis begins with a review of floral volatiles and their role in pollinator attraction and plant speciation. The literature is used to develop a research framework of six testable hypotheses under which we might productively explore the influence of floral volatiles on plant evolution. These hypotheses are then explored in the study system over the following chapters.
A study of pollinator specificity, neutral genetic differentiation and floral chemistry demonstrates that the chemical mimicry crucial to sexual deception is responsible for reproductive isolation and potentially even speciation. Mating system and paternity x analysis provide the first genetic evidence for multiple paternity in orchid broods.
Extensive outcrossing is found to predominate and paternity assignment shows evidence for long distance pollen flow supporting the hypothesis that sexual deception promotes outcrossing and so minimizes the potentially the deleterious effects of selfing.
Lastly, an innovative new method is developed for tracking wasps in the field. Application of this technique to a population of orchid-pollinating wasps reveals detailed information about their movement and mating behaviour. The findings support the conclusion that sexual deception is a superb adaptive solution to the problem flowers face of simultaneously attracting pollinators before persuading them to leave quickly.
CONTENTS
Chapter 1 Integrating floral scent, pollination ecology and population genetics. 1
Summary 2
Introduction 3
Population genetics in the context of floral volatiles 6
Future directions 22
Figures 26
Chapter 2 Introduction to Chiloglottis: a model system for plant–pollinator studies. 30
Figures 35
Chapter 3 Pollinator specificity and strong pre–pollination reproductive isolation in sympatric sexually deceptive orchids. 37
Abstract 38
Introduction 39
Methods 42
Results 50
Discussion 53
Figures 60
Chapter 4 Multiple paternity and outcrossing in self-compatible clonal orchids. 67
Abstract 68
Introduction 69
Methods 72
Results 79 Discussion 83
Figures 91
Chapter 5 Microdot technology for individual marking of small arthropods.
102
Abstract 103
Introduction 104
Materials and methods 106
Results 108
Discussion 109
Figures 112
Chapter 6 Short term but not long term patch avoidance in an orchid- pollinating solitary wasp 113
Abstract 114
Introduction 115
Methods 117
Results 123
Discussion 125
Figures 130
Concluding remarks 133
References 137
Appendices 169
Appendix I 169
Appendix II 173
Appendix III 175
Chapter 1 1
CHAPTER 1
INTEGRATING FLORAL SCENT, POLLINATION ECOLOGY AND
POPULATION GENETICS.
Neozeleboria monticola with Chiloglottis valida
Chapter 1 2
SUMMARY
1. Floral scent is a key factor in the attraction of pollinators. Despite this, the role of
floral scent in angiosperm speciation and evolution remains poorly understood.
Modern population genetic approaches when combined with pollination ecology
can open new opportunities for studying the evolutionary role of floral scent.
2. A framework of six hypotheses for the application of population genetic tools to
questions about the evolutionary role of floral scent is presented. When floral
volatile chemistry is linked to pollinator attraction we can analyse questions such
as: Does floral volatile composition reflect plant species boundaries? Can floral
scent facilitate or suppress hybridization between taxa? Can the attraction of
different pollinators influence plant mating systems and pollen-mediated gene
flow? How is population genetic structure indirectly influenced by floral scent
variation?
3. The application of molecular tools in sexually deceptive orchids has confirmed that
volatile composition reflects species boundaries, revealed the role of shared floral
odour in enabling hybridization, confirmed that the sexual attraction mediated by
floral odour has implications for pollen flow and population genetic structure and
provided examples of pollinator-mediated selection on floral scent variation.
Interdisciplinary studies to explore links between floral volatile variation, ecology
and population genetics are rare in other plant groups.
4. Ideal study systems for future floral scent research that incorporate population
genetics will include closely related taxa that are morphologically similar,
sympatric and co-flowering as well as groups that display wide variation in
pollination mechanisms and floral volatiles. Chapter 1 3
INTRODUCTION
Overview
It has been hypothesized that the immense diversity of the angiosperms can be attributed in greater part to biotic pollination (Grant 1994; Galen 1999; Johnson 2006). When floral volatiles play a key role in governing the attraction of pollinators they, like other floral traits, are likely to be subject to pollinator-mediated selection. It follows therefore that studying floral volatile variation promises to greatly improve understanding of the evolution of flowering plants. Despite this importance, as noted by various authors, floral scent as a trait is poorly understood and has been studied less thoroughly than other floral characters, such as visual cues like colour or morphology (Azuma, Thien & Kawano 1999;
Ervik, Tollsten & Knudsen 1999; Dufaÿ, Hossaert-McKey & Anstett 2004; Raguso 2008b;
Waelti et al. 2008). The relative lack of studies of natural variation in plant volatiles is likely due to several factors. These may include on one hand a bias by chemists who may not realize the relevance of chemical variation and a bias by pollination biologists to more obvious visual characters (Raguso 2008c). It may also simply be that studying floral volatile variation requires a rare combination of inter-disciplinary skills in chemistry and biology and specialized equipment not routinely found in an ecology laboratory.
Another notable gap in the literature is the lack of studies that integrate volatile chemistry with pollination biology and genetics. The literature on floral scent is dominated by a focus on biosynthetic mechanisms and descriptive studies of variation across taxa. Yet, the ever expanding toolkit of molecular methods for investigating population genetic variation, plant mating systems and gene flow remains to be applied to volatiles. Taking advantage of these tools and technology will allow floral volatile research to be linked to plant reproductive biology and the selective forces of population-level differentiation and speciation. Questions about the consequences of specialized pollination on pollen flow within and among populations, the influence of different guilds of pollinators on pollen Chapter 1 4 dispersal and mating system, and the influence of floral scent on reproductive isolation are all open to inquiry by employing these molecular tools.
The overarching goal of this review is to highlight the benefits of—and provide a framework for—combining volatile, ecological and genetic knowledge to illuminate the evolutionary role of floral volatiles as pollinator attractants. First we provide a brief background on the study of floral volatiles as specific pollinator attractants. We then explore six hypotheses that integrate floral volatiles, genetics and pollination ecology. We highlight studies that address components of these hypotheses and we point to relevant gaps in knowledge. Finally, we consider ‘ideal study systems’ for interdisciplinary research. We believe that exciting new insights in the functioning of floral scent in pollination systems await future investigations that combine research in plant volatile chemistry, pollination ecology and population genetics.
A focus on the link between volatile variation and functional specialization
While selection imposed by pollinators has likely played a key role in influencing the patterns of floral scent compound variation across the angiosperms (Knudsen & Tollsten
1993), floral scent is not limited in its role to pollinator attraction. Floral volatiles may fulfil other adaptive functions (Piechulla & Pott 2003; Raguso 2008c), for example as repellents (Omura, Honda & Hayashi 2000; Kessler, Gase & Baldwin 2008) or physiological protection against abiotic stresses (Dudareva et al. 2006; Knudsen et al. 2006). As well as these roles outside pollination, qualitative and quantitative variation in floral volatile profile can be driven by both phylogeny and environment.
Phylogeny can constrain a flower’s biosynthetic repertoire in that the potential for evolution of floral scent profiles is contingent on the variation already present within a lineage. Furthermore, phylogenetic inertia may conserve volatile compound production in the absence of an apparent, present biological function so that floral odour compound Chapter 1 5 variation may be inherited regardless of pollinator preference. The potential for plant secondary compounds to carry phylogenetic information has long been recognised and once was perceived as a promising source of data for botanical systematists (Alston,
Mabry & Turner 1963). Floral volatile variation is now being shown to be too labile an evolutionary character to be useful for constructing phylogeny. Nonetheless, some phylogenetic patterns have been recognised in Nicotiana (Raguso et al. 2006), Cypripedium
(Barkman 2001), and the Nyctaginaceae (Levin, McDade & Raguso 2003) and a recent study on Ophrys orchids has demonstrated utility of chemotaxonomic analysis of non- pollinator-attractive floral volatile compounds in reconstructing phylogeny in that genus
(Gögler et al. 2009).
Intraspecific variation in floral odour blends may arise through genetic drift under relaxed selection, introgression of traits through hybridization, pleiotropic effects of biosynthetic pathways, or environmentally driven phenotypic plasticity (Raguso 2008c). However, identifying the basis of intraspecific floral volatile variation for specific cases in nature remains difficult (Ackerman, Melendez-Ackerman & Salguero-Faria 1997; Azuma, Toyota
& Asakawa 2001; Knudsen 2002; Schlumpberger & Raguso 2008).
In light of this complex odour variation, Raguso (2001) concluded that the evolution of floral scent is a “mosaic product of biosynthetic pathway dynamics, phylogenetic constraints, and balancing selection due to pollinator and florivore attraction”. Because of this, pollination biologists studying fragrance face the challenge of sorting ‘signal’— bioactive compounds potentially subject to pollinator mediated selection—from ‘noise’ of other types of chemical variation (Raguso 2008c). Approaches for addressing this challenge are discussed in more detail later.
While floral odour chemical composition has been documented for hundreds of species
(Knudsen et al. 2006), the links between distinct signals and receivers - odour and pollinator - are known for only a very small subset of these studied species, most of which involve specialized relationships. Notwithstanding the predominance of generalized over Chapter 1 6 specialized plant-pollinator systems and much discussion of their relative importance to floral evolution (see Waser et al. 1996; Waser 1998; Waser 2001; Johnson 2006; Ollerton,
Armbruster & Vazquez 2006; Ollerton et al. 2009) this limited background knowledge necessitates an intentional restriction of this review to ‘functionally specialized’ systems where the links between odour and pollinator are (at present) more likely to be meaningfully investigated.
The “functional group” concept classifies pollinators into groups that behave similarly with respect to the flowers they visit (Fenster et al. 2004). Plants that are pollinated by a functional group comprised of a single pollinator species occupy the highly specialized end of the specialist-generalist pollination spectrum, while a plant species pollinated exclusively by a functional group such as nocturnal moths, while still classed as functionally specialized, would occupy a less extreme position on the spectrum. Fenster et al. (2004) hypothesise that the selection pressure exerted by a functional group of pollinators is responsible for ‘pollination syndromes’, the occurrence of suites of traits adapted to particular modes of pollination, e.g. white, fragrant flowers of nocturnal moth- pollinated plants.
The literature on floral-fragrance mediated plant-pollinator interactions is dominated by functionally specialized cases and therefore our review necessarily reflects this bias.
POPULATION GENETICS IN THE CONTEXT OF FLORAL VOLATILES
In specialized pollination systems the distinct ecology and behaviour of a plant’s functional group of pollinators will influence pollination rates, mating system and the extent and distribution of pollen movement. Thus when pollinator specificity is due (at least in part) to floral fragrance (Waelti et al. 2008) and specific pollinators influence these fundamental elements of plant reproductive ecology, floral volatile variation may be expected to influence the genetic composition and evolution of plant populations. For Chapter 1 7 example, changes in floral volatiles that lead to pollinator switching may result in changing patterns of gene flow which in turn may influence the population genetic structure of the species. Presently, these potentially important links between plant volatiles and population genetic structure are poorly understood for any pollination system.
Perhaps the most thorough application of molecular tools in a floral-volatile context has been in the study of sexually-deceptive terrestrial orchid pollination systems in Europe and Australia. Sexual deceit pollination relies on the mimicry of sex-pheromones to attract pollinators (Schiestl 2005) and is characterized by highly specific plant-pollinator relationships which have long been proposed to act as an ethological isolating mechanism between sympatric taxa (Paulus & Gack 1990; Grant 1994). Because this system has benefited from significant collaborative effort between ecologists, chemists and geneticists, a number of examples throughout this review will be drawn from this system.
Where possible we also provide examples from other systems, although in many cases complete data are lacking. For example, some cases of well characterized volatile variation have limited supporting pollination ecology and no genetic data. In other cases, genetic patterns and pollination biology are well characterized, but knowledge of volatile composition and variation is lacking. It should be stressed that the bridges between floral scent chemistry, pollination biology and genetics can be built from any direction. Genetics can be revealing when applied to systems for which volatiles and pollinators have been well studied, while establishing population genetic patterns for a poorly studied species may provide evidence of interesting pollination phenomena.
Below, we discuss in turn six basic hypotheses we believe to have utility in the study of floral scent variation within taxa (Table 1). At the most basic level the hypotheses could apply to a study-system of sister taxa for which floral odour plays a key role in the attraction of contrasting functionally specialized pollinator guilds (Figure 1). Under the scenario in Figure 1, the interaction between floral volatiles, pollination ecology and population genetics can operate in different ways, and at different levels (species, Chapter 1 8 populations, individuals). Combining knowledge of volatile variation with ecological and genetic knowledge is essential to fully address these hypotheses. Below we consider the six hypotheses in turn.
Hypothesis 1: Volatiles and specific pollination
The first hypothesis: ‘Pollinator specificity is due to the distinct volatile composition of the floral fragrance blend’ is perhaps the most critical and most difficult to satisfactorily test.
The aims here are threefold: to determine pollinator specificity, to confirm distinct volatile composition and to obtain evidence for the activity of floral volatile components as pollinator attractants. In most cases to date, even for this seemingly straight-forward hypothesis, there are few studies that have investigated all three lines of evidence.
Despite the modest but expanding bank of information on floral volatile variation within and among species (Knudsen et al. 2006), definitive proof of the links between such variation and associated pollinators remain scarce. Demonstrating the link between specific volatiles and specific pollinators normally requires multiple lines of evidence.
Ideally this will include field observations or experiments demonstrating a role for floral odour in pollinator attraction followed by evaluation of physiologically active constituents, characterization and synthesis of those compounds and behavioural testing via bioassay to confirm activity in pollinators (Schiestl & Marion-Poll 2002; Schiestl & Peakall 2005;
Franke et al. 2009). Genetic confirmation that pollinator specificity is associated with distinct plant entities can be important, particularly for morphologically similar taxa.
Even before identification of active floral volatile constituents, behavioural experiments should be the first step in examining the importance of scent in pollinator attraction. For example, Okamoto, Kawakita & Kato (2007) confirmed the role of floral scent in the obligate nursery-pollinated Glochidion using choice experiments. When the pollinating
Epicephala moths were exposed to air flowing from a pair of bags via a Y-tube they Chapter 1 9 responded only to the air from bags containing flowers of their host Glochidion species.
There was no response to flowers of a co-ocurring non-host Glochidion species or to empty control bags. In this way, behavioural bioassays provide elegant and powerful tools for confirming that volatiles play a key role in pollinator attraction.
Once a role for floral odour in pollination attraction has been confirmed, gas chromatography with electroantennographic detection (GC-EAD) offers a powerful tool for identifying the compounds detected by pollinators from the myriad of volatile variation that may be produced by a flower. This method combines gas chromatography
(to separate the blend into single constituents) with electroantennographic detection that allows the physiologically active compounds detected by insects to be determined. GC-
EAD active compounds are usually subsequently identified by GC-MS and other diagnostic procedures (Schiestl & Marion-Poll 2002). Complementary to the chemical analysis approach of GC-EAD are field or lab-based bioassays to determine the biological activity of putative floral signals. These behavioural experiments typically use synthetic versions of identified floral volatiles in order to show pollinator attraction in vivo. As well as demonstrating the role of fragrance in pollinator attraction, experiments for confirming attractant activity are crucial for distinguishing signals of attraction from the background chemical mosaic which may include chemical signals with other functions, for example those involved in repelling herbivores or plant defensive signalling (Raguso 2008c).
The well-studied sexually deceptive orchids of Australia and Europe represent two systems in which GC-EAD, GC-MS and other analytical procedures have identified volatile organic compounds whose activity as attractant in the field has been subsequently confirmed through multiple lines of experimental evidence. The attraction of pollinators to specific odour bouquets in the sexually deceptive orchid genus Ophrys has been explored for several species and more than 50 pollinator-active compounds have been discovered
(Schiestl et al. 1999; Ayasse et al. 2000; Stökl et al. 2005; Ayasse 2006; Paulus 2006). Some of these compounds have been identified as commonly occurring molecules including Chapter 1 10 esters, aldehydes, alkanes or alkenes, with specificity determined by ratios of the various compounds that mimic sexual signals of female Andrena bees (Ayasse et al. 2000; Schiestl
& Ayasse 2002; Stökl et al. 2005). The important role of specific components for controlling pollinator specificity has been demonstrated by behavioural experiments utilizing synthetic components of the floral bouquet (Ayasse et al. 2000; Ayasse et al.
2003). In Australian sexually deceptive Chiloglottis orchids pollinated by male thynnine wasps, a previously undiscovered class of natural products (rather than blends of common compounds) provides the chemical basis for pollinator specificity (Schiestl et al. 2003;
Franke et al. 2009). The first of these compounds to be described, “chiloglottone 1” (2- ethyl-5-propylcyclohexan-1,3-dione), was confirmed as both the female thynnine wasp sex pheromone and the orchid pollinator attractant (Schiestl et al. 2003). Subsequent study has revealed other chemical variants of this new class of compounds are involved as specific attractants in other orchid pollinator interactions (Franke et al. 2009).
The euglossine bee-pollinated neotropical orchids have probably the longest history of study into odour-mediated pollinator specificity. These orchids lack nectar and attract male euglossine bee pollinators by floral odour. The bees accumulate volatile compounds from the orchids and other floral and non-floral sources to build complex and often species-specific bouquets presumably to attract mates (Kimsey 1980; Eltz, Roubik &
Lunau 2005; Eltz, Ayasse & Lunau 2006). In a landmark paper, Dressler (1968) explored hypotheses on the role of odour in pollination, specificity, reproductive isolation and speciation in euglossine bee-pollinated orchids. Surprisingly, despite the early start, much remains to be learnt about this system. A number of euglossine-pollinated orchid species have been the subject of floral volatile analysis and these studies indicate that specific pollinator attraction is probably conferred by a complex cocktail of chemical components
(Hills, Williams & Dodson 1972; Whitten & Williams 1992; Cancino & Damon 2007).
Behavioural tests with synthetic chemicals have demonstrated the attractiveness of individual components of the floral blend and the influence of volatile mixtures on attractivity (Williams & Dodson 1972; Ackerman 1983). These findings have only recently Chapter 1 11 been augmented by GC-EAD and electroantennography (EAG) which provide evidence that both antennal and central nervous system processes play a role in the specific attraction of bees to odours (Schiestl & Roubik 2003; Eltz & Lunau 2005; Eltz et al. 2006). It is apparent we are only beginning to understand the role of floral volatiles in this complex and diverse tropical orchid-pollinator interaction.
The number of plant-pollinator relationships for which floral volatiles have been identified and demonstrated as attractants is small (Table 2) and dominated by deceptive and obligate nursery-pollination systems. Outside sexual-deception, perhaps one of the best examples of an integrated approach to linking volatiles to pollinator is provided by
Brodmann et al. (2008), who showed volatiles emitted by the flowers of the orchid
Epipactis helleborine to illicit the attraction of social Vespula wasps in a study that combined tests of natural and synthetic compounds in behavioural assays with GC-EAD confirmation and identification of active floral volatile compounds. Not to be neglected in future studies are those pollination systems outside of these close plant-pollinator relationships where specialization is less extreme. For example, pollination by oligolectic bees (which collect pollen from only one plant species or genus) is common in some parts of the world (Proctor, Yeo & Lack 1996; Dotterl et al. 2005) yet the floral cues involved in these relationships have barely been studied.
Hypothesis 2: Volatiles and species boundaries
Despite the early discovery that different floral odour composition reflected the delineation of species and their respective pollinators in Catasetum orchids (Hills et al.
1972), there appear to be few studies that evaluate the role of distinct volatile composition in maintaining species boundaries. We therefore draw on our own ongoing studies to provide an example. An interesting feature of the Australian sexually deceptive
Chiloglottis orchids is the high frequency of temporally and spatially co-flowering congeneric species (Bower 1996; Peakall et al. 1997; Mant et al. 2005a; Mant, Peakall & Chapter 1 12 Weston 2005c). One pair of species that can be found co-flowering is C. valida and an undescribed species morphologically similar to C. jeanesii (hereafter C. aff jeanesii).
Chiloglottis valida is known to attract its male thynnine pollinator Neozeloboria monticola with the single volatile compound, chiloglottone 1 (2-ethyl-5-propylcyclohexan-1,3-dione)
(Schiestl et al. 2005). Similarly, evidence including GC-EAD, GC-MS and bioassays with synthetic compound has revealed that C. aff jeanesii attracts its undescribed pollinator (a
Neozeloboria species in the impatiens species complex) by a structural isomer of chiloglottone 1, called chiloglottone 3 (2-butyl-5-methylcyclohexan-1,3-dione) (Franke et al. 2009; Peakall unpublished). Field studies have shown no cross-attraction between the two compounds and their respective pollinators.
To test whether chemical composition corresponds with species boundaries we applied
GC-MS with selective ion monitoring (SIM, reduces the detection threshold several orders of magnitude and provides the most sensitive measurement of a compound’s presence or absence) of single orchid labella to identify the active compound in flowers from mixed populations of the two taxa. The results of the floral chemistry indicated that our own diagnosis of species based on morphology and conducted in the field was frequently incorrect (Peakall unpublished). Subsequently, chloroplast DNA analysis with the taxa defined only by their chemical composition revealed extensive genetic differentiation between these chemically-defined taxa (Ebert, Hayes & Peakall 2009a). Thus, this study confirms the hypothesis that distinct floral volatile composition should reflect taxonomic boundaries between morphologically similar species when volatiles function as specific pollinator attractants.
The combination of floral volatile analysis and population genetic analysis can sometimes provide unexpected insights into the nature of species boundaries. Mant et al. (2005b) investigated the patterns of odour and genetic variation among several species of sexually deceptive Ophrys. The odour analysis indicated a previously unknown or cryptic taxon that was characterized by distinct odour composition. Remarkably, this entity was not Chapter 1 13 distinct genetically, at least at the level of the nuclear micosatellite loci investigated, nor were its non-active odour compounds distinct from those of related Ophrys species. This discovery in Ophrys may well represent an incipient taxon in the early phases of pollinator mediated speciation with as yet little or no accumulated differentiation evolving at neutral traits not under selection. It now remains to be experimentally confirmed in the field that the odour differences are directly linked to specific pollinators and that hybridization is minimal or absent as a consequence. This example thus represents a case where floral volatile and genetic knowledge is in hand, but now requires further integration with pollination ecology.
Exploring the role of floral scent in taxonomic boundaries (hypothesis 2) may be approached from different directions. Volatile studies indicating strong odour differences among taxa should seek to integrate ecological and genetic data. Similarly, genetic studies revealing unexpected taxonomic boundaries, particularly among closely related sympatric taxa, should consider whether floral odour variation could be linked to specific pollination and reproductive isolation.
Hypothesis 3: Volatiles, hybridization and population genetics
When floral odour is the major determinant of pollinator specificity, changes or variation in floral odour could break down specificity and increase the frequency of interspecific pollen transfer, thereby promoting hybridization and introgression. Alternatively, when hybridization is detected in otherwise highly specific systems, it is of interest to investigate the role floral odour may or may not play in enabling hybridization.
A direct but unexpected link between floral odour and hybridization can be found in
Chiloglottis orchids. Extreme pollinator specificity is the norm in these orchids with putative hybrids rarely reported (Peakall et al. 1997). One exception is Chiloglottis X pescottiana which, when described, was hypothesized to be a hybrid between C. Chapter 1 14 trapeziformis and C. valida. Allozyme based genetic analysis subsequently confirmed the hybrid status of this taxon (Peakall et al. 1997). GC-EAD analysis and field bioassays further confirmed that both orchid parents employ the same single volatile compound, chiloglottone, to attract their respective and phylogenetically distinct pollinators (Schiestl et al. 2003; Schiestl et al. 2005). By virtue of this shared compound, hybridization between the two taxa (due to pollinator sharing) can occur when their usual geographically and altitudinally separate ranges occasionally overlap (Peakall et al. 2002). This case provides an example of the power of combining floral volatile analysis with ecological and genetic methods to better understand the role of floral volatiles in hybridization.
There are more documented cases of hybridization and introgression in the sexually deceptive Ophrys orchids of Europe compared to Australian sexually deceptive orchids
(Soliva & Widmer 2003; Mant et al. 2005b). This may be due, at least in part, to the differences associated with the chemical basis for pollinator attraction in Ophrys. The volatile blend of Australian sexually deceptive orchids is typically characterized by one, two or three unique active compounds (Mant et al. 2002; Schiestl et al. 2003; Schiestl et al.
2005; Franke et al. 2009) while in some of the better-studied Ophrys taxa, specific pollinator attraction is based not on a single chemical odour compound but on emission of distinct blends or ratios of several commonly occurring hydrocarbon compounds (Schiestl et al. 1999; Stökl et al. 2005). Hybridization due to a breakdown of pollinator specificity may occur between some Ophrys more frequently because variation in blends and ratios could result in a floral bouquet more closely resembling that of a sympatric species (Stökl et al. 2008).
An interesting example of marked scent differences among species is found in the genus
Silene (Caryophyllaceae). Waelti and co-workers (2008) investigated floral odour in white and red campions (Silene latifolia and Silene dioica respectively) which are known to be interfertile and to co-occur in parts of their range. GC-MS of floral headspace samples showed distinct odour differences in the relative amounts of biologically active volatile Chapter 1 15 compounds. In a field experiment the biologically active benzenoid phenylacetaldehyde
(which dominated the scent of S. dioica and contributed strongly to the odour difference between species) was applied to inflorescences of both species to make floral fragrance more similar. Transfer of fluorescent dye (a pollen analogue) was higher in plots containing scent-manipulated flowers than control plots of unmanipulated inflorescences.
Thus, odour differences reduce the potential for gene flow between these species demonstrating the importance of odour for reproductive isolation. This work on volatile variation among species was further supported by an in-depth genetic study of natural hybrid zones among the same two Silene species providing new and detailed insights into the evolutionary role of introgression and hybridization more generally (Minder,
Rothenbuehler & Widmer 2007; Minder & Widmer 2008).
The extensive experimental inter-disciplinary work on the Silene system illustrates the interpretive power of integrating floral fragrance analysis, population genetics and pollination ecology. Furthermore, while pollination in this system appears to be functionally specialized, it does not represent a case of extreme specialization like sexual deception or nursery pollination. It is therefore apparent that many other less specialized systems that involve related co-flowering taxa may be candidates for exploring the links between volatiles, hybridisation and population genetics. Such systems may offer the opportunity to investigate whether hybridization is more or less common in species with distinct floral volatile blends and whether attraction due to floral fragrance is reduced or maintained in F1 hybrids between taxa with different floral scents.
Hypothesis 4: Volatiles, pollination and plant mating systems
Differences among distinct pollinators in behaviour and abundance can be expected to influence the way pollen is moved within and among individual plants and populations.
Therefore, in those systems where floral scent governs pollination specificity, either through innate attraction, floral constancy (see Wright and Schiestl in this feature) or Chapter 1 16 filtering visitor composition, it is likely that floral scent variation will indirectly play a key role in moderating plant mating systems (the degree of selfing versus outcrossing). To our knowledge there are no studies that have directly linked variation in floral volatile composition to plant mating systems.
A study by Brunet and Sweet (2006), although not directly linked to plant volatiles, provides a rare example of the application of genetic methods for testing the effects pollinators have on the populations they service. This study investigated the effect of different insect pollinators on outcrossing rates in the Rocky Mountain columbine,
Aquilegia coerulea; a protandrous, self-compatible herb. Many hours of pollinator observations at eight natural populations over three years revealed considerable variation in the relative proportions of different pollinator species among populations. Outcrossing rate estimates, achieved by analysis of seed at five allozyme loci, showed an increase in outcrossing rate with hawkmoth abundance. No effect on mating system was detected for any other pollinator group. This appears to be one of the first studies directly linking different pollinators to outcrossing rates. One explanation for the high outcrossing rates achieved by hawkmoth pollination was that hawkmoths reduced geitnogamous selfing
(self-pollination between flowers on the same plant) by preferring to visit female-phase flowers before male-phase flowers. Alternatively, hawkmoths may simply be more effective pollinators. Consequently, A. coerulea populations with low hawkmoth abundance might experience pollen transfer limitation and higher rates of autogamous selfing as reproductive assurance.
In one of the few other examples of studies explicitly investigating the influence of distinct pollinators on plant mating systems, Whelan, Ayre and Benyon (2009) examined pollination by birds and honey bees in an Australian shrub, Grevillea macleayana. In their experiment they caged some inflorescences to exclude vertebrate pollinators and included in the study one population known to have a high rate of outcrossing. Birds were found to not only deposit more pollen per visit than bees in the high outcrossing population, they Chapter 1 17 also moved longer distances between plants and visited fewer inflorescences on a single plant.
Other relevant insights into how pollinators affect mating system have emerged from research on floral specialization in bees. Pollinator effectiveness, ‘the single-visit contribution by a flower-visitor to the reproductive fitness of a plant’, was compared among specialist bees and generalist pollinators of Knautia arvensis by Larsson (2005).
While specialist bees deposited more pollen per visit (higher pollinator effectiveness), their impact on overall pollination success was moderated by their lower abundance relative to generalist pollinators. Thus, higher reproductive success and outcrossing might result when specialist bees are in high abundance. By contrast, when generalist pollinators are in high abundance pollen limitation through wasted interspecific pollen transfer may occur. Pollinators may therefore have an impact on plant mating systems, and floral scent may indirectly influence plant mating system through its interaction with pollinator fauna.
Floral scent chemistry may influence plant mating systems by influencing pollinator behaviour. The study of Kessler et al. (2008) on the floral fragrance attractant, benzyl acetone, and the nectar-borne repellent nicotine present in the flowers of self-compatible
Nicotiana attenuata offers a novel example. Field experiments with transgenic plants deficient for benzyl acetone synthesis, nicotine synthesis or both demonstrated that outcrossing rates were highest in wild-type plants. This may be due to moderation of the attraction by benzyl acetone by the repellent nicotine that limited the time pollinators spent at any one flower and maximised total number of flower visits.
Even low rates of outcrossing can provide benefits to plants with life-histories such as those with low rates of recruitment (Raguso 2008a). As such, a plant’s mating system can be an important factor in the evolution of plant populations. Determining the role pollinators and their behaviour play in moderating plant mating systems and the extent to which this is mediated by floral volatile variation will no doubt provide interesting insights into the evolution of floral scent. Chapter 1 18
Hypothesis 5: Volatiles, pollinators and pollen flow
The spatial patterns of pollen movement determine neighbourhood size and inbreeding rates (Mitchell et al. 2009) and are critical in understanding important evolutionary processes such as population differentiation and speciation. It is well established that pollinator behaviour can control the pattern and extent of pollen dispersal (Richards
1986). For example, nocturnal moth pollinators of Silene alba transport a fluorescent dye pollen analogue further on average than bees (Young 2002). Floral volatiles therefore, through their attraction of different pollinators and influence on pollinator behaviour, could exert an indirect influence on pollen movement within and between plant populations.
In sexually deceptive orchids (where the strong relationship between floral odour and specific pollinator has been repeatedly demonstrated) pollinator behaviour and movements may be controlled by optimal mate seeking strategies potentially leading to quite different patterns of pollen flow compared with other pollination systems (Peakall &
Beattie 1996). Two ecological approaches have been taken to investigate pollen flow in sexually deceptive orchids: mark-recapture of pollinators to infer potential pollen movement, and direct measurements of pollen flow by tracking the movements of coloured pollen. In the Australian Caladenia tentaculata, longer distance pollen flow is promoted by the male thynnine pollinator’s avoidance of visits to more than one flower in a patch. Pollen movements approximate a linear rather than a leptokurtic distribution
(mean distance - 17 m; maximum: 58 m) and mirror movements detected by mark- recapture of the pollinator (Peakall 1990). In Drakaea glyptodon mark-recapture of male thynnine pollinators suggests pollen flow could exceed 130 m (Peakall 1990).While near- neighbour pollination may be avoided in male thynnine pollination systems, pollen flow distances will be bounded by the mate search area. Therefore, the type of pollinator exploited by a sexually deceptive orchid may constrain the maximum pollen flow distance. Chapter 1 19 A mark-recapture study of Colletes cunicularius, a bee pollinator of Ophrys, revealed that individual male bees patrol a specific and restricted portion of the total nesting area in search for mates (mean-recapture distances of 5 m, max 50 m). This behaviour may be expected to limit rather than promote long-distance pollen flow in Ophrys orchids (Peakall
& Schiestl 2004). By contrast, while presently unknown, longer distance pollen movements may well occur in those plant species visited by foraging female Colletes bees.
It appears intuitively reasonable that in the classic euglossine trapline pollination (Janzen
1971) and perhaps in some fig-wasp pollination systems (Nason, Herre & Hamrick 1998) that long range volatile mediated attraction of pollinators will result in long-range pollinator movement and likely long distance pollen flow. If so, such cases will demonstrate a clear link between volatiles and pollen flow. There is little enough research examining and comparing landscape-level gene flow for different pollinators (Mitchell et al. 2009) let alone drawing the link to floral volatiles. Such research, while technically challenging, is now very achievable and will be best realised by combining volatile knowledge, pollination ecology (e.g. pollinator mark-recapture) and genetics (e.g. paternity analysis).
Hypothesis 6: Volatiles and population genetic structure
If different distinct pollinators can be expected to influence both plant mating systems
(hypothesis 4) and the patterns and extent of pollen flow at the population scale
(hypotheses 5), it follows that any differences in plant mating system and pollen flow may in turn influence population genetic structure - the patterns and extent of genetic variation within and among populations. In this way, floral volatiles may have indirect interactions on population genetic structure through their interaction with pollinator fauna.
Hughes et al. (2007) have explored the potential impact of bird versus fly pollinators on the population genetic structure of two South African species of Streptocarpus. Lower Chapter 1 20 levels of genetic differentiation (based on both nuclear and chloroplast DNA analysis) were detected in the sunbird pollinated S. dunii compared to its long-tongued fly pollinated congener S. primulifolius. This was attributed to the greater vagility and wider distribution of the sunbird that likely facilitates greater population connectivity than that possible by fly pollination. While it was recognised that this conclusion may be confounded by differences in habitat between the two study species, this study highlights a potential impact of pollinator behaviour on population genetic structure. Although no information on floral volatile differences was reported it has been noted by others that ornithophilous flowers often have little odour in comparison to other biotic pollination systems (Knudsen et al. 1993; Levin, Raguso & McDade 2001; Raguso et al. 2003). Thus floral volatile differences may indirectly contribute to the population genetic differences between the species.
Population genetic structure in plants is determined by the interaction of multiple factors including mating system, gene flow (both contemporary and historic) by pollen and seed, as well as past population events such as bottlenecks, local extinction and range expansions. A major challenge in linking plant volatile variation to population genetic structure is the need to disentangle these multiple factors. This potential for population genetic structure to be driven by multiple factors can also cloud determination of cause and effect when studying its links to phenotype (e.g. floral scent) and gene flow. The closely related species Clarkia breweri and C. concinna partially overlap in range and conform to the parent-offspring style of rapid speciation due to extreme selection in ecologically marginal populations well characterised in the genus (Lewis 1962).
Furthermore, the derivative species, C. breweri, in contrast to its unscented progenitor, C. concinna shows a recent evolution of floral scent production and moth pollination in a largely unscented genus (Raguso & Pichersky 1995). Given that strong selection can dramatically reduce effective population size, with the accompanying founder effects it is conceivable that traits such as floral scent might experience rapid change, elevating rare Chapter 1 21 alleles for floral phenotypes to high frequency by chance within the same genomes as traits under strong selection for fitness (for example drought tolerance).
Given this complexity, careful study design is required in order to be able to definitively identify pollinator-mediated selection as a driver of between-species floral volatile differences. Mant et al. (2005b) compared floral odour variation in both putatively selected pollinator-active compounds and non-pollinator-active floral volatile compounds within and among Ophrys species. Population genetic data for neutral markers was also obtained for the same set of samples. In order to enable a meaningful and comparable contrast between odour (both active and non-active components) and genetic data, (Mant et al. 2005b) adapted the Analysis of Molecular Variance (AMOVA) framework for the analysis of odour. Although initially developed for molecular data, this procedure can be applied to the hierarchical analysis of variance for any data set that can be input as a pairwise individual by individual distance matrix. The study found significant floral odour differentiation among allopatric populations within species, among allopatric species and among sympatric species. Active odour compounds were more strongly differentiated among allopatric conspecific populations than non-active compounds. In marked contrast, there was limited population or species level population genetic differentiation. It was concluded that the strong odor differentiation but lack of genetic differentiation among sympatric taxa indicated selection imposed by the distinct odour preferences of different pollinating species. This conclusion was reinforced by the low genetic differentiation observed within species that suggested large effective population sizes and therefore little opportunity for genetic drift to account for the observed patterns (Mant et al. 2005b). The methods developed and executed in this study may serve as a model for future studies that seek to explore the direct or indirect links between floral odour variation and population genetic structure.
Chapter 1 22
FUTURE DIRECTIONS
In this review we have explored six hypotheses that directly or indirectly link floral volatile variation, pollinator ecology and population genetics. We have highlighted gaps in knowledge and demonstrated that new evolutionary insights can be achieved by combining these often separate fields of research. In this final section we briefly explore
‘ideal study systems’ for multidisciplinary study of the evolutionary role of floral scent. We recommend that targeting these ideal study systems will greatly accelerate discovery and understanding.
A major hurdle in the application of genetic techniques to studying wild populations is the availability of suitable genetic markers (Ebert & Peakall, 2009). The development of genetic markers is costly and time-consuming thus plant groups for which genetic methods have already been developed will offer more tractable systems for population genetic analysis (see Table S2 in Supporting Information). For example, the Solanaceae and Rosaceae as well as having a good record of floral volatile study have undergone much genetic research effort (equivalent to other economically important plant families) and as such provide good targets for identifying ideal study systems.
Ideal systems for studying the pollination consequences of floral fragrance variation must contain controls for other confounding abiotic and biotic variables (Table 3). Abiotic factors that could influence gene flow between taxa (and potentially contribute to reproductive isolation) include geographic distance or vicariant barriers, microhabitat selection and phenology of flowering. The effects of geographical and temporal isolating mechanisms can be eliminated by studying sympatric, co-flowering taxa which must by their nature share a common pollinator pool.
Biotic factors that may confound the detection of floral scent mediated processes include all forms of floral phenotypic variation other than floral volatiles. Grant (1994) identified various forms of mechanical and ethological isolation that rely on floral morphology. As isolating mechanisms the differential placement of pollen on a pollinator’s body, a flower’s Chapter 1 23 physical exclusion of a class of pollinator and the enhancement of foraging efficiency through a pollinator’s behavioural tendency to flower constancy all rely on distinct floral morphologies whether it be colour or floral anatomy. Therefore, studying morphologically similar taxa or experimentally inducing morphological similarity will be desirable.
Furthermore, cases may exist where floral volatile fragrance could operate synergistically with visual signals to determine functional specialization (Raguso 2004; Knudsen et al.
2006). Disentangling the respective effects of visual and fragrance cues in synergistic attraction will be a major experimental challenge, therefore studying systems of morphologically similar taxa removes a layer of complexity by limiting variation in floral advertisement to volatile variation.
Sympatric, co-flowering, morphologically similar (ideally cryptic) taxa provide the best chance of studying the biological and evolutionary significance of floral volatile variation and distinguishing adaptive processes from neutral ones. With the exception of extreme cases of convergent evolution, taxa that conform to this description will tend to be closely related. Studying closely related taxa will also provide the corollary benefit of controlling for phylogenetic constraint (limitations to present phenotypic variation imposed by a lineage’s phylogenetic history). Furthermore, eliminating the potential isolating effects of divergent flowering times, geography and morphology focuses on those systems in which pollinator-mediated selection upon floral volatiles is at its strongest (Knudsen 1999;
Knudsen et al. 2006). Indeed, closely related animal pollinated plants that co-flower are likely candidates for divergent floral odour.
These characteristics of the ideal study system may appear to be difficult to find, but examples of such systems that meet or approximate these criteria already exist in the literature (see Table S1). While these studies have tended to focus on sexually deceptive or obligate nursery pollination other examples from outside these specialized relationships are well known. One promising example is illustrated by Knudsen (1999) who analyzed the floral volatile composition of eight co-flowering, sympatric species of Geonoma palms Chapter 1 24 with similar floral morphologies and found that the species could be separated on the basis of their distinctive floral fragrance chemistry. Geonoma is described as
“taxonomically difficult” and little is known about their pollinators (Knudsen 1999).
Clearly, this potential case for floral volatile driven reproductive isolation represents an excellent candidate for application of genetic tools to delineating species boundaries and assessing the occurrence of hybridization.
Although we have narrowed our focus to specialized pollination systems in this review, and indeed recommend such systems for future studies, we expect there are many other cases outside extreme specialization where volatiles are the key to the attraction of different pollinators. Descriptive floral volatile assays across related species have demonstrated that certain volatile profiles or constituents are associated with certain pollinators. For example, Dobson (1997) and coworkers analyzed headspace collections by GC-MS for nine species of Narcissus and found two groupings of fragrance chemistry: one group that emit typical moth-attracting compounds and are pollinated by lepidopteran insects and another group that lack these compounds and are pollinated exclusively by insects other than Lepidoptera. Similar findings have been found among nine species of Nicotiana in which hawkmoth-pollinated species emit “nitrogenous compounds, benzenoid esters and/or terpenoid alcohols” while hummingbird-pollinated species lacked these compounds (Raguso et al. 2003). An extensive survey of floral odour composition across 20 species in three genera of Nyctaginaceae also found evidence for compounds characteristic of hawkmoth-pollination (Levin et al. 2001). These types of studies that employ wide-ranging assays of plant volatiles across related species with contrasting pollination mechanisms, when coupled with genetic investigations, will undoubtedly enable new insights into the evolutionary implications of fragrance.
It is apparent that there is much to learn about the way in which floral scent interacts directly and indirectly within complex biological systems. Integrating modern population genetic tools with traditional pollination ecology will open up new possibilities for Chapter 1 25 understanding the operation and evolution of these systems. We have shown here that this integrative approach has already proved valuable in enabling new insights into plant- pollinator interactions. Without doubt both a challenge and a key to future progress will be the effective execution of interdisciplinary collaboration in strategically targeted study systems. Table 1: A research framework for integration of floral volatile, pollination ecology and population genetic research.
Hypothesis Volatile Knowledge Ecological Knowledge Genetic Knowledge 1. Pollinator Confirmation of Ecological evidence for Genetic confirmation of specificity is due to a distinct volatile pollinator specificity distinct entities (odour distinct volatile composition associated with specific types, subspecies, composition of the activity as volatile composition species) floral blend attractant
2. Distinct volatile Distinct volatile Ecological evidence for Genetic evidence that composition reflects composition among reproductive isolation volatile composition species boundaries species matches genetically distinct species
3. Hybridization is Shared volatile Morphological and Genetic confirmation of due to sharing of key components ecological evidence for hybridization volatile components hybridization of the floral blend
4. Different Confirmation that Ecological evidence for Genetic evidence for pollinators will distinct volatile behavioural differences differences in plant influence the plant components attract among distinct mating system mating system different pollinators (selfing vs pollinators outcrossing)
5. Different specific Confirmation that Ecological evidence for Paternity analysis or pollinators distinctly distinct volatile different patterns of other genetic evidence influence the extent of components attract pollen movement by for differences in the pollen flow different different pollinators extent of pollen flow pollinators
6. Different Confirmation that Ecological evidence for Measures of genetic pollinators will distinct volatile different mating variation and influence male components attract systems and/or differentiation within contribution to different patterns of pollen and among populations population genetic pollinators movement by different (ideally for nDNA and structure pollinators cpDNA)
Table 2: Functionally specialized pollination systems for which floral fragrance compounds have been characterized and linked to specific pollinators.
Characterization Experimental confirmation
GC-MS etc GC-EAD Bioassay - floral Bioassay – System material synthetics or extracts Figs - fig-wasp (Ware et al. 1993; (Gibernau et al. (Grison-Pige et al. mutualism Grison, Edwards & 1998) among 2002a) Hossaert-McKey others 1999; Grison-Pige et al. 2002a; Grison-Pige et al. 2002b) among others
Sexually (Schiestl et al. (Schiestl et al. (Bower 1996; (Schiestl et al. deceptive 2003; Schiestl et al. 2003; Schiestl et Schiestl et al. 2005; Franke et al. Chiloglottis 2005; Franke et al. al. 2005) 2003; Bower 2009) orchids 2009) 2006)
Sexually (Ayasse et al. 2000; (Schiestl et al. (Schiestl et al. (Schiestl et al. deceptive Ayasse et al. 2003; 1999; Ayasse et al. 1999; Ayasse et al. 1999; Ayasse et al. Ophrys orchids Stokl et al. 2005) 2000; Ayasse et al. 2000; Ayasse et al. 2000; Ayasse et al. 2003; Stokl et al. 2003) 2003; Gögler et al. 2005; Gögler et al. 2009) 2009) Euglossine bees (Hills et al. 1972; (Schiestl & Roubik (Williams & (Dodson et al. and neotropical Whitten & 2003; Eltz & Lunau Dodson 1972) 1969; Williams & orchids Williams 1992; 2005) Dodson 1972; Cancino & Damon Ackerman 1983) 2007)
Glochidion – (Okamoto et al. (Okamoto et al. Epicephala 2007) 2007) moth mutualism
Yucca – yucca (Svensson et al. moth 2005) mutualism
Dead-horse (Stensmyr et al. (Stensmyr et al. (Stensmyr et al. (Stensmyr et al. arum 2002) 2002) 2002) 2002) Willows and (Dotterl et al. (Dotterl et al. (Dotterl et al. oligolectic bees 2005; Fussel et al. 2005) 2005) 2007)
Epipactis (Brodmann et al. (Brodmann et al. (Brodmann et al. (Brodmann et al. helleborine and 2008) 2008) 2008) 2008) Vespula social wasps
Table 3: Potential abiotic and biotic factors that may confound studies of the consequences of floral fragrance variation. Targeting ideal study systems that minimize these confounding factors will improve our understanding of the links between floral volatiles, pollination ecology and population genetics.
Confounding factor Control Geography (allopatry) Sympatry
Flowering time Co-flowering (temporal isolation) Visual cues (ethological Morphologically isolation) cryptic/similar taxa
Mechanical isolation Morphologically cryptic/similar taxa
Phylogenetic constraint Closely related/ sister taxa
Post-zygotic isolation Closely related/sister taxa (in the absence of chromosomal anomaly)
Figure 1: A schematic showing potential links between floral volatiles and population genetics in specialized pollination systems. The flowers represent two hypothetical sister taxa for which floral scent determines pollination by contrasting functionally specialized pollinator groups.
Evolutionary lineages are represented by black lines and there is a trend towards finer taxonomic scale at the top of the diagram. Grey arrows depict the direction of gene flow.
Chapter 2 30 CHAPTER 2
INTRODUCTION TO CHILOGLOTTIS: A MODEL SYSTEM FOR
PLANT–POLLINATOR STUDIES.
Labellum detail of Chiloglottis aff. jeanesii
Like any good advertising, flowers appeal to the core preferences and prejudices of the audience they seek to entice. In this way, all biotically pollinated flowers are etched with the physical manifestations of an animal’s cognitive and behavioural world (Chittka and
Thomson 2001; Raguso 2004; Schaefer and Ruxton 2009). This adaptation of flowers to pollinators and the influence pollinators in turn have on the gene flow and reproduction in plants provides a complex and multifaceted field that has long inspired the attention of biologists (Sprengel 1793a; Darwin 1876). The orchids have long been recognized for Chapter 2 31 providing some of the most spectacular and extreme examples of floral adaptation
(Darwin 1877; Micheneau et al. 2009). Among these, the sexual mimicry employed by sexually deceptive orchids surely ranks as one of the finest examples of the exploitation of animals for plant reproduction.
In the grasses and leaf litter of south–east Australia’s wetter forests lays one such example of sexual deception. Chiloglottis is a genus of approximately 30 species of orchids (Bower
2006) and represent one of the best studied examples of sexual deception and floral mimicry. Typical of sexually deceptive orchids more generally, Chiloglottis orchids attract male hymenopterans through chemical mimicry of a female sex pheromone (Schiestl
2005). This attraction culminates in the provocation of mating behaviour, or
‘pseudocopulation’ at the flower which often results in the pickup or deposition of pollen
(Peakall 1990; Schiestl and Schluter 2009).
The attraction of pollinators in Chiloglottis is also typical of sexual deception generally in that it is has been shown to be highly specific; with most plant–pollinator relationship being one–to–one (Paulus and Gack 1990; Peakall 1990; Bower 2006; Bower and Brown
2009; Peakall et al. 2010b). This specific attraction is due to the chemical mimicry of their
Neozeleboria wasp pollinators’ species specific sex pheromones (Schiestl et al. 2003;
Peakall et al. 2010b). The diversity of available pollinators (Griffiths et al. 2011), and strong pre–pollination reproductive isolation imposed by specific pollinator attraction enables two or more species to co–occur and co–flower. Among sympatric species, the chemistry of pollinator attraction is almost always different, suggesting that chemical change coupled with pollinator mediated selection has been a major driver of speciation
(Bower 1996; Mant et al. 2002; Johnson 2006; Bower and Brown 2009; Peakall et al.
2010b).
The accumulation of chemical, ecological, genetic and pollinator studies on this genus, as well as in other systems of sexual deception (Steiner et al. 1994; Schiestl et al. 1999), makes Chiloglottis an ideal system for exploring the causes and consequences of floral Chapter 2 32 adaptation, pollinator behaviour and specificity for plant mating, gene flow and ultimately speciation. Since floral volatiles appear to play a pre–eminent role in this system,
Chiloglottis could serve as a model system for understanding the interaction between floral volatile chemistry, pollinator behaviour and the evolutionary dynamics of populations (Whitehead and Peakall 2009).
The subjects of study in this thesis are the species Chiloglottis valida and C. aff. jeanesii.
Typical of their genus, they are clonal, terrestrial herbs whose single flower is borne on a stem about 5 cm above a pair of simple leaves. C. valida is widespread in the moist gullies and sheltered slopes of the central and southern ranges of south-east Australia. C. aff. jeanesii is known from two regions embedded within the range of C. valida. The two species are morphologically very similar (Figure 1), flower at the same time of year and are interfertile (Chapter 3). The pollinator of C. valida is the thynnine wasp Neozeleboria monticola while C. aff. jeanesii attracts a species of N. sp. (impatiens) recently shown to be paraphyletic (Griffiths et al. 2011)(Figure 2). Specific pollinator attraction is achieved through the emission of distinct semiochemicals, C. valida emitting chiloglottone 1 and C. aff jeanesii emitting chiloglottone 3 (Peakall et al. 2010b). The two taxa therefore fulfil the criteria set out in Chapter 1 for study systems ideal for the integration of floral volatile chemistry, pollinator ecology and population genetics.
The previous chapter established a research framework and a set of testable hypotheses under which we might productively explore the influence of floral volatiles on plant evolution (see Chapter 1, Table 1). Below I set out how the studies in the following chapters use Chiloglottis as a model for research that integrates across sub–disciplines to address some of these hypotheses (Whitehead and Peakall 2009) on the role of floral volatiles in shaping plant–pollinator interactions and ultimately plant evolution.
Chapter 2 33 Chapter 3: Strong prepollination reproductive isolation in sympatric sexually deceptive orchids is driven by floral scent.
While previous research has already established the role of chiloglottones in pollinator attraction (Hypothesis 1)(Schiestl et al. 2003; Peakall et al. 2010b), this study implements a detailed investigation of the nature of pollinator specificity via behavioural experiments with synthetic chemicals and real flowers. Variable neutral genetic markers are then used to investigate Hypothesis 2; that the floral chemical variation driving pollinator attraction is linked to species boundaries. This aspect is further explored through detailed study into the strength and nature of reproductive isolation in these taxa. This feeds into Hypothesis
3, concerning hybridization. We demonstrate the circumstances under which hybridization is possible and assess the evidence for its occurrence in nature. Finally, we touch on Hypothesis 6 which is concerned with measuring overall genetic structure and relating this back to specific pollinator attraction. For this we present data on the patterns of population genetic structure across two populations for both species.
Chapter 4: Multiple paternity and outcrossing in self-compatible clonal orchids.
Here we take an in depth view of gene flow in this system. Through genetic analysis of naturally pollinated orchid capsules, we address Hypothesis 4 and characterize the mating system of both orchids. Hypothesis 5 is addressed through a paternity analysis of these orchid broods to infer pollen flow distances. We compare the two species for resulting measures of pollen flow and discuss results in relation to pollinator behaviour and the field of plant mating and paternity more broadly.
Chapter 2 34 Chapter 5 & 6: Mark-release-recapture studies of behaviour in a sexually deceptive orchid pollinating wasp.
Here we develop and implement a novel marking technique to study wasp movement and examine the resulting data under a new analytical framework for conducting detailed investigations into movement ecology. The behaviour of wasp pollinators can profoundly influence mating systems and gene flow and the way these interactions feed back into floral evolution underpins the questions raised by Hypotheses 4-6. We discuss the potential influence wasp mating, learning and dispersal behaviour might have on orchid populations and species.
Figure 1: Pollination in Chiloglottis. (a) Neozeleboria sp. (impatiens) with pollinia and
Chiloglottis aff. jeanesii. (b) Pollinia deposited in the visit pictured above. (c) Neozeleboria
monticola and Chiloglottis valida.
Figure 2: Illustration of the close sympatry found in C. aff. jeanesii (foreground) and C. valida (background). Chapter 3 37
CHAPTER 3
POLLINATOR SPECIFICITY AND STRONG PRE–POLLINATION
REPRODUCTIVE ISOLATION IN SYMPATRIC SEXUALLY
DECEPTIVE ORCHIDS.
Neozeleboria monticola with Chiloglottis valida
Chapter 3 38
ABSTRACT
The nature of reproductive boundaries between closely related or incipient taxa is of fundamental importance to understanding speciation. In plants, reproductive isolation usually arises from a combination of pre– and post–pollination factors that limit the free exchange of genetic variation between taxa. The sympatric orchid taxa Chiloglottis valida and C. aff jeanesii, while morphologically difficult to distinguish, attract unique thynnine wasp pollinators via distinct floral volatile chemistry. Here we use a multidisciplinary approach to estimate the strength of potential isolating barriers between the two taxa.
First, behavioural experiments with synthetic versions of floral volatile compounds show very strong pollinator specificity mediated by scent chemistry. Field experiments with flowers back up strong pollinator specificity but artificially placing the flowers in close proximity shows the potential for occasional hybrid pollination. Genetic analysis at microsatellite and chloroplast loci shows significant genetic differentiation between taxa along floral scent boundaries however we show that the two taxa are interfertile, displaying evidence of F1 hybrid vigour. A Bayesian clustering approach to look for evidence of hybrids in nuclear DNA genotypes failed to detect any introgression in a fine scale sample of 581 individuals from one site of sympatry. This collective assessment of reproductive barriers shows floral scent mediated pollinator specificity to be the only detectable barrier to hybridization.
Chapter 3 39
INTRODUCTION
Central to the generation and maintenance of new species is reproductive isolation inhibiting the free exchange of genetic variation between diverging lineages (Coyne and
Orr 2004). Isolating barriers to reproduction act as sequential filters through which genes must pass to be successfully transferred between taxa. Understanding the relative importance and interaction of isolating barriers has thus been a major goal in studies of speciation (Ramsey et al. 2003; Coyne and Orr 2004; Martin and Willis 2007; Lowry et al.
2008; Widmer et al. 2008).
Reproductive barriers are classified by their order of appearance in the life cycle, with early-acting barriers exerting a proportionately larger effect on overall reproductive isolation. For plants these include the prepollination barriers of geographic isolation, temporal isolation and floral isolation by mechanical and ethological means (Grant 1949;
Rieseberg and Willis 2007; Lowry et al. 2008; Widmer et al. 2008; Schiestl and Schluter
2009). Post–pollination barriers then include pollen competition, pollen-pistil interactions and gametic incompatibilities—the final barriers in prezygotic isolation. Postzygotic barriers are later acting and include embryo abortion, hybrid sterility and depressed hybrid fitness (Dobzhansky 1937; Mayr 1942; Coyne and Orr 2004; Rieseberg and Willis
2007).
Floral isolation has long been credited as a key driver in the radiation of flowering plants in that pollinator shifts or divergent specialization on different pollinators can in theory lead to rapid and strong reproductive isolation (Stebbins 1970; Grant 1994; Johnson 2006; van der Niet and Johnson 2012). However, a recent emphasis on quantifying the relative importance of the range of isolating barriers has uncovered a more synergistic interaction of early and late acting barriers, the influence of other extrinsic sources of natural selection and the recognition that pollinator isolation seldom acts alone as a reproductive barrier. For example, while Mimulus lewisii and M. cardinalis attract distinct pollinators with high fidelity (bumble-bee and hummingbird pollination respectively), they occur Chapter 3 40 sympatrically in only part of their range and are otherwise ecogeographically separated in altitude. Further, interspecific crosses result in lower seed set and F1 hybrids exhibit depressed germination (Ramsey et al. 2003). Findings such as this demonstrate that understanding the contribution of the full range of pre and post–mating isolating barriers is crucial for understanding speciation in plant taxa (Ramsey et al. 2003; Martin and Willis
2007; Lowry et al. 2008).
The orchids should be a major focal point for examining pollinator-mediated selection and speciation because they are speciose, display a massive range of floral variation and generally have both weak post–pollination reproductive barriers and highly specific pollinator relationships (Tremblay et al. 2005; Cozzolino and Scopece 2008; Schiestl and
Schluter 2009). Yet despite these facts and very early work in pollination biology’s infancy
(Darwin 1877) the orchids remain surprisingly underrepresented in studies of speciation
(Peakall 2007). Among the orchids, perhaps one of the more extreme forms of pollinator specificity is found in sexually deceptive orchids which attract their pollinators via mimicry of sexual signals (Schiestl 2005). After attraction, male insect pollinators are duped into receiving or transferring pollen when mating routines are stimulated by chemical and tactile cues of the flower. Accumulated evidence shows that the sexual attraction of male pollinators is achieved primarily through chemical mimicry of species- specific sex pheromones (Schiestl et al. 1999; Ayasse et al. 2003; Schiestl et al. 2003;
Peakall et al. 2010a). This gives rise to the potential for strong ethological isolation driven by the interactions of differences in floral scent and pollinator behaviour to drive speciation in these groups (Jersakova et al. 2006; Johnson 2006; Peakall et al. 2010a; Xu et al. 2012).
Studies of reproductive isolation in sexually deceptive orchids (and orchids in general) have largely focused on prezygotic barriers. Only three studies in the single European sexually deceptive genus Ophrys have quantified post–pollination isolation barriers;
Scopece et al. (2007) and Xu et al. (2011) both found very weak post–pollination Chapter 3 41 reproductive isolation in Ophrys species as measured by fruit set and embryo counts of artificially crossed hybrids. These studies however were presumably not able to measure any later-acting post–pollination barriers due to the inability to measure hybrid fitness later than the seed stage. This is a problem peculiar to the orchids whose seeds require a specific obligate mycorrhizal symbiont to initiate germination. Xu et al. (2011) addressed potential later acting barriers through ploidy analysis which inferred no karyotype incompatibility and AFLP analysis of genetic divergence identifying a very low number of putative hybrids. In a different approach, Cortis et al. (2009) took advantage of a natural hybrid zone to identify known hybrids and measure their fitness. The results showed incomplete mechanical isolation and lower fruit and seed set in natural F1 hybrids.
The emerging evidence suggests then that reproductive isolation in sexually deceptive orchids is not necessarily controlled solely by floral isolation. Other mechanisms such as post–pollination barriers (Cortis et al. 2009) and the potential for interspecific gene flow
(Peakall et al. 1997; Soliva and Widmer 2003; Gögler et al. 2009; Stökl et al. 2009) might therefore play a role in some systems. Concurrently, research into obligate nursery pollinated figs—once hypothesized to be a pollinator-driven radiation of taxa through exclusive host specificity—has found increasing evidence for occasional breakdowns in pollinator specificity resulting in hybridization (Machado et al. 2005). If we accept the importance of rare hybridization events or ‘leaky’ barriers (Martin and Willis 2007;
Widmer et al. 2008) in shaping species, as well as the importance of early and late acting barriers, a study of plant reproductive isolation should ideally quantify all pre– and post– pollination barriers with sufficient power to detect rare events likely to be important over evolutionary timescales.
As yet, the relative influence of the wide range of reproductive barriers has been studied for very few plant species (Rieseberg and Willis 2007) and has not been quantified for sexually deceptive orchids outside the genus Ophrys. There is therefore a requirement for studies of reproductive isolation between closely related species to build our Chapter 3 42 understanding of speciation in sexually deceptive systems and flowering plants generally.
In light of this, we carried out a detailed study into the mode and strength of pre– and post–pollination reproductive barriers in two sympatric species of Chiloglottis orchids.
Our first aim was to describe and quantify pre–pollination reproductive isolation through floral chemistry analysis and behavioural experiments with both flowers and synthetic versions of the floral volatiles they emit. Secondly, we investigate post–pollination barriers by symbiotic orchid germination of experimental crosses. We can then make the first measurements of the contribution postzygotic isolation via orchid hybrid seedling fitness makes to reproductive isolation. Lastly, we quantified genetic divergence between the taxa at microsatellite loci and used population genetics data in a thorough hybrid analysis designed to detect rare hybridization events.
METHODS
Study species
The subjects of this study are the terrestrial orchids Chiloglottis valida and C. aff. jeanesii.
These taxa are typical of their genus in that they form clonal colonies in the understory of wet forest habitats in south eastern Australia. Colonies have been shown to extend up to
11m (Chapter 4). Chiloglottis valida is known to attract its male thynnine pollinator
Neozeloboria monticola with the single volatile compound, chiloglottone 1 (Schiestl &
Peakall 2005). Similarly, evidence including GC-EAD, GC-MS and bioassays with synthetic compound has revealed that C. aff. jeanesii attracts its undescribed pollinator (a
Neozeloboria species in the impatiens species complex) by a structural isomer of chiloglottone 1, called chiloglottone 3 (Whitehead and Peakall 2009; Peakall et al. 2010a).
These species make a particularly compelling study for floral isolation in that the only known populations of C. aff. jeanesii are sympatric with C. valida, the range of the former embedded within the latter, making them truly ecogeographically sympatric. Cross Chapter 3 43 amplification of microsatellite primers as well as the occurrence of a flowering hybrid between C. valida and C. trapeziformis (Peakall et al. 1997) leads us to assume that F1 hybrids (if they exist) would be viable and detectable in the field. These species therefore fulfil the criteria set out in Whitehead & Peakall (2009) for systems ideal for investigating the influence of floral scent and pollinator behaviour on population genetics and evolution.
Namely, they should be closely related, sympatric, coflowering, morphologically similar and interfertile.
Further, work by Peakall et al. (2010) has shown that despite morphological similarity and range overlap, these two species are unlikely to be sister taxa yet represent closely related phylogenetically distinguishable clades in the Chiloglottis phylogeny. The advantage of this is that we knew a priori that these taxa were divergent at cpDNA loci, allowing us to investigate introgression at nuclear loci — an approach not possible in more closely related taxa without distinct chloroplast lineages. This also allowed us to question whether genetic incompatibilities had accumulated over time subsequent to the initiation of isolating barriers.
Genetic and chemical sampling
Two levels of sampling were undertaken. First, a wide scale sample of individuals of both taxa from two known sites of sympatry was made in order to delineate taxon boundaries with respect to floral scent, diagnostic chloroplast markers and nuclear population genetic differentiation. This first wide scale sample was used as a validation of our diagnostic cpDNA marker and definitive reference dataset for divergence in nDNA and chemotype.
Second, we made an extensive fine scale sample of these taxa at a single site of sympatry in
Kanangra-Boyd National Park, NSW, designed to detect rare hybrids. Here we mapped a
2m2 grid onto a 100m × 24m quadrat and in 2007 collected one leaf sample per grid square. Where possible, we collected from a flowering specimen, however because only a fraction of plants flower in any one season (Peakall et al. 1997) many individuals were Chapter 3 44 only collected as a leaf making taxonomic diagnosis by morphology or floral chemistry impossible. For these unidentified samples we determined taxa by a diagnostic cpDNA agarose assay. Over subsequent flowering seasons (2008-2011) tissue collections and pollination data were collected from flowering individuals on the site over several days each season.
For floral volatile analysis, single labellum extracts were made in 200uL of HPLC grade dichloromethane for five minutes. This was carried out within 48 hours of collection.
Labella were retained for morphological analysis and a single leaf was taken and stored frozen for DNA extraction.
Floral chemistry
To confirm that floral volatile composition corresponds with species boundaries we applied gas chromatographic analysis with mass spectrometry (GC-MS) with selective ion monitoring (SIM) of single orchid labella to identify the active compound in flowers from mixed populations of the two taxa. GC-MS of the extracts was performed on an Agilent
Technologies 6890N GC coupled with a 5973 Mass Selective Detector (Agilent
Technologies, USA) equipped with a SGE BP21 column (30 m x 0.25 mm x 0.25 μm) connected directly to the MS detector. For each sample, 1-4 µl of concentrated extract was injected splitless with the inlet at 250°C, held at 40°C for 1 min, then programmed at 10°C per min to 230°C and held for 15 mins. Helium served as the carrier gas at a flow of 2 ml/min. Our SIM method reduced the detection threshold several orders of magnitude and provides the most sensitive measurement of a chiloglottone presence or absence from single orchid labella (Whitehead and Peakall 2009). This allowed chiloglottone compounds to be detected in the majority of samples from single flowers.
Chapter 3 45 Obtaining genetic data
DNA was extracted from freeze-dried leaf samples using a QIAGEN Plant-Mini kit as per the manufacturer’s instructions. Samples were genotyped at six microsatellite loci chosen for maximum resolution from a set of 12 initial candidates (Flanagan et al. 2006). PCR conditions and genotyping protocol are outlined in Appendix II. We also haplotyped each sample at two diagnostic cpDNA loci. These three-primer reactions each targeted a diagnostic indel and when visualized on agarose gave either one or two bands depending on taxon. PCR conditions for the chloroplast haplotyping followed protocols in Ebert et al.
(2009).
Analysing nuclear genetic data
Unless otherwise stated, GENALEX 6.5 (Peakall and Smouse 2006, 2012) was used for all genetic data analysis. For both the wide scale and fine scale sample we discarded samples amplifying at less than four microsatellite loci and after identifying clones retained only one sample in any such match. We assessed genetic differentiation within and between populations and taxa for our wide scale sample by calculating F-statistics and standardized F-statistics (Meirmans 2006) via an Analysis of Molecular Variance (AMOVA)
(Excoffier et al. 1992; Peakall et al. 1995).
Pollinator fidelity and pollinator isolation
To experimentally test the influence of floral volatiles on specific pollinator attraction and the potential for interspecific pollination we carried out pollinator choice tests with synthetic chemicals as well as mixed arrays of flowers from both taxa. To confirm volatile- mediated pollinator specificity in the absence of other floral cues synthetic pheromone baiting was carried out in a method similar to Peakall et al. (2010a). At two sites where both taxa of orchid and pollinator were known to occur we presented baits of Chapter 3 46 chiloglottone 1, butyl chiloglottone, a one-to-one blend as well as a DCM
(dichloromethane) control. All synthetic pheromones were dissolved in HPLC grade DCM and volumes applied to pin heads were a standardised 20uL. Baits were placed in a row perpendicular to any prevailing wind, each one separated by approximately 20cm and their order randomised between trials. We scored the relative attractiveness of volatile compounds by recording the behaviour of visiting wasps in four minute trials. Behavioural responses scored were: approaches if a wasp flew towards or lingered in proximity to the bait, landings if the wasp landed on the bait or the ground below the bait and copulations if the copulatory routine including abdominal probing was observed. The two different species were identified on the wing by different colouration and abdominal patterns
(Chapter 2, Figure 1).
In the field it is not uncommon to find mixed colonies of both species of orchid. We investigated the plausibility of heterospecific pollination and limits to pollinator fidelity by manipulating the distance between flowers of different species. We simulated the extreme scenario of both species’ flowers contacted at the stem as well as interspecies separation of 20-40 cm. We did this by translocating adult plants into three vials with water: one C. valida, one C. aff. jeanesii and a combination vial containing one plant of each taxon. As in trials with synthetic pheromones, these three vials were presented in randomised arrays in choice trials of four minutes in which we scored wasp responses to each individual flower in the array. If no wasp responses were scored for the first minute of the trial, the trial was abandoned and a new one restarted in a different location. Flowers used in the trials were later labellum extracted for floral volatile analysis and DNA collections were made for genotyping.
Post–pollination isolation: Hand crosses and symbiotic germination
In order to assess postzygotic compatibility within and among taxa we conducted hand crosses with plants collected from the field. After dehiscence of capsules, seeds were Chapter 3 47 stained in acetocarmine with gelatine and 600 seeds per capsule were counted for presence or absence of an embryo. We inoculated oatmeal agar plates (Clements and
Ellyard 1979) with one of three Tulasnella fungal isolates originally isolated from three taxa in the genus Chiloglottis. Fungal isolates used in germination trials were 07061.III.7 from C. reflexa, CV0627.II.2 from C. valida, and 06082.II.1 from C. aff. jeanesii and are described in more detail in Roche et al. (2010). Following fungal colonization, we sowed seed onto the plates following the protocol outlined in Roche et al. (2010); four plates per cross, two plates of one randomly chosen isolate and one each of the other two isolates.
Plates were checked every two weeks until no further plates germinated. Growth in protocorms tended to plateau with many seedlings stalled at very early stages in germination. When plates were no longer growing, germination success was scored under a dissecting microscope for each plate. We did this by superimposing on the plate a 1 cm2 grid with numbered cells. From a list of randomly generated numbers we then counted 12 different cells on each plate, scoring germination according to the scale set out in (Ramsay et al. 1986).
Calculating reproductive isolation
We calculated the strength of individual reproductive barriers after the general form of
Ramsey et al. (2003): RI = 1-(Pheterospecific Pconspecific) where P is some measure of fitness.
To quantify the contribution of pollinator fidelity to reproductive isolation, RIpollinator was estimated from the visitation data recorded in our pollinator fidelity flower choice experiments. We first converted count data for each of the three behavioural responses
(approach, land, copulate) for each wasp species into observed probabilities of conspecific interactions (pollinator fidelity) and heterospecific interactions (breakdown in fidelity).
Heterospecific pollination requires not only that a wasp pollinator visits the “wrong” orchid, but also that the same wasp visits the “correct” orchid species before or after that event. The estimated probability of heterospecific pollination must therefore be the Chapter 3 48 product of conspecific and heterospecific behaviour probabilities, while conspecific pollination probability is the square of the conspecific behaviour probability. For each behavioural response we obtained the mean probability of conspecific or heterospecific interactions across the two wasp species and calculated an RI estimate using the formula
2 adapted from Ramsey et al. (2003): RIpollinator = 1-( Pheterospecific × Pconspecific Pconspecific ). A single RIpollinator is then obtained as the mean RIpollinator across behaviours. Because the experiment simulated an extreme form of sympatry by presenting a mixed-species pair of flowers at 0cm separation, we calculated two RIpollinator estimates—one for the whole experiment including the mixed pair and two single flowers, the other a more realistic scenario based solely on data from the single species flowers presented at 20-40 cm apart.
We estimated a post–pollination, prezygotic component to reproductive isolation using data on successful fruit set in hybrid and pure hand crosses. For this we recorded the number of capsules that swelled and were carried through to dehiscence as opposed to capsules that remained stunted or withered before maturity. Following Scopece et al
(2007) the formula used was RIpost–poll-prezygotic = 1– (% fruit set heterospecific / % fruit set conspecific).
Both seed viability and seedling growth were measured to estimate separate components of post–pollination postzygotic isolation. Postzygotic isolation was quantified separately by the formula RIpost–zygotic = 1-( Pheterospecific Pconspecific) where P is the mean proportion of seed bearing embryos in viability counts or mean growth index across plates in germination trials.
Total reproductive isolation was estimated through a linear sequence of calculations starting with the latest acting barrier, following the formula RI = pre + (1 – pre) × post
(Moyle et al. 2004). Where RIpost–pollination values were less than zero due to hybrid performance exceeding that of conspecific crosses, we converted these values to zero for the total isolation calculation. Due to isolating barriers’ sequential nature each isolating mechanism can only act on gene flow from preceding barriers, giving later acting barriers Chapter 3 49 the power to increase but not decrease isolation. Incorporating negative values in total RI calculations reduces total isolation which is therefore illogical.
Nuclear DNA analysis for hybrid detection
To identify hybrids in our fine scale sample we implemented the Bayesian clustering analysis for admixture analysis, NEWHYBRIDS (Anderson and Thompson 2002). This method provides the advantage of not requiring any a-priori designation of allele frequencies for the groups of interest. NEWHYBRIDS assumes the sample is from an admixed population of two parental species and that pure parent species, F1, F2 and respective backcrosses can all be represented in the sample. Through a Markov chain
Monte Carlo simulation the program then divides the posterior probability between six genotype classes (two purebreds, F1, F2 and two backcrosses to the respective parent species). The threshold probability value (TQ) to assign a sample to any one class is arbitrary with higher thresholds naturally being more conservative. Because the detection of backcross classes requires more loci and is better achieved in populations of greater divergence we chose to focus here on detecting F1 hybrids because if hybridization is occurring, F1s should be the most common hybrid class. We used TQ values of 0.9, 0.75 and 0.5 and in detecting F1 hybrids we used them as cut offs for the posterior probability value given strictly for the F1 genotype class as well as a combined value summed from probabilities in all non-pure genotype classes.
In order to assess performance of the method on our data, we simulated hybrid genotypes based on observed parental allele frequencies in the subset of samples from KBNP taxonomically defined by floral volatile and chloroplast haplotype data. To do this, 10 population pairs composed of pure parental pools for each taxon (n = 1000) were created by drawing from the observed allele frequency distribution for the respective taxon. These pure parental populations then underwent one round of simulated random mating to produce three offspring populations of n = 1000; two pure populations and one F1 hybrid Chapter 3 50 population. These simulations were performed in a customized version of GENALEX, using the simulation framework outlined in Banks and Peakall (2012). NEWHYBRIDS analysis was then run on these offspring populations with no prior information offered. We measured power to detect hybrids and pure parental taxon status for simulated genotypes by the proportion of correctly assigned individuals over the actual number of individuals in that class (Vähä and Primmer 2006; Burgarella et al. 2009). This was calculated under both hybrid detection criteria described above for varying levels of TQ. Finally, we assessed type I error by the number of pure individuals wrongly assigned hybrid status over the total number of individuals of pure parental stock.
Preliminary runs, varying choice of priors and number of sweeps during and after burn-in, made no discernable difference to the results. We chose to run all analyses with “Jeffreys- like” priors for both the mixing proportions and allele frequencies with a burn in of 50 000 sweeps preceding 100 000 sweeps. We ran two analyses on our real genotype data, one without any prior information and another that specified pure taxon status to those samples for which we had chemistry and haplotype data to define taxon.
RESULTS
Floral chemistry and taxon genetic divergence
Our wide-scale sample showed strict correlation between cpDNA markers and GC-MS results for floral volatile profile. All samples showed internal consistency between diagnostic cpDNA haplotype markers and the inferred species diagnosis invariably correlated with C. valida volatile samples matching to chiloglottone 1 and C. aff. jeanesii matching chiloglottone 3. This dataset when divided into cpDNA/chemistry defined groups showed very low genetic divergence among taxa (FRT=0.073, P=0.001) and marginally higher divergence among populations within taxa (FSR=0.092, P=0.001)(Table
1). Chapter 3 51 Pollinator specificity and floral isolation
In synthetic chemical choice trials we scored 11 trials resulting in visits for N. impatiens and 23 trials with visits for N. monticola (Figure 1). Consistent with previous studies
(Peakall et al. 2010a), we demonstrated that N. impatiens was attracted to chiloglottone 3 and not to chiloglottone 1 nor a 1:1 blend of the two semiochemicals. N. monticola was attracted to chiloglottone 1, did not respond to chiloglottone 3 but showed a greater tendency than N. impatiens to approach and land on the 1:1 blend. As expected, we scored no responses to the DCM control.
Bioassays with live plants generated data for 20 trials with N. impatiens and 17 trials with
N. monticola visits (Figure 2). Both pollinators displayed strong pollinator constancy despite the presentation of a close heterospecific pair of flowers. N. impatiens made a higher proportion of its total visits to C. aff. jeanesii in isolation (0.640) rather than in combination (0.337). When N. impatiens did visit the combination it only landed on C. valida once. N. monticola by contrast made a greater proportion of its total visits to the C. valida in heterospecific combination (0.548) than in isolation (0.356). N. monticola also showed a higher proportion of breaks in pollinator constancy than N. impatiens as seen in visits to C. aff. jeanesii in combination (0.081). As the heterospecific pairing we created for this experiment was an extreme scenario, we calculated pollinator isolation for the entire dataset including the 0 cm heterospecific separation (RIpollinator =0.940) and separately for the single flowers presented at more natural 20-40 cm separation (RIpollinator =0.978).
It is important to note that successful pollen pick up and deposition requires the wasp to align with the column and attempt copulation with the labellum such that movements bring them in contact with the column. It is therefore worth examining copulation data in isolation. Copulation behaviour observed in flower choice tests showed very strong pollinator constancy: N. impatiens did not display copulatory behaviour on C. valida at all while N. monticola was recorded twice copulating with C. aff. jeanesii in the heterospecific pair out of a total of 19 recorded instances of pseudocopulation. Resulting estimates for Chapter 3 52 pollinator isolation when just considering copulations were RIpollinator 0.924 and RIpollinator
=1.00 for 0cm and 20-40cm heterospecific distance respectively.
Post–pollination reproductive isolation
To estimate post–pollination, prezygotic isolation we recorded the ratio of fruit set for
2010 hand crosses and found no substantial difference (overlapping standard error bars) in successful fruit set between conspecific (87.5%, n=18) and heterospecific artificial crosses (84%, n=28). Using these ratios of fruit set we determined RIpost–poll-prezygotic=0.04.
Seed viability as measured by embryo counts (Figure 3) showed that the mean proportion of embryo-bearing seed for conspecific crosses (0.818, SE = 0.022, n = 23) was similar to
(overlapping error bars) viability of heterospecific crosses (0.850, SE = 0.026, n = 28). The resulting component of postzygotic reproductive isolation was a calculated RIviability=-0.039
Growth of symbiotically cultured orchid seed was measured as an estimate of fitness for postzygotic reproductive isolation (Figure 4). Consistent with other orchid symbiotic germination trials we found wide variation within and among crosses in seed germination rates (Batty et al. 2001; Swarts et al. 2010). Our data was strongly non-normal in distribution due to the very large variance and high number of zeroes which necessitated non-parametric tests. Using Kruskal-Wallis tests we looked for interactions of isolate with taxon and cross type and found isolate had no significant effect on growth index for any of these treatments of the data. We found evidence for higher fitness in hybrids with overall mean growth index for hybrids of 127.97 (SE = 23.99, n = 66) compared to mean growth index of 44.94 (SE = 12.50, n =47) for within taxon outcrosses and this higher growth index for hybrid seed was significant (Kruskal-Wallis test H=5.52, df=1, P=0.02).
Reproductive isolation due to F1 seedling fitness was a negative value, reflecting hybrid vigour, RIgermination=-1.27.
Chapter 3 53 Hybrid detection
Analyses of simulated data showed an average of 95% of F1 hybrids were correctly assigned under a relaxed threshold (TQ ≥ 0.5) and this dropped to 82% under the strict threshold (TQ ≥ 0.9)(Figure 5). Basing assignments on strict F1 probabilities versus rolling all non-pure probabilities into one value did not substantially affect the power to detect hybrids. The range in power of correctly assigning simulated C. valida and C. aff. jeanesii under varying TQ was 88-95% and 96-99% respectively. With the exception of one C. valida out of 20 000 simulated individuals, misassigned pure individuals were wrongly assigned to the F1 genotype class. Type I error was very low with a range 0.027-0.006 across relaxed to strict TQ.
Despite the power of our dataset demonstrated by simulations and collection of 571 unique multilocus genotypes, hybrid detection analysis of our fine scale sample failed to assign a single sample to the F1 genotype class. The maximum posterior probability for a sample in the F1 genotype class was just 0.046. There were however a small number of samples which assigned to one of the other non-parental genotype categories (F2, backcrosses). The proportion of these varied depending on whether or not chemistry and haplotype data were used as prior information in the analysis (Table 4). When chemistry was included as a prior, the number of inferred non-pure genotypes was below the number expected under alpha=0.05 as determined by simulation. When we included no such prior, the proportion of non–pure samples was slightly above that expected under alpha=0.05.
DISCUSSION
A major goal in the study of speciation is estimating the timing and relative contribution of individual barriers to reproductive isolation (Rieseberg and Willis 2007; Lowry et al.
2008). As such, to understand the influence pollinators might have on the speciation process in plants we need to understand how pollinator specificity interacts with the Chapter 3 54 mechanisms of reproductive isolation (Kay and Sargent 2009). Although sexually deceptive orchids may offer one of the best candidates for pollinator–driven speciation
(Johnson 2006; Peakall et al. 2010a), comparative studies that combine estimates of floral isolation in the context of other isolating barriers and rigorous tests for hybridization are lacking.
Here we demonstrate that plant volatile chemistry, by controlling pollinator specificity, provides the majority contribution to reproductive isolation in this pair of sympatric taxa.
Furthermore, floral volatile composition delineates taxonomic boundaries that are supported by chloroplast and nuclear genetic analysis. Pollinator experiments show that artificially positioning flowers in close proximity can result in occasional breakdown of pollinator specificity. The potential for introgression is not impeded by post–pollination reproductive barriers which we find to be effectively non-existent. Despite this and the finding of hybrid vigour in germination of artificially produced F1 offspring, a comprehensive and rigorous genetic screen of nuclear DNA variation failed to detect any
F1 hybrids, a result we attribute to very strong pollinator mediated reproductive isolation.
Does floral scent delineate taxonomic boundaries?
Strict correlation between floral volatile composition and chloroplast haplotype suggests that floral scent reflects taxonomic boundaries. This finding gives us confidence that cpDNA diagnosis of taxon is an accurate proxy for morphology and floral volatile profile in samples collected without floral material. Our wide scale sample divided into chemistry/cpDNA defined taxa showed low genetic differentiation between taxa, a similar magnitude of divergence observed between populations within species (Table 2). The two not mutually exclusive hypotheses for this pattern are that these taxa are experiencing ongoing contemporary gene flow or have experienced a divergence too recent for neutral genetic differentiation to accumulate. The same pattern has been found in other sexually Chapter 3 55 deceptive species (Xu et al, 2012) and we suggest that in the absence of contemporary gene flow (discussed below) the more likely explanation is one of recent divergence.
Pollinator specificity
Behavioural assays with synthetic chemicals are an effective way of controlling for the influence of other cues (such as morphology or colour) in pollinator attraction. Our results for synthetic chemical choice trials mirror results previously found in Peakall et al (2010) which showed pollinator specificity to be controlled by a limited suite of floral volatiles.
The successful attraction of N. monticola to chiloglottone 1 and N.sp. (impatiens) to chiloglottone 3 can be inhibited to a large degree by blending with the alternative chiloglottone. N. monticola appeared to display a higher tolerance to the blend than did N. sp. impatiens (Figure 1).
By carrying out equivalent behavioural assays with live plants we were able to test for breakdowns in pollinator fidelity under an extreme version of sympatry. Floral choice tests showed patterns of attraction similar to those obtained through synthetic choice trials however we did observe a higher proportion of copulation attempts on real flowers compared to synthetic chemicals (Figure 2). This may support the importance of visual and tactile cues at close range for illiciting copulatory behaviour (Schiestl 2005; Gaskett
2011). By presenting flowers in immediate proximity it was possible to cause an occasional breakdown in pollinator specificity for the N. monticola, however most of the time the wasp appeared able to discern the source of the semiochemical that drew it to the mixed pair. As seen in the synthetic trials, N. monticola showed a higher tendency to visit a blended volatile stimulus than N. sp. impatiens and this could be evidence of a stronger deterrent effect of heterospecific volatiles to N. sp. impatiens than N. monticola. We calculated all possible interspecies pairwise distances between samples from our fine scale sample. The results show that just 93 out of a total of 72664 possible heterospecific interplant distances were less than 2 m (the limit of accuracy in our mapping of the site). Chapter 3 56 The coflowering of both species within centimetres of each other is thus exceedingly rare but not impossible.
Chiloglottis and Ophrys might provide examples of “genic speciation”, whereby reproductive traits of large effect and simple genetic control might give rise to reproductive isolation (Lexer and Widmer 2008). These results showing variation in the tolerance of specific pollinators to blended floral volatiles suggests that if floral scent is controlled by few genes of large effect, mutation in scent genes leading to blends might still attract pollinators and provide a “bridge” to speciation or divergence.
What are the relative contributions of isolating barriers?
C. aff. jeanesii’s known range is entirely embedded within the larger range of C. valida and coflowering populations are the norm. Geography, habitat and phenology are therefore irrelevant as potential isolation barriers. The strongest reproductive barrier we found evidence for is floral isolation driven by the specific behavioural response of pollinators to floral scent. We calculated RIpollinator to be in the range of 0.94-1.00 depending on the criteria we used (Table 3). Under the conditions naturally most likely to result in heterospecific pollination we estimate isolation based on interplant sympatry scale of 20-
40 cm separation and require that pollinators attempt copulation to achieve pollination and the resulting RIpollinator = 1. Rare interspecific pollination is however possible if the scale of sympatry is 0cm separation between flowers, where RIpollinator = 0.924. It is important to note that our RIpollinator estimate is a proxy calculated on probabilities derived from wasp visits rather than observed matings per se (see Martin and Willis 2007; Lowry et al. 2008; Xu et al. 2011). Recent paternity analysis of natural pollination (Chapter 4) found no interspecific pollination, and if we used these observed results in the same calculation RIpollinator would be 1. Thus, Chiloglottis orchids remain one of the strongest cases for pollinator-driven speciation exemplified by the tight pollinator specificity resulting in floral isolation demonstrated here. Chapter 3 57 Discovering hybrid vigour in seedling growth of F1 hybrids was unexpected and if sustained through the lifecycle should facilitate interspecific gene flow (Lowry et al. 2008).
It is difficult to infer what the effects of this heterosis might be however as the overall fitness of hybrids will depend on their viability and how they respond to the selection pressures in nature. F1 hybrids of C. valida and C. trapeziformis are capable of both flowering and vegetative reproduction in the field, however no backcrossing or F2s have been detected. It might be likely then that C. valida × C. aff. jeanesii hybrids can grow to maturity but further studies such as in situ seed baiting would be required to confirm this and estimate extrinsic postzygotic isolation as a result of hybrids’ response to natural selection. The observed absence of post–pollination barriers between C. valida and C. aff. jeanesii is consistent with other studies of sexually deceptive orchids (Scopece et al. 2007;
Xu et al. 2011) and is perhaps not surprising given that C. valida can successfully hybridize with the more phylogenetically distant C. trapeziformis (Peakall et al. 1997; Peakall et al.
2010a).
Despite having been discussed as a potential contributor to orchid speciation (Cozzolino and Widmer 2006; Otero and Flanagan 2006; Roche et al. 2010; Waterman et al. 2011), mycorrhizal specificity has been largely ignored in studies that quantify reproductive isolation in orchids. In this study, our symbiotic germination trials showed no significant effect of isolate on germination rate whether an orchid was cross-fostered or germinated on its native symbiont. The role of mycorrhizal specificity as a potential reproductive barrier to seed germination should be addressed in future studies of orchid reproductive isolation.
Is there evidence for interspecific gene flow in the population?
Despite demonstrating the potential for rare heterospecific pollination and the absence of strong post–pollination barriers, rigorous population genetic analysis found no evidence for hybridization between these species. Bayesian admixture analysis failed to assign a Chapter 3 58 single sample to F1 genotype class out of a population sample of 485 unique multilocus genotypes acquired from 581 individual samples. Simulations using real allele frequencies showed that even under the most relaxed assignment criteria with highest type I error, we should expect close to a 95% chance of detecting hybrids. Despite these relaxed criteria however, we found the number of non-pure genotypes to be within the realms of type I error when using a chemistry prior, and barely above type I error when offering no such information. The increased number and greater magnitude of non-F1 non-pure class probabilities in the no-prior analysis suggests that low power might be causing a very low number of samples to not assign to any one class. We expect with more loci these samples would group with one or the other parent taxon. As stated above, if F1s were indeed capable of reaching maturity as they are in the more phylogenetically distant C. valida × C. trapeziformis cross, then clonality should allow them to persist for a long time in the population.
Conclusion
A thorough study of pre and post–pollination barriers to reproduction have shown that genetic boundaries in these sympatric orchid taxa are maintained through specific attraction of distinct pollinators. We have attempted to address the challenge of measuring postzygotic hybrid fitness in orchids beyond the seed stage by carrying out the first study to include symbiotic germination trials in quantifying reproductive isolation.
Postzygotic barriers usually require time to accumulate (Dobzhansky 1937; Lowry et al.
2008) and despite some anagenetic change in these two lineages we find that intrinsic postzygotic isolation barriers have yet to form between these taxa. This is backed up by the very shallow divergence in microsatellite allele frequencies which we can ascribe to a recent split given the evidence we have for a lack of contemporary gene flow.
Growing hybrids through to adulthood is a major challenge but will be required to test lifetime hybrid fitness and the maintenance of barriers with any sustained introgression. Chapter 3 59 The question of what chemical phenotype F1 hybrids might show is of special interest as this could tell us not only about attractiveness of hybrids to pollinators but also about the genetic control of floral volatile synthesis. If chiloglottones are a polygenic trait we might expect volatile composition of hybrids to be an intermediate of the two parents which we have shown to be unlikely to illicit pollinator response (Figure 1). Simple genetic control by a single gene however, might result in a phenotype very similar to the parents and so remain attractive to pollinators. Despite not having this information about adult F1 hybrids, we have shown that wasp behaviour could be an effective isolating barrier without the need for depressed hybrid fitness. Later barriers to introgression such as reduced attractiveness, fertility or survival of hybrids might exist however they are not necessary to explain the extreme pollinator specificity, population genetic patterns and absence of hybrids demonstrated by this study.
Table 1: Population genetic statistics for populations within species
Taxon Population n Ho (SE) He (SE) F (SE) C. aff. jeanesii Kanangra-Boyd 50 0.578 0.094 0.592 0.092 0.040 0.034 Tallaganda 31 0.676 0.033 0.727 0.030 0.067 0.038 C. valida Kanangra-Boyd 32 0.662 0.034 0.774 0.041 0.129 0.076 Tallaganda 13 0.712 0.059 0.759 0.021 0.068 0.059
Table 2: AMOVA results for genetic differentiation within and between taxa. Probability values were determined through 999 iterations.
Source df Est. Var. % Among taxa 1 0.181 7% Among pops within taxa 2 0.208 8% Among indiv within pops 122 0.197 8% Within individuals 126 1.909 76% Total 251 2.495 100%
Level Value P
FRT 0.073 0.001 F’RT 0.340 -
FSR 0.090 0.001 F’SR 0.308 -
FST 0.156 0.001 F 0.093 0.001 IS
Table 3: Cumulative estimates of reproductive isolation for C. valida and C. aff. jeanesii. Pollinator isolation as estimated by flower choice experiments varies depending on the criteria for heterospecific pollination (separation in space and requisite behaviour).
Isolating >20cm, >20cm, 0cm, all 0cm, barrier all visits copulations visits copulations Ecogeographic 0.000 0.000 0.000 0.000 Phenology 0.000 0.000 0.000 0.000 Pollinator 0.978 1.000 0.941 0.924 Fruit set 0.040 0.040 0.040 0.040 Seed viability -0.039 -0.039 -0.039 -0.039 Germination -1.270 -1.270 -1.270 -1.270 TOTAL 0.979 1.000 0.943 0.927
Figure 1: Behavioural responses of two wasp pollinators to a choice of synthetic pheromone baits: Chiloglottone-3, Chiloglottone-1 and a 1:1 blend of the two semiochemicals.
Figure 2: Behavioural responses of two wasp pollinators to a choice of flowers: Chiloglottis aff. jeanesii and C. valida presented in isolation or in combination.
Figure 3: Grand mean percentage of seeds bearing embryos per cross type. (CAJ = Chiloglottis aff. jeanesii, CVA = C. valida)
Figure 4: Symbiotic germination success (growth index averaged among plates) across different cross classes. (CAJ = Chiloglottis aff. jeanesii, CVA = C. valida, n = number of plates total / number of unique crosses)
Figure 5: The proportion of accurate assignments of simulated data in NEWHYBRIDS under different threshold values of Q (TQ). Dashed lines show the proportion of hybrids successfully assigned to the F1 genotype category (black dashed) or any non-pure parental genotype category (grey dashed) equal to the inverse of type II error. Solid lines show the proportion of simulated C. valida (grey) and C. aff. jeanesii (black) successfully assigned to their respective pure parental genotype category. The proportion of simulated pure parental population being assigned to a non-pure genotype category (type I error) is shown by the dotted grey line.
Table 4: Proportion of samples assigned to a non-parental genotype class from two NEWHYBRIDS analyses of a mixed C. valida and C. aff. jeanesii sample (n = 517). Alpha levels for the various Q thresholds (TQ) are determined from simulations. Analyses were run with and without chemically defined samples being assigned a-priori to a taxon.
TQ 0.5 0.75 0.9 Alpha 0.027 0.014 0.006 Chemistry prior 0.005 0.000 0.000 No chemistry prior 0.037 0.016 0.009
Chapter 4 67
CHAPTER 4
MULTIPLE PATERNITY AND OUTCROSSING IN SELF-
COMPATIBLE CLONAL ORCHIDS.
Neozeleboria monticola with Chiloglottis valida
Chapter 4 68
ABSTRACT
The movement of pollen is fundamental to plant mating systems, the nature of population genetic structure as well as reproductive isolation and hybridization. Despite the fact that most plants’ fruits contain multiple seeds the subject of multiply sired broods through multiple pollen deposition has been the subject of very little research. Here we use microsatellite markers to reconstruct and infer paternity for the naturally–pollinated capsules of two clonal, self–compatible orchids. Individual offspring for Chiloglottis valida and C. aff jeanesii were acquired through symbiotic culture of seeds collected over three seasons. With reference to the genotype of the mother and a fine scale background population map of candidate fathers we quantified rates of outcrossing and pollen movements. We found these orchids’ mating system to be characterized by extensive outcrossing despite clonality. We found an unexpected departure from strict correlated paternity with the average C. valida and C. aff jeanesii capsules being sired by 1.27 and
1.44 pollen donors respectively. Combined median pollen flow distance from 21 successful paternity assignments was 14m and ranged 0 – 291 metres. The results support sexual deception as a driver of outcrossing and supply evidence for longer distance pollen flow than previously predicted for this system. This is also the first study to genetically confirm the operation of multiple paternity in orchid fruits.
Chapter 4 69
INTRODUCTION
The rich diversity of plant mating systems has been a vibrant field of study for evolutionary biologists since Darwin’s early work (1876, 1877). Floral morphology, inflorescence architecture, self-compatibility, pollinator behaviour and flowering phenology are some examples of the many factors that interact to determine where a plant sits on the spectrum between obligate selfing and obligate outcrossing. One aspect of plant mating that has until recently been largely ignored is the causes and effects of multiple paternity (Bernasconi 2004). While the study of multiple paternity or polyandry in animals is rich with theory and empirical study (Slatyer et al. 2011), surprisingly little work has been carried out on the patterns of multiple mating in flowering plants.
As most plants’ fruits contain multiple seeds, a single flower may give rise to multiply sired broods via two non-exclusive paths. First, multiple pollinator visits might occur sequentially on the same flower during its receptive phase. How long a flower is open relative to the probabilities of pollinator visitation then governs the opportunity seeds have to be sired by ‘sequential deposition’ from the diversity of the pollen pool (Campbell
1998; Mitchell et al. 2005). Second, pollinators can frequently bear mixed pollen loads, composed of pollen grains from the most recent donors visited in the foraging sequence.
Thus, in a single pollinator visit ‘simultaneous deposition’ of a mixed pollen load can result in fertilization by pollen from the same plant, close relatives and unrelated donors
(Dudash and Ritland 1991; Mitchell et al. 2005; Karron et al. 2012). Pollen carryover, i.e. the retention of a single donor’s pollen on a pollinator over subsequent flower visits, is key to simultaneous deposition (Waser and Price 1984; Holmquist et al. 2012). By increasing mate diversity and in turn increasing the proportion of outcrossed progeny, pollen carryover can therefore result in fitness benefits for multiply sired fruits (Johnson and
Nilsson 1999; Karron et al. 2006). As well as benefits intrinsic to an increased probability of outcross fertilization, other benefits to pollen donor diversity include increased quality and quantity of offspring (Marshall and Ellstrand 1986; Montalvo 1992) as well as the Chapter 4 70 potential for ovules to ‘bet-hedge’ across numerous mates of varying quality (Jennions and
Petrie 2000; Karron et al. 2012).
Self-compatible hermaphrodites are ideal study systems for investigating multiple paternity in plant mating (Ellstrand 1984; Epperson and Clegg 1987; Karron and Marshall
1990; Montalvo 1992; Campbell 1998; Mitchell et al. 2005; Karron et al. 2012). Yet despite orchids being both mostly self-compatible hermaphrodites and providing some of the most sophisticated examples of floral adaptation, as a group they have been neglected in studies of plant mating.
The great majority of orchids package their pollen in pollinia—aggregations of pollen grains that are removed and deposited by pollinators as a unit. Because of this, it is widely assumed orchid capsules will usually only be sired by one pollen donor. To our knowledge, the only published genetic paternity analysis of orchid offspring strongly supports this expectation reporting strict correlated paternity—all seeds in a capsule were pollinated by a single donor (Trapnell and Hamrick 2005; Trapnell and Hamrick 2006). However, based on marked pollen analysis, it is recognized that pollen carryover has the potential to promote multiple paternity in some species with sectile or soft pollinia (Peakall 1989;
Peakall and Beattie 1996; Johnson and Nilsson 1999; Johnson and Edwards 2000; Johnson et al. 2005; Tremblay et al. 2005; Harder and Johnson 2008). Despite this, only two studies have specifically quantified pollen carryover for orchid pollinia; one a field study where ants were found to transport a single pollinium across up to 20 flowers (Peakall 1989;
Peakall and Beattie 1991), and a greenhouse study of two species found that during a foraging sequence the median pollen grain visited more than one flower (Johnson and
Nilsson 1999).
One reason for the paucity of orchid mating studies is that studies of pollen flow and postzygotic paternal fitness not only require heritable, informative genetic markers such as microsatellites, but also require sufficient offspring DNA in order to assay loci for Chapter 4 71 paternity. For most orchids, miniscule seeds pose the challenge of minute DNA yields and obligate mycorrhizal association prevents easy germination of offspring in the lab.
Here we take advantage of established methods for orchid symbiotic culture and previously developed molecular tools to investigate paternity and mating in two self- compatible sexually deceptive orchids in the genus Chiloglottis. These orchids attract male wasp pollinators through sexual mimicry of the female’s sex pheromone—a deceptive strategy that has been shown to promote high outcrossing rates, long distance pollen flow
(Peakall 1990; Peakall and Beattie 1996; Peakall and Schiestl 2004) and efficient pollination (Scopece et al. 2010). Pollinator flight distances and behaviour are important factors in Chiloglottis pollination because these orchids form dense clonal colonies that can produce multiple flowers in a season. This sets up a conflict between clonal growth and outcrossing; the possibility of geitenogamy within clonal patches, which decreases in probability with the number of flowers in a clone (Snow et al. 1996; Vallejo-Marín et al.
2010). These orchids have soft pollinia that can break up on deposition leading to the possibility of carryover. Further, the pollinators for this group have been observed bearing multiple pollinia at a time. Given that flowers stay open for approximately 24 hours after pollination, these orchids thus have the potential to be fertilized by multiple fathers through mixed pollen loads brought by the same pollinator or by sequential deposition.
This study focused on a single site monitored over several seasons of natural pollination.
We aimed to characterize the mating system beyond outcrossing-selfing in two sympatric sexually deceptive orchids. Specifically, we aimed to answer the following questions:
1) Is there any evidence of inbreeding depression in the germination of selfed seed
versus outcrossed seed?
2) What are rates of outcrossing in these two clonal species?
3) What is the typical mate diversity for individual fruits?
4) Can we infer paternity for progeny and use this to trace the extent of gene flow?
Chapter 4 72
METHODS
Study species and sampling
The genus Chiloglottis contains some 30 species of terrestrial orchids that form clonal colonies in the understory of wet forest habitats in south eastern Australia (Peakall et al.
2010a). While colonies may be very dense in ramets, only a small proportion of these flower in any given year. The anatomy of their flowers makes them particularly useful for pollination study as their anthers and stigma are easily inspected for pollen removal or deposition. Each flower bears two pairs of curved, soft, mealy pollinia. Chiloglottis valida and C. aff. jeanesii are typical members of their genus, occur in sympatry and are strongly reproductively isolated by their attraction of different specific pollinator species (Chapter
3). This study focuses on one site of sympatry, Kanangra-Boyd National Park, where both species grow intermingled in large numbers.
The study site occupies a gentle slope in tall alpine eucalyptus forest. At the outset of the study we mapped a 100×24 m transect (running roughly north-south) over the most abundant section of the population. This transect was bounded on the east by a creek line and west-facing slope and to the south by a change in vegetation community - both features were not favourable for Chiloglottis as few to none were found in these areas. The west boundary is marked by a vehicle track, the other side of which is lies flat habitat supporting a low density of orchid colonies. To the north of the transect the slope continues with colonies of orchids sporadically distributed in lower numbers for over 400 m.
In 2007 we took a fine-scale sample of the transect of one individual leaf per 2 m2 grid square. This was done to establish population allele frequency distributions, estimate maximum clone size and provide a background database of candidate genotypes for paternity analysis. In the years 2007 – 2009 we thoroughly surveyed all plants on the transect for pollen removal or deposition over the course of several visits per season. For every plant detected with pollen removal, a corresponding leaf sample was taken for DNA. Chapter 4 73
If plants had been pollinated we estimated the number of pollinia deposited on the flowers that were still fresh and then transplanted them to pots and relocated them to a growth cabinet to avoid losses from herbivory. At the onset of capsule dehiscence we collected the seed with a corresponding maternal leaf sample. During the seasons 2010 and 2011 we continued to collect data on pollen removal and deposit, continued taking leaf samples for potential pollen donors, but we did not collect further plants or capsules. Additionally, we made collections of all major colonies on the periphery of the study site to expand our population sample of the site to an area of 160×80 m. At the end of the study, we therefore had material from three seasons of naturally pollinated seed capsules and their mothers, five seasons of all observed pollen donors, a systematic transect-wide sample, collections from all colonies observed on the periphery of the transect as well as a small number of naturally pollinated capsules and leaf samples from elsewhere in the national park
(ranging from 300 to over 5000 m away).
Orchid fungal culture and seed germination
Chiloglottis orchids like Australian terrestrial orchids in general obligately require mycorrhizal infection for seed germination. The procedures adopted here for orchid fungal culture and seed germination follow Roche et al. (2010). Briefly, orchid seed was surface sterilized and dispersed onto oatmeal agar plates (Clements and Ellyard 1979) inoculated and colonized by one of three Tulasnella fungal isolates; 07061.III.7 from C. reflexa, CV0627.II.2 from C. valida, and 06082.II.1 from C. aff. jeanesii. Three plates per capsule were sowed, one of each isolate. Plates were stored in the dark at 20° and germination was monitored every 2 weeks. When protocorms showed signs of slowed or halted growth we collected orchid protocorms greater than approximately 1mm diameter for genetic analysis.
For most orchid species symbiotic germination remains challenging even with the appropriate fungal symbiont as rates of germination can be wildly variable and often very Chapter 4 74 low (Batty et al. 2001; Swarts et al. 2010). Further, the development of protocorms beyond the early stages requires subculture into different media as well as changes in light and gas exchange regime. Therefore, here we only aimed to germinate large numbers of seed to a stage suitable for DNA extraction.
Hand crosses
To assess self-compatibility and evidence for inbreeding depression we performed a series of hand pollinations of virgin plants including self and outcrosses. Seed from hand-crossed capsules was germinated symbiotically in the manner described above and germination success was scored under a dissecting microscope for each plate. We did this by overlaying on the plate a 1 cm2 grid with numbered cells. From a list of 12 different randomly generated numbers for each plate we then counted the corresponding set of cells on each plate. Germination was scored according to the scale set out in Ramsay et al
(1986).
Initial analyses of the data showed no significant influence of isolate on germination, so this variable was dropped from the analysis. We tested differences in germination index between outcross and self treatments with a normal approximate Wilcoxon rank sums test in. Because symbiotic orchid germination is highly variable with many plates not germinating, we also reduced the data to binary germination / no germination and tested differences in percentage of plates germinated by χ2 contingency test. Statistical tests were carried out in the software package JMP (SAS, Cary, NC, USA).
DNA extraction and microsatellite amplification
Leaf tissue was freeze-dried and extracted as described in Chapter 3. Due to the difficulty in germinating large numbers of orchid protocorms, we only focused genotyping effort on capsules that germinated more than 12 individual protocorms of a size large enough for Chapter 4 75
DNA extraction. Protocorms were freeze-dried directly after picking off the agar plate. Due to the high numbers and small size we carried out a 96-well plate DNA extraction (see
Appendix II). As the amount of tissue in early stage protocorms is very small, the resulting
DNA yields were also very small and it became clear we needed to trade off the number of loci genotyped against the amount of DNA required for reliable amplification and repeats if necessary. We therefore carried out a genotyping screen of a mixed-taxon pool of 200 leaf samples from the transect at 12 microsatellite loci described in Flanagan et al (2006).
Using GENCLONE2.0 (Arnaud-Haond and Belkhir 2007) we then plotted the plateau of multilocus genotype recovery with each combination of loci (Figure 1) determining a minimum of six loci of highest resolution necessary for recovery of all unique multilocus genotypes (MLG). Resulting exlusion probability with one parent known was 0.995 and
0.963 and probability of identity was 2.55 x 10-7 and 2.96 x 10-5 for C. valida and C. aff jeanesii respectively. We then ranked loci in order of information content as measured by effective number of alleles and chose the six most informative loci to genotype in further analyses.
Protocol for amplification by polymerase chain reaction and scoring of genotype is outlined in Appendix II. In the event of mismatches with maternal genotypes, or failure to amplify, offspring were re-run until the error was rectified or the DNA stock was exhausted. Each mother was run a minimum of two times per locus to ensure accurate genotypes against which to calculate protocorm genotyping error rates and reconstruct paternity.
Locus power and clone detection
We limited our clone analysis to samples that amplified at five or more loci. Identification of unique MLGs was undertaken with GENCLONE2.0 (Arnaud-Haond and Belkhir 2007).
The resolution of our six locus battery for detection of MLGs was assessed in
GENCLONE2.0 by plotting the recovery of MLGs with each possible combination of loci. Chapter 4 76
Samples with only five genotyped loci were able to be assigned to potential matches by locus subsampling. This consisted of six analyses per taxon – one for each possible five locus combination. Samples with missing data that found a match in a subsample analysis were designated clones if they occurred within a 12 m radius.
Protocorm genotyping error rate
The dataset comprised of protocorms genotyped at more than four loci was used to estimate microsatellite genotyping error rates. By reference for each protocorm to its known maternal genotype, we could estmiate a rate of mistyping for each locus. This was calculated for a locus by obtaining the proportion of samples in which a maternal allele was not recovered: Pmis. Because at a given locus, each diploid offspring represents two chances to recover the maternal allele, our rate of genotyping error = Pmis /2.
Mating system analysis
To estimate the degree of outcrossing we constructed progeny arrays for all mothers of naturally pollinated capsules and their protocorms for whom we obtained genotypes at four or more loci. We divided the data set by taxon and estimated single and multilocus outcrossing rates (ts and tm), biparental inbreeding (tm - ts) and correlation of paternity (rp) in the package MLTR version 3.4 (Ritland 1989; Ritland 2002). The analyses were conducted under a Newton-Raphson iteration routine with pollen and ovule allele frequencies assumed equal. Standard errors were obtained via 1000 bootstrap iterations resampled at the family level. As well as overall taxon analyses, we calculated outcrossing rate for each naturally pollinated family (Ritland and Jain 1981).
Sibship and paternity reconstruction
While orchids provide technical challenges in obtaining genotypic data for progeny arrays, they are ideally suited to the use of reconstruction in assigning paternity. This is because Chapter 4 77 pollen is transferred in aggregate via the pollinia meaning siblings within an orchid capsule have a high correlation of paternity. That is, most if not all ovules have been fertilised by the same pollen donor. If more than one pollinium fertilizes a brood, the aggregation of pollen ensures that full sibs will still be common within a fruit. If the mother’s genotype is known, the paternal genotype can then be deduced by subtracting the maternal alleles at a locus across a series of offspring. Mendelian inheritance ensures that the probability of recovering the diploid paternal genotype at a locus asymptotically approaches one as the number of offspring assayed increases (Jones et al. 2010).
Data for our paternity and sibship analysis was limited to samples and progeny arrays for which we had a minimum of five loci scored. This was because reductions in resolution with fewer loci can lead to spurious matches when assigning paternity. Further, offspring with fewer successfully genotyped loci were likely to be very small protocorms with low resulting DNA concentrations, raising the probability of genotyping error at loci that were amplified.
For each progeny array we first attempted to assign sibship and paternity through reconstruction and exclusion (Ellstrand 1984). This was estimated manually after extracting all unambiguous paternal alleles at each locus for a given progeny array using the software package GENALEX 6.5 (Peakall and Smouse 2006, 2012). We calculated the minimum number of fathers needed to sire a brood by the count of unambiguous paternal alleles. Our criteria for accepting a multiply sired capsule required more than two unambiguous paternal alleles to be recovered at more than one locus. In the case where only one locus showed more than two unambiguous paternal alleles, this was accepted as evidence for multiple sires only if every unambiguous paternal allele at that locus was recovered in more than one individual offspring. These conservative criteria therefore avoided artificial inflation of parent numbers due to possible genotyping errors and were parsimonious in not assigning more fathers than strictly necessary to explain any one progeny array. Chapter 4 78
Unambiguous paternal alleles were then used in exclusion of paternal candidates. This was done in a modified version of the TWOGENER and multilocus match routines in the
GENALEX 6.5 software package. The routine took unambiguous paternal alleles as described above for sibship reconstruction and then for each offspring, excluded candidate fathers that did not match the recovered paternal alleles.
As reconstruction and exclusion are not tolerant of errors or missing data, we also carried out paternity and sibship analysis in COLONY version 2.0 (Jones and Wang 2010). The package uses a maximum likelihood algorithm to infer sibship groups and reconstruct and assign paternity while tolerating genotyping errors. The prime limitations of the maximum likelihood approach however is that it cannot incorporate the solid and parsimonious assumption of correlated paternity into the analysis. The analysis therefore risks false inference of full-sibs with common alleles as separate pollination events and can overestimate the minimum number of sires (Sefc and Koblmuller 2009). To test the program’s sensitivity in delineating full sib groups within mixed progeny arrays we analysed simulated orchid genotype data. A simulated population sample for each taxon was drawn from the observed allele frequency distribution for our loci at the site. These simulated individuals then went through one round of random mating under one of three levels of correlated paternity: 1, 0.9 and 0.5. We limited progeny arrays to 12 offspring
(the resulting mean total genotypes acquired in our real data) and analysed 10 families under each level of correlated paternity. We also looked at the accuracy of strict parsimonious reconstruction by simulating 100 progeny arrays per taxon, per rate of correlated paternity and measuring the rate of correct estimation of minimum number of fathers.
Initial trial runs on simulated data showed that the number of correct paternity assignments and full-sib inferences made by COLONY dropped when multiple families were analysed in the same run. This and the steep rate of increase in computation time with increasing numbers of offspring analysed limited us to running analyses one family at Chapter 4 79 a time (Wang and Santure 2009). COLONY analyses on real and simulated data sets were run with an allelic dropout rate and genotyping error rate per locus of 0.0001 and 0.01 respectively. Allele frequencies supplied in COLONY runs were calculated from parent populations after clones were removed from the data sets. The mating system assumed male and female polygamy and allowed inbreeding. Full-likelihood analysis was set to a medium run length, medium precision with a complexity prior on sibship size that penalized pedigrees resulting in different fathers for every offspring.
RESULTS
While pollination efficiency varied widely over four seasons of data collection, when averaged across all seasons both species showed very similar rates of pollinia depositions to removals (C. valida = 0.80, C. aff. jeanesii = 0.85)(Table 1). We collected 46 C. valida and
48 C. aff jeanesii over three seasons. After maturing to dehiscence, symbiotic germination and genotyping of resulting protocorms we had sufficient genotype data for 22 and 29 fruits for the two respective taxa.
Hand crosses
Total numbers of plates counted for hand crosses were C. valida: 27/29; C. aff. jeanesii:
20/21 for outcross and self matings respectively. These were divided among a number of crosses in each category C. valida: 8/10; C. aff. jeanesii: 7/8 for outcross and self matings respectively. Germination was highly variable (germination index range 0 – 400) and generally very few protocorms per plate reached stage 3. In self pollinated capsules for C. valida we detected reduced germination by germination index (Kruskal-Wallis test
H=14.439, df=1, P = 0.0002) and percentage of plates germinated (Pearson’s χ2=15.412, df=1, P < 0.0001) (Figure 3). Differences for either metric of germination success were not Chapter 4 80 significant for C. aff. jeanesii although we did observe an overall lower percentage of germination compared to C. valida.
Clonality
For the mapped transect, adult leaf tissue collections across the five years totalled 268 C. valida and 313 C. aff. jeanesii samples for which a minimum of five loci were successfully amplified. Genotypic resolution of our loci plateaued at between five and six loci but the curve approached maximum resolution faster for C. valida than in C. aff. jeanesii (Figure 1) which is indicative of lower allelic variation in the latter taxon. In both taxa however, the maximally informative combination of five loci recovered the same number of genotypes as with the full six locus battery.
All possible five-locus subsamples of the data were reanalysed including samples with missing data and this assigned 16 out of 41 missing-data samples to clonal matches.
Samples with missing data that could not be matched to a clone in subsampling were assumed to represent unique MLGs. The total number of unique MLGs recovered for the study site was 199 (74% of samples) and 251 (80% of samples) for C. valida and C. aff. jeanesii respectively. The distribution of clone sizes showed the extent of most clones to be two metres or less and we observed maximum clone sizes of 6 and 11 m for C. valida and
C. aff. jeanesii respectively (Figure 2). The density of clones was estimated during the 2007 season by counting leaves and we frequently recorded more than 40 ramets per m2. Single clones at that density could well exceed several hundred ramets however the number of flowers in dense clones was typically fewer than 5-10. It should be noted that colonies need not be composed of a single genet and that multiple flowers within a colony might represent different genotypes.
Chapter 4 81
Population genetics and mating system
From a total of 94 collected seed capsules sowed onto 376 agar plates, we collected, extracted and genotyped a total of 825 individual protocorms of which 671 amplified successfully at four or more loci and 629 at five or more loci. These were spread across a total of 28 and 31 progeny arrays for C. valida and C. aff. jeanesii respectively. We divided each taxon’s genotypic data into two generations and separately analysed baseline population genetic statistics for each of the four resulting populations (Table 4).
Most loci for the four populations showed a statistically significant departure from Hardy-
Weinberg equilibrium due to a heterozygote deficit. This is consistent with the expectation of some selfing (Flanagan et al. 2006) however we cannot rule out that this might be caused by a low frequency of null alleles. For protocorm DNA we quantified rate of error by reference to the maternal genotype and found an error rate of 0.015 averaged across taxa and loci.
Both species displayed very high levels of outcrossing in naturally pollinated fruits, tm =
1.001 (0.048) and 0.924 (0.049) for C. valida and C. aff. jeanesii respectively (Table 2). The difference between single and multilocus outcrossing estimates showed similar and small rates of biparental inbreeding in both taxa despite clonality. We found a departure from strict correlated mating (rp = 1) in both taxa with rp values of 0.647 and 0.765 for C. valida and C. aff. jeanesii respectively (Table 2) suggesting an appreciable level of multiple paternity.
Paternity simulations
For simulated matings we found that strict parsimonious reconstruction correctly inferred the minimum number of fathers at 94% of the time for dual paternity at 0.5 correlated paternity and single paternity (Table 3). This dropped dramatically to 66% or 54% for the two taxa when correlated paternity was set at 0.9. This demonstrates the sensitivity of Chapter 4 82 sibship reconstruction to offspring numbers and proportional representation in the brood.
When paternity is skewed and progeny arrays are limited in size, the number of progeny sired by the minor contributor will frequently be below levels critical for recovering all paternal alleles across loci.
A similar pattern of accuracy of paternal assignment was found in COLONY under the same three levels of correlated paternity (Figure 4). Under single paternity, COLONY assigned the correct father to simulated offspring with 80-90% accuracy. Accuracy dropped slightly when paternity was shared equally between two fathers but the minimum value still exceeded 75%, suggesting an average of six offspring per father is sufficient to assign paternity at this level of confidence. Again, correlated paternity of 0.9 produced the least reliable results and accuracy dropped below 50%. Without simulating a more full range of offspring numbers under varying levels of correlated paternity we cannot say where the optimum for our data lies, however the simulations presented here suggest that the power to assign paternity that is lost when a brood paternity is divided can be compensated for by an increase in the size of the progeny array. Minimum six offspring for any one given father is probably close to the 80% confidence threshold here.
Paternity
We found 27% (C. valida) and 41% (C. aff. jeanesii) of naturally pollinated capsules required a minimum number of two unique fathers to explain the genetic patterns in the progeny array (Table 5). There was one capsule for which three fathers were inferred and all others were inferred as full-sib single paternity broods. This level of multiple paternity was backed up by observations of stigmatic pollen loads frequently exceeding a single pollinia (Table 5).
We inferred paternity by two methods; strict manual exclusion and error tolerant maximum-likelihood analysis in COLONY requiring a probability cut-off value of 0.80. Chapter 4 83
Between these two techniques we were able to confidently infer paternity for 8 out of 22
C. valida and 13 out of 29 C. aff. jeanesii progeny arrays. Eight of these progeny arrays with inferred paternity showed evidence of more than one father and in two cases we were able to assign paternity to minor contributors (<50% paternal contribution). For progeny arrays with inferred paternity we were able to use paternal shares to calculate the effective number of fathers (KE). We did not find a significant association between multiple paternity (as measured by minimum number of fathers) and successful paternal assignment of a progeny array (Pearson’s χ2=3.607, df=1, P = 0.0575).
Pollen flow distances inferred from paternity assignment ranged from 0-291 m. Median pollen flow distance differed largely between the two species (C. valida = 20m, C. aff. jeanesii = 5m) due to the detection of one long distance movement of 291m in the former and a high number of selfs in the latter. Median pollen flow distance for the taxa combined was 14m.
DISCUSSION
While studies that characterize the mate diversity of animal broods are common, comparatively few plant species have been equivalently studied. Collecting data on multiple paternity in plants is therefore fundamental for enabling us to apply a common theoretical and analytical framework to the evolutionary ecology of both groups
(Bernasconi 2004; Karron et al. 2012). Despite presenting some of the most compelling examples of floral adaptation, paternity studies in orchids remain scarce due to the unique challenges of their biology. Here we address these challenges by collecting mating data on two self-compatible sexually deceptive orchid species. We found a mating system characterized by predominant outcrossing, despite clonality. In addition, we found evidence of moderate multiple paternity with 41% (C. aff. jeanesii) and 27% (C. valida) of naturally pollinated capsules requiring a minimum of two fathers to explain the progeny Chapter 4 84 array. This study therefore provides the first genetically confirmed evidence for a departure from strict correlated mating in orchids.
Outcrossing, selfing and clonality
Of the 581 individuals collected over the four years of genetic sampling we found most
(450) represented a distinct genet (C. valida: 74%; C. aff. jeanesii: 80%). Coupled with our finding that the majority of clones in both taxa did not exceed 2 m in maximum size
(Figure 2), these data suggest that clones are structured on a fine scale in both species.
Despite clonality that has the potential to lead to selfing, analysis of progeny arrays from naturally pollinated fruits found that outcrossing predominated for both species. However, the slightly lower outcrossing rate in C. aff. jeanesii (Tm=0.924) compared to C. valida
(Tm=1.001) (Table 2) indicates that mating between within clones or between relatives might be more common in this species. The lower level of heterozygosity in the adult population of C. aff. jeanesii as well as the increased fixation index observed in the offspring generation supports this finding, suggesting long term low levels of selfing are normal in this species (Table 4).
We observed a significant difference in germination between selfed and outcrossed progeny for C. valida (Figure 3), which is indicative of the operation of inbreeding depression at this early life stage. Because we relied on germinated seed for our outcrossing measures our estimates of outcrossing might therefore be subject to an upward bias. As a result, outcrossing rates here should be interpreted as an estimate of realized gene flow, not just pollen transfer. There may be an element of self pollination that we were unable to detect due to our reliance here on germination however the observed departure from Hardy-Weinberg equilibrium (Table 4) in the adult C. valida suggests some selfing carries through to maturity. Chapter 4 85
The germination stage inbreeding depression observed in C. valida was not observed in C. aff jeanesii (Figure 3). This asymmetry between taxa in relative germination success of selfed versus outcross offspring could be due to different levels of ongoing inbreeding. In selfing populations deleterious alleles are more often exposed to selection in the homozygote state. There is therefore more opportunity for selection to purge deleterious alleles from the population while in outcrossing populations these alleles are more commonly masked in the heterozygote state. This leads to an accumulation of genetic load and resulting inbreeding depression (Charlesworth and Charlesworth 1987; Dole and
Ritland 1993). From this theory, we expect an inverse relationship between the level of selfing in a population and the strength of inbreeding depression, which is precisely the pattern observed in our data with C. aff. jeanesii showing larger estimates of selfing but no inbreeding depression.
Outcrossing has been implicated as an adaptive advantage of sexual deception (Nilsson
1992; Peakall and Beattie 1996; Johnson and Nilsson 1999; Cozzolino and Widmer 2005).
Clonal growth in Chiloglottis, however, may create a conflict between the benefits of vegetative reproduction and the costs of increased selfing. In Chiloglottis, occasional selfing can occur through autogamy; pollination within a flower due to vigorous pollinator behaviour (Peakall and Beattie 1996) or automatic dislodgment of the pollinia to stigma
(Peakall et al. 1997), and geitonogamous selfing between flowers within a clone.
Geitonogamy increases in probability with an increasing number of flowers in the clone
(Vallejo-Marín et al. 2010), while autogamous selfing rates remain the same.
The extent to which clonality affects selfing then depends on pollinator behaviour and the probability of wasp flights that cross the clonal boundary. Studies in three different wasps related to the pollinators for these species show average flight distances in the range of
15–32m (Peakall 1990; Peakall and Beattie 1996; Whitehead and Peakall 2012). If we infer similar behaviour for N. monticola and N. sp (impatiens) this suggests that average distance moved between flowers would typically exceed the clone size found in this study Chapter 4 86
(Figure 2). Furthermore, behavioural routines involved in mate search make it likely that wasps will leave the vicinity of an orchid patch after deception (Peakall 1990; Peakall and
Beattie 1996; Alcock 2000)(Chapter 6) reducing the chances of geitonogamous selfing.
Levels of multiple paternity
The prediction that soft pollinia (such as that in Chiloglottis) might promote multiple paternity was supported here in our analysis of 51 naturally pollinated orchid capsules.
Estimates of stigmatic pollen loads that we made upon collection of pollinated flowers generally correlated well with estimates of multiple paternity from genetic analysis of progeny arrays. In C. valida 27% and C. aff. jeanesii 41% of capsules were sired by more than one father with the average progeny array being explained by 1.27 and 1.44 minimum fathers respectively. As a contrast, the only other paternity study in orchids found strict correlated paternity in 459 progeny arrays from 15 populations (Trapnell
Hamrick 2006). While for orchids this result represents a high diversity of pollen donors, aggregated pollen and pollen limitation in this system produced a diversity of sires per fruit far lower than plants with monad pollen grains (eg. Mimulus average 4.9 outcross sires per fruit (Karron et al. 2006), Ipomopsis average 4.4 sires per fruit (Campbell 1998),
Silene population averages 3.8-6.1 (Teixeira and Bernasconi 2007)) but a higher rate of within fruit multiple paternity than observed in Asclepias (1.9% of fruits); the only other pollinium bearing group outside orchids (Broyles and Wyatt 1990).
We found, as in previous work (Sefc and Koblmuller 2009), that COLONY overestimated sire numbers with respect to minimum number of fathers required to explain a progeny array. COLONY is based on a Maximum Likelihood algorithm that incorporates allele frequencies and Mendelian ratios meaning that two offspring sharing an allele which is common might not necessarily be assigned to the same sibship. While we did run analyses under the “complexity prior”, COLONY still produced a number of sibship divisions such as one off selfed offspring, in progeny arrays that parsimonious reconstruction found a Chapter 4 87 minimum of one father sufficient. This is why we favoured inferring the number of sibships in a progeny array via parsimonious reconstruction. Our method may have resulted in occasional underestimate of multiple paternity in some instances due to minority paternal shares not being picked up in small progeny arrays (Sefc and
Koblmuller 2009). However because pollen loads are moved en masse in few pollination events and flowers close within 24 hours, the assumption of low father numbers and correlated paternity will always be a realistic position of parsimony for orchids in particular.
It has been shown that marker polymorphism is the most critical factor in determining sibships (Sefc and Koblmuller 2009). Our clonal analysis shows that resolution of our marker set plateaued before reaching six markers so we are confident that marker power is not an issue here for sibship reconstruction. The limiting factor in our study was most likely the number of samples obtainable per progeny array. Increasing the number of samples per progeny array increases the chances of recovering all paternal alleles at a locus as well as, in the case of multiple paternity, alleles from minority-share fathers. The influence of offspring number on reconstruction has been examined by others (Sefc and
Koblmuller 2009), but more detailed simulations relevant to orchids are required to test the collective influence of progeny array size, marker polymorphism and error rates on the reconstruction of sibships.
Despite these caveats we have confidently inferred the presence of multiple paternity and provided parsimonious estimates of sire number. As yet, we are unable to determine whether the mechanism of multiple paternity is predominantly simultaneous or sequential deposition. The observation of wasps bearing pollen loads that vary from a fragment to three or more pollinia (personal observation) suggests that simultaneous deposition is definitely a possibility. Given that flowers are open for approximately 24 hours post-pollination, there is also opportunity for sequential deposition. Observations Chapter 4 88 made in the 2011 field season found one case of distinct pollen deposits made over two separate events within a 24 hour period, confirming that sequential deposition does occur.
Paternity assignment and pollen flow
By a combination of maximum likelihood and manual reconstruction methods we successfully assigned paternity to offspring in 21 out of 51 progeny arrays combined across taxa. Only one of these assigned fathers sired more than flower (two flowers) indicating that no genotype was of substantially higher male fitness than any other successful sire. The combined distribution of pollen movements inferred from paternity assignment ranged 0-291 m with a median of 14 m. This was not dissimilar from the distribution and average movements found in related orchid-pollinating wasps; 15-27 m in N. cryptoides (Whitehead and Peakall 2012), 32 m in Zaspilothynnus trilobatus (Peakall
1990) and 17 m in two species of Thynnoides pollinating Caladenia tentaculata (Peakall and Beattie 1996). The occasional long range dispersal (in excess of 100 m) seen here is also characteristic of these wasp flight/orchid pollen displacement distributions.
Why are some progeny unable to be assigned?
Given the demonstrated resolution of our loci (Figure 1), and our sampling of every known pollen donor or recipient on the site over five seasons, we were surprised not to obtain a higher number of paternal assignments. One possible reason for this was undoubtedly the difficulty of obtaining sufficiently large progeny arrays via symbiotic germination. This was a problem further exacerbated by the extent of multiple paternity which quickly diminishes available information for any single father for a given progeny array size. A contingency test showed no significant effect of multiple paternity on our ability to assign paternity here, however our P-value (0.0575) was very low and would drop further if we had underestimated the minimum number of fathers at all, which is likely. Although Chapter 4 89 simulations showed high accuracy in paternity assignment for 12-offspring arrays (even in cases where paternity was evenly split between two fathers) power dropped very quickly when minority share fathers were introduced at correlated paternity 0.9 (Figure 4).
Where we did assign fathers to multiply sired capsules, these generally showed a departure from even paternity as seen in KE (Table 5), which should approach the number of pollen donors (K) as the proportional share in paternity approaches equality
(Bernasconi 2004). If pollen donors usually contribute unequally to the progeny array, as seen in other studies of paternity and carryover (Karron et al. 2006), then this might explain why we were unable to assign paternity for many of our arrays.
Another explanation for not assigning fathers to all offspring may be genotyping error.
Although we made efforts to ensure the quality of our data by genotyping mothers at least twice, rerunning or discarding ambiguous samples or those with too much missing data we cannot rule out the error rate that will always be inherent in microsatellite genotyping of low quality DNA samples. In particular, errors such as dropout and null alleles can lead to false exclusion of a real parent and these types of errors are inflated with low DNA concentrations such as those yielded by orchid protocorms (Dakin and Avise 2004).
Interestingly, in studies of the movement of N. cryptoides (a closely related species of similar size to the wasps studied in this system) we observed hints of a bimodal dispersal behaviour whereby wasps commonly remained in a localised area (<15 m) for days before or after making less common long distance flights across the study site (Chapter 6). It is very difficult to capture data describing the long-tail of these kind of dispersal curves however pollen flow from outside the study site (greater than 150 m) could explain why we were unable to assign a significant number of offspring. Pollen carryover also enhances long distance pollen flow as pollen grains from a single pollinium remain on a wasp for a longer time and well beyond the first flower visit. Our sampling design was made in light of flight distances known from N. cryptoides and other wasp studies (Peakall 1990; Peakall and Beattie 1996; Whitehead and Peakall 2012) which showed high site fidelity and short Chapter 4 90 movements. Mitchell et al (2009) note however that using pollinator flight distance as a proxy for pollen dispersal can result in underestimates of gene movement, especially in the presence of pollen carryover (Broyles and Wyatt 1991).
Conclusion
Although comprising a fundamental component of reproductive success, seed paternity has received much less attention than studies that quantify outcrossing–selfing rates.
Here, as predicted, we have found that pollination via sexual deception promotes very high rates of outcrossing. Further, pollinator behaviour is such that selfing is usually avoided despite clonality. We found extensive multiple paternity in this system which is probably a joint result of pollen carryover and sequential deposition creating mixed stigmatic pollen loads. The fitness implications of multiple paternity in this system are uncertain and future studies on the potential for pollen competition and offspring quantity and quality in multiply sired capsules are needed to address this. Comparative studies across orchids could also consider whether fitness benefits to pollen carryover and multiple paternity might be an adaptive driver in the evolution of pollinia morphology.
An unexpected finding is the support here for substantially longer pollen flow distances than previously thought. These small insects have been shown to display high site fidelity
(Whitehead and Peakall 2012), however occasional long distance movements coupled with pollen carryover might result in pollen flow in excess of 150 m. These findings together reinforce and build on our understanding that sexual mimicry is a superb pollination strategy for maximizing outcrossing and promoting long-distance pollen dispersal.
Table 1: Summary statistics on observed pollination, plants collected and seed families
successfully germinated during the study.
20071 2008 2009 2010 2011 TOTAL
Chiloglottis valida
Pollen removals - 19 30 11 33 93
Pollen deposits - 11 34 7 22 74
Pollination efficiency - 0.58 1.13 0.64 0.67 0.80
Plants collected 9 9 28 - - 46
Seed families germinated / 1 7 14 - - 22
genotyped
Chiloglottis aff. jeanesii
Pollen removals - 33 16 18 42 109
Pollen deposits - 23 17 17 36 93
Pollination efficiency - 0.70 1.06 0.94 0.86 0.85
Plants collected 14 20 14 - - 48
Seed families germinated / 5 16 8 - - 29
genotyped
1: Field work in 2007 occurred later than peak flowering so that most pollinated flowers had already closed and began to swell. This prevented accurate measures of pollen removal from being made that year.
Figure 3: Results of symbiotic germination trials of hand cross seed. Points depict mean growth index with bars for standard error. Shaded bars depict percentage of plates that
showed germination.
Figure 1: Resolution (unique multilocus genotypes recovered) for each possible locus
combination. The boxes are bounded by the most and least informative combinations of
loci while the inner line is the mean value for all combinations of n loci.
Figure 2: Frequency histogram of maximum clone size in metres for Chiloglottis valida and C. aff. jeanesii.
Table 2: Mating system parameters for two sympatric Chiloglottis species.
ts (SD) tm (SD) tm-ts rp
C. valida 0.925 (0.030) 1.001 (0.048) 0.076 0.647 (0.074)
C. aff. jeanesii 0.874 (0.070) 0.924 (0.049) 0.050 0.765 (0.075)
tm, multilocus outcrossing rate; ts, average single locus outcrossing rate; SD based on
bootstrap procedure; tm-ts, biparental inbreeding; rp, correlated mating within maternal progeny arrays.
Table 3: Proportion of correct estimates for minimum number of fathers in simulated data.
Both strict parsimonious reconstruction and maximum-likelihood COLONY derived
estimates are reported over a range correlated paternity.
Correlation of paternity
0.5 0.9 1.0
C. valida
Reconstruction 0.99 0.66 1.00
COLONY 0.60 0.40 0.30
C. aff. jeanesii
Reconstruction 0.94 0.54 1.00
COLONY 0.40 0.30 0.50
Figure 4: Accuracy of paternity assignment in COLONY for simulated data across three levels of correlated paternity for C. aff. jeanesii (open circles) and C. valida (closed circles).
Correlated paternity is calculated as the average ratio of contribution of two fathers in 10
replicate broods of 12 offspring. (a): Overall proportion of the family with correctly assigned paternity. (b): Proportion of assignments given probability > 0.8 actually correct. Table 4: Descriptive population genetic statistics for six microsatellite loci in adult and
offspring cohorts of two sympatric Chiloglottis orchid species.
CV0420 CV0423 CV0425 CV0430 CV0432 TRI03 Mean SE
C. valida adults (n = 216)
F 0.031 0.002 0.131 0.034 0.074 0.219 0.084 0.033
Ho 0.577 0.746 0.736 0.740 0.685 0.662 0.691 0.027
He 0.595 0.748 0.847 0.765 0.740 0.847 0.757 0.038
HWE *** * *** ns *** ***
C. aff. jeanesii adults (n = 262)
F 0.030 -0.005 0.058 0.295 0.066 0.085 0.088 0.043
Ho 0.705 0.533 0.611 0.126 0.650 0.748 0.562 0.092
He 0.727 0.531 0.649 0.179 0.696 0.817 0.600 0.093
HWE *** *** *** *** ns ***
C. valida offspring (n = 304)
F 0.111 -0.026 0.031 0.000 0.140 0.157 0.069 0.032
Ho 0.546 0.864 0.709 0.686 0.661 0.726 0.700 0.040
He 0.625 0.829 0.741 0.681 0.761 0.869 0.752 0.037
HWE *** ** *** ns *** ***
C. aff. jeanesii offspring (n = 325)
F 0.022 0.025 0.111 0.461 0.102 0.001 0.120 0.071
Ho 0.803 0.464 0.592 0.054 0.629 0.813 0.536 0.097
He 0.715 0.467 0.626 0.053 0.683 0.791 0.579 0.087
HWE ns ns ** *** ** *** ns=not significant, * P<0.05, ** P<0.01, *** P<0.001
Table 5: Paternity and mating system analysis for 57 naturally pollinated Chiloglottis seed
families. Pollen load provides an estimate of number of pollinia on the mother plant’s stigma and was not able to be scored for some mothers. Minimum number of fathers are determined by
manual paternal genotype reconstruction unless otherwise noted. Each positive paternal assignment is supplied with the method by which it was obtained (R: manual reconstruction, C:
COLONY), the proportion of the brood sired by that father and the inferred pollen movement.
Single (ts) and multi-locus (tm) outcrossing as reported by MLTR are reported with standard
errors.
Mother Progeny Pollen Min # of Inferred Method Effective Proportion Pollen ts (SE) tm (SE)
>5 loci load fathers father fathers (KE) sired distance (m) Chiloglottis aff. jeanesii CV2431 13 1 2° Self* R/C 1.742 0.69 0 0.131 (0.079) 0.193 (0.239) CV2496 6 1-2 1° Self* C 1.000 1.00 0 0.413 (0.212) 0.343 (0.384) CV2441 14 1 1§ Self* R/C 1.000 1.00 0 0.096 (0.162) 0.518 (0.194) CV2261 7 1 1§ CV2260* R/C 1.000 1.00 0 0.783 (0.075) 1.03 (0.006) CV2500 14 1 1 CV2981* R 1.000 1.00 0 1.077 (0.023) 1.002 (0.001) CV2451 14 1 1 CV2518* R/C 1.153 0.93 0 0.609 (0.079) 0.971 (0.048) CV2440 11 1 2§ CV3054* R/C 1.424 0.82 5 1.393 (0.116) 0.983 (0.046) CV2534 13 >2 1 CV3141* C 1.550 0.23 10 1.012 (0.081) 1.027 (0.001) CV2519 17 2 1 CV2716 R 1.410 0.82 14 0.648 (0.087) 0.936 (0.070) CV2528 6 >2 1§ CV2260* R/C 1.000 1.00 15 1.059 (0.097) 1.006 (0.003) CV2452 9 - 1§ CV2918* C 1.000 1.00 15 0.821 (0.067) 0.982 (0.020) CV2518 7 2 1§ CV3170* R/C 1.000 1.00 16 0.93 (0.049) 1.001 (0.001) CV2474 12 1-2 1 CV3011 R/C 1.385 0.83 23 0.942 (0.072) 0.993 (0.010) CV2453 10 1-2 1§ 1.412 (0.066) 1.026 (0.005) CV2450 9 2 1§ 1.036 (0.018) 1.016 (0.006) CV2253 8 - 1§ 0.577 (0.160) 1 (0.001) CV2499 13 - 2 0.917 (0.033) 1.006 (0.005) CV2477 14 1 2 1.106 (0.063) 1.013 (0.003) CV2475 18 >2 2 1.204 (0.023) 1.021 (0.005) CV2427 13 - 2 0.835 (0.096) 1.018 (0.001) CV2256 4 1-2 2 1.226 (0.128) 1.02 (0.007) CV2520 12 2 2 0.698 (0.092) 0.927 (0.062) CV2490 10 1-2 1 1.314 (0.017) 1.014 (0.005) CV2429 13 - 1 0.628 (0.089) 1.003 (0.001) CV2428 12 1-2 1 1.018 (0.059) 1.016 (0.002) CV2426 14 - 1 1.235 (0.026) 1.011 (0.001) CV2425 2 >2 1 1.225 (0.074) 1.045 (0.014) CV2257 12 - 1 1.008 (0.004) 1 (0.001) CV2249 12 - 1 0.936 (0.118) 1.01 (0.002) CV2436^ 6 >2 1§ Self* C 1.000 1.00 0 0.24 (0.135) 0.902 (0.115) Chiloglottis valida CV2454 3 1 1 Self* R/C 1.000 1.00 0 0.312 (0.049) 1.001 (0.001) CV2461 7 2 1§ CV3162 R/C 1.000 1.00 8 0.466 (0.136) 0.936 (0.059) CV2448 15 1-2 1§ CV2535* R/C 1.000 1.00 10 0.899 (0.056) 1 (0.001) CV2432 19 - 3° CV3055* R/C 2.756 0.47 14 0.807 (0.045) 1.002 (0.001) CV2709* R/C 0.26 36 CV2766* R/C 0.26 69 CV2551 11 - 1§ CV3034* R/C 1.000 1.00 19 0.604 (0.081) 1.001 (0.001) CV2254 4 >2 2§ CV3080* R/C 1.600 0.75 21 0.815 (0.258) 0.687 (0.274) CV2471 8 <1 1° CV2493* R/C 1.000 1.00 33 0.767 (0.110) 1 (0.001) CV2531 4 1 1° CV2557* C 1.000 1.00 291 0.937 (0.176) 1.004 (0.002) CV2521 15 2 1§ 0.212 (0.101) 1.001 (0.001) CV2479 8 <1 1§ 0.987 (0.052) 1 (0.001) CV2508 8 1-2 1§ 0.971 (0.055) 1 (0.001) CV2442 7 1-2 1§ 0.946 (0.088) 1.001 (0.001) CV2511 11 >2 2 0.996 (0.040) 1 (0.001) CV2509 14 >2 2 0.832 (0.079) 1.001 (0.001) CV2535 12 >2 2 1.13 (0.111) 0.848 (0.147) CV2498 14 >2 2 0.955 (0.053) 1 (0.001) CV2522 18 >2 2 0.818 (0.055) 1 (0.001) CV2529 13 1-2 2 1.305 (0.028) 1.004 (0.001) CV2466 16 1 2 0.891 (0.043) 1 (0.001) CV2523 12 - 1 0.876 (0.043) 1 (0.001) CV2486 10 <1 1 0.788 (0.105) 1.002 (0.001) CV2479 16 2 1 1.145 (0.029) 1 (0.001) CV2270^ 4 1 1§ Self* C 1.000 1.00 0 0.249 (0.261) 0.972 (0.025) CV2237^ 10 >2 1° 1.143 (0.031) 1.001 (0.001) CV2445^ 14 1-2 2 1.011 (0.004) 1 (0.001) CV2437^ 16 >2 2 0.445 (0.090) 0.919 (0.069) CV2540^ 15 1-2 1 0.872 (0.041) 1.002 (0.001) ^: Sample not collected on the transect §: Manual reconstruction and COLONY in agreement °: COLONY-derived estimate. Because genotyping error quickly inflates minimum inferred fathers via strict reconstruction, we take the error-tolerant COLONY measure if lower. *: COLONY assignment probability > 0.8. Chapter 5 102
CHAPTER 5
MICRODOT TECHNOLOGY FOR INDIVIDUAL MARKING OF SMALL
ARTHROPODS.
Neozeleboria cryptoides bearing personalized microdot tag Chapter 5 103
ABSTRACT
1. Individual mark-release-recapture is an important method for gathering data on
insect movement but is limited by the constraints of tagging small insects with
individual information.
2. Microdots, originally developed for covert security applications, are small polymer
discs (0.5 mm diameter) bearing up to 26 characters of information and have
potential as an alternative to the larger bee tags. Here we test microdots for the
individual marking of a 9 mm parasitoid wasp.
3. We individually marked 505 wasps. Recapture rate was 24% of individuals over
189 recapture events for which 84% retained legible microdot labels. Movement
ranged from 0-161 m with mean displacement 21.2 ± 2.7 m. A captive survival
experiment showed no difference in lifespan between marked and unmarked
wasps.
4. This study shows that microdots can provide an effective, durable, low-cost
method for individually tagging small insects. The technique offers new
opportunities by greatly expanding the capability for individually marking small
insects, shifting the minimum size below that of bee tags – the only other
manufactured option for individualized miniature marking.
Chapter 5 104
INTRODUCTION
Marking insects in the field is essential for quantifying basic demographic properties such as population size, survival and movement – data essential for the management of both pest and beneficial insects in natural and agricultural systems. For pest control, measurements of pest insect dispersal are essential for establishing the scale over which management should operate, identifying suitable habitat and assessing connectivity between populations (Mahroof et al. 2010), while quantifying population densities and survivorship allows effective monitoring of management outcomes. Data on the movement of introduced control agents for biological control of pests or weeds is also crucial for predicting their spread from release points (Corbett and Rosenheim 1996; Bianchi et al.
2009; Moerkens et al. 2010), monitoring their populations and predicting impacts on non- target species (Rudd and McEvoy 1996; Chapman et al. 2009). Conservation of beneficial insects requires dispersal data to delimit management units and the movement of pollinator species is also becoming increasingly important for restoration of pollinator services in degraded ecosystems (Menz et al. 2010) and assessing the risks of engineered genes escaping from genetically modified crops (Ellstrand 2003).
The most common approach to quantifying insect movement is the suite of sampling and analysis techniques encompassed by mark-recapture. For these studies, marking techniques typically take one of two approaches (Hagler and Jackson 2001). Mass-marking techniques using paints, proteins, trace elements, fluorescent powders or dyes to quickly mark very large numbers of insects have been employed extensively for at least 75 years
(Hagler & Jackson 2001 and references therein) however they cannot provide individual level data. This limits their application to marking or release at certain points in the landscape and measuring diffusion from that point. The alternative approach is individualized marking, such as numbered bee tags, which offers the flexibility to mark individuals at any time or place in the field and avoids the unnatural population densities created by mass release (Bancroft and Smith 2005). Detailed individual level data provides the opportunity to resample the same individuals over time as well as measure individual Chapter 5 105 traits, such as sex or size, which might affect survival and movement (Bancroft and Smith
2005). Marking insects with individual tags is also usually non-destructive, unlike techniques such as trace element marking (Berry et al. 1972, Hagler & Jackson 2001).
Despite these advantages, individual marking techniques are usually more time consuming and labor intensive than mass-marking and are limited to larger insects able to physically bear the chosen tag.
This restriction on size has been the focus of various attempts to adapt technology to tracking individual insects. The ongoing development of harmonic radar and radio technology for tracking insects has proved promising (Mascanzoni and Wallin 1986; Riley et al. 1996; O’Neal 2004; Sword et al. 2005; Wikelski et al. 2010), but the technology remains costly and still limited to larger insects (O’Neal 2004). Passive RFID chips have now miniaturized sufficiently to tag bees (Pahl et al. 2011), paper-wasps (Sumner et al.
2007) and ants (Robinson et al. 2009), however the chips are required to come into close proximity with the chip reader which makes them unsuitable for some systems and potentially expensive (Moreau et al. 2011).
Here we describe a novel, simple and cost effective approach for individualized mark- release-recapture of small insects that combines traditional marking methods with new microdot technology. Microdots, developed for covert security and authentication applications, are polymer discs manufactured to bear a unique identifying code. For insect marking they are analogous to bee tags and the labels developed for chrysomelid beetles
(Piper 2003). Bee tags with diameter 2.5 mm are obviously limited to insects with a single segment larger than 2.5 mm, which makes them unsuitable for use on most insects smaller than around 10 mm total length, and especially for insects with slender bodies. The small size of microdots (0.5 mm diameter, 25 µg) not only expands the lower size limit for individual insect marking but should make them less disruptive to insect flight than larger tags. Further advantages include the capacity to carry more information than bee tags with customizable flexibility, such that researchers may have any combination of up to 26 Chapter 5 106 characters inscribed on them. To test microdots in an experimental mark-release- recapture study we applied them to quantify the movement of the parasitoid thynnine wasp Neozeleboria cryptoides (Smith) (Hymenoptera, Tiphiidae).
The biology of the thynnines is virtually unknown, but it is thought that they parasitize wood or soil dwelling larvae of ants and beetles (Ridsdill Smith 1970; Brown 1998), which may make them a potential biological control agent for scarab larvae (Ridsdill Smith
1970). The females are wingless and dwell in the soil, only surfacing to mate with males.
The male wasps are strong fliers, 9 mm average length and weigh 16 mg (Schiestl 2004). A
2008 pilot study marking individuals with correction fluid, showed that despite being highly mobile, N. cryptoides rarely moved more than 25 m between recaptures with a maximum displacement recorded of 85 m (Whitehead, unpublished data).
Here we report the development and implementation of a novel application of microdot technology as a method for individually marking small wasps. As proof of concept, we apply the technique to measuring individual movement in a population of Neozeleboria cryptoides. We assess the efficacy, benefits and limitations of the technique and conclude by considering new opportunities for application of the method.
MATERIALS AND METHODS
Mark-release-recapture
Field work was conducted in open Eucalyptus woodland in Canberra, Australia, chosen for the known abundance of Neozeleboria cryptoides. We attracted male wasps with techniques previously reported (Schiestl et al. 2003; Schiestl and Peakall 2005; Peakall et al. 2010c), whereby patrolling male N. cryptoides wasps were attracted to synthetic chiloglottone1 (Franke et al. 2009), when presented on a plastic pin-head dummy and captured in a net. In order to implement microdots for mark-release-recapture we needed to both attach the dots and recognize recaptures. Correction fluid and nail varnish have Chapter 5 107 been used widely in insect marking studies without toxicity (White 1970; Peakall 1990;
Piper 2003; St Pierre et al. 2005), therefore they were both used as a colour mark and adhesive in this study. The use of coloured adhesives was important for quickly recognizing recaptures and providing a background contrast against which to read the microdot codes. We painted wasps with one of six coloured adhesives to which microdots were applied before the adhesive dried (Figure 1). The microdots were applied with a plastic pipette-tip to which they adhered when the tip was statically charged on clothing.
We ordered 100 uniquely numbered batches manufactured by DataDot Technology
(www.datadotdna.com, Frenchs Forest, Australia), with each batch containing 1000 microdots, costing approximately $15 (AUD) per batch . For the six coloured adhesives used we had 100 unique codes for individual marking giving a total of 600 unique labels and capacity to tag several thousand wasps.
Data collection was confined to periods of wasp activity which were necessarily warm
(>19° C), fine days without strong wind, resulting in a sampling window of between 1100 and 1700 hours on any suitable day. Phase one (mark phase; 28 Aug 2009 – 14 Sep 2009) of the study involved capture, marking, release and recapture of marked wasps at nine fixed bait stations (two four-point rectangular arrays 15 x 30 m separated by approximately 85 m and one point 20 m from the nearest array; forming a polygon 299 m2). This was intended to quickly establish a large proportion of marked individuals from known locations. Phase two (roam phase; 15 Sep 2009 – 25 Sep 2009) consisted of random sampling across the site in the form of timed recapture trials. The roam phase sampling area was delineated by a habitat boundary of grassland where there were few wasps. A single recapture trial lasted for five minutes with all males approaching the synthetic pheromone captured and scored as either an unmarked capture or a recapture event where there was evidence of a prior mark adhering to the cuticle. Microdots on recaptured wasps were read with a handheld 50x pen-microscope (ProSciTech, Kirwan,
Australia) and position was recorded with a GPS unit. The baits were transported between trials in a sealed container. Throughout both phases of the study, wasps that were Chapter 5 108 recaptured carrying adhesive remnants but lost microdots were refreshed with a new paint and microdot combination before re-release.
Survival experiment and statistics
In order to assess potential toxicity of the marking technique on wasps we kept both marked (n = 51) and unmarked (n = 27) wasps in standard lab conditions fed on sugar solution. The two groups were housed together and each day for 40 days we recorded data on mortality. Difference in survivorship curves between the treatment and control groups was tested by two Wilcoxon test methods implemented in R; a log-rank test that assumes constant proportional hazard between groups and a Gehan-Breslow weighted Wilcoxon test which weights deaths earlier in time more heavily (Klein & Moeschberger, 2003). This latter test we felt appropriate for detecting immediate or short term deleterious effects of marking the wasps. Approximate abundance was estimated by fitting a Poisson log-normal mark-resight model to the recapture data as described by McClintock and White (2009).
This was implemented in MARK 2.0 (White & Burnham, 1999) assuming an open population and no heterogeneity in capture probability among individuals (Whitehead, unpublished).
RESULTS
We marked a total of 505 individual wasps in 26 person/hours over nine days during the mark phase of the study. The roam phase effort totaled 17.2 person/hours in nine days over 214 recapture trials. In total, we recorded 189 recapture events through both phases of the study. Of these, 30 (16%) recaptures were of wasps which had lost their microdot and could not be individually identified. The remaining 159 recapture events for which we recorded individual data were broken down into: 87 individuals recorded for single recaptures, 31 individuals recaptured twice, four individuals recaptured three times and a Chapter 5 109 single individual recaptured four times. Altogether, there were 123 individual wasps recaptured representing 24% of the total marked pool of animals.
As an indicator of survival, mean time between first and last captures for all recaptured individuals was 8.8 days (SD = 5.9) with range 1-24 days. Release to recapture distances varied from 0-161 m and 0-150 m for the mark and roam phases respectively. Mean displacement was 15.5 ± 3.4 m, n = 81, median = 0 for mark phase recaptures and 27 ± 4.1 m, n = 78, median = 14.8 m for roam phase recaptures. The captive survival experiment resulted in similar mean survival (± SE) for marked (11.8 ± 2.1 days) and unmarked wasps
(13.7 ± 1.5 days) and survival curves did not differ significantly (logrank χ2 = 0.1, df = 1, p
= 0.763; Wilcoxon Gehan statistic = 3.419, p = 0.181). The Poisson log-normal model derived population size for the beginning of the mark phase was estimated to be 3454 (±
302) individuals. This is broadly within the range of expectation given the average proportion of marked to unmarked captures of 0.12 (SD = 0.23) and the total marked insect pool of 505.
DISCUSSION
The technique allowed us to uniquely label 505 individual wasps in the field at a rate of 20 individuals per person hour. This rate of marking includes the time taken to attract, net and mark wasps, record data and move between subsequent baiting sites. It is therefore a reasonably labor-intensive process; however marking with any unique identifying tag (eg bee tags, RFID chips) will require similar effort. Durability was demonstrated by 189 recapture events, with 84% of recaptured insects bearing a legible microdot label and legible recaptures collected up to 24 days after initial marking. While the lifespan or flight time of adult wasps in the wild is unknown, the availability of wasps is restricted to 4-5 weeks per year suggesting, along with the results of this study, flight times for individuals of 3-4 weeks. The overall mean survival for captive wasps was 12.5 days after capture and one marked individual remained alive at day 40 when the study was ceased. Wasp Chapter 5 110 movement documented in this study was in the range predicted by a 2008 pilot study that used only correction fluid and nail varnish to mark individuals (Whitehead, unpublished data) which suggests the additional treatment of microdot attachment has not affected behaviour. This study relies on a sexual response in order to recapture males and the rate of recapture (24%) suggests that marking has not negatively impeded wasp mate-search behaviour. Due to the largely fossorial lifestyle of the female thynnine it is rare to find mating wasps in the field, but in 2009 the only observed coupling was with a marked male showing that marking does not prevent subsequent successful mating.
This study shows that microdots can provide a simple, relatively cost effective and useful method of individualized marking. Microdots could enable mark-release-recapture studies in a wider range of insects than presently possible, shifting the minimum size range below that of bee tags—the only other manufactured option for miniaturized marking. Their small size makes them ideal for any small invertebrate with a hard surface larger than about 1 mm diameter for attachment. Limitations at this size range involve the method of adhesion as glues and paints can be difficult to apply with such precision and smaller areas of application will likely require adhesives of higher visibility. The loss of some microdots (16% of recaptures were missing microdots) shows that there is scope for using adhesives that improve label retention. Microdots attached with nail varnish were disproportionately represented in the total of cases where microdots were lost (half of losses but only one sixth of marked insects), therefore correction fluid proved to be the most effective adhesive used here. The tendency of these materials to flake or chip from insect cuticles has been noted before (Wineriter & Walker 1984) but possible alternatives might be found in the range of cyanoacrylate glues already employed to attach harmonic radar tags to insects (Boiteau et al. 2009) provided the glue can be blended with colour to aid detection and readability of the microdot.
Pest management, insect conservation and land management of natural and agro- ecosystems provide many scenarios for employing miniaturized individual tagging. Chapter 5 111 Microdots provide another tool to explore insect movement in these contexts, especially for parasitoid wasps and pollinating bees and flies smaller than the honeybee for which this study serves as proof of concept. Future tests are needed to evaluate applications for marking more challenging species such as burrowing insects or those with hairy cuticles.
There may also be opportunities to utilize this technology outside insects. Aquatic invertebrates such as molluscs and crustaceans with hard shells could be tagged if water- proof adhesives are used. Small seeds could potentially be labeled with microdots in order to trace dispersal and the technology might even be able to be exploited as a pollen analogue in studies of pollen dispersal. Effective markers should be durable, low-cost, non- toxic, easily applied, clearly identifiable and avoid interfering with the animal’s behaviour
(Hagler and Jackson 2001). We conclude that microdots promise an effective new marking technique for any future study employing individual tags.
Figure 1: Anterior of male Neozeleboria cryptoides thorax showing attached microdot.
Image credit: © Commonwealth Scientific and Industrial Research Organisation Chapter 6 113
CHAPTER 6
SHORT TERM BUT NOT LONG TERM PATCH AVOIDANCE IN AN
ORCHID-POLLINATING SOLITARY WASP
Neozeleboria cryptoides on the head of a pin laced with synthetic pheromone
Chapter 6 114
ABSTRACT
The success of exploitative attraction of insect pollinators to rewardless flowers may depend on a constrained capacity for learning. In the case of sexually deceptive orchids, the extent to which pollinators can avoid dishonest signals through learning or adaptation is poorly known. We used field experiments with synthetic pheromone baits in concert with novel miniaturized marking techniques to investigate patterns of behavior and movement in Neozeleboria cryptoides, the wasp pollinator of the sexually deceptive orchid Chiloglottis trapeziformis. In trials of 4 minutes and 60 minutes duration, visitation rates to synthetic sex pheromone declined rapidly after the first minute and remained low, suggesting short term avoidance. Using spatially explicit capture recapture models we then assessed if wasps maintained this avoidance for more than 24 hours. Among our four competing behavioral models, the best supported model was one which showed an increase in detection probability at a location for wasps that had previously been caught at that location. Therefore, we found no evidence for long term patch avoidance. If spatial learning underpins the short term avoidance we observed, then this information appears not to be retained beyond 24 hours. The typical patterns of N. cryptoides movement (range = 0 – 161 m, median = 14.8) coupled with short term patch avoidance likely promote outcrossing in the clonal, self- compatible orchid it pollinates.
Chapter 6 115
INTRODUCTION
In insects, learning is increasingly being recognized for its importance to many behaviors (Dukas, 2008). Spatial orientation (Collett, 1992), foraging (Dukas and
Real, 1991; Vet, 1995; Menzel, 1999; Cunningham et al., 2004) and mating (Dukas,
2006) have all been shown to be influenced by learning. Although the field of insect learning is dominated by studies of honeybees (von Frisch, 1953; Takeda, 1961;
Menzel, 1999) and Drosophila (Quinn et al., 1974; Dudai et al., 1976; Liu et al.,
1999), other groups such as moths (Cunningham et al., 2004; Goyret et al., 2008;
Riffell et al., 2008) and wasps (Vet, 1995; Olson et al., 2003; Sheehan and Tibbetts,
2011) have shown the ability to learn under various scenarios.
The operation of learning in insects may be of considerable importance in the pollination of deceptive flowers—rewardless flowers that attract their pollinators through mimicry of a resource, such as food or a mate (Dafni, 1984). This is because the ability of a pollinator to learn to distinguish between honest and deceptive signals may strongly influence fitness in terms of flower visits, pollen removal and pollen movement. The influence of learning on pollination efficiency has been studied in some food deceptive systems where it has been found that food deceptive flowers typically attract naïve insects (Nilsson, 1980) and learning therefore might underpin the maintenance of some floral polymorphisms
(Smithson and MacNair, 1997; Gigord et al., 2001). Sexual deception, in which mate seeking male insects are lured to flowers through chemical and visual mimicry of receptive females, has by contrast been largely neglected in studies of pollinator learning. This is perhaps surprising given the bank of research on mechanisms of pollinator attraction and the adaptive significance underlying the repeated Chapter 6 116 evolution and maintenance of this pollination strategy (Cozzolino and Widmer,
2005; Schiestl, 2005; Gaskett, 2010; Xu et al., 2012).
One behavior that has been observed across several species of sexually deceived hymenopterans is site-specific patch avoidance. In the field, the numbers of males responding to experimentally presented flowers drops sharply in only a few minutes (Peakall, 1990; Peakall and Beattie, 1996; Alcock, 2000; Ayasse et al.,
2000; Peakall and Schiestl, 2004; Peakall et al., 2010) and may indicate that male wasps learn to avoid the source of deceptive sex pheromone. This behavior is compatible with the hypothesis that sexual deception promotes outcrossing
(Peakall and Beattie, 1996) and may also impact individual fitness in populations of exploited pollinators (Wong and Schiestl, 2002). However, until now the retention of this site aversion has never been investigated over a period longer than several hours (Peakall, 1990).
Here we use field experiments and capture-mark-recapture (CMR) to investigate short and long term patterns of behavior and movement in the solitary wasp
Neozeleboria cryptoides, the specific pollinator of the sexually deceptive orchid
Chiloglottis trapeziformis. Sex pheromone attraction in this species is achieved by a single compound chiloglottone 1 (Schiestl et al., 2003; Franke et al. 2009) and since elucidation of this novel compound, synthetics have been used to probe wasp behavior and orchid-pollinator interactions through manipulative field experiments (Wong and Schiestl, 2002; Wong et al., 2004; Schiestl and Peakall,
2005). This species is therefore an ideal system in which to further test hypotheses about orchid-pollinator interactions. Furthermore, by employing recently developed microdot tagging techniques, we can access detailed individual-level Chapter 6 117 data previously unobtainable due to the small size (~10 mm length) of these wasps (Whitehead and Peakall, 2012).
The goal of this study was to implement a field study with the specific aims to quantify short term patch avoidance in N. cryptoides and use individual based CMR models to assess whether or not there is sustained long term patch avoidance.
Patch avoidance for a period longer than 24 hours might indicate the operation of memory or learning of the deceptive stimulus. The ability to uniquely mark individuals allows us to employ recently developed spatially explicit capture- recapture models (Efford, 2004; 2011a) to investigate avoidance behaviors in the field, whereas traditionally these kinds of behaviors are studied with manipulative experiments in a lab setting (Robacker et al., 1976; Gaskett et al., 2008). The tools for studying spatial avoidance under the CMR analytical framework are also well established, for example testing learning models has long been of interest to practitioners of CMR studies who often trap intelligent birds or mammals capable of modifying their behavior as a learned response to being trapped (Seber, 1970;
Royle et al., 2011). In addition to establishing the timeline of avoidance in deceived wasps we also take advantage of detailed field data on wasp movements to explore the implications of wasp behavior on pollen movement and outcrossing in the sexually deceptive orchids they pollinate.
METHODS
Field work was conducted in an open eucalyptus woodland in Canberra, Australia, chosen for the known abundance of Neozeleboria cryptoides. The populations studied were orchid-naïve, in that orchids known to attract N. cryptoides were not present in the area. We attracted male wasps with techniques previously reported Chapter 6 118 (Schiestl et al., 2003; Schiestl and Peakall, 2005; Peakall et al., 2010), whereby patrolling male N. cryptoides wasps were attracted to synthetic chiloglottone 1
(Franke et al., 2009) when presented on a plastic pin-head dummy and subsequently captured in a net. Data collection was confined to periods of wasp activity which were necessarily warm (>19° C), fine days without strong wind, resulting in a sampling window of between 1000 and 1700 hours on any suitable day. We collected data for various components of the study over different years; wasp attraction curves were described from data collected in 2007, 2008 and 2011 while the CMR study took place in 2009.
Wasp attraction curves
In order to measure and confirm short-term patch avoidance in N. cryptoides we carried out four minute and 60 minute baiting trials. A total of 166 four minute trials were carried out over multiple days at two sites approximately 10 km apart.
We gathered data in September/October 2007/2008 for 54 and 112 trials respectively. A single trial consisted of the presentation of the synthetic pheromone bait and counting approaches, landings and attempted copulations each minute for four minutes. If wasps had not responded in the first minute of a trial, the trial was abandoned and a new trial was started in a different location.
Three replicate 60 minute trials were carried out at two points. Pheromone baits were presented and left in position for 60 minutes. During this time, the frequency of wasp approaches and landings were scored together for each minute up to ten minutes and for every five minutes thereafter. Wasps were captured, marked and released throughout these trials. The work took place on three consecutive days between 21 Sep 2011 and 23 Sep 2011, with one trial at each bait station per day. Chapter 6 119 Capture-mark-recapture
Individual wasps were attracted with pheromone baits and marked with microdots (DataDot Tech, Australia), secured to the insect’s thorax with either nail varnish or correction fluid as described in Whitehead and Peakall (2012).
Microdots, originally developed for covert security applications, are small polymer discs (0.5 mm diameter) bearing up to 26 characters of information providing effective miniature individualized marking for small insects. Phase one (‘Mark- phase’; 28 Aug 2009 – 14 Sep 2009) of the study involved capture, marking, release and recapture of marked wasps at nine fixed bait stations (two four-point rectangular arrays 15 × 30 m separated by approximately 85 m and one point 20 m from the nearest array; forming a polygon 299 m2). This was intended to quickly establish a large proportion of marked individuals sampled from the same set of locations each day. Phase two (‘Roam-phase’; 15 Sep 2009 – 25 Oct 2009) consisted of timed recapture trials carried out at random locations across the entire site. The roam phase sampling area was delineated by a habitat boundary of grassland where there were few wasps. A single recapture trial lasted for five minutes with all males approaching the synthetic pheromone captured and scored as either an unmarked capture or a recapture event where there was evidence of a prior mark adhering to the cuticle. Microdots on recaptured wasps were read with a handheld 50x pen-microscope (ProSciTech, Kirwan, Australia) and position was recorded with a GPS unit. The baits were transported between trials in a sealed container.
Chapter 6 120 Information theoretic analysis and model selection
The information theoretic (IT) approach to making biological inferences has recently garnered significant attention in the evolution and ecology literature as an alternative to traditional null hypothesis testing (Johnson and Omland, 2004;
Stephens et al., 2007; Grueber et al., 2011; Symonds and Moussalli, 2011). Briefly, model selection is based in likelihood theory and investigates multiple and competing hypotheses (models) by fitting them to the observed data. Models can be compared and ranked by their relative support of the data after which the most informative or ‘best fit’ models can be used to make predictions and generate estimates for parameters of interest. Although model selection has relatively recently come to wider use within behavioral ecology (see special issue 1:65
Behavioral Ecology and Sociobiology), its established use in the analysis of CMR ecological data provides a rich and accessible domain of data analysis (Johnson and
Omland, 2004).
Relatively new among CMR data analyses are spatially explicit models which provide the innovation of modelling the probability of detecting an individual at any point in time as a joint function of the location of the trap or “detector” as well as the location of the individual (Efford, 2004; Borchers, 2010). The need for a spatial model incorporating trap location data arises from the simple fact that animals located close to traps will generally be more likely to be caught in those traps (Borchers, 2010). In addressing this issue, spatially explicit capture recapture (SECR) models overcome heterogeneity in capture probability caused by animal movement and the edge effects of a limited sampled area (Efford, 2004;
Borchers and Efford, 2008). Chapter 6 121 In this study, trap location or ‘detector’ data for our spatial capture-recapture analysis included XY coordinates (UTM) of mark-phase bait stations and every roam-phase timed recapture trial. While the two-phase sampling strategy is effective for collecting flight distance data, the mixed sampling strategy could confound parameter estimation. To avoid this we designated each sampling day of the mark-phase as a sampling occasion (n = 9 days) and condensed the roam- phase into a single occasion. Trap-specific individual capture histories were then constructed for each tagged wasp for the entire study of nine mark-phase sampling occasions and a terminal (10th) recapture occasion. Each wasp’s capture history is thus composed of a sequence of binary detected/not detected data points across
10 occasions in addition to corresponding spatial coordinates for each detection.
We allowed detection probability (g0) and the scale over which an individual’s probability of capture declines (σ) to vary on this terminal occasion (as a result of changed sampling regime) via a two-level time covariate between occasions one to nine and occasion ten.
We tested our data against four competing models, each constituting a unique hypothesis on the way that wasps might respond to a localised unrewarding sex pheromone stimulus. The null model assumes no behavioural response to pheromone baits throughout the study and applies uniform detection probabilities among individuals throughout occasions one to nine, allowing a different detection probability on the final occasion via the time covariate. A ‘simple learned response’ model allows a change in wasp behaviour after being captured at any pheromone bait once. This differs from the null model only in allowing a step change in detection probability for individuals following first capture. A ‘trap-specific learned response’ model incorporates a site-sensitive change of behaviour for individuals after first capture. It allows an individual’s detection probability to Chapter 6 122 change after first capture but only for that particular detector. In other words, individuals captured in a certain location have a changed detection probability as a response, but only for that location as in the case of wolverines tending to return to camera-trap baits once discovered (Royle et al., 2011). The final model was a ‘trap- specific transient response’ which was similar to the ‘trap-specific learned response’ but only considered the location of the individual in the last occasion it was seen. This model tests the hypothesis that wasp behavior changes after trapping at a specific site, but only for that site and not for the duration of the study. For all learned response models, detection probabilities can increase (‘trap- happy’) or decrease (‘trap-shy’) as a response to being trapped. Intuitively, if wasps are avoiding specific deceptive sources of sex pheromone after first attraction, the best fitting model should be either a permanent or transient trap- specific learned response with a decreased detection probability for a specific trap location following capture at that trap.
We ran the competing models described above in secr 2.3.1 (Efford, 2011b) (the package’s supporting documentation provides more detail on the models) under the default half-normal detection function. This fits a half-normal distribution to decline in probability of detection with distance from the home range centre according to scale of home-range (σ; modelled as constant) (Efford, 2004). We then compared the AICc among models (Burnham, 2002; Symonds and Moussalli,
2011) to determine the model best fitting the data.
At present, open population SECR models with mortality are not available and the
SECR analysis here assumes a demographically closed population model (no births, deaths, migration). Therefore we intentionally minimized the extent of violation of this assumption by carrying out the CMR study over a short time frame. Further, Chapter 6 123 the two-level time covariate has the effect of rolling any mortality or migration between mark and roam phases into the g0 and σ for roam phase. Possible violations of closure in this study will manifest as underestimated capture probabilities (Kendall, 1999). Over the short time scale of our study however, this bias does not affect the comparison of capture probabilities necessary for assessing the evidence for long-term avoidance—the key objective of this component of the study.
RESULTS
Wasp attraction curves for N. cryptoides showed the rapid decline in responses with time typical of site specific avoidance observed in other species (Peakall,
1990; Peakall and Beattie, 1996; Alcock, 2000) (figure 1). Results for four minute trials show a sharp decline in response over time for landings, copulations and total responses. However approaches did not appear to decline as rapidly. In six 60 minute trials a steep decline was observed over the first 10 minutes of bait presence after which a low level of response was maintained for the duration. A total of 105 wasps were marked during these two trials but we recorded only five recaptures, two of which occurred at the same location within the same trial.
We marked a total of 505 individual wasps during the mark phase of the CMR study. Over both phases of the study 123 (24%) individuals were recaptured in
189 recapture events. Individual recapture frequencies ranged from one recapture
(92 individuals), twice (31 individuals), three times (four individuals) to four times
(one individual). The mean time between first and last captures was 8.8 days with range 1-24 days. We resampled the exact same location within a day seven times Chapter 6 124 during the study and no individuals were recaptured at the same site within the same day.
Mark-recapture distances varied from 0-161 m and 0-150 m for the mark and roam phases respectively (figure 2). Mean displacement was 15.5 m ±3.4, n=81, median = 0 for mark phase recaptures and 27 m ±4.1, n=78, median = 14.8 for roam phase recaptures. The rate of movement between captures ranged between
0-149 m/day with a mean of 7.5 m/day +-1.6 m/day. In order to qualitatively show independence between observed displacements and sampling regime we superimposed the distribution of observed displacements over the distribution of all possible observed displacements. We computed the latter distribution from the pairwise distances between all sampling points throughout the study (figure 2).
The distribution of observed wasp flights was leptokurtic with strong positive skew which differed substantially from the roughly bimodal, less strongly skewed distribution of all possible recorded flight distances due to sample design.
Among our four competing behavioral models, the model best fitting the data, with an AIC weight of 0.99 (table 1), was unambiguously the trap-specific learned response. This model predicts a step-change in detection probability for an individual at a location if that individual had previously been detected at that location. In comparison with this model, none of the other three models attracted any support. Interestingly, the chance of encountering an individual at a specific trap approximately doubled if that individual had previously been caught at that trap, as shown by the estimated detection probabilities (table 2).
Chapter 6 125
DISCUSSION
We found that the site-specific patch avoidance behavior noted in studies of other sexually deceived thynnine wasps (Peakall, 1990; Handel and Peakall, 1993;
Peakall and Beattie, 1996; Alcock, 2000; Peakall and Schiestl, 2004) is present in the short-term behavioral response of N. cryptoides to synthetic pheromone baits.
The refractory period for patch avoidance typically lasted an entire day in that just two repeat visits were recorded for an individual wasp on the same day out of a grand total of 194 recapture events. This is concordant with a CMR study of the thynnine wasp Zaspilothynnus trilobatus which found no revisits to the same site within three hour observation periods (Peakall, 1990).
Unexpectedly, the results of the CMR study showed a large proportion of net-zero displacements among recaptures. Many of these occurred on the day after the individual was recorded at a location, although individuals could also be found within two metres of a previous capture for up to 12 days. This suggests that short term patch avoidance is confined to a refractory period of less than 24 hours. Thus, if underpinned by learning, any retention of spatial information might not exceed this period.
Our SECR analysis specifically investigated long term patch avoidance by fitting competing models for behavioral responses to being captured at pheromone baits.
We found unambiguously that the most appropriate model (among the four we tested) was the trap specific learned response which predicts that wasp capture probability increases for a specific location after being caught at that location
(table 1). The resulting estimates for detection probability, conditional on whether or not a wasp had been caught at a specific location, showed an approximate two- fold increase in probability of capture at a location if a wasp had been caught there Chapter 6 126 previously (table 2). In other words, N. cryptoides displays a trap-happy response to being caught. We can therefore rule out long term patch avoidance (trap-shy response) which would produce the opposite trend in estimated detection probabilities.
The findings of a trap-happy response were unexpected. However one likely reason for this finding is an interaction between wasp home-range size and the scale of our sampling strategy. On a given day a wasp may respond to, and then subsequently avoid, the source of a sex pheromone signal that is not accompanied by successful mating. Given the median movements of approximately 0m and 15m for the two CMR phases respectively, it is clear that a proportion of the population have relatively static home ranges which they patrolled daily. Sampling in the mark-phase limited data collection to nine points and the minimum distance between these points was 15 metres. Thus, if a N. cryptoides daily patrol frequently does not exceed 15 metres, recapture events would necessarily be limited to occurring at the site of previous capture.
Site specific patch avoidance has been interpreted as evidence for learning rather than sensory habituation (Peakall, 1990; Bower, 1996; Wong and Schiestl, 2002;
Gaskett, 2010) because another pulse of visitors can be initiated by moving the pheromone source just a few metres (Bower, 1996; Ayasse et al., 2000). Our observation that approaches did not decline with other behaviors is also consistent with learned avoidance of the dummy after first contact. However as mentioned above, if spatial learning underpins site specific patch avoidance then this information is not retained beyond 24 hours. Although short term patch avoidance is consistent with movement patterns expected under the operation of spatial learning, it is important to point out that there are other explanations that might Chapter 6 127 provide the same observed pattern. One explanation might be that short term patch avoidance is a fixed behavioral routine that might not involve learning at all.
Limited behavioral plasticity via fixed routines, triggered by specific stimuli could be sufficient for a wasp mating system to operate. Further, these behavioral routines could evolve or persist if courtship and mating signals remained consistent across generations (Dukas, 2008). From the perspective of a deceptive plant pollinated via dishonest signalling, exploitation of innate behavioral pathways might also be less easily circumvented via pollinator learning and this may promote evolutionary stability. Experiments that compare movement before and after being sexually deceived with a control group that are not exposed to the stimulus could provide the evidence necessary to distinguish learning from alternative explanations.
One potential influence on behavior and learning that we did not test, however, was the continuity of signal. Our pheromone signals were made intermittently throughout the day and discontinued at night—a scenario probably similar to calling females, but less like orchids where no diurnal differences in chiloglottone levels have been detected (R Peakall, submitted manuscript). If continuous signal could increase the length of site avoidance, which it conceivably might, then we would expect fewer revisits to orchids by the same individual that the results here suggest.
Short term avoidance of a sustained sex pheromone signal has also been shown to operate in response to females (Wong and Schiestl, 2002; Goh and Morse, 2010) and extracts of females (Robacker et al., 1976) (rather than orchid mimics or synthetic sex pheromones). This behavior is also known in a braconid (Goh and
Morse, 2010) and an ichneumonid (Robacker et al., 1976), wasps distantly related Chapter 6 128 to the thynnines in this and other related studies. Thus, it is evident that this behavior is not an evolved response to any costs of sexual deception. Interestingly, wasps for which this behavior is known all exhibit both scramble competition and a polygynous mating system leading to extreme male-bias in their operational sex ratio. The adaptive significance of patch avoidance behavior in these wasps remains unknown but these commonalities in mating system suggest that avoidance of patches recently hosting a sexual signal might be an inherent part of optimal mate search strategy in solitary wasps (Goh and Morse, 2010).
Importantly, irrespective of the underlying basis of this behavior, site-specific avoidance of deceptive flowers is hypothesized to promote outcrossing and prevent inbreeding in the plant by reducing pollen transfers within flowers
(autogamous selfing), between flowers on a plant (geitnogamous selfing), within clones or between closely-related individuals in a patch (Nilsson, 1992; Peakall and Beattie, 1996). Typical of the genus, C. trapeziformis is a self-compatible, solitary flowered, colony-forming terrestrial orchid. To result in an outcross, pollen must therefore traverse the clone size boundary which, for Chiloglottis, is estimated at approximately 5m on average (Peakall, unpublished). Given the average extent of movements in the range of 15-27 m observed here, typical pollen movement should exceed clone size. Average movements recorded here are in a similar range as those found for the thynnine pollinators of Drakaea glyptodon
(Peakall, 1990) (32 m) and Caladenia tentaculata (Peakall and Beattie, 1996) (17 m). Furthermore, occasional long range dispersal as seen in the long tail of the skewed distribution would promote long range pollen flow in excess of 100 m
(figure 2). Chapter 6 129 These results lead us to predict that N. cryptoides mate search movement, coupled with temporary site specific patch avoidance, will promote very high levels of outcrossing in the clonal self-compatible orchid it pollinates. Exploiting patch avoidance behavior of the thynnine wasp could even be considered a form of flower visitor behavioral optimization as seen in other pollination systems. For example, tobacco and cycad plants have been shown to moderate flower visitor behavior through synergistic attraction and repellent mechanisms (Terry et al.,
2007; Kessler et al., 2008). In this way, sexual attraction and an innate avoidance routine might work synergistically to enhance pollen export and avoid near neighbour pollinations. Additionally, through constrained capacity for learning, inflexibility in innate behavior could promote the maintenance of sexual deception.
Table 1: Comparison of four spatially explicit capture-recapture models for N. cryptoides
detection probability. The best fitting model for the data (highest AIC weight (AICwt), lowest
AIC) is a trap-specific learned response. The transient trap-specific learned response, null model
and simple learned trap response attract substantially less AIC support.
logLik AICc dAICc AICwt
Trap specific learned response -956.66 1923.45 0.00 0.99
Transient trap specific learned -961.11 1932.35 8.90 0.01 response
Null model -965.27 1938.62 15.17 0.00
Simple learned trap response -965.21 1940.55 17.10 0.00
Table 2: Estimated detection probability (standard error) under trap-specific learning model as
a function of capture history and phase of the study. The results indicate a substantial increase
in an individual’s detection probability at a specific trap if that individual had been trapped
there previously. The effect was consistent across phases of the study.
Not detected in Previous detection trap X in trap X
Phase 1 0.00409 (0.00066) 0.00943 (0.00151)
Phase 2 0.00047 (0.00007) 0.00108 (0.00017)
Figure 1: a) Mean number of Neozeleboria cryptoides responding to stationary synthetic
pheromone bait over one hour interval (n = 6). b) Mean number of N. cryptoides behavioral
responses to synthetic baits over 4 minute trials (n = 166). Behaviors were scored as
approaches if a wasp flew towards or lingered in proximity to the bait, landings if the wasp
wasp landed on the bait or the ground below the bait and copulations if the copulatory routine including abdominal probing was observed. Total is the sum of all the above observed behaviors
and bars represent one standard error from the mean.
Figure 2: Frequency histogram of displacements recorded from Neozeleboria cryptoides recaptures over both the mark phase (sampling restricted to
nine points) and roam phase (214 randomly placed recapture trials). The black line represents the proportional distribution of all possible pairwise
distances between sampling locations over the entire study. *: median displacement class Concluding Remarks 133
CONCLUDING REMARKS
Boyd River Fire Trail, running parallel to study site. Kanangra Boyd National Park, NSW. 2010.
Concluding Remarks 134
Future studies in the evolutionary biology of pollination, orchids and Chiloglottis.
The nature of scientific inquiry is such that the value of a work lies not only in the answers we gain, but also in the questions the research opens up. Here I highlight some promising new areas of investigation that future work has the potential to illuminate.
Chapter 4 highlighted the lack of studies addressing the male proportion of reproductive success. Although this aspect of plant mating has received increasing attention over the last 20 years (Karron et al. 2003; Bernasconi et al. 2004; Johnson et al. 2005; Trapnell and
Hamrick 2006), there remains much scope for an increasing application of paternity analysis to measure male components of fitness. Experimental pollination studies, targeted at strategic study systems where trait heritability is estimable, offer potential to tell us much about the selective forces imposed by pollinators on floral evolution.
One critical floral trait for which we lack estimates of heritability is floral scent
(Parachnowitsch et al. 2012). This is in part because making quantitative assessments of floral volatiles is difficult, especially when they are at trace levels such as in Chiloglottis.
What is more, environmental conditions undoubtedly strongly influence their variability observed in nature. Nonetheless, Parachnowitsch et al. (2012) have recently experimentally confirmed for the first time that floral scent can be under stronger selection than other floral traits—a condition we predict in Chiloglottis. Quantifying individual level variation in scent traits is critical to opening up experimental avenues such as those described above.
As pointed out in Chapter 4, there is still much progress to be made in the development of methods for the genetic inference of sibship in progeny arrays. Dividing sibships of multiple sires into full– and half–sib groups remains a surprisingly difficult analytical challenge (Berger-Wolf et al. 2007; Ashley et al. 2009; Sefc and Koblmuller 2009; Jones et al. 2010). Despite much work there is yet no benchmark method by which to achieve this Concluding Remarks 135 over the range of mating systems without the most demanding standards of data quantity and quality. No doubt the rapidly increasing availability of genomic data will be critical to solving this problem, although the quality of sequencing reads currently generated from these techniques will need improvement (see Appendix III)
While the focus of this work has been with pollination, not to be forgotten is the influence of seed dispersal on orchid gene flow, population structure and evolution. Tracking the dispersal of the orchids’ minute dust–like seeds is another of the great challenges in orchid research. Despite some recent progress (Trapnell et al. 2004; Jersáková and Malinová
2007; Hamrick and Trapnell 2011), orchid seed dispersal curves remain largely unknown, especially for terrestrial species. More data to solidify the link between orchid seed dispersal and metapopulation gene flow are required to estimate effective population sizes, explore adaptive versus neutral models of evolution (Tremblay et al. 2005) and ultimately understand the forces underlying the generation of one of the Earth’s most diverse plant families.
It has been 10 years since the discovery that Chiloglottis trapeziformis attracted
Neozeleboria cryptoides with compounds that were until that time unknown to science
(Schiestl et al. 2003). Since then we have learned much, and continue to learn about the evolution, ecology and genetics of these orchids. With experimental hybrid crosses
(Chapter 3) and advances in genomic techniques becoming increasingly available, the tools required to elucidate the underlying genetic mechanisms behind floral volatile synthesis in Chiloglottis are now available. Exciting discoveries in this area promise to reveal important mechanisms underlying speciation in this fascinating case study of floral evolution.
Pollination biology in the genomic age
There are few other structures in nature from which evolution has generated such wide diversity as the flower or inflorescence (Barrett 2003). The complex web of relationships Concluding Remarks 136 between this variation in floral traits, pollinator cognition and behaviour, plant reproduction, population genetics, speciation and evolution has inspired and continues to inspire generations of biologists. Pollination biologists today stand at the centre of a radiation of scientific disciplines, each one fast developing as a powerful and mature field of inquiry in its own right. This diversity of study includes topics such as the chemistry and physiology of plant metabolites, the genetic architecture underlying plant traits, neutral genetic approaches to studying population processes and the neuroethology of pollinating insects. These areas are all rich and accessible domains full of tools that are increasingly accessible to the pollination biologist. Underlying these is the basic pollination ecology of careful observation, experimental method and rigorous data collection in which we find the legacy of the birth of our field (Sprengel 1793b; Darwin
1876, 1877). Looking to the future, we can expect to see rapid advancement in genomics and genetic modification opening up yet more new tools that will create further exciting opportunities to explore the ongoing relationship between plants, pollinators and evolution.
137
REFERENCES
Ackerman, J.D. (1983) Specificity and mutual dependency of the orchid-euglossine bee interaction. Biological Journal of the Linnean Society 20, 301-314.
Ackerman, J.D., Melendez-Ackerman, E.J. and Salguero-Faria, J. (1997) Variation in pollinator abundance and selection on fragrance phenotypes in an epiphytic orchid.
American Journal of Botany 84, 1383-1390.
Alcock, J. (2000) Interactions between the sexually deceptive orchid Spiculaea ciliata and its wasp pollinator Thynnoturneria sp (Hymenoptera : Thynninae). Journal of Natural
History 34, 629-636.
Allendorf, F.W., Hohenlohe, P.A. and Luikart, G. (2010) Genomics and the future of conservation genetics. Nature Reviews Genetics 11, 697-709.
Alston, R.E., Turner, B.L. and Mabry, T.J. (1963) Perspectives in chemotaxonomy. Science
142, 545-552.
Anderson, E.C. and Thompson, E.A. (2002) A model-based method for identifying species hybrids using multilocus genetic data. Genetics 160, 1217-1229.
Arnaud-Haond, S. and Belkhir, K. (2007) GENCLONE: a computer program to analyse genotypic data, test for clonality and describe spatial clonal organization. Molecular
Ecology Notes 7, 15-17.
Ashley, M.V., Caballero, I. C., Chaovalitwongse, W., Dasgupta, B., Govindan, P., Sheikh, S.I. and Berger-Wolf, T.Y. (2009) KINALYZER, a computer program for reconstructing sibling groups. Molecular Ecology Resources 9, 1127-1131. 138 Ayasse, M. (2006) Floral scent and pollinator attraction in sexually-deceptive orchids.
Biology of Floral Scent (eds N. Dudareva and E. Pichersky), pp. 219-242. CRC Press, Boca
Raton, FL.
Ayasse, M., Schiestl, F.P., Paulus, H.F., Ibarra, F. and Francke, W. (2003) Pollinator attraction in a sexually deceptive orchid by means of unconventional chemicals.
Proceedings of the Royal Society of London B: Biological Sciences 270, 517-522.
Ayasse, M., Schiestl, F.P., Paulus, H.F., Lofstedt, C., Hansson, B., Ibarra, F. and Francke, W.
(2000) Evolution of reproductive strategies in the sexually deceptive orchid Ophrys sphegodes: How does flower-specific variation of odor signals influence reproductive success? Evolution 54, 1995-2006.
Azuma, H., Thien, L.B. and Kawano, S. (1999) Floral scents, leaf volatiles and thermogenic flowers in Magnoliaceae. Plant Species Biology 14, 121-127.
Azuma, H., Toyota, M. and Asakawa, Y. (2001) Intraspecific variation of floral scent chemistry in Magnolia kobus DC. (Magnoliaceae). Journal of Plant Research 114, 411-422.
Bancroft, J.S. and Smith, M.T. (2005) Dispersal and influences on movement for
Anoplophora glabripennis calculated from individual mark-recapture. Entomologia
Experimentalis et Applicata 116, 83-92.
Banks, S.C. and Peakall, R. (2012) Genetic spatial autocorrelation can readily detect sex- biased dispersal. Molecular Ecology 21, 2092-2105.
Barkman, T.J. (2001) Character coding of secondary chemical variation for use in phylogenetic analyses. Biochemical Systematics and Ecology 29, 1-20.
Barrett, S.C.H. (2003) Mating strategies in flowering plants: the outcrossing-selfing paradigm and beyond. Philosophical Transactions of the Royal Society of London. Series B:
Biological Sciences 358, 991-1004. 139 Batty, A.L., Dixon, K.W., Brundrett, M. and Sivasithamparam, K. (2001) Constraints to symbiotic germination of terrestrial orchid seed in a mediterranean bushland. New
Phytologist 152, 511-520.
Berger-Wolf, T.Y., Sheikh, S.I., DasGupta, B., Ashley, M.V., Caballero, I.C., Chaovalitwongse,
W. and Putrevu, S.L. (2007) Reconstructing sibling relationships in wild populations.
Bioinformatics 23, I49-I56.
Bernasconi, G. (2004) Seed paternity in flowering plants: an evolutionary perspective.
Perspectives in Plant Ecology Evolution and Systematics 6, 149-158.
Bernasconi, G., Ashman, T.L., Birkhead, T.R., Bishop, J.D.D., Grossniklaus, U., Kubli, E.,
Marshall, D.L., Schmid, B., Skogsmyr, I., Snook, R.R., Taylor, D., Till-Bottraud, I., Ward, P. I.,
Zeh, D.W. and Hellriegel, B. (2004) Evolutionary Ecology of the Prezygotic Stage. Science
303, 971-975.
Berry W.L, Stimmann M.W. and Wolf W.W. (1972) Marking of native phytophagous insects with rubidium: a proposed technique. Annals of the Entomological Society of America 65,
236–38
Bianchi, F., Schellhorn, N.A. and Van Der Werf, W. (2009) Predicting the time to colonization of the parasitoid Diadegma semiclausum: The importance of the shape of spatial dispersal kernels for biological control. Biological Control 50, 267-274.
Boiteau, G., Meloche F., Vincent C. and Leskey, T.C. (2009) Effectiveness of glues used for harmonic radar attachment and impact on survival and behavior of three insect pests.
Environmental Entomology 38, 168-175.
Borchers, D. (2010) A non-technical overview of spatially explicit capture–recapture models. Journal of Ornithology:1-10.
Borchers, D.L. and Efford, M.G. (2008) Spatially explicit maximum likelihood methods for capture–recapture studies. Biometrics 64, 377-385. 140 Bower, C.C. (1996) Demonstration of pollinator-mediated reproductive isolation in sexually deceptive species of Chiloglottis (Orchidaceae: Caladeniinae). Australian Journal of
Botany 44, 15-33.
Bower, C.C. (2006) Specific pollinators reveal a cryptic taxon in the bird orchid, Chiloglottis valida sensu lato (Orchidaceae) in south-eastern Australia. Australian Journal of Botany 54,
53-64.
Bower, C.C. and Brown, G.R. (2009) Pollinator specificity, cryptic species and geographical patterns in pollinator responses to sexually deceptive orchids in the genus Chiloglottis: the
Chiloglottis gunnii complex. Australian Journal of Botany 57, 37-55.
Bräutigam, A. and Gowik, U. (2010) What can next generation sequencing do for you? Next generation sequencing as a valuable tool in plant research. Plant Biology 12, 831-841.
Brodmann, J., Twele, R., Francke, W., Holzler, G., Zhang, Q.H. and Ayasse, M. (2008) Orchids mimic green-leaf volatiles to attract prey-hunting wasps for pollination. Current Biology
18, 740-744.
Brown, A. and Allard, R. (1970) Estimation of the mating system in open pollinated maize populations using isozyme polymorphisms. Genetics 66, 133-145.
Brown, G.R. (1998) Revision of the Neozeleboria cryptoides species group of thynnine wasps (Hymenoptera : Tiphiidae): Pollinators of native orchids. Australian Journal of
Entomology 37, 193-205.
Browning, S.R. and Browning, B.L. (2010) High-resolution detection of identity by descent in unrelated individuals. The American Journal of Human Genetics 86, 526-539.
Broyles, S.B. and Wyatt, R. (1990) Paternity analysis in a natural population of Asclepias exaltata: multiple paternity, functional gender, and the "pollen-donation hypothesis".
Evolution 44, 1454-1468. 141 Broyles, S.B. and Wyatt, R. (1991) Effective pollen dispersal in a natural population of
Asclepias exaltata: the influence of pollinator behavior, genetic similarity, and mating success. The American Naturalist 138, 1239-1249.
Brunet, J. and Sweet, H.R. (2006) Impact of insect pollinator group and floral display size on outcrossing rate. Evolution 60, 234-246.
Burgarella, C., Lorenzo, Z., Jabbour-Zahab, R., Lumaret, R., Guichoux, E., Petit, R.J., Soto, A. and Gil, L. (2009) Detection of hybrids in nature: application to oaks (Quercus suber and Q. ilex). Heredity 102, 442-452.
Burnham, K.P. and Anderson, D.R. (2002) Model selection and multimodel inference, 2nd edn, ed. New York: Springer.
Campbell, D. (1998) Multiple paternity in fruits of Ipomopsis aggregata (Polemoniaceae).
American Journal of Botany 85, 1022.
Cancino, A.D., Damon, A. (2007) Fragrance analysis of euglossine bee pollinated orchids from Soconusco, south-east Mexico. Plant Species Biology 22, 129-134.
Chapman, A.V., Kuhar, T.P., Schultz, P.B. and Brewster, C.C. (2009) Dispersal of
Trichogramma ostriniae (Hymenoptera: Trichogrammatidae) in potato fields.
Environmental Entomology 38, 677-685.
Charlesworth, D. and Charlesworth, B. (1987) Inbreeding depression and its evolutionary consequences. Annual Review of Ecology and Systematics 18, 237-268.
Chittka, L. and Thomson, J.D. (2001) Cognitive ecology of pollination. Animal behaviour and floral evolution. , Cambridge University Press, Cambridge, UK.
Claesson, M.J., Wang, Q.O., O'Sullivan, O., Greene-Diniz, R., Cole, J.R., Ross, R.P. and O'Toole,
P.W. (2010) Comparison of two next-generation sequencing technologies for resolving 142 highly complex microbiota composition using tandem variable 16S rRNA gene regions.
Nucleic Acids Research 38, e200.
Clements, M.A. and Ellyard, R.K. (1979) The symbiotic germination of Australian terrestrial orchids. . American Orchid Society Bulletin 48, 810 – 816.
Collett, T.S. (1992) Landmark learning and guidance in insects. Philosophical Transactions:
Biological Sciences 337, 295-303.
Corbett, A. and Rosenheim, J.A. (1996) Quantifying movement of a minute parasitoid,
Anagrus epos (Hymenoptera: Mymaridae), using fluorescent dust marking and recapture.
Biological Control 6, 35-44.
Cortis, P., Vereecken, N.J., Schiestl, F.P., Lumaga, M.R.B., Scrugli, A. and Cozzolino, S. (2009)
Pollinator convergence and the nature of species' boundaries in sympatric Sardinian
Ophrys (Orchidaceae). Annals of Botany 104, 497-506.
Coyne, J.A. and Orr, H.A. (2004) Speciation, Sinauer, Sunderland, MA.
Cozzolino, S. and Scopece, G. (2008) Specificity in pollination and consequences for postmating reproductive isolation in deceptive Mediterranean orchids. Philosophical
Transactions of the Royal Society B: Biological Sciences 363, 3037-3046.
Cozzolino, S. and Widmer, A. (2005) Orchid diversity: an evolutionary consequence of deception? Trends in Ecology and Evolution 20, 487-494.
Cozzolino, S. and Widmer, A. (2005) The evolutionary basis of reproductive isolation in
Mediterranean orchids. Taxon 54, 977-985.
Cozzolino, S. and Widmer, A. (2006) Response to Otero and Flanagan: Orchid diversity - beyond deception. Trends in Ecology and Evolution 21, 65-66.
Cozzolino, S., D'Emerico, S. and Widmer, A. (2004) Evidence for reproductive isolate selection in Mediterranean orchids: karyotype differences compensate for the lack of 143 pollinator specificity. Proceedings of the Royal Society of London Series B-Biological Sciences
271, S259-S262.
Cunningham, J.P., Moore, C.J., Zalucki, M.P. and West, S.A. (2004) Learning, odour preference and flower foraging in moths. Journal of Experimental Biology 207, 87-94.
Dafni, A. (1984) Mimicry and deception in pollination. Annual Review of Ecology and
Systematics 15, 259-278.
Dakin, E.E. and Avise, J.C. (2004) Microsatellite null alleles in parentage analysis. Heredity
93, 504-509.
Dalca, A.V. and Brudno, M. (2010) Genome variation discovery with high-throughput sequencing data. Briefings in Bioinformatics 11, 3-14.
Darwin, C.R. (1876) The effects of cross and self-fertilisation in the vegetable kingdom, John
Murray, London, UK.
Darwin, C.R. (1877) The various contrivances by which orchids are fertilised by insects, John
Murray, London, UK.
Davey, J.W. and Blaxter, M.L. (2011) RADSeq: next-generation population genetics.
Briefings in Functional Genomics 9, 416-423.
Davey, J.W., Hohenlohe, P.A., Etter, P.D., Boone, J.Q., Catchen, J.M. and Blaxter, M.L. (2011)
Genome-wide genetic marker discovery and genotyping using next-generation sequencing. Nature Reviews Genetics 12, 499-510.
Day-Williams, A.G. and Zeggini, E. (2011) The effect of next-generation sequencing technology on complex trait research. European Journal of Clinical Investigation 41, 561-
567.
Dobzhansky, T. (1937) Genetics and the origin of species, Columbia Univ. Press, New York. 144 Dodson, C.H., Dressler, R.L., Hills, H.G., Adams, R.M. and Williams, N.H. (1969) Biologically active compounds in orchid fragrances. Science 164, 1243-1249.
Dole, J. and Ritland, K. (1993) Inbreeding depression in two Mimulus taxa measured by multigenerational changes in the inbreeding coefficient. Evolution 47, 361-373.
Doorduin, L., Gravendeel, B., Lammers, Y., Ariyurek, Y., Chin-A-Woeng, T. and Vrieling, K.
(2011) The complete chloroplast genome of 17 individuals of pest species Jacobaea vulgaris: SNPs, microsatellites and barcoding markers for population and phylogenetic studies. DNA Research 18, 93-105.
Dotterl, S., Fussel, U., Jurgens, A. and Aas, G. (2005) 1,4-Dimethoxybenzene, a floral scent compound in willows that attracts an oligolectic bee. Journal of Chemical Ecology 31, 2993-
2998.
Dressler, R.L. (1968) Pollination by euglossine bees. Evolution 22, 202-210.
Dudai, Y., Jan, Y.N., Byers, D., Quinn, W.G. and Benzer, S. (1976) dunce, a mutant of
Drosophila deficient in learning. Proceedings of the National Academy of Sciences 73, 1684-
1688.
Dudareva, N., Negre, F., Nagegowda, D.A. and Orlova, I. (2006) Plant volatiles: Recent advances and future perspectives. Critical Reviews in Plant Sciences 25, 417-440.
Dudash, M.R. and Ritland, K. (1991) Multiple paternity and self-fertilization in relation to floral age in Mimulus guttatus (Scrophulariaceae). American Journal of Botany 78, 1746-
1753.
Dufaÿ, M., Hossaert-Mckey, M. and Anstett, M.-C. (2004) Temporal and sexual variation of leaf-produced pollinator-attracting odours in the dwarf palm. Oecologia 139, 392-398.
Dukas, R. (2006) Learning in the context of sexual behaviour in insects. Animal Biology 56,
125-141. 145 Dukas, R. (2008) Evolutionary biology of insect learning. Annual Review of Entomology 53,
145-160.
Dukas, R. and Real, L.A. (1991) Learning foraging tasks by bees: a comparison between social and solitary species. Animal Behaviour 42, 269-276.
Ebert, D. and Peakall, R. (2009) Chloroplast simple sequence repeats (cpSSRs): technical resources and recommendations for expanding cpSSR discovery and applications to a wide array of plant species. Molecular Ecology Resources 9, 673-690.
Ebert, D., Hayes, C. and Peakall, R. (2009) Chloroplast simple sequence repeat markers for evolutionary studies in the sexually deceptive orchid genus Chiloglottis. Molecular Ecology
Resources 9, 784-789.
Efford, M. (2004) Density estimation in live-trapping studies. Oikos 106, 598-610.
Efford, M.G. (2011) secr - spatially explicit capture-recapture in R version 2.3.1.
Department of Zoology, University of Otago, Dunedin.
Ekblom, R. and Galindo, J. (2011) Applications of next generation sequencing in molecular ecology of non-model organisms. Heredity 107, 1-15.
Ellstrand, N.C. (1984) Multiple paternity within the fruits of the wild radish, Raphanus sativus. The American Naturalist 123, 819-828.
Ellstrand, N.C. (2003) Current knowledge of gene flow in plants: implications for transgene flow. Philosophical Transactions of the Royal Society B-Biological Sciences 358,
1163-1170.
Eltz, T. and Lunau, K. (2005) Antennal response to fragrance compounds in male orchid bees. Chemoecology 15, 135-138.
Eltz, T., Ayasse, M. and Lunau, M. (2006) Species-specific antennal responses to tibial fragrances by male orchid bees. Journal of Chemical Ecology 32, 71-79. 146 Eltz, T., Roubik, D.W. and Lunau, M. (2005) Experience-dependent choices ensure species- specific fragrance accumulation in male orchid bees. Behavioral Ecology and Sociobiology
59, 149-156.
Epperson, B.K. and Clegg, M.T. (1987) First-pollination primacy and pollen selection in the morning glory, Ipomoea purpurea. Heredity 58, 5-14.
Ervik, F., Tollsten, L. and Knudsen, J.T. (1999) Floral scent chemistry and pollination ecology in phytelephantoid palms (Arecaceae). Plant Systematics and Evolution 217, 279-
297.
Excoffier, L., Smouse, P.E. and Quattro, J.M. (1992) Analysis of molecular variance inferred from metric distances among DNA haplotypes - application to human mitochondrial-DNA restriction data. Genetics 131, 479-491.
Fenster, C.B., Armbruster, W.S., Wilson, P., Dudash, M.R. and Thomson, J.D. (2004)
Pollination syndromes and floral specialization. Annual Review of Ecology, Evolution, and
Systematics 35, 375-403.
Ferrera, V., Arroyo, J., Castro, S. and Navarro, L. (2012) Unusual heterostyly: style dimorphism and self-incompatibility are not tightly associated in Lithodora and Glandora
(Boraginaceae). Annals of Botany, In Press, doi:10.1093/aob/mcr222
Flanagan, N.S., Ebert, D., Porter, C., Rossetto, M. and Peakall, R. (2006) Microsatellite markers for evolutionary studies in the sexually deceptive orchid genus Chiloglottis.
Molecular Ecology Notes 6, 123-126.
Franke, S., Ibarra, F., Schulz, C.M., Twele, R., Poldy, J., Barrow, R.A., Peakall, R., Schiestl, F.P. and Francke, W. (2009) The discovery of 2,5-dialkylcyclohexan-1,3-diones as a new class of natural products. Proceedings of the National Academy of Sciences 106, 8877-8882. 147 Fussel, U., Dotterl, S., Jurgens, A. and Aas, G. (2007) Inter- and intraspecific variation in floral scent in the genus Salix and its implication for pollination. Journal of Chemical
Ecology 33, 749-765.
Galen, C. (1999) Why do flowers vary? Bioscience 49, 631-640.
Gardner, M.G., Fitch, A.J., Bertozzi, T. and Lowe, A.J. (2011) Rise of the machines – recommendations for ecologists when using next generation sequencing for microsatellite development. Molecular Ecology Resources 11, 1093-1101.
Gaskett, A.C. (2011) Orchid pollination by sexual deception: pollinator perspectives.
Biological Reviews 86, 33-75.
Gaskett, A.C., Winnick, C.G. and Herberstein, M.E. (2008) Orchid sexual deceit provokes ejaculation. American Naturalist 171, 206-212.
Gibernau, M., Hossaert-Mckey, M., Frey, J. and Kjellberg, F. (1998) Are olfactory signals sufficient to attract fig pollinators? Ecoscience 5, 306-311.
Gigord, L.D.B., Macnair, M.R. and Smithson, A. (2001) Negative frequency-dependent selection maintains a dramatic flower color polymorphism in the rewardless orchid
Dactylorhiza sambucina (L.) Proceedings of the National Academy of Sciences of the United
States of America 98, 6253-6255.
Gögler, J., Stökl, J., Sramkova, A., Twele, R., Francke, W., Cozzolino, S., Cortis, P., Scrugli, A. and Ayasse, M. (2009) Menage a trios; two endemic species of deceptive orchids and one pollinator species. Evolution (in press) DOI: 10.1111/j.1558-5646.2009.00712.x
Goh, M.Z. and Morse, D.H. (2010) Male mate search for female emergence sites by a parasitic wasp. Animal Behaviour 80, 391-398. 148 Goyret, J., Pfaff, M., Raguso, R. and Kelber, A. (2008) Why do Manduca sexta feed from white flowers? Innate and learnt colour preferences in a hawkmoth. Naturwissenschaften
95, 569-576.
Grant, V. (1949) Pollination systems as isoltaing mechanisms in angiosperms. Evolution 3,
82-97.
Grant, V. (1994) Modes and origins of mechanical and ethological isolation in
Angiosperms. Proceedings of the National Academy of Sciences 91, 3-10.
Griffiths, K.E., Trueman, J.W.H., Brown, G.R. and Peakall, R. (2011) Molecular genetic analysis and ecological evidence reveals multiple cryptic species among thynnine wasp pollinators of sexually deceptive orchids. Molecular Phylogenetics and Evolution 59, 195-
205.
Grison-Pige, L., Bessiere, J.M. and Hossaert-Mckey, M. (2002a) Specific attraction of fig- pollinating wasps: Role of volatile compounds released by tropical figs. Journal of Chemical
Ecology 28, 283-295.
Grison-Pige, L., Hossaert-Mckey, M., Greeff, J.M. and Bessiere, J.M. (2002b) Fig volatile compounds - a first comparative study. Phytochemistry 61, 61-71.
Grison, L., Edwards, A.A. and Hossaert-Mckey, M. (1999) Interspecies variation in floral fragrances emitted by tropical Ficus species. Phytochemistry 52, 1293-1299.
Grueber, C.E., Nakagawa, S., Laws, R.J. and Jamieson, I.G. (2011) Multimodel inference in ecology and evolution: challenges and solutions. Journal of Evolutionary Biology 24, 699-
711.
Hagler, J.R. and Jackson, C.G. (2001) Methods for marking insects: Current techniques and future prospects. Annual Review of Entomology 46, 511-543. 149 Hajibabaei, M., Shokralla, S., Zhou, X., Singer, G.A.C. and Baird, D.J. (2011) Environmental barcoding: a next-generation sequencing approach for biomonitoring applications using river benthos. PLoS ONE 6, e17497.
Hamrick, J.L. and Trapnell, D.W. (2011) Using population genetic analyses to understand seed dispersal patterns. Acta Oecologica 37, 641-649.
Handel, S.N. and Peakall, R. (1993) Thynnine wasps discriminate among heights when seeking mates: tests with a sexually deceptive orchid. Oecologia 95, 241-245.
Harder, L.D. and Johnson, S.D. (2008) Function and evolution of aggregated pollen in angiosperms. International Journal of Plant Sciences 169, 59-78.
Hawkins, R.D., Hon, G.C. and Ren, B. (2010) Next-generation genomics: an integrative approach. Nature Reviews Genetics, 11, 476-486.
Hills, H.G., Williams, N.H. and Dodson, C.H. (1972) Floral fragrances and isolating mechanisms in the genus Catasetum (Orchidaceae). Biotropica 4, 61-76.
Hodges, S.A. and Derieg, N.J. (2009) Adaptive radiations: From field to genomic studies.
Proceedings of the National Academy of Sciences 106, 9947-9954.
Holmquist, K.G., Mitchell, R.J. and Karron, J.D. (2012) Influence of pollinator grooming on pollen-mediated gene dispersal in Mimulus ringens (Phrymaceae). Plant Species Biology
27, 77-85.
Hudson, M.E. (2008) Sequencing breakthroughs for genomic ecology and evolutionary biology. Molecular Ecology Resources 8, 3-17.
Hughes, M., Moller, M., Edwards, T.J., Bellstedt, D.U. and De Villiers, M. (2007) The impact of pollination syndrome and habitat on gene flow: a comparative study of two
Streptocarpus (Gesneriaceae) species. American Journal of Botany 94, 1688-1695. 150 Janzen, D.H. (1971) Euglossine bees as long-distance pollinators or tropical plants. Science
171, 203-205.
Jennions, M.D. and Petrie, M. (2000) Why do females mate multiply? A review of the genetic benefits. Biological Reviews 75, 21-64.
Jersáková, J. and Malinová, T. (2007) Spatial aspects of seed dispersal and seedling recruitment in orchids. New Phytologist 176, 237-241.
Jersakova, J., Johnson, S.D. and Kindlmann, P. (2006) Mechanisms and evolution of deceptive pollination in orchids. Biological Reviews 81, 219-235.
Johnson, J.B. and Omland, K.S. (2004) Model selection in ecology and evolution. Trends in
Ecology and Evolution 19, 101-108.
Johnson, S.D. (2006) Pollinator-driven speciation in plants. In Ecology and Evolution of
Flowers (Ed, Harder, L. D., Barrett, S.C.H.) Oxford University Press, New York.
Johnson, S.D. and Edwards, T.J. (2000) The structure and function of orchid pollinaria.
Plant Systematics and Evolution 222, 243-269.
Johnson, S.D. and Nilsson, L.A. (1999) Pollen carryover, geitonogamy, and the evolution of deceptive pollination systems in orchids. Ecology 80, 2607-2619.
Johnson, S.D. and Steiner, K.E. (2000) Generalization versus specialization in plant pollination systems. Trends in Ecology and Evolution 15, 140-143.
Johnson, S.D., Neal, P.R. and Harder, L.D. (2005) Pollen fates and the limits on male reproductive success in an orchid population. Biological Journal of the Linnean Society 86,
175-190.
Jones, A.G., Small, C.M., Paczolt, K.A. and Ratterman, N.L. (2010) A practical guide to methods of parentage analysis. Molecular Ecology Resources 10, 6-30. 151 Jones, O.R. and Wang, J. (2010) COLONY: a program for parentage and sibship inference from multilocus genotype data. Molecular Ecology Resources 10, 551-555.
Jørgensen, T., Kjaer, K.H., Haile, J., Rasmussen, M., Boessenkool, S., Andersen, K., Coissac, E.,
Taberlet, P., Brochmann, C., Orlando, L., Gilbert, M.T.P. and Willerslev, E. (2012) Islands in the ice: detecting past vegetation on Greenlandic nunataks using historical records and sedimentary ancient DNA Meta-barcoding. Molecular Ecology 28, 1980-1988.
Karron, J.D. and Marshall, D.L. (1990) Fitness consequences of multiple paternity in wild radish, Raphanus sativus. Evolution 44, 260-268.
Karron, J.D., Ivey, C.T., Mitchell, R.J., Whitehead, M.R., Peakall, R. and Case, A.L. (2012) New perspectives on the evolution of plant mating systems. Annals of Botany 109, 493-503.
Karron, J.D., Mitchell, R.J. and Bell, J.M. (2006) Multiple pollinator visits to Mimulus ringens
(Phrymaceae) flowers increase mate number and seed set within fruits. American Journal of Botany 93, 1306-1312.
Karron, J.D., Mitchell, R.J., Holmquist, K.G., Bell, J.M. and Funk, B. (2003) The influence of floral display size on selfing rates in Mimulus ringens. Heredity 92, 242-248.
Kay, K.M. and Sargent, R.D. (2009) The Role of Animal Pollination in Plant Speciation:
Integrating Ecology, Geography, and Genetics. Annual Review of Ecology, Evolution, and
Systematics 40, 637-656.
Kendall, W.L. (1999) Robustness of closed capture-recapture methods to violations of the closure assumption. Ecology 80, 2517-2525.
Kessler, D., Gase, K. and Baldwin, I.T. (2008) Field experiments with transformed plants reveal the sense of floral scents. Science 321, 1200-1202.
Kimsey, L.S. (1980) The behaviour of male orchid bees (Apidae, Hymenoptera, Insecta) and the question of leks. Animal Behaviour 28, 996-1004. 152 Knudsen, J.T. (1999) Floral scent differentiation among coflowering, sympatric species of
Geonoma (Arecaceae). Plant Species Biology 14, 137-142.
Knudsen, J.T. (2002) Variation in floral scent composition within and between populations of Geonoma macrostachys (Arecaceae) in the western Amazon. American Journal of Botany
89, 1772-1778.
Knudsen, J.T. and Tollsten, L. (1993) Trends in floral scent chemistry in pollination syndromes: floral scent composition in moth-pollinated taxa. Botanical Journal of the
Linnean Society 113, 263-284.
Knudsen, J.T., Eriksson, R., Gershenzon, J. and Stahl, B. (2006) Diversity and distribution of floral scent. Botanical Review 72, 1-120.
Larsson, M. (2005) Higher pollinator effectiveness by specialist than generalist flower- visitors or unspecialized Knautia arvensis (Dipsacaceae). Oecologia 146, 394-403.
Lerner, H.R.L. and Fleischer, R.C. (2010) Prospects for the use of next-generation sequencing methods in ornithology. The Auk 127, 4-15.
Levin, R.A., Mcdade, L.A. and Raguso, R.A. (2003) The systematic utility of floral and vegetative fragrance in two genera of Nyctaginaceae. Systematic Biology 52, 334-351.
Levin, R.A., Raguso, R.A. and Mcdade, L.A. (2001) Fragrance chemistry and pollinator affinities in Nyctaginaceae. Phytochemistry 58, 429-440.
Lewis, H. (1962) Catastrophic selection as a factor in speciation. Evolution 16, 257-271.
Lexer, C. and Widmer, A. (2008) The genic view of plant speciation: recent progress and emerging questions. Philosophical Transactions of the Royal Society B-Biological Sciences
363, 3023-3036.
Liu L, Wolf R, Ernst R, Heisenberg M, 1999. Context generalization in Drosophila visual learning requires the mushroom bodies. Nature 400:753-756. 153 Lowry, D.B., Modliszewski, J.L., Wright, K.M., Wu, C.A. and Willis, J.H. (2008) The strength and genetic basis of reproductive isolating barriers in flowering plants. Philosophical
Transactions of the Royal Society B-Biological Sciences 363, 3009-3021.
Machado, C.A., Robbins, N., Gilbert, M.T.P. and Herre, E.A. (2005) Critical review of host specificity and its coevolutionary implications in the fig/fig-wasp mutualism. Proceedings of the National Academy of Sciences of the United States of America 102, 6558-6565.
Mahroof, R.M., Edde, P.A., Robertson, B., Pucketfe, J.A. and Phillips, T.W. (2010) Dispersal of
Rhyzopertha dominica (Coleoptera: Bostrichidae) in different habitats. Environmental
Entomology 39, 930-938.
Mant, J., Bower, C.C., Weston, P.H. and Peakall, R. (2005a) Phylogeography of pollinator- specific sexually deceptive Chiloglottis taxa (Orchidaceae): evidence for sympatric divergence? Molecular Ecology 14, 3067-3076.
Mant, J., Peakall, R. and Schiestl, F.P. (2005b) Does selection on floral odor promote differentiation among populations and species of the sexually deceptive orchid genus
Ophrys? Evolution 59, 1449-1463.
Mant, J., Peakall, R. and Weston, P.H. (2005c) Specific pollinator attraction and the diversification of sexually deceptive Chiloglottis (Orchidaceae). Plant Systematics and
Evolution 253, 185-200.
Mant, J.G., Schiestl, F.P., Peakall, R. and Weston, P.H. (2002) A phylogenetic study of pollinator conservatism among sexually deceptive orchids. Evolution 56, 888-898.
Marshall, D.L. and Ellstrand, N.C. (1986) Sexual selection in Raphanus sativus: experimental data on nonrandom fertilization, maternal choice, and consequences of multiple paternity. The American Naturalist 127, 446-461. 154 Martin, N.H. and Willis, J.H. (2007) Ecological divergence associated with mating system causes nearly complete reproductive isolation between sympatric Mimulus species.
Evolution 61, 68-82.
Mascanzoni, D. and Wallin, H. (1986) The harmonic radar: a new method of tracing insects in the field. Ecological Entomology 11, 387-390.
Mayr, E. (1942) Systematics and the origin of species, Columbia Univ. Press, New York.
Meirmans, P.G. (2006) Using the AMOVA framework to estimate a standardized genetic differentiation measure. Evolution 60, 2399-2402.
Menz, M.H.M., Phillips, R.D., Winfree, R., Kremen, C., Aizen, M.A., Johnson, S.D. and Dixon,
K.W. (2010) Reconnecting plants and pollinators: challenges in the restoration of pollination mutualisms. Trends in Plant Science 16, 4-12.
Menzel, R. (1999) Memory dynamics in the honeybee. Journal of Comparative Physiology A:
Neuroethology, Sensory, Neural, and Behavioral Physiology 185, 323-340.
Metzker, M.L. (2010) Sequencing technologies - the next generation. Nature Reviews
Genetics 11, 31-46.
Micheneau, C., Johnson, S.D. and Fay, M.F. (2009) Orchid pollination: from Darwin to the present day. Botanical Journal of the Linnean Society 161, 1-19.
Minder, A.M. and Widmer, A. (2008) A population genomic analysis of species boundaries: neutral processes, adaptive divergence and introgression between two hybridizing plant species. Molecular Ecology 17, 1552-1563.
Minder, A.M., Rothenbuehler, C. and Widmer, A. (2007) Genetic structure of hybrid zones between Silene latifolia and Silene dioica (Caryophyllaceae): evidence for introgressive hybridization. Molecular Ecology 16, 2504-2516. 155 Mitchell, R.J., Irwin, R.E., Flanagan, R.J. and Karron, J.D. (2009) Ecology and evolution of plant-pollinator interactions. Annals of Botany 103, 1355-1363.
Mitchell, R.J., Karron, J.D., Holmquist, K.G. and Bell, J.M. (2005) Patterns of multiple paternity in fruits of Mimulus ringens (Phrymaceae). American Journal of Botany 92, 885-
890.
Moerkens, R., Leirs, H., Peusens, G. and Gobin, B. (2010) Dispersal of single- and double- brood populations of the European earwig, Forficula auricularia: a mark-recapture experiment. Entomologia Experimentalis et Applicata 137, 19-27.
Montalvo, A.M. (1992) Relative success of self and outcross pollen comparing mixed- and single- donor pollinations in Aquilegia caerulea. Evolution 46, 1181-1198.
Moreau, M., Arrufat, P., Latil, G. and Jeanson, R. (2011) Use of radio-tagging to map spatial organization and social interactions in insects. Journal of Experimental Biology 214, 17-21.
Moyle, L.C., Olson, M.S. and Tiffin, P. (2004) Patterns of reproductive isolation in three
Angiosperm genera. Evolution 58, 1195-1208.
Nason, J.D., Herre, E.A. and Hamrick, J.L. (1998) The breeding structure of a tropical keystone plant resource. Nature 391, 685-687.
Nilsson, L.A. (1980) The pollination ecology of Dactylorhiza sambucina (Orchidaceae).
Botaniska Notiser 133, 367 - 385.
Nilsson, L.A. (1992) Orchid pollination biology. Trends in Ecology and Evolution 7, 255-
259.
O’Neal, M.E., Landis, D. A., Rothwell, E., Kempel, L. and Reinhard, D. (2004) Tracking insects with harmonic radar: a case study. American Entomologist 50, 7. 156 Okamoto, T., Kawakita, A. and Kato, M. (2007) Interspecific variation of floral scent composition in Glochidion and its association with host-specific pollinating seed parasite
(Epicephala). Journal of Chemical Ecology 33, 1065-1081.
Ollerton, J., Alarcon, R., Waser, N.M., Price, M.V., Watts, S., Cranmer, L., Hingston, A., Peter,
C.I. and Rotenberry, J. (2009) A global test of the pollination syndrome hypothesis. Annals of Botany 103, 1471-1480.
Ollerton, J., Armbruster, W.S. and Vazquez, D.P. (2006) The ecology and evolution of specialized and generalized pollination. Plant-pollinator interactions: from specialization to generalization. (eds N.M. Waser and J. Ollerton), pp.19-22 University of Chicago Press,
Chicago.
Olson, D.M., Rains, G.C., Meiners, T., Takasu, K., Tertuliano, M., Tumlinson, J.H., Wäckers,
F.L. and Lewis, W.J. (2003) Parasitic wasps learn and report diverse chemicals with unique conditionable behaviors. Chemical Senses 28, 545-549.
Omura, H., Honda, K. and Hayashi, N. (2000) Floral scent of Osmanthus fragrans discourages foraging behavior of cabbage butterfly, Pieris rapae. Journal of Chemical
Ecology 26, 655-666.
Otero, J.T. and Flanagan, N.S. (2006) Orchid diversity -- beyond deception. Trends in
Ecology and Evolution 21, 64-65.
Ouborg, N.J., Pertoldi, C., Loeschcke, V., Bijlsma, R. and Hedrick, P.W. (2010) Conservation genetics in transition to conservation genomics. Trends in Genetics 26, 177-187.
Pahl, M., Zhu, H., Tautz, J. and Zhang, S. (2011) Large scale homing in honeybees. PLoS ONE
6(5): e19669. doi:10.1371/journal.pone.0019669
Parachnowitsch, A.L., Raguso, R.A. and Kessler, A. (2012) Phenotypic selection to increase floral scent emission, but not flower size or colour in bee-pollinated Penstemon digitalis.
New Phytologist 195, 667-675. 157 Paulus, H.F. (2006) Deceived males – pollination biology of the Mediterranean orchid genus Ophrys (Orchidaceae). Journal Europaischer Orchideen 38, 121-177.
Paulus, H.F. and Gack, C. (1990) Pollinators as prepollinating isolation factors – evolution and speciation in Ophrys (Orchidaceae). Israel Journal of Botany 39, 43-79.
Peakall, R. (1989) A new technique for monitoring pollen flow in orchids. Oecologia 79,
361-365.
Peakall, R. (1990) Responses of male Zaspilothynnus trilobatus (Turner) wasps to females and the sexually deceptive orchid it pollinates. Functional Ecology 4, 159-167.
Peakall, R. (2007) Speciation in the Orchidaceae: confronting the challenges. Molecular
Ecology 16, 2834-2837.
Peakall, R. and Beattie, A.J. (1991) The genetic consequences of worker ant pollination in a self-compatible, clonal orchid. Evolution 45, 1837-1848.
Peakall, R. and Beattie, A.J. (1996) Ecological and genetic consequences of pollination by sexual deception in the orchid Caladenia tentactulata. Evolution 50, 2207-2220.
Peakall, R. and Schiestl, F.P. (2004) A mark-recapture study of male Colletes cunicularius bees: implications for pollination by sexual deception. Behavioral Ecology and Sociobiology
56, 579-584.
Peakall, R. and Smouse, P.E. (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes 6, 288-295.
Peakall, R. and Smouse, P.E. (2012) GenAlEx 6.5: Genetic analysis in Excel. Population genetics software for teaching and research - an update. Bioinformatics In press.
Peakall, R., Bower, C.C., Logan, A.E. and Nicol, H.I. (1997) Confirmation of the hybrid origin of Chiloglottis x pescottiana (Orchidaceae : Diurideae). 1. Genetic and morphometric evidence. Australian Journal of Botany 45, 839-855. 158 Peakall, R., Ebert, D., Poldy, J., Barrow, R.A., Francke, W., Bower, C.C. and Schiestl, F.P.
(2010) Pollinator specificity, floral odour chemistry and the phylogeny of Australian sexually deceptive Chiloglottis orchids: implications for pollinator-driven speciation. New
Phytologist, 188, 437-450.
Peakall, R., Jones, L., Bower, C.C. and Mackey, B.G. (2002) Bioclimatic assessment of the geographic and climatic limits to hybridisation in a sexually deceptive orchid system.
Australian Journal of Botany 50, 21-30.
Peakall, R., Smouse, P.E. and Huff, D.R. (1995) Evolutionary implications of allozyme and
RAPD variation in diploid populations of dioecious buffalograss Buchloe dactyloides.
Molecular Ecology 4, 135-148.
Piechulla, B. and Pott, M.B. (2003) Plant scents - mediators of inter- and intraorganismic communication. Planta 217, 687-689.
Piper, R.W. (2003) A novel technique for the individual marking of smaller insects.
Entomologia Experimentalis et Applicata 106, 155-157.
Proctor, M., Yeo, P. and Lack, A. (1996) A Natural History of Pollination. Harper Collins
Publishers, London.
Quinn, W.G., Harris, W.A. and Benzer, S. (1974) Conditioned behavior in Drosophila melanogaster. Proceedings of the National Academy of Sciences 71, 708-712.
Raguso, R.A. (2001) Floral scent, olfaction, and scent driven foraging behaviour. Cognitive ecology of pollination; Animal behaviour and floral evolution. (eds L. Chittka and J.D.
Thomson). Cambridge University Press, Cambridge.
Raguso, R.A. (2004) Flowers as sensory billboards: progress towards an integrated understanding of floral advertisement. Current Opinion in Plant Biology 7, 434-440. 159 Raguso, R.A. (2008a) Plant science - The "invisible hand" of floral chemistry. Science 321,
1163-1164.
Raguso, R.A. (2008b) Start making scents: the challenge of integrating chemistry into pollination ecology. Entomologia Experimentalis Et Applicata 128, 196-207.
Raguso, R.A. (2008c) Wake up and smell the roses: the ecology and evolution of floral scent. Annual Review of Ecology, Evolution, and Systematics 39, 549-569.
Raguso, R.A. and Pichersky, E. (1995) Floral volatiles from Clarkia breweri and C. concinna
(Onagraceae) – recent evolution of floral scent and moth pollination. Plant Systematics and
Evolution 194, 55-67.
Raguso, R.A., Levin, R.A., Foose, S.E., Holmberg, M.W. and Mcdade, L.A. (2003) Fragrance chemistry, nocturnal rhythms and pollination "syndromes" in Nicotiana. Phytochemistry
63, 265-284.
Raguso, R.A., Schlumpberger, B.O., Kaczorowski, R.L. and Holtsford, T.P. (2006)
Phylogenetic fragrance patterns in Nicotiana sections Alatae and Suaveolentes.
Phytochemistry 67, 1931-1942.
Ramsay, R.R., Sivasithamparam, K. and Dixon, K.W. (1986) Patterns of infection and endophytes associated with Western Australian orchids. Lindleyana 1, 203 – 214.
Ramsey, J., Bradshaw, H.D. and Schemske, D.W. (2003) Components of reproductive isolation between the monkeyflowers Mimulus lewisii and M. cardinalis (Phrymaceae).
Evolution 57, 1520-1534.
Richards, A.J. (1986) Plant Breeding Systems. Chapman and Hall, London
Ridsdill Smith, T.J. (1970) The biology of Hemithynnus hyalinatus (Hymenoptera:
Tephiidae), a parasite on scarabaeid larvae. Journal of the Australian Entomological Society
9, 183-195. 160 Rieseberg, L. and Willis, J. (2007) Plant speciation. Science 317, 910-914.
Riffell, J.A., Alarcon, R., Abrell, L., Davidowitz, G., Bronstein, J.L. and Hildebrand, J.G. (2008)
Behavioral consequences of innate preferences and olfactory learning in hawkmoth- flower interactions. Proceedings of the National Academy of Sciences 105, 3404-3409.
Riley, J.R., Smith, A.D., Reynolds, D.R., Edwards, A.S., Osborne, J.L., Williams, I.H., Carreck,
N.L. and Poppy, G.M. (1996) Tracking bees with harmonic radar. Nature 379, 29-30.
Ritland, K. (1989) Correlated matings in the partial selfer Mimulus guttatus. Evolution 43,
848-859.
Ritland, K. (1996) A marker-based method for inferences about quantitative inheritance in natural populations. Evolution 50, 1062-1073.
Ritland, K. (2002) Extensions of models for the estimation of mating systems using n independent loci. Heredity 88, 221-228.
Ritland, K. (2011) Evolutionary potential in the wild: more than meets the eye. Molecular
Ecology 20, 3494-3495.
Ritland, K. and Jain, S. (1981) A model for the estimation of outcrossing rate and gene frequencies using n independent loci. Heredity 47, 35-52.
Robacker, D.C., Weaver, K.M. and Hendry, L.B. (1976) Sexual communication and associative learning in the parasitic wasp Itoplectis conquisitor (Say). Journal of Chemical
Ecology 2, 39-48.
Robinson, E., Richardson, T., Sendova-Franks, A., Feinerman, O. and Franks, N. (2009)
Radio tagging reveals the roles of corpulence, experience and social information in ant decision making. Behavioral Ecology and Sociobiology 63, 627-636.
Roche, S.A., Carter, R.J., Peakall, R., Smith, L.M., Whitehead, M.R. and Linde, C.C. (2010) A narrow group of monophyletic Tulasnella (Tulasnellaceae) symbiont lineages are 161 associated with multiple species of Chiloglottis (Orchidaceae): Implications for orchid diversity. American Journal of Botany 97, 1313-1327.
Royle, J.A., Magoun, A.J., Gardner, B., Valkenburg, P. and Lowell, R.E. (2011) Density estimation in a wolverine population using spatial capture–recapture models. The Journal of Wildlife Management 75, 604-611.
Rudd, N.T. and McEvoy, P.B. (1996) Local dispersal by the cinnabar moth Tyria jacobaeae.
Ecological Applications 6, 285-297.
Schaefer, H.M. and Ruxton, G.D. (2009) Deception in plants: mimicry or perceptual exploitation? Trends in Ecology and Evolution 24, 676-685.
Schiestl, F.P. (2004) Floral evolution and pollinator mate choice in a sexually deceptive orchid. Journal of Evolutionary Biology 17, 67–75.
Schiestl, F.P. (2005) On the success of a swindle: pollination by deception in orchids.
Naturwissenschaften 92, 255-264.
Schiestl, F.P. and Ayasse, M. (2002) Do changes in floral odor cause speciation in sexually deceptive orchids? Plant Systematics and Evolution 234, 111-119.
Schiestl, F.P. and Marion-Poll, F. (2002) Detection of physiologically active flower volatiles using gas chromatography coupled with electroantennography. Molecular Methods of Plant
Analysis Volume 21, Analysis of Taste and Aroma (eds J.F. Jackson, H.F. Linskens and R.
Inman), pp. 173-198. Springer, Berlin.
Schiestl, F.P. and Peakall, R. (2005) Two orchids attract different pollinators with the same floral odour compound: ecological and evolutionary implications. Functional Ecology 19,
674-680.
Schiestl, F.P. and Schluter, P.M. (2009) Floral isolation, specialized pollination, and pollinator behavior in orchids. Annual Review of Entomology 54, 425-446. 162 Schiestl, F.P., Ayasse, M., Paulus, H.F., Lofstedt, C., Hansson, B.S., Ibarra, F. and Francke, W.
(1999) Orchid pollination by sexual swindle. Nature 399, 421-422.
Schiestl, F.P., Peakall, R., Mant, J.G., Ibarra, F., Schulz, C., Franke, S. and Francke, W. (2003)
The chemistry of sexual deception in an orchid-wasp pollination system. Science 302, 437-
438.
Schiestl, F.P., Roubik, R. (2003) Odor compound detection in male euglossine bees. Journal of Chemical Ecology 29, 253-257.
Schlumpberger, B.O. and Raguso, R.A. (2008) Geographic variation in floral scent of
Echinopsis ancistrophora (Cactaceae); evidence for constraints on hawkmoth attraction.
Oikos 117, 801-814.
Scopece, G., Cozzolino, S., Johnson, S.D. and Schiestl, F.P. (2010) Pollination efficiency and the evolution of specialized deceptive pollination systems. American Naturalist 175, 98-
105.
Scopece, G., Musacchio, A., Widmer, A. and Cozzolino, S. (2007) Patterns of reproductive isolation in Mediterranean deceptive orchids. . Evolution 61, 2623-2642.
Seber, G.A.F. (1970) The effects of trap response on tag recapture estimates. Biometrics 26,
13-22.
Sefc, K.M. and Koblmuller, S. (2009) Assessing parent numbers from offspring genotypes: the importance of marker polymorphism. Journal of Heredity 100, 197-205.
Sheehan, M.J. and Tibbetts, E.A. (2011) Specialized face learning is associated with individual recognition in paper wasps. Science 334, 1272-1275.
Shimizu, K.K., Kudoh, H. and Kobayashi, M.J. (2011) Plant sexual reproduction during climate change: gene function in natura studied by ecological and evolutionary systems biology. Annals of Botany 108, 777-787. 163 Slatyer, R.A., Mautz, B.S., Backwell, P.R.Y. and Jennions, M.D. (2011) Estimating genetic benefits of polyandry from experimental studies: a meta-analysis. Biological Reviews 87, 1-
33.
Smithson, A. and MacNair, M.R. (1997) Negative frequency-dependent selection by pollinators on artificial flowers without rewards. Evolution 51, 715-723.
Snow, A.A., Spira, T.P., Simpson, R. and Klips, R.A. (1996) The ecology of geitonogamous pollination. . In Floral biology: studies on floral evolution in animal-pollinated plants.(Eds,
Lloyd, D. G. and Barrett, S. C. H.) Chapman and Hall, New York, pp. 191–216.
Soliva, M. and Widmer, A. (2003) Gene flow across species boundaries in sympatric, sexually deceptive Ophrys (Orchidaceae) species. Evolution 57, 2252-2261.
Sprengel, C.K. (1793) Das entdeckte Geheimniss der Natur im Bau und in der Befruchtung der Blumen, Vieweg, Berlin.
St Pierre, M.J., Hendrix, S.D. and Lewis, C.K. (2005) Dispersal ability and host-plant characteristics influence spatial population structure of monophagous beetles. Ecological
Entomology 30, 105-115.
Stebbins, G.L. (1970) Adaptive radiation of reproductive characteristics in angiosperms, I:
Pollination mechanisms. Annual Review of Ecology and Systematics 1, 307-326.
Steiner, K.E., Whitehead, V.B. and Johnson, S.D. (1994) Floral and pollinator divergence in
2 sexually deceptive South African orchids. American Journal of Botany 81, 185-194.
Stensmyr, M.C., Urru, I., Collu, I., Celander, M., Hansson, B.S. and Angioy, A.M. (2002)
Rotting smell of dead-horse arum florets - These blooms chemically fool flies into pollinating them. Nature 420, 625-626.
Stephens, P.A., Buskirk, S.W. and del Rio, C.M. (2007) Inference in ecology and evolution.
Trends in Ecology and Evolution 22, 192-197. 164 Stökl, J., Paulus, H., Dafni, A., Schulz, C., Francke, W. and Ayasse, M. (2005) Pollinator attracting odour signals in sexually deceptive orchids of the Ophrys fusca group. Plant
Systematics and Evolution 254, 105-120.
Stökl, J., Schlueter, P.M., Stuessy, T.F., Paulus, H.F., Assum, G. and Ayasse, M. (2008) Scent variation and hybridization cause the displacement of a sexually deceptive orchid species.
American Journal of Botany 95, 472-481.
Stökl, J., Schluter, P.M., Stuessy, T.F., Paulus, H.F., Fraberger, R., Erdmann, D., Schulz, C.,
Francke, W., Assum, G. and Ayasse, M. (2009) Speciation in sexually deceptive orchids: pollinator-driven selection maintains discrete odour phenotypes in hybridizing species.
Biological Journal of the Linnean Society 98, 439-451.
Sumner, S., Lucas, E., Barker, J. and Isaac, N. (2007) Radio-tagging technology reveals extreme nest-drifting behavior in a eusocial insect. Current Biology 17, 140-145.
Svensson, G.P., Hickman, M.O., Bartram, S., Boland, W., Pellmyr, O. and Raguso, R.A. (2005)
Chemistry and geographic variation of floral scent in Yucca filamentosa (Agavaceae).
American Journal of Botany 92, 1624-1631.
Swarts, N.D., Sinclair, E.A., Francis, A. and Dixon, K.W. (2010) Ecological specialization in mycorrhizal symbiosis leads to rarity in an endangered orchid. Molecular Ecology 19,
3226-3242.
Sword, G.A., Lorch, P.D. and Gwynne, D.T. (2005) Insect behaviour: Migratory bands give crickets protection. Nature 433, 703-703.
Symonds, M.R.E. and Moussalli, A. (2011) A brief guide to model selection, multimodel inference and model averaging in behavioural ecology using Akaike's information criterion. Behavioral Ecology and Sociobiology 65, 13-21.
Takeda, K. (1961) Classical conditioned response in the honey bee. Journal of Insect
Physiology 6, 168-179. 165 Teixeira, S. and Bernasconi, G. (2007) High prevalence of multiple paternity within fruits in natural populations of Silene latifolia, as revealed by microsatellite DNA analysis.
Molecular Ecology 16, 4370-4379.
Terry, I., Walter, G.H., Moore, C., Roemer, R. and Hull, C. (2007) Odor-mediated push-pull pollination in cycads. Science 318, 70.
Trapnell, D.W. and Hamrick, J.L. (2005) Mating patterns and gene flow in the neotropical epiphytic orchid, Laelia rubescens. Molecular Ecology 14, 75-84.
Trapnell, D.W. and Hamrick, J.L. (2006) Floral display and mating patterns within populations of the neotropical epiphytic orchid, Laelia rubescens (Orchidaceae). American
Journal of Botany 93, 1010-1018.
Trapnell, D.W., Hamrick, J.L. and Nason, J.D. (2004) Three-dimensional fine-scale genetic structure of the neotropical epiphytic orchid, Laelia rubescens. Molecular Ecology 13,
1111-1118.
Tremblay, R.L., Ackerman, J.D., Zimmerman, J.K. and Calvo, R.N. (2005) Variation in sexual reproduction in orchids and its evolutionary consequences: a spasmodic journey to diversification. Biological Journal of the Linnean Society 84, 1-54.
Vähä, J.P. and Primmer, C.R. (2006) Efficiency of model-based Bayesian methods for detecting hybrid individuals under different hybridization scenarios and with different numbers of loci. Molecular Ecology 15, 63-72.
Vallejo-Marín, M., Dorken, M.E. and Barrett, S.C.H. (2010) The ecological and evolutionary consequences of clonality for plant mating. Annual Review of Ecology, Evolution, and
Systematics 41, 193-213. van der Niet, T. and Johnson, S.D. (2012) Phylogenetic evidence for pollinator-driven diversification of angiosperms. Trends in Ecology andamp; Evolution 27, 353-361. 166 Varshney, R.K., Nayak, S.N., May, G.D. and Jackson, S.A. (2009) Next-generation sequencing technologies and their implications for crop genetics and breeding. Trends in
Biotechnology 27, 522-530.
Vet, L.E.M., Lewis, W. J. and Cardé, R. T. (1995) Parasitoid foraging and learning. In:
Chemical Ecology of Insects (Cardé RT, Bell, W.J., ed). New York: Chapman and Hall; 65-
101. von Frisch, K. (1953) The dancing bees; an account of the life and senses of the honey bee.
New York: Harcourt, Brace and World.
Waelti, M.O., Muhlemann, J.K., Widmer, A. and Schiestl, F.P. (2008) Floral odour and reproductive isolation in two species of Silene. Journal of Evolutionary Biology 21, 111-121.
Wang, J. and Santure, A.W. (2009) Parentage and sibship inference from multilocus genotype data under polygamy. Genetics 181, 1579-1594.
Ware, A.B., Kaye, P.T., Compton, S.G. and Vannoort, S. (1993) Fig volatiles – their role in attracting pollinators and maintaining pollinator specificity. Plant Systematics and
Evolution 186, 147-156.
Waser, N.M. (1998) Pollination, angiosperm speciation, and the nature of species boundaries. Oikos 82, 198-201.
Waser, N.M. (2001) Pollinator behaviour and plant speciation: looking beyond the
"ethological isolation" paradigm. Cognitive ecology of pollination: animal behaviour and floral evolution (eds L. Chittka and J.D. Thomson), pp. 318-315. Cambridge University
Press, Cambridge.
Waser, N.M. and Price, M.V. (1984) Experimental studies of pollen carryover: effects of floral variability in Ipomopsis aggregata. Oecologia 62, 262-268. 167 Waser, N.M., Chittka, L., Price, M.V., Williams, N.M. and Ollerton, J. (1996) Generalization in pollination systems, and why it matters. Ecology 77, 1043-1060.
Waterman, R.J., Bidartondo, M.I., Stofberg, J., Combs, J.K., Gebauer, G., Savolainen, V.,
Barraclough, T.G. and Anton, P. (2011) The effects of above- and belowground mutualisms on orchid speciation and coexistence. The American Naturalist 177, E54-E68.
Whelan, R.J., Ayre, D.J. and Beynon, F.M. (2009) The birds and the bees: pollinator behaviour and variation in the mating system of the rare shrub Grevillea macleayana.
Annals of Botany 103, 1395-1401.
White, E.G. (1970) A self-checking coding technique for mark-recapture studies. Bulletin of
Entomological Research 60, 303-307.
White, G. C. and Burnham, K.P. (1999) Program MARK: Survival estimation from populations of marked animals. Bird Study 46 Supplement, 120–138.
Whitehead, M.R. and Peakall, R. (2009) Integrating floral scent, pollination ecology and population genetics. Functional Ecology 23, 863-874.
Whitehead, M.R. and Peakall, R. (2011) Microdot technology for individual marking of small arthropods. Agricultural and Forest Entomology 14, 171-175.
Whitten, W.M. and Williams, N.H. (1992) Floral fragrances of Stanhopea (Orchidaceae)
Lindleyana 7, 130-153.
Widmer, A., Lexer, C. and Cozzolino, S. (2008) Evolution of reproductive isolation in plants.
Heredity 102, 31-38.
Wikelski, M., Moxley, J., Eaton-Mordas, A., Lopez-Uribe, M.M., Holland, R., Moskowitz, D.,
Roubik, D.W. and Kays, R. (2010) Large-range movements of neotropical orchid bees observed via radio telemetry. PLoS ONE 5, e10738. 168 Williams, N.H. and Dodson, C.H. (1972) Selective attraction of male euglossine bees to orchid floral fragrances and its importance in long distance pollen flow. Evolution 26, 84-
95.
Wineriter, S.A. and Walker, T.J. (1984) Insect marking techniques: Durability of materials.
Entomological News 95, 117-123.
Wong, B.B.M. and Schiestl, F.P. (2002) How an orchid harms its pollinator. Proceedings:
Biological Sciences 269, 1529-1532.
Wong, B.B.M., Salzmann, C. and Schiestl, F.P. (2004) Pollinator attractiveness increases with distance from flowering orchids. Proceedings: Biological Sciences 271, S212-S214.
Xu, S., Schlüter, P.M. and Schiestl, F.P. (2012) Pollinator-driven speciation in sexually deceptive orchids. International Journal of Ecology 2012, doi:10.1155/2012/285081
Xu, S.Q., Schluter, P.M., Scopece, G., Breitkopf, H., Gross, K., Cozzolino, S. and Schiestl, F.P.
(2011) Floral isolation is the main reproductive barrier among closely related sexually deceptive orchids. Evolution 65, 2606-2620.
Young, H.J. (2002) Diurnal and nocturnal pollination of Silene alba (Caryophyllaceae).
American Journal of Botany 89, 433-440.
169 APPENDICES
APPENDIX I
Supplementary Table 1: Studies for which volatiles have been implicated in pollinator specificity in morphologically similar, co-flowering, sympatric intra-generic taxa. 170
Plant genus Species studied Sympatric Coflowering Morphology Specific pollinator Interfertile Ref G. macrostachys Mart. Flies, beetles, bees (inc. Knudsen 1999 Plant Geonoma var. macrostachys Yes Yes Very similar Euglossine) Not reported Sp. Biol. G. macrostachys Mart. Pop. 'small' Yes Yes Very similar Not reported Not reported G. stricta (Poit.) Kunth var. stricta Yes Yes Very similar Not reported Not reported G. stricta (Poit.) Kunth var. piscicauda (Dammer) Hend. Yes Yes Very similar Not reported Not reported Geonoma sp. Yes Yes Very similar Not reported Not reported G. maxima (Poit.) Kunth Yes Yes Very similar Not reported Not reported G. triglochin Burret Yes Yes Very similar Not reported Not reported G. brongniartii Mart. Yes Yes Very similar Not reported Not reported Similar, petal Sphingid and noctuid Waelti et al 2008 J. Silene S. latifolia Yes Yes colour difference moths Hybrids may occur Evol. Biol. Similar, petal Bumblebees, syrphids, S. dioica Yes Yes colour difference butterflies, muscid flies. Hybrids may occur Interfertile, no recorded natural Ebert et al. 2009 Chiloglottis C. valida Yes Yes Cryptic Neozeleboria monticola hybrid Mol. Ecol. Resources Interfertile, no recorded natural C. aff. jeansii Yes Yes Cryptic Neozeleboria impatiens hybrid Not Hybrids may occur, Terry et al 2004 Macrozamia M. machinii reported Yes Similar Tranes spp. weevils none reported Plant Syst. Evol. With M. Hybrids may occur, M. lucida macleayi Yes Similar Cycadothrips chadwicki none reported With M. Hybrids may occur, M. macleayi lucida Yes Similar Cycadothrips chadwicki none reported Schiestl and Ayasse 2002 Plant Syst. Ophrys O. fusca Yes Yes Similar Andrena nigroaenea Interfertile Evol. O. bilunulata Yes Yes Similar Andrena flavipes Interfertile 171
Stokl et al 2008 Am Ophrys O. iricolor Yes Yes Similar Andrena morio Interfertile J. Bot. O. lupercalis Yes Yes Similar Andrena nigroaenea Interfertile With F. Grison-Pige et al Ficus F. fulva condensa Yes* N/A Blastophaga compacta Not reported 2002 J. Chem. Ecol. With F. F. condensa fulva Yes* N/A Ceratosolen constrictus Not reported Not F. deltoidea reported Yes* N/A Wiebesia sp. Not reported Not F. microcarpa reported Yes* N/A Eupristina verticillata Not reported Not Okamoto et al 2007 Glochidion G. acuminatum Yes reported Very similar Epicephala sp. Not reported J. Chem. Ecol. Yes (not G. Not G. rubrum obovatum) reported Very similar Epicephala sp. Not reported Yes (not G. Not G. obovatum rubrum) reported Very similar Epicephala sp. Not reported Not G. lanceolatum Yes reported Very similar Epicephala sp. Not reported Not G. zeylancium Yes reported Very similar Epicephala sp. Not reported Dobson et al 1997 Not Biochem. Syst. And Narcissus N. cuatracasasii Yes reported Similar Unknown Not reported Ecol. Not N. assoanus Yes reported Similar Moths Not reported 172
Supplementary Table2: Results of search for NCBI accessions, Web of Sciences hits for terms, Mol. Ecol. Notes/Resources publications and
published chloroplast (cp) and nuclear (n) genome sequences for the top ten most researched families for volatiles (above the line) and
four economically important families (below the line)
Genomes NCBI ISI WoS search terms Mol Ecol Notes/Resources Family Nucleotide EST "Molecular marker" "Primer*" "Microsatellite" cp n Orchidaceae 15863 7978 51 101 36 12 1 0 Araceae 2038 4308 32 70 5 3 1 0 Arecaceae 4778 20224 51 97 37 9 0 0 Magnoliaceae 1153 24132 13 13 18 2 1 0 Rosaceae 13383 413480 652 1061 336 38 0 0 Cactaceae 1615 0 13 18 9 5 0 0 Rutaceae 2004 231067 190 283 43 3 1 0 Solanaceae 38603 785211 1681 2236 150 4 6 0 Caryophyllaceae 5650 609 40 91 36 5 0 0 Nyctaginaceae 403 8 2 8 0 0 0 0 Brassicaceae 252035 2644499 1060 1720 168 14 12 1 Poaceae 451173 5047765 3899 2950 1141 17 9 2 Convolvulaceae 3600 83157 53 112 12 1 5 0 Fabaceae 47811 1115221 1445 1618 334 30 4 0 173
APPENDIX II
Laboratory protocols
PCR Protocol
PCR amplification was performed in a 25 μl reaction mix consisting of 2.5 μl of 10×PCR buffer, 1.5 μl of 25mM MgCl2, 1.7 μl dNTPs (2 μM), 1.7 μl untailed primer (2 μM), 0.4 μl
M13 primer (2 μM), 0.8 μl 1xBSA (10μg/mL), 0.8 μl flourescent labelled M13 label
(2 μM), 13.5 μl double distilled water, 0.1 μl Taq (5 U/μl) and 2 μl of template DNA. PCR amplification was performed using a Corbett PC-960C cooled thermal cycler and negative controls were run for all amplifications. Amplification of the microsatellite fragments was conducted with an activation step at 94 °C for 3 min followed by 2 cycles of denaturation at 94 °C for 30 s, annealing at 66 °C for 40 s (stepping down 3 °C every 2 successive cycles), and extension at 72 °C for 70 s, with a further 32 cycles of denaturation at 94 °C for 30 s, annealing at 54 °C for 40 s, and extension at 72 °C for
70 s, followed by a final extension step at 72 °C for 35 mins.
174
Protocorm DNA Extraction Protocol
1. Freeze dry protocorms overnight in screwcap tubes.
2. Grind in QIAGEN tissue mill with glass beads, 2 × 20 seconds, at speed 5.
3. Add 100 μl of QIAGEN API and 1 μl RNAse. Mix and incubate at 65°C for 15 minutes with
intermittent inversion.
4. Spin down, move 100 μl of each sample into wells of a 96-well plate, add 50 μl AP2. Mix.
Chill on ice for 5 minutes.
5. Spin plate for 12 minutes at 4000 rpm. Transfer 100 μl of the supernatant to a new plate
with 100 μl of 100% Isopropyl alcohol.
6. Invert 50 times then spin for 12 minutes at 4000rpm.
7. Decant supernatant, blot. Add 150 μl of fresh 70% ETOH. Invert 10 times then spin for
12 minutes at 4000rpm.
8. Decant, blot. Spin inverted at 200rpm for 10 seconds.
9. Dry plate at room temperature. Resuspend overnight at 65°C 30 μl TE.
175
APPENDIX III
The following text was co-authored with Rod Peakall. A revised version of this was published in the paper:
Karron JD, Ivey CT, Mitchell RJ, Whitehead MR, Peakall R, Case AL, (2012). New perspectives on the evolution of plant mating systems. Annals of Botany 109:493-503.
Plant mating system studies in the era of next-generation sequencing
A generation ago research on plant mating systems blossomed with the widespread availability of codominant genetic markers (Brown and Allard, 1970), facilitating studies of outcrossing rates and patterns of paternity. Today plant evolutionary biologists are adapting to another revolution in molecular technology, next generation sequencing (NGS), which again opens up fresh areas of inquiry. With NGS technologies, researchers are increasingly able to genotype hundreds of markers within populations of non-model organisms, and probe the genetic architecture of population and individual traits critical to answering important questions about plant mating systems. The benefits of NGS are already being seen across a wide range of fields including genomics (Hawkins et al., 2010), genetics (Metzker, 2010), human disease
(Day-Williams and Zeggini, 2011), crop genetics and breeding (Varshney et al., 2009). There is much excitement about the potential for NGS to benefit fields closely allied to plant mating systems including genomic ecology and evolutionary biology (Hudson, 2008), molecular ecology
(Ekblom and Galindo, 2011) and conservation genetics (Allendorf et al., 2010; Ouborg et al.,
2010). What opportunities might NGS offer plant mating system studies?
176
Rapid cost-effective marker development for mating system and paternity analysis
One far-reaching benefit of NGS that is already being realised across multiple fields is rapid, cost-effective genome–wide discovery of variable genetic markers such as microsatellites (Dalca and Brudno, 2010; Davey et al., 2011). Gardner et al. (2011) offer a helpful procedural guide for ecologists planning to use NGS for marker development. As yet, limitations on read-length have prevented NGS from being suitable for mass genotyping of individual level variation in these markers. However as the technology develops it should be possible to routinely sequence fragments large enough to traverse microsatellite regions and associated flanking sequence in order to quickly and cheaply genotype multiple loci across multiple individuals in any species.
Thus NGS will enable for the first time cost-effective paternity analysis in large natural populations. This capability could even enable determination of pollen donor representation in the pollen load on a single stigma. Comparison to realized paternity for entire seed families could lead to exciting insights into pollen competition and reproductive incompatibilities.
Contributions of NGS for phylogenetic analysis
The mapping of mating system traits onto a robust phylogeny can reveal important evolutionary insights. For example, Ferrera et al. (2012) discovered phylogenetic evidence for the independent evolution of heterostyly and incompatibility in the Boraginaceae, contradicting the expectation of strong linkage between these traits. Unfortunately, a lack of phylogenetically-informative markers often prevents the construction of robust phylogenies, the first step in this process. This is particularly true in recently-evolved groups that may offer the most potential for exploring dynamic evolutionary processes (Hodges and Derieg, 2009;
Peakall et al. 2010).
NGS can assist phylogenetic analysis in at least two ways. First, rapid, low cost phylogenetically informative genetic markers can be found via transcriptome sequencing that targets coding genes (Ekblom and Galindo, 2011). Genome-wide phylogenetic analysis can be achieved using 177
Restriction-site Associated DNA sequencing (RAD-Seq) that offers the potential for detecting
1000’s of polymorphisms simultaneously across multiple individuals in a single NGS run (Davey and Blaxter, 2011; Davey et al., 2011).
Candidate gene discovery
Beyond marker development per se, NGS offers exciting new opportunities for fast tracking candidate gene discovery (Bräutigam and Gowik, 2010). While the genetic bases of self-incompatibility (SI) mechanisms and the loss of SI are increasingly well understood in model systems such as Arabidopsis (see Shimizu et al., 2011 for review), NGS may open the door for rapid progress in candidate SI gene discovery in non-model systems. Similarly, we anticipate NGS will facilitate progress on the genetic basis of sex determination in plants, which is central to our understanding of the evolution of separate sexes. Additionally, identifying the loci underlying variation in mating system traits is becoming increasingly accessible outside model systems through developments in NGS. Restriction-site associated DNA sequencing techniques can generate a wealth of data across the entire genome for a sample of individuals in a population. When individual sequences within the population sample are grouped by some measured phenotype, loci linked to the measured trait can be identified (Davey and Blaxter,
2011). Theory for marker-based inferences of the heritability of quantitative traits is already well-developed (Ritland, 1996) and in conjunction with the developing measures of relatedness for population data gleaned from massive batteries of markers (Ritland, 2011; Browning and
Browning, 2010), the underlying genetic architecture of floral traits is now within reach.
Species ID by DNA barcoding for investigating pollinator interactions and networks
By pre-amplifying known informative chloroplast markers, such as those used for DNA barcoding , it should be possible to elucidate composition of pollen mixtures carried by 178 pollinating insects. Massive parallelization makes NGS particularly well-suited to detecting low- copy variants in mixed DNA samples (Lerner and Fleischer, 2010) and the technology is already being successfully exploited to characterise entire communities such as benthic macroinvertebrates (Hajibabaei et al., 2011), gut microbiota (Claesson et al., 2010) and historical vegetation in ancient sediments (Jørgensen et al., 2011).
Wilson et al. (2010) recently demonstrated the genetic identification of plants from an assemblage of 16 species by genetic analysis of the pollen grains carried by Hawaiian bees. We suggest that by combining this approach with NGS at several plant DNA barcoding loci there is enormous potential to disentangle more complicated pollination networks. In concert with quantitative data from relative read numbers this approach could provide insights into the proportional representation of species within pollen loads and identify heterogeneity of pollen placement on different parts of a pollinator’s body. Comparison of pollen loads with the background plant community could tell us much about pollinator specialization/generalization.
A quantitative approach could also provide insights into male fitness within species. Lastly, questions on stigma-blockage and interspecific pollen flow could be probed by sequencing species composition of stigmatic pollen loads.
Rapid chloroplast genome sequencing
With next generation sequencing capability the entire chloroplast genome is becoming an increasingly accessible tool for use in both population and community studies. For example,
Doorduin et al. (2011) reported the complete chloroplast genome sequence of 17 individuals within a single pest plant species, Jacobaea vulgaris. Although only recently recognised for its potential to reveal genetic variation within species (see Ebert and Peakall, 2009 for review), when combined with predominantly uniparental transmission, genetic variation at the chloroplast is ideally suited for teasing apart pollen and seed-mediated gene flow. Studies that 179 make use of individual levels of cpDNA variation will also shed new light on patterns of both maternity and paternity in dispersed seeds and juvenile plants in natural populations.
There can be little doubt that NGS offers powerful and exciting new opportunities for advancing the field of plant mating systems, however studies that employ these techniques will still require not only careful, elegant experimental designs but also carefully chosen insightful questions.