Pollination Ecology and Demography of a Deceptive Orchid

Home , Anacamptis, Cephalanthera rubra, Cypripedioideae, Dactylorhiza sambucina, Myrmecia urens, Neotinea

POLLINATION ECOLOGY AND DEMOGRAPHY OF A DECEPTIVE ORCHID

Ryan P. Walsh

A Dissertation

Submitted to the Graduate College of Bowling Green State University in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

August 2013

Committee:

Helen Michaels, Advisor

Timothy J. Murnen Graduate Faculty Representative

Karen Root

Moira van Staaden

Randy Mitchell

ii ABSTRACT

Helen Michaels, Advisor

This dissertation is focused on three main questions addressing the reproductive and demographic effects of pollen limitation, seed predation and deceit pollination in the food deceptive orchid Cypripedium candidum. We conducted two hand pollination field experiments to quantify pollen limitation and inbreeding. Both studies showed strong pollen limitation, with supplemental hand pollinations increasing fruit set in 2009 by 41% and 2011 by 30-35%. Taller plants in the study were more likely to be pollinated, while all other size variables did not influence pollination or fruit set. The 2011 study demonstrated a reduction in seed mass in selfed capsules by 63%. We found high levels of fruit predation in 2009 with 73% of the fruit experiencing pre-dispersal seed predation resulting in an 89% reduction in seed mass. Of the size variables analyzed, shorter plants were more likely to be attacked by weevils. In a nectar addition study we manipulated plants to provide a nectar reward, dyed their pollinia for tracking and compared their reproduction against control plants with no reward. Nectar reward, which had no effect on fruit production, however did result in a nearly threefold increase in selfing.

Approximately 26% of non-rewarding C. candidum pollination events result in selfing, while the addition of nectar increased selfing to 78%. Selfed seed capsules had a decreased seed mass as demonstrated in the previous experiment. Finally, we conducted a four year demographic study and produced matrix models that estimated the population growth rate at λ = 1.01 under an average of 22% pre-dispersal seed predation. Elasticity values of the models indicated the stasis and growth of one-flowered individuals to be the most important factors to the population growth iii rate. A model simulating the effects of nectar addition with the average rate of seed predation resulted in λ = 0.99. These studies demonstrate the complex reproductive dynamics of deceptive plants and provide evidence suggesting the evolution of deceit pollination is driven by multiple factors, including predation and decreased fecundity from selfing.

iv ACKNOWLEDGMENTS

I would like to thank my advisor, Dr. Helen Michaels for mentoring me through both undergraduate and graduate school and dragging me begrudgingly into the world of plant pollinator interactions. I also thank my committee, Dr. Karen Root, Dr. Moira van Staaden and

Dr. Randy Mitchell for their feedback throughout my dissertation.

Furthermore I would like to thank the Michaels lab: Mike Plenzler, Jennifer Shimola, and

Jacob Meier for their feedback through endless versions of this dissertation. Also, thank you to my undergraduate field assistant, Paige Arnold, who worked tirelessly in less than ideal conditions, and was monumental in helping me finish the field research. Furthermore I thank

SETGO for funding Paige and my research through the undergraduate research program.

I want to thank the Ohio Department of Natural Resources for allowing me to work at

Resthaven Wildlife area and particularly Jennifer Windus for her input into the orchid, streamlining of the permit process and valuable data on the C. candidum population at

Resthaven. v

TABLE OF CONTENTS

Page

CHAPTER I. GENERAL INTRODUCTION ...... 1

Introduction ...... 1

Figures ...... 5

References ...... 7

CHAPTER II. THE ROLE OF POLLINATOR LIMITATION AND SEED PREDATION

ON THE DECEPTIVE ORCHID CYPRIPEDIUM CANDIDUM ...... 9

Abstract ...... 9

Introduction ...... 10

Materials and Methods ...... 14

Results ...... 19

Discussion ...... 21

References ...... 26

Tables and Figures ...... 31

CHAPTER III. EXAMINING THE ROLE OF FOOD DECEPTION ON

REPRODUCTION IN THE DECEPTIVE ORCHID CYPRIPEDIUM CANDIDUM ..... 38

Abstract ...... 38

Introduction ...... 39

Materials and Methods ...... 44

Results ...... 47

Discussion ...... 49

References ...... 54 vi

Tables and Figures ...... 57

Appendix ...... 64

CHAPTER IV. THE DEMOGRAPHIC IMPLICATIONS OF SEED PREDATION AND

DECEPTION IN THE DECEPTIVE ORCHID CYPRIPEDIUM CANDIDUM

MUHL. EX WILLD...... 65

Abstract ...... 65

Introduction ...... 66

Materials and Methods ...... 71

Results ...... 77

Discussion ...... 80

References ...... 86

Tables and Figures ...... 90

CHAPTER V. CONCLUSIONS ...... 96

Discussion ...... 93

References ...... 100 vii

LIST OF TABLES

Table Page

CHAPTER II

1 MANOVA of the Effect of the Pollination Treatment on Percent Abortion, Percent

Capsules Produced and Average Seed Mass...... 31

2 The Effect of Number of Flowers, Number of Stems and the Average Height of

Flowering Stems on the Initial (A) and Final (B) Fruit Sets of Open and Hand

Pollinated Plants...... 31

3 The Effect of Treatment, Number of Flowers, Number of Stems, Number of Fruits

and the Average Height of Flowering Stems on Rate of Predation Across All

Treatments...... 32

4 The Effect of Hand Outcross, Hand Selfed and Open Pollination Treatments on

Initial Fruit Set, Final Fruit Set, Abortion Rate and Seed Mass...... 32

5 The Effect of Hand Outcross, Hand Selfed and Open Pollination Treatments on

Initial and Final Fruit Set for the 2011 Study...... 33

6 The Effect of Hand Outcross, Hand Selfed and Open Pollination Treatments on

Seed Mass for the 2011 Study...... 33

7 The Effect of Hand Outcross, Hand Selfed and Open Pollination Treatments on

Abortion Rates for the 2011 Study...... 34

CHAPTER III

1 The Effect of Nectar Addition, Quadrat, Number of Stems, Number of Flowers and

the Distance of the Three Nearest Non-Orchid, Nectar-Producing Neighbors on the viii

Percentage of Pollinia, Percent Self Pollinia, and Percent Outcross Pollinia

Received ...... 57

2 The Effect of Nectar Addition, Quadrat, Number of Stems, Number of Flowers and

Distance of the Three Nearest Non-Orchid, Nectar-Producing Neighbors on

Percent Fruit Set, and Percent Fruit Abortion...... 58

3 The Effect of Self and Outcross Pollinia Received, Number of Stems and Number

of Flowers on the Seed Mass of a Capsule...... 58

CHAPTER IV

1 Modified fecundity rates used in the demographic models ...... 90

2 The Effect of Previous Year’s Number of Stems, Number of Flowers and Percent

Capsules (Number of Capsules/Number of Flowers) on the Number of Flowers

Produced the Following Year...... 90

3 The Effect of Previous Year Number of Stems, Number of Flowers and Percent

Capsules on the Probability of Becoming Dormant in the Following Year...... 91 ix

LIST OF FIGURES

Figure Page

CHAPTER I

1 Cypripedium candidum Distribution Map ...... 5

2 Population Map of Prairie at Resthaven Wildlife Area ...... 6

CHAPTER II

1 The Effect of Hand and Open Pollination Treatments on the Percent Initial and

Final Fruit Set With For 2009 ...... 35

2 Percent Capsules at Maturity For Open Pollinated Plants Compared With the

Mean Height of Flowering Stems...... 35

3 Percent Capsules Preyed Upon Compared With, Mean Height of Flowering

Stems of All Plants in the Study ...... 36

4 The Effects of Selfed, Outcrossed and Open Pollination Treatments on

Initial and Final Fruit Set ...... 36

5 The effects of open, outcrossed and self-pollination treatments on seed mass

(g) per fruit ...... 37 x

CHAPTER III

1 The Effect of the Nectar Treatment on the Percentage of Pollinia Received

Blocked by Quadrat ...... 59

2 The Effect of the Nectar Treatment on the Percent of Flowers That Received

Self Pollinia, Blocked by Quadrat ...... 59

3 The Effect of the Nectar Treatment on the Percent of Flowers That Received

Outcross Pollinia, Blocked by Quadrat ...... 60

4 The Effect of the Nectar Treatment on the Percentage of Flowers Setting

Fruit, Blocked by Quadrat ...... 61

5 The Effect of the Nectar Treatment on the Percentage of Flowers Aborted,

Blocked by Quadrat ...... 62

6 The Effect of Self and Outcross Pollinia on the Seed Mass of the Resulting

Capsules...... 63

CHAPTER IV

1 Life Cycle Diagram Illustrating the Stages Used in the Demographic Model...... 92

2 Mean Percent Capsules Produced Per Plant Based on the Number of

Flowers with Standard Errors...... 93

3 The Average Number of Stems, Flowers, and Capsules Compared Between

Years...... 94

4 Average Residence Time in Years for Each Stage of the Model...... 95 1

CHAPTER I: GENERAL INTRODUCTION

Insect pollinated plants are reproductively limited in part by available resources, attraction of suitable pollinators and receipt of pollen (Wesselingh 2007). Increases in plant size, such as the number of flowering stems, or the height of the plant, have been repeatedly shown to increase pollination and seed maturation (Peakall 1989; Aragón and

Ackerman 2004; Mitchell et al. 2004; Li et al. 2011). Plants in which reproduction is limited by the receipt of pollen are referred to as “pollen limited”. Pollen limitation can be defined as the difference between natural seed production and seed production in individuals receiving supplemental pollen. Populations are considered pollen limited if the average natural seed production is significantly less than the average seed production of individuals receiving supplemental pollen (Knight 2003). Decreased fecundity from pollen limitation may drive floral trait selection through impacts on the life-time fitness of individuals.

In addition to increased plant, flower or display size, plants have evolved a multitude of floral forms in order to maximize pollinator efficiency (Barrett 2002). One such adaptation is deceit pollination, a pollination strategy in which the flower provides floral cues indicating a reward while not actually providing that reward (Faegri and va der Pijl 1971; Dafni 1987; Cozzolino and Widmer 2005). Deceit pollination has been shown to decrease pollination when compared to rewarding plants (see Johnson et al.

2004); however, it has been hypothesized that this decrease in pollination is offset with decreased intra-individual self pollination, or geitonogamy (Smithson 2002; Johnson et al. 2004; Kindlmann & Jersáková 2006; Sun et al. 2009). 2

Post-pollination factors such as seed predation also impact reproduction and long- term population persistence by reducing the number of offspring produced after pollination (Louda and Potvin 1995; Russell et al. 2010). Many seed predators rely on developing ovules and seeds to feed their offspring (Cariveau et al. 2004). Seed predators may be attracted to the same floral resource cues as pollinators, resulting in conflicting selective pressures (Stephens and Myers 2012).

Demographic models are often utilized in order to understand how pre and post- pollination factors affect population dynamics. Demographic modeling is a technique used to examine population level variation over space and time (Kéry and Gregg 2004).

In order to analyze demographic data, matrix models are often employed, (Lefkovitch

1965; Caswell 1989), which organize the data into age, size or stage specific rates and incorporate survival, growth and reproduction into linear equations (Rooney and Gross

2003; Jacquemyn et al. 2007) in order to calculate the growth rate of the population (λ).

Demographic models may be used to predict the long-term persistence of a threatened species (Population Viability Analyses) and to understand the lifetime effects of complex ecological interactions.

The genus Cypripedium, commonly referred to as the temperate Lady’s Slipper

Orchids, is a genus of food deceptive orchids that occurs throughout most of the Northern

Hemisphere. The state of Ohio has four native Cypripedium species that differ in number of populations and size of the populations. The white Lady’s Slipper Orchid,

Cypripedium candidum, occurs in prairies throughout the United States and Canada

(Figure 1). Due to the large scale conversion of prairie to agriculture, as well as poaching, over the course of the 20th century C. candidum populations have declined 3 throughout their range. In Ohio, C. candidum occurs in only two geographically disjunct populations. The Northern population, Resthaven Wildlife Area, contains more than

6000 individuals, whereas the recently discovered Southern population, Edge of

Appalachia, contains less than 80. Despite the large individual size of the Resthaven population, a total of only two populations make this species perilously susceptible to extirpation from Ohio due to stochastic events. According to NatureServe (2013), throughout the rest of its range C. candidum is listed as extirpated, threatened, or potentially threatened.

Furthermore, the life-history characteristics of C. candidum (long-lived, slow to reach reproductive maturity, and low net fecundity), limit C. candidum reproduction. In contrast to most angiosperm seeds, C. candidum seed contains no endosperm and many orchids have been found to have a limited seedbank (Whigham et al. 2006). Due to the lack of endosperm, germination requires the presence of a symbiotic mycorrhizal fungus and total time between germination and emergence can be as long as four years (Kull

1999).

Although the data obtained in this experiment is likely to be relevant to C. candidum in other areas and Cypripedium in general, this study focuses on the northern

Ohio population of C. candidum,. In order to understand the relationship between deceit pollination, pollen limitation and pre-dispersal seed predation in the deceptive orchid,

Cypripedium candidum, I completed the following studies in three separate sub- populations at Resthaven Wildlife area (Figure 2):

1: The role of pollinator limitation and seed predation on the reproductive success of the deceptive orchid, Cypripedium candidum 4

2: Examining the role of food deception on reproduction in the deceptive orchid

Cypripedium candidum:

3: The demographic implications of granivory and deception in the rewardless orchid,

Cypripedium candidum Muhl ex. Willd.

In the second chapter we determined whether the attraction of a suitable pollinator and successful receipt of pollen is a limiting factor to C. candidum reproduction. We examined the effect of pollen quality on seed production by conducting a hand pollination experiment using both outcross and self pollen. Additionally, we examined the role of post-pollination seed predation and its effect on the net fecundity of C. candidum.

The third chapter focused on how deceit pollination affects overall fitness as well as selfing and outcrossing rates. In this study we artificially created rewarding plants by adding nectar to half of the study individuals to determine whether food rewards alter pollinator behavior leading to selfing. Additionally we color coded each pollinia using histochemical dyes to track pollinia movement and quantify the presence or absence of selfing events. to determine whether food rewards alter pollinator behavior leading to selfing.

In the fourth chapter we constructed a population demographic model using four years of demographic data. Using the demographic model as a starting point, we adjusted the fecundity rates of the model in order to assess the population level impacts of seed predation and deceit pollination on C. candidum.

Figures

Figure 1. Range map for the Small White Lady’s Slipper Orchid, Cypripedium candidum, color coded with its conservation status (Natureserve 2013).

Chapter 1 Chapter 2 Chapter 3

Figure 2. Population map of prairie at Resthaven Wildlife Area. C. candidum sub- populations denoted by numbers. Colored stars indicate populations where study chapters took place.

Literature Cited

Aragón, S., & Ackerman, J. D. (2004). Does flower color variation matter in deception pollinated Psychilis monensis (Orchidaceae)? Oecologia, 138(3), 405–13.

Barrett, S. C. H. 2002. The evolution of plant sexual diversity. Nature Reviews Genetics 3: 275-284.

Cariveau, D., Irwin, R. E., Brody, A. K., Garcia-Mayeya, L. S., & Von der Ohe, A. (2004). Direct and indirect effects of pollinators and seed predators to selection on plant and floral traits. Oikos, 104(1), 15–26.

Caswell, H. (1989). Analysis of life table response experiments I. Decomposition of effects on population growth rate. Ecological Modelling, 46(3-4), 221–237.

Cozzolino, S., & Widmer, A. (2005). Orchid diversity: an evolutionary consequence of deception? Trends in ecology & evolution, 20(9), 487–94.

Dafni, A., & Ivri, Y. (1979). Pollintation ecology of, and hybridization between, Orchis coriophora L. and O. collina SOL. EX RUSS. (Orchidaceae) in Israel. New Phytologist, 83(1), 181–187.

Faegri, K., & Van der Pijl, L. (1971). The Principles of Pollination Ecology. Pergamon.

Jacquemyn, H., De Meester, L., Jongejans, E., & Honnay, O. (2012). Evolutionary changes in plant reproductive traits following habitat fragmentation and their consequences for population fitness. Journal of Ecology, 100(1), 76–87.

Johnson, S. D., Peter, C. I., & Agren, J. (2004). The effects of nectar addition on pollen removal and geitonogamy in the non-rewarding orchid Anacamptis morio. Proceedings. Biological sciences / The Royal Society, 271(1541), 803–9.

Knight, T. M. (2003). Floral density, pollen limitation, and reproductive success in Trillium grandiflorum. Oecologia, 137(4), 557–63.

Knight, T. M. (2004). The effects of herbivory and pollen limitation on a declining population of Trillium grandiflorum. Ecological Applications, 14(3), 915–928.

Kull, T. (1999). Cypripedium calceolus L. Journal of Ecology, 87(5), 913–924.

Kéry, M., & Gregg, K. B. (2004). Demographic analysis of dormancy and survival in the terrestrial orchid Cypripedium reginae. Journal of Ecology, 92(4), 686–695. 8

Li, P., Huang, B. Q., Pemberton, R. W., Luo, Y. B., & Cheng, J. (2011). Floral display influences male and female reproductive success of the deceptive orchid Phaius delavayi. Plant Systematics and Evolution, 296(1-2), 21–27.

Louda, S. M., & Potvin, M. A. (1995). Effect of Inflorescence-Feeding Insects on the Demography and Lifetime of a Native Plant. Ecology, 76(1), 229.

Mitchell, R. J., Karron, J. D., Holmquist, K. G., & Bell, J. M. (2004). The influence of Mimulus ringens floral display size on pollinator visitation patterns. Functional Ecology, 18(1), 116–124.

NatureServe. 2013. NatureServe Explorer: An online encyclopedia of life [web application]. Version 7.1. NatureServe, Arlington, Virginia. Available http://www.natureserve.org/explorer. (Accessed: April 3, 2013 ).

Peakall, R. (1989). The unique pollination ofLeporella fimbriata (Orchidaceae): Pollination by pseudocopulating male ants (Myrmecia urens, Formicidae). Plant Systematics and Evolution, 167(3-4), 137–148.

Rooney, T. P., & Gross, K. (2003). A demographic study of deer browsing impacts on Trillium grandiflorum. Plant Ecology, 168, 267–277.

Russell, F. L., Rose, K. E., & Louda, S. M. (2010). Seed availability and insect herbivory limit recruitment and adult density of native tall thistle. Ecology, 91(10), 3081– 3093.

Smithson, A. (2002). The consequences of rewardlessness in orchids: reward- supplementation experiments with Anacamptis morio (Orchidaceae). American Journal of Botany, 89(10), 1579–87.

Stephens, A. E. A., & Myers, J. H. (2012). Resource concentration by insects and implications for plant populations. Journal of Ecology, 100(4), 923–931.

Sun, H.-Q., Cheng, J., Zhang, F.-M., Luo, Y.-B., & Ge, S. (2009). Reproductive success of non-rewarding Cypripedium japonicum benefits from low spatial dispersion pattern and asynchronous flowering. Annals of botany, 103(8), 1227–37.

Wesselingh, R. A. (2007). Pollen limitation meets resource allocation: towards a comprehensive methodology. The New phytologist, 174(1), 26–37.

Whigham, D. F., O’Neill, J. P., Rasmussen, H. N., Caldwell, B. A., & McCormick, M. K. (2006). Seed longevity in terrestrial orchids – Potential for persistent in situ seed banks. Biological Conservation, 129(1), 24–30.

CHAPTER II: THE ROLE OF POLLINATOR LIMITATION AND SEED PREDATION ON THE REPRODUCTIVE SUCCESS OF THE DECEPTIVE ORCHID, CYPRIPEDIUM CANDIDUM

Abstract

For many species of conservation significance, multiple factors limit reproduction. This research examines the contributions of plant size, pollen limitation and seed predation as factors limiting reproduction in the White Lady’s Slipper Orchid, Cypripedium candidum

Muhl ex Willd. Cypripedium spp. require pollinator visitation as the relatively complex floral anatomy prevents passive selfing. While increased floral display size can serve to attract additional pollinators, C. candidum’s multiple synchronously flowering stems may promote geitonogamous selfing and also increase seed predator attraction. C. candidum are preyed upon by the Orchid Weevil, Stethobaris ovate, which are known to consume embryos of Cypripedium spp as well as other orchid species. Pollen limitation, seed predation and reproductive trade-offs were examined in a field experiment in a population (n > 6000) over two flowering seasons (2009 and 2011). In the 2009 study, 36 pairs of plants, matched by flower number, received either supplemental hand or open pollination. Plants were scored for fruit maturation, mass of seeds, and seed predation rates. Pollen supplementation increased proportion of flowers maturing fruit, with 87% fruit set for flowers receiving added pollen compared to only 46% for those receiving only natural pollinator service. Taller inflorescences had higher initial and final fruit set, while other measures of plant size were not associated with female function. Total seed set was also limited by seed predation, with 73% of the capsules affected, while shorter stems were more likely to be preyed upon. Number of fruits had no effect on predation rates. A similar 2011 experiment with the addition of a self-pollination treatment found

10 that both initial and final fruit set were higher in the self and outcross hand pollination treatments than in the open pollinated treatment. However, seed mass was significantly higher in both open pollinated and outcross hand pollination treatments compared to those hand self-pollinated. This research suggests a potential mechanism driving the outcrossing hypothesis, which explains the benefits of deceptive orchid pollination via increased outcrossing

Introduction

Understanding the complex dynamics between plants, pollinators and seed predators, and how these interactions affect plant reproduction is critically important to understanding the ecology and evolution of plant reproduction (Irwin and Brody 2011).

Despite the importance of understanding the effects of both mutualists and antagonists on plant reproduction, relatively few have examined these roles simultaneously (Strauss and

Irwin 2004; Abdala-Roberts et al. 2009; Burkhardt et al. 2009; Carlson and Holsinger

2010). Simultaneous interactions between multiple species may extend beyond the expected outcomes of simple pairwise interactions (Straus and Irwin 2004). Both mutualists and antagonists may select flowering stems based on display size, where larger floral displays provide a source of concentrated resources for pollinators (Peakall 1989;

Brody and Mitchell 1997; Mitchell et al. 2004) or seed predators (Stephens and Myers

2012). Our goal was to assess the effects of floral display size on both pollination and seed predation and understand how these biotic factors influence reproductive success.

Maximizing pollen transfer efficiency has greatly shaped the evolution of a multitude of floral forms and functions in angiosperms (Barrett 2002). Increased floral

11 display size is expected to increase pollinator attraction and visitation (Peakall 1989;

Aragón and Ackerman 2004; Li et al. 2011). In an experimental manipulation of

Mimulus ringens floral display size, Mitchell et al. (2004) found that bumble bee pollinators strongly responded to large floral display size by probing more flowers in sequence on plants with large numbers of flowers. Larger floral displays also increase the visibility of the plant, thereby theoretically increasing the attraction of pollinators from greater distances (Kindlmann & Jersáková 2006). On a population level, synchronous flowering effectively increases floral display size increasing local resource density for the pollinator while simultaneous increasing competition for pollinators from the plant perspective.. However, plants in lower density populations may be more pollen limited due to lack of available pollinators (Knight 2003; Dauber et al. 2010).

Pollen limitation can be defined as the difference between open pollinated

(pollination with no human interference) seed production and seed production in individuals receiving supplemental pollen. Populations are considered pollen limited if the average open pollinated seed production is significantly less than the average seed production of individuals receiving supplemental pollen (Knight 2003). When supplemental pollen is provided, saturating the stigmatic surface, a plant should on average produce as many seeds as ovules present as long as other resources aren’t limitingpresented with an overabundance of a pollen resource, individuals on average should produce a maximum amount of seeds up to the point that other resources are limiting reproduction (Ashman et al. 2004). The standard method of assessing pollen limitation is to compare the fruit set between open pollinated plants and plants that have had all of their flowers hand pollinated (Wesselingh 2007). Ideally, all flowers on an

12 experimental plant should be pollinated to avoid redirection of resources from non- pollinated flowers, thereby skewing any information on fruit or seed set (Knight et al.

2006; Wesselingh 2007). Haig and Westoby (1988) suggested that if viewed from an evolutionary perspective, in a population of plants that are perpetually pollen limited, natural selection should favor plants that allocate more resources towards pollen attraction. At the same time, in a resource-limited environment, increases in resources dedicated to pollen attraction will necessarily decrease resources available for seed development. Therefore plants should reach an equilibrium that maximizes ovule fertilization while minimizing resources devoted to pollen attraction.

Selfing events, such as geitonogamy, may result in both inbreeding depression and pollen discounting (Johnson and Nilsson 1999). While increased floral display size may increase reproduction by attracting more pollinators, it may also substantially increase geitonogamy (de Jong et al. 1993; Barret and Harder 1996; Snow et al. 1996;

Galloway et al. 2002). Geitonogamy may reduce female function by reducing the number or quality of the offspring through inbreeding depression, but may also impact male function by reducing the quantity of pollen available for export to other plants, also known as pollen discounting (Johnson et al. 2004).

After successfully receiving pollen, plants face further pressures in the form of seed predation. Pre-dispersal seed predation can play an important role in determining fecundity and long-term population persistence (Louda and Potvin 1995; Russell et al.

2010). Chronic, high levels of seed predation may limit population growth by reducing fecundity, while stochastic predation rates can play a more diffuse, but equally important role in population dynamics (Kolb et al. 2007). Just as pollinators are often attracted to

13 large floral displays, seed predators may be attracted to the accompanying large ovule resource (Stephens and Myers 2012) as many seed predators rely on developing ovules to feed their offspring (Cariveau et al. 2004). Therefore, dense floral resources may attract mutualist pollinators, while the accompanying dense fruit resources attract antagonist seed predators, creating conflicting selective pressures (Louda and Potvin 1995; Ågren et al. 2008). A variable predatory environment coupled with a stochastic pollination environment could strongly select for Burd’s (1995) model of overproduction of ovules as a means of bet-hedging.

The goal of this study was to assess the effects of plant floral display size on both pollination and seed predation, and understand how these biotic factors influence reproductive success in the long-lived, highly specialized deceptive orchid, Cypripedium candidum. At our study site C. candidum may produce 1-12 single flowered stems per plant. We hypothesized that while an increase in floral display should increase pollinator visitation, thereby increasing fruit set, it could also increase geitonogamy, resulting in an increase in fruit abortion, decreased fruit maturation, or offspring fitness. However, the potential increased pollination and initial fruit set may provide an attractive resource concentration for seed predators, subsequently reducing plant fecundity (Light and

MacConail1 2011; Stephens and Meyers 2012). In order to address these hypotheses, we asked the following questions: 1. Does pollen addition increase fruit set, indicating pollen limitation? 2. How much does seed predation reduce plant fecundity? 3. Are indicators of larger plant size (height of plant, numbers of flowers, fruits and leaves) associated with greater open pollination and seed predation rates? And 4. How does pollen quality (i.e. self vs. outcross pollen) affect fruit abortion and maturation rates? We conducted two

14 pollen limitation experiments over the course of the 2009 and 2011 field seasons in two separate patches, examining the effects of plant size on pollinator limitation and seed predation in 2009, and the effect of pollen quality in a seed predator exclusion experiment in 2011.

Materials and Methods

Study Species and Site

Cypripedium are deciduous, terrestrial orchids with growth emerging from a subterranean rhizome (Stoutamire 1967). Cypripedium candidum Muhl ex Willd., the

Small White Lady’s Slipper, has yellow-green lateral sepals and petals with a white labellum spotted with purple (Stoutamire 1967), and occurs in calcareous prairies as well as fens and limestone barrens (Cusick 1980). C. candidum is highly dependent on full sun in open areas and, as with many prairie species, populations begin to decline with the invasion of woody plants (Curtis 1946).

The primary field site for this study is located at Resthaven Wildlife Area in

Castallia, Ohio (GPS coordinates available upon request). Historically C. candidum existed in at least 7 Ohio counties (OSU Herbarium, BGSU Herbarium). Currently

Resthaven Wildlife Area is one of only two locations in Ohio where C. candidum remains (Pers. Comm. ODNR). Resthaven has a large, actively managed prairie area

(approximately 900 hectares) with wooded areas intertwined. The prairie is maintained through controlled burns in early spring approximately every three years (J. Windus,

ODNR pers. com.). Large, thriving populations of C. candidum (total N ~ 6000) exist in the calcareous soil and actively managed regions of the prairie.

The weevil, Stethobaris ovate, is a known seed predator of Cypripedium spp. and other temperate orchid genera, with reports of adult weevils feeding on emerging shoots and flower buds (Light and MacConaill 2011). Adult weevils emerge in early spring along with Cypripedium shoots, and oviposit in developing fruits and possibly stems, resulting in either fruit abortion or near total loss of the developing embryos (Light and

MacConaill 2011). Predation rates on C. parviflorum in Canada have varied from 32%-

53% depending on climate and availability of fruit resources (Light and MacConaill

2002). Little is known about the life history of the organism, although it is suspected they may complete two life cycles within a growing season (pers. comm. M. Light).

2009 Study

In early May 2009, three 60 m line transects were established within a patch of C. candidum (n>250) and sampled at 5 m intervals (n = 36). The population density of C. candidum at Resthaven WA was previously estimated at 3.26 plants per meter2 (range =

1-9 plants/m2; SD = 2.68) (Walsh Ch. 3). Initial placement of transects through the population was randomly selected. For each sampling point, the plant closest to the transect was selected and a second plant of equal size (number of stems and flowers) was selected on the other side of the transect within 0.5 m. Plants with flowers that had already opened prior to setup of the experiment, or had any pollinia removed or deposited were excluded from the study. One of the paired plants was randomly chosen to receive a hand pollination treatment with pollinia from a different population at Resthaven at least 100m away, while the other member of the pair was open pollinated. The number of flowering stems, number of total stems, number of leaves per stem and the height of each

16 flowering stem (to the nearest 0.1 cm) were also recorded. Because the site was burned approximately 1 month prior to sampling, surrounding vegetation was sparse during the flowering period, precluding the collection of surrounding vegetation data. Vegetative cover at Resthaven is dominated by Andropogon gerardii, Sorghastrum nutans and

Silphium terebinthinaceum, with Viola spp., Sisyrinchium montanum, and Fragaria virginiana in flower along with C. candidum.

Each flower received a single pollinium from a mixed batch of pollinia gathered earlier the same day. After supplemental hand pollination the flowers were bagged with mesh (mesh opening size=3mm x 3mm) to prevent accidental removal of the pollinium while allowing ample room for seed predators to enter. Mesh bags were removed after all flowers in the experiment had dehisced (approximately 2 weeks) to allow weevil predation and oviposition. Capsule development was recorded 1 month after floral dehiscence (June) as well at maturity in August when all capsules were collected and scored for insect damage. Fruit abortion was scored as the number of fruits initiated in

May minus the number matured in August. Fruit abortion rate was then calculated by dividing the number of fruits aborted by the total number of fruits produced. Mature capsules were dried at 60°C for three days prior to obtaining the masses for the capsule and seeds. Capsules were scored as preyed upon when circular insect exit holes, approximately 1mm in diameter, were observed on the mature capsules. In addition to the presence of an exit hole, weevil body parts (possibly molts) and a lack of seeds provided clear evidence for weevil predation.

2011 Study

In May 2011 prior to flower opening, we set up a second pollen limitation study at a different nearby population of similar size to the 2009 study at Resthaven Wildlife

Area. In this experiment, a hand self-pollination treatment was added, stem and flower number were controlled, and fruits were protected from weevil predation in order to obtain enough fruits for analysis of the effect of pollen quality on fruit set and seed mass.

Five 50m transects were randomly located across the population. For this study, only 3- flowered, 3-stemmed individuals were chosen to control for size of the individuals, to correspond with the three treatment levels, and to limit within-year resource reallocation issues that might arise if non-experimental flowers set fruit. At 5m intervals along each transect the nearest 3-flowered, 3-stemmed individual was tagged and the flower buds were initially wrapped in mesh to prevent visitation. Each stem was randomly assigned one of three treatments: hand self-pollinated, hand pollinated with outcross pollen from a different population (>100m away), or open pollination. Stems were marked using colored wire wrapped around the base of the stem to differentiate the respective treatment levels on a given plant. Plants were checked daily for flower opening and treatments were applied when the stigma became receptive. Open pollinated stems had their mesh bags removed as soon as the flowers opened, while hand-outcross and hand-selfed flowers were re-bagged after receiving their appropriate treatments. Following floral dehiscence the pollinator exclusion bags were removed. Initial fruit set was scored two weeks after flower dehiscence, with green, enlarged fruits scored as pollinated and yellow, shrunken or missing fruits scored as a failed pollination. All fruits at this time were covered in dialysis tubing, secured at both ends of the fruit to exclude insect

18 damage. Fruit abortion was scored 4 weeks after flower dehiscence as well as at the end of the study (August 2011). Fruits were collected 3 months after flower dehiscence, dried at 65°C for 48 hours and seed mass was weighed to the nearest hundred-thousandth of a gram on a Mettler AE-240 scale (Mettler-Toledo Inc.).

Data Analysis

Data was analyzed using JMP v.9.0.2 (SAS Institute, California, U.S.A.). For the

2009 study, MANOVAs examining the effects of plant size on fruit set and seed predation were initially performed, and then each variable was examined using a

Generalized Linear Model with an identity link function to examine the effect of treatment, number of stems, fruit abortion and number of flowers on fruit set and seed production. The proportions of fruit set and predation were arcsin square root transformed. The effect of number of flowers, number of stems, number of fruits and treatment on the probability of predation was examined using a Generalized Linear

Model (identity link function) with n = 62 plants.

In the 2011 study, MANOVAs examining the effect of pollination treatments on initial fruit set, final fruit set and seed mass were performed followed by a one-way

ANOVAs to test for differences amongst treatment groups on fruit set, abortion rates and final seed mass.

Results

2009 Study

Plants in the study had an average of 2.06 flowers (SE = 0.14, range 1-7) flowers and 3.05 stems (SE = 0.23, range 1-9) per individual. Mature flowering stems had an average height of 22.6 cm (SE = 0.47), ranging between 15.1 cm and 40.9 cm. The average number of leaves on a plant varied little, with an average of 3.2 leaves per stem

(SE = 0.01, range 3-4). Leaf length and width were not measured in this study based on prior work (Walsh 2008) showing little variation in sizes of these variables between individuals.

Analysis of initial fruit set, measured 1 month after floral dehiscence, indicated a significant increase in early fruit set for plants receiving supplemental pollen compared to open pollinated flowers (p <0.0001; Figure 1). The MANOVA on final percent fruit set, seed mass and abortion rates indicated a significant treatment effect (Wilks λ = 0.347, p <

0.0001) (Table 1). Plants that received supplemental pollen had an initial fruit set of

89%, while plants allowed to open pollinate had substantially lower initial fruit set

(Figure 1). Final fruit set, measured at three months post floral dehiscence, also indicated significant pollen limitation (p<0.0001, Figure 1). Study plants receiving supplemental pollen set capsules 87% of the time, while only 46% of open pollinated plants set fruit

(Figure 1). Post-pollination processes such as abortion did not influence reproduction and did not differ between treatments (t = 1.21, p = 0.88).

Analysis of initial fruit set indicated that floral display size, average height of flowering stems and number of stems did not influence this component of reproduction

(Table 2) while there was a significant effect of treatment (p < 0.0001). Number of

20 leaves did not vary amongst treatment plants and was omitted from the analyses. When examining the final fruit set, the average height of a flowering stem did significantly influence final fruit set (p = 0.03, Table 2) along with treatment (p < 0.0001). When examining open pollinated plants separately, taller stems were more likely to set more

2 fruit (F1,35 = 7.0, p = 0.012, R = 0.17, Figure 2). Mean number of flowers and number of stems had no significant impact on final fruit production (Table 2).

The analysis of the effect of size variables on predation rates showed the average height of flowering stems was the only size variable to significantly explain probability of predation (p = 0.001, Table 3), with taller stems less likely to be preyed upon (F1,61 =

10.1, p = 0.002, R2 = 0.14, Figure 3). Treatment, number of flowers, number of fruits and number of stems had no effect on predation rates (Table 3). Of the flowering stems that set fruit, 73% were preyed upon. Fruits suffering predation had poor seed production, as predated capsules had 89% less seed mass than capsules that were not preyed upon.

2011 Study

A MANOVA of initial fruit set, final fruit set and abortion rates indicated a significant difference between the three pollination treatments (Wilks λ = 0.409, p =

0.0011, Table 4). Initial fruit set, scored two weeks after floral dehiscence significantly differed amongst pollination treatments (F2,89 = 7.73, p = 0.0008, Figure 4, Table 5).

Plants receiving hand self pollinations had a significantly higher fruit set than the open- pollinated plants (p<0.05 Figure 4).

Final fruit set, scored 12 weeks after floral dehiscence, was affected by the pollination treatments (F2,89 = 5.73, p = 0.0046, Figure 4, Table 5). The self and outcross

21 hand-pollinations produced more fruits (self mean fruits per flower = 0.46, outcross =

0.4) than the open pollinated plants (mean = 0.1), while the fruit set of selfed and outcrossed treatments were similar (Figure 4). Final seed mass showed a significant difference between treatment groups (p<0.0001, F2,28 = 18.72, Figure 5, Table 6). The open and outcross treatments produced significantly more average seed mass per plant than the self pollinated treatments (open = 0.027g, outcross = 0.026g), while seed masses of the open pollinated and outcross hand pollination treatments were similar (Figure 5).

There was no significant difference in abortion rates amongst treatments (Table 7). A power analysis indicated that 81 replicates would have been needed in order to reach a significance level of p < 0.05.

Discussion

This study documents strong evidence of pollen limitation in a deceptive orchid over two flowering seasons on different individuals in different populations. In the 2009 study we observed a relatively high fruit set from open pollination (46%) while fruit set was greatly reduced (16.6%) for plants in the 2011 study. The 2011 fruit set more closely matches fruit set from open pollination in other related species such as 10.5 % in C. calceolus (Kull 1998) and 5-13% in C. acaule (O’Connell and Johnston 1998). The 2009 study took place immediately after a controlled burn of the prairie, suggesting that both increased visibility to pollinators and the surge of nutrients from the burn contributed to a higher than normal fruit set. Hand pollinated fruit set was consistent between study years, with at least 40% of the flowers setting fruit when supplemental pollen was provided. Orchid species have been routinely shown to demonstrate consistent pollen

22 limitation across multiple years (Snow and Whigham 1989; Calvo 1990; Primack and

Hall 1990; Ackerman and Montalvo 1990; Dudash and Fenster 1997). Furthermore, deceptive orchid species have been found to produce only half as many fruits as their non-deceptive counterparts (Johnson and Bond 1997; Neiland and Wilcock 1998;

Tremblay et al. 2004; Jersáková et al. 2006).

Pollen limitation may consist of two components, pollen quantity and pollen quality (Aizen and Harder 2007). In plant systems producing normal, dust-like pollen, inadequate saturation of the stigmatic surface may result in only partial pollination.

Orchids on the other hand, with pollen aggregated into sac-like pollinia containing large amounts of pollen, will have a more binary response: the flower is either pollinated with full seed set per fruit or not. Knight et al. (2005) defined pollen limitation as:”…when plants produce fewer fruits and/or seeds than they would with adequate pollen receipt.”

Although Aizen and Harder (2007) make the argument that pollen supplementation often involves high quality outcross pollen that could inflate seed production estimates, our

2011 study found that manipulation of pollen quality in this system did not significantly increase fruit production, although it did have a significant effect on seed mass. Studies from both years show no significant difference in seed mass/fruit between open pollinated and hand-outcrossed flowers, suggesting that many of the open pollinated flowers were the product of out-crossed pollen.

The 2009 study specifically addressed the effect of plant and floral display size on successful fruit maturation. We saw no effect of number of flowers, number of stems or number of leaves on the receipt of pollen or overall fruit maturation. The height of a flowering stem was the only measured size variable that significantly explained fruit set

23 across treatments. The height of the flowering stem also increased pollination and fruit production in a closely related species C. acaule (O’Connell and Johnston 1998). Given the tall grass prairie environment in which C. candidum occurs, a taller flower would be more visible to pollinators through the vegetation and therefore be more likely to receive pollinator servicing. Furthermore, taller flowering stems have increased sun exposure creating a warmer environment for pollinators in the spring, and increased photosynthetic rates. Numerous studies have shown that an increase in floral display and plant size increases pollen receipt and fruit maturation (Peakall 1989; Aragón and Ackerman 2004;

Meléndez-Ackerman and Ackerman 2001;Mitchell et al. 2004; Li et al. 2011), contrary to our fruit maturation results. The population we worked in was very dense and the large numbers of closely spaced individuals may have negated any effect of floral display size on fruiting success.

The presence of a deceptive pollination system may explain why our results with

C. candidum are contrary to the other literature. Although other plants with larger floral displays attract pollinators from greater distances (Sih and Baltus 1987; Hessing 1988), the absence of a reward may discourage C. candidum pollinators from subsequent pollination events on the same plant and select against any increase in pollination and fruit set that would otherwise be seen in a large multi-flowered rewarding plant.

Jersáková et al. (2006) cite numerous examples of deceptive orchids having reduced geitonogamy. Other studies using nectar supplementation in deceptive orchids have found dramatic increases in self-pollination when a reward is added (Johnson et al. 2004;

Jersáková et al. 2006; Walsh Ch.2). In this situation, a taller stem may have the highest probability of attracting a pollinator, but the pollinator quickly leaves the plant after

24 receiving no compensation for its efforts. Although the deceptive system will result in increased pollen limitation compared to a rewarding flower, as seen in this study and numerous other studies on deceptive orchids (Tremblay et al. 2004; Jersáková et al.

2006), it may be more advantageous to produce fewer, but higher quality fruits than producing large numbers of lower quality (selfed) offspring. Indeed, our 2011 study showed a dramatic decrease in seed mass on flowering stems receiving supplemental selfed pollen when compared to plants that received supplemental outcrossed pollen or open pollination.

Seed predation heavily reduced total reproductive output of the population with

73% of all capsule producing stems with predated fruit. All capsules appeared to be attacked by the same insect, identified as a weevil in the Stethobaris genus previously reported to prey upon Cypripedium fruit (Light and MacConaill 2011). Our analysis revealed a significant increase in weevil predation in shorter flowering stems. As with pollination, other size variables such as number of flowers, number of stems and number of leaves had no effect on predation rates. Weevil predators feed upon the leaves, flowers and developing capsules of many Cypripedium spp. and oviposit in the maturing capsules (Light and MacConaill 2011). Our data suggests these weevil predators do not respond to the number of fruits on a plant, but instead utilize shorter stems that are possibly easier to access. Greater potential risk of predation as well as the increased energy required to climb taller stems may explain the weevil preference for shorter flowering stems. Furthermore, in a tall grass prairie, shorter flowering stems will be surrounded by the previous years’ detritus and current year’s vegetative growth, potentially providing greater cover from predators. Weevil foraging and oviposition in

25 this habitat may further promote selection towards taller flowering stems that are both more visible to pollinators and offer less cover for seed predators. Complex pollinator/seed predator interactions have been documented by others (Strauss and Irwin

2004), although instances of conflicting pressures seem to greatly outnumber instances of unidirectional pressure.

Our study quantifies a unique plant, pollinator, and seed predator interaction in a deceptive orchid system. Fruit set increased with taller stems, but other traditional indicators of plant attractiveness such as floral display size, numbers of stems and numbers of leaves did not affect pollination. Seed predators may be attracted to more easily reached resources that are sheltered by surrounding vegetation and less apparent to their invertebrate predators (Marquis 1992; Palo et al. 1993). Additionally this study demonstrates greater seed mass from outcrossed hand and open pollination events, suggesting that most pollination in C. candidum arises predominately from outcrossing.

These data suggest a potential mechanism driving the outcrossing hypothesis, which explains the benefits of deceptive orchid pollination via increased outcrossing (Dafni and

Ivri 1979; Nilsson 1983; Ackerman 1986; Johnson and Nilsson 1999; Jersáková et al.

2006). However, while there have been several studies examining the effect of nectar addition on deceptive orchids, there has been no study to date explicitly testing the demographic consequences of the deceptive pollination strategy in any plant. Further research confirming increased outcrossing rates as a result of deceit pollination, as well as understanding the demographic consequences of the strategy, are needed to understand the evolution and maintenance of this unique reproductive syndrome.

Literature Cited

Abdala-Roberts, L., Parra-Tabla, V., Salinas-Peba, L., & Herrera, C. M. (2009). Noncorrelated effects of seed predation and pollination on the perennial herb Ruellia nudiflora remain spatially consistent. Biological Journal of the Linnean Society, 96(4), 800–807.

Ackerman, J.D. (1986). Mechanisms and evolution of food-deceptive pollination systems in Orchids. Lindleyana, 1, 108–113.

Ackerman, J D., & Montalvo, A. M. (1990). Short- and long-term limitations to fruit production in a tropical Orchid. Ecology, 71(1), 263–272.

Ågren, J., Ehrlén, J., & Solbreck, C. (2008). Spatio-temporal variation in fruit production and seed predation in a perennial herb influenced by habitat quality and population size. Journal of Ecology, 96(2), 334–345.

Aizen, M. A., & Harder, L. D. (2007). Expanding the limits of the pollen-limitation concept: Effects of pollen quantity and quality. Ecology, 88(2), 271–281.

Aragón, S., & Ackerman, J. D. (2004). Does flower color variation matter in deception pollinated Psychilis monensis (Orchidaceae)? Oecologia, 138(3), 405–13.

Ashman, T.-L., Knight, T. M., Steets, J. A., Amarasekare, P., Burd, M., Campbell, D. R., Dudash, M. R., et al. (2004). Pollen limitation of plant reproduction: Ecological and evolutionary causes and consequences. Ecology, 85(9), 2408–2421.

Barrett, S. C. H. (2002). The evolution of plant sexual diversity. Nature Reviews Genetics 3: 275-284.

Barret, S.C.H., Harder, L.D. (1996). Ecology and evolution of plant mating. Trends Ecol Evol, 11, 73-79.

Brody, A.K. and Mitchell, R.J. (1997). Effects of experimental manipulation of inflorescence size on pollination and pre-dispersal seed predation in the hummingbird pollinated plant Ipomopsis aggregata. Oecologia. 110, 86-93.

Burd, M. (1995). Pollinator behavioural responses to reward size in Lobellia deckenii: No escape from pollen limitation of seed set. Journal of Ecology, 83(5), 865–872.

Burkhardt, A., Delph, L. F., & Bernasconi, G. (2009). Benefits and costs to pollinating, seed-eating insects: the effect of flower size and fruit abortion on larval performance. Oecologia, 161(1), 87–98.

Calvo, R. N. (1990). Inflorescence size and fruit distribution among individuals in three Orchid species. American Journal of Botany, 77(10), 1378–1381.

Cariveau, D., Irwin, R. E., Brody, A. K., Garcia-Mayeya, L. S., & von der Ohe, A. (2004). Direct and indirect effects of pollinators and seed predators to selection on plant and floral traits. Oikos, 104(1), 15–26.

Carlson, J. E., & Holsinger, K. E. (2010). Natural selection on inflorescence color polymorphisms in wild Protea populations: The role of pollinators, seed predators, and intertrait correlations. American Journal of Botany, 97(6), 934–44.

Curtis, J. T. (1946). Use of mowing in management of white ladyslippers. Journal of Wildlife Management, 10, 303–306.

Cusick, A. W. (1980). Cypripedium candidum Muhl ex Willd. Ohio Heritage Database.

Dafni, A., & Ivri, Y. (1979). Pollination ecology of, and hybridization between, Orchis coriophora L. and O. collina Sol. ex Russ. (Orchidaceae) in Israel. New Phytologist, 83(1), 181–187.

De Jong, T.J., Waser, N.M., Klinkhamer, P.G.L. (1993) Geitonogamy: the neglected side of selfing. Trends Ecol Evol, 8, 321-325.

Dudash, M. R., & Fenster, C. B. (1997). Multiyear study of pollen limitation and cost of reproduction in the iteroparous Silene virginica. Ecology, 78(2), 484–493.

Galloway, L.F., Cirigliano, T., Gremski, K. (2002). The contribution of display size and dichogamy to potential geitonogamy in Campanula Americana. International Journal of Plant Sciences, 163(1), 133-139.

Grindeland, J. M., N., S., & Ims, R. A. (2005). Effects of floral display size and plant density on pollinator visitation rate in a natural population of Digitalis purpurea. Functional Ecology, 19(3), 383–390. doi:10.1111/j.1365-2435.2005.00988.x

Haig, D., & Westoby, M. (1988). On limits to seed production. The American Naturalist, 131(5), 757–759.

Hessing, M. B. (1988). Geitonogamous pollination and its consequences in Geranium caespitosum. American Journal of Botany, 75(9), 1324–1333.

Irwin, R. E., & Brody, A. K. (2011). Additive effects of herbivory, nectar robbing and seed predation on male and female fitness estimates of the host plant Ipomopsis aggregata. Oecologia, 166(3), 681–92.

Jersáková, J., Johnson, S. D., & Kindlmann, P. (2006). Mechanisms and evolution of deceptive pollination in orchids. Biological reviews of the Cambridge Philosophical Society, 81(2), 219–35.

JMP, Version 9. SAS Institute Inc., Cary, NC, 1989-2012.

Johnson, S. D., & Bond, W. J. (1997). Evidence for widespread pollen limitation of fruiting success in Cape wildflowers. Oecologia, 109(4), 530–534.

Johnson, S. D., & Nilsson, L. A. (1999). Pollen carryover, geitonogamy, and the evolution of deceptive pollination systems in orchids. Ecology, 80(8), 2607–2619.

Johnson, Steven D, Peter, C. I., & Agren, J. (2004). The effects of nectar addition on pollen removal and geitonogamy in the non-rewarding orchid Anacamptis morio. Proceedings. Biological sciences / The Royal Society, 271(1541), 803–9.

Kolb, A., Ehrlen, J., & Eriksson, O. (2007). Ecological and evolutionary consequences of spatial and temporal variation in pre-dispersal seed predation. Perspectives in Plant Ecology, Evolution and Systematics, 9(2), 79–100.

Kindlmann, P., & Jersáková, J. (2006). Effect of floral display on reproductive success in terrestrial orchids. Folia Geobotanica, 41(1), 47–60.

Knight, T. M. (2003). Floral density, pollen limitation, and reproductive success in Trillium grandiflorum. Oecologia, 137(4), 557–63.

Knight, T. M., Steets, J. A., & Ashman, T.-L. (2006). A quantitative synthesis of pollen supplementation experiments highlights the contribution of resource reallocation to estimates of pollen limitation. American journal of botany, 93(2), 271–7.

Knight, T. M., Steets, J. A., Vamosi, J. C., Mazer, S. J., Burd, M., Campbell, D. R., Dudash, M. R., et al. (2005). Pollen limitation of plant reproduction: Pattern and process. Annu. Rev. Ecol. Syst., 36, 467–497.

Light, M. H. S., & MacConaill, M. C. (2002). Climatic influences on flowering and fruiting of Cypripedium parviflorum var. pubescens. in: Kindlmann P, Willems JH, Whigham DH (eds) Trends and fluctuations and underlying mechanisms in terrestrial orchid populations, Backhuys Publishers, Leiden, The Netherlands, pp. 85-97.

Light, M. H. S., & MacConaill, M. C. (2011). Potential impact of insect herbivores on Orchid conservation. European Journal of Environmental Sciences, 1(2).

Louda, S. M., & Potvin, M. A. (1995). Effect of Inflorescence-Feeding Insects on the Demography and Lifetime of a Native Plant. Ecology, 76(1), 229.

Marquis, R. J.. 1992. Selective impact of herbivores. – In: Fritz, R. S. and Simms, E. L. (eds), Plant resistance to herbivores and pathogens, ecology, evolution and genetics. Univ. Chicago Press, pp. 301–325.

Meléndez-Ackerman, E. J., & Ackerman, J. D. (2001). Density-dependent variation in reproductive success in a terrestrial orchid. Plant Systematics and Evolution, 227(1- 2), 27–36.

Mitchell, R. J., Karron, J. D., Holmquist, K. G., & Bell, J. M. (2004). The influence of Mimulus ringens floral display size on pollinator visitation patterns. Functional Ecology, 18(1), 116–124.

Neiland, M. R. M., & Wilcock, C. C. (1998). Fruit set, nectar reward, and rarity in the Orchidaceae. Am. J. Botany, 85(12), 1657–.

Nilsson, L. A. (1983). Mimesis of bellflower (Campanula) by the red helleborine orchid Cephalanthera rubra. Nature, 305(5937), 799–800.

O’Connell, L. M., & Johnston, M. O. (1998). Male and female pollination success in a deceptive orchid, a selection study. Ecology, 79(4), 1246–1260.

Palo, T. R., Gowda, J., & Högberg, P. (1993). Species height and root symbiosis, two factors influencing antiherbivore defense of woody plants in East African savanna. Oecologia, 93(3), 322–326.

Peakall, R. (1989). The unique pollination of Leporella fimbriata (Orchidaceae): Pollination by pseudocopulating male ants (Myrmecia urens, Formicidae). Plant Systematics and Evolution, 167(3-4), 137–148.

Primack, R. B., & Hall, P. (1990). Costs of reproduction in the Pink Lady’s Slipper Orchid: A four-year experimental study. The American Naturalist, 136(5), 638–656.

Russell, F. L., Rose, K. E., & Louda, S. M. (2010). Seed availability and insect herbivory limit recruitment and adult density of native tall thistle. Ecology, 91(10), 3081– 3093.

Sih, A., & Baltus, M.-S. (1987). Patch size, pollinator behavior, and pollinator limitation in catnip. Ecology, 68(6), 1679–1690.

Snow, A.A., Spira, T.P., Simpson, R., Klips, R.A. (1996). The ecology of geitonogamous pollination. Pages 191–216 in DG Lloyd, SCH Barrett, eds. Floral biology. Chapman & Hall, New York.

Snow, A. A., & Whigham, D. F. (1989). Costs of flower and fruit production in Tipularia discolor (Orchidaceae). Ecology, 70(5), 1286–1293.

Stephens, A. E. A., & Myers, J. H. (2012). Resource concentration by insects and implications for plant populations. Journal of Ecology, 100(4), 923–931.

Stoutamire, W. P. (1967). Flower biology of the Lady’s-Slippers (Orchidaceae: Cypripedium). The Michigan Botanist, 6(4), 159–175.

Strauss, S. Y., & Irwin, R. E. (2004). Ecological and evolutionary consequences of multispecies plant-animal interactions. Annual Review of Ecology, Evolution, and Systematics, 35, 435–466.

Thomson, J. D., & Thomson, B. A. (1989). Dispersal of Erythronium grandiflorum pollen by bumblebees: Implications of gene flow and reproductive success. Evolution, 43(3), 657–661.

Tremblay, R. L., Ackerman, J. D., Zimmerman, J. K., & Calvo, R. N. (2004). Variation in sexual reproduction in orchids and its evolutionary consequences: a spasmodic journey to diversification. Biological Journal of the Linnean Society, 84(1), 1–54.

Walsh, R. P. (2008) Cypripedium (Orchidaceae) Hybridization Along a Prairie/Woodland Ecotone: Evidence for Hybridization using Morphology and Genetics. M.S. Thesis [Bowling Green, (OH)]: Bowling Green State University.

Wesselingh, R. A. (2007). Pollen limitation meets resource allocation: towards a comprehensive methodology. The New Phytologist, 174(1), 26–37.

Tables and Figures

Table 1. MANOVA of the effect of the pollination treatment on percent abortion, percent capsules produced and average seed mass in 2009. Total df = 68.

Source Value F df Prob>F Full Model 0.14 2.77 3 0.0493* Intercept 34.3 653.5 3 <.0001* Treatment 0.14 2.77 3 0.0493*

Table 2. The effect of number of flowers, number of stems and the average height of flowering stems on the initial (A) and final (B) fruit sets.

Source DF L-R χ2 Prob> χ2 A Full Model 4 21.2 0.0003* Treatment 1 17.4 <0.0001* # of Flowers 1 0.074 0.785 # of Stems 1 1.18 0.27 Avg. Height 1 3.49 0.061 B Full Model 4 23.75 <0.0001* Treatment 1 4.33 0.037* # of Flowers 1 0.69 0.40 # of Stems 1 0.02 0.88 Avg. Height 1 4.33 0.03*

Table 3. The effect of treatment, number of flowers, number of stems, number of capsules and the average height of flowering stems on rate of predation across all treatments.

Source DF L-R χ2 Prob> χ2 Full Model 5 16.7 0.005* Treatment 1 0.08 0.77 # capsules 1 1.40 0.23 # of Flowers 1 0.28 0.59 # of Stems 1 0.133 0.71 Avg. Height 1 10.0 0.0016*

Table 4. The effect of hand outcross, hand selfed and open pollination treatments on initial fruit set, final fruit set abortion rate and seed mass for the 2009 study. Total df = 26.

Source Value F df Prob>F Full Model 0.40 3.23 8 <0.0053* Intercept 20.47 117.7 4 <0.0001* Treatment 0.40 3.23 8 <0.0053*

Table 5. The effect of hand outcross, hand selfed and open pollination treatments on initial and final fruit set for the 2011 study.

Source DF SS MS F Ratio Prob>F Initial Fruit Set Treatment 2 3.28 1.64 7.73 0.0008* Error 87 18.5 0.21 C. Total 89 21.78 Final Fruit Set Treatment 2 2.28 1.14 5.73 0.0046* Error 87 17.3 0.19 C. Total 89 19.6

Table 6. The effect of hand outcross, hand selfed and open pollination treatments on seed mass for the 2011 study.

Source DF SS MS F Ratio Prob>F Treatment 2 0.0007 0.0003 18.7 0.0001* Error 26 0.0005 0.00002 C. Total 28 0.0012

Table 7. The effect of hand outcross, hand selfed and open pollination treatments on abortion rates for the 2011 study.

Source DF SS MS F Ratio Prob>F Treatment 2 0.288 0.144 1.79 0.172 Error 87 7.00 0.080 C. Total 89 7.28

1 0.9 0.8 0.7 0.6 0.5 0.4

% Fruit Set 0.3 0.2 0.1 0 Hand Open Hand Open Initial Fruit Set Final Fruit Set

Figure 1. The effect of hand and open pollination treatments on the percent initial and final fruit set with standard error for 2009. Initial: T-test, t = -4.11, n = 36, p>0.001; Final: T-test, t = -4.30, n = 36, p>0.001.

1.2

0.8

0.6

% Fruit Set 0.4

0.2

0 15 20 25 30 35 40 45 Mean Height of Flowering Stems

Figure 2. Percent capsules at maturity for open pollinated plants compared with the mean height of flowering stems in the 2009 study. Linear regression, % Fruit set = -0.344071 + 2 0.0378733*avg. height flowering stems, F1,35 = 7.0, p = 0.012, R = 0.17.

1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2

% Capsules Preyed Upon 0.1 0 15 20 25 30 35 40 Mean Height of Flowering Stems

Figure 3. Percent capsules preyed upon compared with, mean height of flowering stems of all plants in the 2009 study. Linear regression, % capsules preyed upon = 1.5739182 - 2 0.0334288*avg. height flowering stems, F1,61 = 10.12, p = 0.023, R = 0.144.

0.8 A 0.7 0.6 AB C 0.5 C 0.4 0.3 B % Fruit Set 0.2 D 0.1 0 Self Out Open Self Out Open Initial Fruit Set Final Fruit Set

Figure 4. The effects of selfed, outcrossed and open pollination treatments on initial and final fruit set in the 2011 study. N = 30. Treatments separated by a different letter are significantly different. Initial fruit set: F2,89 = 7.73, p<0.001; Final fruit set: F2,89 = 5.73, p = 0.004.

0.035

0.03 B B 0.025

0.02 A 0.015

Seed Mass (g) 0.01

0.005

0 Self Out Open

Figure 5. The effects of open, outcrossed and self-pollination treatments on seed mass (g) per fruit in the 2011 study. N = 29. Treatments separated by a different letter are 2 significantly different. F2,28 = 18.72, p<0.001 R = 0.59.

Chapter III: EXAMINING THE ROLE OF FOOD DECEPTION ON

REPRODUCTION IN THE DECEPTIVE ORCHID CYPRIPEDIUM CANDIDUM

Abstract

Attraction of pollinators in deceit-pollinated species often relies on producing a conspicuous floral display. Cypripedium candidum is a deceptive orchid that provides no nectar reward to floral visitors, but produces a variable number of single-flowered stems, presumably depending on age and available resources. Increasing the number of flowering stems may increase visibility to pollinators, but in-turn may increase within plant selfing. The lack of a nectar reward may discourage pollinators from foraging on multiple flowers on a single plant, while encouraging pollinia dispersal. Understanding the factors limiting pollination and how deceit pollination influences reproductive success will provide important insight into the evolution of deception in flowering plants.

In order to address the role of deception in C. candidum, a nectar addition experiment was conducted from May (flowering) through August (fruit dehiscence) of

2011. Six 100m transects were established across a population at the study site. At 20m intervals along each transect a randomized block design was employed that randomly assigned the four closest plants to the transect point to receive one of four histochemical dyes. Two individuals selected to receive nectar received 2µl of 25% sucrose solution in the labellum of the flower, while the other two individuals received no additional nectar.

Plants were scored for pollen receipt as well as removal every two days throughout the flowering period. Additionally, the density of surrounding nectar producing plants was assessed by measuring the distance to the three nearest nectariferous plants. Number of 39 fruits produced, fruit mass and fruit abortion were scored at the end of the experiment.

The addition of nectar significantly increased (p<0.0001) self-pollination by nearly 3x, while plants not receiving nectar had significantly greater (p<0.0001) numbers of non- self pollinia deposited on the stigmatic surface. The color of dye used in the experiment had no affect on any variables measured. There was a non-significant (p = 0.0645) trend for the non-nectar plants to set more fruit, while presence of nectar did not effect the number of pollinia removed. This study supports the outcrossing hypothesis that states the deceit pollination strategy evolved in order to decrease geitnogamy, by demonstrating increased selfing rates in the presence of a nectar reward and a reproductive cost to selfing.

Introduction

Plants have evolved many strategies to maximize pollen transfer efficiency

(Grindeland et al. 2005) including the evolution of large floral displays. Large floral displays serve to attract more visitors (Mitchell et al. 2004; Kindlmann and Jersáková

2006) by providing a larger visual presence and a greater concentration of potential resources. Increased floral display size can arise through increases in flower size, increases in the number of flowers or synchronous flowering. Increasing the floral display size should increase attractiveness to pollinators and visitation rate (Peakall 1989;

Aragón and Ackerman 2004; Ackerman and Ackerman 2001; Li, Huang et al. 2011).

While producing multiple flowers or inflorescences on a single plant increases visitation, 40 it may simultaneously increase the probability of intra-individual self-pollination, or geitonogomy (Mitchell et al. 2004; Kindlmann and Jersáková 2006; Sun et al. 2009).

Deceptive pollination, specifically food deception, is a pollination strategy in which the flower provides floral cues indicating a food reward while not actually providing that reward (Faegri and va der Pijl 1971; Cozzolino and Widmer 2005). Non- rewarding flowers are found in 146 genera from 33 flowering plant families (Jersáková et al. 2009), but are most prevalent in the Orchidaceae family. Of the approximately 7500 food deceptive plant species, approximately 6500 belong to the family Orchidaceae

(Renner 2005), where nearly one-third of all species employ the food deceptive strategy

(Cozzolino and Widmer 2005). Deceptive pollination has fascinated biologists since its discovery by Sprengel in 1793 and Darwin’s subsequent treatise on Orchid pollination

(1862). While research over the past 200 years has exhaustively identified food deceptive species, the role deception plays in the reproduction of these plants has remained enigmatic (Schiestl 2005). A popular hypothesis for the role of food deception in plant reproduction states that in multi-flowered individuals, food deception serves to encourage pollinators to leave the individual, resulting in decreased self-pollination

(Schiestl 2005; Jersáková et al. 2006).

Orchid species relying on a food deceptive strategy often rely on newly emergent or otherwise inexperienced insects for pollinator services (Jersáková et al. 2006). As discussed by Johnson et al. (2003), the presence of large numbers of rewardless species presents a conundrum in understanding the evolution of orchid pollination. Numerous studies have shown increased pollination and fruiting success in rewarding orchids (see

Johnson et al. 2003). In order to account for reduced visitation and pollination, some 41 have hypothesized that deception promotes outcrossing and reduces geitonogamous pollination because lack of a reward causes pollinators to leave non-rewarding patches

(Schiestl 2005; Dressler 1981), referred to as the outcrossing hypothesis (Jersáková et al.

2006). Additionally, Sun et al. (2009) and Kindlmann & Jersáková (2006) as well as many others (Dressler 1981; Smithson 2002; Johnson, Peter et al. 2004) have postulated that deceptive orchids with multiple stems avoid the problem of increased geitonogomy because bee visitors tend to visit fewer flowers on the same plant when no reward is provided (Nilsson 1980; Nilsson 1984).

Various other hypotheses have been proposed to explain the costs and benefits of the deceptive pollination strategy (Schiestl 2005). Aside from larger floral displays, many deceit-pollinated plants may also benefit from the close proximity of other nectar providing plants (Jersáková et al. 2006, Johnson et al. 2003). In this case, a flower may be discovered via the instinctive foraging behavior of a nectar-seeking insect while avoiding the physiological cost of producing nectar and the potential reproductive cost of geitonogamous pollination. It has also been suggested that deceptive orchid pollination is a type of Batesian mimicry between the deceptive orchid and neighboring nectar producing plants (Gumber and Kunze 2001). Schiestl (2005) further suggests that deception may encourage pollen flow over longer -distances. Working with the food deceptive orchid Anacamptis morio, Johnson and Nilsson (1999) found that nectar supplementation increased average foraging time on an individual plant and self- pollination.

Understanding the role of deception in orchid evolution requires a detailed examination of both male and female function. Geitonogamy serves to reduce female 42 function but also male function by reducing the quantity of pollen available for export to other plants, also known as pollen discounting (Johnson et al. 2004). Peakall (1989) was first to examine the fate of orchid pollinia using histochemically stained pollinia. At this time, the fate of pollen has been directly surveyed in relatively few orchid species (see review in Kropf & Renner 2008). Maximum pollinia transport distances have varied widely from 6.9m in an Andrena pollinated species (Kropf and Renner 2008) to 76m in a hawkmoth-pollinated species (Nilson et al. 1992). Even fewer studies have directly addressed the role of deception through nectar addition studies (Johnson and Nilsson

1999; Smithson and Gigord 2001; Smithson 2002; Johnson et al. 2004; Jersáková and

Johnson 2006). The addition of nectar to non-rewarding species in these studies resulted in increased pollinator foraging times, increased pollinia removal and increased rates of self-pollen deposition, further supporting the outcross hypothesis (Jersáková and Johnson

2006). Previous studies have focused primarily on orchids pollinated by Bombus spp.

(Johnson and Nilsson 1999; Smithson and Grigord 2001; Smithson 2002; Johnson et al.

2004) as well as moths (Johnson and Nilsson 1999) and flies (Jersáková and Johnson

2006). No study at this point has examined the response of small bees and wasps, such as

Andrenidae and Halictidae, to nectar addition. Furthermore, all previous studies have focused on plants within one subfamily, Orchidoideae and only two tribes, Orchidae and

Diseae.

This paper seeks to examine how deception influences reproduction in the orchid

Cypripedium candidum (subfamily Cypripidiodieae). The Cypripedium genus has been proposed as a model system for the study of Orchid reproduction, conservation and the evolution of deception due to its unique trap-like floral design, the presence of deception 43 throughout the entire genus, consistent pollinator limitation, low amount of pollinator fidelity and the abundance of self-incompatible species (Bernhardt and Edens-Meier

2010). Although several studies have examined the role of nectar addition on deceptive orchid species, none have specifically focused on the Cypripedium genus and these unique attributes. This study assessed the role of deception in the pollination of multiple- stemmed flowering perennials using the slipper orchid, Cypripedium candidum

(subfamily Cypripedioideae). All species within the Cypripedium genus are deceit pollinated (Bernhardt and Edens-Meier 2010). Although C. candidum has evolved a non- rewarding flower that should reduce within-flower selfing, the pollinator limitation observed in this species (Walsh Chapter 1) is likely to drive selection towards individuals that produce multiple flowering stems with synchronous flowering. This situation may increase attractiveness to pollinators, while at the same time increasing the probability of between-flower selfing (geitonogomy). We hypothesize pollination by deceit increases fitness by decreasing intra-individual pollen transfer and increasing pollen export. This study specifically addresses the following questions: 1. How does the presence of a nectar reward affect pollinia deposition? 2. How does nectar presence affect pollinia transport distances? 3. Does the presence of nectar increase geitonogamous pollination? and 4.

How do surrounding nectar-producing co-flowering species influence orchid pollination?

We expect the presence of a nectar reward will increase foraging time on a plant, and therefore predict an increase the probability of geitonogomous pollination (Kindlmann and Jersáková 2006). Furthermore, on plants without added nectar, pollen removal and deposition should be dependent on attractiveness to the pollinator and therefore display size. The presence and proximity of near-by nectar producing plants should draw 44 pollinators into the vicinity and thereby increase fruit set in the study orchid (Juillet et al.

2007).

Materials and Methods

Study Species and Site

Cypripedium are deciduous, terrestrial orchids with growth emerging from a subterranean rhizome (Stoutamire 1967). Cypripedium candidum Muhl ex Willd., the

Small White Lady’s Slipper, have yellow-green lateral sepals and petals with a white labellum spotted with purple (Stoutamire 1967), and occur in calcareous prairies as well as fens and limestone barrens (Cusick 1980). C. candidum is highly dependent on full sun in open areas and, as with many prairie species, populations begin to decline with the invasion of woody plants (Curtis 1946). C. candidum as with all Cypripedium spp. do not produce a nectar reward. Pollinators are predominately Andrenidae and Halictidae bees (Bowles 1983).

The study was conducted in May-August 2011 at Resthaven Wildlife Area in

Castallia, Ohio (GPS coordinates available upon request). Resthaven is one of two locations in Ohio where C. candidum occurs and has an estimated 6000 plants in an actively managed prairie area (approximately 900 hectares) with wooded areas intertwined. The prairie is maintained through controlled burns approximately every three years in the early spring (J. Windus, ODNR pers. com.). Resthaven has 16 populations of C. candidum of which we were given permission to work in a total of 4.

This study was carried out in the largest of those populations that contained roughly 250 individuals. 45

To understand the role of deception in C. candidum, we conducted a nectar addition experiment. Six 100m transects were distributed equally across a population at the study site with the initial transect location randomly chosen. At 20m intervals along each transect, four similar-sized plants closest to the transect within a 1 m2 quadrat were selected by finding the plant closest to the center of the quadrat, and then locating three other plants within the quadrat with the same number of flowers and stems. The 20m interval between quadrats was chosen based on previously reported pollination distances of 6-8m in similar pollinator assemblages of andrenids and halictids (Kropf & Renner

2008). At each group of four plants, a randomized block design was employed to randomly assign which of four histochemical dyes (Neutral Red, Gentian Violet, Fast

Green and Methylene Blue, all at 1% solution) were applied to a plant’s pollinia, and two plants were selected to receive nectar or not. A total of 100 plants, 50 plants per nectar treatment, were examined during this experiment, consisting of 4 plants within each quadrat across a total of 25 quadrats. Fifty individuals selected for nectar addition received 2µl of 25% sucrose solution in the labellum of the flower. Nectar was removed and replaced every two days. Following the framework established by Johnson et al.

(2004), 1-2 µl of histochemical dye of one of the four different colors were injected into the pollinia of study orchids (total N = 100) to color code the pollinia. The other 50 individuals received no additional nectar but had stain injected into their pollinia.

Plants were scored every 2 days for pollen receipt or removal from the stigmatic surface throughout the flowering period (15 days). The presence of pollinia containing either non-dyed pollinia or dye of a different color on a stigma was scored as an outcrossing event. Presence of deposited pollinia dyed with the same color initially 46 applied to a plant’s pollinia was scored as a selfing event. Additionally, we assessed the proximity of surrounding nectar-producing co-flowering plants by measuring and identifying the distance to the three nearest nectiferous plants. The plants in flower at the time of the study and included in this analysis were: Maianthemum racemosum, Fragaria vesca, Viola spp., and Hypoxis hirsuta. Voucher specimens of the listed species are deposited in the Bowling Green State University herbarium. Pollinia removal was scored every 2 days. If pollinia had been removed all plants within a 10m radius were searched for the dyed pollinia. The distance to the nearest likely (dyed with the appropriate color) pollen source plant was measured and recorded for all deposited pollinia. Capsule development was recorded 1 month after floral dehiscence (June) as well at maturity in

August when all capsules were collected and scored for insect damage. Fruit abortion was scored as the number of fruits initiated in May minus the number matured in August.

Fruit abortion rate was then calculated by dividing the number of fruits aborted by the total number of fruits produced. Mature capsules were dried at 60°C for three days prior to obtaining the masses for the capsule and seeds. seed mass was weighed to the nearest hundred-thousandth of a gram on a Mettler AE-240 scale (Mettler-Toledo Inc.).

Analyses were performed using JMP for Mac v.9.0.2 (SAS Institute, California,

USA). The dependent variables, receipt of self, out-cross pollinia and total pollinia received, percent fruit set (number of fruits divided by number of flowers), and fruit abortion were checked for normality and the percent data was transformed using an arc- sine square root transformation. Each dependent variable was analyzed using a Standard

Least Squares ANOVA with quadrat, number of stems, number of flowers, treatment and the average distance to the three nearest nectar-producing plants as covariates. The effect 47 of selfed or outcross pollinia, number of flowers and number of stems on the seed mass of plants that set fruit was checked for normality and analyzed using a Standard Least

Squares ANOVA.

Results

Plants included in the study had an average of 3.01 (SE = 0.13) stems and 2.42

(SE = 0.06) flowers per plant. The population density of C. candidum at Resthaven

Wildlife Area (WA) was previously estimated at 3.26 plants per meter2 (range = 1-9 plants/m2; SD = 2.68) (Walsh Ch. 3). Of the total of 484 pollinia that were stained, 103

(21.3 %) were removed during this experiment. Pollinia receipt of plants receiving nectar and those not receiving nectar did not differ (F1,99 = 0.06, p = 0.81; Table 1, Figure 1).

We found a significant effect of nectar addition on the receipt of self pollinia,( F1,99 =

12.57, p = 0.0007; Table 1). Of the plants receiving nectar, 78% of pollinia deposited on the stigmatic surface of a plant’s flowers were selfed pollinia, while plants that did not receive nectar received selfed pollinia only 26% of the time (Figure 2). Additionally, the presence of nectar significantly reduced the probability of a plant receiving outcrossed pollen (F1,99 = 15.6, p = 0.0002,; Table 1). Of the plants receiving nectar, 22% of the pollinia were outcrossed, while the plants not receiving nectar had outcrossed pollen deposited on their pollinia 74% of the time (Figure 3). An increase in number of flowers produced by a plant was also significantly associated with increased total receipt of pollinia (F1,99 = 4.22; p = 0.043) and receipt of self pollinia (F1,99 = 4.17; p = 0.044), but not receipt of outcross pollinia (F1,99 = 0.91; p = 0.34) (Table 1). Color of the dye, 48 transect and quadrat location of the study plants did not significantly affect fruit set or the receipt of pollinia (Appendix A).

Finding the distance travelled for pollinia proved to be difficult as only 8 of the

103 removed pollinia were recovered that were not part of a selfing event. Interestingly of the 8, all originated from non-nectar treated plants, and all were on the stigmatic surface of another plant. Of the recovered pollinia, the average distance travelled was

40.06 cm (SD = 5.8), while the furthest distance was 75.6 cm.

Fruit production was not significantly affected by nectar addition (F1,99 = 2.89; p =

0.09), number of stems (F1,99 = 0.05; p = 0.81), or number of flowers (F1,99 = 3.27; p =

0.07); however, there was a trend for plants with nectar added to produce fewer fruits

(F1,99 = 2.82, p = 0.09; Table 2; Figure 4). A power analysis revealed a required sample size of n = 100 in order to reach the p = 0.05 threshold. Overall, of the 72 flowering stems that matured fruit during the experiment, 46 were from outcross pollinations, while

26 were from self pollinations. Fruit abortion was higher in nectar treated plants (F1,99 =

4.04, p = 0.049; Figure 5) but was not affected by the number of stems (F1,99 = 0.52, p =

0.47), or the number of flowers (F1,99 = 0.0, p = 0.99) (Table 2). The average seed mass for capsules of outcrossed paternity was 0.033g, which was significantly larger (F1,71 =

4.7, p = 0.033; Table 3) than the 0.016 g for selfed capsules (Figure 6).

The proximity of surrounding nectar-producing plants had no effect on the total number of pollinia received, or numbers of self or outcross pollinia received (Table 1) and showed no significant effect on fruit production (Table 2). The mean distance of a nectar producing plant from a flowering C. candidum was 21.8 cm (SE = 1.23).

Discussion

This study demonstrates clear evidence of the outcrossing hypothesis, which states that deceptive pollination evolved primarily to increase outcrossing while decreasing geitonogomous pollination (Schiestl 2005; Jersáková et al. 2006). In support of this hypothesis we observed a natural rate of self pollination of 26% compared to a selfing rate of 78% in the supplemental nectar treatment. Reproduction in non-rewarding orchids is often significantly reduced when compared to their rewarding counterparts

(Calvo 1993; Tremblay et al. 2005; Jersáková et al. 2006), suggesting there must other reproductive factors, such as increased offspring fitness, that maintain the deceptive strategy. We did not see an effect of treatment on fruit set between the two treatments although we did find a non-significant trend towards increased fruit set in plants that received nectar.

Our data showed no difference in the rate of pollinia receipt between plants supplemented with nectar and those that were not. This result is contrary to most deception literature that shows increases in pollination when nectar is present (Johnson et al. 2004; Tremblay et al. 2005). However, concurrent research (Walsh Chapter 1) showed a strong effect of flower height on pollination rates, which may have had an over- riding influence and may explain why nectar addition had no overall effect on increasing pollination rates. Other research in the closely related Cypripedium acaule (O’Connell and Johnston 1998) saw similar effects of floral height. Although this study found no significant effect of number of flowers and number of stems on proportion of fruits produced, we did find increased total pollinia receipt as well as self pollinia receipt with 50 increased flower number. Furthermore, in such a relatively dense patch of orchids (3.26 plants/m2), it is possible many of the pollinators previously have learned to avoid C. candidum, reducing any effects of nectar addition (Nillson 1980; Fritz 1990; Smithson and MacNair 1997; Ferdy et al. 1998) .

Pollen discounting, or the reduction in pollen available for outcross as a result of selfing events (Holsinger and Thomson 1994) may play an important role in explaining the advantage of deceptive pollination. In our study, we observed significant pollen discounting within the nectar addition treatment. Of the flowers in the nectar addition treatment that received pollinia, 74% of the pollinia came from the same plant. Not only does this dramatically increase the chances of inbreeding depression, but it potentially reduces male fitness by decreasing the availability of pollinia that can outcross. Similar results were seen in the deceptive orchid Anacamptis morio where Johnson et al. (2004) estimated that nectar production would result in 39.5% of the removed pollen being deposited on a self-stigma but only 8.6% if nectar were not provided. In the case of orchids, specifically Cypripedium candidum, each flower only produces two pollinia, potentially exacerbating pollen discounting further.

Self-pollination often leads to decreased fitness through reduced seed set or seed quality (Tremblay 2005) or even fruit abortion. We saw no impact of nectar addition on fruit production, but did see an increase in fruit abortion when plants were supplemented with nectar indicating potential inbreeding depression. Other studies have shown similar trends in fruit set with nectar addition in deceptive plants. Johnson et al. (1999) noted no increase in fruit set when nectar was added to Orchis morio and Platanthera chlorantha as did Ackerman (1981) in Calypso bulbosa, Smithson and Grigord (2001) in Barlia 51 robertiana and Smithson (2002) in Anacampstis morio. This trend has also been noted outside the Orchidaceae (Burd 1995; López-Portillo et al. 1993). We also saw a strong negative effect of selfing on seed mass, supporting data previously obtained in the same system through a study examining the effects of hand selfing, hand outcross and natural pollination treatments on fruit set and seed mass (Walsh Chapter 1). We observed significant ovule discounting, or reduction in number of ovules available for outcrossing due to fertilization from self pollen (Barrett 2002; Verdú et al. 2006). In this study, plants receiving selfed pollen had an average of 50% lower seed mass than those receiving outcrossed pollen. Similar results in seed mass reduction and increased fruit abortion were noted in the deceptive orchid Disa pulchra when nectar was added (Jersáková and

Johnson 2006). A 50% reduction in seed mass and increase in fruit abortion, multiplied over the long life of a perennial plant such as C. candidum would potentially translate to a significant reduction in life-time fitness of the plant.

Although we had a very small number of recovered pollinia, most were found at distances less than 1m, suggesting the small body size of the pollinators limits the potential pollinia dispersal distance. Previous studies have found pollinia transport distances to vary widely with pollinator species (Nilson et al. 1992; Kropf and Renner

2008) and most have focused on larger pollinators. Kropf and Renner (2008) found a maximum transport distance of 6.9m for relatively large Andrena bees averaging 8-17 mm long. In contrast pollinators caught during preliminary trials in our field site averaged 4-6 mm long. We suspect pollinia and pollinator size are the primary contributing factors to pollinia dispersal distance and most likely differ between orchid species and locations. 52

Pollination by deceit has been an area of intense interest to evolutionary biologists, and plants in the Orchidaceae have often been a focus in this research

(Cozzolino and Widmer 2005), with nearly 1/3 of all orchids employing some type of deceptive strategy (Dafni and Bernhardt 1990). Our study strongly supports the outcrossing hypothesis in that the addition of nectar severely reduced overall plant fitness. We saw virtually no increase in pollination with the addition of nectar.

Furthermore, we observed nearly a 50% reduction in seed mass when plants were selfed, and more than double the number of fruit abortions with the nectar treatment. The presence of nectar in Cypripedium candidum led to strong ovule discounting without increasing fruit production. Based on this study, and our previous work (Walsh Chapter

1) we can conclude that although the deceptive pollination strategy may entail a certain amount of pollen limitation, the resulting increase in female fitness far outweighs the potential reduction in fruit set over the long life of the plant. Our data shows that when nectar production is present in the C. candidum system, plants produce half as many seeds on average compared to the deceptive system. This does not account for reductions in seed viability and offspring fitness that may come with increased self-pollination and inbreeding depression (Borba et al. 2001; Smithson 2006; Ferdy et al. 2011). Orchids that are subject to selfing events have been shown to exhibit inbreeding depression

(Smithson 2006; Borba et al. 2001), especially in deceptive orchids that have evolved with very low levels of selfing (Ferdy et al. 2001; Jersáková et al. 2006).

It seems at first counterintuitive that nectar deception, which leads in most cases to lower fruit production, would be maintained so consistently within the Orchidaceae family (Gill 1989). Our study demonstrates a strong cost to both male and female fitness 53 that has no doubt been a driving factor to the maintenance of food deception in

Cypripedium. Johnson et al. (2004) noted that it was difficult to understand how deceit was maintained in plants with low pollen receipt (<10%), however our data suggests that nectar addition does not affect total pollinia receipt or fruit set, thereby making the increase in offspring quality the driving factor in the maintenance of deceit pollination.

All studies on deceit in Orchidaceae have examined the effect over a narrow period of time, not completely considering the effects of this strategy over the considerable lifetime of most orchid species. Further studies incorporating what is known about deceit pollination, along with other variables affecting reproduction such as seed predation and dormancy over longer periods of time are needed in order to account for the costs and benefits of deception over the life of individuals.

Literature Cited

Barrett, S. C. H. (2002). Sexual interference of the floral kind. Heredity, 88(2), 154–9.

Bernhardt, P., & Edens-Meier, R. (2010). What We Think We Know vs. What We Need to Know About Orchid Pollination and Conservation: Cypripedium L. as a Model Lineage. The Botanical Review, 76(2), 204–219.

Borba, E., Semir, J., Shepherd, G. (2001). Self-incompatibility, Inbreeding Depression and Crossing Potential in Five Brazilian Pleurothallis (Orchidaceae) Species. Annals of Botany, 88(1), 89–99.

Bowles, M. L. (1983). The tallgrass prairie Orchids Platanthera leucopaea (Nutt.) Lindl and Cypripedium candidum Muhl. ex Willd.: Some aspects of their status, Biology and Ecology, and implications toward management. Natural Areas Journal, 3(4), 14–37.

Cozzolino, S., & Widmer, A. (2005). Orchid diversity: an evolutionary consequence of deception? Trends in ecology & evolution, 20(9), 487–94.

Darwin, C. (1862). On the various contrivances by which British and foreign orchids are fertilized by insects. London: John Murray.

Faegri, K., & Van der Pijl, L. (1971). The Principles of Pollination Ecology. Pergamon.

Ferdy, J. B., Gouyon, P. H., Moret, J., & Godelle, B. (1998). Pollinator behavior and deceptive pollination: learning process and floral evolution. The American naturalist, 152(5), 696–705.

Ferdy, J.-B., Loriot, S., Sandmeier, M., Lefranc, M., & Raquin, C. (2001). Inbreeding depression in a rare deceptive orchid. Canadian Journal of Botany, 79(10),1181- 1188.

Jersáková, J., & Johnson, S. D. (2006). Lack of floral nectar reduces self-pollination in a fly-pollinated orchid. Oecologia, 147(1), 60–8.

Jersáková, J., Johnson, S. D., & Jurgens, A. (2009). Plant-Environment Interactions- Signaling and Communication in Plants. (F. Baluska, Ed.)Signaling and Communication in Plants (pp. 223–246). Berlin, Heidelberg: Springer Berlin Heidelberg. 55

Jersáková, J., Johnson, S. D., & Kindlmann, P. (2006). Mechanisms and evolution of deceptive pollination in orchids. Biological reviews of the Cambridge Philosophical Society, 81(2), 219–35.

Johnson, S. D., & Nilsson, L. A. (1999). Pollen carryover, geitonogamy, and the evolution of deceptive pollination systems in orchids. Ecology, 80(8), 2607–2619.

Juillet, N., Gonzalez, M. A., Page, P. A., & Gigord, L. D. . (2007). Pollination of the European food-deceptive Traunsteinera globosa (Orchidaceae): the importance of nectar-producing neighbouring plants. Plant Systematics and Evolution, 265(1-2), 123–129.

Kindlmann, P., & Jersáková, J. (2006). Effect of floral display on reproductive success in terrestrial orchids. Folia Geobotanica, 41(1), 47–60.

Kropf, M., & Renner, S. S. (2005). Pollination success in monochromic yellow populations of the rewardless orchid Dactylorhiza sambucina. Plant Systematics and Evolution, 254(3-4), 185–197.

Kropf, Matthias, & Renner, S. S. (2008). Pollinator-mediated selfing in two deceptive orchids and a review of pollinium tracking studies addressing geitonogamy. Oecologia, 155(3), 497–508.

Miller, T. E. X., Tenhumberg, B., & Louda, S. M. (2008). Herbivore-mediated ecological costs of reproduction shape the life history of an iteroparous plant. The American naturalist, 171(2), 141–9.

Mitchell, R. J., Karron, J. D., Holmquist, K. G., & Bell, J. M. (2004). The influence of Mimulus ringens floral display size on pollinator visitation patterns. Functional Ecology, 18(1), 116–124.

Nilsson, L. A., Rabakonandrianina, E., & Pettersson, B. (1992). Exact tracking of pollen transfer and mating in plants. Nature, 360(6405), 666–668.

O’Connell, L. M., & Johnston, M. O. (1998). Male and female pollination success in a deceptive orchid, a selection study. Ecology, 79(4), 1246–1260.

Schiestl, F. P. (2005). On the success of a swindle: pollination by deception in orchids. Die Naturwissenschaften, 92(6), 255–64. 56

Smithson, A, & Gigord, L. D. (2001). Are there fitness advantages in being a rewardless orchid? Reward supplementation experiments with Barlia robertiana. Proceedings. Biological sciences / The Royal Society, 268(1475), 1435–41.

Smithson, Ann. (2002). The consequences of rewardlessness in orchids: reward- supplementation experiments with Anacamptis morio (Orchidaceae). American journal of botany, 89(10), 1579–87.

Smithson, A. (2006). Pollinator limitation and inbreeding depression in orchid species with and without nectar rewards. The New phytologist, 169(2), 419–30.

Verdú, M., González-Martínez, S. C., Montilla, A. I., Mateu, I., & Pannell, J. R. (2006). Ovule discounting in an outcrossing, cryptically dioecious tree. Evolution, 60(10), 2056-2063.

Tables and Figures

Table 1. The effect of nectar addition, quadrat, number of stems, number of flowers and the distance of the three nearest non-orchid, nectar-producing neighbors on the percentage of total pollinia (R2 = 0.31), percent self pollinia (R2 = 0.41), and percent outcross pollinia received (R2 = 0.38). Percentage pollinia received calculated as the number of pollinia received divided by the number of flowers on a given plant.

Source DF Sum of Mean F Prob>F

Squares Square Ratio Nectar 1 0.01 0.06 0.81 Quadrat 24 5.31 1.16 0.3 # of Stems 1 0.003 0.01 0.89 % Total # of Flowers 1 0.8 4.22 0.043* Pollinia 3 Nearest 1 0.0005 0.002 0.95 Received Neighbors Model 28 6.3 0.22 1.18 0.27 Error 71 13.48 0.189 Total 99 19.79 Nectar 1 1.13 12.57 0.0007* Quadrat 24 3.14 1.45 0.11 # of Stems 1 0.001 0.01 0.9 # of Flowers 1 0.37 4.17 0.044* % Self Pollinia 3 Nearest 1 0.003 0.03 0.85 Received Neighbors Model 28 4.61 0.16 1.82 0.021* Error 71 6.39 0.09 Total 99 11.01 Nectar 1 1.35 15.6 0.0002* Quadrat 24 2.29 1.1 0.36 # of Stems 1 0.0004 0.005 0.94 % Outcross # of Flowers 1 0.079 0.91 0.34 Pollinia 3 Nearest 1 0.006 0.07 0.78 Received Neighbors Model 28 3.91 0.13 1.61 0.05* Error 71 6.15 0.08 Total 99 10.07

Table 2. The effect of nectar addition, quadrat, number of stems, number of flowers and distance of the three nearest non-orchid, nectar-producing neighbors on percent fruit set (R2 = 0.34), and percent fruit abortion (R2 = 0.42).

Source DF Sum of Mean F Prob>

Squares Square Ratio F Nectar 1 0.33 2.89 0.09 Quadrat 24 3.61 1.29 0.19 # of Stems 1 0.006 0.05 0.81 # of Flowers 1 0.37 3.27 0.07 % Fruit Set 3 Nearest 1 0.00002 0.0002 0.98 Neighbors Model 28 4.26 0.15 1.31 0.17 Error 71 8.23 0.11 Total 99 12.5 Nectar 1 0.77 4.04 0.049* # of Stems 1 0.10 0.52 0.47 # of Flowers 1 0 0 0.99 % Fruit Abortion Model 3 0.85 0.28 1.48 0.23 Error 48 9.22 0.19 Total 51 10.07

Table 3. The effect of pollinia source, number of stems and number of flowers on the seed mass of a capsule, R2 = 0.067.

Source DF Sum of Mean Square F Ratio Prob>F Squares Pollinia Source 1 0.004 4.09 0.47* # of Stems 1 0.0001 0.17 0.67 # of Flowers 1 8.82E-09 0 0.99 Model 3 0.004 0.001 1.63 0.18 Error 68 0.068 0.001 Total 71 0.07

0.6

0.5

0.4

0.3

0.2 % Pollinia Received 0.1

0 No Nectar Nectar

Figure 1. The effect of the nectar treatment on the proportion of pollinia received blocked by quadrat. F1,99 = 0.04, p = 0.83.

1 0.9 0.8 0.7 0.6 0.5

Pollen 0.4 0.3 0.2

% Flowers Receiving Self 0.1 0 nectar non-nectar

Figure 2. The effect of the nectar treatment on the percent of flowers that received self pollinia, blocked by quadrat. F1,99 = 15.76, p = 0.0002 .

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 % Outcrossed Pollina Received 0 nectar non-nectar

Figure 3. The effect of the nectar treatment on the percentage of flowers receiving outcross pollinia, blocked by quadrat. F1,99 = 18.12, p > 0.0001.

0.45

0.4

0.35

0.3

0.25

0.2 % Fruit Set 0.15

0.1

0.05

0 No Nectar Nectar

Figure 4. The effect of nectar treatment on the percentage of flowers setting fruit, blocked by quadrat. F1,99 = 2.82, p = 0.09. Mean (No Nectar = 0.36, Nectar = 0.25; SE

0.048)

0.6

0.5

0.4

0.3

0.2 % Fruit Abortion

0.1

0 No Nectar Nectar

Figure 5. The effect of nectar addition on the percentage of fruits aborted, blocked by quadrat. F1,99 = 4.58, p = 0.035. Mean (No Nectar = 0.20, Nectar = 0.42; SE = 0.07)

0.04

0.035

0.03

0.025

0.02

0.015 Seed Mass (g) 0.01

0.005

0 self out

Figure 6. The effect of self and outcross pollinia on the seed mass of the resulting capsules. F1,71 = 4.7, p = 0.033. Mean self = 0.016g, SE = 0.0007; Mean outcross =

0.033g, SE = 0.005

Appendix

Appendix A. Standard Least Squares analysis of the effect of transect, quadrat, and plant number on percent fruit set.

Source df SS MS F P

Model 28 4.34 0.15 1.35 0.15

Error 71 8.15 0.11

C. Total 99 12.5

Effect Tests

Source df SS F p

Transect/Quadrat 24 3.27 1.18 0.28

Color 3 0.84 2.45 0.07

Tag Number 1 0.03 0.31 0.57

CHAPTER IV: THE DEMOGRAPHIC IMPLICATIONS OF SEED PREDATION

AND DECEPTION IN THE REWARDLESS ORCHID, CYPRIPEDIUM

CANDIDUM MUHL EX. WILLD.

Abstract

Attempts to conserve threatened and endangered populations are often hindered by a lack of available data. Many long-lived clonal plants, and specifically terrestrial orchids, have faced decades of population decline and habitat loss, yet very little is known about their complete life history. Furthermore, few studies have examined the role of seed predation and food deceptive pollination in a large, thriving population.

Cypripedium candidum is a non-rewarding orchid that provides no nectar reward to floral visitors. To examine the demography of this threatened terrestrial orchid, in 2009 we established two-100m transects that were sampled at 5m intervals in a large, actively managed population at Resthaven Wildlife Area in Castalia, Ohio. At each point a 1m x

1m quadrat was used to select plants for monitoring. Demographic characteristics of each plant (number of stems, number of flowering stems, presence/absence in years 2, 3, and four, pollinia receipt, numbers of fruit, seeds produced and capsules preyed upon) were recorded throughout this four-year study. Each plant was also classified into a stage, based on size and reproduction (juvenile, 1 flowered individuals and multiple flowered individuals). Demographic events and states (seedling emergence, size class, reproductive state, and dormancy) were recorded. Demographic models examined the effects of seed predation, simulated selfing from nectar addition and a combination of the two on the population growth rate. The demographic data was used to build a matrix 66 model of the population to determine the important demographic stages and transitions for C. candidum.

Reproductive effort varied from year to year, with an average change of +/- 1.6 flowers year to year. Approximately 5% of individuals in the study were observed dead or dormant in the following year. An average of 51% (SD = 0.44, range = 20 - 65%) of flowers set fruit over the study; however, there was significant variation in yearly capsule set. Seedling recruitment was low, with an average of 0.1 seedling per m2 over the study. Although variable across years, the number of stems and flowers produced in a given year significantly predicted the number of flowers produced in the following year.

Furthermore, increased fruit set in year t was associated with increased dormancy in the following year. Population growth rate (λ) was 1.01 without seed predation or nectar addition. Fecundity reductions due to increased selfing with nectar addition and seed predation each reduced λ by 0.01, while the combination of both factors reduced λ below

1 to 0.99. This study provides evidence that the presence of deceptive pollination and the resulting increased fecundity is important in population maintenance and expansion of deceptive plants when multiple variables that affect fitness such as seed predation and inbreeding depression are considered.

Introduction

In order to successfully reproduce, many flowering plants rely on animal vectors to transfer pollen from one individual to the other. Plants have evolved numerous mechanisms to facilitate this transfer (Peakall 1989; Aragón and Ackerman 2004; Barrett

2002; Grindeland et al. 2005; Li et al 2011). Despite the large variety of adaptations seen 67 in angiosperms, successful pollination is still a highly stochastic process (Price et al.

2005; 2008). The limitation placed on reproduction by the stochastic nature of pollen receipt is referred to in the literature as pollen limitation. Due to pollen limitation’s role in reducing plant fecundity and population survival, numerous studies and literature reviews have documented pollen limitation in natural plant populations (e.g. Aizen et al.

2007; Brys et al. 2008; Cariveau et al. 2004; Harder and Aizen 2010; and Knight 2004).

Constraints on reproduction such as pollen limitation, and the resulting decreased fecundity of an individual would be expected to have long-term consequences on population dynamics. However, the consequences of pollination limitation for demography and life history have to be examined in light of other selective forces that influence reproductive allocation, survival and defense.

The post-pollination fate of ovules plays an equally important role in plant reproduction and population dynamics. Pre-dispersal seed predators are prevalent in flowering plant populations and are often highly influential in individual and population fitness (Cariveau et al. 2004; Crawley 1992; Louda and Potvin 1995). Damage to developing seeds not only results in lost fitness to the individual, but also represents a large amount of lost resources that were allocated to producing the flowers, fruits and ovules. Even in the absence of predators, seeds must be deposited in an appropriate area to assure germination and growth of the new individual. Perhaps no other plant family demonstrates the need for perfect seed germination conditions as does the family

Orchidaceae. Orchid seeds lack nutritive endosperm and therefore rely on symbiosis with mycorrhizal fungi to provide nutrition. This additional challenge in orchid 68 recruitment creates an even more unpredictable recruitment environment (Jacquemyn et al. 2007).

Adult whole-plant dormancy, in which entire individuals fail to produce above ground photosynthetic and reproductive structures through one or multiple growing seasons, is common throughout many long-lived terrestrial plants (Lesica and Steele

1994; Shefferson et al. 2005). During the dormant period, plants sustain themselves through built up reserves although the actual mechanisms by which this occur are unclear

(Wells 1981; Kull 2002). Dormancy has been studied extensively in the Orchid genus

Cypripedium (Primack and Stacy 1998; Shefferson 2001, 2003, 2006, 2009; Kery and

Gregg 2004) in which prolonged dormancy is relatively common (Kull 2002). Frequency of dormancy has been shown to be related to climactic factors (Shefferson et al. 2001;

Kery and Gregg 2004), or occurring in response to suboptimal early season growth conditions (Reintal et al. 2010). Furthermore, it has been suggested that high rates of fruiting in a previous year may increase the probability of dormancy in the following year

(Primack and Stacy 1998; Lesica and Crone 2007). Although some have observed a reduction in size following a dormant period (Hutchings 1987; Willems and Melser 1998,

Shefferson et al. 2003; Shefferson and Tali 2007), others have found no such effect, with the dormant individual breaking dormancy and returning at a size equal to what it was prior to the dormant period (Lesica and Crone 2007; Jäkäläniemi et al 2011).

Floral adaptations such as food deceptive pollination can influence the reproductive capabilities of a plant. Deceptive pollination can be defined as the presence of floral cues indicating a reward while failing to provide that reward (Faegri and va der

Pijl 1971; Dafni 1987; Cozzolino and Widmer 2005). Non-rewarding flowers are found 69 in 146 genera from 33 flowering plant families (Jersáková et al. 2009), but are most prevalent in the Orchidaceae family where nearly one-third of all species employ the food deceptive strategy (Cozzolino and Widmer 2005). It has been hypothesized that the deceptive pollination strategy reduces geitonogamous pollination, or intra-individual self- pollination, by discouraging pollinators from visiting multiple flowers on the same plant

(Jersáková et al. 2006; Schiestl 2005; Dressler 1981). However, this strategy has been shown to increase pollinator limitation resulting in decreased fruiting success when compared to other closely related nectar providing plants (see Johnson et al. 2003).

Demographic models are often utilized to understand how pre and post- pollination factors affect population dynamics. Demographic modeling is a technique often used to examine population level variation over space and time (Kéry and Gregg

2004). To analyze demographic data, matrix models are often employed, (Lefkovitch

1965; Caswell 1989), which organize the data into age, size or stage specific rates and incorporate survival, growth and reproduction into linear equations (Rooney and Gross

2003; Jacquemyn et al. 2007) to calculate the growth rate of the population (λ). In addition to calculating population growth rate, perturbation analyses may be used to determine the stage transitions that have the greatest effect on population growth rate

(Caswell 2000). Population demographic models can prove invaluable in understanding long term dynamics of endangered or threatened populations through the use of

Population Viability Analyses (PVA) that quantitatively assess the probability of extinction (Boyce 1992; Brook et al. 2000; Menges 2005; Nicolè et al. 2005). Despite the usefulness of demographic models in understanding the consequences of evolutionary and ecological factors, there are no existing models examining the role of food deception 70 on population dynamics. Food deception is expected to reduce within-plant selfing

(geitonogamy) at the expense of increased pollen limitation (Dafni 1987; Cozzolino and

Widmer 2005) but the population level impacts of this expected tradeoff between increased offspring fitness alongside potential fecundity reductions have not been examined for food deceptive pollination systems. Furthermore, most studies examining the demographics of Cypripedium spp. focus on small, at risk populations (Shefferson

2001, 2003, 2006, 2009; Kery and Gregg 2004; Nicolè et al. 2005), but very few have looked at demographic parameters within a large, thriving population. Previous work

(Walsh Chapter 3) has demonstrated increased selfing in the presence of a nectar reward with no reduction in pollination.

The goal of this study was to examine the impacts of seed predation and deceit pollination on the population growth rate, elasticities and stage distribution of C. candidum. Specifically, we first constructed a baseline model that examined the deterministic population growth rate (λd) with the weighted average of seed predation across four years in order to understand the population dynamics of C. candidum within a large population. We then constructed a model that removed all seed predation from the population to assess the demographic costs of predation by comparing fecundities to a situation where predation is excluded. Our high seed predation estimate came from the

2009 season during which the prairie population was managed through a controlled burn.

The high seed predation recorded that year is within the range of seed predation estimates found by Light and MacConaill (2002) in the closely related C. parviflorum var. pubescens. To assess the demographic advantages of deceit pollination on the population under realistic population conditions, we constructed a model using fecundity values 71 obtained in Walsh (Chapter 3) reflecting reproductive rates in the presence of a nectar reward with the weighted average seed predation levels in the baseline model. A previous study found nectar addition resulted in a 22% increase in selfing (Walsh Chapter

3) and selfing reducing seed mass by 48% (Walsh Chapter 2). Finally, we combined the nectar reward data along with the average seed predation to construct a model reflecting a scenario in which C. candidum produced a nectar reward and was subjected to the average seed predation rate.

Materials and Methods

Study System

Cypripedium plants are deciduous, terrestrial orchids with growth emerging from a subterranean rhizome (Stoutamire 1967). Cypripedium candidum Muhl ex Willd., the

Small White Lady’s Slipper, occurs in calcareous prairies as well as fens and limestone barrens (Cusick 1980). C. candidum is highly dependent on full sun in open areas and, as with many prairie species, populations begin to decline with the invasion of woody plants

(Curtis 1946). C. candidum, like many other slipper orchids, employs a deceptive pollination strategy that utilizes floral cues to attract pollinators while providing no nectar reward in return for visitation.

The weevil, Stethobaris ovata is a known seed predator of Cypripedium spp. and other temperate orchid genera, with reports of adult weevils feeding on emerging shoots and flower buds (Light and MacConaill 2011). Adult weevils emerge in early spring along with Cypripedium shoots, and oviposit in developing fruits and possibly stems, resulting in either fruit abortion or near total loss of the developing embryos (Light and 72

MacConaill 2011). Predation rates on C. parviflorum have varied from 32%-53% depending on climate and availability of fruit resources (Light and MacConaill 2002).

Little is known about the life history of the organism although it is suspected they may complete two life cycles within a growing season (pers. comm. M. Light).

The primary field site for this study was located at Resthaven Wildlife Area in

Castallia, Ohio (GPS coordinates available upon request). Resthaven is one of only two locations in Ohio where C. candidum remains (Pers. Comm. ODNR). Resthaven has a large, actively managed prairie area (approximately 900 hectares) with wooded areas intertwined. The prairie is maintained through controlled burns approximately every three years in the early spring (J. Windus, ODNR pers. com.). Large, thriving populations of C. candidum (total N ~ 6000, measured in 2001) exist in the calcareous soil and actively burned regions of the prairie. Dominant vegetation in the prairie includes Andropogon gerardii, Sorghastrum nutans, and Silphium terebinthinaceum.

Demographic Monitoring

To examine the demography of C. candidum, in 2009 we began a four-year study yielding three transition matrices that were used to calculate a weighted average transition matrix representing the study period. We established three-60m transects that were sampled at 5m intervals, where a focal plant was selected and a 1m x 1m grid was used to monitor the surrounding area of the focal plant for seedling emergence. All plants within the grid were marked with numbered metal tags to facilitate identification in subsequent years. Demographic characteristics of each focal plant (number of stems, number of flowers per plant, presence/absence in years 2, 3 and 4, number of seed 73 capsules and number of seeds produced) were recorded at the time of flowering (May) and at approximate fruit maturity (August). Multiple stems within 15cm of each other were considered a single individual (Shefferson 2006). In subsequent years new seedlings within the grid were recorded and marked to empirically sample recruitment rates indirectly within the population. Seed packets were constructed using glassless slide mounts and 35 µm plankton netting (Rasmussen and Whigham 1993). One hundred seeds were placed into each seed packet and four seed packets were sown 3cm deep at the corner of every three sampling locations (total seed packets = 96). Seed packets showed no germination over the three-year period. However estimates of seed survival could be obtained by measuring the number of seeds that were destroyed within the seed packets during a given year by retrieving the seed packet, washing with distilled water and examining the seed under a microscope. Seed packets were returned to their marked locations following examination. Seed was considered non-viable when either completely missing or when the embryo was no longer present. Over the course of the study, an average of 53% of the seeds within the seed packet were destroyed in a given year, leading us to assume 47% remain viable in the seed bank from year to year.

Germination and protocorm transition rates were estimated using the average number of emergent seedlings found during a given year divided by the number of seeds produced.

This rate was distributed across a four-year period (one year to germinate and three years as a protocorm) to estimate the probability of a seed becoming a juvenile plant. Data obtained was compared to previous studies on orchid seed germination (Rasmussen and

Whigham 1993) and Cypripedium reproduction to confirm they are realistic (Shefferson et al. 2001; Kery and Gregg 2004). Due to the short duration of this experiment in 74 relation to the long life-span of the study species and known extended dormancy in the genus (Shefferson et al. 2001; Kery and Gregg 2004), we conducted all simulations twice, once based on the assumption that all plants failing to emerge in a given year were dormant and a second time assuming all plants not emerging were dead. This approach allows us to simultaneously examine the effect of dormancy on population growth rate and establish realistic parameters for the model.

Stages and Parameter Estimations of the Model

Each plant was classified into a stage based on size, age and reproduction. Stages were selected to be biologically meaningful and to limit variability within stages.

Additionally, all above ground stages have an accompanying dormant stage. The life cycle (Figure 1) is composed of nine stages: seeds, two protocorm stages, juveniles, dormant juveniles, 1-flowered individuals, dormant 1-flowered individuals, 2+-flowered individuals, and dormant 2+-flowered individuals. To construct biologically meaningful adult stage classes we used a generalized linear model to examine the relationship between percent capsules (# of capsules/# of flowers) and the number of flowering stems

(Figure 2) in JMP v. 9.0.1 (SAS Institute, U.S.A.). We used a post hoc contrast analysis to compare differences in reproduction between 1 and multiple flowered individuals.

Based on this analysis, one-flowered individuals produced significantly higher fruit set than two-flowered individuals, and were therefore classified separately in our stage classifications.

The vital rate for seed production per plant was calculated by determining the mass of 100 seeds to five decimal places using a Mettler AE240 scale (Mettler-Toledo 75

Inc., USA) and estimating the number of seeds in an average capsule from that mass.

Our calculations indicated an average C. candidum seed capsule from an outcross event contains approximately 10,239 seeds (n = 30 from 30 separate randomly chosen capsules; mean = 10,239±451; range = 9788-10,690), an average that falls within other estimations of Cypripedium spp. seed counts in the literature (Kull 1999; Nicolé et al. 2005). Total fecundity for each stage was then estimated by multiplying the estimate of total number of seeds produced in a single capsule by the weighted average of number of capsules produced in that stage.

Model Analyses

The data collected was used to build a matrix model of the stage structure of the population (García, Goñi et al. 2009). Model simulations were performed in RAMAS

Metapop (v. 5, Applied Biosciences, NY, USA) in a deterministic model that assumes consistent vital rates and populations at equilibrium. Due to the small number of transition periods along with the long-life span of the study species, a deterministic model was deemed the most conservative. The model was calculated using no demographic or environmental stochasticity, and a starting population size of n = 72. Perturbation analyses were performed with Metapop (v. 5, Applied Biosciences, NY, USA) in order to estimate sensitivities and elasticities amongst size classes.

To examine the effects of seed predation, pollen limitation and nectar production on population dynamics, four additional analyses were performed where we modified the fecundity rates of reproductive size classes. Fecundity estimates (Table 1) were obtained from previous research on the effects of pollen limitation and seed predation (Walsh Ch. 76

2) and the effects of nectar addition (Walsh Ch. 3) on fitness in C. candidum. The

Baseline Model uses the weighted average transition matrix in which fecundity includes the effects of observed 22% seed predation (averaged over the four years of the study).

The No Seed Predation model was calculated by removing the effect of 22% seed predation on the fecundity estimate to specifically examine the effect of seed predation on population dynamics. The High Seed Predation Model uses fecundity estimates under a 42% reduction in seed production as measured in Walsh (Ch. 1), which was obtained during the 2009 burn year. This model provides insight into the effects of high seed predation rates similar to the levels seen by Light and MacConaill (2002), as well as potential repercussions from management. To assess the demographic impacts of the deceptive pollination strategy, the Nectar Addition Model incorporates data showing a

48% reduction in seed mass when fruits were selfed, which affects 78% of all fruits produced by selfing when nectar is supplied (Walsh Ch. 2). The Nectar + Seed Predation

Model combines the 22% reduction in seed production due to seed predation along with fecundity changes from selfing if nectar was provided to simulate this alternative scenario and understand the relationship between deceptive pollination and other post-pollination factors that limit reproduction.

Factors Affecting Future-Year Size and Dormancy

We analyzed the effects of a previous year’s production (number of stems, number of flowers and % capsules) on number of flowers in the following year using

Generalized Linear Models (GLM’s). Multivariate logistic regressions were used to examine the relationship between previous year growth and dormancy. Comparisons 77 between years of the number of stems, number of flowers and number of capsules produced were analyzed using an ANOVA with a Tukey-Kramer HSD to compare means between groups. Count data was log transformed while percent capsules data was arc- sine square root transformed to normalize the data. Date was analyzed using JMP v.9.0.2

(SAS Institute, California, U.S.A.).

Results

The first year of the study, 2009, was the only year that followed a controlled burn and, as might be expected, had the highest average number of stems, flowers, fruits, but also had highest seed predation. The mean number of stems and number of flowers were the lowest in 2011 (x̅ = 1.96, SE = 0.27 and x̅ = 1.1, SE = 0.17 respectively), while the mean number of capsules produced was equally lowest in both 2010 and 2011 (x̅ =

0.25, SE = 0.11). The number of stems, flowers and capsules did not differ in all other years of the study (Figure 3). Averaged over all years in the study (2009-2012; N=232), focal plants produced an average of 2.6 stems (SD = 2.07), 1.75 flowers (SD = 1.33) and

0.84 capsules (SD = 0.99) per plant. The average number of plants dormant during the study was 24 (SD = 7.81) or 10.3 %, but was variable with 19 plants dormant in 2010, 33 plants dormant in 2011 and 20 plants dormant in 2012. The number of stems and flowers produced was lowest in 2011 (F3,231 = 3.23, p = 0.02; F3,231 = 6.20, p = 0.0005 respectively), with the lowest fruit set produced occurring in both 2010 and 2011 (F3,231 =

33.45, p<.0001) (Figure 3).

Throughout the study, flower number increased with increased numbers of stems produced the previous year (Table 2). For the 2010-2011 and 2011-2012 time periods, 78 numbers of flowers was also significantly related to increased flower production in the following year (Table 2). The percentage of capsules produced had no effect on future year flower production. Conversely, in both the 2009-2010 and 2011-2012 time periods, the proportion of capsules per flower produced by a plant increased the probability of the plant going dormant the following season (Table 3). However, for the 2011-2012 time period, plants that had more stems were less likely to go dormant in the following year

(Table 3).

Based on the weighted average matrix obtained over the four year study period, and assuming all non-emergent plants were dormant and the weighted average seed predation = 22% (SD=6.8), we estimated the deterministic growth rate of the Baseline

Model, to be λ = 1.015. When we removed non-emergent plants (previously assumed dormant) from the population (assuming death) the population growth rate declined to λ

= 0.9136. The removal of seed predation, representing a population without this post- pollination selective pressure, and modeling all missing plants as dormant, resulted in only a slight change to the growth rate with λ = 1.0174. The removal of dormant plants from that simulation resulted in a population growth rate comparable to the Baseline

Model with λ = 0.9248 Elasticity values for the two models when dormancy was assumed were the same, with most of the elasticity of λd attributed to stasis (58%) or growth (11%) of one flowered individuals within the population. Fecundity had a total elasticity value of 3.3%, whereas seed stasis, germination and progression through the protocorm stages accounted for a total of 13.8% of the elasticity of λd. Sensitivity values were heavily weighted towards seed germination and survival due to the high fecundity values in relation to the small survivorship values. When plants were scored as dead 79 instead of dormant, elasticity values differed substantially. In the Baseline model 62% of all variation in the model was explained by stasis of one-flowered individuals and 15% by increases in size from one-flowered to multiple flowered stages. When granivory was removed from the model, 75% of the variation in the model was attributable to germination and progression through the protocorm stages, with 21% attributed to maintenance in the seed bank.

The inclusion of high seed predation, associated with the 2009 burn year, into the dormancy model reduced fecundity in all reproductive size classes by 48% and resulted in λd = 1.0037. When non-emergent plants were assumed dead, the growth rate dropped to λd = 0.90. Elasticity values in the dormancy model changed slightly from the Baseline

Model with seed stasis and germination playing a more important role (14.8%). When death was assumed, elasticity values were once again focused on the stasis of one- flowered individuals (62%) and growth to a multi-flowered individual (15.6%).

Estimation of the effect of selfing from nectar production produced fecundity values identical to those in the Baseline Models, λd = 1.0053 (when dormancy assumed), or λd =

0.90 (if death assumed) while elasticities remained unchanged.

When both average seed predation and selfing due to nectar addition were included in the model, λd=0.9961 (dormancy assumed) and λd = 0.87 (death assumed), indicating declining populations. Additionally in the dormancy model, elasticity values for the transition of dormant, 1-flowered individuals to emerge from dormancy rose substantially (15.1% versus 0% in all other models). Furthermore the elasticity value for a dormant 1-flowered individual to remain in stasis increased from a value of 0 in all other models to 31.0%. Seed stasis and germination represented 4.6% of the elasticity on 80

λd. Average residence times remained the same through all models (Figure 4).

Individuals spent the most time in the one flowered individual stage (3.23 yrs.) as well as the multi-flowered dormant stage (3.03 yrs.). Elasticity values of the models assuming non-emerged individuals had died were once again focused on stasis and growth of one- flowered individuals (65%).

Discussion

Plant size played an important role in predicting future year flowering in C. candidum. For all transition periods, the number of vegetative stems in the previous year was associated with the number of flowers in the following year, a trend seen in other perennial plants (Ollerton and Lack 1998; Gross 1981), although not observed in the closely related Cypripedium acaule (Primack and Hall 1990). The number of flowers was also predictive of the number of flowers in the following year except for the 2009 -

2010 transition period. It should be noted that a controlled burn was conducted by the land managers at the beginning of the field study (early spring 2009) and may have influenced growth and reproduction in that year by increasing light levels, changing soil nutrient levels, and increasing visibility of flowers or fruits to both pollinators and seed predators. The relationship between number of flowers between years is consistent with similar studies in the literature where flowering Cypripedium individuals were more likely to return to a flowering state in the following year than small non-flowering individuals were (Falb and Leopold 1993; Shefferson et al. 2003; Kéry and Gregg 2004).

Our dormancy estimates, indicating a 15% probability of single flowered plants going dormant, and a 42% probability of a multi-flowered individual going dormant, 81 were similar to estimates reported for other Cypripedium spp. studies (Shefferson et al.

2001; Nicolé et al. 2005). Non-flowering juveniles had a 33% probability of going dormant in a given year. The probability of a flowering plant to become dormant the following year falls within the range found by Shefferson et al. (2001) in C. parviflorum var. parviflorum (0.21 ± 0.23), and Nicolé et al. (2005) in C. reginae (0.32 ± 0.024) providing some validation to our model. Size, number of flowers and number of fruits produced in the previous year had variable effects on dormancy in the following year.

For both the 2009-2010 and 2011-2012 transition years, plants that set large proportions of capsules were more likely to become dormant in the following year. The 2010 - 2011 transition period may be somewhat anomalous due high rainfall and flooding of the field site during the 2011 flowering period, which led to lower than normal pollination and fruit set.

Literature on the effect of size and flowering on future year dormancy in

Cypripedium spp. has been contradictory. Shefferson (2003) reports vegetative plants are most likely to go dormant in C. calceolus, while Kéry and Gregg (2004) found that above ground plants were likely to remain the same size regardless of flowering. Other studies on Cypripedium spp. either downplay the importance of dormancy all together (Kull

1995) or do not report differences in dormancy rates between flowering and non- flowering individuals (Nicolé et al. 2005). In our study, all models that considered non- emergent plants as dead instead of dormant decreased λd by approximately 0.1, demonstrating the importance of dormancy in the long-term stability of C. candidum populations. Furthermore the elimination of dormancy from the model focused the majority of the elasticity on the stasis of one-flowered individuals, indicating the 82 importance of this size class on long-term population persistence. It is probable that other factors such as weather play an important role in dormancy in Cypripedium. Both

Kéry and Gregg (2004) and Shefferson (2001) have reported an effect of weather on dormancy and stage transitions. Further studies focused on effects of latitude, weather or climatic changes on vital rates are needed to fully understand the relationships between demography, dormancy and climate in long-lived perennials.

Seed predation in the baseline model was estimated at 22% over the four-year study period resulting in a population growth rate of 1.010. Removal of seed predation from the model produced a slight increase in λ of 0.007. High seed predation in this study (estimated at 42% from previous work done on C. candidum, Walsh Ch. 1) reduced population growth rate (λ) by only ~0.01 compared to average seed predation and no seed predation, but increased elasticity values for seed stasis and germination by 1%.

The increased emphasis on seed germination and survival was seen throughout the model when population fecundity was modified, leading us to conclude that reductions in the numbers of seeds within a population lead to a proportional increase in the importance of the extant seeds surviving and germinating. Our seed packet study following the methods of Rasmussen and Whigham (1993) conducted in tandem with this demographic monitoring revealed that 53% of seeds within a seed packet are destroyed on average per year. Although seed bank estimates have been historically difficult to obtain due to the small size of orchid seeds, previous research has shown variable rates of seed persistence, with Whigham et al. (2006) finding very high rates (>85%) of seed persistence and variable germination in Aplectrum, Corallorhiza, Liparis, and Tipularia. Other studies on a wide array of additional orchids have found seed persistence either lacking or very 83 low (van der Kinderen, 1995, Zelmer and Currah 1997; Batty et al. 2000; McKendrick et al. 2000). Additional variables with unknown effects such as climate, associated species, spatial relationships, soil pH and chemistry, as well as habitat quality most likely play a role in orchid seed persistence and should be explicitly examined in the future.

Management of the prairie through prescribed burns may have had significant effects on the number of fruits produced as well as the number of fruits predated.

Increase in fruit set may be due to increased nutrient availability, increased light and increased pollinator visibility after a burn. Although we found seed predation increased substantially during a burn year, this may be due to the increase in food available for the seed predators. Non-burn years saw the lowest number of flowers, stems and fruits produced, indicating a positive effect of the prescribed burn. Based on our data, the current practice of burning every three years is recommended in order to increase flowering and fruit set within the population. Furthermore, due to the strong emphasis our models had on the stasis and growth of one-flowered individuals, land managers can use censuses of one-flowered individuals before and after a burn as an indicator of the health of the population. Achieving populations that maintain a large number of one- flowered individuals should be the primary objective of management activities.

Research on deceptive pollination has heavily focused on orchids due to the high number of orchids (nearly 1/3 of all orchids) that employ this strategy (Dafni and

Bernhardt 1990). However, most studies have focused on the short-term effects, such as impacts on fruit set and geitonogamy. Using previously obtained data on fecundity rates with and without nectar reward, we were able to examine the demographic consequences of the deceptive pollination strategy. Our previous study showed no increase in fruit set 84 with the addition of nectar, an overall increase of geitonogamy, and a 50% reduction in seed mass for capsules resulting from a selfing event. These data, when evaluated in a demographic framework, strongly support the outcrossing hypothesis in that the addition of nectar reduced overall plant fitness by 22% and therefore the population growth rate.

Based on this study, and our previous work (Walsh Chapter 1) we concluded that, although the deceptive pollination strategy may contribute to the measured 42% pollen limitation, the resulting 22% decrease in female fitness due to selfing, outweighs the potential reduction in fruit set over the long life of the plant. Furthermore, these population growth rate estimates are conservative in that we only estimated a reduction in number of seeds produced and did not take into account reductions in seedling viability and offspring survival that may result from self pollination. Orchids that are subject to selfing events have been shown to exhibit inbreeding depression (Smithson 2006; Borba et al. 2001), especially in deceptive orchids that have evolved with very low levels of selfing (Ferdy et al. 2001; Jersáková et al. 2006). Although increased selfing due to deceptive pollination did not reduce λ below 1.0 by itself, the addition of seed predation did result in λ<1.0, indicating a declining population.

Although deception plays an integral role in Cypripedium reproduction, it should be viewed as a piece of the overall puzzle of a stochastic germination and predation environment. We suggest that the deceptive pollination strategy is uniquely tailored to conditions in which seed germination and survival are consistently low and seed predation is consistently high. In order to test this hypothesis, further research should be performed to compare demographic vital rates, seed germination and predation between groups of deceptive and non-deceptive orchids. This study demonstrates the importance 85 of understanding the demographic consequences of both mutualists involved in deceptive pollination and antagonists causing seed predation, where the increased fecundity resulting from the deceit strategy plays a critical role in population maintenance and expansion when other post-pollination selective factors are considered.

Literature Cited

Aragón, S., & Ackerman, J. D. (2004). Does flower color variation matter in deception pollinated Psychilis monensis (Orchidaceae)? Oecologia, 138(3), 405–13.

Barrett, S. C. H. 2002. The evolution of plant sexual diversity. Nature Reviews Genetics 3: 275-284.

Batty, A. L., Dixon, K. W., & Sivasithamparam, K. (2000). Soil seed-bank dynamics of terrestrial orchids. Lindleyana, 15(4), 227–236.

Calvo, R. N., & Horvitz, C. C. (1990a). Pollinator limitation, cost of reproduction, and fitness in plants: A transition-matrix demographic approach. The American Naturalist, 136(4), 499–516.

Caswell, H. (1989). Analysis of life table response experiments I. Decomposition of effects on population growth rate. Ecological Modelling, 46(3-4), 221–237.

Caswell, H. (1996). Analysis of life table response experiments II. Alternative parameterizations for size- and stage-structured models. Ecological Modelling, 88(1- 3), 73–82.

Falb, D. L., & Leopold, D. J. (1993). Population dynamics of Cypripedium candidum Muhl. ex Willd., small white ladyslipper, in a western New York fen. Natural Areas Journal, 13(2), 76-86.

Grindeland, J. M., Sletvold, N., & Ims, R. A. (2005). Effects of floral display size and plant density on pollinator visitation rate in a natural population of Digitalis purpurea. Functional Ecology, 19(3), 383–390.

Gross, K.L. (1981). Predications of frate from rosette size in four “biennial” plant species: Verbascis Thapsus, Oenothera biennis, Daucus carota, and Tragopogon dubis. Oecologia, 48(2), 209-213.

Hutchings, M. 1999. Clonal plants as cooperative systems: benefits in heterogeneous environments. Plant Species Biology, 14, 1–10.

Horvitz, C. C., & Schemske, D. W. (2002). Effects of plant size, leaf herbivory, local competition and fruit production on survival, growth and future reproduction of a neotropical herb. Journal of Ecology, 90(2), 279–290.

Janette A. Steets, Tiffany M. Knight, & Tia‐Lynn Ashman. (2007). The Interactive Effects of Herbivory and Mixed Mating for the Population Dynamics of Impatiens capensis.

Jäkäläniemi, A., Crone, E. E., Närhi, P., & Tuomi, J. (2011). Orchids do not pay costs at emergence for prolonged dormancy. Ecology, 92(7), 1538–1543.

Kesler, H. C., Trusty, J. L., Hermann, S. M., & Guyer, C. (2008). Demographic responses of Pinguicula ionantha to prescribed fire: a regression-design LTRE approach. Oecologia, 156(3), 545–57.

Knight, T. M. (2004). The effects of herbivory and pollen limitation on a declining population of Trillium grandiflorum. Ecological Applications, 14(3), 915–928.

Kull, T. (1999). Cypripedium calceolus L. Journal of Ecology, 87(5), 913–924.

Kéry, M., & Gregg, K. B. (2004). Demographic analysis of dormancy and survival in the terrestrial orchid Cypripedium reginae. Journal of Ecology, 92(4), 686–695.

Kéry, M., & Gregg, K. B. (2007, March 30). Demographic estimation methods for plants with dormancy. Animal Biodiversity and Conservation.

Lesica, P., & Crone, E. E. (2007). Causes and consequences of prolonged dormancy for an iteroparous geophyte, Silene spaldingii. Journal of Ecology, 95(6), 1360-1369.

Light, M. H. S., & MacConaill, M. C. (2011). Potential impact of insect herbivores on Orchid conservation. European Journal of Environmental Sciences, 1(2).

Menges, E. S. (1990). Population Viability Analysis for an Endangered Plant. Conservation Biology, 4(1), 52–62.

Menges, E. S. (2000). Population viability analyses in plants: challenges and opportunities. Trends in Ecology & Evolution, 15(2), 51–56.

Nicolè, F., Brzosko, E., & Till-Bottraud, I. (2005). Population viability analysis of Cypripedium calceolus in a protected area: longevity, stability and persistence. Journal of Ecology, 93(4), 716–726. 88

Ollerton, J., Lack, A. (1998). Relationships between flower phenology, plant size and reproductive success in shape Lotus corniculatus (Fabaceae). Plant Ecology, 139(1), 35-47.

Peakall, R. (1989). The unique pollination of Leporella fimbriata (Orchidaceae): Pollination by pseudocopulating male ants (Myrmecia urens, Formicidae). Plant Systematics and Evolution, 167(3-4), 137–148.

Primack, R.B., & Hall, P. (1990). Costs of reproduction in the pink lady’s slipper orchid: a four-year experimental study. American Naturalist, 136(5), 638-656.

Rasmussen, H. N., & Whigham, D. F. (1993). Seed ecology of dust seeds in situ: A new study technique and its application in terrestrial orchids. American Journal of Botany.

Rooney, T. P., & Gross, K. (2003). A demographic study of deer browsing impacts on Trillium grandiflorum. Plant Ecology, 168, 267–277.

Schutzenhofer, M. R., Valone, T. J., & Knight, T. M. (2009). Herbivory and population dynamics of invasive and native Lespedeza. Oecologia, 161(1), 57–66.

Shefferson, R. P., Sandercock, B. K., Proper, J., & Beissinger, S. R. (2001). Estimating dormancy and survival of a rare herbaceous perennial using mark-recapture models. Ecology.

Shefferson, R. P., J. Proper, S. R. Beissinger, and E. L. Simms. (2003). Life history trade- offs in a rare orchid: the costs of flowering, dormancy, and sprouting. Ecology, 84, 1199– 1206.

Shefferson, R. P. (2006). Survival costs of adult dormancy and the confounding influence of size in lady’s slipper orchids, genus Cypripedium. Oikos, 115(2), 253–262.

Shefferson, R. P., & Simms, E. L. (2007). Costs and benefits of fruiting to future reproduction in two dormancy-prone orchids. Journal of Ecology, 95(4), 865–875.

Shefferson, R. P., & Tali, K. (2007). Dormancy is associated with decreased adult survival in the burnt orchid, Neotinea ustulata. Journal of Ecology, 95(1), 217-225.

Van der Kinderen, G. (1995). A method for the study of field germinated seeds of terrestrial orchids. Lindleyana, (10), 68–73.

Willems, J. H., and C. Melser. 1998. Population dynamics and life-history of Coeloglossum viride (L.) Hartm.: an endangered orchid species in The Netherlands. Botanical Journal of the Linnean Society, 126, 83–93. 89

Zelmer, C. D., & Currah, R. S. (1997). Symbiotic germination of Spiranthes lacera (Orchidaceae) with a naturally occurring endophyte. Lindleyana, 142–148.

Tables and Figures

Table 1. Modified fecundity rates used in the demographic models. Baseline model refers to the original model including 22% seed predation averaged over the four years of the study. The no seed predation model was calculated by removing the effect of 22% seed predation from the fecundity estimate. The high seed predation model reflects a 42% reduction in seed production as measured in Walsh (Ch. 1). Nectar addition reflects a 48% reduction in seed mass when fruits were selfed and 78% of all fruits being selfed when nectar is supplied (Walsh Ch. 2). Nectar + seed predation is a combination of 22% reduction in seed production due to seed predation as well as selfing via the addition of nectar. Model 1 Flowered Fecundity 2+ Flowered Fecundity Baseline 3759 9133 No Seed Predation 6009 14600 High Seed Predation 3485 8468 Nectar Addition 3759 9133 Nectar + Seed Predation 2932 2992

Table 2. The effect of previous year’s number of stems, number of flowers and percent capsules (number of capsules/number of flowers) on the number of flowers produced the following year. A positive direction indicates increased flowering with an increase in the variable. Whole model tests: 2009-2010 p = 0.0001, 2010-2011 p = 0.032, 2011-2012 p < 0.001.

Years Variable DF L-R χ2 Prob> χ2 Direction 2009-2010 # Stems 1 6.83 0.0089* + # Flowers 1 0.004 0.94 % Capsules 1 0.13 0.71 2010-2011 # Stems 1 3.91 0.047* + # Flowers 1 7.77 0.005* + % Capsules 1 2.85 0.09 2011-2012 # Stems 1 4.58 0.032* + # Flowers 1 11.29 0.0008* + % Capsules 1 0.01 0.89

Table 3. The effect of previous year number of stems, number of flowers and percent capsules on the probability of becoming dormant in the following year. Negative directions indicate increased dormancy with an increase in the variable. Whole model test 2009-2010 p = 0.163, R2 = 0.06; 2010-2011 p = 0.209, R2 = 0.06; 2011-2012 p = 0.0002, R2 = 0.2.

Years Variable DF L-R χ2 Prob> χ2 Direction 2009-2010 # Stems 1 0.65 0.41 # Flowers 1 0.47 0.48 % Capsules 1 3.84 0.05* - 2010-2011 # Stems 1 0.26 0.60 # Flowers 1 0.44 0.50 % Capsules 1 2.85 0.09 2011-2012 # Stems 1 9.91 0.0016* + # Flowers 1 0.29 0.58 % Capsules 1 6.62 0.01* -

Figure 1. Life cycle diagram illustrating the stages used in the demographic model. Numbers correspond to weighted average transition probabilities between stages calculated over the four year study. Question marks indicate data missing from the experiment and were coded as “0” in the model. See table 1 for fecundity values (**) used in the models.

1.6 1.4 1.2 1 0.8 0.6

% Capsules 0.4 0.2 0 1 2 3 4 5 6 Number of Flowers

Figure 2. Mean percent capsules produced per plant based on the number of flowers with standard errors. Percent capsules calculated as the number of capsules produced by a plant divided by the number of flowers produced by that individual. The differences between groups were analyzed using a Generalized Linear model (df=5, prob>χ2=0.043) with a post hoc contrast analysis. The mean percent capsules between one and two flowered individuals differed significantly (prob>χ2=0.001). There was no significant difference between two or more flowered individuals (prob>χ2=0.18).

3.5 A AB AB 3

2.5 B A A A 2 A 1.5 B A

0.5 B B

0 2009 2010 2011 2012 2009 2010 2011 2012 2009 2010 2011 2012 # Stems # Flowers # Capsules

Figure 3. The average number of stems (F3,231=3.23, p=0.02), flowers (F3,231=6.20, p=0.0005) and capsules (F3,231=33.45, p<.0001) compared between years. Comparison of means using a Tukey-Kramer HSD; levels not connected by the same letter are significantly different at p<0.05.

3.5 3 2.5 2 1.5 1 0.5 Average Residence (Yrs.) 0 S0 P1 P2 P3 J A1 A2 JD A1D A2D Stage

Figure 4. Average residence time in years for each stage of the model. Average residence time did not vary with modified fecundity values. S = seeds, P = protocorm stages, J = juvenile, A = adult stage, D = dormancy stages.

CHAPTER V: GENERAL SUMMARY AND CONCLUSIONS

This dissertation focused on examining the role of pollen limitation, seed predation and deceit pollination on the reproduction and demographics of the Small

White Lady’s Slipper Orchid, Cypripedium candidum. Specifically the following questions were addressed in the three main chapters of the dissertation: 1) Are C. candidum pollen limited, does the quality of pollen affect reproduction and what factors affect pollination?; 2) What is the rate of pre-dispersal seed predation within the population and what plant or floral cues influence rates of seed predation?; 3) How does deceit pollination affect reproduction and rates of geitonogamy?; 4) What is the population growth rate and demographic characteristics of C. candidum in a large, healthy population?; and 5) What is the effect of seed predation and deceit pollination on the population demographics of C. candidum?

Plant mating systems are often composed of interrelated factors including mutualists, such as pollinators, and antagonists such as seed predators and other herbivores. Despite the importance of understanding the effects of both mutualists and antagonists on a plant breeding system, relatively few have examined these interactions simultaneously (Strauss and Irwin 2004; Abdala-Roberts et al. 2009; Burkhardt et al.

2009; Carlson and Holsinger 2010). Additionally, plant size may play an important role in attracting both mutualist pollinators and antagonist seed predators. Increased floral display size is expected to increase pollinator attraction and visitation (Peakall 1989;

Aragón and Ackerman 2004; Li et al. 2011), while also attracting seed predators to the accompanying large ovule resource (Stephens and Myers 2012). Through two hand 97 pollination experiments we found C. candidum plants to be consistently pollen limited and that plant height played an important role in both pollination and seed predation. In this system taller plants were more likely to be pollinated while shorter plants were more likely to be attacked by seed predators. We also found a strong reduction in seed mass associated with selfing events.

In Chapter III we examined the effects of deceit pollination on rates of inbreeding, pollen removal, deposition, and fruit production. Deceptive pollination, specifically food deception, is a pollination strategy in which the flower provides floral cues indicating a food reward while not actually providing that reward (Faegri and va der Pijl 1971;

Cozzolino and Widmer 2005). Non-rewarding flowers are found in 146 genera from 33 flowering plant families (Jersáková et al. 2009), but are most prevalent in the

Orchidaceae family. By experimentally manipulating plants to provide a nectar reward and tracking the fate of pollinia using histochemical dyes we were able to measure selfing and outcrossing rates in response to the presence or absence of a nectar reward. There was no significant difference in fruit production between the two treatments; however plants receiving the nectar treatment had nearly three times the rate of selfing compared to non-nectar treated plants.

In Chapter IV we explored the demography of C. candidum in a large thriving population over a four-year period of time. Demographic modeling is a technique often used to examine population level variation over space and time (Kéry and Gregg 2004).

Despite the usefulness of demographic models in understanding the consequences of evolutionary and ecological factors, there are no existing models examining the role of food deception on population dynamics. Our population data indicated an average rate of 98 seed predation over a four year period to be 22%. The population growth rate, including the 22% seed predation, was estimated at λ = 1.01. When including the effects of nectar production on the fecundity of plants, as measured in Chapter III, the population growth rate dropped to λ = 0.99, providing further evidence of the impact of the deceptive pollination strategy on plant reproduction.

This research demonstrates the multiple factors that affect reproduction in the complex mating system of C. candidum. Although there is a high degree of pollen limitation in the population, possibly due to the absence of a pollinator reward, there is a relatively low amount of inbreeding. The addition of nectar to the system increases inbreeding substantially and directly affects individual fecundity. Seed predation rates are variable from year to year, but serve as an additional selective pressure on the population. Although seed predation or nectar addition alone does not bring the population growth rate below 1, a combination of the two results in a declining population. The uniqueness of this research lies in the examination of multiple factors that affect reproduction and how these factors come together to impact population demographics. In the Cypripedium system, we found no negative effect of the deceit pollination strategy. The addition of nectar did not increase fruit production, nor did it increase the number of pollinia removed from a plant. The evolution of deceit pollination appears to be driven primarily by the consistently high outcrossing rates it enforces and the decrease in plant fecundity under self-pollination. Further research examining the effects of self-pollination on both deceptive and non-deceptive orchids, in tandem with outside selective pressures such as seed predation, should be conducted in order to further 99 understand whether the deceptive pollination strategy has evolved primarily as a response to decreased fecundity from selfing.

100

Literature Cited

Aragón, S., & Ackerman, J. D. (2004). Does flower color variation matter in deception pollinated Psychilis monensis (Orchidaceae)? Oecologia, 138(3), 405–13.

Cozzolino, S., & Widmer, A. (2005). Orchid diversity: an evolutionary consequence of deception? Trends in ecology & evolution, 20(9), 487–94.

Faegri, K., & Van der Pijl, L. (1971). The Principles of Pollination Ecology. Pergamon.

Kéry, M., & Gregg, K. B. (2004). Demographic analysis of dormancy and survival in the terrestrial orchid Cypripedium reginae. Journal of Ecology, 92(4), 686–695.

Peakall, R. (1989). The unique pollination of Leporella fimbriata (Orchidaceae): Pollination by pseudocopulating male ants (Myrmecia urens, Formicidae). Plant Systematics and Evolution, 167(3-4), 137–148.

Strauss, S. Y., & Irwin, R. E. (2004). Ecological and evolutionary consequences of multispecies plant-animal interactions. Annual Review of Ecology, Evolution, and Systematics, 35, 435–466.

Stephens, A. E. A., & Myers, J. H. (2012). Resource concentration by insects and implications for plant populations. Journal of Ecology, 100(4), 923–931.