This thesis has been submitted in fulfilment of the requirements for a postgraduate degree (e.g. PhD, MPhil, DClinPsychol) at the University of Edinburgh. Please note the following terms and conditions of use:

This work is protected by copyright and other intellectual property rights, which are retained by the thesis author, unless otherwise stated. A copy can be downloaded for personal non-commercial research or study, without prior permission or charge. This thesis cannot be reproduced or quoted extensively from without first obtaining permission in writing from the author. The content must not be changed in any way or sold commercially in any format or medium without the formal permission of the author. When referring to this work, full bibliographic details including the author, title, awarding institution and date of the thesis must be given.

Ant-pollinator interactions in Turnera velutina: ecological costs and evolutionary consequences for the -plant pollination

Nora Villamil Buenrostro

January 2019

A thesis submitted for the degree of Doctor of Philosophy University of Edinburgh 2019

LAY SUMMARY

Ant-plants recruit aggressive to defend themselves against herbivores. Herbivores, eat plants voraciously. Plants have evolved different mechanisms to defend themselves against herbivores. Some are poisonous: like coffee or tobacco; some are prickly: like cacti or nettles. But others, persuade mercenaries - mostly ant guards - offering them sugary rewards and housing, and hence are called ant-plants. So, in exchange for protection services, these plants provide ants with food and nesting sites. However, most ant-plants are pollinated by , different from ants, and they also reward them with food, such as nectar. But guarding ants are usually not great at discriminating between friends and foes that visit the plant. So, pollinators may be weary when visiting the flowers of ant-plants. This plant problem is called ant-pollinator conflict. My PhD research tries to answer: Do ants and pollinators fight directly or over nectar? Does having efficient bodyguards scare pollinators away and how does this affect the plants? Can plants bribe ants away with extrafloral sugary rewards? I found that ants and bees do not fight over nectar, but they may fight directly. Pollinators spend less time visiting the flowers where the most aggressive ants are. Bodyguards scare away pollinators affecting flowers’ sex life. By reducing the time pollinators spend inside flowers, ants help the plants avoid mating with themselves (self-pollinating). Self-pollination can make plant seeds grow poorly, like haemophilia, an illness that made the royals die from bleeding and was caused by them marrying with their relatives. However, to reduce ant-pollinator conflicts, plant bribe ants away from the flowers providing them extrafloral sugary rewards on the outskirts of the flowers. Understanding how guarding ants affect the pollination biology and fruit production of ant-plants is crucial for all of us, since we deeply care about our morning toast, sweet or savoury, and our passionfruit juice to down the toast. ABSTRACT

Ant-plants recruit ants to defend them against herbivores, but most of them also require pollinator for successful seed set. Interactions between patrolling ants and pollinators on ant-plants have received relatively little attention. Negative ant- pollinator interactions are expected for several reasons. First, ants and pollinators benefit from plant investment in different functions (defence and reproduction, respectively) which can lead to plant trade-offs. Second, although more aggressive ants are likely to be better defenders, they may also deter pollinators, affecting plant fitness. However, ant-plants may have mechanisms to manage ant-pollinator interactions, maximising the benefits from their services whilst minimising the costs.

I used field experiments to investigate the ecological costs and evolutionary consequences of ant patrolling for the pollination biology of a facultative ant-plant, the Mexican endemic Turnera velutina (Passifloraceae), addressing the following questions: a) Which are the most aggressive and best ant defenders of T. velutina? b) Is there direct (pollinator deterrence) or indirect (nectar trade-offs) ant- pollinator conflict? c) What are the ecological costs and evolutionary consequences of myrmecophily for the host plant pollination and mating system? d) Do ant-plants have adaptations to cope with both mutualists, avoiding conflict?

Answers: a) sp. ants was detected as parasitic non-defenders, and the remaining ant species were ranked as: Capmponotus planatus < Crematogaster sp. < Paratrechina longicornis < Brachymyrmex sp. < Dorymyrmex bicolor. b) I found evidence for direct but not for indirect conflict. c) Ant patrolling affected pollinator visit duration, pollen loads, outcrossing rates, and male fitness, leading to negative effects on pollinator foraging efficiency, but such changes had positive effects for plant fitness increasing outcrossing and male fitness. d) Extrafloral nectar also serves to bribe ants away from reproductive structures during the crucial pollination period, reducing the probability of ant-occupation of flowers, reducing ant-pollinator conflict, and increasing plant reproductive success.

DECLARATION

I declare that this thesis has been composed solely by myself and that it has not been submitted, in whole or in part, in any previous application for a degree. Except where stated otherwise by reference or acknowledgment, the work presented is entirely my own.

14 January 2019 Nora Villamil Buenrostro

ACKNOWLEDGEMENTS

First and foremost, thanks to the Mexican people. Their tax contributions, despite the Mexican government’s well-known skimming of public resources, funded my postgraduate studies through a Conacyt Graduate Scholarship. A real privilege for which I will be always grateful.

Thanks to my supervisors, both past and surviving ones. My PhD was not a smooth glide, but rather a jumpy rollercoaster with constant changes. With almost as many supervisory teams as years in the PhD. Two locations and three different projects down the line, this was not a smooth journey, but a very enrichening one. Although I cursed them at the time, now I appreciate every single one of those transitions.

Thanks to Graham, my longest standing official supervisor. Thanks for the opportunity provided, despite the abnormal starting circumstances. I am very grateful for your openness and risk-prone behaviour. Being willing to take a random person with a half-started project which did not align with your current projects and for which there was no budget, is quite a bet I shall always be grateful for. Without that opportunity this PhD would not have been possible, and I have thoroughly enjoyed it. I am very glad to have been ‘associated to’ your lab. Thank you as well for your knowledge and patience, and meticulous teachings on how to use whiskey to proofread a paper for the 500th time. For the nourishing discussions on hypothesis and biological implications. Thank you for all the space and freedom you gave me to shift my project in any direction I wanted to. But above all, thanks for being always open to receiving critical views, even those that were not coated with Britishness. Thanks for resisting all the meetings, half of which I know felt like boxing matches. I have learned so much and it has been all good fun.

Thanks, and many of them, to Jarrod. Your contribution towards the successful completion of this thesis was enormous and scattered in all dimensions. Thanks for your kind disposition to take me on board, the statistical advice and for being the wizard that turned weeks of problems into solutions in an R spell. Thanks for the patience, empathy, but above all, the modesty with which you share your knowledge and time. Thanks also for employing me as a field assistant despite my spatial awareness and visa handicaps. The salary was essential for the writing up months, but the outdoorsy break was too!

To Karina all my gratitude for being the longest standing unofficial supervisor, bearing with me for over a decade. Thanks for the advice, for always being there for me ‘en las buenas y en las malas’. Welcoming, year after year, project after project, helping me get out of every ditch I got myself into. For giving me the courageous spell to exit Harry-Potterland. Your never-ending enthusiasm, your encouragement for the good ideas, and your honest discouragement of the crazy ones have been invaluable. Thanks for believing in me even when not even I did so. Your support has been so big that it spread across the ocean!

Roosa, long time no see but still many thanks for the starting push, the coffee room meetings where you patiently listened to and unravelled messy ideas and enhanced them with accurate suggestions. Thanks for showing me the best sandwich shop in town, the yoga lessons and the opportunity of a fresh, new start.

William and Cicely, thanks for the multiple rescue missions expanding from the advisory realm to the emergency couch, or the great plans Ebola destroyed. Your passions for tropical botany are contagious and were inspiring and sparkling bubbles in the flattest and darkest of times.

Many thanks to Alex Twyford and his lab for kindly being my foster lab, even though I struggle to comprehend any genomics! Thanks to Susie Johnston and Patrick Walsh for their support.

I am grateful to my examiners, Ally Phillimore and Scott Armbruster, who brought upon themselves this heavy document.

To the Stone lab, past and present members, a great crowd! To the LIPA lab, a place of endless comradery. To Sofía and Xochitl for a decade of side-by-side growth, fun and friendship. To all the other Turneritos and non- Turneritos.

Thanks to the volunteers who helped out during fieldwork: Sofía Ochoa, Xochitl Damián and Paulina Zedillo, Alberto Bernal, Arturo Martínez, Rosa Ríos, Iñigo Silva, Adriana Fournier, Jorge Comensal, Ana Urrutia, Aura Orloff, and to Helene Köck who helped processing pollen samples.

Thanks as well to the staff at CICOLMA Field Station for the palapa chats, the best beans in the world, the natural remedies for any ailment and much more.

To my friends, Licha, Fabs y Chiva for being throughout time and space, the cathartic/revitalising chat and the Friday memes that fuelled me many a times. To Andresito, Orne, Santi, Jonás and Paula for sharing the love and hatred of science, graduate days, and so much more. For giving me strength, unconditional support, courage and advice, and keeping sane and cosy many a times. To Iñigo for the epistolary joy and to Jorgito for the encouragement to write. To both for the shared love of whiskey, mezcal, sotol, nature, and literature, all healthy vitamins for life. To the Oxford pals. Clementico and the alternative coffee breaks, the walk through the cow field on the day of doom, the last minute-midnight dinners, the travels and adventures had and to come. Shimmering and the bargain shopping and deep life chats at crazy night hours. Louise and her kindness with my poor English, my crow-penguins, and her delicious elderflower cordial. To the Edinburgh pals. Cristina and Richard, Nikola and Jovana, Lisa, Andrés, Stevie, Maarit, Doris, Mafe and Gytis, Christine, Mario and Karen, Myria and Guille. For the feasts, the pub nights, the walks in the woods, the dancing nights out, the ears lent for science or cursing… Edinburgh would have been wetter and more dreich without youse. Words will always be meagre to describe how lucky I am of having such good friends.

To Tom, for the complicity, solidarity, gigantic support, patience, advice, love and what not. Without a doubt, the best finding of this whole PhD!

To my mother and sister. Karla, for being always there for scheming, pranks, travels, or calm advice. Despite being younger you have always been cannier and more sensible.

Mamá, thanks for supporting me to study as far as I wished to, since the early days… I could not have gotten here on an empty stomach or a missing notebook. This degree would not have been possible without you. Thanks for your generosity, your kindness and your love.

TABLE OF CONTENTS

CHAPTER 1 ...... 6

GENERAL INTRODUCTION ...... 6 INTRODUCTION ...... 7 Defining ant-plants (myrmecophily) ...... 7 Facultative vs. obligate mutualisms ...... 7 Indirect interactions: a simultaneous approach ...... 8 Ants as agents for floral traits ...... 10 RESEARCH AIMS AND QUESTIONS ...... 12 CHAPTERS OVERVIEW ...... 12 Chapter 2: Ant aggressivity ...... 12 Chapter 3: Ant-Pollinator conflict ...... 13 Chapter 4: Ecological costs ...... 13 Chapter 5: Extrafloral nectaries as ant distractors ...... 14 STUDY SYSTEM AND SITES ...... 15 The study system in brief ...... 15 The study sites in brief ...... 18 Conservation of coastal sand dunes in Mexico: an essay ...... 20

CHAPTER 2 ...... 28

ANT AGGRESSIVITY AND PROTECTION EFFECTIVENESS ...... 28 INTRODUCTION ...... 29 Quantifying aggressivity: a brief summary ...... 32 QUESTIONS & HYPOTHESES ...... 36 MATERIALS AND METHODS ...... 36 Study site ...... 36 Study system ...... 36 Anti-herbivore assays ...... 37 QUANTIFYING AGGRESSIVITY ...... 40 Maximum aggressivity score ...... 40 Likelihood of aggressive behaviours ...... 41 Ant recruitment ...... 41 Response time ...... 42 Repellence time ...... 42 Protection effectiveness ...... 42 RESULTS ...... 44 Quantifying aggressivity ...... 45 Ant recruitment, response and repellence time ...... 49 Collinearity plots between aggressivity variables ...... 52 Maximum aggressivity score ...... 54 Protection Effectiveness Index ...... 54 DISCUSSION ...... 57 How to quantify aggressivity? ...... 57

1

. Which are the most aggressive ant species? ...... 60 Are ant traits good predictors of aggressivity? ...... 62 Protection effectiveness varies between hosts ...... 64 The more, the merrier ...... 65 ACKNOWLEDGEMENTS ...... 66

CHAPTER 3 ...... 67

ANT-POLLINATOR CONFLICT ...... 67 INTRODUCTION ...... 68 QUESTIONS & AIMS ...... 71 MATERIALS AND METHODS ...... 72 Study site ...... 72 Study system ...... 72 FIELDWORK METHODS ...... 73 Mutualist activity curves: patrolling ants and pollinators ...... 73 Nectar secretion and pollen deposition curve ...... 73 Direct conflict ...... 75 Indirect conflict ...... 76 STATISTICAL METHODS ...... 77 Mutualist activity and reward secretion curves ...... 77 Timing of daily activity and secretion peaks ...... 78 Direct conflict ...... 78 Indirect conflict ...... 79 RESULTS ...... 80 Mutualist activity, reward secretion, and pollen deposition curves ...... 80 Direct conflict ...... 84 Indirect conflict ...... 84 DISCUSSION ...... 89 Mutualist activity, reward secretion and pollen deposition ...... 89 Direct conflict ...... 91 The effect of ants’ deterrence can vary: ...... 92 Does anti-herbivore match anti-pollinator? ...... 92 Indirect conflict ...... 93 CONCLUSIONS ...... 95 ACKNOWLEDGEMENTS ...... 96

CHAPTER 4 ...... 97

ECOLOGICAL AND EVOLUTIONARY COSTS OF ANT PATROLLING ...... 97 INTRODUCTION ...... 98 QUESTIONS & AIMS ...... 102 HYPOTHESES AND PREDICTIONS ...... 103 Pollinator composition and behaviour ...... 103 Pollinator visitation and plant mating system ...... 104 Plant fitness ...... 104 MATERIALS AND METHODS ...... 105 Study site ...... 105 Study system ...... 105

2 . Ants exclusions and experimental setup ...... 106 Composition ...... 110 Visitation frequency ...... 111 Visit duration ...... 111 Pollinator behaviour ...... 112 Pollen dyes ...... 113 Plant mating system ...... 114 Male plant fitness ...... 115 RESULTS ...... 116 Floral visitors and potential pollinators ...... 116 Pollinator visitation: composition ...... 117 Pollinator visitation: frequency ...... 119 Pollinator visitation: visit duration...... 119 Pollinator visitation: behaviour ...... 119 Plant mating system ...... 120 Plant fitness ...... 124 DISCUSSION ...... 128 Effects of ant patrolling on pollinator visitation dynamics: Composition, visitation frequency and duration ...... 128 Effects of ant patrolling on plant mating system ...... 132 Effects of ant patrolling on plant fitness...... 135 Pollinator efficiency and net ant effects ...... 136 Conclusions ...... 138

CHAPTER 5 ...... 140

THE DISTRACTION HYPOTHESIS OF EXTRAFLORAL NECTAR ...... 140 INTRODUCTION ...... 141 QUESTIONS & AIMS ...... 145 HYPOTHESES AND PREDICTIONS ...... 146 MATERIALS AND METHODS ...... 146 Study site and system ...... 146 FIELDWORK METHODS ...... 146 Surveys of ants inside flowers ...... 146 Experimental manipulation of EFN secretion ...... 147 Impacts of EFN secretion on fitness ...... 148 STATISTICAL ANALYSES ...... 149 Surveys of ants in flowers ...... 149 Ecological consequences of EFN secretion ...... 150 Impacts of EFN secretion on plant fitness ...... 152 Exploring the responses and effects of different ant species ...... 152 Effect sizes ...... 153 RESULTS ...... 153 Surveys of ants in flowers ...... 153 Floral visitors ...... 154 Ecological consequences of EFN removal ...... 157 Impacts of EFN secretion on plant fitness ...... 158 Comparison of patterns across spatio-temporal scales ...... 162 Ant species-specific responses to clogging ...... 162 and effects on pollinators ...... 162 DISCUSSION ...... 164

3 . The Distraction Hypothesis ...... 164 Exploring ant species-specific effects ...... 167 The spatio-temporal scale of the Distraction Hypothesis ...... 169 Implications for ant and pollinator foraging strategies ...... 170 Conclusions ...... 171

SUPPLEMENTARY ...... 173

MATERIAL 1 ...... 173 FIELDWORK METHODS ...... 173 STATISTICAL ANALYSES ...... 174 RESULTS AND DISCUSSION ...... 176 Reduction of EFN secretion ...... 176 Effect sizes ...... 178 Comparing effect sizes and spatio-temporal scales ...... 179

SUPPLEMENTARY ...... 185

MATERIAL 2 ...... 185 STATISTICAL ANALYSES ...... 185 Effects of clogging on ant species ...... 185 Effects of ant species on pollinators ...... 186 RESULTS AND DISCUSSION ...... 187 Effects of clogging on ant species ...... 187 Effects of ant species on pollinators ...... 188

CHAPTER 6 ...... 192

GENERAL DISCUSSION ...... 192 SUMMARY OF KEY FINDINGS ...... 193 SOME LOOSE ENDS ...... 195 Natural history ...... 195 Is the corolla a natural ant shield for pollinators? ...... 195 Networks ...... 196 BROADER IMPLICATIONS ...... 197 Herbivore vs. pollinator susceptibility to ants ...... 197 Are Brachymyrmex ants the best guards by being pollinator friendly? ...... 198 Ant patrolling, outcrossing rates, and plant selective fruit abortion ...... 199 Ant patrolling & dioecy ...... 201 Herkogamy, outcrossing and ant aggressivity ...... 202

REFERENCES ...... 205

NATURAL HISTORY SUPPLEMENTARY NOTES ...... 222 Behavioural observations catalogue ...... 222

PUBLICATIONS ARISING FROM THIS THESIS ...... 230

4

THE START...

On not writing poetry By Ellie McDonald

This is yer Muse talkin. Ye’re on yer final warnin. Nae mair sclatchin I the kitchen, nae mair hingin out the washin, nae mair stour soukin. This is yer Muse talkin. Fae the wyste paper basket. […]

5

CHAPTER 1

GENERAL INTRODUCTION

Ant-plants Turnera velutina Coastal sand dunes

6

Chapter 1: General introduction

INTRODUCTION

Defining ant-plants (myrmecophily) Mutualisms are defined as interspecific cooperative interactions in which two species are reciprocally exploited, resulting in a net benefit for both partners (Begon et al. 1986, Bronstein 2001). Plants interact with a vast range of animals in mutualistic or antagonistic relationships (Herrera and Pellmyr 2009). Animals that interact with plants can also interact with each other. In fact, to deter herbivores, some plant species recruit ants that predate on herbivores (Belt 1874, Janzen 1966, Bentley 1977). Extrafloral nectaries (EFN), domatia and food bodies have been described as mechanisms by which plants attract ants, by providing nesting sites or nutrients (Rosumek et al. 2009). In return, ants attack herbivores, prune climber vines and prevent fungal and microbial infestation on plant tissues (Bentley 1977, Rosumek et al. 2009). This barter of food and lodging products in exchange for protection services between ants and plants is called myrmecophily, and the plants are often called ant-plants (Rico- Gray and Oliveira 2007a, Herrera and Pellmyr 2009).

Facultative vs. obligate mutualisms Ant-plant mutualisms can be classified as obligate or facultative depending on the tightness of the interaction, the rewards provided, the specificity of the mutualisms and the number of ant species involved. Plants in obligate mutualisms (myrmecophytes) offer domatia and food bodies which are specific food rewards and nesting sites. Obligate mutualisms often involve only one specialised ant species that tends to nest and feed exclusively on the host plant (Rico-Gray and Oliveira 2007a, Herrera and Pellmyr 2009, Rosumek et al. 2009, Quintero et al. 2013). In contrast, plants in facultative associations (myrmecophiles) tend to be patrolled by many ant species, from different

7 Chapter 1: General introduction subfamilies (Rico-Gray and Oliveira 2007a), that do not nest on the plant and show low fidelity to the food association (Kersch and Fonseca 2005). Domatia offer a more specific reward and ants show stronger fidelity to domatia providing hosts than EFN-hosts (Rosumek et al. 2009). Despite exceptions (see Kautz et al. 2009 for specialisation against cheating ants), extrafloral nectaries are general sugar source available for any to harvest. Consequently, EFN- based mutualisms tend to be less specialised and specific than those mediated by domatia or food bodies (Rico-Gray and Oliveira 2007b). In fact, providing only EFN is one of the criteria to define an ant-plant mutualism as facultative and the host plant a myrmecophile (Quintero et al. 2013).

Patrolling ants from obligate systems are expected to be more aggressive and better guards than those in facultative systems, because ants in obligate mutualisms are defending their nesting sites and colony, as well as their food source. In contrast, ants from facultative systems are only defending part of their food sources. Differences in their protection efficiency were confirmed by several meta-analyses (Chamberlain and Holland 2009, Rosumek et al. 2009, Trager et al. 2010) revealing that obligate ant-plants suffered, on average, three times more damage when ants were excluded than their facultative counterparts (Rosumek et al. 2009). So far, no obligate ant-plants associations have been reported in temperate regions; whilst in the tropics 70% of the mutualisms are facultative and 30% are obligate, with EFN being the most frequent ant attractor (Rosumek et al. 2009).

Indirect interactions: a simultaneous approach Although huge progress has been made in understanding different plant-animal interactions through the last decades, most research on plant-animal interactions has been done by uncoupling the community and considering paired interactions (e.g., plant-herbivore, plant-pollinator, plant-fungus) (Strauss 1997, Herrera 2000). This pairwise approach, although convenient, is

8 Chapter 1: General introduction artificial and oversimplifies nature (Herrera 2000) since plants interact with pollinators, herbivores, herbivore predators and pathogens, among others, simultaneously or sequentially throughout their lifetime (Armbruster 1997). All these interactions jointly affect plants at many levels, from anatomy to physiology, development, ecology and evolution. Furthermore, plants interactions with a certain animal group or guild may also affect relationships with other groups or guilds (Armbruster 1997, Strauss and Irwin 2004, Adler 2008). By simultaneously investigating more than one interaction it is possible to assess the relative importance of each interacting organism for plant fitness (Herrera 2000) as well as the benefits plants provide for each of the interactors.

A direct interaction is defined as a pairwise relationship between a plant and another living organism mediated by a plant trait; indirect interactions are defined as those involving specialised plant traits, and mediated by a third species (Adler 2008). There are few publications on the effects of indirect interactions on evolutionary ecology (Herrera et al. 2002, Irwin et al. 2004, Strauss and Irwin 2004, Takayuki 2005), and indirect interactions have received little attention in herbivore-plant-pollinator systems and have been almost unexplored regarding ant-plant-pollinator interactions (but see: Assunção et al. 2014, Alves-Silva et al. 2015, Dáttilo et al. 2016, Del-Claro et al. 2016, Del-Claro et al. 2018).

In order to successfully grow and reproduce most plants must attract mutualists and deter antagonists. Beneficial and prejudicial interactions are not a cut-and-dry determination. The involvement of a third species is likely to influence pairwise plant-animal interactions. A growing amount of evidence now suggests that pollinator attracting traits and herbivore resistance traits are not independent (Strauss 1997, Strauss et al. 1999, Herrera 2000, Mothershead and Marquis 2000, Strauss and Irwin 2004, Takayuki 2005, Buchanan and Underwood 2013). The negative effects of herbivores on plant fitness can be

9 Chapter 1: General introduction direct (by reducing available resource due to lost photosynthetic area) or indirect (by deterring pollinators or affecting floral or mating system traits). Nonetheless, many studies have implicitly assumed that the effects of herbivores on plant fitness are due to direct effects (Rodríguez-Robles et al. 1992, Johnson et al. 1995, Conner and Rush 1996, Strauss et al. 1996), and little attention has been paid to indirect effects of herbivory on plant fitness. Some of the few studies on indirect effects of herbivores on pollination highlight the importance that induced defences have on pollination biology (Lohman et al. 1996, Kessler and Halitschke 2009, Kessler et al. 2011, Campbell and Kessler 2013), emphasising how pollination services and effectiveness can be influenced by herbivory (Takayuki 2005).

Features that originally evolved to serve one function, can later co-opt to serve the other one, such as anti-herbivore resins that develop into pollinator rewards (Armbruster 1997). The evolution of plant-pollinator interactions is a great example of this: Flowers – a key trait in angiosperms massive adaptive radiation – are nowadays a means to lure pollinators but might have evolved as a defensive mechanism to protect pollen and ovules from herbivores (Mulcahy 1979). Enclosing the reproductive structures in modified leaves (carpels) that later evolved into ovaries, petals, and sepals provided the basis for natural selection to act upon (Willmer 2011). Since some visitors were less detrimental than others, and some carpels were more protective than other, coevolution shaped plant-herbivores interactions into plant-pollinator relationships (Rico- Gray and Oliveira 2007b).

Ants as agents for floral traits In many species plant reproductive traits seem to be the result of conflicting selective pressures exerted by mutualists and antagonists (Herrera 2000). Pollinators are not always the selective agents for floral traits and the selective pressures exerted by pollinators cannot account for all the variation observed in

10 Chapter 1: General introduction floral traits (Galen 1999, Parachnowitsch and Caruso 2008). Several studies have shown that herbivores indirectly modulate the nature, strength and consequences of plant-pollinator interactions (Lohman et al. 1996, Strauss 1997, Strauss et al. 1999, among others) and modify plant mating systems (Leege and Wolfe 2002, Adler 2008, Kessler and Heil 2011). Defensive traits might directly deter pollinators and therefore select for less defended phenotypes; pollinators could also be selecting for highly defended plants when preferring undamaged individuals (Adler 2008). It has been extensively shown that leaf herbivory reduces plant fitness by deterring pollinators (Strauss et al. 1996, Strauss 1997, Mothershead and Marquis 2000, Poveda et al. 2003). Actively patrolled plants are better defended and therefore less damaged, which that might attract more pollinators, but, the presence of ants could also imply pollinator deterrence, and this has been poorly studied.

Although the overall effect of ants on plant fitness has been demonstrated as positive (Romero and Koricheva 2011), ant patrolling can affect pollinator dynamics and plant mating systems in a variety of ways revealing interesting community ecological dynamics. Defensive traits, such as aggressive ant patrollers, might directly deter pollinators and therefore select for less defended phenotypes; but pollinators could also be selecting for highly defended plants when preferring undamaged individuals (Adler 2008).

Ant-pollinator interactions are likely to impact pollen transfer, influencing pollination success. Ants might deter pollinators, causing pollen limitation in animal pollinated ant-plants. If this is the case, it would be expected that selfing mechanisms could be selected in plants as a reproductive insurance to compromised animal vectors. Although selfing reduces genetic variability in populations, increasing the risk of inbreeding depression, it has been suggested that a switch from an allogamous to an autogamous mating system might be beneficial in marginal environments, since it provides a back-

11 Chapter 1: General introduction up, reproductive insurance when mates or pollen vectors are scarce (Campbell and Kessler 2013). The same logic can be applied when pollinators are in short supply because of ants: those ant-plants capable of selfing are expected to have a better outcome. On the other hand, ant-pollinator interaction may be beneficial for plant fitness if ants increase pollen transfer efficiency by pollinator relocation. In this case, mechanisms that promote outcrossing would be benefitted. More studies are required to gain a better understanding on the outcome of ant-pollinator interactions, their effects on plants and the adaptations plants might have developed to promote or ease these relationships.

Yet, these are only hypotheses. The relative importance of mutualists and antagonists in shaping plant traits is mostly unknown given that studies rarely identify the agents of selection on plant traits (Parachnowitsch and Caruso 2008). Ants have been reported as agents selecting for floral traits in Polemonium viscosum (Galen 1999, Galen and Cuba 2001) and Erica halicacaba (Turner et al. 2012). However, neither of these species are ant-plants and so ants in these cases ants are herbivorous, prejudicial , rather than beneficial herbivore predators. To my knowledge, the potential of mutualistic ants as selective agents upon floral traits has not been explored.

RESEARCH AIMS AND QUESTIONS

The main aim of this thesis is to investigate the ant-pollinator interactions on a facultative ant-plant, and assess the ecological and evolutionary consequences of ant patrolling on the host plant pollination biology.

CHAPTERS OVERVIEW Chapter 2: Ant aggressivity The net outcome in facultative mutualisms depends heavily on the identity of the partner species involved, the quality of the rewards provided, and the

12 Chapter 1: General introduction aggressivity and protection effectiveness of the patrolling ant species (Raine et al. 2004, Ness et al. 2006, Del‐Claro and Marquis 2015, Fagundes et al. 2017). In this chapter I quantified the aggressivity of the most abundant ant species patrolling on T. velutina.

Chapter 3: Ant-Pollinator conflict During anthesis, conflict between ants and pollinators is expected because aggressive ants inside the flowers might directly attack pollinators or make the flowers unattractive due to high predation risk and low rewards. Plant trade- offs in resource allocation to reproduction or defence is another reason why we expect conflicting interactions between pollinators and ants. Plants have a limited amount of resources they must distribute amongst different physiological needs, such and growth, defence, reproduction, etc. Resources allocated to one of these functions are no longer available for the others, and so trade-offs can occur. Pollinators would benefit from resources invested in reproduction, whilst ants benefit from plant investment in defence.

This chapter explores whether ants and pollinators have a conflicting relationship expressed as direct conflict in which the presence of ants inside flowers hinder pollination, and as indirect conflict over plant reward allocation to each mutualistic guild. Comparing how plants invest in each mutualist’s reward and whether reallocation is possible are two key steps towards understanding ant-pollinator interactions.

Chapter 4: Ecological costs Negative interactions between predator ants and pollinators are theoretically expected, and have been confirmed in several species (Rico-Gray and Oliveira 2007, Frederickson 2009, Palmer et al. 2010). Those studies that have addressed ant-pollinator conflict show contrasting evidence: ants can either increase

13 Chapter 1: General introduction fitness (Altshuler 1999) or reduce it (Ness 2006). Changes in pollinator composition, frequency and duration of the visits have been hypothesized as potential factors driving the positive or negative outcomes of ant-pollinator interactions for plant reproduction (Altshuler 1999, Ness et al. 2009). Pollinators’ responses to ant threats, may have different consequences for the host reproduction and mating systems. Yet, none of these have been expressly tested or observed in an adequate experiment where pollinator visitation, frequency, and duration of the visits is recorded along with plant fitness associated to plants with and without defensive ants patrolling. In this chapter we used an experimental approach to test the effects of patrolling ants on pollinator visitation frequency, duration, behavior, and their cascading effects on the host plant mating system and fitness.

Chapter 5: Extrafloral nectaries as ant distractors Previous findings show that EFN secretion volume and sugar content peak when leaves are associated with flowers, compared to EFN secretion in leaves associated buds and fruits (Villamil 2017). I tested if this increased secretion during anthesis could be part of a mutualist management strategy by which the plant lures ants outside the flowers, avoiding the costs of ant-pollinator interference without giving up the protection. This suggested function of EFN is known as the Distraction Hypothesis, and despite being proposed in 1878 (Kerner 1878), it had not been adequately tested (but see: Wagner and Kay 2002, Galen 2005, Chamberlain and Holland 2008 and further discussion in Chapter 5). In this chapter I tested whether EFN secretion was distracting ants from going inside the flowers, whilst luring them to patrol the vicinity of flowers defending them against florivores using an experimental design in which we prevented EFN secretion.

14 Chapter 1: General introduction

STUDY SYSTEM AND SITES

In this thesis, all data chapters (Chapters 2-5) contain research conducted on the same study plant (Turnera velutina) investigated at two different locations in Mexico (La Mancha, Veracruz and Troncones, Guerrero). Since all these chapters are self-contained research pieces, each will have a paragraph describing essential information about the study system and site, and although tailored to the specific needs and context of each chapter, these sections may be slightly repetitive. Apologies to the reader. However, here I contrasted the main differences between the study sites, providing a justification as to why I explore both sites.

The study system in brief Turnera velutina C. Presl (Passifloraceae: Turneroideae) is a Mexican endemic perennial shrub that grows in coastal sand dune scrubs and in tropical dry forests, from sea level to 1300 m of altitude (Arbo 2005). This species has alternate, petiolated, lanceolate-shaped, and pubescent leaves bearing paired cup-shaped extrafloral nectaries on the abaxial blade surface (Arbo 2005, Villamil et al. 2013). Flowers are axillary, yellow, pentamerous, bisexual, actinomorphic, self-compatible, and last only one day opening from three to five hours. Five floral nectaries are located at the base of the receptacle, between each pair of petals (Torres-Hernández and Rico-Gray 2000, Benitez-Vieyra et al. 2010). Although plants flower year round, reproductive output peaks during the rainy, summer season (Arbo 2005). Fruits mature in 10-20 days (Ochoa-López 2013) and reproductive structures from several ontogenetic stages (buds, flowers, and fruits) may be found in adjacent leaves within the same branch.

T. velutina is an entomophilous species, whose pollinators include lepidopterans (Aphrissa statira, Phoebis phillea, Anteos maerula, and

15 Chapter 1: General introduction unidentified members of the Hesperiidae family), hymenopterans (Apis mellifera, Agapostemon sp., Auglochlora sp., Ceratina sp., Auglochloropsis sp., Thygater sp., and Lassioglossum sp.), dipterans (Bombyliidae flies) have also been reported as pollinators (Fig. 2.2; Benitez-Vieyra et al. 2010, Ramos-Castro 2013, Sosenski et al. 2016). Besides displaying physical and chemical defenses and tolerance to cope with herbivore damage, this species also relies on myrmecophily to defend against herbivores (Villamil et al. 2013, Ochoa-López et al. 2015). Euptoteia hegesia Cramer (Lepidoptera: Nymphalidae) is the main foliar herbivore, and its activity peaks from June to August (Fig. 1.1; Cuautle and Rico-Gray 2003), although it can be found all year round. The foliar extrafloral nectar is made of sucrose, glucose and fructose in similar percentages (Elias et al. 1975) and is consumed by at 7-13 ant species (Fig. 2.1; Cuautle et al. 2005, Zedillo-Avelleyra 2017), and two species (Torres-Hernández and Rico-Gray 2000). Besides defending the leaves, some of these ants also disperse the seeds (Cuautle and Rico-Gray 2003).

This species was initially described as Turnera ulmifolia var. velutina (C. Presl) Urb. and recorded as a neotropical species distributed from Florida to Argentina, including several Caribbean islands (Schappert and Shore 1995, Arbo 2005). However, the was later amended by Arbo, who identified the Mexican specimens as a separate species and renamed it as Turnera velutina C. Presl (Arbo 2005). Turnera velutina has been recorded only in Mexico in the provinces of Chiapas, Guerrero, Jalisco, Estado de México, Michoacán, Oaxaca, Tabasco, Tamaulipas and Veracruz. Because of this taxonomic amendment many studies on this species have been published under the name of Turnera ulmifolia L. (C. Presl) Urb.

16 Chapter 1: General introduction

Figure 1.1. Some of the ant species patrolling Turnera velutina plants. a) Two ants (ca) harvesting seeds (se) from a T. velutina fruit (fe); b) Brachymyrmex sp.; c) Camponotus mucronatus; d) Camponotus bicolor; e) bicolor; f) gracilis; g) Cephalotes.

Figure 1.2. Main herbivores of Turnera velutina. Left: Euptoieta claudia; right Euptoieta hegesia.

17 Chapter 1: General introduction

Figure 1.3. Main pollinators of T. velutina. Top row: Flower and Apis mellifera. Middle row: lepidopterans. Bottom row: Native bees, Diptera (Bombyliidae); and wasp.

The study sites in brief Two different populations of T. velutina were studied because the array of pollinators differs greatly between both populations (Fig. 1.4). The European honeybee Apis mellifera was introduced in Mexico in the 1800’s (Guzmán-Novoa et al. 2011) and is the dominant pollinator at La Mancha, Veracruz (Ramos- Castro 2013, Sosenski et al. 2016) where most of my PhD research has taken place. The population in Troncones, Guerrero, differs from La Mancha, in that plants are visited by a more diverse array of native insects, mostly butterflies. Furthermore, the aggressivity of patrolling ants differ between locations, with ants from Troncones reported as more aggressive than those at La Mancha (Zedillo and Boege pers. comm.). Increased ant aggressivity creates a larger

18 Chapter 1: General introduction potential for direct ant-pollinator conflict. Such features made Troncones an interesting place to study ant-pollinator interactions and investigate how different pollinators (social and solitary) interact with the ant assemblage.

Figure 1.4. Map of Mexico indicating with stars the study sites: Troncones, Guerrero and La Mancha, Veracruz.

19 Chapter 1: General introduction

Conservation of coastal sand dunes in Mexico: an essay

On doing research on a sandy patch…

During my PhD, the most biologically interesting field site I visited was located in Troncones, a seaside village in the state of Guerrero, on the southern coast of Mexico: A vibrant and effervescent area in all respects. Just a few hundred kilometres west along the coast from the legendary beach of Acapulco, a prime holiday destination for 1950´s cinema divas and jet-setters, Troncones lies 380 km south-west of the historic city of Iguala, where (at least) 43 Ayotzinapa College students – training to be rural teachers - were killed on 26th September 201412. Although officially these students were mistakenly killed by the organised crime, it is common knowledge that Mexican army and police forces were involved. Furthemore, recent revelations, sparsely-covered in the mainstream media, suggest that the slaughter was requested by a drug dealer held for trial in the US, whose mercenaries coordinated with the Mexican military and police forces to kill and disappear students heading towards a political rally (although this remains as unofficial news; withheld information by the USA DEA agency)3456. Regardless of Marylin Monroe and Elizabeth Taylor or the 43 missing students (and the DEA’s secrets), the sandy patches of Troncones were of great interest to me. These patches host a large array of the

1 VICE News Documentary. The Missing 43: Mexico’s Disappeared Students. Link: https://www.youtube.com/watch?v=0jt-urgNN3A 2 The New Yorker. 30 September 2015. Francisco Goldman. Mexico’s missing forty-three: one year, many lies, and a theory that might make sense. Link: https://www.newyorker.com/news/news- desk/mexicos-missing-forty-three-one-year-many-lies-and-a-theory-that-might-make-sense 3 El Universal. 12 April 2018. Link: http://www.eluniversal.com/internacional/5994/jefes-cartel- dieron-ordenes-desde-chicago-caso-estudiantes-ayotzinapa 4 Aristegui Noticias. 12 April 2018. Link: https://aristeguinoticias.com/1204/mexico/mensajes-entre- guerreros-unidos-muestran-debilidad-de-verdad-historica-del-caso-ayotzinapa-centro-pro/ 5 Aristegui Noticias. 20 April 2018. Link: https://aristeguinoticias.com/2004/mexico/la-dea-oculto- informacion-al-gobierno-mexicano-sobre-caso-ayotzinapa-maria-idalia-gomez/ 6 The only English document I could find on the issue: https://unredacted.com/2018/04/16/transcripts-of-intercepted-cell-phones-open-new-lines-of- investigation-in-ayotzinapa-case/

20 Chapter 1: General introduction native insects that pollinate the plant I study, Turnera velutina: a relative of the passion-fruit. However, my study site stretched between two signposts reading “Se vende”; hence, my field site was on sale.

This sandy patch hosts an ecosystem called coastal sand dune or coastal scrub, which is located at sea-level, just a skip and a jump away from the sea! With the soil being mostly sand, the rainfall quickly runs through, making for a rather dry and water-restricted soil and habitat, even within the humid tropics. You can tell where this ecosystem starts and where the sandy beach ends because the sand in these scrubs is held in place by the roots of vegetation. Precisely because of this, these habitats are often called stabilised dunes, as opposed to the nearby sunbathing-type of dunes, which are mobile and shaped by the wind.

The first settlers of these sandy cords are prostrate plants like Ipomoea, the genera of sweet potato. Little by little, these shy sweet potato strings recruit other non-woody species, and then shrubs, until mixed grasslands cover all the sand dunes. The shrubs gather in thickets that recruit more species and form richer communities that spread inland, away from the sea and towards the tropical forest.

Until recently, the 1st of March 20187 to be precise, there were no laws or regulations that protected coastal ecosystems in the infamous Mexican LGEEPA: the General Law for Ecological Equilibrium and Environmental Protection. Despite being a country with over 800,000 ha of coastal sand dunes, all we had were six incongruent and mutually cancelling criteria regarding the

7 Federal Official Gazette (DOF:23/04/2018). Decree by which several dispositions within the General Law for Ecological Equilibrium and Environmental Protection are amended or added. Diario Oficial de la Federación (DOF:23/04/2018) Decreto por el que se reforman y adicionan diversas disposiciones de la Ley General del equilibrio Ecológico y la Protección al Ambiente.

21 Chapter 1: General introduction development requirements on Mexican beaches in a lower ranked law, called the Official Mexican Norm (NOM) (NMX-AA-120-SCFI-2006).

Duna is the Spanish word for dune, and a Norm is a lowly ranked law within the legal hierarchy of Mexican Law. Searching for duna within the NMX- AA-120-SCFI-2006 norm which regulates the quality of Mexican beaches provides just eleven hits. Seven of those hits comprise the paltry and incongruent criteria that do not protect sand dunes, but “establish the requirements and specifications on the quality and sustainability of the beaches”:

1. Native coastal sand dune vegetation shall not be removed, and the beaches must display adequate information regarding the efforts towards sand dune protection. (section 5.4.g) 2. There shall not be any infrastructure upon the coastal sand dunes. (section 5.3.d) [So far, so good, but…] 3. Any infrastructure that is to be built upon sand dunes, shall be built 5 metres behind the second cord of dunes. In those places where the presence of dunes is not easily identifiable due to size or because these have been removed, such infrastructure should be built behind the stripe of existing vegetation. If such stripe exists. (section 5.9.b) [Further down the norm states that…] 4. It is forbidden to remove any vegetation from the sand dunes. (section 5.9.d) 5. Beach accesses through the dunes should be wooden skywalks built using appropriate techniques to prevent erosion, allowing constant influx of users into the dune without deteriorating the sand dunes. (section 5.9.e)

22 Chapter 1: General introduction

6. No vehicle shall be allowed to circulate or park upon the beach or the sand dunes. Except for those vehicles involved in public services such as cleaning, bin collection, security, or towing vehicles. (section 5.5.i)

In March 2018 the Congress finally approved the petition to include the concept “coastal ecosystems” in the LGEEPA. Finally, ecosystems such as mangroves, wetlands, marshes, intertidal zones, coastal sand dunes, coastal lagoons, macroalgae, coral reefs, marine prairies, cenotes, oases, floodable rainforests, and rocky cliffs, were included in the Environmental Protection Legislation! Given the richness of these habitats, and that they are amongst the most lucrative spaces for touristic development, it is hard to believe that the long-standing lack of regulation for coastal environments was a coincidence, or a by-product of the overwhelmingly busy and understaffed Mexican Environment Council.

And so, back from fieldwork at Troncones. Counting is done, fieldwork is over, and I return to the lab desk in Auld Reekie. As I am writing the findings from Troncones I am left with a list of interesting wee puzzles to solve. Ecological nooks and crannies I am desperate to pry into. Ant gossip I am desperate to get! But… all of a sudden I am assaulted by a humongous concern: What if this patch finally gets sold?

What if someone builds there another holiday house? One more addition to that long tail of holiday rentals which are not inhabited, but rented or occupied for only a few weeks per year.

What if when I go to do the follow-up experiments I dream of, instead of this unregulated patch of trees, hives, plants, beetles, ants and butterflies, I find - a fence. A fence lined-up beautifully with cacti and Aloe vera plants, a couple of magueyes sprinkled here and there. A fence of breath-taking plants, especially

23 Chapter 1: General introduction when they decide to flower, which they will certainly do thanks to the caring hand of the local gardener. Him who gets paid per month what the owner probably charges the tourists for a night stay. And when not in flower, these majestic plants are really green, and really thorny, providing a natural discouragement from getting too close to the wooden planks lining the fence from inside. Behind the fence lays an enormous pool, surrounded by an orange clay-tiled patio and white, comfy sofas on which to sunbathe. In the corner, over a bar table, Margarita glasses hang upside down from a black, wrought-iron rail. The thick, blue, blown-glass rims of these harebell cocktail glasses stare into the infertile, orange clay tiles. Upstairs, a balcony under a thatch-roofed palapa will make for the perfect holiday let picture and title.

One more holiday let to go up on Airbnb under the sexy name of: “Casa Prana Yama: Oceanfront Balcony”, or “Casa Joya del Mar: Luxurious beachfront villa”, or perhaps as “La mariposa: Luxury beachfront villa”. All of these names aiming to be the most appealing punchline by grasping different avenues of uniqueness. But all these ads are in fact strikingly similar, sharing a key feature: the beachfront or oceanfront offer. A dangerous punchline, or shall I say a destructive one. The beachfront or oceanfront is, in ecological terms, the stabilised sand dunes ecosystem, an overlooked but heavily endangered habitat worldwide and particularly in Mexico.

Marketing alongside the capitalistic tourist culture has made many of us dream and drool of a beach holiday at a tropical, sandy beach with powdery white sands and warm seas. Many dream of an all-inclusive resort, with a choice of infinity pools, each with a bar, looking out over the ocean. Others dream of a modest hotel, but with the essential air-con, or maybe a very low-budget, low- impact ‘beachfront’ thatched-roof hut. Yet, all of these facilities are built upon coastal habitats. All of them. Mostly on stabilised sand dunes. No matter how

24 Chapter 1: General introduction modest, luxurious, environmentally-friendly or wasteful, posh or tacky, or hipster- all of these facilities displace the coastal ecosystems.

On top of these threats comes the icing of the cake, with plenty of room left for a cherry. Lack of protection regulations; developer’s interests: these threats have just budded, like bacteria, doubling themselves overnight, with the cheeky approval of the so-called Biodiversity Law. Approved in early April 2018 by a minimal quorum in an under-attended Congress; against the wishes and loud voices of hundreds of scientists, artists, NGO’s, and citizens8. Nonetheless, the ‘biodiversity' law was supported by the Mexican Green party, which is in bed with the ruling party (the PRI). The only ´Green´ party in the world that promotes mining and industrial activities inside nature reserves. With this law the Mexican Green party is pushing to deregulate and legalise the exploitation of Nature Reserves in a megadiverse country, such as Mexico.

In Troncones, Guerrero, with the “Se vende” posts marking its east and west limits, facing the sea to the south, and with its northern back against a Tarmacked road, lies one of the last strongholds of coastal sand dunes in Troncones that has not been built upon. Yet. But given the immense pressure for holiday houses in this area, the clock is ticking louder and louder every second. Meanwhile, while I spend days and weeks sorting out the databases, finding the right model, getting plots just right, and writing up my findings, I suddenly realise the hourglass has been dripping dry on the corner of my desk.

8 Gloria Leticia Díaz. 2 de abril 2018. Activistas y académicos alertan sobre privatización de los recursos naturales con Ley de Biodiversidad. Revista Proceso. Referencia Ley de Biodiversidad. Redacción de Aristegui Noticias. 3 de abril de 2018. https://aristeguinoticias.com/0304/mexico/falta- pulir-ley-de-biodiversidad-atenta-contra-derechos-de-pueblos-indigenas-prd/

25 Chapter 1: General introduction

On doing research in a sandy patch, watching it slip away, grain by grain out of your hands as you carry on counting leaves and critters. Keep calm and count critters? Impossible to keep calm!

I am the sole author of all the text used here, with comments on earlier drafts from Jorge Comensal and Thomas Godfrey.

This essay has been published in Spanish as: Dunas costeras. Un ecosistema codiciado, soslayado y

amenazado. Revista de la Universidad de México. Daños

26 Chapter 1: General introduction

‘Scotland small?’ By Hugh MacDiarmid

Scotland small? Our multiform, our infinite Scotland small? Only as a patch of hillside may be a cliché corner To a fool who cries ‘Nothing but heather!’ where in September another Sitting there and resting and gazing around Sees not only the heather but blaeberries With bright green leaves and leaves already turned scarlet, Hiding ripe blue berries; and amongst the sage-green leaves Of the bog-myrtle the golden flowers of the tormentil shining; And on the small bare places, where the little Blackface sheep Found grazing, milkworts blue as summer skies; And down in neglected peat-hags, not worked Within living memory, sphagnum moss in pastel shades Of yellow, green, and pink; sundew and butterwort Waiting with wide-open sticky leaves for their tiny winged prey; And nodding harebells vying in their colour With the blue butterflies that poise themselves delicately upon them; And stunted rowans with harsh dry leaves of glorious colour. ‘Nothing but heather!’ ̶ How marvelously descriptive! And incomplete!

27 Chapter 2: Ant aggressivity CHAPTER 2

ANT AGGRESSIVITY AND PROTECTION EFFECTIVENESS

Which are the most aggressive ant species?

28

Chapter 2: Ant aggressivity INTRODUCTION

Mutualisms, defined as reciprocal exploitation that results in a net benefit for both partners (Begon et al. 1986), are ubiquitous in nature and involve organisms from every kingdom (Bronstein 2001). Given their frequency there are many opportunities for exploitation of mutualisms (Bronstein 2001, Raine et al. 2004), in which one partner (now defined as a parasite) exploits resources offered by a second partner but, but without being mutually exploited by their partner, e.g., without also providing a net benefit for its partner (Bronstein 2001). And so, the mutual exploitation with a net benefit for both parts is broken.

Ant-plant mutualisms are also widespread, with over 4000 angiosperm species bearing EFN to attract ants, plus more species that attract ants via other rewards such as food bodies or nesting sites (Keeler 2014). In tropical rainforests approximately one third of woody dicots, herbs and vines produce ant-luring structures (Davidson and Cook 2008), and ants represent 50% of (Davidson and Cook 2008). EFN is a general resource, easily accessible to all agents encountering it (Carroll and Janzen 1973, but for exceptions see: Kautz et al. 2009). EFN-bearing plants are thus usually associated with multiple ant species, leading to facultative and diffuse interactions, in contrast with obligate mutualisms which are biunivocal interactions, usually involving specialised partners (Rico-Gray and Oliveira 2007, Herrera and Pellmyr 2009, Melián et al. 2009, Dáttilo et al. 2016). On average, facultative ant-plants are visited by 6-9 ant species, whilst obligate ant-plants are visited by one or occasionally two ant species (Rico-Gray and Oliveira 2007, Rosumek et al. 2009). The diffuse nature of many ant-plant associations limits the potential for tight coevolutionary interactions between partners, and these mutualisms are often associated with great variation in the net benefits received

29 Chapter 2: Ant aggressivity from their ant partners (Itioka et al. 2000, Apple and Feener Jr 2001, Ness et al. 2006, Fagundes et al. 2017).

Several factors can influence the net outcome of facultative ant-plant interactions and these have been classified by Del‐Claro and Marquis (2015) as internal (the identity of the particular species involved in the interaction) or external (abiotic factors that affect at least one partner). Internal variation is driven by differences among species in the quality (effectiveness) and quantity (frequency and abundance) of their interactions (Ness et al. 2006, Alves-Silva et al. 2015, Fagundes et al. 2017) and can often be stronger than external, abiotic factors such as fire (Del‐Claro and Marquis 2015).

Although, on average, ant species within facultative mutualisms are less aggressive than those in obligate mutualisms (Rosumek et al. 2009), interacting with multiple ant partners may be advantageous for plant for several reasons (Fagundes et al. 2017). First, if found simultaneously, ant species may differ in their anti-herbivore strategy, and so defensive success may increase with the number of partners (for exception on host sharing: Raine et al. 2004, Heil et al. 2009, Palmer et al. 2010). Redundant species, can have a synergic effect and enhance overall protection levels at a lower colony cost. Second, redundant species can replace absent protector species under different conditions, as these diffuse mutualists are variable in time and space.

Furthermore, despite the abundance of facultative mutualisms (Keeler 2014), the diversity of their associated species (Rosumek et al. 2009, Fagundes et al. 2017), the accessibility of EFN (Carroll and Janzen 1973, but for exceptions see: Kautz et al. 2009) and the consequent ease to cheat in such mutualisms (Bronstein 2001), very few examples of cheating species have been documented. According to

30 Chapter 2: Ant aggressivity Raine et al. (2004), until 2004 only four ant species had been detected as parasites (Janzen 1975, Young et al. 1996, Yu and Pierce 1998, Gaume and McKey 1999, Stanton et al. 1999), although now at least fifteen cases have been added to the list (Raine et al. 2004, Kautz et al. 2009, Del‐Claro and Marquis 2015, Fagundes et al. 2017). This low frequency of cheaters amongst facultative ant mutualists may reflect a biological truth, or represent a research underestimation bias deriving from the fact that few studies have actually estimated the protecting benefits of individual ant species.

Functional variation among ant species in apparent mutualisms must be kept in mind when using the abundance of ants to gauge the level of protection experienced by a plant. Although the number of ants recruited may be an estimate of plant attractiveness, ant identity is crucial for estimation of defensive benefits (Box 2.1). Many studies assume that all ant species collecting extrafloral nectar defend plants against herbivores, although very few studies have tested the protection services (Box 2.1) provided by each ant species. When aggressivity is evaluated, ant guards are often ranked based on a qualitative approach, using observations of clearly aggressive behaviours, such as biting and attacking, or ambiguous behaviours, such as recruitment as indicators of aggressivity (Ness et al. 2006). However, a quantitative approach is a more accurate estimate that reduces subjectivity and misinterpretation of aggressive behaviours (Ness et al. 2006, Fagundes et al. 2017), as for instance, it is easier to detect a biting attack than other attacking modes such as acid spraying. Hence, a currency to objectively measure the protection effectiveness (Box 2.1) of ants – such as herbivore damage prevented – in terms directly linked to plant benefits is crucial. But even fewer have quantified these parameters (but see: Apple and Feener Jr 2001, Raine et al. 2004, Ness et al. 2006, Del‐Claro and Marquis 2015, Fagundes et al. 2017).

31 Chapter 2: Ant aggressivity The protection effectiveness of an ant species is influenced by its taxonomic identity, which influences multiple morphological and behavioural traits including aggressivity (Del‐Claro and Marquis 2015, Fagundes et al. 2017), frequency of interaction with the host plant (Apple and Feener Jr 2001, Cuautle and Rico-Gray 2003, Ness et al. 2006), strength of recruitment, effectiveness in capturing or repelling a herbivore (Itioka et al. 2000, Raine et al. 2004, Ness et al. 2006, Fagundes et al. 2017), the plant ontogenetic stage and nectar availability (Villamil et al. 2013, Lange et al. 2017), and environmental conditions (Chamberlain and Holland 2008, Del‐Claro and Marquis 2015). The most effective ant protectors would be those ants that patrol often, and are likely to find, capture or deter herbivores from the host plant.

Box 2.1 Three key definitions

AGRESSIVITY Refers to the antagonistic behaviours displayed by ants. Particularly the probability of attacking, harassing, or displaying other behaviours to deter a visitor and the severity of such response. This is an estimate of ant behaviour and may play a role in other interactions such as pollinator deterrence.

PROTECTION EFFECTIVENESS Refers to the anti-herbivore capacity of ants and their ability to prevent or reduce plant damage by herbivores. This is an estimate of their benefits to the plant (although it inherently involves ant behavioural traits) and a proxy of their quality as plant defenders.

DEFENSIVE QUALITY The defensive quality is a qualitative, categorical and descriptive assessment of the overall ability of ants to prevent or reduce plant damage (e.g. good, bad, effective, non-effective, non-defenders).

32 Chapter 2: Ant aggressivity Quantifying aggressivity: a brief summary

Few studies have assessed the defensive quality of different ant species associated to ant-plants and even less have experimentally quantified the aggressivity of different ant species (Apple and Feener Jr 2001, Raine et al. 2004, Ness et al. 2006, Del‐Claro and Marquis 2015, Fagundes et al. 2017). Some studies have assessed ant aggressivity and responses when their host plant is artificially damaged (Christianini and Machado 2004, Romero and Izzo 2004, Zedillo-Avelleyra 2017). However, experimental quantification of ant aggressivity via in situ assays involving real herbivores on a host plant is, in my opinion, the best and most realistic method to standardise and quantify aggressive behaviours by defensive mutualists. Some studies using herbivores use surrogates that may not usually damage the plants but are experimentally convenient. Examples include termite workers (Apple and Feener Jr 2001, Fagundes et al. 2017), diverse local caterpillars and orthopterans (Fiala et al. 1989), or caterpillars of the generalist herbivore Manduca sexta (Ness et al. 2006). However, the degree to which ant aggressivity to surrogate herbivores mimics natural conditions remains unknown and is a cause for concern. My experiment is one of the few, perhaps the first one, to use natural herbivores of a host myrmecophile. The metrics and calculations of papers that, in my opinion, have best quantified ant aggressivity in an ant-plant context are summarized in Box 2.2.

33 Chapter 2: Ant aggressivity

Box 2.2 Aggressivity quantification methods

FIALA Fiala et al. (1989) evaluated protection effectiveness by placing a diverse array of local herbivore surrogates (several lepidopteran caterpillar species, orthopterans, beetles and bugs) and eggs on Macaranga leaves. They assessed the protection effectiveness of the ant species Crematogaster bornbeensis in four congeneric host plants (M. triloba, M. hosei, M. hypoleuca, M. hulletti). They evaluated ant protection using only a qualitative aggressivity metric: attack and repellence rates (Fiala et al. 1989).

NESS Ness et al. (2006) estimated the differences in the quality and quantity of defensive services provided by the four most common ant visitors to Ferocactus wislizeni when larvae of the generalist lepidopteran herbivore Manduca sexta were placed on the host plant.

They multiplied the quantity of the service by its quality to quantify the non-linear relationship between both components. They used Michaelis-Menton equations to describe this relationship, arguing that defensive behaviours behave in a similar fashion to enzyme saturation dynamics.

They used ant per capita effectiveness as the quality metric, defined as the number of workers required to remove an experimentally placed Manduca sexta caterpillar on the host plant. The quantity metric was the number of ants that visited the plant and the frequency of visitation (Ness et al. 2006).

They suggest that the quantitative and qualitative components of an interaction are often not independent because any bodyguards beyond those needed to remove all herbivores may be redundant (Ness et al. 2006). Based on this assumption they predict that as the quantity of an interaction increases, the quality will often decline. Consequently, the quantity and quality components are inversely proportional and linked via a non-linear relationship (Ness et al. 2006).

34 Chapter 2: Ant aggressivity

Box 2.1 continued. Aggressivity quantification methods

FAGUNDES Fagundes et al. (2017) quantified ant aggressivity in the most comprehensive study so far, by placing termites (Sintermes sp.) as surrogate herbivores on 20 plants for each of 10 host plant species.

They quantified the protection effectiveness of the ant community within the whole plant community in a riparian grassland, producing detailed quantitative outputs with a network approach that included every ant species associated with every plant species.

Aggressivity was quantified as the product of a qualitative and a quantitative component. The qualitative metric was based on ant behaviour and effects on the termites. They recorded defence as successful in their assay when ants found, captured or repelled the termite, and a failure when the ants found but ignored the termite. The quantitative component was the inverse of the ant response time, multiplied by the repellence efficiency within that essay. Since they placed multiple termites per assay, they could produce a repellence efficiency estimate per assay. PE = QNC * QLC PE = protection effectiveness QNC = number of ants recruited QLC = 1/time * repellence efficiency

Based on the QNC and QLC components, ant species were classified into four functional groups: a) highly effective protectors: high recruitment and quick removal of herbivores on most plant species b) low effective protectors: high recruitment and quick removal of herbivores on a few plant species c) ineffective protectors: low recruitment, low detection and removal of herbivores d) Non-protectors: no recruitment, no detection or removal of herbivores.

35 Chapter 2: Ant aggressivity QUESTIONS & HYPOTHESES

Here I tested and quantified the aggressivity of six species within the ant community associated with T. velutina at La Mancha, Veracruz using larvae of the natural herbivore Euptoieta hegesia (Lepidoptera), to answer the following questions:

(i) Do ants vary in their aggressivity displays and in their ability to repel herbivores (protection effectiveness)? If so, which are the most aggressive ant species?

(ii) Does any observed variation in aggressivity correlate with previous observations of ant-herbivore interactions for the same or related ant taxa?

My null hypothesis was that ant species do not differ in their aggressivity display or protection effectiveness.

MATERIALS AND METHODS Study site All experiments and observations were conducted at La Mancha, Veracruz in coastal sand dunes within the CICOLMA Field Station in the Gulf of Mexico (19˚36’N, 96˚22’W, elevation <100 masl).

Study system Turnera velutina (Passifloraceae) is a Mexican endemic perennial shrub (Arbo 2005) that is also a myrmecophile (Cuautle and Rico-Gray 2003) that grows in lowland coastal environments. It establishes a facultative mutualism with up to 13 ant

36 Chapter 2: Ant aggressivity species (Cuautle et al. 2005, Zedillo-Avelleyra 2017). Extrafloral nectar is provided in paired cup-shaped glands located at the bottom of the leaf blade or on the leaf petiole, on the underside of the leaves. T. velutina is a self-compatible, monomorphic herkogamous species that requires pollinators for adequate seed production (Sosenski et al. 2016). Although it flowers year-round, flowering peaks in summer (Cuautle et al. 2005). Flowers are bisexual, last one day, and are - pollinated (Sosenski et al. 2016). As rewards for pollinators, flowers offer pollen and floral nectar that is easily accessible at the base of the yellow, pentamerous, bell- shaped corolla. The main pollinators at La Mancha are honeybees, and the main herbivores are the caterpillars of the butterfly Euptoieta hegesia (Nymphalidae), although caterpillars of the congeneric butterfly Euptoieta claudia, grasshoppers (Acrididae) and hemipterans also feed on the plants (Cuautle and Rico-Gray 2003, Villamil-Buenrostro 2012, Ochoa-López et al. 2018). The fruits are dry septicidal capsules that dehisce to expose the seeds, which are dispersed by ants. The seeds bear an elaiosome rich in fatty acids and lipids, which needs to be removed to allow imbibition and germination (Ochoa-López 2013). Myrmechochorus ants swarm onto newly open fruits, carry the seeds towards their nests and detach the elaiosome for consumption (Cuautle et al. 2005, Salazar-Rojas et al. 2012). Anti-herbivore assays Larvae of Euptoieta hegesia were reared and maintained in greenhouse conditions under mesh exclusions feeding on Turnera velutina plants from multiple ontogenetic stages (Rebollo Hernández 2018). For the aggressivity assays, larvae were kept in 0.5 L plastic containers with a tulle mesh window lid for ventilation. Containers were lined with wet filter paper and fresh T. velutina foliage was provided for food and shelter. The frontal width of the head capsule of larvae was measured and larvae were classified in three size groups: small (1.58±0.054 mm), medium (1.77±0.027 mm), and large (2.77±0.019 mm). Only large larvae were used in aggressivity assays.

37 Chapter 2: Ant aggressivity

In the field, adult plants patrolled by a given ant species were selected haphazardly and the numbers of leaves, buds, flowers, and fruits of a haphazardly selected branch were counted. The species identity and number of patrolling ants on that branch was also recorded. One larva was placed on a vegetative leaf (i.e. not associated with any reproductive structure) of the selected branch and the behavioural response of ants and the larva was observed for 10 minutes. As ant behaviours I recorded evasion, attending, recruiting, attacking and repelling behaviours, and the time to these events. As larval behaviours I recorded relocation (moving to another part of the plant) and drop-off (falling / launching themselves from the plant), and the time to these events, as described below.

Ant behaviours:

o Ignore / Not find / Avoid: Did the ants ever reach the leaf with the larva? If found, did ants ignore the larva? In ignore / fail to find behaviours ants neither respond to the larva nor do they remain on the leaf with the larva. Avoidance behaviours are those in which ants find the larva but then move away from the leaf or branch with the larvae and do not return during the assay, and neither patrolled nor recruited more ants to the same leaf or branch. (Categorical (yes/no) variable)

o Response time: How long after it was placed was the larva found by ants? (Continuous variable)

o Patrol: Once the larva has been found, do ants patrol around it or on the leaf bearing it? (Categorical (yes/no) variable)

o Recruitment: Did the first ant to find a larva recruit more ants? If so, then the number of ants recruited. (Continuous variable)

38 Chapter 2: Ant aggressivity

o Attack: Did the ants attack the larva? Attack can be either by bites, direct harassment (such as antennation or stinging), or by spraying the larvae through the acidopore (see further details in Results). (Categorical (yes/no) variable)

o Time to first attack: How long after the larva was placed on the plant was it attacked? (Continuous variable)

o Repelling: Did the ants cause the larvae to abandon the leaf or the plant? (Categorical (yes/no) variable)

Larval behaviours:

o Relocation: Did the larva abandon the leaf where it was initially placed and moved to another leaf within the same plant?

o Time to relocation: How long after placement did the larva relocate? o Drop-off: Did the larvae drop itself off the plant as a consequence of ant interference?

o Time to drop-off: How long after placement did the larva drop-off? o Foraging: Was the larvae observed feeding on any plant structure?

If a larva was attacked on one assay, they were not used again in another assay that same day. If they were not attacked after an assay, larvae were used again for a maximum of four assays in a single day. After a trial, larvae were placed back into their individual containers at the greenhouse. Larval containers were washed and the wet lining was changed every night, foliage was replenished as required. These assays were conducted between 22 October – 4 November, 2016. All statistical analyses were done in R software. Mixed models were fitted using the ‘lme4’ (Bates et al. 2016) and MCMCglmm (Hadfield 2010), and post-hoc model comparisons were fitted using the ‘multcomp’ package (Hothorn et al. 2008), unless otherwise specified.

39 Chapter 2: Ant aggressivity QUANTIFYING AGGRESSIVITY

I quantified ant aggressivity and protection effectiveness by assessing ant responses to larvae placed on their host plant by evaluating several metrics in each assay. These metrics were the maximum aggressivity score displayed per assay, the likelihood of each species to display each of the different behaviours defined in response to a larva on their host plant, the number of ants recruited, the time to first response and the time to herbivore repellence. Furthermore, I combined several of these metrics into a Protection Effectiveness Index, allowing us to evaluate and compare the aggressivity display and anti-herbivore efficiency of the different ant species tested.

Maximum aggressivity score I assigned each behaviour a number between 0-5, in rank of increasing apparent aggressivity: 0 – Ignore / Not find 1 – Find 2 – Patrol 3 – Recruit 4 – Attack 5 - Repel Based on this score, I assigned every assay a score reflecting the maximum level of aggressivity displayed by the patrolling ants. The effect of ant species identity on the maximum aggressivity score was analysed using a Poisson mixed model, with maximum aggressivity score as the response variable and ant species identity fitted as a fixed effect. Plant identity and assay identity were fitted as random effects, along with an observation-level random effect (OLRE) to account for overdispersion (Hinde 1982). Post hoc Tukey model comparisons were used to test for significant differences between all species.

40 Chapter 2: Ant aggressivity Likelihood of aggressive behaviours The effect of ant species identity of the likelihood with which ants displayed each of the aggressive behaviours recorded when encountering a larva on their host plant, was evaluated using binomial mixed models. These likelihoods are hereafter referred to as efficiencies for each behaviour recorded, such as attacking or patrolling efficiency.

Species varied greatly in their likelihood of behaviours, with some always displaying a given behaviour whilst others never displayed it, causing an extreme category problem (Hauck-Donner effect) in this data for recruiting, attacking, and repelling behaviours. Because of this, I analysed these models using the ‘MCMCglmm’ R package (Hadfield 2010), fitting binomial Bernoulli models with a ‘probit’ link function. I fitted one model for each behaviour independently (ignore/not find, patrol, recruit, attack, repel). I fitted the presence or absence of each behaviour as the response variable and ant species as a fixed effect. Plant identity and assay identity were fitted as random effects. Wald tests were used as post hoc contrasts to detect significant differences between species. However, I decided to plot only the mean from the raw data, as standard errors and confidence intervals from the model outputs are extremely wide due to the Hauck-Donner effect. Ant recruitment The effect of ant species on the number of ants recruited per assay was analysed using a Poisson mixed model. The number of ants was fitted as the response variable, and ant species as a fixed effect. Plant identity, assay identity and OLRE were fitted as random effects. Post hoc Tukey comparisons were used to detect significant differences between species in the number of ants recruited.

41 Chapter 2: Ant aggressivity Response time The effect of ant species on the ants’ response time per assay was analysed using a Poisson mixed model. For those assays in which ants did not ignore the larva, the number of seconds after which ants found the larva was fitted as the response variable, and ant species as a fixed effect. Plant identity, assay identity and OLRE were fitted as random effects. Post hoc Tukey comparisons were used to detect significant differences between species in the response time.

Repellence time The effect of ant species on the time it took for a larva to move away from the leaf where it was originally placed was analysed using a Poisson mixed model. Repellence included larvae relocating themselves to another leaf, or larvae dropping-off entirely from the plant. For those assays in which ant interference lead larvae repellence, the number of seconds after which the larvae was repelled was fitted as the response variable, and ant species as a fixed effect. Plant identity, assay identity and OLRE were fitted as random effects. Post hoc Tukey comparisons were used to detect significant differences between species in the time to repellence.

Protection effectiveness Ievaluated the Protection Effectiveness (PE) of different ant species by creating an index combining several of the previously described defensive metrics as follows: = ×

= 𝑃𝑃𝑃𝑃 𝑄𝑄𝑄𝑄𝑄𝑄 𝑄𝑄𝑄𝑄𝑄𝑄 1 = 𝑄𝑄𝑄𝑄𝑄𝑄 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 ×𝑜𝑜𝑜𝑜 (4𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎× 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟+𝑟𝑟𝑟𝑟𝑟𝑟5𝑟𝑟× );

𝑄𝑄𝑄𝑄𝑄𝑄 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 where PE is the protection𝑅𝑅𝑅𝑅𝑅𝑅 𝑅𝑅effectiveness𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 index; QNC is the quantitative component of the index and is the number of ants recruited; QLC is the qualitative part of the index which includes the Response time (the time between larval placement and ant detection), and the attack and repel score combined. Attack and Repel are the

42 Chapter 2: Ant aggressivity Bernoulli variables for presence (1) or absence (0) of either behaviour, each multiplied by their aggressive score, respectively. I decided to include the score for attack and repellence in order to account for differences in attack and repellence efficiency between species, as some species may be aggressive attackers but not effective at repelling larvae from this particular species and instar.

Differences between ant species in their PE score were initially analysed using a Poisson mixed model. The PE index values resulted in a continuous variable between 0-4 that did not follow any usual distribution. Because of this, I recalculated the PE by removing the response time component ( =

× (4 + 5 )) in order to obtain an integer𝑃𝑃𝑃𝑃 𝑁𝑁number𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 that𝑜𝑜𝑜𝑜 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 was 𝑟𝑟𝑟𝑟fitted𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟as𝑟𝑟𝑟𝑟 𝑟𝑟the response𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 variable,𝑅𝑅𝑅𝑅𝑅𝑅 𝑅𝑅and𝑅𝑅 ant species identity was fitted as a fixed effect. The reciprocal of response time (Rt-1) was fitted in the model as a log-transformed offset to account for this variable whilst maintaining Poisson data in the response variable. However, this statistical approach was abandoned as the model could not run, producing unconstrained predicted values that lead to the arising of infinite numbers, despite having a log link function that is intended to constrain predicted values. Probably these problems are inherent to the unusual distribution of the PE data. Hence, I finally analysed the data using non-parametric statistics. I used a Kruskal-Wallis test to determine whether Protection Effectiveness scores differed among ant species. I used Dunn tests as post hoc model comparisons to find significant differences between species pairs.

Ant species were quantitatively ranked according to the values of PE. Furthermore, I quantitatively classified them into four categories depending on their high, low or null values of QNC and QLC (Fagundes et al. 2017). Those ant species with high values of QNC and QLC were considered highly effective defenders and highly aggressive species. Species with low QNC and high QLC were

43 Chapter 2: Ant aggressivity considered as low effective defenders and aggressive species. Species with low QNC and low QLC were considered as ineffective defenders and non-aggressive species. Finally, species with a PE value of zero were considered as parasitic species.

RESULTS

At La Mancha I observed ten ant species interacting with T. velutina from four families and quantified and ranked the aggressivity of the six most frequent species (Table 1).

Table 1. Ants species reported in Turnera velutina at CICOLMA, Veracruz (Zedillo-Avelleyra, 2015) and observed in these studies. Species whose aggressivity was tested are marked with an asterisk (*).

Family Sting Acidopore Species Patrolling habit

Present Dorymyrmex bicolor Gregarious *

rinae (slit) Vestigial Dolichode Camponotus planatus Loner *

Camponotus claviscapus Not observed Present

Camponotus mucronatus Loner - (circular) Camponotus novogranadensis Loner - Absent

Formicinae Brachymyrmex sp. Gregarious * Paratrechina longicornis Gregarious *

Myrmicocrypta sp. Not observed

Present Monomorium ebenimum Gregarious - Cephalotes Loner/few * retracted Present but but Present Crematogaster Gregarious *

Pseudomyrmex gracilis Loner - Absent Pseudomyrmex sp.1 Not observed well Present meciinae Pseudomyrmex sp.2 Not observed Pseudomyr

44 Chapter 2: Ant aggressivity Quantifying aggressivity Ant species differed significantly in their recruiting, attacking and repelling efficiencies, but not in their herbivore finding efficiency or patrolling likelihoods (Table 2). Behavioural differences were very consistent between some species, with certain species always displaying a given behaviour, whilst others never displayed it. For instance, Cephalotes ants never attacked the larvae, whilst Dorymyrmex bicolor always did (Figs. 1,2).

Figure 1. Frequency of assays with each of the maximum aggressive behaviour scores per ant species, when encountering a larva on their host plant.

Cephalotes sp. ants were the least aggressive ants, never observed attacking or repelling any of the larvae, and ignoring larvae in 25% of the essays (Table 2; Figs, 1-2). Even when they found the larvae, they only patrolled around the leaf and were observed recruiting in only one assay (Fig. 1).

45 Chapter 2: Ant aggressivity Except for Cephalotes, all ant species recruited fellow ants in all assays where they found the larva (Fig. 1,2, Table 2). Highly aggressive species prone to attacking the larvae found were not always the most efficient larvae repellers (Fig. 2, Table 2). Further details regarding behavioural observations of ants are available in the Supplementary Natural History Notes: Ants. Although C. planatus were efficient at larvae repellence, repelling 56% of larvae they found (Fig. 1) their repellence efficiency predicted by the model was only 32% because they seem to be very inefficient at detecting the larvae on the plant they are patrolling, ignoring or failing to find 23% of the larvae. Hence, they ranked lower on the maximum aggressivity score (Table 2). Paratrechina longicornis had an average finding efficiency of 87%, and often engage in aggressive behaviours with the larvae found. Yet, with a repellence efficiency of only 15% this species is the least efficient at repelling larvae among ant species found to have this effect (excluding Cephalotes) (Table 2, Fig. 1). Crematogaster sp. had an average finding efficiency of 82%, and always engaged in aggressive behaviours once they found the larvae. Despite attacking the larvae in 75% of the assays, they had a low repellence efficiency of 54% (Table 2). Brachymyrmex sp. had an average finding efficiency of 87%, and almost always engage in aggressive behaviours with the found larvae (88%), and attacking the larvae in 87% of the assays, resulting in a very high repelling efficiency of 81% of the larvae found. Dorymyrmex bicolor ants were the most efficient species at larvae detection, finding 100% of the larvae placed on their plant, and always engaging in some sort of aggressive defensive behaviour, and attacking the larvae in 98% of the assays. They were also the most efficient at repellence, repelling 95% of the caterpillars found.

46 Chapter 2: Ant aggressivity

Figure 2. Per species probability of ants displaying different behaviours when encountering a larva on their host plant. Dots represent the proportion of the data (positive/total events), the P value displayed indicates the significance of the effect of species on each behaviour, and different letters indicate significant differences between species (P < 0.05).

47 Chapter 2: Ant aggressivity

Table 2. Model estimates for the effect of ant species on the likelihood of ants displaying different defensive behaviours.

Fixed Random Posterior Lower Upper Response N DIC X2 P-value effects effects mean 95% CI 95% CI -05 Ignore/Not Ant species 93 70.10 3.21 0.66 Assay 8.1 8.37 27.6 find Plant 1.73 2.82-07 7.80 Patrol Ant species 93 101.89 8.20 0.14 Assay 5.37 4.14-06 33.81 Plant 5.19 9.81-05 30.59 Recruit Ant species 93 90.11 21.71 0.0005 *** Assay 0.98 4.26-08 4.46 Plant 1.58 1.23-05 5 Attack Ant species 93 86.78 19.36 0.001 *** Assay 2.41 2.04-07 8.93 Plant 4.51 0.006 15.38 Repel Ant species 93 79.58 13.74 0.01 * Assay 2.79 0.0001 11.38 Likelihood of behaviours Plant 7.43 0.44 21.09

48

Chapter 2: Ant aggressivity Ant recruitment, response and repellence time I found significant differences between ant species in the number of ants recruited after finding a Euptoieta hegesia caterpillar on their host plant (Fig. 3a, Tables 3 & 5). Cephalotes had the smallest ant recruitment, recruiting on average less than an ant per assay, contrasting with Brachymyrmex ants which recruited on average 24 ants (Table 3 & 5, Fig. 3). Based on the post hoc Tukey contrasts I can split ant species into three groups with significant differences in the number of ants recruited towards herbivores: high recruiters (Brachymyrmex sp. Dorymyrmex sp. Crematogaster sp. and Paratrechina longicornis: >5 ants), low recruiters (Camponotus planatus: <5 ants), and non-recruiting species (Cephalotes sp.: < 1 ant) (Tables 3, Fig. 3a). This and the following two analyses on response and repellence time only included ants which did engage in any response; those that ignored or did not find the larvae were excluded from these analyses, but accounted for in the analyses regarding likelihood of behaviours and protection effectiveness estimates.

Ant species differed significantly in their response time (Tables 3 & 5), with the ant community being split into two groups: slow responders, finding larvae in >20 sec. (Cephalotes, Camponotus planatus, Crematogaster sp. and Paratrechina longicornis), and fast responders finding larvae in <20 sec. (Brachymyrmex sp. and Dorymyrmex sp.) (Fig. 3b). The time it took for larvae to be repelled also differed significantly between ant species (Table 3 & 5), with the ant community being split into three significantly different groups: Cephalotes ants who were never seen repelling a larva; slow repelling species managed to repel a larva in >120 sec. (Camponotus planatus, Crematogaster sp. and Paratrechina longicornis), and fast repelling species that achieved this in <120 sec (Brachymyrmex sp. and Dorymyrmex sp.) (Fig. 3c).

49

Chapter 2: Ant aggressivity

Figure 3. a) Ant recruitment, b) response time and c) time to larval repellence times for different ant species after encountering a larva placed on their host plant. Different letters indicate significant differences between species (P < 0.05).

50 Chapter 2: Ant aggressivity

Table 3. Model predictions for the number of ants recruited, ants’ time to first response and larval time to repellence after ant interference in assays for different ant species interacting with Turnera velutina.

Ant recruitment Response time Repellence time Ant species Estimate SE Prediction Estimate SE Prediction Estimate SE Prediction (ants) (sec) (sec) Cephalotes sp. -0.29 0.44 0.74 ± 1.5 4.54 0.46 93.76 ± 1.58 Camponotus planatus 1.44 0.33 4.22 ± 1.4 3.72 0.36 41.35 ± 1.43 5.16 0.26 175.69 ± 1.29 Paratrechina longicornis 2.26 0.39 9.58 ± 1.47 3.29 0.35 27.08 ± 1.43 3.42 0.40 30.58 ± .149 Crematogaster sp. 1.67 0.47 5.31 ± 1.61 3.92 0.39 50.65 ± 1.49 5.06 0.31 158.06 ± 1.37 Brachymyrmex sp. 3.18 0.40 24 ± 1.5 2.97 0.32 19.67 ± 1.37 3.79 0.28 44.60 ± 1.32 Dorymyrmex bicolor 2.43 0.39 11.35± 1.48 2.84 0.33 17.14 ± 1.39 3.79 0.24 44.32 ± 1.27

51

Chapter 2: Ant aggressivity Collinearity plots between aggressivity variables I explored the correlation between different metrics used to assess the aggressivity displays and protection effectiveness of the ant species associated with T. velutina using Pearson correlations. The number of ants recruited was negatively and exponentially correlated with response time (Fig. 4a). Assays with the greatest number of ants recruited also occurred in a relatively shorter period of time, suggesting that the high recruiters are also fast recruiters (Fig. 4a).

Ignoring efficiency (the likelihood to ignore or not find a larva) was positively and significantly correlated with response time (r = 0.33, t(5,77)= 3.07, P = 0.002), so ants more likely to ignore or fail to find the larva also had longer response times (Fig. 4b). Repellence efficiency was significantly and negatively correlated with response time (r = -0.35, t(5,77)= 3.35 , P < 0.001) and with repellence time (r =

-0.32, t(5,36)= 2.07, P = 0.008), implying that more efficient repelling ants were faster at finding the larva and removing it in a reduced period of time (Fig. 4c,f). Repellence efficiency was significantly and positively correlated with the number of ants recruited (r = 0.23, t(5,90)= 2.33, P = 0.02), implying that more ants are more efficient at repelling a given herbivore (Fig. 4d). Repellence efficiency was also positively and significantly correlated with attacking efficiency (r = 0.89, t(5,91)= 19.08, P < 0.0001), implying that attacking efficiency is in fact an excellent predictor of ant taxon defensive quality and efficiency at repelling herbivores. However, I also found that some species may be very prone to attack but have lower repelling efficiency than less aggressive species, as is the case for Paratrechina longicornis and Camponotus planatus ants, and hence I decided to include both attack and repelling behaviours in the aggressivity and protection effectiveness index, aiming to obtain a more comprehensive metric of both aspects.

52

Chapter 2: Ant aggressivity

Figure 4. Correlations between the different aggressivity and defensive variables assessed for the ant community interacting with T. velutina. For significant correlations, the Pearson correlation coefficient (r) and the associated P-value are indicated. a) Response time vs. number of ants recruited; b) Probability of ignoring a larva vs. response time; c) Repellence efficiency vs. response time; d) Repellence efficiency vs. number of ants recruited; e) Repellence efficiency vs. attacking efficiency; f) Repellence efficiency vs. time to repellence.

53 Chapter 2: Ant aggressivity Maximum aggressivity score Ant species differed significantly in their maximum aggressivity score (Table 5) allowing me to split the ant community into two groups: defensive and non- defensive ants, with only Cephalotes in the latter category (Fig. 5f, Table 4-5).

Protection Effectiveness Index The quantitative (QNC) and qualitative (QLC) components used to calculate PE were not significantly correlated (r = 0.19, t(5,77)= 1.77, P = 0.08; Fig. 5a). The

2 protection effectiveness index differed significantly between ant species (X (5, 92) = 23.68, P = 0.0002). This index split the ant community into two groups, here termed defenders and non-defender species, with Cephalotes identified as the only non- defender ant species (Fig. 5b). Ant species were classified according to their high or low QNC and QLC values as follows: Dorymyrmex bicolor and Brachymyrmex ants had high QNC and QLC values, and hence were classified highly effective defenders and highly aggressive species (Fig. 5c). Paratrechina longicornis had a high QNC but low QLC value, and hence was classified as a low aggressive and low effective defender, whilst Camponotus planatus and Crematogaster ants were considered ineffective defenders. Finally, Cephalotes ants were identified as parasitic species because of their null QNC and QLC values.

54 Chapter 2: Ant aggressivity Table 4. Model predictions of maximum aggressivity score for different ant species interacting with Turnera velutina.

Maximum aggressivity Ant species Estimate SE Prediction (ants) Cephalotes sp. 0.16 0.23 1.18 ± 1.26 Camponotus planatus 1.06 0.13 2.90 ± 1.14 Paratrechina longicornis 1.14 0.15 3.14 ± 1.16 Crematogaster sp. 1.29 0.17 3.64 ± 1.18 Brachymyrmex sp. 1.38 0.14 3.99 ± 1.16 Dorymyrmex bicolor 1.53 0.13 4.65 ± 1.47

a) b)

c) d)

Figure 5. Metrics used to assess the aggressivity levels and protection effectiveness of different species associated with T. velutina. a) Correlations between the quantitative and qualitative components per assay; b) Protection effectiveness index; c) Average quantitative and qualitative protection effectiveness components per species; d) Maximum aggressivity score.

55 Chapter 2: Ant aggressivity

Table 5. Model estimates for the effect of ant species on recruitment, repellence, and protection effectiveness metrics.

Fixed Random Response N AIC LRT P-value Variance effects effects

Ants Ant species 92 650.4 37.60 4.53-07 *** Assay 8.05-09 recruited Plant 0.50 OLRE 1.07 Response Ant species 79 816.3 16.69 0.005 ** Assay 0.91 time Plant 0.12 OLRE 0.16 Repellence Ant species 38 431.1 18.28 0.001 ** Assay 0.49 time Plant 0.00 Recruitment and time and Recruitment OLRE 0.11 -06 -09

Maximum Ant species 93 376.4 34.23 2.13 *** Assay 5.05 aggressivity Plant 0.02 OLRE 4.81-09 Protection Ant species df X2 = 23.68 0.0002 *** Effectiveness 5,92 Aggressivity

56

Chapter 2: Ant aggressivity

DISCUSSION

For six ant species associated with T. velutina I estimated the likelihood of ants ignoring, patrolling, recruiting, attacking and repelling a caterpillar of Euptoieta hegesia placed on their host plant. I used two metrics to assess the aggressivity and defensive level of the ant species: the maximum aggressivity score and the protection effectiveness index. Both metrics converged in the same result, splitting the ant community into defender and non-defender species, with the only non-defender species being Cephalotes sp. A more detailed picture of the protection and aggressivity of each species was achieved using the qualitative and quantitative components classifying species into four categories: highly effective defenders and highly aggressive species (Dorymyrmex bicolor and Brachymyrmex ants), lowly aggressive and lowly effective defender (Paratrechina longicornis), ineffective defenders (Camponotus planatus and Crematogaster ants), and parasitic species (Cephalotes).

How to quantify aggressivity? Overall, Fagundes et al. (2017) proposed that the number of plants visited (interaction frequency), ant recruitment (patrolling intensity) and the rate of herbivore removal (aggressivity), but not the time spent by ants foraging on the plant (foraging time), are important components of ant foraging behaviour and regulate protection effectiveness of ant species. I agree with this, except that I consider that time to first response or time to finding is a crucial aspect of protection effectiveness, as voracious herbivores can cause severe damage in a short period of time. Hence, I think this should be part of the protection effectiveness metric.

In contrast to Fagundes et al. (2017), I did not evaluate assays as an overall success or failure, but evaluated the presence or absence of each behaviour to produce a more comprehensive picture of ant aggressivity modes and frequency

57

Chapter 2: Ant aggressivity within the species associated with T. velutina. Fagundes et al. (2017) had multiple termites per plant, and so could estimate the removal percentage per assay and then average such estimates per species. I did not have enough larvae to do this, and so I produced an estimate per species for each behaviour. This is a limitation of this study; however, an advantage of this study is being the first to use the main natural herbivore of the ant-plant, and that it provides a comprehensive picture of ant defensive and patrolling behaviour.

Ness et al. (2006) suggest that the quantitative and qualitative components of an interaction are often not independent, and hence as the quantity of an interaction increases, the quality will often decline. In particular, they suggest that excessive bodyguards beyond those needed to remove all herbivores may be redundant, and so quantity and quality are inversely proportional (Ness et al. 2006). However, I am unsure about this statement and think it may be related to the way these authors defined it. It is possible to envision very effective mutualists that are also very frequent and abundant, and the opposite scenario, with very infrequent and scarce mutualists that are also very inefficient, and a linked scenario in which per capita effectiveness is low, but mutualists are very abundant and overall, are a very effective species due to large recruitment and frequencies.

When quantifying the aggressivity of defensive mutualists, it is crucial to account for differences in the behavioural ecology of each species, reflected in differences in patrolling habits. Ness et al. (2006) showed that relatively few highly aggressive ants can provide similar levels of protection to milder but more abundant ants, particularly if the milder ants are more likely to discover the herbivore due to a higher patrolling frequency. I agree with Ness et al. (2006) in that the interaction frequency (patrolling frequency) is a crucial factor to be accounted for. Patrolling frequency estimates are likely to change the interpretation of PE values, because an average guard may become highly

58 Chapter 2: Ant aggressivity efficient if it patrols very frequently, being likely to find herbivores much faster than a highly aggressive species, with infrequent and lone patrolling habits.

Del‐Claro and Marquis (2015) showed that ant species that recruited more foragers were also those that visited more shrubs (higher frequency of interaction) and more shrub species. Yet there was no difference in the time to find the herbivore between species that interacted with a high or low number of plants. These findings lead to the following questions: • Is per capita protection effectiveness a better or worse estimate than the PE metric used here? • Is the number of ants required to repel a larva an important factor in ant-based defence?

Given that ants are social organisms, if the colony is big and the ants are capable of quickly recruiting more ants, then the number of ants required to repel a larva becomes a less important factor. I think the PE estimate should include time, which may be a more crucial factor than a per capita estimate as voracious herbivores can eat a flower or a leaf in less than 10 minutes. Furthermore, constant patrolling by a mild ant species may also deter adults from ovipositing on the plant. Ant patrolling is a key factor in deterring oviposition by E. hegesia on more heavily patrolled plants of T. velutina (Ochoa- López et al. 2018). Older T. velutina plants rely more heavily on biotic defences, whilst younger stages rely on physical and chemical defences (Villamil et al. 2013, Ochoa-López et al. 2015) and butterflies oviposited more frequently on younger stages, with less nutritional quality but also with lower ant patrolling (Ochoa-López et al. 2018).

59 Chapter 2: Ant aggressivity Which are the most aggressive ant species? Behavioural differences were very consistent between some species, with certain species always displaying a given behaviour, whilst others never displayed it (Figs. 1,2).

Despite finding efficiencies (the inverse of ignoring rates) between ant species ranging from 75-100%, no significant differences were detected between species (Fig. 2, Table 5). The same pattern was true for patrolling efficiency, which ranged from 35-100% but no significant differences were detected between species. This suggests that even very infrequent patrolling levels by poor defenders are sufficient for ants to notice a herbivore or plant disturbance.

Recruitment differed significantly between species and the efficiency range between species was even greater, from 1% in Cephalotes, to 96-98% in D. bicolor and Brachymyrmex ants, respectively. Recruitment efficiencies were significantly different for every single species, suggesting this may be a species- specific trait, although further research investigating whether the recruiting efficiency of a given ant species varies depending on the host plant it occupies would shed light on plant traits driving ant behaviour.

The likelihood of attacking displayed by ant species was not always directly correlated with their repellence efficiency (Fig. 4e). This was an interesting and useful finding that helped discriminate between the aggressive and the defensive efficiency traits of ant species. Incapacitated Manduca sexta larvae were either killed or knocked off F. wislizeni plants by the guarding ants (Ness et al. 2006). This same behaviour was observed in T. velutina, where guarding ants would induce E. hegesia larvae to relocate or drop-off the plant, and killed the larvae that did not do so.

Cephalotes ants were inefficient finders, the slowest species to find the larvae, with low ant recruitment and no evidence of aggressive behaviours,

60 Chapter 2: Ant aggressivity leading to attacking and repelling efficiency levels of zero (Figs. 1-2). Cephalotes had a null PE index and had the lowest maximum aggressivity score (Fig. 5). All these traits suggest this is a parasitic species that harvests EFN without providing any defensive services. These findings in T. velutina are consistent with previous evidence of other Cephalotes species being non-defenders on Ouratea spectabilis Engl. (Ochnaceae) (Byk and Del-Claro 2010) and Banisteriopsis malifolia (Alves-Silva et al. 2015) in the Brazillian cerrado, and across many host plants within a riparian grassland community in Brazil (Fagundes et al. 2017). Del‐Claro and Marquis (2015) also suggested that the weak protective levels in plants with a mixed array of ant species was due to the predominance of Cephalotes pusillus ants in those ant mixes. C. pusillus are ineffective defenders against herbivores consuming pollen and nectar without reciprocating with defensive services for the plant (Byk and Del-Claro 2010). Overall, I suggest that Cephalotes may be a parasitic that exploits the host plants without providing any guarding services, although additional evidence is required to confirm this hypothesis.

In T. velutina, C. planatus ants did attack herbivores, biting and spraying the larvae in the assays, suggesting that in this associations they are guarding ants, despite their low protection effectiveness (Figs. 1, 2). However, in Acacia mayana, Camponotus planatus ants were never seen providing any protective service, leading to the conclusion that they were true exploiters (Raine et al. 2004). Yet, non-aggressive ants can still reduce herbivory rates through non- consumptive effects or cleaning services, such as egg removal, although these roles as plant cleaners and egg removers have yet to be tested.

The maximum aggressivity score and the PE index both allowed us to discriminate between defensive and non-defensive species, and detect Cephalotes as a parasitic species (Fig.4), with both metric showing similar degrees of resolution. However, these metrics did not have enough resolution

61 Chapter 2: Ant aggressivity to rank defensive species according to their aggressivity and protection effectiveness. Splitting the PE index into its quantitative and qualitative components allowed a higher resolution analysis of ant species (Fig. 5c).

Although PE index and the maximum aggressivity score could only split the ant community into defender mutualists and cheating species, they are useful metrics because they produce a quantitative estimate. Such numerical measurement of the protective benefits provided by each ant species can later be used in other analyses, investigating the drivers and consequences of such ant behavioural responses. However, mapping the quantitative (QNC) and qualitative (QLC) components of ant aggressivity displays produced a more detailed description of individual species protection effectiveness. Despite being qualitative and not producing a concrete numerical output, the cartesian QNC- QLC mapping generated a hierarchy of ant aggressivity and defenders’ effectiveness, discriminating between highly efficient, lowly efficient, ineffective defenders and parasitic species (Fig. 5a,c).

Based on all the metrics and behaviours described here, I conclude that the hierarchy of ant aggressivity and protection effectiveness for T. velutina at La Mancha, Veracruz is as follows: Dorymyrmex bicolor > Brachymyrmex sp. > Paratrechina longicornis > Camponotus planatus > Crematogaster sp. > Cephalotes sp..

Are ant traits good predictors of aggressivity? Attack mode: sting vs. spray It has been suggested that ants that sting are less aggressive and inferior defenders than those that deploy chemical defences (Davidson et al. 1988). However, this does not seem to be the case in Acacia mayana where spraying Camponotus planatus ants aggressively outcompeted Pseudomyrmex

62 Chapter 2: Ant aggressivity ferrugineus stinging ants (Raine et al. 2004). As plant defenders this does not seem to the case either amongst the guarding partners of T. velutina, since biting attacks (by Dorymyrmex or Brachymyrmex) resulted in higher larvae repellence efficiency levels than spraying attacks (by Crematogaster or Camponotus ants).

Patrolling habits I suggest that significant differences in the numbers of ants recruited revealed by the assay with Euptoieta hegesia larvae may reflect inherent differences in behavioural ecology and patrolling habits among ant taxa. Despite C. planatus ants having a higher herbivore repelling efficiency than P. longicornis (Fig. 2), the latter ranked higher in the Protection Effectiveness index (Fig. 5b) by having a higher recruitment size (Table 4, Fig. 3). I acknowledge that the PE index calculated here may underestimate lone patrolling species with low recruitment sizes. However, lone patrollers will visit plant sections less frequently because fewer ants will take longer to patrol a whole plant than a numerous group of patrolling ants. If this is true, this PE index may be correctly reflecting the quality of ants as defenders by penalizing infrequent patrollers who may take longer to find a herbivore. In ants associated with Peixotoa tomentosa (Malpighiaceae), species with higher recruitment numbers also had higher frequencies of interaction (visited more shrubs) and from a larger number of species. Yet there was no differences in the time to find the herbivore between species that interacted with a high or low number of plants (Del‐Claro and Marquis 2015). However, this data do not have enough resolution to show a robust pattern on the correlation between the likelihood of ants finding a herbivore and their response time. Having one herbivore per assay per plant, I can only estimate finding efficiency (1-ignorance) at a species level, and not per assay. Other studies using termites as herbivores instead of the natural herbivores managed larger replicate sizes and efficiency estimates per assay (Fagundes et al. 2017). Although studies with this experimental design sacrifice on natural history realism, they achieve more replication and increased

63 Chapter 2: Ant aggressivity statistical resolution and would be an ideal follow-up experiment to answer whether lone patrollers are less efficient at finding herbivores than gregarious ant species.

Ant size Small ants have often been regarded as poor defenders (Fiala et al. 1989, Cuautle and Rico-Gray 2003, Cuautle et al. 2005). Although the defensive effectiveness of Crematogaster borneensis was questioned mostly based on their small size and their lack of a functional sting, worker ants were observed biting intruders and displayed very strong attaching abilities with high repellence efficiency (Fiala et al. 1989). In T. velutina (Cuautle and Rico-Gray 2003) also suggested that larger ants with larger mandibles would provide greater defensive benefits to the plants, but this hypothesis was not supported by my quantitative results of aggressivity. Dorymyrmex bicolor and Brachymyrmex sp. ant species had the smallest body sizes but the highest aggressivity displays and repellence efficiency levels (Figs. 2, 4, 5). I advocate avoiding a priori judgements on the ant traits that may drive their protection effectiveness, and instead suggest quantitative assessments using surrogate or naturally encountered herbivores.

Protection effectiveness varies between hosts Ant recruitment and herbivore removal efficiency are key aspects in determining the protection effectiveness of ants, and these traits have been thought as intrinsic to each ant species, but recent studies show the opposite. The protection effectiveness of a given ant species is likely to differ depending on the host plant it is occupying (Apple and Feener Jr 2001, Fagundes et al. 2017). Defensive attributes can be up- or downregulated by external factors such as the quality of rewards provided (Grover et al. 2007, Fagundes et al. 2017). In a comprehensive community study on the effect of individual ant species on different host plants, Fagundes et al. (2017) found that ant protection was positively associated with sugar concentration of nectar, which explained the

64 Chapter 2: Ant aggressivity variation in leaf damage among plant species. This demonstrates that ant protection varies among ant species and is enhanced by the plant investment in nectar reward (Fagundes et al. 2017).

Plants providing better nectar quality are likely to interact with more aggressive and territorial ants, reducing the frequency of commensal and parasitic species (Grover et al. 2007, Koptur et al. 2015, Fagundes et al. 2017) and such investment may result in plant benefits via reduction in damaged leaf area (Fagundes et al. 2017). Since increased ant aggressivity increases the reproductive success of the host plant, it may be advantageous for plants to produce more sugar-rich nectar to attract more and better ant guards. Although this may be limited, as more abundant and aggressive partners may also result in ecological costs such as repellence of other natural predators like (Cuautle and Rico-Gray 2003), coccinellid beetles, or spiders (Koptur et al. 2015), and may also deter pollinators (Ness 2006, Assunção et al. 2014). Despite this, the effects of EFN availability, sugar, prey and other resources on ant foraging patterns have rarely been studied (but see: Apple and Feener Jr 2001, Grover et al. 2007, Fagundes et al. 2017).

The more, the merrier Herbivore traits, such as behaviour or morphology may affect their susceptibility to attacks. Ants are generally more effective against small, and less sclerotised herbivores, with caterpillars being more susceptible than beetles (Fiala et al. 1989). In Macaranga plants, Crematogaster borneensis was more effective at deterring small, non-sclerotised animals, such as caterpillars, opposed to beetles. In T. velutina the main herbivores are the larvae of Euptoieta butterflies, which are not sclerotised and hence are susceptible to ant attacks. Curculionidae and Staphylinidae beetles have also been recorded as florivores in T. velutina (pers. obs.) and these beetles do not seem to be susceptible to ant

65 Chapter 2: Ant aggressivity attacks as their abundance did not vary between control and ant excluded plants (See Natural History Notes for further details). Their highly sclerotised bodies make them highly resistant to ant biting or acid spraying attacks. Nonetheless, Pseudomyrmex gracilis ants have been seen actively and exclusively hunting Staphylinidae beetles and thrips from the flowers (pers. obs., further details in Natural History Notes). If ant species differ in their effectiveness against different kinds of herbivores, an apparently redundant array of patrolling ant species as those in facultative mutualisms may be beneficial for plants.

ACKNOWLEDGEMENTS

Thanks to Roberto Rebollo who kindly donated the larvae colony for the aggressivity experiment, to J. Hadfield who provided valuable inputs towards the statistical analyses. Thanks to R. Ríos who helped helped out during fieldwork.

.

66

CHAPTER 3

ANT-POLLINATOR CONFLICT

Nectar trade-offs and pollinator deterrence

The weekend naturalist By Tom Buchan

My humanoid friend, myself, a limited animal In love with the planet Escaping across the dumb topographies of Assynt With maps and a compass Taking incorrect fixes on anonymous Bens Staring into bog-pools

Entertaining myself with half-formulated notions of a non-utilitarian character and applying my ragbag of ecological data to flowers which I recognize absentmindedly as if they were old friends whose names I’ve forgotten […]

67

Chapter 3: Ant-Pollinator conflict INTRODUCTION

Extrafloral nectaries, domatia and food bodies are all means by which ant-plants (comprising facultative species termed myrmecophiles and obligate ones termed myrmecophytes) (Rosumek et al. 2009; Del-Claro et al. 2016) attract and support ants by providing nesting sites or nutrients (Rico-Gray & Oliveira 2007; Rosumek et al. 2009). In return, ants attack herbivores, prune climbing vines and prevent fungal and microbial infestation on plant tissues (Bentley 1977; Rosumek et al. 2009). This mutualistic interaction is termed myrmecophily.

Interactions involving two or more types of mutualists of a single host are common in nature, but multispecies interactions are much less studied than pairwise and intraguild mutualisms (Strauss 1997; Tscharntke & Hawkins 2002; Strauss & Irwin 2004; Adler 2008; J. et al. 2009; Melián et al. 2009; Koptur et al. 2015). To date, most research on plant-animal interactions has focused on pairwise relationships (e.g., plant-herbivore, plant-pollinator, plant-fungus) in isolation from the community in which they are embedded (Strauss 1997; Herrera 2000; Dáttilo et al. 2016; Del-Claro et al. 2018). This pairwise approach necessarily oversimplifies reality (Herrera 2000) since plants interact sequentially or simultaneously with each of pollinators, herbivores, herbivore predators and pathogens, (Armbruster 1997). Furthermore, plant interactions with one partner or guild can also affect relationships with other groups or guilds (Armbruster 1997) and alter outcomes from mutualistic to antagonistic (Strauss 1997; Strauss et al. 1999; Herrera 2000; Strauss & Irwin 2004; Del-Claro et al. 2016). As a result, a growing number of studies are focusing on multispecies and multitrophic interactions (Melián et al. 2009; Fontaine et al. 2011; Nahas et al. 2012; Pineda et al. 2013; Dáttilo et al. 2016). It might be expected, for example, that the presence of predatory ants can influence pollinators, with top-down effects on plant fitness. This makes ant-plants, which rely on ants for defence against herbivores and on pollinators for seed set seeds, a model tritrophic

68 Chapter 3: Ant-Pollinator conflict system in which to explore the dynamics of multispecies and multitrophic interactions.

Here, I focus on disentangling ant-pollinator interactions that occur when both mutualists share a host plant. Previous work has revealed evidence of ant-pollinator conflict in such systems (Yu & Pierce 1998; Stanton et al. 1999; Gaume et al. 2005; Ness 2006; Palmer & Brody 2007; Frederickson 2009; Stanton & Palmer 2011; Malé et al. 2012; Assunção et al. 2014; LeVan et al. 2014; Hanna et al. 2015). Ant-pollinator conflict is expected for two main reasons. Firstly, both mutualists share with their host interest in different plant functions. Pollinators benefit from plant resource allocation to reproduction (i.e., flowers, floral nectar (FN) and pollen), whilst ants benefit from allocation to growth and defence (i.e., vegetative structures bearing extrafloral nectar (EFN) or domatia) (Yu & Pierce 1998; Frederickson 2009; Palmer et al. 2010). This could derive in a conflict mediated by plant rewards, known as indirect conflict. Floral and extrafloral nectar share sugar as a common currency, providing potential for a trade-off and also a means of quantifying investment in each. Secondly, because as ant guards actively defend their host plant as a means of protecting food and/or nesting sites, they may also repel or attack pollinators (Ness 2006; Stanton & Palmer 2011; Chamberlain & Rudgers 2012), and this drawback of patrolling is known as direct conflict.

Castration is an extreme example of direct ant-pollinator conflict in which guarding ants destroy or consume the reproductive meristems, floral buds or flowers of their host plant (Yu & Pierce 1998; Stanton et al. 1999; Gaume et al. 2005; Palmer & Brody 2007; Frederickson 2009; Malé et al. 2012). Such castrating behaviour inevitably reduces availability of flowers and hence floral rewards for pollinators. It has been suggested that the ultimate cause of castration by patrolling ants is that it promotes reallocation of plant resources from reproduction to growth (Yu & Pierce 1998; Frederickson 2009; Malé et al.

69 Chapter 3: Ant-Pollinator conflict 2012), and hence increasing the availability of resources on which ant colonies depend. In ant species that are obligate inhabitants of ant-plants, colony size is limited by the number of domatia (Fonseca 1993; Fonseca 1999; Orivel et al. 2011), which is positively correlated with plant investment in growth. Resource allocation strategies towards these two mutualists should be approached in a linked way, as plants interact with both mutualists simultaneously and the presence of one mutualist may increase or decrease the presence of the other. Also, because investment in one may come at the cost on investment in the other, affecting plants investment in the other mutualists reward, positively via linkage, or negatively via trade-offs.

Even in those species that do not castrate their host, ants’ aggressive behaviours might threaten and deter pollinators, compromising plant reproduction (Ness 2006; Assunção et al. 2014; LeVan et al. 2014). Avoidance of such direct conflict has been suggested to explain plant architecture or behaviours that reduce spatial (Raine et al. 2002; Malé et al. 2015; Martínez- Bauer et al. 2015) or temporal overlap of ant guards and pollinators (Gaume & McKey 1999; Gaume et al. 2005; Nicklen & Wagner 2006; Ohm & Miller 2014; Malé et al. 2015), and repelling compounds which exclude ants from flowers when pollen is released (Junker et al. 2007; Willmer et al. 2009; Ballantyne & Willmer 2012).

Extrafloral nectar is a key resource mediating multispecies interactions in many plant communities, and plants bearing extrafloral nectaries comprise up to a third of species in some biomes (Dyer & Phyllis 2002; Rudgers & Gardener 2004; Davidson & Cook 2008; Dyer 2008), particularly in tropical dry forests, savannas and cerrados (Rico-Gray & Oliveira 2007; Assunção et al. 2014). The importance of ant-plants as food resources for mutualists in a given plant community is enhanced if these plants also secrete FN and pollen for pollinators. Management of ant-pollinator conflict in such a way that the crucial

70 Chapter 3: Ant-Pollinator conflict services provided by both mutualist groups are maintained is thus likely to be part of the adaptive landscape of many plant species. Ant-plants with extrafloral and floral nectaries represent an excellent system to test for trade-offs in resource allocation, competition amongst mutualistic guilds, and assess plant strategies that minimise direct and indirect conflicts between their mutualists. To my knowledge, no study has addressed both direct and indirect ant- pollinator conflict in a single ant-plant system.

QUESTIONS & AIMS

Here I tested for direct and indirect ant-pollinator conflict on a Mexican endemic ant-plant, Turnera velutina. In particular, I assessed whether Turnera velutina reduces the potential for conflict through the daily timing of FN and EFN release. I also tested for potential indirect (nectar-mediated) and direct (deterrence) conflicts between ants and pollinators. I addressed the following specific questions:

(i) What are the daily timings of nectar reward secretion, ant activity, and floral visitation?

(ii) Does the presence of patrolling ants deter pollinators from the flowers? (Direct conflict)

(iii) Do ant species vary in their impacts on potential deterrence to pollinators? (Direct conflict)

(iv) Are T. velutina plants able to re-allocate extrafloral nectar resources into floral nectar resources (increasing reward availability to flower visitors)? (Indirect conflict)

71 Chapter 3: Ant-Pollinator conflict MATERIALS AND METHODS Study site Field experiments were conducted in coastal sand dunes at the CICOLMA Field Station in La Mancha, Veracruz, in the Gulf of Mexico (19°36’ N, 96°22’W, elevation < 100 m). The climate is warm sub-humid, with a rainy season during the summer (June to September), an annual precipitation of 1100 – 1500 mm, and a mean annual temperature ranging between 24 and 26 °C (Travieso-Bello & Campos 2006). Experiments were carried out in November 2014 at four sites with high densities of Turnera velutina (Passifloraceae). Greenhouse experiments were conducted in a shade house at CICOLMA.

Study system Turnera velutina is an endemic perennial shrub (Arbo 2005) and myrmecophile (Cuautle & Rico‐Gray 2003). At La Mancha, T. velutina is patrolled by at least seven ant species (Cuautle et al. 2005; Zedillo-Avelleyra 2017) and its main herbivores are the specialist caterpillars of the butterfly Euptoieta hegesia (Nymphalidae). Extrafloral nectar is provided in paired cup-shaped glands located at the bottom of the leaf blade at the junction with the leaf petiole, on the underside of the leaves (Fig. 1a, Elias et al. 1975; Villamil et al. 2013). Although it flowers year-round, flowering peaks during summer (Cuautle et al. 2005). Flowers last one day, are animal-pollinated (Sosenski et al. 2016) and provide FN at the base of the corolla (Fig. 1). Honeybees (Apis mellifera) are the dominant pollinators at La Mancha, accounting for 94% of the visits (Sosenski et al. 2016) and collect both pollen and floral nectar, but the role of other floral visitors as effective pollinators is yet to be investigated. There is no spatial segregation of patrolling ants and floral visitors in Turnera velutina since flower buds emerge from the axillary meristems of leaves bearing extrafloral nectaries (Villamil 2017). Nor are they segregated in daily time, since, EFN volume and sugar content both peak during anthesis (Villamil 2017).

72 Chapter 3: Ant-Pollinator conflict

Figure 1. Flower of Turnera velutina (top) with arrows indicating the location of the floral nectaries at the base of the corolla, between the petals, and leaves (bottom) with arrows indicating the location of extrafloral nectaries, on the underside of the leaves.

FIELDWORK METHODS Mutualist activity curves: patrolling ants and pollinators I quantified daily activity of patrolling ants and flower visitors on T. velutina by surveying flowers (n=120 plants, n=1604 flowers;) and their associated leaves at all four La Mancha sites in November 2014. I observed all open flowers for 2 minutes every hour throughout anthesis (0800–1300 hours), with one observer at each site. Every hour, I counted the total number of floral visitors and ants patrolling extrafloral nectaries across all flowers within a site. I sampled the same sites over multiple days. Since these are one-day flowers, I considered each site-day as a replicate (n = 10 site-days; 43.23 ± 2.89 flowers/site-day), and incorporated site-and-day effects into the statistical modelling (see below).

Nectar secretion and pollen deposition curve Nectar secretion and pollen deposition data were collected from 4 sites within CICOLMA over 5 consecutive days in September 2015. I visited a single site per day (with one exception which was visited twice) and sampled FN and EFN secretion rate and pollen deposition from 1 flower per plant for 5 plants per site (n=20 plants). Flowers were sampled every hour during the anthesis period

73 Chapter 3: Ant-Pollinator conflict (0800-1300 hours). Flowers were bagged with tulle bags before anthesis and FN and EFN were collected every hour during anthesis. The first collection was taken as soon as the corolla was fully open. Flowers were re-bagged between measurements to avoid nectar consumption and a masking tape band with Tanglefoot was applied on the stem below the flower to exclude ants for the duration of the experiment. FN was extracted using a 1 μl capillary inserted into each of the five nectaries in a single flower (Minicaps Disposable capillaries, Hirschmann Laborgerate, In 20°C, ISO 7550, R<0.5%, CV<1.0%, Germany). EFN was also collected from the glands using 1 μl capillaries. Nectar volume was measured using a digital calliper (Mitutoyo Digimatic) and sugar concentration was determined in °Brix (g sucrose per 100 g solution) using a 0-50°B hand held refractometer (Reichert, Munich, Germany). To obtain all of the sugar from extrafloral nectaries, the glands were washed with 2 μl of deionized water using a 0-5 µl micropipette and the sugar concentration of the wash was quantified using the refractometer. Total sugar content in FN and EFN was calculated from volume and °Brix values according to Comba et al. (1999) using the formula: where s is the sugar content (µg), v is the nectar or wash volume (µl), and d is the density of a sucrose solution at a concentration C (g of sucrose per 100 g solution) as read on the refractometer. The density was obtained according to Comba et al. (1999) using the formula:

A different flower from each of these twenty plants was chosen and marked, but never bagged. I collected one stigma every hour, to sample the pollen deposited one, two and three hours post-anthesis. A small proportion of flowers contained a fourth stigma, contributing to a small dataset for four hours post anthesis. Stigmas were stored individually in labelled Eppendorf tubes. Stigmas were individually mounted on slides for fresh squash glycerin preparations. Slides were sealed using nail polish and kept in a fresh and dry environment at 22°C. Each slide was labeled with the site, day, hour, and plant identity. The number

74 Chapter 3: Ant-Pollinator conflict of pollen grains on each stigma was counted under a microscope, and changes in numbers at each time interval plotted to generate the pollen deposition curve.

Direct conflict To test whether non-ant flower visitors detect and avoid flowers with ants, visitation was observed in flowers with and without dead ants for 4 flowers on each of 40 plants (n=160 flowers). Plants > 80 cm in height and with at least four flowers, were haphazardly selected and flowers bagged at 0700 before anthesis. Once the corollas had opened, three ant corpses from a single ant species (either Dorymyrmex bicolor, Brachymyrmex sp., or Paratrechina longicornis) were placed inside each of three flowers / plant. A fourth flower was left as an ant- free control. Ant corpses were placed on the inner surface of the petals in each flower. These species were chosen because they were amongst the most abundant species (see Results), displayed patrolling behaviours on T. velutina, and were detected as potentially differing in their effects on Apis mellifera pollinators by a previous study (Villamil et al. submitted). Additional flowers were removed from the plant to control for floral display across all individuals. Ants were killed in 50% ethanol, which was allowed to evaporate for 30 minutes at ambient temperature (28-35°C) from all samples to prevent ethanol vapours from influencing pollinator behaviour. Flowers containing dead ants were observed for 20 minutes, recording pollinator identity, visit frequency and duration, and associated pollinator behaviours. Observations were conducted during October-November 2016 with four simultaneous observers working collecting data from different plants.

When assessing the effects of ants inside the flowers on pollinator visitation I only considered visits by Apis mellifera, since this is the dominant T. velutina pollinator in this population (Sosenski et al. 2016) and accounted for 91% of the visits in this experiment. I classified honeybee behaviours as ‘inspection’ or ‘contact’. Inspection behaviours are defined as those displayed by the pollinators at the floral headspace before or without landing on the flower comprised either

75 Chapter 3: Ant-Pollinator conflict approaching or hovering over a particular flower. Contact behaviours were those that occurred inside the flower, between landing and take-off, and comprised foraging on pollen or nectar resources or standing on the petals, anthers or stigmas.

Indirect conflict To test possible trade-offs in plant resources between FN and EFN I conducted a greenhouse experiment on 72 plants from 18 different maternal lines (2-4 siblings per maternal family, generated from field individuals). Plants were kept under greenhouse conditions in a shade house located within CICOLMA field station. Plants were grown in 1L plastic pots with a substrate of local soil and vermiculite (50:50) and watered every other day (for rearing details see: Ochoa- López et al. 2015). During the experiment, plants were watered every night, and extrafloral nectaries were sprayed with water to wash away any nectar secretion from previous days and to prevent fungal infections on the glands. Pre-anthesis buds were bagged either the night before or during the morning before anthesis to prevent any nectar theft by unexpected insects that may occasionally enter the nursery.

Pairs of maternal siblings were chosen and randomly assigned to either control or experimental groups. The extrafloral nectaries in all leaves of experimental plants were clogged by applying Mylin transparent textile paint in the nectary cup, as in Villamil et al. (in press). Extrafloral nectaries from control plants were left intact and droplets of the same paint were applied on the leaf blade above the glands to match any unintended effects on plants across treatments. Very young floral buds were marked, and their extrafloral nectaries clogged from their emergence, throughout their development, and until anthesis. The droplets on the extrafloral nectaries or leaf blades were checked daily and replenished when required, especially when a new leaf emerged in order to guarantee uniform and continuous application of the clogging treatment across all leaves.

76 Chapter 3: Ant-Pollinator conflict

FN secretion was measured before the clogging treatment was applied, and once again after the young clogged/marked buds became flowers. The aim was to test whether flowers that were unable to secrete EFN invested more sugar resources in floral nectar. FN was collected from control and treatment flowers between 1300 and 1500 h using a 1µl microcapillary pipette. Nectar volume and total sugar mass was estimated and calculated as described above.

STATISTICAL METHODS All statistical analyses were conducted in R version 3.23 (R Core Team 2016). All mixed effects models were fitted using the ‘lme4’ R package (Bates et al. 2016) and post hoc Tukey comparisons were fitted using the ‘multcomp’ R package (Hothorn et al. 2008), unless stated otherwise.

Mutualist activity and reward secretion curves To test whether ant or pollinator activity changed over daily time, I fitted a Poisson mixed model with either the number of ants patrolling or the number of floral visitors as the response variable. I fitted time of day as a fixed effect, with linear and quadratic terms to detect non-linear activity patterns over time. The number of flowers per site was fitted as a log-transformed offset to control for floral display, since I recorded visitor counts per site rather than counts per individual flower (see fieldwork methods). Flowers of T. velutina last for a single day, and because multiple flowers were sampled on a given site on a given day, I fitted site identity as a random effect to account for differences between site and day variation in variables that could influence ant abundance, such as resource availability, ant diversity, or the abundance/proximity of ant nests. I also included an observation-level random effect (OLRE) where each data point receives a unique level of a random effect to control for overdispersion (Hinde 1982).

77 Chapter 3: Ant-Pollinator conflict To test if FN and EFN secretion changed over the anthesis period I fitted a Poisson mixed model independently for each nectar type, using the sugar mass (µg) value rounded to the nearest integer as the response variable. Nectar sugar content is usually estimated in µg, and I report raw data in such units. However, to facilitate model convergence I re-scaled the response variable (sugar content) from µg to g, and rounded it to the next integer to better fit a Poisson distribution. I fitted time of day as a fixed effect, with linear and quadratic terms to detect non-linear activity patterns over time. I fitted plant identity as a random effect, and, included an observation-level random effect.

Timing of daily activity and secretion peaks I computed the time at which mutualist activity and nectar secretion reached their maximum by calculating the time at which the slope (i.e., the differential of the fitted model with respect to time) for each variable is zero and then solving for hour, as follows:

Direct conflict The effect of different ant species on the visitation frequency was tested using Poisson mixed effects models. I fitted the number of visitors as the response variable, and ant species inside the flower (Control, Dorymyrmex bicolor, Brachymyrmex sp., or Paratrechina longicornis) was fitted as a fixed effect. Because these are one day flowers, plant-day identity was chosen as a random effect to control for individual variation in floral and extrafloral nectar investment. I also included an observation-level random effect. Post hoc Tukey comparisons were used to test differences in visit duration between the four treatments.

78 Chapter 3: Ant-Pollinator conflict

I tested whether ant species inside the flower differed in their effect on the likelihood with which a pollinator displayed an inspection behaviour using a binomial mixed model. The presence or absence of inspection behaviours was coded as the response variable and ant species was fitted as a fixed effect. As random effects I fitted the plant-day identity, and the visitor identity. For those visits where the pollinator displayed an inspection behaviour, I fitted the proportion of time per visits spent displaying inspection behaviours using a Gaussian mixed model with Logit transformation for data normality. Ant species was included as a fixed effect, and plant-day identity, and the visitor identity were fitted as random effects.

Finally, differences in the duration of inspection or contact behaviours in flowers containing different ant species were analysed using Gamma mixed models. Mixed effects models were fitted independently for each behaviour, but using the same model structure fitting visit duration per flower as the response variable. Ant species inside the flower was fitted as a fixed effect. Plant-day identity was chosen as a random effect to control for individual variation in floral and extrafloral investment, and daily weather variations. I also included an observation-level random effect. Post hoc Tukey comparisons were used to test differences in visit duration between the four treatments.

Indirect conflict For each plant, I estimated the difference in FN produced before and after the extrafloral nectaries were clogged as follows: .

Where DFN is the difference, PostFN is the FN production after the extrafloral nectaries secretion was prevented and PreFN is the FN production before the extrafloral nectaries were clogged. Differences in volume and sugar content between control and experimental flowers were tested using mixed effects models. Both variables had normal distributions and so Gaussian mixed effects models were fitted using the same model structure: The clogging treatment was

79 Chapter 3: Ant-Pollinator conflict fitted as a fixed effect, and the maternal family was fitted as a random effect to independently explain variation in both nectar volume and sugar content.

RESULTS Mutualist activity, reward secretion, and pollen deposition curves Activity curves show that both patrolling ants and floral visitors were most active within the first two hours post-anthesis (Fig. 2), although visitation by potential pollinators peaked on average over an hour before ant activity (9 min post-anthesis for potential pollinators, 90 min post-anthesis for ant patrolling; Table 1).

On average, a flower and its associated leaf secreted a total of 2815 ± 767 µg of sugar via floral and extrafloral nectar throughout the 4.5 hour anthesis period. The total sugar content in FN was 149 ± 19.3 µg of sugar, whilst total extrafloral sugar was 2665 ± 765 µg. Thus, the relative sugar contributions of floral and extrafloral nectar in a leaf-flower module were 5.3% and 94.7%, respectively.

Floral and extrafloral nectaries of T. velutina are both able to quickly replenish nectar after experimental removal by non-destructive sampling of the same flowers over the entire anthesis period (Fig. 2a). FN and EFN are secreted simultaneously during anthesis and with the highest amount of sugar content secreted during the first two hours of anthesis (Fig. 2a), although their secretion peaks predicted from model estimates are slightly mismatched (Table 1). EFN secretion peaked 68 minutes after the first collection, which was taken as soon as flowers were fully open, whilst FN peaked 23 min after the first collection. The timing of peaks in secretion of the two types of nectar matches that for the mutualists that harvest each resource. Peak pollen deposition occurred at the beginning of anthesis and steadily declined over time (Fig. 2c). Hence, pollen

80 Chapter 3: Ant-Pollinator conflict deposition data were analysed using a linear model without fitting hour as a squared term and I did not estimate timing of daily maxima using model derivations (Table 1).

I recorded 1535 ant visitors from nine ant species patrolling extrafloral nectaries of T. velutina at CICOLMA (Table 2). Dorymyrmex bicolor, Paratrechina longicornis, and Brachymyrmex spp. accounted for 68.5% of the total ants observed, and 77.35% of the patrolling ants, after excluding Monomorium spp. that were never observed displaying patrolling behaviours on T. velutina and are mostly parasitic consumers of FN and EFN (lestobiotic) (Ettershank 1966; Bolton 1987).

81 Chapter 3: Ant-Pollinator conflict

a) b)

c) Figure 2. Timing of a) reward secretion (n = 20 flowers-leaves), b) mutualist activity (n = 1604 flowers; n = 10 sites/day) and c) pollen deposition (n = 20 flowers) in Turnera velutina during the anthesis period (08:30 – 12:30), showing on the right raw data from field observations (mean ± se). Red circles show floral nectar and the activity of pollinators in flowers, whilst green triangles show extrafloral nectar and the activity of ants at extrafloral nectaries.

82 Chapter 3: Ant-Pollinator conflict Table 1. Model statistics for the timing of mutualists activity, reward secretion and pollen deposition in Turnera velutina. Estimates at which the maxima for mutualist activities and reward production are reached are shown in the last three columns. Peak hour is the time in hours estimated, mpa shows minutes post-anthesis when the maxima is reached, and time of day indicates an approximation of when that activity is likely to occur, although this varies depending on the season and the time of sunrise.

Maxima estimates Fixed Random Peak mpa Time Response N Estimate LRT P-value Variance SD effects effects hour of day Patrolling Log(flowers) 33 0.804 7.967 0.004 *** OLRE 0.221 0.471 ants Hour 33 0.088 0.148 0.699 Site ID 0.462 0.680 1.5 90 09:30 (number) Hour2 33 -0.028 0.201 0.653

activity Floral visitors Log(flowers) 42 0.507 2.619 0.105 OLRE 0.097 0.311 Mutualist (number) Hour 0.040 0.036 0.849 Site ID 0.991 0.991 0.15 9 08:09 Hour2 -0.131 3.838 0.05 *

Floral sugar Hour 78 0.1445 0.1019 0.7495 OLRE 0 0

0.39 23 08:23 (µg) Hour2 78 -0.1818 1.3222 0.2502 Plant ID 0 0

Extrafloral Hour 78 0.4647 1.0070 03156 OLRE 0.2425 0.4925

Reward 1.13 68 09:08 secretion sugar (µg) Hour2 78 -0.2053 1.8949 0.1686 Plant ID 0.4013 0.6335

Pollen OLRE 0.1049 0.3240 Hour 44 0.3240 21.38 3.73-06 *** deposition Site ID 0.0957 0.3094 (number of grains)

83 Chapter 3: Ant-Pollinator conflict Direct conflict I recorded 991 floral visitors, of which 907 (91.5%) were by the honeybee Apis mellifera. Of the remaining 8.5%, 61 visits were by native bees, 11 by flies (Diptera), 10 visits by Lepidoptera, one visit by a beetle (Coleoptera) and one by a wasp () (Table 2).

Neither the presence of ants inside flowers nor their identity had any significant effect on the number of honeybees visiting the flowers (Fig. 3a, Table 3). The presence of the most aggressive ant species, Dorymyrmex bicolor, increased the likelihood of a pollinator displaying inspection behaviours by 20% (Fig. 3b), and increased by 12% the proportion of time per visit spent displaying inspection behaviours rather than foraging or pollinating the flower (Fig. 3c). Finally, the presence of Dorymyrmex bicolor and Paratrechina longicornis inside the flowers halved the duration of contact visits compared to control flowers without ants, or to flowers with Brachymyrmex sp. ants inside.

Indirect conflict Clogging extrafloral nectaries on the leaves associated with newly emerged floral buds had no effect on their FN volume (LRT (1, 49) = 0.21; P = 0.64, Table 3, Fig.

4b) or sugar content (LRT (1, 49) = 0.087; P = 0.77, Table 3, Fig. 4a). Differences in

FN volume and sugar content (FNpost-treatment – FNpre-treatment) were positive in plants under both control and clogged treatments (Fig. 4).

84 Chapter 3: Ant‐Pollinator conflict

Figure 3. The effect of dead ants inside the flowers of Turnera velutina on different aspects of the behaviour of visiting honeybees: A) visitation frequency, B) the probability of displaying inspection behaviours, C) the duration of contact visits (time spent inside the flower), and D) proportion of time spent inspecting the flowers per visit bout (hovering over the floral head space). Ant species are arranged in order of increasing aggressivity and names are abbreviated: C = control with no ants, 1P are Paratrechina longicornis, 2M are Brachymyrmex sp., and 3D are Dorymyrmex bicolor ants. (n = 40 plants; n = 160 flowers). Different letters indicate significant differences between species (P<0.05).

85 Chapter 3: Ant-Pollinator conflict

Table 2. Abundance and identity of the ants recorded patrolling extrafloral nectaries during the 2014 census and the floral visitors recorded during the direct conflict experiment in 2016 on Turnera velutina plants.

Visitor Patrolling Taxon Percentage Subfamily s Paratrechina 487 31.72 Gregarious longicornis

Dorymyrmex bicolor 421 27.42 Gregarious Monomorium 270 17.58 Myrmicinae Gregarious ebenium Camponotus 184 11.98 Formicinae Loner planatus Brachymyrmex sp. 73 4.75 Formicinae Gregarious Camponotus 60 3.90 Formicinae Loner mucronatus Crematogaster sp. 23 1.49 Myrmicinae Gregarious Camponotus

Ants at nectaries extrafloral 15 0.97 Formicinae Loner novogranadensis Pseudomyrmex Loner and 2 0.13 Pseudomyrmicinae gracilis very rare

Apis mellifera 907 91.5 Native bees 61 6.05 Diptera 11 1.10 Lepidoptera 10 1 Coleoptera 1 0.1 Floral visitors Floral Wasps 1 0.1

86

Chapter 3: Ant-Pollinator conflict

A) B)

Figure 4. Effect of clogging on nectar secretion and sugar re-allocation in Turnera velutina for nectar volume (A) and sugar content (B), between control (C) and treatment (T) plants (means ± se; n = 50 plants). Values shown are a difference in floral nectar, defined as: [FN post-treatment – FN pre-treatment].

87 Chapter 3: Ant-Pollinator conflict

Table 3. Model statistics for the experiments testing indirect (nectar-mediated) and direct (pollinator deterrence) ant-pollinator conflict.

Fixed Random Response Distribution N LRT P-value Variance SD effects effects

Floral nectar Clogging Gaussian 50 0.0813 0.7754 Family 0.6987 0.8359 (µl) treatment

conflict Floral nectar Clogging Indirect Indirect Gaussian 50 0.21006 0.6467 Family 138.5 11.77 (µl) treatment

e-08 Ants in OLRE 2.77 0.0001 Number of visitors Poisson 95 1.1848 0.7567 flowers Plant 0.144 0.3803

Likelihood of being ID e-10 e-05 Ants in 8.279 2.877 alerted Binomial 373 10.715 0.01337 * visitor flowers 0.595 0.7715 Plant

Proportion of time ID per visit spent Gaussian Ants in 0.0009313 0.03052 112 7.5728 0.055 * visitor displaying inspection (logit) flowers 0.0201 0.14198

Direct conflict behaviours Plant OLRE Duration of presence 0.1202 0.3468 Ants in ID behaviours Gamma 307 392.37 2.2e-16 *** 0.001587 0.03984 flowers visitor (sec) 1.928e-10 0.001389 Plant

88 Chapter 3: Ant-Pollinator conflict DISCUSSION

My findings show that both ants and pollinators are active while flowers are open (Fig. 2b), that FN and EFN are simultaneously secreted (Fig. 2a), and that pollen deposition occurs when ants are actively patrolling (Fig. 2c). Consequently, in T. velutina guarding ants and pollinators operate in close spatial and temporal proximity, implying that direct and indirect conflict could occur in this system. I found, however, no evidence for indirect, nectar- mediated, conflict between ants and pollinators, since plants did not reallocate resources towards floral nectar, even when 95% of the sugar investment during anthesis, which is in EFN, is prevented (Fig 2). I found evidence for direct conflict, as the presence of dead individuals of patrolling ant species inside flowers was associated with both higher frequency of inspection behaviours in potential pollinators, and reduced visit duration (time spent inside flowers) (Fig. 3). Taken together, these effects increased handling time per flower and reduced pollinator foraging efficiency. Nonetheless, this result was obtained under an experimental setting using dead ants placed inside flowers. Further studies are required to test (i) the effect of living patrolling ants on pollinator visitation, and (ii) the impact of any such effects on plant fitness. The latter are crucial to understanding whether there is ongoing selection on Turnera velutina to manage direct ant-pollinator conflict.

Mutualist activity, reward secretion and pollen deposition Floral and extrafloral nectar were secreted simultaneously and rapidly replenished in T. velutina, especially during the first two hours post anthesis (Fig. 2). Replenishment is a general feature amongst EFN secretion (Pacini et al. 2003; Pacini & Nepi 2007; Pacini & Nicolson 2007). In fact, I am unaware of any report documenting extrafloral nectaries incapable of replenishing secretion after consumption (Pacini et al. 2003; Pacini & Nepi 2007; Pacini & Nicolson 2007; Escalante-Pérez & Heil 2012; Orona‐Tamayo et al. 2013; Heil 2015).In contrast, species vary in whether FN is replenished or not (for details on floral

89 Chapter 3: Ant-Pollinator conflict nectar dynamics see: Pacini et al. 2003; Nicolson et al. 2007; Willmer 2011). I suggest that rapid resupply is crucial in short-lived flowers, such as the one-day flowers of Turnera species, because it makes the flower attractive again for another visit, potentially increasing pollen transfer, pollen deposition, and seed set. Interestingly, Dutton et al. (2016) reported no FN resupply in flowers from three congeners (Turnera ulmifolia, Turnera subulata, and Turnera joelii), when sampling FN in the morning and afternoon, finding no FN secretion in the afternoon collection. These observations show either variation in nectar resupply within closely related species displaying similar floral biology, or perhaps suggest that shorter term dynamics in the nectar supply of the other three species were not detected by the sampling methods used. The latter highlights the need to test floral resupply within short time scales, because over long sampling intervals, flowers can both secrete and reabsorb nectar (Pacini et al. 2003; Nicolson et al. 2007; Willmer 2011). Finding no standing crop when a flower is resampled in the afternoon does not rule out replenishment after emptying in the morning, and then reabsorption later in the day (Kearns & Inouye 1993).

Model-based estimation of the daily timing of the maxima of nectar secretion suggest that floral nectar secretion peaks a few minutes after the corolla is fully open (Table 3) and 45 min before peak EFN sugar secretion (Fig. 2a). This represents a slight mismatch in the timing of rewards for ants and pollinators, which may underlie the 85 min mismatch in estimated peaks of ants and floral visitor activity (Table 1). To my knowledge, temporal segregation in the activity of ants and pollinators has been reported only for obligate (myrmecophytic) species patrolled or tended by a single ant species at a time (Humboltia brunonis (Fabaceae; Gaume et al. 2005), Hirtella physophora (Chrysobalanaceae; Malé et al. 2015), and Opuntia imbricata (Cactaceae; Ohm & Miller 2014)). In some specialised systems, ant and pollinator activity occurs in close proximity and simultaneously, but conflict is prevented by ant-repellent

90 Chapter 3: Ant-Pollinator conflict floral volatiles (Vachellia zanzibarica (Fabaceae; Willmer & Stone 1997)) and Vachellia hindsii (Fabaceae; Raine et al. 2002)). On the other hand, temporal overlap in ant activity at extrafloral nectaries and pollinator visitation to flowers has been reported for facultative ant-plants associated with many ant species simultaneously (Vachellia constricta (Fabaceae; Nicklen & Wagner 2006), Acacia myrtifolia, (Acacia sensu stricto, Fabaceae; Martínez-Bauer et al. 2015), and Heteropterys pteropetala (Malpighiaceae; Assunção et al. 2014)). My results add Turnera velutina to the list of facultative myrmecophiles with synchronised ant and pollinator activity (Fig. 2). This synchronous myrmecophile vs. segregated myrmecophyte pattern is consistent with evidence that ants in obligate mutualisms are more aggressive and better defenders (Chamberlain & Holland 2009; Rosumek et al. 2009; Trager et al. 2010), but may impose greater ecological costs on host pollination. I suggest that temporal segregation of mutualists and/or ant repellent floral volatiles are alternative strategies that reduce such costs. Further studies on the timing of pollinator, ant visitation and ant aggressivity in a wider range of systems are required to test the temporal component of this hypothesis.

Direct conflict I showed that dead ants inside the flowers of T. velutina have an impact on honeybee behaviour (Fig. 3). Ant presence was correlated with shorter honeybee flower visits (Fig. 3c), an increase in the proportion of visitors displaying inspection behaviours, and increased duration of inspection behaviours per visiting bout (Fig. 3). I interpret longer inspection behavior to indicate increased caution in the bees (as previously assumed by: Altshuler 1999; Ness 2006; Junker et al. 2007; Assunção et al. 2014; Cembrowski et al. 2014). My findings are consistent with work on Heteropterys pteropetala in which plastic ants inside flowers negatively affected pollination (Assunção et al. 2014). Results for H. pteropetala differ from ours in that the bees pollinating H. pteropetala showed significantly reduced visitation rates to flowers containing plastic ants. In

91 Chapter 3: Ant-Pollinator conflict contrast, honeybees in T. velutina did not visit less frequently flowers containing ant corpses, compared to control flowers (Fig. 3a). In both systems, ants feeding at extrafloral nectaries did not hinder pollination (Assunção et al. 2014) or T. velutina (Villamil et al., unpublished data). This suggests that while pollinators avoid ants in flowers, plants may have evolved other mechanisms to prevent ants from intruding flowers, resulting in ant-pollinators encounters being rare.

Although experiments that place ant cues on flowers can tell us about the response of pollinators to ants, they must be interpreted with caution as an indicator of current ant-pollinator conflict. Firstly, because ants may rarely enter flowers (Villamil et al. submitted). Secondly, by placing such ant cues in flowers I may be violating existing ant-excluding or ant-repelling plant mechanisms (Willmer 2011). Thirdly, in contrast to such experimental treatments, ants do not naturally remain constantly in the flowers for long periods (Assunção et al. 2014), and only a low proportion of flowers may be occupied at any one time. For instance, in T. velutina only 10% of the flowers are occupied by ants (Villamil et. al., submitted).

The effect of ants’ deterrence can vary: Does anti-herbivore match anti-pollinator? Although some studies have documented variation among ant species in aggression towards herbivores (Ness 2006; Miller 2007; Ohm & Miller 2014), little is known about the effect of different patrolling ant species with varying levels of aggressivity on pollinator visitation (Ness 2006; Miller 2007; LeVan et al. 2014; Ohm & Miller 2014). Nonetheless, a positive correlation between the level of defence provided and the level of pollinator deterrence they exert has often been assumed since ant traits involved in defence (patrolling activity and aggressivity) are likely to be the same as those involved in pollinator deterrence (Ohm & Miller 2014). Bees, like any animal, tend to forage in a way that maximises the net benefit of each foraging trip (Stephens & Krebs 1986; Jones 2010; Cembrowski et al. 2014). When foraging in ant-plants, this benefit might

92 Chapter 3: Ant-Pollinator conflict be maximised if foragers avoid flowers or patches where predation risk is high (Dukas 2001; Dukas & Morse 2003; Ness 2006; Jones & Dornhaus 2011; Assunção et al. 2014), as could be the case when encountering ant species that attack pollinators. Some ants also consume FN and pollen, and such plants may represent high risk foraging environments with low net rewards for pollinators (Ness 2006). Shorter of fewer visits to such flowers may be a pollinator strategy to maximise foraging efficiency by avoiding flowers, plants, or patches with high predation risk (Jones & Dornhaus 2011).

In T. velutina, the most aggressive ant guard, Dorymyrmex bicolor, had the strongest effect on pollinator behaviour (Fig. 3), while Brachymyrmex sp. ants inside the flowers did not reduce the duration of pollinator visits. The least effective anti-herbivore ant species, Paratrechina longicornis (Villamil unpublished data), halved the duration of pollinator visits (Figs. 3). In Ferocactus wislizeni, plants tended by Solenopsis xyloni, the most aggressive ant species, had fewer and shorter pollinator visits (Ness 2006). Such differences are consistent with pollinator sensitivity to ant aggressiveness. In contrast, although ant exclusion in Opuntia imbricata significantly increased pollinator visitation, there were no differences in impacts associated with different ant species (Ohm & Miller 2014), and no evidence that the more aggressive guard (Liometopum apiculatum) had a stronger deterring effect on pollinators (Ohm & Miller 2014). Mapping whether the level of ant aggressivity towards herbivores correlates positively with the ecological costs on pollination via pollinator deterrence remains unknown (but see: Ness 2006; Miller 2007; LeVan et al. 2014; Ohm & Miller 2014), and should be tested, not assumed.

Indirect conflict My experimental approach found no evidence for a trade-off in sugar investment in extrafloral and floral nectar in T. velutina. I conclude that there is no indirect nectar-mediated conflict between guarding ants and pollinators in

93 Chapter 3: Ant-Pollinator conflict Turnera velutina, and that pollinators do not obtain greater rewards when rewards for patrolling ants are eliminated. I found only two previous studies testing indirect, nectar-mediated ant- pollinator conflict by quantifying sugary rewards (FN and EFN) to both mutualists (Chamberlain & Rudgers 2012; Dutton et al. 2016). Previous work on other Turnera species by Dutton et al. (2016) found evidence of a trade-off in two of the three Turnera species tested; removing EFN decreased FN and vice versa in T. ulmifolia and T. subulata, but not in T. joelii. Interestingly, both Turnera species in which trade-offs were detected by Dutton et al. (2016) invested equally in FN and EFN, whilst T. joelli (which showed no trade-off) invested more in EFN (Dutton et al. 2016). The same pattern holds for T. velutina, a species with an asymmetric investment towards EFN, which accounts for 95% of the sugar allocation per leaf-flower module, and where I found no trade-off or resource reallocation from EFN to FN (Fig. 4). Unfortunately data on FN and EFN volume and sugar content was not reported for the cotton species (Chamberlain & Rudgers 2012).

One possible reason for lack of a trade-off is that sugar is not a limiting resource for the plant. If so, there would be no reason to expect dynamic reallocation. Estimates of the metabolic costs of nectar secretion vary, and while some studies suggest low metabolic costs (O'Dowd 1979: EFN accounts for 1% of the total energy invested per leaf), others indicate investment of up to 37% of daily photosynthesis in floral nectar (Southwick 1984; Pyke 1991). A second reason, which applies in particular to comparative cross-species analyses rather than experimental manipulations, is that investment in both forms of nectar may be influenced by other aspects of life history strategy. Chamberlain and Rudgers (2012) found no significant negative correlations between extrafloral nectary and floral traits in a comparative analysis across cotton (Gossypium) species, and correlations were significantly positive in 11 of 37 cotton species. Foliar extrafloral nectary volume was positively associated with plant

94 Chapter 3: Ant-Pollinator conflict investment in floral nectar, rejecting the hypothesis of trade-offs among investments in pollinators versus bodyguards in Gossypium. Several potential mechanisms underlying the positive correlations between FN and EFN have been proposed, including pleiotropy, genetic, physiological or ecological linkage (Chamberlain & Rudgers 2012). The pleiotropy or genetic linkage hypothesis could be tested using genome sequencing (Chamberlain & Rudgers 2012). Positive correlations could also arise from physiological or ecological linkage. Traits such as FN and EFN may be physiologically linked. However, the fact that in Gossypium FN volume was most strongly correlated with foliar EFN volume, but FN was weakly correlated with bracteal EFN volume (Chamberlain & Rudgers 2012) questions the physiological linkage hypothesis since bracteal and floral nectaries are spatially closer than floral and extrafloral nectaries, but they are not strongly correlated. I suggest that lack of any trade-off could also indicate that FN and EFN may be phenotypically integrated as a functional module for mutualist attraction. Although formal analyses are required to test this hypothesis, I think it is a strong possibility since T. velutina leaves are phenotypically integrated modules in which leaf economics, defensive and morphological traits covary and are ecologically linked (Damián et al. 2018). Whatever the drivers of these positive correlations may be, available evidence suggests that plants may experience fewer investment trade-offs among different functional traits than previously assumed.

CONCLUSIONS

To my knowledge, trade-offs between extrafloral and floral nectar traits have been studied in 41 species from two genera: 37 Gossypium species (Chamberlain & Rudgers 2012), and four Turnera species, including this study (Dutton et al. 2016; Villamil 2017, Fig. 2). Negative correlations or evidence for trade-offs have been found in only two of these species: T. ulmifolia and T. subulata (Dutton et al. 2016), representing less than 5% of the species studied. Although many more studies are required to shed light on quantitative trends of floral and extrafloral

95 Chapter 3: Ant-Pollinator conflict investment in plants, trade-offs between floral and extrafloral seem infrequent. On the other hand, evidence of direct conflict with patrolling ants reducing pollinator visitation frequency and duration, inducing inspection behaviours and increasing foraging time has been widely reported (Rudgers & Gardener 2004; Ness 2006; Chamberlain & Rudgers 2012; Malé et al. 2012; Assunção et al. 2014; Koptur et al. 2015; Malé et al. 2015; Martínez-Bauer et al. 2015). I suggest that these two issues are not isolated, and hypothesise that positive correlations between FN and EFN investment in ant-plants may be a plant strategy to compensate or lure pollinators to apparently risky flowers.

ACKNOWLEDGEMENTS

Thanks to S. Ochoa-López who kindly donated the plants for the greenhouse experiment, X. Damián-Domínguez who provided insightful thoughts for the experimental design and to J. Hadfield who provided valuable inputs towards the statistical analyses. Thanks to the volunteers who helped out during fieldwork: A. Bernal, A. Martínez, R. Ríos, I. Silva, and A. Fournier, to the staff at CICOLMA station, to H. Köck who helped processing pollen samples, and to R. Pérez-Ishiwara for logistic support.

I am the sole author of all the text used here, with comments on earlier drafts from Graham Stone and Karina Boege.

This chapter has been published:

Villamil et al. 2018. Ant-Pollinator conflict results in Pollinator deterrence but no nectar trade-offs. Frontiers of Plant Sciences. 14 Aug, 2018.

96

Chapter 5: The Distraction Hypothesis of EFN

CHAPTER 4

ECOLOGICAL AND EVOLUTIONARY COSTS OF ANT PATROLLING

How costly are guarding ants for pollination?

“Bees and humans alike have their criteria for selection: symmetry and sweetness in the case of the bee, heft and nutritional value in the case of the potato-eating human.”

“The bees! The bees will let themselves be lured into the most ridiculous positions, avidly nosing their way like pigs through the thick purple brush of a thistle, rolling around helplessly in a single peony’s blonde Medusa thatch of stamens- they remind me of Odysseus’s crew in thrall to Circe.”

Michael Pollan The Botany of desire

97

Chapter 4: Ecological costs INTRODUCTION

Mutualisms are traditionally defined as interactions from which both partners benefit. However, mutualisms involve a combination of costs and benefits (Begon et al. 1986, Bronstein 2001, Herrera and Pellmyr 2009). Hence, mutualisms should be viewed as reciprocal exploitation relationships where the benefits outweigh the costs (Begon et al. 1986). Identifying and quantifying costs and benefits are essential to understanding the magnitude and direction of the selective pressures acting upon the participants (Herre et al. 1999, Ohm and Miller 2014) and the evolutionary stability of the interaction (Palmer et al. 2010). In order to address these costs, we need to tackle mutualisms with a broader approach, moving away from pairwise species interactions and into a multispecies community approach, considering several interactions simultaneously (Strauss 1997, Herrera 2000, Dáttilo et al. 2016, Del-Claro et al. 2018).

Patrolling ants in ant-plant mutualisms defend plants from herbivores, pathogens, fungi, or encroaching vegetation (Bentley 1977a, Beattie 1985, Martin and Doyle 2003). However, engaging with patrolling ants may also impose costs on the plant, especially if these guards deter other mutualists (Koptur et al. 2015), although the net outcome has been demonstrated as beneficial in the subset of species in which this had been quantified (Romero and Koricheva 2011). Ant-plant mutualisms represent an ideal system to explore the costs and benefits of mutualisms from a multispecies approach because, on one hand ants are mutualists that benefit plants by decreasing herbivory (Chamberlain and Holland 2009, Rosumek et al. 2009, Trager et al. 2010, Romero and Koricheva 2011), and on the other ants can act as antagonists by consuming floral structures (Stanton et al. 1999, Gaume et al. 2005, Frederickson 2009), deterring pollinators (Assunção et al. 2014, Villamil et al. 2018) or other predatory arthropods (such as ladybirds, spiders, or wasps) that also benefit the plant (Jones and Koptur 2015).

98 Chapter 4: Ecological costs

The negative effects that competition between different pollinators for floral resources may impose on plant fitness are smaller compared to the negative effects of competition between pollinators and other plant mutualists that are not pollinators, such as bodyguards, may have on plant fitness (Cembrowski et al. 2014, Dutton et al. 2016). Negative interactions between predator ants and pollinators are theoretically expected, and have been confirmed in several species as an evolutionary stable strategy (Rico-Gray and Oliveira 2007b, Frederickson 2009, Palmer et al. 2010). Plant trade-offs in resource allocation to reproduction or defence may be one cause of conflicting interactions between pollinators and ants (Dutton et al. 2016, Villamil et al. 2018). Ant aggressivity and the predation risk it represents for pollinators is another source of conflict (Romero and Koricheva 2011, Villamil et al. 2018).

Aggressivity to herbivores and reductions in herbivory rates are the most commonly used criteria to evaluate the benefits and protection efficiency (herbivore removal and damage reduction rates) of ant mutualists (Chamberlain and Holland 2009, Rosumek et al. 2009, Trager et al. 2010, Romero and Koricheva 2011). However, these indicators do not reveal the ecological costs of such defensive mutualists, and in fact considering only these criteria can underestimate costs associated with myrmecophily that have been demonstrated by several studies (Dukas 2001, Gaume et al. 2005, Ness 2006, Goncalves-Souza et al. 2008, Frederickson 2009, Romero and Koricheva 2011, Stanton and Palmer 2011, Dutton and Frederickson 2012, Ohm and Miller 2014, Jones and Koptur 2015, Malé et al. 2015). In fact, ant aggressivity may be a double-edged sword that underlies the core ecological costs and benefits of myrmecophily. More aggressive ants are likely to be better defenders against herbivores, but also pose a higher predation risk to other mutualistic guilds such non-ant predators of herbivores or pollinators due either to direct attacks or

99 Chapter 4: Ecological costs non-consumptive effects (Ness 2006, Ohm and Miller 2014, Jones and Koptur 2015, Villamil et al. 2018).

Non-consumptive effects of predation are defined as changes in prey traits or behaviours when under predation risk (Preisser et al. 2005, Sheriff and Thaler 2014) and their magnitude can be similar to, or higher than, that of direct consumptive effects (Preisser et al. 2005, Clinchy et al. 2013, Sheriff and Thaler 2014). Even low levels of predation risk may impose strong selection on pollinators to develop detection mechanisms and anti-predatory behaviours, given that the cost of failing to detect predators includes death (Dukas 2001, Abbott and Dukas 2009). Several mechanisms of predator detection have been documented in pollinators. Pollinators recognise parts of predator bodies, such as crab spider forelimbs (Goncalves-Souza et al. 2008), detect past predation events through remains of previously preyed insects (Romero et al. 2011) or, in the case of Apis mellifera, communicate predation zones with waggle dance communication (Abbott and Dukas 2009). Predation risk also has strong effects influencing pollinator behaviour as shown by a meta-analysis (Romero et al. 2011).

A meta-analyses showed that although anti-predatory responses in pollinators depend on a combination of both pollinator and predators traits (Romero et al. 2011), non-consumptive effects of predators on pollinators can have important consequences for plant-pollinator interactions and the reproductive success of both partners (Romero et al. 2011, Clinchy et al. 2013). When predators interfered in plant-pollinator mutualism, they decreased plant fitness by 17% as reported in a meta-analyses by Romero and Koricheva (2011). Yet, predator-pollinator interactions in flowers are still poorly known and heavily understudied compared to predator-herbivore interactions in plants (Romero and Koricheva 2011).

100 Chapter 4: Ecological costs Many studies have investigated the effects of guarding ants on the reproductive outcome of the host plant. Four meta-analyses have compiled this information (Chamberlain and Holland 2009, Rosumek et al. 2009, Trager et al. 2010, Romero and Koricheva 2011), revealing that only few studies have addressed the ecological costs of ants. Most of the studies on the effects on ants on plant fitness have focussed on their effects via herbivore deterrence, overlooking the effects via pollinator deterrence (Romero and Koricheva 2011). In fact, up to 2011, only four studies had addressed the indirect interactions between ants and pollinators (Romero and Koricheva 2011) compared to over 31 papers on their effects via herbivore deterrence (Trager et al. 2010). And so, we emphasize the importance and need for a multispecies approach in ecological interactions research.

Although ants can increase plant fitness by decreasing herbivory and florivory, ants may also affect plant fitness by disrupting plant–pollinator mutualisms (Ness 2006, Goncalves-Souza et al. 2008, Ohm and Miller 2014). However, mutualisms are only maintained throughout evolutionary time if their fitness benefits are greater than their costs (Herre et al. 1999, Stanton and Palmer 2011). The overall net effect of predators, such as ants, on plant fitness was positive, increasing plant fitness by 32%, indicating that effects via herbivores were stronger than effects via pollinators (Romero and Koricheva 2011).

Due to ant threats, pollinators may respond in different ways which will have different consequences on the host plant mating system and fitness. Those studies that have addressed ant-pollinator conflict show contrasting evidence regarding the effects of ants on pollinator visitation and their consequences on plant mating systems. Through direct attacks or non-consumptive effects, aggressive ants can reduce pollinator visitation frequency and duration (Ness 2006), and reduce plant fitness by eating flower buds or making pollen inviable

101 Chapter 4: Ecological costs (Stanton et al. 1999, Frederickson 2009, Stanton and Palmer 2011, Dutton and Frederickson 2012, Malé et al. 2015), leading to decrease the number of fruits, seeds or seed mass (Ness 2006). Ness (2006) attributes lower seed set to a three- fold reduction in visitation frequency of Ferocactus wislizenii plants tended by aggressive ants, however he did not test this hypothesis. Guarding ants can also increase plant fitness despite disrupting plant-pollinator mutualisms. Due to ant threats, pollinators may spend less time per flower and could potentially visit more flowers, promoting more efficient pollen transfer (Altshuler 1999). Increased fruit set in the presence of ants in Psychotria limonensis was interpreted as a beneficial outcome of ant-pollinator interactions, due to a higher rates of movement between flowers, leading to enhanced outcrossing (Altshuler 1999), although this has not either been tested. Previous experiments on the ant-plant T. velutina have shown that ant corpses placed inside flowers reduce pollinator visit duration (Villamil et al. 2018). However, such an experimental setup may differ from natural circumstances since flower occupation by ants is a rare event, and live ants (in contrast to dead ones) do not remain immobile in the flowers for long periods. Overall, changes in pollinator assemblage composition, visit frequency and duration have been hypothesized to drive positive or negative impacts of ant-pollinator interactions on plant reproduction (Altshuler 1999, Ness 2006). As yet, none of these have been specifically tested or observed in an adequate experiment where pollinator visitation, frequency, and duration of the visits is recorded along with the fitness of plants with and without defensive ant patrolling.

QUESTIONS & AIMS

We estimated the ecological and fitness consequences of myrmecophily on the pollination biology of Turnera velutina (Passifloraceae) by contrasting plants patrolled by free roaming, living ants with ant-excluded plants. We quantified the effects of ant patrolling on pollinator community composition, visitation

102 Chapter 4: Ecological costs rates, behavioural consequences on floral visitors, and their consequences on the plant mating system, pollen transferring rates and dynamics, and plant fitness. We addressed the following questions:

(i) What is the effect of ant patrolling on pollinator visitation? (pollinator assemblage composition, visitation frequency, duration, and behaviour)

(ii) Does ant patrolling affect the host plant mating system?

(iii) Does ant patrolling affect pollen transfer dynamics?

(iv) Does ant patrolling affect plant male fitness?

HYPOTHESES AND PREDICTIONS Pollinator composition and behaviour Pollinator responses to predation risk can vary between taxa. Pollinator body size and lifestyle (social or solitary) have been proposed as a source of variation. Smaller pollinators have been predicted to be more vulnerable to predation risk (Dukas and Morse 2003), whilst contrasting predictions have been made regarding lifestyle. On one hand, social hymenopterans can influence naïve colony members to forage away from dangerous flowers, such as Apis mellifera does by waggle dance communications (Abbott and Dukas 2009), leading to predictions of stronger avoidance behaviours in social species compared to solitary bees. But, social hymenopterans may also be under weaker selection pressures to evolve avoidance behaviours since death of a non-reproductive worker has lower inclusive fitness consequences than the death of a solitary bee female (Clark and Dukas 1994). We predicted ant patrolling would induce changes in pollinator composition by deterring non-aggressive, solitary pollinators resulting in more visits by non-aggressive pollinators to plants without ants.

103 Chapter 4: Ecological costs Pollinator visitation and plant mating system Ant patrolling can affect pollinator dynamics and plant mating systems in a variety of ways as described in the following envisioned scenarios and the predictions associated with them.

1) Pollinators are completely deterred by the presence of ants and avoid these plants leading to reduced visitation frequency, pollinator limitation, and reduced seed set and plant fitness. 2) Pollinators are partially deterred and are forced to relocate frequently to avoid being attacked by ants, leading to a reduction in visit duration. If pollinators spend less time per flower they could potentially visit more flowers, promoting more efficient pollen transfer and increasing male fitness. If this scenario is true, there are two possible outcomes for the plant mating system: a. If pollinators relocate to another plant, this would promote higher outcrossing rates, increasing the genetic diversity of the seeds and enhancing the advantages of sexual reproduction for the host plant. b. If pollinators relocate to another flower within the same plant, this would promote geitonogamy (intra-plant pollination), which is a form of selfing that does not enhance genetic diversity, representing an intermediate cost for pollination of the host plant.

Plant fitness The effects of patrolling ants on host plant fitness have been shown to vary. Romero and Koricheva (2011) detected several factors explaining the variation in the effect of predators on plant fitness via pollinator deterrence. We highlight three of these factors relevant to the ecology of T. velutina: predator hunting mode, habitat domain, and plant rewards. 1. Active hunters, such as ants, increased plant fitness whilst sit-and-wait predators, such as spiders, did not.

104 Chapter 4: Ecological costs 2. Predators with a broader territory, such as wasps and ants, had negative effects on plant fitness, whereas those with a narrow habitat domain, such as bugs, ladybirds and parasitoids had positive effects. 3. Predator effects were positive on plants that offered rewards such as extrafloral nectar, and negative for plants without rewards.

Based on these findings, the predictions for the effect of patrolling ants on T. velutina’s fitness via pollinator deterrence are contrasting. On the one hand, the active hunting mode of ants and the presence of EFN rewards in T. velutina predict ants would have a positive effect on fitness. But on the other hand, the wide territorial domain of ants predicts a negative effect. Patrolling ants in T. velutina increase plant fitness via herbivory suppression (Cuautle and Rico-Gray 2003), and we predicted that their effects on plant fitness via pollinator deterrence would be negative, decreasing fitness.

MATERIALS AND METHODS Study site Field experiments were conducted at Troncones, Guerrero on the southern Pacific coastline of Mexico (17°47’ N, 101° 44’ W, elevation < 50 m). The climate is warm sub-humid, with a rainy season during the summer (mid-May to November), 1025 mm annual precipitation, and a mean annual temperature range of 25.4-28.0 °C (Zedillo-Avelleyra 2017). The habitat has been described as secondary vegetation of tropical low forest (Zedillo-Avelleyra 2017), which becomes coastal sand scrub vegetation as it approaches the sea.

Study system Turnera velutina (Passifloraceae) is a Mexican endemic perennial shrub (Arbo 2005) that is also a myrmecophile (Cuautle and Rico-Gray 2003) which grows in lowland, coastal environments. It establishes a facultative mutualism with up to

105 Chapter 4: Ecological costs 13 ant species (Cuautle et al. 2005, Zedillo-Avelleyra 2017), 10 of which are present in Troncones (Zedillo-Avelleyra 2017). Extrafloral nectar is provided in paired cup-shaped glands located at the bottom of the leaf blade or on the leaf petiole, on the underside of the leaves. T. velutina is a self-compatible, monomorphic herkogamous species that requires pollinators for adequate seed production (Sosenski et al. 2016). Although it flowers year-round, flowering peaks during summer (Cuautle et al. 2005) and flowers are bisexual, last one day, and are insect-pollinated (Sosenski et al. 2016). As rewards for pollinators, flowers offer pollen and floral nectar that is easily accessible at the base of the yellow, pentamerous, bell-shaped corolla. At Troncones, native butterflies are the dominant flower visitors of T. velutina, followed by the introduced European honeybee (Apis mellifera). Native bees, wasps, and occasionally flies also visit flowers at Troncones. Apis mellifera has been confirmed as a pollinator for T. velutina (Ramos-Castro 2013, Sosenski et al. 2016) but the pollination efficiency of other floral visitors still remains to be confirmed.

Ants exclusions and experimental setup Twelve large plants fitting the following criteria were haphazardly chosen: plants producing six or more flowers per day, surrounded by other T. velutina plants, and semi-isolated plants (not within dense patches) and with few vegetation links to other plants. Within these twelve plants, pairs of nearby plants with similar height and foliage abundance were chosen and marked as either control (N=6) or experimental (N=6). The vegetation surrounding these plant pairs was trimmed or cleared in order to isolate these plants from direct contact with any surrounding vegetation. Plants within a pair were located < 2 m apart, and each pair of plants was separated from any other pair by at least 10 m, to reduce the chances of dyed pollen from a neighbour pair confounding our results. These plant pairs were used to assess the effect of ant patrolling on pollinator visitation, plant mating system and fitness.

106 Chapter 4: Ecological costs Ants were excluded from experimental plants using TanglefootTM, a highly sticky insect-trapping resin, whilst control plants had no ant exclusions. A band of masking tape with the sticky side outwards was wrapped around the base of the trunk or branch from which ants will be excluded. A ring of TanglefootTM was applied only on the masking tape band, carefully avoiding spreading it onto plant tissue as it blocks stomata, killing the plant. A 2 cm wide masking tape band is enough for small twigs; thicker bands were required for larger branches. In shrubs with multiple stem branching from the base, exclusions were applied on every stem. All vegetation links to other plants were trimmed, and if required, parts of the bush were tied with string to contain and effectively isolate them, destroying any points of contact that ants could use to climb onto the plant. TanglefootTM was also applied to string ties to prevent ants from accessing the plants by crawling up the ties. Exclusions were checked daily and TanglefootTM was replenished if required.

To assess the effect of ant patrolling on pollen transfer and its consequences on the rate of selfing, geitonogamy, and outcrossing, the anthers of plants with and without ant patrolling were differentially dyed. We used a focal-plant, focal-flower experimental design. Each focal plant, either control or ant-excluded, was surrounded by other neighbouring plants. The anthers of focal plants were dyed (red, blue, green or purple), and flowers in neighbouring (non-focal) plants remained undyed (naturally yellowish-orange). Within a focal plant, a focal flower was selected and its anthers were dyed differently from the anthers in the rest of the flowers from that plant, hereafter referred to as satellite flowers.

Every morning at 0700 before flowers opened, six flower buds per plant (a focal bud surrounded by five satellite buds) were bagged using tulle bags to exclude visitors. All additional pre-anthesis buds were removed to standardise floral display across experimental plants. Once the corollas were fully open,

107 Chapter 4: Ecological costs anthers were dyed and the flowers were re-bagged until the dye dried and anthers had dehisced. Pollinator observations were conducted on all flowers (focal and satellite) and stigmas from focal flowers were collected at the end of the anthesis period to count pollen grains received. To assure a minimum common supply of allogamous pollen across all flower pairs, 10-12 flowers from other plants around each control-exclusion plant pair were also bagged before anthesis and remained bagged until the focal and satellite flowers were un- bagged for observation. All statistical analyses were conducted in R version 3.5 (R Core Team 2016). All mixed effects models were fitted using ‘lme4’ R package (Bates et al. 2016) and post-hoc Tukey comparisons were tested using the ‘multcomp’ R package (Hothorn et al. 2008).

108 Chapter 4: Ecological costs a)

b)

Anthers dyed

Pistils + pollen

squashes

c)

Figure 1. Experimental design and methods. a) diagram of the focal-plant, focal-flower experimental design. Empty pentagons represent focal plants with or without ants, coloured circles represent focal and satellite flowers with dyed anthers, and yellow pentagons represent neighbouring plants with undyed anthers (naturally yellow); b) syringes with the dyes, dyed anthers, stigmas bearing dyed pollen grains, and stigma squash slides; c) pollen flow diagram.

109 Chapter 4: Ecological costs Visitation Pollinator visitation to all flowers (1 focal + 5 satellites) on control and ant excluded focal plants was recorded to estimate the effect of ant patrolling on pollinator visitation. Once the dyed anthers had dried and dehisced, bags were removed from all flowers from one focal plant and its surrounding un-dyed flowers. Every focal plant was observed for 40 min split into two rounds of 20 min. The first round was conducted immediately after bag removal when flowers had a full pollen and nectar load. The second round started ~90 min after the first round was completed. Unobserved flowers remained bagged until their first observation round started to ensure all flowers had a full pollen and nectar load during the first round of observations. We recorded the identity, frequency, duration, and behaviour of floral visitors and visits, as detailed below.

Composition To estimate the overall abundance of flower visitors by each taxonomic group, observations from control and ant-excluded plants were pooled together and the percentage of visits by each taxon was calculated. Ants and beetles were recorded as floral visitors but were not regarded as potential pollinators (see Results for further details). Potential pollinators (hereafter pollinators) comprised the following taxonomic groups: Apis mellifera, native bees, Lepidoptera, Diptera, and wasps. Bees were split into Apis mellifera and native bees because we were interested in contrasting the visitation patterns of the native bee species with those of the introduced European honeybee (Apis mellifera). Within each of these taxonomic groups, differences in the total number of visitors between control and ant excluded plants were assessed using a Pearson Chi-squared test. Since Apis mellifera and Lepidoptera jointly accounted for 94% of all visitors (Table 1) we considered only these taxonomic groups for all further analyses, excluding native bees, wasps and flies.

110 Chapter 4: Ecological costs Visitation frequency Visitation frequency was defined as the number of interactions between a floral visitor or potential pollinator and any of the flowers from a focal plant during the 40 min of observation (visits/40 min/plant). We recorded visitor identity and so the same visitor could account for more than one visit if, for instance, it landed, hovered, and landed again in a flower. Visitors were not followed, and visits to other surrounding plants were not recorded. We stopped recording a visitor whenever it left the plant and escaped our vision field. Visitors abundance was estimated as the number of individual visitors per taxa (individuals per plant).

The effect of ant patrolling on visitation frequency was tested using a Poisson mixed model, fitting the number of visits per plant as the response variable. Ant exclusion treatment, pollinator type (Apis mellifera or Lepidoptera), and the interaction between both factors were fitted as fixed effects. Plant identity and day were fitted as a random effect, along with an observation level random effect (OLRE) to account for overdispersion (Hinde 1982). Post hoc Tukey model comparisons were used to detect statistically significant differences between the two types of pollinators under control or ant exclusion treatments.

Visit duration Visit duration is the time a pollinator (Apis mellifera or butterflies) spent interacting with a flower. The effect of ant patrolling on visit duration was tested using a Poisson mixed model, fitting the mean visit duration per individual visitor as the response variable. Ant exclusion treatment, pollinator type (Apis mellifera or Lepidoptera), and the interaction between both factors were fitted as fixed effects. Plant identity, day, and an OLRE were fitted as random effects. Post hoc Tukey model comparisons were used to detect statistically significant differences between the two types of pollinators under control or ant exclusion treatments.

111 Chapter 4: Ecological costs Pollinator behaviour Two types of pollinator behaviours were identified: ‘inspection’ and ‘contact’, following (Villamil et al. 2018, Chapter 3). Inspection behaviour is defined as a pollinator approaching to, or hovering over, a flower without landing on it. Contact behaviours are defined as all behaviours made within the flower between landing and take-off, and included foraging on pollen or nectar, or resting, basking or grooming on the petals, anthers or stigmas.

The effect of ant patrolling on the likelihood of pollinators displaying inspection behaviours as defined above was tested using a binomial mixed model. For all Apis mellifera and butterflies observed, the presence or absence of inspection behaviours was fitted as the response variable. Ant exclusion treatment, pollinator type (Apis mellifera or Lepidoptera), and the interaction between both factors were fitted as fixed effects. Plant identity, pair identity, and day were fitted as random effects. Post hoc Tukey comparisons were used to detect statistically significant differences between all four groups (Apis mellifera or butterflies, at control or ant exclusion plants).

The effect of ant patrolling on pollinator deterrence was tested using a binomial mixed model. Pollinator deterrence is here defined as the absence of contact behaviours following an inspection behaviour, where pollinators displayed only inspection behaviours, approaching or hovering over, but without landing inside the flower. For every pollinator observed displaying inspection behaviours, the presence or absence of contact behaviours was fitted as the response variable. Ant exclusion treatment, pollinator type (Apis mellifera or Lepidoptera), and the interaction between both factors were fitted as fixed effects. Plant identity, pair identity, and day were fitted as random effects. Post hoc Tukey comparisons were used to detect statistically significant differences between all four groups (Apis mellifera or butterflies, at control or ant exclusion plants).

112 Chapter 4: Ecological costs The effect of ant patrolling on the duration of each type of behaviour was tested using a Poisson mixed model, splitting visitor observations into inspection or contact behaviours. The total duration of each behaviour (inspection or contact) per visitor was fitted as the response variable. Ant exclusion treatment, pollinator type, (Apis mellifera or Lepidoptera), behaviour (inspection or contact) and paired interactions between these variables (Treatment*Type of pollinator, Treatment*Behaviour, Type of pollinator*Behaviour) were fitted as fixed effects. Plant identity, pair identity, day, and an OLRE were fitted as random effects. Post hoc Tukey model comparisons were fitted to test statistically significant differences between all eight groups (inspection and contact behaviours in Apis mellifera or butterflies, under control or ant exclusion plants).

Pollen dyes Anthers were dyed once the corolla opened completely (~0800-0815), but before pollen was released (anther dehiscence). Anthers, one by one, were embedded in a droplet of dye until soaked. This was done carefully, to avoid staining any other floral structures, such as the filaments, pistils or petals. Once dyed, flowers were bagged again until anthers dehisced and the pollen released was dry. Purple dye was prepared by diluting commercially available methyl violet with ethanol 70% [50:50], the green dye used was commercially available Green S without dilution. Red (safranin) and blue (methylene blue) were donated by Manuel Ochoa from LIPA, UNAM. These dyes were prepared by diluting 1 g of either dye in 15.5 ml of distilled water, and then mixing this solution with ethanol 95% [50:50]. Previous studies showed that dyeing Turnera velutina anthers in these colours had no effect on pollinator visitation (Ochoa Sánchez 2016). Towards the end of the anthesis period (11:30), pistils from focal flowers were collected in Eppendorf tubes and mounted on a slide as a glycerine squash. Pistils were collected before corollas closed in order to prevent petals touching

113 Chapter 4: Ecological costs the stigma, which could damage and collapse the stigmatic hairs, interfering with the squashes pollen counts. Slides were sealed using nail polish and kept in a fresh and dry environment, with controlled room temperature (22°C) (Kearns and Inouye 1993, Ochoa Sánchez 2016). Selfing, outcrossing, and geitonogamy rates on plants with and without ant patrolling were estimated by counting red, blue, green, purple, and yellow (un-dyed) pollen grains from focal flower stigmas under a light microscope at 10x or 40x using phase contrasts Ph3 and Ph4 for colour discrimination.

Plant mating system Pollen load, the total number of pollen grains received per stigma, was counted from stigma squash slides. The effect of ant patrolling on stigma pollen load was tested using a Poisson mixed model, fitting the total number of pollen grains per stigma as the response variable. Ant exclusion treatment was fitted as a fixed effect, plant identity, pair identity, day, and an OLRE were fitted as random effects.

The effect of ant patrolling on pollen transfer and its consequences on plant mating system (selfing, geitonogamy, and outcrossing) was assessed by counting differentially dyed pollen grains. Pollen grains received from the same flower (selfing), from another flower within the same plant (geitonogamy), from the reciprocal pair plant, or another plant (outcrossing) were counted on focal- flower stigmas. The number of pollen grains from each origin (selfing, geitonogamy or outcrossing) was divided by the pollen load per stigma to determine the proportion of pollen from each mating system. Proportional data were transformed to normality using the logit transformation as follows:

Logit pollen proportion = log( ). Infinite numbers resulting

𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 from impossible quotients were1− 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝replaced𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 by zeros. The effect of ant patrolling on the mating system was tested using a linear mixed model, fitting the proportion of pollen as the response variable. Ant exclusion treatment, mating

114 Chapter 4: Ecological costs system (outcrossing, geitonogamy or selfing), and the interaction between both factors were fitted as fixed effects. Plant identity, pair identity, and day were fitted as random effects. Post hoc Tukey comparisons were used to test statistically significant differences between the proportion of pollen grains from each type of mating system in plants with and without ants.

The pollen flow dynamics were analysed using six categories to describe the mating system and the pollen origin (MSPO), dictating from where to where the pollen was moved. The effect of ant patrolling on pollen flow dynamics was tested using a Poisson mixed model, fitting as the response variable the number of pollen grains received and donated to and from every possible pollen source identifiable in this experiment (same flower selfing, other flower from the same plant: geitonogamy, reciprocal pair plant (outcrossing pair), or another plant from the same species (outcrossing unknown)). Ant exclusion treatment, pollen origin-destination, and the interaction between both factors were fitted as fixed effects. Plant identity, pair identity, day and an OLRE were fitted as random effects. Post hoc Tukey comparisons were used to test statistically significant differences between these groups.

Male plant fitness The number of pollen grains donated per flower was used as an estimate for male plant fitness and quantified based on pollen grains from each flower/colour deposited on focal stigmas. The total number of pollen grains from satellite flowers was divided by five, to obtain the mean number of pollen grains donated by each satellite flower within the plant. The total number of pollen grains from the reciprocal pair plant was divided by six to obtain the mean number of pollen grains donated by each flower from the reciprocal plant pair.

115 Chapter 4: Ecological costs The effect of ant patrolling on male fitness was estimated using a Poisson mixed model fitting as the response variable the number of pollen grains donated per flower in control or ant-excluded plants. Ant exclusion treatment (control or ant exclusion) was fitted as a fixed effect and plant identity, pair identity, day, and an OLRE were fitted random effects. Post hoc Tukey comparisons were used to test statistically significant differences between these groups.

The effect of ant patrolling on the destination of the pollen grains donated per flower was tested using a Poisson mixed model. The number of pollen grains was fitted as the response variable. Ant exclusion treatment, pollen donor-destination, and the interaction between both factors were fitted as fixed effects. We only contrasted the number of pollen grains donated by focal or satellite flowers, as every plant had only six flowers, since we controlled floral display in our experimental design. Pollen donated by unknown plants was excluded as the number of donor flowers was unknown, and hence cannot estimate pollen grains donated per flower. Plant identity, pair identity, day, and an OLRE were fitted as random effects. Post hoc Tukey comparisons were used to test statistically significant differences between these groups.

RESULTS Floral visitors and potential pollinators We recorded 40 h of observations of 360 flowers on 12 plants over five days. These flowers received 967 floral visitors over the observation period, which included Apis mellifera, several morphospecies of butterflies, native bees, beetles, ants, wasps, and flies (Table 1). Only one individual ant was observed inside one of the flowers in the ant exclusion treatment, and three individual ants were observed inside flowers on control plants. A total of 108 beetles were

116 Chapter 4: Ecological costs recorded inside flowers, with 54 in each of control and ant-excluded flowers (see Supplementary Natural History Notes). Nonetheless, ants and beetles were excluded from the list of potential pollinators and from any further analyses because their role as pollinators of T. velutina is dubious (for ants see:Rico-Gray and Oliveira 2007b, Dutton and Frederickson 2012) (for beetles see: Kerner 1878, Krupnick and Weis 1999, Rudgers and Gardener 2004, Takayuki 2005). Beetles were excluded as they are usually regarded as florivores since they eat pollen, anthers (Baldock et al. 2011), and are nectar robbers in T. velutina (see Supplementary Natural History Notes). Beetles are also commonly excluded from pollination studies because their visit durations are extremely long and would otherwise inflate and confound the behavioural patterns of other floral visitors. All of these factors are true in T. velutina where the duration of beetle visits ranged from 30 – 1200 sec, whilst the duration for other visitors ranged from 1 - 95 sec.

We recorded 853 individual visitors from taxa regarded as potential pollinators (hereafter pollinators), because they were observed contacting male and female plant sexual organs, although further experiments testing their efficacy as pollen vectors are still required. Visitors so defined comprised Lepidoptera, Apis mellifera, native bees, wasps, and Diptera (Table 1). Because lepidopterans and Apis mellifera accounted for more than 80% of all floral visitors and > 94 % of potential pollinators (Table 1), we considered only these two taxa in further analyses regarding pollinator frequency, duration, and behaviour. Pollinator visitation: composition The total number of pollinators from each taxonomic group visiting T. velutina flowers did not differ significantly between control and ant-excluded plants (X2 = 1.42, df = 4, P = 0.84; Fig. 2).

117 Chapter 4: Ecological costs Table 1. Relative abundance of floral visits by and visitors by floral visitors and potential pollinators. Visits consider the number of times a visitors contacted the floral organs, whilst visitors consider the number of individuals contacting flowers.

Visits Visitors Number Floral Pollinators Number Floral Pollinators Taxon visitors (%) visitors (%) (%) (%) Lepidoptera 436 45.09 51.11 192 42.01 55.65 Apis mellifera 370 38.26 43.38 132 28.88 38.26 Native bees 31 3.21 3.63 8 1.75 2.31 Wasps 10 1.03 1.17 8 1.75 2.31 Diptera 6 0.62 0.70 5 1.09 1.44 Coleoptera 110 11.38 108 23.63 Ants 4 0.41 4 0.87 TOTAL 967 100% 100% 457 100% 100%

Figure 2. Composition of floral visitors visiting Turnera velutina. (a) Overall number of visitors per taxa in control (blue) and ant excluded (red) plants; and (b) community composition comparisons of floral visitors to plants with and without ant patrolling.

118 Chapter 4: Ecological costs Pollinator visitation: frequency Visitation frequency did not differ significantly between pollinator types, and ant exclusion treatment had no significant effect on visitation frequency, regardless of the pollinator taxa considered as shown by the non-significant interaction term (Fig 3a, Table 1).

Pollinator visitation: visit duration Pollinator taxa differed in visit duration, but there was no significant effect of ant exclusion treatment on visit duration. The ant exclusion x pollinator taxon interaction on visit duration was significant (Fig. 4b, Table 1). The duration of Apis mellifera visits had a significant two-fold increase in ant-excluded plants (Z = 2.45, P = 0.05; Table 1; Fig. 4b), whilst butterfly visit duration was not significantly affected by ant exclusion (Z = 1.07, P = 0.70; Table 1; Fig. 4b).

Pollinator visitation: behaviour The likelihood of displaying inspection behaviours differed significantly between pollinator taxa (Table 1, Fig. 3c), with butterflies being on average 15% more likely to display inspection behaviours than Apis mellifera (Fig. 3c). However, the ant exclusion treatment did not affect the likelihood of pollinators displaying inspection behaviours, regardless of the pollinator taxon (Apis mellifera or butterflies), as indicated by the non-significant interaction term (Table 1, Fig. 3c).

The percentage of pollinators deterred from a plant differed significantly between pollinator taxa, with, on average, three times more butterflies deterred than Apis mellifera (butterflies: 27%, Apis mellifera: 8.5%; Fig. 3d, Table 1). Ant exclusion treatment had no significant effect (Fig. 3d, Table 1). The ant exclusion x pollinator taxon interaction was significant (Table 1), with exclusion having a

119 Chapter 4: Ecological costs positive effect on butterflies, increasing deterrence, but a negative effect on Apis mellifera, decreasing deterrence (Table 1, Fig. 3d).

When the duration of a given pollinator’s visit was split between inspection and contact behaviours, we found that the effect of ant exclusion treatment on visit duration differed between pollinator taxa, and behaviours (Fig. 3e, Table 1), as shown by both significant interaction terms (Ant exclusion treatment x pollinator taxon; pollinator taxon x behaviour; Table 1). However, the effect of ant exclusion treatment did not differ between behaviours as shown by the non-significant interaction between ant exclusion treatment and behaviour (Table 1, Fig. 3d). Ant exclusion significantly increased the duration of Apis mellifera contact visits (Z = 2.96, P = 0.05), increasing the time bees spent inside flowers. However, ant exclusion did not significantly affect the time butterflies spent inside the flowers (contact behaviours: Z = -2.34, P = 0.25), or the duration of inspection behaviours by either pollinator (Apis mellifera: Z = - 0.48, P = 0.99; butterflies: Z = -1.73, P = 0.65). Pollinators spent longer displaying contact behaviours than inspection behaviours, regardless of the ant exclusion treatment. This trend did not differ between Apis mellifera and butterflies, as demonstrated by the post-hoc Tukey contrasts (Fig. 3e).

Plant mating system Pollen load per stigma was significantly increased in stigmas from ant- excluded plants, (Fig. 4a, Table 1), with ant excluded stigmas receiving 150 additional pollen grains compared to stigmas from ant-patrolled flowers (Control: 85 ± 12; Exclusion: 240 ± 32).

120 Chapter 4: Ecological costs a) b) c)

d) e) 50% b 25 Figure 3. Effects of ant patrolling c on (a) pollinator visitation 40% 20 frequency, (b) visit duration, and Behaviour

b Presence pollinator behaviours (c-e) 30% 15 b Alert affecting (c) the display of alert b Duration of visits of (s Duration

Visitors deterred Visitors ab b behaviours, (d) the likelihood of 10 Tx 20% deterrence, and (e) the time spent a C T displaying alert (triangles) or 5 10% contact behaviours (circles) by a a a a Apis mellifera (Am) and native 0 0% butterflies on Turnera velutina Am Butte Am Butterfly flowers on control plants with ant Type of pollinators Type of pollinator patrolling (blue), and ant excluded plants (red).

121 Chapter 4: Ecological costs The proportion of pollen grains in each mating system category differed significantly both within and between plant treatments (Fig. 4b, Table 1). The effect of ant exclusion on the proportion of pollen grains differed between mating systems, as shown by the significant ant exclusion x mating system interaction (Fig. 5, Table 1). In fact, excluding ants from plants significantly reduced outcrossing by half and geitonogamy by 33-fold, whilst tripling selfing rates (Fig. 5).

The effect of ant exclusion on pollen flow differed significantly between the MSPO (categories used to describe pollen flow dynamics), as shown by the significant interactions between exclusion treatment and MSPO. Ant exclusion increased by four times the number of selfing pollen grains received by stigmas, doubled the number of allogamous pollen grains received from neighbouring, un-dyed flowers, and reduced to a third the number of geitonogamous pollen grains.

Flowers received significantly more pollen grains than the number of pollen than we could trace as donated by each of the dyed flowers to the focal stigmas. However, our donation counts are under-estimated in a ratio of 1:22 as we only counted donations to focal stigmas, and not to any other flowers, whilst accounting for ≥ 22 donors (one selfing flower + five geitonogamous flowers + six flowers from the reciprocal pair + ≥ 10 neighbouring, un-dyed flowers). Therefore, the only symmetrical estimate of received:donated pollen grains is the comparison between reciprocal pair plants, as in that case the ratio is equal 1:6, with six flowers donating pollen to one stigma in both directions. However, ant exclusion had no effect on the number of pollen grains received and donated by flowers between reciprocal pair plants (OUT_PAIR_REC: control vs. exclusion: Z = 0.31, P = 1.00; Fig. 4c).

122 Chapter 4: Ecological costs a) b)

c) MSPO TRONCONES Pollen grains per plant

b

150

100 e Tx

Pollen grains Pollen C T

d 50

a a c c c c c 0 SELF_REC GEIT_REC OUT_PAIR_REC OUT_PAIR_DON OUT_UNK_REC Mating system and Pollen origin

Figure 4. Effects of ant patrolling on the (a) pollen load, (b) mating system rates and (c) pollen flow (origin and destination) in Turnera velutina showing mean ± se for control (blue) and ant-excluded (red) plants.

123 Chapter 4: Ecological costs Plant fitness Male fitness, assessed as the number of pollen grains donated per flower, was positively and significantly affected by ant exclusion, increasing the number of pollen grains donated per flower from 27.2 ± 6.54 in ant-patrolled plants to 163 ± 23.9 in ant-excluded plants. Ant exclusion, mating system, and the interaction between these factors all had significant effects on the number of pollen grains donated per flower to different destinations (Table 1). The vast majority of the pollen donated per flower arrived on the same flower’s stigma as selfing pollen, regardless of the exclusion treatment (Fig. 5b). The effect of ant exclusion on the number of pollen grains donated differed between mating systems and pollen destinations. Ant exclusion significantly increased the number of pollen grains donated to the same flower’s stigma and significantly decreased the number of pollen grains donated to other flowers of the same plant, but had no significant effect on the number of pollen grains donated to the reciprocal pair plant. a) b)

Figure 5. Effects of ant patrolling on male fitness, showing (a) average number pollen grains fathered per flower, and (b) the destination of the pollen grains donated per flower,(mean ± se) for control (blue) and ant excluded (red) plants.

124 Chapter 4: Ecological costs

Table 1. Model statistics testing the costs of ant patrolling on Turnera velutina’s pollination biology, including ecological, behavioural, mating system, and fitness consequences.

Random Response Fixed effects N LRT P-value Variance SD effects

Ant exclusion 90 0.64 0.42 Plant 0.00 0.00 Visitation Pollinator taxon 3.26 0.07 Day 0.113 0.336 frequency (Am & butterflies) OLRE 0.108 0.328 Ant exclusion x Taxon 0.75 0.78 Ant exclusion 90 1.60 0.20 Visitor taxon Plant 5.48-10 2.34-05 Duration 8.36 0.0038 ** (Am & butterflies) Day 9.97-10 3.15-05 OLRE 0.681 0.825

Ecological consequences Ant exclusion x Taxon 7.76 0.0053 **

Likelihood of Ant exclusion 326 1.64 0.19 inspection Plant 0 0 Visitor taxon Day 0 0 (Binomial) 10.01 0.001 ** (Am & butterflies) Pair 0 0 Ant exclusion x Taxon 1.03 0.30 Likelihood of Ant exclusion 326 1.56 0.211 avoidance Plant 0 0 Visitor taxon Day 0.118 0.343 (Binomial) 14.71 0.0001 *** (Am & butterflies) Pair 0.051 0.226 Behavioural consequences Behavioural Ant exclusion x Taxon 5.05 0.024 *

125 Chapter 4: Ecological costs

Random Response Fixed effects N LRT P-value Variance SD effects

Ant exclusion 652 0.84 0.35 Visitor taxon Plant 1.09-06 0.0010 11.31 0.0007 *** (Am & butterflies) Pair 2.58-07 0.00050 Duration Day 1.77-07 0.00042 Behaviour 331.44 2.2-16 *** (per behaviour) OLRE 1.70 1.30 Ant exclusion x Taxon 5.40 0.02 * Ant exclusion x 0.90 0.34 Behaviour

Behavioural consequences Behavioural Taxon x Behaviour 29.90 4.54-08 ***

Plant 0.12 0.34 Pollen load Pair 0.03 0.18 Ant exclusion 72 9.19 0.002 ** Day 0.16 0.41 OLRE 0.52 0.72 Mating system Ant exclusion 336 0.08 0.77 Plant 3.92-06 0.001 Mating system 24.67 4.37-06 *** Pair 0.17 0.41 Ant exclusion x system consequences system 75.92 2.2-16 *** Day 0.53 0.73 Mating system OLRE 6.46 2.54

Mating Mating Ant exclusion 336 0.58 0.44 Plant 0.0006 0.077 Pollen flow Pollen origin 108.95 2-16 *** Pair 0.22 0.83

126 Chapter 4: Ecological costs

(mating system, Day 0.69 0.47 Ant exclusion x received and 92.71 2.2-16 *** OLRE 4.25 2.06 Pollen origin donated)

Random Response Fixed effects N LRT P-value Variance SD effects

Male fitness Ant exclusion 72 9.56 0.001 ** Plant 0.12 0.34 (Poisson) Pair 0.03 0.18 Day 0.16 0.41 OLRE 0.52 0.72 Destination of Ant exclusion 200 3.93 0.04 * Plant 0.16 0.40 donated pollen Pair 0.004 0.06 (Poisson) Pollen origin 69.69 7.33-16 *** Fitness consequences Fitness Day 0.44 0.67 Ant exclusion x 77.33 2.2-16 *** OLRE 2.99 1.72 Pollen origin

127 Chapter 4: Ecological costs DISCUSSION

This study provides a comprehensive picture of the interaction between myrmecophily and pollination by showing the ecological and behavioural effects of ant patrolling and associated changes in plant mating systems and fitness consequences. Experimental evidence in this system had shown potential for direct ant-pollinator conflict (Villamil et al. 2018). Despite this, and contrary to our expectations, excluding ants from plants did not affect pollinator composition (Fig. 2) or visitation frequency (Fig. 3a). Ant exclusion did not affect either the probability of pollinators being deterred (Fig. 3d), the probability of pollinators displaying inspection behaviours (Fig. 3c), or the duration of inspection behaviours (Fig. 3e). However, ant exclusion had a positive effect on visit duration (Fig. 3e), pollen load, outcrossing rates (Fig. 4), and male fitness (Fig. 5). Ant exclusion increased the time Apis mellifera spent inside flowers (Fig. 3b). Furthermore, ant exclusion increased the pollen load on stigmas by 150%, switched the mating system from outcrossing to selfing (Fig. 4b), and increased by 150% the number of pollen grains sired per flower. The increase in the time Apis mellifera spent inside flowers may be the mechanism underlying the increased selfing rates observed in plants under the ant exclusion treatment. The positive effect of ant exclusion on the duration of contact visits by Apis mellifera may also be responsible for the increase in male fitness if longer visits allow Apis mellifera to collect and transport more pollen grains. Although ant patrolling had negative effects on pollinator visitation behaviours and foraging efficiency, such behavioural changes seems to result in positive effects on plant fitness promoting outcrossing and increasing male fitness.

Effects of ant patrolling on pollinator visitation dynamics: Composition, visitation frequency and duration In Troncones, butterflies were slightly more frequent visitors than Apis mellifera (Fig. 3a), accounting for 51% of the pollinator visits, whilst Apis mellifera accounted for 43% of the visits (Table 1). Our results are consistent with

128 Chapter 4: Ecological costs previous studies suggesting that butterflies are more frequent visitors, and bees are better pollinators (Barrios et al. 2016), although the pollinator efficiency of each taxon in T. velutina remains to be investigated. Such differences in the abundance of native butterflies and A. mellifera are an interesting contrast with other populations of T. velutina where A. mellifera is the dominant pollinator accounting for 94% of floral visits (Sosenski et al. 2016, Villamil et al. 2018). Such contrast and the higher frequency of native pollinators make Troncones an ideal population to study the evolutionary ecology of plant-ant-pollinator interactions around T. velutina.

Other non-pollinator animals interacting with flowers can reduce flower attractiveness for pollinators and alter pollinator visitation dynamics, resulting in lower pollen donation, lower pollen deposition and reduced seed set (Krupnick and Weis 1999). In ant-plant systems patrolling ants can deter pollinators, decreasing visitation frequency and duration (Ness 2006). The predation risk that patrolling ants represent for pollinators can also reduce flower attractiveness, increase flower avoidance rates and decrease pollinator visitation frequency and duration (Romero et al. 2011). Anti-predatory responses in pollinators vary depending on the pollinator and predator taxa involved (Romero et al. 2011) but few studies have documented how different floral visitors respond to ant patrolling (Ness 2006, Ohm and Miller 2014, Carper et al. 2016) and how different ant partners affect pollinators (Ness 2006, Miller 2007, Ohm and Miller 2014, Villamil et al. 2018).

We are only aware of three systems in which the effect of different ant partners on pollinators has been examined: two Cactaceae species and Turnera velutina (Passifloraceae). In Opuntia imbricata ant patrolling significantly decreased pollinator visitation, but there were no differences between ant species (Ness 2006, Miller 2007, Ohm and Miller 2014). In Ferocactus wislizeni plants patrolled by the most aggressive bodyguards received less and shorter

129 Chapter 4: Ecological costs pollinator visits and produced less and lighter seeds (Ness 2006). In T. velutina the identity of the ant partners had no effect on visitation frequency, but the presence of the most aggressive bodyguard inside flowers increased by 20% the likelihood of pollinators displaying inspection behaviours, halved the time Apis mellifera spent inside flowers, and reduced pollinator foraging efficiency by increasing by 12% the proportion of time spent displaying inspection behaviours (Villamil et al. 2018).

Several hypotheses have been made regarding the effect of different ant partners on pollinators, and many more studies have addressed this question (Romero et al. 2011). Pollinator body size and lifestyle (social or solitary) have been proposed as a source of variation, predicting smaller and solitary pollinators as more vulnerable to predation risk (Clark and Dukas 1994, Abbott and Dukas 2009). In Troncones, native bee and butterfly species are solitary, non-aggressive floral visitors, whilst Apis mellifera are introduced, colonial and aggressive visitors. We predicted social and aggressive pollinators would be less susceptible to ant patrolling. However, these predictions were only partially supported by our results. Contrary to our initial predictions, we found no changes in pollinator composition between control and ant excluded plants (Fig. 2). Contrary to our initial predictions, only the introduced, aggressive, social pollinators (Apis mellifera) showed behavioural changes by increasing visit duration in ant excluded plants (Fig. 3). However, consistent with our predictions, non-aggressive solitary pollinators showed higher likelihood of anti-predatory behaviours. We found significant differences between solitary and social pollinators in avoidance and inspection behaviours rates, although no significant differences were found between pollinator taxa in response to ant exclusion (Fig. 3, Table 1). Apis mellifera and butterflies differed in their likelihood of avoidance and inspection behaviours, with butterflies more likely to avoid flowers (e.g. approach or hover over a flower without landing in it) (Fig.

130 Chapter 4: Ecological costs 3d), and also more likely to display inspection behaviour (Fig. 3c), regardless of the ant exclusion treatment.

Our results on pollinator composition are consistent with the general pattern revealed in a meta-analysis by Romero et al. (2011) showing that pollinator lifestyle (social vs. solitary) is not a good predictor of anti-predatory sensitivity. Overall, our findings suggest that the predation risk represented by ant patrolling in T. velutina is not strong enough to affect pollinator composition or increase the natural avoidance and inspection rates of pollinators, but provides evidence highlighting the different responses between pollinator’s lifestyles. Solitary pollinators are more inclined to display inspection behaviours and avoid flowers more often, regardless of ant patrolling. This suggests they are more sensitive and perhaps more susceptible to predation risk than social pollinators.

In T. velutina we found that ant patrolling halved the time Apis mellifera spent inside flowers, from 7.98 ± 1.69 sec in control plants to 16.04 ± 2.83 sec in ant-excluded plants. Although the effect of ant exclusion on visit duration differed between pollinator taxa since ant patrolling did not affect the time butterflies spend inside flowers (Table 1, Fig. 3b). Our results are comparable with those for F. wislizeni plants, in which ant patrolling decreased the duration of the visits, with foraging bouts being 20 s longer on plants without ants (Ness 2006).

We found no significant effect of ant exclusion on pollinator visitation in either butterflies or Apis mellifera (Fig. 3a). A meta-analysis on the effects of predation risk on pollinators found a non-significant trend for reduced visitation frequency in solitary pollinators, and reduced duration in social species (Romero et al. 2011). This result was consistent with our findings in T. velutina, where ant patrolling had a non-significant negative trend on the

131 Chapter 4: Ecological costs visitation frequency of butterflies (solitary pollinators) (Fig. 3a, Table 1) and a significant negative effect on the time Apis mellifera bees spent inside flowers (Fig. 3b, Table 1). These results suggest that both pollinators are capable of detecting ant patrolling as a predation risk but differ in their anti-predatory response behaviours. The non-significant trend on visitation frequencies of solitary pollinators in our data is consistent with the overall data pattern suggesting that the effect of ant patrolling on solitary pollinators may be smaller than the effect on social pollinators. Perhaps because solitary pollinators are more risk-averse, the effect of ant exclusion is smaller than the effect on social pollinators who seem to be more risk-prone. We believe this reflects a real biological pattern, rather than a spurious result although larger sample sizes may be required to detect a significant effect.

Effects of ant patrolling on plant mating system Although the deleterious effects of selfing and geitonogamy are well known (Waser and Price 1991, de Jong et al. 1992, Lloyd 1992), it is less well known the extent to which bisexual flowers suffer from them because it requires partitioning pollination events into intraflower (hereafter selfing), intraplant (hereafter geitonogamy), and interplant (hereafter outcrossing) (de Jong et al. 1992, Schoen and Lloyd 1992, de Jong et al. 1993, Eckert 2000, Wu et al. 2018), a methodological approach rarely undertaken (Wu et al. 2018). Furthermore, the full decomposition of pollen on stigmas into donor components (intraflower, intraplant, interplant) has rarely been performed, and its importance remains underappreciated (de Jong et al. 1993, Wu et al. 2018). For these reasons in this study we assessed the effects of ant patrolling on plant mating systems and male fitness decomposing pollen flow by origin and fate and considering selfing, geitonogamy, and outcrossing.

In this self-compatible species, ant exclusion shifted the plant mating system from predominantly outcrossing to predominantly selfing, and reduced

132 Chapter 4: Ecological costs geitonogamy (Fig. 4b). In bisexual flowers self-pollination can be mediated by pollinators (Wu et al. 2018). We suggest that ant patrolling reduced selfing rates in the bisexual flowers of T. velutina by affecting pollinator visitation behaviour, reducing the time bees spent inside the flowers.

Ant exclusion in T. velutina significantly increased the stigmatic pollen load by 150%, (Fig. 4a) and had strong impacts on the direction of pollen transfer, affecting the donor and destination of pollen grains being transferred (Fig. 4c). The number of pollen grains exchanged (donated and received) between the reciprocal pair plants was not significantly different between control and ant excluded plants (Fig. 4c, Table 1). Although ant patrolling significantly decreased the number of pollen grains sired per flower (Fig. 5a), it did not affect the transfer ratio of received:donated pollen grains (Fig, 4c). This is probably because ant patrolling also decreased pollen load, and so, the received:donated pollen ratio was maintained.

When contrasting our results from the mating system analyses (which show the proportion of pollen from different origins received by a stigma) with data on pollen flow (showing counts of pollen grains from different origins) it becomes evident that the increase in pollen load under ant exclusion is driven by selfing and allogamous pollen from other undyed, surrounding, T. velutina plants. This comparison also reveals that the number of selfing pollen grains on ant excluded stigmas is much higher, and this drives the change in the mating system from outcrossing to selfing. Yet, ant exclusion also increased the absolute number of allogamous pollen grains received by the stigmas.

Higher numbers of selfing pollen grains on focal stigmas from ant excluded plants (Fig. 4c) can be explained by bees staying longer inside the flower (Figs. 3d), foraging on pollen and nectar and hence increasing the transfer of pollen from the anthers to the stigmas of the same flower. Longer

133 Chapter 4: Ecological costs visit durations to ant excluded plants also increase the chances of bees depositing allogamous pollen grains (interplant or outcrossing) attached to their bodies. This would explain the increased number of outcrossing pollen grains from unknown, undyed, donors that we observed on ant excluded plants (Fig. 4c). The significant increase in the number of geitonogamous pollen grains received by ant patrolled stigmas (Fig. 4c) is consistent with the pollinator relocation hypothesis, which proposes that ant patrolling mildly deters pollinators, causing them to move to a nearby flower and hence increasing the rate of geitonogamy in the plant (Altshuler 1999, Romero and Koricheva 2011).

Our results suggest that ants contribute to maintain an outcrossing mating system in this self-compatible species, which has a beneficial effect on the quality of the progeny by promoting genetic diversity, especially given that this is a self-compatible species (Sosenski et al. 2016). The magnitude of the effect of ant patrolling on outcrossing that we found in T. velutina (Fog. 4) was comparable to the effect of distyly on outcrossing in Fagopyrum (Wu et al. 2018). Ant patrolling in T. velutina and distyly in Fagopyrum promoted a significant two-fold increase in outcrossing rates. Selfing accounted for 71.3% of the pollen load on the homostylous buckwheat Fagopyrum tataricum, which was twice as much as the proportion of self-pollen found on stigmas of the distylous congener Fagopyrum esculentum (35.3% - 39.6%), as expected from the greater degree of herkogamy in the distylous species (F. esculentum) (Wu et al. 2018). Plants are known to have a diversity of physical, mechanical, and chemical traits to avoid selfing and promote outcrossing, but we suggest that ant patrolling may by another mechanism to promote outcrossing in ant-plants, although further research on the mating systems of ant-plants is required.

134 Chapter 4: Ecological costs Effects of ant patrolling on plant fitness In order to fully understand the effect of biotic interactions on plant fitness, both female and male fitness components should be evaluated since the magnitude and direction of the effects of mutualists and antagonists may differ between plant male and female reproductive functions (Schaeffer et al. 2013, Carper et al. 2016). Male fitness is expected to increase as the number of mates reached by the pollen increases. Pollinator behaviour has strong impacts on pollen transfer, with longer visits increasing pollen export and pollen deposition on the stigma (Carper et al. 2016). Consequently, since male fitness is more constrained by the number of mates than female fitness, we expect male fitness to be more susceptible to changes in pollinator behaviour (Krupnick and Weis 1999, Schaeffer et al. 2013, Carper et al. 2016). To date, very little has been done to assess the effect of ant patrolling on female and male reproductive functions. Our experimental design allowed us to trace the consequences of ant patrolling on pollen transfer dynamics, estimate pollen transfer on a per flower basis, and estimate the effect of ant patrolling on male reproductive function on a per flower and per plant basis.

In T. velutina ant exclusion significantly increased the number of pollen grains donated per flower by a six-fold (Fig. 5a) suggesting guarding ants may be hindering male fitness. The full decomposition of donated pollen grains by fate revealed that most of this trend is driven by pollen donated towards selfing (Fig. 5b), which contrasts with stigmas from ant patrolled flowers which were dominated by outcrossing pollen (Fig. 5b). Our results are consistent with other ant-plant case studies. Ant patrolling had negative consequences on the female fitness component in Opuntia imbricata, reducing seed count per fruit by 30% and seed mass by 16%, although these effects varied between ant species (Ohm and Miller 2014). Plants tended by Crematogaster opuntiae ants had lower seeds per fruit, but not those tended by Liometopum apiculatum ants, suggesting that L. apiculatum benefits plant fitness by decreasing the proportion of fruits infested by the pre-dispersal seed predator (Ohm and Miller 2014). Ant

135 Chapter 4: Ecological costs patrolling in H. physophora had negative effects on female and male plant fitness: first, by castrating the host by consuming floral buds, and second, by deterring pollinators, reducing pollen transfer and causing pollinator limitation which reduced fruit set in those buds that escaped ant predation (Malé et al. 2012).

Overall, the effect of ant patrolling on T. velutina’s fitness differs between male and female functions. On the one hand, guarding ants reduced male fitness by reducing the number of pollen grains donated per flower, but on the other hand they may increase female fitness by promoting outbred seeds and so increasing the quality of the offspring. These findings exemplify how other non- pollinators insects interacting with plants and its pollinators may have different effects on male and female plant reproduction, and highlight the importance of considering both fitness components.

Pollinator efficiency and net ant effects The contribution of a given pollinator to plant fitness is determined by its visitation frequency and its efficiency at carrying and depositing pollen grains on conspecific stigmas (Ne'eman et al. 2010, Willmer 2011 pp. 279-287, King et al. 2013). Pollination efficiency is a function of several morphological and behavioural traits (Ollerton et al. 2007, Ne'eman et al. 2010, Willmer 2011pp. 279- 287, Willmer and Finlayson 2014, Barrios et al. 2016, Willmer et al. 2017). The relationship between size and structure of floral visitors’ body and mouthparts with floral morphology are crucial factors determining their efficiency as pollinators (Armbruster 1985, 1988, Ordano et al. 2008, Barrios et al. 2016). In the self-incompatible plant Angadenia berteroi (Apocynaceae) which relies on pollinators to set fruit, lepidopterans were the most common visitors although very inefficient pollinators that seldom carry pollen because their long, thin tongues reach the nectar but miss the plant sexual organs and consequently their visits did not lead to fruit set (Barrios et al. 2016). This contrasted with long-tongued bees who were less frequent visitors, but better pollen carriers

136 Chapter 4: Ecological costs whose visits lead to fruit set because they push their mouthparts against the anthers to reach for nectar, collecting more pollen on their wide and shorter tongues (Barrios et al. 2016). In fact, butterflies have been reported as poor pollinators that act mostly as nectar thieves in other systems (Venales and Barrows 1985). We observed a similar pattern in T. velutina with butterflies being more abundant than bees, but where changes in the behavioural patterns of bees in ant excluded plants (Fig. 3) seem to be driving changes in plant mating systems (Fig. 4) and fitness (Fig. 5). Based on these observations we suggest that butterflies may be less efficient pollinators than Apis mellifera in T. velutina, although further studies are required to test this hypothesis.

The effect of defensive mutualists on the host plant pollination and fitness can vary depending on whether predators deter efficient pollinators or a subset of inefficient visitors (Romero and Koricheva 2011). Guarding ants had negative effects on plant fitness when they attacked efficient pollinators in Ficus pertusa (Moraceae: Bronstein 1991) and Opuntia imbricata (Cactaceae:Ohm and Miller 2014), but had positive effects in Banistriopsis malifolia (Malpighiaceae:Alves‐Silva et al. 2013) with wasps efficiently protecting flowers from herbivory without interfering with pollinators. Yet, there are still many cases in which their effect is unknown because deterrence has not been assessed for all pollinators, as in Heteropterys pteropetala (Malpighiaceae), where only large bees (Centris and Epicharis) are efficient pollinators, whilst small bees are mostly thieves; but it still remains unknown whether patrolling ants deter all or only a certain subset of these bees (Assunção et al. 2014). Even when deterrence is known for different floral visitors, the effects of patrolling ants on plant fitness may be unclear if the pollinator efficiency of those floral visitors has not been determined, as is the case of T. velutina.

Contrary to our initial prediction, the effect of ant patrolling on T. velutina via pollinator deterrence may benefit plant fitness by reducing

137 Chapter 4: Ecological costs pollinator visit duration, which promoted pollinator relocation and lead to reduced selfing rates and higher outcrossing rates. Although T. velutina has self- compatible hermaphroditic flowers, it has a suite of traits to favour outcrossing such as being a monomorphic, herkogamous species with pollinator rewards and attraction features (Sosenski et al. 2016). Although ant patrolling does reduce pollen load and male fitness, far from having an ecological cost on the host plant pollination and mating system, ant patrolling seems to be another mechanism contributing to promote outcrossing in T. velutina.

Conclusions This study provides a comprehensive picture of the interaction between myrmecophily and pollination by showing the ecological and behavioural effects of ant patrolling on pollinators that affect plant mating systems and lead to fitness consequences. Ant patrolling in T. velutina decreased pollinator visit duration, increased pollen deposition to stigmas, maintained outcrossing as the predominant mating system, and decreased male fitness, but assured outbred progeny. However, its effects on plant male and female fitness varied. Overall, we conclude ant patrolling is not an ecological cost for the host plant reproductive outcome as it promotes outcrossing in this self-compatible species, which has a beneficial effect on the progeny by promoting genetic diversity. This study contributes towards understanding the potential of plant mating systems to be shaped and evolve in response to ant-pollinator interactions and considers pollination from a complex multispecies approach, as product of interactions several simultaneous interactions.

Plant mating system seems to be the main determinant of population genetic structure (Loveless and Hamrick 1984, Vekemans and Hardy 2004) and changes in pollinator availability are likely to affect the evolution of plant mating systems and hence, population genetic structures (Opedal et al. 2017).

138 Chapter 4: Ecological costs Differences in ant aggressivity between T. velutina populations (Zedillo- Avelleyra 2017) could be affecting plant mating systems differently. Consequently, it would be worth exploring the effect of ant aggressivity on shaping long-term mating system traits such as herkogamy and the genetic structure of plant populations.

139

Chapter 5: The Distraction Hypothesis of EFN

CHAPTER 5

THE DISTRACTION HYPOTHESIS OF EXTRAFLORAL NECTAR

Does EFN distract ants from entering flowers and reduce ant-pollinator conflict?

140

Chapter 5: The Distraction Hypothesis of EFN INTRODUCTION

Ant-plants recruit ants by providing nesting sites and/food resources, and benefit from ant-mediated reduction in damage by herbivores and pathogens (Janzen 1966, Bentley 1977a, Chamberlain and Holland 2009, Herrera and Pellmyr 2009, Rosumek et al. 2009, Trager et al. 2010). The most widespread reward produced by ant-plants is extrafloral nectar (EFN), a key food resource for ants (Rudgers and Gardener 2004, Ochoa Sánchez 2016) that increases individual survivorship (Fisher et al. 1990), colony growth rate and reproductive output (Byk and Del-Claro 2011). Guarded plants in turn show increased somatic growth (biomass or leaf production) and reproductive output (Heil et al. 2001a, Chamberlain and Holland 2009, Rosumek et al. 2009, Trager et al. 2010, Escalante-Pérez and Heil 2012).

Most ant-plants are angiosperms (Keeler 2014), and many require the services of animal pollen vectors for seed set (Bentley 1977b, Torres-Hernández and Rico-Gray 2000, Raine et al. 2002, Díaz-Castelazo et al. 2005, Rico-Gray and Oliveira 2007b, Ballantyne and Willmer 2012, Dutton et al. 2016, Villamil et al. 2018, amongst many other studies documenting animal-pollinated ant-plants), making ants and pollinators likely to co-occur on a given host plant. This raises the possibility of several types of direct and/or indirect conflicts between ants and pollinators. First, indirect conflict can arise if there is a trade-off between plant allocation of resources to reproduction (which benefits pollinators) versus investment in growth and defence (which benefits ant guards) (Bazzaz et al. 1987). Plants that do not reproduce grow faster and develop larger resource- acquiring and producing organs (roots and leaves) (Frederickson 2009), leading to indirect conflict between ants and pollinators over plant resources and rewards (Afkhami et al. 2014, Dutton et al. 2016). In extreme cases of ant- pollinator conflict, ants actively increase plant investment towards growth and defence by castrating their host plant through consumption of floral meristems (Frederickson 2009, Palmer et al. 2010) or mature inflorescences (Izzo and

141 Chapter 5: The Distraction Hypothesis of EFN Vasconcelos 2002). Second, ants may enter flowers and consume floral nectar without providing pollination services, providing no benefit to the plant and potentially reducing the attractiveness of flowers to effective pollinators (Rico- Gray and Oliveira 2007b). Third, ant visits to flowers may reduce pollen viability by depositing antimicrobial substances that decrease pollen germination rates, and hence decrease male fitness for the plant (Wagner 2000, Dutton and Frederickson 2012). Finally, ants may attack or intimidate pollinators directly (Wagner and Kay 2002, Willmer et al. 2009, Villamil et al. 2018), reducing flower visitation rates (Ness 2006, Lach 2008) or duration (Villamil et al. 2018). One hundred and forty years ago, Anton Joseph Kerner, an Austro-Hungarian botanist, wrote:

“Of all the wingless insects it is the widely dispersed ants that are most unwelcome guests to flowers. And yet are they the very ones which have the greatest longing for the nectar, as numberless observations sufficiently show.” (Kerner 1878, p21)

Whilst ants may be unbidden floral visitors, they are also effective bodyguards (Bentley 1977a, Chamberlain and Holland 2009, Rosumek et al. 2009, Trager et al. 2010), which may represent an ecological trade-off for ant- plants. Given that ant guards can have both costs and benefits for different aspects of plant fitness, we expect natural selection to have acted on ant-plant traits to minimise the negative impacts of ants relative to the protection they provide, ameliorating the negative consequences of ant-pollinator antagonism for plant fitness (Raine et al. 2002). A wide range of mechanisms have been interpreted as achieving this by reducing ant visitation to flowers during anthesis, including physical barriers (Harley 1991, Galen 1999, Galen and Cuba 2001, Carlson and Harms 2007, Willmer 2011), chemical repellents (Willmer and Stone 1997, Junker et al. 2007, Agarwal and Rastogi 2008, Junker and Blüthgen

142 Chapter 5: The Distraction Hypothesis of EFN 2008, Willmer et al. 2009, Ballantyne and Willmer 2012) or bribes (Kerner 1878, Willmer 2011, Martínez-Bauer et al. 2015). Physical barriers include spiny or hairy surfaces on the outside of the corolla or on floral pedicels that prevent tarsi from gripping and so hinder ant walking (Willmer 2011), and waxy or sticky plant secretions that prevent ants from climbing (Harley 1991). Bracts around the calyx can act as a water trap, creating a pool of water or mucilage that prevents ants and other small insects from crawling into the flowers (Carlson and Harms 2007). The shape of the flower may itself stop ants from entering flowers: pendant, thin, and constricted corollas are effective ant-excluding morphologies (Galen 1999, Galen and Cuba 2001, Willmer et al. 2009). Several species produce ant-repelling flowers (Junker et al. 2007, Agarwal and Rastogi 2008, Junker and Blüthgen 2008, Willmer et al. 2009) and furthermore, ant repellence may be concentrated in specific floral parts such as petals (Ness 2006, Ballantyne and Willmer 2012), or pollen and anthers (Willmer and Stone 1997, Ghazoul 2001, Raine et al. 2002, Willmer et al. 2009, Ballantyne and Willmer 2012). Finally, some species may entice ants away from flowers by offering alternative sugary rewards outside the flowers, using EFN as a distraction or bribe (Wagner and Kay 2002, Galen 2005, Chamberlain and Holland 2008, Willmer 2011). During the nineteenth century, Kerner (1878) suggested that EFN in plants with floral nectar might serve to distract ants from visiting the flowers.

“Any insects that creep along the stem must, if they would get at the flower, of necessity pass over this disk with its drop of nectar; thus what they would have found, in the flower, is already offered to them here in rich abundance. The creeping insects are not fastidious. They are content with that which is first offered, and so do not trouble themselves to climb farther up to the flowers. […] I do not therefore hesitate to interpret all nectar-glands that are found on leaves

143 Chapter 5: The Distraction Hypothesis of EFN as means of protection against the unwelcome, because unprofitable, visits of creeping insects.” (Kerner 1878, p137-139)

The idea that EFN could distract non-pollinating insects away from flowers and so reduce any disruption of pollination is known as the Distraction Hypothesis (Wagner and Kay 2002, Chamberlain and Holland 2008, Holland et al. 2011). In many ant-plants, flowers and extrafloral nectaries are in close proximity (Weber and Keeler 2013, Keeler 2014) and several species secrete extrafloral nectar only or predominantly during the flowering period (Bentley 1977b, Chamberlain and Holland 2008, Holland et al. 2010, Villamil-Buenrostro 2012, Villamil et al. 2013, Falcão et al. 2014, Dutton et al. 2016, Villamil 2017). For example, extrafloral nectaries on leaves associated with flowers of the Mexican ant-plant Turnera velutina (Passifloraceae) secrete more nectar with higher sugar content than extrafloral nectaries on leaves bearing buds and fruits (Villamil 2017). This increase in EFN secretion during anthesis is compatible with the Distraction Hypothesis, in that EFN secretion near flowers could lure ants that might otherwise enter flowers seeking floral nectar. However, the same floral behaviour and the frequent proximity of extrafloral nectaries to reproductive structures can also be explained by the Optimal Defence Theory (ODT), which predicts that plants should focus defensive investment on highly vulnerable and valuable tissues for plant fitness, such as flowers, fruits and seeds (Stamp 2003). Finally, it is possible that high EFN secretion on flowering shoots in myrmecophiles could fulfil both distracting and protective roles, simultaneously keeping ants out of flowers but promoting their patrolling around reproductive tissues to deter herbivores.

While the defensive role of ant recruitment through EFN secretion has been widely demonstrated (Chamberlain and Holland 2009, Rosumek et al. 2009, Trager et al. 2010), the Distraction Hypothesis has not been adequately

144 Chapter 5: The Distraction Hypothesis of EFN tested. To our knowledge, since Kerner proposed it in 1878 only three experimental studies have been performed and all have rejected it (Wagner and Kay 2002, Galen 2005, Chamberlain and Holland 2008). However, none of these studies were carried out in an ecologically realistic setting (a point that we address further in the Discussion).

QUESTIONS & AIMS

Here we use experimental manipulation of EFN secretion during anthesis in a Mexican endemic plant, Turnera velutina, to test the Distraction Hypothesis under natural conditions, improving on previously reported experimental designs. We evaluated the potential ecological and fitness consequences of the Distraction Hypothesis addressing the following questions:

(i) How often are flowers occupied by ants and how many ants are found in them?

(ii) Does preventing EFN secretion affect the number of ants patrolling extrafloral nectaries, the number of ants inside the flowers or the number of pollinators visiting flowers?

(iii) Does preventing EFN secretion increase the probability of a flower being occupied by ants?

(iv) Does the number of ants at extrafloral nectaries or inside flowers affect pollinator visitation?

(v) Does preventing EFN secretion affect plant fitness?

145 Chapter 5: The Distraction Hypothesis of EFN HYPOTHESES AND PREDICTIONS

If the Distraction Hypothesis is true, we predict that experimental elimination of extrafloral nectar secretion should: 1) reduce ant visitation to extrafloral nectaries, 2) increase the numbers of ants inside flowers, 3) increase the proportion of flowers occupied by ants, 4) leading to decreased levels of floral visitation by pollinators, and 5) a reduction in plant fitness.

MATERIALS AND METHODS Study site and system All experiments and observations were conducted in the stabilised coastal sand dunes at the CICOLMA Field Station in La Mancha, Veracruz, in the Gulf of Mexico. Within this population, we selected four sites with high densities of Turnera velutina (Passifloraceae). At La Mancha, Turnera velutina establishes facultative mutualisms with at least thirteen ant species (Cuautle et al. 2005, Zedillo-Avelleyra 2017) and its main herbivores are caterpillars of a butterfly, Euptoieta hegesia (Nymphalidae). Extrafloral nectar is provided in paired cup- shaped glands located on the underside of either the leaf blade and petiole (Fig. 1). Although it flowers year-round, flowering peaks during summer (Cuautle et al. 2005). Flowers last one day, are insect-pollinated (Sosenski et al. 2016) and have a yellow, pentamerous, campanulate corolla with nectar easily accessible at the base. Honeybees are the dominant pollinators (Apis mellifera) at La Mancha, accounting for 94% of visits (Sosenski et al. 2016, Villamil et al. 2018).

FIELDWORK METHODS Surveys of ants inside flowers We quantified ant occupancy in flowers of T. velutina by surveying 1604 flowers across four sites within CICOLMA in November 2014. Flowers at each site were

146 Chapter 5: The Distraction Hypothesis of EFN observed for 2 minutes every hour throughout the whole anthesis period (0830– 1230 hours), with one observer at each site. We estimated the proportion of flowers occupied by ants, and the total number of ants across occupied flowers within a site. Flowers were sampled at the same site over multiple days, for 10 site-and-day combinations. Since these are one-day flowers, we considered each site-day as a replicate (n = 10 site-days), with site-and-day effects incorporated into our statistical modelling (see below).

Experimental manipulation of EFN secretion To test the Distraction Hypothesis we experimentally clogged extrafloral nectaries to prevent nectar secretion and compared ant and pollinator behaviours on paired shoots with and without EFN secretion. This experiment was conducted over 5 days during November 2014. Early on each day of the experiment, a pair of neighbouring, unopened floral buds within a plant were marked as either control or clogged treatments (n=216 flowers; n = 108 pairs, n = 108 plants). EFN secretion on clogged treatment leaves was eliminated by sealing the nectary cup with a droplet of transparent acrylic textile paint (Mylin dimensional, Mexico). On control treatment leaves we applied similarly sized droplets of the same textile paint a couple of millimetres above the gland (Fig. 1), controlling for any effects of the acrylic paint itself. Pilot tests confirmed that the paint totally prevented EFN secretion and also that the paint did not deter ants or pollinators. We recorded the frequency and identity of ants (to genera or species level following Zedillo-Avelleyra 2017) and other insects visiting each flower pair and the associated extrafloral nectaries for two minutes every hour during anthesis (0830–1230 hours). Simultaneous observations were performed at each of three sites by different observers. For brevity, we refer to non-ant flower visitors as pollinators, while recognising that the efficacy of visits by all species mentioned in contributing to seed set in T. velutina remains to be demonstrated.

147 Chapter 5: The Distraction Hypothesis of EFN Based on the results from the clogging experiment described above (from now on referred to as the short-term experiment), we conducted a follow-up experiment in which treatment duration and spatial scale were both increased by a factor of 10, using paired branches and focusing on one flower on each control or clogged branch, rather than paired flowers on the same branch. We refer to this experiment from now on as the long-term experiment (See Supplementary material 1 for further details). The extrafloral nectaries of all ten leaves on the clogged treatment branches were sealed as described above, and the treatment was maintained for ten days (Fig. 1). Our hypothesis was that increasing both the temporal and spatial scales of our treatment would result in a larger experimental effect size. However, a comparison of the results from the short-term and long-term experiments showed that ants respond at a smaller scale (Supplementary material 1: Table S2), and we therefore focus on the results of the short-term experiment and highlight differences in results for the longer- term, larger-scale experiment where these are relevant to the Distraction Hypothesis. Full results for the long-term experiment are provided in the Supplementary material 1.

Impacts of EFN secretion on fitness To quantify the impact of clogging EFN secretion and the Distraction Hypothesis on plant fitness we collected the fruits resulting from experimental flowers (control and clogged) at which pollinator visitation was observed. We recorded whether those flowers developed into fruits with seeds or whether they were aborted, and counted the number of seeds per fruit. All fruits were collected at least one week post-anthesis, at which stage retained fruits can be distinguished from aborted fruits, and developing seeds can be counted distinguishing viable from unviable seeds, even if still immature.

148 Chapter 5: The Distraction Hypothesis of EFN

Figure 1. Images showing a) an apex of Turnera velutina bearing an apical flower bud and two lateral fruits, b) the location of extrafloral nectaries on the underside of a leaf, c) a comparison of clogged and control leaves, and d) the spatial arrangement of the short-term and long-term experiments.

STATISTICAL ANALYSES All statistical analyses were conducted in R version 3.23 (R Core Team 2016). Mixed effects models were fitted using ‘lme4’ (Bates et al. 2016) or ‘MCMCglmm’ (Hadfield 2010) R packages.

Surveys of ants in flowers To test if the proportion of flowers with ants inside them changed over the anthesis period, we fitted a binomial mixed model with time of day as a fixed effect. Flowers of T. velutina last for a single day, and because multiple flowers

149 Chapter 5: The Distraction Hypothesis of EFN were sampled on a given site on a given day, we fitted site identity as a random effect to account for differences between site and day variation in variables that could influence ant abundance, such as resource availability, ant diversity, or the abundance of ant nests. Tukey post hoc comparisons were used to test differences between hours using the ‘multcomp’ R package (Hothorn et al. 2008). To test if the number of ants inside occupied flowers changed over the anthesis period we fitted a Poisson mixed model, using the number of ants inside flowers per site as the response variable and fitted as fixed effects time of day as a linear and as a quadratic term. The number of flowers occupied by ants was fitted as a log-transformed offset to control for ant density in flowers, which is likely to decrease in sites with more flowers occupied by ants, since we recorded counts per site rather than counts per individual flower (see fieldwork methods). Time of day was fitted as a linear and as a quadratic term to investigate the shape of the activity pattern of ants in flowers relationship between the number of ants inside flowers through the day. We fitted site identity as a random effect to account for variation that could influence ant abundance (as detailed above). We also included an observation-level random effect where each data point receives a unique level of a random effect to control for overdispersion (Hinde 1982). Tukey post hoc comparisons were used to test differences between hours using the ‘multcomp’ R package (Hothorn et al. 2008).

Ecological consequences of EFN secretion Five mixed effects models (i-v) were fitted to test the ecological consequences of the Distraction Hypothesis. Because all of these models had the same random effects structure unless otherwise specified, we detail the random effects first and then describe the fixed effects for each model. Flower identity was fitted as a random effect to account for repeated hourly

150 Chapter 5: The Distraction Hypothesis of EFN observations. Because this experiment had a paired experimental design we fitted flower pair identity as a random effect to control for between-pair variation in floral and extrafloral investment. We also included an observation- level random effect where each data point received a unique level of a random effect to control for overdispersion. We fitted the following models, and have structured our results following the same order:

(i) To test the effect of nectary clogging on the number of ants we fitted a Poisson mixed effects model using number of ants as the response variable. Ant location (at extrafloral nectaries or in flowers), treatment, and the interaction between these two factors were fitted as fixed effects. Tukey tests were conducted to test differences between the number of ants at extrafloral nectaries or flowers under control or clogged gland conditions.

(ii) To test whether preventing EFN secretion by clogging the glands increased the probability of flower occupancy by ants, we fitted a binomial mixed effect model with the presence or absence of ants in a flower as a response variable. Clogging treatment was fitted as a fixed effect. The observation-level random effect was omitted.

(iii) To test the effect of clogging EFN secretion on pollinator visitation we fitted a Poisson mixed model using the number of pollinators as the response variable and treatment as the only fixed effect.

(iv) To test the effect that the total number ants had on pollinator visitation (regardless of their location in flowers or at extrafloral nectaries), we fitted a Poisson mixed model using number of pollinators as the response variable. As fixed effects we fitted the total number of ants, and treatment to test whether treatment affected pollinator visitation in a way that was unlinked to the number of ants.

151 Chapter 5: The Distraction Hypothesis of EFN

(v) To test if the location (inside flowers or at extrafloral nectaries) and number of ants had an effect on pollinator visitation, we fitted a Poisson mixed model. The number of pollinators was fitted as the response variable, whilst treatment, number of ants in flowers, and number of ants at extrafloral nectaries were fitted as fixed effects.

Data from the long-term experiment were analysed following a similar model structure reported for the ecological consequences models (S.i-v) (see Supplementary material).

Impacts of EFN secretion on plant fitness (vi) To test the effect of clogging on fruit abortion rates, we fitted a binomial mixed model, with clogging treatment as the fixed effect and pair identity as a random effect.

(vii) For those fruits that developed seeds, we tested the effect of clogging on the number of seeds by fitting a Poisson mixed model. Clogging was fitted as a fixed effect and as random effects we fitted pair identity and an observation- level random effect to account for overdispersion.

Exploring the responses and effects of different ant species We fitted additional models aiming to explore differences between ant species in their response to clogging and in their effects on pollinator visitation. We investigated whether ant species differed in their response to clogging (see model (S.vi) in Supplementary material), and whether different ant species patrolling the plants or inside the flowers differed in their effect on pollinator visitation (see models (S.vii, S.viii) in Supplementary material). These models allowed us to: (1) estimate the effects of individual ant species on pollinator visitation; (2) account for plants occupied by multiple ant species, and (3) capture the variation in ant abundance within a given ant species.

152 Chapter 5: The Distraction Hypothesis of EFN Effect sizes Cohen d effect sizes for all models were calculated using the likelihood ratio tests (LRT) statistics from each model. To test whether increasing the duration and scale of the clogging treatment by a factor of 10 had a larger effect on the number of ants and pollinators, we estimated the ratio of change in the effect size between the short- and long-term experiment for each type of visitor (See Supplementary material).

RESULTS Surveys of ants in flowers We observed 10 ant species from four subfamilies interacting with T. velutina: Dorymyrmex bicolor (Dolichoderinae), Camponotus planatus, Camponotus mucronatus, Camponotus novogranadensis, Brachymyrmex sp., and Paractrechina longicornis (Formicinae), Cephalotes sp., Crematogaster sp., and Monomorium ebenimum (Myrmicinae), and Pseudomyrmex gracilis (Pseudomyrmicinae). We observed ants associated with plants to vary spatially, we assume due to variation in the proximity of nests of different species. Though not formally quantified, we observed apparent differences in ant behaviour among species. Some species patrolled individually, such as Camponotus planatus, C. mucronatus and C. novo-granadensis, while others were gregarious, such as Dorymyrmex bicolor, Brachymyrmex sp., and Paratrechina longicornis. Monomorium ebenimum probably provides no guarding services to T. velutina since they have only been observed consuming floral nectar and not patrolling elsewhere. In addition, this species belongs to a worldwide genus of floral nectar thieves (Ettershank 1966, Bolton 1987). Feeding preferences also vary among ant species, from opportunistic carnivores such as Pseudomyrmex gracilis found inside flowers hunting for thrips and beetles, to omnivores such as

153 Chapter 5: The Distraction Hypothesis of EFN Crematogaster sp. that harvest elaiosomes attached to T. velutina´s seeds (S. Ochoa-López pers. comm., Dec. 2014).

Across all four sites and over all time intervals, surveys of the frequency and abundance of ants inside flowers revealed that 9.30 ± 0.19 % of the flowers within a site were occupied by ants, with an average density of 2 ± 0.28 ants/occupied flower. The low proportion of flowers (Fig. 2a) with low numbers of ants (Fig. 2b) was constant throughout the anthesis period (Tables 1,2). The number of ants inside flowers did not vary significantly through daily time and we found no statistical support for any quadratic effect (Table 2).

Floral visitors All but one of 202 visits to T. velutina flowers by non-ant visitors were made by other insects (Table 1). The only exception was a single visit by a hummingbird (Trochilidae). Of the insect visits, 90.5% were by honeybees, Apis mellifera, with most of the remainder by native bees and butterflies.

154 Chapter 5: The Distraction Hypothesis of EFN a)

b)

Figure 2. a) Proportion of flowers with ants inside them and b) number of ants per flower throughout the anthesis period (mean ± SE per site) in hourly observations (n = 42 observations, from 10 sites).

155

Chapter 5: The Distraction Hypothesis of EFN

Table 1. Taxonomic identities of floral visitors recorded in the short-term clogging experiment. Taxa with the epithet ‘sp.’ were identified only to genus, but all the individuals belong to the same morphospecies.

Taxon Number of visitors Subfamily At EFN In flowers Dorymyrmex bicolor 373 10 Dolichoderinae Brachymyrmex sp. 342 74 Formicinae Paratrechina longicornis 166 3 Formicinae Camponotus planatus 128 1 Formicinae

Camponotus mucronatus 41 4 Formicinae Camponotus sp. 49 5 Formicinae Camponotus 3 1 Formicinae novogranadensis Crematogaster sp. 26 1 Myrmicinae Ants at EFN Cephalotes sp. 16 0 Myrmicinae Monomorium ebenium 58 11 Myrmicinae Pseudomyrmex gracilis Pseudomyrmicina 18 0 e Unidentified ants 13 0 ? Apis mellifera 183 Native bees (Apoidea) 12 Diptera 1 Lepidoptera 4

Floral visitors Floral Wasps 1 Hummingbird 1

156

Chapter 5: The Distraction Hypothesis of EFN Ecological consequences of EFN removal Numbers of patrolling ants were significantly affected by EFN treatment (clogged versus control), ant location, and the interaction between these factors (Fig. 3a; Table 2). Ten times more ants were found patrolling extrafloral nectaries (1.49 ± 0.079 ants) than were found inside flowers (0.14 ±0.02 ants) (Fig. 3a; Table 2), regardless of treatment (control: Z = -17.23, P < 0.001; clogged: Z = -14.03, P < 0.001). The effect of eliminating EFN secretion on the number of ants differed between extrafloral nectaries and flowers, resulting in a significant decline in numbers of ants patrolling extrafloral nectaries (Z = -4.22, P < 0.001), but no significant change in the numbers of ants observed inside flowers (Z = 1.05, P = 0.705; Fig. 3a; Table 2). The percentage of flowers occupied by ants increased significantly from 6.1% under the control treatment to 9.7% when extrafloral nectaries were clogged (Table 2).

Numbers of flower visitors were not significantly affected by the elimination of EFN secretion (Fig. 3a; Table 2), nor was there any significant interaction between visitor numbers and the total number of ants (Table 2). When ant abundance was partitioned by location on the plant (at extrafloral nectaries or in flowers), neither the number of ants patrolling extrafloral nectaries nor the number of ants inside a flower had a significant effect on the number of flower visitors (Table 2).

In all five models used to analyse the short-term experiment (one leaf, one day), differences between individual plants (captured by the pair random effect) explained the largest proportion of variation in the numbers of ants and pollinators (Table 2). Differences between individual flowers (captured by the flower random effect) or random variation between observations (captured by the OLRE random effect) explained smaller proportions of variation in the numbers of ants or pollinators (Table 2).

157 Chapter 5: The Distraction Hypothesis of EFN

Figure 3. Mean numbers of visitors to flowers of Turnera velutina (mean ±1 SE) recorded in hourly surveys during 2 minutes of observation per flower for the short-term experiment. Clogged treatment flowers had secretion of EFN prevented by clogging the associated extrafloral nectaries. The short-term experiment involved prevention of EFN secretion associated with one flower for one day (see Fig.1). Red circles represent ants at extrafloral nectaries; blue triangles represent ant in flowers, and green squares represent pollinators.

Impacts of EFN secretion on plant fitness Clogging had a marginally significant effect (P = 0.059) on fruit abortion, increasing by 12 % the probability of abortion in flowers associated with leaves in which EFN had been clogged (Fig. 4a, Table 2). Despite the P-value being marginally significant, clogging had a considerable Cohen d effect size (Cohen 1988) on fruit abortion (Tables 2,3). However, clogging had no effect on the number of seeds per fruit (Fig. 4b, Table 2) with a small Cohen d effect size between treatments (Cohen 1988) (Table 3). a)

158 Chapter 5: The Distraction Hypothesis of EFN

b)

Figure 4. Effects of clogging the extrafloral nectaries on a) the probability of fruit abortion (mean ± se), and b) the number of seeds (mean ± se) produced by Turnera velutina.

159 Chapter 5: The Distraction Hypothesis of EFN

Table 2. Estimates and likelihood ratio test results for statistical models used to test ant occupation of flowers, and the ecological and plant fitness consequences of clogging EFN secretion on Turnera velutina. Ant location stands for the number of ants patrolling EFN or inside flowers. The values highlighted in bold are statistically significant (* = P < 0.05; ** = P < 0.01, *** = P < 0.001; NS = non-significant, P > 0.05), OLRE stands for observation-level random effect.

Fixed P- Random Variance SD Experiment Model Response Estimate LRT effects value effects Proportion of flowers with Time of day -0.0127 -0.017 0.98 Site 0.29 0.53 ants inside Surveys of ants in log (Flowers Number of 0.6362 18.55 1.64-05 *** flowers with ants) Site 5.09-09 0 ants per Time of day 0.3288 1.54 0.21 OLRE 0.20 0.45 flower Time of day2 -0.0771 1.12 0.28

Clogging -0.5314 12.42 0.0004 *** Flower 0.39 0.62 i) Number of Ant location -0.2772 647.09 2.2-16 *** Pair 1.04 1.01 ants Clogging × 0.7656 13.28 0.0002 *** OLRE 0.28 0.53 Ant location Clogging: Proportion of Ecological flower Flower 0.38 0.62 ii) Clogging 0.7669 4.61 0.031 * consequences occupied by Pair 5.81 2.41 ants Flower 0 0 Number of iii) Clogging -0.0749 0.33 0.56 Pair 0.89 0.94 pollinators OLRE 0.01 0.12 160

Chapter 5: The Distraction Hypothesis of EFN

Number of Flower 2.02-08 1.42-05 iv) ants inside Clogging 0.4413 3.41 0.06 ~ Pair 4.36 2.08 flowers OLRE 0.72 0.85 Flower 0 0 Number of Clogging -0.0947 0.52 0.46 iv) Pair 0.86 0.92 pollinators Total ants -0.0569 1.97 0.16 OLRE 0.01 0.13 Clogging -0.1336 1.01 0.31 Flower 1.53-8 1.23-05 v) Number of Ants at EFN -0.0967 4.20 0.04 * Pair 0.83 0.91 pollinators Ants in 0.1897 1.63 0.20 OLRE 0.019 0.13 flowers

Clogging -0.1848 0.220 Flower 0.023 vi) Number of Ants at EFN -0.1219 0.099 Pair 1.228 pollinators Ants in Ant -0.1309 0.512 0.020 flowers species

Fruit abortion * Pair 4-14 2-7 Clogging: vi) Clogging 0.6059 3.54 0.05 Impacts on Number of Pair 0.12 0.35 plant fitness vii) Clogging -0.1512 1.241 0.265 seeds OLRE 0.84 0.92

161 Chapter 5: The Distraction Hypothesis of EFN

Comparison of patterns across spatio-temporal scales In contrast to our prediction, increasing the duration and spatial scale of the clogging treatment by a factor of 10 did not result in larger effect sizes on ant behaviours (Table 3). In fact, the long-term clogging experiment had less impact on ant patrolling than the short-term clogging experiment, resulting in smaller effect sizes on numbers of ants at extrafloral nectaries, numbers of ants inside flowers, and the proportion of flower occupancy by ants (Table 3). The impact of preventing EFN secretion on the number of inside flowers changed from positive at a short-term, local scale to negative in the long-term, branch-scale experiment (10 leaves, 10 days) (Table 3).

Ant species-specific responses to clogging and effects on pollinators Although ant species explained only 0.21 % of the variation in the number of ants inside flowers (Model S.vi in Table S3), there was variation between ant species in responses to clogging (Fig. S4a). Brachymyrmex sp. ants were the most abundant ants found inside flowers, as shown by ant surveys (Table 1) and predicted by the non-zero-overlapping effects in the model (Fig. S4a). While effect estimates vary for other ant species, confidence intervals for all taxa other than Brachymyrmex sp. overlap with zero (see Table 1 for rank order of abundance inside flowers and Fig. S4a for likelihood of response to clogging; see Supplementary material for further details). Activity by individual ant taxa at extrafloral nectaries had very small effects on pollinator visitation, as shown by the small estimates (model S.vii, Table S4), although their effects were precisely estimated by our models, as indicated by narrow variation around these estimates (Fig. S4b). In contrast, the effects of activity by individual ant taxa inside flowers on pollinator visitation of individual ant species inside flowers could not be precisely estimated from our data, as indicated by the large variation associated with these estimates (model S.viii, Fig. S4c).

162

Chapter 5: The Distraction Hypothesis of EFN

Table 3. Comparison of the effect sizes of the short and long-term clogging experiments on the number of visits per visitor type.

Short term experiment: Long term experiment: Effect size Clog 1 day 1 leaf Clog 10 days 10 leaves Model difference Effect Effect Fixed effect d Fixed effect d ratio size size i) Ants in EFN -0.2865 Small Ants in EFN -0.2489 Small -13% i) Ants in flowers +0.070 Small Ants in flowers -0.0387 ns -155%

Flowers occupied by Flowers occupied ii) +0.146 Small +0.011 ns -132% ants by ants

iii) Pollinators +0.0503 ns Pollinators +0.0698 ns +38%

iv) Clogging -0.0488 ns Clogging +0.1053 ns +315% iv) Total ants -0.0915 ns Total ants +0.2065 Small +325%

v) Clogging -0.0676 ns Clogging +0.1245 ns +284% v) Ants in EFN -0.1314 ns Ants in EFN +0.3054 Medium +332% v) Ants in flowers -0.0893 ns Ants in flowers -0.1555 ns -74%

163

Chapter 5: The Distraction Hypothesis of EFN DISCUSSION The Distraction Hypothesis Plants face a potential trade-off between the benefits they receive from ants patrolling their leaves and flowers and the costs associated with this (Altshuler 1999, Assunção et al. 2014, Dutton et al. 2016). In Turnera velutina the presence of the most aggressive ants inside flowers increases the likelihood of pollinators displaying alert behaviours and reduces the time honeybees spend inside the flowers (Villamil et al. 2018). To reduce the costs without waiving the protective benefits, several authors have hypothesised that plants should evolve mechanisms that minimise ant access to floral structures and pollinators, whilst recruiting them to the vicinity in order to reduce herbivore damage (Willmer and Stone 1997, Martínez-Bauer et al. 2015). Two current theories – the Distraction Hypothesis and Optimal Defence Theory (ODT) - are compatible with the commonly observed location of extrafloral nectaries close to valuable and vulnerable reproductive structures. The Distraction Hypothesis specifically predicts that EFN secretion draws ant guards away from flowers in such a way that ant-pollinator conflict is reduced (Kerner 1878). The Distraction Hypothesis has been widely overlooked, with only three studies addressing it since its proposal in 1878. We briefly outline these studies below, highlighting aspects of their experimental design that contrast with our approach, and we summarise the extent to which our results match predictions of the Distraction Hypothesis and ODT.

Wagner and Kay (2002) tested the Distraction Hypothesis using sticks as artificial plants, and identical plastic caps as artificial floral (primary) or extrafloral (additional) nectaries. Sticks with additional nectar sources did not attract more ants, but reduced the number of ants at primary sources. They concluded that additional (extrafloral) nectar sources did not increase ant recruitment, but distracted ants from the primary, floral nectar sources (Wagner and Kay 2002). These results differ from studies conducted on natural

164 Chapter 5: The Distraction Hypothesis of EFN plants (Bentley 1976, Shenoy et al. 2012, Villamil et al. 2013) and from our findings (Fig. 3), which show that increased EFN results in increased ant visitation. Furthermore, the plastic caps used by Wagner and Kay (2002) to simulate floral (primary) and extrafloral (additional) nectaries were morphologically identical and equally accessible, but neither assumption is met in natural EFN-bearing species (Escalante-Pérez and Heil 2012, Keeler 2014). Therefore, no robust conclusions about the Distraction Hypothesis can be drawn from this experimental design.

In 2005, Galen tested the Distraction Hypothesis on Polemonium viscosum, a plant species without extrafloral nectaries. Extrafloral nectaries were simulated by trimming the petals, anthers and pistils from some flowers, leaving only the calyx and toral disk that bears the floral nectaries to simulate extrafloral nectaries (Galen 2005). Control inflorescences contained only intact flowers, whilst inflorescences with simulated extrafloral nectaries contained intact flowers plus trimmed flowers simulating extrafloral nectaries (Galen 2005). Intact flowers in the EFN-simulation inflorescences had higher ant visitation than flowers from control inflorescences, and Galen saw this result as rejecting the Distraction Hypothesis. However, rather than testing the Distraction Hypothesis, we suggest that this experiment tested the effect of total floral nectar availability on ant recruitment, and the effect of removing floral parts on ant visitation to flowers. By trimming the corolla and sexual organs, Galen facilitated ant access to the flower. Previous studies on P. viscosum demonstrated that corolla morphology effectively excludes ants from flowers (Galen 1999, Galen and Cuba 2001). Furthermore, artificial damage (trimming) is a confounding factor because it triggers plant induced defences (Heil et al. 2001b, Ness 2003, Heil 2008, Ballaré 2011) that strongly affect floral and extrafloral nectar secretion (Ness 2003, Radhika et al. 2008, Heil 2011, 2015). Consequently, higher ant visitation to intact flowers in the EFN-simulation

165 Chapter 5: The Distraction Hypothesis of EFN inflorescences may have been a response to the trimming of neighbouring flowers.

Finally, Chamberlain and Holland (2008) tested The Distraction Hypothesis on Pachycereus schottii, a senita cactus bearing extrafloral nectaries. They found higher rates of ant visitation on flowers from plants where EFN had been experimentally removed, as the Distraction Hypothesis would predict. However, in contrast to our clogging treatment in T. velutina, Chamberlain and Holland’s EFN-elimination treatment consisted of removing EFN secreting structures (buds, flowers and fruits). Hence, as in Galen’s (2005) study, the increase in flower-ant interactions observed on EFN-removal plants could be an ant response to the artificial damage inflicted by removing reproductive structures.

We tested the Distraction Hypothesis in a field population of T. velutina, a species bearing extrafloral nectaries, by experimentally manipulating EFN availability without inducing artificial damage to plant structures. If the Distraction Hypothesis is true and EFN distracts ants from entering the flowers, we predicted that elimination of EFN by clogging extrafloral nectaries should result in: 1) decreased numbers of ants patrolling extrafloral nectaries; 2) increased numbers of ants inside flowers; 3) an increase in the proportion of flowers occupied by ants, leading to 4) a reduction in the numbers of pollinators visiting the flowers, and 5) a reduction in plant fitness. If EFN secretion has evolved to reduce herbivore damage to flowers by increasing ant activity in their proximity, as predicted by ODT, then we expect elimination of EFN to result in patterns compatible with predictions 1 and 5 above, with the difference that reduced plant fitness should be caused by increased floral herbivory rather than ant-associated reduction of visitation. However, ODT does not make predictions 2, 3 and 4.

166 Chapter 5: The Distraction Hypothesis of EFN Our results support the Distraction Hypothesis with predictions one, three, and five being met. We found that clogging EFN secretion reduced the number of ants patrolling extrafloral nectaries by 30 % (prediction 1), increased the likelihood of flower occupation by ants by 3.6 % (prediction 3), and increased the likelihood of fruit abortion by 12 % (prediction 5). However, we found no significant increase in the number of ants inside flowers (prediction 2), or reduction in pollinator visitation (prediction 4) when extrafloral nectaries were clogged (Tables 1-2). Support for prediction 3 (increased flower occupation by ants), and reduction in plant fitness through increased rates of fruit abortion (rather than damage to flowers; Fig. 4) are both specific to the Distraction Hypothesis. We therefore conclude that our results represent the first experimental support for this hypothesis obtained under field conditions.

Exploring ant species-specific effects We found ten ant species interacting with T. velutina, representing a diverse mosaic of partners that may differ in their response to clogging and on their effects on pollinators. Evolution of plant mechanisms that reduce plant pollinator conflict could be driven by interactions with one or more of these species. We expect plants to evolve phenotypes that favour ant taxa that are both effective guards and that have minimal net negative impacts, including interference with pollinators. Despite the potential for variation in effects across ant taxa, we found that ant species and the interaction between clogging and ant species had a negligible effect on the number of ants found inside flowers (Table S4) and on pollinator visitation (Table S5).

The estimates of the effects that individual ant species inside flowers have on pollinator visitation in model (S.vii) are imprecise (Fig. S4c) for two main reasons: first, ants rarely occupy flowers (Fig. 2, Table 2), and second, the scarce variation in ant species composition between flowers, caused by

167 Chapter 5: The Distraction Hypothesis of EFN Brachymyrmex sp. being the dominant ant taxa inside flowers, resulting in few observations for other taxa (Table 1). Because of these two reasons, our data has little variation and model estimates are thus driven largely by uncertainty and further experiments are required to test the effect different ant species inside flowers have on pollinator visitation. (Villamil et al. 2018). Finally, although models S.vi – S.viii do not elucidate the effects of specific ant species, they demonstrate quantitatively that lumping ant species together and testing the Distraction Hypothesis on the ant community associated with T. velutina is an adequate approach given the constraints of our dataset imposed by the biology of this system (see Supplementary material for further details).

Little is known about the extent to which positive and negative impacts of ant taxa are correlated and whether ant species that are threatening for herbivores (hence, highly defensive species) are also threatening for pollinators (hence, ecologically costly via pollinator deterrence) (but see: Ness 2006, Miller 2007, LeVan et al. 2014, Villamil et. al. 2018 in press, Ohm and Miller 2014, Villamil et al. 2018). Large-bodied and eusocial pollinators such as Apis mellifera have been assumed to be less susceptible than smaller solitary bees or other non-eusocial pollinators to ant attacks and more prone to visit flowers patrolled by aggressive ants (Queller 1989, Gadagkar 1990, Brechbühl et al. 2010a, Brechbühl et al. 2010b, Romero et al. 2011).

Our findings suggest that ant species vary in their deterrent effect on Apis mellifera bees, (Fig. S4). Qualitative patterns show that the presence of the most aggressive ant species, Dorymyrmex bicolor, inside flowers and at extrafloral nectaries have, on average, a negative effect on pollinators. These results are consistent with experimental findings demonstrating that placing dead Dorymyrmex bicolor ants inside flowers of T. velutina induced alert behaviours in Apis mellifera, reduced visit duration, and increased handling time per flower leading to a decreased in pollinator foraging efficiency (Villamil et al. 2018).

168 Chapter 5: The Distraction Hypothesis of EFN However, A. mellifera honeybees are introduced pollinators, and further work is required to assess the effect of the ant community on pollinator assemblages dominated by native, smaller-bodied, solitary pollinators.

The spatio-temporal scale of the Distraction Hypothesis Based on the relatively small effect sizes of the short-term leaf-scale experiment (Tables 1-2), we hypothesised that clogging the glands of only one leaf for one day was perhaps too local and short-term a treatment to detect a measurable effect. In the long-term experiment we therefore increased both the spatial scale and duration of the EFN-removal treatment by a factor of ten, expecting to obtain larger effect sizes overall. However, in contrast to our prediction, the long-term clogging experiments had smaller effect sizes on ant patrolling (Table S2). For example, the short-term clogging experiment had a in 13 % greater effect size in reducing numbers of ants patrolling extrafloral nectaries, 155 % greater effect size increasing the numbers of ants inside flowers, and 132 % greater increase of ant occupancy of flowers than the long-term experiment (Fig. 3b; Tables 1-2). Hence, we can robustly conclude that clogging the glands of only one leaf for one day is not too local and short-term a treatment. In fact, leaf-day is the scale at which we detected an effect of clogging and our experimental evidence showed that ant foraging behaviour responds to reward availability over a this spatio-temporal scale. The non-provision of a whole branch for ten days is a rather unnatural setting for ants, or may resemble a low-rewarding plant (Lemus Domínguez 2014).

Our results suggest that in T. velutina EFN-mediated ant distraction is a mutualist management strategy that acts at a local and short-term scale. This makes adaptive sense because plant structures vary in their vulnerability to herbivores and sensitivity to both benefits and costs of ant guards over similar local and short-term scales (Bentley 1977b, Willmer and Stone 1997, Falcão et al.

169 Chapter 5: The Distraction Hypothesis of EFN 2014, Villamil 2017). From the plant’s perspective, protection needs changes at a very small spatial and temporal scale (Bentley 1977b, Willmer and Stone 1997, Falcão et al. 2014, Villamil 2017) because in T. velutina, buds, flowers and fruits indeed occur in close proximity on the same shoot, and develop from bud to young fruit in only three days. Flowers are suggested to be the most vulnerable structure due to their soft and exposed water-rich tissues, whilst buds and fruits are protected by the sepals or the exocarp, respectively (Villamil 2017). Previous work has shown that EFN secretion in T. velutina is greatest at the flower stage, with glands in the associated leaf secreting ten times more sugar than glands associated with fruit, and 40% more sugar than glands associated with buds (Villamil 2017). This pattern of investment is compatible with both ODT (McKey 1979, Rhoades 1979, Stamp 2003, Ochoa-López et al. 2015, Ochoa-López et al. 2018) and the Distraction Hypothesis (for reduction of negative ant-pollinator interactions).

From the ant’s perspective, adjustment of foraging patterns at a local scale could maximise net sugar gain (Schilman and Roces 2006). The rapid transition from bud to fruit in T. velutina means that secretion by individual glands can vary substantially over consecutive days since EFN secretion varies greatly throughout this transition (Villamil 2017). Consequently, for the ants, missing the extrafloral nectaries of leaves associated with flowers means missing a bountiful reward.

Implications for ant and pollinator foraging strategies We suggest that ants associated with T. velutina learn the location of highly rewarding EFN glands by monitoring variation in rewards within a single day, rather than relying on cues from previous days – a pattern compatible with demonstrated ability of ants to learn spatial and temporal scales of food rewards (Jackson and Morgan 1993, Robinson et al. 2005, Jackson and Ratnieks 2006).

170 Chapter 5: The Distraction Hypothesis of EFN There is also evidence that at least some pollinating insects can respond to similarly local variation in ant activity. Bees in other systems are known to use ant scents to discriminate and avoid heavily patrolled flowers, preventing harassment (Cembrowski et al. 2014) and we suggest that bees visiting T. velutina may also use olfactory cues to reduce their visitation of ant-occupied flowers.

The local foraging decisions we propose and the effects of within-plant variation in EFN availability on ants and pollinators should be seen as occurring against a backdrop of significant between-plant variation in EFN rewards. Differences between individual plants, and not between branches or flowers, explained a large part of the variance in both numbers of ants and pollinators at both experimental scales (Tables 1, S1). It is possible that plant-level variation in nectar availability underlies the positive correlation between numbers of patrolling ants and pollinator visitation observed in the long-term clogging treatment (Table S2), with each mutualist guild independently selecting more rewarding plants. Plant level variation in EFN rewards could have many causes, including phenotypic plasticity (Ochoa-López et al. 2018), genetic variation in floral (Ramos-Castro 2013) and extrafloral nectar (Ochoa-López 2013, Ochoa- López et al. 2015, Ochoa-López et al. 2018), and other variables such as plant age (Villamil et al. 2013, Ochoa-López et al. 2018), size, floral display (Ramos-Castro 2013), proximity to a nest or hive (Cuautle and Rico-Gray 2003, Cuautle et al. 2005), plant vigour or soil fertility (Yamawo et al. 2012, Dattilo et al. 2013).

Conclusions Our findings on flower occupancy by ants at a leaf-day scale support the Distraction Hypothesis suggesting extrafloral nectar secretion during anthesiscan bribe ants away from flowers and reduce ant occupancy of flowers. However, clogging EFN secretion did not result in a significant increase in ant abundance within flowers. This suggests that distraction via EFN secretion is

171 Chapter 5: The Distraction Hypothesis of EFN neither the only nor the strongest mechanism in mutualist management by T. velutina. Further research is required to understand why ants rarely visit the flowers of T. velutina, and which mechanisms may be keeping ants outside these accessible, nectar producing flowers. Perhaps other mechanisms such as floral ant repellents (Willmer et al. 2009, Ballantyne and Willmer 2012) reinforce ant exclusion. Differences in chemical composition or sugar concentration between floral and extrafloral nectars may also underlie observed ant foraging preferences. Further studies on a range of ant-plants are required to assess the wider significance of EFN-mediated ant distraction in amelioration of ant- pollinator conflict.

I am the sole author of all the text used here, with comments on earlier drafts from Graham Stone, Karina Boege, Sarah Heath and Richard Whittet. Cristina Rosique kindly drew the illustration for Figure 5.1. Jarrod Hadfield contributed valuable statistical advice.

This chapter has been accepted at Journal of Ecology.

172 Chapter 5: The Distraction Hypothesis of EFN SUPPLEMENTARY MATERIAL 1

Long-term clogging experiment (10 leaves, 10 days)

FIELDWORK METHODS

Based on the results from the short-term experiment, we conducted a follow-up experiment in which treatment duration and spatial scale were both increased by a factor of 10, using paired branches and focusing on one flower on each control or clogged branch, rather than paired flowers on the same branch (Fig.

1). In each of three sites, we selected 15 T. velutina plants bearing at least four apical flowering branches, each with ten leaves and reproductive structures (n

= 45 plants). Two branches per plant were marked as control and two as clogged treatments using different string colours. We chose two branches to increase the probability of having one pair of focal flowers with each of the experimental treatments per plant per day. This experiment was conducted over 5 days during

September 2015 (08 - 12 Sept.) for a total of 80 pairs of control and blocked flowers (n = 160 flowers). The extrafloral nectaries of all ten leaves on the clogged treatment branches were clogged as described in the methods for the short-term experiment, and the treatment was maintained for ten days. Glands

173 Chapter 5: The Distraction Hypothesis of EFN were checked each day, and the paint was replenished if required, especially after heavy rains overnight. If new leaves emerged on a branch during the experiment, paint droplets were applied on or above their extrafloral nectaries according to their respective treatment (Fig. 1). Every morning all plants were checked and those bearing at least one pair of flowers with control or clogged treatment were observed following the same observational protocol as for the short-term experiment. Simultaneous observations were performed at each of three sites by different observers.

STATISTICAL ANALYSES

Five mixed effects models with the same structure as those used to analyse the results from the short-term clogging experiment were fitted to analyse the long- term experiment data. Because all of these models had the same random effects structure, unless otherwise specified, we detail the random effects first and then describe the fixed effects for each model. In the long-term experiment, branch identity was fitted as a random effect to account for repeated hourly observations (Table S1). Because this experiment had a paired experimental design we fitted branch pair identity as a random effect to control for between- pair variation in floral and extrafloral investment. We also included an observation-level random effect where each data point received a unique level

174 Chapter 5: The Distraction Hypothesis of EFN of a random effect to control for overdispersion. We fitted the following models, and have structured our results following this order:

(S.i) To test the effect of nectary clogging on the number of ants we fitted a

Poisson mixed effects model using number of ants as the response variable. Ant location (at extrafloral nectaries or in flowers), treatment, and the interaction between these two factors were fitted as fixed effects. Tukey post-hoc tests were conducted in this model to find differences between the number of ants at extrafloral nectaries or flowers under control or clogged gland conditions.

(S.ii) To test whether preventing EFN secretion by clogging the glands increased the probability of a flower being occupied by ants, we fitted a binomial mixed effect model. We used the presence or absence of ants in a flower as a response variable, and fitted treatment as a fixed effect. The random effects remained the same, except for the observation-level random effect which we omitted.

(S.iii) To test the effect of clogging EFN secretion on pollinator visitation we fitted a Poisson mixed model using the number of pollinators as the response variable and treatment as the only fixed effect.

(S.iv) To test the whether the total number ants, regardless of their location, had an effect on pollinator visitation we fitted a Poisson mixed model using number of pollinators as the response variable. As fixed effects we fitted the

175 Chapter 5: The Distraction Hypothesis of EFN total number of ants, and treatment to test whether treatment affected pollinator visitation in a way that was unlinked to the number of ants.

(S.v) To test if the location (inside flowers or at extrafloral nectaries) and number of ants affect pollinator visitation, we fitted a Poisson mixed model. The number of pollinators was fitted as the response variable, whilst treatment, number of ants in flowers, and number of ants at extrafloral nectaries were fitted as fixed effects.

Cohen d effect sizes for all models were calculated using the LRT statistics from each model. To test whether increasing the duration and scale of the clogging treatment by a factor of 10 had a larger effect on the number of ants and pollinators, we estimated the ratio of change in the effect size between the short- and long-term experiment for each type of visitor (Table S2).

RESULTS AND DISCUSSION Reduction of EFN secretion (S.i) EFN treatment (clogged versus control) and ant location both had significant effects on the number of ants, but with no significant interaction between these factors (Fig. S1b; Table S1). There were significant differences between the numbers of ants patrolling the nectaries and those inside flowers over both EFN treatments (Table S1). Almost seven times more ants were found patrolling extrafloral nectaries (1.53 ± 0.087) than were found inside the flowers

176 Chapter 5: The Distraction Hypothesis of EFN (0.22 ± 0.034) (Fig. S1b; Table S1), regardless of EFN treatment (control: Z = -

14.04, P < 0.001; clogged: Z = -11.41, P < 0.001). In contrast to the results from the short-term experiment, eliminating EFN secretion had a negative effect on the number of ants at both locations (Fig. S1b; Table S3). Clogging significantly reduced the number of ants at extrafloral nectaries (Z = -3.043, P = 0.011; Table

S2), but had no significant effect on the number of ants in flowers (Z = -0.477, P

= 0.96; Table S2) (Fig. S1b).

(S.ii) The percentage of flowers occupied by ants was 11% in both treatments, hence eliminating EFN secretion from all leaves did not increase ant visitation to flowers (Table S1).

(S.iii) Eliminating EFN secretion had no significant effect on the number of pollinators (Fig. S1b; Table S1).

(S.iv) The number of pollinators was significantly and positively correlated with the total number of ants found patrolling either the extrafloral nectaries or inside the flowers, regardless of the treatment (Tables S1, S2).

(S.v) Furthermore, when ant abundance was partitioned by their location on the plants between extrafloral nectaries and flowers, the number of ants patrolling extrafloral nectaries was significantly positively correlated with the number of pollinators (Table S1), whilst the number of ants inside a flower was significantly

177 Chapter 5: The Distraction Hypothesis of EFN negatively correlated with the number of pollinators (Table S2). These results contrast strongly with patterns at the single flower scale.

In models (S.i)-(S.iii) used to analyse the long-term experiment data, differences between plants (captured by the plant random effect) explained more of the variation in the numbers of ants or pollinators than differences between branches (captured by the branch random effect) or random variation between observations (captured by the OLRE random effect) (Table S1). In models (S.iv, S.v) random variation between observations explained a larger proportion of the variation in the numbers of ants or pollinators, followed by plant identity (Table S1). Branch identity was the random effect that consistently explained least variation in the numbers of ants or pollinators (Table S1).

Effect sizes Increasing the duration and spatial scale of the clogging treatment by a factor of 10 did not result in larger overall effect sizes as we had predicted (Table S2).

In fact, the direction of some effects was reversed when the scale was increased.

For example, the impact of preventing EFN secretion on the number of ants inside flowers changed from positive at a short-term, local scale to negative in the long-term, branch-scale experiment (Table S2). The effect size of total ant numbers on pollinator activity changed from negative to positive when we increased the scale of the treatment, as expected. The same pattern was observed for the effect size of ants patrolling extrafloral nectaries on pollinator

178 Chapter 5: The Distraction Hypothesis of EFN activity. However, the effect size of ants inside flowers on pollinator activity was reduced when we increased the scale of the treatment.

Comparing effect sizes and spatio-temporal scales Here we provide further discussion regarding other comparisons between the short and long-term clogging experiments that are beyond the scope of

Distraction Hypothesis.

The effect of the total number of ants and the number of ants at EFN on pollinators was three times greater in the long-term scale experiment, as expected. Clogging extrafloral nectaries from a whole branch caused a 13 % greater decrease in the number of ants patrolling the EFN glands than clogging

EFN glands from only one leaf. Since most ants are found patrolling EFN glands, this decreased the total ant numbers, and such reduction in ant patrolling may be driving the increase in pollinator visitation. However, the negative effect of ants inside flowers on pollinator visitation was reduced in the long-term experiment, supporting the idea that the Distraction Hypothesis operates at a leaf, daily scale.

The long-term experiment showed that numbers of ants inside flowers have a significant negative effect on the number of pollinators, as predicted by the Distraction hypothesis (Tables S1, S2). But the same experiment revealed two positive effects that are not predicted by the Distraction Hypothesis:

179 Chapter 5: The Distraction Hypothesis of EFN between the total number of ants and pollinators, and between the number of ants patrolling extrafloral nectaries and pollinators (Tables S1-S2).

The positive effect of the number of ants on pollinator visitation may well represent the outcome of reward-based foraging decisions at the plant scale by both mutualists, leading to selection of more rewarding plants. Having all extrafloral nectaries within a branch clogged may resemble a low-rewarding plant, as explained in the main text. The positive relationship between ants and pollinators at a plant-scale (Table S2), may imply that plants of T. velutina able to recruit large numbers of ants do not experience a trade-off in pollinator visitation. Further experiments are required to test whether the ecological costs of ant patrolling on pollinator visitation differ between plants with high or low reward secretion.

180 Chapter 5: The Distraction Hypothesis of EFN

Figure S1. Effect of clogging the extrafloral nectaries on the number of visitors recorded in hourly surveys during 2 min of observation per flower (mean ± se) a) for the short-term experiment where one leaf was clogged for one day; and b) for the long- term experiment where ten leaves within a branch were clogged for ten days. Circles represent ants at extrafloral nectaries; triangles represent ant in flowers, and squares represent pollinators. Plot a) is also reported in the main text but has been repeated here to facilitate comparison with b).

181 Chapter 5: The Distraction Hypothesis of EFN

Figure S2. Hourly observation showing the effect of clogging of extrafloral nectaries on the number of visitors (mean ±se) a) for the short-term experiment in which one leaf was clogged for one day, and b) for the long-term experiment in which ten leaves within a branch were clogged for ten days.

182 Chapter 5: The Distraction Hypothesis of EFN

Table S1. Likelihood ratio test results for statistical models used to test different response variables in the experiments. The values highlighted in bold are statistically significant (P > 0.05).

P- Random Variance SD Experiment Model Response Fixed effects df LRT value effects Clogging 1 7.28 0.0069 ** Ant location Branch 0.20 0.45 S.i) 1 395.5 2.2-16 *** Number of ants (EFN or flowers) Plant 0.30 0.55 Clogging × Ant OLRE 0.60 0.77 1 1.79 0.18 location Proportion of Branch 9.87-10 3.14-10 flowers occupied -5 S.ii) Clogging 1 16.49 0.98 by ants Plant 1.07 1.03

1.04-9 3.23-05 Long-term Number of Branch S.iii) Clogging 1 0.63 0.42 0.39 0.62 Clogging pollinators Plant 0.54 0.73 experiment OLRE Clogging 1 1.43 0.23 Branch 0 0 Number of Total number of Plant 0.39 0.63 pollinators 1 5.17 0.02 * S.iv) ants OLRE 0.48 0.69

Clogging 1 2.00 0.15 Branch 0 0 Number of S.v) Ants in EFN 1 10.85 0.0009 *** Plant 0.38 0.62 pollinators Ants in flowers 1 3.78 0.05 * OLRE 0.41 0.64

183 Chapter 5: The Distraction Hypothesis of EFN

Table S2. Comparison of the effects sizes of the short and long-term clogging experiments on the number of visits per visitor type.

Short term experiment: Long term experiment: Effect size Clog 1 day 1 leaf Clog 10 days 10 leaves Model difference Effect Effect Fixed effect d Fixed effect d ratio size size i) Ants in EFN -0.2865 Small Ants in EFN -0.2489 Small -13% i) Ants in flowers +0.070 Small Ants in flowers -0.0387 ns -155%

Flowers occupied by Flowers occupied ii) +0.146 Small +0.011 ns -132% ants by ants

iii) Pollinators +0.0503 ns Pollinators +0.0698 ns +38%

iv) Clogging -0.0488 ns Clogging +0.1053 ns +315% iv) Total ants -0.0915 ns Total ants +0.2065 Small +325%

v) Clogging -0.0676 ns Clogging +0.1245 ns +284% v) Ants in EFN -0.1314 ns Ants in EFN +0.3054 Medium +332% v) Ants in flowers -0.0893 ns Ants in flowers -0.1555 ns -74%

184 Chapter 5: The Distraction Hypothesis of EFN

SUPPLEMENTARY MATERIAL 2

Ant species-specific responses to clogging and effects on pollinators

STATISTICAL ANALYSES

Effects of clogging on ant species (S.vi) To investigate whether ant species differed in their response to clogging we fitted a Poisson mixed model using the number of ants inside the flower as the response variable, and treatment as a fixed effect. As random effects we fitted flower identity, pair identity, an observation-level random effect to deal with overdispersion, ant species, and the interaction between treatment and ant species. This model allowed us to estimate the amount of variance in the number of ants inside flowers explained by ant species, and estimate the effect of clogging on the abundance of each ant species inside flowers.

Although the ant species and the interaction between clogging and ant species explained a negligible amount of variation in the abundance of ants inside flowers, we extracted the random effect estimates for our “species” and “species x clogging” random effects. For the random effect “ant species” in model (S.vi) we extracted the estimates for every level within this random effect and its standard error (Fig. S3a).

185 Chapter 5: The Distraction Hypothesis of EFN Effects of ant species on pollinators To investigate whether different ant species patrolling the plants or inside the flowers differed in their effect on pollinator visitation, we fitted multiple membership mixed models using MCMCglmm (Hadfield 2010). In these models, we fitted the abundance of ants from each species as covariates and treated the regression coefficients as random. This allowed us to estimate the effects that different species of ant have on the number of pollinators.

These statistical tools allow us to account for plants occupied by multiple ant species, and capture the variation in ant abundance within a given species. This model also prevents us from making a priori assumptions of ant attributes that may make certain species more prone to responding to clogging or ant attributes that may have an effect on pollinator visitation. Such a priori assumptions may be flawed due to the lack of previous studies testing this. Instead, these models allow us to detect the magnitude and direction (positive or negative) of the effect each ant species has on pollinator visitation. Using this approach allows us to generate robust evidence-based assumptions of ant species attributes that may enhance or hinder pollinator visitation.

(S. vii) To investigate if the number of ants of different species inside the flowers affected pollinator visitation, we fitted an ordinal multi-membership mixed model. Although the data are count data we chose to treat the number of pollinator visits as ordinal because the number of visits fell into a few categories (0, 1, 2, or 3) and the mean and variance of the counts were equal suggesting that the data would be underdispersed with respect to the Poisson if there was real heterogeneity in pollinator visitation rates. The number of pollinators was fitted as the response variable, whilst treatment, number of ants in flowers, and number of ants at extrafloral nectaries were fitted as fixed effects. As random effects we fitted flower, pair identity, and an observation-level random effect to account for overdispersion. The regression coefficients of the number of ants inside flowers from each species were treated as random additive effects. Fitting

186 Chapter 5: The Distraction Hypothesis of EFN these additive random effects allowed us to estimate the effects that different ant species inside flowers had on pollinator visitation.

(S.viii) To investigate whether the number of ants of different species patrolling the extrafloral nectaries differed in their effect on pollinator visitation, we fitted an ordinal multi-membership mixed model. The number of pollinators was fitted as the response variable, whilst treatment, number of ants in flowers, and number of ants at extrafloral nectaries were fitted as fixed effects. As random effects we fitted flower and pair identity, and an observation-level random effect to account for overdispersion. We also fitted the abundance of ants from each species at the extrafloral nectaries as random additive effects which allowed us to estimate the amount of variance in the number of pollinators explained by ant species, and estimate the effects of patrolling by each ant species on pollinator visitation.

RESULTS AND DISCUSSION

Effects of clogging on ant species (S.vi) We found no significant effect of clogging on the number of ants inside flowers (Table S3), which confirmed the results from model (i) in the short-term experiment. The interaction between clogging and ant species explained no variation in the number of ants inside flowers. Most of the variation was explained by the differences between observations (OLRE: 97 %), followed by pair identity (2.3 %), and finally ant species explaining only 0.21 % of the variation. The estimates of the between-species effects are low and imprecise (Fig. S3a). The imprecision is for two reasons; more common species are sampled more often, whilst rare species are seldom sampled, hence their confidence intervals are larger and estimates less precise. Secondly, because

187 Chapter 5: The Distraction Hypothesis of EFN overall ants inside the flowers is a rare phenomenon (Fig. S1). The qualitative patterns obtained from the “ants species x clogging” effects helped us detect ant species with the greatest differences between control and clogged (Fig. S3a) which would be worth exploring in a controlled experimental way. These species are: Brachymyrmex, Paratrechina longicornis, C. planatus, and C. mucronatus.

Effects of ant species on pollinators (S.vii) The effects of different ant species at extrafloral nectaries on pollinator visitation are precisely estimated by our models, (see SE in Fig. S3b). All effects overlap with zero, revealing a negligible effect of ant species at extrafloral nectaries on pollinator visitation (Table S4). Despite this, we could detect qualitative patterns in this data: Dorymyrmex bicolor ants have a on average, negative effect on pollinator visitation, whilst Camponotus mucronatus and Paratrechina longicornis have, on average, a positive effect on pollinator visitation (Fig. S3b).

(S.viii) The effects of different ant species inside flowers have on pollinator visitation are imprecisely estimated by our models, as indicated by the large SE values associated with these estimates (Fig. S3c). Qualitative patterns in this data suggest that when inside flowers, Dorymyrmex bicolor ants may have a negative effect on pollinators. Such patterns also help us detect ant species whose positive or negative effects on pollinators would be worth exploring in further experimental work. Furthermore, these results are consistent with experimental findings demonstrating that Dorymyrmex bicolor ants placed inside flowers of T. velutina induce alert behaviours in pollinators, reduce visit duration, increasing handling time per flower and reducing foraging efficiency (Villamil et al. 2018).

188 Chapter 5: The Distraction Hypothesis of EFN a)

b)

c)

Figure S3. a) Effect of clogging the extrafloral nectaries on different ant species inside flowers and effects of different ant species on pollinators, and their precision. Effect of ant species (b) patrolling extrafloral nectaries, or (c) inside the flowers on pollinator visitation (mode ± SE).

189 Chapter 5: The Distraction Hypothesis of EFN

Table S3. Mean effect of each ant species patrolling at EFN glands on pollinator visitation using a multiple membership mixed model.

Model Response Fixed effects Estimate LRT P-value Random effects Variance SD

Species 0.20 0.45 Species x -10 6.14 2.47-05 Ants inside Clogging S.vi) Clogging 0.55 0.61 0.43 0 0 flowers Flower id 2.20 1.48 Pair id 90.41 9.50 OLRE

190 Chapter 5: The Distraction Hypothesis of EFN

Table S4. Mean effect of each ant species patrolling at EFN glands on pollinator visitation using a multiple membership mixed model.

Post 95% CI 95% CI Random 95% CI 95% CI Model Response Fixed effect Variance mean lwr upr effects lwr upr Clogging -0.1803 -0.4725 0.0959 Flower id 0.02372 3.83-09 0.08859 Number of Ants at EFN -0.1208 -0.2717 0.0261 Pair id 1.234 0.6468 1.935 s.vii Ants in pollinators -0.1377 -0.5213 0.2409 Ants at EFN 0.02008 2.21-11 0.07075 flowers

Clogging -0.1707 -0.4460 0.1155 Flower id 0.02506 5.3712 0.0969 Number of s.viii Ants at EFN -0.1479 -0.2601 -0.0402 Pair id 1.326 0.7137 2.104 pollinators Ants in Ants in -0.7854 -2.2555 0.3802 2.364 7.284-07 9.139 flowers flowers

191

CHAPTER 6

GENERAL DISCUSSION

Key messages and loose ends

192

Chapter 6: General discussion SUMMARY OF KEY FINDINGS

Chapter 2. Ants aggressivity • I quantified the aggressivity of six ant species interacting with T. velutina, described their behavioural ecology and used qualitative and quantitative components to assess their protection effectiveness. • Cephalotes sp. ants were detected as parasitic non-defenders. • The rest of the species were ranked according to their protection effectiveness as: Capmponotus planatus < Crematogaster sp. < Paratrechina longicornis < Brachymyrmex sp. < Dorymyrmex bicolor.

Chapter 3. Direct and indirect ant-pollinator conflict • I explored ant-pollinator conflict in Turnera velutina, which rewards ants with extrafloral nectar (EFN) and pollinators with floral nectar (FN). • Secretion of FN and EFN, activity of ants and pollinators, and pollen deposition all overlapped in time. • I found evidence of direct conflict, since ants inside flowers altered pollinator behaviour and reduced visit duration, and more aggressive ants had stronger effects on pollinator visitation. • I found no evidence for indirect conflict with no significant difference in the volume or sugar content of FN between control plants and those in which EFN secretion was prevented.

Chapter 4. Ecological costs • I experimentally tested the effects of ant patrolling on pollinator community composition, visitation frequency, duration and, behaviour of floral visitors, and their cascading effects on the plant mating system, pollen transfer dynamics, and plant fitness. • Patrolling did not affect pollinator composition, visitation frequency, or inspection/alert behaviours in pollinators. • Ant patrolling had a significantly decreased visit duration, pollen load on stigmas and male fitness; and significantly increased outcrossing rates.

193 Chapter 6: General discussion • The effects of ant patrolling on pollinator visitation lead to negative effects on pollinator foraging efficiency, but such behavioural changes had positive effects for plant progeny promoting outcrossing.

Chapter 5. EFN as ant distractors • I experimentally tested the Distraction Hypothesis for EFN, which posits that rewarding ants with extrafloral nectar could reduce their visitation of flowers, reducing ant-pollinator conflict while retaining protection of other structures. • Ant activity inside flowers was low throughout all anthesis period. • Preventing EFN secretion reduced ant patrolling at extrafloral nectaries, increased the proportion of flowers occupied by ants, and reduced plant reproductive output through an increase in fruit abortion, but did not affect the numbers of ants or pollinators inside flowers. • These results support the Distraction Hypothesis, suggesting that EFN secretion may bribe ants away from reproductive structures during the crucial pollination period, reducing ant-occupation of flowers.

194 Chapter 6: General discussion SOME LOOSE ENDS

Natural history

The pollination biology of Turnera velutina at Troncones, where visitors are mostly native butterfly species, should be further investigated. So far, the only studies on the pollination biology of this species was conducted at La Mancha (Ramos-Castro 2013, Sosenski et al. 2016), were the introduced honeybee Apis mellifera is the dominant pollinator. Beetles were excluded from the analyses because their role as pollinators is dubious, however, my observations reveal interesting ecological dynamics worthy of further investigation, such as: • The role of beetles as florivores and their impacts on plant fitness • The effect of their presence on pollinator visitation • Their mating behaviour and whether they obtain attractive pheromones from these flowers • Beetle-ant interactions

Is the corolla a natural ant shield for pollinators?

In Heteropterys pteropetala EFN are located at the base of the pedicel that supports several flowers, and corollas are small, less than a fifth of the size of the bees that pollinate it (Assunção et al. 2014). Hence, bees cling to the entire stalk supporting several flowers and their bodies are in close proximity to the EFN, leaving plenty of opportunity for patrolling ants to attack them. In contrast, Turnera flowers are cup-like structures at least three times bigger than honeybees and large enough to fit an entire butterfly inside. Consequently, pollinators can feed on floral resources, and ants on extrafloral resources simultaneously without contacting each other. Pollinators and ants need not be in close proximity whilst foraging, unless ants intrude the flowers. Hence, I

195 Chapter 6: General discussion hypothesise the corolla acts as a natural shield to protect the pollinators, whom are spatially separated from the EFN.

This hypothesis could be experimentally tested by trimming and manipulating corolla shape, without damaging its structural integrity. Although artificial damage is a confounding factor. A potential experiment could have four treatments: (1) Control: with intact flowers; (2) Procedures control: cutting the edge of the petals to induce artificial damage without hugely affecting the corolla’s shape, symmetry, structural integrity or providing an entry for ants (as punch-holes would do); (3) Plucking: plucking the whole petals out from the toral disc, to completely deploy the corolla structure, but aiming to reduce artificial damage and volatile emission; (4) Trimming: trimming the petals just above the sepals to deploy the corolla structure, whilst inducing artificial damage and releasing volatiles. Additionally, it would be worth exploring the pollen volatiles for ant-deterring odours.

Networks

Observational data collected on ant patrolling to different plants could be used to build networks of ant-pollinator species coexisting on a plant. These data would help unravel the effects of specific ants on specific pollinator taxa, building the network of two interactions simultaneously.

Using these same data, we can delve into the community dynamics of ants, studying which ant species can coexist and share a host plant, and which ones do not coexist. This information could shed light on competition dynamics between ant species.

196 Chapter 6: General discussion BROADER IMPLICATIONS Herbivore vs. pollinator susceptibility to ants

Ant aggressivity towards herbivores in T. velutina does not seem to correspond exactly to the threat level they represent for pollinators. The reduction in pollinator visit duration caused by different ant species varying in their aggressivity levels (Chapter 3) was not proportional to the level ant aggressivity against herbivores (Chapter 2). Honeybees spent the same time visiting control flowers without ants as they spent visiting flowers with Brachymyrmex ants. Brachymyrmex sp. ants were the second most aggressive ant species, and more aggressive to herbivores than P. longicornis. Yet, the presence P. longicornis inside flowers had a similar effect size on pollinator visit duration as that of the most aggressive ant species, D. bicolor. This suggests that herbivores and pollinators are not equally vulnerable to all ant species, adding the challenge of considering species-specific ant effects when estimating of the net effect of ant patrolling on plant fitness via herbivore suppression and pollinator deterrence.

In the cholla cacti Opuntia imbricata, Ohm and Miller (2014) found that the weaker defender, imposed more pollination costs than those of an aggressive defender. But these results contrast with previous findings on Ferocactus wizlizeni reporting that better guards also had high anti-pollinator effects (Ness 2006). Our results in T. velutina and those of Ohm and Miller (2014) in cholla cacti contrast with the general hypothesis that more aggressive bodyguards will also have stronger pollinator deterring effects, imposing larger costs for the host plant pollination. This contrasting evidence suggests herbivores and pollinators may not be equally vulnerable to all ant species. Morphological traits such cuticle thickness, and behavioural ones such as foraging locations may affect visitors (herbivores or pollinators) susceptibility to ant attacks. Ant attacking mode may also make certain visitors more susceptible than others. But on the other hand, highlights the need for more

197 Chapter 6: General discussion studies assessing the costs of myrmecophily. Especially those with a community approach since a given ant species may differ in aggressivity display and protection effectiveness depending on the host plant (Fagundes et al. 2017).

Are Brachymyrmex ants the best guards by being pollinator friendly?

I expected the probability of finding different ant species inside the flowers to vary between species, with the least aggressive ant species more often found inside flowers. I predicted that the most aggressive ant species would be patrolling at the extrafloral nectaries, whilst least aggressive ant species would be the ones more often found inside flowers. However, this prediction was not supported by the data. The ant community inside flowers is not a representative sample of the ant community on the EFNs (Chapter 4, 5). Brachymyrmex ants are overrepresented inside the flowers (Chapter 5), despite being the second most aggressive ant species (Chapter 2). Additionally, the effect of Bracymyrmex ants on pollinator visitation were not significantly different from those of control flower without ants (Chapter 3).

This evidence suggests that if T. velutina has ant-repelling mechanisms, these are not deterring one of the most aggressive species from the flowers; alternatively, Brachymyrmex ants may be resistant to the floral repelling mechanisms. This may also imply some degree of specialisation in the T. velutina-Brachymyrmex association. Either ants that are aggressive towards herbivores may not be equally aggressive towards pollinators, resulting in good defensive traits without ecological costs for the host plant pollination. Or, if plants have ant-repelling compounds, Brachymyrmex ants may have evolved resistance against those signals, which allows them to enter flowers. But perhaps both scenarios are true, and restricting Bracymyrmex sp. access to flowers has no fitness benefits because Brachymrmex sp. ants may not represent a cost for pollination. Despite being the second most aggressive ants species (Chapter 2),

198 Chapter 6: General discussion the presence of Bracymyrmex sp. corpses inside flowers did not significantly reduce the time honeybees spent inside flowers, the percentage of visitors displaying inspection behaviours, or the proportion of time spent by pollinators displaying inspecting flowers per visiting bout, compared to control flowers without any ant corpses (Chapter 3). In contrast, the presence of Dorymyrmex bicolor corpses– the other highly aggressive ant species (Chapter 3) –did reduce visit duration and increased inspection behaviours.

It would be interesting to further explore the repellence response induced by flowers on different ant species. For instance, using wiping a petal onto half a Petri dish and placing an ant in the centre and observing the amount of time spent on the flower-wiped vs non-flower-wiped Petri dish section.

Ant patrolling, outcrossing rates, and plant selective fruit abortion

I found evidence for direct ant-pollinator conflict in T. velutina since the presence of dead ants inside flowers induced alert behaviours in pollinators and reduced the duration of visits, under an experimental setting (Chapter 3). The ant exclusion experiment confirmed this in a more realistic setting with live ants instead of ant corpses: ant exclusion resulted in longer pollinator visits and higher selfing rates (Chapter 4). Clogging extrafloral nectaries increased fruit abortion by 12% and decreased ant patrolling (Chapter 5). Hence, it is likely that fruits developed from flowers where extrafloral nectaries were clogged had a higher proportion of selfed seeds than those developed from control flowers.

If this was the case, selective fruit abortion may then explain the higher abortion rates in fruits developed from clogged flowers (predominantly selfing), compared to fruits in the same plant developed from control (predominantly

199 Chapter 6: General discussion outcrossing) flowers. Plants can abort fruits with a higher proportion of selfed seeds, to increase resource allocation in fruits with a higher proportion of outcrossed seeds. Selective fruit abortion linked to pollen origin (selfing vs outcrossing) has been observed in a wide array of plant species (Stephenson 1981, Marshall and Ellstrand 1988, Niesenbaum 1999, Huth and Pellmyr 2000).

Selective abortion in T. velutina may also be linked to pre-dispersal seed predators (Parachnowitsch and Caruso 2008, Sanz-Veiga et al. 2017). There is evidence of pre-dispersal seed predation in T. velutina by the lepidopteran larvae Crocidosema plebejana (Ochoa-López pers. obs.). If the ant community in T. velutina protects plants against pre-dispersal seed predators, then fruits developing in ant excluded plants may have higher infestation levels. Selectively aborting damaged fruits and reallocating those resources into healthy fruits would be a beneficial plant strategy to enhance fitness. The increased rates of fruit abortion in plants with low ant patrolling levels could be due to pre- dispersal seed predators or plant selective abortion linked to pollen origin. Further studies are required to analyse pre-dispersal seed predation and selective fruit abortion in T. velutina in order to demonstrate the suggested mechanisms and its underlying cause.

The effect of ant patrolling on pre-dispersal seed predation could by clarified by characterising the rates of pre-dispersal seed predators in T. velutina, the role of ant patrolling in preventing it, and the rate of fruit abortion linked to pre-dispersal predators. This could be done by excluding ants from certain branches or plants and assess to assess the effect of ant exclusion on pre- dispersal seed predation.

The hypothesis of selective plant abortion linked to pollen origin could be experimentally tested selective fruit abortion in T. velutina by conducting an experiment with two types of manual crosses: Self-pollinating half the flowers

200 Chapter 6: General discussion within a plant, and outcrossing the remaining flowers; and compare fruit abortion rates between treatments.

Ant patrolling & dioecy

In T. velutina ant patrolling had contrasting effects on male and female fitness components. Ant patrolling had a negative effect on male fitness reducing the number of pollen grains sired per flower, but a positive effect on female fitness by reducing the likelihood of fruit abortion and promoting outbred seeds and so increasing the quality of the offspring (Chapter 4). If this trend of opposite effects of ant patrolling on male and female fitness was maintained across most ant-plants, I would expect natural selection to favour different ant patrolling regimes or abundances in each sex on species with unisexual individuals (dioecious, androdioecious or gynodioecious).

I would expect natural selection to favour male plants that reduce or deplete ant patrolling to avoid the costs of ants of pollen exports. Whilst female plants would be under weaker or no selective pressure to deplete ant patrolling, assuming that EFN secretion or ant rewards are not very costly. Although, in dioecious plants, female individuals do not need ant patrolling to promote outcrossing, ant patrolling may reduce fruit abortion to florivores, and so ant patrolling may result in fewer fitness benefits. However, this contrast may be stronger in androdioecious plants, where individuals bearing bisexual flowers may benefit from ant patrolling promoting pollinator relocation, and so enhance outcrossing rates over geitonogamy. Exploring the effects of ant patrolling on male and female fitness components of dioecious ant-plants bearing EFN such as Diospyros, Fraxinus, Ilex, Populus, Salix, Costus, may uncover interesting findings on the role of myrmecophily as a selective pressure for the evolution of ant-plants of mating systems. However, herbivory levels also

201 Chapter 6: General discussion vary between plant sexes (refs), and so these costs should also be considered when estimating the net effects of ant patrolling on the fitness of each plant sex.

Herkogamy, outcrossing and ant aggressivity

Pollinators are usually considered the main selective agents of floral traits (Mothershead and Marquis 2000, Parachnowitsch and Kessler 2010), however other agents interacting with plants have proven to also play important roles in shaping the floral and mating system traits of plants (Rudgers and Gardener 2004, Parachnowitsch and Caruso 2008, Hernández-Cumplido et al. 2010, Kessler et al. 2011, Hernandez‐Cumplido et al. 2016). Ant patrolling has been suggested as a factor influencing pollinator visitation and plant fitness (Altshuler 1999, Ness 2006). In Chapter 4, I showed an example of ant patrolling affecting pollinator visit duration with cascading effects on T. velutina’s mating system and fitness. By affecting pollinator visitation patterns, ant patrolling may impose pollinator limitation or unreliability to the host plant, and so influence plants mating systems further than conceived so far.

Previous studies have demonstrated that outcrossing/selfing, self- compatibility and herkogamy (anther-stigma distance) respond to pollinator reliability (Opedal et al. 2016). Populations with unreliable pollinators showed higher compatibility, lower herkogamy and higher selfing rates, compared to populations with reliable pollinators (Opedal et al. 2016). These results are consistent with the reproductive assurance hypothesis, which suggests that plants with pollinator limitation may develop floral and mating system traits to promote self-pollination, as a means to assure seed set.

T. velutina is a self-compatible species incapable of autonomous pollination, and so requires pollinators to transport pollen grains onto the stigmas in order to set seed (Sosenski et al. 2016). Ant patrolling in T. velutina

202 Chapter 6: General discussion significantly decreased stigmatic pollen load by 150%, (Chapter 4), which may imply negative effects of ant patrolling on pollinator reliability. However, ant patrolling also decreased selfing, which is not consistent with pollinator unreliability. Further studies testing changes in herkogamy in T. velutina populations with varying levels of ant aggressivity would shed light on the long- term evolutionary effects of ant patrolling on this host plant’s mating system. Following the reproductive assurance hypothesis, I would expect populations with more aggressive ants to have lower pollinator reliability, and hence lower herkogamy to promote self-pollination.

203

THE END.

Definition of Poetry By Samuel Taylor Coleridge

“Poetry is not the proper antithesis to prose but to science. Poetry is opposed to science.”

On reading poetry

“If I had to live my life again I would have a rule to read some poetry and listen to some music at least once every week; for perhaps parts of my brain now atrophied could thus have been kept active through use. The loss of these tastes is a loss of happiness and may possibly be injurious to the intellect, and more probably to the moral character, by enfeebling the emotional part of our nature.”

Charles Darwin

204

REFERENCES

Abbott, K. R., and R. Dukas. 2009. Honeybees consider flower danger in their waggle dance. Animal Behaviour 78:633-635. Adler, L. S. 2008. Selection by pollinators and herbivores on attraction and defense. Pages 162-173 in K. J. Tilmon, editor. The evolutionary biology of herbivorous insects. Specialization, speciation and radiation. University California Press, Berkeley. Afkhami, M. E., J. A. Rudgers, and J. J. Stachowicz. 2014. Multiple mutualist effects: conflict and synergy in multispecies mutualisms. Ecology 95:833-844. Agarwal, V. M., and N. Rastogi. 2008. Deterrent effect of a guild of extrafloral nectary‐visiting ant species on Raphidopalpa foveicollis, a major insect pest of sponge gourd, Luffa cylindrica. Entomologia experimentalis et applicata 128:303-311. Altshuler, D. L. 1999. Novel interactions of non-pollinating ants with pollinators and fruit consumers in a tropical forest. Oecologia 119:600- 606. Alves-Silva, E., A. Bächtold, G. J. Barônio, H. M. Torezan-Silingardi, and K. Del-Claro. 2015. Ant–herbivore interactions in an extrafloral nectaried plant: are ants good plant guards against curculionid beetles? Journal of natural history 49:841-851. Alves‐Silva, E., G. J. Barônio, H. M. Torezan‐Silingardi, and K. Del‐Claro. 2013. Foraging behavior of Brachygastra lecheguana (Hymenoptera: Vespidae) on Banisteriopsis malifolia (Malpighiaceae): Extrafloral nectar consumption and herbivore predation in a tending ant system. Entomological Science 16:162-169. Apple, J., and D. Feener Jr. 2001. Ant visitation of extrafloral nectaries of Passiflora: the effects of nectary attributes and ant behavior on patterns in facultative ant-plant mutualisms. Oecologia 127:409-416. Arbo, M. M. 2005. Estudios sistemáticos en Turnera (Turneraceae). III. Series Anomalae y Turnera. Bonplandia 14:115-318. Armbruster, W. 1997. Exaptations link evolution of plant-herbivore and plant- pollinator interactions: a phylogenetic inquiry. Ecology. Armbruster, W. S. 1985. Patterns of character divergence and the evolution of reproductive ecotypes of Dalechampia scandens (Euphorbiaceae). Evolution 39:733-752. Armbruster, W. S. 1988. Multilevel comparative analysis of the morphology, function, and evolution of Dalechampia blossoms. Ecology 69:1746-1761. Assunção, M. A., H. M. Torezan-Silingardi, and K. Del-Claro. 2014. Do ant visitors to extrafloral nectaries of plants repel pollinators and cause an indirect cost of mutualism? Flora-Morphology, Distribution, Functional Ecology of Plants 209:244-249.

205

Baldock, K. C., J. Memmott, J. C. Ruiz-Guajardo, D. Roze, and G. N. Stone. 2011. Daily temporal structure in African savanna flower visitation networks and consequences for network sampling. Ecology 92:687-698. Ballantyne, G., and P. Willmer. 2012. Nectar theft and floral ant-repellence: a link between nectar volume and ant-repellent traits? PLoS One 7:e43869. Ballaré, C. L. 2011. Jasmonate-induced defenses: a tale of intelligence, collaborators and rascals. Trends in plant science 16:249-257. Barrios, B., S. R. Pena, A. Salas, and S. Koptur. 2016. Butterflies visit more frequently, but bees are better pollinators: the importance of mouthpart dimensions in effective pollen removal and deposition. AoB Plants 8. Bates, D., M. Maechler, B. Bolker, S. Walker, R. H. B. Christensen, H. Singmann, B. Dai, G. Grothendieck, P. Green, and M. B. Bolker. 2016. Package ‘lme4’. R Package Version 1.1–10. Bazzaz, F. A., N. R. Chiariello, P. D. Coley, and L. F. Pitelka. 1987. Allocating resources to reproduction and defense. BioScience 37:58-67. Beattie, A. J. 1985. The evolutionary ecology of ant-plant mutualisms. Cambridge University Press. Begon, M., J. L. Harper, and C. R. Townsend. 1986. Ecology. Individuals, populations and communities. Fourth edition edition. Blackwell Scientific Publications, Singapore. Belt, T. 1874. The naturalist in Nicaragua. University Press of the Pacific, Honolulu, Hawaii. Benitez-Vieyra, S., M. Ordano, J. Fornoni, K. Boege, and C. A. DomÍnguez. 2010. Selection on signal–reward correlation: limits and opportunities to the evolution of deceit in Turnera ulmifolia L. Journal of Evolutionary Biology 23:2760-2767. Bentley, B. 1976. Plants bearing extrafloral nectaries and the associated ant community: interhabitat differences in the reduction of herbivore damage. Ecology. Bentley, B. 1977a. Extrafloral nectaries and protection by pugnacious bodyguards. Annual Review of Ecology and Systematics 8:407-427. Bentley, B. L. 1977b. The protective function of ants visiting the extrafloral nectaries of Bixa orellana (Bixaceae). The Journal of Ecology 65:27-38. Bolton, B. 1987. A review of the Solenopsis genus-group and revision of Afrotropical Monomorium Mayr (Hymenoptera: Formicidae). Bulletin of the British Museum (Natural History), Entomology 54:263-452. Brechbühl, R., J. Casas, and S. Bacher. 2010a. Ineffective crypsis in a crab spider: a prey community perspective. Proceedings of the Royal Society of London B: Biological Sciences 277:739-746. Brechbühl, R., C. Kropf, and S. Bacher. 2010b. Impact of flower-dwelling crab spiders on plant-pollinator mutualisms. Basic and Applied Ecology 11:76-82. Bronstein, J. L. 1991. The nonpollinating wasp fauna of Ficus pertusa: exploitation of a mutualism? Oikos:175-186.

206

Bronstein, J. L. 2001. The exploitation of mutualisms. Ecology Letters 4:277- 287. Buchanan, A., and N. Underwood. 2013. Attracting pollinators and avoiding herbivores: insects influence plant traits within and across years. Oecologia 173:473-482. Byk, J., and K. Del-Claro. 2010. Nectar-and pollen-gathering Cephalotes ants provide no protection against herbivory: a new manipulative experiment to test ant protective capabilities. Acta Ethologica 13:33-38. Byk, J., and K. Del-Claro. 2011. Ant–plant interaction in the Neotropical savanna: direct beneficial effects of extrafloral nectar on ant colony fitness. Population Ecology 53:327-332. Campbell, S., and A. Kessler. 2013. Plant mating system transitions drive the macroevolution of defense strategies. Proceedings of the National Academy of Sciences of the United States of America 110:3973-3978. Carlson, J. E., and K. E. Harms. 2007. The benefits of bathing buds: water calyces protect flowers from a microlepidopteran herbivore. Biology Letters 3:405-407. Carper, A. L., L. S. Adler, and R. E. Irwin. 2016. Effects of florivory on plant‐ pollinator interactions: Implications for male and female components of plant reproduction. American Journal of Botany 103:1061-1070. Carroll, C. R., and D. H. Janzen. 1973. Ecology of foraging by ants. Annual Review of Ecology and Systematics 4:231-257. Cembrowski, A. R., M. G. Tan, J. D. Thomson, and M. E. Frederickson. 2014. Ants and ant scent reduce bumblebee pollination of artificial flowers. The American Naturalist 183:133-139. Chamberlain, S. A., and J. N. Holland. 2008. DENSITY‐MEDIATED, CONTEXT‐DEPENDENT CONSUMER–RESOURCE INTERACTIONS BETWEEN ANTS AND EXTRAFLORAL NECTAR PLANTS. Ecology 89:1364-1374. Chamberlain, S. A., and J. N. Holland. 2009. Quantitative synthesis of context dependency in ant-plant protection mutualisms. Ecology 90:2384-2392. Chamberlain, S. A., and J. A. Rudgers. 2012. How do plants balance multiple mutualists? Correlations among traits for attracting protective bodyguards and pollinators in cotton (Gossypium). Evolutionary Ecology 26:65-77. Christianini, A. V., and G. Machado. 2004. Induced biotic responses to herbivory and associated cues in the Amazonian ant‐plant Maieta poeppigii. Entomologia experimentalis et applicata 112:81-88. Clark, C. W., and R. Dukas. 1994. Balancing foraging and antipredator demands: an advantage of sociality. The American Naturalist 144:542- 548. Clinchy, M., M. J. Sheriff, and L. Y. Zanette. 2013. Predator‐induced stress and the ecology of fear. Functional Ecology 27:56-65. Cohen, J. 1988. Statistical power analysis for the behavioral sciences. 2nd. Hillsdale, NJ: erlbaum.

207

Comba, L., S. A. Corbet, A. Barron, A. Bird, S. Collinge, i. N. Miyazak, and M. Powell. 1999. Garden Flowers: Insect Visits and the Floral Reward of Horticulturally-modified Variants. Annals of Botany 83:73-86. Conner, J. K., and S. Rush. 1996. Effects of flower size and number on pollinator visitation to wild radish, Raphanus raphanistrum. Oecologia 105:509-516. Cuautle, M., and V. Rico-Gray. 2003. The effect of wasps and ants on the reproductive success of the extrafloral nectaried plant Turnera ulmifolia (Turneraceae). Functional Ecology 17:417-423. Cuautle, M., V. Rico-Gray, and C. Díaz-Castelazo. 2005. Effects of ant behaviour and presence of extrafloral nectaries on seed dispersal of the Neotropical myrmecochore Turnera ulmifolia L.(Turneraceae). Biological Journal of the Linnean Society 86:67-77. Damián, X., J. Fornoni, C. A. Domínguez, K. Boege, and J. Baltzer. 2018. Ontogenetic changes in the phenotypic integration and modularity of leaf functional traits. Functional Ecology 32:234-246. Dáttilo, W., N. Lara-Rodríguez, P. Jordano, P. R. Guimarães, J. N. Thompson, R. J. Marquis, L. P. Medeiros, R. Ortiz-Pulido, M. A. Marcos-García, and V. Rico-Gray. 2016. Unravelling Darwin's entangled bank: architecture and robustness of mutualistic networks with multiple interaction types. Proc. R. Soc. B 283:20161564. Dattilo, W., V. RICO‐GRAY, D. J. Rodrigues, and T. J. Izzo. 2013. Soil and vegetation features determine the nested pattern of ant–plant networks in a tropical rainforest. Ecological Entomology 38:374-380. Davidson, D. W., and S. C. Cook. 2008. Tropical arboreal ants: linking nutrition to roles in rainforests ecosystems.in W. P. Carson and S. A. Schnitzer, editors. Tropical forest community ecology. Wiley-Blackwell, United Kingdom. Davidson, D. W., J. T. Longino, and R. R. Snelling. 1988. Pruning of host plant neighbors by ants: an experimental approach. Ecology 69:801-808. de Jong, T. J., N. M. Waser, and P. G. Klinkhamer. 1993. Geitonogamy: the neglected side of selfing. Trends in ecology & evolution 8:321-325. de Jong, T. J., N. M. Waser, M. V. Price, and R. M. Ring. 1992. Plant size, geitonogamy and seed set in Ipomopsis aggregata. Oecologia 89:310-315. Del-Claro, K., D. Lange, H. M. Torezan-Silingardi, D. V. Anjos, E. S. Calixto, W. Dáttilo, and V. Rico-Gray. 2018. The Complex Ant–Plant Relationship Pages 59-71 in R.-G. V. Dattilo Wesley, editor. Ecological Networks in the Tropics. Springer. Del-Claro, K., V. Rico-Gray, H. M. Torezan-Silingardi, E. Alves-Silva, R. Fagundes, D. Lange, W. Dáttilo, A. A. Vilela, A. Aguirre, and D. Rodriguez-Morales. 2016. Loss and gains in ant–plant interactions mediated by extrafloral nectar: fidelity, cheats, and lies. Insectes Sociaux 63:207-221.

208

Del‐Claro, K., and R. J. Marquis. 2015. Ant Species Identity has a Greater Effect than Fire on the Outcome of an Ant Protection System in B razilian C errado. Biotropica 47:459-467. Díaz-Castelazo, C., V. Rico-Gray, F. Ortega, and G. Angeles. 2005. Morphological and secretory characterization of extrafloral nectaries in plants of coastal Veracruz, Mexico. Annals of Botany 96:1175-1189. Dukas, R. 2001. Effects of perceived danger on flower choice by bees. Ecology Letters 4:327-333. Dukas, R., and D. H. Morse. 2003. Crab spiders affect flower visitation by bees. Oikos 101:157-163. Dutton, E. M., and M. E. Frederickson. 2012. Why ant pollination is rare: new evidence and implications of the antibiotic hypothesis. -Plant Interactions 6:561-569. Dutton, E. M., E. Y. Luo, A. R. Cembrowski, J. S. Shore, M. E. Frederickson, C. M. Caruso, and Y. Michalakis. 2016. Three’s a Crowd: Trade-Offs between Attracting Pollinators and Ant Bodyguards with Nectar Rewards in Turnera. The American Naturalist 188:38-51. Dyer, L. 2008. Tropical tritrophic interactions: nasty hosts and ubiquitous cascades.in W. P. Carson and S. A. Schnitzer, editors. Tropical forest community ecology. Wiley-Blackwell, United Kingdom. Dyer, L., and C. Phyllis. 2002. Temperate versus tropical communities. Pages 67-88 in B. H. Teja Tscharntke, editor. Multitrophic level interactions. Cambridge University Press, Cambridge. Eckert, C. G. 2000. Contributions of autogamy and geitonogamy to self‐ fertilization in a mass‐flowering, clonal plant. Ecology 81:532-542. Elias, T., W. Rozich, and L. Newcombe. 1975. The foliar and floral nectaries of Turnera ulmifolia L. American Journal of Botany 62:570-576. Escalante-Pérez, M., and M. Heil. 2012. Nectar secretion: its ecological context and physiological regulation. Pages 187-219 in B. F. Vivanco J., editor. Secretions and exudates in biological systems. Springer-Verlag, Berlin, Heidelberg. Ettershank, G. 1966. A generic revision of the world Myrmicinae related to Solenopsis and Pheidologeton (Hymenoptera: Formicidae). Australian Journal of Zoology 14:73-171. Fagundes, R., W. Dáttilo, S. Ribeiro, V. Rico-Gray, P. Jordano, and K. Del- Claro. 2017. Differences among ant species in plant protection are related to production of extrafloral nectar and degree of leaf herbivory. Biological Journal of the Linnean Society 122:71-83. Falcão, J., W. Dáttilo, and T. Izzo. 2014. Temporal variation in extrafloral nectar secretion in different ontogenic stages of the fruits of Alibertia verrucosa S. Moore (Rubiaceae) in a Neotropical savanna. Journal of Plant Interactions 9:137-142. Fiala, B., U. Maschwitz, T. Y. Pong, and A. J. Helbig. 1989. Studies of a South East Asian ant-plant association: protection of Macaranga trees by Crematogaster borneensis. Oecologia 79:463-470.

209

Fisher, B. L., L. d. S. L. Sternberg, and D. Price. 1990. Variation in the use of orchid extrafloral nectar by ants. Oecologia 83:263-266. Fonseca, C. 1993. Nesting space limits colony size of the plant-ant Pseudomyrmex concolor. Oikos:473-482. Fonseca, C. R. 1999. Amazonian ant–plant interactions and the nesting space limitation hypothesis. Journal of Tropical Ecology 15:807-825. Fontaine, C., P. R. Guimarães, S. Kéfi, N. Loeuille, J. Memmott, W. H. van Der Putten, F. J. van Veen, and E. Thébault. 2011. The ecological and evolutionary implications of merging different types of networks. Ecology Letters 14:1170-1181. Frederickson, M. E. 2009. Conflict over Reproduction in an Ant‐Plant Symbiosis: Why Allomerus octoarticulatus Ants Sterilize Cordia nodosa Trees. The American Naturalist 173:675-681. Gadagkar, R. 1990. Evolution of eusociality: the advantage of assured fitness returns. Phil. Trans. R. Soc. Lond. B 329:17-25. Galen, C. 1999. Flowers and enemies: predation by nectar-thieving ants in relation to variation in floral form of an alpine wildflower, Polemonium viscosum. Oikos:426-434. Galen, C. 2005. Catching ants with honey: an experimental test of distraction and satiation as alternative modes of escape from flower-damaging ants. Oecologia 144:80-87. Galen, C., and J. Cuba. 2001. DOWN THE TUBE: POLLINATORS, PREDATORS, AND THE EVOLUTION OF FLOWER SHAPE IN THE ALPINE SKYPILOT, POLEMONIUM VISCOSUM. Evolution 55. Gaume, L., and D. McKey. 1999. An ant-plant mutualism and its host-specific parasite: activity rhythms, young leaf patrolling, and effects on herbivores of two specialist plant-ants inhabiting the same myrmecophyte. Oikos 84:130-144. Gaume, L., M. Zacharias, and R. M. Borges. 2005. Ant-plant conflicts and a novel case of castration parasitism in a myrmecophyte. Evolutionary Ecology Research 7:435-452. Ghazoul, J. 2001. Can floral repellents pre‐empt potential ant–plant conflicts? Ecology Letters 4:295-299. Goncalves-Souza, T., P. M. Omena, J. C. Souza, and G. Q. Romero. 2008. Trait‐ mediated effects on flowers: artificial spiders deceive pollinators and decrease plant fitness. Ecology 89:2407-2413. Grover, C. D., A. D. Kay, J. A. Monson, T. C. Marsh, and D. A. Holway. 2007. Linking nutrition and behavioural dominance: carbohydrate scarcity limits aggression and activity in Argentine ants. Proceedings of the Royal Society of London B: Biological Sciences 274:2951-2957. Guzmán-Novoa, E., A. Correa Benítez, L. Espinosa Montaño, and G. Guzmán Novoa. 2011. Colonization, impact and control of Africanized bees in Mexico. Revista Veterninaria México 42:149-178.

210

Hadfield, J. D. 2010. MCMC methods for multi-response generalized linear mixed models: the MCMCglmm R package. Journal of Statistical Software 33:1-22. Hanna, C., I. Naughton, C. Boser, R. Alarcón, K.-L. J. Hung, and D. Holway. 2015. Floral visitation by the Argentine ant reduces bee visitation and plant seed set. Ecology 96:222-230. Harley, R. 1991. The greasy pole syndrome. Huxley, C, R,, Cutler, D, F ed (s). Ant-plant interactions. Oxford Univ. Press: Oxford, etc:430-433. Heil, M. 2008. Indirect defence via tritrophic interactions. The New phytologist 178:41-61. Heil, M. 2011. Nectar: generation, regulation and ecological functions. Trends in plant science 16:191-200. Heil, M. 2015. Extrafloral nectar at the plant-insect interface: a spotlight on chemical ecology, phenotypic plasticity, and food webs. Annual review of entomology 60:213-232. Heil, M., F. Brigitte, M. Ulrich, and K. E. Linsenmair. 2001a. On benefits of indirect defence: short- and long-term studies of antiherbivore protection via mutualistic ants. Oecologia 126. Heil, M., M. González-Teuber, L. W. Clement, S. Kautz, M. Verhaagh, and J. C. S. Bueno. 2009. Divergent investment strategies of Acacia myrmecophytes and the coexistence of mutualists and exploiters. Proceedings of the National Academy of Sciences 106:18091-18096. Heil, M., T. Koch, A. Hilpert, B. Fiala, W. Boland, and K. E. Linsenmair. 2001b. Extrafloral nectar production of the ant-associated plant, Macaranga tanarius, is an induced, indirect, defensive response elicited by jasmonic acid. Proceedings of the National Academy of Sciences 98:1083-1088. Hernández-Cumplido, J., B. Benrey, and M. Heil. 2010. Attraction of flower visitors to plants that express indirect defence can minimize ecological costs of ant–pollinator conflicts. Journal of Tropical Ecology 26:555-557. Hernandez‐Cumplido, J., B. Forter, X. Moreira, M. Heil, and B. Benrey. 2016. Induced Floral and Extrafloral Nectar Production Affect Ant‐pollinator Interactions and Plant Fitness. Biotropica. Herre, E. A., N. Knowlton, U. G. Mueller, and S. A. Rehner. 1999. The evolution of mutualisms: exploring the paths between conflict and cooperation. Trends in ecology & evolution 14:49-53. Herrera, C. 2000. Measuring the effects of pollinators and herbivores: evidence for non-additivity in a perennial herb. Ecology. Herrera, C., M. Medrano, P. Rey, A. Sanchez-Lafuente, M. Garcia, J. Guitian, and A. Manzaneda. 2002. Interaction of pollinators and herbivores on plant fitness suggests a pathway for correlated evolution of mutualism- and antagonism-related traits. Proceedings of the National Academy of Sciences of the United States of America 99:16823-16828. Herrera, C. M., and O. Pellmyr. 2009. Plant Animal Interactions : An Evolutionary Approach. Wiley, Hoboken.

211

Hinde, J. 1982. Compound Poisson regression models. Pages 109-121 in G. R., editor. GLIM 82: Proceedings of the International Conference on Generalised Linear Models. Springer, New York, NY. Holland, J. N., S. A. Chamberlain, and K. C. Horn. 2010. Temporal variation in extrafloral nectar secretion by reproductive tissues of the senita cactus, Pachycereus schottii (Cactaceae), in the Sonoran Desert of Mexico. Journal of Arid Environments 74. Holland, J. N., A. C. Scott, and E. X. M. Tom. 2011. Consequences of ants and extrafloral nectar for a pollinating seed-consuming mutualism: ant satiation, floral distraction or plant defense? Oikos 120. Hothorn, T., F. Bretz, P. Westfall, and R. Heiberger. 2008. Multcomp: simultaneous inference for general linear hypotheses. R Package Version 1.0-3. Huth, C. J., and O. Pellmyr. 2000. Pollen‐mediated selective abortion in yuccas and its consequences for the plant–pollinator mutualism. Ecology 81:1100-1107. Irwin, R. E., L. S. Adler, and A. K. Brody. 2004. The dual role of floral traits: pollinator attraction and plant defense. Ecology 85:1503-1511. Itioka, T., M. Nomura, Y. Inui, T. Itino, and T. Inoue. 2000. Difference in Intensity of Ant Defense among Three Species of Macaranga Myrmecophytes in a Southeast Asian Dipterocarp Forest 1. Biotropica 32:318-326. Izzo, T. J., and H. L. Vasconcelos. 2002. Cheating the cheater: domatia loss minimizes the effects of ant castration in an Amazonian ant-plant. Oecologia 133:200-205. J., M. C., B. Jordi, J. Pedro, and K. Vlastimil. 2009. Diversity in a complex ecological network with two interaction types. Oikos 118:122-130. Jackson, B. D., and E. D. Morgan. 1993. Insect chemical communication: pheromones and exocrine glands of ants. Chemoecology 4:125-144. Jackson, D. E., and F. L. Ratnieks. 2006. Communication in ants. Current biology 16:R570-R574. Janzen, D. H. 1966. Coevolution of mutualism between ants and acacias in Central America. Evolution:249-275. Janzen, D. H. 1975. Pseudomyrmex nigropilosa: a parasite of a mutualism. Science 188:936-937. Johnson, S. G., L. F. Delph, and C. L. Elderkin. 1995. The effect of petal-size manipulation on pollen removal, seed set, and insect-visitor behavior in Campanula americana. Oecologia 102:174-179. Jones, E. I. 2010. Optimal foraging when predation risk increases with patch resources: an analysis of pollinators and ambush predators. Oikos 119:835-840. Jones, E. I., and A. Dornhaus. 2011. Predation risk makes bees reject rewarding flowers and reduce foraging activity. Behavioral Ecology and Sociobiology 65:1505-1511.

212

Jones, I. M., and S. Koptur. 2015. Dynamic extrafloral nectar production: the timing of leaf damage affects the defensive response in Senna mexicana var. chapmanii (Fabaceae). American Journal of Botany 102:58-66. Junker, R., A. Y. Chung, and N. Blüthgen. 2007. Interaction between flowers, ants and pollinators: additional evidence for floral repellence against ants. Ecological Research 22:665-670. Junker, R. R., and N. Blüthgen. 2008. Floral scents repel potentially nectar- thieving ants. Evolutionary Ecology Research 10:295-308. Kautz, S., H. T. Lumbsch, P. S. Ward, and M. Heil 2009. HOW TO PREVENT CHEATING: A DIGESTIVE SPECIALIZATION TIES MUTUALISTIC PLANT-ANTS TO THEIR ANT-PLANT PARTNERS. Evolution 63. Kearns, C. A., and D. W. Inouye. 1993. Techniques for pollination biologists. University Press of Colorado, Niwot, Colorado. Keeler, K. H. 2014. World list of plants with extrafloral nectaries. http://biosci- labs.unl.edu/Emeriti/keeler/extrafloral/Cover.htm. Kerner, A. J. 1878. Flowers and their unbidden guests. C. Kegan Paul & Co. , London. Kersch, M. F., and C. R. Fonseca. 2005. Abiotic factors and the conditional outcome of an ant-plant mutualism. Ecology 86:2117-2126. Kessler, A., and R. Halitschke. 2009. Testing the potential for conflicting selection on floral chemical traits by pollinators and herbivores: predictions and case study. Functional Ecology 23:901-912. Kessler, A., R. Halitschke, and K. Poveda. 2011. Herbivory-mediated pollinator limitation: Negative impacts of induced volatiles on plant-pollinator interactions. Ecology. Kessler, A., and M. Heil. 2011. The multiple faces of indirect defences and their agents of natural selection. Functional Ecology 25. King, C., G. Ballantyne, and P. G. Willmer. 2013. Why flower visitation is a poor proxy for pollination: measuring single‐visit pollen deposition, with implications for pollination networks and conservation. Methods in Ecology and Evolution 4:811-818. Koptur, S., I. M. Jones, and J. E. Peña. 2015. The influence of host plant extrafloral nectaries on multitrophic interactions: an experimental investigation. PLoS One 10:e0138157. Krupnick, G. A., and A. E. Weis. 1999. The effect of floral herbivory on male and female reproductive success in Isomeris arborea. Ecology 80:135- 149. Lach, L. 2008. Floral visitation patterns of two invasive ant species and their effects on other hymenopteran visitors. Ecological Entomology 33:155- 160. Lange, D., E. S. Calixto, and K. Del-Claro. 2017. Variation in extrafloral nectary productivity influences the ant foraging. PLoS One 12:1-13. Leege, L. M., and L. M. Wolfe. 2002. Do floral herbivores respond to variation in flower characteristics in Gelsemium sempervirens (Loganiaceae), a distylous vine? American Journal of Botany 89:1270-1274.

213

Lemus Domínguez, I. 2014. Efecto de la variación intraindividual en el néctar en la interacción entre turnera velutina y sus polinizadores y defensores asociados. Universidad Nacional Autónoma de México (UNAM), México D.F. LeVan, K. E., K.-L. J. Hung, K. R. McCann, J. T. Ludka, and D. A. Holway. 2014. Floral visitation by the Argentine ant reduces pollinator visitation and seed set in the coast barrel cactus, Ferocactus viridescens. Oecologia 174:163-171. Lloyd, D. G. 1992. Self-and cross-fertilization in plants. II. The selection of self- fertilization. International Journal of Plant Sciences 153:370-380. Lohman, D., A. Zangerl, and M. Berenbaum. 1996. Impact of floral herbivory by parsnip webworm (Oecophoridae: Depressaria pastinacella Duponchel) on pollination and fitness of wild parsnip (Apiaceae: Pastinaca sativa L.). American Midland Naturalist:407-412. Loveless, M. D., and J. L. Hamrick. 1984. Ecological determinants of genetic structure in plant populations. Annual Review of Ecology and Systematics 15:65-95. Malé, P.-J. G., C. Leroy, A. Dejean, A. Quilichini, and J. Orivel. 2012. An ant symbiont directly and indirectly limits its host plant’s reproductive success. Evolutionary Ecology 26:55-63. Malé, P.-J. G., C. Leroy, L. Lusignan, F. Petitclerc, A. Quilichini, and J. Orivel. 2015. The reproductive biology of the myrmecophyte, Hirtella physophora, and the limitation of negative interactions between pollinators and ants. Arthropod-Plant Interactions 9:23-31. Marshall, D. L., and N. C. Ellstrand. 1988. Effective mate choice in wild radish: evidence for selective seed abortion and its mechanism. The American Naturalist 131:739-756. Martin, H., and M. Doyle. 2003. P ROTECTIVE A NT-PLANT I NTERACTIONS AS M ODEL S YSTEMS IN E COLOGICAL AND E VOLUTIONARY R ESEARCH. Annual Review of Ecology, Evolution, and Systematics 34. Martínez-Bauer, A. E., G. C. Martínez, D. J. Murphy, and M. Burd. 2015. Multitasking in a plant–ant interaction: how does Acacia myrtifolia manage both ants and pollinators? Oecologia 178:461-471. McKey, D. 1979. The distribution of secondary compounds within plants. Herbivores: their interaction with secondary plant metabolites. Academic Press, New York:55-133. Melián, C. J., J. Bascompte, P. Jordano, and V. Krivan. 2009. Diversity in a complex ecological network with two interaction types. Oikos 118:122- 130. Miller, T. E. X. 2007. Does having multiple partners weaken the benefits of facultative mutualism? A test with cacti and cactus-tending ants. Oikos 116:500-512. Mothershead, K., and R. J. Marquis. 2000. Fitness impacts of herbivory through indirect effects on plant-pollinator interactions in Oenothera macrocarpa. Ecology 81:30-40.

214

Mulcahy, D. L. 1979. The Rise of the Angiosperms: A Genecological Factor. Science 206:20-23. Nahas, L., M. O. Gonzaga, and K. Del‐Claro. 2012. Emergent impacts of ant and spider interactions: herbivory reduction in a tropical savanna tree. Biotropica 44:498-505. Ne'eman, G., A. Jürgens, L. Newstrom‐Lloyd, S. G. Potts, and A. Dafni. 2010. A framework for comparing pollinator performance: effectiveness and efficiency. Biological Reviews 85:435-451. Ness, J. 2003. Catalpa bignonioides alters extrafloral nectar production after herbivory and attracts ant bodyguards. Oecologia 134:210-218. Ness, J., W. Morris, and J. Bronstein. 2006. Integrating quality and quantity of mutualistic service to contrast ant species protecting Ferocactus wislizeni. Ecology 87:912-921. Ness, J., W. Morris, and J. L. Bronstein. 2009. For ant‐protected plants, the best defense is a hungry offense. Ecology 90:2823-2831. Ness, J. H. 2006. A mutualism's indirect costs: the most aggressive plant bodyguards also deter pollinators. Oikos 113:506-514. Nicklen, E. F., and D. Wagner. 2006. Conflict resolution in an ant–plant interaction: Acaciaconstricta traits reduce ant costs to reproduction. Oecologia 148:81-87. Nicolson, S., M. Nepi, and E. Pacini. 2007. Nectaries and Nectar. Springer, Dordrecht, Netherlands. Niesenbaum, R. A. 1999. The effects of pollen load size and donor diversity on pollen performance, selective abortion, and progeny vigor in Mirabilis jalapa (Nyctaginaceae). American Journal of Botany 86:261-268. Ochoa Sánchez, M. A. 2016. El valor funcional de la integración floral en Turnera velutina. Universidad Nacional Autónoma de México (UNAM), México. Ochoa-López, S. 2013. Caracterización de las trayectorias ontogenéticas de la defensa y tolerancia de Turnera velutina y su variación genética. Universidad Nacional Autónoma de México, México. Ochoa-López, S., R. Rebollo, K. E. Barton, J. Fornoni, and K. Boege. 2018. Risk of herbivore attack and heritability of ontogenetic trajectories in plant defense. Oecologia:1-14. Ochoa-López, S., N. Villamil, P. Zedillo-Avelleyra, and K. Boege. 2015. Plant defence as a complex and changing phenotype throughout ontogeny. Annals of Botany 116:797-806. Ohm, J. R., and T. E. Miller. 2014. Balancing anti‐herbivore benefits and anti‐ pollinator costs of defensive mutualists. Ecology 95:2924-2935. Ollerton, J., A. Killick, E. Lamborn, S. Watts, and M. Whiston. 2007. Multiple meanings and modes: on the many ways to be a generalist flower. Taxon 56:717-728. Opedal, Ø. H., E. Albertsen, W. S. Armbruster, R. Pérez‐Barrales, M. Falahati‐ Anbaran, and C. Pélabon. 2016. Evolutionary consequences of ecological

215

factors: pollinator reliability predicts mating‐system traits of a perennial plant. Ecology Letters 19:1486-1495. Opedal, Ø. H., M. Falahati‐Anbaran, E. Albertsen, W. S. Armbruster, R. Pérez‐ Barrales, H. K. Stenøien, and C. Pélabon. 2017. Euglossine bees mediate only limited long‐distance gene flow in a tropical vine. New Phytologist 213:1898-1908. Ordano, M., J. Fornoni, K. Boege, and C. A. Domínguez. 2008. The adaptive value of phenotypic floral integration. New Phytologist 179:1183-1192. Orivel, J., L. Lambs, P.-J. G. Malé, C. Leroy, J. Grangier, T. Otto, A. Quilichini, and A. Dejean. 2011. Dynamics of the association between a long-lived understory myrmecophyte and its specific associated ants. Oecologia 165:369-376. Orona‐Tamayo, D., N. Wielsch, M. Escalante‐Pérez, A. Svatos, J. Molina‐ Torres, A. Muck, E. Ramirez‐Chávez, R. M. Ádame‐Alvarez, and M. Heil. 2013. Short‐term proteomic dynamics reveal metabolic factory for active extrafloral nectar secretion by Acacia cornigera ant‐plants. The Plant Journal 73:546-554. Pacini, E., and M. Nepi. 2007. Nectar production and presentation. Pages 167- 214 in S. Nicolson, M. Nepi, and E. Pacini, editors. Nectaries and Nectar. Springer, Dordrecht, Netherlands. Pacini, E., M. Nepi, and J. Vesprini. 2003. Nectar biodiversity: a short review. Plant Systematics and Evolution 238:7-21. Pacini, E., and S. Nicolson. 2007. Introduction. Pages 1-19 in S. Nicolson, M. Nepi, and E. Pacini, editors. Nectaries and Nectar. Springer, Dordrecht, Netherlands. Palmer, T., D. Doak, M. Stanton, J. Bronstein, E. Kiers, T. Young, J. Goheen, and R. Pringle. 2010. Synergy of multiple partners, including freeloaders, increases host fitness in a multispecies mutualism. Proceedings of the National Academy of Sciences of the United States of America 107:17234-17239. Palmer, T. M., and A. K. Brody. 2007. Mutualism as reciprocal exploitation: African plant‐ants defend foliar but not reproductive structures. Ecology 88:3004-3011. Parachnowitsch, A., and C. Caruso. 2008. Predispersal seed herbivores, not pollinators, exert selection on floral traits via female fitness. Ecology 89:1802-1810. Parachnowitsch, A., and A. Kessler. 2010. Pollinators exert natural selection on flower size and floral display in Penstemon digitalis. The New phytologist 188:393-402. Pineda, A., M. Dicke, C. M. Pieterse, and M. J. Pozo. 2013. Beneficial microbes in a changing environment: are they always helping plants to deal with insects? Functional Ecology 27:574-586. Poveda, K., I. Steffan-Dewenter, S. Scheu, and T. Tscharntke. 2003. Effects of below-and above-ground herbivores on plant growth, flower visitation and seed set. Oecologia 135:601-605.

216

Preisser, E. L., D. I. Bolnick, and M. F. Benard. 2005. Scared to death? The effects of intimidation and consumption in predator–prey interactions. Ecology 86:501-509. Pyke, G. H. 1991. What does it cost a plant to produce floral nectar? Nature 350:58-59. Queller, D. C. 1989. The evolution of eusociality: reproductive head starts of workers. Proceedings of the National Academy of Sciences 86:3224- 3226. Quintero, C., K. E. Barton, and K. Boege. 2013. The ontogeny of plant indirect defenses. Perspectives in plant ecology, evolution and systematics 15:245-254. Radhika, V., C. Kost, S. Bartram, M. Heil, and W. Boland. 2008. Testing the optimal defence hypothesis for two indirect defences: extrafloral nectar and volatile organic compounds. Planta 228:449-457. Raine, N., P. Willmer, and G. Stone. 2002. Spatial structuring and floral avoidance behavior prevent ant-pollinator conflict in a Mexican ant- acacia. Ecology 83:3086-3096. Raine, N. E., N. Gammans, I. J. Macfadyen, G. K. Scrivner, and G. N. Stone. 2004. Guards and thieves: antagonistic interactions between two ant species coexisting on the same ant‐plant. Ecological Entomology 29:345-352. Ramos-Castro, S. E. 2013. Determinación del componente genético de la media y la varianza en las señales y recompensas de las plantas a sus polinizadores. Master in Biological Sciences Dissertation. . Uiversidad Nacional Autónoma de México (UNAM), México, D.F. Rebollo Hernández, R. 2018. Conducta de oviposición, desempeño diferencial y secuestro de glucósidos cianogénicos en orugas de Euptoieta hegesia según el estadio ontognenético de su planta hospedera. . Universidad Nacional Autónoma de México, México. Rhoades, D. F. 1979. Evolution of plant chemical defense against herbivores. Herbivores: their interaction with secondary plant metabolites. Academic Press, New York:3-54. Rico-Gray, V., and P. S. Oliveira. 2007a. The ecology and evolution of ant-plant interactions. University of Chicago Press. Rico-Gray, V., and P. S. Oliveira. 2007b. Mutualism from Antagonism: Ants and Flowers. Pages 85-98 in V. Rico-Gray and P. S. Oliveira, editors. The ecology and evolution of ant-plant interactions. University of Chicago Press, Chicago. Robinson, E. J., D. E. Jackson, M. Holcombe, and F. L. Ratnieks. 2005. Insect communication:‘no entry’signal in ant foraging. Nature 438:442-442. Rodríguez-Robles, J. A., E. J. Meléndez, and J. D. Ackerman. 1992. Effects of display size, flowering phenology, and nectar availability on effective visitation frequency in Comparettia falcata (Orchidaceae). American Journal of Botany 79:1009-1017.

217

Romero, G. Q., P. A. Antiqueira, and J. Koricheva. 2011. A meta-analysis of predation risk effects on pollinator behaviour. PLoS One 6:e20689. Romero, G. Q., and T. J. Izzo. 2004. Leaf damage induces ant recruitment in the Amazonian ant-plant Hirtella . Journal of Tropical Ecology 20:675-682. Romero, G. Q., and J. Koricheva. 2011. Contrasting cascade effects of carnivores on plant fitness: a meta‐analysis. Journal of Animal Ecology 80:696-704. Rosumek, F., F. Silveira, F. de S Neves, N. de U Barbosa, L. Diniz, Y. Oki, F. Pezzini, G. Fernandes, and T. Cornelissen. 2009. Ants on plants: a meta- analysis of the role of ants as plant biotic defenses. Oecologia 160:537- 549. Rudgers, J., and M. Gardener. 2004. Extrafloral nectar as a resource mediating multispecies interactions. Ecology 85:1495-1502. Salazar-Rojas, B., V. Rico-Gray, A. Canto, and M. Cuautle. 2012. Seed fate in the myrmecochorous Neotropical plant Turnera ulmifolia L., from plant to germination. Acta oecologica 40:1-10. Sanz-Veiga, P. A., L. R. Jorge, S. Benitez-Vieyra, and F. W. Amorim. 2017. Pericarpial nectary-visiting ants do not provide fruit protection against pre-dispersal seed predators regardless of ant species composition and resource availability. PLoS One 12:1-18. Schaeffer, R. N., J. S. Manson, and R. E. Irwin. 2013. Effects of abiotic factors and species interactions on estimates of male plant function: a meta‐ analysis. Ecology Letters 16:399-408. Schappert, P. J., and J. S. Shore. 1995. Cyanogenesis in Turnera ulmifolia L. (Turneraceae). I. Phenotypic distribution and genetic variation for cyanogenesis on Jamaica. Heredity 74. Schilman, P. E., and F. Roces. 2006. Foraging energetics of a nectar-feeding ant: metabolic expenditure as a function of food-source profitability. Journal of experimental biology 209:4091-4101. Schoen, D. J., and D. G. Lloyd. 1992. Self-and cross-fertilization in plants. III. Methods for studying modes and functional aspects of self-fertilization. International Journal of Plant Sciences 153:381-393. Shenoy, M., V. Radhika, S. Satish, and R. Borges. 2012. Composition of Extrafloral Nectar Influences Interactions between the Myrmecophyte Humboldtia brunonis and its Ant Associates. Journal of chemical ecology 38:88-99. Sheriff, M. J., and J. S. Thaler. 2014. Ecophysiological effects of predation risk; an integration across disciplines. Oecologia 176:607-611. Sosenski, P., S. Ramos, C. A. Domínguez, K. Boege, and J. Fornoni. 2016. Pollination biology of the hexaploid self‐compatible species Turnera velutina (Passifloraceae). Plant Biology 19:101-107. Southwick, E. E. 1984. Photosynthate Allocation to Floral Nectar: A Neglected Energy Investment. Ecology 65:1775-1779. Stamp, N. 2003. Out of the quagmire of plant defense hypotheses. The Quarterly review of biology 78:23-55.

218

Stanton, M. L., and T. M. Palmer. 2011. The high cost of mutualism: effects of four species of East African ant symbionts on their myrmecophyte host tree. Ecology 92:1073-1082. Stanton, M. L., T. M. Palmer, T. P. Young, A. Evans, and M. L. Turner. 1999. Sterilization and canopy modification of a swollen thorn acacia tree by a plant-ant. Nature 401:578-581. Stephens, D. W., and J. R. Krebs. 1986. Foraging theory. Princeton University Press, Princeton, New Jersey. Stephenson, A. 1981. Flower and fruit abortion: proximate causes and ultimate functions. Annual Review of Ecology and Systematics 12:253-279. Strauss, S. 1997. Floral characters link herbivores, pollinators, and plant fitness. Ecology 78:1640-1645. Strauss, S., Y. , and R. Irwin, E. . 2004. ECOLOGICAL AND EVOLUTIONARY CONSEQUENCES OF MULTISPECIES PLANT-ANIMAL INTERACTIONS. Annual Review of Ecology, Evolution, and Systematics 35:435-466. Strauss, S. Y., J. K. Conner, and S. L. Rush. 1996. Foliar herbivory affects floral characters and plant attractiveness to pollinators: implications for male and female plant fitness. American Naturalist:1098-1107. Strauss, S. Y., D. H. Siemens, M. B. Decher, and T. Mitchell-Olds. 1999. Ecological costs of plant resistance to herbivores in the currency of pollination. Evolution 53:1105-1113. Takayuki, O. 2005. INDIRECT INTERACTION WEBS: Herbivore-Induced Effects Through Trait Change in Plants. Annual Review of Ecology, Evolution, and Systematics 36. Torres-Hernández, L., and V. Rico-Gray. 2000. Effect of nectar-foraging ants and wasps on the reproductive fitness of Turnera ulmifolia (Turneraceae) in a coastal sand dune in Mexico. … Mexicana Nueva Serie. Trager, M. D., S. Bhotika, J. A. Hostetler, G. V. Andrade, M. A. Rodriguez- Cabal, C. S. McKeon, C. W. Osenberg, and B. M. Bolker. 2010. Benefits for plants in ant-plant protective mutualisms: a meta-analysis. PLoS One 5:e14308-e14308. Travieso-Bello, A. C., and A. Campos. 2006. Los componentes del paisaje. Moreno-Casasola P.(Comp.). Entornos Veracruzanos: la costa de La Mancha. Instituto de Ecología AC Xalapa, Ver., México:139-150. Tscharntke, T., and B. Hawkins. 2002. Multitrophic level interactions: an introduction.in B. H. Teja Tscharntke, editor. Multitrophic level interactions. Cambridge University Press, Cambridge. Turner, R., J. Midgley, P. Barnard, R. Simmons, and S. Johnson. 2012. Experimental evidence for bird pollination and corolla damage by ants in the short-tubed flowers of Erica halicacaba (Ericaceae). South african journal of botany 79:25-31.

219

Vekemans, X., and O. J. Hardy. 2004. New insights from fine‐scale spatial genetic structure analyses in plant populations. Molecular ecology 13:921-935. Venales, B., and E. M. Barrows. 1985. Skippers: Pollinators or nectar thieves? Journal of the Lepidopterists Society 39:299-312. Villamil, N. 2017. Why are flowers sweeter than fruits or buds? Variation in extrafloral nectar secretion throughout the ontogeny of a myrmecophile. Biotropica 49:581-585. Villamil, N., K. Boege, and G. Stone. 2018. Ant-pollinator conflict results in pollinator deterrence but no nectar trade-offs. Frontiers in Plant Science 9:1093. Villamil, N., J. Márquez-Guzmán, and K. Boege. 2013. Understanding ontogenetic trajectories of indirect defence: ecological and anatomical constraints in the production of extrafloral nectaries. Annals of Botany 112:701-709. Villamil-Buenrostro, N. 2012. Caracterización de las trayectorias ontogenéticas de la defensa contra herbívoros en Turnera ulmifolia. Bachelor in Biology Dissertation. . Universidad Nacional Autónoma de México, México. Wagner, D. 2000. Pollen viability reduction as a potential cost of ant association for Acacia constricta (Fabaceae). American Journal of Botany 87:711-715. Wagner, D., and A. Kay. 2002. Do extrafloral nectaries distract ants from visiting flowers? An experimental test of an overlooked hypothesis. Evolutionary Ecology Research. Waser, N. M., and M. V. Price. 1991. Reproductive costs of self‐pollination in Ipomopsis aggregata (Polemoniaceae): are ovules usurped? American Journal of Botany 78:1036-1043. Weber, M., and K. Keeler. 2013. The phylogenetic distribution of extrafloral nectaries in plants. Annals of Botany 111:1251-1261. Willmer, P. 2011. Pollination and floral ecology. Princeton University Press, Princeton, US and Oxford, UK. Willmer, P. G., H. Cunnold, and G. Ballantyne. 2017. Insights from measuring pollen deposition: quantifying the pre-eminence of bees as flower visitors and effective pollinators. Arthropod-Plant Interactions 11:411- 425. Willmer, P. G., and K. Finlayson. 2014. Big bees do a better job: intraspecific size variation influences pollination effectiveness. Journal of Pollination Ecology. Willmer, P. G., C. V. Nuttman, N. E. Raine, G. N. Stone, J. G. Pattrick, K. Henson, P. Stillman, L. McIlroy, S. G. Potts, and J. T. Knudsen. 2009. Floral volatiles controlling ant behaviour. Functional Ecology 23:888- 900. Willmer, P. G., and G. N. Stone. 1997. How aggressive ant-guards assist seed- set in Acacia flowers. Nature 388:165-167.

220

Wu, L. Y., F. F. Chang, S. J. Liu, W. Scott Armbruster, and S. Q. Huang. 2018. Heterostyly promotes compatible pollination in buckwheats: Comparisons of intraflower, intraplant, and interplant pollen flow in distylous and homostylous Fagopyrum. American Journal of Botany 105:108-116. Yamawo, A., Y. Hada, and N. Suzuki. 2012. Variations in direct and indirect defenses against herbivores on young plants of Mallotus japonicus in relation to soil moisture conditions. Journal of plant research 125:71-76. Young, T. P., C. H. Stubblefield, and L. A. Isbell. 1996. Ants on swollen-thorn acacias: species coexistence in a simple system. Oecologia 109:98-107. Yu, D. W., and N. E. Pierce. 1998. A castration parasite of an ant–plant mutualism. Proceedings of the Royal Society of London B: Biological Sciences 265:375-382. Zedillo-Avelleyra, P. 2017. Variación genética en la expresión de trayectorias ontogenéticas de la defensa de Turnera velutina Research. Universidad Nacional Autónoma de México, Ciudad de México.

221 222

NATURAL HISTORY SUPPLEMENTARY NOTES

Ants

Behavioural observations catalogue

The behavioural catalogue presented here is a compilation of observations carried out throughout field experiments across years, particularly between 2014-2016, but also includes personal observations recorded prior to such period.

Dorymyrmex bicolor • Actively patrolling in groups 5-6 per branch or apex. Usually observed climbing up from the stem, searching for nectar. Very alert and aggressive, quickly detecting plant disturbances and always responding to them by displaying aggressive behaviours. This species does not share a host plant with other plant species (we only have one record of a Cephalotes ants on a plant occupied by D. bicolor). D. bicolor ants are exclusive patrollers of groups of plants near their nests, usually dominating patches consistently over at least a whole field season, indicating they are highly territorial. Although their location may change between years, their abundance seems constant over years. • Efficient recruiters with small response times, capable of recruiting over 30 ants in a few minutes. • A few go inside the flowers and once inside they patrol petals, walking around the inside of the corolla. These ants deter pollinators, reducing visit duration and preventing visitors from entering the flowers by

222

rapidly patrolling the inside of the petals (video from Troncones available). • Also seen walking on the outside of the corolla. Never seen at the bottom of the corolla consuming floral nectar. However, in Troncones, Guerrero, these ants used the holes bitten by the Curculionidae beetles on the corolla base to consume floral nectar from the outside of the corolla. • When extrafloral nectaries lack visible secretion droplets, they do not bite the glands, but rather explore/touch them with their antennae. • D. bicolor ants secrete high amount of some acid, probably formic acid, when stressed. This can be detected from its citric smell and sour taste. • Not biologist-friendly, these ants attack humans and their bite is painful. Nests easily identifiable through the sit-and-ouch! method.

Camponotus platanus • Active patrolling, climbing up from the stem and patrolling the upper and underside of the leaves. Very rarely seen inside the flowers, and when so, only seen walking on the petals, never seen consuming floral nectar. • Solitary ants, usually one per branch or apex, or even one per plant. Sometimes seen standing immobile. Poor recruiters, seldom recruiting more than two other individuals, which is probably due to their solitary patrolling habit. • Alert and very responsive towards plant disturbance, but do not attack biologists. Their attacking mode mixes bites and also spraying through their acidopore. These ants recoil the gaster, curving their abdomen forward, underneath their thorax, pointing the rear end with the acidopore towards the attack target. Although their main response is recoiling. • When they feel under attack, these ants head towards the underside of the leaves and drop themselves off the plant.

223

• When EFN lack visible secretion droplets, they bite the glands and contact them with their mouthparts open. • Often seen sharing a plant with their congeneric species: C. mucronatus and C. novogranadensis, as well as with Crematogaster, Paratrechina longicornis, Cephalotes, and P. gracilis.

Camponotus novogranadensis • They do not actively patrol the plant and are usually solitary ants (1-2 per plant). They remain (hide) amongst other gregarious ant species. Very agile and quick. When under threat (trying to capture them) they quickly dissociate from the other ants and escape. They are seen foraging nectar, but are inefficient patrollers, recruiters or attackers, which display evasive behaviours when plants are disturbed and hence considered as parasitic species. Their occupation patterns are infrequent and erratic, and are always seen sharing a host plant, suggesting no territoriality.

Camponotus mucronatus • These ants are not active patrollers and are usually seen standing immobile on the underside of petioles and leaf blades. • Not alert or responsive towards plant disturbance, whose main attacking mode seems to be spraying, although seldom seen recoiling the gaster and pointing the rear end with the acidopore towards the attack target. Do not attack biologists. • Solitary ants, usually one per branch or apex, or even one per plant. Very infrequent patrollers across sites and throughout years. Observed almost exclusively on site C, and occasionally site B. Usually located in a few areas per year, but highly unpredictable suggesting they are not territorial. • Not seen foraging on nectar and rarely found inside the flowers.

224

• When they feel under attack, these ants head towards the underside of the leaves and drop themselves off the plant. They display this behaviour much more often than C. planatus and so were considered very shy ants. • They often share a host plant with Brachymyrmex and C. planatus.

Brachymyrmex sp. • Highly aggressive ants actively patrolling through branches, they climb up the stem and look for nectar in EFN associated with buds, flowers, and fruits. Seen on both leaf blade sides (upper side and underside). • Gregarious patrolling, with several ants per branch or apex. They consume floral and extrafloral nectar. • Efficient recruiters with very small response times, capable of recruiting over 30 ants in <120 sec. Although they have an acidopore, these ants were rarely seen recoiling their abdomens against the larvae or other herbivores in the typical spraying behaviour. Their main attacking mode is biting and directly harassing larvae. Docile to biologists, no human attacks recorded. • They were observed inside the flowers deterring Apis mellifera: one ant was enough to prevent the bee from entering the flower, and causing it to overfly the flower in circles. When the ant detected the pollinator, it started patrolling towards the edge of the inner petal, and the bee stopped overflying and flew away (Xochitl Damián pers. obs. 19 Nov. 2014). • These ants are exclusive patrollers, and do not share a host plant with other plant species. During the direct conflict experiment when we planted ant corpses from different species inside flowers, In plants patrolled by Brachymyrmex ants, these ants often removed the experimental ant corpses from other species placed by us inside the flowers to test direct ant-pollinator conflict (Chapter 3). Consequently,

225

these ants seem to be highly territorial, since they can reliably by found patrolling the same plants over several days or weeks. • Although their location changes between years, their abundance seems constant over years.

Monomorium ebenium • These ants seem to be parasitic, nectar thieves as were never observed patrolling, recruiting or attacking. Usually observed directly on the nectaries – floral and extrafloral. Their abundance fluctuates greatly between years, but when abundant, they are commonly spotted inside flowers, at the base of the corolla, consuming floral nectar from the junction between petals where floral nectaries are located. • Small, gregarious ants, always seen in large groups (6-12 or more). Perhaps the ant with the smallest body size (<2 mm total length). • They do not interfere with pollinator visitation. Apis mellifera and other native bees will visit the flowers M. ebenimum ants inside. • Observed sharing a host plant or branch with Paratrechina longicornis and Brachymyrmex; but not with C. planatus.

Cephalotes • These ants are not active patrollers and are usually seen standing immobile on the leaf blades. • Not alert or responsive towards plant disturbance. Not aggressive, seen patrolling but rarely recruiting and never seen attacking or repelling any herbivore. They do not attack biologists. • Solitary ants, usually one per branch or apex, or even one per plant. Very infrequent patrollers across sites and throughout years. Usually located in a few areas per year, but highly unpredictable suggesting they are not territorial.

226

• Rarely seen foraging on extrafloral nectar and almost never found inside the flowers. • This ant species has the prototypical evasive behaviour: when detecting any sort of plant disturbance they will move away from it, without engaging with it at all. • When they feel under attack, these ants lower themselves onto the leaves, by slightly flattening themselves, presumably through bending their legs in order to get a better grip onto the leaf blade. And sometimes head towards the underside of the leaves, where they also ground themselves to the leaf blade. • Observed sharing a host plant with M. ebenium, Crematogaster sp., C. planatus, and D. bicolor. • On a personal note, these were by far my personal favourite ant species, and the same opinion was shared almost by consensus with my fellow colleagues from the LIPA Lab in Mexico. Nicknamed ‘tanquecito’ due to its robust appearance and metallic dark grey colour, resembling a wee war tank. Under the microscope these ants are also amazing! Honouring even more their nickname, with the exoskeleton covered in spines and their thick plates accentuated. I was utterly disappointed to find out that this charismatic species is a well-known parasitic genus in facultative ant-plant mutualisms (Byk and Del-Claro 2010). •

Crematogaster • Actively patrolling the apex of the plants, especially those bearing mature fruits. Rarely seen inside the flowers, but frequently observed harvesting extrafloral nectar and seeds. This species has been associated with seed dispersion, seen collecting the seeds as soon as the fruit dehisces, harvesting seeds and bringing them down to the ground, and removing the elaiosome enabling the seed to germinate. Gregarious patrolling, usually 5-10 ants per apex and many more (30+) swarming around a

227

newly dehisced fruit. They are very quick at detecting dehisced fruits, usually faster than biologists and fruits must be bagged or we are outran by these ants. • Their attacking mode mixes bites and also spraying through their acidopore. These ants recoil the gaster, curving their abdomen forward, underneath their thorax, pointing the rear end with the acidopore towards the attack target. • Their abundance fluctuates between years, but these ants are reliably found within the same area throughout weeks or a whole field season, indicating territoriality. • Observed sharing a host plant with Cephalotes ants.

Pseudomyrmex gracilis • Only seen in specific areas within a dune site (Parche C). These ants seem to be opportunistic patrollers and show a very erratic plant occupation pattern. Ants may be found patrolling a few neighbouring plants for two or three consecutive days at most, but then will not be found in those or other plants for several days. They tend to occupy a very low number of plants, in fact, rarely more than four. They are lone patrollers, one ant per plant. They are usually seen patrolling the flowers, swiftly dashing in and out of them, hunting for thrips or beetles (Staphylinoidea) inside the flowers. Not seen consuming floral or extrafloral nectar, not seen either deterring pollinators or any other visitor. We are unsure about their status as defenders, commensals or parasitic. Species within this genus, including this species, are often obligate mutualists of ant-plants, nesting within the plant and displaying very aggressive behaviours towards any plant visitor (including humans). However, they were never observed displaying aggressive behaviours on T. velutina and were always rather shy, escaping the plant after the least disturbance. They are very infrequent and have remained infrequent throughout years. • Often seen sharing a host plant with C. planatus.

228

Beetles

Beetles in T. velutina’s flowers at Troncones, Guerrero were observed usually in paired numbers ranging from 1 to 10 beetles, with a mean abundance of 3.18 ± 0.39 beetles per flower. Beetles were observed engaging in several behaviours: piercing the base of the corolla and stealing floral nectar, walking constantly around the inner petal surface, sometimes climbing up the reproductive organs (often the anthers), and mating inside the flowers. As they walked around, beetles left visible traces in the inner surface of the petals, especially when they walked whilst mating. These petal marks could be used in further studies to trace the presence and behaviour of beetles in flowers, even if they are not directly observed. Beetles spent very long periods inside the flowers with often the same individuals being there before our observation round started, and remaining there until the end of our observation round (range: 30 sec – 20 min). Preliminary analyses show a positive correlation between the numbers of beetles and the numbers of ants. This could imply that ants are either recruited to deter beetles, or that ants are attracted to flowers with beetles. Based on observations of ants using the corolla holes created by beetles to drink floral nectar from outside the corolla, I think there may be potential for a facilitation loop between ants and beetles, where beetles open the flowers and ants later use these pre-made holes to access floral nectar. Since ants were rarely observed inside flowers and were never observed consuming floral nectar, I hypothesise T. velutina flowers may have ant-repelling compounds to deter ants from inside the corolla, which also prevent ants from consuming floral nectar from inside the corolla. However, floral nectar does not seem to be unpalatable to ants, although further experiments outwidth the scope of this thesis are required to unravel these dynamics.

229

PUBLICATIONS ARISING FROM THIS THESIS

230 ALAMBIQUE

DUNAS COSTERAS: Las primeras colonizadoras de estos cordo- [Lo cual suena muy bien, pero…] UN ECOSISTEMA CODICIADO, nes son plantas rastreras, como las del géne- (5.9.b) Toda aquella infraestructura que se ro Ipomoea, al que pertenece el camote. Poco desee construir debe de ubicarse por lo menos SOSLAYADO Y AMENAZADO a poco, los tímidos listones de camotes reclu- 5 m por detrás del segundo cordón de dunas. En Nora Villamil Buenrostro tan otras plantas sin leño (herbáceas y pas- sitios donde la presencia de dunas no sea fácil- tos) y luego arbustos, hasta formar pastizales mente identificable, ya sea por su tamaño o por que recubren por completo la arena. Los ar- que hayan sido removidas, en caso de existir la bustos se reúnen en núcleos de matorrales que franja de vegetación permanente se utilizará se enriquecen con más y más especies, avan- ésta, para ubicar detrás la infraestructura de- zando tierra adentro hasta formar una selva seada. Las dunas costeras de Troncones, Guerrero, se encuen- baja conforme las dunas se alejan del mar. [Sin embargo, también señala que…] tran 284 km al oeste de Acapulco y albergan un amplio Hasta hace muy poco, el primero de marzo (5.4.g) No debe removerse la vegetación na- repertorio de insectos nativos que polinizan la planta de 2018 para ser precisos,2 no existía ley o re- tiva de la duna costera, y debe existir señaliza- Turnera velutina, una prima del sabroso maracuyá a la gulación alguna que protegiera estos ecosis- ción en la playa acerca de las medidas que se que comencé a estudiar durante la carrera de biología temas costeros en la reputada Ley General del realizan para su protección. en la UNAM. A pesar de su enorme valor biológico, Tron- Equilibrio Ecológico y la Protección al Am- (5.9.e) Los accesos a las playas a través de du- cones está en grave peligro: dos letreros que enmar- biente (LGEEPA). A pesar de ser un país con nas se deben realizar por medio de andadores can la región indican que “se vende”. 808,711 hectáreas de dunas costeras las úni- de madera que se construyan utilizando técni- Las dunas costeras son un ecosistema que, como su cas legislaciones que teníamos con respecto cas apropiadas que eviten su erosión y permi- nombre lo indica, se localiza a unos cuantos pasos del a los desarrollos en playas mexicanas eran seis tan el paso constante de los usuarios a la playa mar.1 Como el suelo está formado principalmente de criterios contradictorios en varios aspectos y sin el deterioro de la duna. arena, la lluvia se filtra con rapidez al subsuelo, lo que soslayados en una legislación menor, la Nor- (5.5.i) No se permitirá que ningún tipo de ve- reseca el terreno y hace del agua un recurso escaso, ma Oficial Mexicana (NOM). hículo circule o se estacione sobre la playa o aunque las dunas se ubiquen en regiones lluviosas y Al buscar la palabra duna en la norma NMX- sobre las dunas, a excepción de aquellos que tropicales. AA-120-SCFI-2006, que regula el certificado de prestan servicios públicos de limpia, vehículos Uno puede distinguir dónde comienza el ecosiste- calidad de las playas mexicanas, se encuen- de seguridad y aquellos de remolque de em- ma de las dunas costeras y termina la playa porque en tran sólo 11 menciones. Siete de ellas con- barcaciones. él la arena está sujeta por las raíces de la vegetación, forman los escuetos e incongruentes criterios mientras que las dunas de la playa son móviles. Gracias dentro del cuerpo de la norma, que no prote- En marzo de 2018 el Congreso finalmente a la cobertura vegetal, los cordones o crestas de las du- ge las dunas, sino que “establece los requisi- aprobó la petición para incluir en el artículo nas costeras se estabilizan y éstas dejan de ser trans- tos y especificaciones de sustentabilidad de 28 de la LGEEPA el concepto de “ecosistemas formadas por la acción del viento, a diferencia de los calidad de las playas”; algunos de ellos son costeros”, como los manglares, humedales, caprichosos montículos de arena donde los vacacionis- los siguientes: franjas intermareales, dunas costeras, lagu- tas se asolean. nas costeras, macroalgas, arrecifes de coral, (5.3.d) No debe existir infraestructura en las pastos marinos, fondos marinos o bentos y dunas costeras. las costas rocosas. Dado que estos espacios 1 M.L. Martínez, P. Moreno-Casasola, I. Espejel, O. Jiménez-Orocio, D. Infante figuran entre los más lucrativos para el desa- Mata y N. Rodríguez-Revelo, Diagnóstico de las dunas costeras de México, Conafor/Semarnat, 2014, 350 pp. rrollo turístico, es difícil creer que esta lagu- 2 'LDULR2ƼFLDOGHOD)HGHUDFLµQ, “Decreto por el que se reforman y na en la legislación sea un descuido produci- adicionan diversas disposiciones de la Ley General del Equilibrio ◀ Dunas costeras en Veracruz. Foto: Nora Villamil Buenrostro Ecológico y la Protección al Ambiente”, 23 de abril de 2018. do por la sobrecarga laboral de la Secretaría

124 ALAMBIQUE 125 DUNAS COSTERAS lenta aprobación de la Ley de Biodiversidad3 aprobada a principios de abril de 2018 con un raquítico quórum en el Congreso, muy por debajo del mínimo necesario; un puñado de legisladores negligentes y sordos ante las vo- ces de cientos de científicos, artistas, varias ONG y ciudadanos que se oponían. Fue una ley apoyada por el Partido Verde Ecologista de México (PVEM) en contubernio con el partido gobernante, el Partido Revolucionario Institu- cional (PRI). El PVEM es el único partido “ver- de” en el mundo que promueve actividades mineras e industriales en reservas y parques naturales, y a través de esta ley el PVEM está apostando por legalizar la explotación de re- servas naturales en un país megadiverso como lo es México. En Troncones, Guerrero, con los letreros

Pig Photo, Vista aérea, 2018 de “se vende” como límites al este y oeste, el mar al sur, y un sendero ya pavimentado al de Medio Ambiente y Recursos Naturales y de todo cuando deciden florecer gracias a la bue- trata de una cualidad destructiva. Este “fren- norte, yace uno de los últimos bastiones de los legisladores mismos. na mano del jardinero local, quien segura- te al mar” o “sobre la playa” significa, en tér- dunas costeras de la zona en los que aún no Así, tiempo después del trabajo de campo mente gana en un mes lo que el propietario minos ecológicos, estar ubicado en el ecosis- se ha edificado. Y así, mientras yo ordeno las en Troncones y mientras escribo mis hallaz- cobra a un vacacionista por noche. Tras la bar- tema de las dunas costeras estabilizadas, un bases de datos obtenidos en Troncones y bus- gos sobre las dinámicas ecológicas de este da tal vez habrá una enorme alberca rodeada hábitat soslayado y gravemente amenazado co el modelo estadístico más adecuado y la sitio, me asalta una inmensa preocupación: por un patio con losetas de barro anaranjado a nivel mundial y en particular en México. gráfica más puntual para reflejar las interac- ¿Qué tal si ese terreno arenoso finalmente y cómodos camastros blancos para asolear- La mercadotecnia y la cultura vacacional ciones biológicas que ahí suceden, el reloj de se vende? ¿Qué tal si alguien construye allí se. Sobre la barra de la cantina penderán bo- del capitalismo nos hacen fantasear con unas arena que hay sobre mi escritorio acusa el paso otra casa de vacaciones, una más en la larga cabajo copas para margaritas, cuyos pétalos vacaciones de playa en una costa tropical de amenazador del tiempo para las dunas cos- fila de casas de playa en renta a lo largo de la de vidrio soplado apuntarán, infértiles, a las arenas finas y mares tibios. Sin importar qué teras. Así es investigar en un terreno areno- costa? Casas no habitadas, sino rentadas u losetas de barro. tan modestas, lujosas, ecoturísticas, ecológicas so, viéndolo escurrirse, grano tras grano, en- ocupadas tan sólo unas cuantas semanas al Esta propiedad se subirá al portal de Airbnb o derrochadoras, “fresas”, corrientes o “hips- tre mis dedos, mientras cuento hojas, bichos año. ¿Qué tal si para cuando vuelva a realizar bajo el seductor encabezado: “Casa Joya del ters” sean las instalaciones del hospedaje que y dinámicas que tienen fecha de expiración. mi ensoñado siguiente experimento, en lu- Mar: Beautiful new beachfront villa” para atraer anhelemos, todas ellas desplazan los ecosis- ¿Serena y contando? ¡Imposible serenarse! gar de este reducto lleno de árboles, panales, turismo internacional, o quizá “La Mariposa: temas costeros, amenazados de antemano por plantas, escarabajos, hormigas y mariposas, lujosa villa sobre la playa”. Todos estos nom- la ya descrita falta de regulación y protec- encontrara yo una barda? Una barda impe- bres, que compiten por ser los más singulares ción legal. 3 Gloria Leticia Díaz, “Activistas y académicos alertan sobre cablemente alineada con cactus, sábilas y ma- y atractivos, son extremadamente similares y Por otro lado, estas amenazas se han acre- privatización de los recursos naturales con Ley de Biodiversidad”, gueyes salpicados por aquí y por allá. Una bar- comparten un elemento clave: la oferta de centado recientemente, multiplicándose cual Proceso, 2 de abril de 2018. Aristegui Noticias, 3 de abril de 2018. https://aristeguinoticias.com/0304/mexico/falta-pulir-ley-de- da de plantas que te dejan sin aliento, sobre una locación frente al mar, sobre la playa. Se bacterias de un día para otro, tras la trucu- biodiversidad-atenta-contra-derechos-de-pueblos-indigenas-prd/

ALAMBIQUE 126 DUNAS COSTERAS ALAMBIQUE 127 DUNAS COSTERAS Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs ORIGINAL RESEARCH published: 14 August 2018 doi: 10.3389/fpls.2018.01093 infestation on plant tissues (Bentley, 1977; Rosumek et al., 2009). Malé et al., 2012). Such castrating behavior inevitably reduces This mutualistic interaction is termed myrmecophily. availability of flowers and hence floral rewards for pollinators. Interactions involving two or more types of mutualists of a It has been suggested that the ultimate cause of castration by single host are common in nature, but multispecies interactions patrolling ants is promoting reallocation of plant resources from are much less studied than pairwise and intraguild mutualisms reproduction to growth (Yu and Pierce, 1998; Frederickson, (Strauss, 1997; Tscharntke and Hawkins, 2002; Strauss and 2009; Malé et al., 2012), and hence increases the availability of Ant-Pollinator Conflict Results in Irwin, 2004; Adler, 2008; Melián et al., 2009; Koptur et al., resources on which ant colonies depend. In ant species that are 2015). To date, most research on plant-animal interactions has obligate inhabitants of ant-plants, colony size is limited by the Pollinator Deterrence but no Nectar focused on pairwise relationships (e.g., plant-herbivore, plant- number of domatia (Fonseca, 1993, 1999; Orivel et al., 2011), pollinator, plant-fungus) in isolation from the community in which is positively correlated with plant investment in growth. Trade-Offs which they are embedded (Strauss, 1997; Herrera, 2000; Dáttilo Consequently, resource allocation strategies toward these two et al., 2016; Del-Claro et al., 2018). This pairwise approach mutualists should be approached in a linked way because plant Nora Villamil 1*, Karina Boege 2 and Graham N. Stone 1 necessarily oversimplifies reality (Herrera, 2000) since plants investment toward growth may come at the cost of investment to interact sequentially or simultaneously with each of pollinators, reproduction, and vice versa. And so, plant investment in rewards 1 Ashworth Laboratories, Institute of Evolutionary Biology, University of Edinburgh, Edinburgh, United Kingdom, 2 Instituto de herbivores, herbivore predators and pathogens (Armbruster, for each mutualist reward may be affected, either positively Ecología, Departamento de Ecología Evolutiva, Universidad Nacional Autónoma de México, Mexico City, Mexico 1997). Furthermore, plant interactions with one partner or via linkage, or negatively via trade-offs. Furthermore, because guild can also affect relationships with other groups or guilds plants interact with both mutualists simultaneously the presence Direct and indirect negative interactions between ant guards and pollinators on ant-plants (Armbruster, 1997) and alter outcomes from mutualistic to of one mutualist may increase or decrease presence of the are expected for two reasons. First, aggressive ants may deter pollinators directly. antagonistic (Strauss, 1997; Strauss et al., 1999; Herrera, 2000; other. Strauss and Irwin, 2004; Del-Claro et al., 2016). As a result, a Even in those species that do not castrate their host, Second, pollinators benefit from plant investment in reproduction whilst ants benefit growing number of studies are focusing on multispecies and ants’ aggressive behaviors might threaten and deter pollinators, from plant investment in indirect defense, and resource allocation trade-offs between multitrophic interactions (Melián et al., 2009; Fontaine et al., compromising plant reproduction (Ness, 2006; Assunção et al., these functions could lead to indirect conflict. We explored the potential for ant-pollinator 2011; Nahas et al., 2012; Pineda et al., 2013; Dáttilo et al., 2014; LeVan et al., 2014). Avoidance of such direct conflict has conflict in a Mexican myrmecophile, Turnera velutina, which rewards ants with extrafloral 2016). It might be expected, for example, that the presence of been suggested to explain plant architecture or behaviors that predatory ants can influence pollinators, with top-down effects reduce spatial (Raine et al., 2002; Malé et al., 2015; Martínez- nectar and pollinators with floral nectar. We characterized the daily timing of ant and on plant fitness. This makes ant-plants, which rely on ants for Bauer et al., 2015) or temporal overlap of ant guards and pollinator activity on the plant and used experiments to test for direct and indirect conflict defense against herbivores and on pollinators for seed set, a pollinators (Gaume and Mckey, 1999; Gaume et al., 2005; Nicklen between these two groups of mutualists. We tested for direct conflict by quantifying model tritrophic system in which to explore the dynamics of and Wagner, 2006; Ohm and Miller, 2014; Malé et al., 2015), pollinator responses to flowers containing dead specimens of aggressive ant species, multispecies and multitrophic interactions. and the presence of ant-repelling compounds which exclude ants relative to unoccupied control flowers. We assessed indirect conflict by testing for the Here, we focus on disentangling ant-pollinator interactions from flowers when pollen is released (Junker et al., 2007; Willmer that occur when both mutualists share a host plant. Previous et al., 2009; Ballantyne and Willmer, 2012). existence of a trade-off in sugar allocation between ant and pollinator rewards, evidenced Edited by: work has revealed evidence of ant-pollinator conflict in such Extrafloral nectar is a key resource mediating multispecies Massimo Nepi, by an increase in floral nectar secretion when extrafloral nectar secretion was prevented. systems (Yu and Pierce, 1998; Stanton et al., 1999; Gaume interactions in many plant communities, and plants bearing University of Siena, Italy Secretion of floral and extrafloral nectar, activity of ants and pollinators, and pollen et al., 2005; Ness, 2006; Palmer and Brody, 2007; Frederickson, extrafloral nectaries comprise up to a third of species in Reviewed by: deposition all overlapped in daily time and peaked within the first 2 h after flowers opened. 2009; Stanton and Palmer, 2011; Malé et al., 2012; Assunção some biomes (Dyer and Phyllis, 2002; Rudgers and Gardener, Kleber Del-Claro, We found evidence of direct conflict, in that presence of ants inside the flowers altered et al., 2014; LeVan et al., 2014; Hanna et al., 2015). Ant- 2004; Davidson and Cook, 2008; Dyer, 2008), particularly in Federal University of Uberlandia, Brazil pollinator conflict is expected for two main reasons. Firstly, tropical dry forests, savannas and cerrados (Rico-Gray and Rubén Torices, pollinator behavior and reduced visit duration, although visit frequency was unchanged. both mutualists share with their host interest in different plant Oliveira, 2007; Assunção et al., 2014). The importance of Estacion Experimental de Zonas We found no evidence for indirect conflict, with no significant difference in the volume Aridas (EEZA), Spain functions. Pollinators benefit from plant resource allocation to ant-plants as food resources for mutualists in a given plant reproduction (i.e., flowers, floral nectar (FN) and pollen), whilst community is enhanced if these plants also secrete FN and *Correspondence: or sugar content of floral nectar between control plants and those in which extrafloral ants benefit from allocation to growth and defense [i.e., vegetative pollen for pollinators. Management of ant-pollinator conflict Nora Villamil nectar secretion was prevented. The presence of ants in flowers alters pollinator behavior structures bearing extrafloral nectar (EFN) or domatia] (Yu in such a way that the crucial services provided by both [email protected] in ways that are likely to affect pollination dynamics, though there is no apparent trade- and Pierce, 1998; Frederickson, 2009; Palmer et al., 2010). This mutualist groups are maintained is thus likely to be part off between plant investment in nectar rewards for pollinators and ant guards. Further Specialty section: could result in a conflict mediated by plant rewards, known as of the adaptive landscape of many plant species. Ant-plants This article was submitted to studies are required to quantify the effect of the natural abundance of ants in flowers on indirect conflict. Floral and extrafloral nectar share sugar as a with extrafloral and floral nectaries represent an excellent Functional Plant Ecology, pollinator behavior, and any associated impacts on plant reproductive success. common currency, providing potential for a trade-off and also system in which to test for trade-offs in resource allocation, a section of the journal a means of quantifying investment in each. Secondly, because competition amongst mutualistic guilds, and assess whether plant Frontiers in Plant Science Keywords: Turnera velutina, ant-plant, floral nectar, extrafloral nectar, resource allocation, indirect interactions, ant guards actively defend their host plant as a means of strategies minimize direct and indirect conflicts between their Myrmecophily Received: 20 April 2018 protecting food and/or nesting sites, they may also repel or attack mutualists. To our knowledge, no study has addressed both Accepted: 05 July 2018 pollinators (Ness, 2006; Stanton and Palmer, 2011; Chamberlain direct and indirect ant-pollinator conflict in a single ant-plant Published: 14 August 2018 INTRODUCTION and Rudgers, 2012), and this drawback of ant guards is known as system. Citation: direct conflict. Here we tested for direct and indirect ant-pollinator conflict Villamil N, Boege K and Stone GN Castration is an extreme example of direct ant-pollinator on a Mexican endemic ant-plant, Turnera velutina. In particular, (2018) Ant-Pollinator Conflict Results Extrafloral nectaries, domatia, and food bodies are all means by which ant-plants (comprising myrmecophiles and myrmecophytes) (Rosumek et al., 2009; Del-Claro et al., 2016) attract and conflict in which guarding ants destroy or consume the we assessed whether Turnera velutina reduces the potential in Pollinator Deterrence but no Nectar reproductive meristems, floral buds or flowers of their host for conflict through the daily timing of FN and EFN release. Trade-Offs. Front. Plant Sci. 9:1093. support ants by providing nesting sites or nutrients (Rico-Gray and Oliveira, 2007; Rosumek et al., doi: 10.3389/fpls.2018.01093 2009). In return, ants attack herbivores, prune climbing vines and prevent fungal and microbial plant (Yu and Pierce, 1998; Stanton et al., 1999; Gaume We also tested for potential indirect (nectar-mediated) and et al., 2005; Palmer and Brody, 2007; Frederickson, 2009; direct (deterrence) conflicts between ants and pollinators. We

Frontiers in Plant Science | www.frontiersin.org 1 August 2018 | Volume 9 | Article 1093 Frontiers in Plant Science | www.frontiersin.org 2 August 2018 | Volume 9 | Article 1093 Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs addressed the following specific questions: (i) What are the in a single flower (Minicaps Disposable capillaries, Hirschmann vapors from influencing pollinator behavior. Flowers containing daily timings of nectar reward secretion, ant activity, and Laborgerate, In 20◦C, ISO 7550, R<0.5%, CV<1.0%, Germany). dead ants were observed for 20 min, recording pollinator identity, floral visitation? (ii) Does the presence of patrolling ants deter EFN was also collected from the glands using 1 μl capillaries. visit frequency and duration, and associated pollinator behaviors. pollinators from the flowers? (iii) Do ant species vary in their Nectar volume was measured using a digital caliper (Mitutoyo Observations were conducted during October-November 2016 deterrence for pollinators? (iv) Are T. velutina plants able to re- Digimatic) and sugar concentration was determined in ◦Brix with four simultaneous observers collecting data from different allocate extrafloral nectar resources into floral nectar resources (g sucrose per 100 g solution) using a 0–50◦Bhandheld plants. (increasing reward availability to flower visitors)? refractometer (Reichert, Munich, Germany). To obtain all of the When assessing the effects of ants inside the flowers on sugar from extrafloral nectaries, the glands were washed with 2 pollinator visitation we only considered visits by Apis mellifera, μl of deionized water using a 0–5 μl micropipette and the sugar since this is the dominant T. velutina pollinator in this MATERIALS AND METHODS concentration of the wash was quantified using the refractometer. population (Sosenski et al., 2016) and accounted for 91% of Study Site and System Total sugar content in FN and EFN was calculated from volume the visits in this experiment. We classified honeybee behaviors and ◦Brix values according to Comba et al. (1999) using the as “inspection” or “contact.” Inspection behaviors comprised Field experiments were conducted in coastal sand dunes at the formula: either approaching or hovering over a particular flower without CICOLMA Field Station in La Mancha, Veracruz, in the Gulf ◦  ◦  landing. Contact behaviors were those that occurred inside the of Mexico (19 36 N, 96 22 W, elevation < 100 m). The climate s = dvC/100 flower, between landing and take-off, and comprised foraging on is warm sub-humid, with a rainy season during the summer pollen or nectar resources or standing on the petals, anthers or (June to September), an annual precipitation of 1,100–1,500 mm, ◦ μ stigmas. and a mean annual temperature ranging between 24 and 26 C where s is the sugar content ( g), v isthenectarorwashvolume μ (Travieso-Bello and Campos, 2006). Experiments were carried ( l), and d is the density of a sucrose solution at a concentration Indirect Conflict out in November 2014 at four sites with high densities of C (g of sucrose per 100 g solution) as read on the refractometer. To test possible trade-offs in plant resources between FN and Turnera velutina (Passifloraceae). Greenhouse experiments were The density was obtained according to Comba et al. (1999) using EFN we conducted a greenhouse experiment on 72 plants conducted in a shade house at CICOLMA. the formula: from 18 different maternal lines (2–4 siblings per maternal Turnera velutina is an endemic perennial shrub (Arbo, = 2 + + family, generated from field individuals). Plants were kept 2005) and myrmecophile (Cuautle and Rico-Gray, 2003). At La d 0.0000178C 0.003791C 0.9988603 under greenhouse conditions in a shade house located within Mancha, T. velutina is patrolled by at least seven ant species CICOLMA field station. Plants were grown in 1L plastic pots (Cuautle et al., 2005; Zedillo-Avelleyra, 2017) and its main A different flower from each of these twenty plants was chosen with a substrate of local soil and vermiculite (50:50) and watered herbivores are the specialist caterpillars of the butterfly Euptoieta and marked, but never bagged. We collected one stigma every every other day (for rearing details see: Ochoa-López et al., hegesia (Nymphalidae). Extrafloral nectar is provided in paired hour, to sample the pollen deposited 1 h, 2 h, and 3 h post- 2015). During the experiment, plants were watered every night, cup-shaped glands located at the bottom of the leaf blade at the anthesis. A small proportion of flowers contained a fourth stigma, and extrafloral nectaries were sprayed with water to wash away junction with the leaf petiole, on the underside of the leaves contributing to a small dataset for 4 h post anthesis. Stigmas any nectar secretion from previous days and to prevent fungal (Figure 1A, Elias et al., 1975; Villamil et al., 2013). Although were stored individually in labeled Eppendorf tubes. Stigmas infections on the glands. Pre-anthesis buds were bagged either the it flowers year-round, flowering peaks during summer (Cuautle were individually mounted on slides for fresh squash glycerin night before or during the morning before anthesis to prevent any et al., 2005). Flowers last one day, are animal-pollinated (Sosenski preparations. Slides were sealed using nail polish and kept in a ◦ nectar theft by unexpected insects that may occasionally enter the et al., 2016) and provide FN at the base of the corolla (Figure 1). fresh and dry environment at 22 C. Each slide was labeled with FIGURE 1 | Flower of Turnera velutina (top) with arrows indicating the location shade house. Honeybees (Apis mellifera) are the dominant pollinators at La the site, day, hour, and plant identity. The number of pollen of the floral nectaries at the base of the corolla, between the petals, and leaves Pairs of maternal siblings were chosen and randomly assigned Mancha, accounting for 94% of the visits (Sosenski et al., 2016) grains on each stigma was counted under a microscope, and (bottom) with arrows indicating the location of extrafloral nectaries, on the to either control or experimental groups. The extrafloral and collect both pollen and floral nectar, but the role of other underside of the leaves. changes in numbers at each time interval plotted to generate the nectaries in all leaves of experimental plants were clogged floral visitors is yet to be investigated as effective pollinators. pollen deposition curve. by applying Mylin transparent textile paint in the nectary There is no spatial segregation of patrolling ants and floral visitors Direct Conflict cup. Extrafloral nectaries from control plants were left intact in Turnera velutina since flower buds emerge from the axillary incorporated site-and-day effects into our statistical modeling To test whether non-ant flower visitors detect and avoid flowers (unclogged) and droplets of the same paint were applied on meristems of leaves bearing extrafloral nectaries (Villamil, 2017). (see below). with ants, visitation was observed in flowers with and without the leaf blade above the glands to match any unintended Furthermore, EFN volume and sugar content are higher at the dead ants for 4 flowers on each of 40 plants (n = 160 flowers). effects on plants across treatments. Very young floral buds flower stage than for either buds or fruits (Villamil, 2017). Nectar Secretion and Pollen Deposition Curve Plants > 80 cm in height were haphazardly selected and flowers were marked, and their extrafloral nectaries clogged from their Nectar secretion and pollen deposition data were collected from bagged at 0700 before anthesis. Once the corollas had opened, emergence, throughout their development, and until anthesis. Fieldwork Methods 4 sites within CICOLMA over 5 consecutive days in September three ant corpses from a single ant species (either Dorymyrmex The droplets on the extrafloral nectaries or leaf blades were Mutualist Activity Curves: Patrolling Ants and 2015. We visited a single site per day (with one exception which bicolor, Brachymyrmex sp., or Paratrechina longicornis)were checked daily and replenished when required, especially when Pollinators was visited twice) and sampled FN and EFN secretion rate and placed inside each of three flowers/plant. A fourth flower was a new leaf emerged in order to guarantee uniform and We quantified daily activity of patrolling ants and flower visitors pollen deposition from 1 flower per plant for 5 plants per site left as an ant-free control. Ant corpses were placed on the inner continuous application of the clogging treatment across all on T. velutina by surveying flowers (n = 120 plants, n = 1,604 (n = 20 plants). Flowers were sampled every hour during the surface of the petals in each flower. These species were chosen leaves. flowers) and their associated leaves at all four La Mancha sites in anthesis period (0800–1300 h). Flowers were bagged with tulle because they were amongst the most abundant species (see FN secretion was measured before the clogging treatment was November 2014. We observed all open flowers for 2 min every bags before anthesis and FN and EFN were collected every Results), displayed patrolling behaviors on T. velutina,andwere applied, and once again after the young clogged/marked buds hour throughout anthesis (0800–1300 h), with one observer at hour during anthesis. The first collection was taken as soon detected as potentially differing in their effects on Apis mellifera became flowers. The aim was to test whether flowers that were each site. Every hour, we counted the total number of floral as the corolla was fully open. Flowers were re-bagged between pollinators by a previous study (Villamil et al. unpublished data). unable to secrete EFN invested more sugar resources in floral visitors and ants patrolling extrafloral nectaries across all flowers measurements to avoid nectar consumption and a masking tape Additional flowers were removed from the plant to standarise nectar. FN was collected from control and treatment flowers within a site. We sampled the same sites over multiple days. band with Tanglefoot was applied on the stem below the flower to floral display across all individuals. Ants were killed in 50% between 1300 and 1500 h using a 1 μl microcapillary pipette. Since these are one-day flowers, we considered each site-day as exclude ants for the duration of the experiment. FN was extracted ethanol, which was allowed to evaporate for 30 min at ambient Nectar volume and total sugar mass was estimated and calculated areplicate(n = 10 site-days; 43.23 ± 2.89 flowers/site-day), and using a 1 μlcapillaryinsertedintoeachofthefivenectaries temperature (28–35◦C) from all samples to prevent ethanol as described above.

Frontiers in Plant Science | www.frontiersin.org 3 August 2018 | Volume 9 | Article 1093 Frontiers in Plant Science | www.frontiersin.org 4 August 2018 | Volume 9 | Article 1093 Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs

Statistical Methods the flower (Control, Dorymyrmex bicolor, Brachymyrmex sp., or All statistical analyses were conducted in R version 3.23 (R Paratrechina longicornis) was fitted as a fixed effect. Because Core Team 2016). All mixed effects models were fitted using these are 1 day flowers, plant-day identity was chosen as a the ‘multcomp’ R package (Bates et al., 2016)andpost hoc random effect to control for individual variation in floral and Tukey comparisons were fitted using the ‘multcomp’ R package extrafloral nectar investment. We also included an observation- (Hothorn et al., 2008), unless stated otherwise. level random effect. Post hoc Tukey comparisons were used to test differences in visit duration between the four treatments. Mutualist Activity and Reward Secretion Curves We tested whether ant species inside the flower differed in To test whether ant or pollinator activity changed over daily their effect on the likelihood with which a pollinator displayed time, we fitted a Poisson mixed model with either the number an inspection behavior using a binomial mixed model. The of ants patrolling or the number of floral visitors as the response presence or absence of inspection behaviors was coded as the variable. We fitted time of day as a fixed effect, with linear and response variable and ant species was fitted as a fixed effect. quadratic terms to detect non-linear activity patterns over time. As random effects we fitted the plant-day identity, and the The number of flowers per site was fitted as a log-transformed visitor identity. For those visits where the pollinator displayed offset to control for floral display, since we recorded visitor an inspection behavior, we fitted the proportion of time per visit counts per site rather than counts per individual flower (see spent displaying inspection behaviors using a Gaussian mixed fieldwork methods). Flowers of T. velutina last for a single day, model with logit transformation for data normality. Ant species and because multiple flowers were sampled on a given site on a was included as a fixed effect, and plant-day identity, and the given day, we fitted site identity as a random effect to account visitor identity were fitted as random effects. FIGURE 2 | Timing of (A) reward secretion (n = 20 flowers-leaves), (B) mutualist activity (n = 1,604 flowers; n = 10 sites/day) and (C) pollen deposition (n = 20 for differences between site and day variation in variables that Finally, differences in the duration of inspection or contact flowers) in Turnera velutina during the anthesis period (08:30–12:30), showing raw data from field observations (mean ± se). Red circles show floral nectar and the activity of pollinators in flowers, whilst green triangles show extrafloral nectar and the activity of ants at extrafloral nectaries. could influence ant abundance, such as resource availability, behaviors in flowers containing different ant species were ant diversity, or the abundance/proximity of ant nests. We also analyzed using Gamma mixed models. Mixed effects models were included an observation-level random effect (OLRE) where each fitted independently for each behavior, but using the same model TABLE 1 | Model statistics for the timing of mutualists activity, reward secretion and pollen deposition in Turnera velutina. data point receives a unique level of a random effect to control structure fitting visit duration per flower as the response variable. P for overdispersion (Hinde, 1982). Response Fixed effects N Estimate LRT -value Random Variance SD Maxima estimates Ant species inside the flower was fitted as a fixed effect. Plant-day effects To test if FN and EFN secretion changed over the anthesis identity was chosen as a random effect to control for individual period we fitted a Poisson mixed model independently for each variation in floral and extrafloral investment, and daily weather Peak mpa Time of μ hour day nectar type, with sugar mass ( g) as the response variable. Nectar variations. We also included an observation-level random effect. sugar content is usually estimated in μg, and we report our raw Post hoc Tukey comparisons were used to test differences in visit Mutualist activity Patrolling ants Log(flowers) 33 0.804 7.967 0.004 *** OLRE 0.221 0.471 1.5 90 09:30 data in such units. However, to facilitate model convergence duration between the four treatments. (number) Site ID 0.462 0.680 μ we re-scaled our response variable (sugar content) from gto Hour 33 0.088 0.148 0.699 g, and rounded it to the next integer to better fit a Poisson Indirect Conflict Hour2 33 −0.028 0.201 0.653 distribution. We fitted time of day as a fixed effect, with linear For each plant, we estimated the difference in FN produced Floral visitors Log(flowers) 42 0.507 2.619 0.105 OLRE 0.097 0.311 0.15 9 08:09 and quadratic terms to detect non-linear activity patterns over before and after the extrafloral nectaries were clogged as follows: (number) Site ID 0.991 0.991 time. We fitted plant identity as a random effect, and, included Hour 0.040 0.036 0.849 DFN = PostFN − PreFN,whereD-FN is the difference, PostFN is an observation-level random effect. the FN production after the extrafloral nectaries secretion was Hour2 −0.131 3.838 0.05 * Reward secretion Floral sugar Hour 78 0.1445 0.1019 0.7495 OLRE 0 0 0 0 0.39 23 08:23 prevented and PreFN is the FN production before the extrafloral Timing of Daily Activity and Secretion Peaks μ nectaries were clogged. Differences in volume and sugar content ( g) Plant ID We computed the time at which mutualist activity and nectar Hour2 78 −0.1818 1.3222 0.2502 secretion reached their maximum by calculating the time at between control and experimental flowers were tested using mixed effects models. Both variables had normal distributions Extrafloral sugar Hour 78 0.4647 1.0070 03156 OLRE 0.2425 0.4925 1.13 68 09:08 which the slope (i.e., the differential of the fitted model with (μg) Plant ID 0.4013 0.6335 and so Gaussian mixed effects models were fitted using the same respect to time) for each variable is zero and then solving for hour, Hour2 78 −0.2053 1.8949 0.1686 model structure: The clogging treatment was fitted as a fixed as follows: Pollen deposition Hour 44 0.3240 21.38 3.73−06 *** OLRE 0.1049 0.3240 effect, and the maternal family was fitted as a random effect to (number of grains) Site ID 0.0957 0.3094 = β ∗ + β ∗ 2 independently explain variation in both nectar volume and sugar y hour hour hour2 hour content. Estimates at which the maxima for mutualist activities and reward production are reached are shown in the last three columns. Peak hour is the time in hours estimated, mpa shows dy minutes post-anthesis when the maxima is reached, and time of day indicates an approximation of when that activity is likely to occur, although this varies depending on the season = β + 2β 2 ∗hour hour hour and the time of sunrise. *P < 0.05; ***P < 0.001. dhour = β + β ∗ 0 hour 2 hour2 hour RESULTS − β hour = On average, a flower and its associated leaf secreted a total by non-destructive sampling of the same flowers over the β hour Mutualist Activity, Reward Secretion, and ± μ 2 hour2 of 2,815 767 g of sugar via floral and extrafloral nectar entire anthesis period (Figure 2A). FN and EFN are secreted β throughout the 4.5 h anthesis period. The total sugar content in simultaneously during anthesis and with the highest amount =−1∗ hour Pollen Deposition Curves hour β FN was 149 ± 19.3 μg of sugar, whilst total extrafloral sugar was of sugar content secreted during the first 2 h of anthesis 2 hour2 Activity curves show that both patrolling ants and floral visitors were most active within the first 2 h post-anthesis (Figure 2), 2,665 ± 765 μg. Thus, the relative sugar contributions of floral (Figure 2A), although their secretion peaks predicted from Direct Conflict although the visitation peak by potential pollinators predicted and extrafloral nectar in a leaf-flower module were 5.3 and 94.7%, model estimates are slightly mismatched (Table 1). EFN secretion The effect of different ant species on the visitation frequency from model estimates was on average over an hour before the respectively. peaked 68 min after our first collection, which was taken as soon was tested using Poisson mixed effects models. We fitted the predicted peak for ant activity (9 min post-anthesis for potential Floral and extrafloral nectaries of T. velutina are both as flowers were fully open, whilst FN peaked 23 min after the first number of visitors as the response variable, and ant species inside pollinators, 90 min post-anthesis for ant patrolling; Table 1). able to quickly replenish nectar after experimental removal collection. The timing of peaks in secretion of the two types of

Frontiers in Plant Science | www.frontiersin.org 5 August 2018 | Volume 9 | Article 1093 Frontiers in Plant Science | www.frontiersin.org 6 August 2018 | Volume 9 | Article 1093 Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs nectar matches that for the mutualists that harvest each resource. Indirect Conflict Peak pollen deposition occurred at the beginning of anthesis and Clogging extrafloral nectaries on the leaves associated with steadily declined over time (Figure 2C). Hence, pollen deposition newly emerged floral buds had no effect on their FN volume data were analyzed using a linear model without fitting hour as a [LRT (1,49) = 0.21; P = 0.64, Table 3, Figure 4B]orsugar squared term and we did not estimate timing of daily maxima content [LRT (1,49) = 0.087; P = 0.77, Table 3, Figure 4A]. using model derivations (Table 1). Differences in FN volume and sugar content (FNpost−treatment − We recorded 1,535 ant visitors from nine ant species FNpre−treatment) were positive in plants under both control and patrolling extrafloral nectaries of T. velutina at CICOLMA clogged treatments (Figure 4). (Table 2). Dorymyrmex bicolor, Paratrechina longicornis,and Brachymyrmex spp. accounted for 68.5% of the total ants observed, and 77.35% of the patrolling ants, after excluding DISCUSSION Monomorium spp. that were never observed displaying patrolling behaviors on T. velutina and are mostly parasitic consumers Our findings show that both ants and pollinators are active of FN and EFN (lestobiotic) (Ettershank, 1966; Bolton, while flowers are open (Figure 2B), that FN and EFN 1987). are simultaneously secreted (Figure 2A), and that pollen deposition occurs when ants are actively patrolling (Figure 2C). Consequently, in T. velutina guarding ants and pollinators Direct Conflict operate in close spatial and temporal proximity, implying that We recorded 991 floral visitors, of which 907 (91.5%) were by the direct and indirect conflict could occur in this system. We honeybee Apis mellifera. Of the remaining 8.5%, 61 visits were found, however, no evidence for indirect, nectar-mediated, by native bees, 11 by flies (Diptera), 10 visits by Lepidoptera, one conflict between ants and pollinators, since plants did not visit by a beetle (Coleoptera) and one by a wasp (Hymenoptera) reallocate resources toward floral nectar, even when 95% of (Table 2). the sugar investment during anthesis, which is in EFN, is Neither the presence of ants inside flowers nor their prevented (Figure 2). We found evidence for direct conflict, identity had any significant effect on the number of honeybees as the presence of dead individuals of patrolling ant species visiting the flowers (Figure 3A, Table 3). The presence of the inside flowers was associated with both higher frequency most aggressive ant species, Dorymyrmex bicolor, increased the of inspection behaviors in potential pollinators, and reduced likelihood of a pollinator displaying inspection behaviors by 20% visit duration (time spent inside flowers) (Figure 3). Taken (Figure 3B), and increased by 12% the proportion of time per together, these effects increased handling time per flower and visit spent displaying inspection behaviors rather than foraging reduced pollinator foraging efficiency. Nonetheless, this result or pollinating the flower (Figure 3C). Finally, the presence of was obtained under an experimental setting using dead ants Dorymyrmex bicolor and Paratrechina longicornis inside the placed inside flowers. Further studies are required to test (i) flowers halved the duration of contact visits compared to control the effect of living patrolling ants on pollinator visitation, flowers without ants, or to flowers with Brachymyrmex sp. ants and (ii) the impact of any such effects on plant fitness. The inside. latter are crucial to understanding whether there is ongoing

TABLE 2 | Abundance and identity of the ants recorded patrolling extrafloral nectaries during the 2014 census and the floral visitors recorded during the direct conflict experiment in 2016 on Turnera velutina plants.

Taxon Visitors Percentage Subfamily Patrolling FIGURE 3 | The effect of dead ants inside the flowers of Turnera velutina on different aspects of the behavior of visiting honeybees: (A) visitation frequency, (B) the percentage of visitors of displaying inspection behaviors, (C) the duration of contact visits (time spent inside the flower), and (D) proportion of time spent Ants at extrafloral nectaries Paratrechina longicornis 487 31.72 Formicinae Gregarious inspecting the flowers per visit bout (hovering over the floral head space). Ant species are arranged in order of increasing aggressivity and names are abbreviated: C, Dorymyrmex bicolor 421 27.42 Dolichoderinae Gregarious control with no ants; 1P, Paratrechina longicornis;2M,Brachymyrmex sp.; 3D, Dorymyrmex bicolor ants. (n = 40 plants; n = 160 flowers). Monomorium ebenium 270 17.58 Myrmicinae Gregarious Camponotus planatus 184 11.98 Formicinae Loner Brachymyrmex sp. 73 4.75 Formicinae Gregarious selection on Turnera velutina to manage direct ant-pollinator secretion after consumption (Pacini et al., 2003; Pacini and Camponotus mucronatus 60 3.90 Formicinae Loner conflict. Nepi, 2007; Pacini and Nicolson, 2007; Escalante-Pérez and Crematogaster sp. 23 1.49 Myrmicinae Gregarious Heil, 2012; Orona-Tamayo et al., 2013; Heil, 2015). In contrast, Camponotus novogranadensis 15 0.97 Formicinae Loner Mutualist Activity, Reward Secretion and species vary in whether FN is replenished or not (for details on Pseudomyrmex gracilis 2 0.13 Pseudomyrmicinae Loner and very rare Pollen Deposition floral nectar dynamics see: Pacini et al., 2003; Nicolson et al., Floral visitors Apis mellifera 907 91.5 Floral and extrafloral nectar were secreted simultaneously and 2007; Willmer, 2011). We suggest that rapid resupply is crucial Native bees 61 6.05 rapidly replenished in T. velutina, especially during the first 2 h in short-lived flowers, such as the one-day flowers of Turnera Diptera 11 1.10 post anthesis (Figure 2). Replenishment is a general feature of species, because it makes the flower attractive again for another Lepidoptera 10 1 EFN secretion (Pacini et al., 2003; Pacini and Nepi, 2007; Pacini visit, potentially increasing pollen transfer, pollen deposition, Coleoptera 1 0.1 and Nicolson, 2007). In fact, we are unaware of any report and seed set. Interestingly, Dutton et al. (2016) reported no FN Wasps 1 0.1 documenting extrafloral nectaries incapable of replenishing resupply in flowers from three congeners (Turnera ulmifolia,

Frontiers in Plant Science | www.frontiersin.org 7 August 2018 | Volume 9 | Article 1093 Frontiers in Plant Science | www.frontiersin.org 8 August 2018 | Volume 9 | Article 1093 Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs

TABLE 3 | Model statistics for the experiments testing indirect (nectar-mediated) and direct (pollinator deterrence) ant-pollinator conflict. ant activity at extrafloral nectaries and pollinator visitation to Does Herbivore Deterrence Match flowers has been reported for facultative ant-plants associated Response Distribution Fixed effects N LRT P-value Random Variance SD Pollinator Deterrence? effects with many ant species simultaneously [Vachellia constricta Although some studies have documented variation among ant (Fabaceae; Nicklen and Wagner, 2006), Acacia myrtifolia, species in aggression toward herbivores (Ness, 2006; Miller, Indirect conflict Floral nectar Gaussian Clogging treatment 50 0.0813 0.7754 Family 0.6987 0.8359 (Acacia sensu stricto,Fabaceae;Martínez-Bauer et al., 2015), 2007; Ohm and Miller, 2014), little is known about the effect of μ ( l) and Heteropterys pteropetala (Malpighiaceae; Assunção et al., different patrolling ant species with varying levels of aggressivity Floral nectar Gaussian Clogging treatment 50 0.21006 0.6467 Family 138.5 11.77 2014)]. Our results add Turnera velutina to the list of facultative on pollinator visitation (Ness, 2006; Miller, 2007; LeVan et al., (μl) myrmecophiles with synchronized ant and pollinator activity 2014; Ohm and Miller, 2014). Nonetheless, a positive correlation Direct conflict Number of visitors Poisson Ants in flowers 95 1.1848 0.7567 OLRE 2.77e−08 0.0001 (Figure 2). This synchronous myrmecophile vs. segregated between the level of defense provided and the level of pollinator Plant 0.144 0.3803 myrmecophyte pattern is consistent with evidence that ants in e−10 e−05 deterrence they exert has often been assumed since ant traits Likelihood of being Binomial Ants in flowers 373 10.715 0.01337 * ID visitor 8.279 2.877 obligate mutualisms are more aggressive and better defenders alerted Plant 0.595 0.7715 involved in defense (patrolling activity and aggressivity) are likely (Chamberlain and Holland, 2009; Rosumek et al., 2009; Trager Proportion of time per Gaussian (logit) Ants in flowers 112 7.5728 0.055 * ID visitor 0.0009313 0.03052 to be the same as those involved in pollinator deterrence (Ohm visit spent displaying Plant 0.0201 0.14198 et al., 2010), but may impose greater ecological costs on host and Miller, 2014). Bees tend to forage in a way that maximizes inspection behaviors pollination. We suggest that temporal segregation of mutualists the net benefit of each foraging trip (Stephens and Krebs, Duration of presence Gamma Ants in flowers 307 392.37 2.2e−16 *** OLRE ID 0.1202 0.3468 and/or ant repellent floral volatiles are alternative strategies 1986; Jones, 2010; Cembrowski et al., 2014). When foraging in behaviors (sec) visitor Plant 0.001587 0.03984 that reduce such costs. Further studies on the timing of e−10 ant-plants, this benefit might be maximized if foragers avoid 1.928 0.001389 pollinator, ant visitation and ant aggressivity in a wider range flowers or patches where predation risk is high (Dukas, 2001; *P < 0.05; ***P < 0.001. of systems are required to test the temporal component of this Dukas and Morse, 2003; Ness, 2006; Jones and Dornhaus, 2011; hypothesis. Assunção et al., 2014), as could be the case when encountering ant species that attack pollinators. Some ants also consume FN and pollen, and such plants may represent high risk foraging Direct Conflict environments with low net rewards for pollinators (Ness, 2006). We showed that dead ants inside the flowers of T. velutina Shorter or fewer visits to such flowers may be a pollinator have an impact on honeybee behavior (Figure 3). Ant presence strategy to maximize foraging efficiency by avoiding flowers, was correlated with shorter honeybee flower visits (Figure 3C), plants, or patches with high predation risk (Jones and Dornhaus, an increase in the proportion of visitors displaying inspection 2011). behaviors, and increased duration of inspection behaviors In T. velutina, the most aggressive ant guard, Dorymyrmex per visiting bout (Figure 3). We interpret longer inspection bicolor, had the strongest effect on pollinator behavior (Figure 3), behavior to indicate increased caution in the bees (as previously while Brachymyrmex sp. ants inside the flowers did not reduce assumed by: Altshuler, 1999; Ness, 2006; Junker et al., 2007; the duration of pollinator visits. The least effective anti-herbivore Assunção et al., 2014; Cembrowski et al., 2014). Our findings ant species, Paratrechina longicornis (Villamil unpublished data), are consistent with work on Heteropterys pteropetala in halved the duration of pollinator visits (Figure 3). In Ferocactus which plastic ants inside flowers negatively affected pollination wislizeni, plants tended by Solenopsis xyloni,themostaggressive (Assunção et al., 2014). Results for H. pteropetala differ ant species, had fewer and shorter pollinator visits (Ness, from ours in that the bees pollinating H. pteropetala showed 2006). Such differences are consistent with pollinator sensitivity significantly reduced visitation rates to flowers containing to ant aggressiveness. In contrast, although ant exclusion in FIGURE 4 | Effect of clogging on nectar secretion and sugar re-allocation in Turnera velutina for nectar volume (A) and sugar content (B), between control (C) and plastic ants. In contrast, honeybees in T. velutina did not Opuntia imbricata significantly increased pollinator visitation, treatment (T) plants (means ± se; n = 50 plants). Values shown are a difference in floral nectar, defined as: [FN post-treatment–FN pre-treatment]. visit flowers containing ant corpses less frequently than there were no differences in impacts associated with different control flowers (Figure 3A). In both systems, ants feeding ant species (Ohm and Miller, 2014), and no evidence that at extrafloral nectaries did not hinder pollination (Assunção the more aggressive guard (Liometopum apiculatum)hada Turnera subulata,andTurnera joelii), when sampling FN in the a few minutes after the corolla is fully open (Table 3)and et al., 2014; Villamil et al., unpublished data). This suggests stronger deterring effect on pollinators (Ohm and Miller, 2014). morning and afternoon, finding no FN secretion in the afternoon 45 min before peak EFN sugar secretion (Figure 2A). This that while pollinators avoid ants in flowers, plants may have Whether the level of ant aggressivity toward herbivores correlates collection. These observations show either variation in nectar represents a slight mismatch in the timing of rewards for ants evolved other mechanisms to prevent ants from entering positively with the ecological costs on pollination via pollinator resupply within closely related species displaying similar floral and pollinators, which may underlie the 85 min mismatch in flowers, resulting in only rare encounters between ants and deterrence remains unknown (but see: Ness, 2006; Miller, 2007; biology, or perhaps suggest that shorter term dynamics in the estimated peaks of ants and floral visitor activity (Table 1). pollinators. LeVan et al., 2014; Ohm and Miller, 2014), and should be tested, nectar supply of the other three species were not detected by To our knowledge, temporal segregation in the activity of Although experiments that place ant cues on flowers can not assumed. the sampling methods used. The latter highlights the need to ants and pollinators has been reported only for obligate tell us about the response of pollinators to ants, they must test floral resupply within short time scales, because over long (myrmecophytic) species patrolled or tended by a single ant be interpreted with caution as an indicator of current ant- sampling intervals, flowers can both secrete and reabsorb nectar species at a time (Humboltia brunonis (Fabaceae; Gaume pollinator conflict. Firstly, because ants may rarely enter flowers Indirect Conflict (Pacini et al., 2003; Nicolson et al., 2007; Willmer, 2011). Finding et al., 2005), Hirtella physophora (Chrysobalanaceae; Malé et al., (Villamil et al. submitted). Secondly, by placing such ant cues Our experimental approach found no evidence for a trade-off in no standing crop when a flower is resampled in the afternoon 2015), and Opuntia imbricata (Cactaceae; 2014)). In some in flowers we may be violating existing ant-excluding or ant- sugar investment in extrafloral and floral nectar in T. velutina. does not rule out replenishment after emptying in the morning, specialized systems, ant and pollinator activity occurs in close repelling plant mechanisms (Willmer, 2011). Thirdly, in contrast We conclude that there is no indirect nectar-mediated conflict and then reabsorption later in the day (Kearns and Inouye, proximity and simultaneously, but conflict is prevented by to such experimental treatments, ants do not naturally remain in between guarding ants and pollinators in Turnera velutina,and 1993). ant-repellent floral volatiles [Vachellia zanzibarica (Fabaceae; the flowers for long periods (Assunção et al., 2014), and only a that pollinators do not obtain greater rewards when rewards for Model-based estimation of the daily timing of the maxima Willmer and Stone, 1997)] and Vachellia hindsii (Fabaceae; low proportion of flowers may be occupied at any one time. For patrolling ants are eliminated. of nectar secretion suggest that floral nectar secretion peaks Raine et al., 2002). On the other hand, temporal overlap in instance, in T. velutina only 10% of the flowers are occupied by We found only two previous studies testing indirect, nectar- ants (Villamil et al., submitted). mediated ant-pollinator conflict by quantifying sugary rewards

Frontiers in Plant Science | www.frontiersin.org 9 August 2018 | Volume 9 | Article 1093 Frontiers in Plant Science | www.frontiersin.org 10 August 2018 | Volume 9 | Article 1093 Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs

(FN and EFN) to both mutualists (Chamberlain and Rudgers, suggests that plants may experience fewer investment trade-offs REFERENCES by extrafloral nectar: fidelity, cheats, and lies. Insectes Soc. 63, 207–221. 2012; Dutton et al., 2016). Previous work on other Turnera among different functional traits than previously assumed. doi: 10.1007/s00040-016-0466-2 species by Dutton et al. (2016) found evidence of a trade- Adler, L. S. (2008). “Selection by pollinators and herbivores on attraction and Dukas, R. (2001). Effects of perceived danger on flower choice by bees. Ecol. Lett. defense,” in The Evolutionary Biology of Herbivorous Insects Specialization, 4, 327–333. doi: 10.1046/j.1461-0248.2001.00228.x off in two of the three Turnera species tested; removing EFN CONCLUSIONS Speciation and Radiation, ed K. J. Tilmon (Berkeley, CA: University California Dukas, R., and Morse, D. H. (2003). Crab spiders affect flower visitation decreased FN and vice versa in T. ulmifolia and T. subulata, Press), 162–173. by bees. Oikos 101, 157–163. doi: 10.1034/j.1600-0706.2003.12 but not in T. joelii. Interestingly, both Turnera species in which To our knowledge, trade-offs between extrafloral and floral Altshuler, D. L. (1999). Novel interactions of non-pollinating ants with 143.x trade-offs were detected by Dutton et al. (2016) invested equally nectar traits have been studied in 41 species from two genera: pollinators and fruit consumers in a tropical forest. Oecologia 119, 600–606. Dutton, E. M., Luo, E. Y., Cembrowski, A. R., Shore, J. S., Frederickson, M. E., in FN and EFN, whilst T. joelli (which showed no trade-off) 37 Gossypium species (Chamberlain and Rudgers, 2012), and doi: 10.1007/s004420050825 Caruso, C. M., et al. (2016). Three’s a crowd: trade-offs between attracting Arbo, M. M. (2005). Estudios sistemáticos en Turnera (Turneraceae). III. series pollinators and ant bodyguards with nectar rewards in Turnera. Am. Nat. 188, invested more in EFN (Dutton et al., 2016).Thesamepattern four Turnera species, including this study (Dutton et al., 2016; anomalae y turnera. Bonplandia 14, 115–318. Available online at: http://hdl. 38–51. doi: 10.1086/686766 holds for T. velutina, a species with an asymmetric investment Villamil, 2017, Figure 2). Negative correlations or evidence for handle.net/11336/30558 Dyer, L. (2008). “Tropical tritrophic interactions: nasty hosts and ubiquitous toward EFN, which accounts for 95% of the sugar allocation trade-offs have been found in only two of these species: T. Armbruster, W. (1997). Exaptations link evolution of plant-herbivore and plant- cascades,” in Tropical Forest Community Ecology, eds W. P. Carson, and S. A. per leaf-flower module, and where we found no trade-off or ulmifolia and T. subulata (Dutton et al., 2016), representing less pollinator interactions: a phylogenetic inquiry. Ecology 78, 1661–1672. doi: 10. Schnitzer (Chichester: Wiley-Blackwell), 275–293. resource reallocation from EFN to FN (Figure 4). Unfortunately than 5% of the species studied. Although many more studies 1890/0012-9658(1997)078[1661:ELEOPH]2.0.CO;2 Dyer, L., and Phyllis, C. (2002). “Temperate versus tropical communities,” Assunção, M. A., Torezan-Silingardi, H. M., and Del-Claro, K. in Multitrophic Level Interactions, ed B. H. Teja Tscharntke (Cambridge: data on FN and EFN volume and sugar content were not are required to shed light on quantitative trends of floral and (2014). Do ant visitors to extrafloral nectaries of plants repel Cambridge University Press), 67–88. reported for the cotton species (Chamberlain and Rudgers, extrafloral investment in plants, trade-offs between floral and pollinators and cause an indirect cost of mutualism? Flora-Morphol. Elias, T., Rozich, W., and Newcombe, L. (1975). The foliar and 2012). extrafloral seem infrequent. On the other hand, evidence of Distrib.Funct.Ecol.Plants209, 244–249. doi: 10.1016/j.flora.2014.0 floral nectaries of Turnera ulmifolia L. Am. J. Bot. 62, 570–576. One possible reason for lack of a trade-off is that sugar direct conflict with patrolling ants reducing pollinator visitation 3.003 doi: 10.1002/j.1537-2197.1975.tb14085.x frequency and duration, inducing inspection behaviors and Ballantyne, G., and Willmer, P. (2012). Nectar theft and floral ant-repellence: Escalante-Pérez, M., and Heil, M. (2012). “Nectar secretion: its ecological context is not a limiting resource for the plant. If so, there would a link between nectar volume and ant-repellent traits? PLoS ONE 7:e43869. and physiological regulation,” in Secretions and Exudates in Biological Systems. be no reason to expect dynamic reallocation. Estimates of the increasing foraging time has been widely reported (Rudgers and doi: 10.1371/journal.pone.0043869 Signaling and Communication in Plants, Vol 12, eds J. Vivanco and F. Baluška metabolic costs of nectar secretion vary, and while some studies Gardener, 2004; Ness, 2006; Chamberlain and Rudgers, 2012; Bates, D., Maechler, M., Bolker, B., Walker, S., Christensen, R. H. B., Singmann, H., (Berlin; Heidelberg: Springer), 187–219. suggest low metabolic costs (O’Dowd, 1979: EFN accounts Malé et al., 2012, 2015; Assunção et al., 2014; Koptur et al., 2015; et al. (2016). “Package ‘lme4’,” i n R Package Version 1, 1–10. Ettershank, G. (1966). A generic revision of the world Myrmicinae related to for 1% of the total energy invested per leaf), others indicate Martínez-Bauer et al., 2015). We suggest that these two issues are Bentley, B. (1977). Extrafloral nectaries and protection by pugnacious bodyguards. Solenopsis and Pheidologeton (Hymenoptera: Formicidae). Aust. J. Zool. 14, Annu. Rev. Ecol. Syst. 8, 407–427. doi: 10.1146/annurev.es.08.110177.0 73–171. doi: 10.1071/ZO9660073 investment of up to 37% of daily photosynthesis in floral not isolated, and hypothesize that positive correlations between 02203 Fonseca, C. (1993). Nesting space limits colony size of the plant-ant nectar (Southwick, 1984; Pyke, 1991). A second reason, which FN and EFN investment in ant-plants may be a plant strategy to Bolton, B. (1987). A review of the Solenopsis genus-group and revision of Pseudomyrmex concolor. Oikos 67, 473–482. doi: 10.2307/3545359 applies in particular to comparative cross-species analyses rather compensate or lure pollinators to apparently risky flowers. Afrotropical Monomorium Mayr (Hymenoptera: Formicidae). Bull. Br. Mus. Fonseca, C. R. (1999). Amazonian ant–plant interactions and the than experimental manipulations, is that investment in both Entomol. 263–452. nesting space limitation hypothesis. J. Trop. Ecol. 15, 807–825. forms of nectar may be influenced by other aspects of life Cembrowski, A. R., Tan, M. G., Thomson, J. D., and Frederickson, M. E. (2014). doi: 10.1017/S0266467499001194 DATA AVAILABILITY STATEMENT Ants and ant scent reduce bumblebee pollination of artificial flowers. Am. Nat. Fontaine, C., Guimarães, P. R., Kéfi, S., Loeuille, N., Memmott, history strategy. Chamberlain and Rudgers (2012) found no 183, 133–139. doi: 10.1086/674101 J., Van Der Putten, W. H., et al. (2011). The ecological and significant negative correlations between extrafloral nectary and The raw data supporting the conclusions of this manuscript will Chamberlain, S. A., and Holland, J. N. (2009). Quantitative synthesis of context evolutionary implications of merging different types of networks. floral traits in a comparative analysis across cotton (Gossypium) be made available by the authors upon request. dependency in ant-plant protection mutualisms. Ecology 90, 2384–2392. Ecol. Lett. 14, 1170–1181. doi: 10.1111/j.1461-0248.2011.01 species, and correlations were significantly positive in 11 of 37 doi: 10.1890/08-1490.1 688.x cotton species. Foliar extrafloral nectary volume was positively Chamberlain, S. A., and Rudgers, J. A. (2012). How do plants balance Frederickson, M. E. (2009). Conflict over reproduction in an ant-plant symbiosis: AUTHOR CONTRIBUTIONS multiple mutualists? Correlations among traits for attracting protective why allomerus octoarticulatus ants sterilize cordia nodosa trees. Am. Nat. 173, associated with plant investment in floral nectar, rejecting bodyguards and pollinators in cotton (Gossypium). Evolut. Ecol. 26, 65–77. 675–681. doi: 10.1086/597608 the hypothesis of trade-offs among investments in pollinators NV conceived the ideas, all authors designed the experiments. doi: 10.1007/s10682-011-9497-3 Gaume, L., and Mckey, D. (1999). An ant-plant mutualism and its host-specific versus bodyguards in Gossypium. Several potential mechanisms NV conducted fieldwork, processed, and analyzed the data. NV Comba, L., Corbet, S. A., Barron, A., Bird, A., Collinge, S., Miyazak, I. parasite: activity rhythms, young leaf patrolling, and effects on herbivores underlying the positive correlations between FN and EFN have led the writing of this paper, with critical inputs from KB and GS. N., et al. (1999). Garden flowers: insect visits and the floral reward of of two specialist plant-ants inhabiting the same myrmecophyte. Oikos 84, horticulturally-modified variants. Ann. Bot. 83, 73–86. doi: 10.1006/anbo.1998. 130–144. doi: 10.2307/3546873 been proposed, including pleiotropy, and genetic, physiological 0798 Gaume, L., Zacharias, M., and Borges, R. M. (2005). Ant-plant conflicts and a novel or ecological linkage (Chamberlain and Rudgers, 2012). The FUNDING Cuautle, M., and Rico-Gray, V. (2003). The effect of wasps and ants on the case of castration parasitism in a myrmecophyte. Evol. Ecol. Res. 7, 435–452. pleiotropy or genetic linkage hypothesis could be tested using reproductive success of the extrafloral nectaried plant Turnera ulmifolia Hanna, C., Naughton, I., Boser, C., Alarcón, R., Hung, K. L. J., and genome sequencing (Chamberlain and Rudgers, 2012). Positive Thanks to the Mexican taxpayers who funded NV via a (Turneraceae). Funct. Ecol. 17, 417–423. doi: 10.1046/j.1365-2435.2003.00732.x Holway, D. (2015). Floral visitation by the Argentine ant reduces bee correlations could also arise from physiological or ecological CONACYT PhD scholarship. Research costs were funded by Cuautle, M., Rico-Gray, V., and Díaz-Castelazo, C. (2005). Effects of ant behavior visitation and plant seed set. Ecology 96, 222–230. doi: 10.1890/14-0 Davis Trust Research Grant awarded to NV (2016), Wellcome- and presence of extrafloral nectaries on seed dispersal of the Neotropical 542.1 linkage. Traits such as FN and EFN may be physiologically myrmecochore Turnera ulmifolia L. (Turneraceae). Biol. J. Linnean Soc. 86, Heil, M. (2015). Extrafloral nectar at the plant-insect interface: a spotlight on linked. However, the fact that in Gossypium FN volume was BBSRC Insect Pollinator Initiative: Urban Pollinators: their 67–77. doi: 10.1111/j.1095-8312.2005.00525.x chemical ecology, phenotypic plasticity, and food webs. Annu. Rev. Entomol. most strongly correlated with foliar EFN volume, but FN was ecology and conservation (BB/I000305/1) awarded to GS, and Damián, X., Fornoni, J., Domínguez, C. A., Boege, K., and Baltzer, J. (2018). 60, 213–232. doi: 10.1146/annurev-ento-010814-020753 weakly correlated with bracteal EFN volume (Chamberlain and PAPIIT-UNAM IN211314 grant awarded to KB. Ontogenetic changes in the phenotypic integration and modularity of leaf Herrera, C. (2000). Measuring the effects of pollinators and herbivores: Rudgers, 2012) questions the physiological linkage hypothesis functional traits. Funct. Ecol. 32, 234–246. doi: 10.1111/1365-2435.12971 evidence for non-additivity in a perennial herb. Ecology 81, 2170–2176. Dáttilo, W., Lara-Rodríguez, N., Jordano, P., Guimarães, P. R., Thompson, J. N., doi: 10.2307/177105 since bracteal and floral nectaries are spatially closer than floral ACKNOWLEDGMENTS Marquis, R. J., et al. (2016). Unravelling Darwin’s entangled bank: architecture Hinde, J. (1982). “Compound Poisson regression models,”in GLIM 82: Proceedings and extrafloral nectaries, but they are not strongly correlated. and robustness of mutualistic networks with multiple interaction types. Proc. R. of the International Conference on Generalised Linear Models (New York, NY: Wesuggestthatlackofanytrade-offcouldalsoindicatethat Thanks to S. Ochoa-López who kindly donated the plants for the Soc. B 283:20161564. doi: 10.1098/rspb.2016.1564 Springer), 109–121. FN and EFN may be phenotypically integrated as a functional greenhouse experiment, X. Damián-Domínguez who provided Davidson, D. W., and Cook, S. C. (2008). “Tropical arboreal ants: linking nutrition Hothorn, T., Bretz, F., Westfall, P., and Heiberger, R. (2008). “Multcomp: module for mutualist attraction. Although formal analyses are insightful thoughts for the experimental design and to J. Hadfield to roles in rainforests ecosystems,” in Tropical Forest Community Ecology, eds Simultaneous Inference for General Linear Hypotheses R Package Version, 1.0-3.” W. P. Carson, and S. A. Schnitzer. (Chichester: Wiley-Blackwell), 334–348. Jones, E. I. (2010). Optimal foraging when predation risk increases required to test this hypothesis, we think it is a strong possibility who provided valuable inputs toward the statistical analyses. Del-Claro, K., Lange, D., Torezan-Silingardi, H. M., Anjos, D. V., Calixto, E. S., with patch resources: an analysis of pollinators and ambush since T. velutina leaves are phenotypically integrated modules in Thanks to the volunteers who helped out during fieldwork: A. Dáttilo, W., et al. (2018). “The complex ant–plant relationship within tropical predators. Oikos 119, 835–840. doi: 10.1111/j.1600-0706.2009.17 which leaf economics, defensive and morphological traits covary Bernal, A. Martínez, R. Ríos, I. Silva, and A. Fournier, to the staff ecological networks,” in Ecological Networks in the Tropics, eds W. Dáttilo, V. 841.x and are ecologically linked (Damián et al., 2018). Whatever the at CICOLMA station, to H. Köck who helped processing pollen Rico-Gray (Cham: Springer), 59–71. Jones, E. I., and Dornhaus, A. (2011). Predation risk makes bees reject rewarding samples, and to R. Pérez-Ishiwara for logistic support. Del-Claro, K., Rico-Gray, V., Torezan-Silingardi, H. M., Alves-Silva, E., Fagundes, flowers and reduce foraging activity. Behav. Ecol. Sociobiol. 65, 1505–1511. drivers of these positive correlations may be, available evidence R., Lange, D., et al. (2016). Loss and gains in ant–plant interactions mediated doi: 10.1007/s00265-011-1160-z

Frontiers in Plant Science | www.frontiersin.org 11 August 2018 | Volume 9 | Article 1093 Frontiers in Plant Science | www.frontiersin.org 12 August 2018 | Volume 9 | Article 1093 Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs Villamil et al. Ant-Pollinator Conflict: Deterrence but no Nectar Trade-Offs

Junker, R., Chung, A. Y., and Blüthgen, N. (2007). Interaction between flowers, Palmer, T. M., and Brody, A. K. (2007). Mutualism as reciprocal exploitation: Willmer, P. G., and Stone, G. N. (1997). How aggressive ant-guards Conflict of Interest Statement: The authors declare that the research was ants and pollinators: additional evidence for floral repellence against ants. Ecol. african plant-ants defend foliar but not reproductive structures. Ecology 88, assist seed-set in Acacia flowers. Nature 388, 165–167. doi: 10.1038/ conducted in the absence of any commercial or financial relationships that could Res. 22, 665–670. doi: 10.1007/s11284-006-0306-3 3004–3011. doi: 10.1890/07-0133.1 40610 be construed as a potential conflict of interest. Kearns, C. A., and Inouye, D. W. (1993). Techniques for Pollination Biologists. Pineda, A., Dicke, M., Pieterse, C. M., and Pozo, M. J. (2013). Beneficial microbes Yu, D. W., and Pierce, N. E. (1998). A castration parasite of an ant– Niwot, CO: University Press of Colorado. in a changing environment: are they always helping plants to deal with insects? plant mutualism. Proc. R. Soc. London B Biol. Sci. 265, 375–382. Copyright © 2018 Villamil, Boege and Stone. This is an open-access article distributed Koptur, S., Jones, I. M., and Peña, J. E. (2015). The influence of host plant extrafloral Funct. Ecol. 27, 574–586. doi: 10.1111/1365-2435.12050 doi: 10.1098/rspb.1998.0305 under the terms of the Creative Commons Attribution License (CC BY). The use, nectaries on multitrophic interactions: an experimental investigation. PLoS Pyke, G. H. (1991). What does it cost a plant to produce floral nectar? Nature 350, Zedillo-Avelleyra, P. (2017). Variación Genética en la Expresión de distribution or reproduction in other forums is permitted, provided the original ONE 10:e0138157. doi: 10.1371/journal.pone.0138157 58–59. doi: 10.1038/350058a0 Trayectorias Ontogenéticas de la Defensa de Turnera velutina Master author(s) and the copyright owner(s) are credited and that the original publication LeVan, K. E., Hung, K.-L. J., Mccann, K. R., Ludka, J. T., and Holway, D. A. (2014). Raine, N., Willmer, P., and Stone, G. (2002). Spatial structuring and floral in Biological Sciences Research. Universidad Nacional Autónoma de in this journal is cited, in accordance with accepted academic practice. No use, Floral visitation by the Argentine ant reduces pollinator visitation and seed avoidance behavior prevent ant-pollinator conflict in a Mexican ant-acacia. México. distribution or reproduction is permitted which does not comply with these terms. set in the coast barrel cactus, Ferocactus viridescens. Oecologia 174, 163–171. Ecology 83, 3086–3096. doi: 10.1890/0012-9658(2002)083[3086:SSAFAB]2.0. doi: 10.1007/s00442-013-2739-z CO;2 Malé, P.-J. G., Leroy, C., Dejean, A., Quilichini, A., and Orivel, J. Rico-Gray, V., and Oliveira, P. S. (2007). “Mutualism from Antagonism: Ants (2012). An ant symbiont directly and indirectly limits its host plant’s and Flowers,” in The Ecology and Evolution of Ant-plant Interactions,edsV. reproductive success. Evol. Ecol. 26, 55–63. doi: 10.1007/s10682-011-9 Rico-Gray,P.S.Oliveira(Chicago,IL:UniversityofChicagoPress),85–98. 485-7 Rosumek, F., Silveira, F., De S Neves, F., De U Barbosa, N., Diniz, L., Oki, Y., Malé,P.-J.G.,Leroy,C.,Lusignan,L.,Petitclerc,F.,Quilichini,A.,andOrivel,J. et al. (2009). Ants on plants: a meta-analysis of the role of ants as plant biotic (2015). The reproductive biology of the myrmecophyte, Hirtella physophora, defenses. Oecologia 160, 537–549. doi: 10.1007/s00442-009-1309-x and the limitation of negative interactions between pollinators and ants. Rudgers, J., and Gardener, M. (2004). Extrafloral nectar as a resource mediating Arthropod Plant Interact. 9, 23–31. doi: 10.1007/s11829-014-9352-x multispecies interactions. Ecology 85, 1495–1502. doi: 10.1890/03-0391 Martínez-Bauer, A. E., Martínez, G. C., Murphy, D. J., and Burd, M. Sosenski, P., Ramos, S., Domínguez, C. A., Boege, K., and Fornoni, J. (2016). (2015). Multitasking in a plant–ant interaction: how does Acacia Pollination biology of the hexaploid self-compatible species Turnera velutina myrtifolia manage both ants and pollinators? Oecologia 178, 461–471. (Passifloraceae). Plant Biol. 19, 101–107. doi: 10.1111/plb.12518 doi: 10.1007/s00442-014-3215-0 Southwick, E. E. (1984). Photosynthate allocation to floral nectar: a Melián, C. J., Bascompte, J., Jordano, P., and Krivan, V. (2009). Diversity in a neglected energy investment. Ecology 65, 1775–1779. doi: 10.2307/19 complex ecological network with two interaction types. Oikos 118, 122–130. 37773 doi: 10.1111/j.1600-0706.2008.16751.x Stanton, M. L., and Palmer, T. M. (2011). The high cost of mutualism: effects of Miller, T. E. X. (2007). Does having multiple partners weaken the benefits of four species of East African ant symbionts on their myrmecophyte host tree. facultative mutualism? A test with cacti and cactus-tending ants.Oikos116, Ecology 92, 1073–1082. doi: 10.1890/10-1239.1 500–512. doi: 10.1111/j.2007.0030-1299.15317.x Stanton, M. L., Palmer, T. M., Young, T. P., Evans, A., and Turner, M. L. (1999). Nahas, L., Gonzaga, M. O., and Del-Claro, K. (2012). Emergent impacts of ant and Sterilization and canopy modification of a swollen thorn acacia tree by a spider interactions: herbivory reduction in a tropical savanna tree. Biotropica plant-ant. Nature 401, 578–581. doi: 10.1038/44119 44, 498–505. doi: 10.1111/j.1744-7429.2011.00850.x Stephens, D. W., and Krebs, J. R. (1986). Foraging Theory. Princeton, NJ: Princeton Ness, J. H. (2006). A mutualism’s indirect costs: the most aggressive University Press. plant bodyguards also deter pollinators. Oikos 113, 506–514. Strauss, S. (1997). Floral characters link herbivores, pollinators, and plant fitness. doi: 10.1111/j.2006.0030-1299.14143.x Ecology 78, 1640–1645. doi: 10.1890/0012-9658(1997)078[1640:FCLHPA]2.0. Nicklen, E. F., and Wagner, D. (2006). Conflict resolution in an ant–plant CO;2 interaction: acaciaconstricta traits reduce ant costs to reproduction. Oecologia Strauss, S. Y., and Irwin, R. E. (2004). Ecological and evolutionary consequences of 148, 81–87. doi: 10.1007/s00442-006-0359-6 multispecies plant-animal interactions. Annu.Rev.Ecol.Evol.Syst.35, 435–466. Nicolson, S., Nepi, M., and Pacini, E. (2007). Nectaries and Nectar. Dordrecht: doi: 10.1146/annurev.ecolsys.35.112202.130215 Springer. Strauss, S. Y., Siemens, D. H., Decher, M. B., and Mitchell-Olds, T. Ochoa-López, S., Villamil, N., Zedillo-Avelleyra, P., and Boege, K. (2015). Plant (1999). Ecological costs of plant resistance to herbivores in the currency defence as a complex and changing phenotype throughout ontogeny. Ann. Bot. of pollination. Evolution 53, 1105–1113. doi: 10.1111/j.1558-5646.1999.tb0 116, 797–806. doi: 10.1093/aob/mcv113 4525.x Ohm, J. R., and Miller, T. E. (2014). Balancing anti-herbivore benefits Trager, M. D., Bhotika, S., Hostetler, J. A., Andrade, G. V., Rodriguez- and anti-pollinator costs of defensive mutualists. Ecology 95, 2924–2935. Cabal, M. A., Mckeon, C. S., et al. (2010). Benefits for plants in doi: 10.1890/13-2309.1 ant-plant protective mutualisms: a meta-analysis. PLoS ONE 5:e14308. Orivel, J., Lambs, L., Mal,é, P. J. G., Leroy, C., Grangier, J., Otto, T., doi: 10.1371/journal.pone.0014308 et al. (2011). Dynamics of the association between a long-lived understory Travieso-Bello, A. C., and Campos, A. (2006). “Los componentes del paisaje,” in myrmecophyte and its specific associated ants. Oecologia 165, 369–376. Entornos Veracruzanos: la costa de La Mancha, ed P. Moreno-Casasola, (Comp) doi: 10.1007/s00442-010-1739-5 (México: Instituto de Ecología AC Xalapa, Ver.), 139–150. Orona-Tamayo, D., Wielsch, N., Escalante-Pérez, M., Svatos, A., Molina-Torres, Tscharntke, T., and Hawkins, B. (2002). “Multitrophic level interactions: an J., Muck, A., et al. (2013). Short-term proteomic dynamics reveal metabolic introduction,” in Multitrophic Level Interactions, ed B. H. Teja Tscharntke factory for active extrafloral nectar secretion by Acacia cornigera ant-plants. (Cambridge: Cambridge University Press), 1–7. Plant J. 73, 546–554. doi: 10.1111/tpj.12052 Villamil, N. (2017). Why are flowers sweeter than fruits or buds? variation Pacini, E., and Nepi, M. (2007). “Nectar production and presentation,” in Nectaries in extrafloral nectar secretion throughout the ontogeny of a myrmecophile. and Nectar, eds S. Nicolson, M. Nepi, and E. Pacini (Dordrecht: Springer), Biotropica 49, 581–585. doi: 10.1111/btp.12463 167–214. Villamil, N., Márquez-Guzmán, J., and Boege, K. (2013). Understanding Pacini, E., Nepi, M., and Vesprini, J. (2003). Nectar biodiversity: a short review. ontogenetic trajectories of indirect defence: ecological and anatomical Plant Syst. Evolut. 238:7. doi: 10.1007/s00606-002-0277-y constraints in the production of extrafloral nectaries. Ann. Bot. 112, 701–709. Pacini, E., and Nicolson, S. (2007). “Introduction,” in Nectaries and Nectar,edsS. doi: 10.1093/aob/mct005 Nicolson, M. Nepi, and E. Pacini (Dordrecht: Springer), 1–19. Willmer, P. (2011). Pollination and Floral Ecology. Princeton, NJ; Oxford: Palmer, T., Doak, D., Stanton, M., Bronstein, J., Kiers, E., Young, T., et al. (2010). Princeton University Press. Synergy of multiple partners, including freeloaders, increases host fitness in Willmer, P. G., Nuttman, C. V., Raine, N. E., Stone, G. N., Pattrick, J. G., Henson, a multispecies mutualism. Proc. Natl. Acad. Sci. U.S.A. 107, 17234–17239. K., et al. (2009). Floral volatiles controlling ant behaviour. Funct. Ecol. 23, doi: 10.1073/pnas.1006872107 888–900. doi: 10.1111/j.1365-2435.2009.01632.x

Frontiers in Plant Science | www.frontiersin.org 13 August 2018 | Volume 9 | Article 1093 Frontiers in Plant Science | www.frontiersin.org 14 August 2018 | Volume 9 | Article 1093