TITLE: Diseases Challenges and Defenses in Leaf-Cutting Ants

FACULTY OF SCIENCE UNIVERSITY OF COPENHAGEN

PhD thesis

Sze Huei, Yek

Disease challenges and defences in

leaf-cutting ants

Academic advisor: Jacobus J. Boomsma

FACULTY OF SCIENCE UNIVERSITY OF COPENHAGEN

Disease challenges and defences in leaf-cutting ants

A dissertation submitted to the University of Copenhagen in accordance with the requirements for the degree of the PhD at the Graduate School of Science, Faculty of Science, University of Copenhagen, Denmark to be defended publicly before a panel of examiners

PhD thesis Sze Huei, Yek September 2012

Academic advisor: Jacobus J. Boomsma

Front cover illustration: A Marvel of Ants @ Bence Máté Recipient of Eric Hosking Award Preface

This PhD thesis is the result of three years work carried out at the Centre for Social Evolution, Department of Biology, University of Copenhagen, under supervision of Prof. Jacobus (Koos) Boomsma, with collaborations and assistance from Assoc. Prof. David R. Nash, Assoc. Prof. Annette Jensen, Assis. Prof. Morten Schiøtt, Post-doc Michael Poulsen and Post-doc Sandra Andersen, each of whom played a crucial role in the timely completion of my thesis. Projects in my PhD thesis were supported by the Danish National Research Foundation.

The thesis initially focused on one component of defence in ants – the metapleural glands. Over time, projects were adjusted to take a wider approach related to leaf-cutting ants’ defences. These adjustments do not affect the level of ambition, but merely reflect novel insights and experimental opportunities that could not be foreseen when the initial PhD plan was conceived. I am grateful to be able to present four area of work that explores the dynamics of parasitism and defences that leaf- cutting ants employ to maintain the integrity of this delicate mutualism with their fungal crops.

The thesis comprises of three main parts. The first is a synopsis of the current understanding of parasitism and defence in social insects, focusing particularly on dynamics of the model system leaf-cutting ants. This section also includes the framework and methodology used to address questions pertaining to defence in leaf-cutting ants-fungus mutualism. Second, four chapters of original empirical works, two of which were already published in peer-reviewed journals. The remaining two are in various stages of preparation for publication. Third, a concluding section discussing questions that await further exploration and insights from this thesis that can be applied to a similar model system.

Sze Huei, Yek Copenhagen, September 2012

Table of Contents English summary 1 Dansk resumé 3 Chinese summary 5

SYNOPSIS 7 1. Parasite-host interactions 8 2. Susceptibility of parasites in social insects 9 3. Evolutionary mitigations of parasitism in social insects 10 4. Leaf-cutting ants and their susceptibility to parasitism 11 5. Defences against parasites in social insects 13 6. Methodologies and techniques 17 7. Conceptual frameworks in evolutionary ecology studies 21

RESEARCH CHAPTERS Chapter 1 34-82 Differential gene expression after multiple fungal disease challenges in leaf- cutting ants

Chapter 2 83-91 Regulation and specificity of antifungal metapleural gland secretion in leaf-cutting ants

Chapter 3 92-102 Towards a better understanding of the evolution of specialized parasites of fungus-growing ant crops

Chapter 4 103-137 Interaction specificity between leaf-cutting ant workers and their vertically transmitted cuticular Pseudonocardia bacteria before and after challenges with Escovopsis fungus garden parasites

CONCLUSION & PERSPECTIVES 138-143

ACKNOWLEDGEMENTS 144

CURRICULUM VITAE 146 English summary

The life history of social insects, often characterized by high density of related individuals, increases their susceptibility to parasitism. This thesis focuses on the Acromyrmex leaf-cutting ant, a social insect species that also engages in an obligate mutualism with fungal crops. The ants provide substrate and protections to their fungal crops and in return receive food and nutrition from the fungus. Leaf-cutting ants not only require efficient defence strategies to protect themselves, but also need defence mechanisms that keep their fungal crops free from parasites. This PhD thesis explores the complementary layers of defence, covering (1) innate immunity, (2) antimicrobial glandular secretions, (3) overview of specialized parasites, and (4) antibiotic-producing bacteria of the leaf- cutting ants.

Innate immunity is conserved across insects and vertebrates. Heightened parasite pressures from social living and fungiculture are expected to leave distinct signatures in the innate immune system. These signatures can be studied with differential gene expression analysis. Chapter 1 uses a genome-wide analysis of differential immune-related gene expressions, demonstrating for the first time how innate immunity responds differentially to fungal infections on ants, and to parasitism on the ants’ fungus garden. Fungal infections on ants induce up-regulations of all components of immune pathways (Toll and Imd), whereas parasitism on fungus garden did not, implying that ants utilize different genetic mechanisms for self protection and defences against parasitism on fungus garden.

One defence mechanisms unique to ants is the metapleural gland that produces antimicrobial secretions. This paired gland was a key innovation that allowed ants to diversify while nesting in microbe-ridden terrestrial environments. The leaf-cutting ants have well-developed metapleural glands, and the secretions are hypothesized to maintain ants’ hygiene. Previous studies on Atta leaf- cutting ants show that metapleural glands are not only important in ants’ protection but also important in protecting the fungus garden against parasitism. Chapter 2 demonstrates the ability of Acromyrmex leaf-cutting ants to regulate the glandular secretions in a context dependent manner, and that their glandular secretions are predominantly for ants’ protection, not fungus garden.

The generational propagation (i.e. vertical transmission) of ants’ fungus garden increases the fungal crops’ susceptibility to parasitism. Escovopsis, a specialized parasite of leaf-cutting ants have

1 interacted and partly co-evolved with ant-fungus mutualism for ca. 50 million years. Chapter 3 reviews recent progress in understanding the patterns of specificity of this ant fungus garden- parasite association, using a modified version of Tinbergen’s four categories of evolutionary questions. The review identifies prospective areas for future studies, emphasizing use of next generation sequencing (NGS) technologies and genomic tools to answer some of the unresolved questions in ant fungus garden-parasite dynamics.

Leaf-cutting ants recruit antibiotic-producing bacteria into their medley of defences, to the extent of having specialized morphological adaptations on their cuticle to house these antibiotic-producing bacteria. The diversity of the ant hosts/bacteria associations on the ants has been subject of controversy, with one argument emphasizing long-term co-evolutions of ant hosts/bacteria associations, and another argument emphasizing dynamic recruitments of antibiotic-producing bacteria from surrounding environments. Chapter 4 used a cross-fostering approach to examine ant hosts/bacteria partner fidelity and their manifestations in prophylactic defences against Escovopsis parasitism. The study revealed no ant host/bacteria genotype-genotype differences in bacteria colonization on the ants, but showed differences in prophylactic defence action, implying that although non-native bacteria are able to colonize ants’ cuticle, native bacteria were more efficient in defences against Escovopsis parasitism.

The constant parasitism pressure from social living and fungiculture have shaped diverse and complementary defences in leaf-cutting ants. From genes to phenotypes, the four chapters of the thesis highlight several components of these defence strategies unique to leaf-cutting ants.

2 Dansk resumé

Sociale insekters livshistorie er ofte kendetegnet ved høj tæthed af beslægtede individer, hvilket øger modtageligheden overfor parasitisme. Denne afhandling fokuserer på det sociale insekt bladskærermyren Acromyrmex, der lever i en obligat mutuallisme med dyrkede svampe. Myrerne leverer substrat og beskyttelse til deres svampeafgrøder, i retur forsyner svampen myrene med mad og næring. Bladskærermyrer har udover effektive forsvarsstrategier til at beskytte dem selv, også forsvarsmekanismer der holder deres svampeafgrøder fri for parasitisme. Denne PhD-afhandling udforsker disse komplimenterende forsvar og omhandler (1) den medfødte immunitet, (2) antimikrobielle kirtelsekretioner, (3) et overblik over specialiserede parasitter samt (4) antibiotikaproducerende bakterier hos bladskærermyrerne.

Medfødt immunitet, eller cellulær immunitet, er bevaret indenfor både insekterne og hvirveldyrene. Social livsstil og dyrkning af svampekultur medfører en forhøjet mængde af parasitter som forventes at have særskilte signaturer i det medfødte immunforsvar. Signaturerne kan studeres med differrential genekspressionsanalyse. Kapitel 1 gør brug af en genom analyse af differentiel immunrelateret genekspression og demonstrerer for første gang hvordan medfødt immunitet reagerer forskelligt på svampeinfektioner af myrer og parasitisme af myrenes svampehave. Svampeinfektioner på myrer forårsager en opregulering af alle komponenter af immunvejene (Toll og Imd), dette gør parasitisme af svampehaven derimod ikke. Dette indebærer, at myrer bruger forskellige genetiske mekanismer for selvforsvar, og forsvar af parasitisme af svampehaven.

En af forsvarsmekanismerne der er unik for myrer er metapleuralkirtlen, som producerer antimikrobielle sekreter. Denne parrede kirtel betragtes som en nøgleinnovation der muliggjorde myrenes diversifikation med reder i mikrobebefængte terrestriske miljøer. Især bladskærermyrerne har veludviklede metapleuralkirtler og det menes at sekreterne har spillet en rolle i myrenes hygiejniske pleje. Tidligere studier af bladskærermyrerne Atta har vist, at metapleuralklirtler ikke kun var vigtige for myrenes beskyttelse, men også i beskyttelsen af svampehaven mod parasitisme. Kapitel 2 demonstrerer Acromyrmex’s evne til at regulere sine kirtelsekreter afhængigt af miljø og at disse kirtelsekreter hovedsageligt er for myrenes egen beskyttelse, ikke svampehavens.

Nedarvningen af svampehaven over generationer (d.vs. den vertikale transmission) øger svampeafgrødernes modtagelighed overfor parasitisme. I over ca. 50 millioner år har

3 bladskærermyrerne interageret, og til dels udviklet sig sammen den specialiserede parasit Escovopsis. Kapitel 3 giver et overblik over de seneste fremskridt i forståelsen af specificiteten af denne svampehave-parasit association ved brug af en tilpasset udgave af Tinbergen’s fire kategorier af evolutionære spørgsmål. Litteraturstudiet identificerer potentielle indsatsområder for fremtidige studier, og understreger brugen af ”next-generation” sekventering (NGS) teknologier samt genom værkøjer til at besvare nogle af de endnu uløste spørgsmål vedrørende dynamikkerne mellem svampehave og parasit.

Bladskærermyrer rekrutterer antibiotikaproducerende bakterier ind i deres potpourris af forsvar, at de har udviklet specialiserede morfologiske tilpasninger i deres kutikula til at huse disse bakterier. Diversiteten af myrevært/bakterie associationer på myrene har været et kontroversielt emne, hvor én skole argumenterer for langvarig coevolution af myrevært/bakterie associationer, og en anden skole har lagt vægt på den dynamiske rekruttering af antibiotikaproducerende bakterier fra det omgivende miljø. Kapitel 4 bruger en ”cross-forstering” tilgang til at undersøge myrevært/bakterie troskab og dennes manifestation i profylaktiske forsvar mod Escovopsis-parasitisme. Studiet viste ingen tegn på myrevært/bakterie genotype-genotype forskelle i bakteriernes kolonisering af myrerne, men viste dog forskelle i hvordan det profylaktiske forsvar blev anvendt. Sidstnævnte indebærer, at omend fremmede bakterier er i stand til at kolonisere myrenes kutikula, så er hjemmehørende bakterier mere effektive i forsvaret mod Escovopsis-parasitisme.

Det konstante parasitisme tryk i forbindelse med en social livsstil og svampekulturer har skabt et diverst og komplimenterende sæt af forsvarsmekanismer hos bladskærermyrerne. De fire kaptiler i denne afhandling fremhæver flere af disse komponenter, fra gener til fænotyper, som er unikke for bladskærermyrerne.

4 Chinese summary

社会性昆虫的生活史中重要的特征是聚集大量的紧密联系的个体组成社会群体，从而提高对寄生生物的抵抗力。本文研究的主要对象是切叶蚁，它是一种与真菌互利共生的社会性昆虫。在这一共生系统中，真菌为蚁群提供食物和养分，而切叶蚁为真菌的提供生长基质，同时切叶蚁高效的防御策略和防御机制在保护它自己的同时也保护了真菌免受寄生生物的侵害。本博士论文将从多个层次探讨切叶蚁的防御系统，包括切叶蚁的自然免疫功能，抗菌腺分泌物，专性寄生生物以及切叶蚁与细菌结合体。

由于自然免疫或者细胞免疫功能在昆虫和脊椎动物中是很保守的，所以群居生活和真菌生长过程中的大量寄生物会给免疫系统带了强大的压力，因而会在免疫系统中留下痕迹。这些痕迹可以通过检测基因的表达变化的方法来进行分析。第一章采用全基因组范围中免疫相关基因的差别表达分析方法，第一次比较了蚁群和真菌共生系统中被寄生生物感染与未被感染的菌圃中两类群蚂蚁免疫系统响应差别。比较结果发现，未被感染的真菌和蚁群共生诱导蚂蚁免疫通路Toll和Imd中所有组成部分的基因表达上调，而寄生生物入侵的真菌苗圃中并没有这种现象。这意味着切叶蚁在与真菌共生中的自我保护和防御寄生物生物时利用的遗传学（基因调控）机制是不同的。

蚂蚁独特的防御系统之一是具有能分泌抗生素的元胸膜腺。作为一种关键的进化特征，这种成对的腺体能让蚂蚁在不同微生物环境中筑巢生活。尤其是切叶蚁，它具有高度发达的元胸膜腺，而且有假说认为它的元胸膜腺的分泌物维持了蚂蚁的清洁卫生。之前的研究也发现切叶蚁的元胸膜腺不仅对蚂蚁本身的保护机制很重要，同时它还能够保护菌圃免受寄生生物的侵害。第二章主要讨论切叶蚁不同的情况下调节元胸膜腺的分泌的能力，结果发现切叶蚁元胸膜腺分泌物主要是对切叶蚁本身起到保护作用而不是对真菌苗圃。

菌圃中蚂蚁的世代遗传增加了真菌对寄生物的敏感性。Escovopsis是专性寄生于切叶蚁真菌苗圃中的寄生生物，它与切叶蚁真菌共生系统具有一定程度上交互作用并协同进化了大约5 亿年前。第三章围绕修订的Tinbergen’s的四类进化问题，综述了对理解这种菌圃- 寄生生物复合系统的寄生生物专性格局方面的研究进展。本综述阐明了未来研究的重点，并强调下一代基因组测序技术和基因组研究工具在未来研究中的重要性。

切叶蚁的混合防御系统中还包括了寄生于蚂蚁并能产生抗生素的细菌。切叶蚁自身通过外皮在一定程度上的特化来适应这种细菌的生存。蚂蚁中宿主和寄生细菌的专一性一直是一个具

5 有争议的话题。一些学者认为，宿主和细菌长期协同进化的结导致细菌专一寄生在宿主上，而另一些学者则认为是蚂蚁能在环境中动态补充不同的细菌。第四章采用cross- fostering方法研究宿主和拮抗细菌的协作紧密度以及他们在预防寄生生物Escovopsis中的表现。本研究揭示了宿主/细菌基因型在不同来源的细菌群落中没有差异，但是在蚂蚁防御系统中却存在差别。这暗示着尽管外来细菌可以在蚂蚁表皮生存，但是本地菌落在防御系统中抵抗Escovopsis的效用更高。

综上所述，来自与群居生活和真菌共生中持续的外来寄生生物的入侵压力使得切叶蚁形成了非常多样和互补的防御体系，以上四章研究内容由基因到表型系统研究了切叶蚁防御系统的不同组成部分。

6 SYNOPSIS

A typical habitat for Acromyrmex leaf-cutting ants, in Costa Rica

Photo credit: Janni Larsen

Disease Challenges & Defences in Leaf-Cutting Ants

SYNOPSIS

1. Parasite-host interactions Parasitism is a common mode of life, and has evolved multiple times over the course of evolution (Combes 2001). Parasite-host interactions can be viewed as a special case of predator-prey interactions where the parasite benefits at the expense of their host (Raffel et al. 2008). The parasites employ numerous strategies to reduce their hosts’ fitness, such as parasitic castration (Lafferty & Kuris 2009), modification of host behaviour (Poulin 2010), and sometimes the effect of parasite-host interactions modifies food web structure of their hosts (Hernandez & Sukhdeo 2008). A classic example of parasitic castration was documented in flukes (Microphallus pseudopygmaeus) that chemically castrates its snail host (Onoba aculeus), resulting in the castrated snail becoming larger than normal (Gorbushin & Levakin 1999). In response to parasitism, host defence strategies are as numerous, exemplified by the documented behavioural, morphological, and chemical defences (for example, see Gross 1993 for a review of defence strategies in herbivorous insects against parasitoids). Long-term parasite-host interactions often lead to adaptations and counter- adaptations against each other, resembling an evolutionary arms race (Dawkins & Krebs 1979). In evolutionary arms races, parasites evolve to optimize host exploitation, whereas hosts evolve to minimize parasite-induced fitness loss (Ebert & Hamilton 1996). Most parasites hold the advantage of having faster generation time than their hosts, resulting in hosts having fewer chances to adapt than their parasites do over evolutionary time (Kaltz & Shykoff 1998). Despite these advantages, the outcomes of parasite-host arms races do not necessarily only favour the parasites, with co-accomodation between parasite-host that may or may not end up as co-speciation between species (Brooks 1987). Parasite-host arms races lead to sustained oscillations in gene frequencies for both parasite and host populations (Dawkins & Krebs 1979). As parasites track host gene frequencies over long periods of time, such interactions often leads to the phenomenon of parasites being better adapted at infecting sympatric rather than allopatric host populations (Kaltz & Shykoff 1998). This phenomenon can be explained with the Red

Disease Challenges & Defences in Leaf-Cutting Ants

Queen hypothesis (Lively 1996) where the dynamics of sustained oscillations in gene frequencies are reminiscent of the Red Queen’s remark to Alice in Through the Looking Glass (Carroll 1872) - “it takes all the running you can do, to keep in the same place”. The central implications of the Red Queen hypothesis is that hosts can keep abreast of faster evolving parasites by having genetically heterogeneous offspring through sexual reproduction, thus creating new defence mechanisms (Hamilton 1980; West et al. 1999).

2. Susceptibility to parasitism in social insects Social insects are among the most dominant and prolific life-forms on earth (Wilson 1990). The most familiar examples of social insects are ants, bees, and wasps (order Hymenoptera), and termites (order Isoptera). Recently, ambrosia beetles Austroplatypus incompertus (Kent & Simpson 1992), gall-making aphids (Stern & Foster 1996), and thrips (Chapman et al. 2002) have also evolved sociality. The classical social insects such as ants and termites often form large colonies with thousands of colony members, and are characterized by overlapping generations, cooperative care of young and reproductive division of labour, which often includes sterility or near sterility of the overwhelming majority of colony members (Wilson 1971). Social living increases foraging efficiency, cooperative brood care, colonizing and competitive abilities, defence from enemies, and the ability to adaptively modify the environment (Rosengaus & Maxmen 2010; Wilson 1975). On the downside, social living increases the vulnerability to parasitism, due to two factors: (1) physical proximity of individuals increase the chance of encountering parasites introduced by other colony members, and (2) high colony genetic relatedness and susceptibility characteristics facilitates the spread of parasites within the colony (Schmid-Hempel 1995). Indeed, social insects are often host to a wide variety of parasites, ranging from viruses to parasitoids which cause a multitude of symptoms that decrease the fitness of the colony in one way or another (for a review of parasites in social insects, see Schmid-Hempel 1995).

Disease Challenges & Defences in Leaf-Cutting Ants

3. Evolutionary mitigations of parasitism in social insects To mitigate against the harmful effects of parasites associated with social living, social insects employ means that increase morphological- and genetic-heterogeneity as defence mechanisms, partly described by the Red Queen hypothesis (see 1. parasite-host interactions). There are two well documented ways for social insects to increase heterogeneity in their colony: (1) multiple matings that generate genetic heterogeneity (Tooby 1982; Sherman et al. 1988), and (2) division of labour and/or caste specialization that generate morphological heterogeneity (Hölldobler & Wilson 1990; Keller 1995; Tooby 1982). Multiple matings are widespread taxonomically among social insects, commonly found in many lineages in the order Hymenoptera where the majority of social insects (ants, bees, and wasps) reside (Hughes et al. 2008). For multiple matings to evolve in response to parasitism, the predicted effect should be a reduction of parasitism due to increased genetic heterogeneity among offspring and an associated fitness gain (Tooby 1982; Sherman et al. 1988). This has been nicely illustrated in the bumblebee Bombus terrestris where the intensity and prevalence of the most common parasite trypanosome (Crithidia bombi) were documented to decrease with the increasing levels of colony genetic heterogeneity (Baer & Schmid-Hempel 2001). Morphological heterogeneity can defend against parasites by “expressing different phenotypes, as the result of developmental plasticity and this may interfere with parasitism in much the same way as tissue differentiation does˝ (Tooby 1982, p. 573). Caste specialisation not only increases morphological heterogeneity, but individuals of different castes might also differ in physiological attributes such as metabolic rate (Calabi & Porter 1989), hormonal titers (Suzzoni et al. 1980), and behavioural specialisation (Oster & Wilson 1978) that may further influence susceptibility to specific parasites. An example of behavioural specialisation was found in Atta leaf-cutting ants, where hitch- hiking of small workers on leaf-fragments carried by large workers significantly reduces the success of parasitic phorid flies depositing eggs in the head capsule of the large workers (Feener & Moss 1990).

Disease Challenges & Defences in Leaf-Cutting Ants

4. Leaf-cutting ants and their susceptibility to parasitism Leaf-cutting ants are a subgroup of the fungus-growing ants (family Formicidae, tribe Attini) in the genera Atta and Acromyrmex that comprises of species that cut and harvest living plants. They are distributed in the New World, approximately between latitudes 33˚N and 44˚S (Weber 1982). The leaf-cutting ants harvest fresh leaves as substrates for their obligate mutualist fungus garden, which is the major food source for the ants (Weber 1972) and in exchange the ants provide the fungus garden with nourishments as well as protection from garden pathogens (Currie et al. 1999; Mueller et al. 2005; figure 1). Leaf-cutting ants have clonally propagated the same fungal lineage belonging to the family Lepiotaceae (Agaricales: Basidiomycota) for at least 23 million years (Chapela et al. 1994). During nuptial flight, the foundress queen carries a tiny piece of fungal inocula in her mouth from her natal nest that she will use to start her own fungiculture in a newly founded colony. This mode of propagation ensures partner fidelity between leaf- cutting ants and their fungal cultivar (Weber 1972). The disadvantage of clonal propagation is the lack of genetic heterogeneity. As a result, the leaf-cutting ants fungal cultivar is more susceptible to parasites and pathogens compared to other modes of propagation, like sexual propagation (Hamilton et al. 1990). The leaf-cutting ants’ fungiculture is analogous to human mono-agriculture system, whereby large and genetically homogenous crops are extremely susceptible to devastating epidemics (Mueller 2002). Indeed, numerous parasites, either saprophytic microfungi (Rodrigues et al. 2008) or specialized parasite (Currie et al. 1999) have been found to be prevalent and negatively impacting the fungus garden of leaf-cutting ants.

11 Ants housed antibiotic- secreting bacteria on their cuticle for defence against Chapter 4: ant host/bacteria Escovopsis parasitism delity and defence against Escovopsis parasitism

Chapter 1: innate immunity and prophylactic behaviour of leaf-cutting ants Black yeast parasitized the antibiotic-secreting bacteria on the ants cuticle

Ants harvest fresh leaves as substrates for their Specialized fungus garden parasite obligate mutualist Escovopsis directly consume fungal cultivar the fungus garden

Chapter 3: review of Escovopsis

Chapter 2: metapleural glandular regulations of ants against parasitism

Figure 1: An illustration of leaf-cutting ant-fungus mutualisms and defence against parasitism. Black texts explains the dynamics, whereas red texts indicates the ant-fungus defence components that were being examined in this PhD thesis. Illustrations modi ed from Caldera et al. 2009.

Disease Challenges & Defences in Leaf-Cutting Ants

5. Defences against parasites in social insects Parasitic defence mechanisms in social insects include the innate immune system in individuals, mechanisms of social immunity that often involve allo-groomings (Cremer & Sixt 2009), the application of glandular and/or bacteria-derived antibiotics, and behavioural parasite removal (reviewed in Cremer et al. 2007). Due to the vulnerability of the fungal cultivar, the leaf-cutting ants not only have to defend their colony but also their fungus garden from parasitism, thus requiring more comprehensive defence mechanisms. This thesis aims to examine the interplay of parasite defence mechanisms in Acromyrmex leaf-cutting ants, focusing on the innate immune system (chapter 1), glandular secretions (chapter 2), the dynamics of specialized fungus garden parasites (chapter 3) and bacteria-derived protection of the ants (chapter 4). The below sections start with a general overview of defence mechanisms in social insects, followed by a more focused overview on leaf-cutting ants, the model system of this PhD thesis, and conclude with how results from my experimental work contributes to the current understanding of defence mechanisms in ant-fungus mutualisms.

5.1 Innate immunity in social insects Innate immune pathways of insects share an overall architecture and specific orthologous components with the innate immune system of vertebrates (Beutler 2004), suggesting a shared root of these immune pathways over hundreds of millions of years (Vilmos & Kurucz 1998). Hence, it’s not surprising that many parallels were found between the innate immune response of insects and vertebrates, such as secretion of antimicrobial peptides, phagocytosis, melanization, and the enzymatic degradation of pathogens (Hultmark 2003). Social insects, contrary to solitary insects, have different selection pressures because of their collective effort against parasitism (social immunity; Cremer et al. 2007) and their vastly different life-histories (Boomsma et al. 2005). Innate immunity of social insects receives little attention despite extensive studies on other aspects of social immunity (see Cremer et al. 2007). Although correlations of social living and heightened parasite pressure with changes in innate immunity have been

Disease Challenges & Defences in Leaf-Cutting Ants documented, the mechanism by which parasitic pressure affects innate immune pathways is unclear. For example, honeybee Apis mellifera have fewer immune genes compared to solitary insects (Weinstock et al. 2006; Evans et al. 2006), suggesting reliance on social effort in defence mechanisms rather than individual innate immune system. Also, several immune genes in ants and honeybees showed higher rates of evolution as measured against fruit fly Drosophila immune genes (Viljakainen et al. 2009), consistent with the heightened parasite pressures expected in social insects. A survey of innate immunity of Atta cephalotes (Suen et al. 2011) and Acromyrmex echinatior leaf-cutting ants (Lumi Viljakainen, personal communications) revealed intact immune signaling pathways, with fewer immune response genes compared to solitary insects, the results resembling what is already known from the honeybee Apis mellifera’s innate immune system (Evans et al. 2006). However, the leaf- cutting ants’ innate immune system would have to contend with heightened parasite pressure from both social living, and clonal fungus garden (see 4. Leaf-cutting ants and their susceptibility to parasitism). Surveys of innate immune pathways gives an indication of the selection signatures that are associated with social evolution and heightened parasite pressure, but only the study of differential gene expression (transcriptomics) would tell us how genes relate to the innate immunity are being expressed. Taking advantage of the A. echinatior leaf-cutting ant genome (Nygaard et al. 2011), chapter 1 uses a next generation sequencing technique - RNA-Seq technologies (Wang et al. 2009; see 6. Methodologies and techniques) to examine how innate immunity genes are differentially expressed in response to parasitism on social living and parasitism on the clonal fungus garden respectively, adding insights into gene-level expressions on the ant-fungus mutualism defence mechanisms.

5.2 Glandular secretions and applications of antimicrobial compounds Aside from the genetic components of innate immunity, antimicrobial glandular secretions constitute most common forms of external defence among terrestrial insects, particularly well developed in the order Hymenoptera where many social insects are

Disease Challenges & Defences in Leaf-Cutting Ants found (Whitman et al. 1990). An ant worker can have more than 20 major glands over her body, essentially a walking glandular battery (Billen 1991) with the secretions consisting of extraordinarily complex mixtures of natural products (Blum & Brand 1972). Although studies on social insect glands mostly focus on the aspects of chemical communications (e.g. Wilson 1965; Jackson & Morgan 1993; Ayasse et al. 2001), in many cases it is likely that glandular secretions originally functioned as defensive compounds and that their communicative function developed secondarily (Blum & Brand 1972). Among the numerous exocrine glands, ants have a unique pair of metapleural glands that primarily produces antimicrobial secretions effective against diverse microbes such as yeast, bacteria, and fungi (reviewed in Yek & Mueller 2010). Homologues of metapleural glands are unknown from other insect lineages, hence its unique presence and antimicrobial properties suggests a critical role of this gland in the ants’ origination and ecological success (Hölldobler & Wilson 1990). Metapleural glands were present ancestrally (Bolton 2003) and subsequently and independently lost in diverse lineages (reviewed in Yek & Mueller 2010). These losses occur most frequently in males, arboreal nesting lineages, and social parasites, all of which are consistent with reduced requirement for hygienic defences, suggesting a correlated association of parasite pressures in the functional diversification of this gland in ants. The metapleural gland secretions of leaf-cutting ants consist mainly of acids of various chain lengths (Ortius-Lechner et al. 2000) displaying a general antimicrobial activity against various fungi and bacteria (Bot et al. 2002). The general antimicrobial effects also extend to keeping fungus-gardens free from parasites (Powell & Stradling 1986). For dispersion of glandular secretions, leaf-cutting ants groom the metapleural gland opening with their legs and then apply to contaminated areas (Fernendez-Marin et al. 2003), and the rate of dispersion can be adjusted by increasing the grooming rate during microbial infection (Fernández-Marín et al. 2006). Having the prior findings of grooming-regulated glandular dispersion, Chapter 2 addresses the questions of internal regulation of glandular secretions during microbial infections. My experimental results indicated internal regulation in quantity and different sensitivities to fungal species in the chemical components of the glandular secretions. Chapter 2 complements studies of external regulation (Fernández-Marín et al. 2009)

Disease Challenges & Defences in Leaf-Cutting Ants providing further evidence in the possible trade-offs in glandular usage and bacteria- derived antibiotics between two closely related attine ants (Acromyrmex and Atta) that shared similar fungus-farming mutualisms, and yet differed in several important aspects of their defence mechanisms against parasitism (Fernández-Marín et al. 2009).

5.3 Bacteria-derived antibiotics Bacteria-derived antibiotics are commonly used in human medicine and agricultural application against parasites and pathogens (Clardy et al. 2006). Insects appear to have ‘discovered’ this mode of protection far earlier than humans have. However, it is only in recent times that bacteria-derived antibiotics in insects came to light. Antibiotic-producing bacteria had been found to be associated with fungus-growing ants (Currie & Scott 1999), European beewolves (Kaltenpoth et al. 2005), and southern pine beetles (Scott et al. 2008). All of these examples show long-term associations between the hosts and their antibiotic-producing bacteria to the extent that hosts have specially adapted morphology to house and provide nutrition to their bacteria (Currie et al. 2006). In human medicine and agricultural practices, the efficacy of the bacteria-derived antibiotics is continuously being threatened by the emergence of resistant parasite strains – superbugs (Davies 2007) that arise due to selective pressure imposed by the applied antibiotics (Cohen 2000). Similarly, the target parasites would evolve resistant against the naturally occurring antibiotic producing-bacteria, and that would create selective pressure on the bacteria for the evolution of novel defences. However, the presence and impact of such resistant parasites have not been evaluated in natural systems. The ant-fungus mutualisms of leaf-cutting ants are parasitized by microfungi in the genus Escovopsis that attack and consume the ants’ fungus garden (Hypocreales: Ascomycota) (Reynolds & Currie 2004; reviewed in Yek et al. 2012 aka. Chapter 3). To protect the fungus garden against Escovopsis parasitism, leaf-cutting ants used a wide array of defensive strategies (reviewed in Yek et al. 2012), with bacteria-derived antibiotics as one of the major defence components. The leaf-cutting ants have specialized morphological adaptations on their cuticles that house antibiotic-producing

Disease Challenges & Defences in Leaf-Cutting Ants filamentous bacteria (Actinomycetales: Pseudonocardiaceae), and in return the antibiotic- producing bacteria produce antibiotics that target the Escovopsis parasitism on the fungus garden (Currie et al. 2003). The long-term antagonistic interactions between Escovopsis parasitism and Pseudonocardia bacteria defence likely resulted in the Pseudonocardia-derived antibiotics evolving in parallel with Escovopsis parasites. In turn, the antibiotics derived from Pseudonocardia against Escovopsis are expected to impose a selective pressure on Escovopsis to evolve resistance. In Chapter 4, we examined Pseudonocardia’s association with their ant hosts, by examining specificity of the ant host/Pseudonocardia genotype association and the antagonistic interactions between Escovopsis parasitism and Pseudonocardia defence. The genotype by genotype association was found to be stable in the study populations, but other studies suggest that the ants readily acquire different bacterial symbionts from the environment potentially with novel antibiotic metabolites (Mueller et al. 2008; Haeder et al. 2009). Using a cross-fostering approach, we show that the host/symbiont genotype does not affect bacterial growth on the ant, but does affect the defence efficiency against Escovopsis parasitism, implying a stable ant host/Pseudonocardia genotype association that could not be measured by the rate/pattern of bacterial colonization on the ant hosts but quantifiable through the prophylactic defences against Escovopsis parasitism.

6. Methodologies and techniques Scientific process involves formulation of questions, making conjectures (hypotheses), or predictions, testing and analysis (Jarrard 2001). This PhD thesis addresses questions under the umbrella of disease challenges and defences in leaf-cutting ants-fungus mutualism, spanning differential genes expression in innate immunity (chapter 1), mechanistic regulation in antimicrobial glandular secretions (chapter 2), evolutionary dynamics in host-parasite interactions (chapter 3) and host/defensive-symbiont fidelities and efficiencies (chapter 4). Methodology, a body of techniques for investigating phenomena, complement the scientific process by acquiring new knowledge or correcting and integrating previous knowledge.

Disease Challenges & Defences in Leaf-Cutting Ants

The breadth of questions addressed in this thesis requires interdisciplinary methodologies. Three interdisciplinary methodologies and techniques are being used extensively in this PhD thesis: 1. a cutting edge next-generation sequencing technique that is quickly replacing the traditional Sanger sequencing technology (Martinez & Nelson 2010), 2. a gold standard microbiology agar diffusion test (Kerr 2004) that has been used since the 1950s (Bauer et al. 1959), and 3. a cross-fostering approach, one of the most elegant techniques to uncover genetic and/or environmental effects on a particular behavioural trait (Ankerl & Pereboom 1974).

6.1 Next-generation sequencing technology Next-generation sequencing (NGS) technology is increasingly being favored over the well-established Sanger sequencing method because NGS technology sequences DNA quickly and cheaply, in short segments. The technology and shorter segments needing to be sequenced reduces the costs and increases capacity at an unprecedented rate (Martinez & Nelson 2010). However, several technical drawbacks such as short read length, lack of paired end reads, and quality problems have limited the effective use of NGS (Martinez & Nelson 2010). Realizing these shortcomings, manufacturers are driven to continuously improve the read length and implement paired end methods, thus allowing even wider applications of NGS in basic research in evolutionary ecology (Hudson 2008). Among the NGS technologies, RNA-sequencing (RNA-Seq) is one of the most complex and fastest growing NGS applications in biology research (Costa et al. 2010). RNA-Seq experiments can accurately determine expression levels of specific genes, differential splicing, and allele-specific expression of transcripts, which in turn address many basic biological and mechanistic questions (Costa et al. 2010), such as photoperiodic regulation in the switch of the reproductive mode in pea aphids (Le Trionnaire et al. 2009), and detection of oral and systemic diseases through saliva sequencing (Palanisamy & Wong 2010). Taking advantage of the improvements and benefits of NGS technologies, Chapter 1 spear-headed the use of RNA-Seq in the study of defence mechanisms on ant-fungus mutualisms, providing the first insights into the differential expressions of innate immune defences in social living and fungus-farming.

18 Paper disk with antimicrobials

Zone of inhibition

Bacteria growth

Resistant to antimicrobials

Figure 2: An example of disk di usion test, where the paper disks were impregnated with several di erent antimicrobial agents to be tested against a strain of bacteria. Clear area (zone of inhibition) was caused by antimicrobial agents inhibiting growth of bacteria, whereas absence of zone of inhibition (ZOI) denotes bacteria being resistant to the antimicrobials agents tested. The size of ZOI is the measure of susceptibility of the bacteria being tested, the larger the ZOI, the more susceptible the bacteria is to the antimicrobial agent being tested.

Disease Challenges & Defences in Leaf-Cutting Ants

6.2 Agar diffusion test The agar diffusion test, also known as Kirby-Bauer disk-diffusion method (Bauer et al. 1959) is an in vitro means of measuring the effect of an antimicrobial agent against bacteria. The bacterium in question is grown evenly over an agar culture plate. Paper disks, impregnated with antimicrobial compounds to be tested, are then placed on the surface of the agar, and antimicrobial compounds will diffuse from the filter paper disks into the agar. If these antimicrobial compounds are effective against the bacteria, no bacteria colonies will grow in the area where the antimicrobial compounds diffuse to. This area is refers to the zone of inhibition (ZOI), and the size of ZOI is used as a measure of the compound’s effectiveness, the larger the ZOI around the paper disks, the more effective the antibiotic compounds are (Bauer et al. 1959; figure 2). In clinical microbiology laboratories, agar diffusion test is the standard method used for performing susceptibility testing (Kerr 2004). Other than this application, agar diffusion tests have also found use in species identification (Sobczak 1985; Rivera- Sánchez et al. 2005), bacteria resistance testing (Fukazawa et al. 1999), and discovery of new antimicrobial compounds (Fuchs et al. 1987; Jones et al. 2010). In chapter 2, we “borrow” the idea of agar diffusion testing to measure the susceptibilities of antimicrobial agents in the ants’ glandular secretions against fungi grown on culture plates. This simple test allowed us to test the sensitivity of antimicrobial glandular secretions against different fungus species, hence informing us on the specificity and regulation of the antimicrobial glandular secretions against fungus infections.

6.3 Cross-fostering approach Cross-fostering approach is a technique where offspring are taken from their genetic parents and reared by unrelated foster parents. This approach is widely used to investigate genetic and/or environmental influences for a particular behavioural trait. If cross- fostered offspring show a behavioural trait similar to their genetic parents and dissimilar from their foster parents, the behaviour is concluded to have genetic basis. In contrast, if the offspring develop behavioural traits dissimilar to their genetic parents and similar to

Disease Challenges & Defences in Leaf-Cutting Ants their foster parents, environmental factors were concluded to play a more dominant role over genetic components (Alcock 2009). In many of the cross-fostering studies, such as mate choice and imprinting (Slagsvold et al. 2002), kin recognition (Mateo & Holmes 2004), and inheritance of alcohol abuse in humans (Cloninger et al. 1981), both genetic and environmental influences were identified. This elegant approach has not escaped the notice of ant’ biologists, as evidenced by a few studies that use the cross-fostering approach to investigate host-symbiont genotype matching in immune defence (Armitage et al. 2011), and genetic and/or environmental influences in nestmates recognition (Van Zweden et al. 2010). Chapter 4 used a cross-fostering approach to test the effect of genotype-genotype fidelity of antibiotic-producing Pseudonocardia bacteria symbionts and their hosts, and how these fidelities affect the prophylactic defences against Escovopsis parasitism.

7. Conceptual frameworks in evolutionary ecology studies A conceptual framework is a type of intermediate theory that attempts to connect all aspects of inquiry (e.g., problem definition, purpose, literature review, methodology, data collection and analysis) (Pettigrew et al. 2001). For two of the chapters in this thesis, conceptual frameworks were deemed useful to complement the experimental and empirical studies. Chapter 2 used a modified immune defence framework in evolutionary ecology (Schmid-Hempel & Ebert 2003) to argue for the presence of an additional cuticular immune system in leaf-cutting ants. Chapter 3 used a well-establish framework in the field of ethology (Tinbergen 1963) to organize, synthesize, conceptualize, and mapped potential study areas that will add most value in the dynamics of specialized parasite-host interactions in the ant-fungus mutualisms.

7.1 The immune defence chart The study of immune defence from an evolutionary ecology perspective resolves along two major conceptual avenues. One avenue explores the costs of immune defence, and the second concentrates on the dynamics of host-parasite interactions (Schmid-Hempel & Ebert 2003). Both avenues are based on the central idea of adaptation through evolution

Disease Challenges & Defences in Leaf-Cutting Ants by natural selection, i.e. fitness maximization principles. However, in spite of the obvious common principle, the two major research avenues have not had too much overlap (Ebert & Hamilton 1996; Gemmill & Read 1998), to the detriment of a comprehensive theory. Schmid-Hempel & Ebert (2003) conceptualized these two avenues into a ‘defence chart’ that attempt to understand the dynamic roles of specific responses within an entire defence cascade and to analyse how costs constrain the evolution of different defence components. The conceptualization of the two main avenues has greatly aided researchers to adopt this unified view in various aspects of immune defence studies, such as prophenoloxidase-activating system in invertebrates (Cerenius & Söderhäll 2004), evolution of innate immune system in worms (Caenorhabditis elegans) (Schulenburg et al. 2004), and testosterone-mediated immune functions and male life histories (Muehlenbein & Bribiescas 2005). Experimental results from chapter 2 and synthesis from previous literature (see Yek & Mueller 2010) led us to consider the adaptations documented in leaf-cutting ants metapleural gland under new light, to view various aspects of metapleural gland defence as components in the ‘defence chart’ (figure 4 in chapter 2). We hope the immune defence chart will motivate future researchers to adopt an evolutionary ecology perspective to the study of the metapleural gland defence system.

7.2 Tinbergen’s four questions Tinbergen’s four questions, named after Nikolaas Tinbergen, are complementary categories of explanations for a behavioural phenomenon. According to Tinbergen’s four questions, an integrative understanding of a behaviour must include both proximate and ultimate analysis, as well as an understanding of both phylogenetic and/or developmental history and the operation of current mechanisms (Alcock 2009). As an example, when asked about the function of sight in animals, one explanation is to help find food and avoid danger (adaptation). Under Tinbergen’s four questions, the function of sight has three additional explanations: sight is caused by a series of evolutionary adaptations over time (phylogeny), sight occurs through the mechanics of the eye and connections to the

Disease Challenges & Defences in Leaf-Cutting Ants brain (causation), and sight is the process of an individual’s development (ontogeny). Although the answers are different, they are all consistent with each other in addressing the question of ‘sight’. Tinbergen’s four questions framework has been used in structuring and understanding a wide range of studies, spanning from human infant crying (Zeifman 2001), parasitoid oviposition decisions (Charnov & Skinner 1985), to making conservation choices (Buchholz 2007). We added another application of Tinbergen’s four questions to the field of parasite-host (Escovopsis-attine ant fungus) interactions. In addition to structuring the different aspects of current understanding on parasite-host interactions, we use the framework to identify the most prospective areas for future studies (Yek et al. 2012).

Disease Challenges & Defences in Leaf-Cutting Ants

8. References

Alcock, J. 2009. Animal Behavior: An Evolutionary Approach 9th Edition, Sinauer Associates, Inc.

Ankerl, G. & Pereboom, D. 1974. Scientific methods in ethology. Science 185: 814.

Armitage, S.A.O., Broch, J.F., Fernández-Marín, H., Nash, D.R. & Boomsma, J.J. 2011. Immune defense in leaf-cutting ants: a cross-fostering approach. Evolution 65: 1791- 1799.

Ayasse, M., Paxton, R.J. & Tengö, J. 2001. Mating behavior and chemical communication in the order Hymenoptera. Annual Review of Entomology 46: 31-78.

Baer, B. & Schmid-Hempel, P. 2001. Unexpected consequences of polyandry for parasitism and fitness in the bumblebee, Bombus terrestris. Evolution 55: 1639-1643.

Bauer, A.W., Perry, D.M. & Kirby, W.M. 1959. Single-disk antibiotic-sensitivity testing of staphylococci: an analysis of technique and results. AMA Archives of Internal Medicine 104: 208-216.

Beutler, B. 2004. Innate immunity: an overview. Molecular Immunology 40: 845-859.

Billen, J.P.J. 1991. Ultrastructural organization of the exocrine glands in ants. Ethology Ecology Evolution 1: 67-73.

Blum, M.S. & Brand, J.M. 1972. Social insect pheromones: their chemistry and function. Integrative and Comparative Biology 12: 553-576.

Bolton, B. 2003. Synopsis and classification of Formicidae. Memoirs of the American Entomological Institute 71: 1-370.

Boomsma, J.J., Schmid-Hempel, P. & Hughes, W.O.H. 2005. Life histories and parasite pressure among the major groups of social insects. In M. Fellowes, G. Holloway, & J. Rolff, eds. Insect Evolutionary Ecology. Royal Entomological Society of London, pp. 139-175.

Bot, A.N.M., Ortius-Lechner, D., Finster, K., Maile, R. & Boomsma, J.J. 2002. Variable sensitivity of fungi and bacteria to compounds produced by the metapleural glands of leaf-cutting ants. Insectes Sociaux 49: 363-370.

Brooks, D.R. 1987. Analysis of host-parasite coevolution. International Journal for Parasitology 17: 291-297.

Disease Challenges & Defences in Leaf-Cutting Ants

Buchholz, R. 2007. Behavioural biology: an effective and relevant conservation tool. Trends in Ecology & Evolution 22: 401-407.

Calabi, P. & Porter, S.D. 1989. Worker longevity in the fire ant Solenopsis invicta: ergonomic considerations of correlations between temperature, size and metabolic rates. Journal of Insect Physiology 35: 643-649.

Caldera, E.J., Poulsen, M., Suen, G. & Currie, C.R. 2009. Insect symbioses: a case study of past, present, and future fungus-growing ant research. Environmental Entomology 38: 78-92.

Carroll, L. 1872. Through the looking glass, and what Alice found there. Macmillan, pp. 224.

Cerenius, L. & Söderhäll, K. 2004. The prophenoloxidase-activating system in invertebrates. Immunological Reviews 198: 116-126.

Chapela, I.H., Rehner, S.A., Schultz, T.R. & Mueller, U.G. 1994. Evolutionary history of the symbiosis between fungus-growing ants and their fungi. Science 266: 1691-1694.

Chapman, T.W., Kranz, B.D., Bejah, K-L., MOrris, D.C., Schwarz, M.P. & Crespi, B.J. 2002. The evolution of soldier reproduction in social thrips. Behavioral Ecology 13: 519-525.

Charnov, E.L. & Skinner, S.W. 1985. Complementary approaches to the understanding of parasitoid oviposition decisions. Environmental Entomology 14: 383-391.

Clardy, J., Fischbach, M.A. & Walsh, C.T. 2006. New antibiotics from bacterial natural products. Nature Biotechnology 24: 1541-1550.

Cloninger, C.R., Bohman, M. & Sigvardsson, S. 1981. Inheritance of alcohol abuse. Cross-fostering analysis of adopted men. Archives of General Psychiatry 38: 861- 868.

Cohen, M.L. 2000. Changing patterns of infectious disease. Nature 406: 762-767.

Combes, C. 2001. Parasitism: the ecology and evolution of intimate interactions; translated by Isaure de Buron & Vincent A. Connors, University of Chicago Press.

Costa, V., Angelini, C., De Feis, I. & Ciccodicola, A. 2010. Uncovering the complexity of transcriptomes with RNA-Seq. Journal of Biomedicine Biotechnology 2010: 853- 916.

Cremer, S., Armitage, S.A.O. & Schmid-Hempel, P. 2007. Social immunity. Current Biology 17: R693-R702.

Disease Challenges & Defences in Leaf-Cutting Ants

Cremer, S. & Sixt, M. 2009. Analogies in the evolution of individual and social immunity. Philosophical Transactions of the Royal Society B Biological Sciences 364: 129-142.

Currie, C.R., Poulsen, M., Mendenhall, J., Boomsma, J.J. & Billen, J. 2006. Coevolved crypts and exocrine glands support mutualistic bacteria in fungus-growing ants. Science 311: 81-83.

Currie, C.R., Mueller, U.G. & Malloch, D. 1999. The agricultural pathology of ant fungus gardens. Proceedings of the National Academy of Sciences of the United States of America 96: 7998-8002.

Currie, C.R. & Scott, J.A. 1999. Fungus-growing ants use antibiotic-producing bacteria to control garden parasites. Nature 398: 701-705.

Currie, C.R., Bot, A.N.M. & Boomsma, J.J. 2003. Experimental evidence of a tripartite mutualism: bacteria protect ant fungus gardens from specialized parasites. Oikos 1: 91-102.

Davies, J. 2007. Microbes have the last word. Molecular Biology 9: 302-317.

Dawkins, R. & Krebs, J.R. 1979. Arms races between and within species. Proceedings of the Royal Society B Biological Sciences 205: 489-511.

Ebert, D. & Hamilton, W.D. 1996. Sex against virulence: the coevolution of parasitic diseases. Trends in Ecology & Evolution 11: 79-82.

Evans, J.D., Aronstein, K., Chen, Y.P., Hetru, C., Imler, J-L., Jiang, H., Kanost, M., Thompson, G.J., Zou, Z. & Hultmark, D. 2006. Immune pathways and defence mechanisms in honey bees Apis mellifera. Insect molecular biology 15: 645-56.

Feener Jr., D.H. & Moss, K.A.G. 1990. Defense against parasites by hitchhikers in leaf- cutting ants: a quantitative assessment. Behavioral Ecology and Sociobiology 26: 17-29.

Fernendez-Marin, H., Zimmermann, J.K. & Wcislo, W.T. 2003. Nest-founding in Acromyrmex octospinosus (Hymenoptera, Formicidae, Attini): demography and putative prophylactic behaviors. Insectes Sociaux 50: 304-308.

Fernández-Marín, H., Zimmerman, J.K., Rehner, S.A. & Wcislo, W.T. 2006. Active use of the metapleural glands by ants in controlling fungal infection. Proceedings of the Royal Society B Biological Sciences 273: 1689-1695.

Fernández-Marín, H., Zimmerman, J.K., Nash, D.R., Boomsma, J.J. & Wcislo, W.T. 2009. Reduced biological control and enhanced chemical pest management in the evolution of fungus farming in ants. Proceedings of the Royal Society B Biological Sciences 276: 2263-2269.

Disease Challenges & Defences in Leaf-Cutting Ants

Fuchs, P.C., Barry, A.L., Jones, R.N., Thornsberry, C., Ayers, L.W., Gavan, T.L. & Gerlach, E.H. 1987. CI-934, a new difluoroquinolone: in vitro antibacterial activity and proposed disk diffusion test interpretive criteria. Diagnostic Microbiology and Infectious Disease 6: 185-192.

Fukazawa, K., Seki, M. & Sugano, K. 1999. Antimicrobial resistance testing of H. pylori epsilometer test and disk diffusion test. Nippon Rinsho 57: 76-80.

Gemmill, A.W. & Read, A.F. 1998. Counting the cost of disease resistance. Trends in Ecology & Evolution 13: 8-9.

Gorbushin, A.M. & Levakin, I.A. 1999. The effect of Trematode parthenitae on the growth of Onoba aculeus, Littorina saxatilis and L. obtusata (Gastropoda: Prosobranchia). Journal of the Marine Biological Association of the United Kingdom 79: 273-279.

Gross, P. 1993. Insect behavioral and morphological defenses against parasitoids. Annual Review of Entomology 38: 251-273.

Haeder, S., Wirth, R., Herz, H. & Spiteller, D. 2009. Candicidin-producing Streptomyces support leaf-cutting ants to protect their fungus garden against the pathogenic fungus Escovopsis. Proceedings of the National Academy of Sciences of the United States of America 106: 4742-4746.

Hamilton, W.D., Axelrod, R. & Tanese, R. 1990. Sexual reproduction as an adaptation to resist parasites. Proceedings of the National Academy of Sciences of the United States of America 87: 3566-3572.

Hamilton, W.D. 1980. Sex versus non-sex versus parasite. Oikos 35: 282-290.

Hernandez, A.D. & Sukhdeo, M.V.K. 2008. Parasites alter the topology of a stream food web across seasons. Oecologia 156: 613-624.

Hudson, M.E. 2008. Sequencing breakthroughs for genomic ecology and evolutionary biology. Molecular Ecology Resources 8: 3-17.

Hughes, W.O.H., Oldroyd, B.P., Beekman, M. & Ratnieks, F.L.W. 2008. Ancestral monogamy shows kin selection is key to the evolution of eusociality. Science 320: 1213-1216.

Hultmark, D. 2003. Drosophila immunity: paths and patterns. Current Opinion in Immunology 15: 12-19.

Hölldobler, B. & Wilson, E.O. 1990. The Ants. Harvard University Press.

Disease Challenges & Defences in Leaf-Cutting Ants

Jackson, B.D. & Morgan, E.D. 1993. Insect chemical communication: pheromones and exocrine glands of ants. Chemoecology 4: 125-144.

Jarrard, R.D. 2001. Scientific methods. Experimental Techniques 22: 236.

Jones, R.N., Ross, J.E. & Rhomberg, P.R. 2010. MIC quality control guidelines and disk diffusion test optimization for CEM-101, a novel fluoroketolide. Journal of Clinical Microbiology 48: 1470-1473.

Kaltenpoth, M., Göttler, W., Herzner, G. & Strohm, E. 2005. Symbiotic bacteria protect wasp larvae from fungal infestation. Current Biology 15: 475-479.

Kaltz, O. & Shykoff, J.A. 1998. Local adaptation in host–parasite systems. Heredity 81: 361-370.

Keller, L. 1995. Parasites, worker polymorphism, and queen number in social insects. American Naturalis 145: 842-847.

Kent, D.S. & Simpson, J.A. 1992. Eusociality in the beetle Austroplatypus incompertus (Coleoptera: Curculionidae). Naturwissenschaften 79: 86-87.

Kerr, J. 2004. Manual of Clinical Microbiology, 8th Edition. Journal of Clinical Pathology 57: 111.

Lafferty, K.D. & Kuris, A.M. 2009. Parasitic castration: the evolution and ecology of body snatchers. Trends in Parasitology 25: 564-572.

Lively, C.M. 1996. Host-parasite coevolution and sex. BioScience 46: 107-114.

Martinez, D.A. & Nelson, M.A. 2010. The next generation becomes the now generation. PLoS Genetics 6: 3.

Mateo, J.M. & Holmes, W.G. 2004. Cross-fostering as a means to study kin recognition. Animal Behaviour 68: 1451-1459.

Muehlenbein, M.P. & Bribiescas, R.G. 2005. Testosterone-mediated immune functions and male life histories. American Journal of Human Biology 17: 527-558.

Mueller, U.G. 2002. Ant versus fungus versus mutualism: ant-cultivar conflict and the deconstruction of the attine ant-fungus symbiosis. The American Naturalist 160: S67-S98.

Mueller, U.G., Dash, D., Rabeling, C. & Rodrigues, A. 2008. Coevolution between attine ants and actinomycete bacteria: a reevaluation. Evolution 62: 2894-2912.

Disease Challenges & Defences in Leaf-Cutting Ants

Mueller, U.G., Gerardo, N.M., Aanen, D.R., Six, D.L. & Schultz, T.R. 2005. The evolution of agriculture in insects. Annual Review of Ecology Evolution and Systematics 36: 563-595.

Nygaard, S., Zhang, G., Schiøtt, M., Li, C., Wurm, Y., Hu, H., Zhou, J., Ji, L., Qiu, F., Rasmussen, M., Pan, H., Hauser, F., Krogh, A., Grimmelikhuijzen, C.J.P., Wang, J. & Boomsma, J.J. 2011. The genome of the leaf-cutting ant Acromyrmex echinatior suggests key adaptations to advanced social life and fungus farming. Genome Research 21: 1339-1348.

Ortius-Lechner, D., Maile, R., Morgan, E.D. & Boomsma, J.J. 2000. Metapleural gland secretion of the leaf-cutter ant Acromyrmex octospinosus: new compounds and their functional significance. Journal of Chemical Ecology 26: 1667-1683.

Oster, G.F. & Wilson, E.O. 1978. Caste and ecology in the social insects. Princeton University Press.

Palanisamy, V. & Wong, D.T. 2010. Transcriptomic analyses of saliva. Methods In Molecular Biology 666: 43-51.

Pettigrew, K.E., Fidel, R. & Bruce, H. 2001. Conceptual frameworks in information behavior. Annual Review of Information Science and Technology 35: 43-78.

Poulin, R. 2010. Parasite manipulation of host behavior: an update and frequently asked questions. Advances in the Study of Behavior 41: 151-186.

Powell, R.J. & Stradling, D.J. 1986. Factors influencing the growth of Attamyces bromatificus, a symbiont of attine ants. Transactions of the British Mycological Society 87: 205-213.

Raffel, T.R., Martin, L.B. & Rohr, J.R. 2008. Parasites as predators: unifying natural enemy ecology. Trends in Ecology & Evolution 23: 610-618.

Reynolds, H.T. & Currie, C.R. 2004. Pathogenicity of Escovopsis weberi: the parasite of the attine ant-microbe symbiosis directly consumes the ant-cultivated fungus. Mycologia 96: 955-959.

Rivera-Sánchez, R., Flores-Paz, R., Hicks-Gómez, J.J. & Arriaga-Alba, M. 2005. A new test of disk-diffusion to help in staphylococci identification. Archives of Medical Research 36: 70-74.

Rodrigues, A., Bacci, M., Mueller, U.G., Ortiz, A. & Pagnocca, F.C. 2008. Microfungal “weeds” in the leafcutter ant symbiosis. Microbial Ecology 56: 604-614.

Disease Challenges & Defences in Leaf-Cutting Ants

Rosengaus, R.B. & Maxmen, A.B. 2010. Disease resistance: a benefit of sociality in the dampwood termite Zootenmopsis angusticollis (Isoptera: Termopsidae). Behavioral Ecology 44: 125-134.

Schmid-Hempel, P. 1995. Parasites and social insects. Apidologie 26: 255-271.

Schmid-Hempel, P. & Ebert, D. 2003. On the evolutionary ecology of specific immune defence. Trends in Ecology & Evolution 18: 27-32.

Schulenburg, H., Kurz, C.L. & Ewbank, J.J. 2004. Evolution of the innate immune system: the worm perspective. Immunological Reviews 198: 36-58.

Scott, J.J., Oh, D-C., Cetin Yuceer, M., Klepzig, K.D., Clardy, J. & Currie, C.R. 2008. Bacterial protection of beetle-fungus mutualism. Science 322: 63.

Sherman, P.W., Seeley, T.D. & Reeve, H.K. 1988. Parasites, pathogens, and polyandry in social Hymenoptera. American Naturalis 131: 602-610.

Slagsvold, T., Hansen, B.T., Johannessen, L.E. & Lifjeld, J.T. 2002. Mate choice and imprinting in birds studied by cross-fostering in the wild. Proceedings of the Royal Society B Biological Sciences 269: 1449-1455.

Sobczak, H. 1985. A simple disk-diffusion test for differentiation of yeast species. Journal of Medical Microbiology 20: 307-316.

Stern, D.L. & Foster, W.A. 1996. The evolution of soldiers in aphids. Biological Reviews 71: 27-79.

Suen, G., Teiling, C., Li, L., Holt, C., Abouheif, E., Bornberg-Bauer, E., Bouffard, P., Caldera, E.J., Cash, E., Cavanaugh, A., Denas, O., Elhaik, E., Favé, M-J., Gadau, J., Gibson, J.D., Graur, D., Grubbs, K.J., Hagen, D.E., Harkins, T.T., Helmkampf, M., Hu, H., Johnson, B.R., Kim, J., Marsh, S.E., Moeller, J.A., Muñoz-Torres, M.C., Murphy, M.C., Naughton, M.C., Nigam, S., Overson, R., Rajakumar, R., Reese, J.T., Scott, J.J., Smith, C.R., Tao, S., Tsutsui, N.D., Viljakainen, L., Wissler, L., Yandell, M.D., Zimmer, F., Taylor, J., Slater, S.C., Clifton, S.W., Warren, W.C., Elsik, C.G., Smith, C.D., Weinstock, G.M., Gerardo, N.M. & Currie, C.R. 2011. The genome sequence of the leaf-cutter ant Atta cephalotes reveals insights into its obligate symbiotic lifestyle. PLoS Genetics 7: 11.

Suzzoni, J.P., Passera, L. & Strambi, A. 1980. Ecdysteroid titre and caste determination in the ant, Pheidole pallidula (Hymenoptera: Formicidae). Experientia Basel 36: 1228-1229.

Tinbergen, N. 1963. On aims and methods in Ethology. Zeitschrift Für Tierpsychologie 20: 410-433.

Disease Challenges & Defences in Leaf-Cutting Ants

Tooby, J. 1982. Pathogens, polymorphism, and the evolution of sex. Journal of Theoretical Biology 97: 557-576.

Le Trionnaire, G., Francis, F., Jaubert-Possamai, S., Bonhomme, J., De Pauw, E., Gauthier, J-P., Haubruge, E., Legeai, F., Prunier-Leterme, N., Simon, J-C., Tanguy, S. & Taguet, D. 2009. Transcriptomic and proteomic analyses of seasonal photoperiodism in the pea aphid. BMC Genomics 10: 456.

Viljakainen, L., Evans, E.D., Hasselmann, M., Rueppell, O., Tingek, S. & Pamilo, P. 2009. Rapid evolution of immune proteins in social insects. Molecular Biology and Evolution 26: 1791-801.

Vilmos, P. & Kurucz, E. 1998. Insect immunity: evolutionary roots of the mammalian innate immune system. Immunology Letters 62: 59-66.

Wang, Z., Gerstein, M. & Snyder, M. 2009. RNA-Seq: a revolutionary tool for transcriptomics. Nature Reviews Genetics 10: 57-63.

Weber, N.A. 1982. Fungus ants. In H. R. Hermann, ed. Social insects Vol. 4. Academic Press, pp. 255-363.

Weber, N.A. 1972. Gardening ants, the attines. The American Philosophical Society.

Weinstock, G.M., Robinson, G.E., Gibbs, R.A., Worley, K.C., Evans, J.D., Maleszka, R., Robertson, H.M., Weaver, D.B., Beye, M., Bork, P., Elsik, C.G., Hartfelder, K., Hunt, G.J., Zdobnov, E.M., Amdam, G.V., Bitondi, M.M.G., Collins, A.M., Cristino, A.S., Michael, H., Lattorff, G., Lobo, C.H., Moritz, R.F.A., Nunes, F.M.F., Page, R.E., Simes, Z.L.P., Wheeler, D., Carninci, P., Fukuda, S., Hayashizaki, Y., Kai, C., Kawai, J., Sakazume, N., Sasaki, D., Tagami, M., Albert, S., Baggerman, G., Beggs, K.T., Bloch, G., Cazzamali, G., Cohen, M., Drapeau, M.D., Eisenhardt, D., Emore, C., Ewing, M.A., Fahrbach, S.E., Fort, S., Grimmelikhuijzen, C.J.P., Hauser, F., Hummon, A.B., Huybrechts, J., Jones, A.K., Kadowaki, T., Kaplan, N., Kucharski, R., Leboulle, G., Linial, M., Littleton, J.T., Mercer, A.R., Richmond, T.A., Rodriguezas, S.L., Rubin, E.B., Sattelle, D.B., Schlipalius, D., Schoofs, L., Shemesh, Y., Sweedler, J.V., Velarde, R., Verleyen, P., Vierstraete, E., Williamson, M.R., Ament, S.A., Brown, S.J., Corona, M., Dearden, P.K., Dunn, W.A., Elekonich, M.M., Fujiyuki, T., Gattermeier, I., Gempe, T., Hasselmann, M., Kage, E., Kamikouchi, A., Kubo, T., Kunieda, T., Lorenzen, M., Milshina, N.V., Morioka, M., Ohashi, K., Overbeek, R., Ross, C.A., Schioett, M., Shippy, T., Takeuchi, H., Toth, A.L., Willis, J.H., Wilson, M.J., Gordon, K.H.J., Letunic, I., Hackett, K., Jane Peterson, J., Felsenfeld, A., Guyer, M., Solignac, M., Agarwala, R., Cornuet, M., Monnerot, M., Mougel, F., Reese, J.T., Vautrin, D., Gillespie, J.J., Cannone, J.J., Gutell, R.J., Johnston, J.S., Eisen, M.B., Iyer, V.N., Iyer, V., Kosarev, P., Mackey, A.J., Solovyev, V., Souvorov, A., Aronstein, K.A., Bilikova, K., Chen, Y.P., Clark, A.G., Decanini, L.I., Gelbart, W.M., Hetru, C., Hultmark, D., Imler, J-L., Jiang, H., Kanost, M., Kimura, K., Lazzaro, B.P., Lopez, D.L., Simuth, J., Thompson, G.J.,

Disease Challenges & Defences in Leaf-Cutting Ants

Zou, Z., De Jong, P., Sodergren, E., Csürös, M., Milosavljevic, A., Osoegawa, K., SRichards, S., Shu, C-L., Duret, L., Elhaik, E., Graur, D., Anzola, J.M., Campbell, K.S., Childs, K.L., Collinge, D., Crosby, M.A., Dickens, C.M., Grametes, L.S., Grozinger, C.M., Jones, P.L., Jorda, M., Ling, X., Matthews, B.B., Miller, J., Mizzen, C., Peinado, M.A., Reid, J.G., Russo, S.M., Schroeder, A.J., St Pierre, S.E., Wang, Y., Zhou, P., Jiang, H., Kitts, P., Ruef, B., Venkatraman, A., Zhang, L., Aquino-Perez, G., Whitfield, C.W., Behura, S.K., Berlocher, S.H., Sheppard, W.S., Smith, D.R., Suarez, A.V., Tsutsui, N.D., Wei, X., Wheeler, D., Havlak, P., Li, B., Liu, Y., Jolivet, A., Lee, S., Nazareth, L.V., Pu, L-L., Thorn, R., Stolc, V., Newman, T., Samanta, M.J., Tongprasit, W.A., Claudianos, C., Berenbaum, M.R., Biswas, S., De Graaf, D.C., Feyereisen, R., Johnson, R.M., Oakeshott, J.G., Ranson, H., Schuler, M.A., Muzny, D., Chacko, J., Davis, C., Dinh, H., Gill, R., Hernandez, J., Hines, S., Hume, J., Jackson, L., Kovar, C., Lewis, L., Miner, G., Morgan, M., Nguyen, N., Okwuonu, G., Paul, H., Santibanez, J., Savery, G., Svatek, A., Villasana, D. & Wright, R. 2006. Insights into social insects from the genome of the honeybee Apis mellifera. Nature 443: 931-949.

West, S.A., Lively, C.M. & Read, A.F. 1999. A pluralist approach to sex and recombination. Science 12: 1003-1012.

Whitman, D., Blum, M., Alsop, D., Evans, D., Schmidt, J. & Brown, J. 1990. Allomones: chemicals for defense. Insect Defenses Adaptive Mechanisms and Strategies of Prey and Predators 482.

Wilson, E.O. 1965. Chemical communication in the social insects. Science 149: 1064- 1071.

Wilson, E.O. 1971. The Insect Societies. Harvard University Press.

Wilson, E.O. 1975. Sociobiology: The New Synthesis. Harvard University Press.

Wilson, E.O. 1990. Success and dominance in ecosystems: the case of the social insects Ecology Institute, CABI, pp. 104.

Yek, S.H. & Mueller, U.G. 2010. The metapleural gland of ants. Biological Reviews 91: 201-224.

Yek, S.H., Boomsma, J.J. & Poulsen, M. 2012. Towards a better understanding of the evolution of specialized parasites of fungus-growing ant crops. Psyche Article ID 239392.

Zeifman, D.M. 2001. An ethological analysis of human infant crying: answering Tinbergen’s four questions. Developmental Psychobiology 39: 265-285.

Van Zweden, J.S., Brask, J.B., Christensen, J.H., Boomsma, J.J., Linksvayer, T.A. & D'Ettorre, P. 2010. Blending of heritable recognition cues among ant nestmates

Disease Challenges & Defences in Leaf-Cutting Ants

creates distinct colony gestalt odours but prevents within-colony nepotism. Journal of Evolutionary Biology 23: 1498-1508.

33 CHAPTER 1

Differential gene expression after multiple disease challenges in leaf-cutting ants

Ants continuously tend the fungus-garden, keeping their cultivar free from parasites

Photo credit: Mark Moffett

Manuscript in preparation

34 Differential gene expression after multiple fungal disease challenges in leaf-cutting ants

Sze Huei Yek, Jacobus J. Boomsma, Morten Schiøtt

Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark

Authors for correspondence: Morten Schiøtt ([email protected])

35 Abstract Social insect in general and leaf-cutting ants in particular have increased selection pressures on their innate immune system due to their social lifestyle and monoclonality of the symbiotic fungus cultivar. As this symbiosis is obligate for both parties, prophylactic behavioral defenses against infections are expected to increase either ant survival or fungus garden survival, but also to possibly trade-off when infections are specific. We examined the effectiveness of prophylactic behaviors and modulations of innate immune defense reactions by a combination of infection bioassays and genome-wide transcriptomic studies (RNA-Seq), using an ant pathogen (Metarhizium brunneum) and a garden pathogen (Escovopsis weberi) and administering infections both directly and indirectly (via the symbiotic partner). Upon detection of pathogen conidia, ant workers responded by increasing both general activity and specific behaviors (self-grooming, allo- grooming, garden-grooming) towards the most relevant targets, independent of the pathogen encountered. This trend was also evident in the patterns of gene expression changes. Both direct and indirect infection with M. brunneum induced a general up- regulation of gene expression, including a number of well-known immune-related genes. In contrast, direct infection of the fungus garden by E. weberi induced an overall down- regulation of ant gene expression whereas indirect infection did not, suggesting that increased activity of ants to remove fungus garden infections is costly and involves trade- offs with the activation of other physiological pathways.

Keywords Innate immunity; symbiosis; mutualism; RNA-Seq, Metarhizium; entomopathogenic fungus; Escovopsis; fungal crop parasite

36 INTRODUCTION Social insects colonies are characterized by dense aggregations of individuals that are generally highly related. These characteristics facilitate the transmission of disease and thus make social insect colonies particularly vulnerable to parasites (Boomsma, Schmid- Hempel & Hughes 2005; Schmid-Hempel 2006). Among the social insects, the fungus- growing ants experience particular disease challenges, because in addition to densely packed and genetically homogeneous colony members, they maintain clonal mutualistic fungus gardens (Mueller et al. 2010) that also need to be kept free of disease (Bot et al. 2002; Poulsen et al. 2002). However, despite this high expected disease pressure, no specialized diseases of these ants are known and, although a wide variety of saprophytic weeds can be found in the fungus garden, only one specialized garden disease, Escovopsis, appears to exist (Currie et al. 2003; Rodrigues et al. 2008; Yek, Boomsma & Poulsen 2012). Ants and termites with long-lived colonies have been under particularly strong selection to keep their nest environments free of disease, which may indirectly have inhibited the evolution of virulent specialized diseases (Boomsma, Schmid-Hempel & Hughes 2005; Hughes, Pierce & Boomsma 2008). However, how ants in general and leaf-cutting ants in particular achieve their highly efficient defenses against pathogens is only partly understood. Several defenses may have helped leaf-cutting ants to combat the spread of germs. First, these ants have unusually large metapleural glands (Hughes, Bot & Boomsma 2010), which have many characteristics reminiscent of an extra cuticular immune system (Yek et al. 2012). Second, most attine ants maintain actinomycete bacteria in specially adapted cuticular crypts and tubercles (Currie et al. 2006), which assist in the control of Escovopsis infections (Currie et al. 2003; Taerum et al. 2007; Yek, Boomsma & Poulsen 2012). Both these defenses are expressed by individual ants, but may become collective social defenses by the ways in which they are applied (e.g. Little et al. 2006), and it is generally believed that collective defenses of this kind have very significant functions across the social insects (Cremer, Armitage & Schmid-Hempel 2007; Cremer & Sixt 2009). Studies of innate immune systems of social insects (e.g. bumblebees, Mallon, Loosli & Schmid-Hempel 2003; honeybees, Weinstock et al. 2006; other ants,

37 Viljakainen & Pamilo 2008) have shown that trade-offs between general and specific immune responses may occur. In leaf-cutting ants, innate phenoloxidase (PO) and ProPO immune responses have been documented (Armitage & Boomsma 2010) in addition to caste- and sex-specific encapsulation responses (Baer et al. 2005; Baer, Armitage & Boomsma 2006). Variation in these responses generally suggests that such defenses are costly so their expression can trade-off with other vital functions in these social insects as well. The innate immune system of insects is generally very efficient in detecting and eliminating potentially harmful microorganisms. Pathogen recognition receptors bind and recognize non-self cells, which leads to the activation of signaling pathways such as Toll and IMD, resulting in the expression of antimicrobial effectors (Lazzaro 2008; Lemaitre & Hoffmann 2007; Levashina 2004). In non-social insects such as Drosophila fruit flies (Christophides et al. 2002) and Anopheles mosquitoes (Levashina 2004) the ample availability of reference genomes has facilitated immune system studies in recent years, and to a lesser extent the same is true for honeybees (Weinstock et al. 2006). Initial analyses of ant genomes revealed that all the components of the innate immune system already known from Drosophila are present (Bonasio et al. 2010; Gadau et al. 2011). Recent work has increasingly focused on the expression of multiple immune-genes responding to either pathogens or endosymbiont (Erler, Popp & Lattorff 2011; Ratzka et al. 2011), but genome-wide expression studies focusing on innate immune responses have not been attempted. The present study exploits the recent publication of a reference genome of the leaf-cutting ant Acromyrmex echinatior (Nygaard et al. 2011; see also Suen et al. 2011 for a reference genome of the related leaf-cutting ant Atta cephalotes) to do a large scale gene expression study by subjecting A. echinatior leaf-cutting ants to two types of pathogens: 1. The generalist fungus Metarhizium brunneum that infects ants, and 2. The specialized fungus Escovopsis weberi that infects fungus gardens. We administered infections both directly (targeting the known host) and indirectly (targeting the mutualistic partner), and did RNA-seq-based transcriptomics of whole ants to assess which genes were up- and down-regulated relative to controls. We supplemented these data by a set of replicate infection experiments to assess the efficiency of infections (the

38 extent to which ants died or not, and the loss of fungus garden mass) and behavioral responses such as self-grooming, allo-grooming, garden-grooming and general level of activity.

MATERIALS AND METHODS Fungal infection experiments Three representative colonies of Acromyrmex echinatior, collected in Gamboa, Panama in 2005 (Ae226B) and 2006 (Ae363 and Ae376), were used in the experiments. These colonies were similar in size, with ca. 2 liter of fungus garden and 100-200 major garden workers being easily available at each of our consecutive bouts of sampling. The colonies were kept under standardized conditions in a climate room at 25˚C and 70% RH in a 12 h light:dark cycle, with a diet of fresh bramble leaves (Rubus fruticosus), apple and dry rice. To avoid age- or caste-specific variation, we used only major garden workers of approximately the same intermediate age class, i.e. we avoided both light-colored callows and very dark old foraging workers. This implied that the ants varied very little in the amount of cuticular actinomycete bacteria, i.e. score 0 – 3 on the scale of Poulsen et al. (2003), typical for large workers that are about to start their forager careers. The entomopathogenic fungus Metarhizium brunneum (Bischoff, Rehner & Humber 2009; formerly M. anisopliae) was chosen as disease agent that has direct negative impacts on A. echinatior workers (Baer et al. 2005; Hughes, Eilenberg & Boomsma 2002; Hughes & Boomsma 2004), whereas Escovopsis weberi, the specialized parasite that attacks the fungal cultivar of the ants, was chosen as disease agent that has direct negative impacts on gardens but not on A. echinatior workers (Reynolds & Currie 2004; Yek, Boomsma & Poulsen 2012). M. brunneum (KVL 04-57) was obtained from the stock collection of the Department of Agriculture and Ecology, Faculty of Life Sciences, University of Copenhagen, whereas the E. weberi strain was isolated from Acromyrmex nests in Gamboa by Hermogenes Fernández-Marín in early 2010, and identified based on morphological characters (Muchovej & Della Lucia 1990). For each colony, we established five sub-colonies in transparent plastic boxes (7.4 X 7.4 X 3.1 cm3) that each contained 10 major garden workers and 1 g of fungus garden. The infection assays were set up as 2 by 2 factorial designs (figure 2f), with: 1. M.

39 brunneum infections on ants (MbAnts), 2. M. brunneum infections on fungus garden fragments (MbFungus), 3. E. weberi infections on ants (EwAnts), 4. E. weberi infections on fungus garden fragments (EwFungus), and 5. Untreated ants and fungus garden (Control). We used a fine paint brush to transfer fungal conidia (asexual spores) either to the ants or onto the fungus garden. For each infection, we used conidia taken from a 1 cm2 piece of pure fungal culture, grown on potato dextrose agar (PDA) medium, which corresponds to 2.1 X 108 conidia/ml for M. brunneum and 9.2 X 107 conidia/ml for E. weberi. The viability of conidia was checked by measuring their germination on plates with the same medium, which produced germination rates of >90% for M. brunneum and E. weberi. From 10 minutes after infections until the second hour, we recorded grooming behaviors and the number of times they occurred for all workers observed in a 3.5 cm diameter field of view under a stereomicroscope (6.5X), producing four observations of 10 minutes each with five minutes intervals between them. We observed four main behaviors (self-grooming, allo-grooming, garden-grooming, and immobility inside the fungus garden). The grooming behaviors were characterized by a sequence of behaviors starting with scraping antennae and legs with the forelegs and licking the forelegs with the mouth parts (Bot et al. 2001), upon which the ants started to remove fungal conidia from either their own body (self-grooming) or the body of a nestmate (allo-grooming). In contrast, “garden-grooming” started with antennating a garden fragment, followed by extending the maxillae and labium to grasp the piece, and lifting it from the fungus garden matrix to pull it through the mouth parts (Bot et al. 2001). Ants were either active or stayed motionless for long periods of time hidden under garden fragments or just sitting on the fungus garden, which we categorized as “immobility inside fungus garden”. Due to the rather low number of occurrences of grooming behavior at each time point, we pooled the 4 observation for subsequent analysis. 48 hours after infections, both A. echinatior workers and the fungus garden fragments were killed with liquid nitrogen and stored at -80˚C, after which RNA was isolated from the A. echinatior ants. To verify whether our infections had indeed realized significant disease pressure, we set up a parallel (i.e. replicate) set of sub-colonies to monitor differential ant survival and possible changes in garden biomass (scored as

40 garden loss) over the course of 10 days after infection. The cadavers of any ants that died during the experiment were surface sterilized with 70% ethanol followed by 96% ethanol to remove surface conidia, placed in petri dishes with wet paper towel and monitored for the appearance of the typical Metarhizium conidiophores on the surface of the ants (see Yek et al. 2012 for further details).

RNA isolation RNA from the experimental A. echinatior workers was isolated using the RNeasy Plant Mini Kit (Qiagen) with modifications from a standard protocol. RNA extractions and sequencing were replicated across the three experimental colonies (Ae226B, Ae363 and Ae376). For each colony and treatment 10 individuals were pooled and used for RNA extraction, giving a total of 15 RNA samples. For consistency, RNA extractions of experimental samples from the same colony were carried out in parallel. Each sample (10 major A. echinatior workers, approximately 100 μg) was transferred to a 2 ml extraction tube that contained 750 μl RLT buffer, 7.5 μl β-mercaptoethanol and 1 ceramic bead. Tubes with ants were disrupted in a Fastprep instrument at level 6 for 30 seconds, followed by centrifugation at 13.500g for 1 minute. The homogenized samples were then transferred to Qiashredder tubes, followed by centrifugation at 13.500 for 2 minutes. The liquid without any precipitate was transferred to a phase lock tube that had been centrifuged at 13.500g for 30 seconds. 700 μl phenol/CHCl3/iso-amy-alcohol (25:24:1) pH 8 was added to the phase lock tube (5 Prime), and the resulting liquid was gently mixed by turning the tubes upside down for 1 minute. The mixtures in the phase lock tube were incubated at room temperature for 3 minutes, after which tubes were centrifuged at 20.000g for 30 minutes in a cooling centrifuge at 20˚C. The upper phase was poured into a clean tube and half the volume 96% ethanol was added, thoroughly mixed and immediately transferred to a Qiagen column. The column was centrifuged at 8000g for 30 seconds and the flow through discarded. 450 μl RW1 buffer was added into the column, centrifuged at 8000g for 30 seconds and flow through discarded. Then, 80 μl DNAseI (10 μl DNAse diluted in 70 μl buffer) was added to the column and incubated at room temperature for 15 minutes, after which 450 μl RW1 buffer was added followed by centrifugation at 8000g for 30 seconds and removal of the flow through. 500 μl RPE

41 buffer was then added, centrifuged at 8000g for 30 seconds and flow through was discarded. Finally, 500 μl RPE buffer was added, centrifuged at 8000g for 2 minutes and the collecting tube with flow through discarded. The Qiagen column was then refitted with a new collection tube and centrifuged at 20.000g for 1 minute, after which the new collection tube was discarded. The Qiagen column was then refitted with a clean

Eppendorf tube, 50 μl RNase free H2O was added onto the column and left to incubate for 1 minute, after which the column was centrifuged at 8000g for 1 minute. The flow through was then transferred back into the column and incubated for 1 minute. Finally, the Qiagen column was centrifuged again at 8000g for 1 minute. The integrity of the RNA sample (collected now in the Eppendorf tube) was confirmed by agarose gel electrophoresis and total RNA quantity and purity were determined spectrophotometrically.

Library preparations and sequencing RNA-seq library construction and sequencing was carried out by BGI Hong Kong as described in Nygaard et al. (2011). The 200bp short-insert libraries were sequenced using Illumina HiSeq 2000 with 90 bp paired-end sequencing.

STATISTICAL ANALYSES Infection assays Survival of inoculated ants over 10 days was analyzed with a proportional-hazards regression model, with colonies, fungal treatments and their interactions as main effects. Post-hoc pairwise differences between colonies and fungal treatments were based on risk- ratio tests, with the significance level being Bonferroni adjusted to correct for multiple comparisons. The proportions of ants sporulating were compared using G-tests of goodness-of-fit for heterogeneity. Differences in the occurence of grooming behaviours were analyzed with one-way ANOVA, testing for differential effects of fungal treatments, with each experimental colony providing a data point consisting of the pooled observations. In the case of overall significance, post-hoc multiple comparison Tukey tests were performed to determine which fungal treatments had a significant effect. All statistical analyses were performed with JMP software (version 9.02, SAS institute).

RNA-Seq analyses Analysis of the RNA-Seq data was carried out with the Tuxedo pipeline (Bowtie, TopHat and Cufflinks programs) available through the web-based bioinformatics platform Galaxy (http://usegalaxy.org/). Raw reads from five experimental samples (MbAnts, MbFungus, EwAnts, EwFungus and Control) replicated over three experimental colonies (Ae226B, Ae363 and Ae376) were first processed with FASTQ Groomer to convert the clean Illumina reads to the FASTQ format (Blankenberg et al. 2010). The reads of respective treatments were then aligned to the A. echinatior genome (Genome build 3.9, Nygaard et al. 2011) using TopHat (Trapnell et al. 2009). We then used Cufflinks to measure the gene expression for the transcripts in the test samples. Due to the presence of multiple treatments and replications, we used Cuffmerge to merge the assemblies generated by Cufflinks. Finally, Cuffdiff identified genes for which the expression level was significantly different in a test sample relative to control (Trapnell et al. 2012). The distributions of significantly differentially expressed genes were compared using chi- square goodness-of-fit test, with the null hypothesis of computed genes from each experimental treatment having the same frequencies. Similarly, we examined the directionality of gene regulations from each experimental treatment with the null hypothesis of equal distribution of number of genes that were up- and down-regulated. The profiles of these significantly differentially expressed genes were visualized with Venn diagrams, drawn using free web-based software (http://bioinfogp.cnb.csic.es/tools/venny/index.html). The correlations of direct and indirect infections between pathogens (Metarhizium and Escovopsis) and targets (ants and fungus garden) were carried out using bivariate equal variance principal component analysis (PCA). The distributions of genes were first plotted to identify outliers. Any

genes with log2 values of > 12 were excluded from the PCA analyses as these substantially weakened the correlations and skewed the directionalities of the pair-wise comparisons, and were unlikely to represent real expression responses.

Functional annotation of immune-related genes

43 We grouped immune-related genes into four categories on the basis of their known or inferred molecular functions: ‘recognition genes’ that encode pathogen surveillance proteins (for e.g. peptidoglycan recognition proteins and Gram-negative binding proteins); ‘signaling genes’ that encode proteins in immune-related signaling pathways (for e.g. Toll, IMD, JAK-STAT and JNK), ‘effector genes’ that encode proteins that directly inhibit pathogen growth and survival (for e.g. antimicrobial peptides), and “stress-related genes” that have known functional roles in responses to external and biotic stimuli (Jesenberger & Jentsch 2002; Ratzka et al. 2011; Sackton et al. 2007). Any such broad functional classification is necessarily somewhat subjective as some proteins could plausibly have been assigned to multiple categories (for instance, some recognition proteins also initiate signal transduction), but this classification serves the purpose of facilitating the overall interpretation of differences in gene expression. The 375 significantly differentially expressed genes that we obtained after Cuffdiff computations (see above) were analyzed with Blast2GO (Conesa & Götz 2008) that performs functional annotation of sequences based on the gene ontology (GO) vocabulary. Genes that have functional notations related to immunity such as “cell death, response to stress, response to external stimulus, response to biotic stimulus” were tentatively assigned to respective immunity roles. ImmunoDB, a database containing immune-related gene families of sequenced insects such as Drosophila melanogaster, Anopheles gambiae, and Aedes aegypti (Waterhouse et al. 2007) was also used to query our 375 differentially expressed genes. Furthermore, we added a list of well-annotated honeybee immune genes (Lumi Viljakainen, personal communications) in our query in the hope of identifying as many immune-related genes as possible in our dataset. These queries were carried out mainly using the BLAST family of search functions (www.ncbi.nlm.nih.gov). The visualizations of immune-related genes in their respective experimental treatments and categories were created using the gplots package in R (http://www.R-project.org).

RESULTS Survival and prophylactic responses to fungal infections

44 The survival of ants (figure 1a) differed significantly between the fungal treatments (Effect likelihood ratio (LR) χ2 = 37.89, df = 4, P < 0.0001), but there was no significant interaction between colonies and fungal treatment type (LR χ2 = 10.89, df = 8, P = 0.2079). Ants infected with M. brunneum consistently suffered higher mortality than control ants and ants infected directly (MbAnts) suffered significantly higher mortality than ants infected indirectly (MbFungus) (Wilcoxon χ2 = 16.16, df = 1, P < 0.0001). On the other hand, mortalities of ants directly or indirectly infected with E. weberi did not differ significantly to those of control ants, confirming previous findings of E. weberi not being harmful to ants (Reynolds & Currie 2004; Yek et al. 2012). Direct infection of E. weberi on the fungus garden (EwFungus) did not cause ants to suffer significantly higher mortality than did indirect infection (EwAnts) (Wilcoxon χ2 = 0.83, df = 1, P = 0.3626). As expected infection treatments with Escovopsis and controls never produced sporulating cadavers, but 100% and 90% of the respective ant cadavers from the direct and indirect infections with Metarhizium produced characteristic conidiophores after 10

days (GHet = 97.60, df = 14, P < 0.0001; figure 1b). Survival patterns differed among the three colonies (LR χ2 = 22.21, df = 2, P < 0.0001; figure 1a). Colony Ae363 appeared to be more robust compared to colonies Ae226B and Ae376B, as Metarhizium-induced mortality came later and with a lower proportion of dead ants sporulating (figure 1a-b).

The occurrences of allo-grooming (F2,4 = 6.00, P = 0.0100), garden-grooming

(F2,4 = 6.99, P = 0.0060), and immobility inside fungus garden (F2,4 = 13.18, P = 0.0005) were significantly different between fungal treatments. Ants spend significantly more time being immobile in the control treatments compared to any of the conidia infection treatments (figure 2a). Ants also performed significantly more allo-grooming when fungal conidia were directly applied on them rather than indirectly via the fungus garden (figure 2b), and they performed more garden-grooming when fungal conidia were applied on the fungus garden rather than on the ants (figure 2c). There were no differences in the

occurrences of self-grooming (F2.4 = 2.36, P = 0.1228; figure 2d) and no significant

changes in garden biomass before and after fungal treatments (F2,4 = 1.18, P = 0.3766; figure 2e), although the infection of gardens induced somewhat elevated losses compared to controls and infections of ants. Thus, both pathogens induced the expected behavioral responses, but the Escovopsis infections were controlled by the ants and caused little or

45 no lasting damage (in terms of garden loss), whereas the Metarhizium infections were 100% lethal for the ants.

Overall gene expressions For each of the 15 RNA-seq samples about 10 million usable reads were generated, and for more than 85 % of these, the sequence and it’s mate pair could be mapped to the genome within an appropriate distance (Table 1). The RNA-seq data were mapped to the genome sequence using Tophat, and the mapping data were subsequently analyzed by Cufflinks to assemble transcripts and to measure gene expression in fragments per kilobase of exon per million fragments of mapped genome sequence (FPKM), a normalized measure of read density that allows gene expression levels to be compared both between experimental samples and within colony replications (Mortazavi et al. 2008). This resulted in 24,330 genes in the MbAnts treatment, 22,215 genes in the MbFungus treatment, 21,653 genes in the EwAnts treatment, 21,660 genes in the EwFungus treatment, and 21,313 genes in the controls. To assess the dynamics of gene expression across the fungal treatments, we examined the differences in gene expression levels between the fungal treatments relative to controls using Cuffdiff, and found in total 375 genes (ca. 2%) that appeared to show significant expression changes, i.e. p values significantly greater than FDR (False Discovery Rates) after using Bonferroni corrections for multiple testing (Supplementary table 1). The expression patterns of these genes in the different fungal treatments are summarized in figure 3. Overall, 172 genes were differentially expressed in MbAnts, 109 genes in MbFungus, 74 genes in EwAnts and 149 genes in EwFungus treatments, of which only 7 genes (1.9 %) overlapped for all the fungal treatments (figure 3a). Of these differentially expressed genes, 239 were up-regulated (positive log2 values) and 144

genes were down-regulated (negative log2 values) (figure 3b-c). We found a statistically significant effect of fungal treatments on the distribution of genes in the up-and down- regulated categories (χ2 = 94.40, df = 3, P < 0.0001). The MbAnts treatment caused significantly more genes to be up-regulated than down-regulated (141 vs. 31 genes; χ2 = 70.35, df = 1, P < 0.0001), whereas the opposite pattern was found in the EwFungus experiment (44 up-regulated vs. 105 down-regulated genes; χ2 = 24.93, df = 1, P <

46 0.0001). The two indirect infection experiments showed intermediate values, with MbFungus having more or less equal amounts of up- and down-regulated genes (57 vs. 52; χ2 = 0.23, df = 1, P = 0.6320) and EwAnts having more up- than down- regulated genes (50 vs. 24; χ2 = 9.13, df = 1, P = 0.0025). The least unique gene expression profile was found in the EwAnts treatment, in which only 32 % (24 out of 74) of the genes were not found in any other experiment (Figure 3a), consistent with expectation as this treatment should be least dangerous for the ant-garden symbiosis, and thus induce relatively non-specific responses.

Direct and indirect fungal infections Metarhizium infections onto the ants were administered directly (MbAnts) or indirectly through the fungus garden (MbFungus). These alternative routes of infection represent a dosage difference as direct application on ants (MbAnts) will always imply contact with more M. brunneum conidia, compared to ants encountering conidia through application on the fungus garden (MbFungus) because conidia on the fungus garden can be either groomed off by ants or avoided (Reber et al. 2011). Direct infections with M. brunneum (MbAnts) resulted in an overall up-regulation of differentially expressed genes with a

median log2 value of 3.32, whereas indirect infections of M. brunneum via the fungus garden (MbFungus) resulted in a much lower degree of overall up-regulation (median

log2 value 0.32) (figure 3b). Limiting the analyses to immune-related genes, direct infections with M. brunneum (MbAnts) likewise resulted in overall up-regulation of genes (median log2 value 2.68) while indirect M. brunneum infection (MbFungus)

resulted in a weaker up-regulation of genes (median log2 value 0.64) (figure 4). Escovopsis infections of fungus gardens were also administered directly (EwFungus) or indirectly via the ants (EwAnts) and thus also represent dosage differences, as ant infections (EwAnts) will only partially be passed onto gardens because Escovopsis conidia will be removed by grooming and ants with Escovopsis conidia will avoid contact with the fungus garden (Currie & Stuart 2001). Direct infections with E. weberi on fungus gardens (EwFungus) resulted in an overall down-regulation of

differentially expressed ant genes (median log2 value -0.96), whereas indirect infections of E. weberi through the ants (EwAnts) resulted in an overall up-regulation of

47 differentially expressed genes (median log2 value 2.11), suggesting that the physiological responses to these two Escovopsis treatments were very different (figure 3b-c). Limiting the analyses to immune-related genes resulted in the same general pattern (figure 4) although indirect infections of E. weberi (EwAnts) resulted in a much weaker up-

regulation of genes (median log2 value 0.27) than when all genes were taken into account. In order to compare gene expression patterns between treatments we plotted the

log2 values for genes that had a significant expression change in at least one of the two compared treatments (figure 5; plotting the highlighted values in supplementary table 1). As expected from the consistent mortality patterns (figure 1a), direct and indirect

Metarhizium infections were significantly positively correlated (ANOVA F1,226 = 46.27, P < 0.0001) whereas direct and indirect Escovopsis infections were only weakly

positively correlated (ANOVA F1,182 = 4.44, P = 0.0364) and with only direct infection having a consistent down-regulation effect (figure 5a-b). Figure 5a further indicates that direct Metarhizium infections have a stronger up-regulating and weaker down-regulating effect than indirect infections, as the fitted correlation axis had a slope > 1. Focusing on comparisons between targets (figure 5c–d), infections on both ants and fungus garden were significantly positively correlated (ants, ANOVA F1,193 = 57.52, P < 0.0001; fungus

garden ANOVA F1,217 = 207.62, P < 0.0001). It is clear that both Metarhizium and Escovopsis infections induce up-regulation of many expressed genes, but that the degree of up-regulation is higher in the former than in the latter type of infection (slope correlation axes in figure 5c > 1). The opposite pattern was observed when comparing infections administrated via fungus gardens (figure 5d) where most differentially expressed ant genes were down-regulated. Also here the slope of the fitted correlation axis was > 1, consistent with E. weberi being a direct pathogen of fungus gardens and M. brunneum not.

Differentially expressed immune-related genes For functional annotation of immune-related genes, we focused on the 375 genes with significant expression changes (figure 3). Cross-checking with several databases, such as Blast2GO, ImmunoDB and comparing with honeybee immune genes, we identified 57 immune-related genes that were differentially expressed in one or more of the four

48 treatments: 8 ‘recognition genes’ that are likely involved in encoding pathogen surveillance proteins, 36 ‘signaling genes’ that are likely involved in encoding proteins in immune-related signaling pathways, 4 ‘stress-related genes’ that likely have known functional roles in responses to external and biotic stimuli, and 9 ‘effector genes’ that are likely involved in encoding proteins that directly inhibit pathogen growth and survival (figure 4, supplementary table 2). Direct infections of M. brunneum (MbAnts) resulted in an overall up-regulation of

immune-related genes with a median log2 value of 2.68 (figure 4). This trend was experienced in all four gene categories, with only three genes being down-regulated, mainly from the signaling category (figure 4, supplementary table 2). The indirect M. brunneum infection (MbFungus) likewise resulted in an overall up-regulation of immune- related genes (median log2 value 0.64) although this was mainly caused by up-regulation of effector genes, while signaling genes were slightly up-regulated and recognition genes were unaffected (figure 4). Only 12 immune genes were found to be affected in the indirect Escovopsis infection experiment (EwAnts). Overall, there was a slight tendency

towards up-regulation (median log2 value of 0.27) mainly caused by signaling genes, while stress response genes were unaffected and only one recognition and one effector gene were affected, the former being down-regulated and the latter being up-regulated (figure 4). Direct Escovopsis infection on fungus garden (EwFungus) was the only

treatment resulting in an overall down-regulation of immune-related genes (median log2 value -0.60), a trend that was observed in three of the four gene categories, with only the effector gene category being up-regulated and for only two genes.

DISCUSSION Survival and prophylactic responses to fungal infections Leaf-cutting ants are extremely effective at recognizing dangerous material, such as conidia of entomopathogenic fungi on themselves (Jaccoud, Hughes & Jackson 1999) and on their fungus gardens (Currie & Stuart 2001). As soon as fungal conidia (asexual spores) make contact with the ants’ bodies or direct environments, they trigger a rapid response in the form of increased activity such as self-grooming, activating otherwise immobile ants in the fungus garden (figure 2a & d). This is consistent with ants generally

49 engaging in intensive prophylactic behaviors such as various forms of grooming to rid themselves and nestmates of alien fungal conidia (Currie & Stuart 2001; Reber et al. 2011). The prophylactic responses of ants to fungal conidia, especially allo-grooming and garden-grooming seems to be location specific, with conidia on the ant cuticle (MbAnts and EwAnts) triggering more allo-grooming and conidia on the fungus garden (MbFungus and EwFungus) triggering more garden-grooming (figure 2b-c). Regardless of whether the Metarhizium conidia that we applied were directly infecting the ants or indirectly through the fungus garden, the infection doses were sufficient to cause 90-100% mortalities (figure 1b). We can thus infer that the immune systems of ants in the parallel experiment where we obtained gene expression data were indeed severely and lethally challenged. However, the mortality rate was significantly lower in the indirect treatment (MbFungus) than in the direct treatment (MbAnts) (figure 1a), and we would thus expect a less pronounced immune gene response in the indirect treatment. The correlation slope > 1 in figure 5a suggests that this was indeed the case. The infections with Escovopsis induced prophylactic responses (figure 2a-d) even though they are harmless for the ants (figure 1a-b) (See also Yek et al. 2012), but these prophylactic responses were highly efficient in preventing Escovopsis infection, as garden masses were only slightly and not significantly affected (figure 2e). As the large workers in our experiments had very few actinomycetes on their cuticle, we believe that this mostly reflects that non-stressed colonies can control Escovopsis infections via efficient grooming and killing conidia in their infrabuccal pellets (Little et al. 2003). The harmlessness of Escovopsis conidia toward ants would make it reasonable to infer that the innate immune system of the ants was not severely challenged by our Escovopsis treatments. This is supported by our finding that the expression of only very few immune genes was up-regulated by the application of Escovopsis conidia to either the ants or their fungus garden (figure 4).

Overall patterns of differentially expressed genes Gene expression is one of the most fundamental and comprehensive levels at which immune responses to fungal infections can be examined. Gene expression measurements will elucidate which biochemical pathways are activated or deactivated by the infection

50 and thereby give important information about the physiological mechanisms governing immunity (see Bonizzoni et al. 2011). For each of our four fungal infection experiments as well as the controls, we measured gene expression levels in the ants using RNA-seq technique. Overall we identified 375 genes that were differentially expressed in one or more of the treated colonies when compared to the controls. Of these genes ca. 2/3 were significantly up-regulated and ca. 1/3 were significantly down-regulated across our fungal challenge experiments, with relatively few genes overlapping between the fungal treatments (figure 3). This indicates that fundamentally different sets of genes were involved in direct and indirect Metarhizium and Escovopsis infections. Comparing treatments using different pathogens on the same target (figure 5c-d) revealed relatively similar slopes of gene expression change, but with up-regulation being typical for challenges of the ants and down-regulation for challenges of the fungus. A similar correlation was also observed when Metarhizium was used to infect either the ants or the fungus gardens (figure 5a). In contrast, the responses to treatments using Escovopsis to infect either the ants or the fungus gardens showed only a very weak correlation (figure 5b), indicating that direct and indirect Escovopsis infections induce different physiological responses. The same pattern was observed when only immune genes were considered (figure 4). In this analysis the two treatments involving infection of ants (MbAnts and EwAnts) clustered together and the two treatments involving infections of fungus garden fragments clustered together. These findings suggest that the physiological responses of the ants are relatively unspecific, depending more on the target of infection than on the specific pathogen used. Finding little specificity may imply that even though Escovopsis is harmless to the ants, it still elicits an initial physiological response similar to the one induced by infection with the lethal pathogen Metarhizium. However, by comparing the changes in immune gene expression in the two experiments it is evident that the similarities are mostly due to expression patterns of signaling genes, while changes in effector gene expression are much more different. It seems, therefore, that only the upstream immune pathway genes are activated by Escovopsis infections, but that the downstream effector genes are not activated because Escovopsis is unable to harm the ants.

51 Expression patterns of immune-defense genes In our immune-related gene expression analysis, we observed that each of the immune response phases had representative genes that were differentially expressed across our fungal treatments (figure 4; supplementary table 2). Besides confirming the well-known genetic components that are involved in innate immunity, our differential gene expression data also generated a high number of uncharacterized (hypothetical) proteins that might play important roles in cellular innate immunity. However, until further experimental evidence can be obtained, we refrain from including these putative immune-related genes in our discussions. Immune responses begin with molecular recognition of microbial surface molecules producing immune signals. We observed eight differentially regulated ’recognition‘ genes in our experimental data, of which genes involved in direct Metarhizium infection (MbAnts) were overall up-regulated whereas the few recognition genes that respond to direct and indirect Escovopsis infection (EwFungus and EwAnts) were down-regulated (figure 4). The up-regulation of ’recognition‘ genes observed in the direct M. brunneum infection treatment indicates that the innate immune system reacted to the successful penetration of the fungal conidia into the ants’ hemolymph, but that this was apparently less so when Metarhizium infections happened more slowly via indirect infections on the garden (figure 1a & 5a). As the ants were successfully infected by the indirect infection (figure 1b), the apparent lack of response in expression of recognition genes by this treatment could be because it was mostly taking place after the 48 hours when the samples for RNA-seq analysis were collected. Among the significantly up-regulated ‘recognition’ genes was a member of the

Dscam (Down syndrome cell adhesion molecule) family (log2 fold change 0.54) and a ß-

1,3-glucan-binding molecule (log2 fold change 1.40). The Dscam family contains a hyper-variable member (Dscam-hv) that can produce thousands of isoforms via alternative splicing, and together with ß-1,3-glucan-binding has been implicated in bacterial recognition in Drosophila (Armitage et al. 2012; Hoffmann 2003). The up- regulated Dscam gene in the present study is an ortholog of one of the other family members that is not believed to be hypervariable (Armitage et al. 2012). The fact that this

52 gene is up-regulated after infection may indicate that non-hypervariable Dscam genes may also play a role in immune defence. Immune signals produced by molecular recognition are modulated and/or transduced through four well-known pathways (Toll, IMD, JAK/STAT and JNK) before activating effector mechanisms. We expected to find an overall expression pattern similar to the recognition genes category, as the signaling cascades ought to be triggered by molecular recognition. We thus included genes implicated in all four pathways in our ‘signaling’ data set, and found that these genes were up-regulated after direct infections with Metarhizium (MbAnts) and to a lesser extent in indirect infections with Escovopsis (EwAnts) and Metarhizium (MbFungus). In contrast, infections of the fungus garden with Escovopsis caused a down-regulation of signaling genes, indicating that alternative signaling pathways may have been activated instead. Current immunology theory suggests that fungal and bacterial challenges activate Toll and IMD pathways, whereas JAK/STAT and JNK pathways are activated after bacterial challenges and septic injury such as wounding, suggesting a correlation between cellular immunity and stress responses (Hoffmann 2003; Silverman et al. 2003). To elucidate any correlations between these stress and immune responses we functionally categorized an additional gene set as being ‘Stress-related’. Only four such putative ‘Stress-related’ genes were identified, which seems to imply that stress and immune responses were not correlated in our experimental design, contrary to immune challenge studies in other insects (Erler, Popp & Lattorff 2011; Ratzka et al. 2011; Xu & James 2012). This could perhaps be due to the ‘wounding’ treatments that were customarily used to induce cellular innate immune responses in these studies, making it more likely to find correlations between stress and immune gene expression. However, the lack of a separate wounding treatment in our experiment and the scarcity of stress-related genes make it difficult to further validate any of these possible explanations. Immune effectors are required to target and neutralize the source of the immune signals. Across the fungal treatments, effector genes were consistently up-regulated and most pronouncedly so in the MbAnts and MbFungus experiments (figure 4). Looking at the specific effector genes across the four treatments showed that the MbAnts and MbFungus treatments up-regulated three of the same effectors, while the EwAnts and

53 EwFungus treatments had one common effector up-regulated. Despite the low number of genes, this might indicate a good correlation between the pathogen used in the experiment and the effector genes changing expression due to the infection. Among the effector molecules the antimicrobial peptides (AMP) are the best studied effector molecules of insect immune systems. In our data, two AMPs (abaecin and hymenoptaecin) were up-regulated, and represented some of the most highly up-regulated genes in our data set, with log2 values ranging from 2.4 to 8.4, indicating ca. 5-300 fold change of differential expression (supplementary table 2). The highly up-regulated abaecin and hymenoptaecin were mostly found in the Metarhizium infections, both in direct- (MbAnts) and indirect-infections (MbFungus), suggesting that abaecin and hymenoptaecin may in fact be able to eliminate Metarhizium infections when conidia doses are lower. Abaecin and hymenoptaecin were first isolated in the honey bee (Casteels et al. 1990; Casteels et al. 1993) and are both characterized as antibacterial response peptides. Our data thus suggests that abaecin and hymenoptaecin might have a broader target spectrum across the eusocial Hymenoptera and that their responses are not restricted to bacterial infections. In contrast, the antimicrobial peptide defensin (Viljakainen & Pamilo 2005) was not found to be significantly up-regulated in any of the fungal infections, although the gene sequence was readily identified among the transcripts. This may indicate that defensin in contrast to abaecin and hymenoptaecin is not involved in immune responses towards fungal pathogens. Interestingly, the direct and indirect infections with Metarhizium show much higher similarity in the expression responses of effector genes than in the responses of genes more upstream in the immune pathways (recognition and signaling genes). One explanation for this difference may be that even though the same immune pathway genes were actually up-regulated in the MbFungus experiment, the latter up-regulation was weaker because infection was indirect and thus delayed. However, signal transduction pathways might amplify relatively weak stimuli into stronger physiological signals, which may explain that our indirect infections led to lower increases in gene expression than the direct infections. The normal directionality of such amplification could imply even less affected expression levels of upstream genes in the specific pathways, while the expression of downstream effector genes is more profoundly affected so these expression

54 changes are more likely to reach statistical significance. A process like this might explain that recognition and signaling immune genes (figure 4; supplementary table 2) had parallel changes in expression, with just lower and non-significant levels after indirect infection treatments. It might also explain that correlation axes in figure 5 had slopes higher than 1, except in the Escovopsis treatment where the direct (EwFungus) and indirect (EwAnts) infections might have tended to switch on/off alternative physiological pathways, so that expression levels were no longer strongly correlated. Given that similar immune responses have been elicited in the MbAnts and MbFungus experiments, we may hypothesize that the many non-overlapping differentially expressed genes in the two experiments represent genes that are not directly involved in the immune response, but instead in general hygienic house-keeping, which differs depending on whether the fungus garden or the ants are the target of infection (figure 2).

Fungus garden disease and ants immune response Direct Escovopsis infections of fungus gardens induced a general down-regulation of gene expression in the ants, which was also evident when only immune genes were considered. The high number of down-regulated genes could be due to trade-offs in physiological pathways, for example inactivation of innate immunity so that other physiological pathways involved in neurological or locomotive processes could be prioritized because they play a key role in sustaining prophylactic behaviors (garden- grooming, weeding, metapleural gland grooming and infrabuccal pellet productions; Fernández-Marín et al. 2009). This would then possibly imply that other genes could be over-expressed to negate the negative effect of Escovopsis infections on fungus gardens (Reynolds & Currie 2004). A recent study in another leaf-cutting ant from the same area in Panama, Atta colombica, indicated that individual immune defenses are costly, and Escovopsis infections in Acromyrmex can induce an increase in the covers of cuticular actinomycete bacteria (Currie, Bot & Boomsma 2003) at considerable metabolic cost (Poulsen et al. 2003). This would imply that even though Escovopsis infections are harmless for the ants, they might induce significant changes in metabolic allocation patterns because they threaten the ants by eliminating their vital fungus garden symbiont. A similarily pronounced down-regulation of genes was not observed when Escovopsis

55 conidia were applied to the ants, or when Metarhizium conidia were applied to the fungus garden. There was, however, a very strong correlation of gene expression patterns between the EwFungus and the MbFungus treatments (figure 5d), indicating that similar physiological responses may have taken place in the ant bodies, with Escovopsis infections on gardens soliciting stronger up- or down-regulations than Metarhizium infections on fungus gardens (slope in figure 5d > 1). We hypothesize that this may be related to the increased garden-grooming taking place in both treatments. The fact that the relatively drastic down-regulation of gene expression in the ants is only taking place in the EwFungus experiment suggests that it is a specific reaction to infection of the fungus-garden by Escovopsis, the occurrence of which might also be communicated by the fungus garden to the ants by yet unknown signals (North, Jackson & Howse 1999). Further targeted experiments will be needed to elucidate the possible mechanisms behind Escovopsis induced changes in gene regulation in A. echinatior fungus-farming ants.

Acknowledgements We thank Louise Lee Munk for technical assistance in rearing the fungal species, Hermogenes Fernández-Marín for making the Escovopsis strain available, and Lumi Viljakainen for sharing honeybee immune gene and analyses suggesting the immune gene categorization. All authors were supported by a grant from the Danish National Research Foundation.

References

Armitage, S.A.O., Freiburg, R.Y., Kurtz, J. & Bravo, I.G. 2012. The evolution of Dscam genes across the arthropods. BMC Evolutionary Biology 12: 53.

Armitage, S.A.O. & Boomsma, J. J. 2010. The effects of age and social interactions on innate immunity in a leaf-cutting ant. Journal of Insect Physiology 56: 780-787.

Baer, B., Krug, A., Boomsma, J.J. & Hughes, W.O.H. 2005. Examination of the immune responses of males and workers of the leaf-cutting ant Acromyrmex echinatior and the effect of infection. Insectes Sociaux 52: 298-303.

Baer, B., Armitage, S.A.O. & Boomsma, J.J. 2006. Sperm storage induces an immunity cost in ants. Nature 441: 872-875.

Bischoff, J.F., Rehner, S. A. & Humber, R. A. 2009. A multilocus phylogeny of the Metarhizium anisopliae lineage. Mycologia 101: 512-530.

Blankenberg, D., Von Kuster, G., Coraor, N., Ananda, G., Lazarus, R., Mangan, M., Nekrutenko, A. & Taylor, J. 2010. Galaxy: a web-based genome analysis tool for experimentalists. In M. Frederick Ausubel et al. eds. Current Protocols in Molecular Biology, Chapter 19: 10.1-21.

Bonasio, R., Zhang, G., Ye, C., Mutti, N.S.,Fang, X., Qin, N., Donahue, G., Yang, P., Li, Q., Li, C., Zhang, P., Huang, Z.,Berger, S.L., Reinberg, D.,Wang, J. & Liebig, J. 2010. Genomic comparison of the ants Camponotus floridanus and Harpegnathos saltator. Science 329: 1068-1071.

Bonizzoni, M., Dunn, W.A., Campbell, C. L., Olson, K.E., Dimon, M.T., Marinotti, O. & James, A.A. 2011. RNA-seq analyses of blood-induced changes in gene expression in the mosquito vector species, Aedes aegypti. BMC Genomics 12: 82.

57 Boomsma, J.J., Schmid-Hempel, P. & Hughes, W.O.H. 2005. Life histories and parasite pressure across the major groups of social insects. In M. Fellowes, G. Holloway, & J. Rolff, eds. Insect Evolutionary Ecology. Royal Entomological Society, pp. 139-176.

Bot, A.N.M., Currie, C.R., Hart, A.G. & Boomsma, J.J. 2001. Waste management in leaf- cutting ants. Ethology Ecology Evolution 13: 225-237.

Casteels, P., Ampe, C., Riviere, L., Van Damme, J., Elicone, C., Fleming, M., Jacobs, F. & Tempst, P. 1990. Isolation and characterization of abaecin, a major antibacterial response peptide in the honeybee (Apis mellifera). The Federation of European Biochemical Societies Journal 187: 381-386.

Casteels, P., Ampe, C., Jacobs, F. & Tempst, P. 1993. Functional and chemical characterization of Hymenoptaecin, an antibacterial polypeptide that is infection- inducible in the Honeybee (Apis mellifera). The Journal of Biological Chemistry 268: 7044-7054.

Christophides, G.K., Zdobnov, E., Barillas-Mury, C., Birney, E., Blandin, S., Blass, C., Brey, P.T., Collins, F.H., Danielli, A., Dimopoulos, G., Hetru, C., Hoa, N. T., Hoffmann, J.A., Kanzok, S.M., Letunic, I., Levashina, E.A., Loukeris, T.G., Lycett, G., Meister, S., Michel, K., Moita, L.F., Müller, H-M., Osta, M.A., Paskewitz, S.M., Reichhart, J-M., Rzhetsky, A., Troxler, L., Vernick, K.D., Vlachou, D., Volz, J., von Mering, C., Xu, J., Zheng, L., Bork, P. & Kafatos, F.C. 2002. Immunity-related genes and gene families in Anopheles gambiae. Science 298: 159-65.

Conesa, A. & Götz, S. 2008. Blast2GO: A comprehensive suite for functional analysis in plant genomics. International Journal of Plant Genomics Article ID 619832.

Cremer, S., Armitage, S.A.O. & Schmid-Hempel, P. 2007. Social immunity. Current Biology 17: R693-R702.

Cremer, S. & Sixt, M. 2009. Analogies in the evolution of individual and social immunity. Philosophical Transactions of the Royal Society Biological Sciences 364: 129-142.

Currie, C.R. & Stuart, A.E. 2001. Weeding and grooming of pathogens in agriculture by ants. Proceedings of the Royal Society B: Biological Sciences 268: 1033-9.

Currie, C.R., Wong, B., Stuart, A.E., Schultz, T.R., Rehner, S.A., Mueller, U.G., Sung, G-H., Spatafora, J.W. & Straus, N.A. 2003. Ancient tripartite coevolution in the attine ant-microbe symbiosis. Science 299: 386-388.

58 Currie, C.R., Poulsen, M., Mendenhall, J., Boomsma, J.J. & Billen, J. 2006. Coevolved crypts and exocrine glands support mutualistic bacteria in fungus-growing ants. Science 311: 81-83.

Currie, C.R., Bot, A.N.M. & Boomsma, J.J. 2003. Experimental evidence of a tripartite mutualism: bacteria protect ant fungus gardens from specialized parasites. Oikos 1: 91-102.

Erler, S., Popp, M. & Lattorff, H.M.G. 2011. Dynamics of immune system gene expression upon bacterial challenge and wounding in a social insect (Bombus terrestris). PLoS ONE 6: 8.

Gadau, J., Helmkampf, M., Nygaard, S., Roux, J., Simola, D.F., Smith, C.R., Suen, G., Wurm, Y. & Smith, C.D. 2011. The genomic impact of 100 million years of social evolution in seven ant species. Trends in Genetics 28: 14-21.

Hoffmann, J.A. 2003. The immune response of Drosophila. Nature 426: 33-38.

Hughes, D.P., Pierce, N.E. & Boomsma, J.J. 2008. Social insect symbionts: evolution in homeostatic fortresses. Trends in Ecology & Evolution 23: 672-677.

Hughes, W.O.H. & Boomsma, J.J. 2004. Let your enemy do the work: within-host interactions between two fungal parasites of leaf-cutting ants. Proceedings of the Royal Society B: Biological sciences 271: S104-S106.

Hughes, W.O.H., Bot, A.N.M. & Boomsma, J.J. 2010. Caste-specific expression of genetic variation in the size of antibiotic-producing glands of leaf-cutting ants. Proceedings of the Royal Society B: Biological Sciences 277: 609-615.

Hughes, W.O.H., Eilenberg, J. & Boomsma, J.J. 2002. Trade-offs in group living: transmission and disease resistance in leaf-cutting ants. Proceedings of the Royal Society B: Biological Sciences 269: 1811-1819.

Jaccoud, D.B., Hughes, W.O.H. & Jackson, C.W. 1999. The epizootiology of a Metarhizium infection in mini-nests of the leaf-cutting ant Atta sexdens rubropilosa. Entomologia Experimentalis et Applicata 93: 51-61.

Jesenberger, V. & Jentsch, S. 2002. Deadly encounter: ubiquitin meets apoptosis. Nature Reviews Molecular Cell Biology 3: 112-121.

59 Lazzaro, B.P. 2008. Natural selection on the Drosophila antimicrobial immune system. Current Opinion in Microbiology 11: 284-289.

Lemaitre, B. & Hoffmann, J. 2007. The host defense of Drosophila melanogaster. Annual Review of Immunology 25: 697-743.

Levashina, E.A. 2004. Immune responses in Anopheles gambiae. Insect Biochemistry and Molecular Biology 34: 673-678.

Little, A.E.F., Murakami, T., Mueller, U.G. & Currie, C.R. 2003. The infrabuccal pellet piles of fungus-growing ants. Ethology Ecology Evolution 90: 558-562.

Little, A.E.F., Murakami, T., Mueller, U.R. & Currie, C.R. 2006. Defending against parasites: fungus-growing ants combine specialized behaviours and microbial symbionts to protect their fungus gardens. Biology letters 2: 12-16.

Mallon, E.B., Loosli, R. & Schmid-Hempel, P. 2003. Specific versus nonspecific immune defense in the bumblebee, Bombus terrestris L. Evolution 57: 1444-1447.

Mortazavi, A.,Williams, B.A., McCue, K., Schaeffer, L. & Wold, B. 2008. Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nature Methods 5: 621-628.

Muchovej, J.J. & Della Lucia, T.M. 1990. Escovopsis, a new genus from leaf cutting ant nests to replace Phialocladus nomem invalidum. Mycotaxon 37: 191-195.

Mueller, U.G., Scott, J.J., Ishak, H.D., Cooper, M. & Rodrigues, A. 2010. Monoculture of leafcutter ant gardens. PLoS ONE 5: p7.

North, R.D., Jackson, C.W. & Howse, P.E. 1999. Communication between the fungus garden and workers of the leaf-cutting ant, Atta sexdens rubropilosa, regarding choice of substrate for the fungus. Physiological Entomology 24: 127-133.

Poulsen, M., Bot, A.N.M., Currie, C.R. & Boomsma, J.J. 2002. Mutualistic bacteria and a possible trade-off between alternative defence mechanisms in Acromyrmex leaf-cutting ants. Insectes Sociaux 49: 15-19.

Poulsen, M., Bot, A.N.M., Currie, C.R., Nielsen, M.G. & Boomsma, J.J. 2003. Within- colony transmission and the cost of a mutualistic bacterium in the leaf-cutting ant Acromyrmex octospinosus. Functional Ecology 17: 260-269.

60 Ratzka, C., Liang, C., Dandekar, T., Gross, R. & Feldhaar, H. 2011. Immune response of the ant Camponotus floridanus against pathogens and its obligate mutualistic endosymbiont. Insect Biochemistry and Molecular Biology 41: 529-536.

Reber, A.,Purcell, J., Buechel, S.D., Buri, P. & Chapuisat, M. 2011. The expression and impact of antifungal grooming in ants. Journal of evolutionary biology 24: 954-64.

Reynolds, H.T. & Currie, C.R. 2004. Pathogenicity of Escovopsis weberi: The parasite of the attine ant-microbe symbiosis directly consumes the ant-cultivated fungus. Mycologia 96: 955-959.

Rodrigues, A., Bacci, M., Mueller, U.G., Ortiz, A. & Pagnocca, F.C. 2008. Microfungal “weeds” in the leafcutter ant symbiosis. Microbial Ecology 56: 604-614.

Sackton, T.B., Lazzaro, B.P., Schlenke, T.A., Evans, J.D., Hultmark, D. & Clark, A.G. 2007. Dynamic evolution of the innate immune system in Drosophila. Nature Genetics 39: 1461-1468.

Schmid-hempel, P. 2006. Parasitism and life history in social insects. In V.E. Kipyatkov ed. Life cycles in social insects: Behaviour, Ecology and Evolution, pp.37-48.

Silverman, N.,Zhou, R., Erlich, R.L., Hunter, M., Bernstein, E., Schneider, D. & Maniatis, T. 2003. Immune activation of NF-kappaB and JNK requires Drosophila TAK1. The Journal of Biological Chemistry 278: 48928-48934.

Taerum, S.J., Cafaro, M.J., Little, A.E.F., Schultz, T.R. & Currie, C.R. 2007. Low host- pathogen specificity in the leaf-cutting ant-microbe symbiosis. Proceedings of the Royal Society B: Biological Sciences 274: 1971-1978.

Trapnell, C.,Pachter, L. & Salzberg, S.L. 2009. TopHat Manual. Bioinformatics 25: 1105-1111.

Trapnell, C., Roberts, A., Goff, L., Pertea, G., Kim, D., Kelley, D.R., Pimentel, H., Salzberg, S.L., Rinn, J.L. & Pachter, L. 2012. Differential gene and transcript expression

61 analysis of RNA-seq experiments with TopHat and Cufflinks. Nature Protocols 7: 562- 578.

Viljakainen, L. & Pamilo, P. 2005. Identification and molecular characterization of defensin gene from the ant Formica aquilonia. Insect Molecular Biology 14: 335- 338.

Viljakainen, L. & Pamilo, P. 2008. Selection on an antimicrobial peptide defensin in ants. Journal of Molecular Evolution 67: 643-52.

Waterhouse, R.M., Kriventseva, E.V., Meister, S., Xi, Z., Alvarez, K.S., Bartholomay, L.C., Barillas-Mury, C., Bian, G., Blandin, S., Christensen, B.M., Dong, Y., Jiang, H., Kanost, M.R., Koutsos, A.C., Levashina, E.A., Li, J., Ligoxygakis, P., Maccallum, R.M., Mayhew, G.F., Mendes, A., Michel, K., Osta, M.A., Paskewitz, S., Shin, S.W., Vlachou, D., Wang, L., Wei, W., Zheng, L., Zou, Z., Severson, D.W., Raikhel, A.S., Kafatos, F.C., Dimopoulos, G., Zdobnov, E.M. & Christophides, G.K. 2007. Evolutionary dynamics of immune-related genes and pathways in disease-vector mosquitoes. Science 316: 1738- 1743.

62 Mizzen, C., Peinado, M.A., Reid, J.G., Russo, S.M., Schroeder, A.J., St Pierre, S.E., Wang, Y., Zhou, P., Jiang, H., Kitts, P., Ruef, B., Venkatraman, A., Zhang, L., Aquino-Perez, G., Whitfield, C.W., Behura, S.K., Berlocher, S.H., Sheppard, W.S., Smith, D.R., Suarez, A.V., Tsutsui, N.D., Wei, X., Wheeler, D., Havlak, P., Li, B., Liu, Y., Jolivet, A., Lee, S., Nazareth, L.V., Pu, L-L., Thorn, R., Stolc, V., Newman, T., Samanta, M.J., Tongprasit, W.A., Claudianos, C., Berenbaum, M.R., Biswas, S., De Graaf, D.C., Feyereisen, R., Johnson, R.M., Oakeshott, J.G., Ranson, H., Schuler, M.A., Muzny, D., Chacko, J., Davis, C., Dinh, H., Gill, R., Hernandez, J., Hines, S., Hume, J., Jackson, L., Kovar, C., Lewis, L., Miner, G., Morgan, M., Nguyen, N., Okwuonu, G., Paul, H., Santibanez, J., Savery, G., Svatek, A., Villasana, D. & Wright, R. 2006. Insights into social insects from the genome of the honeybee Apis mellifera. Nature 443: 931-949.

Xu, J. & James, R.R. 2012. Temperature stress affects the expression of immune response genes in the alfalfa leafcutting bee, Megachile rotundata. Insect Molecular Biology 21: 269-280.

Yek, S.H., Boomsma, J.J. & Poulsen, M. 2012. Towards a better understanding of the evolution of specialized parasites of fungus-growing ant crops. Psyche, Article ID: 239392.

Yek, S.H., Nash, D.R., Jensen, A.B. & Boomsma, J. J. 2012. Regulation and specificity of antifungal metapleural gland secretion in leaf-cutting ants. Proceedings of the Royal Society B: Biological Sciences doi: 10.1098/rspb.2012.1458.

63 Table 1. Samples used for RNA-seq, the number of high quality (clean) reads for each sample and the number of clean reads that mapped to the genome within a proper distance of their mapped mate pair.

Mapped and properly Sample Treatment Colony Clean reads paired reads D1 MbAnts Ae363 12,431,574 11,401,788 D2 MbFungus Ae363 12,138,055 11,305,658 D3 EwAnts Ae363 11,180,006 10,322,488 D4 EwFungus Ae363 11,853,276 10,914,108 D5 Control Ae363 12,505,366 11,320,934 E1 MbAnts Ae376 10,894,885 9,289,812 E2 MbFungus Ae376 11,684,876 10,470,204 E3 EwAnts Ae376 11,552,533 10,549,868 E4 EwFungus Ae376 10,971,843 9,935,394 E5 Control Ae376 10,226,125 8,938,540 F1 MbAnts Ae226B 12,411,699 11,387,804 F2 MbFungus Ae226B 12,050,785 10,535,894 F3 EwAnts Ae226B 12,025,652 10,907,438 F4 EwFungus Ae226B 12,079,257 10,655,842 F5 Control Ae226B 12,328,936 11,047,388

64 Figures Figure 1. (a) Survival curves of A. echinatior workers in the three experimental colonies subjected to the experimental fungal treatments or controls: MbAnts (Metarhizium brunneum infections of ants), MbFungus (M. brunneum infections of fungus garden), EwAnts (Escovopsis weberi infections of ants), and EwFungus (E. weberi infections of fungus garden). Controls were untreated ants and fungus garden fragments together. Survival of ants was monitored for 10 days following fungal infections. (b) The proportion of ant cadavers from the experimental treatments that showed characteristic Metarhizium sporulation three weeks after fungal infections. None of the Escovopsis infections or controls showed characteristic Metarhizium sporulation.

Figure 2. Behavioral responses (a-d), garden loss (e), and overall design (f) in experiments testing the effects of direct and indirect Metarhizium and Escovopsis infections: (a) immobile ants inside fungus gardens, (b) allo-grooming, (c) garden- grooming, and (d) self-grooming are presented as boxplots, where the bottom and top represent the minimum and maximum number of occurrences and the middle band of the box the median. (e) Garden loss recorded at the end of the survival experiment (10 days) as percentage (i.e. initial garden mass minus final garden mass). Photo credits: fungus garden by Mark Moffett, and Acromyrmex worker by Alex Wild.

Figure 3. Venn diagrams showing the overlapping profiles of genes expressed after fungal infections relative to controls: (a) overall profile of all 375 differentially expressed genes, (b) profile of the 239 up-regulated genes, and (c) profile of the 144 down- regulated genes. Tables beside the Venn diagrams give the total number of genes (N), the median, mean, and standard deviations (SD) of the log2 values of the respective fungal treatments. See supplementary table 2 for the data on which these diagrams and summary statistics are based. Note that the numbers of (b) and (c) do not necessarily add up to the numbers in (a), which is due to the same genes sometimes being up-regulated in some treatments but down-regulated in other treatments.

Figure 4. Expression profiles of immune-related genes in four categories: ’recognition’, ’signaling’, ’stress-related’, and ’effector’ across the four fungal treatments relative to controls. The red/blue colors indicate the intensity of up (red) and down (blue) gene regulations. Hierachical ordering grouped MbAnts-EwAnts and MbFungus-EwFungus immune gene expressions as most similar to each other. The table at the bottom of the figure gives the median value of the log2 fold changes with bracketed figures indicating the number of significantly over- or under-expressed genes per category.

Figure 5. Scatter plots showing pairwise comparisons of gene expression levels in different experiments, using genes that show significantly changed expression in at least one of the involved experiments. (a) direct and indirect Metarhizium infections, (b) direct and indirect Escovopsis infections, (c) fungal conidia on ants, and (d) fungal conidia on fungus garden. Dotted horizontal and vertical lines represent the null hypotheses of no change in expression (i.e. log2 value = 0). Red lines are the best-fit correlations based on the differentially regulated genes, with their correlation coefficient (r) and slope values

65 given in the scatter plots. Values along the axes represent log2 fold-changes of up- regulations (positive values) or down-regulations (negative values), cut off at ± 12 to prevent outliers (1.5-7.6% of the 375 genes) affecting the analyses.

66 Supplementary tables

Supplementary table 1. The log2 values of genes that were differentially expressed across the four fungal treatments: MbAnts (Metarhizium brunneum infections of ants), MbFungus (M. brunneum infections of fungus garden fragments), EwAnts (Escovopsis weberi infections of ants), and EwFungus (E. weberi infections of fungus garden fragments). The highlighted cells denote genes that were significantly differentially expressed, i.e. having p values significantly greater than False Discovery Rate (FDR) after using Bonferroni corrections for multiple testing (see text for details).

Supplementary table 2. The log2 values of immune-related genes across the four fungal treatments, categorized based on known or supposed immune-related roles (see text for details). The highlighted cells denote genes that were significantly differentially expressed.

67 Figure 1

Proportion of ants with conidia Proportion of ants surviving Proportion of ants surviving Proportion of ants surviving 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 1 0 0 1 0 1 1 (a) (b) 10 9 8 7 6 5 4 3 2 1 10 9 8 7 6 5 4 3 2 1 10 9 8 7 6 5 4 3 2 1 bnsMFnu wnsEFnu Control EwFungus EwAnts MbFungus MbAnts

Ae363 Control EwAnts EwFungus MbAnts MbFungus Ae376 Ae226B

Ae363

Experimental Treatments Ae376

Ae226B Time afterinfection(days) Colony Ae376 Colony Ae363 Colony Ae226B

68 Figure 2

(a) Immobility inside fungus garden (b) Allo-grooming (c) Garden-grooming 45 3.5 12 A AB A 3.0 A 10 40 2.5 ABC 8 35 AB B 2.0 6 B B 1.5 30 B 4 Number of occurences Number of occurences 1.0 Number of occurences BC C

25 2 0.5

B B B 20 0 0 MbAnts MbFungus EwAnts EwFungus Control MbAnts MbFungus EwAnts EwFungus Control MbAnts MbFungus EwAnts EwFungus Control Treatment Treatment Treatment

(d) Self-grooming 9 100 (e) Garden loss (f) 2 by 2 experimental design A 8 90 Metarhizium Escovopsis AB AB 80 7 infection infection AB 70 6 A A 60 5 A 50 Direct 4 A A 40 3 Garden loss (%) Garden 30 2 Number of occurences 20 B Indirect 1 10

0 0 MbAnts MbFungus EwAnts EwFungus Control MbAnts MbFungus EwAnts EwFungus Control Treatment Treatment

69 Figure 3

(a) overall pro le of expressed genes MbAnts MbFungus EwAnts EwFungus Median 3.32 0.32 2.11 -0.96 Mean 4.40 -0.07 1.96 -1.97 SD 5.37 3.61 5.79 4.40 N 172 109 74 149

(b) pro le of up-regulated genes MbAnts MbFungus EwAnts EwFungus Median 4.15 0.88 3.83 0.43 Mean 5.77 1.86 4.45 1.09 SD 4.95 2.80 4.59 3.03 N 141 57 50 44 Percentage 82% 52% 67% 30%

(c) pro le of down-regulated genes MbAnts MbFungus EwAnts EwFungus Median -1.67 -1.09 -1.72 -1.62 Mean -1.8 -2.19 -3.22 -3.25 SD 1.17 3.19 4.46 4.26 N 31 52 24 105 Percentage 18% 48% 33% 70%

70 Figure 4 60 0 5 t n 4 0 u o 0 3 C 2 0 0 1 0 -20 -15 -10 -5 0 5 10 log2 value

1 2 3 4 5 Recognition (1-8) 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Signaling (9-44) 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 Stress-related (45-48) 47 48 49 50 51 52 53 54 55 56 57 EwFungus MbFungus EwAnts MbAnts

EwAnts Recognition -0.90(2) 0(0) -10.58(1) 4.79(5) 0.32(10) 0.90(12) Signaling -0.62(14) 0.37(13) median log value 0.11(2) 2 Response to stress -0.16(2) 0(0) 0.46(1) (gene number) 0.32(2) 2.43(5) 0.28(1) 5.25(5) 771171 Overall -0.60(20) 0.64(20) 0.27(12) 2.68(23) Figure 5

-12 -10 MbAnts_direct infection 10 12 -8 -6 -4 -2 4 6 8 0 2 -12 -10

MbAnts_direct infection 10 12 -12 -8 -6 -4 -2 0 2 4 6 8 -12 (c) Ants -10 (a) Metarhizium -10 -8 -8

-6 -6 EwAnts_indirect infection MbFungus_indirect infection -4 -4 -2 -2 Direct and indirect infectionswiththesamepathogen andindirect Direct 0 Direct and indirect infectionsofthesametarget andindirect Direct 0 2 2 slope=1.44 slope=1.66 4 r =0.48 4 r =0.41 6 6 8 8 10 10 12 12 -12 -10 -12 -10 EwFungus_direct infection 10 12 EwFungus_direct infection 10 12 -8 -6 -4 -2 -8 -6 -4 -2 0 2 4 6 8 0 2 4 6 8 -12 -12 (d) Fungus gardens (b) Escovopsis -10 -10 slope=1.41 slope=1.15 -8 -8 r =0.70 r =0.15 -6 -6

MbFungus_indirect infection EwAnts_indirect infection -4 -4 -2 -2 0 0 2 2 4 4 6 6 72 8 8 10 10 12 12 Supplementary table 1 test_id MbAnts MbFungus EwAnts EwFungus XLOC_000190 -1,02218 -1,36682 -0,938618 -1,12793 XLOC_000327 1,74373 1,59431 1,58413 1,01963 XLOC_000727 -0,530233 -4,25959 1,18398 -0,247919 XLOC_001002 0 0 0 0,124521 XLOC_001014 0 0 0 -0,320517 XLOC_001639 0 0,301699 0,0163327 0,29898 XLOC_001706 -0,885432 1,23999 -0,674364 2,59647 XLOC_002065 0 0 0 0,876618 XLOC_002084 -2,28947 4,29733 -2,30963 3,37261 XLOC_002308 -3,0094 -1,84388 -1,3948 -1,1637 XLOC_002312 0,810856 0,684294 0,259844 1,30927 XLOC_002789 1,23159 1,17733 0,253584 0,972932 XLOC_002819 1,09591 0 0,605443 0,949996 XLOC_003147 0,82205 0,572098 0,154188 1,58422 XLOC_003152 0,190494 1,04099 0,121102 0,436686 XLOC_003544 0 0,619444 0 0,578263 XLOC_003545 -2,50856 0,32583 -0,586519 0,387601 XLOC_003622 -0,359405 -0,763848 -0,297209 -0,317094 XLOC_003665 -0,665323 -2,06448 -0,695362 -0,752951 XLOC_003960 -0,366328 -0,402471 -0,11102 -0,426727 XLOC_003971 -2,03731 0,383355 -1,23831 0,832942 XLOC_004002 0,447101 0,802944 0,516018 0,445349 XLOC_004008 0,255502 0,586791 0,472359 0,258019 XLOC_004154 1,99712 1,06958 1,43518 2,99649 XLOC_004156 2,0441 1,65162 0,883807 2,02716 XLOC_004161 1,15565 0,451281 0,632233 1,08629 XLOC_004188 -0,18494 -0,606395 -0,212494 -0,167727 XLOC_004193 -0,45243 -0,688971 -0,467373 -0,21137 XLOC_004706 -0,385748 -0,685397 -0,40799 -0,385859 XLOC_004708 -1,08328 -0,355927 -0,276065 -0,277074 XLOC_004929 -2,73923 0,194203 -0,976296 0,775585 XLOC_004994 0 0,508275 0 -0,0304764 XLOC_005015 -2,90433 0,0298091 0,18239 0,362716 XLOC_005457 0 -0,705488 -0,177089 -0,17052 XLOC_005467 -0,257897 -0,324919 -0,260305 -0,225694 XLOC_005472 0 0,726944 1,04516 2,23343 XLOC_005487 0 0,768117 0 0,631274 XLOC_006153 -0,542329 -2,67221 0,355204 -0,683739 XLOC_006220 0,933368 1,82841 2,89535 2,88835 XLOC_006711 0,785022 0,460468 0,459809 0,863804 XLOC_006718 0,0380324 0,362135 0,115171 0,112588 XLOC_006722 -2,8234 -0,447854 0 1,34547 XLOC_006776 -0,368594 -1,95326 -0,681079 -0,790211 XLOC_006777 -0,128329 -0,639831 -0,643712 -0,498794 XLOC_006782 0,142177 3,0307 0,361385 4,06555 XLOC_006794 0,913708 0,219109 0,225583 0,861348 XLOC_006803 -2,78903 0 -2,15829 1,68561 XLOC_006857 0 0,387819 0 0,567182 XLOC_007479 2,44048 2,75526 2,73243 5,74109 XLOC_007658 -2,73072 0 0,839906 -2,28115

73 test_id MbAnts MbFungus EwAnts EwFungus XLOC_007667 0 0,0170113 0 0,1425 XLOC_007721 0 -0,703033 -0,318771 -0,412555 XLOC_007722 0 -1,06259 0 0 XLOC_008307 -0,247824 -0,435177 -0,507603 -0,0623144 XLOC_008473 0 0 -0,196869 -0,264615 XLOC_008501 -3,07189 0,459347 -1,38073 0,110541 XLOC_009229 0 -0,62412 0 -0,0680116 XLOC_009538 0 -0,299414 0 -0,368911 XLOC_009776 -4,72197 -2,4312 -0,903172 -0,847224 XLOC_009952 -6,38624 -1,82204 0,0596626 1,09421 XLOC_010446 -3,52597 0,994805 -1,28089 1,47473 XLOC_010447 -3,18703 1,09084 -4,08087 1,69803 XLOC_010604 -9,17199 -0,452224 -0,528555 1,00298 XLOC_010605 -11,1608 0,307316 0,0398641 0,538879 XLOC_010607 -10,3167 20 -3,43929 20 XLOC_010608 -20 -20 0 0 XLOC_010612 0 0 0 0,275224 XLOC_010645 -5,1032 -5,76727 -4,02303 -5,30808 XLOC_010908 0,410231 0 0,711554 0,516361 XLOC_010913 0 0 0 -0,492391 XLOC_011262 -2,98822 -2,4727 -3,05596 0,27334 XLOC_011602 2,72452 0,089566 1,29012 1,63724 XLOC_011646 -3,34594 -1,34444 -1,5666 -0,648137 XLOC_011659 -0,355283 -1,19174 -0,681337 -0,905672 XLOC_011922 -2,46592 -1,11418 -0,740783 -0,277655 XLOC_012251 -7,77997 1,05862 -5,1142 20 XLOC_012332 0,422304 -0,236111 0,170631 0,881581 XLOC_013125 -0,933386 -0,342989 -3,57906 -0,893399 XLOC_013136 -3,28792 0,0627707 -2,63021 -0,164592 XLOC_013613 0 0,502973 0,29469 -0,0233224 XLOC_013624 0,814682 1,26663 0,848164 1,71206 XLOC_013737 -0,102589 0,200916 -8,59017 20 XLOC_013738 4,39935 4,54201 -0,905985 8,82825 XLOC_013746 0 1,50614 0 3,40111 XLOC_013747 -2,63964 2,1126 -0,152508 3,05525 XLOC_013755 0 0 0 -0,34896 XLOC_013761 2,9595 3,88739 -0,3262 5,72541 XLOC_013771 -0,161381 -0,477491 -0,16579 -0,244511 XLOC_013928 0 -0,562095 0 0 XLOC_014185 -1,78606 -0,409545 -0,346916 0,499338 XLOC_014191 -3,32868 -1,46108 -0,198223 -0,284271 XLOC_015252 -0,475404 -0,52656 -0,343395 -0,466832 XLOC_015314 0 0 0 0,247269 XLOC_015319 -0,019098 0,978788 0,310479 0,488518 XLOC_015702 -5,56003 0,473812 -2,39992 0,533369 XLOC_015703 -4,84614 1,64976 -1,69437 1,24694 XLOC_015704 -5,28676 2,44187 -2,07596 20 XLOC_015705 -1,31291 1,45722 -0,467954 2,53247 XLOC_015706 -20 -20 -20 0 XLOC_015707 -3,35343 0,437076 -2,32958 3,64742

74 test_id MbAnts MbFungus EwAnts EwFungus XLOC_015777 -2,00753 2,47441 0,552582 5,40848 XLOC_016497 0,717903 0,183361 0,196665 1,62338 XLOC_016499 -0,510405 -1,27133 0,016085 -0,429544 XLOC_016538 0,470232 0 0 0,58688 XLOC_016593 -2,40364 -4,05974 -1,9588 -0,821187 XLOC_016594 -0,728176 -2,99423 -0,795773 -0,112564 XLOC_016598 -0,29509 -0,656576 -0,478485 -0,250315 XLOC_016777 -3,1552 -3,47655 -6,08311 20 XLOC_016985 -3,43263 -1,05378 -1,15364 0,893759 XLOC_017020 -1,17744 -1,30466 -0,332066 -0,353115 XLOC_017097 0 0 0 0,468848 XLOC_017112 0,504068 0,16812 0,243623 0,789337 XLOC_017254 0,246491 0 0,0172308 0,973623 XLOC_017464 -0,0820737 -0,252476 -0,0822953 -0,169939 XLOC_017477 0,0647358 -0,00994813 -0,279436 -0,42786 XLOC_017485 1,16751 0,437017 0,566192 0,821119 XLOC_017627 -0,322956 -0,763281 0 -0,902502 XLOC_018560 -0,439168 -0,673676 -0,338431 -0,381153 XLOC_019123 -4,21188 -4,98276 -1,37431 -0,393814 XLOC_019129 -0,0684948 -0,343346 -0,0699541 0,43803 XLOC_019131 -1,13855 0 0 0,0436517 XLOC_019133 0 0 0 0,917683 XLOC_019503 -1,64097 2,21017 -0,893575 3,29417 XLOC_019504 0,893837 2,42201 0,224987 20 XLOC_019505 -0,0665939 1,85752 -0,336006 6,42622 XLOC_019636 -3,54617 -0,618931 0 -0,336583 XLOC_019802 0,570375 0 0 0,517473 XLOC_019807 -2,73884 0 0 -0,49565 XLOC_019848 -0,514518 -0,91331 -0,335751 -0,600295 XLOC_020002 -0,13947 -0,640952 -0,945079 -0,220969 XLOC_020024 0 -2,10712 -0,153208 0,0880542 XLOC_020053 0,593418 -0,155612 0,356052 0,991063 XLOC_020168 0,721558 0,234724 0,461297 0,30893 XLOC_020314 -4,40136 -0,337423 -0,958966 0,618907 XLOC_020640 -0,538199 -0,16159 -0,178281 -0,316271 XLOC_020644 -7,75451 -3,36002 -1,13396 -1,2584 XLOC_020685 0,851264 -0,79545 -3,68614 7,88059 XLOC_020686 2,75331 0 -0,128172 11,2194 XLOC_020691 -20 0 0 0 XLOC_020709 -3,06349 -0,657842 -0,688287 -0,227142 XLOC_020716 -3,16161 -0,85991 -0,584425 0,517469 XLOC_020720 -0,806094 0 -0,312258 0,00652287 XLOC_020748 -8,10964 -1,96903 -5,08702 20 XLOC_020749 -20 -20 0 0 XLOC_020750 2,55447 1,38869 0,374312 11,2351 XLOC_020752 -6,90542 -4,10021 -1,70284 0,235183 XLOC_021020 -8,4446 -5,30417 -1,14859 -1,1776 XLOC_021041 -2,61888 -0,0540086 -0,284509 0,463398 XLOC_021056 -6,96853 -3,60394 -1,35581 2,37351 XLOC_021207 0,713963 -0,239578 0,239999 1,45142

75 test_id MbAnts MbFungus EwAnts EwFungus XLOC_021242 0,282496 0,668382 0 0,442828 XLOC_021585 0,834401 0,334578 0,669753 1,12206 XLOC_021760 -4,56438 -1,20624 -1,01248 -0,645756 XLOC_021830 0,761503 1,28365 0,158274 0,964236 XLOC_022101 2,25015 1,79006 1,41241 2,56226 XLOC_022207 -6,97209 -0,681587 -1,74919 -0,731925 XLOC_022258 0,102954 0,386786 0,248045 0,646538 XLOC_022260 0 0 0,272072 0,12916 XLOC_022845 -4,99965 -1,28321 -0,836894 -0,892701 XLOC_022846 -5,35075 2,02558 -0,585889 0,240182 XLOC_022867 -0,135064 -0,72563 -0,484733 -0,22983 XLOC_022888 -8,12983 -2,30375 -3,05128 0,267617 XLOC_023079 -0,835951 -0,647354 -1,37309 -1,84347 XLOC_023192 -6,84786 -2,27709 -4,58484 -0,81238 XLOC_023289 -0,821587 -0,283395 -0,0505015 0,919535 XLOC_023713 -3,8533 -1,00155 -1,43856 -0,122139 XLOC_023761 0,386913 0 0,291758 0,159893 XLOC_023842 0 0 0,371521 0,316197 XLOC_024311 -0,871727 -1,00845 -0,627094 -0,466305 XLOC_025282 1,3742 0,504875 0,693318 1,40419 XLOC_025338 1,73867 1,73994 0,695307 1,90395 XLOC_025455 -0,62264 -0,761823 -0,774161 -1,21835 XLOC_025911 -7,52635 -4,0081 -4,6206 -0,266988 XLOC_025945 0 0 0 0,823244 XLOC_026786 -2,55835 0,60512 -0,643594 0,491953 XLOC_026824 2,57383 0,393104 -4,63396 1,15076 XLOC_026830 -0,177016 0 -0,398202 -0,409905 XLOC_027288 -1,17427 0,377124 0,329809 0,13082 XLOC_027294 0 0,204767 -0,274179 -0,21387 XLOC_027397 -0,358862 0 -0,271158 -0,289051 XLOC_027562 -2,89403 0,127785 -0,767278 0,282167 XLOC_027822 -7,03539 -0,616781 -5,20462 -1,03081 XLOC_027830 1,40832 1,38064 1,88707 1,55576 XLOC_027946 -3,26573 1,25509 -1,96741 0,186562 XLOC_028394 -1,07832 0,070978 -3,38803 0,304169 XLOC_028671 1,83359 1,58062 1,39262 2,58938 XLOC_028702 -0,713528 -0,387365 0 0,17712 XLOC_028959 -0,254885 -0,368551 -0,123757 -0,205491 XLOC_028962 -3,19126 -0,47103 -0,628579 -0,399579 XLOC_028980 0 0 2,14517 2,55726 XLOC_029023 6,02108 20 20 0,33367 XLOC_029273 0 -0,802967 -0,461865 -0,295646 XLOC_029280 -0,0908148 0 -0,274798 -0,518097 XLOC_029392 2,54725 10,0369 0,92257 10,5646 XLOC_029636 -1,86691 2,28083 -0,575258 6,50078 XLOC_029643 -0,272894 0 0 -0,538164 XLOC_029814 0 0 0 -0,36995 XLOC_029832 0 -0,369478 -0,386014 -0,394848 XLOC_029891 -0,633945 -1,10556 -0,402565 -0,572966 XLOC_030417 -3,45055 -0,131444 -0,738876 0,265304

76 test_id MbAnts MbFungus EwAnts EwFungus XLOC_030432 0 0 -0,0426882 -0,616434 XLOC_030470 0 0 0,523362 0,22476 XLOC_030480 0 0 0 -0,302275 XLOC_030692 0 0 0 0,290274 XLOC_031058 0,483457 0,605342 0,3172 0,4019 XLOC_031103 0 -0,447557 -0,218907 0 XLOC_031476 -0,408499 -0,609954 -0,365029 -0,349652 XLOC_032035 -1,3736 0,0306524 0,0743834 0,330526 XLOC_032182 2,92482 -0,402745 10,5853 -1,25077 XLOC_032327 0 0 0 -0,313086 XLOC_032451 0,624898 0,376097 0,280093 0,156866 XLOC_032463 0 0,660491 0 0,204847 XLOC_032483 1,45071 2,04407 0,486354 1,41669 XLOC_032637 0 0,345217 0 -0,362053 XLOC_032700 0 0 0 -0,78764 XLOC_033008 -3,12883 -1,20932 -0,440381 XLOC_033019 -0,369794 0,750196 0,169515 -0,0471138 XLOC_033612 -7,37152 -0,385619 -5,26276 20 XLOC_033614 -7,72314 20 -8,92297 20 XLOC_033615 -7,88187 -0,705936 -9,28475 -1,16241 XLOC_033616 -20 -20 -20 0 XLOC_033698 -0,261799 0 0 -0,0768551 XLOC_034001 0,383904 0,758918 0,384157 0,668591 XLOC_034008 3,483 1,45005 3,65291 8,61517 XLOC_034092 -6,47175 -0,290229 -4,11544 0,200033 XLOC_034094 -0,471915 0,476467 -2,51941 2,33288 XLOC_034109 -8,45792 0,32818 -6,68247 0,780376 XLOC_034642 -3,28915 -0,617947 0,0487514 0,233351 XLOC_034643 -5,76832 -2,64324 -0,780719 -0,0200852 XLOC_034672 0 0,743604 0 0,255721 XLOC_035099 0,952778 0,927301 1,05327 0,594711 XLOC_035158 0 0 0,432703 1,12133 XLOC_035441 -2,76284 2,67745 -0,977204 4,55114 XLOC_035453 -10,6419 0,205754 -1,08133 1,5446 XLOC_035454 -6,41831 -0,931224 -2,17316 -0,318302 XLOC_035464 -4,88191 1,45536 -0,425834 2,62577 XLOC_035594 0 0 -0,356274 -0,502566 XLOC_035595 -0,385736 -0,56388 -0,180307 -0,680514 XLOC_035644 0 1,47586 0 1,92943 XLOC_035687 0,497504 0,827947 0 0,443403 XLOC_035696 0,190224 0,523168 0,725858 1,32108 XLOC_035705 -8,91936 -2,11511 -2,92341 -5,31284 XLOC_035706 -7,45387 -0,400892 -0,315468 -0,774869 XLOC_035976 -0,783188 -0,145913 -4,87505 1,20071 XLOC_036038 1,04379 0,801953 0,802072 1,16508 XLOC_036039 0,481572 0,312949 0,23349 0,28146 XLOC_036059 0,267815 0 -0,0173282 0,346063 XLOC_036073 0,513418 0,116891 0,185148 0,613042 XLOC_036094 0,275101 0,883506 0,439459 0,157156 XLOC_036204 0,968399 0,417799 0,463084 1,07929

77 test_id MbAnts MbFungus EwAnts EwFungus XLOC_036375 0,237098 0,765811 0,470229 0,509749 XLOC_036862 -2,18254 -0,263771 2,70065 3,97013 XLOC_036870 -3,86359 1,82127 -0,570344 2,55563 XLOC_036871 -1,15241 2,55862 -1,48368 20 XLOC_036873 0,635949 2,8016 0,108822 3,35417 XLOC_036875 -2,79558 3,01179 0,330395 7,75966 XLOC_036876 -5,3377 1,43385 -0,100138 2,79173 XLOC_036891 0 0 0 -0,381148 XLOC_037151 -0,49975 -1,56167 -0,40817 -0,505594 XLOC_037631 0,778173 0,808056 0 0,882017 XLOC_037745 -2,68446 -0,238992 -0,475496 0,0058212 XLOC_037794 -3,34758 1,15152 -0,455816 1,12151 XLOC_038240 -4,14615 0,62926 -1,448 1,55198 XLOC_038651 1,0293 2,49478 0,920064 3,76208 XLOC_038656 0 -0,303612 -0,150818 -0,206791 XLOC_038660 0,966297 1,29301 0,769485 1,46772 XLOC_038959 2,18005 2,81015 -0,421544 10,992 XLOC_038960 3,36807 3,27284 -1,95138 4,84026 XLOC_038961 2,85726 4,90932 -1,24984 7,80448 XLOC_039014 0,946513 2,10339 0,62364 1,45739 XLOC_039020 0,43446 0,648414 0,407362 0,250536 XLOC_039026 -1,81418 1,22249 -0,261028 2,58513 XLOC_039511 0 -0,656911 0 0 XLOC_039915 -1,42319 20 -4,29109 20 XLOC_039916 -1,63033 3,32834 -6,01218 6,81293 XLOC_039946 0,397431 0 0,30681 0,673855 XLOC_040097 0 0 0 -0,435405 XLOC_040166 -1,31068 -1,94329 -0,836304 -1,1088 XLOC_040584 -0,416301 -0,485486 -0,662766 -0,667076 XLOC_040854 -6,46579 -1,25176 -1,1831 -1,34258 XLOC_040855 -1,74482 -0,717231 -6,4314 -0,753985 XLOC_040856 -0,256857 -0,832965 -4,93229 0,564628 XLOC_040872 -2,29633 0 -1,66061 1,5166 XLOC_040914 -8,60933 -0,946555 0 -0,470958 XLOC_041107 -4,0293 0,136747 -0,451615 -0,111524 XLOC_041155 -2,70776 0,612091 0,024441 0,468964 XLOC_041966 0,328751 -0,448235 0 0,611374 XLOC_041977 1,60738 2,27612 0,0405401 3,23961 XLOC_041979 -9,12628 -0,734878 -4,16785 -0,00666968 XLOC_041980 0 2,18927 -1,93471 3,40025 XLOC_042787 0 -20 0 -20 XLOC_042956 -3,80638 1,56295 -4,38736 1,87348 XLOC_043061 0,368779 0,664746 0,240408 0,162328 XLOC_043105 0 0 -0,156609 -0,300863 XLOC_043377 0,804542 0,934457 -2,20784 2,38972 XLOC_043405 0 -0,00767876 0 0,588145 XLOC_044174 -1,3955 -0,518476 -0,28378 -0,0747783 XLOC_044275 -4,26543 2,00897 -0,0753375 2,54239 XLOC_044276 -5,57 1,73824 0,671293 2,00007 XLOC_044277 -5,98583 20 -0,397901 1,50581

78 test_id MbAnts MbFungus EwAnts EwFungus XLOC_044319 -20 0 0 -20 XLOC_044321 -11,5926 20 -2,54991 -0,243887 XLOC_044323 -1,09329 -0,0424939 -0,71261 -0,394253 XLOC_044348 1,67386 1,10012 0,713152 1,81444 XLOC_044349 -5,54501 -1,38237 -2,42046 -0,3597 XLOC_044361 -7,9816 1,39925 0,828647 0,142772 XLOC_044712 0,100804 0,314513 1,86665 0,820098 XLOC_044713 -9,73777 20 -6,01255 -0,241416 XLOC_044853 -0,171104 -0,510403 0 0 XLOC_044949 -0,275399 -0,508133 -0,493102 -0,746659 XLOC_045518 0 -0,881869 0 0 XLOC_045642 -0,460633 -0,754693 -0,390511 -0,446588 XLOC_045880 -1,46502 2,21847 -0,320355 4,59746 XLOC_045881 -5,92003 0,301719 -2,52684 20 XLOC_046088 -5,25081 -0,331951 -0,13091 -0,276661 XLOC_046175 -0,711819 -2,65424 -1,5298 -1,096 XLOC_046201 -20 -20 -20 0 XLOC_046204 1,08495 0,563515 0,926218 1,23836 XLOC_046220 0 0 -0,263196 -0,433523 XLOC_046780 -2,79988 0,372478 -0,131332 1,04321 XLOC_046915 -3,63119 -0,389553 -0,11651 -0,115876 XLOC_046978 4,09375 3,09778 3,98953 5,348 XLOC_047177 -0,10716 1,26803 0,286175 0,961912 XLOC_047188 0,165516 0,264239 0,102234 0,0616477 XLOC_047280 0 -0,449323 0 -0,296534 XLOC_047343 -3,68039 -0,669703 -0,361702 -0,339388 XLOC_047345 -10,5773 20 -0,21236 0,878626 XLOC_047346 -10,8931 0,744447 -1,85308 0,888015 XLOC_047347 -20 -20 -20 -20 XLOC_047348 -20 0 -20 -20 XLOC_047351 -7,84867 0,0414182 -2,30525 -1,61827 XLOC_047354 -1,31393 -0,536533 0,0566072 -0,214874 XLOC_047607 0 -0,375246 -0,385588 0 XLOC_047611 -0,962339 -0,563231 -0,521077 -0,253009 XLOC_047641 -4,67349 -2,67526 -0,834494 -0,755139 XLOC_047696 0,0482648 0 0 -0,744317 XLOC_048023 0,197144 -0,207394 -0,0630919 -0,525839 XLOC_048831 -2,70786 0,744402 -1,04942 0,451323 XLOC_048890 0 0 -0,212503 -0,346975 XLOC_048952 -0,404684 1,27619 0,258789 2,15116 XLOC_048984 -3,69306 3,02072 -4,73749 5,00416 XLOC_048985 0,986307 0,298933 0,619868 1,82949 XLOC_048987 -4,43718 2,60055 -4,26952 20 XLOC_049399 0 -0,880504 0 0 XLOC_049715 -2,19541 0,597689 -2,9365 0,772698 XLOC_049849 -4,79521 -3,01033 -1,99408 -0,395334 XLOC_049880 -2,2284 -0,841453 -0,347245 0,275585 XLOC_049910 0 -0,357048 -0,224981 -0,062906 XLOC_049913 2,34951 1,56205 2,38781 2,52266 XLOC_049929 0,576253 0,251305 0,588705 0

79 test_id MbAnts MbFungus EwAnts EwFungus XLOC_049965 -0,677738 0,716739 -0,785072 1,56061 XLOC_049999 2,09665 0,0760913 0,897114 2,61089 XLOC_050356 -6,38819 -2,05469 0,600386 0,644498 XLOC_050357 -6,76654 20 -0,0827641 20 XLOC_050376 -1,53722 1,14163 0 2,3711 XLOC_050427 -20 0 0 0 XLOC_050429 -6,28991 -1,6582 -4,48455 0,728035 XLOC_050433 -20 -20 -20 0 XLOC_050434 -20 0 0 0 XLOC_050435 -11,1384 -5,94923 -3,32295 -0,289177 XLOC_050474 1,2952 0,910636 0,463235 1,22313 XLOC_050485 0 -0,942577 -2,06783 -0,710146 XLOC_050602 1,50626 2,1842 9,16929 2,9467 XLOC_050700 -0,781868 -1,03548 -0,233277 -0,238665 XLOC_051310 -3,12085 -2,18681 -1,05257 -0,439945 XLOC_051640 0 0 -0,277021 -0,320882 XLOC_051787 0 0,535217 0 -0,215027 XLOC_051813 0,265374 0,471367 0,0414128 -0,107801 XLOC_052008 2,58288 5,6439 3,10421 20 XLOC_052012 3,57041 8,42301 5,97983 9,68714 XLOC_052077 0,875089 -0,501082 0,220768 1,52553 XLOC_052179 -20 -20 -20 0 XLOC_052242 -8,86097 0,388717 -4,44431 20 XLOC_052259 0 -0,582803 -0,220851 -0,33821 XLOC_052319 -3,31455 1,99267 -0,931364 2,64278

80 Supplementary table 2

di_tset eneG seman noisnecsA rebmun yrogetaC sugnuFwEstnAwEsugnuFbMstnAbM XLOC_017254 Feline leukemia virus subgroup C receptor-related protein 2 EGI66947.1 Recognition -0.246491 0 -0.0172308 -0.973623 XLOC_020640 Down syndrome cell adhesion molecule-like protein (DSCAM) EGI65995.1 Recognition 0.538199 0.16159 0.178281 0.316271 esanegyxoonom-4 nireficuL549520_COLX esanegyxoonom-4 1.15546IGE noitingoceR 0 0 442328.0-0 3940FPU281230_COLX 1.53446IGE noitingoceR 77052.13585.01-547204.028429.2- nietorp gnidnib-naculg-3,1-ateB471440_COLX nietorp 1.33595IGE noitingoceR 3877470.087382.0674815.05593.1 itehtopyh543740_COLX lac nietorp 2.87231_15G 1.22685IGE noitingoceR 3775.01 02- 626878.0-63212.0 1.87231_15G nietorp lacitehtopyh643740_COLX nietorp 1.87231_15G 1.22685IGE noitingoceR 510888.0-80358.1744447.0-1398.01 nitcel epyt-c948940_COLX nitcel 1.39775IGE noitingoceR 433593.080499.133010.312597.4 48 nietorp regnif cniZ226300_COLX regnif nietorp 48 1.35407IGE gnilangiS 490713.0902792.0848367.0504953.0 5 esaretseidohpsohp retseidohpsohporecylg evitatuP800400_COLX retseidohpsohporecylg esaretseidohpsohp 5 1.72916IGE gnilangiS 910852.0-953274.0-197685.0-205552.0- esanikid retaw ,edineleS607400_COLX retaw esanikid 1.68007IGE gnilangiS 704.0793586.0847583.0 958583.099 rotcaf aimekuel citapeH807400_COLX aimekuel rotcaf 1.72916IGE gnilangiS 470772.0560672.0729553.082380.1 2 nietorp gniniatnoc-niamod nixelP764500_COLX gniniatnoc-niamod nietorp 2 1.40896IGE gnilangiS 496522.0503062.0919423.0798752.0 niclacugeR497600_COLX 1.32496IGE gnilangiS 843168.0-385522.0-901912.0-807319.0- XLOC_009229 XLOC_009229 Signaling 0 0.62412 0 0.0680116 rtomyhC206110_COLX 2-nispy 1.36386IGE gnilangiS 21092.1-665980.0-25427.2- - 42736.1 51 esanik esanik esanik nietorp detavitca-negotiM316310_COLX nietorp esanik esanik esanik 51 1.36876IGE gnilangiS 0 4223320.096492.0-379205.0- XLOC_015319 Phosphorylase b kinase gamma catalytic chain, skeletal muscle isoform EGI67480.1 Signaling 0.019098 -0.978788 -0.310479 -0.488518 2htM rotpecer delpuoc nietorp-G835610_COLX delpuoc rotpecer 2htM 1.99176IGE gnilangiS 232074.0- 0 0 5.0- 8 886 2-epyt esanegordyhed AoC-lycayxordyh-3790710_COLX esanegordyhed 2-epyt 1.37966IGE gnilangiS 0 0 848864.0-0 906 nietorp regnif cniZ726710_COLX regnif nietorp 906 1.15866IGE gnilangiS 182367.0659223.0 0 205209.0 01A nietorp gnidnib-tnarodo evitatuP413020_COLX gnidnib-tnarodo nietorp 01A 1.51266IGE gnilangiS 3.063104.4 709816.0-669859.032473 19450_15G nietorp lacitehtopyh847020_COLX nietorp 19450_15G 1.89066IGE gnilangiS 02-20780.530969.146901.8 Xnik esanik nietorp-eninoerht/enires evitatuP057020_COLX nietorp-eninoerht/enires esanik Xnik 1.10166IGE gnilangiS 1532.11-213473.0-96883.1-74455.2- 2PC rotcaf noitpircsnarT242120_COLX rotcaf 2PC 1.01956IGE gnilangiS 283866.0-694282.0- 828244.0-0 esayl-ammag eninoihtatsyC852220_COLX esayl-ammag 56IGE 1.075 gnilangiS 835646.0-540842.0-687683.0-459201.0- elbbuts esanietorp enireS648220_COLX esanietorp elbbuts 1.01456IGE gnilangiS 281042.0-988585.085520.2-57053.5 1k6 054P emorhcotyC768220_COLX 054P 1k6 1.73456IGE gnilangiS 38922.0337484.036527.0460531.0 XLOC_023842 Insulin-like growth factor-binding protein complex acid labile chain EGI65161.1 Signaling 0 0 -0.371521 -0.316179 tsae esaetorp enireS207820_COLX esaetorp retsae 1.80836IGE gnilangiS 63783.0825317.0 5 0 21771.0- XLOC_029832 High affinity cAMP-specific 3',5'-cyclic phosphodiesterase 7A EGI63493.1 Signaling 0 0.369478 0.386014 0.394848 30_COLX lacitehtopyh1542 nietorp 43880_15G 1.93826IGE gnilangiS 668651.0-390082.0-790673.0-898426.0- elzteaps nietorp607530_COLX elzteaps 1.95916IGE gnilangiS 968477.0864513.0298004.078354.7 esatahpsohp nietorp-ninoerht/enireS679530_COLX esatahpsohp 1.29207IGE gnilangiS 17002.1-50578.4319541.0881387.0 03101_15G nietorp lacitehtopyh268630_COLX nietorp 03101_15G 1.66516IGE gnilangiS 31079.3-56007.2-177362.045281.2 84101_15G nietorp lacitehtopyh378630_COLX nietorp 84101_15G 1.48516IGE gnilangiS 71453.3-228801.0-6108.2-949536.0- nietorp 1C kciP-nnameiN544730_COLX 1C nietorp 1.49316IGE langiS gni 2128500.0-694574.0299832.064486.2 73901_15G nietorp lacitehtopyh115930_COLX nietorp 73901_15G 1.73806IGE gnilangiS 0 119656.0 0 0 SP-ateb nirgetnI649930_COLX SP-ateb 1.18606IGE gnilangiS 134793.0- 0 558376.0-18603.0- nietorp gnitavitca-ohR gnidnib-nitcA485040_COLX gnitavitca-ohR nietorp 1.31506IGE gnilangiS 670766.0667266.0684584.0103614.0 71pS nietorp ecafrus mrepS022640_COLX ecafrus nietorp 71pS 1.78885IGE gnilangiS 0 0 325334.0691362.0 1 nietorp gniniatnoc-staeper GF dna niamod PAG-frA706740_COLX niamod dna GF gniniatnoc-staeper nietorp 1 IGE 1.20585 gnilangiS 0 885583.0642573.0 0 emyzne gnitrevnoc-editpep citeruirtan lairtA584050_COLX citeruirtan gnitrevnoc-editpep emyzne 1.04775IGE gnilangiS 0 641017.038760.2775249.0 dg esaetorp enireS206050_COLX esaetorp dg 1.58636IGE gnilangiS 7649.2-92961.9-2481.2-62605.1-

81 test_id Gene names Ascension number Category MbAnts MbFungus EwAnts EwFungus XLOC_027397 Protein SPT2-like protein EGI64266.1 Response to stress 0.358862 0 0.271158 0.289051 XLOC_036073 Sorbitol dehydrogenase EGI61782.1 Response to stress -0.513418 -0.116891 -0.185148 -0.613042 XLOC_045642 RNA-binding protein Musashi-like protein Rbp6 EGI59067.1 Response to stress 0.460633 0.754693 0.390511 0.446588 XLOC_051787 hypothetical protein G51_14528 EGI57459.1 Response to stress 0 -0.535217 0 0.215027 XLOC_001014 Dynamin EGI71098.1 Effector 0 0 0 0.320517 XLOC_004994 Uncharacterized protein C21orf2 EGI70001.1 Effector 0 -0.508275 0 0.0304764 XLOC_007721 XLOC_007721 Effector 0 0.703033 0.318771 0.412555 XLOC_009776 Abaecin EGI68768.1 Effector 4.72197 2.4312 0.903172 0.847224 XLOC_021020 Hymenoptaecin.1 EGI65979.1 Effector 8.4446 5.30417 1.14859 1.1776 XLOC_021056 Hymenoptaecin.2 EGI65979.1 Effector 6.96853 3.60394 1.35581 -2.37351 XLOC_032035 hypothetical protein G51_08717 EGI62937.1 Effector 1.3736 -0.0306524 -0.0743834 -0.330526 XLOC_046088 hypothetical protein G51_12949 EGI58837.1 Effector 5.25081 0.331951 0.13091 0.276661 XLOC_051640 APAF1-interacting protein-like protein EGI57504.1 Effector 0 0 0.277021 0.320882

82 CHAPTER 2

Regulation and specificity of antifungal metapleural gland secretion in leaf-cutting ants

An Acromyrmex worker is a walking chemical factory, producing complex mixture of secretions

Photo credit: Alex Wild

Published in Proceedings of the Royal Society B

83 Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

Proc. R. Soc. B doi:10.1098/rspb.2012.1458 Published online

Regulation and specificity of antifungal metapleural gland secretion in leaf-cutting ants Sze Huei Yek1,*, David R. Nash1, Annette B. Jensen2 and Jacobus J. Boomsma1 1Department of Biology, Centre for Social Evolution, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark 2Department of Agriculture and Ecology, Centre for Social Evolution, University of Copenhagen, Thorvaldsensvej 40, 1871 Frederiksberg, Denmark Ants have paired metapleural glands (MGs) to produce secretions for prophylactic hygiene. These exocrine glands are particularly well developed in leaf-cutting ants, but whether the ants can actively regulate MG secretion is unknown. In a set of controlled experiments using conidia of five fungi, we show that the ants adjust the amount of MG secretion to the virulence of the fungus with which they are infected. We further applied fixed volumes of MG secretion of ants challenged with constant conidia doses to agar mats of the same fungal species. This showed that inhibition halos were significantly larger for ants challenged with virulent and mild pathogens/weeds than for controls and Escovopsis- challenged ants. We conclude that the MG defence system of leaf-cutting ants has characteristics reminiscent of an additional cuticular immune system, with specific and non-specific components, of which some are constitutive and others induced. Keywords: prophylactic hygiene; pathogens; virulence; induced and constitutive immunity

1. INTRODUCTION led to significantly enlarged MGs in Atta and Acromyrmex The metapleural glands (MGs), paired structures at the leaf-cutting ants [13], where the secondary evolution of posterolateral margin of the mesosoma, are found only queen multiple-mating has also allowed considerable in ants and are one of the defining apomorphies of the genetic variation for MG size to be maintained [15,16]. family Formicidae [1,2]. The secretions of MGs may Studies on MG secretions in attine ants have shown have many functions, but the production of anti- that the production of these secretions is metabolically microbial compounds for general nest sanitation is the costly [17], that they contain both a diversity of carboxylic most widespread and general [3–6]. The evolution of acids of various chain lengths and proteinaceous com- glands that produce secretions with generalized prophy- pounds [18,19], and that their functioning may be lactic functions seems a logical adaptation for ants, subject to metabolic trade-offs [20,21]. However, specific because colonies are often densely packed and individuals tests of antimicrobial function have remained rare, mainly interact continuously with the nest substrate (usually soil) due to the technical difficulties of extracting the very and food (often cadavers), where micro-organisms small quantities of these glandular secretions [9,18,19]. abound [7,8]. Despite their likely importance for the Hence, the antimicrobial role of MG secretions has evolution and diversification of the ants, few studies usually been inferred rather than measured, or has have investigated the details of MG function. The taxo- focused on single compound testing rather than con- nomic coverage of these studies is scattered, and most dition-dependent variation in natural MG secretions [5]. have primarily focused on ants with exceptionally derived MG secretions were believed to mainly spread pas- MG functions [9,10]. sively over the ant cuticle [22], but a recent study by The leaf-cutting ants (Attini) are a partial exception to Fernańdez-Marıń et al. [23] showed that active acqui- this dearth of information, as MG reservoir size can be sition of small quantities of secretion with the front legs measured relatively easily, and these sizes show interesting followed by focused grooming frequently occurs across allometries across worker castes [11–13]. The anti- the fungus-growing ants. Most bacterial, viral and proto- microbial role of MG secretions in leaf-cutting ants is zoan disease propagules must be ingested by ants to potentially of particular significance, because these ants become infective, but pathogenic fungi enter the ant have to protect both themselves against entomopatho- hosts via the cuticle [8]. This has led to the expectation gens, and their mutualistic fungus gardens against that MG secretions in leaf-cutting ants may function pre- parasites and competitors [14], which appears to have dominantly as an antifungal defence, a notion that was corroborated by Fernańdez-Marıń et al.[21,23], showing that a number of attine species increase their active * Author for correspondence ([email protected], [email protected]). grooming rate with MG secretion after being exposed to Electronic supplementary material is available at http://dx.doi.org/ fungal conidia (asexual spores). The ants also apply 10.1098/rspb.2012.1458 or via http://rspb.royalsocietypublishing.org. MG secretions to fungus garden infections, after which

Received 25 June 2012 Accepted 2 August 2012 1 This journal is q 2012 The Royal84 Society Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

2S.H.Yeket al. Cuticular immune system of ants the compromised but treated mycelial debris is stored in period were surface-sterilized by rinsing with 70 per cent the infrabuccal pocket (a cavity just behind the ant ethanol followed by 96 per cent ethanol to avoid fungal mouthparts), where any remaining conidia are likely to growth from external contaminants. The surface-sterilized be killed by the mixture of MG and labial gland secretions ants were placed in a plastic pot lined with damp filter paper. before debris-pellets are discarded [23,24]. However, it Cadavers were inspected after 10 days to score the presence has remained unknown whether any infection-induced or absence of hyphal growth on the cuticular surface, and to increase in grooming with MG secretions is accompanied check whether these hyphae produced conidia of the fungus by an increase in quantity or adjustment in the specific with which the ants were inoculated (sporulation data in potency of the secretions. electronic supplementary material, table S2). The fungal The present study examined whether the MG secretions inoculations and controls were carried out simultaneously so of Acromyrmex octospinosus leaf-cutting ant workers can be that ant mortality assessment and cadaver inspections were adjusted according to the fungal conidia to which the ants completed within two weeks from the start of the experiments. are exposed, using five species of fungi representing three The workload of the experiments was such that the treat- broad categories of threat (entomopathogen, generalist ments could only be handled in parallel by same person (the saprophyte and fungus garden pathogen). first author) for two colonies at a time. The experiments with colonies Ao273 and Ao404 were therefore carried out six months before the experiments with colonies Ao482 and 2. MATERIAL AND METHODS Ao492. This reduced error variance as the variance across (a) Fungal inoculation experiments colonies could be partialled out in subsequent statistical Four representative colonies of Acromyrmex octospinosus, col- analyses (where it turned out to be insignificant). lected in Gamboa, Panama, in 2005 (colony Ao273) and 2007 (colonies Ao404, Ao482 and Ao492), were used in the (b) Extraction and quantification of metapleural experiments. Colonies were picked because of their similarity glands secretions in size, with approximately 1 litre of fungus garden and To examine the quantitative responses of MG secretion in 50–100 large garden workers being easily available at each of inoculated ants, we set up a second experiment structured the consecutive bouts of sampling. The colonies were kept as two-way repeated measures factorial design, to measure under standardized conditions in a climate room at 258Cand the amount of MG secretions from inoculated and control 70 per cent relative humidity at the University of Copenhagen. ants for each of the five different fungal species. To inocu- To avoid age- or caste-specific variation in the composition late ants with each of the five fungal species and obtain of the MG secretions, we used only large garden workers of sufficient amounts of the glandular secretions after these approximately the same intermediate age class [25]. treatments, we took out 30 major workers from each Two entomopathogenic fungi (Beauveria bassiana and colony at a time and divided these randomly into six Metarhizium brunneum—previously called Metarhizium groups of five individuals to be inoculated with the five anisopliae, but now distinguished as a sibling species [26]), different kinds of conidia or to serve as controls. two saprophytic fungi with the potential of causing low The inoculation procedure was as mentioned before, but pathogenicity (Aspergillus niger and Gliocladium virens) and a the ants were killed by freezing in liquid nitrogen 12 h after specialized parasite that attacks the fungal cultivar of inoculation, and gland extractions were carried out immedi- the ants (Escovopsis weberi ) were chosen as disease agents. ately after killing. Dipping in liquid nitrogen serves two Beauveria bassiana (KVL 03-90) and M. brunneum (KVL purposes in glandular extractions: (i) to prevent passive 04-57) were obtained from the stock collection of the flow or active grooming of MG secretions during the extrac- Department of Agriculture and Ecology, University of tion process, and (ii) to halt the chemical reactions of MG Copenhagen, whereas A. niger (DSM 1957) and G. virens secretions so that all secretions collected in the reservoir (a (DSM 1963) were purchased from Deutsche Sammlung sclerotized atrium where glandular secretions were stored von Mikroorganismen und Zellkulturen GmbH. The [6]) represented the same state (after fungal inoculations E. weberi strain was collected from an A. octospinosus nest in or control solution application) at the time of extraction. Gamboa by Hermogenes Fernańdez-Marıń in early 2010, The amount of secretion was measured by piercing the and identified based on morphological characters [27]. MG reservoir (externally visible as the bulla [28]) with a We first tested the pathogenicity and virulence of very fine insect-mounting needle, after which a graduated these five fungal species on A. octospinosus workers by inocu- 10 ml Hamilton syringe was inserted to extract accumula- lating ants with fungal conidia and monitoring their survival. ted secretion. Secretions from both reservoirs of the paired From each colony, we collected a random sample of 60 major glands were extracted in the same syringe (i.e. pooled), and workers, of which we inoculated half and used the other half their volume recorded (MG quantity data in electronic sup- as controls. Inoculations were conducted by gently grasping plementary material, table S3). To ensure that all secretions individuals with a pair of sterilized soft forceps, and pipetting present were extracted with our method, we dissected the 2 ml of fungal (approx. 107 conidia ml21) suspension in 0.05 reservoir from a sample of ants after glandular extractions per cent Triton-X (to avoid conidia clumping) onto the pro- with a fine razor blade and examined the reservoir under podeum. Inoculated individuals were placed separately into the microscope. These reservoirs were found to be empty plastic pots (diameter 2.5 cm, height 4 cm), where they and dry [17]. Similar dissections were carried out on a were maintained at 258C with an ad libitum supply of 10 sub-sample of ants prior to glandular extractions, and these per cent sucrose water. The other half of the workers (control reservoirs were found to be either wet or with secretions group) were treated with 2 ml 0.05 per cent Triton-X solution ‘oozing’ from the punctured point. Extraction and quantifi- applied in the same way. Subsequently, ant mortality was cation procedures were done for all five individuals of the assessed daily for 14 days (survival data in electroic sup- same treatment group in quick succession (usually spanning plementary material, table S1). Ants that died during this 10–20 min). Secretions extracted from the same treatment

Proc. R. Soc. B 85 Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

Cuticular immune system of ants S. H. Yek et al.3 for all five individuals were deposited as droplets on a steri- secretion under unchallenged conditions [17], so we assumed lized microscope slide, after which 10 ml of the pooled that we would be able to measure any increases of secretion sample of MG secretions was collected and immediately dis- volume relative to such controls. solved in 1 ml of pentane, capped to prevent evaporation, To test the differences in inhibitive efficiency between MG and used for testing the quality of these MG secretions. secretions from ants inoculated with different fungal species, we used a repeated-measures multivariate analysis of variance (c) Testing quality of metapleural gland secretions (MANOVA), with halo area of fungal bioassay species as the after challenging with fungal conidia dependent variable set, and fungal species and colony as The efficacy of MG secretions in inoculated and control independent variables. This allowed us to evaluate whether ants was measured with in vitro inhibition assays measuring the identity of inoculated fungal species had an effect on zone of inhibition. All five fungal species were grown on potato the inhibitive activity of MG secretions, and whether any dextrose agar (PDA) plates. Conidia (asexual spores borne on such effect was specific to the fungus tested. hyphae) were harvested, cleaned and diluted to approxima- Totest the differences in inhibitive efficiency between boiled tely 105 conidia ml21, a concentration low enough to prevent MG secretions and non-heated MG secretions, we performed a fungal overgrowth that could mask inhibition halos. Prior to Student’s t-test for the available comparisons. Analyses were using the fungal species for these inhibitive bioassays, assess- carried out using JMP software (v. 9.02, SAS institute). ments of conidia germination were carried out to ensure that the fungal conidia were viable. All fungal species had more than 90 per cent germination (M. brunneum 99%; B. bassiana 3. RESULTS 96%; A. niger 97.8%; G. virens 93.6%; E. weberi 91.5%). The survival of ants differed significantly between the To achieve an even fungal growth, 1 ml volumes of fungal fungal treatments (likelihood-ratio, LR: x2 ¼ 654.18, conidia suspensions were pipetted onto the surface of PDA d.f. ¼ 5, p , 0.0001), with treatment explaining approxi- plates and evenly spread out using a Drigalski spatula. mately 60 per cent of the variation in survivorship 2 PDA plates were divided into six equal segments, and a (R LR ¼ 0.597). Ants inoculated with the entomopatho- disc of sterilized filter paper (5 mm diameter) was placed at genic fungi M. brunneum and B. bassiana consistently the centre of each segment. About 10 ml of the pooled MG suffered greater mortality than those inoculated with the secretion extracted from ants after inoculation was dissolved control solution. Ants inoculated with A. niger also con- in 1 ml pentane (see §2b) and 10 ml was then pipetted onto sistently suffered greater mortality than those inoculated the paper discs. The sixth segment was used as a control with the control solution, but significantly lower mortality and received 10 ml of pentane. To examine whether the inhi- compared with ants inoculated with both entomopatho- bitive effects of MG secretions were due to enzymatic genic fungi (figure 1). Mortalities of ants inoculated activities, we performed a sub-sample of inhibition assays with G. virens and E. weberi did not differ significantly (MG secretions from ants treated with M. brunneum and from mortalities of ants inoculated with the control sol- control solution) by incubating the MG secretions in a ution (figure 1), confirming previous findings of water bath at 1008C for 5 min before application to denature E. weberi conidia not being harmful to ants [29]. The the proteinaceous components. overall survival of ants differed significantly between the All agar plates were incubated at 258C for 24 h, after four experimental colonies (LR: x2 ¼ 14.47, d.f. ¼ 3, p ¼ which the halos in the fungal mats caused by the inhibitive 0.0023), reflecting resistance variation between colonies. action of the applied secretions were measured (see panel There was also a significant interaction between colony in figure 3). The zone-of-inhibition assay was replicated on and fungal treatment (LR: x2 ¼ 36.41, d.f. ¼ 15, p ¼ ten PDA plates for each fungal species (inhibitive halo 0.0015), reflecting small but consistent differences in how radius data in electronic supplementary material, table S4). each colony responded to each fungus (figure 1), although colony and the interaction between colony and treatment (d) Statistical analyses each explained less than 5 per cent of the variance in survival 2 The survival of inoculated ants over 14 days was analysed (R LR ¼ 0.020 and 0.049, respectively). The proportion of using a proportional-hazards model, with colonies, fungal dead ants sporulating differed significantly between the treatments and their interaction as main effects. Surviving fungal treatments (LR: x2 ¼ 150, d.f. ¼ 5, p , 0.0001), individuals were included as censored cases. Post hoc pairwise but not between experimental colonies (LR: x2 ¼ 8.58, differences between colonies and fungal treatments were based d.f. ¼ 3, p ¼ 1.0000). No sporulation was detected for G. on risk-ratio tests, with the significance level being adjusted virens, E. weberi and control treatments, except minor con- with the Bonferroni procedure to correct for multiple compari- tamination from unknown saprophytic fungi that also sons. The proportion of sporulating ants was analysed using a occurred in the controls and were therefore ignored. How- generalized linear model with binomial error structure, with ever, all cadavers from ants exposed to B. bassiana and fungal treatments and colonies as main effects. M. brunneum, and (depending on colony) between 33 and Differences in the amounts of MG secretion produced by 63 per cent of the cadavers of ants exposed to A. niger, ants inoculated with different fungal conidia were analysed produced characteristic conidia. using two-way ANOVA, testing for differences between both Workers from the four colonies did not differ signifi- colonies and fungal treatments, with each ant providing a cantly in the quantities of MG secretion that their large data point consisting of pooled secretion volumes of its two workers produced (F3,120 ¼ 0.08, p ¼ 0.97), nor in their MGs. In the case of overall significance, post hoc multiple- respective responses to the different fungal treatments comparison Tukey’s tests were performed to examine which (F15,120 ¼ 0.77, p ¼ 0.70), so data were pooled for pres- treatments made ants produce significantly more MG entation in figure 2. There were, however, significant secretion. Approximately one-third of the worker MG reser- differences in the amount of MG secretion produced fol- voirs have previously been reported as being filled with lowing the different inoculation treatments (F5,120 ¼

Proc. R. Soc. B 86 Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

4S.H.Yeket al. Cuticular immune system of ants

(a)(colony Ao273 b) colony Ao404 1.0 0.9 A 0.8 A 0.7 A A A 0.6 A 0.5 0.4

proportion of 0.3 ants surviving 0.2 B 0.1 C C C B 0 C

proportion of 0.3 ants surviving 0.2 B B 0.1 CDD CC 0 123456789101112131415 0123456789101112131415 time after application (days) time after application (days) Figure 1. Survival of ants from four Acromyrmex octospinosus colonies (a–d) challenged with conidia of a range of fungal species or a control solution. Different symbols represent fungal conidia challenges or controls: filled circle, an entomopathogenic fungus Metarhizium brunneum; open circle, an entomopathogenic fungus Beauveria bassiana; filled triangle, a potentially mild insect pathogen Aspergillus niger; open triangle, a likely garden weed Gliocladium virens; filled diamond, an ant cultivar parasite Escovopsis weberi; and open diamond, a control solution. Fungal challenges and controls marked with the same letter did not differ significantly in their survival (post hoc Tukey’s test using Bonferroni correction for multiple treatments).

19.45, p , 0.0001), with B. bassiana and M. brunneum 10 treatments eliciting about twice as much MG secretion Beauveria bassiana compared with the controls. Treatments with A. niger, 0 G. virens and E. weberi elicited significantly less secretion 10 than M. brunneum and B. bassiana, and did not differ Metarhizium brunneum significantly from the controls (Tukey’s tests; figure 2). 0 Across all five treatment and control groups, we extracted 10 on average 3.6 + 0.1 ml(mean+ s.e.) of secretion from Aspergillus niger y 120 workers MG reservoirs. From the control treatments, 0 10 we extracted 2.3 + 0.1 ml of secretion from 20 worker Gliocladium virens frequenc MG reservoirs. Using the scale of MG reservoir content 0 developed by Poulsen et al. [17], we estimated that these 10 corresponded to approximately one-third-filled reservoirs Escovopsis weberi on average, whereas the 40 M. brunneum- and B. bassiana- 0 challenged ants had 4.8 + 0.3 mland4.8+ 0.3 mlof 10 secretion in their respective reservoirs, indicating that they control were approximately two-third-filled on average. The 40 0 ants inoculated with A. niger and G. virens had 3.7 + 01234567 0.1 ml and 3.0 + 0.2 ml of secretion in their respective reser- empty volume of MG extract (µl) full voirs, which amounts to them being roughly half-filled. The Figure 2. Histograms showing the frequency distributions of 20 E. weberi inoculated ants had 2.8 + 0.2 ml of secretion in the volume of metapleural gland secretion extracted from their reservoirs on average, which was not significantly workers (n ¼ 20) exposed to each of the five fungi, and con- different from the control ants. In our total sample, four trol treatments. Because there was no effect of colony on ants (3%) had an empty reservoir for unknown reasons: volume of secretion, frequency distributions were drawn one each from the M. brunneum, B. bassiana, G. virens and pooling individuals from all four colonies. E. weberi treatment group (figure 2). Using our volume approximation, four workers had their reservoir almost fully filled with secretion (approx. 6.5 ml)—three from the higher antifungal activity compared with MG secretions B. bassiana and one from the M. brunneum treatment taken from E. weberi and control treatments. There were group (figure 2)—indicating an induced response to no significant between-colony differences in efficacy of infection with entomopathogenic fungi. MG secretions or interactions between colony and fungal There were highly significant differences in the treatment, indicating that the response was colony-inde- efficiency of antifungal activity of MG secretions from pendent. Across colonies, however, there were significant ants treated with different fungal species (table 1 and differences in the sensitivity of the different fungi to MG figure 3). Secretions taken from ants inoculated with secretions (table 1), with G. virens being somewhat more B. bassiana, M. brunneum, A. niger and G. virens exhibited sensitive and B. bassiana less sensitive to MG secretions

Proc. R. Soc. B 87 Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

Cuticular immune system of ants S. H. Yek et al.5

3.0

2.5 A A 2.0 AB A A Aspergillus AB A niger Metarhizium 1.5 B B brunneum control B Escovopsis B weberi B 1.0 C Gliocladium halo radius (mm) Beauveria virens C C bassiana 0.5 C

0 M. brunneum B. bassiana A. niger G. virens E. weberi bioassay Figure 3. Antifungal activity of metapleural gland secretions extracted from worker ants exposed to six treatments and tested under five types of fungal bioassays, measured as the width of the halo in a zone-of-inhibition assay (mean + 95% CI). Meta- pleural gland secretions from ants treated with entomopathogenic fungi (Metarhizium brunneum and Beauveria bassiana) and saprophytic/weedy/mildly pathogenic fungi (Aspergillus niger and Gliocladium virens) had significantly higher antifungal activities than those treated with the fungus garden parasite (Escovopsis weberi ) and the controls. Within each bioassay, clusters of treatments marked with the same letter did not differ significantly in the size of halo (post hoc Tukey’s test using Bonferroni correction for multiple bioassays). To the right, an example PDA bioassay is shown, with the halo radius for B. bassiana marked.

Table 1. Repeated-measures MANOVA, testing the inhibitory efficiency of metapleural gland secretions of major workers of A. octospinosus (dependent variable) across five fungal infections and controls, after inoculation with the same fungal species (treatment) and with four different colonies as independent factors. For within-bioassays, we give approximate F statistics based on Wilks’s lambda; for between bioassays, we report exact F statistics. source F d.f. p within treatments bioassay (plated fungi) 9.37 4,93 ,0.0001 bioassay Â treatment 4.78 20,309.4 ,0.0001 bioassay Â colony 1.44 12,246.35 0.145 bioassay Â colony Â treatment 0.67 60,365.25 0.970 between treatments treatment 224.28 5,96 ,0.0001 colony 0.09 3,96 0.965 treatment Â colony 1.13 15,96 0.341 than the other species (figure 3). There was also a signifi- 4. DISCUSSION cant interaction between the species of fungus to which (a) Functional plasticity in metapleural gland worker ants were exposed and the sensitivity of the responses fungus to the MG secretions of those ants (table 1), Infections with virulent M. brunneum and B. bassiana ento- which was primarily due to a relatively constant response mopathogens triggered both more MG secretion and more of the bioassay fungi to MG secretions for controls and efficient inhibition per unit of MG secretion. These results ants treated with E. weberi, and a different pattern of effi- are consistent with those of Bot et al. [5], who showed that cacy of MG secretions from ants inoculated with M. both conidia and hyphae of these pathogens are inhibited brunneum or B. bassiana (figure 3). similarly by seven different classes of chemical compounds There were no significant differences in the inhibitive from the MG secretions of A. octospinosus. The mild patho- efficiency between boiled MG secretions and non-heated gen A. niger has an interesting intermediate position, as MG secretions (t ¼ 0.94, d.f. ¼ 1, p ¼ 0.350 for boiled infections resulted in more potent MG secretions (figure 3) control versus control; t ¼ 20.77, d.f. ¼ 1, p ¼ 0.443 for but without increasing the secretion volume (figure 2). boiled secretions treated with M. brunneum inoculations This may be related to conidia germination, but not hyphal versus non-heated secretions treated with M. brunneum growth, of both A. niger and G. virens being less efficiently inoculations), indicating that the antifungal ingredients inhibited by several compounds in the MG secretion than were not enzymatic. This result concurs with an earlier conidia of the two virulent entomopathogens [5]. study of MG secretion antibiotic properties in Myrmecia While previous authors [28,31] concluded that ant MGs gulosa ants [30]. seemed to be somewhat disappointingly simple, because no

Proc. R. Soc. B 88 Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

6S.H.Yeket al. Cuticular immune system of ants

indoleacetic acid (IAA) in Acromyrmex increased efficiency in neutralizing disease (function to be confirmed) threats to own survival in Acromyrmex (present study)

phenylacetic acid (PAA) in Atta specific increased efficiency in neutralizing disease (function to be confirmed) threats to self and garden-symbiont in Atta [21,23] constitutive

induced acidic (low pH) metapleural gland secretion active dispersion of metapleural gland secretion for general antimicrobial hygiene [3,5] through MG-grooming of self and garden in Atta [21] (Acromyrmex and Atta) [9,19]

baseline secretory rates resulting in approx. half- to two-third-filled MG-reservoirs resulting from one third-filled MG-reservoirs [17]

non-specific increased rates of metapleural gland secretion passive flow of metapleural gland secretion in response to disease threat to self in Acromyrmex over the cuticle (Acromyrmex and Atta) [6,28] (present study)

Figure 4. Diagrammatic representation of the analogies between generally established immune system functions and metapleural gland defences in Acromyrmex and Atta leaf-cutting ants, with references to published studies where they were available. The upper-left quadrat refers to hypothetical inferences of specific chemicals having constitutive hygiene functions in leaf-cutting ants that remain to be tested for their specific efficacy. muscles were present to directly regulate the emergence of of studies on leaf-cutting ants [5,33], and are probably secretion to the cuticular surface, our present data suggest due to the abundant presence of organic acids in MG that the production of MG secretion is remarkably plastic secretions [19]. and appears to be both quantitatively (figure 2) and quali- The second component in this quadrat is that at least tatively (figure 3) adjusted to specific fungal infection some secretion is normally present in the MGs of all threats. Given these conditional responses, the question ants, although the amount in the reservoirs of unchal- now seems almost the reverse: why have behaviourally con- lenged ants varies, probably as a function of nutrition trolled muscular release mechanisms for MG secretion not and age [19], so that some ants (3% in our study) end been favoured by selection? up having reservoirs that are scored as empty [3,17]. Our results suggest that the MGs of leaf-cutting ants There is thus almost always secretion present for the may function as analogues of a simple prophylactic ants to work with, even under benign circumstances, immune system that operates on the cuticular surface and this constitutive defence amounts to about one- of ant workers. This hypothesis is based on the evidence third of the maximal holding capacity of the MG provided here that MG secretions have both general and reservoirs, on average [17](figure 2). specific functions, and that their production is partly The third component is the direct supply of MG constitutive and partly induced. We will discuss this secretion to the external environment via the rather hypothesis below, using a modified version of a framework large opening of the reservoir [6]. The lack of any auton- diagram from Schmid-Hempel & Ebert [32] that maps omous or behaviourally controlled muscles to control immune defences along perpendicular specificity and release of the secretion [28] indicates a constitutive and inducibility axes (figure 4). We also include our current non-specific defence component of the MGs. understanding of MG chemistry and function in Atta leaf-cutting ants, which are the sister genus of Acromyrmex, (c) Induced, non-specific defence components but different in several key life-history traits. We use this The bottom-right quadrat of figure 4 summarizes the two analogy with the immune system proper because it helps known mechanisms by which the application of MG to interpret our findings in a coherent conceptual frame- secretion for non-specific defence can be induced. The work emphasizing evolutionary adaptation, not because first is active grooming. The location of the MG openings we would claim that the mechanisms and metabolic path- just above the hind-legs allows secretions to be picked ways involved are similar or even related to those of the up by leg movements as soon as they emerge. This form normal immune system of these insects. of dispersing MG secretion [22] is particularly employed when leaf-cutting ants are challenged with fungal (b) Constitutive, non-specific defence components conidia [21,23]. of leaf-cutting ant metapleural glands The frequency of MG grooming and the target of appli- The bottom-left quadrat of figure 4 summarizes the cation is known to vary between Acromyrmex and Atta, with most basal antimicrobial MG functions, which were the Atta substantially increasing their MG grooming rate and first to be discovered [3]. General antimicrobial activity expanding their grooming targets to the garden, and probably stems from the high acidity (pH 2.5–4) of Acromyrmex maintaining much lower MG grooming rates MG secretions [3,9,30]. This acidity and its pH-reducing and primarily targeting the brood [21]. This is consistent effects in fungus gardens are well documented in a series with Atta workers relying more on MG grooming and less

Proc. R. Soc. B 89 Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

Cuticular immune system of ants S. H. Yek et al.7 on antibiotic production by mutualistic actinomycete bac- (e) Constitutive, specific defence components teria, whereas the opposite combination appears to apply The upper-left quadrat in figure 4 largely represents a in Acromyrmex [21] (see §4d for more details). ‘black box’, but there is enough evidence for this defence component to hypothesize that this aspect of the immune (d) Induced, specific defence components system analogy is also likely to be important in fungus- The top-right quadrat of figure 4 summarizes two known growing ants [5,19]. There are major compounds of MG mechanisms in Acromyrmex and Atta leaf-cutting ants by secretions that are likely to qualify as specific constitutive which MG secretion appears to be differentially induced defences, which are different in Atta and Acromyrmex in depending on the type of challenge. The fact that spite of other compounds being present in the MG A. octospinosus workers significantly increase the quantity of secretions of both genera [6]. The prime candidate com- MG secretion only after being challenged with directly lethal pound in Atta is phenylacetic acid, which makes up 72 to insect pathogens such as B. bassiana and M. brunneum indi- 80 per cent of the total MG secretion, but is absent in the cates that the ants are able to classify pathogens according to MG secretion of Acromyrmex. However, Acromyrmex MG the degree of threat to themselves and their nest-mates. The secretion has indoleacetic acid (IAA) as a major component same results (figure 2) suggest that this recognition process is (24–25% of total secretion), whereas this compound is a continuum, rather than an all-or-nothing response, as the found only in trace amounts in the MG secretion of Atta. amount of MG secretion was also elevated somewhat after Further work will be needed to unravel the specific targets challenges with G. virens and A. niger. of these major components. The role of IAA seems particu- Interestingly, induction specificity also included larly intriguing, as this compound is a well-known plant increases in potency of the MG secretion, and here growth hormone [36], but initial tests have shown that it responses were generally similar for all fungal infections does not inhibit fungal conidia or hyphae [5]. except Escovopsis, in spite of the variation in threat It has been repeatedly found that in addition to chemi- across these challenges (figure 3). This underlines that cal compounds that make MG secretion acidic, there is a the MG secretion responses are analogous to immune significant fraction of proteinaceous compounds of system functioning, because challenges both increase unknown identity in the MG secretions of leaf-cutting overall investment in defence (cf. lymphocytes) and ants [3,18]. However, the results of our boiling assay indi- induce the production of specific defence agents (cf. anti- cate that any such proteins do not seem to have the bodies) after the kind of challenge has been identified specific functions found in other insects [37–39]. This [34]. There was difference in how the fungal species underlines that secretions of the MGs of leaf-cutting responded to MG secretions that were induced by par- ants might have a combination of adaptive traits that are ticular fungal conidia (table 1), which suggests that specific for this clade of ants only. there is potential for changing MG secretions to target We thank Louise Lee Munk Larsen for technical assistance in particular threats. However, we found no match between rearing the fungal species and performing the infection which fungal species had induced a particular MG assays, Michael Poulsen for sharing his skills on working secretion and the susceptibility of that species to the with Escovopsis weberi, and the reviewers for their secretion (figure 3). From our heat-denaturing testing comments. All authors were supported by a grant from the and an earlier study [30], we infer that proteinaceous/ Danish National Research Foundation. enzymatic compounds are unlikely to be important as antifungal ingredients. Hence, the identities of compounds that make MG secretion more potent after infections remain unknown, but will be interesting targets REFERENCES for future research. 1Ho¨lldobler, B. & Wilson, E. O. 1990 The ants. The specific defence responses obtained in our present Cambridge, MA: Harvard University Press. study are similar to those obtained by Fernańdez-Marıń 2 Bolton, B. 2003 Synopsis and classification of Formici- et al. [21,23], who showed that MG grooming rates in dae. Memoirs Am. Entomol. Inst. 71, 1–370. Atta colombica were elevated when inoculated with conidia 3 Maschwitz, U., Koob, K. & Schildknecht, H. 1970 Ein of Metarhizium and Escovopsis but not by inert talcum beitrag zur funktion der metathoracaldru¨se der ameisen. powder controls. Whether this response is mediated by J. Insect Physiol. 16, 387–404. (doi:10.1016/0022-1910 (70)90180-0) quantitative and/or qualitative change in MG secretion 4 Beattie, A. J., Turnbull, C. L., Hough, T. & Knox, R. B. is unknown, but it seems reasonable to infer that an 1986 Antibiotic production: a possible function for the increase in grooming rate would only make sense if at metapleural glands of ants (Hymenoptera: Formicidae). least the amount of available MG secretion would also Ann. Entomol. Soc. Am. 79, 448–450. be proportionally increased. There is a significant differ- 5 Bot, A. N. M., Ortius-Lechner, D., Finster, K., Maile, R. & ence between Atta and Acromyrmex in that MG grooming Boomsma, J. J. 2002 Variable sensitivity of fungi and bac- in the former also has a major function for infection control teria to compounds produced by the metapleural glands in the fungus garden, whereas this is not so in the latter of leaf-cutting ants. Insectes Soc. 49, 363–370. (doi:10. [21]. This is consistent with Acromyrmex maintaining cul- 1007/PL00012660) tures of actinomycete bacteria on their exoskeleton to 6 Yek, S. H. & Mueller, U. G. 2010 The metapleural gland of ants. Biol. Rev. Camb. Phil. Soc. 91, 201–224. (doi:10. control Escovopsis infections, whereas Atta has abandoned 1111/j.1469-185X.2010.00170.x) this form of biological control (possessed by most basal 7 Schmid-Hempel, P. 1995 Parasites and social insects. attine ants [35]) in exchange for chemical control via MG Apidologie 26, 255–271. (doi:10.1051/apido:19950307) grooming [21]. It is therefore not surprising that Atta 8 Boomsma, J. J., Schmid-Hempel, P. & Hughes, W. O. H. has inducible specific MG defences against Escovopsis and 2005 Life histories and parasite pressure across the major Acromyrmex has not. groups of social insects. In Insect evolutionary ecology

Proc. R. Soc. B 90 Downloaded from rspb.royalsocietypublishing.org on August 23, 2012

8S.H.Yeket al. Cuticular immune system of ants

(eds M. Fellowes, G. Holloway & J. Rolff), pp. 139–176. 23 Fernańdez-Marıń, H., Zimmerman, J. K., Rehner, S. A. & Wallingford, UK: CABI. Wcislo, W.T.2006 Active use of the metapleural glands by 9 Maschwitz, U. 1974 Vergleichende Untersuchungen zur ants in controlling fungal infection. Proc. R. Soc. B 273, Funktion der Ameisenmetathorakaldru¨se. Oecologia 16, 1689–1695. (doi:10.1098/rspb.2006.3492) 303–310. (doi:10.1007/BF00344739) 24 Little, A. E. F., Murakami, T., Mueller, U. G. & Currie, 10 Jaffe, K. & Puche, H. 1984 Colony-specific territorial C. R. 2006 Defending against parasites: fungus-growing marking with the metapleural gland secretion in the ants combine specialized behaviours and microbial ant Solenopsis geminata (Fabr). J. Insect Physiol. 30, symbionts to protect their fungus gardens. Biol. Lett. 2, 265–270. (doi:10.1016/0022-1910(84)90126-4) 12–16. (doi:10.1098/rsbl.2005.0371) 11 Wilson, E. O. 1980 Caste and division of labor in leaf- 25 Ortius-Lechner, D., Petersen, H. C., Boomsma, J. J., cutter ants (Hymenoptera: Formicidae: Atta). Behav. Maile, R. & Morgan, E. D. 2003 Lack of patriline-specific Ecol. Sociobiol. 7, 143–156. (doi:10.1007/BF00299520) differences in chemical composition of the metapleural 12 Bot, A. N. M. & Boomsma, J. J. 2011 Variable meta- gland secretion in Acromyrmex octospinosus. Insectes Soc. pleural gland size-allometries in Acromyrmex leafcutter 50,113–119.(doi:10.1007/s00040-003-0640-1) ants (Hymenoptera: Formicidae). J. Kans. Entomol. Soc. 26 Bischoff, J. F., Rehner, S. A. & Humber, R. A. 2009 A 69, 375–383. multilocus phylogeny of the Metarhizium anisopliae line- 13 Hughes, W. O. H., Pagliarini, R., Madsen, H. B., age. Mycologia 101, 512–530. (doi:10.3852/07-202) Dijkstra, M. B. & Boomsma, J. J. 2008 Antimicrobial 27 Muchovej, J. J. & Della Lucia, T. M. 1990 Escovopsis, defense shows an abrupt evolutionary transition in the a new genus from leaf cutting ant nests to replace fungus-growing ants. Evolution 62, 1252–1257. (doi:10. Phialocladus nomem invalidum. Mycotaxon 37, 191–195. 1111/j.1558-5646.2008.00347.x) 28 Bot, A. N. M., Obermayer, M. L., Ho¨lldobler, B. & 14 North, R. D., Jackson, C. W. & Howse, P. E. 1997 Boomsma, J. J. 2001 Functional morphology of the meta- Evolutionary aspects of ant-fungus interactions in leaf- pleural gland in the leaf-cutting ant Acromyrmex cutting ants. Trends Ecol. Evol. 12, 386–389. (doi:10. octospinosus. Insectes Soc. 48, 63–66. (doi:10.1007/ 1016/S0169-5347(97)87381-8) PL00001747) 15 Villesen, P., Murakami, T., Schultz, T. R. & Boomsma, J. 29 Reynolds, H. T. & Currie, C. R. 2004 Pathogenicity of J. 2002 Identifying the transition between single and mul- Escovopsis weberi: the parasite of the attine ant-microbe tiple mating of queens in fungus-growing ants. symbiosis directly consumes the ant-cultivated fungus. Proc. R. Soc. Lond. B 269, 1541–1548. (doi:10.1098/ Mycologia 96, 955–959. (doi:10.2307/3762079) rspb.2002.2044) 30 Mackintosh, J. A., Trimble, J. E., Beattie, A. J., Veal, 16Hughes,W.O.H.,Bot,A.N.M.&Boomsma,J.J.2010 D. A., Jones, M. K. & Karuso, P. H. 1995 Antimicrobial Caste-specific expression of genetic variation in the size of mode of action of secretions from the metapleural antibiotic-producing glands of leaf-cutting ants. Proc. R. gland of Myrmecia gulosa (Australian bull ant). Soc. B 277, 609–615. (doi:10.1098/rspb.2009.1415) Can. J. Microbiol. 41, 136–144. (doi:10.1139/m95-018) 17 Poulsen, M., Bot, A. N. M., Nielsen, M. G. & Boomsma, 31 Schoeters, E. & Billen, J. 1993 Anatomy and fine-structure J. J. 2002 Experimental evidence for the costs and hygie- of the metapleural gland in Atta (Hymenoptera, nic significance of the antibiotic metapleural gland Formicidae). Belgian J. Zool. 123, 67–75. secretion in leaf-cutting ants. Behav. Ecol. Sociobiol. 52, 32 Schmid-Hempel, P. & Ebert, D. 2003 On the evolution- 151–157. (doi:10.1007/s00265-002-0489-8) ary ecology of specific immune defence. Trends Ecol. Evol. 18 Nascimento, R. R., Schoeters, E., Morgan, E. D., 18, 27–32. (doi:10.1016/S0169-5347(02)00013-7) Billen, J. & Stradling, D. J. 1996 Chemistry of metapleural 33 Papa, F. & Papa, J. 1982 Etude de l’activite microbiologi- gland secretions of three attine ants, Atta sexdens rubropilosa, que dans les nids d’Acromyrmex octospinosus Reich en Atta cephalotes,andAcromyrmex octospinosus (Hymenoptera: Guadeloupe. Bull. Soc. Pathol. Exotiq. 75, 404–414. Formicidae). J. Chem. Ecol. 22, 987–1000. (doi:10.1007/ 34 Vilmos, P. 1998 Insect immunity: evolutionary roots of BF02029949) the mammalian innate immune system. Immunol. Lett. 19 Ortius-Lechner, D., Maile, R., Morgan, E. D. & 62, 59–66. (doi:10.1016/S0165-2478(98)00023-6) Boomsma, J. J. 2000 Metapleural gland secretion of the 35 Cafaro, M. J. & Currie, C. R. 2005 Phylogenetic analysis leaf-cutter ant Acromyrmex octospinosus: new compounds of mutualistic filamentous bacteria associated with and their functional significance. J. Chem. Ecol. 26, fungus-growing ants. Can. J. Microbiol. 51, 441–446. 1667–1683. (doi:10.1023/A:1005543030518) (doi:10.1139/w05-023) 20 Poulsen, M., Bot, A. N. M., Currie, C. R. & Boomsma, 36 Schildknecht, H. & Koob, K. 1971 Myrmicacin, the J. J. 2002 Mutualistic bacteria and a possible trade-off first insect herbicide. Angew. Chem. Int. Ed. Engl. 10, between alternative defence mechanisms in Acromyrmex 124–125. (doi:10.1002/anie.197101241) leaf-cutting ants. Insectes Soc. 49, 15–19. (doi:10.1007/ 37 Bulet, P. 1999 Antimicrobial peptides in insects; struc- s00040-002-8271-5) ture and function. Dev. Comp. Immunol. 23, 329–344. 21 Fernańdez-Marıń, H., Zimmerman, J. K., Nash, D. R., (doi:10.1016/S0145-305X(99)00015-4) Boomsma, J. J. & Wcislo, W. T. 2009 Reduced biological 38 Duman, J. & Horwath, K. 1983 The role of hemolymph control and enhanced chemical pest management in the proteins in the cold tolerance of insects. Annu. Rev. evolution of fungus farming in ants. Proc. R. Soc. B Physiol. 45, 261–270. (doi:10.1146/annurev.ph.45. 276, 2263–2269. (doi:10.1098/rspb.2009.0184) 030183.001401) 22 Brown, W. L. 1968 An hypothesis concerning the 39 Pelosi, P. & Maida, R. 1995 Odorant-binding proteins in function of the metapleural gland in ants. Am. Nat. insects. Comp. Biochem. Physiol. B Biochem. Mol. Biol. 102, 188–191. (doi:10.1086/282536) 111, 503–514. (doi:10.1016/0305-0491(95)00019-5)

Proc. R. Soc. B 91 CHAPTER 3

Towards a better understanding of the evolution of specialized parasites of fungus-growing ant crops

Escovopsis quickly overtake the fungus garden in the absence of ants

Photo credit: www.dimijianimages.com

Published in Psyche

92 Hindawi Publishing Corporation Psyche Volume 2012, Article ID 239392, 10 pages doi:10.1155/2012/239392

Review Article Towards a Better Understanding of the Evolution of Specialized Parasites of Fungus-Growing Ant Crops

Sze Huei Yek, Jacobus J. Boomsma, and Michael Poulsen

Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark

Correspondence should be addressed to Sze Huei Yek, [email protected]

Received 5 October 2011; Accepted 12 December 2011

Academic Editor: Alain Lenoir

Copyright © 2012 Sze Huei Yek et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Fungus-growing ants have interacted and partly coevolved with specialised microfungal parasites of the genus Escovopsis since the origin of ant fungiculture about 50 million years ago. Here, we review the recent progress in understanding the patterns of speciﬁcity of this ant-parasite association, covering both the colony/population level and comparisons between phylogenetic clades. We use a modiﬁed version of Tinbergen’s four categories of evolutionary questions to structure our review in complementary approaches addressing both proximate questions of development and mechanism, and ultimate questions of (co)adaptation and evolutionary history. Using the same scheme, we identify future research questions that are likely to be particularly illuminating for understanding the ecology and evolution of Escovopsis parasitism of the cultivar maintained by fungus-growing ants.

1. Introduction The clonal propagation of cultivars through vertical transmission [2, 13] predicts ancient association and con- 1.1. The Attine Fungus-Growing Ants. Fungus-growing ants gruence between the ant and fungal cultivar phylogenies. (Hymenoptera: Formicidae: Attini) form a monophyletic High degrees of congruence have indeed been found at tribe of primarily tropical ants that obligately depend on the deep phylogenetic levels in both higher [14, 15]and fungal cultivars (Agaricales: mostly Lepiotacea: Leucoco- lower attines [12, 14]. However, the phylogenetic inter- prineae). The ants provide the fungus with optimal growth action specificity breaks down within, and occasionally conditions, and in return, the fungus serves as the main between, ant genera and their cultivar strains, indicating that food source for the ants [1, 2]. The symbiosis between switches and/or reacquisitions of new garden cultivars have fungus-growing ants and their fungi originated about 50 occurred (e.g., [12, 16–19]). While the higher-attine fungi million years ago [3–6] from a single ancestor that was no longer persist outside the symbiosis, lower attine fungi most likely a generalist forager [3]. Subsequently, the Attini have free-living close relatives, which is likely to facilitate have diversified to encompass approximately 230 described gene flow and reacquisitions of symbionts [12]. Cultivar species, distributed across 14 ant genera [4, 7]. Colonies switches can be induced in the laboratory, including the of fungus-growing ants are typically founded by a single formation of chimeric gardens [20–22]. However, consistent queen, who carries a piece of the fungus cultivar from with predictions from host-symbiont conflict theory [23], her natal colony in the infrabuccal pocket [8] during her mature individual colonies appear to consistently maintain nuptial flight [9–11]. The Attini are divided into the “higher a single fungus clone, at least in leaf-cutting ants (Atta and attine” and the phylogenetically basal “lower attine” genera Acromyrmex) where this has been best studied [24, 25]. based on their cultivars [5]. Lower attine cultivars are largely Clonally propagated monoculture crops are expected unmodified and resemble free-living Leucocoprini, whereas to be particularly prone to infection with parasites and higher attine cultivars (including those of leaf-cutting ants) pathogens [26], because they represent an attractive resource are highly derived [12]. that should be easy to exploit. This “Red Queen” logic 93 2 Psyche

[27, 28] assumes that parasites and hosts are involved in of association specificity is apparent for the lower attine evolutionary arms races, in which unpredictable genetic genus Cyphomyrmex, where subclades of a single Escovopsis heterogeneity, due to sexual recombination, is the most morphotype (pink) are phylogenetic congruent with corre- powerful defence against parasites that have short generation sponding clades of cultivar host genotypes [41]. In the higher times relative to their hosts [28–30]. Single asexual cultivar attine ants, Trachymyrmex and Sericomyrmex are infected by clones thus seem to represent a liability for the farming specific Escovopsis parasites that are phylogenetically distinct symbiosis [25]thatneedstobeovercomebyactiveprotection from the two clades that parasitize Atta and Acromyrmex by the host ants (see below). Colony-level monoculture does leaf-cutting ants [37, 38]. Within the leaf-cutting ant genera, not imply population-wide monoculture, as is often the case Escovopsis infections are nonspecific [38], confirming the in modern human crops that are vulnerable to disease. With high degree of ant-cultivar specificity of all extant leaf-cut- the possible exception of some species, there is likely to be ting ants to a single species of Attamyces symbiont [42]. considerable strain diversity across neighbouring colonies [16, 18] that should discourage the spread of infections between colonies. 1.3. Defence Strategies against Escovopsis . Fungus-growing ants, especially the leaf-cutting ants, have elaborate prophylactic fungus grooming and weeding behaviours to keep 1.2. Specialised Coevolved Parasites. Microfungal parasites in their cultivar free from parasites [44, 45]. In Acromyrmex, the genus Escovopsis (anamorphic Hypocreales) have been minor workers are particularly efficient at restricting spore known for more than a century to overgrow fungus gardens germination [45], and major workers appear to recruit minor of laboratory colonies [1, 13, 31], but the formal status of workers to infected sites, thereby potentially increasing the Escovopsis as a disease was confirmed only just over a decade efficiency of disease suppression [46]. If spores manage ago when Currie et al. [32] showed that Escovopsis fulfils to escape the attention of minor workers and germinate, Koch’s four defining postulates [33] for causative disease major workers appear to perform the task of removing agents. This included evidence that Escovopsis (i) is found in infected garden pieces (weeding) [45]. Task specialization abundance in diseased but infrequently in apparently healthy between castes thus appears to make hygienic policing more colonies, (ii) can be isolated from diseased colonies, (iii) efficient in general, which has been proposed to be normally can cause disease when colonies are artificially infected, and ffi (iv) can be reisolated from diseased experimental colonies su cient for eliminating generalist fungal parasites, but not [32]. It was also shown that Escovopsis has a directly negative for completely eradicating Escovopsis infections [44]. impact on the ant cultivar [32, 34, 35] through the secretion To control Escovopsis infections, fungus-growing ants of compounds that break down the cultivar mycelium [36]. may also use metapleural gland secretions, which contain As fungus-growing ants rely on healthy fungus gardens for an array of compounds with antibiotic properties [48, 69]. growth and reproduction, this implies that Escovopsis is a In a seminal study, Fernandez-M´ ar´ın et al. [47]described potentially serious threat to ant fitness [35]. highly coordinated and challenge-specific foreleg movements Deep-level phylogenetic congruence has been found along the metapleural gland opening (metapleural gland between the fungus-growing ants, their cultivars, and Escov- grooming), which allowed Atta and Acromyrmex ants to opsis parasites, suggesting a long history of codivergence precisely target the application of antibiotic secretion to their within the attine agricultural systems [37]. However, cophy- gardens. In combination with metapleural gland grooming, logenies at lower levels appear to be punctuated with occa- fungus-growing ants utilize their infrabuccal pocket (located sional host switching of the parasites [38], consistent with in the oral cavity) as a further filtering and sterilising device. ongoing arms races [37], although null hypotheses of genetic After grooming, the ants collect Escovopsis spores in this drift in isolated parasite populations can usually not be pocket, where they are sterilised by an as-of-yet unknown dismissed. mechanism (potentially metapleural gland compounds), Even within ant genera, there is some evidence for ant- after which the infrabuccal pellet is expelled on the colony cultivar-Escovopsis pairing specificity. Four morphologically refuse pile [47, 49, 70]. and genetically distinct Escovopsis types parasitize the cul- The cuticle of major garden workers is often covered tivars maintained by Apterostigma, a basal fungus-growing with a thick white growth of Actinobacteria [50, 51], which ant genus [39]. These have so far been categorised as produce antimicrobial compounds that aid in the protection “brown,” “yellow,” “pink,” and “white,” but are genetically of the fungal cultivar from Escovopsis [49–51, 64, 71]and distinct and likely different subspecies or species (cf. [39]). possibly other parasites [65]. These beneficial Actinobacteria Even within these groupings, there is evidence for speci- are reared by the ants and housed in cuticular crypts, tuber- ficity: “pink” Escovopsis appears to infect only G3 cultivars cles, or other modifications associated with subcuticular and(rare)“white”Escovopsis only G2 cultivars, whereas exocrine glands [52]. Most work on the Actinobacteria has “brown” and “yellow” Escovopsis commonly coinfect G2 focused on specifically associated lineages of Pseudonocardia cultivars (cf. Table 1 in [39]). Current evidence suggests [51, 55]. Pseudonocardia appears to be vertically transmitted that these pathogen lineages display patterns of phyloge- by default [50], but phylogenetic evidence indicates that netic congruence with their fungal host [39], maintained events of horizontal transfer and incorporation of free-living by chemotaxis and host resistance in nonnative (i.e., not Pseudonocardia to the symbiosis have occurred [55–57]. naturally occurring) combinations [40]. A similar scenario Recent studies have further shown that other Actinobacteria

94 Psyche 3 genera (mainly Streptomyces) are often also present [57– 2. Using Tinbergen’s Four Quadrats to Structure 61], but their degree of specificity with the symbiosis is Attine-Escovopsis Research less clear. There is little doubt that cuticular Actinobacteria cultures serve active defence functions in the symbiosis, but Nikolaas Tinbergen was a Dutch ethologist and ornithologist clarifying the relative importance of predominantly vertically who received a Nobel Prize in Physiology or Medicine in transmitted Pseudonocardia and horizontally transmitted 1973 together with Karl von Frisch and Konrad Lorenz other defensive microbes will need much further work. for their joint work on the organization and elicitation of Ant cultivars, the hosts of Escovopsis parasitism, are individual and social behaviour in animals [74]. Tinbergen’s able to launch defences themselves by secreting chemical four categories of evolutionary questions were originally compounds that suppress Escovopsis growth. This has been developed to obtain an integrated explanation for animal tested in the Apterostigma and Cyphomyrmex [41, 43], where behaviour, based on complementary understanding of prox- antifungal compounds secreted by the cultivar appeared to imate mechanisms (1) and ontogenetic developments (2), be more effective in suppressing the growth of Escovopsis as well as ultimate selection forces resulting in adaptive strains that are unknown to infect them in nature, but less evolution of individuals (3) and long-term evolutionary effective against their native Escovopsis strains [41, 43]. Such change of populations or higher-level clades (phylogenetic cultivar responses towards novel Escovopsis strains might history) (4) [75]. Tinbergen’s framework has since been result in limitations for Escovopsis host switching outside used in many research programs throughout the life sciences the agricultural system that they are adapted to. Overall, [76–78] but has, to our knowledge, not been applied to therefore, the defences of the ants, the Actinobacteria, and host-parasite interactions. For the purpose of the present the cultivar appear to reinforce each other in suppressing paper, we modify Tinbergen’s framework to encompass Escovopsis infection and proliferation within attine ant a classification of questions that have been (Table 1), or fungus gardens (see e.g., Figure 10.1 in [22]). could be (Figure 1), addressed to better understand the evolutionary ecology of attine ant-Escovopsis interactions. Table 1 summarizes how studies available so far can 1.4. Trade-Offs between Alternative Defence Functions. Over be grouped into Tinbergen four quadrats framework. This the course of millions of years of selection on the interaction was relatively straightforward for the ultimate questions of between fungus-growing ants and Escovopsis,different ant adaptive evolution and phylogenetic history, but not always genera have diversified in their specific utilization and for the proximate ontogeny and mechanism categories, be- combination of alternative defence mechanism to reduce the cause available research tools have so far not allowed much impact of Escovopsis. This has been best studied in species of understanding of the (epi)genetics behind developmental the leaf-cutting ant genera Atta and Acromyrmex. Escovopsis pathways and phenotypic plasticity. It is, therefore, also ar- infections appear to be more prevalent in Acromyrmex guable that the questions addressed in our ontogeny and than Atta colonies [35], possibly due to differences in the mechanism categories are rather ambiguous, in being both efficiency of alternative defensive strategies. First, the chem- technologically challenging and relatively imprecise in their ical compounds in the metapleural glands differ between fit to a single Tinbergen quadrat. We nonetheless felt that Acromyrmex than Atta,reviewedin[53], making it inevitable making a first attempt to structure a research agenda was that compounds with different antimicrobial properties are worthwhile and have chosen to group questions of Escovopsis produced (cf. [48]). Second, Actinobacteria are abundant in specialization in the ontogeny quadrat and questions of Acromyrmex and essentially absent in Atta [54]. Third, the cultivar utilization and defences by the ants and fungal sym- rate of metapleural gland grooming differs in a contrasting bionts in the mechanism quadrat. In the sections below, we manner, with Atta increasing grooming rates after Escovopsis utilise these groupings to formulate how new experimental infection and Acromyrmex maintaining a constantly low rate work, combined with the increasing availability of genome of metapleural gland grooming [54]. sequences, may allow novel insights in Escovopsis parasitism. Differences in metapleural gland chemistry, grooming rate, and Actinobacteria coverage indicate that trade-offs 3. Tinbergen’s Ontogeny Quadrat between these alternative defensive strategies are likely, conceivably because these defences are known to be costly 3.1. Escovopsis Recognition of Cultivars. In vitro assays have [72, 73]. Different defences may target the same parasite, but shown that Escovopsis canrecognizenativecultivarhosts with different modes of action. For example, in Acromyrmex, through chemotaxis, followed by directed growth of the par- metapleural gland secretions kill Escovopsis spores but show asite towards the cultivar, the secretion of parasite enzymes limited effect on hyphae [48], while Actinobacteria secretions breaking down cultivar cells, and absorption of cultivar cell suppress hyphal growth but do not kill spores [64]. A similar contents [36]. In contrast, Escovopsis is not able to utilize scenario has been proposed for two other genera of higher nonnative cultivar strains and can even be inhibited by them attine ants, Trachymyrmex and Sericomyrmex,ascertain [41, 43]. The mechanisms and genes underlying parasite species from the former genus have abundant Actinobacteria differentiation between native and nonnative host cultivars cover and low frequencies of metapleural gland grooming, remain unknown, that is, we neither know the identity or the while Sericomyrmex has very few Actinobacteria and a higher evolution of the chemicals (what does Escovopsis recognize?) frequency of metapleural gland grooming [72]. nor the genes coding for the chemicals produced and their

95 4 Psyche

Table 1: Overview of available studies on Escovopsis virulence in gardens of fungus-growing ants, and our assortment of these studies into the four Tinbergen quadrats.

Quadrat Study focus References Ontogeny Pathology, impact, and prevalence [32, 34, 35] Genetic and chemical basis of Escovopsis recognition of cultivars [36, 38–41, 43] Mechanism Ant behavioural defences [44–47] Chemical defences [47, 48] Actinobacteria defences [49–52] Cultivar defences [40, 41, 43] Phylogeny Population-level specificity [38–41, 43] Cross-phylogeny specificity [38, 51] Adaptation Susceptibility/resistance to metapleural gland compounds ([48], reviewed in [53, 54]) Degree of Actinobacteria specificity [55–63] Susceptibility/resistance to Actinobacteria secretions [50, 58, 60, 61, 64, 65] Host cultivar use [38–41, 43]

Objects of explanation

Development/Historical Single form What organisms need to function and Progression in current form why those functions arose otnO g ne y sinahceM m (i) Escovopsis recognition of cultivars Proximate (ii) Genetic basis for Escovopsis (i) Escovopsis transmission between How organisms work by describing their recognition by the ants colonies developmental and functional traits (iii) Trade-oﬀsbetween alternative (ii) Colony-level virulence defences hP y ol g ne y Adap noitat

Questions (i) Origin anddiversiﬁcation of the (i) Evolutionary potential of Evolutionary association Escovopsis as a parasite How evolution has shaped organisms to (ii) Phylo-geographic patterns, (ii) Evolutionary consequences of acquire their extant forms coevolutionary interactions, and host-parasite interactions dispersal

Figure 1: Tinbergen’s four quadrat framework applied to evolutionary questions about Escovopsis parasitism of fungus-farming ant crops. Ontogeny refers to the description of development, from DNA to progressive phenotype, mechanism refers to the physiological and cellular processes that organisms have available, phylogeny refers to the idiosyncratic evolutionary history of a lineage, and adaptation refers to traits that acquired their extant function because of specific selective advantages, modified from [66–68]. evolutionary history. Ongoing genome sequencing of culti- have left signatures of enhanced dN/dS ratios compared to vars and Escovopsis,aswellasefforts to isolate the chemicals housekeeping and neutral genes, while nonsignificant dN/dS involved, will thus allow considerable progress to be made. ratios would make the nonadaptive null hypothesis more Two evolutionary explanations for the maintenance of likely. In general, it seems unlikely that Escovopsis popu- Escovopsis-cultivar utilization patterns seem possible. The lations are highly structured (see also below), but solid nonadaptive explanation would hold that Escovopsis strains empirical evidence on this is lacking. (or species) would be subject to consistent genetic drift in isolated populations, so that they would lose adaptations to 3.2. Genetic Basis for Escovopsis Recognition by the Ants. Ants allopatric hosts by chance. The alternative adaptive expla- are able to discriminate between Escovopsis and other fungi nation would hold that populations are mostly panmictic, and behave accordingly [44, 45]. Natural selection in the ant so that genes coding for innovative pathogen traits and host is expected to select for genes involved in the recognition defensive recognition and resistance traits of cultivars would and removal of Escovopsis from the fungus garden, as this is tend to coevolve. If so, Escovopsis would track cultivar predicted to provide a selective advantage. Further, Escovopsis evolution in continuous, but variable, arms races reminiscent has the potential to be much more virulent than any general ofageographicmosaicofcoevolution[79]. If the latter is the fungal weeds of attine ant colonies, at least in the higher case, expectations are that positive selection on specific gene attine system where virulence has been studied, implying complexes (e.g., recognition or resistance genes) will likely stronger selection on Escovopsis recognition pathways in the

96 Psyche 5 ants compared to pathways mediating the recognition of is unlikely to be the case for Escovopsis because it sporulates weed fungi. The genetic basis of Escovopsis recognition has inside colonies and has wet spores [35]. The mechanism of not been explored, but genomic tools will make this possible Escovopsis transmission, therefore, continues to be enigmatic, intheyearstocome[80, 81]. For example, two leaf-cutting with untested hypotheses of commensal garden arthropods ant genomes are now published [82, 83]andasequenced vectoring spores between colonies, or foraging ants picking Escovopsis genome will soon follow (anonymous reviewer, up spores via encounters outside the nest as reasonable personal communication), providing the tools necessary for leads [41, 84]. Both mechanisms could be further facilitated such new approaches to studying behavioural recognition by attine colonies nesting in each others close proximity. mechanisms. Recognition of, and concomitant behavioural Culture-based attempts to isolate Escovopsis from potential responses to, Escovopsis infection are faster and last longer vectors are, therefore, needed for a better understanding of than the response to general fungal pathogens [44, 47], transmission modes. Expectations are that Escovopsis is more leading to the prediction of higher levels of recognition likely to be transmitted between colonies by commensal gene expression in the presence of Escovopsis.However,itis arthropods. This is so, because foragers presumably rarely, if conceivable that the mechanism of recognition of Escovopsis ever, enter other colonies, and are therefore unlikely to pick and other fungi by the ants does not differ but that responses up Escovopsis spores from nonnative infected colonies, and do, so that it is rather genes underlying behavioural removal becauseworkersareefficient at recognizing and removing responses that are differentially expressed. Escovopsis spores from their cuticle (e.g., [85, 86]). In contrast, commensal arthropods moving between colonies 3.3. Trade-Offs between Alternative Defences. Defences are not expected to have evolved such avoidance behaviours against Escovopsis include behavioural removal (including towards Escovopsis. self- and allo-grooming), glandular secretions, cultivar defensive compounds against nonnative Escovopsis,andcom- 4.2. Colony-Level Virulence. The within-nest dynamics of pounds with antibiotic properties derived from Actinobacte- Escovopsis infections remain a frontier awaiting exploration. ria. These defences all involve interactions on the ant cuticle Escovopsis can coexist with other nonmutualistic filamentous and are expected to require coordinated interactions to fungi within colonies without colonies displaying signs of avoid negative interference. In Acromyrmex octospinosus, infection [62, 87–89]. However, it is not known if infection the metapleural gland secretions do not appear to harm the sets out shortly after Escovopsis introduction, or if Escovopsis Actinobacteria, so that both defences can be freely expressed spores remain dormant in the colonies until an outbreak [54]. Further, complementarity is expected to maximize cost- of mycelial growth is triggered by external factors. To begin benefit ratios of defences as well as to avoid redundancies. to understand these dynamics, two essential questions need It is conceivable that differences in Actinobacteria cover to be addressed. Firstly, we need a better understanding between closely related ant species, such as A. octospinosus of the level of metabolically active spores and hyphae of and A. echinatior [72], reflect more recent adjustments Escovopsis in normally functioning and apparently healthy (trade-offs) in the relative importance of defences between colonies. This could be obtained through quantitative PCR the species. approaches, so that amounts of Escovopsis biomass and Explorations of defence trade-offshaveonlybeendone levels of metabolic activity, measured as gene expression, in some higher attines, leaving questions of this kind can be estimated. Ideally, this should be explored over time unexplored in most of the fourteen extant fungus-growing to also determine temporal variation. Only when we have ant genera. We propose that utilizing the phylogenetic a better idea of such dynamics, we can begin to explore framework of structural modifications over the course of the the role of the ants in mediating these threats. Secondly, association between fungus-growing ants and Actinobacteria if spore-dormancy is the rule, work should address what [52] would offer a good basis for future work to understand factors trigger within-colony outbreaks. One approach that the dynamics of defence components across the attine could potentially address this is long-term field surveys of tribe. The relative usage of metapleural gland grooming natural colonies to better understand the interplay between and the chemistry of glandular and bacterial secretions ecological fluctuations, (e.g., temperature, rainfall, and food in Acromyrmex/Trachymyrmex versus Atta/Sericomyrmex availability), intrinsic factors (e.g., loss of queen, imbalance exemplify how such comparative approaches can be insight- of worker to garden ratio, and emergence of reproductives), ful [54]. However, considering the vast diversity of cultivar and infection dynamics. usage, Actinobacteria communities, substrate choice, and ant life-history traits, it is conceivable that defence strategies and 5. Tinbergen’s Phylogeny Quadrat trade-offs in unstudied attine ants might be different from those found in the higher attines. 5.1. Origin and Diversification of the Association. The apparent presence of Escovopsis throughout the fungus-growing 4. Tinbergen’s Mechanism Quadrat ants suggests that an ancestral Escovopsis was present as a parasite in the first ant cultivars that were domesticated ca. 4.1. Escovopsis Transmission between Colonies. The success 50 million years ago (cf. [37, 90]). However, an alternative of parasitism is tightly linked to the transmission frequency scenario is that Escovopsis parasitism originated shortly after between host colonies [84]. The most common transmission the early attine ants had become irreversibly committed to for fungal spores is passive dispersal through the air, but this farming. The latter would indicate that Escovopsis parasitism

97 6 Psyche was not merely a passive carry-over process, but that the explainable if their sticky spores would use vectors for long highly peculiar garden phenotype of domesticated fungi distance dispersal that are not available to dispersing ants. It created a novel niche to parasites like Escovopsis. Finding that would be tempting to speculate that other arthropods living Escovopsis parasitism would also occur in free-living relatives in attine nests might have this vector function, but examples of lower attine garden symbionts would make an origin of such long distance flyers vectoring spores are presently predating ant fungiculture more likely, but several lines of lacking. Alternatively, wind dispersal of small leaf fragments indirect evidence suggest that the “new garden niche” model with Escovopsis spores would also seem a realistic mechanism is more likely to apply. First, Actinobacteria cultures on the for parasite populations to become less viscous than host cuticle of attine ants arose also shortly after these ants became populations. Future studies addressing relative dispersal farmers [52], and it would be hard to imagine that the origin efficiencies of partners in the attine ant symbiosis would of this costly biocontrol habit was not somehow related to seem most informative if they could span geographic areas Escovopsis infections. Second, the impact of Escovopsis on that would be large enough to include natural barriers that fungus-growing ant cultivars is likely to be particularly high would differentially affect Escovopsis spores and dispersing because colonies keep a high density of cultivar mycelium ant queens transmitting fungus-garden symbionts. without sufficient own defences. Third, it is striking that the only clade of attine ants that secondary developed a radically 6. Tinbergen’s Adaptation Quadrat different and much less conspicuous garden phenotype, the yeast-rearing Cyphomyrmex, have secondary lost Escovopsis 6.1. Evolutionary Potential of Escovopsis as a Parasite. As as a parasite [4]. already mentioned, Escovopsis has probably persisted as a To date, there are two described species of Escovopsis, parasite of fungus-growing ant gardens since the origin of with E. weberi from a Brazilian Atta species thought to be the ant fungiculture 50 million years ago [4, 37]. If that is so, monotypic species of the genus [91]. Later, a morphological- “Red Queen” like arms races with the ant and fungal hosts ly distinct E. aspergilloides was isolated from Trachymyrmex may have at least periodically occurred, so that genetic ruthae in Trinidad [92]. Both large scale (cf. [37, 38]) diversity of the parasite is likely to be substantial [26–29]. and lower-level lineage diversity [39–41] are considerable, However, the sexual “teleomorph” of Escovopsis has never suggesting that there are more Escovopsis species associated been observed so that Escovopsis may not have sexual re- with fungus-growing ants. Molecular species delineation production, similar to many other Ascomycetes [94]. Lack α based on conserved genes such as EF-1 and 18S rRNA is of sex would not necessarily preclude the integration and unlikely to distinguish lineages that diverged recently, so that exchange of genetic material between different anamorphous more sensitive marker studies are needed. Recent multilocus mycelia within nests, provided that coinfections occur with sequence analyses (MLSAs) have provided the opportunity to some frequency. This is because asexual Ascomycetes can estimate divergence dates for crucial nodes in phylogenetic ff undergo genetic exchange between strains after hyphal trees [4, 19] and would thus also o er novel insights when merging (anastomosis) and parasexual heterokaryosis (the applied to an Escovopsis phylogeny [37]. Approaches of this exchange of cell nuclei) [95]. If such exchanges lead to kind will ultimately allow conclusions about the origin of mitotic crossovers, then there is potential for recombination Escovopsis parasitism (before or after attine ants became between genetically different strains [95]. It will be very farmers) and the rates of Escovopsis evolutionindifferent interesting to investigate whether the Escovopsis genome still host clades. shows signs of such genetic recombination. 5.2. Phylo-Geographic Patterns, Coevolutionary Interactions, The presence of coinfections within individual nests is and Dispersal. Coevolutionary theory predicts that geno- a prerequisite for such genetic exchanges. Both Atta and typic and phenotypic variation across the geographic range Acromyrmex leaf-cutting ants appear to frequently harbour of a host-parasite association can lead to parasite adapta- genetically distinct Escovopsis strains, including ones appear- tions to locally available host genotypes, while becoming ing in two separate phylogenetic clades [63]. Similarly, in the maladapted to nonnative genotypes [93]. A prerequisite for paleoattine genus Apterostigma, fungus gardens are infected such coevolutionary interactions is that host populations are by four distinct Escovopsis morphotypes “brown,” “yellow,” genetically structured, so that gene flow between populations “pink,” and “white” [39]. This implies the potential for ex- remains limited [93]. In fungus-growing ants, only a single change of genetic material between coinfecting strains within study has attempted to explore such coevolutionary dynam- colonies. By explicitly addressing this question, we could gain ics (in the ant species Apterostigma dentigerum [43]). This insight both in the dynamics of coinfections (e.g., facilitation, showed the presence of six distinct host genotype clusters inhibition, the role of the order of infection precedence) across Central America, while structuring was essentially within colonies and in the putative species status of different absent in the parasite, indicating that Escovopsis genotypes Escovopsis morphotypes. are not tightly tracking those of the host [43]. We would expect that other fungus-growing ant- 6.2. Evolutionary Consequences of Host-Parasite Interactions. cultivar-Escovopsis interactions will mirror the findings in A common question in the evolutionary study of host-patho- Apterostigma, since cultivars are vertically transmitted by gen interactions is whether coevolutionary arms races are al- default while Escovopsis is horizontally transmitted. There- most continuous or relatively rare. This is partly because of fore, the population structure in Escovopsis could be the difficulty of testing such dynamics when exploring

98 Psyche 7 biological systems in real time. Fungus-growing ants have 7. Conclusions evolved extensive complementary defences to deal with Esco- vopsis, but the parasite nevertheless prevails at relatively high Since the discovery of Escovopsis parasitism of fungus- population-level frequencies, ranging from 27–75%, de- growing ants less than 15 years ago, we have obtained a pending on the ant genus and geographic location (e.g., [32, broad understanding of prevalence, impact, role, and coevo- 88]). This finding suggests that Escovopsis continues to exert lution of the parasite with the attine ant-fungus symbiosis. selection pressure on the ant hosts, potentially leading to Nevertheless, many fundamental questions remain unan- concomitant changes in ant defences. All this is suggestive of, swered, including the origin of the host-parasite association, but not decisive evidence for, antagonistic coevolution (cf. its presence and potential role outside attine ant nests, [96]). parasite transmission between colonies, and within-colony The efficiency of behavioural defences (grooming/weed- disease dynamics. We know that Escovopsis is attracted to ing) in attine ants is known to have an impact on the specific ant cultivars in some cases, but the generality of virulence of Escovopsis [44, 45]. Under a coevolutionary sce- this phenomenon and the underlying recognition mech- nario, expectations are that Escovopsis has exerted selection anisms are unknown. Several defences against Escovopsis pressures on the ants to optimize their behavioural response are known, including prophylactic behaviours, metapleural towards native parasite strains. Such a scenario would predict gland grooming and compounds, and Actinobacteria sym- that infections with (avirulent) nonnative strains of the bionts, which all contribute to reducing the impact of Escov- parasite would not elicit the same efficient response from the opsis. However, we know little about the context-specific ffi ants. Similarly, if metapleural gland grooming behaviour and e ciency of these alternative and complementary defences, chemistry have been shaped by coevolutionary interactions and only in some cases do we have a crude understanding of ff with Escovopsis, then we would expect that the grooming the potential trade-o s involved. More detailed phylogenetic rate and the chemical secretion cocktail would be adapted to studies of the association specificity of ants, fungal cultivars, inhibit native parasite strains more than nonnative strains. Escovopsis, and Actinobacteria are needed to improve our The coevolutionary patterns arising from interactions interpretations of reciprocal interactions observed. Although between Escovopsis and the Actinobacteria are inevitably the Tinbergen framework did not allow us to do full justice to different from those between Escovopsis and direct defences the complexity of this host-parasite interaction, we feel that by the ants. Two, perhaps nonmutually exclusive, scenarios it does provide a useful structuring device for the research derived from Red-Queen dynamics in relation to Acti- agenda that will be required to make further progress in nobacteria defence have been proposed. The first scenario understanding this unique genus of crop-pests of fungus- suggests that Actinobacteria in the genus Pseudonocardia growing ants. evolve in response to antibiotic resistance evolving in Esco- vopsis. Evidence supporting the potential for this to be the Acknowledgments case comes mainly from observations of variation in the propensities of different Pseudonocardia-derived antibiotics, The authors thank Panagiotis Sapountzis, Jelle van Zweden, including the presence of Escovopsis strains that are resistant Eric Caldera, and Jeremy Thomas Poulsen for comments [55, 64]. Phenotypic variation is a prerequisite for such on an early draft of this manuscript, and three anonymous dynamics to be maintained, as this is what natural selection reviewers for valuable comments and suggestions. S. H. Yek can act on. However, no studies have as yet shown that and J. J. Boomsma were supported by the Danish National changes in Pseudonocardia genes for antibiotic production Research Foundation, and M. Poulsen by The Danish do indeed change in response to Escovopsis susceptibility. Council for Independent Research, Natural Sciences (FNU). A second possible scenario is that attine ants frequently acquire strains of bacteria from the environment that have novel antibiotic properties against Escovopsis, be it either References Pseudonocardia [55] or other Actinobacteria [57, 61, 65]. Evidence for such acquisitions comes from survey data [1]F.W.Møller,Die Pilzgarten¨ Einiger Sudamerikanischer¨ Ameis- showing that free-living Pseudonocardia are phylogeneti- en, Gustav Fischer, Jena, Germany, 1893. cally interspersed with ant-associated clades [55], and that [2] N. A. Weber, Gardening Ants: The Attines, American Philo- additional Actinobacteria with antibiotic properties (mainly sophical Society, Philadelphia, Pa, USA, 1972. Streptomyces) can be obtained from the ant cuticles or [3] U.G.Mueller,T.R.Schultz,C.R.Currie,R.M.M.Adams,and gardens of colonies. We expect that characterizations of the D. Malloch, “The origin of the attine ant-fungus mutualism,” The Quarterly Review of Biology, vol. 76, no. 2, pp. 169–197, antibiotic profiles produced by the major clades of Pseudono- 2001. cardia that associate with fungus-growing ants will clarify [4] T. R. Schultz and S. G. Brady, “Major evolutionary transitions the role that these alternative acquisition mechanisms have in ant agriculture,” Proceedings of the National Academy of Sci- played in maintaining a successful Pseudonocardia-defence ences of the United States of America, vol. 105, no. 14, pp. 5435– against Escovopsis. Further studies will also benefit from a 5440, 2008. more explicit emphasis on exploring how and to what extent [5] T. R. Schultz and R. Meier, “A phylogenetic analysis of the such horizontal acquisitions of novel Actinobacteria occur, fungus-growing ants (Hymenoptera: Formicidae: Attini) and whether they have a selective advantage for ant colony based on morphological characters of the larvae,” Systematic fitness. Entomology, vol. 20, no. 4, pp. 337–370, 1995.

99 8 Psyche

[6]J.K.Wetterer,T.R.Schultz,andR.Meier,“Phylogenyof Mutualism in Microbial Symbiosis,J.F.WhiteandM.S.Torres, fungus-growing ants (tribe attini) based on mtDNA sequence Eds., CRC Press, Boca Raton, Fla, USA, 2009. and morphology,” Molecular Phylogenetics and Evolution, vol. [23] S. A. Frank, “Models of parasite virulence,” The Quarterly Re- 9, no. 1, pp. 42–47, 1998. view of Biology, vol. 71, no. 1, pp. 37–78, 1996. [7]C.KlingenbergandC.R.F.Brandao,˜ “Revision of the fungus- [24] M. Poulsen and J. J. Boomsma, “Mutualistic fungi control crop growing ant genera Mycetophylax Emery and Paramycetophy- diversity in fungus-growing ants,” Science, vol. 307, no. 5710, lax Kusnezov rev. stat., and description of Kalathomyrmex n. pp. 741–744, 2005. gen. (Formicidae: Myrmicinae: Attini),” Zootaxa, no. 2052, pp. [25] U. G. Mueller, J. J. Scott, H. D. Ishak, M. Cooper, and A. 1–31, 2009. Rodrigues, “Monoculture of leafcutter ant gardens,” PLoS One, [8] T. Eisner and G. M. Happ, “The infrabuccal pocket of a formi- vol. 5, no. 9, Article ID e12668, 2010. cine ant: a social filtration device,” Psyche, vol. 69, pp. 107–116, [26]K.ClayandP.X.Kover,“TheRedQueenHypothesisand 1962. plant/pathogen interactions,” Annual Review of Phytopathol- [9] B. Holldobler¨ and E. O. Wilson, The Ants, Springer Verlag, ogy, vol. 34, pp. 29–50, 1996. Berlin, Germany, 1990. [27] W. D. Hamilton, “Sex versus non-sex versus parasite,” Oikos, [10] H. von Ihering, “Die anlage neuer colonien und pilzgarten¨ vol. 35, no. 2, pp. 282–290, 1980. bei,” Atta Sexdens. Zoologischer Anzeiger, vol. 21, pp. 238–245, [28]L.T.Morran,O.G.Schmidt,I.A.Gelarden,R.C.ParrishII, 1898. and C. M. Lively, “Running with the Red Queen: host-parasite [11] M. Autuori, “La fondation des societ´ es´ chez le fourmis cham- coevolution selects for biparental sex,” Science, vol. 333, no. pignnonnistes du genre “Atta” (Hym. Formicidae),” in L’In- 6039, pp. 216–218, 2011. stinct dans le Comportement des Animaux et de L’Homme,M. [29] R. J. Ladle, “Parasites and sex: catching the red queen,” Trends Autuori et al., Ed., Masson & Cie, Paris, France, 1956. in Ecology and Evolution, vol. 7, no. 12, pp. 405–408, 1992. [12] U. G. Mueller, S. A. Rehner, and T. R. Schultz, “The evolution [30] J. Jokela, M. F. Dybdahl, and C. M. Lively, “The maintenance of agriculture in ants,” Science, vol. 281, no. 5385, pp. 2034– of sex, clonal dynamics, and host-parasite coevolution in a 2038, 1998. mixed population of sexual and asexual snails,” The American [13] N. A. Weber, “Fungus-growing ants,” Science, vol. 153, no. Naturalist, vol. 174, no. 1, pp. S43–S53, 2009. 3736, pp. 587–604, 1966. [31] G. Stahel and D. C. Geijskes, “Weitere untersuchungen uber¨ [14] I. H. Chapela, S. A. Rehner, T. R. Schultz, and U. G. Mueller, nestbau und gartenpilz von Atta cephalotes L. und Atta sexdens “Evolutionary history of the symbiosis between fungus- L. (Hym. Formicidae),” Revista Entomology, vol. 12, pp. 243– growing ants and their fungi,” Science, vol. 266, no. 5191, pp. 268, 1940. 1691–1694, 1994. [32] C. R. Currie, U. G. Mueller, and D. Malloch, “The agricultural [15] G. Hinkle, J. K. Wetterer, T. R. Schultz, and M. L. Sogin, “Phy- pathology of ant fungus gardens,” Proceedings of the National logeny of the attine ant fungi based on analysis of small Academy of Sciences of the United States of America, vol. 96, no. subunit ribosomal RNA gene sequences,” Science, vol. 266, no. 14, pp. 7998–8002, 1999. 5191, pp. 1695–1697, 1994. [33] R. Koch, “Ueber¨ den augenblicklichen stand der bakteriolo- [16] A. S. Mikheyev, U. G. Mueller, and J. J. Boomsma, “Population gischen choleradiagnose,” Zeitschrift fur¨ Hygiene und Infek- genetic signatures of diffuse co-evolution between leaf-cutting tionskrankheiten, vol. 14, no. 1, pp. 319–338, 1893. ants and their cultivar fungi,” Molecular Ecology, vol. 16, no. 1, [34] C. R. Currie, The ecology and evolution of a quadripartite sym- pp. 209–216, 2007. biosis: examining the interactions among attine ants, fungi, and [17] A. S. Mikheyev, T. Vo, and U. G. Mueller, “Phylogeography of actinomycetes, Ph.D. thesis, Department of Botany, University post-pleistocene population expansion in a fungus-gardening of Toronto, 2000. ant and its microbial mutualists,” Molecular Ecology, vol. 17, [35] C. R. Currie, “Prevalence and impact of a virulent parasite on no. 20, pp. 4480–4488, 2008. a tripartite mutualism,” Oecologia, vol. 128, no. 1, pp. 99–106, [18] A. M. Green, U. G. Mueller, and R. M. M. Adams, “Extensive 2001. exchange of fungal cultivars between sympatric species of [36] H. T. Reynolds and C. R. Currie, “Pathogenicity of Escovopsis fungus-growing ants,” Molecular Ecology,vol.11,no.2,pp. weberi: the parasite of the attine ant-microbe symbiosis 191–195, 2002. directly consumes the ant-cultivated fungus,” Mycologia, vol. [19] A. S. Mikheyev, U. G. Mueller, and P. Abbot, “Comparative 96, no. 5, pp. 955–959, 2004. dating of attine ant and lepiotaceous cultivar phylogenies [37] C. R. Currie, B. Wong, A. E. Stuart et al., “Ancient tripartite co- reveals coevolutionary synchrony and discord,” The American evolution in the attine ant-microbe symbiosis,” Science, vol. Naturalist, vol. 175, no. 6, pp. E126–E133, 2010. 299, no. 5605, pp. 386–388, 2003. [20] R. Sen, H. D. Ishak, T. R. Kniffin,andU.G.Mueller,“Con- [38] S. J. Taerum, M. J. Cafaro, A. E. F. Little, T. R. Schultz, and C. struction of chimaeric gardens through fungal intercropping: R. Currie, “Low host-pathogen specificity in the leaf-cutting a symbiont choice experiment in the leafcutter ant Atta texana ant-microbe symbiosis,” Proceedings of the Royal Society B, vol. (Attini, Formicidae),” Behavioral Ecology and Sociobiology, vol. 274, no. 1621, pp. 1971–1978, 2007. 64, no. 7, pp. 1125–1133, 2010. [39] N. M. Gerardo, U. G. Mueller, and C. R. Currie, “Complex [21] R. M. M. Adams, U. G. Mueller, A. K. Holloway, A. M. Green, host-pathogen coevolution in the Apterostigma fungus-grow- and J. Narozniak, “Garden sharing and garden stealing in ing ant-microbe symbiosis,” BMC Evolutionary Biology, vol. 6, fungus-growing ants,” Naturwissenschaften, vol. 87, no. 11, pp. article 88, 2006. 491–493, 2000. [40] N. M. Gerardo, S. R. Jacobs, C. R. Currie, and U. G. Mueller, [22] M. Poulsen, A. E. F. Little, and C. R. Currie, “Fungus-growing “Ancient host-pathogen associations maintained by specificity ant-microbe symbioses: using microbes to defend beneficial of chemotaxis and antibiosis,” PLoS Biology,vol.4,no.8,pp. mutualisms within symbiotic communities,” in Defensive 1358–1363, 2006.

100 Psyche 9

[41] N. M. Gerardo, U. G. Mueller, S. L. Price, and C. R. Currie, [57] U. G. Mueller, D. Dash, C. Rabeling, and A. Rodrigues, “Co- “Exploiting a mutualism: parasite specialization on cultivars evolution between attine ants and actinomycete bacteria: a within the fungus-growing ant symbiosis,” Proceedings of the reevaluation,” Evolution, vol. 62, no. 11, pp. 2894–2912, 2008. Royal Society B, vol. 271, no. 1550, pp. 1791–1798, 2004. [58] S. Haeder, R. Wirth, H. Herz, and D. Spiteller, “Candicidin- [42] A. S. Mikheyev, U. G. Mueller, and P. Abbot, “Cryptic sex producing Streptomyces support leaf-cutting ants to protect and many-to-one coevolution in the fungus-growing ant their fungus garden against the pathogenic fungus Escovopsis,” symbiosis,” Proceedings of the National Academy of Sciences of Proceedings of the National Academy of Sciences of the United the United States of America, vol. 103, no. 28, pp. 10702–10706, States of America, vol. 106, no. 12, pp. 4742–4746, 2009. 2006. [59] C. Kost, T. Lakatos, I. Bottcher, W. R. Arendholz, M. [43] N. M. Gerardo and E. J. Caldera, “Labile associations between Redenbach, and R. Wirth, “Non-specific association between fungus-growing ant cultivars and their garden pathogens,” The filamentous bacteria and fungus-growing ants,” Naturwis- ISME Journal, vol. 1, no. 5, pp. 373–384, 2007. senschaften, vol. 94, no. 10, pp. 821–828, 2007. [44] C. R. Currie and A. E. Stuart, “Weeding and grooming of [60] J. Barke, R. F. Seipke, S. Gruschow¨ et al., “A mixed community pathogens in agriculture by ants,” Proceedings of the Royal of actinomycetes produce multiple antibiotics for the fungus Society B, vol. 268, no. 1471, pp. 1033–1039, 2001. farming ant Acromyrmex octospinosus,” BMC Biology, vol. 8, [45]D.Abramowski,C.R.Currie,andM.Poulsen,“Castespecial- no. 1, article 109, 2010. ization in behavioral defenses against fungus garden parasites [61] I. Schoenian, M. Spiteller, M. Ghaste, R. Wirth, H. Herz, and in Acromyrmex octospinosus leaf-cutting ants,” Insectes Sociaux, D. Spiteller, “Chemical basis of the synergism and antagonism vol. 58, no. 1, pp. 65–75, 2010. in microbial communities in the nests of leaf-cutting ants,” [46] A. Gerstner, M. Poulsen, and C. R. Currie, “Recruitment of Proceedings of the National Academy of Sciences of the United minor workers for defense against a specialized parasite of States of America, vol. 108, no. 5, pp. 1955–1960, 2011. Atta leaf-cutting ant fungus gardens,” Ethology, Ecology and [62] A. Rodrigues, F. C. Pagnocca, M. Bacci Jr., M. J. A. Hebling, O. Evolution, vol. 23, no. 1, pp. 61–75, 2011. C. Bueno, and L. H. Pfenning, “Variability of non-mutualistic filamentous fungi associated with Atta sexdens rubropilosa [47] H. Fernandez-Mar´ ´ın, J. K. Zimmerman, S. A. Rehner, and W. nests,” Folia Microbiologica, vol. 50, no. 5, pp. 421–425, 2005. T. Wcislo, “Active use of the metapleural glands by ants in [63] S. J. Taerum, M. Cafaro, and C. R. Currie, “Presence of multi- controlling fungal infection,” Proceedings of the Royal Society parasite infections within individual colonies of leaf-cutter B, vol. 273, no. 1594, pp. 1689–1695, 2006. ants,” Environmental Entomology, vol. 39, no. 1, pp. 105–113, [48] A. N. M. Bot, D. Ortius-Lechner, K. Finster, R. Maile, and J. J. 2010. Boomsma, “Variable sensitivity of fungi and bacteria to com- [64] M. Poulsen, D. Erhardt, A. E. F. Little et al., “Variation in pounds produced by the metapleural glands of leaf-cutting Pseudonocardia antibiotic defence helps govern parasite-in- ants,” Insectes Sociaux, vol. 49, no. 4, pp. 363–370, 2002. duced morbidity in Acromyrmex leaf-cutting ants,” Environ- [49] A. E. F. Little, T. Murakami, U. G. Mueller, and C. R. Currie, mental Microbiology Reports, vol. 2, no. 4, pp. 534–540, 2010. “Defending against parasites: fungus-growing ants combine [65] R. Sen, H. D. Ishak, E. Dora, S. E. Dowd, E. Hong, and U. G. specialized behaviours and microbial symbionts to protect Mueller, “Generalized antifungal activity and 454-screening of their fungus gardens,” Biology Letters, vol. 2, no. 1, pp. 12–16, Pseudonocardia and Amycolatopsis bacteria in nests of fungus- 2006. growing ants,” Proceedings of the National Academy of Sciences [50] C. R. Currie, J. A. Scott, R. C. Summerbell, and D. Malloch, of the United States of America, vol. 106, no. 42, pp. 17805– “Erratum: fungus-growing ants use antibiotic-producing bac- 17810, 2009. teria to control garden parasites,” Nature, vol. 398, pp. 701– [66] R. M. Nesse, “Tinbergen’s Four Questions Organized,” 2000, 704, 1999. http://nesse.us. [51] C. R. Currie, A. N. M. Bot, and J. J. Boomsma, “Experimental [67] N. Tinbergen, “On aims and methods of ethology,” Zeitschrift evidence of a tripartite mutualism: bacteria protect ant fungus fur¨ Tierpsychologie, vol. 20, no. 4, pp. 410–433, 1963. gardens from specialized parasites,” Oikos, vol. 101, no. 1, pp. [68] E. Mayr, “Adaptation and selection,” Italian Journal of Zoology, 91–102, 2003. vol. 48, pp. 66–77, 1981. [52] C. R. Currie, M. Poulsen, J. Mendenhall, J. J. Boomsma, and [69] D. Ortius-Lechner, R. Maile, E. D. Morgan, and J. J. Boomsma, J. Billen, “Coevolved crypts and exocrine glands support “Metapleural gland secretion of the leaf-cutter ant Acro- mutualistic bacteria in fungus-growing ants,” Science, vol. 311, myrmex octospinosus: new compounds and their functional no. 5757, pp. 81–83, 2006. significance,” Journal of Chemical Ecology,vol.26,no.7,pp. [53] S. H. Yek and U. G. Mueller, “The metapleural gland of ants,” 1667–1683, 2000. Biological Reviews, vol. 91, no. 3-4, pp. 201–224, 2010. [70] A. E. F. Little, T. Murakami, U. G. Mueller, and C. R. Currie, [54] H. Fernandez-Mar´ ´ın, J. K. Zimmerman, D. R. Nash, J. J. “The infrabuccal pellet piles of fungus-growing ants,” Ethol- Boomsma, and W. T. Wcislo, “Reduced biological control ogy, Ecology and Evolution, vol. 90, no. 12, pp. 558–562, 2003. and enhanced chemical pest management in the evolution of [71]D.C.Oh,M.Poulsen,C.R.Currie,andJ.Clardy,“Denti- fungus farming in ants,” Proceedings of the Royal Society B, vol. gerumycin: a bacterial mediator of an ant-fungus symbiosis,” 276, no. 1665, pp. 2263–2269, 2009. Nature Chemical Biology, vol. 5, no. 6, pp. 391–393, 2009. [55] M. J. Cafaro, M. Poulsen, A. E. F. Little et al., “Specificity in [72] M. Poulsen, A. N. M. Bot, M. G. Nielsen, and J. J. Boomsma, the symbiotic association between fungus-growing ants and “Experimental evidence for the costs and hygienic significance protective Pseudonocardia bacteria,” Proceedings of the Royal of the antibiotic metapleural gland secretion in leaf-cutting Society B, vol. 278, no. 1713, pp. 1814–1822, 2011. ants,” Behavioral Ecology and Sociobiology, vol. 52, no. 2, pp. [56] M. Poulsen, M. Cafaro, J. J. Boomsma, and C. R. Currie, “Spec- 151–157, 2002. ificity of the mutualistic association between actinomycete [73] M. Poulsen, A. N. M. Bot, C. R. Currie, M. G. Nielsen, and J. bacteria and two sympatric species of Acromyrmex leaf-cutting J. Boomsma, “Within-colony transmission and the cost of ants,” Molecular Ecology, vol. 14, no. 11, pp. 3597–3604, 2005. a mutualistic bacterium in the leaf-cutting ant Acromyrmex

101 10 Psyche

octospinosus,” Functional Ecology, vol. 17, no. 2, pp. 260–269, [92] K. A. Seifert, R. A. Samson, and I. H. Chapela, “Escovopsis 2003. aspergilloides, a rediscovered Hyphomycete from leaf-cutting [74] D. A. Dewsbury, “The 1973 Nobel prize for physiology or ant nests,” Mycologia, vol. 87, no. 3, pp. 407–413, 1995. medicine: recognition for behavioral science?” American Psy- [93] J. N. Thompson, The Geographic Mosaic of Coevolution,The chologist, vol. 58, no. 9, pp. 747–752, 2003. University of Chicago Press, Chicago, Ill, USA, 2005. [75] J. R. Krebs and N. B. Davies, An Introduction to Behavioural [94] D. H. Jennings and G. Lysek, Fungal Biology: Understanding Ecology, Blackwell Publisher, 1993. the Fungal Lifestyle, Bios Scientific Publishers, Guildford, UK, [76] E. L. Charnov and S. W. Skinner, “Complementary approaches 1996. to the understanding of parasitoid oviposition decisions,” [95] R. D. Tinline and B. H. MacNeill, “Parasexuality in plant path- Environmental Entomology, vol. 14, no. 4, pp. 383–391, 1985. ogenic fungi,” Annual Review of Phytopathology, vol. 7, pp. [77] D. M. Zeifman, “An ethological analysis of human infant 147–168, 1969. crying: answering Tinbergen’s four questions,” Developmental [96] A. Buckling and P. B. Rainey, “Antagonistic coevolution be- Psychobiology, vol. 39, no. 4, pp. 265–285, 2001. tween a bacterium and a bacteriophage,” Proceedings of the [78] R. Buchholz, “Behavioural biology: an effective and relevant Royal Society B, vol. 269, no. 1494, pp. 931–936, 2002. conservation tool,” Trends in Ecology and Evolution, vol. 22, no. 8, pp. 401–407, 2007. [79] J. N. Thompson, “Specific hypotheses on the geographic mosaic of coevolution,” The American Naturalist, vol. 153, no. S5, pp. S1–S14, 1999. [80] M. E. Hudson, “Sequencing breakthroughs for genomic ecology and evolutionary biology,” Molecular Ecology Resources, vol. 8, no. 1, pp. 3–17, 2008. [81] A. Rokas and P.Abbot, “Harnessing genomics for evolutionary insights,” Trends in Ecology and Evolution,vol.24,no.4,pp. 192–200, 2009. [82] S. Nygaard, G. J. Zhang, M. Schiøtt et al., “The genome of the leaf-cutting ant Acromyrmex echinatior suggests key adaptations to advanced social life and fungus farming,” Genome Research, vol. 21, pp. 1339–1348, 2011. [83] G. Suen, C. Teiling, L. Li et al., “The genome sequence of the leaf-cutter ant Atta cephalotes reveals insights into its obligate symbiotic lifestyle,” PLoS Genetics, vol. 7, no. 2, Article ID e1002007, 2011. [84] R. M. Anderseon and R. M. May, “The population dynamics of microparasites and their invertebrate hosts,” Philosophical Transactions of the Royal Society of London B, vol. 291, no. 1054, pp. 451–524, 1981. [85] C. R. Currie, “A community of ants, fungi, and bacteria: a multilateral approach to studying symbiosis,” Annual Review of Microbiology, vol. 55, pp. 357–380, 2001. [86] H. Fernandez-Mar´ ´ın, J. K. Zimmerman, and W. T. Wcislo, “Nest-founding in Acromyrmex octospinosus (Hymenoptera, Formicidae, Attini): demography and putative prophylactic behaviors,” Insectes Sociaux, vol. 50, no. 4, pp. 304–308, 2003. [87] F. J. Richard and C. Errard, “Hygienic behavior, liquid- foraging, and trophallaxis in the leafcutting ants, Acromyrmex subterraneus and Acromyrmex octospinosus,” Journal of Insect Science, vol. 9, article 63, 2009. [88] A. Rodrigues, M. Bacci Jr., U. G. Mueller, A. Ortiz, and F. C. Pagnocca, “Microfungal “weeds” in the leafcutter ant symbiosis,” Microbial Ecology, vol. 56, no. 4, pp. 604–614, 2008. [89] A. Rodrigues, U. G. Mueller, H. D. Ishak, M. Bacci Jr., and F. C. Pagnocca, “Ecology of microfungal communities in gardens of fungus-growing ants (Hymenoptera: Formicidae): a year-long survey of three species of attine ants in Central Texas,” FEMS Microbiology Ecology, vol. 78, no. 2, pp. 244–255, 2011. [90] E. Caldera, M. Poulsen, G. Suen, and C. R. Currie, “Insect symbioses—a case study of past, present, and future fungus- growing ant research,” Environmental Entomology, vol. 38, no. 1, pp. 78–92, 2009. [91] J. J. Muchovej and T. M. Della Lucia, “Escovopsis, a new genus from leaf cutting ant nests to replace Phialocladus nomem invalidum,” Mycotaxon, vol. 37, pp. 191–195, 1990.

102 CHAPTER 4

Interaction specificity between leaf-cutting ant workers and their vertically transmitted cuticular Pseudonocardia bacteria before and after challenges with Escovopsis fungus garden parasites

Photo credit: Janni Larsen

Manuscript in preparation

103 Interaction specificity between leaf-cutting ant workers and their vertically transmitted cuticular Pseudonocardia bacteria before and after challenges with Escovopsis fungus garden parasites

Sandra B. Andersen, Sze Huei Yek, David R. Nash, Jacobus J. Boomsma

Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark.

Corresponding authors: Sandra B. Andersen ([email protected]) and Sze Huei Yek ([email protected])

Keywords: attine ant mutualism, cross-fostering, prophylactic defences, host-symbiont coevolution

104

Abstract The complex symbiosis of the fungus growing attine ants presents an ideal system to address questions of symbiont specificity and stability. The ants rear a monoculture of fungal cultivar and have evolved an additional mutualistic relationship with actinomycete bacteria to defend their crop against the fungal parasite Escovopsis. The leaf-cutting ant Acromyrmex echinatior has been found to associate primarily with Pseudonocardia bacteria, with ant colonies apparently maintaining the same bacterial strain over a lifetime, yet it has been suggested that the ants may readily associate with other bacteria than their native symbiont. To understand the dynamics of this symbiosis, we used a cross-fostering approach, with colonies harbouring two different Pseudonocardia strains to examine colony versus strain specificity in ants/symbiont associations by measuring bacterial growth and efficiency in disease defence against Escovopsis. We found that Pseudonocardia readily colonize ants irrespective of the host/symbiont match and that growth is primarily affected by the host genotype. The efficiency of Escovopsis removal was however negatively affected by the presence of a non-native strain of Pseudonocardia and ants appeared to take longer time to develop into foragers. In particular, one of the strains appeared superior in Escovopsis defence but only when associated with the native host. In contrast, the other strain reached a higher final cover. These results suggest the presence of complex trade-offs in the host-symbiont relationship and that coevolution between ants, bacteria and Escovopsis plays a role in maintaining specificity.

105 Introduction Symbiotic interactions between species can range from mutualistic to parasitic, depending on the levels of cooperation and conflict expressed between the interactors (Queller & Strassmann 2009). A range of traits, such as mode of transmission, genetic diversity, and the probability of repeated interactions, have been suggested to shape the outcome of symbioses (Herre et al. 1999). The complex symbiosis of the fungus-growing ants, centered on the interaction between attine ant colonies and monocultures of fungal cultivar has become a model for testing the mechanisms of interaction stability (Mueller 2002). The genetic homogeneity of the symbiont is maintained by recognition mechanisms of the cultivar (Poulsen & Boomsma 2005) and the ant hosts (Ivens et al. 2009), and by vertical transmission (Wheeler 1910), which together secured substantial adaptive radiation and ecological expansion for more than 50 MY (Schultz & Brady 2008). This evolutionary success also resulted in a range of other lineages of organisms joining the symbiosis, covering the entire spectrum from being additional mutualists to parasites (Currie 2001; Little & Currie 2007). As is the case with human crop- monocultures, the fungus garden is prone to attack by pests, with species of the hypocrealean genus Escovopsis specializing on the attine fungal cultivar (Reynolds & Currie 2004; Rodrigues et al. 2008). To protect their food source the majority of attine ants grow antibiotic-producing actinomycete bacteria on their cuticle (Currie et al. 2006). The specificity of the association of the bacteria and ants, in addition to the specificity of the antibiotics against Escovopsis, is however the subject of an ongoing debate (see Barke et al. 2011; Boomsma & Aanen 2009; Cafaro et al. 2011; Mueller et al. 2008; Mueller 2012; Sen et al. 2009). The Acromyrmex leaf-cutting ants are one of the two most advanced genera of fungus-growing ants. Together with the genus Atta, they are characterized by the ability to harvest fresh plant material as fungal substrate and to realize high growth rates of their garden symbiont belonging to the single species Leucocoprinus gongylophorus (Mikheyev et al. 2010). In contrast to Atta, however, they have maintained the ancestral attine ant habit of growing filamentous actinomycete bacteria on their cuticle, which have been shown to help control Escovopsis in an experimental study (Currie et al. 2003). In the best studied species, A. echinatior and A. octospinosus, Pseudonocardia bacteria

106 dominate cuticular communities (Andersen et al. in review; Poulsen et al. 2005) and the build-up of the bacterial cover of large workers follows a characteristic pattern starting with inoculation shortly after hatching (scale 1; Poulsen et al. 2003a), culminating with almost full cover (scale 12) after 2-3 weeks, and decline to ca. scale 3 cover by the time workers start their foraging careers about 7 weeks after hatching. The Pseudonocardia strains isolated from Panamanian Acromyrmex appear to be vertically transmitted across generations of colonies and among age cohorts of workers within colonies, a process that may be facilitated by bacterial fragments in the fungus garden (Andersen et al. in review; Currie et al. 1999; Poulsen et al. 2003a). After ca. a month the bacterial cover is concentrated on the laterocervical plates of the propleura. This implies that older large workers rarely lose their actinomycete symbionts, but that prevalence on their cuticles has become low by the time they start being foragers (up till scale 3; see above). How the bacterial dynamics are controlled is not yet well understood but the initial growth to obtain a full cover has been suggested to be supported by available nutrients on the ant cuticle and subsequently by metapleural gland secretions (Poulsen et al. 2003b) whereas the laterocervical plate cover on mature workers is apparently supported by secretions of other cuticular glands (Currie et al. 2006). Acromyrmex ants have been observed to use bacteria-derived antibiotic secretions for Escovopsis defence (Currie & Scott 1999) and to induce at least some renewed bacterial growth when their colonies are faced with an experimental Escovopsis challenge (Currie et al. 2003). However, the mechanisms that secure the acquisition of specific cuticular actinomycetes remain poorly understood. Also these aspects have been the subject of debate, as other actinomycetes than Pseudonocardia have been isolated, both from Acromyrmex and from other fungus-growing ants, and the antibiotics produced by these bacteria are not always effective against Escovopsis and may suppress other potential pathogens (Barke et al. 2010; Seipke et al. 2011; Sen et al. 2009). A recent study has also shown that the actinomycete cover can serve as a general barrier against infections of callow (newly eclosed) workers with entomopathogens (Mattoso et al. 2011). The A. echinatior population of Gamboa, Panama is well studied, with two strains of Pseudonocardia known to be present on ant cuticles and colonies maintaining their original strain for more than 10 years in the laboratory (Andersen et al. in review;

107 Poulsen et al. 2005). When ant pupae are cross-fostered in a different colony than their own, they readily obtain a bacterial cover following eclosion, but a single large scale experiment (in terms of number of ants involved) has indicated that the bacteria apparently grew less well on cross-fostered ants and reached a less abundant cover on mature ants (Armitage et al. 2011). These findings suggested that there might be some degree of specificity in the association between single host colonies and strains of the actinomycete symbiont. If generally true, such specificity would be consistent with some form of coevolution between Acromyrmex ant-hosts and actinomycete symbionts. The aim of the present study was to address these issues in a larger experimental set-up with a fully balanced pupal cross-fostering approach in which strain identities were known, bacterial growth rates were measured more systematically, and behavioural reactions after a controlled challenge with Escovopsis were monitored. Using four A. echinatior source colonies with two previously identified Pseudonocardia strains (Andersen et al. in review), we created a full factorial design of subcolonies where pupae of large workers were raised in an environment with either their own Pseudonocardia strain, the same strain but from another source colony or another strain from two other source colonies. The growth-rate part of the experiments allowed us to shed light on the extent to which specificity factors might constrain the efficiency of bacterial inoculations that are not strictly vertical, whereas the Escovopsis challenge part of the experiment would offer insight into the extent to which horizontally transmitted strains might compromise disease defense efficiency and thus potentially affect ant fitness. If there would be specificity in average bacterial growth rate per colony or prophylactic behaviour after infection, we would expect the best performance from ants eclosing in their own colony, followed by ants eclosing in colonies with the same actinomycete strain (horizontal inoculation with an almost identical strain). Ants inoculated with an alien symbiont (horizontal inoculation with a genetically different strain that lives in symbiosis with other colonies in the population, but is almost always mutually exclusive in single colonies) would then be expected to do worst. We found evidence that colonies varied in cuticular bacterial growth rates irrespective of host/symbiont match but that the final cover depended both on the symbiont strain and an interaction with the rearing environment. Ants with novel acquisitions of bacterial strains

108 were also less effective in handling Escovopsis infections, independent of the variation between colonies, showing that the host/symbiont match is important for the cost/benefit ratio of this mutualism.

Materials and methods Cross-fostering A full factorial cross-fostering experiment was set up with four source colonies of Acromyrmex echinatior collected in Gamboa, Panama. Colonies Ae.150 and Ae.488 harboured the same dominant Pseudonocardia strain ‘Ps1’, where colonies Ae.160 and Ae.331 harboured a different dominant Pseudonocardia strain ‘Ps2’ (Andersen et al. in review; Poulsen et al. 2005). Subcolonies were created with 2 g of fungus garden (small brood items were not removed), 25 small workers, 15 medium workers and 4 mature pupae of large workers. The fungus garden was placed in a small petri dish (∅ 5.5 cm) covered by a lid with a hole at the side, allowing ant movement into a larger petri dish (∅ 9 cm) designated as foraging area where moist cotton and bramble leaves were supplied regularly. In this experimental set-up, large worker pupae eclosed in subcolonies containing fungus and workers from: 1. Their own colony; 2. A colony with the same Pseudonocardia strain, and 3. One of two colonies harbouring the other Pseudonocardia strain. Each combination was replicated twice resulting in a total of 16 different combinations and 32 subcolonies (Fig. 1). The colonies were checked daily and when a large worker eclosed from the pupal stage it was marked with nail-polish between the pronotal spines and the date noted following Poulsen et al. (2003a). We used four different colours in order to individually identify each ant. The date of eclosure was noted as ‘Day 1’ and the ventral and dorsal side of the ants were photographed on day 4 (if bacterial growth was visible), 8, 12, 16 and for some individuals also on day 20. The corresponding ventral and dorsal photographs were printed, the ant identity and date noted, and the bacterial cover scored blindly according to the scale (1-12) of Poulsen et al. (2003a). Some subcolonies were infected with Escovopsis (see below) before all individuals were 16 days old, because the four pupae eclosed up to 8 days apart.

109 Prophylactic behaviour following Escovopsis infection When either all individuals were fully covered in bacteria (scale 12) or all ants were at least 16 days old, the fungus gardens of subcolonies were infected with Escovopsis weberi. This difference in procedure was unavoidable due to differences between pupae within subcolonies in the timing of bacterial growth, and limitations in the number of sub-colonies that could be handled by a single observer in a day. To account for a potential effect of the bacterial cover on the control of Escovopsis infection, the cover of each ant was scored as described above on the day of infection. For each infection, Escovopsis conidia were taken from 1 cm2 of pure E. weberi culture, grown on potato dextrose agar (PDA) medium, which corresponds to 1.3-2.5 x 107 dry conidia. The E. weberi strain that we used was isolated from Acromyrmex nests in Gamboa, Panama by H. Fernández-Marín in early 2010, and identified based on morphological characters (Muchovej & Della Lucia 1990). The conidia were collected from the culture media with a fine paintbrush and transferred to the fungal cultivar by brushing evenly over the fungal surface. The viability of E. weberi conidia was checked by measuring the germination of conidia on plates, and was found to be > 90% at the time of infections. During the first hour after infection, behavioural repertoires expressed by the four marked large workers were observed for 5 min in a 3.5 cm diameter field of view of a stereomicroscope (6.3× magnification) with 5 min intervals between each observation, resulting in 6 time points for the first hour after infection. We focused on behaviours associated with garden tending, but noted down nine behaviours in total. Behaviours were recorded in 30s time blocks, and if a marked individual was observed to continuously perform the same behaviour within any 30s block, that behaviour was recorded as being performed once, whereas it was recorded twice if performed in two 30s blocks etc. Activities that were performed for less than 30 s were not recorded. This gave us the total duration of continuous behaviours, while ignoring transient behaviours, and meant that the total duration of behaviours differed between subcolonies. If marked individuals were outside the fungus garden (and hence out of the field of view) for the whole of any 30s block, we recorded this as “outside fungus garden” without recording what these workers were actually doing. Behavioural repertoires were recorded again 48 and 72 hrs after

110 infection, using the same time interval (5 min observation, 5 min break), resulting in 18 time points of 5 min for each subcolony during the three days of observations. The nine different behaviours performed after the Escovopsis infection of the fungus garden were: self-grooming, garden-grooming, garden-weeding, immobile inside fungus garden, allo-grooming, carrying eggs/larvae/pupae, tending eggs/larvae/pupae, trophallaxis, and garden manuring. We defined grooming behaviours as scraping antennae and legs with forelegs, licking forelegs with mouth parts (Bot et al. 2001), followed by application of forelegs to different locations. If the target of grooming was a worker’s own body, we categorized it as ‘self-grooming’, if the target was a nestmate, we categorized it as ‘allo-grooming’, and if the target was eggs, larvae or pupae, we categorized it as ‘tending eggs/larvae/pupae’. ‘Garden-grooming’ was scored when the ants antennated the fungus garden, extended their maxillae and labium to grasp a piece of the garden matrix, and closed them to retract and raise the fungal fragment off the garden matrix while pulling it through their mouth parts (Currie & Stuart 2001). We categorized ‘garden weeding’ as actions involving the physical removal of garden pieces, whereby the ants detached a piece of fungus garden from the garden matrix by rocking laterally, side to side, while pulling. Once the piece was detached from the garden, the ants picked it up and carried it to the dump (Currie & Stuart 2001). Often, ants stayed motionless for prolonged periods of time hidden under or sometimes on the fungus garden, a behaviour that we categorized as “immobile in fungus garden”. ‘Trophallaxis’ was characterized as transfer of liquid between two workers (Richard & Errard 2009) and ‘manuring’ as gaster curving followed by production of a fecal droplet, picking it up with the mandibles and depositing it onto the fungus garden (Fernandez-Marin et al. 2003). We categorized the physical retrieval and active moving of eggs/larvae/pupae by the ants as ‘carrying eggs/larvae/pupae’. Finally, if the ants were in the foraging arena or dump site, we noted that as ‘outside fungus garden’.

Final bacterial cover Following the 3 days of behavioural observations after Escovopsis infection the fungus garden of the subcolonies was replaced with 2 g of fresh fungus garden including ca. 25 medium workers and 50 small workers from the same source colony that they eclosed in,

111 ensuring that the subcolonies had enough fresh fungus garden and work force to continue maintenance of the fungal cultivar. The bacterial cover of each marked large worker was estimated two weeks after fungal cultivar infection. Again, the ants were photographed and the bacterial cover scored blindly. Subsequently the ants were frozen at -20 °C for PCR analysis of the Pseudonocardia strains.

PCR analysis of Pseudonocardia strains From each of the subcolonies one individual was used for analysis of the Pseudonocardia strain identity. This procedure served to confirm that the large workers acquired the Pseudonocardia strain of the source colony they eclosed in. Samples from the colonies Ae.227 and Ae.266 used in the cross-fostering experiment by Armitage et al. (2011) were included, to allow direct comparison with the result of the present study. For isolation of Pseudonocardia, the laterocervical plates were dissected off the ants under a stereomicroscope and DNA extracted with the DNEasy Blood and Tissue kit following the manufactures instructions (Invitrogen, Hilden, Germany). Part of the elongation factor EF-1A gene was amplified with the Pseudonocardia specific primers 52F and 920R following Poulsen et al. (2005) in 20 μL PCR reactions with the AmpliTaq Gold kit (Applied Biosciences, New Jersey, USA) at the conditions 95° for 4 min followed by 40 cycles of 95° for 30 s, 62° for 50 s and 72° for 2 min with a final extension step of 72° for 10 min. PCR products were purified with an MSB Spin PCRapace kit (Invitek, Berlin, Germany) and sequenced by Eurofins MWG Operon (Ebersberg, Germany). The sequences were trimmed and compared in Sequencher 4.7 and the consensus sequences of both strains compared to known sequences with a NCBI GenBank BLAST search.

Data analyses The rate with which the ants achieved maximum bacterial cover was calculated in Excel by fitting a logistic growth model. This approach was used as it reflected growth of the bacterial population on the ants after eclosion, which levelled off to the maximum “carrying capacity” on the worker ants’ surface after a few weeks. Although the bacterial cover was measured on an ordinal scale, the fitted growth rate was treated as a continuous variable here and in the following analyses. The bacterial cover was estimated for each

112 rt individual at each time step (t), and then a logistic growth equation (Cover = KP0e / K rt +P0(e -1)) was fitted to the data, with K (the “carrying capacity”) set to the maximum cover (scale 12), and the value of r (the intrinsic rate of increase in bacterial cover) estimated for each individual using iterative least squares executed in the solver add-in of

Microsoft Excel 2011. The single value of P0 (the bacterial cover at day 1) that provided the best fit to the data across all individuals was also estimated iteratively, and found to be 2.67×10-4. Growth rates were Box-Cox transformed to maximize normality and homogeneity of variance for analysis. The rate of increase in bacterial cover and the final cover were analyzed in four different ways. The effect of the source colony of the fungus garden and the pupae was first tested in a two-way ANOVA, and then subcolonies were classified in three different ways and subjected to one-way ANOVA to test any overall differences between the treatments: subcolonies were classified as: 1. ‘own’ (8 subcolonies), ‘same’ (8 subcolonies) or ‘alien’ (16 subcolonies) describing pupae reared either in their own garden, a garden with the same Pseudonocardia strain or a garden with the other strain, or 2. ‘same’ or ‘alien, indicating whether pupae were reared in a garden with the same Pseudonocardia strain or a garden with the other strain and 3. according to the specific combination of the Pseudonocardia strain of the fungus garden and the native strain of the workers. Analyses were carried out using JMP 9.0.3 for Macintosh. The achieved power of the analysis was analysed post-hoc with G*Power 3.1.4 (Faul et al. 2009). Since the different behavioural responses to Escovopsis infections were not independent, as only one behaviour could be performed during each 30 s time period by each worker, and the performance of several behaviours was highly correlated (supplementary table 1), behavioural responses were converted to behavioural profiles using principal component analysis. Principal components were calculated using covariances, based on the proportion of 30 s time periods in which each behaviour was performed. Variation in the first four principal components was examined using ANCOVA with time (days) and rearing environment as main effects, and median Pseudonocardia cover infection as a covariate. Due to the substantial garden loss in the fungus garden from source colony Ae. 150 at the time of Escovopsis infection, we excluded these subcolonies (i.e. three subcolonies from either ‘own’ or ‘alien’ treatments)

113 from the analyses. The proportion of time spent outside the fungus garden was analysed separately using a Generalized Linear Model (GLM) with binomial errors, correcting for overdispersion. Once again the GLM examined the effects on time spent outside the garden of time (days), rearing environments, and median Pseudonocardia cover and their interactions, excluding the three subcolonies that had significant garden losses at the start of the infection experiments (supplementary table 2). When significant differences were found across the rearing environments, the effect was further analysed by looking for an effect of the Pseudonocardia strains Ps1 and Ps2.

Results Confirmation of strain identity It was possible to amplify and sequence the Pseudonocardia EF-1A gene from 29 out of the 32 experimental subcolonies and from 2 samples (one each from colony Ae.227 and Ae.322 used by Armitage et al. (2011)). This confirmed that all ants obtained the expected Pseudonocardia strain matching that of the fungus garden they eclosed in, and showed that Ae.227 and Ae.322 harboured Pseudonocardia strain Ps1 and Ps2, respectively. These were 99% similar to GenBank acc.no. DQ098127 (Ps1) and identical to GenBank acc.no. DQ098133 (Ps2), which were sampled in the same area by Poulsen et al. (2005). Individuals from the remaining five colonies either had a very low bacterial cover or PCR amplification was not achieved for unknown reasons.

Cross-fostering All 128 pupae eclosed to large workers but nine (7%) died shortly after eclosion. These mortalities were all from the same source colony, were equally distributed between the different subcolonies, and most likely caused by desiccation of pupae during subcolony setup, when pupae had to be isolated from their surrounding fungus material for somewhat variable periods of time. The dead workers were replaced with new mature pupae in all cases except one where the other three pupae in the subcolony had eclosed much earlier. In addition, one newly eclosed worker from another source colony did not develop properly, remained immobile with deformed legs, and was removed from the experiment. Five dead medium nursing workers were also replaced during the experiment.

114 Three subcolonies lost more than half of their fungus garden in the course of the experiment, but the decreased garden mass did not affect the growth of Pseudonocardia

cover (One-Way ANOVA, F1,125 = 0.1416, p = 0.7073). However, due to the likely effect of these decreased garden mass at the time of Escovopsis infection experiments, these subcolonies were excluded from the behaviour analyses (see above). Poulsen et al. (2003a) reported a highly specific ontogeny of the bacterial cover on newly eclosed Acromyrmex large workers, and designed a scale to score this variation from 1 to 12. In our experiments, the bacterial growth on the ants did not entirely follow this pattern, as bacterial cover on the dorsal side in some cases preceded growth on the laterocervical plates. These deviating individuals were observed in all treatments. In order to standardize scaling, we scored bacterial covers primarily based on the growth on the dorsal side. We observed the same general increase in bacterial cover over the first two weeks, with a subsequent decrease, consistent with Poulsen et al. (2003b) (Fig. 2A). To calculate the rate of bacterial growth on the ants, the cover was scored on Day 1, 4, 8, 12, 16 and 20. Bacterial cover at Day 20 was only estimated for 21 of the 32 subcolonies, as the remaining were already experimentally infected with Escovopsis at this stage, but for these subcolonies the bacterial cover was not significantly different from the cover at Day 16 within each fungus garden/pupae combination (Wilcoxon Signed Rank test, W = 7, p = 0.436; Fig. 2A). When comparing the rate at which maximum bacterial cover was achieved between all 16 combinations of fungus gardens and pupae, there was a marginally insignificant difference between receiving colonies in bacterial growth rate

(ANOVA on Box-Cox transformed rates; F3,110 = 2.53, p = 0.060), with growth rate of bacteria in colony Ae.331 being significantly higher than that in colony Ae.160 (Tukey HSD test, see Fig. 2B), and a significant effect of colony from which the pupae were

derived (F3,110 = 5.00, p = 0.003; ants from colonies Ae.150 and Ae.331 supporting lower growth rates than those from colony Ae.488), but there was no significant interaction

term (F9,110 = 1.09, p = 0.376). No differences in the rate of acquiring cover were found when testing for an effect of overall classification of subcolonies (‘own, same or alien’

fungus garden, F2,123 = 0.172, p = 0.843; ‘same or alien’ fungus garden, F1,123 = 0.093, p

= 0.761; ‘Ps1/Ps1, Ps1/Ps2, Ps2/Ps2 or Ps2/Ps1, F3,122 = 0.345, p = 0.793). There was also no statistical difference in the growth rate variances (untransformed data) between

115 categories of bacterial matching (Levene’s test, classified as ‘own, same, alien’: F2,123 =

0.939, p = 0.394; classified as ‘same, alien’: F1,123 = 0.504, p = 0.479). To estimate the probability of detecting a difference in the growth rate between the combinations in the given set-up we performed a power analysis in the program G*Power, finding the effect size that would be detectable at an 80% likelihood for the analysis of the effect of the rearing environment. It was found that the variance between groups had to be 0.25-0.3 times greater than the variance within groups for a difference to be detected, which is classified as a small-medium effect size. Differences in bacterial growth rate between the treatments should thus have been detected with the given sample size. The bacterial cover on the day of Escovopsis infection was quantified for all ants and the median coverage for each subcolony calculated. A significantly higher variation in bacterial cover was found for the subcolonies classified as ‘alien’ compared to those classified as ‘own’ and ‘same’ although the median cover was not significantly different, suggesting that bacterial growth on ants reared in the ‘alien’ environment were more sensitive to variables other than the host-symbiont match, such as the bacterial cover of nestmates etc. (Levene’s test for unequal variances, F2,29 = 3.815, p = 0.0263; Welch’s

ANOVA allowing for unequal variances, F2,18.506 = 0.6165, p = 0.551).

Prophylactic behaviour following Escovopsis challenge The proportion of time spent outside the fungus garden was examined using a GLM with binomial errors. There was a significant effect of rearing environment on time spent outside the fungus garden (Likelihood ratio χ2 = 6.16, df = 2, p = 0.046), where ants in the ‘own’ environment spent more time outside the fungus garden than ants reared in the ‘same’ and ‘alien’ environments (Tukey HSD test). All other effects were insignificant (see supplementary table 2). Further analysis within rearing environments revealed that there was a significant effect of the specific Pseudonocardia strain, with workers derived from colonies with strain Ps2 spending a greater proportion of time outside the fungus garden (Likelihood ratio χ2 = 10.8, df = 1, p = 0.001; Fig. 3A). The nine behaviours that we observed in the fungus garden (Fig. 4) were reduced to a smaller set of uncorrelated behavioural profiles using Principal Component Analysis

116 (PCA). The first four principal components (PC1-4) explained 48.0%, 24.0%, 10.5% and 9.78% of the variation in all behaviours respectively, and were judged to capture all significant variation in behaviour based on a scree-plot and the Joliffe criterion. PC1 captured the highly negatively correlated behaviours “Self grooming” (higher values) and “Immobile in fungus garden” (lower values)(Fig. 3B), while PC2 primarily captured “Garden grooming” (Fig. 3C), PC3 “Allo-grooming”, and PC4 “Garden weeding”. For more details of the PCA analysis see supplementary figure 1. Variation in the behavioural profiles represented by PC1 to PC4 were then analyzed using separate ANCOVAs with time (days) and rearing environment as main effects, and median Pseudonocardia cover as a covariate, plus all their interactions. The minimal adequate model for each behavioural profile was found by backward elimination of insignificant terms. Behavioural profiles are summarised in Fig. 3 and Fig. 4. For PC1,

the minimal adequate model contained only the main effect “environment” (F2,83 = 4.51, p = 0.014) and the covariate “median bacterial cover” (F1,83 = 7.71, p = 0.007), with workers covered in an alien strain of bacteria, and those with lower bacterial cover showing greater immobility and less self-grooming behaviour. For PC2, the minimal adequate model only contained the interaction between the main effect “environment”

and the covariate “median bacterial cover” (F1,84 = 4.16, p = 0.019), with a negative relationship between level of bacterial cover and garden-grooming behaviour for those ants bearing the bacteria from their own colony, and a positive relationship for the other two groups (“same” and “alien”). There were no significant predictors of PC3, while for PC4, the minimal adequate model only contained the main effect “time”, with ants examined on the first day after infection showing more garden weeding behaviour than those examined on subsequent days. For full details of the behavioural analysis, see supplementary tables 3-6. For those behavioural profiles that were found to be correlated with rearing environment, further analysis of the Pseudonocardia strain effect revealed that PC1 was higher for workers derived from colonies with strain Ps1, which represented

more self-grooming and less immobility for this group (F1,81 = 4.84, p = 0.031; Fig. 3B), but there were no significant correlations between strain and PC2 (representing fungus- grooming; Fig. 3C).

117 Final bacterial cover The final bacterial cover was quantified two weeks after Escovopsis infection. Given the experimental setup, with pupae within a subcolony eclosing at different days and the time of infection with Escovopsis varying between subcolonies, there were differences in how old the ants were at this time. There was however no correlation between the days since eclosion and the final bacterial cover (R < 0.01, p = 0.988). The estimate of the final bacterial cover was used to test whether there were significant differences between the cross-fostering combinations. This was found to be the case, with a significant effect of

pupal source colony and garden source colony (Two-Way ANOVA, garden: F3,106 =

11.216, p < 0.0001; pupae: F3,106 = 5.544, p = 0.0014; interaction: F9,106 = 1.314, p = 0.230), with subcolonies with garden from colony Ae.331 and Ae.150 having a significantly higher bacterial cover (Fig. 5A) and pupae from colony Ae.331 having a higher cover than pupae from Ae.160 and Ae.488, and those from Ae.150 being intermediate (Fig. 5B). No differences in the final bacterial cover were found between the different classifications of subcolonies (One-Way ANOVAs, p > 0.33 in all cases), i.e. the rearing environment alone apparently did not affect the final cover. For comparison with the results of Armitage et al. (2011) we also tested the effect of the native strain (Ps1 and Ps2) of the fungus garden and pupal source, which gave a non-significant trend towards

ants with Ps2 as native strain having a higher cover (One-Way ANOVA, garden: F3,118 = 1.711, p < 0.168, Fig. 5C). No correlation was found between the bacterial cover at the time of infection and two weeks after infection (R < 0.01, F = 0.139, p = 0.710) but the majority of ants experienced a drop in bacterial cover (mean difference of the two estimates of 5.53 ± 3.02 SD scale units).

Discussion In the present study we asked whether the apparent high specificity between Acromyrmex ants and strains of Pseudonocardia bacteria is maintained by faster growth of their vertically acquired “native” strain or a more efficient defence against Escovopsis, the only known specialized parasite of fungus gardens (Reynolds & Currie 2004). Differential growth would imply at least some degree of fine-tuned coevolution between

118 the ant glandular secretions that feed the bacterial cultures, but not necessarily that the ants recognize their dominant symbiont. Differential efficiency in controlling Escovopsis in cross fostering experiments would imply that it is the interaction between ant behavior and specific Pseudonocardia strains, rather than variation between ant colonies or Pseudonocardia strains per se, that would determine control efficiency. Evidence for such effects would indicate some form of coadaptation between vertically transmitted partners. In the sections below, we will evaluate our findings in light of this hypothetical framework.

Differential growth rates of Pseudonocardia strains on cross-fostered workers We did not find support for the hypothesis that the initial bacterial growth on the ants is dependent on whether the vertical transmission match between host and symbiont is maintained. After eclosion the ants readily obtained the Pseudonocardia strain of the fungus garden or nursing workers they were exposed to. No overall correlations between the bacterial cover growth rate and type of subcolonies (‘own’, ‘same’, ‘alien’) were found, even though ants from one colony (Ae.488) and some of the pupae/garden combinations reached maximum bacterial cover faster than others, consistent with findings by Poulsen et al. (2003a,b). Our results thus indicate that there were differences between source colonies in bacterial growth rate, but that these conditions by themselves were not responsible for maintaining host/symbiont specificity. The relatively high overall variance between individuals suggest that uncontrolled factors also affect bacterial growth rates. Bacterial cover of the other ants in the fungus garden, potentially including other bacteria than Pseudonocardia, and variation in the extent to which standard laboratory provisioning of colonies may not have quite the same effect on genetically different ants and fungus gardens would seem realistic possibilities. Also the fact that Acromyrmex pupae are variably covered in hyphae of the fungus garden prior to eclosure (Armitage et al. 2012) might affect the conditions for bacterial growth on newly eclosed workers. Clearer effects of cross fostering were found when analysing the final bacterial covers, estimated two weeks after infection. A previous study by Armitage et al. (2011) showed that there was a significant effect of both the source colony and cross-fostering of

119 ants on the final bacterial cover of mature foragers, which normally has a score of ca. 3. The effect was subtle but significant because of the large sample size of >900 ants, but with only two source colonies it was not possible to infer how general this effect was. We would have liked to compare that result with the outcome of our present four-colony experiment, but our estimates of the final cover ended up higher (mean score of 4.87 ± 2.11 SD), suggesting that the ants had not yet reached their final low cover typical of mature foragers by the time we terminated our experiment. The observed higher cover of ants from colony Ae.150 and Ae.331 may thus be confounded by these individuals achieving their maximum cover slower, consistent with pupae from Ae.488 being significantly faster and reaching their final low cover later. We did however confirm that the colonies of the Armitage et al. (2011) study harboured two different strains, so that result could indeed have been due to a mismatch. While not significant, the trend of ants harbouring Ps2 as native strain reaching a higher cover (Fig.5B) was thus consistent with Fig. 2 in Armitage et al. (2011) where colony Ae.266 with Ps2 ended up with a higher cover than colony Ae.227 with Ps1. Overall data thus support that there are strain and colony specific effects on the final bacterial cover of large Acromyrmex workers, with one of the strains maintaining a higher cover when grown on its native host, consistent with at least some degree of co-adaptation and host-specificity in bacterial growth. However, it remains to be seen how important the degree of native strain cover actually is for mature workers that have reached the foraging stage, as these workers are unlikely to still participate in fungus garden disease defence and transmission of Pseudonocardia to newly eclosed workers.

Behavioural changes induced by cross-fostering Callow workers with extensive bacterial cover have been suggested to be particularly important for defence against Escovopsis in the fungus garden (Currie et al. 2003). Such defences will normally be intricately connected to the expression of behaviours, as they will somehow need to vector the bacterial antibiotics towards their disease targets. For example, full cover callow workers have been found to often sit still in the lower parts of a fungus garden, where problems with Escovopsis are most likely (Abramowski et al. 2010). Our cross-fostering experiments allowed us to explicitly analyse the possible

120 interactions between carrying own, same, and alien Pseudonocardia strains and several prophylactic behaviours, while adjusting for intrinsic variation across ant families and fungus garden symbionts. These results confirmed the importance of callow workers with a high Pseudonocardia cover for Escovopsis defence as we found positive correlations between Pseudonocardia cover and self- and garden-grooming behaviour, as well as increased garden weeding behaviour immediately after infection compared with the following two days (Fig. 4). Most of these grooming behaviours, such as self-grooming, garden-grooming and garden weeding have also previously been shown to be important for the successful control of Escovopsis infections (Currie & Stuart 2001). We found differences in the proportion of time that marked ants spent in the fungus garden depending on the environment in which they were reared, with ants reared in fragments of their own colony spending more time outside the fungus garden than ants reared in alien nests, whether they shared the same Pseudonocardia strain or not. There was also an effect of natal Pseudonocardia strain, with workers derived from colonies with strain Ps2 spending more time outside the fungus garden. This indicates some overall colony-specific behaviour, but it is not possible to infer whether this is related to defence against Escovopsis, or whether it is related to other aspects of the integration of cross-fostered workers. For the nine behaviours recorded within the fungus garden, variations in behavioural profiles (based on principal component analysis) that included self-grooming, garden-grooming, and immobility inside the fungus garden were significantly associated with the rearing environment experienced by worker ants and their own level of Pseudonocardia cover. Ants with ‘alien’ and low bacterial cover spent more time immobile inside the fungus garden and less time performing self-grooming (alien > same = own), as did those with natal Pseudonocardia strain Ps2, indicating a strain-specific response. Analyses of fungus-grooming behaviours (PC2) also showed interesting colony- and strain-specific effects, as ants with the Ps1 strain spent less time on grooming, but only when they were reared with their own bacterial and fungal symbiont. These differences can be interpreted as the ants either “knowing” the efficiency of their native Pseudonocardia symbiont, or that there is an actual difference in the efficiency of the antibiotic defence dependent on the host symbiont match, resulting in non-self symbionts

121 requiring more labour to achieve the same efficiency. In addition the data show a clear difference between the two Pseudonocardia strains. Ants from all rearing environments were found to garden-weed their infected fungus garden the most on the day of infection, with proportion of time spent garden-weeding decreasing on days 2 and 3, consistent with the rapid detection and removal of contaminated fungus garden pieces. The increase in garden weeding response after detecting Escovopsis infection in their fungus garden fragment is consistent with earlier findings (Abramowski et al. 2010; Fernández-Marín et al. 2006; Fernández-Marín et al. 2009), showing that this ability is not affected by any transmission match between host and bacterial symbiont. The inferences above seem consistent with the overall levels of worker activity inside fungus gardens and with the rate at which newly eclosed workers proceed towards forager tasks (Little et al. 2006). Ants with alien Pseudonocardia spent more time inactive in the fungus garden, whereas ants with matching Pseudonocardia “graduated” more rapidly into their forager careers, which may be somehow related to a lower degree of familiarity to the other ants and fungus garden material that these cross-fostered ants interact with. Although much further work will be needed to unravel these complex patterns of ontogenetic development, our results suggest that the hypothesis that ants inoculated with non-native Pseudonocardia both need to work harder as callows and take longer to become foragers is worth further explicit testing.

Conclusions The general challenges (Mueller et al. 2008; Sen et al. 2009) of our initially naïve understanding of the interaction between cuticular actinomycetes and fungus-growing ants has initiated useful discussion and incisive further work (Barke et al 2011, Scheuring & Yu 2012; Seipke et al. 2011). However, our present study and Andersen et al. (in review) clearly suggest that it is too early to dismiss the coevolution model between Pseudonocardia (and possibly other actinomycetes) and specific lineages of attine ants. While the fact that different Pseudonocardia strains can readily colonize the ant cuticle are in line with an environmental acquisition model, we presented clear evidence that this is not without a cost to the host ants when facing an infectious disease supporting a role of coevolution between host and symbionts. The two Pseudonocardia strains have been

122 found to stably co-occur at similar frequencies in the Panamanian host population (Andersen et al. in review; Poulsen et al. 2005). The present study provides insight in the trade-offs for the ant host that may facilitate this coexistence: whereas the Ps2 strain may achieve a higher final cover on the ants the Ps1 strain was more efficient in Escovopsis defence. Future studies on ant/Pseudonocardia symbiont specificity should address these trade-offs, taking into account the correlations between ant and fungus garden genotype in the prophylactic defence against Escovopsis parasitism.

Acknowledgements We thank Louise Lee Munk Larsen for technical assistance in rearing fungal species, Michael Poulsen for sharing his skills on working with Escovopsis weberi. All authors were supported by a grant from the Danish National Research Foundation.

References

Abramowski, D., Currie, C.R. & Poulsen, M. 2010. Caste specialization in behavioral defenses against fungus garden parasites in Acromyrmex octospinosus leaf-cutting ants. Insectes Sociaux 58: 65-75.

Andersen, A.B., Hansen, L.H. & Boomsma, J.J. in review. Deconstructing a disease- defense symbiosis: specificity and stability of Acromyrmex-Pseudonocardia associations in changing environments.

Armitage, S.A.O., Broch, J.F., Fernández-Marín, H., Nash, D.R. & Boomsma, J.J. 2011. Immune defense in leaf-cutting ants: a cross-fostering approach. Evolution 65: 1791- 1799.

Armitage, S.A.O., Fernández-Marín, H., Wcislo, W.T. & Boomsma, J.J. 2012. An evaluation of the possible adaptive function of fungal brood covering by attine ants. Evolution 66, 6: 1966-1975

Barke, J., Seipke, R.F., Grüschow, S., Heavens, D., Drou, N., Bibb, M.J., Goss, R.J.M., Yu, D.W. & Hutchings, M.I.. 2010. A mixed community of actinomycetes produce multiple antibiotics for the fungus farming ant Acromyrmex octospinosus. BMC biology 8: 109.

Barke, J. Seipke, R.F., Yu, D.W. & Hutchings, M.I. 2011. A mutualistic microbiome: How do fungus-growing ants select their antibiotic-producing bacteria? Communicative Integrative Biology 4: 41-43.

123 Boomsma, J.J. & Aanen, D.K. 2009. Rethinking crop-disease management in fungus- growing ants. Proceedings of the National Academy of Sciences of the United States of America 106: 17611-17612.

Bot, A.N.M., Currie, C.R., Hart, A.G. & Boomsma, J.J. 2001. Waste management in leaf- cutting ants. Ethology Ecology Evolution 13: 225-237.

Cafaro, M.J., Poulsen, M., Little, A.E.F., Price, S.L., Gerardo, N.M., Wong, B., Stuart, A.E., Larget, B., Abbot, P. & Currie, C.R. 2011. Specificity in the symbiotic association between fungus-growing ants and protective Pseudonocardia bacteria. Proceedings of the Royal Society B: Biological Sciences 278: 1814-1822.

Currie, C.R. & Scott, J.A. 1999. Fungus-growing ants use antibiotic-producing bacteria to control garden parasites. Nature 398: 701-705

Currie, C.R., Mueller, U.G. & Malloch, D. 1999. The agricultural pathology of ant fungus gardens. Proceedings of the National Academy of Sciences of the United States of America 96: 7998-8002.

Currie, C.R. 2001. A community of ants, fungi, and bacteria: a multilateral approach to studying symbiosis. Annual Review of Microbiology 55: 357-380.

Currie, C.R. & Stuart, A.E. 2001. Weeding and grooming of pathogens in agriculture by ants. Proceedings of the Royal Society B: Biological Sciences 268: 1033-1039.

Currie, C.R., Bot, A.N.M. & Boomsma, J.J. 2003. Experimental evidence of a tripartite mutualism: bacteria protect ant fungus gardens from specialized parasites. Oikos 1: 91-102.

Currie, C.R., Poulsen, M., Mendenhall, J., Boomsma, J.J. & Billen, J. 2006. Coevolved crypts and exocrine glands support mutualistic bacteria in fungus-growing ants. Science 311: 81-83.

Faul, F., Erdfelder, E., Buchner, A. & Lang, A-G. 2009. Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behavior Research Methods 41: 1149-1160.

Fernández-Marín, H., Zimmerman, J.K. & Wcislo, W.T. 2003. Nest-founding in Acromyrmex octospinosus (Hymenoptera, Formicidae, Attini): demography and putative prophylactic behaviors. Insectes Sociaux 50: 304-308.

124 Fernández-Marín, H., Zimmerman, J.K., Nash, D.R., Boomsma, J.J. & Wcislo, W.T. 2009. Reduced biological control and enhanced chemical pest management in the evolution of fungus farming in ants. Proceedings of the Royal Society B: Biological Sciences 276: 2263-2269.

Herre, E., Knowlton, N., Mueller, U.G. & Rehner, S.A. 1999. The evolution of mutualisms: exploring the paths between conflict and cooperation. Trends in Ecology & Evolution 14: 49-53.

Ivens, A.B.F., Nash, D.R., Poulsen, M. and Boomsma, J.J. 2009. Caste-specific symbiont policing by workers of Acromyrmex fungus-growing ants. Behavioral Ecology 20: 378-384.

Little, A.E.F. & Currie, C.R. 2007. Symbiotic complexity: discovery of a fifth symbiont in the attine ant-microbe symbiosis. Biology Letters 3: 501-504.

Mattoso, T.C., Moreira, D.D.O. & Samuels, R.I. 2011. Symbiotic bacteria on the cuticle of the leaf-cutting ant Acromyrmex subterraneus subterraneus protect workers from attack by entomopathogenic fungi. Biology Letters 8: 461-464.

Mikheyev, A.S., Mueller, U.G. & Abbot, P. 2010. Comparative dating of attine ant and lepiotaceous cultivar phylogenies reveals coevolutionary synchrony and discord. The American Naturalist 175: E126-E133.

Muchovej, J.J. & Della Lucia, T.M. 1990. Escovopsis, a new genus from leaf cutting ant nests to replace Phialocladus nomem invalidum. Mycotaxon 37: 191-195.

Mueller, U.G. 2002. Ant versus fungus versus mutualism: ant-cultivar conflict and the deconstruction of the attine ant-fungus symbiosis. The American Naturalist 160: S67-S98.

Mueller, U.G., Dash, D., Rabeling, C. & Rodrigues, A. 2008. Coevolution between attine ants and actinomycete bacteria: a reevaluation. Evolution 62: 2894-2912.

Mueller, U.G. 2012. Symbiont recruitment versus ant-symbiont co-evolution in the attine ant-microbe symbiosis. Current Opinion in Microbiology, in press: 1-9.

Poulsen, M. & Boomsma, J.J. 2005. Mutualistic fungi control crop diversity in fungus- growing ants. Science 307: 741-744.

125 Poulsen, M., Cafaro, M., Boomsma, J.J. & Currie, C.R. 2005. Specificity of the mutualistic association between actinomycete bacteria and two sympatric species of Acromyrmex leaf-cutting ants. Molecular Ecology 14: 3597-3604.

Poulsen, M., Bot, A.N.M., Currie, C.R., Nielsen, M.G. & Boomsma, J.J. 2003a. Within- colony transmission and the cost of a mutualistic bacterium in the leaf-cutting ant Acromyrmex octospinosus. Functional Ecology 17: 260-269.

Poulsen, M., Bot, A.N.M. & Boomsma, J.J. 2003b. The effect of metapleural gland secretion on the growth of a mutualistic bacterium on the cuticle of leaf-cutting ants. Naturwissenschaften 90: 406-409.

Quelle, D. & Strassmann, J. 2009. Beyond society: the evolution of organismality. Philosophical transactions of the Royal Society of London Series B Biological Sciences 364: 3143-3155.

Reynolds, H.T. & Currie, C.R. 2004. Pathogenicity of Escovopsis weberi: the parasite of the attine ant-microbe symbiosis directly consumes the ant-cultivated fungus. Mycologia 96: 955-959.

Richard, F-J. & Errard, C. 2009. Hygienic behavior, liquid-foraging, and trophallaxis in the leaf-cutting ants, Acromyrmex subterraneus and Acromyrmex octospinosus. Journal of Insect Science 9: 63.

Rodrigues, A., Bacci, M., Mueller, U.G., Ortiz, A. & Pagnocca, F.C. 2008. Microfungal “weeds” in the leafcutter ant symbiosis. Microbial Ecology 56: 604-614.

Scheuring, I. & Yu, D.W. 2012. How to assemble a beneficial microbiome in three easy steps. Ecology Letters in press: doi: 10.1111/j.1461-0248.2012.01853.x.

Schultz, T.R. & Brady, S.G. 2008. Major evolutionary transitions in ant agriculture. Proceedings of the National Academy of Sciences of the United States of America 105: 5435-5440.

Seipke, R.F., Barke, J., Brearley, C., Hill, L., Yu, D.W., Goss, R.J.M. & Hutchings, M.I. 2011. A single Streptomyces symbiont makes multiple antifungals to support the fungus farming ant Acromyrmex octospinosus. PLoS ONE 6: 8.

Sen, R., Ishak, H.D., Estrada, D., Dowd, S.E., Hong, E. & Mueller, U.G. 2009. Generalized antifungal activity and 454-screening of Pseudonocardia and Amycolatopsis bacteria in nests of fungus-growing ants. Proceedings of the National Academy of Sciences of the United States of America 106: 17805-17810.

Wheeler 1910. Ants: their structure, development and behaviour (3rd edition). Columbia University Press, Chapter 18, pp. 329.

126

Figure legends Figure 1. Cross-fostering experiment set-up. Fungus garden and pupae from four source colonies were used to set up 32 subcolonies in a duplicated (4x4) full-factorial design. Source colony identity is indicated by colour with two shades of blue used for colonies with Pseudonocardia strain 1 and two shades of red used for colonies with Pseudonocardia strain 2. Pupae from each source colony eclosed in two replicate subcolonies with the garden of their own colony (e.g. light blue pupae in light blue garden), a garden from a colony with the same Pseudonocardia strain (e.g. dark blue garden) and a garden from two colonies with the other Pseudonocardia strain (e.g. pink and red gardens). Drawing of pupae from http://etc.usf.edu/clipart.

Figure 2. Bacterial growth rates on ant cuticles. A. Boxplot of the bacterial cover scores of all individuals from all subcolonies from day 4 – 20 and 2 weeks after Escovopsis infection; quantile boxplots, with bottom and top of boxes representing the 25th and 75th percentiles, the middle band the median, and the lower and upper whiskers the minimum and maximum covers scored. At day 1 all individuals had a score of 0. B. Bacterial cover scores with the solid lines showing the median cover and the dashed lines indicating the 25% and 75% quantiles for all individuals in the two replicate subcolonies for each treatment. For each plot the source colony of the fungus garden and the pupae is noted, in addition to the identity of the native Pseudonocardia strain. Bold-lined boxes denote crossing of the ‘own’ strain, unshaded boxes denote crossing of the ‘same’ strain, and shaded boxes denote crossing of the ‘alien’ strain.

Figure 3. Performance of behaviours grouped by Pseudonocardia strains and environments. A. Time spent outside the fungus garden is presented as the proportion of time spent performing this behaviour while B-C. Behaviours that are plotted with the first and second principal axes obtained from PCA of behaviours. All data are plotted as quantitle boxplots, with bottom and top of boxes being 25th and 75th percentiles, the middle band the median, and the lower and upper whiskers the minimum and maximum frequencies recorded. Ants with strain Ps2 reared in their native colony spent more time outside the fungus garden and performed less garden-grooming compared to the other

127 categories, while self-grooming and immobility were performed similarily (for the ‘alien’ environment categories are labelled with first the strain of the fungus garden followed by the native strain of the pupae).

Figure 4. Behavioural profiles presented as pie charts showing the proportion of time spent on each behaviour. The y-axis shows the distributions of median bacterial cover of ants, the major x-axis shows the distribution of behaviours for ants in their ‘own’, ‘same’ and ‘alien’ rearing environments, and the minor x-axis shows the number of days post- infection on which behaviours were recorded (infection took place on day 1). Different colours denote the nine different behaviours recorded. Self- and garden-grooming were the most commonly performed behaviours, with garden-weeding mainly observed on day 1 of infections.

Figure 5. Final bacterial cover of cross-fostered ants. The final bacterial covers of the experimental ants were estimated two weeks after infection with Escovopsis. There were significant differences in the final cover of ants hatching in fungus gardens from different source colonies, with those exposed to gardens fragments from Ae.150 and Ae.331 having a higher cover than those hatching in garden fragments of colonies Ae.488 and Ae.160, as indicated by the letters a and b showing the grouping following Tukey’s post hoc tests (A) and between pupae from the different source colonies, with those from Ae.331 having a higher cover than those of Ae.488 and Ae.160 with Ae.150 being intermediate (B). C. There was non-significant tendency (p < 0.168) towards ants from a source colony with strain Ps2 reared in a fungus garden with strain Ps2 to have a higher final cover. Box plots are the same as in Fig. 3.

Supplementary table and figure legends Supplementary table 1. The correlations matrix between 9 behaviours, excluding three subcolonies that suffered significant garden losses at the time of the fungus garden infection experiment. The correlations were calculated based on behavioural profiles of Principal Component Analysis (PCA), using covariances methods (see methods text for details).

128

Supplementary table 2. Results of a Generalized Linear Model on “time spent outside the fungus garden” with binomial errors, correcting for over-dispersion, testing the effect of time (days), rearing environment, median Pseudonocardia cover and their interactions. We report the Log-Ratio (L-R), degrees of freedom (df) and probability value of the effect tests. * denotes significant effects with p < 0.05.

Supplementary table 3. Results of stepwise ANCOVA on the first principal with time (days) and rearing environment as main effects, and median Pseudonocardia cover as covariate, excluding three subcolonies that had substantial garden mass losses at the time of the infection experiment. We report the F ratios, degrees of freedom (df) and probability values of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignificant effects. Significant effects (p < 0.05) are given in bold.

Supplementary table 4. Results of stepwise ANCOVA on the second principal with time (days) and rearing environment as main effects, and median Pseudonocardia cover as covariate, excluding three subcolonies that had substantial garden mass losses at the time of the infection experiment. We report the F ratios, degrees of freedom (df) and probability value of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignificant effects. Significant effects (p < 0.05) are given in bold.

Supplementary table 5. Results of stepwise ANCOVA on the third principal with time (days) and rearing environment as main effects, and median Pseudonocardia cover as covariate, excluding three subcolonies that had substantial garden mass losses at the time of the infection experiment. We report the F ratios, degrees of freedom (df) and probability value of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignificant effects. Significant effects (p < 0.05) are given in bold.

129 Supplementary table 6. Results of stepwise ANCOVA on the fourth principal with time (days) and rearing environment as main effects, and median Pseudonocardia cover as covariate, excluding three subcolonies that had substantial garden mass losses at the time of the infection experiment. We report the F ratios, degrees of freedom (df) and probability value of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignificant effects. Significant effects (p < 0.05) are given in bold.

Supplementary figure 1. Loading plots of the first four principal components (PC1-4) from a principal component analysis of the nine behaviours observed within fungus gardens. For each plot, the loading of that behaviour on each pair of PCs is plotted on a scale ranging from -1 to 1. Behaviours that have the highest loadings for each PC are shown in bold, with the imputation that variation in those behaviours is well-captured by the principal component.

130 Figure 1

Own Same Strange

Pseudonocardia strain 1 Pseudonocardia strain 2

131 Figure 2 A 12

cover 8

Pseudonocardia 2

day 4 day 8 day 12 day 16 day 20 Time since eclosion

14 daysinfection post

B Colony from which pupae were collected Ae.150 Ae.488 Ae.160 Ae.331 Ps1 Ps2

12 10 Ae.150 8 6 4 2 Ps1 0

12 Colony into whi cover 10 Ae.488 8 6 4

2 c h pupae were introduced

Pseudonocardia 0 12 10

8 Ae.160 6 4 2 Ps2 0 12 10 Ae.331 8 6 4 2 0 4 8 12 16 4 8 12 16 4 8 12 16 4 8 1 2 16 132 Time since eclosion (days) Figure 3

0.6 A 0.5

0.4

0.3

0.2 Proportion of time 0.1 outside the fungus garden

0.0

0.2 B

0.1

0.0 Self-grooming → PC1 -0.1

-0.2

← Immobile -0.3

-0.2 C

-0.1

0.0 PC2 0.1 Garden-grooming → 0.2

0.3 own Ps1 own Ps2 same Ps1 same Ps2 alien Ps1/Ps2 alien Ps2/Ps1 Specific Environment 133 Figure 4

Own Same Foreign Day Day Day 1 2 3 1 2 3 1 2 3

Median cover score Self grooming Garden grooming 7 Weeding Immobile Allogrooming Carrying larvae or eggs 6 Tending larvae or eggs Trophallaxis Manuring

134 Figure 5

12 a a 11 A B 10 a ab b 9 b b 8 7 6 5 b 4 3 2 Cover 2 weeks after infection 2 weeks Cover 1 150 160 331 488 150 160 331 488 Source colony fungus garden Source colony pupae

12 11 C 10 9 8 7 6 5 4 3 2 Cover 2 weeks after infection 2 weeks Cover 1 Ps1/Ps1 Ps1/Ps2 Ps2/Ps1 Ps2/Ps2 135 Strain fungus garden/pupae Supplementary tables

Table 1: The correlations matrix between 9 behaviours, excluding three subcolonies that suffered significant garden losses at the time of fungus garden infection experiment. The correlations were calculated based on behavioural profiles of Principal Component Analysis (PCA), using covariances method. Carrying Tending Garden- Immobile inside eggs/larvae/p eggs/larvae/pu Self-grooming grooming Weeding fungus garden Allo-grooming upae pae Trophallaxis Manuring Self-grooming 1.000 -0.253 0.049 -0.560 -0.140 -0.157 0.079 -0.013 -0.136 Garden-grooming -0.253 1.000 -0.174 -0.312 -0.141 -0.088 -0.004 -0.178 -0.001 Weeding 0.049 -0.174 1.000 -0.289 -0.265 0.032 0.130 -0.030 0.082 Immobile inside fungus garden -0.560 -0.312 -0.289 1.000 -0.080 -0.138 -0.335 -0.021 0.002 Allo-grooming -0.140 -0.141 -0.265 -0.080 1.000 0.105 -0.305 0.091 0.071 Carrying eggs/larvae/pupae -0.157 -0.088 0.032 -0.138 0.105 1.000 0.001 -0.011 -0.003 Tending eggs/larvae/pupae 0.079 -0.004 0.130 -0.335 -0.305 0.001 1.000 -0.020 0.013 Trophallaxis -0.013 -0.178 -0.030 -0.021 0.091 -0.011 -0.020 1.000 -0.003 Manuring -0.136 -0.001 0.082 0.002 0.071 -0.003 0.013 -0.003 1.000

Table 2: Results of Generalized Linear Model on 'time spent outside fungus garden' with binomial error, correcting for overdispersion, testing the effects of time (days), rearing environment, median Psuedonocardia cover and their interactions. We reported the Log-Ratio (LR), degrees of freedom (df) and probability value of the effect tests. * denotes significant effect with p < 0.05.

Source LR df p Cover (median) 0.3996 1 0.5273 )syad( emiT )syad( 2609.2 2 8332.0 Environment 6.1624 2 0.0459* emiT X revoC X emiT 2041.0 2 3239.0 Cover X Environment 4.8810 2 0.0871 Time X Environment 0.6376 4 0.9588 Cover X Time X Environment 0.3329 4 0.9876

Table 3: Results of stepwise ANCOVA on the first principal component with time (days) and rearing environment as main effects, and median Pseudonocardia cover as co-variate, excluding three subcolonies that had substantial garden mass losses at the time of infection experiment. We report the F ratios, degrees of freedom (df) and probability value of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignifivant effects. Significant effects ( p < 0.05) are marked in bold.

Source F ratio df P Cover (median) 7.707 1, 83 0.007 )syad( emiT )syad( 772.0 ,2 18 957.0 tnemnorivnE 705.4 ,2 38 410.0 emiT X revoC X emiT 691.0 ,2 18 228.0 Cover X Environment 0.649 2, 81 0.525 Time X Environment 0.692 4, 79 0.600 Cover X Time X Environment 0.594 4, 79 0.668

Table 4: Results of stepwise ANCOVA on the second principal component with time (days) and rearing environment as main effects, and median Pseudonocardia cover as co-variate, excluding three subcolonies that had substantial garden mass losses at the time of infection experiment. We report the F ratios, degrees of freedom (df) and probability value of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignifivant effects. Significant effects ( p < 0.05) are marked in bold.

Source F ratio df P Cover (median) 1.002 1, 83 0.320 )syad( emiT )syad( 102.2 ,2 28 711.0 tnemnorivnE 369.1 ,2 28 741.0 emiT X revoC X emiT 584.1 ,2 28 232.0 Cover X Environment 4.159 2, 84 0.019 Time X Environment 1.454 4, 80 0.224 Cover X Time X Environment 2.215 4, 80 0.075

Table 5: Results of stepwise ANCOVA on the third principal component with time (days) and rearing environment as main effects, and median Pseudonocardia cover as co-variate, excluding three subcolonies that had substantial garden mass losses at the time of infection experiment. We report the F ratios, degrees of freedom (df) and probability value of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignifivant effects.

Source F ratio df P Cover (median) 0.054 1, 85 0.817 )syad( emiT )syad( 437.1 ,2 48 624.0 tnemnorivnE 546.0 ,2 48 546.0 emiT X revoC X emiT 70000.0 ,2 48 889.0 Cover X Environment 0.374 2, 84 0.689 Time X Environment 0.003 4, 82 0.917 Cover X Time X Environment 0.005 4, 82 0.804

Table 6: Results of stepwise ANCOVA on the fourth principal component with time (days) and rearing environment as main effects, and median Pseudonocardia cover as co-variate, excluding three subcolonies that had substantial garden mass losses at the time of infection experiment. We report the F ratios, degrees of freedom (df) and probability value of the effect tests for removal from the minimal adequate model for significant effects, and for addition to the model for insignifivant effects. Significant effects ( p < 0.05) are marked in bold.

Source F ratio df P Cover (median) 0.099 1, 83 0.754 )syad( emiT )syad( 325.6 ,2 48 200.0 tnemnorivnE 151.2 ,2 28 221.0 emiT X revoC X emiT 335.0 ,2 28 985.0 Cover X Environment 1.133 2, 82 0.327 Time X Environment 0.424 4, 80 0.791 Cover X Time X Environment 0.400 4, 80 0.808

136 PC4 PC3 PC2 Immobile Immobile Immobile Allogrooming Carrying larvaeoreggs Allogrooming Trophallaxis Allogrooming Manuring Tending larvaeoreggs Trophallaxis Manuring Manuring Trophallaxis PC1 Weeding Garden grooming Carrying larvaeoreggs Garden grooming Carrying larvaeoreggs Self grooming Garden grooming Tending larvaeoreggs Weeding Weeding Self grooming Tending larvaeoreggs Self grooming Garden grooming Garden grooming Carrying larvaeoreggs Carrying larvaeoreggs Allogrooming Tending larvaeoreggs Manuring Manuring Tending larvaeoreggs Allogrooming PC2 Immobile Weeding Weeding Trophallaxis Trophallaxis Immobile Self grooming Self grooming

Garden grooming Tending larvaeoreggs Self grooming Immobile PC3 Supplementary gure 1 Supplementary Weeding Manuring Carrying larvaeoreggs Trophallaxis Allogrooming 137 CONCLUSION

AND

PERSPECTIVES

Leaf-cutting may be the obvious trait that give these ants their name, but it is just the rst step to the whole ant-fungus mutualism

Photo credit: Janni Larsen

138 CONCLUSION AND PERSPECTIVES Social insect populations are known for their optimal defences against invaders, with most research questions focus on identifying trade-offs in physiological mechanisms that balance the costs and benefits. In social insects, trade-offs are often in the form of prophylactic defences such as various grooming behaviours, adjusted to the levels of threat and complexities of sociality. In this thesis, we built from the foundation of prophylactic defences and take a few steps further by incorporating techniques from different scientific disciplines to aid in explore insect-fungus mutualism, insect-bacteria mutualism, and host-parasite interactions. Below I will conclude with questions that arise from these thesis chapters and present perspectives on how insights can be transferable to a similar model system.

1st chapter Techniques involving whole-genome sequencing are beginning to revolutionize the study of ecology and evolution by providing massively parallel sequencing at affordable cost, making genome and transcriptome sequencing possible for more projects and species. We took advantage of this technology and the availability of the Acromyrmex genome to examine trade-offs in the form of prophylactic defences and gene expressions, topics that have never been integrated into a single comprehensive project. We showed for the first time how increased general prophylactic behaviours affect the specific gene expression responses to pathogens of different kinds. The immune system was activated differentially when ants encountered a threat to self or to fungal symbiont, demonstrating that although prophylactic response to both threats were similar, the physiological pathways involved were vastly different. Our results of the genome-wide gene expression and especially on immune- related gene expression revealed candidate genes that potentially play important roles in immune defence against parasites that either harm the ants or their fungal symbiont, hence important in maintaining integrity of ant-fungus mutualism. These genes warrant further studies. Now that the sequences of these genes are known, conventional tools such as qPCR or proteomics re-construction would be useful for elucidating the identities and functions of these genes. Furthermore, the extent of alternative splicing would be

139 another interesting topic to address in the future with the current datasets. In other known model systems such as human and fruit flies, a way to increase the diversity in immune defences against parasites is achieved partly by gene regulation via alternative splicing. Similar gene expression mechanisms are expected to take place even more commonly in social insects as their need for defence extend to symbionts associated with their nests. Hence, genomic research would gain importance as more social insect genomes are becoming available.

2nd chapter An ant can be viewed as a walking chemical factory. A typical ant has 20+ major exocrine glands on her body that produce a complex and highly diverse mixture of chemicals essential for various aspects of her life, such as foraging, colony communication, discrimination against intruders, and mating. One exocrine gland unique to ants is the metapleural gland, an exocrine gland that secretes antimicrobial compounds. Due to the lack of muscle attachments to the glands, it has always been viewed as a passive organ. Through laboratory-based microbiology testing and works from others, we found parallels between ant metapleural gland action and an immune system, which have specific and non-specific components that are either always active or inducible. Our interpretation of an exocrine gland as an immune system, a notion that has never been suggested will likely spark controversy but is necessary to advance the field. Future research that adopt this notion would benefit most in identifying chemical components that have properties reminiscent to an immune system: general antimicrobial compound, specific antimicrobial compounds, compounds that are always active (i.e. constitutive) or compounds that are inducible, which can be identified by a combination of infection assays and chemical analyses such as high-performance liquid chromatography (HPLC). These compounds have undergone selection function in its current effective form and would potentially be of value to human health industry.

3rd chapter Symbionts that co-evolved with insects, reminiscent to crops cultivated by human, are an easy target for parasitism. Over the course of long-term mutualism, the fungal symbiont

140 gradually lost abilities to defend itself against parasites and disease, instead relying on ant partner for protection. For example, when ants are removed from their fungus garden the specialized fungal parasite Escovopsis quickly overwhelms the garden. The parasite is often present in the garden but is controlled by the ants prophylactic behaviours and antibiotics produced by bacteria housed on their cuticular surface. 50 years of research on Escovopsis parasitism in fungus-growing ant crops is timely reviewed in chapter 3. Our Escovopsis review adopted the Tinbergen-approach of asking questions, which served as a comprehensive way of addressing both proximate questions of development and mechanisms, and ultimate questions of co-adaptation and evolutionary history. Using this approach, we identified future questions that can close the current ‘gaps’ in the understanding of ecology and evolution of Escovopsis parasitism. Surprisingly, most of the basic questions such as origin of the association, its presence and potential role outside ant nests, transmission mode between colonies, and within colony dynamics had been neglected. As these questions are variable, we urge future researchers to be creative in choosing their methodologies. Many of these fundamental questions can be addressed with long-term monitoring of field colonies, augmented with laboratory-based infection assays. The advent of next-generation sequencing techniques, on the other hand, will aid in elucidating the origin of the host-parasite associations.

4th chapter Insect recruiting antibiotic-producing bacteria as a form of defence has only recently been recognized. One such system that leads this area of research is the attine ant- Pseudonocardia bacteria association, hypothesized to be the most important line of defence against Escovopsis parasitism. However, the specificity of the association of ants-bacteria and the specificity of antibiotics against Escovopsis are often examined separately, hence generating pieces of puzzles but rarely gave an overview of the interactions, resulting in conflicting views and debate in this field. We bridge the two independent questions by first examining the specificity of the ant-bacteria association, followed by testing the efficiencies of these associations against Escovopsis parasitism. Leaf-cutting ants always harbor a dominant strain of Pseudonocardia bacteria on their cuticular surface, although other bacteria species can be detected. This tight

141 association is expected to be the outcome of parasitism pressure. Although cross-fostered ants can be induced to adopt host bacteria cover through competitive dominance, ant- bacteria match varies in their Escovopsis defence with native ant-bacteria performed better than non-native ant-bacteria. Incorporation of Escovopsis parasitism in experiments is crucial in future studies as Escovopsis parasitism affect the ant-bacteria association outcome. Rather than focusing on specificity-related questions, our work provides an example of how combining parasitism can gave an overall understanding to the integrity of ant-bacteria association.

Perspectives While the experiments in this thesis were solely conducted on Acromyrmex leaf-cutting ants, the insights from these experimental results are relevant to other systems. One example is the Sirex-Amylostereum-Deladenus (woodwasp-fungus-parasite) interactions. A female Sirex woodwasp deposit her eggs by drilling a hole into the trunk of pine tree. During oviposition, she also deposits a piece of Amylostereum fungal cultivar that will degrade cellulose into nutrients that can be consumed directly by the growing Sirex larvae. Recently, cellulose-degrading bacteria were found to be associated with the Sirex- Amylostereum association. The efficiency of Sirex-Amylostereum symbiosis imposes a great threat to commercial and natural pine plantations across the globe. Parasitic nematode Deladenus have been employed globe-wide to control the woodwasp-fungus populations. The parasites, transmitted by adult Sirex, are detrimental to the symbiont fungus Amylostereum. Due to the woodwasp pest status, the questions that are directly relevant in biocontrol are the most pressing. The genomes of these three partners are now being sequenced. With the availabilities of these genomes, infection bioassays and genome-wide screening similar to chapter 1 can now be carried out, to answer questions such as virulence and resistance in parasitism. Exocrine glands secretions are important in the oviposition and subsequent colonization of Amylostereum into the pine tree host. Although there are scattered studies that look at single compound production, the insights from chapter 2 imply that glandular productions that are important in maintaining integrity of mutualism are often more complex and context-dependent. Culture-based and chemical analyses of host choice and

142 nutrient availabilities would aid in elucidating the roles these exocrine glands play in the mutualism. Parasitic nematode Deladenus is not the only parasite in the Sirex- Amylostereum model system, although it is often being taken for granted in biocontrol studies as the most effective parasite. It is timely to have a synthesis of parasite diversity and their effect (or lack thereof) in this commercially important pest system, identifying the ‘gaps’ in Deladenus studies and at the same time increasing the knowledge of parasite diversity that could be potential biocontrol agents. Finally, the acquisition of cellulose- degrading bacteria would be similar as the acquisition of antibiotic-producing bacteria in ant-fungus mutualism, involving trade-offs of cost and benefit. Rather than inventing the conflicting views and debate on woodwasp-bacteria association and specificity of enzymes in cellulose-degradation, these two questions should be examined in parallel as the specificity of woodwasp-bacteria association would depend on the efficiencies of the bacteria to degrade cellulose.

143 Acknowledgements

I offer my gratitude to my supervisor, Prof. Jacobus (Koos) Boomsma, who have supported me throughout the thesis. I am not the best student and have caused several occasions of high-blood pressure and headaches to Koos. Yet, Koos still offers his support to the very end. Louise Lee Munk Larsen, a fine technician and good friend who is instrumental in all my experiments that involved ‘fungi’. Three out of four chapters of my thesis have your fine touch in it. This thesis belongs to Louise as much as it is to me. Michael Poulsen for cementing the first publication of my PhD thesis and introducing me to the concept of dropbox sharing. I am now a staunch dropbox fan. Thank you! Morten Schiøtt for your crucial molecular skills in RNA extractions and open-door policy. I feel comfortable approaching you anytime and asking stupid questions. Thank you for your patience in the whole collaborative process. I could not ask for a better collaborator. Annette Jensen for entry to LIFE and entomopathogenic fungi, also for introducing me into a fascinating field that I will pursue more in my post-doc. I have yet much to learn and hope to continue the learning process with you in the future. David Nash for all the statistical advice offered freely. I couldn’t imagine writing a manuscript now without consulting you first on the proper statistical analyses, and especially thank you for your patience in answering the same questions over and over again. Sandra Andersen for trusting in my Escovopsis infection skills and exchange of ideas. I appreciate the comradeship of long hours in the ant room. This is truly a collaborative process that I enjoyed. Henrik de Fine Licht, Anne Andersen, Janni Larsen and Marlene Stürup for being supportive and especially fun companies in the same office space. Because of all of you, I look forward to come to work just so I can spent some time gossiping and laughing with the real Danes – a true cultural immersion. Backbones to CSE, Slyvia Mathiasen and Bettina Markussen have made the work and stay in Copenhagen possible and enjoyable. Finally, to all the CSE members, there is no way I can remember all of your names (but I can definitely remember your faces and kindness). Thank you! Writer Development Course 3, 2012, provide avenue for working environment where majority of my PhD thesis were written. Sarah Haas for teaching us to be in charge of our writing. I got lots done in the short duration of the course, and discover joy in writing amid the stress. My writer group members: Andreas Næsby, Helle B Klenow, Karin Peschardt and Sophie Nissen, for the writing comradeships. You are my sisters/brother in arms in the thesis writing process. I look forward to every writer group meeting. My only regret is that I won’t be able to attend our future writer group meetings.

During the thesis writing process, Marlene Stürup, Vicky Huang, Julia Ng, Sean Byars, and Rachelle Adams patiently went through and proof-read the thesis at various stages, especially

144 Acknowledgements

Vicky Huang reading through the whole thesis and correcting all my grammar mistakes. Peter Jørgensen translated the English summary to ‘meaningful’ Danish resumé. My father (Yek Nai Ging) did the first Chinese translation but due to his too scholastic language, Xu Siaoting and Zhang Goujie make some editing for more popular reading. I would like to specially thank Xu Siaoting for not only editing the Chinese summary but also feeding me almost daily in the last week of thesis writing. Thank you for being with me all the way to the finish line!

Yogacentralen, being my family in Denmark, have kept me mentally sane and physically fit during my stay in Copenhagen. Jack Davis and Peter Huber, as my surrogate parents in Denmark have make yoga and sometimes food available to me at 7 days per week, rain or shine! My Iyengar teachers Lone Kristensen, Ulla Nielsen, and Mette Vulpius for teaching me to take charge of my body and introducing me to the toys/props of yoga. My Vinyasa/Ashtanga teacher/practice partner Tanya Lee Markul for your ever positive personality and heart-opening practice. Go chocolate! Not to mention my Kundalini teacher Pia Saranpreet Kaur for all the ‘breath of fire’ and introduction to harmonium playing.

My parent (Liew Goat Khoon and Yek Nai Ging) have not asked me to fulfill the typical Chinese daughter responsibilities, which allow me to pursue my ‘dreams’ of globe tripping in the name of science. My siblings (Sze Geen, Joo Liang and Zu Jie) ease all my worries and responsibility of being a ‘good’ Chinese daughter. I thank all of you for that, and I sincerely wish I am not too big a disappointment to all of you.

145 CURRICULUM VITAE

Sze Huei Yek Skaffervej 11, 4 tv 2400 København NV +45-24 46 86 69 [email protected] or [email protected]

EDUCATION

2009-2012 PhD, Centre for Social Evolution, University of Copenhagen (expected) 2006-2009 MA, Section of Ecology, Evolution and Behavior, University of Texas at Austin 2004-2006 Grad. Diploma, School of Tropical Zoology, James Cook University, Australia 1997-2001 B.Sc. Biology and Chemistry (Campbell University, U.S.A, Malaysia campus, 2001)

______

RELEVANT WORKING EXPERIENCE

2009-2012 PhD fellow at University of Copenhagen, Denmark 2006-2008 Teaching assistants for undergraduate courses at University of Texas at Austin 2001-2004 Research officer in Tropical Biodiversity at Forestry Research Institute of Malaysia (FRIM) and World Wide Fund Malaysia (WWFM)

______

CONFERENCES

2010 IUSSI, North European Winter Meeting, Oral presentation 2010 IUSSI VXI Congress, Poster presentation 2006 IUSSI XV Congress, Poster presentation 2005 Joint conferences of Australian Entomological Society/Invertebrate Biodiversity and Conservation/Society of Australian Systematic Biologists, Oral presentation

______

PUBLICATIONS

S.H. Yek, D.R. Nash, A.B. Jensen and J.J. Boomsma. 2012b. Regulation and specificity of antifungal metapleural gland secretion in leaf-cutting ants. Proceedings of The Royal Society B: Biological Sciences doi: 10.1098/rspb.2012.1458.

S. H. Yek, J. J. Boomsma and M. Poulsen. 2012a. Towards a better understanding of the evolution of specialized parasites of fungus-growing ant crops. Physce Article ID: 239392.

S.H. Yek and U.G. Mueller. 2010 The Metapleural Gland of Ants. Biological Reviews 91: 201-224.

146 S.H. Yek, S. Williams, C. Burwell, S. Burwell, S. Robson and R. Crozier. 2009 Ground dwelling ants as surrogates for establishing conservation priorities in the Australian Wet Tropics. Journal of Insect Science 12: 1-12

F. Ito, R. Hashim, S.H. Yek, E. Kaufman, T. Akino and J. Billen. 2004 Spectacular batesian mimicry in ants. Naturwissenschaften 91: 481-484

M. Maruyama, S.H. Yek, R. Hashim and F. Ito. 2003 A new myrmecophilous species of Drusilla (Coleoptera, Staphylinidae, Aleocharinae) from Peninsular Malaysia, a possible Batesian mimic associated with Crematogaster inflata. Japan Journal of Systematic Entomology 9: 267-275.

______

HONORS AND AWARDS

Dorothea Bennett Memorial Graduate Fellowship, 2008

Terrell H. Hamilton Endowed Graduate Fellowship, 2007

Research award, New South Wales Ecological Society, 2005

______

PEER-REVIEWING EXPERIENCE

Reviewed for Microscopy Research and Technique

______

LANGUAGES

Mandarin mother tongue, fluent in English, conversation-level in Cantonese and Bahasa, basic in Japanese, 5-classes worth of Danish, planning to learn Gurmukhi and Afrikaans…

147