FACULTY OF SCIENCE FFACULTY A C U L T Y OOFF SSCIENCE C I E N C E UNIVERSITY OF COPENH AGEN UNIVERSITY OF COPENH AGEN

Dynamics of ant-microbial interactions Coevolution along the parasitism-mutualism continuum

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

Sandra Breum Andersen December 2011

Academic advisors: David P. Hughes & Jacobus J. Boomsma 2 Preface

This PhD thesis is the result of three years of work at the Centre for Social Evolution at University of Copenhagen, pleasantly interrupted by shorter periods of fieldwork in Panama and Brazil and a stay at Penn State University USA, under supervision of David Hughes and Jacobus (Koos) Boomsma. I was funded by a grant from the Faculty of Science, University of Copenhagen.

My project proposal was originally aimed at elucidating the role of Ophiocordyceps fungal symbionts in leaf‐cutting ants, however I gradually became less and less convinced that these fungi actually had a role to be elucidated and we decided to make the scope of the thesis broader, including a variety of different microbial symbionts of ants. I am thus very grateful to be able to present a thesis including work on three exciting systems of microbial symbioses with ants.

The thesis is comprised of a synopsis of the current understanding of symbiotic interactions, which provides the theoretical framework for the following four chapters of original empirical works, prepared for publication. A short concluding section aims at putting the obtained results into a broader perspective.

Sandra Breum Andersen

3

4 Table of Contents

SUMMARIES...... 7 SYNOPSIS...... 11 INTRODUCTION TO SYMBIOTIC INTERACTIONS ...... 13 Evolution of interactions –cooperation and conflict ...... 13 Adaptations to symbiotic life...... 15 Complexity of multi-species interactions...... 17 THESIS OBJECTIVES...... 20 The social insect as hosts and symbiotic partners...... 20 The model systems ...... 22 Techniques and fieldwork...... 23 Chapter outlines ...... 25 REFERENCES...... 26 CHAPTER 1...... 31 DECONSTRUCTING A DISEASE-DEFENCE SYMBIOSIS: SPECIFICITY AND STABILITY OF ACROMYRMEX-PSEUDONOCARDIA ASSOCIATIONS IN CHANGING ENVIRONMENTS ...... 31 CHAPTER 2...... 65 DYNAMIC WOLBACHIA PREVALENCE IN ACROMYRMEX LEAF-CUTTER ANTS: POTENTIAL FOR A NUTRITIONAL SYMBIOSIS ...... 65 CHAPTER 3 ...... 97 DISEASE DYNAMICS IN A SPECIALIZED PARASITE OF ANT SOCIETIES ...... 97 CHAPTER 4...... 127 ...... 127 HOST SPECIFICITY OF PARASITE MANIPULATION –ZOMBIE ANT DEATH LOCATION IN THAILAND VS. BRAZIL...... 127 CONCLUSIONS AND PERSPECTIVES ...... 137 PICTURES ...... 145 ACKNOWLEDGEMENTS ...... 147 CURRICULUM VITAE...... 149

5 6 SUMMARIES

ENGLISH DANSK

7 8 Summary

The life history of social insects, with division of labour, cooperative brood care and overlapping generations, affects the strategies of the associated symbionts. A high density of related individuals could be an open invitation to recurrent disease epidemics, but the complementary layers of social and individual immunity efficiently protect the society. In this thesis the interaction between ants and three different microbial symbionts are dealt with, covering the spectrum from parasitism to mutualism and something in between.

Two chapters focus on the leaf‐cutting ants native to South and Central America that are (in)famous for their ability to defoliate vegetation surrounding their colonies, making them the dominant herbivores of the region. The leaf‐cutter ants are the most advanced of the fungus‐growing ants and this is what makes them capable of living on leaves: in underground chambers the ants farm fungus that degrade the substrate the ants bring it, in return letting the ants feed on the fungus. The association between the ants and the fungus are the corner stones in an intriguing multi‐trophic interaction involving also other fungi and bacteria. One of these bacterial partners called Pseudonocardia grow on the cuticle of some leaf‐cutter ants, visible to the naked eye as a white patch on the ants’ chest, and used by the ants as antibiotic factories employed against a parasite of the fungus garden. The diversity of the bacteria on the ants has been the subject of some controversy. We found a low diversity with only one strain of bacteria dominating on the ants. By comparing ants collected in the field and in the lab from the same colonies we show that this association is highly stable, even after 10 years in the lab and exposure to many other bacteria. Another bacterial partner is of the genus Wolbachia and these are live inside the ants’ tissues. Wolbachia are found associated with a wide range of insects and typically as a reproductive parasite, yet what they may do in the leaf‐ cutter ants is not well understood. Our work show that the ants are found in great numbers in sterile workers and surprisingly also extracellularly in the gut, suggesting a new potential role in the ant’s nutritional system. In contrast to these likely helpful bacteria carpenter ants in tropical regions of the world are attacked by parasitic fungi of the genus Ophiocordyceps. When infected, the ants are manipulated into leaving their colony and die in ‘graveyards’, biting under leaves. The last two chapters of the thesis deals with the disease pressure of this parasite experienced by the ant colonies, which is found to be surprisingly low, and how the social structure of the host apparently has shaped the life‐strategy of the parasite into iteroparity. In addition, differences in the manipulation was found between species in Thailand and Brazil, likely reflecting variation in host behaviour and environmental parameters.

Together, the four chapters highlights different ways in which the symbionts of ants have adapted to the social structure of the host.

9 Resumé

De sociale insekters livshistorie, med opdeling af arbejdsopgaver, fælles pleje af afkom og overlappende generationer, påvirker strategierne hos de forskellige symbionter der associerer med kolonien. En høj tæthed af nært beslægtede individer kan synes at være en åben invitation til gentagne epidemiske sygdomsudbrud, men de kompletterende lag af hhv. individuel og social immunitet beskytter effektivt kolonien. Denne afhandling ser på hvordan tre forskellige symbionter interagerer med myrer, dækkende hele spektret fra mutualisme til parasitisme og noget ind i mellem.

To kapitler fokuserer på bladskærermyrerne fra Syd og Central Amerika, som er berømte og berygtede for deres evner udi defoliering af vegetationen omkring deres kolonier, med en sådan effektivitet at de er de dominerende herbivorer i regionen. Bladskærermyrerne er de højst udviklede af de svampedyrkende myrer, og heri ligger forklaringen på hvordan de kan leve af blade: I underjordiske kamre dyrker myrerne en svamp og denne nedbryder bladene for myrerne der til gengæld for lov at spise af svampen. Mutualismen mellem myrerne og svampen er grundlaget for en avanceret fler‐laget symbiose der involverer flere forskellige svampe og bakterier. Én af disse bakterielle partnere kaldet Pseudonocardia gror på myrernes exoskelet og kan ses med det blotte øje som en hvid plet på myrernes bryst. Bakterierne bruges af myrerne som mobile antibiotikafabrikker som kan aktiveres hvis en parasitisk svamp invaderer deres svampehave. Diversiteten af disse bakterier har længe været noget kontroversiel. Vores studium fandt en lav diversitet hvor kun én slags bakterie dominerer. Ved at sammenligne myrer der var indsamlet i felten og i laboratoriet fra de samme kolonier viser vi at associationen er meget stabil, selv efter op til 10 år i laboratoriet omgivet af kolonier med andre bakterier har myrerne de samme som i felten. En anden slags bakterier af slægten Wolbachia lever inde i myrernes væv. Wolbachia findes i et bredt udvalg af insekter, oftest som reproduktive parasitter, men hvordan de påvirker bladskærer myrerne er ikke klart. Vores studium viser at bakterierne findes i vævet i højt antal i de ellers sterile arbejder myrer og overraskende nok også ekstra cellulært i dele af tarmen. Dette kunne tyde på at bakterierne spiller en rolle i myrernes fordøjelses system. I modsætning til de tilsyneladende hjælpsomme bakterier angribes ’tømrer’ myrer i verdens tropiske egne af parasitiske svampe af slægten Ophiocordyceps. Inficerede myrer manipuleres til at forlade deres koloni og dø i ’kirkegårde’ fastbidt i blade. De sidste to kapitler i afhandlingen fokuserer på hvilken effekt svampen har på myrekoloniens tilstand, hvilken konkluderes at være overraskede lav, og hvordan myreværtens sociale struktur har påvirket svampen livsstrategi i retning af iteroparitet. Dertil kommer en sammenligning af selve myremanipulationen mellem arter fra Brasilien og Thailand, hvor en forskel blev fundet, formentlig afhængig af forskelle i myrernes adfærd og miljømæssige variabler.

Tilsammen afdækker de fire kapitler forskellige aspekter af symbionters tilpasning til en social myre vært.

10 SYNOPSIS

11 12 INTRODUCTION TO SYMBIOTIC INTERACTIONS The term symbiosis is defined as a close association of two different species over an extended period of time, and covers a continuous spectrum from parasitism to mutualism, as popularized in 1879 by the German scientist De Bary (Sapp 1994). The importance of symbioses cannot be overestimated from an ecological, evolutionary or economical perspective. The inter‐species cooperation of mutualisms represents major evolutionary transitions, allowing some associations to achieve ecological dominance. Prime examples are the role of mitochondria in the evolution of the eukaryotic cell following bacterial endosymbiosis (Margulis 1993; Gray et al. 2001), the mutualism between corals and zooxanthellae algae as the foundation of the productive and diverse tropical reefs (Muscatine 1990), and the leafcutter ants farming fungus, the primary herbivores in South‐ and Central America (Schultz & Brady 2008). Similarly, the majority of plants are dependent on mycorrhizal fungi (Smith & Read 2008), and efficient insect pollination is crucial to many types of plants, including human crops (Losey & Vaughan 2006). The impact of parasites is no less impressive, with a cautious estimate of 20‐50% of all extant species employing this life strategy (Poulin & Morand 2000). Contrary to mutualists, parasites decrease the fitness of their host, making them capable of controlling host population densities, and thereby indirectly increasing biodiversity and species coexistence (Hudson et al. 2006). However when targeting agriculture and humans directly parasites are a major cost and burden, as illustrated by the estimated 247 million annual cases of malaria induced fever, caused by the mosquito vectored parasite, resulting in 881.000 human deaths a year (WHO world malaria report 2008, www.who.int/malaria).

Evolution of interactions –cooperation and conflict The symbiosis spectrum thus ranges from conflict to cooperation between host and symbiont, involving the exploitation of resources and services, such as protection or transport. The cost‐benefit ratio for each partner will determine the outcome on the parasitism‐mutualism scale, but quantifying this ratio is difficult as the currencies and exchange rates may be far from obvious (Herre et

13 al. 1999). This is further complicated by its context dependent nature, as e.g. the value of a given symbiont‐provided resource is dependent on its availability in the environment. A useful framework to understand these dynamics, and predict the outcome of an interaction, is the separation of cooperation and conflict into two parameters instead of simply viewing them as opposites (Queller & Strassmann 2009). Intuitively, when cooperation is high and conflict low a mutualistic relationship is expected while the opposite results in a parasitism. There may however be great differences between symbioses in the levels of both conflict and cooperation. Both the yucca‐yucca moth and the fig tree‐fig wasp interactions are highly evolved pollination mutualisms, yet the experienced conflict is expected to be lower in the latter, as the transfer of host pollen dependent on mature wasp offspring align host‐symbiont interests to a greater degree. Also, a study of populations of Polistes wasps differing in levels of predation and social parasitism showed that while conflict between hosts and parasites were always high, cooperation by nest defence between them in high predation areas lowered the overall cost of parasitism for the host (Lorenzi & Thompson 2011). Separating cooperation and conflict also allows for a better understanding of specific adaptations of the interactors. A detailed comparative study of related species of the above mentioned fig wasp‐fig tree mutualisms for example revealed different degrees of conflict across the systems, by finding a correlation between the presence of wasp cheating and host sanctions with the evolution of active, in stead of passive, pollination by the wasps (Jandér & Herre 2010). This framework may thus make it easier to explain how and why we e.g. sometimes see high levels of cooperation in spite of extended conflict (Queller & Strassmann 2009). In addition, the emerging field of synthetic mutualisms, which employs genetically engineered microbes, provides a novel way of studying the evolution of mutualisms by varying the costs and benefits (reviewed by Xavier 2011).

An important factor in determining the outcome of an interaction is the mode of symbiont transmission (Bull 1994). If vertically transmitted, the interests of the host and symbiont are aligned to some extent, as symbiont propagation depends

14 on the host reproducing successfully. This is well represented by the Attine ant symbiosis, where founding queens vertically transmit the fungal cultivar, and the ants restrict horizontal transmission by suppressing fungal sporulation (Mueller et al. 2004). In contrast, horizontal symbiont transmission may create a conflict of interests with the host, making a parasitic relationship more likely. Ophiocordyceps fungi that manipulate their ant hosts are extreme examples of this, as the host not only needs to die in order for the parasite to reproduce but also at the right location (Andersen et al. 2009). However, transmission mode does not always predict the outcome, as illustrated by the mutualism between termites and their horizontally transmitted fungal cultivar (Aanen et al. 2009) and the vertically transmitted Wolbachia bacteria that parasites many insect taxa (Werren et al. 2008).

Even in mutualistic relationships where the two partners achieve a net benefit from the interaction, costs are paid and getting more for less will always be a desirable strategy (Herre et al. 1999). Explaining the stability of mutualisms is thus considered one of the current major challenges of evolutionary biology. Conditions mediating such stability are suggested to be 1) partner fidelity ensuring consistency in the association over evolutionary time and limiting interactions to genetically similar symbionts (Herre et al. 1999), 2) active partner choice (Sachs et al. 2004) and 3) sanctions against non‐cooperatives (Kiers et al. 2003).

Adaptations to symbiotic life When previously free‐living species come to interact closely in a symbiosis, it will have evolutionary consequences at multiple levels. Interactions can range from loose and facultative to intimate and obligate, with close coevolution expected among partners partaking in obligate associations. The nature of the interaction has long been thought to be important for the type of selection experienced, where parasitism should tend to cause negative frequency‐ dependent selection of host‐symbiont genotypes, favouring diversity and recombination, while the opposite should serve to stabilize mutualisms by

15 maintaining traits beneficial to the interaction (Sachs et al. 2011). Accumulating molecular data however show frequent exceptions to this pattern, with apparent similar rates of evolution in parasitic and mutualistic species and e.g. homologous genes used for establishing host interactions between related mutualistic and parasitic species (Sachs et al. 2011). The similarities between mutualistic and parasitic relationships were also highlighted in a recent study of bacterial endosymbionts of weevil beetles, showing how the growth and spread of the mutualistic bacteria were strictly controlled by host immune genes also employed as part of the general immune defence (Login et al. 2011). Ultimately symbiotic relationships can cause gene loss, if the interaction provides a predictable level of services that leaves the host’s or symbiont’s own provision of these superfluous (Moran 2007; McCutcheon & Moran 2012). This is true for both parasitic and mutualistic relationships. Examples include the extensive reductions in genome size and occurrence of pseudogenes in parasitic, intracellular bacteria like Rickettsia and Mycobaterium, that appear to depend on the host’s metabolism for essential biosynthesis pathways (Andersson & Andersson 1999). The Russian doll‐like mutualisms between a bacterium within a bacterium within a mealybug is another example of genome reduction; neither bacterial endosymbiont harbours an intact pathway for the biosynthesis of essential amino acids and are thus completely dependent on each other and potentially also the host for completion of the necessary steps in the production of these (McCutcheon & von Dohlen 2011).

In mutualistic relationships the interests of the partners are to a large extent aligned. Often observed adaptations are the evolution of housing structures in the host for the symbionts, such as mycangia in bark beetles (Six & Klepzig 2004) or bacterial pouches in ants (Billen & Buschinger 2000). Evolution of novel behaviours are also common, such as weeding and grooming of fungal gardens in leafcutter ants (Currie & Stuart 2001). Above the species level, the mutualistic union of organisms with complementary traits may create a compound organism with a novel, more complex phenotype for selection to work on (Moran 2007; Zilber‐Rosenberg & Rosenberg 2008). As such, symbiosis can be a highly efficient way to acquire novel metabolic capabilities and allow for new niches to be

16 exploited and potentially dominated. Lichens illustrate this well, as the symbiotic union of fungi and cyanobacteria have allowed them to thrive in habitats with extreme conditions where few other organisms can survive (e.g. Ruibal et al. 2009). In parasitic relationships the partners’ interests have opposite direction; while the host attempt to avoid infection the parasite wants to exploit its resources. This may lead to an arms race, as described by the ‘Red Queen hypothesis’ (Van Valen 1973). The host and parasite co‐evolve, by respectively improving defence mechanisms versus increasing infectivity and transmissibility, but the net outcome is effectively the same. An interesting consequence of parasite adaptation may be host manipulation (Thomas et al. 2005), which will be dealt with in more detail later (Chapter 3 & 4).

Complexity of multi-species interactions While much of the early literature on symbioses has focused on interactions between two different species in isolation, it is becoming increasingly clear that this is a practical yet unfortunate oversimplification. The environment in which the interaction takes place is comprised of both abiotic parameters, such as temperature and precipitation, and the other species living there, e.g. providing food or predation or additional symbiotic contributions. These parameters may vary across a species range, potentially creating different selective pressures on the symbiotic interaction (Thompson 2005; Thompson 2010), as in the above‐ mentioned example of social parasites of wasps cooperatively defending the nest of their host in the presence of predators (Lorenzi & Thompson 2011). More studies are emerging that take this diversity into account. An impressive long‐ term study by Palmer et al. (2010) on the ant‐acacia symbiosis highlights the importance of the synergistic effects of a multi‐species interaction. Acacia trees associate with different ant species over their lifetime, and the associations with a given ant species has previously been characterized on a scale from parasitism to mutualism. However, by following the assemblage of symbionts during host ontogeny a more nuanced and surprising result was found. For example, associating with – what was considered to be – a castrating parasite at an early, premature, life stage was more beneficial than being without any ants at all. The

17 succession of symbionts over the entire lifetime of a long‐lived host thus revealed dynamics not detectable by short‐term observations, a phenomenon likely to be relevant in other systems as well. Also in clearly parasitic relationships the composition of the symbionts matters. The two main types of interactions are at the level of exploitation of host resources and circumvention of the host immune system. Multiple parasite species may either compete for the same limited resource, potentially increasing virulence, or facilitate each other by having complementary needs. Likewise can infection by one parasite impede the co‐infection of another by upregulating the host immune response, or facilitate multiple infections by weakening the host (Pedersen & Fenton 2007). Such dynamics were observed in a study of four different parasites in natural populations of voles (Telfer et al. 2010). Strong correlations were found between the different parasites’ prevalence, where some facilitated infection by another species, whereas others lowered host susceptibility to a specific parasite, resulting in an intricate web of parasite interactions.

Not only will a symbiotic association often be affected by the actions of other species, diversity of the symbionts below species level may also be of great importance. A host can associate with different genotypes or strains of symbionts on a temporal and spatial scale. In mutualisms it is believed to be in the interest of the host, and/or symbiont, to keep symbiont diversity low, or at least keep different genotypes separated in time or space. This will theoretically allow for kin selection to increase cooperation among symbionts and limit unproductive competition. In the attine ant system, the fungal cultivar is kept as a monoculture (at the fungus level, not mentioning the various microbes in the garden (Pinto‐Tomás et al. 2009)) and the fungus actively competes with unrelated fungi by suppressing their growth and affecting the ants’ abilities to utilize them (Poulsen & Boomsma 2005). When a host is infected by multiple species or strains it is suggested that competition among them likely will increase virulence (Bull 1994), as e.g. found in double infections by the trematode Schistosoma mansonii in snails (Davies et al. 2002). This need not be the case however, as e.g. observed by Massey et al.

18 (2004) in double bacterial infections of caterpillars. Here bacterial species interacting in the host lowers the overall virulence by killing each other with bacteriocins, without harming the host. The two species are capable of coexisting in the host because of a spatially structured environment, creating distinct niches for each species. Applying social evolution theory to such microbial interactions gives very interesting results and predictions, by linking the necessity for cooperation among bacteria to the production of many virulence factors. Cooperative individuals are prone to exploitation by so‐called cheaters, that free‐load on the public goods made available by the cooperative effort. Artificial introduction of such cheaters into a population of pathogenic bacteria may thus lower virulence (Brown et al. 2009; Rumbaugh et al. 2009). With improved molecular tools, such as next generation sequencing, we are only beginning to appreciate the true with‐in host parasite diversity. Using this technique an extended parasite diversity within malaria‐infected humans was thus discovered recently, with unknown implications for the disease development (Juliano et al. 2010).

When studying symbioses it is thus of outmost importance to recognize the different types of interactions, i.e. at the level of the symbiont (inter‐ and intraspecific interactions), the host‐symbiont association, and among hosts. Add to this the influence of the abiotic environment. This potentially leaves us with the ungrateful job of attempting to disentangle highly complex systems, but only by appreciating these dynamics may the true nature of symbiotic associations be understood.

19 THESIS OBJECTIVES

The social insect as hosts and symbiotic partners The focus of this thesis is symbioses between ants and microbes. Ants belong to the social insects, a group dominated by the social Hymenoptera and the termites, and defined by group living with overlapping generations, cooperative brood care by relatives and division of labour with reproductive and non‐ reproductive individuals (Crozier & Pamilo 1996). The ants and the termites in particular have succeeded in creating enormous and long‐lived societies playing a large functional role in their given ecosystems, with the ants estimated to make up 15‐20% of the terrestrial animal biomass (Schultz 2000). The stable, clean and protected environment of the colony makes it an attractive host, however it also provides a challenge affecting the strategies of the associated symbionts. The high density of related individuals within the colony makes it particularly prone to parasite attack, was it not for the extended social immune system that complements the individual immunity. The social immunity of the colony consists of several lines of defence, firstly avoiding encounters with parasites, and secondly limiting parasite intake, establishment, and spread within the colony. This is achieved through behaviours, such as grooming, compartmentalization of the nest and labour division (Cremer et al. 2007). In addition it has been suggested that the features of the society as a host will select for less virulent parasites, but also less beneficial mutualists (Boomsma et al. 2005; Hughes et al. 2008). The homeostatic colony can be viewed as analogous to a tree, long‐lived and modular, with individual workers being dispensable as the leaves of the tree. Thus, while some diseases may be detrimental to the individual ant only few will kill off an entire colony. On the other hand, the mutualistic associates may be less efficient, e.g. by attracting their own parasites and maintaining selfish interests.

The list of known ant symbionts is extensive and covers the whole spectrum from parasites to mutualists with obligate to facultative associations (Kronauer & Pierce 2011). Of the most conspicuous interactions is the tending of hemipterans for sugar exudates (Styrsky & Eubanks 2007), the mutualistic and

20 parasitic relationships with caterpillars (Pierce et al. 2002) and the fungus farming of Attine ants (Schultz & Brady 2008). In addition, there exists numerous ant‐plant symbioses (e.g. Oliveira & Freitas 2004) such as the above‐mentioned interactions with Acacia trees (Palmer et al. 2010) and the creation of monoculture ‘devil gardens’, where ants protect the resources of their host plants by the application of herbicides against non‐host invaders (Frederickson et al. 2005). Extensive molecular screenings is also revealing a great diversity of microbial symbionts (reviewed by Zientz et al. 2005). Among the first to be discovered where Blochmannia bacteria in Camponotus ants, which most likely play a nutritional role. Subsequently a range of other bacterial species has been found to be nutritional mutualists of ants, thereby allowing them to survive on nutritionally unbalanced diets (Cook & Davidson 2006). The function of other bacterial groups is less well understood such as the newly discovered Entomoplasmatales bacteria in army ants (Funaro et al. 2011) and the intracellular bacteria Wolbachia that are, in addition to being common in insects in general, found in a broad range of ant species (Wenseleers et al. 1998; Russell et al. 2009; Russell 2012).

The overall objective of this thesis is to elucidate the implications of heterogeneity on the interactions between a social host and its microbial symbionts. Variation in inherited traits is one of the required cornerstones of evolution, and in symbiotic systems variation may occur between hosts and among symbionts within and between hosts. In addition, environmental heterogeneity influence the outcome of an interaction and for symbionts the environment is comprised of the host and in some cases the external environment. In chapter 1 and 2 the focus will be on heterogeneity in the shape of symbiont diversity with‐in and between hosts. If multiple symbiont strains inhabit the same host there is the potential for both conflict and cooperation among strains, provided that the symbionts overlap temporally and spatially. Such interactions at the symbiont level will likely affect the nature of the host‐ symbiont relationship. In contrast, chapter 3 and 4 deal with how heterogeneity in and between hosts affects symbiont strategy. The behaviour of the ant host, with well‐protected brood and efficient grooming, in addition to temporal and

21 spatial structure in the ants movement, results in a heterogeneous host population where only a proportion of the colony members are available for certain symbionts. In addition to the focus on these different aspects of host and symbiont heterogeneity, the systems studied also span the parasitism‐mutualism continuum, allowing the exploration of how different symbionts deals with the challenges of a social ant host.

The model systems Chapter 1 and 2 in this thesis concern the monophyletic group of fungus farming Attine ants, which are found in South and Central America and extending into the Southern US. The symbiosis between the Attine ants and their fungal cultivar is to a large degree obligate (always for the ants but only for the fungal symbiont of higher attines), and believed to have originated more than 50 mill years ago. Today more than 230 fungus growing ant species are recognized, divided into five major agricultural systems. The phylogeny of the Attine ants corroborate with the substrate utilized by the fungal cultivar, from lower attines that use dry organic matter to the leaf‐cutting ants of Atta and Acromyrmex, the only genera relying solely on fresh plant material (Schultz & Brady 2008). The fungal monoculture farmed by the ants is however challenged by parasites, in particular the fungus Escovopsis (Currie et al. 1999a). The majority of the Attine ants use antibiotic secretions from bacteria growing on their cuticle to protect their crop (Fernandez‐Marin et al. 2009). These bacteria were first believed to be waxy exudates of the ants’ cuticle but were subsequently identified as actinomycete bacteria (Currie et al. 1999b). They are housed in special structures on the cuticle and apparently supported by gland secretions by the ant (Currie et al. 2006). The bacterial diversity and colony specificity of the cuticular bacteria of Acromyrmex echinatior is the focus of chapter 1. In addition to the complex diversity of symbionts mentioned above, the Acromyrmex leafcutter ants harbour a variety of Wolbachia endosymbiotic bacteria (Van Borm et al. 2002; Van Borm et al. 2003; Frost et al. 2010). While Wolbachia are often found to be reproductive parasites, their effects in ants are not well understood and may likely vary between species (Wenseleers et al. 1998; Wenseleers et al. 2002).

22 Chapter 2 focuses on the dynamics of multiple strain infections of Wolbachia in Acromyrmex octospinosus.

In contrast to the likely mutualistic associates of leafcutter ants covered in chapter 1 and 2, chapter 3 and 4 deals with a lethal parasite, the fungus Ophiocordyceps camponoti­rufipedis that infects and manipulate the behaviour of Camponotus rufipes ants in Brazil. Ants infected with Ophiocordyceps fungi exhibit a remarkable extended phenotype of the parasite in that they are manipulated to behave in concordance with the interests of the fungus. Infected ants will thus leave their colony and bite onto vegetation to allow for parasite development and transmission. The phenomenon arised more than 48 million years ago, as documented by bitemarks on fossil leaves (Hughes et al. 2011). These Ophiocordyceps parasites have been recognized for a long time since descriptions by Wallace in 1859 (referred in Hughes et al 2011) and also form the basis of myths by indigenous Amazonian people. The occurrence of ‘graveyards’ with particularly high densities of dead infected ants was first described from Brazil (Evans & Samson 1982) and more recently Thailand (Pontoppidan et al. 2009). In chapter 3 the consequences of such ant graveyards in regards to the experienced disease pressure on the ant colony and the trade‐ offs between parasite transmission and survival is investigated, while chapter 4 compares the parasitc manipulation in two related host‐parasite systems.

Techniques and fieldwork I have used a variety of techniques to obtain the results presented in this thesis. In chapter 1 I studied the bacterial diversity on ant cuticles by 454 sequencing in the lab of Dr. Lars Hestbjerg KU. The main challenge was obtaining an ideal extraction of bacterial DNA, representing the natural diversity but avoiding ‘contamination’ by intracellular bacteria. I attempted blotting the bacteria of the ant cuticle with moist cotton but limited material obtained and the risk of cotton contamination made this unpractical. In stead, whole cuticle plates were dissected from the ants and attached tissue carefully removed. The sequencing

23 was performed by Karin Vestberg and analyzed with technical assistance by Sanne Nygaard. In chapter 2, I used real‐time quantitative PCR (RT‐qPCR) to measure the density of bacterial symbionts across ant life‐stages by targeting a highly variable gene and standardizing for host cell number by amplification of a single‐ copy host gene. RT‐qPCR is a powerful technique for accurately quantifying gene copy number, and was performed in the lab of Prof. Cornelis Grimmelikhuijzen KU with the advice of Dr. Tom Gilbert and Michael Williamson. To visualize the bacteria inside the ants I used histology and fluorescence in situ hybridization (FISH), which was challenging because of the hard cuticle of the adult ants. Moreover, the resin usually used for embedding ants was not compatible with FISH but another material proved useful. Embedding and sectioning was done in the lab of Aase Jespersen with assistance from Lisbeth Haugkrogh. FISH is typically performed with short oligonucleotide probes that, in the case of bacteria, target the gene 16S because of its high copy number in the cell. I however wished to visualize the location of two bacterial strains that did not differ in their 16S sequence. Using in vitro transcription I thus generated fluorescently tagged RNA probes that targeted a single‐copy variable gene (RING‐FISH, Zwirglmaier et al. 2004), but unfortunately the probes did not work successfully. 16S targeted FISH was performed at the veterinary lab of the Technical University of Copenhagen of Mette Boye with advice from Mette Boye, Marianne Rasmussen, Joanna Amenuvor and Annie Ravn Pedersen.

During my PhD I have been fortunate to do fieldwork in some extraordinary places. Field collection of Acromyrmex leaf‐cutter ants for chapter 1 & 2 were done in Gamboa, Panama in collaboration with the Smithsonian Tropical Research Institute in 2010. Collection of Camponotus rufipes ants infected with Ophiocordyceps camponoti­rufipedis for chapter 3 & 4 took place in Mata do Paraíso, Brazil in collaboration with the Federal University of Viçosa in 2011 with visiting professor Harry Evans and co‐supervisor Dr. David Hughes. Subsequent data analyses were done at Penn State University USA, in collaboration with Dr. Matt Ferrari, hosted by David Hughes.

24 Chapter outlines Chapter 1: The majority of attine ants harbour actinomycete bacteria in crypts on their cuticle. The antibiotic production of these is believed to be the main defence of the ants against the fungus Escovopsis, a specialized parasite of their crop. The first analyses suggested that there was only one strain of bacteria within each colony. This view has been challenged by other studies. We addressed the question of bacterial community diversity and host specificity by 454 sequencing analyses of cuticular bacteria of ants kept in the lab for up to 10 years, and samples from the same colonies collected in the field over 17 years. We find that the cuticular diversity is dominated by Pseudonocardia bacteria with only one strain pr. colony and two strains in the population. Chapter 2: Wolbachia bacteria are found as intracellular symbionts in many insects, often as reproductive parasites but sometimes as mutualists. Many ants also harbour Wolbachia, but the consequences of these infections are not well understood. We used quantitative PCR and fluorescence in situ hybridization to study the dynamics of Wolbachia infections across different life stages in workers of Acromyrmex octospinosus leaf‐cutter ants, which harbour multiple Wolbachia strains. The non‐reproducing workers of Acromyrmex were found to contain high densities of Wolbachia, and our data suggest that the different Wolbachia strains interact, competing in the immature stages. We hypothesize that this is because the different strains occupy the same tissues early in host life, while they specialize on different tissue types in the adult workers. The presence of large amounts of extracellular bacteria in the crop of the gut and in the fecal droplets suggests that Wolbachia in Acromyrmex potentially function as a nutritional mutualist. Chapter 3: Fungal ant parasites of the genus Ophiocordyceps manipulate host behaviour to ensure that the host die in an appropriate location for parasite growth and reproduction. The existence of graveyards with a high density of dead ants suggests that the parasite is highly virulent. We measured the effective parasite pressure at the ant colony level by studying parasite life‐stage distribution within graveyards. This data was used to model and explore the life‐ history trade‐offs experienced by the parasite in the challenge of targeting a social host. We conclude that few parasites are in an infective stage. The well‐ defended host requires long‐term persistence in the environment of the parasite, resulting in an iteroparous strategy, but the correspondingly slow development attracts an array of hyperparasites, resulting in low parasite pressure.

Chapter 4: Parasite manipulation of host behaviour is expected to be more or less fine‐tuned depending on the systems characteristics. We compare parasite extended phenotypes in two systems of Ophiocordceps fungi infecting Camponotus ants. Infected ants in Thailand have been observed to die within a narrow spatial range and experimental manipulation suggested that this is where parasite growth is optimal, even though it comes at a trade‐off with transmission of propagules. We compare this with newly obtained data from a related system in Brazil, where dead infected ants are found in a wider spatial range. We suggest that this is the result of both host and environmental differences between the systems.

25 REFERENCES

Aanen D.K., de Fine Licht H.H., Debets A.J.M., Kerstes N.A.G., Hoekstra R.F. & Boomsma J.J. (2009). High symbiont relatedness stabilizes mutualistic cooperation in fungus‐growing termites. Science, 326, 1103‐1106. Andersen S.B., Gerritsma S., Yusah K.M., Mayntz D., Hywel‐Jones N.L., Billen J., Boomsma J.J. & Hughes D.P. (2009). The life of a dead ant: the expression of an adaptive extended phenotype. The American Naturalist, 174, 424‐ 433. Andersson J.O. & Andersson S.G.E. (1999). Insights into the evolutionary process of genome degradation. Curr Opin Genet Dev, 9, 664‐671. Billen J. & Buschinger A. (2000). Morphology and ultrastructure of a specialized bacterial pouch in the digestive tract of Tetraponera ants (Formicidae, Pseudomyrmecinae). Arthropod Structure & Development, 29, 259‐266. Boomsma J.J., Schmid‐Hempel P. & Hughes W.O.H. (2005). Life histories and parasite pressure across the major groups of social insects. In: Insect evolutionary ecology (eds. Fellowes MDE, Holloway GJ & J. R). Royal Entonomological Society. Brown S.P., West S.A., Diggle S.P. & Griffin A.S. (2009). Social evolution in micro‐ organisms and a Trojan horse approach to medical intervention strategies. Philos T R Soc B, 364, 3157‐3168. Bull J.J. (1994). Perspective ‐ Virulence. Evolution, 48, 1423‐1437. Cook S.C. & Davidson D.W. (2006). Nutritional and functional biology of exudate‐ feeding ants. Entomologia Experimentalis et Applicata, 118, 1‐10. Cremer S., Armitage S.A.O. & Schmid‐Hempel P. (2007). Social immunity. Current Biology, 17, R693‐R702. Crozier R.H. & Pamilo P. (1996). Evolution of social insect colonies. Sex allocation and kin selection. Oxford University Press. Currie C.C. & 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., Mueller U.G. & Malloch D. (1999a). The agricultural pathology of ant fungus gardens. PNAS, 96, 7998‐8002. 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., Scott J.A., Summerbell R.C. & Malloch D. (1999b). Fungus‐growing ants use antibiotic‐producing bacteria to control garden parasites. Nature, 398, 701‐704. Davies C.M., Fairbrother E. & Webster J.P. (2002). Mixed strain schistosome infections of snails and the evolution of parasite virulence. Parasitology, 124, 31‐38. Evans H.C. & Samson R.A. (1982). Cordyceps species and their anamorphs pathogenic on ants (Formicidae) in tropical forest ecosystems I. The Cephalotes (Myrmycinae) complex. British Mycological Society, 79, 431‐ 453. Fernandez‐Marin H., Zimmerman J.K., Nash D.R., Boomsma J.J. & Wcislo W.T. (2009). Reduced biological control and enhanced chemical pest

26 management in the evolution of fungus farming in ants. Proceedings of the Royal Society B: Biological Sciences, 276, 2263‐2269. Frederickson M.E., Greene M.J. & Gordon D.M. (2005). 'Devil's gardens' bedevilled by ants. Nature, 437, 495‐496. Frost C.L., Fernández‐Marín H., Smith J.E. & Hughes W.O.H. (2010). Multiple gains and losses of Wolbachia symbionts across a tribe of fungus‐growing ants. Mol Ecol, 19, 4077‐4085. Funaro C.F., Kronauer D.J.C., Moreau C.S., Goldman‐Huertas B., Pierce N.E. & Russell J.A. (2011). Army ants harbor a host‐specific clade of Entomoplasmatales bacteria. Appl Environ Microb, 77, 346‐350. Gray M.W., Burger G. & Lang B.F. (2001). The origin and early evolution of mitochondria. Genome Biology, 2, 1018.1‐1018.5. Herre E.A., Knowlton N., Mueller U.G. & Rehner S.A. (1999). The evolution of mutualisms: exploring the paths between conflict and cooperation. Trends in Ecology and Evolution, 14, 49‐53. Hudson P.J., Dobson A.P. & Lafferty K.D. (2006). Is a healthy ecosystem one that is rich in parasites? Trends Ecol Evol, 21, 381‐385. Hughes D.P., Pierce N.E. & Boomsma J.J. (2008). Social insect symbionts: evolution in homeostatic fortresses. Trends Ecol Evol, 23, 672‐677. Hughes D.P., Wappler T. & Labandeira C.C. (2011). Ancient death‐grip leaf scars reveal ant‐fungal parasitism. Biology Letters, 7, 67‐70. Jandér K.C. & Herre E.A. (2010). Host sanctions and pollinator cheating in the fig tree‐fig wasp mutualism. Proceedings of the Royal Society B: Biological Sciences, 277, 1481‐1488. Juliano J.J., Porter K., Mwapasa V., Sem R., Rogers W.O., Ariey F., Wongsrichanalai C., Read A. & Meshnick S.R. (2010). Exposing malaria in‐host diversity and estimating population diversity by capture‐recapture using massively parallel pyrosequencing. Proceedings of the National Academy of Sciences. Kiers E.T., Rousseau R.A., West S.A. & Denison R.F. (2003). Host sanctions and the legume‐rhizobium mutualism. Nature, 425, 78‐81. Kronauer D.J.C. & Pierce N.E. (2011). Myrmecophiles. Current Biology, 21, R208‐ R209. Login F.H., Balmand S., Vallier A., Vincent‐Monégat C., Vigneron A., Weiss‐Gayet M., Rochat D. & Heddi A. (2011). Antimicrobial peptides keep insect endosymbionts under control. Science, 334, 362‐365. Lorenzi M.C. & Thompson J.N. (2011). The geographic structure of selection on a coevolving interaction between social parasitic wasps and their hosts hampers social evolution. Evolution, no‐no. Losey J.E. & Vaughan M. (2006). The economic value of ecological services provided by insects. BioScience, 56, 311‐323. Margulis L. (1993). Symbiosis in cell evolution. Freeman. Massey R.C., Buckling A. & Ffrench‐Constant R. (2004). Interference competition and parasite virulence. P Roy Soc Lond B Bio, 271, 785‐788. McCutcheon J.P. & Moran N.A. (2012). Extreme genome reduction in symbiotic bacteria. Nat Rev Micro, 10, 13‐26. McCutcheon J.P. & von Dohlen C.D. (2011). An interdependent metabolic patchwork in the nested symbiosis of mealybugs. Current biology : CB. Moran N.A. (2007). Symbiosis as an adaptive process and source of phenotypic complexity. P Natl Acad Sci USA, 104, 8627‐8633.

27 Mueller U.G., Poulin J. & Adams R.M.M. (2004). Symbiont choice in a fungus‐ growing ant (Attini, Formicidae). Behav. Ecol., 15, 357‐364. Muscatine L. (1990). The role of symbiotic algae in carbon and energy flux in reef corals. Coral Reefs, 25, 75‐87. Oliveira P.S. & Freitas A.V.L. (2004). Ant‐plant‐herbivore interactions in the neotropical cerrado savanna. Naturwissenschaften, 91, 557‐570. Palmer T.M., Doak D.F., Stanton M.L., Bronstein J.L., Kiers E.T., Young T.P., Goheen J.R. & Pringle R.M. (2010). Synergy of multiple partners, including freeloaders, increases host fitness in a multispecies mutualism. P Natl Acad Sci USA, 107, 17234‐17239. Pedersen A.B. & Fenton A. (2007). Emphasizing the ecology in parasite community ecology. Trends Ecol Evol, 22, 133‐139. Pierce N.E., Braby M.F., Heath A., Lohman D.J., Mathew J., Rand D.B. & Travassos M.A. (2002). The ecology and evolution of ant association in the Lycaenidae (Lepidoptera). Annu Rev Entomol, 47, 733‐771. Pinto‐Tomás A.A., Anderson M.A., Suen G., Stevenson D.M., Chu F.S.T., Cleland W.W., Weimer P.J. & Currie C.R. (2009). Symbiotic nitrogen fixation in the fungus gardens of leaf‐cutter ants. Science, 326, 1120‐1123. Pontoppidan M.‐B., Himaman W., Hywel‐Jones N.L., Boomsma J.J. & Hughes D.P. (2009). Graveyards on the move: the spatio‐temporal distribution of dead Ophiocordyceps‐infected ants. PLoS ONE, 4, e4835. Poulin R. & Morand S. (2000). The Diversity of Parasites. The Quarterly Review of Biology, 75, 277‐293. Poulsen M. & Boomsma J.J. (2005). Mutualistic fungi control crop diversity in fungus‐growing ants. Science, 307, 741‐744. Queller D.C. & Strassmann J.E. (2009). Beyond society: the evolution of organismality. Philos T R Soc B, 364, 3143‐3155. Ruibal C., Gueidan C., Selbmann L., Gorbushina A.A., Crous P.W., Groenewald J.Z., Muggia L., Grube M., Isola D., Schoch C.L., Staley J.T., Lutzoni F. & de Hoog G.S. (2009). Phylogeny of rock‐inhabiting fungi related to Dothideomycetes. Stud Mycol, 64, 123‐1337. Rumbaugh K.P., Diggle S.P., Watters C.M., Ross‐Gillespie A., Griffin A.S. & West S.A. (2009). Quorum sensing and the social evolution of bacterial virulence. Current Biology, 19, 341‐345. Russell J.A. (2012). The ants (Hymenoptera: Formicidae) are unique and enigmatic hosts of prevalent Wolbachia (Alphaproteobacteria) symbionts. Myrmecological News, 16, 7‐23. Russell J.A., Goldman‐Huertas B., Moreau C.S., Baldo L., Stahlhut J.K., Werren J.H. & Pierce N.E. (2009). Specialization and geographic isolation among Wolbachia symbionts from ants and Lycaenid butterflies. Evolution, 63, 624‐640. Sachs J.L., Essenberg C.J. & Turcotte M.M. (2011). New paradigms for the evolution of beneficial infections. Trends Ecol Evol, 26, 202‐209. Sachs J.L., Mueller U.G., Wilcox T.P. & Bull J.J. (2004). The Evolution of Cooperation. The Quarterly Review of Biology, 79, 135‐160. Sapp J. (1994). A history of symbiosis. Oxford University Press. Schultz T.R. (2000). In search of ant ancestors. Proceedings of the National Academy of Sciences, 97, 14028‐14029.

28 Schultz T.R. & Brady S.G. (2008). Major evolutionary transitions in ant agriculture. Proceedings of the National Academy of Sciences, 105, 5435‐ 5440. Six D.L. & Klepzig K.D. (2004). Dendroctonus bark beetles as model systems for studies on symbiosis. Symbiosis, 37, 207‐232. Smith S.E. & Read D.J. (2008). Mycorrhizal symbiosis. 3 edn. Academic Press. Styrsky J.D. & Eubanks M.D. (2007). Ecological consequences of interactions between ants and honeydew‐producing insects. P Roy Soc B­Biol Sci, 274, 151‐164. Telfer S., Lambin X., Birtles R., Beldomenico P., Burthe S., Paterson S. & Begon M. (2010). Species interactions in a parasite community drive infection risk in a wildlife population. Science, 330, 243‐246. Thomas F., Adamo S. & Moore J. (2005). Parasitic manipulation: where are we and where should we go? Behavioural Processes, 68, 185‐199. Thompson J. (2010). Four central points about coevolution. Evolution: Education and Outreach, 3, 7‐13. Thompson J.N. (2005). The Geographic Mosaic of Coevolution. University of Chicago Press. Van Borm S., Billen J. & Boomsma J.J. (2002). The diversity of microorganisms associated with Acromyrmex leafcutter ants. BMC Evol Biol, 2, 9. Van Borm S., Wenseleers T., Billen J. & Boomsma J.J. (2003). Cloning and sequencing of wsp encoding gene fragments reveals a diversity of co‐ infecting Wolbachia strains in Acromyrmex leafcutter ants. Molecular Phylogenetics and Evolution, 26, 102‐109. Van Valen L. (1973). A new evolutionary law. Evolutionary theory, 1, 1‐30. Wenseleers T., Ito F., Van Borm S., Huybrechts R., Volckaert F. & Billen J. (1998). Widespread occurrence of the micro‐organism Wolbachia in ants. Proceedings of the Royal Society of London Series B: Biological Sciences, 265, 1447‐1452. Wenseleers T., Sundström L. & Billen J. (2002). Deleterious Wolbachia in the ant Formica truncorum. Proceedings of the Royal Society of London. Series B: Biological Sciences, 269, 623‐629. Werren J.H., Baldo L. & Clark M.E. (2008). Wolbachia: master manipulators of invertebrate biology. Nature reviews Microbiology, 6, 741‐751. Xavier J.B. (2011). Social interaction in synthetic and natural microbial communities. Mol Syst Biol, 7. Zientz E., Feldhaar H., Stoll S. & Gross R. (2005). Insights into the microbial world associated with ants. Arch Microbiol, 184, 199‐206. Zilber‐Rosenberg I. & Rosenberg E. (2008). Role of microorganisms in the evolution of animals and plants: the hologenome theory of evolution. FEMS Microbiology Reviews, 32, 723‐735. Zwirglmaier K., Ludwig W. & Schleifer K.H. (2004). Recognition of individual genes in a single bacterial cell by fluorescence in situ hybridization ‐RING‐ FISH. Molecular Microbiology, 51, 89‐96.

29 30

CHAPTER 1

DECONSTRUCTING A DISEASE- DEFENCE SYMBIOSIS: SPECIFICITY AND STABILITY OF

ACROMYRMEX-

PSEUDONOCARDIA ASSOCIATIONS IN CHANGING ENVIRONMENTS

31 32

Sandra B. Andersen1, Lars H. Hansen2 and Jacobus J. Boomsma1

1Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark

2Molecular Microbial Evolution, Department of Biology, University of Copenhagen, Sølvgade 83H, 1307 Copenhagen, Denmark

Corresponding authors: Sandra B. Andersen ([email protected]) and Jacobus J. Boomsma ([email protected])

Keywords: attine ant mutualism, Pseudonocardia, bacterial community, 454 pyrosequencing

33 Abstract

Fungus‐growing (attine) ants live in a complex multi‐trophic symbiosis that involves both fungal and bacterial partners. Among these are Actinobacteria of the genus Pseudonocardia that are maintained on the ant cuticle to produce antibiotics, primarily against a parasitic fungus of their garden symbiont. This association has been assumed to be a hallmark of evolutionary stability, but this notion has recently been challenged by culturing and sequencing data indicating that the actinobacterial cultures can be relatively diverse and variable. We used 454 pyrosequencing of a region of the 16S rDNA gene to estimate the diversity of the bacterial community on the cuticle of the leaf‐cutting ant Acromyrmex echinatior and some other sympatric fungus‐growing ants from Gamboa, Panama. Cuticular bacterial cultures tend to be concentrated on the ventral side of the ant thorax, so we restricted our sampling to include only the laterocervical plates and pronotum with abundant bacterial cover. We used both field and lab samples of the same colonies, the latter after colonies had been kept under laboratory conditions for up to 10 years. We show that the cuticular bacterial communities are highly colony‐specific and stable over time. Colonies always had a single dominant Pseudonocardia strain and only two such strains were found in the Gamboa population across 17 years of sampling, confirming an earlier study using elongation factor 1α (EF-1α). A number of other Actinobacteria were identified at low densities in some samples, but no consistent patterns were observed. We suggest several directions in which future studies may shed light on the interaction‐specificity of this symbiosis, which apparently can be very different across species and allopatric populations of fungus‐growing ants.

34 Introduction

Questions of conflict and cooperation are fundamental for understanding the evolutionary stability of genomes, societies and interspecific mutualisms (Bourke, 2011; Burt and Trivers, 2006; Herre et al., 1999). This is particularly apparent when considering multi‐species symbioses consisting of a mixture of mutualistic and parasitic partners with potentially diverging fitness interests, such as microbial gut‐communities (Qin et al., 2010), nitrogen fixing bacteria of legumes (Kiers et al., 2003) and microparasites of wild voles (Telfer et al., 2010). The prevailing view of the last decade has been that parasitism and mutualism are opposites of a continuum of reciprocal exploitation (Herre et al., 1999) as phylogenies often show transitions in both directions (Sachs and Simms, 2006) and even local populations of the same metapopulation may represent examples of net win‐win and win‐loose interactions (Hochberg et al., 2000; Thompson, 1999). While there is now considerable consensus about the evolutionary phenomenology of the parasite‐mutualist spectrum of symbioses, the debate on the relative importance of genetic and phenotypic mechanisms for maintaining evolutionary stability of mutualistic interactions is ongoing. Mutualisms that involve life‐time commitments between a single host and symbiont strain tend to have efficient competitive exclusion mechanisms to maintain symbionts in monoculture (Aanen et al., 2009; Poulsen and Boomsma, 2005), so that symbionts can remain maximally cooperative (Frank, 1996). However, a larger range of mutualisms have multiple symbiont strains and continue to acquire new ones throughout life. Such dynamics may either be an unavoidable liability for the host or a potential asset in allowing a more flexible community composition of symbionts. Two main mechanisms have been suggested to secure long term evolutionary stability of these interactions: host screening of symbionts for their performance before they are admitted (Archetti, 2010; Weyl et al., 2010), and host sanctions against underperforming symbionts (Kiers and Denison, 2008; Kiers et al., 2003). These mechanisms differ in relative importance between mutualisms, and they make different predictions as to what kind of host adaptations for controlling symbiont diversity we should expect to find.

35 The symbiosis between fungus‐growing (attine) ants and their microbial symbiont community is an intriguing example of a complex mutualism involving multiple partners. The ants farm basidiomycete fungi in underground nest chambers, and are completely dependent on this life‐style, which evolved ca. 50 million years (Schultz and Brady, 2008). The cultivar is reared as a monoculture, partly controlled by antagonistic behaviour of the ants and resident fungus towards unrelated strains (Bot et al., 2001; Poulsen and Boomsma, 2005; Ivens et al., 2008). In contrast to these effective measures to prevent competition with related, but genetically different symbionts, the fungus garden is relatively vulnerable to infections by the specialized parasitic ascomycete fungus Escovopsis (Currie et al., 1999a). To meet these challenges, the ants employ a range of behavioural (Currie and Stuart, 2001), chemical (Fernández‐Marín et al., 2009) and biological (Currie et al., 1999b) control measures. The latter are often achieved by the use of antibiotic compounds from actinomycete bacteria, housed in specialized structures on the ant cuticle (Currie et al., 2006). Similar to the symbiosis between the ants and their crop fungus, the association with the cuticular actinomycetes has also been thought to represent an ancient co‐evolved mutualism characterized by a single strain of bacteria in each ant colony (Cafaro and Currie, 2005; Poulsen et al., 2005), vertical transmission with newly eclosed workers and virgin queens obtaining the bacteria from their sisters and the fungus garden (Poulsen et al., 2003b), and coevolution between the ant hosts and the bacteria (Cafaro et al. 2005). However, controversy has arisen in recent years over the extent to which: 1. ants and bacteria are sufficiently faithful to each other to make coevolution likely, 2. the actinobacterial antibiotics are specifically targeted towards Escovopsis and 3. the growth‐form of a colony’s bacteria is indeed monocultural (Barke et al., 2010; Cafaro et al., 2011; Mueller et al., 2008; Sen et al., 2009). These studies showed that multiple actinomycete bacteria can be isolated from single fungus‐ growing ants, but also that specific Pseudonocardia lineages are consistently associated with genera of attine ants in spite of having free‐living close relatives (Cafaro et al., 2011). The initial studies of the diversity of the cuticular bacteria of attine ants were done by culturing and sequencing isolates, which produced monocultures

36 of what was ultimately identified as Pseudonocardia bacteria (Cafaro and Currie, 2005; Currie et al., 2003). This approach was criticized for its use of selective media, limiting the growth of other bacterial species that might also have been present (Mueller et al., 2008). Later culturing studies indeed found a greater diversity, including other actinomycetes such as Streptomyces and Amycolatopsis (Barke et al., 2010; Haeder et al., 2009; Kost et al., 2007; Sen et al., 2009). However, the major overall limitation of culturing methods to infer diversity is that only a sub‐sample of the true diversity may be able to grow in the chosen agar‐plate conditions, and that no measure of the relative abundance of cuticular bacteria under natural conditions is obtained. These issues can be addressed with next generation sequencing, allowing for a potentially unbiased picture of true diversity and species richness of bacterial symbionts. The first surveys with 454 sequencing included three fungus‐growing ant genera, Trachymyrmex (4 ants from one lab nest), Cyphomyrmex (4 ants from one lab nest) and Mycocepurus (4 ants from 2 lab nests) and showed that the ants carried multiple Pseudonocardia species in addition to a wide range of other bacteria (Sen et al., 2009). However, a result like this is not necessarily surprising when extracts of whole ants are sequenced, so that also bacteria living in or passing through the gut become included. As actinomycetes and other bacteria occur in great diversity in soil and plant material, it may be that such surveys have limited relevance for addressing the specificity and function of cuticular bacteria in attine ants. Another potential problem is that ants that have been kept in the laboratory may have secondarily acquired bacteria that they would not associate with in the field. In the present study we revisit the questions of diversity, specificity and stability of ant cuticular bacterial communities (point 1 and 3 above), focusing on the leaf‐cutting ant Acromyrmex echinatior that was not included in the study by Sen et al. (2009). This species has been relatively well studied for its phenotypic associations with actinomycete bacteria, which revealed that the growth pattern of the bacteria on the cuticle of large workers of these ants is very predictable: Following eclosure the laterocervical plates (the ventral thorax area) are quickly colonized to produce a characteristic bacterial bloom and entire ants tend to become almost completely covered over the two following

37 weeks (Poulsen et al., 2003a). After this peak density, the bacterial bloom gradually disappears, until again only the laterocervical plates are covered ca. a month later (Poulsen et al., 2003a). This location appears to be particularly adapted to harbouring the bacterial growth because it is speckled with cuticular tubercles that each are supplied with secretions of tiny subcuticular glands (Currie et al., 2006). While detailed morphological adaptations such as tubercles to feed actinomycetes are consistent with a long history of bacterial domestication and coevolution (Currie et al. 2006), this does not necessarily imply that there has been strict co‐cladogenesis. On one hand, the open ‘external’ location of the cuticular crypts should make it relatively easy for environmental bacteria to invade, which could lead to considerable symbiont diversity, as e.g. in the zooxanthellae of corals (Knowlton & Rohwer, 2003) and rhizobial bacteria of legumes (Kiers et al., 2008). On the other hand, the ants should remain under strong selection to make their glandular secretions so specific that they preferentially enhance the growth of bacterial cultures that produce useful antibiotics, be it Pseudonocardia or other lineages (Boomsma and Aanen, 2009). To advance our understanding of the true nature and diversity of the bacterial cultures on the cuticle of advanced leaf‐cutting ants, we set out to obtain a precise culture‐independent estimate of the bacterial diversity on the laterocervical plates and adjacent parts of the pronotum of A. echinatior and to compare these estimates across field and lab samples of the same colonies.

Materials & Methods

Ant sampling To assess the cuticular bacterial diversity, large worker ants of two age categories were sampled from lab colonies of Acromyrmex echinatior. First, the laterocervical plates and pronotum of callow nurse workers were dissected. These ants had relatively pale cuticles that were completely covered in bacteria as is typical for large workers of A. echinatior ca. 2 weeks after eclosing (bacterial cover scales 10‐12, Poulsen et al., 2003a). The attached soft tissue was carefully

38 removed from the internal sides of these cuticular fragments to minimize the presence of Wolbachia endosymbiotic bacteria that are abundant in the thoracic muscles (Andersen et al., submitted). From the same colonies and two additional ones, we also sampled an older ant that only had visible bacterial growth on the laterocervical plates. These were ants with darker cuticles, representing scales 1‐ 3 (Poulsen et al., 2003a), i.e. the final stage of bacterial cover that is typical for foragers. The first set of samples will be referred to as category C (callow) samples (n = 17, one ant each from 17 different colonies) and the second as category M (mature) samples (n = 19; Table 1). The two categories were chosen to assess whether bacterial diversity changes with ant age, i.e. whether bacterial diversity would be higher on callow workers that are entirely covered in bacteria compared to mature workers or the other way around. In addition, the collection of multiple samples from 17 colonies (and 19 colonies overall), allowed us to ask whether there are within and between colony differences in bacterial community composition. All ants were sampled from within the fungus garden of colonies that had been collected in Gamboa, Panama between 2001 and 2011 and subsequently kept in four culture rooms in Copenhagen at ca. 25 ˚C and 70% relative humidity, each containing multiple colonies of different attine ants. Throughout their laboratory ‘tenure’, colonies received the same locally collected bramble leaves, fruit fragments and dry rice. For 11 of the sampled lab colonies field collected samples of mature workers were available, stored in 96% ethanol at ‐20 °C in tubes containing multiple workers from the same colony. One mature worker from each colony was oven‐dried and the laterocervical plates dissected. Freezer samples from another six field‐collected ants from A. echinatior colonies analysed in the study by Poulsen et al. (2005) were also included, to represent the two clusters of Pseudonocardia strains identified in that early study where the elongation factor 1α (EF-1α) gene was sequenced. These samples are referred to as category F (field; Table 1). In addition to the A. echinatior ants, six lab samples representing four other species of attine ants from the same study site, sharing their culture rooms with some of the sampled Acromyrmex colonies, were collected. These six samples contained the laterocervical plates of two samples of Trachymymex zeteki (three individuals pooled to compensate for body masses being smaller

39 than in A. echinatior), two samples of Cyphomyrmex costatus (five individuals pooled in each sample), C. longiscapus (three individuals pooled) and Acromyrmex volcanus (Table 1).

DNA extraction The dissected cuticular fragments were placed individually in sterile 2 ml screw lid tubes with 0.1 mm glass beads (MO BIO laboratories, Inc.) and DNA extracted with a MasterPure DNA purification kit (Epicentre Technologies), which extracts Gram‐negative and Gram‐positive bacteria with about equal efficiency (Rantakokko‐Jalava and Jalava, 2002). In short, 300 µl tissue lysis buffer was added and the bacterial membranes disrupted in a FastPrep machine for 45 s. at 4.5 speed. Three µl of Proteinase K (Invitrogen) was added followed by >25 min incubation at 65 °C with frequent vortexing. The samples were cooled and precipitated according to the manufacturer’s instructions and the DNA eluted in 35 µl TE buffer.

Amplification of 16S rDNA by PCR and tag­encoded FLX 454 pyrosequencing Bacterial DNA was amplified with the general bacterial primers 341F/806R spanning the hypervariable region V3 (Masoud et al. 2011). PCR was performed in a final volume of 20 µl with 4 µl 5x Phusion HF buffer, 0.4 µl 10 mM dNTP mixture, 0.2 µl Phusion Hot Start DNA Polymerase (Finnzymes), 1 µl of each primer (10 µM), 1 µl 10x diluted template and water at the conditions: 98 ˚C for 30s, followed by 35 cycles of 98 ˚C for 5 s, 56 ˚C for 20 s and 72 ˚C for 20 s, and a final extension at 72 ˚C for 5 min. The samples were moved directly to ice and run on a 1% agarose gel containing EtBr for 50 min. The specific bands were cut and purified from the gel using the Montage DNA gel extraction kit (Millipore). To each sample A and B adaptors for emPCR and pyrosequencing were added together with a sample‐unique tag in an additional PCR. This procedure was performed as above except that the forward primer was replaced with 59 differently tagged forward primers and with only 15 cycles in the PCR. The first 36 of these samples were provided with an A adaptor, LinA_341F_1 – 36, and the reverse primer with a B adaptor LinB_806R, whereas the last 23 samples received a B adaptor, LinB_341F_58‐80 and the reverse primer with an A

40 adaptor LinA_806R (Masoud et al. 2011). The PCR product was run on a gel and purified as described above, after which the DNA concentration was quantified using a Quant‐iT dsDNA High‐Sensivity Assay Kit and a Qubit fluorometer (Invitrogen). Amplicons were mixed to ensure an equal representation of each sample and two one‐region 454 sequencing runs were performed on a GS FLX Titanium PicoTiterPlate (70X75) using a GS FLX Titanium Sequencing Kit XLR70 according to the manufacturer's instructions (Roche). The A and B tagged samples were prepared for two separate 454 sequencing runs, where the first included all the lab samples of A. echinatior and the second run all the field samples and the other species of attine ants.

Data analyses The data of the two runs were analysed using the QIIME pipeline (http://qiime.sourceforge.net/index.html#, Caporaso et al., 2010). The raw sequence data of the two runs were trimmed separately using the default settings (minimum average quality score = 25, minimum/maximum length 200/1000 bp, removal of forward primer) and sorted by sample ID. The resulting FASTA files were compiled and operational taxonomic units (OTUs) were picked with the default settings using the ‘uclust’ algorithm based on 97% similarity, and a representative sequence for each OTU was selected using the ‘first’ algorithm. The identified OTUs were aligned with PyNAST, after which names were assigned to OTUs with the default RDP classifier. This allowed for an OTU table and OTU heatmap to be constructed. In addition, the alpha‐diversity for each sample was computed and rarefaction curves of the diversity index Chao‐1 (Chao 1984) and the observed species number, as a function of simulated sequencing effort, generated in Qiime with the default settings. Wolbachia is a prevalent endosymbiont of Acromyrmex ants (Frost et al., 2010; Van Borm et al., 2001; Andersen et al., submitted) and is not expected to be relevant for cuticular bacterial diversity, so all OTUs classified as Wolbachia were removed from the data set. As different depths of sequencing were achieved across the samples, the percent prevalence of a given OTU in each sample was calculated to minimize bias towards proportionally rare OTUs being represented by many sequences in some deeply sequenced samples. These proportions were

41 used to focus the further analyses on biologically relevant bacteria, by conservatively narrowing the OTU collection to OTUs with > 5% prevalence in at least one sample. The identified Pseudonocardia OTUs were manually validated against five high‐quality sequences of different Pseudonocardia species representing the diversity of the genus with Genbank accession numbers EF114314, AJ252833, AJ252827, AJ249206 and AJ252822 (Mueller et al., 2010) in Sequencher 4.7. All gaps and ambiguous base pairs not found in these five sequences were removed from the OTUs.

Statistical analyses Bacterial diversity was further analysed in JMP 9.0.2 for Mac OSX. Community composition was assessed by principal component analysis, using the first principal component as the ordering variable in a subsequent two‐way hierarchical clustering with Ward’s minimum variance method. The diversity within each sample after the exclusion of Wolbachia OTUs was estimated by the Simpson’s index 1‐D calculated as 1 ‐ ∑(n/N)2 with n being the frequency of a given OTU in a sample and N the total frequency of the OTUs with a prevalence > 5% in the sample. The diversity index was compared between samples with a Two‐Way ANOVA with Tukey‐Kramer HSD post hoc testing.

Results

Data quality, read distribution and Pseudonocardia diversity In the first run with 36 samples 266520 sequences out of 446202 passed the quality controls. The distribution of reads per sample was highly skewed, as one sample contributed 20.5% of the sequences (sample Ae.480M, Fig.1A). The remaining samples contributed on average 6052 ± 2635 SD sequences. In the second run with 23 samples 202375 sequences out of 504794 passed the quality control with a more even distribution of 8796 ± 3093 SD sequences per sample (Fig. 1A). Some of the sequences contained a high proportion of Wolbachia OTUs, which were removed to focus on the cuticular bacterial diversity (Fig. 1A, red bars). Rarefaction curves showed that a deeper sequencing likely would have

42 revealed more rare species (Fig. 1B, including Wolbachia OTUs), but that the data provide a reliable picture of the total diversity as estimated by the Chao‐1 index, as they include all common OTUs (Fig. 1C, including Wolbachia OTUs). After removal of the Wolbachia sequences the 2678 identified OTUs amounted to 2491. When including only the OTUs that contributed > 5% of the total sequences in at least one sample the diversity was narrowed down to 35 OTUs, comprising on average 84 ± 6% SD of the sequences of each sample. A total of eight Pseudonocardia OTUs dominated the bacterial communities, with a mean overall prevalence of 70% of the total sequences in the A. echinatior samples. As the diversity of this genus was the focus of the study, these eight OTUs were validated against high‐quality Pseudonocardia sequences to identify potential errors introduced in the sequencing or subsequent analysis. The majority of errors were found in highly conserved regions and validation revealed that the eight Pseudonocardia OTUs identified in the analysis in reality only represented four different Pseudonocardia strains, two found on Acromyrmex and Cyphomyrex ants, one on A. volcanus and one on C. longiscapus. On A. echinatior, the two strains were generally not found to co‐occur on individual ants or in single colonies. A BLAST search suggested that also the two Amycolatopsis OTUs found on the Trachymyrmex ants only represented one actual strain. The identical OTUs were thus collapsed and the distribution recalculated.

Other bacteria Three samples stood out by containing a large proportion of Archaea bacteria (26%, Ae.153M), Cyanobacteria/chloroplast DNA (41%, Ae.480M) and Enterobacteriaceae (Ae.342F), which likely reflects contamination of an unknown origin. Some of the samples contained Pseudomonas OTUs, but these were only identified in field samples and lab samples of lower attines from the second 454 run, suggesting that they may represent contaminants from either the DNA extraction or a PCR step. In addition to the removed Wolbachia bacteria, also some other OTUs were suspected to be of ant soft‐tissue origin. There were two Rhizobiales OTUs reaching frequencies of 35 and 53% in one of the T.zeteki and the C. longiscapus

43 samples, respectively, with a 98‐99% sequence similarity to bacteria previously identified as gut symbionts of various ant species (e.g. GenBank acc.no. FJ477647; Russell et al. 2009), three other Rhizobiales OTUs with closest matches to environmental samples, and an Entomoplasmataceae OTU that closely matched bacteria found in association with ant guts (93‐94% similarity, e.g. GenBank acc.no. HM996870; Funaro et al., 2011; see also Russell et al., 2009; Stoll et al., 2007). In addition, OTUs from the Burkholderiales and Xanthomonadales were identified, orders which also have been found previously in association with ants (Russell et al. 2009). These OTUs were only identified in a minority of the samples and may result from contamination from the oesophagus during dissection. However, as their origin was not confirmed they were included in the analysis. Other OTUs represented Chitinophaga (Crenotrichaceae, Sphingobacteria) and Actinobacteria in Rubrobacteraceae (not identified to genus level), Aeromicrobium (Nocardioidaceae), Intrasporangiaceae (not identified to genus level), Microbacteriaceaea (not identified to genus level) and Amycolatopsis (Pseudonocardiaceae). None of the Actinobacteria were identified as belonging to the genus Streptomyces (Pseudonocardiaceae), which has previously been found to be associated with attine ants (Haeder et al., 2009; Kost et al., 2007; Barke et al., 2010).

Multivariate analyses of community composition Principal component analysis was performed on the four Pseudonocardia and the single Amycolatopsis OTU. The first principal component, explaining 53.1% of the variation between samples, was saved and used to subsequently order the samples in a hierarchical clustering analysis using Ward’s minimum variance method. The analysis visualizes the clusters of samples with similar bacterial communities in a heatmap (Fig. 2). Five clusters of attine ant samples were identified, two with A. echinatior samples and a single C. costatus sample (cluster 1 and 2), one with Trachymyrmex samples (cluster 3), one with a mixture of A. echinatior, Cyphomyrmex and A. volcanus samples (cluster 4), and one comprised of a single A. echinatior sample (cluster 5, Fig.2). Cluster 1 and 2 were each dominated by a single Pseudonocardia OTU. Cluster 1 was comprised of all samples from 10 different colonies (C, M and/or F) and the field‐collected

44 sample from Ae.263. Sample Ae.153M was, however, placed as its own group next to cluster 1 because of the high proportion of Archaea bacteria. Cluster 2 was comprised of all samples from five different colonies and the lab collected sample(s) from Ae.342 and Ae.356, in addition to Ae.263C, Ae.280C and Ae.480C. The field sample of Ae.342 formed its own cluster (5, Fig.2) as this sample was dominated by Enterobacteriaceae bacteria, likely of soft ant tissue origin. The only other bacteria present were the Pseudonocardia OTU from cluster 2, confirming that both field and lab samples of this colony still harboured the same cuticular community. Sample Ae.263F and Ae.263C clustered in each of the two main clusters, suggesting a complete change in bacterial community from the field to the lab. However, whether the sampled ants actually came from the same colony remains to be checked with microsattelite analysis of the host DNA. The fifth cluster contained a mixture of samples, where Ae.26F, Ae.263M, Ae.280F and Ae.356F were apparently grouped together because of slightly higher levels of suspected contaminants from Rhizobiales and Pseudomonadales, while these samples otherwise clearly belonged in cluster 2. Ae.480M was placed on its own within cluster 4 because of the high proportion of the Cyanobacteria/chloroplast OTU but the Pseudonocardia OTU clearly placed it in cluster 2. The field sample from Ae.480 and seven other samples from cluster 5 stood out by being unusually diverse. Ae.480F had low prevalences for all OTUs, with the 30 most prevalent OTUs only contributing 62% of the sequences of this sample, potentially indicating low sequence quality. The only Pseudonocardia OTU identified was that of cluster 2, placing the sample in the same cluster as the lab samples of the same colony, even though other Actinomycetales bacteria also contributed to the diversity as did some Betaproteobacteria from Burkholderiales. The bacterial communities of the two samples from Ae.406 were more complex as they were comprised of Pseudonocardia primarily belonging to cluster 1 but having also some representatives of cluster 2, in addition to significant amounts of Actinomycetales bacteria from Intrasporangiaceae and Nocardioidaceae respectively. The Nocardiaceae OTU(s), identified as Aeromicrobium, was also found in Ae.24F and Ae.220M, which were otherwise dominated by the

45 Pseudonocardia from cluster 1 and 2 respectively. A slightly higher proportion of potentially non‐cuticular bacteria from Rhizobiales may again partly explain the placement of the C. longiscapus and C. costatus samples in the mixed cluster 4, in addition to various Gammaproteobacteria and Sphingobacteria of unknown function and importance. The most prevalent Pseudonocardia OTU on C. costatus belonged to cluster 2, while the C. longiscapus sample carried another unique Pseudonocardia and the Amycolatopsis from the Trachymyrmex colonies in smaller amounts. The A. volcanus sample was interesting as it harboured the two, otherwise not co‐occuring, main Pseudonocardia OTUs, in addition to an OTU specific to this sample, all in about equal amounts.

Samples from six field colonies previously analysed by Poulsen et al. (2005) were included in the analyses. One of these was assigned to cluster one (Ae.33F), which is likely also where Ae.24F belonged, three were assigned to cluster 2 (Ae.47F, Ae.26F, Ae.112F), while the last (Ae.44F) could not be confidently placed in either of the clusters. The finding of two Pseudonocardia OTUs and the clustering of Ae.33F with Ae.24F, and Ae.47F with Ae.26F and Ae.112F thus replicated the results of Poulsen et al. (2005), a study that only used the EF-1α gene, further corroborating the validity of the two main Pseudonocardia clusters in our Gamboa population.

Phylogenetic placement of the Amycolatopsis and Pseudonocardia OTUs The two sampled colonies of Trachymyrmex ants primarily carried Amycolatopsis Actinobacteria with a 100% identity to various environmental samples, representing 91% and 52% of the respective sequences. The same Amycolatopsis was also found to a minor degree on C. longiscapus (3%). One of the T. zeteki samples contained 37% Rhizobiales bacteria, which are likely contaminants from the ant tissues not properly removed during dissection of these small ants, suggesting that the Amycolaptopsis bacteria dominate on the cuticle. It was not possible to compare the entire Amycolatopsis sequence to those obtained by Sen et al. (2009) from Mycocephorus attine ants, as a different region of the 16S rDNA was targeted. However a 166 bp overlap between the OTU and GenBank acc.no. FJ948128 showed only 93.4% similarity. No Pseudonocardia OTUs were found at

46 a prevalence > 5% in the Trachymyrmex samples. Cafaro et al. (2011) found Pseudonocardia on T. cf. zeteki but it is not clear whether this is the same ant species as that of the present study. The four Pseudonocardia OTUs were tentatively assigned to the six known Pseudonocardia clades associated with attine ants described by Cafaro et al. (2011). The OTU from cluster 1 was identical to Pseudonocardia strains from Cafaro’s clade VI isolated from Acromymex and a few Trachymyrmex species, while the OTU from cluster 2 was most similar (99%) to strains from Cafaro’s clade IV and V isolated from primarily Apterostigma, Trachymyrmex and Acromyrmex ants. The OTU from A. volcanus was also most similar (99%) to strains found in clade IV while the OTU from Cyphomyrmex was most similar (99%) to strains in Cafaro’s clade III isolated from Cyphomyrmex, Trachymyrmex and Mycetarotes. The two dominant Pseudonocardia OTUs from A. echinatior had 98% sequence similarity but the alignment introduced four gaps, making the two OTUs clearly distinguishable.

Comparing diversity between sample types and clusters The bacterial diversity on A. echinatior cuticles as estimated with the Simpson’s index was compared between callow workers from the lab and mature workers from the lab and the field and between the two main clusters by a Two‐Way ANOVA. Only samples that had been placed in either cluster 1 or 2 by the hierarchical clustering analysis (Fig. 1) were included. There was no significant difference in the diversity index between the callow and mature individuals sampled from the same cluster, but the diversity of lab callow workers from cluster 1 was significantly lower than that of mature and field samples from cluster 2 (F5,32 = 5.09, p ≤ 0.05; Table 3). A One‐Way ANOVA with Tukey‐Kramer post‐hoc testing including all A. echinatior samples found an overall lower

diversity in the samples of callow workers (F2,50 = 10.12, p ≤ 0.05). When only looking at the lab collected samples there was no correlation between how long the ants had been kept in the lab (measured as years since collection) and their cuticular bacterial diversity (Lin. reg. cluster 1: R2 = 0.057, p = 0.92; cluster 2: R2 = 0.11, p = 0.17; all samples: R2 = 0.0087, p = 0.59).

47 Of the Acromyrmex colonies from the lab that were assigned to either cluster 1 or 2, nine had been kept in climate rooms primarily with leaf‐cutter ants (room ‘Acro 3’ and room ‘Atta’) and eight together with a variety of lower attines and non‐leafcutter attine ants (room ‘Acro 1’ and room ‘Q’; Table 4). There was no statistically significant correlation between which room the colony had been kept in and which cluster its bacterial

2 community belonged to (Likelihood ratio test, χ 1,17 = 1.55, p = 0.08). A suggestive tendency towards colonies with a cluster 1 community to have shared rearing rooms with only leaf‐cutter ants and colonies with a cluster 2 community to have shared rooms with other attine ants is expected to be merely coincidental, as the colonies were placed in these rooms in no particular order and all retained the bacteria that they had in the field.

Discussion

Bacterial diversity on the ant cuticle Multivariate analyses of the bacterial communities identified two Pseudonocardia OTUs that dominated the cuticular diversity of Acromyrmex echinatior but almost never co‐occured in the same colonies or on the same ants (The exception being the field and lab samples from Ae.263 placed in different clusters and sample Ae.406M and F, see Results; Fig. 2). This confirms the conclusions drawn from studies that employed culturing and direct sequencing of only one Pseudonocardia strain pr. colony (e.g. Poulsen et al., 2005). The association between A. echinatior colonies and their Pseudonocardia strain was apparently very stable, with colonies that had been kept in the lab for up to ten years in each other’s close proximity, retaining their original strains in all cases but one, in spite of ample opportunities for horizontal transmission. The single exception will need to be checked to see whether it indeed were ant samples from the same colony sampled in the field and the lab. The colonies used in the present study had been collected from the same field site over a period of 17 years with both strains being sampled across that period, confirming the stable co‐occurrence of both strains in the Gamboa

48 Acromyrmex population. Analysis of less conserved genes than the 16S rDNA region analyzed here may reveal a greater diversity within and between colonies, but sequencing of the more variable gene EF-1α also found just two Pseudonocardia strains on A. echinatior and its sister species A. octospinosus in the study area (Poulsen et al. 2005), so we consider the results of the two combined studies to be robust. Other Actinomycetales bacteria were found to be inconsistently and rather sparsely associated with the attine ants that we analysed. In four samples (three from the field and one from a lab colony) OTUs from Microbacteriaceae and Nocardiaceae reached densities of 12‐23%. The closest BLAST search matches to these OTUs were different environmental isolates. Relatively low densities of Pseudonocardia bacteria were found in these samples, but whether the greater occurrence of the other actinomycetes is the cause or the effect of a limited Pseudonocardia cover on these respective ants remains to be explored. The possible ecological importance of these and other non‐Pseudonocardia OTUs remains unknown, but co‐occurrence of bacterial species on the ant cuticle has been hypothesized to be facilitated by the sharing of resistance genes against bacteriocins (Barke et al., 2011). No Streptomyces bacteria were observed in any of our lab or field samples. This contrasts with what has been found on A. octospinosus workers from Trinidad and Tobago after culturing washes from field‐samples (Barke et al., 2010), from the same species in French Guiana by rubbing ant cuticles against agar plates (Kost et al., 2007), from fungus gardens of A. echinatior and A. octospinosus, and from to A. volcanus workers from the same site as the samples of our present study, Gamboa in Panama (Haeder et al., 2009). This may reflect geographical differences or that these bacteria were located on other parts of the cuticle (than the laterocervical plates and pronotum), or in the fungus garden as suggested by Haeder et al. (2009). Streptomyces may therefore not be associated with A. echinatior or, if only found at low frequency in the field, have been lost from the cuticle during storage in ethanol. The Streptomyces previously found associated with ants were hypothesized to be regularly acquired from the soil (Barke et al., 2010), suggesting that they may also be lost more easily without the possibility of reacquisition in the lab.

49

Community composition across life­stages and environments No difference was found in the diversity index between mature workers in the lab and field, suggesting that the transition does not drastically affect the diversity of cuticular symbionts. The slightly lower diversity found on the callow workers from cluster 1 is likely the result of a higher proportion of Wolbachia sequences in these samples, as a larger surface of cuticle and thus potentially more soft tissue was included in the samples of callow workers. Fewer sequences overall would thus limit the probability of sequencing rarer OTUs. Horizontal transmission between colonies and species under lab conditions could not clearly be inferred from our present data. A. echinatior colonies with communities from each Pseudonocardia cluster are apparently capable of maintaining their original Pseudonocardia symbiont even when kept in close proximity (see above). The Trachymyrmex species and C. longiscapus carried an entirely different genus of Actinobacteria, Amycolatopsis, in addition to a specific Pseudonocardia OTU in the case of C. longiscapus. There was extensive overlap in the actinomycete communities of A. echinatior and C. costatus but it was not possible to evaluate whether this had also been the case in the field. A. volcanus was interesting in apparently harbouring three different Pseudonocardia OTUs in similar prevalences. A. volcanus has only rarely been encountered at the study site and is typically found nesting in wood cavities that are usually in the canopy. The foraging workers are clearly distinct from those of A. echinatior and A. octospinosus by being almost black and covered in white bacterial growth throughout the mature stage. How and why this growth is sustained remains unclear, but maintaining high degrees of bacterial bloom throughout older workers may well change the bacterial interactions and on the ant cuticle and potentially enable co‐occurrence of otherwise segregating OTUs. The potential for colony‐level adaptations to the cuticular bacterial community was not directly addressed in this study, but our results indicate that this is a possibility. The ca. 50:50 distribution of the two Pseudonocardia strains in the population across time suggests that some form of balancing selection may act on the Pseudonocardia community composition. Cross‐fostering experiments have recently shown that

50 the bacteria grow less well on newly eclosed ants developing in another colony than their own, although the Pseudonocardia of the different colonies were not identified (Armitage et al. 2011). Such an approach in colonies with known bacterial composition could be used in further attempts to elucidate whether and how the ants manage and control their interactions with Pseudonocardia and other actinomycete bacteria.

Conclusions In the present study we exploited the potential of next generation sequencing to obtain precise estimates of the diversity of bacterial communities on the cuticle of Panamanian Acromyrmex leaf‐cutting ants. We show that diversity per colony is relatively low and always revolves around a single dominant strain of Pseudonocardia, consistent with earlier findings by Poulsen et al. (2005). The relatively low diversity of actinomycetes, in comparison with estimates by Sen et al. (2009), Kost et al. (2007) and Barke et al. (2010), may be due to our rather precise targeting of the cuticular region with the most explicit concentration of glands feeding the actinomycete symbionts. However, even after taking substantial care to include only cuticular bacteria it was difficult to avoid contaminations with other bacteria such as thorax muscle Wolbachia and gut bacteria. This suggests that the use of extracts of whole ants as employed by Sen et al. (2009) may not tell us much about symbiotically relevant diversity of bacteria on the cuticle. In addition, we discovered that the slightly higher error rates in pyrosequencing, compared to Sanger sequencing, offered considerable challenges in data analysis, which could only be resolved by manually separating very closely related strains. These manual validations of OTUs produced considerably lower diversity estimates than those obtained from the automatic analyses of the raw sequence data. We conclude, therefore, that the Acromyrmex­ Pseudonocardia symbiosis in our study population combines intriguing characteristics of long‐term interaction‐ specificity with indications that horizontal acquisition of other Actinobacteria may indeed happen at low frequencies. It is thus too early to dismiss that co‐evolutionary interactions may be frequent, even though they will only broadly result in co‐cladogenesis, consistent with findings by Cafaro et al. (2010).

51

Acknowledgements

We would like to thank Karin Vestbjerg for assistance in the lab, Panagiotis Sapountzis for computer assistance, Sanne Nygaard and Martin Asser Hansen for bioinformatic advice and Søren Sørensen for access to the 454 pyrosequencing facilities. SBA was funded by a PhD Scholarship from the Science Faculty of the University of Copenhagen. SBA and JJB were further supported by the Danish National Research Foundation.

52 References

Aanen, D.K., de Fine Licht, H.H., Debets, A.J.M., Kerstes, N.A.G., Hoekstra, R.F., Boomsma, J.J., 2009, High symbiont relatedness stabilizes mutualistic cooperation in fungus‐growing termites. Science 326, 1103‐1106. Archetti, M., 2010, Contract theory for the evolution of cooperation: The right incentives attract the right partners. J Theor Biol 269, 201‐207. Barke, J., Seipke, R., Gruschow, S., Heavens, D., Drou, N., Bibb, M., Goss, R., Yu, D., Hutchings, M., 2010, A mixed community of actinomycetes produce multiple antibiotics for the fungus farming ant Acromyrmex octospinosus. Bmc Biol 8, 109. Barke, J., Seipke, R., Yu, D., Hutchings, M., 2011, A mutualistic microbiome. How do fungus‐growing ants select their antibiotic‐producing bacteria? Communicative & Integrative Biology 4, 41‐43. Boomsma, J.J., Aanen, D.K., 2009, Rethinking crop‐disease management in fungus‐growing ants. P Natl Acad Sci USA 106, 17611‐17612. Bourke, A., 2011, Principles of social evolution. Oxford University Press. Bot A.N.M., Rehner S.A. & Boomsma J.J. (2001). Partial incompatibility between ants and symbiotic fungi in two sympatric species of Acromyrmex leaf‐ cutting ants. Evolution, 55, 1980‐1991. Burt, A., Trivers, R., 2006, Genes in conflict: The biology of selfish genetic elements. Harvard University Press. Cafaro, M.a.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. Cafaro, M.J., Currie, C.R., 2005, Phylogenetic analysis of mutualistic filamentous bacteria associated with fungus‐growing ants. Canadian Journal of Microbiology 51, 441‐446. Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., Fierer, N., Pena, A.G., Goodrich, J.K., Gordon, J.I., Huttley, G.A., Kelley, S.T., Knights, D., Koenig, J.E., Ley, R.E., Lozupone, C.A., McDonald, D., Muegge, B.D., Pirrung, M., Reeder, J., Sevinsky, J.R., Turnbaugh, P.J., Walters, W.A., Widmann, J., Yatsunenko, T., Zaneveld, J., Knight, R., 2010, QIIME allows analysis of high‐throughput community sequencing data. Nat Meth 7, 335‐336. Chao A. (1984). Nonparametric estimation of the number of classes in a population. Scandinavian Journal of Statistics, 11, 265‐270. Currie, C.C., 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., 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., Scott, J.A., Summerbell, R.C., Malloch, D., 1999, Fungus‐growing ants use antibiotic‐producing bacteria to control garden parasites. Nature 398, 701‐704.

53 Currie, C.R., Scott, J.A., Summerbell, R.C., Malloch, D., 2003, corrigendum: Fungus‐ growing ants use antibiotic‐producing bacteria to control garden parasites. Nature 423, 461‐461. Douglas, A.E., 2010, The symbiotic habit. Princeton University Press. 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. Frank, S.A., 1996, Models of parasite virulence. Q Rev Biol 71, 37‐78. Frost, C.L., Fernández‐Marín, H., Smith, J.E., Hughes, W.O.H., 2010, Multiple gains and losses of Wolbachia symbionts across a tribe of fungus‐growing ants. Mol Ecol 19, 4077‐4085. Funaro, C.F., Kronauer, D.J.C., Moreau, C.S., Goldman‐Huertas, B., Pierce, N.E., Russell, J.A., 2011, Army ants harbor a host‐specific clade of Entomoplasmatales bacteria. Appl Environ Microb 77, 346‐350. 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 106, 4742‐4746. Herre, E.A., Knowlton, N., Mueller, U.G., Rehner, S.A., 1999, The evolution of mutualisms: exploring the paths between conflict and cooperation. Trends in Ecology and Evolution 14, 49‐53. Hochberg, Gomulkiewicz, Holt, Thompson, 2000, Weak sinks could cradle mutualistic symbioses – strong sources should harbour parasitic symbioses. Journal of Evolutionary Biology 13, 213‐222. Ivens A.B.F., Nash D.R., Poulsen M. & Boomsma J.J. (2009). Caste‐specific symbiont policing by workers of Acromyrmex fungus‐growing ants. Behavioral Ecology, 20, 378‐384. Kiers, E.T., Denison, R.F., 2008, Sanctions, cooperation, and the stability of plant‐ rhizosphere mutualisms. Annual Review of Ecology, Evolution, and Systematics 39, 215‐236. Kiers, E.T., Rousseau, R.A., West, S.A., Denison, R.F., 2003, Host sanctions and the legume‐rhizobium mutualism. Nature 425, 78‐81. Knowlton N. & Rohwer F. (2003). Multispecies microbial mutualisms on coral reefs: the host as a habitat. The American Naturalist, 162, S51‐S62. Kost, C., Lakatos, T., Böttcher, I., Arendholz, W.‐R., Redenbach, M., Wirth, R., 2007, Non‐specific association between filamentous bacteria and fungus‐ growing ants. Naturwissenschaften 94, 821‐828. Masoud W., Takamiya M., Vogensen F.K., Lillevang S., Al‐Soud W.A., Sørensen S.J. & Jakobsen M. (2011). Characterization of bacterial populations in Danish raw milk cheeses made with different starter cultures by denaturating gradient gel electrophoresis and pyrosequencing. International Dairy Journal, 21, 142‐148. Mueller, U., Ishak, H., Lee, J., Sen, R., Gutell, R., 2010, Placement of attine ant‐ associated Pseudonocardia in a global Pseudonocardia phylogeny (Pseudonocardiaceae, Actinomycetales): a test of two symbiont‐ association models. Antonie van Leeuwenhoek 98, 195‐212.

54 Mueller, U.G., Dash, D., Rabeling, C., Rodrigues, A., Brown, J., 2008, Coevolution between Attine Ants and Actinomycete Bacteria: A Reevaluation. Evolution 62, 2894‐2912. Palmer, T.M., Doak, D.F., Stanton, M.L., Bronstein, J.L., Kiers, E.T., Young, T.P., Goheen, J.R., Pringle, R.M., 2010, Synergy of multiple partners, including freeloaders, increases host fitness in a multispecies mutualism. P Natl Acad Sci USA 107, 17234‐17239. Poulsen, M., Boomsma, J.J., 2005, Mutualistic fungi control crop diversity in fungus‐growing ants. Science 307, 741‐744. Poulsen, M., Bot, A.N.M., Boomsma, J.J., 2003a, The effect of metapleural gland secretion on the growth of a mutualistic bacterium on the cuticle of leaf‐ cutting ants. Naturwissenschaften 90, 406‐409. Poulsen, M., Bot, A.N.M., Currie, C.R., Nielsen, M.G., Boomsma, J.J., 2003b, Within‐ colony transmission and the cost of a mutualistic bacterium in the leaf‐ cutting ant Acromyrmex octospinosus. Functional Ecology 17, 260‐269. 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. Mol Ecol 14, 3597‐3604. Poulsen, M., Erhardt, D.P., Molinaro, D.J., Lin, T.‐L., Currie, C.R., 2007, Antagonistic Bacterial Interactions Help Shape Host‐Symbiont Dynamics within the Fungus‐Growing Ant‐Microbe Mutualism. PLoS ONE 2, e960. Qin, J., Li, R., Raes, J., Arumugam, M., Burgdorf, K.S., Manichanh, C., Nielsen, T., Pons, N., Levenez, F., Yamada, T., Mende, D.R., Li, J., Xu, J., Li, S., Li, D., Cao, J., Wang, B., Liang, H., Zheng, H., Xie, Y., Tap, J., Lepage, P., Bertalan, M., Batto, J.‐M., Hansen, T., Le Paslier, D., Linneberg, A., Nielsen, H.B., Pelletier, E., Renault, P., Sicheritz‐Ponten, T., Turner, K., Zhu, H., Yu, C., Li, S., Jian, M., Zhou, Y., Li, Y., Zhang, X., Li, S., Qin, N., Yang, H., Wang, J., Brunak, S., Dore, J., Guarner, F., Kristiansen, K., Pedersen, O., Parkhill, J., Weissenbach, J., Bork, P., Ehrlich, S.D., Wang, J., 2010, A human gut microbial gene catalogue established by metagenomic sequencing. Nature 464, 59‐65. Rantakokko‐Jalava, K., Jalava, J., 2002, Optimal DNA isolation method for detection of bacteria in clinical specimens by broad‐range PCR. J Clin Microbiol 40, 4211‐4217. Russell, J.A., Moreau, C.S., Goldman‐Huertas, B., Fujiwara, M., Lohman, D.J., Pierce, N.E., 2009, Bacterial gut symbionts are tightly linked with the evolution of herbivory in ants. Proceedings of the National Academy of Sciences 106, 21236‐21241. Sachs, J.L., Simms, E.L., 2006, Pathways to mutualism breakdown. Trends Ecol Evol 21, 585‐592. Schultz, T.R., Brady, S.G., 2008, Major evolutionary transitions in ant agriculture. Proceedings of the National Academy of Sciences 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, e22028. 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, ‐.

55 Stoll, S., Gadau, J., Gross, R.O.Y., Feldhaar, H., 2007, Bacterial microbiota associated with ants of the genus Tetraponera. Biological Journal of the Linnean Society 90, 399‐412. Telfer, S., Lambin, X., Birtles, R., Beldomenico, P., Burthe, S., Paterson, S., Begon, M., 2010, Species interactions in a parasite community drive infection risk in a wildlife population. Science 330, 243‐246. Thompson, J.N., 1999, Specific hypotheses on the geographic mosaic of coevolution. The American Naturalist 153, S1‐S14. Van Borm, S., Wenseleers, T., Billen, J., Boomsma, J.J., 2001, Wolbachia in leafcutter ants: a widespread symbiont that may induce male killing or incompatible matings. Journal of Evolutionary Biology 14, 805‐814. Weyl, E.G., Frederickson, M.E., Yu, D.W., Pierce, N.E., 2010, Economic contract theory tests models of mutualism. Proceedings of the National Academy of Sciences 107, 15712‐15716. Zhang, M.M., Poulsen, M., Currie, C.R., 2007, Symbiont recognition of mutualistic bacteria by Acromyrmex leaf‐cutting ants. ISME J 1, 313‐320.

56 Figure legends

Figure 1 Data obtained by 454 sequencing. (A): Histograms showing the number of sequences obtained for each sample after quality control for all sequences (blue bars) and after Wolbachia OTUs had been removed (red bars). The sampled A. echinatior colonies are arranged along the x‐axis in chronological order of collection, spanning a period of 17 years (1994 for Ae.24F to 2011 for Ae.529), with the other attine ant species added towards the right. The name tags representing the samples from the first run are in full colour while the samples from the second run are in lighter shades. (B, C): Rarefaction curves of sequencing depth (including Wolbachia OTUs). The curves each represent an individual sample and show the observed number of OTUs (B) and the Chao‐1 index of diversity (C) as a function of simulated sequencing effort.

Figure 2 Hierarchical clustering of the bacterial communities on the cuticle of Acromyrmex echinatior and other attine ants. On the x‐axis are the 30 most common OTUs, with the different taxonomic groups highlighted. The dendrograms illustrate clustering in the samples and not phylogenetic relationship between the OTUs. On the y‐axis are the sampled ants, with the frequency of each OTU illustrated by a color scale from blue (= zero to low frequency) over gray to red (= high frequency to complete dominance). The colour of the sample IDs towards the left and the branches of the dendrogram indicate the annotation obtained by hierarchical clustering analysis; the clusters are numbered as marked on the left side of the sample names. Five clusters were identified; cluster 1 and 2 were each dominated by one Pseudonocardia out that was not found in the other cluster (the tentative placement of these two OTUs in the clades of the phylogeny by Cafaro et al. (2011) is marked beneath the cluster number). Cluster 3 contained the Trachymymex samples, which harboured Amycolatopsis bacteria, while cluster 4 contained a mixture of samples grouped together because of a higher non‐Pseudonocardia diversity. Some of this diversity could be explained by the presence of OTUs suspected to reflect contaminations rather than cuticular diversity. For these nine samples the

57 suggested placement in either cluster 1 or 2 is marked by matching the colour of the sample name tag to that of cluster 1 (green) or 2 (blue). Cluster 5 was comprised of a single sample, Ae.342F, with a high prevalence of Enterobacteriaceae; the Pseudonocardia OTU of this sample indicated that it should be placed in cluster 2.

Figure S1 (only accessible via the pdf version of this file) The OTU heatmap showing the distribution of the 2678 OTUs obtained from the ant cuticle samples and the number of sequences that each sample contributed to each OTU.

58

Figures

Figure 1

59 Figure 2

60 Tables

Table 1 Colony ID, sample tag (C = callow workers for which both the laterocervical plates and the pronotum were dissected; M = mature workers for which only the laterocervical plates were dissected; F = mature workers from field colonies), collection year (Gamboa, Panama), and the name of the ant rearing room in which the sampled colony had been kept (Atta, Acro 1, Acro 3, Q). Ae. = A. echinatior, Av. = A. volcanus, Tz = T. zeteki, Cc. = C. costatus, Cl. = C. longiscapus

Colony ID Collection year Categories Name of ant sampled room in the lab Ae.24 1994 F ­ Ae.26 1994 F ­ Ae.33 1996 F ­ Ae.44 1996 F ­ Ae.47 1996 F ­ Ae.112 2000 F ­ Ae.150 2001 C, M, F Atta Ae.153 2001 C, M Atta Ae.160 2002 C, M, F Acro 1 Ae.220 2004 M Acro 3 Ae.263 2004 C, M, F Acro 3 Ae.280 2004 C, M, F Acro 3 Ae.282 2004 C, M, F Acro 1 Ae.322 2006 C, M, F Acro 3 Ae.331 2007 C, M Acro 1 Ae.335 2007 C, M Atta Ae.342 2007 C, M, F Acro 3 Ae.356 2008 C, M, F Acro 3 Ae.406 2009 M, F Acro 3 Ae.420 2009 C, M Acro 3 Ae.480 2010 C, M, F Acro 1 Ae.488 2010 C, M Acro 1 Ae.505 2011 C, M Q Ae.528 2011 C, M, F Q Ae.529 2011 C, M Q Av.520 2011 M Q

Tz.15‐02 2010‐2 2010 3 x M Q

Tz.022‐0509 2009 3 x M Acro 1

Cc.RMMA100603‐04 2004 5 x M Q

Cc.011‐170507 2007 5 x M Acro 1 Cl 3 x M Acro 1

61 Table 2 Phylogenetic distribution of the 35 OTUs that had prevalences of > 5% in at least one sample after exclusion of Wolbachia OTUs. The eight Pseudonocardia OTUs and two Amycolatopsis OTUs were subsequently collapsed into four and one, respectively, following manual validation against reference sequences. For each group, the mean contributions to the 59 samples are given as percentages. The Amycolatopsis OTUs were only identified in Trachymyrmex and C. longiscapus samples.

Phylum Order / Family / Genus Number of Mean % OTUs contribution per sample ± SD Actinobacteria Pseudonocardia 8 66 ± 26% Intrasporangiaceae 2 1 ± 3% Microbacteriaceae 1 2 ± 3 % Nocardioidaceae 3 3 ± 5% Amycolatopsis 2 2 ± 14% Rubrobacteracea 1 0 ± 1% Unknown 1 0 ± 1% Proteobacteria Rhizobiales 5 2 ± 10% Pseudomonadales 2 1 ± 3% Xanthomonadaceae 2 1 ±5% Rhodospirillales 1 0 ± 2% Burkholderiales 1 0 ± 2% Enterobacteriaceae 1 1 ±8% Bacteroidetes Chitinophagaceae 2 0 ± 2% Tenericutes Entomoplasmataceae 1 0 ± 3% Euryarchaeota Methanosaetaceae 1 0 ± 3% (Archaea) Cyanobacteria/ 1 2 ± 6% cloroplast

62 Table 3 Results of a Two‐Way ANOVA with the Simpson diversity index of cuticular bacteria in A. echinatior samples from cluster 1 and 2 as dependent variable and callow workers from the lab, mature workers from the lab, and mature workers from the field as categories.

Dependent Source F d.f. P variable Diversity Cluster (1 or 2) 2.77 2,32 0.106 index 1‐D Ant category (C, M or F) 4.82 1,32 0.0148 Ant category * cluster 3.46 2,32 0.0437

Table 4 The species distribution of attine ants across the culture rooms in which the colonies had been kept for 1‐10 years before their cuticular bacterial communities were sampled.

Species/rearing room Acro 1 Acro 3 Q Atta Only leaf‐cutting ants (3 Atta* 17 21 19 24 species, 4 Acromyrmex species) Higher non‐leaf‐cutting attine ants Ca. 50 0 Ca. 35 0 (2 Sericomyrmex species, 3 Trachymyrmex species) Lower attine ants (4 Cyphomyrmex Ca. 50 Ca. 20 (colony Ca. 20 0 species, 2 Apterostigma species) boxes covered in plastic) * Atta species do not rear actinomycete bacteria on their cuticle (Fernández‐ Marín et al. 2009)

63 64 CHAPTER 2

DYNAMIC WOLBACHIA PREVALENCE IN ACROMYRMEX LEAF- CUTTER ANTS: POTENTIAL FOR A NUTRITIONAL SYMBIOSIS

65 66 Dynamic Wolbachia prevalence in Acromyrmex leaf- cutting ants: potential for a nutritional symbiosis

Sandra B. Andersen1*, Mette Boye2, David R. Nash1 & Jacobus J. Boomsma1

1: Centre for Social Evolution, Department of Biology, University of Copenhagen Universitetsparken 15, DK‐2100 Copenhagen, Denmark

2: National Veterinary Institute, Technical University of Denmark, Bülowsvej 27, DK‐1790 Copenhagen, Denmark

*: Corresponding author: [email protected] / +45 26209197

Running title: Wolbachia in Acromyrmex ants

IN REVIEW

67 Abstract

Wolbachia are renowned as reproductive parasites, but their phenotypic effects in eusocial insects are not well understood. We used a combination of qrt‐PCR, fluorescence in situ hybridisation, and laser scanning confocal microscopy to evaluate the dynamics of Wolbachia infections in the leaf‐cutting ant Acromyrmex octospinosus across developmental stages of sterile workers. We confirm that workers are infected with one or two widespread wsp strains of Wolbachia, show that colony prevalences are always 100%, and characterize two rare recombinant strains. One dominant strain is always present and most abundant while another strain only proliferates in adult workers of some colonies and is barely detectable in larvae and pupae. An explanation may be that Wolbachia strains compete for host resources in immature stages while adult tissues provide substantially more niche space. Tissue‐specific prevalences of the two strains differ, with the rarer strain being overrepresented in the adult foregut and thorax muscles. Both strains occur extracellularly in the foregut, suggesting an unknown mutualistic function in worker ant nutrition. Both strains of bacteria are also abundant in the faecal fluid of the ants, suggesting that they may further have extended functional phenotypes in the fungus garden, which the ants manure with their own faeces.

Keywords: Wolbachia, Acromyrmex ants, symbiosis, gut bacteria, fluorescence in

situ hybridization

68

Introduction

Symbiotic interactions span the entire spectrum between mutualism and parasitism, because the respective costs and benefits for hosts and symbionts ultimately determine whether interactions become “win‐win” or “win‐lose” (Bull, 1994; Herre et al., 1999). Vertical transmission typically aligns the reproductive interests of host and symbiont but this transmission mode is neither necessary nor sufficient to keep a symbiotic interaction mutualistic. For example, Wolbachia is usually a vertically transmitted parasite with relatively high virulence (Werren et al., 2008), while Termitomyces, the garden symbiont of fungus‐growing termites, is a horizontally transmitted mutualist with an unusually stable commitment to its hosts (Aanen et al., 2009). In addition, even vertically transmitted mutualists are not permanently evolutionarily stable, as some are known to have been lost over time (Sachs & Simms, 2006).

Symbioses are increasingly known to involve more than two partners (e.g. Palmer et al., 2010). This further complicates the dynamics and selective forces that shape the ultimate nature of these interactions, because cooperation and conflict in such multiple partnerships depend on the interactions between symbionts in addition to those between host and symbionts (Vautrin & Vavre, 2009; Telfer et al., 2010). Such interactions can either have positive or negative effects on the host, but typically require that symbionts have spatially and temporally overlapping niches within hosts.

Communities of bacterial symbionts with complementary roles may produce stable mutualisms when confined to specific host organs or tissues. Examples are the gut pouches of Tetraponera ants that contain multiple highly divergent species of nitrogen fixing bacteria (Van Borm et al., 2002) and the bacteriomes of hemipteran sharpshooters (Homalodisca coagulata) that contain two bacterial species supplying amino acids and vitamins to the host (Wu et al., 2006). However, when it comes to genetic variation among symbionts with similar roles, diversity may be costly for hosts, as within‐host competition often selects

69 for more virulent parasites (Frank, 1996; Davies et al., 2002) or less cooperative mutualists (Herre et al., 1999; Poulsen & Boomsma, 2005).

Wolbachia α‐proteo bacteria are intracellular symbionts in many insects, mites and some nematodes and crustaceans. They often affect host fitness as reproductive parasites by causing cytoplasmic incompatibility, as in Drosophila flies (Bourtzis et al., 1996), Ephestia moths (Lewis et al., 2011) and Nasonia wasps (Tram & Sullivan, 2002). Other Wolbachia cause host parthenogenesis as in Bryobia mites (Weeks & Breeuwer, 2001), male‐killing as in ladybirds and butterflies (Hurst et al., 1999), or feminization as in various isopods (Bouchon et al., 1998). However, in other associations the host has become dependent on these bacteria as nutritional mutualists or reproduction facilitators (Pannebakker et al., 2007: Hosokawa et al., 2010). The default Wolbachia transmission‐mode is vertical, from mother to offspring, but host and symbiont phylogenies often indicate that horizontal transmission occurs frequently enough over evolutionary time to prevent co‐cladogenesis (Werren et al., 2008). Many host species have also been found to carry multiple Wolbachia strains, and in some cases these strains reside in different tissues (e.g. Ijichi et al., 2002).

While a number of thorough case studies have clarified the phenotypic effects of Wolbachia infections in models of solitary invertebrates, rather little progress has been made in understanding the phenotypic effects of similar infections in eusocial insects. Surveys have shown that a wide range of termites are infected, but that eusocial wasps and bees are rarely hosts (Lo & Evans, 2007; Russell, 2012). Many ants are also known to harbour Wolbachia (Wenseleers et al., 1998; Russell et al., 2009; Russell, 2012), but prevalences vary considerably between species, between colonies in populations, and between castes within colonies. Some correlation between mode of colony founding and the likelihood of infection has been suggested, as species that found colonies aided by workers of the parental colony have slightly higher prevalences than species that found colonies by single newly‐mated queens (Wenseleers et al., 1998). Other studies have shown that Wolbachia infections are frequently lost in invasive ants when these populations are compared to their native sister populations or species

70 (Shoemaker et al., 2000; Reuter et al., 2004; Cremer et al., 2008). Female‐biased sex ratios in colonies of eusocial Hymenoptera have also been suggested to be influenced by Wolbachia, but no evidence for this was found in Formica exsecta (Keller et al., 2001) and Formica truncorum (Wenseleers et al., 2002). In the latter case, and in the fire ant Solenopsis invicta, Wolbachia infections may potentially reduce host fitness (Shoemaker et al., 2000), but otherwise the phenotypic consequences of Wolbachia infections for ant hosts have remained enigmatic.

In the present study we use a novel combination of techniques to assess how strain‐specific Wolbachia prevalence varies across different life stages of sterile workers of the fungus‐growing ant Acromyrmex octospinosus. Earlier studies of this ant have indicated that most workers are infected (Van Borm et al., 2001; Frost et al., 2010) and often by multiple strains (Van Borm et al., 2003). By measuring the diversity and tissue distribution of these strains within individual ants we aimed to elucidate the potential for interaction between Wolbachia strains and to evaluate the phenotypic effects of these infections on host fitness. We used quantitative real time PCR (qrt‐PCR) and fluorescence in situ hybridization (FISH) to measure the density of Wolbachia symbionts and the distribution of bacteria among host tissues. After establishing that considerable concentrations of Wolbachia are associated with the ant gut, we used laser scanning confocal microscopy to document this in more detail. Our visualizations of bacteria in ant tissues revealed an unexpected extracellular presence of Wolbachia in the ant gut, which suggests a novel role of Wolbachia in the fungus‐ growing ant symbiosis.

Methods

DNA extraction, sequencing and quantitative PCR Acromyrmex octospinosus colonies were collected in Gamboa, Panama in the period 2004‐2010 (Table 1). DNA was extracted from whole individuals after crushing them with a plastic pestle, and from dissected tissues (DNeasy blood

71 and tissue kit, Qiagen). The wsp primers from Zhou et al. (1998), targeting a surface protein, were used to amplify ca. 560 bp of Wolbachia DNA. The PCR product was cloned (TOPO TA cloning kit, Invitrogen) and 10‐23 clones from a single adult worker from six field collected colonies (102 clones in total) were sequenced by Eurofins MWG Operons (Ebersberg, Germany). Two dominant strains where identified, which were identical to strains previously sequenced from ants (Shoemaker et al., 2000; Van Borm et al., 2001; Van Borm et al., 2003). The sequences were translated to amino acids (http://expasy.org/tools/dna.html) and the four hypervariable regions (HVRs) of wsp (Baldo et al., 2005) were identified using the Wolbachia MLST website (http://pubmlst.org/wolbachia/; Jolley et al., 2004). Following Shoemaker et al., (2000) we called the two dominant strains “WSinvictaA” and “WSinvictaB”. Specific primers for these strains were designed (wspa F: 5’‐ GAAAACTGCTGTGAATGGTC‐3’, wspa R: 5’‐TCCTCCTTTGTCTTTCTC‐3’; wspb F: 5’‐GAAAACTGCTGTGAATGGTC‐3’, wspb R: 5’‐ATTKCAGCATCGTCTTTARCT‐3’) to amplify 167‐170 bp, and the specificity of the primers was checked with direct sequencing. The primers amplified a region where the WSinvictaB strain was 100% identical to the other non‐dominant strains (see results) and it was thus not possible to quantify the presence of these rare additional strains any further.

For analysis of the distribution of Wolbachia strains across different individuals, castes and colonies DNA was extracted from eight colonies sampled in the field and from six colonies reared under lab conditions for >7 months (no colonies were sampled both in the field and in the lab). From each colony eight entire large larvae, pupae and adult workers were sampled (see Table 1 for colony ID and exact sample number). Field colonies were sampled after the annual mating flight, when they were not producing sexuals, to ensure that the large larvae were immature large workers. For analysis of Wolbachia strain distributions across worker tissues, DNA was extracted from dissected thoracic muscle tissues, from three different parts of the gut, and faecal droplets of eight ants from a single lab reared colony (Ao492). Absolute wsp copy numbers were quantified by quantitative real time PCR (qrt‐PCR) using SYBR Premix Ex Taq (Takara Bio Inc.) on the Mx3000P system (Stratagene). Reactions took place in a

72 final volume of 20.5 µl containing 10 µl buffer, 8.8 µl ddH2O, 0.4 µl of each primer (10 µM), 0.4 µl ROX standard and 0.5 µl template DNA. Bacterial measurements were standardised with qrt‐PCR of the single copy ant gene, elongation factor 1α (primers EF‐1α f: 5’ ACGGAAGCTCTGCCCGGTGA‐3’ EF‐1α r: 5’‐ TGGCAGTCAAGCACTGGCGT‐3’), providing an estimate of host cell number, under the assumption that bacterial and ant DNA are preserved and extracted equally well between castes and independent of storage method (in ethanol at ‐ 20°C vs. freshly collected). All PCR reactions consisted of a 2 min denaturation step at 95°C, 35 cycles of 95°C for 30 s, 52°C for 30 s and 72°C for 30 s, followed by a dissociation curve analysis. All samples were replicated in the same run and the mean was used for analysis. Each run also included three negative controls with no added template. The initial template concentration was calculated from a standard curve with PCR product in ten fold dilutions of known concentration, as quantified by nanodrop.

Cross sectioning and embedding Larvae (n = 4), pupae (n = 2) and workers (n = 8) from colonies Ao49a, Ao491 and Ao496 were fixed and embedded following the protocol of Kulzer Technovit 8100 (Heraeus Kulzer, Germany). Tissues were cut to allow penetration of the fixative (2% paraformaldehyde in phosphate‐buffer, pH 7.4) for < 4 h followed by overnight washing in PBS pH 7.4 at 4°C. The tissues were dehydrated in 100% acetone for 1 h at 4°C and infiltrated with Technovit 8100 solution for 6‐10 hours at 4°C, followed by transfer to the embedding solution, agitation for 5 min, and transfer to a plastic mould. Moulds were sealed with plastic foil and left to harden on ice at 4°C overnight. The tissue blocks were cut with a glass knife and sections attached to superfrost plus slides (Menzel‐Gläser, Germany) by heating for 15 min. For whole‐mount laser scanning confocal microscopy, eggs were collected from the fungus garden of an isolated laying queen (n = 5, Ao492) and ant guts were dissected out in fixative, fixed for > 4 h and washed in PBS (Ao492, n = 10).

73 Fluorescence in situ hybridization (FISH) Tissue sections were treated with lysozyme (5 mg/ml) for 30 min at 37°C to increase cell permeabilization (Moter & Göbel, 2000) and dehydrated for 3 min each in 50%, 70% and 100% ethanol prior to hybridization. Slides were hybridized with a 16S rRNA targeted probe specific for Wolbachia and labelled with Cy3 (Wol: 5’‐ CTAACCCGCCTACGCGCC‐3’, from Eurofins MWG Operons, Germany) overnight at 46°C. This was done in 100 µl hybridization buffer (100 mM Tris pH 7.2, 0.9 M NaCl, 0.1% sodium dodecyl sulphate) with 5 ng/μl probe in a Sequenza slide rack (Thermo Shandon, Cheshire, United Kingdom). As a negative control a Cy3 labelled probe targeting the spirochaete bacteria Treponema sp. was used (S‐S‐ Trep DDKL 12‐432: 5’‐CATCTCAAGGTCATTCCC‐ 3’). Slides were then washed with preheated (46°C) hybridisation buffer for 3 x 3 min followed by wash with preheated (46°C) washing buffer (100 mM Tris pH 7.2, 0.9 M NaCl) for 3 x 3 min. Finally the slides were rinsed in water, air dried and mounted with Vectashield (Vector Laboratories Inc., Burlingame, CA) for epifluorescence microscopy using an Axioimager M1 epifluorescence microscope. Images were obtained using an AxioCAM MRm version 3 FireWire monochrome camera (Carl Zeiss, Oberkochen, Germany).

Gut dissections and ant eggs were treated with lysozyme, dehydrated, hybridized and washed as above in an Eppendorf tube and mounted on slides with Vectashield containing DAPI (DAPI stains host nuclei blue and it is thus possible to infer whether bacteria are intra‐ or extracellularly located). These slides were observed and photographed with a Zeiss LSM 710 laser scanning confocal microscope equipped with Zen 2009 software. After some editing, the images were further processed to adjust contrast and crop irrelevant parts using Photoshop CS3 for Mac.

Live/dead bacterial staining To evaluate the occurrence of bacteria in the faecal fluid of the ants, a droplet of ca. 0.5 µl was deposited on a microscope slide by squeezing the ant gaster with forceps (as described in Schiøtt et al., 2010). The bacteria were stained with the

74 BacLight L 13152 live/dead stain (Molecular Probes Inc.), staining live bacteria green (Syto‐9 probe) and dead bacteria (i.e. cells with a compromised membrane) red (propidium iodide). 0.5 µl of each stain was added to each fresh faecal droplet and slides were sealed with a cover slide and incubated in the dark for 15 min, after which slides were analysed using the Axioimager M1 epifluorescence microscope (n = 5).

Results

Identification of Wsp strains using the HVR typing system Previously, Wolbachia phylogenies were primarily based on the highly variable Wsp gene, but this gene later turned out to be unsuitable for inferring phylogenetic relationships, because of its high divergence and recombination rate (Baldo et al., 2005; Baldo et al., 2010). However, Wsp remains a useful marker for identifying different strains, and allowed us to identify four different Wolbachia wsp strains from the six screened colonies. Two were identical to the WSinvictaA and WSinvictaB strains found in Solenopsis invicta (GenBank acc. no. AF243435 and AF243436, Shoemaker et al., 2000) and in three Panamanian Acromyrmex species (Van Borm et al., 2003). All colonies carried the WSinvictaB strain, while only some had the WSinvictaA strain. 3.9% of the sequences were different with colonies Ao483 and Ao493 each yielding an additional strain (GenBank acc. no. to be added) that was 98‐99% similar to a previously identified strain in A. octospinosus (Genbank acc. no AF472561.1). A new strain was obtained from colony Ao496 (GenBank acc. no. to be added), showing 90% similarity to other sequences in GenBank.

The Wsp gene consists of four hypervariable regions (HVRs), each with multiple alleles that have been numbered, alternating with conserved sequences. Recombination typically takes place between the four regions, and HVR typing is a useful way of identifying recombination points (Baldo et al., 2005). All strains were thus further characterized with the HVR system. The WSinvictaA strain of Acromyrmex octospinosus contained the elements 42‐43‐198‐25 and the

75 WSinvictaB strain had the elements 21‐21‐25‐21. The other three strains turned out to be chimeras of the dominant strains, and had HVRs 42‐43‐25‐21 (found in colony Ao496) and 42‐21‐25‐21 (found in colony Ao493 and Ao492, the sequences from each colony were slightly different but translated to the same protein sequence). The fact that recombination was localized between the hypervariable regions, as previously reported for other strains, confirms that these strains are true chimeras and not simply the result of sequencing errors.

qrt­PCR All individuals from all colonies were found to be infected with Wolbachia. qrt‐ PCR showed that the WSinvictaB strain was dominant in all individuals at all life stages (Fig. 1, Table 2). In three of the field‐collected colonies this was the only strain found in measurable amounts, except for two adult workers from one colony that also carried the WSinvictaA strain. This colony (Ao471) had been kept in the lab at the field site in Gamboa, Panama for > 1 month, which may have enhanced the expression of the WSinvictaA strain (see below). In the remaining colonies all adult individuals carried both strains.

Based on the prevalence differences of the WSinvictaA and WSinvictaB strains, colonies were divided into three categories: field collected single infected (FS, n = 3), field collected double infected (FD, n = 5), and lab reared double infected (LD, n = 6). No lab reared colonies showed single infection. The differences in bacterial densities were analysed in JMP 9.0.2 for Mac OSX using a repeated‐ measures ANOVA, as individuals collected from the same colony could not be regarded as independent. There was considerable between‐colony variation in standardised bacterial densities within colonies and castes, with outliers apparent in many combinations, so the geometric mean density per caste per colony was analyzed, as this showed the most homogenous variance of all measures examined. Category was included as the between‐subject effect, and caste and the category by caste interaction included as within‐subject effects. Post‐hoc testing was by paired or unpaired t‐tests for within‐ and between‐ subject effects respectively, with Bonferroni correction based on the total

76 number of tests carried out. Overall there was an increase in total bacterial number with developmental stage, with the bacterial density being significantly higher in pupae than larvae, and significantly higher in workers than pupae. There was also a significant caste by category interaction, due to somewhat different development of bacteria in the different categories. In single infected field colonies, the bacterial density did not vary significantly between castes. In double infected field colonies, the increase was significant between all castes, while it was only significant between larvae and pupae and larvae and adults in lab reared double infected colonies. There was a significant difference in the total number of bacteria between categories, with lab colonies contained slightly higher densities at all life stages (Table 2 and 3, Fig.1).

Looking at the WSinvictaB strain only, the overall pattern was bacterial density increasing from the larval to the pupal stage and remaining at this high level in the adults. There was no significant caste by category interaction, showing that this pattern was the same in each category, and the difference between categories did not quite reach significance (Table 2 and 3, Fig.1).

The highest prevalence of the (non‐dominant) WSinvictaA strain was found in adult workers of FD and LD colonies, where they reached a mean of 29% (± 0.015 SE) of the total bacteria. In the FS colonies, the WSinvictaA strain was not present in measureable amounts, and abundances in the immatures of FD colonies were only slightly (not significantly) higher. The LD colonies carried significantly higher amounts of the WSinvictaA strain in the pupal stages (Table 2 & 3, Fig. 1).

The bacterial estimates obtained from dissections of different tissue types were not standardized with host gene copy number, as the majority of the bacteria were found to be extracellular (see below). As the variance in proportions was very different across tissues, with the faecal droplet material in particular containing either high or low proportions of the WSinvictaA strain (Fig. 2) they were compared pair wise using the Steel‐Dwass non‐parametric test. The proportions of the WSinvictaA strain were significantly higher in the crop and

77 the muscle tissues (44%) compared to the rest, while the rectum (37%) contained significantly more than the whole ant (26%), and the midgut (24%). Because of the high variance in the WSinvictaA proportion in the faecal droplets, the mean proportion in these (29%) was not significantly different from the other tissues.

FISH The FISH analyses showed bacterial colonization of multiple tissue types. In the ant eggs, the bacterial density was highest around one pole (SI Fig. 1). In larvae, the dominant fat body cells were carrying many bacteria and the gut tissue also housed some (Fig. 3A, B). In the pupae the ant cells are diversifying into more tissue types, which were widely infected (e.g. muscle fibres and fat cells, data not shown). This was also the case in the adults, where particularly the muscle cells, fat body and gut tissue harboured many bacteria (Fig. 3C, D). Histology showed a large amount of Wolbachia occurring extracellularly in the crop part of the gut (in 6 out of 8 individuals, Fig. 3C), and this was confirmed by confocal microscopy of whole guts (in 10 out of 10 individuals, Fig. 4). These extracellular bacteria were to a lesser extent also seen in the midgut (SI Fig. 2). No clear identification of Wolbachia in the rectum could be made because of strong autofluorescence of the tissues. As a negative control for unspecific hybridization a probe specific to the bacterium Treponema sp. was used. This showed some unspecific hybridization to the part of the gut leading to the crop and the ileum connecting the midgut and rectum, so hybridization to these tissues by the Wolbachia probe was ignored, as it was possibly unspecific.

The live/dead bacterial staining of faecal droplets showed a high density of living bacteria, but it was not possible with the applied methods to confirm how many of these were Wolbachia.

78 Discussion

Wolbachia prevalence and diversity We found that all individuals and all life stages and colonies were infected with Wolbachia, and vertical symbiont transmission was confirmed by visualization of Wolbachia in ant eggs (SI Fig. 1). This high prevalence contrasts somewhat with previous studies, on the same ant species from the same area, where all colonies were also found to be infected, but individual infection rates were lower (Van Borm et al., 2001, mean infection rate of individuals of 40%; Frost et al., 2010, 81%). The explanation for this may be technical rather than biological, because qrt‐PCR allows for a higher level of detection and the amplification of a shorter DNA fragment (ca. 170 bp in the present study vs. 783 bp by Van Borm et al., 2001), which ensures that even slightly fragmented DNA is amplified.

The two dominant Wolbachia wsp strains that we found have been observed in other ants as well (Shoemaker et al., 2000; Van Borm et al., 2003) and are also very similar to strains in beetles and spiders (e.g. Sintupachee et al., 2006). The WSinvictaB strain was found in all colonies of A. octospinosus while the WSinvictaA strain only occurred in some colonies and never alone. There were thus colony level differences in strain diversity, as either all or no individuals within a colony carried the WSinvictaA strain. The presence of the WSinvictaA strain was not correlated with sampling site, indicating that geographic clustering in the sample population is unlikely (data not shown). However, no good estimates of colony age and colony size upon collection were available, so we cannot directly evaluate whether these variables, which may be important for the development of bacterial infections, had any effect.

In the colonies harbouring both strains we identified two recombinant strains. Although the prevalences of these strains were not assessed by qrt–PCR for all life stages, the low frequencies in three adult ants for which cloning estimates were available suggest that they were rare. Our HVR typing further indicated that these chimera strains arose by recombination of the WSinvictaA and WSinvictaB strains between the first and second HVR, and the second and third

79 HVR, respectively. Recombination may thus be rather frequent but while these recombinant strains may persist in the population they do not appear to be particularly successful. The very occurrence of within host recombination shows that some degree of interaction between the WSinvictaA and WSinvictaB strains occurs. Wsp is a major outer surface protein of Wolbachia and has been suggested to mediate contact with the host cells via its two transmembrane regions that likely interact with the host immune system (Braig et al., 1998; Bazzocchi et al., 2007), so that recombination may affect these recognition processes. Recombination between Wolbachia strains has previously been found in other host species, including ants (Reuter & Keller, 2003).

Bacterial density increases with host age We found an increase in the bacterial load with age, suggesting that Wolbachia thrive in the mature workers. Adult worker ants of the species Formica truncorum were previously found to have lower infection rates than immatures (Wenseleers et al., 2002), which generated the hypothesis that workers may lose infection for reasons that are adaptive for the bacteria, because they are evolutionary dead ends for a reproductive parasite. This appears not to be the case for A. octospinosus. For the dominant WSinvictaB strain, the increase in density occurs between the larval and the pupal stage and prevalence stays at this level in adults, equivalent to what has been found in the Adzuki bean beetle, where Wolbachia is a confirmed reproductive parasite (Ijichi et al., 2002).

The increase in bacterial load could reflect the appearance of new tissue types that the bacteria are able to invade after metamorphosis (see below). In most host‐symbiont interactions, whether parasitic or mutualistic, the host has a clear interest in controlling bacterial growth and dispersal. In Drosophila, the ability to do so appears connected to life‐stage‐specific expression of immunity genes (Samakovlis et al., 1990). In the eusocial honeybee Apis mellifera, phenoloxidase activity (a measure of immune defence) was low in both larvae and pupae, most likely because alternative social immunity mechanisms provide efficient protection of brood (Wilson‐Rich et al., 2008). Such a down‐regulation of the individual immune defence could be of importance for the ability of vertically

80 transmitted symbionts to grow in the immature stages of social insects. If this is also the case for Acromyrmex octospinosis it may partly explain the increase in Wolbachia density with host age, suggesting that the bacteria mostly grow and disperse when host control mechanisms are constrained.

Niche segregation of bacterial symbionts Our results (Fig. 1) indicate that the WSinvictaA strain proliferates mainly in adult individuals. However, when comparing the single and double infected field‐ colonies there is a suggestion of WSinvictaA strain proliferation in adult workers being associated with lower WSinvictaB strain prevalences in the larval and pupal stages. This could reflect some form of scramble competition between WSinvictaA and WSinvictaB strain bacteria in the immature developmental stages. In this hypothetical scenario, the initial degree of dominance of the WSinvictaB strain would then determine the available niche space for the other strain, so that individuals where WSinvictaA strain bacteria remain under the detection limit in the immature stages will only be able to grow very few of them as adults (the observed pattern of singly infected field colonies; Fig. 1). However, when the WSinvictaA strain bacteria for some reason become more abundant already in the immature stages (so that immature field individuals are scored as double infected), they are much more likely to proliferate further in adult workers.

A competitive scenario as outlined above would be most likely when bacterial strains interact in the same host tissues during the larval and pupal stages but, at least partly, segregate into different tissue types in adult workers. Such tissue tropism of Wolbachia strains has previously been observed in Adzuki bean beetles (Ijichi et al., 2002). As our FISH results showed a high density of bacteria in the muscle fibres and the gut we further measured the distribution of Wolbachia strains in these tissues. The qrt‐PCRs of specific tissue types showed that the WSinvictaA strain was significantly more abundant in muscle fibres and in the crop of the gut, relative to later stages in the digestive process (midgut and rectum) and the whole ant (Fig. 2). The muscle tissue is only fully developed in the adult ants and the adult gut is very distinct from the larval gut, being more

81 complex and divided into sections varying morphologically and in pH, enzyme activity and retention time of contents (Erthal Jr. et al., 2004). This corroborates the notion that, although overlapping, the adult Wolbachia niches are somewhat distinct and that they are unlikely to be differentiated earlier in development. However, the substantial overlap in Wolbachia niches within hosts also raises the possibility that these strains may have different functional roles in adult ants.

Interpreting the infection patterns of double infected lab colonies as being consistent with strain competition would imply that resource constraints somehow affect field colonies more than lab colonies. This is reasonable, as lab colonies were being fed regularly with a standard selection of Danish bramble leaves (Rubus sp.), experienced no predation or other hazards while foraging, and generally had large and thriving fungus gardens while living under stable humidity and temperature regimes. All of these lab colonies were double infected and also harboured slightly higher densities of the WSinvictaA strain in the immature life stages compared to the field sampled colonies (Fig. 1). In addition, the total bacterial number in the lab‐reared colonies was slightly higher than that found in the field, suggesting that the bacteria thrive when their hosts experience lab conditions. The finding of two double infected workers in an otherwise single infected colony in the field seems consistent with this interpretation, as this was the only colony that had been kept for more than a month in the field lab in Panama under ad libitum resource conditions before ant samples were collected (Ao471). The presence of the WSinvictaA strain in measurable quantities early on could thus be dependent on colony resource condition, which in the field may be correlated with colony size.

Are Wolbachia new mutualists in the attine fungus­farming symbiosis? The FISH data surprisingly showed that Wolbachia bacteria are abundant in the lumen of the adult worker gut (Fig. 3 & 4). While Wolbachia has been found in gut tissue (Dobson et al., 1999; Ijichi et al., 2002) an extracellular location is highly unusual and has to our knowledge not been documented before (but see Fischer et al., 2011 showing the occasional appearance of extracellular Wolbachia close to ovarian tissue in nematodes). However, this observation is

82 consistent with WSinvictaA and WSinvictaB strain Wolbachia being present in the faecal droplets of Acromyrmex (Fig. 2). The faecal droplets contained a large amount of viable bacteria and positive DNA‐level evidence suggests that at least part of these bacteria were Wolbachia. This combined result therefore indicates that Wolbachia cells are not harmed by digestive gut enzymes, consistent with this environment being their natural niche.

The faecal droplets have unique functions in the fungus‐growing ants. They contain proteins from the fungal garden, which are ingested by the ants but pass undigested through the gut to assist decomposition in newly established fungus garden (Schiøtt et al., 2010). They also play a role in the recognition and elimination of genetically different fungal cultivars that workers may collect (Poulsen & Boomsma, 2005). The various adaptative functions of the faecal droplets to the ant‐fungal symbiosis suggest that there is strong selective pressure on the gut environment and the composition of faecal fluid. This and the atypical location of Wolbachia in the gut lumen and faecal droplets, suggests that Wolbachia in A. octospinosus may have a mutualistic nutritional role in the ant‐fungus cultivation symbiosis. The recent finding of a beneficial role of Wolbachia symbionts in the Western rootworm, larvae of Diabrotica virgifera virgifera, causing the down regulation of defence compounds in the plants that they feed on (Barr et al., 2010), offers an intriguing possible analogue to our present results. Similar to the herbivorous beetle larva, the alliance of leaf‐cutting ants and their fungus garden symbionts also faces challenges from secondary plant defences. Recent work (Schiøtt et al., 2010) has shown that the fungal symbionts of leaf‐cutting ants have convergently evolved an entire set of pectinases that are normally only found in pathogenic fungi that attack live plant hosts, and also these enzymes pass the ant gut unharmed. It therefore seems highly worthwhile to further explore the functional role of Wolbachia in Acromyrmex, both in the worker guts and in the faecal fluid where the bacteria interact with the multiple microorganisms that are now known from attine ant fungus gardens (Pinto‐ Tomás et al., 2009). We note that recent work has also suggested that plant defences may not only be chemical, but also biotic, as leaf‐substrate choice by the

83 ants is affected by the endophytic community of the leaves (Bittleston et al., 2010; Van Bael et al., 2009). Further studies along these lines will also have the potential to elucidate why only some colonies carry the WSinvictaA strain in measurable amounts.

Acknowledgements

We would like to thank Joanna Amenuvor and Annie Ravn Pedersen (DTU), and Lisbeth Haugkrogh and Aase Jespersen (KU), for advice concerning histology and FISH, Morten Schiøtt, Henrik de Fine Licht and Tom Gilbert (KU) for advice on qrt‐PCR, and Panagiotis Sapountzis for comments on the manuscript. SBA was funded by a PhD Scholarship from the Science Faculty of the University of Copenhagen, and SBA, DRN and JJB were supported by the Danish National Research Foundation.

84 References

Aanen, D.K., de Fine Licht, H.H., Debets, A.J.M., Kerstes, N.A.G., Hoekstra, R.F. & Boomsma, J.J. 2009. High symbiont relatedness stabilizes mutualistic cooperation in fungus‐growing termites. Science 326: 1103‐1106. Baldo, L., Lo, N. & Werren, J.H. 2005. Mosaic nature of the Wolbachia surface protein. Journal of Bacteriology 187: 5406‐5418. Baldo, L., Desjardins, C.A., Russell, J.A., Stahlhut, J.K. & Werren, J.H. 2010. Accelerated microevolution in an outer membrane protein (OMP) of the intracellular bacteria Wolbachia. BMC evolutionary biology 10. Barr, K.L., Hearne, L.B., Briesacher, S., Clark, T.L. & Davis, G.E. 2010. Microbial symbionts in insects influence down‐regulation of defense genes in maize. Plos One 5 Bazzocchi, C., Comazzi, S., Santoni, R., Bandi, C., Genchi, C. & Mortarino, M. 2007. Wolbachia surface protein (WSP) inhibits apoptosis in human neutrophils. Parasite Immunology 29: 73‐79. Bittleston, L.S., Brockmann, F., Wcislo, W. & Van Bael, S.A. 2010. Endophytic fungi reduce leaf‐cutting ant damage to seedlings. Biology Letters 7: 30‐32. Bouchon, D., Rigaud, T. & Juchault, P. 1998. Evidence for widespread Wolbachia infection in isopod crustaceans: molecular identification and host feminization. Proceedings of the Royal Society of London. Series B: Biological Sciences 265: 1081‐1090. Bourtzis, K., Nirgianaki, A., Markakis, G. & Savakis, C. 1996. Wolbachia infection and cytoplasmic incompatibility in Drosophila species. Genetics 144: 1063‐1073. Braig, H.R., Zhou, W., Dobson, S.L. & O'Neill, S.L. 1998. Cloning and characterization of a gene encoding the major surface protein of the bacterial endosymbiont Wolbachia pipientis. J. Bacteriol. 180: 2373‐2378. Bull, J.J. 1994. Perspective ‐ Virulence. Evolution 48: 1423‐1437. Cremer, S., Ugelvig, L.V., Drijfhout, F.P., Schlick‐Steiner, B.C., Steiner, F.M., Seifert, B. et al. 2008. The evolution of invasiveness in garden ants. Plos One 3 Davies, C.M., Fairbrother, E. & Webster, J.P. 2002. Mixed strain schistosome infections of snails and the evolution of parasite virulence. Parasitology 124: 31‐38. Dobson, S.L., Bourtzis, K., Braig, H.R., Jones, B.F., Zhou, W.G., Rousset, F. & O'Neill, S.L. 1999. Wolbachia infections are distributed throughout insect somatic and germ line tissues. Insect Biochemistry and Molecular Biology 29: 153‐ 160. Erthal Jr., M., Peres Silva, C. & Ian Samuels, R. 2004. Digestive enzymes of leaf‐ cutting ants, Acromyrmex subterraneus (Hymenoptera: Formicidae: Attini): distribution in the gut of adult workers and partial characterization. Journal of Insect Physiology 50: 881‐891. Fischer, K., Beatty, W.L., Jiang, D., Weil, G.J. & Fischer, P.U. 2011. Tissue and stage‐ specific distribution of Wolbachia in Brugia malayi. PLoS Neglected Tropical Diseases 5: e1174. Frank, S.A. 1996. Models of parasite virulence. Quarterly Review of Biology 71: 37‐78.

85 Frost, C.L., Fernández‐Marín, H., Smith, J.E. & Hughes, W.O.H. 2010. Multiple gains and losses of Wolbachia symbionts across a tribe of fungus‐growing ants. Molecular Ecology 19: 4077‐4085. Herre, E.A., Knowlton, N., Mueller, U.G. & Rehner, S.A. 1999. The evolution of mutualisms: exploring the paths between conflict and cooperation. Trends in Ecology and Evolution 14: 49‐53. Hosokawa, T., Koga, R., Kikuchi, Y., Meng, X.‐Y. & Fukatsu, T. 2010. Wolbachia as a bacteriocyte‐associated nutritional mutualist. Proceedings of the National Academy of Sciences 107: 769‐774. Hurst, G.D.D., Jiggins, F.M., Hinrich Graf von der Schulenburg, J., Bertrand, D., West, S.A., Goriacheva, I.I., Zakharov, I.A., Werren, J.H., Stouthamer, R. & Majerus, M.E.N. 1999. Male killing Wolbachia in two species of insect. Proceedings of the Royal Society of London. Series B: Biological Sciences 266: 735‐740. Ijichi, N., Kondo, N., Matsumoto, R., Shimada, M., Ishikawa, H. & Fukatsu, T. 2002. Internal spatiotemporal population dynamics of infection with three Wolbachia strains in the Adzuki bean beetle, Callosobruchus chinensis (Coleoptera: Bruchidae). Applied Environmental Microbiology 68: 4074‐ 4080. Jolley, K., Chan, M.‐S. & Maiden, M. 2004. mlstdbNet ‐ distributed multi‐locus sequence typing (MLST) databases. BMC Bioinformatics 5: 86. Keller, L., Liautard, C., Reuter, M., Brown, W.D., Sundstrom, L. & Chapuisat, M. 2001. Sex ratio and Wolbachia infection in the ant Formica exsecta. Heredity 87: 227‐233. Lewis, Z., Champion de Crespigny, F.E., Sait, S.M., Tregenza, T. & Wedell, N. 2011. Wolbachia infection lowers fertile sperm transfer in a moth. Biology Letters 7: 187‐189. Lo, N. & Evans, T.A. 2007. Phylogenetic diversity of the intracellular symbiont Wolbachia in termites. Molecular Phylogenetics and Evolution 44: 461‐ 466. Moter, A. & Göbel, U.B. 2000. Fluorescence in situ hybridization (FISH) for direct visualization of microorganisms. Journal of Microbiological Methods 41: 85‐112. Palmer, T.M., Doak, D.F., Stanton, M.L., Bronstein, J.L., Kiers, E.T., Young, T.P., Goheen, J.R. & Pringle, R.M. 2010. Synergy of multiple partners, including freeloaders, increases host fitness in a multispecies mutualism. Proceedings of the National Academy of Sciences of the United States of America 107: 17234‐17239. Pannebakker, B.A., Loppin, B., Elemans, C.P.H., Humblot, L. & Vavre, F. 2007. Parasitic inhibition of cell death facilitates symbiosis. Proceedings of the National Academy of Sciences 104: 213‐215. Pinto‐Tomás, A.A., Anderson, M.A., Suen, G., Stevenson, D.M., Chu, F.S.T., Cleland, W.W., Weimer, P.J. & Currie, C.R. 2009. Symbiotic nitrogen fixation in the fungus gardens of leaf‐cutter ants. Science 326: 1120‐1123. Poulsen, M. & Boomsma, J.J. 2005. Mutualistic fungi control crop diversity in fungus‐growing ants. Science 307: 741‐744. Reuter, M. & Keller, L. 2003. High levels of multiple Wolbachia infection and recombination in the ant Formica exsecta. Mol Biol Evol 20: 748‐753.

86 Reuter, M., Pedersen, J.S. & Keller, L. 2004. Loss of Wolbachia infection during colonisation in the invasive Argentine ant Linepithema humile. Heredity 94: 364‐369. Russell, J.A., Goldman‐Huertas, B., Moreau, C.S., Baldo, L., Stahlhut, J.K., Werren, J.H. & Pierce, N.E. 2009. Specialization and geographic isolation among Wolbachia symbionts from ants and Lycaenid butterflies. Evolution 63: 624‐640. Russell, J.A. 2012. The ants (Hymenoptera: Formicidae) are unique and enigmatic hosts of prevalent Wolbachia (Alphaproteobacteria) symbionts. Myrmecological News 16: 7‐23. Sachs, J.L. & Simms, E.L. 2006. Pathways to mutualism breakdown. Trends in Ecology & Evolution 21: 585‐592. Samakovlis, C., Kimbrell, D.A., Kylsten, P., Engstrom, A. & Hultmark, D. 1990. The immune response in Drosophila ‐ pattern of cecropin expression and biological activity. The Embo Journal 9: 2969‐2976. Schiøtt, M., Rogowska‐Wrzesinska, A., Roepstorff, P. & Boomsma, J. 2010. Leaf‐ cutting ant fungi produce cell wall degrading pectinase complexes reminiscent of phytopathogenic fungi. Bmc Biology 8: 156. Shoemaker, D.D., Ross, K.G., Keller, L., Vargo, E.L. & Werren, J.H. 2000. Wolbachia infections in native and introduced populations of fire ants (Solenopsis spp.). Insect Molecular Biology 9: 661‐673. Sintupachee, S., Milne, J.R., Poonchaisri, S., Baimai, V. & Kittayapong, P. 2006. Closely related Wolbachia strains within the pumpkin arthropod community and the potential for horizontal transmission via the plant. Microbial Ecology 51: 294‐301. Telfer, S., Lambin, X., Birtles, R., Beldomenico, P., Burthe, S., Paterson, S. & Begon, M. 2010. Species interactions in a parasite community drive infection risk in a wildlife population. Science 330: 243‐246. Tram, U. & Sullivan, W. 2002. Role of delayed nuclear envelope breakdown and mitosis in Wolbachia‐induced cytoplasmic incompatibility. Science 296: 1124‐1126. Van Bael, S.A., Fernández‐Marín, H., Valencia, M.C., Rojas, E.I., Wcislo, W.T. & Herre, E.A. 2009. Two fungal symbioses collide: endophytic fungi are not welcome in leaf‐cutting ant gardens. Proceedings of the Royal Society B: Biological Sciences 276: 2419‐2426. Van Borm, S., Wenseleers, T., Billen, J. & Boomsma, J.J. 2001. Wolbachia in leafcutter ants: a widespread symbiont that may induce male killing or incompatible matings. Journal of Evolutionary Biology 14: 805‐814. Van Borm, S., Buschinger, A., Boomsma, J.J. & Billen, J. 2002. Tetraponera ants have gut symbionts related to nitrogen‐fixing root‐nodule bacteria. Proceedings of the Royal Society B­Biological Sciences 269: 2023‐2027. Van Borm, S., Wenseleers, T., Billen, J. & Boomsma, J.J. 2003. Cloning and sequencing of wsp encoding gene fragments reveals a diversity of co‐ infecting Wolbachia strains in Acromyrmex leafcutter ants. Molecular Phylogenetics and Evolution 26: 102‐109. Vautrin, E. & Vavre, F. 2009. Interactions between vertically transmitted symbionts: cooperation or conflict? Trends in microbiology 17: 95‐99.

87 Weeks, A.R. & Breeuwer, J.A.J. 2001. Wolbachia‐induced parthenogenesis in a genus of phytophagous mites. Proceedings of the Royal Society of London. Series B: Biological Sciences 268: 2245‐2251. Wenseleers, T., Ito, F., Van Borm, S., Huybrechts, R., Volckaert, F. & Billen, J. 1998. Widespread occurrence of the micro‐organism Wolbachia in ants. Proceedings of the Royal Society of London Series B: Biological Sciences 265: 1447‐1452. Wenseleers, T., Sundström, L. & Billen, J. 2002. Deleterious Wolbachia in the ant Formica truncorum. Proceedings of the Royal Society of London. Series B: Biological Sciences 269: 623‐629. Werren, J.H., Baldo, L. & Clark, M.E. 2008. Wolbachia: master manipulators of invertebrate biology. Nature reviews Microbiology 6: 741‐751. Wilson‐Rich, N., Dres, S.T. & Starks, P.T. 2008. The ontogeny of immunity: development of innate immune strength in the honey bee (Apis mellifera). Journal of Insect Physiology 54: 1392‐1399. Wu, D., Daugherty, S.C., Van Aken, S.E., Pai, G.H., Watkins, K.L., Khouri, H., Tallon, L.J., Zaborsky, J.M., Dunbar, H.E., Tran, P.L., Moran, N.A. & Eisen, J.A. 2006. Metabolic complementarity and genomics of the dual bacterial symbiosis of sharpshooters. Plos Biology 4: e188. Zhou, W.G., Rousset, F. & O'Neill, S. 1998. Phylogeny and PCR‐based classification of Wolbachia strains using wsp gene sequences. Proceedings of the Royal Society of London Series B­Biological Sciences 265: 509‐515.

88 Figure legends

Figure 1. The density of Wolbachia bacteria in three different life stages (L = larvae, P = pupae and W = workers) in three different colony categories (FD = field collected double infected, FS = field collected single infected and LD = lab‐ reared double infected). Each bar represents the number of Wolbachia cells per host cell, standardized by the copy number of the host gene EF‐1α estimating the total number of host cells, as measured with qrt‐PCR (dark grey = WSinvictaA strain; light grey = WSinvictaB strain). For each caste in each colony the geometric mean was calculated and the depicted value is the mean for each colony category. There is an increase in the amount of bacteria with progressing life stages and lab‐reared colonies have slightly higher numbers than field colonies. The WSinvictaB strain dominates in all life stages and the WSinvictaA strain only proliferates in adults of the double infected colonies.

Figure 2. Box‐plots showing the proportion of Wolbachia strain WSinvictaA in thoracic muscle tissue, three different parts of the ant gut (the crop, midgut and rectum) and faecal droplets in comparison to in whole ant samples. The bacterial load of WsinvictaA and WSinvistaB was measured by qrt‐PCR of individual samples. The central line represents the median proportion, while the box runs from the lower 25% to the upper 75% quartile for each sample, with the whiskers linking the extremes of the data. Letters indicate group‐level differences following the pairwise Steel‐Dwass method. The insert shows the corresponding overview of the ant gut and its surroundings, with the crop being connected to the midgut, where the Malphigian tubules attach, and the ileum connecting the midgut to the rectum.

Figure 3. Fluorescence in situ hybridization of Wolbachia (stained red) in ant tissues. (A) Larval gut showing some hybridization. (B) A larval fat body with many Wolbachia around cell nuclei. (C) Crop (foregut) of an adult large worker ant with extracellular Wolbachia in the lumen. (D) Muscle fibres of an adult worker with some intracellular concentrations of Wolbachia. Scale bar 50 µm (A) and 20 µm (B‐D).

89

Figure 4. Extracellular Wolbachia bacteria in the crop of the ant gut, a highly flexible sac that is slightly deflated so it appears somewhat folded. The LSCM image shows the 3D structure of the crop containing Wolbachia bacteria (stained red). The central image shows a horizontal optical section while the flanking images represent the vertical optical sections. Scale bar 50 µm.

SI Figure 1. Wolbachia bacteria in eggs of Acromyrmex octospinosus, with LSCM images showing the 3D structure of the ant egg containing Wolbachia bacteria around the egg pole (stained red). The central image shows a horizontal optical section while the flanking images represent the vertical optical sections. Scale bar 50 µm.

SI Figure 2. Extracellular Wolbachia in the ant midgut. (A) Wolbachia bacteria stained red against the green autofluorescent tissue. (B) The same bacteria now visualized with a DAPI staining. DAPI stains nuclei blue and the absence of stained host nuclei thus confirms that bacteria are extracellular (Insert shows an example of host nuclei stained with DAPI). Scale bar 50 µm.

90 Figures

Figure 1

Figure 2

91 Figure 3

Figure 4

92 SI Figure 1

SI Figure 2

93 Tables Table 1 Collection data for the colonies of Acromyrmex octospinosus that were used for estimating Wolbachia abundance by qrt‐PCR.

Sample size Date Colony Lab/ Infection

ID Larvae Pupae Workers Field Collection Sampling status

Ao273 8 8 8 lab May‐04 Dec‐10 Double (A/B)

Ao346 5 8 8 lab May‐07 Dec‐10 Double (A/B)

Ao367 8 8 8 lab May‐08 Dec‐10 Double (A/B)

Ao404 16 16 16 lab May‐09 Dec‐10 Double (A/B)

Ao431 8 8 8 lab May‐09 Dec‐10 Double (A/B)

Ao471 8 7 8 field Apr‐10 May‐10 Single (B)

Ao482 4 8 8 field May‐10 May‐10 Double (A/B)

Ao483 8 8 8 field May‐10 May‐10 Double (A/B)

Ao491 7 8 8 field May‐10 May‐10 Single (B)

Ao49a 8 8 8 field May‐10 May‐10 Single (B)

Ao492 8 8 8 lab May‐10 Dec‐10 Double (A/B)

Ao493 8 8 8 field May‐10 May‐10 Double (A/B)

Ao496 8 8 8 field May‐10 May‐10 Double (A/B)

AoClay 8 8 8 field May‐10 May‐10 Double (A/B)

94 Table 2

Standardized bacterial densities (± SE) of strains WSinvictaA and WSinvictaB and their cumulative densities estimated by qrt‐PCR. The geometric mean for each colony was used to calculate the mean for each category. Superscript letters indicate the grouping in Bonferroni‐corrected paired t‐tests within each category, following repeated measures ANOVA analysing the differences between castes (larvae, pupae, adults: within‐subject effect), categories (field single infected, field double infected and lab double infected: between‐subject effect) and their interaction. For WSinvictaA and the total bacteria, letters to the right of each category represent groupings in post‐hoc Bonferroni‐corrected unpaired t‐tests between pairs of categories.

Category WSinvictaA WSinvictaB Total bacteria

Field single infected 0.002 ± 0.0004 larvae A 3.413 ± 0.788 A 3.415 ± 0.788 A Field single infected 0.002 ± 0.0002 A A pupae A 6.196 ± 1.286 A 6.198 ± 1.286 A Field single infected workers 0.279 ± 0.277 A 6.324 ± 1.114 A 6.603 ± 1.002 A Field double infected larvae 0.027 ± 0.016 A 1.898 ± 0.269 A 1.925 ± 0.281 A Field double infected B A pupae 0.134 ± 0.090 A 4.582 ± 0.214 B 4.716 ± 0.272 B Field double infected workers 2.231 ± 0.328 B 6.188 ± 0.925 B 8.419 ± 0.942 B Lab double infected larvae 0.338 ± 0.061 A 3.898 ± 0.262 A 4.236 ± 0.295 A Lab double infected C B pupae 0.780 ± 0.058 B 7.028 ± 0.395 B 7.808 ± 0.398 B Lab double infected workers 2.607 ± 0.107 C 6.994 ± 0.525 B 9.600 ± 0.571 B

95 Table 3 The results of the repeated‐measures ANOVA for the density of WSinvictaA and WSinvictaB and their cumulative amounts estimated by qrt‐PCR, testing the effect of caste and category and the interaction between them. For category (the between‐subject effect), the F‐test shown is exact, while for the within‐subject effects (caste and the caste by category interaction), the F‐test is approximate, and based on Wilk’s λ.

Dependent variable Source F d.f. P

WSinvictaA Category 30.30 2, 11 <0.0001 Caste 58.49 2, 10 <0.0001 Caste × Category 11.35 4, 20 <0.0001

WSinvictaB Category 3.21 2, 11 0.080 Caste 206.2 2, 10 <0.0001 Caste × Category 1.36 4, 20 0.284

Total No. Category 5.46 2, 11 0.023 Caste 273.4 2, 10 <0.0001 Caste × Category 4.68 4, 20 0.008

96 CHAPTER 3

DISEASE DYNAMICS IN A

SPECIALIZED PARASITE OF ANT SOCIETIES

97 98 Disease dynamics in a specialized parasite of ant societies

Sandra B. Andersen1, Matt Ferrari2, Harry C. Evans3,4, Sam L. Elliot4, Jacobus J. Boomsma1. & David P. Hughes5

1Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark

2 Center for Infectious Disease Dynamics, Penn State University, PA 16802, USA

3CAB International, E‐UK, Egham, Surrey, TW20 9TY, UK

4Department of Entomology, Federal University of Viçosa, 36571.000, Viçosa, Brazil

5Department of Entomology and Department of Biology, Penn State University, PA 16802, USA

Authors for correspondence: Sandra Breum Andersen ([email protected]) and David P. Hughes ([email protected]) Co‐authors email addresses: [email protected], [email protected], [email protected] and [email protected]

Statement of authorship SBA and DPH designed the study; SBA, DPH, HCE and SLE collected the data in the field; MF build the model; SBA, DPH and MF analysed the data; SBA, DPH and JJB wrote the manuscript and all authors contributed substantially to revisions.

Running title

Disease dynamics of ant societies

Keywords

Ant, parasite, disease pressure, society, manipulation, entomopathogen

To be submitted January 2012

99 Abstract

Coevolution between ant colonies and their specialized parasites are intriguing, because lethal infections of workers may correspond to tolerable chronic diseases of colonies, but the life‐history adaptations that allow stable coexistence with ant hosts are virtually unknown. We explore the trade‐offs experienced by Ophiocordyceps parasites manipulating ants into dying in nearby graveyards. We used field data from Brazil and Thailand to parameterize and fit a model for the growth rate of graveyards. We show that parasite pressure is much lower than the abundance of ant cadavers suggests and that hyperparasites often castrate Ophiocordyceps. However, once fruiting bodies become sexually mature they appear robust. Such parasite life‐history traits are consistent with iteroparity –a reproductive strategy rarely considered in fungi. We discuss how tropical habitats with high biodiversity of hyperparasites and high spore mortality have likely been crucial for the evolution and maintenance of iteroparity in parasites with low dispersal rates.

100 Introduction

Specialized parasites that interact with a single or narrow spectrum of hosts tend to have fascinating life‐histories, because virulence and defence traits are likely to have been shaped by co‐evolutionary arms races (Poulin 2007; Schmid‐ Hempel 2011). This is particularly true for parasites that have evolved ways to manipulate host behaviour, so that dying hosts express extended phenotypes for the benefit of parasite reproductive success (e.g. Poulin 2010; Hoover et al. 2011). The fungal hypocrealean genus Ophiocordyceps (formerly Cordyceps) is well known for attacking specific hosts from diverse insect orders (Sung et al. 2007). Several lineages have evolved species that attack ants (Evans 1982b) leading to manipulative extended phenotypes, that make infected ants leave their nests to die and disperse spores in ways that serve parasite fitness (Andersen et al. 2009). Ant colonies are peculiar hosts for parasites. Following a high mortality rate at the founding stage, mature colonies are typically long‐lived and experience low extrinsic mortality. The high density of continuously interacting individuals within colonies implies that infection risks are high (Hamilton 1987; Sherman et al. 1988), but also that selection for efficient prophylactic defences has been strong (Schmid‐Hempel 1998; Naug & Camazine 2002; Boomsma et al. 2005). Recent reviews (Cremer et al. 2007; Hughes et al. 2008; Cremer & Sixt 2009) have emphasized that behavioural forms of social immunity are normally very efficient, so that ant parasites pose a limited threat for escalating epidemics within colonies. Thus, even though individual ants may die from infection, disease‐induced colony mortality is low (Hughes et al. 2008). Horizontal disease transmission requires the introduction of parasite propagules to uninfected nests. This process may not be very efficient as territoriality often limits overlap between infected and susceptible colonies (Boomsma et al. 2005) and propagules can often only reach the modest percentage of workers that are out foraging (typically between 10‐25 % of the workers; Mirenda & Vinson 1981; Porter & Jorgensen 1981; MacKay 1985). However, as chronically infected ant colonies tend to be long‐lived, a combination of frequent vertical (nestmate to nestmate) infection and rare

101 horizontal transmission across colonies appears to have secured stable host‐ parasite interactions in ants (Marikovsky 1962; Charney 1969; Schmid‐Hempel 1998: appendix 2, p.291‐324; Yanoviak et al. 2008). Such situations of stable coexistence between hosts and parasites are examples of colony‐level disease tolerance rather than disease resistance (Miller et al. 2005, 2006). Here we set out to examine the dynamics of the interaction between ant hosts and Ophiocordyceps parasites, which all available evidence show is a highly specialized, in tropical forests. The fungus manipulates workers to leave their nest to die close to their host colony in high‐density graveyards that may persist on the same location for years, offering the advantage that mortality rates due to chronic parasitism can be estimated (Evans 1974; Evans & Samson 1984; Sanjuán et al. 2001; Pontoppidan et al. 2009). Apart from the intriguing extended phenotype adaptations that allow the fungus to control ant behaviour, Ophiocordyceps fungi that exploit ants are also unusual in that the major growth phase and all parasite reproduction occurs long after host death. The fruiting body of the parasite has a latency period of at least two weeks before it can reproduce (shoot spores) for the first time, and the fungus secures the dead host‐ ant body so efficiently that it can continue with successive bouts of reproduction without succumbing to decay (Andersen et al. 2009). This implies that Ophiocordyceps have life‐history traits reminiscent of perennial plants, including traits such as age at first reproduction and allocation to current versus future reproduction that have been shaped by selection and are likely to be linked to rates of ageing and investment in somatic repair (Harper 1977). In a classic paper, Charnov and Schaffer (1973) showed that iteroparous life cycles with continuing investment in somatic tissue can only evolve when juvenile mortality is high relative to adult mortality. To our knowledge, the applicability of this logic has never been explicitly tested in fungi (where iteroparous fruiting bodies are rare with the exception of some saprotrophic fungi; Moore et al. 2008), but available natural history data suggest that Ophiocordyceps may well have the appropriate combination of traits for this conceptual framework to apply. Somatic investment to secure continued growth of the fruiting body is substantial in the only species studied in detail so far, O. unilateralis s.l. (Andersen et al. 2009), and Ophiocordyceps spores are fragile in

102 general, and easily killed by UV light and desiccation (Evans et al. 2011). However, the Charnov and Schaffer model would not be supported if, in spite of investments in somatic maintenance, only spores produced shortly after sexual maturity of fruiting bodies would pass on genes to future parasite generations. This seems a distinct possibility because a variety of fungal hyperparasites colonize the developing stalks and fruiting bodies and potentially cause effective castration (Evans 1982a). It is therefore essential to know the relative rates at which fruiting bodies become reproductively dysfunctional in their early phases of development. Applying iteroparity life‐history theory to a specialized host‐parasite interaction such as Ophiocordyceps has interesting additional complications, as within colony transmission success may, paradoxically, limit between‐colony reproductive success of parasites, no matter whether spores are produced directly after sexual maturity of a fruiting body or long after that. Within colony transmission needs a minimum number of dead ants per unit of time and a particular rate of infectivity to maintain a local population of parasites, whereas host colonies need to be large enough to sustain the ensuing level of worker mortality without going extinct. When mortality happens in ‘graveyards’, this would require that these graveyards have a growth rate (i.e. net inflow or ‘birth rate’ of dead infected ants) above one or, in case of long‐term equilibrium with the population of host ants, a growth rate equal to one. In the present study, we used data from previous studies on O. unilateralis in Thailand (Andersen et al. 2009; Pontoppidan et al. 2009) and a new data set from O. camponoti­rufipedis (= O. unilateralis s.l.) from Brazil to parameterize a developmental‐stage‐structured model describing the interaction dynamics between Ophiocordyceps and its host ants. By measuring the distribution of parasite life stages and the occurrence of hyperparasitism within ant graveyards we estimated the realized parasite pressure on the ants. We show that most parasite fruiting bodies are incapable of transmitting infectious propagules because of hyperparasitism, but that iteroparous reproduction appears essential for maintaining marginally positive growth rates in ant graveyards. Our results suggest that slow development of fruiting bodies and iteroparous reproduction are likely to be adaptations that achieve long‐term persistence with host‐ant

103 colonies. This is because infection success of spores is likely to be low when new host ants are difficult to target in both time and space, so that prolonged survival of fruiting bodies increases parasite reproductive success in spite of relatively high costs of hyperparasitism.

Materials and Methods

Fieldwork The common ant Camponotus rufipes is host to the specialized parasite Ophiocordyceps camponoti­rufipedis in the Atlantic rainforests of Brazil (Evans et al. 2011). The ants form large, long‐lived colonies headed by a single queen (monogyny) that have been observed to survive at the same site for more than 10 years (R.F. de Souza & S. Robeiro, personal communication). The ants construct nests of leaves, twigs and soil, typically at the base of trees and often connected to smaller satellite nests, and forage at night along temporally stable trails. The overall distribution of O. camponoti­rufipedis has been studied since 2006 and provided the stimulus for the present focal study, undertaken in February 2011 in Mata do Paraíso, a 400 ha Atlantic rainforest nature reserve in Minas Gerais, Brazil. The forest harbours a high density and diversity of Ophiocordyceps that infect ants, of which O. camponoti­rufipedis is one of the most common (Evans et al. 2011). We identified ants infected with O. camponoti­rufipedis by searching the underside of leaves along a ca. 460 m stretch of forest path and found five graveyards (sensu Pontoppidan et al. 2009) with a high density of dead C. rufipes, each of them situated around a single host ant colony. We marked areas covering approximately the entire graveyard (graveyards 1, 2, 3) or large parts of them (graveyards 4, 5), and tagged all dead infected ants ‐ found typically on the underside of leaves and on twigs (n = 432) ‐ with pink tape around the leaf stem. After death, Ophiocordyceps parasitized ants progress through several developmental stages. For each cadaver we therefore characterized the state of parasite development as being: 1. a freshly killed ant, 2. a cadaver with a parasite stroma (stalk‐like structure that is meant to develop into a mature fruiting

104 body), 3. a cadaver with a mature parasite sexual fruiting body (ascoma), 4. a cadaver at stage 2 or 3, but hyperparasitized by other fungi, or 5. a cadaver whose status could not be identified as it had been damaged by unknown causes and lacked obvious fungal growth (see below for more detailed category descriptions). The coordinates of each graveyard were obtained with a Garmin etrex GPS and mapped in GoogleEarth.

Measuring fungal reproduction To estimate the infectivity of mature parasite fruiting bodies, 15 dead ants with developed sexual reproductive bodies (ascomata) were collected with the leaves they were attached to and suspended on a wooden platform above a microscope slide in the forest close to, but outside, a graveyard of dead infected ants. Spore discharge takes place during the night and microscope slides were therefore checked on the three following mornings for deposition of the highly distinctive spore clouds, which are visible to the naked eye (Evans et al. 2011). After this collection period, the 15 parasitized ants with mature fruiting bodies were brought to the lab, in addition to 16 newly collected dead ants with mature fruiting bodies. All 31 ants were individually attached with Vaseline to the lid of a Petri dish with a microscope slide or agar at the bottom and placed in a dew chamber with 100% relative humidity for 18 hours (18.00 to 12.00). These lab‐generated microscope slides and agar plates were checked for spore deposits every morning for 4‐6 days.

Parameterizing an age­structured model for cadaver­turnover in graveyards From fieldwork in Thailand on a similar system of O. unilateralis s.l. infecting Camponotus leonardi (Andersen et al. 2009) and from observations of newly infected ants in Brazil we estimated the duration of the different parasite life stages. The first freshly killed stage, from death to the first signs of a stroma, takes ca. 4 days, for parasites in the stromal stage we arrived at a wide range of stage‐duration times (7‐30 days), and we inferred that parasites with sexually mature fruiting bodies were at least 20 days old. These ranges are likely to vary across the season, as growth conditions for fungi are probably proportional to rainfall. To remain conservative when parameterizing our model (see below), we

105 therefore used a relatively crude timescale assuming that development from host death to stroma appearance takes one week, and development from stroma to sexual maturity takes four weeks. Adopting the approximate approach described above has the advantage that estimates apply to both Ophiocordyceps unilateralis infections in Thailand and O. camponoti­rufipedis infections in Brazil, allowing us to supplement missing data in our present study. Data from Thailand obtained by following 17 individuals for 10 months with intervals of two months produced approximate figures for the risk of O. unilateralis s.l. becoming hyperparasitized at a given age. We found that 30/31 O. unilateralis s.l monitored for 18 months became infected by hyperparasitic fungi. Based on our field observation we estimated that mature parasites persist for at least 4 weeks before being hyperparasitized (DPH, unpublished data). Finally, we assumed that the accumulation of freshly dead cadavers was proportional to the number of spore‐producing parasites with mature fruiting bodies, as our measurements would show that neither immature stromata nor mature but hyperparasitized fruiting bodies produced any spores. We formalized the "life‐cycle" of parasitized cadavers as illustrated in Figure 2, where arrows represent transitions between life stages. We assumed that the rate of new infections, b, is determined by the availability of live ants and is constant over time (assuming no colony growth or decline). We further assumed that cadavers in the stromal stage would be hyperparasitized at a rate

Ps and cadavers in the mature stage would be hyperparasitized at rate Pm; these would then abort the normal completion of these stages and transfer them to the effectively sterile hyperparasitized stage (see data below). Writing the number of cadavers in each class as a vector N = (Fresh, Stroma, Mature, Hyperparasitized)', we then summarized the cadaver "life cycle" as a population transition matrix given by equation 1:

 e−1 0 b 0    −1 −(0.25+Ps ) 1 − e e 0 0  A =  0 0.25 1− e−(0.25+Ps ) e−(0.25+Pm ) 0  0.25+Ps ( )  P P  0 s 1− e−(0.25+Ps ) m 1− e−(0.25+Pm ) e−0.019  0.25+Ps ( ) 0.25+Pm ( ) 

€ 106 In this matrix (Equation 1) time steps proceed in multiples of one week and development rates are such that the mean time in each class is one week for the freshly killed cadaver class, and four weeks for both the stromal and the sexually mature fruiting body stage. If we assume that the rates of transition between classes in A are constant through time then the long‐term stable stage distribution of the matrix A gives the expected distribution of cadavers observed in each class. To estimate the unknown parameters of A (b, Ps, and Pm), we found the values that minimized the sum of squared differences between the expected stable stage distribution of A and the stage distribution of cadavers observed in the field (Table 2). This allowed us to subsequently estimate the graveyard growth rate, λ, accounting for the unhyperparasitized part of the graveyard, as the dominant eigenvalue of the transition matrix (Leslie 1945; Lefkovitch 1965).

Adding variation in overall growth rates The growth rate of graveyards (λ) and the developmental stage distribution in graveyards depends on the assumed fungal development rate, which in turn depends on temperature and humidity (e.g. Arthurs & Thomas 2001; Arthurs et al. 2001; Hatzipapas et al. 2002). We first explored the sensitivity of the

estimated parameters b, Ps, Pm and λ to effects of variation in the assumed fungal developmental rate. This was done by varying Ophiocordyceps development rates relative to the fitted model by ± 50% from the original fitted value, by incorporating the parameter θ (ranging from 0.5 to 1.5, rather than being fixed at 1) to Equation 1, under the assumption that all parasite life stages would be affected equally. We then set up four alternative versions of Equation 1 to further explore the effects of seasonal variation in the fungal developmental rate on graveyard growth rate and the proportion of the graveyard cadavers that have escaped hyperparasitism in relation to faster (θ > 1) or slower (θ < 1) overall fungal development. We hypothesized that seasonal variation could affect fungal life history in three different ways by: 1) only affecting parasite developmental rate, 2) affecting parasite and hyperparasite developmental rates similarly so they become positively correlated and 3) affecting the inflow rate of new cadavers b by a positive correlation between time spent in the mature life stage

107 and new ants infected. The four alternative scenarios capture the different combinations of these potential effects of variation in fungal development rate. In scenario 1A and 1B the inflow rate of new cadavers is assumed to be uncorrelated with the parasite development rate, while cadaver inflow rate is assumed to be positively correlated with parasite development rate in scenario 2A and 2B. In scenario 1A and 2A hyperparasite development is uncorrelated with parasite development time, while the variables are positively correlated in scenario 1B and 2B (Table 1).

Results

Life stage distribution of dead infected ants in graveyards In the five graveyards we found a total of 432 dead infected ants; 12.5% were fresh, 12.9% carried a stroma, 6.5% were mature, 55.4% were hyperparasitized, and the remaining 12.7% were damaged with no obvious fungal growth (Fig. 1, Table 2). None of the mature parasite fruiting bodies (0/15) dispersed spores at ambient temperature and humidity at the time of our spore collection in the forest. However, after exposure to higher humidity simulating nights of heavy rainfall in the forest, 42% (13/31) of the mature parasite fruiting bodies were shooting spores in the lab.

Model fitting We fitted the stage‐structured graveyard growth model to the observed distributions of parasite life stages and performed simulations assuming that fungal developmental rates ranged from 50% to 150% of our estimated means to evaluate the sensitivity of parameter estimates to the assumed developmental rates. The estimated cadaver inflow rate b was 1.42 new cadavers per mature cadaver, varying from 0.85 to 1.75 across the total range of 0.5 to 1.5 times the mean developmental rate (Fig. 3A). The estimated probability of hyperparasitism among parasites in the stromal stage Ps was 0.55, varying from 0.31 to 0.75 across the total range of 0.5 to 1.5 times the mean developmental

108 rate (Fig. 3B). The estimated probability of hyperparasitism among parasites in

the mature stage Pm was 0.057, and relatively invariant across the range of developmental rates. This suggests that the probability of new hyperparasitism of mature parasites is very limited for the four weeks that the parasite is assumed to spend in this life stage (Fig. 3C). The graveyard growth rate λ for the model showing the best overall fit with all parameter values was 1.07 and had a very small range of variation (1.035 ‐ 1.100) across the range of 0.5 to 1.5 times the mean developmental rate, indicating that the observed stage distribution was consistent with a relatively slow graveyard growth rate (Fig. 3D).

Growth and longevity trade­off Exploring the four different scenarios for implementation of variation in overall fungal development rate (the x‐axis in Fig. 4) showed that, as expected, slow development increases the time spent in the mature, infectious stage and the likelihood of a susceptible ant coming in contact with spores, while fast development increases the likelihood of reaching the mature stage prior to hyperparasitism. In scenario 1A and 1B, where the inflow rate of new cadavers is uncorrelated with the Ophiocordyceps developmental rate, the overall growth rate of the graveyard remains > 1 for the total range of fungal development rates (dashed curve, Fig. 4A; scenario 1A and 1B give the same result, as do 2A and 2B, because they only differ in whether the probability of hyperparasitized individuals remaining in the graveyard is correlated to the parasite development rate or not, which in the model does not affect the graveyard growth rate). However, if faster fungal development also results in lower infectivity due to reduced likelihood of ant encounters (scenario 2A and 2B), then the graveyard growth peaks at even slower rates of fungal development and rapidly declines when developmental rates increase (solid curve, Fig. 4A). The four scenarios differ in the proportion of cadavers that remain free of hyperparasites (Fig. 4B). If parasite and hyperparasite developmental rates affect only the transitions among cadaver categories (1A and B; black and red curves), then faster development rates result in fewer hyperparasitized cadavers (Figure 4B). Faster developmental rates in the hyperparasite lead to faster senescence of hyperparasitized cadavers and thus a greater proportion of

109 unhyperparasitized cadavers (i.e. the difference between scenario 1A and 1B, Figure 4B). If the cadaver inflow rate is positively correlated with the time spent in the mature stage (scenario 2A and 2B, green and blue curves) the proportion of unhyperparasitized cadavers is maximized at slow developmental rates as fast development leads to relatively low replenishment of fresh, unhyperparasitized cadavers.

Discussion

Low density and limited interaction efficiency between infective parasites and susceptible hosts We found that only ca. 6.5% of the O. camponoti­rufipedis fruiting bodies were effectively producing spores, as most dead ants were sterile because they were immature (25.5%), damaged (12.7%) or hyperparasitized (55.4%) by other fungi that are not pathogens of ants. Field and lab trials further indicated that only 42% (13 out of 31 tested) of the apparent fertile fruiting bodies were shooting spores at a particular time interval, illustrating that detailed environmental conditions matter as well. Finally, upon dissection some apparently healthy O. camponoti­rufipedis cadavers were found to be invaded by larvae of small unidentified arthropods (SBA and DPH, unpublished data). This may also have reduced the probability of the parasite reaching maturity and would have moved a number of them to the sterile hyperparasitized category. This demonstrates that most cadavers are not infectious to foraging ants and implies that disease pressure at the colony‐level is much lower than the high numbers of dead graveyard ants suggest. In addition to the low number of infective parasites, only a small percentage of the ant colony members are actually available as targets for Ophiocordyceps spores, as all brood and most workers remain inside the safe nest boundaries, so that only foragers face the risk of being infected (Mirenda & Vinson 1981, Porter & Jorgensen 1981, MacKay 1985). The local interaction‐ interface between parasite and host is therefore limited, so that colony‐level infections can only be stable when graveyards continue to grow until a steady

110 state that maintains host and parasite individuals at relatively constant densities of chronic colony infection. The finding of a graveyard growth rate just above one supports such a scenario. Infecting not just workers of the same colony but also those of other colonies, would require that the founding rate of graveyards exceeds their extinction rate, i.e. that host colonies and the graveyards around them produce enough dead ants over a sufficient number of years to replace themselves by founding ‘offspring’ graveyards around uninfected host colonies. The scale and duration of our study were insufficient to obtain hard data on such population‐ level equilibria, but sheds interesting light on how parasite iteroparity helps to maintain the stability of infections within graveyards. We will therefore first evaluate our present understanding of within‐colony transmission, and then briefly address what kind of studies would be needed to comprehend transmission across colonies.

The logic of iteroparous reproduction in Ophiocordyceps It seems likely that the absence of persistent spores or alternative non‐host reservoirs is crucial for understanding iteroparity in Ophiocordyceps. Many generalist entomopathogenic fungi such as the well‐studied genera Metarhizium and Beauvaria, which are asexual anamorphs of Cordyceps‐like teleomorphs, combine rapid semelparous asexual reproduction with the production of persistent spores. They are also increasingly suspected to have ‘hidden lives’ as endophytes of leaves and in the rhizosphere, with the possibility to produce spores outside the bodies of insect hosts (St. Leger 2008; Vega 2008). This suggests that sexual reproduction in the Ophiocordyceps sexual morphs of hypocrealean fungi has somehow necessitated a life‐history focusing on producing short‐lived spores for a long time rather than longer lived spores for a short time as asexual forms do. Another specialist ant pathogen, the entomophtoralean fungus Pandora infecting Formica wood ants in Europe, also has a semelparous strategy of rapid asexual reproduction (Marikovsky 1962). As with Ophiocordyceps, the Pandora hosts are manipulated into biting vegetation close to ant trails just before dying, but spores develop quickly from conidial mats on the ant body surface. These normally only survive for a few days, but the

111 spores produced towards the end of the ant foraging season tend to be durable (J. Malagocka & A.B. Jensen, personal communication) and may thus play a decisive role in maintaining colony‐level infections year after year. While it appears logical that tropical rainforest habitat does not require dormant resting spores, this does not necessarily imply that spores should be so short‐lived as those of Ophiocordyceps. Protection of spores by e.g. pigmentation has evolved repeatedly in fungi (Butler & Day 1998; Bell et al. 2009) so it seems likely that Ophiocordyceps spores that would have remained viable for weeks rather than days could have evolved. However, selection for increased viability would be unlikely to arise in habitats where frequent torrential rains would tend to wash spores away from the territories of the foraging ants. If that is so, the key question remaining is how parasite iteroparity secures a sufficiently high infection rate with ephemeral spores. Our model offers suggestions for the selection forces that are ultimately responsible for the origin and maintenance of iteroparity in Ophiocordyceps. As outlined in the Introduction, iteroparous reproduction tends to be evolutionary stable when externally imposed juvenile mortality is high relative to adult mortality (Charnov & Schaffer 1973). In addition to short lived spores, also the immature fruiting body stages appear to be highly vulnerable, but here the external factors are biotic rather than abiotic, because hyperparasitism risk is high in the stromal parasite life stage (ca. 55%; Fig. 3B), but expected to be very low in the mature life‐stage (ca. 5.5%; Fig. 3C). This is under the assumption that the mature life stage has a duration of one month; in the field in Thailand mature fruiting bodies were observed to accumulate hyperparasites after this time period but we speculate that spore production may have ceased at this stage. A low risk of hyperparasitism may well be related to mature fruiting bodies expressing a much more efficient immune defence than the rapidly growing stromata, because it takes time for the growing parasite mycelium to compartmentalize the dead host body into specific fungal tissues with complementary roles in protecting the elaborate fruiting body structures that produce the spores (Andersen et al. 2009). Fungal immune defences are poorly understood (but see e.g. Soanes & Talbot 2010), but Ophiocordyceps fungi are known to produce a range of secondary metabolites that may be relevant for

112 maintaining cadavers (Isaka et al. 2005). Thus, the likelihood for spores to survive, infect and produce a stoma is very low, but a fruiting body that has made it towards maturity is worth maintaining for a long time. Though not known in any detail, it is probable based on morphological structures such as stalks and spore producing bodies that iteroparity also occurs in a range of the other Clavicipitaceaeous fungi, such as those infecting lepidoteran and coleopteran larvae and spiders (Sung et al. 2007). While our knowledge of these groups is cursory it is likely that the life‐histories of these parasites are also characterized by a high degree of host specificity and limited contact between hosts and infective spores, and also these fungi appear to have evolved and be most diverse in (sub)tropical regions. This suggests that the evolution and maintenance of iteroparity in these obligate insect pathogens is primarily related to climate and host specificity, and that ant hosts have merely required the additional evolution of the well‐known extended phenotypes that manipulate infected workers to leave their nests. Our model also explored whether and how it matters that the time spent in the mature parasite stage is positively correlated with cadaver inflow rate (scenario 2A and 2B) or not (scenario 1A and 1B). Such correlations are likely to exist because the number of new infected ants should be some monotonously increasing function of the number of spores produced in graveyards, which in turn should be positively correlated with the reproductive life span of mature fruiting bodies. As lifespan would likely trade‐off against development rate, this suite of traits would then also be accompanied by slow development. Our model confirms that positive correlations between fruiting body life span and cadaver inflow rate maximize graveyard growth and the proportion of cadavers that escape being hyperparasitized, provided development is slow (Fig. 4). However, when fruiting body life span and cadaver inflow rate are not correlated, faster rates of development appear to be optimal for parasite reproduction (Scenario 1A and 1B). We believe that correlations between parasite and hyperparasite developmental rates are likely to occur as environmental fluctuations in temperature and humidity may affect different fungal species similarly. However, such correlations may also vary considerably because hyperparasite

113 growth is generally faster than Ophiocordyceps growth, due to the much simpler morphology of the hyperparasites.

Graveyard growth and disease spread across spatial scales Our estimates of graveyard growth rates just above 1 are consistent with colony‐ specific aggregations of dead ants being sustainable, with each mature parasite on average producing slightly more than one new mature parasite. To appreciate the significance of this result it is important to realize that the colony, not the individual ant, is the host for parasites such as Ophiocordyceps (Sherman et al. 1988; Schmid‐Hempel 1998: p. 204). If ant workers die close to their nest and end up infecting their younger siblings, this is equivalent to vertical transmission (Boomsma et al. 2005; Cremer et al. 2007; Hughes et al. 2008). By contrast, true horizontal transmission would then be restricted to spores produced by parasites of one colony infecting workers of another colony. This could either be achieved by rare infected workers dying much further away from their colony than the ‘resident’ graveyard, or to spores produced in graveyards occasionally dispersing over much longer distances. Long distance spore dispersal seems unlikely as Ophiocordyceps spores are heavy and not easily dispersed by wind. It could be possible that vectoring occurs but no evidence is known. However, if horizontal transmission would primarily depend on the movement of the infected ants themselves, this would suggest the intriguing possibility of disruptive selection for both short and (occasionally) long‐distance dispersal of parasite extended phenotypes. Infected ants should then either die very close or relatively far from their colony because ant territories are geographical mosaics with most if not all interactions being restricted to nearest neighbor colonies (Leston 1973; Majer 1993; Blüthgen & Stork 2007). Future studies may address this by estimating the gene flow within and between graveyards and by looking at the genetic diversity of the dead hosts and their parasites.

114 Acknowledgments

We thank Roberto Barreto at the Federal University of Viçosa, Minas Gerais, Brazil for kind hospitality. We are grateful to Anna Mosegaard Schmidt and Raquel Loreto for discussion, Jørgen Eilenberg for comments on an earlier version of the manuscript and Gösta Nachman for suggestions for improvement of the model. SBA was funded by a PhD. Scholarship from the Science Faculty of the University of Copenhagen, and JJB and SBA were supported by the Danish National Research Foundation. HCE and SLE were funded by the Brazilian science foundation (CNPq). DPH was funded by an Outgoing International Marie Curie Fellowship.

115 References

Andersen S.B., Gerritsma S., Yusah K.M., Mayntz D., Hywel‐Jones N.L., Billen J., Boomsma J.J. & Hughes D.P. (2009). The life of a dead ant: the expression of an adaptive extended phenotype. The American Naturalist, 174, 424‐ 433. Arthurs S. & Thomas M.B. (2001). Effects of temperature and relative humidity on sporulation of Metarhizium anisopliae var. acridum in mycosed cadavers of Schistocerca gregaria. Journal of Invertebrate Pathology, 78, 59‐65. Arthurs S.P., Thomas M.B. & Lawton J.L. (2001). Seasonal patterns of persistence and infectivity of Metarhizium anisopliae var. acridum in grasshopper cadavers in the Sahel. Entomologia Experimentalis Et Applicata, 100, 69‐ 76. Bell A.S., Blanford S., Jenkins N., Thomas M.B. & Read A.F. (2009). Real‐time quantitative PCR for analysis of candidate fungal biopesticides against malaria: Technique validation and first applications. Journal of Invertebrate Pathology, 100, 160‐168. Blüthgen N. & Stork N.E. (2007). Ant mosaics in a tropical rainforest in Australia and elsewhere: a critical review. Austral Ecology, 32, 93‐104. Boomsma J.J., Schmid‐Hempel P. & Hughes W.O.H. (2005). Life histories and parasite pressure across the major groups of social insects. In: Insect evolutionary ecology (eds. Fellowes MDE, Holloway GJ & J. R). Royal Entonomological Society. Butler M.J. & Day A.W. (1998). Fungal melanins: a review. Canadian Journal of Microbiology, 44, 1115‐1136. Charney W.P. (1969). Behavioral and morphological changes in carpenter ants harboring Dicrocoeliid metacercariae. American Midland Naturalist, 82, 605‐611. Charnov E.L. & Schaffer W.M. (1973). Life‐history consequences of natural selection: Cole's result revisited. The American Naturalist, 107, 791‐793. 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 B: Biological Sciences, 364, 129‐142. Evans H.C. (1974). Natural control of arthropods, with special reference to ants (Formicidae), by fungi in tropical high forest of Ghana. J Appl Ecol, 11, 37‐ 49. Evans H.C. (1982a). Entomogenous fungi in the tropical forest ecosystems: an appraisal. Ecological Entomology, 7, 47 ‐ 60. Evans H.C. (1982b). Entomogenous fungi in the tropical forest ecosystems: an appraisal. Ecological Entomology, 7, 47‐60. Evans H.C., Elliot S.L. & Hughes D.P. (2011). Hidden diversity behind the zombie‐ ant fungus Ophiocordyceps unilateralis: four new species described from carpenter ants in Minas Gerais, Brazil. PLoS ONE, 6, e17024. Evans H.C. & Samson R.A. (1984). Cordyceps species and their anamorphs pathogenic on ants (Formicidae) in tropical forest ecosystems II. The

116 Camponotus (Formicinae) complex. Transactions of the British Mycological Society, 82, 127‐150. Hamilton W.D. (1987). Kinship, recognition, disease, and intelligence: constraints of social evolution. In: Animal societies: Theories and facts (ed. Ito Y). Japan Scientific Societies Press, pp. 81‐102. Harper J.L. (1977). Population biology of plants. Academic Press, London. Hatzipapas P., Kalosaka K., Dara A. & Christias C. (2002). Spore germination and appressorium formation in the entomopathogenic Alternaria alternata. Mycological Research, 106, 1349‐1359. Hoover K., Grove M., Gardner M., Hughes D.P., McNeil J. & Slavicek J. (2011). A gene for an extended phenotype. Science, 333, 1401. Hughes D.P., Pierce N.E. & Boomsma J.J. (2008). Social insect symbionts: evolution in homeostatic fortresses. Trends Ecol Evol, 23, 672‐677. Isaka M., Kittakoop P., Kirtikara K., Hywel‐Jones N.L. & Thebtaranonth Y. (2005). Bioactive substances from insect pathogenic fungi. Accounts of Chemical Research, 38, 813‐823. Lefkovitch L.P. (1965). The study of population growth in organisms grouped by stages. Biometrics, 21, 1‐18. Leslie P.H. (1945). On the use of matrices in certain population mathematics. Biometrika, 33, 183‐212. Leston D. (1973). The ant mosaic, tropical tree crops and the limiting of pests and diseases. . Pest Articles and News Summaries, 19, 311–341. MacKay W.P. (1985). A comparison of the energy budgets of three species of Pogonomyrmex harvester ants (Hymenoptera: Formicidae). Oecologia, 66, 484‐494. Majer J.D. (1993). Comparison of the arboreal ant mosaic in Ghana, Brazil, Papua New Guinea and Australia ‐ its structure and influence on arthropod diversity. In: Hymenoptera and Biodiversity (eds. LaSalle J & Gauld ID). Wallingford: CAB International, pp. 115‐141. Marikovsky P.I. (1962). On some features of behaviour of the ants Formica rufa L. infected with fungous disease. Insectes Sociaux, 9, 173‐179. Miller M.R., White A. & Boots M. (2005). The evolution of host resistance: Tolerance and control as distinct strategies. J Theor Biol, 236, 198‐207. Miller M.R., White A. & Boots M. (2006). The evolution of parasites in response to tolerance in their hosts: the good, the bad, and apparent commensalism. Evolution Int J Org Evolution, 60, 945‐56. Mirenda J.T. & Vinson S.B. (1981). Division of labour and specification of castes in the red imported fire ant Solenopsis invicta buren. Animal Behaviour, 29, 410‐420. Moore D., Gange A.C., Gange E.G. & Boddy L. (2008). Fruit bodies: their production and development in relation to environment. In: Ecology of saprotrophic Basidiomycetes (eds. Boddy L, Frankland JC & van West P). Academic Press. Naug D. & Camazine S. (2002). The role of colony organization on pathogen transmission in social insects. Journal of Theoretical Biology, 215, 427‐ 439. Pontoppidan M.‐B., Himaman W., Hywel‐Jones N.L., Boomsma J.J. & Hughes D.P. (2009). Graveyards on the move: the spatio‐temporal distribution of dead Ophiocordyceps‐infected ants. PLoS ONE, 4, e4835.

117 Porter S.D. & Jorgensen C.D. (1981). Foragers of the harvester ant, Pogonomyrmex owyheei: a disposable caste? Behavioral Ecology and Sociobiology, 9, 247‐256. Poulin R. (2007). Evolutionary ecology of parasites. Princeton University Press. Poulin R. (2010). Parasite manipulation of host behavior: An update and frequently asked questions. In: Advances in the Study of Behavior (ed. Brockmann HJ). Academic Press, pp. 151‐186. Sanjuán T., Henao L.G. & Amat G. (2001). Distribución espacial de Cordyceps spp. (Ascomycotina: Clavicipitaceae) y su impacto sobre las hormigas en selvas del piedemonte amazónico de Colombia. Revista de Biología Tropical, 49, 945‐955. Schmid‐Hempel P. (1998). Parasites in social insects. Princeton University Press. Schmid‐Hempel P. (2011). Evolutionary Parasitology. Oxford University Press. Sherman P.W., Seeley T.D. & Hudson K.R. (1988). Parasites, Pathogens, and Polyandry in Social Hymenoptera. The American Naturalist, 131, 602‐610. Soanes D.M. & Talbot N.J. (2010). Comparative genome analysis reveals an absence of leucine‐rich repeat pattern‐recognition receptor proteins in the kingdom fungi. PLoS ONE, 5, e12725. St. Leger R.J. (2008). Studies on adaptations of Metarhizium anisopliae to life in the soil. Journal of Invertebrate Pathology, 98, 271‐276. Sung G.H., Hywel‐Jones N.L., Sung J.M., Luangsa‐ard J.J., Shrestha B. & Spatafora J.W. (2007). Phylogenetic classification of Cordyceps and the clavicipitaceous fungi. Stud Mycol, 57, 5‐59. Vega F.E. (2008). Insect pathology and fungal endophytes. Journal of Invertebrate Pathology, 98, 277‐279. Yanoviak S.P., Kaspari M., Dudley R. & Poinar G.J. (2008). Parasite‐induced fruit mimicry in a tropical canopy ant. The American Naturalist 171, 536‐544.

118 Figure Legends

Figure 1. Aerial photo (from GoogleEarth) of the sampling area with the five graveyards marked and the distribution of parasite life stages plotted as pie charts. A total of 432 dead infected ants were encountered, distributed with 41, 35, 44, 149 & 163 individuals, respectively, in graveyard 1‐5.

Figure 2. Idealized parasite life‐cycle. Boxes indicate life stages (fresh, stroma, mature and hyperparasitized) and arrows indicate transitions between stages. New cadavers enter the system with birth rate b and remain in the ‘fresh’ stage for a week on average. They then move to the ‘stroma’ life stage and stay there for an average of four weeks, during which a proportion is lost to the

‘hyperparasitized’ life stage at rate Ps. Those individuals that move to the mature stage spend on average four weeks there, during which a proportion is lost to the

‘hyperparasitized’ life stage at rate Pm. Individuals in the hyperparasitized stage remain on here for an average of 52 weeks before being lost.

Figure 3. Sensitivity analysis of the estimated life‐history parameters based on the stage structured graveyard growth model that fitted the empirical data best. Panel A: the new cadaver inflow rate b, panel B: the hyperparasitism rate in the

stromal life state Ps, panel C: the hyperparasitism rate in the mature life stage Pm, Panel D: the graveyard growth rate λ. The variation in fungal developmental rate from 50% to 150% is plotted along the x‐axes, relative to the average fungal development rate that was estimated from the field data (here represented by the relative value of 1). Dashed lines connect the expected mean for the y‐axis estimate with this overall mean development rate.

Figure 4. Graveyard growth rate λ (panel A) and proportion of cadavers un‐ hyperparasitized (panel B) as a function of developmental rate (x‐axis) for the four different modelled scenarios. The variation in fungal developmental rate from .25% to 400% is plotted along the x‐axes, relative to the average fungal development rate that was estimated from the field data (here represented by the relative value of 1). A. The graveyard growth rate, which only accounts for

119 the unhyperparasitized individuals, is identical in scenario 1A and 1B, and is >1 across all developmental rates but peaks at fast to intermediate developmental rate. Scenario 2A and B are also identical, with negative growth rates at fast development rates but peaks with growth rates >1 at intermediate to slow developmental rates. B. The four scenarios differ in the proportion of un‐ hyperparasitized cadavers across the developmental range. Scenario 1A and 1B have high rates of hyperparasitism at slow developmental rates. Note that as development rates increase, a greater proportion of cadavers escape hyper‐ parasitism in scenario 1B due to the faster senescence of the hyperparasitized cadavers. Scenario 2A and 2B show have increasing rates of hyperparasitism as the developmental rate increases due to the relative decrease in the inflow of new cadavers.

120 Figures

Figure 1

Figure 2

121 Figure 3

122

Figure 4

123

Tables

Table 1 Four alternative scenarios of the impact of variation in fungal development rate for Ophiocordyceps and hyperparasitic fungi. In scenario 1A and B the infection rate (b) is uncorrelated with parasite developmental rate while the rates are correlated in scenario 2A and B, meaning that the time the parasite spends in the mature life stage, as determined by development rate, is positively correlated with the rate at which new infected individuals appear. In scenario 1A and 2A the hyperparasite developmental rate is uncorrelated with parasite development rate while the rates are correlated in scenario 1B and 2B, implying that environmental factors determining fungal growth affect the parasite and hyperparasites in the same way.

124

125 Table 2 Parasite life‐stage distribution across the five graveyards with numbers of sampled cadavers in each of the four categories.

Graveyard Fresh Stroma Mature Hyperparasitized Other Total

1 26 13 5 93 12 149 2 7 6 1 15 6 35

3 4 10 8 17 5 44 4 8 4 2 24 3 41 5 9 23 12 90 29 163

126 CHAPTER 4

HOST SPECIFICITY OF PARASITE MANIPULATION –ZOMBIE ANT DEATH LOCATION IN THAILAND VS. BRAZIL

127 128 Host specificity of parasite manipulation –zombie ant death location in Thailand vs. Brazil

Sandra B. Andersen1 & David P. Hughes2

1Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark 2Department of Entomology and Department of Biology, Penn State University, PA 16802, USA

Invited Addendum to:

Andersen SB, Gerritsma S, Yusah KM, Mayntz D, Hywel‐Jones NL, Billen J, Boomsma JJ, Hughes DP (2009) The life of a dead ant: the expression of an adaptive extended phenotype. The American Naturalist 174:424‐433

Key words: Parasite manipulation, host specificity, zombie ants

Correspondence to:

Sandra Breum Andersen, Email: [email protected] and David P. Hughes, Email: [email protected]

IN PROOF, non‐peer reviewed article

129 Abstract

Recently we presented how Camponotus ants in Thailand infected with the fungus Ophiocordyceps unilateralis are behaviorally manipulated into dying where the conditions are optimal for fungal development. Death incurred in a very narrow zone of space and here we compare this highly specific manipulation with a related system in Brazil. We show that the behavioral manipulation is less fine‐tuned and discuss the potential explanations for this by examining differences in ant host and environmental characteristics.

Text

Parasite manipulation of host behavior is an intriguing example of parasite adaptation. The change of host behavior is considered to be an extended phenotype of the parasite, as it can be explained as an expression of parasite genes in the host phenotype to increase parasite fitness.1‐3 One of the most dramatic examples of a parasite extended phenotype is the manipulation of ant behavior by the fungus Ophiocordyceps unilateralis sl.4,5

In a recent study we showed that the manipulation of the host ants by O. unilateralis sl is highly specific and beneficial to parasite fitness, thereby fulfilling the criteria of an extended phenotype 6. Infected Camponotus leonardi worker ants leave their nest in the canopy and seek out the underside of a leaf in the undergrowth, bite into a leaf vein and die. The parasite then quickly colonizes the ant and grows for > 2 weeks before achieving reproduction. The death location of the ants was found to be far from random: dead infected ants were located 25.20 ± 2.46 SE cm above the ground, where the humidity and temperature were optimal for fungal growth, and on the north‐northwest side of the plant biting onto a vein of the leaf. Parasites in the dead ants relocated from this ‘manipulative zone’ did not grow, confirming the adaptive value of the behavioral change. O. unilateralis was until recently believed to be a globally distributed species, but morphological studies in Brazil revealed a species complex with high host specificity.7 Here we wanted to explore the data on adaptive

130 manipulation further and infer the role of host characteristics, by comparison of the data on the death position from Thailand with that of ants in the related system of Ophiocordyceps camponoti­rufipedis (= unilateralis s.l.) infecting the ant Camponotus rufipes in Brazil. In addition, as hyperparasitism by mycoparasites is extensive and likely very costly for the parasite, we wished to elucidate whether the position where the ant was manipulated to die had an effect on the risk of hyperparasitism.

Fieldwork took place in February 2011 in Mata do Paraíso, a 400 ha Atlantic rainforest nature reserve in Minas Gerais, Brazil. The location and condition of 132 dead C. rufipes ants infected with O. camponoti­rufipedis was registered along a 460 m stretch of forest path. For each dead ant we noted the height above ground, the orientation of the ant (which compass direction the head was pointing) and the parasite life stage as one of the following four categories: 1. a freshly killed ant (n = 26), 2. a cadaver with a stroma (stem‐like structure that is the precursor to the production of a mature fruiting body; n = 28), 3. a cadaver with a mature sexual fruiting body (ascoma; n = 19), 4. a cadaver at stage 2 or 3, but hyperparasitized by other fungi (n = 59). The data were analyzed in JMP 9.0.2 for Mac and PAST 1.80 (available as free download at http://folk.uio.no/ohammer/past/) and compared with the data on height and orientation of dead C. leonardi ants infected with O. unilateralis s.l. obtained in Thailand as reported in Andersen et al.6

At the Brazilian site we found no difference in the height of dead infected ants between the parasite life stages or in the variation around the mean. We suggest that there is no relationship between the height at which ants are manipulated to die and the probability of the fungus reaching maturity or subsequently it self

becoming a host to hyperparasites (Fig.1A, One‐Way ANOVA Brazil: F3,128: 0.329, p = 0.804, Levene’s test of unequal variance: F3,128: 2.390 p = 0.0718). The dead infected ants were found higher up in Brazil than in Thailand and with a greater variance around the mean (Fig.1B, t‐test assuming unequal variances t174.5: ‐

15.506, p ≤ 0.0001, Levene’s test of unequal variance: F1,181: 89.555 p ≤ 0.0001). This may suggest a less fine tuned host manipulation in Brazil, potentially as a

131 consequence of a wider height range of appropriate growth conditions. A temperature and humidity profile by height from the ground was not measured in Brazil, but may likely differ from Thailand, as the forest was denser. While both locations experience a wet and a dry season, the dry season in the Atlantic Rainforests of Minas Gerais, Brazil is correlated with lower temperatures and high humidity 8, which may lower the risk of parasite desiccation. In addition, no pattern in the orientation of the ants was found in Brazil, in contrast to that of the Thai ants (Fig. 2). Note that the direction of the ant head is depicted, in contrast to location around the plant as in Andersen et al. 6, so this is opposite to what we previously reported for Thailand as the majority of ants were facing the plant in Thailand. We do not know whether it is the location around the plant or the direction of the head that is relevant to the ant death location in Thailand, but neither appeared to matter in Brazil. We suggest that this is because infected ants in Brazil likely bite during the night, when the ants are most active, in contrast to at noon in Thailand 9. This would eliminate the potential for using solar cues for orientation, e.g. by shade‐seeking behavior. The ant C. rufipes is the dominant ant species at the field site in Brazil, where they nest on the ground in contrast to the canopy dwelling C. leonardi, which we recorded from over 20 m up in the canopy. How this affects the parasite strategy is unknown, but it may make the host more accessible to the parasite in Brazil, as the dead infected ants are positioned in close proximity to the foraging trails of the host (R. Loreto et al. in preparation). However, the social immunity of the ant host 10 is likely still the main challenge for the parasite, selecting for persistence in the environment to ensure transmission (S.B. Andersen et al. in preparation). In all, the comparison of the two host‐parasite systems suggests that both parasites are highly adapted to their hosts but that environmental and host differences confer different strengths of selective pressure on the specificity of host manipulation.

132 Acknowledgements

We thank Roberto Barreto and Simon Elliot at the Federal University of Viçosa, Minas Gerais, Brazil, and Harry C. Evans at the Federal University of Viçosa and CAB International, Surrey, UK for their kind hospitality. We are grateful to Jacobus J. Boomsma and Raquel Loreto for discussion. SBA was funded by a PhD. Scholarship from the Science Faculty of the University of Copenhagen and DPH was funded by an Outgoing International Marie Curie Fellowship.

133 References

1. Poulin R. Manipulation of host behaviour by parasites: a weakening paradigm? Proc R Soc Lond (Biol) 2000; 267:787‐92. 2. Moore J. Parasites and the behavior of animals. Oxford University Press, 2002. 3. Dawkins R. The extended phenotype. Oxford University Press, 1982. 4. Evans HC, Samson RA. Cordyceps species and their anamorphs pathogenic on ants (Formicidae) in tropical forest ecosystems II. The Camponotus (Formicinae) complex. Transactions of the British Mycological Society 1984; 82:127‐50. 5. Pontoppidan M‐B, Himaman W, Hywel‐Jones NL, Boomsma JJ, Hughes DP. Graveyards on the move: the spatio‐temporal distribution of dead Ophiocordyceps‐infected ants. PLoS ONE 2009; 4:e4835. 6. Andersen SB, Gerritsma S, Yusah KM, Mayntz D, Hywel‐Jones NL, Billen J, Boomsma JJ, Hughes DP. The life of a dead ant: the expression of an adaptive extended phenotype. Amer Nat 2009; 174:424‐33. 7. Evans HC, Elliot SL, Hughes DP. Hidden diversity behind the zombie‐ant fungus Ophiocordyceps unilateralis: four new species described from carpenter ants in Minas Gerais, Brazil. PLoS ONE 2011; 6:e17024. 8. Carmo Pinto Sd, Venâncio Martins S, Barros NFd, Teixeira Dias HC. Produção de serapilheira em dois estádios sucessionais de floresta estacional semidecidual na Reserva Mata do Paraíso, em Viçosa, MG. Revista Árvore 2008; 32:545‐56. 9. Hughes DP, Andersen SB, Hywel‐Jones NL, Himaman W, Billen J, Boomsma JJ. Behavioral mechanisms and morphological symptoms of zombie ants dying from fungal infection. BMC Ecol 2011; 11:13. 10. Cremer S, Armitage SAO, Schmid‐Hempel P. Social immunity. Curr Biol 2007; 17:R693‐R702.

134 Figure Legends

Figure 1 The height above ground of dead infected ants. Panel A shows that there were no significant differences between the height of dead ants in four different life stages in Brazil (fresh, stroma, mature and hyperparasitized (HP); mean height cm ± SD). The death height therefore does not affect which parasites reach maturity and which succumb to hyperparasitic fungi. Panel B shows the difference between the height at which dead infected ants are found in Brazil and Thailand, where the ants die at a greater height in Brazil with a greater variance around the mean (mean height cm ± SD).

Figure 2 The direction of the heads of dead infected ants in Brazil and Thailand. In Brazil there was no pattern in the direction while the dead ants pointed towards South in Thailand. The blue slices show the number of dead ants in a given direction while the red lines indicate the mean head direction and the 95% confidence interval.

135

Figures

Figure 1

Figure 2

136 CONCLUSIONS AND

PERSPECTIVES

137 138 Conclusions and perspectives The chapters of this thesis cover different aspects of heterogeneity in the symbiotic interactions between ants and microbes. While the systems dealt with are quite different there are some overall conclusions to be drawn and new questions to be addressed generated by the findings.

Genetic diversity of symbionts The genetic diversity of bacterial symbionts was addressed in chapter 1 and 2 and in both systems a range of genotypes was found in most hosts. The effects on the hosts of this diversity were however not completely clear from the obtained results. In the case of the Wolbachia bacteria, the overall consequences of infection are still not understood, yet the study presented here suggests that it may be more in the mutualistic range of the symbiosis spectrum than the parasitic, which has been the prevalent view. The unexpected location of the bacteria extracellularly in parts of the ant gut suggests that they may have an undiscovered role in the ant digestion, an inference that holds promise for further studies. The apparent interaction between the strains at certain host life stages, and the potential for the strains to have different phenotypic effects, suggests that the diversity likely will be of importance to the host. The presence of Wolbachia bacteria in the fecal droplets opens the possibility for transmission between workers and not just between queen and offspring. The partial segregation of the two strains found is in addition relevant to screening studies of various insect species, where perhaps only a leg is used for the DNA extraction. If some strains only are located in e.g. gut, they would likely be missed by such a screening approach. In the analysis of the actinomycete community composition on A. echinatior we found a low diversity, with overall just two segregating strains of Pseudonocardia dominating. The results thus confirmed the findings of the initial studies on the issue, by culturing of bacteria. It however also highlighted some of the challenges of the analysis of 454 pyrosequencing data, in both the sample collection step and the data analysis. While only a small fragment of cuticle was collected, with a cover of the target bacteria visible to the naked eye, a range of other confirmed and suspected contaminants from the tissue, and potentially

139 elsewhere, was present in the sequencing output of some of the samples. It is thus not too surprising that other studies sequencing the bacteria of whole ants or culturing washes of ants identify a higher bacterial and actinomycete diversity. More relevant to the questions of cooperation and conflict, and stability of mutualisms, is it that a low diversity was generally found on the laterocervical plates, the area of the cuticle where the ants apparently supply the bacteria with nutrients. This suggests that either the ants or the bacteria are able to control the community composition. The data was analysed in a range of different ways by adjusting the filtering parameters and OTU selection settings, all within the ‘standard’ range with the employed method, yet very variable results were obtained, especially for the number of Pseudonocardia OTUs present. In this case, only the subsequent manual validation against high quality sequences revealed a lower diversity. While the exact number of OTUs from a given genus may not matter in some studies, it was central to this. Given the focus on symbiont diversity in the first two chapters I was also very interested in the diversity of Ophiocordyceps infecting ants, within and between the individual ants in the graveyard. Parasite diversity within the ant could either be a welcome opportunity for recombination or result in competition with suboptimal host exploitation as outcome. Also, with‐in host diversity could open up the possibility of cheater strains, gaining more than their fair share of the reproduction by preferentially locating in the reproductive tissues. Parasite diversity in the graveyard would be equivalent to within host variation (see below), a set‐up often observed to increase parasite virulence by competition for host resources. The autonomy of the individual infected ants may however limit this effect. The dead infected ants, for which the death location data was used in chapter 4, were originally collected to study the effect of genetic diversity on host manipulation. DNA was extracted from whole individuals and the amplification of variable regions of fungal DNA (ITS 1 & 2 and elongation factor 1a) was attempted. While amplification was successful, cloning of the obtained PCR product proved challenging for unknown reasons and when accomplished, a lot of sequence diversity was found, suggesting amplification of a range of other fungi growing on or attached to the cadavers. As many fungal sequences deposited in GenBank, especially of environmental

140 samples, are not accurately named, the identification of Ophiocordyceps was difficult so in combination with limited time it was not possible to address this question further. Hopefully sequencing of pure cultures of Ophiocodyceps will allow for the design of species‐specific primers, which will ease this challenge.

Environmental impacts – field vs. lab colonies In chapter 1 and 2 the bacterial symbionts were compared between field‐ collected colonies and colonies kept in culture rooms at relatively constant conditions in regards to temperature, humidity and food for two months to 10 years. For Wolbachia, the overall density of bacteria was found to be higher in the samples from lab reared colonies compared to those from the field, which was particularly obvious for the rarer strain WSinvictaA. It also worth noting that all colonies sampled from the lab were double infected, in contrast to 63% in the field. While the sample size is not large enough to infer whether all colonies in the lab indeed are double infected it is interesting to speculate on the implications of such a scenario. This could occur if only double infected colonies survived the transition from the field to the lab, because the rarer Wolbachia strain provided some essential service needed in the new environment. Alternatively, it might be the result of what would be considered single infected colonies, if sampled in the field, becoming double infected in the lab. This could either be because WSinvictaA is actually present in all colonies in the field, but repressed in some to immeasurable amounts, while in the lab it is able to achieve a higher density. Such a mechanism could also be the reason that WSinvictaA is found in measurable amounts in the field in some colonies. Another explanation may be that the rarer Wolbachia strain may be vectored between colonies by some unknown mechanism. Unfortunately, the field‐sampled colonies for this study were not brought to the lab and field samples from the lab colonies were only available for three colonies. These have yet to be tested but the sample size either way has to be increased to conclude anything on the matter. The Pseudonocardia community of A. echinatior was found to be incredibly stable under lab conditions, considering its external location and the high proximity of colonies with different strains and species of bacteria. Interactions with Escovopsis are expected to be a major selective factor for the

141 association between attine ants and Pseudonocardia in the field. However once successfully established in the Copenhagen lab, colonies will rarely encounter Escovopsis. The elimination of this threat in the lab may remove the advantage of carrying the bacteria. It could thus be argued that in the absence of Escovopsis the cost:benefit ratio of harboring the bacteria, which are sustained by the ants at a significant cost, should favour the ants losing them entirely when kept in the lab. Unless this maintenance cost is negated by the stable environment and ad libitum feeding in the culture rooms. The continued association between the ants and bacteria could suggest that the ants are not actively controlling the interaction. This raises the interesting question of to what extent the host is capable of controlling the bacterial growth and the mechanisms available to the host for doing so and thus for maintaining a mutualistic interaction. I find the questions of how the association is controlled especially worthy of future studies. The ants apparently support the bacterial growth by nutritional secretions but how is the growth controlled to either just the laterocervical plates or up‐regulated to the whole ant in case of an Escovopsis attack? Actinobacteria in other conditions are found to produce antibiotics when starved (M. Hutchings, personal communication), suggesting that the inducement of growth, to fully cover the ant, and increased antibiotic production may not be easily correlated. Why do newly enclosed ants become fully covered in bacteria that then subsequently are limited to the laterocervical plates? And what role does competition between bacterial strains play in the specificity of the association versus e.g. host control via secretions? The bacteria may however also have yet undiscovered functions in addition to their activity against Escovopsis. Other traits could thus become relevant to the host in the laboratory setting and thereby stabilize the interaction.

The ant colony as a host A well‐recognized consequence of the social structure of an ant colony is that the entire colony, and not the individual ants, is the host of the associated symbionts. In the studies of Wolbachia and Pseudonocardia we found no within colony variation in symbiont diversity. The ants are known to acquire Pseudonocardia bacteria from the fungal garden and other colony members, which would

142 effectively maintain the colony specific community. While Wolbachia are transmitted vertically from the queen, the presence of bacteria in the fecal droplets as mentioned could serve as another route of transmission, primarily within the colony. More intriguing is the aspect of horizontal transmission of Ophiocordyceps, which requires the establishment of a new graveyard around another ant colony. When searching for mature parasites in the field we on a number of occasions found beautiful samples in areas with very low density of dead infected ants, that is, not in graveyards, while these specimens were difficult to locate within graveyards. This lead us to hypothesize that while the majority of ants die within an established graveyard, a few travel a longer distance and in this non‐graveyard area they have a higher chance of reaching maturity by avoiding the hyperparasitism plaguing the high‐density graveyards. If in proximity of another ant colony, a new graveyard may be founded. This would create meta‐population dynamics of graveyards as patches establishing and going extinct over periods of years. In addition, this would give the expectation that within graveyard parasite diversity is low as it would only take one or a few infected ants to initiate and maintain the graveyard.

143

144 Pictures Fieldwork in Brazil. 1) Field‐station in Mata do Paraiso, Minas Gerais, Brazil. The low palm next to the scooter continually attracted infected ants biting into the leaves, in spite of its location in the burning sun. 2) A freshly dead C. rufipes ant with fungal hyphae growing from the joints. 3) Commonly more than one dead ant is found on one leaf within a graveyard. Here one has bit onto another dead infected ant. 4) Dead infected ant biting onto pink tag used to mark another dead ant (photo H.C. Evans).

145

5) At a graveyard with Anna Mosegaard‐Schmidt, dead ants are marked with pink tape, over 40 cadavers were found on this tree alone (Photo D.P. Hughes). 6) Mature parasite in ant biting onto a twig. 7) Dead infected ant covered in hyperparasites.

146 ACKNOWLEDGEMENTS

The completion of this thesis had not been possible without the contributions of a long list of fantastic people.

First and foremost thank you to Koos and David, for in total almost five years of encouraging, challenging and enthusiastic supervision with countless great discussions, shaping me as the scientist I know start feeling like. Thanks to the CSE and CMEC communities and others in the building for an outstanding work environment, providing great discussions on social interactions and even greater actual social interactions: Line Vej Ugelvig, Susanne den Boer, Henrik de Fine Licht, Matthias Fuerst, Lisi Fuerst, Nana Hesler, Maj‐Britt Pontoppidan, Anna Mosegaard Schmidt, Jelle van Zweden, Dóra Huszar, Sämi Schär, Panos Sapountzis, Andreas Kelager, Anne Andersen, Luke Holman, Aniek Ivens, Birgitte Hollegaard, David Nash, Janni Larsen, Jes Søe Pedersen, Luigi Pontieri, Marlene Stürüp, Michael Poulsen, Morten Schiøtt, Patrizia d´Ettorre, Pepijn Kooij, Rachelle Adams, Rasmus Stenbak Larsen, Sanne Nygaard, Sean Byars, Sze Huei Yek, Volker Nehring, Dani Moore, Nick Bos, Sylvia Mathiasen, Charlotte Olsen, Bettina Markussen, Henning Bang Madsen, Ruth Bruus Jakobsen, Rikke Anker Jensen, Jonas Geldman, Irina Levinsky, Michael Borregaard, David Nogues, Anna‐Sofie Steensgaard, Ben Holt, Katie Marske, Christian Hof, Susanne Fritz and everyone I accidently forgot… Special thanks to Line and Henrik for commenting on the synopsis. Thanks to the more or less ‘external’ collaborators: Lars Hestbjerg and Karin Vestbjerg in the Microbial Molecular Biology group at University of Copenhagen; Mette Boye, Joanna Amenuvor and Annie Ravn Pedersen at the Veterinary lab of the Technical University of Copenhagen; Tom Gilbert from the Centre for GeoGenetics at University of Copenhagen; Matt Ferrari and Raquel Loreto at Centre for Infectious Diseases at Penn State University; Harry Evans at CABI, England; Sam Elliot from Federal University of Viçosa, Brazil; Hermogenes Fernandez‐Marín, STRI, Panama; Lisbeth Haugkrogh, Aase Jespersen and Jørgen Lützen from Zoomorphology at University of Copenhagen and Michael Williamson then at the section for Cell and Neurobiology at University of Copenhagen. Thanks to Charissa de Bekker and Roel Fleuren for support and company in encountering the American ‘culture’ of State College PA.

And last but not least many thanks to my family, friends and Patrick for providing the necessary support and distractions to function outside the university as well!

147 148 CURRICULUM VITAE Personal data

Sandra Breum Andersen 27.05 1983 Slotsgade 7, st. tv. 2200 Kbh N Tlf.nr.: 26209197 [email protected] / [email protected]

Education

2008‐ 2012 PhD, Centre for Social Evolution, University of Copenhagen 2006‐2008 MSc Biology, University of Copenhagen 2003‐2006 BSc Biolog, University of Copenhagen 1998‐2001 Mathematical highschool, Virum Gymnasium

Relevant working experience

2011 Penn State University, USA, 2 months hosted by Dr. David Hughes 2011 Fieldwork in Viçosa, Brazil 2008‐2011 Teaching assistant in ’Populationbiology’ at University of Copenhagen. 2008‐2011 NOVO science ambassador. Dissemination of science to school children 2008 Communication course arranged by Swift & Gelinde and DNS. 2008/2011 Fieldwork in Gamboa, Panama with STRI 2007 Fieldwork in Khao‐Chong, Thailand 2006 BSc‐project at University of Queensland, Australien with Prof. Ove Hoegh‐Guldberg, including 6 weeks of fieldwork at Heron Island Research Station. 2002‐2007 Guide at Øresunds Aquarium, University of Copenhagen

Conferences

2011 ESEB, Tübingen, Germany, poster presentation 2010 IUSSI, Copenhagen, Denmark, poster presentation 2010 Evolutionary potential of wild populations, Sønderborg, Denmark 2009 ESEB, Turin, Italy, poster presentation 2008 Social Insect Biology, Oulanka, Finland, oral presentation 2008 Biology of Social Insects, Tartu, Estland, oral presentation 2007 Population and Evolutionary Biology of Fungal Symbionts, Ascona, Switzerland

149 Publications, peer­reviewed

Hughes, D.P., Andersen, S.B., Hywel‐Jones, N., Himaman, W., Billen, J., Boomsma, J.J. 2011: Convulsions and lock‐jaws: behavioral mechanisms and morphological symptoms of zombie ants dying from fungal infection, BMC Evolutionary Biology (11) 1

Andersen, S.B., Vestergaard, M.L., Ainsworth, T.D., Hoegh‐Guldberg, O., Kühl, M. 2010: Acute tissue death (white syndrome) affects the microenvironment of tabular Acropora corals, Aquatic Biology (10) 1

Andersen, S.B., Gerritsma, S., Yusah, K.M., Mayntz, D., Hywel‐Jones, N.L., Billen, J., Boomsma, J.J. & David P. Hughes 2009: The life of a dead ant – the expression of an adaptive extended phenotype, The American Naturalist (174) 3

Publications, non peer­reviewed

Andersen, S.B., Hughes, D.P. 2010: Zombiemyrer på vej til kirkegården. Aktuel Naturvidenskab (4) 26‐29 (Danish popular science jounal)

Funding

2008 NOVO Scholarship for Master’s students, 36.000 DKKR 2007 Oticon foundation grant for fieldwork in Thailand, 2000 DKKR 2006 Oticon foundation, Frimodt‐Heineke foundation and Copenhagen Education grants for BSc project in Australia, in total 15.000 DKKR

Peer­reviewing experience

Reviewed for Journal of Insect Pathology

150