Discrimination Behavior in the Supercolonial Pharaoh Ant

Home , Pharaoh ant, Pseudomyrmex pallidus

FACULTY OF SCIENCE UNIVERSITY OF COPENHAGEN

CENTRE FOR SOCIAL EVOLUTION

DEPARTMENT OF BIOLOGY

PhD thesis

Luigi Pontieri

Discrimination behavior in the

supercolonial pharaoh ant

Academic advisor Jes Søe Pedersen June 2014 faculty of science university of copenhagen centre for social evolution department of biology Danish National Research Foundation Centre for Social Evolution

Discrimination behavior in the supercolonial pharaoh ant

This thesis dissertation has been submitted in accordance with the requirements for the degree of PhD, at the PhD School of The Faculty of Science, University of Copenhagen, Denmark to be defended publicly before a panel of examiners.

by Luigi Pontieri June 2014

Academic advisor Assoc. Prof. Jes Søe Pedersen Cover: artwork of queens, workers and brood of Monomorium pharaonis © Luigi Pontieri Preface

This thesis is the result of a three year PhD project carried out at the Centre for Social Evolution (CSE), Department of Biology, University of Copenhagen in Denmark under the supervision of Assoc. Prof. Jes Søe Pedersen. During my PhD period I spent three months in the laboratory of Assoc. Prof. Timothy Linksvayer, Department of Biology, University of Pennsylvania. I additionally spent a total of three weeks at the Laboratoire d’Ethologie Expérimentale et Comparée (LEEC), Université Paris 13, Sorbonne Paris Cité, hosted by Prof. Patrizia d’Ettorre. The projects in this thesis were funded by CSE via a grant from the Danish National Research Foundation (grant DNRF 57).

The thesis is comprised of a general introduction on the problem of biological invasion with a special focus on invasive ant species, the traits that makes them successful invaders and how genetic diversity can affect the recognition system and the intra- and interspecific interactions that these species present. I then present my model organisms, a summary of the main findings of my projects and suggested future directions. This is followed by four chapters of original empirical work, three of which are in preparation for submission in peer reviewed journals and one is currently submitted.

Luigi Pontieri Copenhagen, June 2014

Contents

Summaries ...... 7

Summary ...... 8

Resumé ...... 10 General introduction ...... 13

Introduced species and the problem of biological invasions ...... 14

Invasive ants ...... 14

The “invasive ant syndrome” ...... 15

The structure of unicolonial populations and the problem of recognition ...... 16

Ant recognition system ...... 18

The effect of genetic diversity on nestmate recognition and supercolony interactions . . . 18

The effect of unicoloniality on disease spread ...... 20

Model Organisms ...... 22 The Pharaoh ant. 22 Metarhizium brunneum . 26

Short summary of the thesis chapters ...... 27

Future perspectives ...... 28

References ...... 30 Chapter 1 ...... 39

Genetic diversity is more important than relatedness for nestmate discrimination in the pharaoh

ant Chapter 2 ...... 63

Unexpected fusion of genetically and chemically divergent colonies of the invasive pharaoh ant Chapter 3 ...... 91

The statistical approach to identify nestmate recognition cues Chapter 4 ...... 115

Ant colonies prefer infected over uninfected nest sites Acknowledgements ...... 137

English, Dansk Summaries 7 Summary

The majority of eusocial insect species live in small, kin structured colonies that are mutually aggressive and rarely interact. By contrast, a restricted group of ant species show a peculiar social organization called unicoloniality, where colonies can grow to vast networks of geographically separated but mutually tolerant nests, also referred to as “supercolonies”. Many unicolonial ants are invasive, as their introduced supercolonies attain huge size and cause severe economic and ecological damage, affecting in particular species composition and functioning of ecosystems. There is therefore an increasing need to understand which factors promote the ecological dominance of these species, and particularly how the discrimination of both conspecifics and heterospecifics (including parasites) might influence structure and ecological success of invasive populations. In this PhD thesis I investigated the discrimination behavior of the invasive pharaoh ant (Monomorium pharaonis) as a model for other invasive and supercolonial ant species. The pharaoh ant is one of the few ant species that can be reared in the laboratory for many generations. Furthermore, the possibility to do controlled crosses of colonies provides the unique opportunity to establish colonies of different genetic composition. These traits make this species a suitable study subject to set up behavioral experiments that aim to investigate which factors, and to which extent, might influence the inter- and intraspecific discrimination abilities of invasive ants. In the first chapter I focused on the nestmate recognition system of pharaoh ants, investigating whether the cues used for discrimination had a genetic origin and how different level of within-colony genetic diversity and relatedness influenced the discrimination abilities of the colonies. I show that, despite a general ability of discriminating nestmates from non-nestmates, the degree of relatedness between colonies did not influence the overall level of discrimination. Furthermore, I found that genetically low-diverse colonies displayed better discrimination abilities than high diverse ones and that low genetically diverse colonies discriminated high diverse non-nestmates better than vice versa. In the second chapter I investigated whether the high genetic differentiation characterizing natural colonies of pharaoh ants is sufficient to prevent unrelated colonies to fuse and how different levelsof genetic similarity shape the outcome. By pairing laboratory colonies in a fusion assay, I show that the majority of unrelated colonies fused despite high initial levels of aggression. Moreover, I also found that the initial aggression was positively correlated with the chemical and genetic distance between colony pairs, further confirming the important role of endogenous cues in the nestmate recognition of this species. The third chapter presents a methodological study on the best procedures for identifying chemical compounds used for nestmate recognition in social insects. We first compiled datasets of cuticular hydrocarbons (CHCs) and aggression between colonies of three species of ants (Formica exsecta, Camponotus aethiops and Monomorium pharaonis) and a simulated dataset. Then, using the available

8 information about the exact cues used for nestmate recognition in F. exsecta, we evaluated the power of different combinations of data transformation and chemical distance calculation in differentiating between true nestmate recognition (NMR) cues and other compounds. We found that particular combinations of statistical procedures are more effective in differentiating NMR cues from other compounds. We also developed a new method for centroid calculation that increased the power of the analysis and can therefore be used in future studies that aim to identify nestmate recognition cues in other species. In the fourth chapter I investigated the nest site preference of pharaoh ant colonies and, specifically, their ability to avoid nests containing infectious pathogens as invasive, supercolonial ants are hypothesized to be particularly prone to disease. Using binary choice tests between three types of nests, I found that migrating colonies surprisingly preferred nest sites containing nestmate corpses overgrown with sporulating mycelium of the generalist fungus Metarhizium brunneum. This unexpected finding can provide new insight into the important co-evolution of social insects and their pathogens.

9 Resumé

Størstedelen af eusociale insektarter lever i små kolonier bestående af beslægtede individer. Kolonierne er typisk aggressive over for hinanden og interagere sjældent. Der findes imidlertid en begrænset gruppe af myrearter som har en speciel social organisation kaldet unikolonialitet, hvor kolonier kan vokse til et enormt koloninetværk, som er geografisk adskilte men ikke aggressive over for hinanden. Dette fænomen kaldes også “superkolonier”. Mange unikoloniale myrer er invasive, fordi indførte superkolonier kan opnå enorme størrelser og påvirker især artssammensætningen og økosystems funktioner, og derved forårsage alvorlig økonomisk samt økologisk skade. Der er derfor, et stigende behov for at forstå, hvilke faktorer der forårsager disse arters økologiske dominans, og især hvordan diskrimination af både artsfæller og ikke artsfæller (herunder parasitter) kan have indflydelse på økologisk succes og strukturen af invasive populationer. I denne ph.d.-afhandling, undersøgte jeg diskriminationsadfærden hos den invasive faraomyre (Monomorium pharaonis) som en model for andre invasive og superkoloniale myrearter. Faraomyren er en af de få myrearter, der kan opdrættes i laboratoriet i mange generationer. Det er tilmed muligt at lave kontrollerede krydsninger mellem kolonier, hvilket giver en unik mulighed for at etablere kolonier med forskellig genetisk sammensætning. Disse egenskaber gør denne art perfekt egnet til at udføre adfærdseksperimenter, der har til formål at undersøge hvilke faktorer der påvirker diskriminationsadfærden mellem både artsfæller og ikke- artsfæller hos invasive myrer. I det første kapitel, fokuserede jeg på bofællegenkendelsessystemet hos faraomyrer. Jeg undersøgte om de kemiske forbindelser der anvendes til diskrimination var påvirket af den genetiske sammensætning, og hvordan forskellige niveauer af genetisk diversitet samt slægtskab påvirket diskriminationsevnen hos kolonierne. Jeg viser, trods en generel evne til at diskriminere bofæller fra ikke-bofæller, at det indbyrdes slægtskab mellem kolonier ikke påvirker den overordnede diskriminationsevne. Desuden har jeg fundet, at kolonier med lav genetisk diversitet viste bedre diskriminationsevnen end kolonier med høj genetisk diversitet, og at kolonier med lav genetisk diversitet diskrimineret ikke-bofæller med høj genetisk diverse bedre end omvendt. I det andet kapitel undersøgte jeg om den høje genetiske differentiering som karakteriserer naturlige kolonier af faraomyrer er tilstrækkelig til at forhindre uafhængige kolonier i at fusionere, og hvordan forskellige niveauer af slægtskab har betydning for resultatet. Ved at sætte laboratorie kolonier sammen i et fusionsforsøg, viser jeg at hovedparten af ubeslægtede kolonier fusionere trods højt aggressionsniveau i starten. Desuden fandt jeg, at aggressionen i starten af forsøget var positivt korreleret med de kemiske og genetiske afstand mellem parrede kolonier, hvilket yderligere bekræfter, at de kemiske forbindelser de selv producere spiller en vigtig rolle for bofællegenkendelse hos denne art. Det tredje kapitel præsenterer en metodisk undersøgelse af de bedste procedurer til at identificere de

10 kemiske stoffer der bruges til bofællegenkendelse hos sociale insekter. Først samlede vi datasæt af kutikulære hydrocarbons (CHCs) og aggression mellem kolonier fra tre myrer arter (Formica exsecta, Camponotus aethiops og Monomorium pharaonis) og en simuleret datasæt. Derefter brugte vi den tilgængelige viden om hvilke stoffer F. exsecta bruger til bofællegenkendelse, og derfra, evaluerede styrken af forskellige kombinationer af datatransformation og kemisk afstandsberegninger. Derved kan man skelne mellem ægte bofællegenkendelses (NMR) forbindelser og andre kemiske forbindelser som ikke har betydning for bofællegenkendelse. Vi fandt, at bestemte kombinationer af statistiske procedurer er mere effektive til at skelne NMR forbindelser fra andre kemiske forbindelser. Vi har også udviklet en ny metode til beregning af centroiden som øget styrken af analysen og som ville kunne anvendes i fremtidige undersøgelser hvis mål er at identificere bofællegenkendelses forbindelser for andre arter. I det fjerde kapitel undersøgte jeg faraomyrekolonier bosteds præferencer og om de specifikt har evnen til at undgå bosteder som indeholder smittefarlige patogener som invasive supercoloniale myrer kan være særligt sårbare over for. Ved at teste hvilket bosted myrekolonierne valgte når de fik valget mellem to ud af tre typer af bo, fandt jeg overraskende, at migrerende kolonier foretrak bosteder som indeholdt døde kroppe af bofæller som var overgroet med sporulerende mycelium af generalist svampen, Metarhizium brunneum. Dette uventede fund kan give ny indsigt i den vigtige co-evolution hos sociale insekter og deres patogener.

General

introduction13 Introduced species and the problem of biological invasions The terms “introduced”, “non-indigenous”, “alien”, “exotic” and “non-native” are often interchangeably used to define a species occurring outside of its native distributional range (past or present) and dispersal potential (i.e. outside the range it occupies naturally or could not occupy without direct or indirect introduction or care by humans). Although the confusion and misuse of the existing terminology have arose a passionate semantic debate (for a review on the topic, see Richardson et al. 2000; Colautti & MacIsaac 2004; Davis 2009), there is a wide consensus about the fact that today’s extensive global trades and passenger movements have drastically increased the frequency and magnitude at which alien species are introduced, intentionally or accidentally, in new areas compared to the past (Perrings et al. 2005). The exponential increase in the number of introduced species has, in parallel, raised the interest of both scientific and public communities on the problem of biological invasions (Richardson & Pyšek 2008). The reasons of this renewed interest lie on the fact that, although most of the introduced organisms die during transit, soon after release or turn out to be harmless (Ricciardi & Cohen 2007), a steadily increasing number of them have been reported to have established successful populations in the area of introduction. Once established, some of these species are able to quickly spread and colonize a huge variety of habitats, causing grave impacts not only on biodiversity and ecosystem function but also on economy, human health and agriculture (Reinthal & Kling 1994; Lounibos 2002; Pimentel et al. 2005; Davis 2009). These organisms are referred to as “invasive” and, as the number of species spread beyond their natural dispersal borders increased with globalization, so have the number of species being described as invasive and the economic and ecological costs associated with their invasions. For example, it has been proposed that invasive species represent the major reason for reduction in biodiversity second only to habitat destruction (Vié et al. 2009). Furthermore, estimated damage and control cost of invasive species in the U.S. alone amount to more than $138 billion annually (Pimentel et al. 2005). Understanding causes and consequences of biological invasions, as well as the ecological, behavioral and environmental factors promoting the spread and the success of invasive species have thus become major goals, both for biodiversity conservation and economic purposes.

Invasive ants Ants represent one of the most devastating, well-studied and costly group of all invasive species (Kenis et al. 2009). Over 200 species of ants have been transported by humans outside of their natural range (McGlynn 1999b). Most of them now present a global distribution and some have become highly destructive invaders (Suarez et al. 2005). The majority of the invasive ants are closely tied to the human-modified habitats, where they are primarily transported and rarely able to establish successful populations in nearby undisturbed areas. Despite being limited to urban landscapes, the invasive impact of these species can be very high, particularly in terms of economy and human health since they are usually found as pest infestations in houses or hospitals (Beatson 1972). In addition to being economically costly in both urban and agricultural areas, a subset of invasive ants is also

14 able to quickly penetrate and dominate natural ecosystems, affecting the composition and/or the abundance of native species both directly and indirectly (Holway et al. 2002). This subset of species has frequently received most of the attention in the literature because of their tremendous ecological impact, and five of them have been even ranked among the worst animal invaders on the planet: the Argentine ant (Linepithema humile), the red imported fire ant (Solenopsis invicta), the big-headed ant (Pheidole megacephala), the little fire ant (Wasmannia auropunctata) and the yellow crazy ant (Anoplolepis gracilipes) (Lowe et al. 2000; Krushelnycky et al. 2010).

The “invasive ant syndrome” Invasive ants represent a tiny fraction (less than 1%) of the approximately 12,500 ant species described so far and exhibit both phylogenetic and morphological diversity (Holway et al. 2002; Suarez et al. 2010). Given that only few ant species become successful invaders while many others fail, several studies have started to investigate whether invasive ants share life-history traits that might predispose them to establish successful populations in novel areas and, if so, whether these characteristics might help scientists to predict and identify new potential invaders (Passera 1994; McGlynn 1999b; Holway et al. 2002). All the invasive ants appear to possess a suite of traits that could collectively be termed an “invasive ant syndrome” and which are thought to have facilitated the introduction of new colonies in distant localities through human- mediated dispersal, their establishment, their rapid spread and the subsequent ecological success (Passera 1994; McGlynn 1999b; Holway et al. 2002; Cremer et al. 2008).

Generalist habits Invasive ants appear to be particularly pre-adapted to live in disturbed environments, where they usually thrive. They have loose nesting requirements and are able to quickly relocate their colonies in response to disturbance in a wide variety of sites (Hölldobler & Wilson 1977; Passera 1994). They are also very opportunistic in terms of food requirements, a trait that enables a more complete use of the available resources. Together, these generalist habits allow invasive ants to be better adapted for transport and for the initial survival in areas where nest sites and food sources might be different from those exploited in the native range.

Polygyny, polydomy, intranidal mating and dependent colony foundation Colonies of invasive ants are also highly polygynous (multiple queens are present in a single nest) and polydomous (colony inhabiting many nests) (Hölldobler & Wilson 1977; Debout et al. 2007). Mating often occurs within the nest and colony reproduction is performed via budding, where a group of workers, brood and sometimes queens depart the natal nest on foot to establish a new nest in the nearby. These two traits are particularly important because they are responsible for the rapid growth and the ease of spread of the colonies in new localities. Intranidal mating removes the need to find a partner in the area of introduction, whereas budding circumvents the dangerous mating flight and the slow initial establishment phase characterizing

15 species that rely on independent colony foundation (Hölldobler & Wilson 1990; Tschinkel 1993). Species that reproduce solely through budding are characterized by low rates of spread and are forced to depend on human-mediated dispersal to reach distant locations. However, once introduced, they can start colonies in a great number and slowly, but steadily, spread from the introduction point and potentially saturating habitats.

Small body size and worker sterility Body size is considered to be another trait invasive ants have in common: compared to ants as a whole, they are small to medium-sized. Furthermore, ants from introduced populations appear to be smaller when compared to their conspecific counterparts from the native range (McGlynn 1999a). By producing smaller workers, colonies of invasive ants can attain bigger sizes and, since numerical superiority determine competitive outcomes in ants generally, easily displace local ant fauna by recruiting more individuals during battles or at food sources (Krushelnycky et al. 2010). Workers in unicolonial species are also completely sterile.

Unicoloniality Lastly, the most striking feature shared by many invasive ants is their tendency to be unicolonial. While the vast majority of ant species form multicolonial populations consisting of discrete, mutually aggressive colonies that recognize behavioral colony boundaries, unicolonial ants form huge cooperative groups – called “supercolonies” – consisting of networks of hundreds or even thousands of highly polygynous nests that exchange individuals and share territories peacefully over extensive areas (Hölldobler & Wilson 1977; Helanterä et al. 2009). Unicoloniality is not unique to invasive ants (some Formica ants, for example, can also show this form of social organization; Bourke & Franks 1995; Holzer et al. 2006), but it appears to be overrepresented in this group (Holway et al. 2002). Moreover, supercolonies formed by some invasive ants can reach astonishing sizes compared to those formed by non-invasive ants. Those characterizing the introduced populations of L. humile and W. auropunctata are probably the best examples, sometimes spanning over hundreds of square kilometers (Tsutsui et al. 2000; Giraud et al. 2002; Le Breton et al. 2003; Corin et al. 2007). Unicoloniality is considered the key attribute responsible for the ecological success of these species because consistently promotes high nest and worker density through the reduction of the costs associated with territoriality. In fact, once colonies cease to engage in intraspecific battles, the energy that might have been directed into fighting is instead put in the production of more ants. This, in turns, allows invasive ants to obtain a decisive advantage over other species both in direct encounters and indirect resource allocation such as foraging (Human & Gordon 1999; Holway et al. 2002; Holway & Suarez 2004; Cremer et al. 2006).

The structure of unicolonial populations and the problem of recognition The term “unicolonial” was coined by Hölldobler and Wilson (1977) to describe “ant species in which no colony boundaries exist and local populations are comprised of networks of intercommunicating

16 aggregations of workers, brood, and fertile queens.... Such species have been referred to as unicolonial, or more precisely as forming unicolonial populations, as opposed to the more frequent multicolonial ant species in which intercolony recognition and aggression are the rule”. Consequently, unicoloniality has generally been used to define species rather than population (Passera 1994; Bourke & Franks 1995; Queller & Strassmann 1998). However, for in the very strictest sense of there being a single colony per population, true unicoloniality might be extremely rare. In fact, there is no species known to consist of a single cooperative unit, but there are unicolonial populations (i.e. the Argentine ant in New Zealand, Corin et al. 2007). Most of the populations originally described as unicolonial in reality appear to consist of multiple, genetically distinct “supercolonies” (Holway et al. 1998; Suarez et al. 1999; Giraud et al. 2002; Pedersen et al. 2006; Abbott et al. 2007; Drescher et al. 2007; Mikheyev & Mueller 2007; Blight et al. 2010). Furthermore, while within each supercolony individuals can freely move between nests without eliciting aggression, the situation drastically changes at the contact zone between two genetically differentiated supercolonies. Here, battles are frequent, prolonged and intense and can lead not only to considerable worker mortality, but also to a complete lack of gene flow between supercolonies (Jaquiéry et al. 2005; Thomas et al. 2005; Thomas et al. 2006). Genetic studies conducted both on native and introduced populations of Argentine ants seem to confirm this view. In fact, in both ranges populations consist of supercolonies in which there is almost no genetic differentiation between nests (workers within a nest are not significantly more related to their nestmates than to workers in other nests of the same supercolony). On the other hand, supercolonies are highly genetically differentiated, suggesting that each supercolony is a close breeding unit (Pedersen et al. 2006; Vogel et al. 2009). The discovery that unicolonial ants can be profoundly aggressive towards conspecifics from genetically different supercolonies but extremely tolerant of unrelated non-nestmates from the same supercolony has led to considerable research on the mechanisms of nestmate discrimination in these species (Suarez et al. 2008). Many studies have focused on the structure of the recognition system and the discriminatory cues used by invasive ants to distinguish group members from non-members (Chen & Nonacs 2000; Liang & Silverman 2000; Buczkowski et al. 2005; Torres et al. 2007). As the expression of aggression seems to be tightly linked to pronounced genetic differences among supercolonies (Tsutsui & Case 2001; Pedersen et al. 2006; Van Wilgenburg et al. 2010), it appears extremely important to investigate whether and how genetics influence the recognition system and the cues used for discrimination. This is crucial because genetic changes occurring during and/or after invasions (i.e. founder events, genetic bottlenecks) might then drastically affect the way in which supercolonies interact, as well as their ecological success (Giraud et al. 2002; Tsutsui et al. 2003; Payne et al. 2004). To understand how genetics might shape supercolony interactions, it is first necessary to examine briefly how ant recognition system operates.

17 Ant recognition system The ability to distinguish group members from non-members is central to the functioning of many biological systems, from multicellular organisms to complex societies. In social insect colonies, where individuals are usually kin, the ability to discriminate between nestmates and non-nestmates is of a fundamental importance because not only prevent colony resources exploitation by hetero and conspecific intruders (Wilson 1975) but also allows group members to gain indirect fitness benefits by ensuring that altruistic acts are directed towards relatives (Hamilton 1964; Crozier & Pamilo 1996). In many social insects, and particularly in ants, the cues for nestmate recognition are primarily encoded in the complex mixture of lipids, mainly hydrocarbons, present on the cuticle of each individual (Akino et al. 2004; Howard & Blomquist 2005; van Zweden & d’Ettorre 2010). These odor labels, which can have a genetic (Stuart 1988) or an environmental origin (Gamboa et al. 1986; Obin & Vander Meer 1988; Buczkowski et al. 2005), are shared among all colony-members by means of grooming, trophallaxis or nest material (Soroker et al. 1995; Bos et al. 2011) thereby leading to the establishment of a specific, constantly updated “Gestalt colony odour” (Crozier & Dix 1979; Provost et al. 1993; Lenoir et al. 1999; van Zweden et al. 2010). A proposed mechanism for nestmate recognition, the so called “phenotype matching” (Lacy & Sherman 1983), involves the matching of the set of cues (e.g a specific pattern of cuticular hydrocarbons) borne by the encountered individual with the neural representation of the colony odour (the “template”) in the discriminating individual. A graded behavioral response (going from acceptance to complete rejection) is then performed according to the degree of cue-template similarity (Vander Meer & Morel 1998; Lenoir et al. 1999) and to the social context where the discrimination process takes place (Reeve 1989).

The effect of genetic diversity on nestmate recognition and supercolony interactions If the odor labels (CHCs) used by ants to distinguish between nestmates and non-nestmates are under genetic control, the level of within colony genetic variation is hypothesized to influence the diversity of such cues and, consequently, the ability of workers to operate a correct discrimination (Keller & Passera 1989). For example, it has been shown that workers from polygynous colonies discriminate to a lesser extent than those of monogynous colonies between nestmate and non-nestmate conspecifics (Breed & Bennett 1987; Morel et al. 1990; Sundström 1997; but see Rosset et al. 2007). It is hypothesized that having multiple reproducing queens increases the genetic diversity of a nest, leading to a considerable variability in worker- produced odor cues. In turn, this can broaden both the mix of cues borne by workers and the internal template characterizing the colony (but see Helanterä et al. 2011). Therefore, workers from polygynous nests might accept conspecific intruders with variable cue profiles, while being themselves rejected by individuals from genetically less diverse colonies, particularly monogynous ones (Vander Meer & Morel 1998). It is worth noting that this sort of “asymmetrical aggression” (i.e. individuals from less genetically diverse colonies are more aggressive towards individuals from more diverse colonies than vice versa) depends only on the level of genetic diversity between the interacting colonies. Hence, this phenomenon

18 can in principle happen also when the competitors are two highly polygynous supercolonies characterized, however, by a different degree of within colony genetic variation (Chapter 1). A polarization in aggression, with low diverse supercolonies being more aggressive towards high diverse supercolonies than vice versa, was indeed found by Tsutsui et al. in the Argentine ant and proposed as a possible mechanism to explain how invasive populations of this species maintain their unicolonial social structure (2003). A selection against rare recognition alleles through a mechanism of positive frequency dependent selection (Crozier paradox, Crozier 1986) appears to be central also in another hypothesis proposed by Giraud et al. about supercolony formation in Europe (“genetic cleansing” hypothesis, Giraud et al. 2002; see Starks 2003). If low diverse colonies have a competitive edge over high diverse ones because they have more chance to win battles, there might be some important ecological consequences. Almost all species introductions lead to a reduction in genetic diversity, since only few individuals are typically introduced as the founders of a new population (founder effect, Mayr 1942). A reduction in genetic diversity at the population level has been documented in the introduced range of some invasive ant species: the Argentine ant Linepithema humile (Tsutsui et al. 2000; Giraud et al. 2002; Vogel et al. 2010), the fire ant Solenopsis invicta (Ross et al. 1996) and the invasive garden ant Lasius neglectus (Ugelvig et al. 2008). For the mechanism described above, these low diverse supercolonies might prevent the establishment of more diverse conspecific propagules in the introduced range as they would be more prone to initiate and win a fight. Furthermore, for the very same effect, an invasive supercolony could become more and more genetically homogeneous and aggressive over time as they go through several secondary introductions (i.e. from established to new population in the introduced area), as every introduction will be characterized by a reduction in genetic diversity. This, in turn, might increase the invasive impact of these bottlenecked populations. The depletion of genetic diversity at the recognition loci following introduction might also increase the rate at which colonies that possess similar chemical and genetic background can fuse (Chapter 2). Fusion of unrelated colonies is particularly common in lower termites (Perdereau et al. 2010; Korb & Roux 2012) and has also been shown in some ant species, particularly after queen loss (Kronauer et al. 2010). Recently it has been shown that the merging of unrelated colonies is also possible in Argentine ant under laboratory conditions, and that the process is modulated by levels of intraspecific aggression and both genetic and phenotypic (chemical) similarity (Vasquez & Silverman 2008). These results lead the authors to speculate that colony fusion might be a mechanism through which this species attained extensive supercolonies in the introduced range, although this hypothesis has been challenged (Helanterä et al. 2009). All the hypotheses described so far base their assumptions on the fact that the cues used for nestmate discrimination are genetically determined and that a reduction in genetic diversity is often accompanied by a parallel reduction in diversity at the recognition loci. The observation that introduced populations of Argentine ants are both genetically and chemically less diverse than their native counterparts appears to be a good, indirect indication that this might be true (Brandt et al. 2009). However, a final prove of these

19 hypotheses can be provided only after the identification of the exact cues used in nestmate recognition, which are actually unknown. Until the end of the 1990s, even the evidence linking hydrocarbons to nestmate recognition were at the best circumstantial, being based on either correlation studies or bioassay using removal and replacement of cuticular compounds by solvent extraction (Lenoir et al. 1999). In recent years an increasing number of studies have started to provide not only the direct evidence needed (Lahav et al. 1999; Wagner et al. 2000; Akino et al. 2004), but they have also shown the differential importance of structural classes of compounds (i.e. linear and methyl-branched alkanes and alkenes) in the nestmate discrimination process (Dani et al. 2001; Dani et al. 2005; Martin et al. 2008a; Martin et al. 2008b; Martin et al. 2012; Krasnec & Breed 2013). Despite the advances in the field, the identification of the exact cues used in nestmate discrimination has remained problematic. The difficulties stem from mainly two reasons: social insects often present a sheer number of hydrocarbons on the cuticle and these vary quantitatively within species (Martin et al. 2008a). Hence, an unequivocal identification of the compounds employed as recognition cues requires not only that each compound is tested both individually and in combination with others, but also in different amount and/or relative proportions. The use of natural or synthetic hydrocarbons allows investigators to manipulate both the amount and the quality of hydrocarbons used in bioassays, but the synthesis of such compounds is costly and sometimes complicated (Sturgis & Gordon 2012). A preliminary step that needs to be taken is therefore to narrow the list of candidates compounds to be tested in such bioassays. This has been mostly done by employing multivariate statistical methods such as discriminate analysis (DA) and principal component analysis (PCA), which however present their own drawbacks (Martin & Drijfhout 2009). Many more statistical approaches can be used to search for likely nestmate recognition compounds (Chapter 3), but few studies have explored these endless possibilities.

The effect of unicoloniality on disease spread As said, unicoloniality provides an extraordinary competitive advantage over other species, allowing invasive ants to quickly dominate the environments they have been introduced to (Holway et al. 2002). On the other hand, some of the features of this peculiar social system can heavily affect the ways in which pathogens and disease spread compared to multicolonial populations and, ultimately, the susceptibility of these societies to epidemics (Ugelvig & Cremer 2012). For example, unicoloniality might lower the horizontal transmission of diseases. This could happen because the free mixing of individuals and the presence of multiple reproducing queens characterizing this social system can lower the relatedness and increase the genetic diversity between nestmates, which will therefore display a different resistance to infections. On the other hand, unicolonial populations are close breeding units and genetically diverse supercolonies fiercely fight against each other at the contact zone (Thomas et al. 2006). Unicoloniality thus prevent the introduction of new genetic material in the supercolony and, as consequence, the amount of genetic diversity possessed by the society is set at the moment of introduction in the new area. If the strength of the genetic bottleneck that often characterizes introduction events has been particularly severe, then the

20 resulting low level of genetic variation can increase the susceptibility of the population to diseases. To this it is necessary to add the fact that unicoloniality drastically increase the contact rates and the density of individuals, which are factors that usually increase the disease transmission rates (Schmid-Hempel 1998). Moreover, the competitive displacement of native species should in addition favor quick adaptation by native parasites. Taken together, these factors appear to drastically increase the risk of disease transmission in unicolonial ants and it is therefore important to study how this species cope with the problem of infections because this can have important implications also for their management (Chapter 4).

21 Model Organisms

The Pharaoh ant

Among all the introduced ants, the pharaoh ant Monomorium pharaonis (Linnaeus, 1758) can probably be considered the most widely human-dispersed (Fig. 1) and, furthermore, the one with the longest introduction history (Wilson & Taylor 1967; Schmidt et al. 2010; Wetterer 2010). Although its native range is still unknown, it has been proposed that this ant originated in the tropical regions of south-east Asia or Africa because these are the only localities where it can be collected both outdoor and inside human structures (Wetterer 2010). In the temperate regions of its global distribution the pharaoh ant is in fact found almost exclusively indoor and heavily depends on humans for transport, food and shelters (Suarez et al. 2005; Buczkowski & Bennett 2009). The pharaoh ant thus represents a typical “tramp” ant as it lives in close association with humans, and relies on them for dispersal to new and distant localities and it thrives in urban environments (Passera 1994). Unlike other invasive ants, this species is not thought to cause significant ecosystem damages such as competitive displacement of native organisms because of its tight indoor life- style (McGlynn 1999b). However, it can have a tremendous impact on economy and human health as it is found as pest infestation in houses and hospitals, where it is also known to act as a vector for diseases (Beatson 1972; Berndt & Eichler 1987).

Fig. 1. Distribution records of M. pharaonis (Wetterer 2010).

The pharaoh ant exhibits all the life-history traits that we previously described as an “invasive ant syndrome”. It is a small ant (~ 2 mm long, Fig. 2) that shows generalist habits in terms of food and nest requirements. Indoor nests are often established in and around household items and readily undergo migrations in

22 response to physical or chemical disturbance, sudden depletion of food supply or overcrowding (Peacock et al. 1955; Passera 1994; Buczkowski & Bennett 2009). Colonies (Fig. 3) are usually polydomic and new nests, established by dependent colony foundation (or “budding”), are highly polygynous and sometimes contain hundreds of fertile queens (Buczkowski & Bennett 2009). Eggs are always queen-laid, as workers have no ovaries and are thus completely sterile (Berndt & Eichler 1987). In addition, pharaoh ants display low levels of intraspecific aggression and, like the majority of other invasive ant species, they are considered to be unicolonial (Hölldobler & Wilson 1990; Helanterä et al. 2009; Schmidt et al. 2010).

In this species there does not appear to be any seasonality in the production of new sexuals, which are induced at intervals of 3-4 months as the queens senesce (Petersen-Braun 1975). However, the production of new reproductives can be induced at any time by removing all the queens from the nest, as workers can rear both males and gynes from the existing brood stages. The same phenomenon can be obtained by removing brood and workers from a colony and rearing them in the absence of queens (Peacock et al. 1955; Edwards 1987, 1991). Although both males and newly-emerged gynes are winged, there is no “nuptial flight” and mating occurs within the natal nest. Despite the intranidal mating, colonies do not appear to suffer from inbreeding depression and therefore it seems unlikely that pharaoh ant possess a single locus complementary sex determination (SL-CDS). These characteristics make the pharaoh ant a very suitable laboratory organism, as its colonies can be kept in captivity for several years and can be easily mated (Peacock & Baxter 1949).

23 © Luigi Pontieri Fig. 3. Three M. pharaonis queens surrounded by brood and workers.

What makes this ant species even more interesting is the fact that both laboratory and field colonies have very low levels of within-colony genetic diversity and are highly genetically divergent (Schmidt et al. 2010). These features, combined with the characteristics described above, has allowed previous researchers (Dr. Anna Schmidt in collaboration with Dr. Timothy Linksvayer) to set up a crossing protocol and perform sequential, multigenerational crosses to obtain a genetically diverse, heterogeneous stock of colonies. A representation of the crossing scheme can be seen in Fig. 4 (adopted from Schmidt 2010). Briefly, eight low diverse, highly genetically divergent colonies were collected either from other laboratories or from the field (colonies U1, U2, U3, U4, U5, U7, U11 and Gh4; see Schmidt et al. 2010 for further details). These original inbred lineages were sequentially crossed following the approach of Hansen and Spulher (1984) to obtain colonies with higher level of genetic diversity. Some of these colonies have been already used to investigate the effect of genetic diversity on disease resistance (Schmidt et al. 2011) and can be used, more generally, to test the possible influence of genetic diversity on a number of behavioral and biological processes in the invasive ants (Chapter 1 and 2).

24 original eight colonies Generation

Each crossing contained 10-20 gynes F4 and 4-15 males

Fig. 4. Representation of the crossing scheme employed to obtain a genetic diverse, heterogeneous stock of M. pharaonis colonies.

The depicted crosses represent only a fraction of the crosses made in each generation. One gyne and one male represent each colony in the parental generation (P), color-coded according to colony. The colors in the filial generations indicate the mixing of genes from the parents. Each cross was made using 10-20 gynes and 4-15 males. (Adopted from Schmidt 2010)

25 Metarhizium brunneum

Metarhizium brunneum (formerly named Metarhizium anisopliae, Bischoff et al. 2009) is a virulent, generalist entomopathogenic fungus known for its ability to parasitize a broad spectrum of both social and non-social insect hosts and commonly used as an experimental parasite of ants (Schmid-Hempel 1998; Hughes et al. 2004; Chapuisat et al. 2007; Ugelvig & Cremer 2007; Bos et al. 2012). The life-cycle of this fungus is as follows: the spores (or conidia) first firmly adhere on the host surface and then germinate (i.e. emergence of the germ tube, or appressoria) and penetrate the host cuticle within 24-48 hours through a combination of enzymatic and mechanical means (Gillespie et al. 2000; Boucias et al. 2012). Once inside the body, the fungus grows by producing hyphal bodies that assimilate the nutrients from the host environment and, in absence of an efficient immunological reaction, the insect usually dies within 5-7 days after the establishment of the infection. The host resources are then converted into greenish, infective conidia (Fig. 5) that are released from the cadaver over time following sporulation (Gottwald & Tedders 1984). Spores dispersal take place by wind, rainsplash or by the activity of the insect (Meyling & Eilenberg 2007) and, once arriving on the cuticle of a new host, they can start the cycle again. Metarhizium brunneum and other species of this genus have been used as biological control agents to manage and prevent infestations of a number of insects such as mosquitoes, locusts and grasshoppers (Hunter et al. 2001; Lomer et al. 2001; Farenhorst et al. 2009). However, the use of this and similar entomopathogenic fungi in controlling invasive ants has not provided so far successful results, probably because of the number of behavioral and chemical defensive mechanisms they exhibit (Williams et al. 2003; Cremer et al. 2007; Zarzuela et al. 2012). Studies on the behavioral strategies that invasive ants adopt to cope with these deadly pathogens can provide not only further insights on disease dynamics in unicolonial systems (Ugelvig & Cremer 2012), but also information about the efficiency of entomopathogenic fungi as control agents of such pests (Chapter 4).

26 Short summary of the thesis chapters

In Chapter 1 we focus on the nestmate recognition system of pharaoh ants. Using two types of aggression bioassays, we investigated whether pharaoh ants were able to discriminate between nestmates and non- nestmates, whether the cues used for discrimination had a genetic origin and how different level of within colony genetic variation and relatedness could eventually influence the discrimination abilities of the colonies. We found that pharaoh ants were able to distinguish nestmates from non-nestmates likely using heritable odor cues, and high levels of within colony genetic diversity overall reduced the discrimination skills. While the degree of relatedness did not affect the overall level of discrimination, whether colonies were related or not played a major role in determining whether discrimination was asymmetrical with respect to the genetic diversity of the opponents. When interacting colonies were related to some extent, we found that the genetically low diverse colonies were discriminating high diverse colonies better than vice versa. The greater discrimination abilities of low diverse colonies can have important implications for the success of invasive populations of this and other invasive ants, which often display reduced levels of genetic diversity. In Chapter 2 we investigated whether the high genetic differentiation that characterize laboratory and field colonies of pharaoh ant is sufficient to prevent unrelated colonies to fuse and how different level of genetic similarity shape the outcome by pairing laboratory colonies in a fusion assay. We found that, despite initial high level of aggression, the majority of unrelated colonies readily fused in absence of barriers preventing their encounter, and that the degree of genetic similarity did not influence the outcome. Interesting, the initial aggression was positively correlated with the chemical and genetic distance between colony pairs, suggesting an important role of endogenous cues in the nestmate recognition system of this species. Why fusion does not occur in natural populations of spatially close colonies could be due to environmental factors. In Chapter 3 we compiled data sets of cuticular hydrocarbons (CHCs) and aggression between colonies of three species of ants (Formica exsecta, Camponotus aethiops and Monomorium pharaonis) and a simulated data set built upon the F. exsecta data set. Then, using the available information about the exact cues used for nestmate recognition in F. exsecta, we searched for likely nestmate recognition (NMR) compounds in the other two species and we evaluated and compared the power of different combinations of data transformation and chemical distance calculation in differentiating between true NMR cues and other compounds. While we could not unequivocally identify specific sets of compounds that might act as recognition cues in C. aethiops and M. pharaonis, we found that particular combinations of statistical procedures are more effective than other in differentiating NMR cues from other compounds. Particularly, we propose that the use of a new method of centroid calculation that we developed (the “global centroid”) can help to identify likely NMR cues.

27 In Chapter 4 we investigated the nest site preference of pharaoh ant colonies and, specifically, their ability to avoid nests containing infectious pathogen as invasive, unicolonial ants might be particularly prone to diseases compared to other ant species. Using binary choice tests between three types of nests, we found that migrating colonies surprisingly prefer nest sites containing nestmate corpses overgrown with sporulating mycelium of the generalist fungus Metarhizium brunneum. While we cannot rule out that the ants are actually manipulated by the pathogen, we propose that the preference for infected nests might be an adaptive strategy by the host to “immunize” the colony against future exposure to the same pathogenic fungus.

Future perspectives

There are several interesting questions that arise from the work conducted in this thesis. In the first two chapters we have shown that pharaoh ants, contrary to the current view (Schmidt et al. 2010), are able to discriminate between nestmates and non-nestmates and that the type of discrimination behavior performed is often influenced by the social and ecological context where the discrimination takes place (see Roulston et al. 2003). Intriguingly, pharaoh ants seem to further discriminate non-nestmates according to their level of genetic diversity and relatedness, and there seems to be a strict link between genetics and the cuticular hydrocarbon profile of the colonies. As the evidence we provide for this relationship are mostly correlative, a natural follow up of these studies is to investigate in greater details the possibility that in pharaoh ants the cues used for nestmate recognition are entirely genetically based. In this respect, the pharaoh ant represents a perfect model system to investigate this question as colonies can be crossed and maintained indefinitely in the laboratory. For example, by crossing laboratory colonies that are highly genetically and chemically divergent, it would be possible to investigate which hydrocarbons are inherited and to which extent. Finally, compounds that show high inheritance might be tested in bioassays to assess whether they really constitute the nestmate recognition signal in this species, and genomic techniques might then help to investigate the genes involved in the production of such cues. Another interesting question arising from our work is why field colonies of pharaoh ants do not seem to merge, although we showed that fusion of even unrelated colonies is possible in laboratory conditions. Is it because, despite the close proximity, colonies do not meet? Are there factors other than genetic diversity, such as colony size and queen number, that prevent merging? Future studies conducted directly in the field might answer these questions. Pharaoh ant showed a preference for nest sites containing infectious corpses when relocating their colonies, but we don’t know whether this choice is an adaptive strategy of the host to immunize the colony or is the result of a manipulation operated by the pathogen to increase its transmission. Future studies might elucidate this aspect. For example, it would be interesting to see whether colonies that move in infected nests become more resistant to subsequent exposure to the same pathogen compared to colonies that avoided infected

28 nests. It would also be interesting to conduct a long term experiment to see how and to which extent this choice can affect survival and growth of the colony. Another interesting topic that deserves future researches is the investigation of which compounds emitted by the fungal spores are perceived by the ants and whether these compounds have a repulsive or an attractive effect. This information might be useful not only to further elucidate host-pathogen interactions, but also to evaluate the possibility to use entomopathogenic fungi as biological control agents of invasive ants.

29 References

Abbott, K. L., Greaves, S. N. J., Ritchie, P. A. & Lester, P. J. 2007. Behaviourally and genetically distinct populations of an invasive ant provide insight into invasion history and impacts on a tropical ant community. Biological Invasions, 9, 453-463. Akino, T., Yamamura, K., Wakamura, S. & Yamaoka, R. 2004. Direct behavioral evidence for hydrocarbons as nestmate recognition cues in Formica japonica (Hymenoptera: Formicidae). Applied Entomology and Zoology, 39, 381-387. Beatson, S. 1972. Pharaoh’s ants as pathogen vectors in hospitals. The Lancet, 299, 425-427. Berndt, K. P. & Eichler, W. 1987. Die Pharaoameise, Monomorium pharaonis (L.) (Hym., Myrmicidae). Mitteilungen aus dem Museum für Naturkunde in Berlin. Zoologisches Museum und Institut für Spezielle Zoologie (Berlin), 63, 3-186. Bischoff, J. F., Rehner, S. A. & Humber, R. A. 2009. A multilocus phylogeny of the Metarhizium anisopliae lineage. Mycologia, 101, 512-530. Blight, O., Renucci, M., Tirard, A., Orgeas, J. & Provost, E. 2010. A new colony structure of the invasive Argentine ant (Linepithema humile) in Southern Europe. Biological Invasions, 12, 1491-1497. Bos, N., Grinsted, L. & Holman, L. 2011. Wax on, wax off: nest soil facilitates indirect transfer of recognition cues between ant nestmates. PLoS ONE, 6, e19435. Bos, N., Lefèvre, T., Jensen, A. B. & d’Ettorre, P. 2012. Sick ants become unsociable. Journal of Evolutionary Biology, 25, 342-351. Boucias, D. G., Lietze, V.-U. & Teal, P. 2012. Chemical signals that mediate insect-fungal interactions. In Biocommunication of Fungi (ed. G. Witzany), pp. 305-336. Dordrecht, The Netherlands: Springer. Bourke, A. F. & Franks, N. R. 1995. Social evolution in ants. Princeton: Princeton University Press. Brandt, M., Van Wilgenburg, E. & Tsutsui, N. 2009. Global‐scale analyses of chemical ecology and population genetics in the invasive Argentine ant. Molecular Ecology, 18, 997-1005. Breed, M. D. & Bennett, B. 1987. Kin recognition in highly eusocial insects. In Kin Recognition in Animals. (ed. D. J. C. Fletcher & C. D. Michener), pp. 243-285. Chichester, New York: John Wiley & Sons. Buczkowski, G. & Bennett, G. 2009. Colony budding and its effects on food allocation in the highly polygynous ant, Monomorium pharaonis. Ethology, 115, 1091-1099. Buczkowski, G., Kumar, R., Suib, S. L. & Silverman, J. 2005. Diet-related modification of cuticular hydrocarbon profiles of the Argentine ant,Linepithema humile, diminishes intercolony aggression. Journal of Chemical Ecology, 31, 829-843. Chapuisat, M., Oppliger, A., Magliano, P. & Christe, P. 2007. Wood ants use resin to protect themselves against pathogens. Proceedings of the Royal Society B: Biological Sciences, 274, 2013-2017.

30 Chen, J. S. C. & Nonacs, P. 2000. Nestmate recognition and intraspecific aggression based on environmental cues in Argentine ants (Hymenoptera: Formicidae). Annals of the Entomological Society of America, 93, 1333-1337. Colautti, R. I. & MacIsaac, H. J. 2004. A neutral terminology to define ‘invasive’ species. Diversity and Distributions, 10, 135-141. Corin, S. E., Abbott, K. L., Ritchie, P. A. & Lester, P. J. 2007. Large scale unicoloniality: the population and colony structure of the invasive Argentine ant (Linepithema humile) in New Zealand. Insectes Sociaux, 54, 275-282. Cremer, S., Armitage, S. A. O. & Schmid-Hempel, P. 2007. Social immunity. Current Biology, 17, R693-R702. Cremer, S., Ugelvig, L. V., Drijfhout, F. P., Schlick-Steiner, B. C., Steiner, F. M., Seifert, B., Hughes, D. P., Schulz, A., Petersen, K. S., Konrad, H., Stauffer, C., Kiran, K., Espadaler, X., d’Ettorre, P., Aktaç, N., Eilenberg, J., Jones, G. R., Nash, D. R., Pedersen, J. S. & Boomsma, J. J. 2008. The evolution of invasiveness in garden ants. Plos One, 3, e3838. Cremer, S., Ugelvig, L. V., Lommen, S. T. E., Petersen, K. S. & Pedersen, J. S. 2006. Attack of the invasive garden ant: aggression behaviour of Lasius neglectus (Hymenoptera: Formicidae) against native Lasius species in Spain. Myrmecologische Nachrichten, 9, 13-19. Crozier, R. H. 1986. Genetic clonal recognition abilities in marine invertebrates must be maintained by selection for something else. Evolution, 40, 1100-1101. Crozier, R. H. & Dix, M. 1979. Analysis of two genetic models for the innate components of colony odor in social Hymenoptera. Behavioral Ecology and Sociobiology, 4, 217-224. Crozier, R. H. & Pamilo, P. 1996. Evolution of Social Insect Colonies: Sex Allocation and Kin Selection. Oxford: Oxford University Press. Dani, F. R., Jones, G. R., Corsi, S., Beard, R., Pradella, D. & Turillazzi, S. 2005. Nestmate recognition cues in the honey bee: differential importance of cuticular alkanes and alkenes. Chemical Senses, 30, 477-489. Dani, F. R., Jones, G. R., Destri, S., Spencer, S. H. & Turillazzi, S. 2001. Deciphering the recognition signature within the cuticular chemical profile of paper wasps.Animal Behaviour, 62, 165-171. Davis, M. A. 2009. Invasion Biology. Oxford: Oxford University Press. Debout, G., Schatz, B., Elias, M. & Mckey, D. 2007. Polydomy in ants: what we know, what we think we know, and what remains to be done. Biological Journal of the Linnean Society, 90, 319-348. Drescher, J., Blüthgen, N. & Feldhaar, H. 2007. Population structure and intraspecific aggression in the invasive ant species Anoplolepis gracilipes in Malaysian Borneo. Molecular Ecology, 16, 1453- 1465.

31 Edwards, J. P. 1987. Caste regulation in the pharaoh’s ant Monomorium pharaonis: the influence of queens on the production of new sexual forms. Physiological Entomology, 12, 31-39. Edwards, J. P. 1991. Caste regulation in the pharaoh’s ant Monomorium pharaonis: recognition and cannibalism of sexual brood by workers. Physiological Entomology, 16, 263-271. Farenhorst, M., Mouatcho, J. C., Kikankie, C. K., Brooke, B. D., Hunt, R. H., Thomas, M. B., Koekemoer, L. L., Knols, B. G. J. & Coetzee, M. 2009. Fungal infection counters insecticide resistance in African malaria mosquitoes. Proceedings of the National Academy of Sciences, 106, 17443-17447. Gamboa, G. J., Reeve, H. K., Ferguson, I. D. & Wacker, T. L. 1986. Nestmate recognition in social wasps: the origin and acquisition of recognition odours. Animal Behaviour, 34, 685-695. Gillespie, J. P., Bailey, A. M., Cobb, B. & Vilcinskas, A. 2000. Fungi as elicitors of insect immune responses. Archives of Insect Biochemistry and Physiology, 44, 49-68. Giraud, T., Pedersen, J. S. & Keller, L. 2002. Evolution of supercolonies: the Argentine ants of southern Europe. Proceedings of the National Academy of Sciences of the United States of America, 99, 6075-6079. Gottwald, T. R. & Tedders, W. L. 1984. Colonization, transmission, and longevity of Beauveria bassiana and Metarhizium anisopliae (Deuteromycotina: Hypomycetes) on Pecan weevil larvae (Coleoptera: Curculionidae) in the soil. Environmental Entomology, 13, 557-560. Hamilton, W. D. 1964. The genetical evolution of social behaviour. I. Journal of Theoretical Biology, 7, 1-16. Hansen, C. & Spuhler, K. 1984. Development of the National Institutes of Health genetically heterogeneous rat stock. Alcoholism: Clinical and Experimental Research, 8, 477-479. Helanterä, H., Lee, Y. R., Drijfhout, F. P. & Martin, S. J. 2011. Genetic diversity, colony chemical phenotype, and nest mate recognition in the ant Formica fusca. Behavioral Ecology, 22, 710-716. Helanterä, H., Strassmann, J. E., Carrillo, J. & Queller, D. C. 2009. Unicolonial ants: where do they come from, what are they and where are they going? Trends in Ecology & Evolution, 24, 341-349. Hölldobler, B. & Wilson, E. O. 1977. The number of queens: an important trait in ant evolution. Naturwissenschaften, 64, 8-15. Hölldobler, B. & Wilson, E. O. 1990. The Ants. Cambridge, Massachusetts: Harvard University Press. Holway, D. A., Lach, L., Suarez, A. V., Tsutsui, N. D. & Case, T. J. 2002. The causes and consequences of ant invasions. Annual Review of Ecology and Systematics, 33, 181-233. Holway, D. A. & Suarez, A. V. 2004. Colony-structure variation and interspecific competitive ability in the invasive Argentine ant. Oecologia, 138, 216-222. Holway, D. A., Suarez, A. V. & Case, T. J. 1998. Loss of intraspecific aggression in the success of a widespread invasive social insect. Science, 282, 949-952.

32 Holzer, B., Chapuisat, M., Kremer, N., Finet, C. & Keller, L. 2006. Unicoloniality, recognition and genetic differentiation in a native Formica ant. Journal of Evolutionary Biology, 19, 2031-2039. Howard, R. W. & Blomquist, G. J. 2005. Ecological, behavioral, and biochemical aspects of insect hydrocarbons. Annual Review of Entomology, 50, 371-393. Hughes, W. O. H., Thomsen, L., Eilenberg, J. & Boomsma, J. J. 2004. Diversity of entomopathogenic fungi near leaf-cutting ant nests in a neotropical forest, with particular reference to Metarhizium anisopliae var. anisopliae. Journal of Invertebrate Pathology, 85, 46-53. Human, K. G. & Gordon, D. M. 1999. Behavioral interactions of the invasive Argentine ant with native ant species. Insectes Sociaux, 46, 159-163. Hunter, D. M., Milner, R. J. & Spurgin, P. A. 2001. Aerial treatment of the Australian plague locust, Chortoicetes terminifera (Orthoptera : Acrididae) with Metarhizium anisopliae (Deuteromycotina : Hyphomycetes). Bulletin of Entomological Research, 91, 93-99. Jaquiéry, J., Vogel, V. & Keller, L. 2005. Multilevel genetic analyses of two European supercolonies of the Argentine ant, Linepithema humile. Molecular Ecology, 14, 589-598. Keller, L. & Passera, L. 1989. Influence of the number of queens on nestmate recognition and attractiveness of queens to workers in the Argentine ant, Iridomyrmex humilis (Mayr). Animal Behaviour, 37, 733-740. Kenis, M., Auger-Rozenberg, M.-A., Roques, A., Timms, L., Péré, C., Cock, M. W., Settele, J., Augustin, S. & Lopez-Vaamonde, C. 2009. Ecological effects of invasive alien insects. Biological Invasions, 11, 21-45. Korb, J. & Roux, E. A. 2012. Why join a neighbour: fitness consequences of colony fusions in termites. Journal of Evolutionary Biology, 25, 2161-2170. Krasnec, M. O. & Breed, M. D. 2013. Colony-specific cuticular hydrocarbon profile inFormica argentea ants. Journal of Chemical Ecology, 39, 59-66. Kronauer, D. J. C., Schöning, C., d’Ettorre, P. & Boomsma, J. J. 2010. Colony fusion and worker reproduction after queen loss in army ants. Proceedings of the Royal Society B: Biological Sciences, 277, 755-763. Krushelnycky, P. D., Holway, D. A., LeBrun, E. G., Lach, L., Parr, C. & Abbott, K. 2010. Invasion processes and causes of success. Ant ecology, 245-260. Lacy, R. C. & Sherman, P. W. 1983. Kin recognition by phenotype matching. American Naturalist, 489- 512. Lahav, S., Soroker, V., Hefetz, A. & Vander Meer, R. K. 1999. Direct behavioral evidence for hydrocarbons as ant recognition discriminators. Naturwissenschaften, 86, 246-249.

33 Le Breton, J., Chazeau, J. & Jourdan, H. 2003. Immediate impacts of invasion by Wasmannia auropunctata (Hymenoptera: Formicidae) on native litter ant fauna in a New Caledonian rainforest. Austral Ecology, 28, 204-209. Lenoir, A., Fresneau, D., Errard, C. & Hefetz, A. 1999. Individuality and colonial identity in ants: the emergence of the social representation concept. In Information Processing in Social Insects (ed. C. Detrain, J. Deneubourg & J. Pasteels), pp. 219-237. Basel: Birkhäuser. Liang, D. & Silverman, J. 2000. ”You are what you eat”: Diet modifies cuticular hydrocarbons and nestmate recognition in the Argentine ant, Linepithema humile. Naturwissenschaften, 87, 412-416. Lomer, C. J., Bateman, R. P., Johnson, D. L., Langewald, J. & Thomas, M. 2001. Biological control of locust and grasshoppers. Annual Review of Entomology, 46, 667-702. Lounibos, L. P. 2002. Invasions by insect vectors of human disease. Annual Review of Entomology, 47, 233-266. Lowe, S., Browne, M., Boudjelas, S. & De Poorter, M. 2000. 100 of the world’s worst invasive alien species: a selection from the Global Invasive Species Database: The Invasive Species Specialist Group New Zealand. Martin, S. J. & Drijfhout, F. P. 2009. How reliable is the analysis of complex cuticular hydrocarbon profiles by multivariate statistical methods?Journal of Chemical Ecology, 35, 375-382. Martin, S. J., Helanterä, H. & Drijfhout, F. P. 2008a. Colony-specific hydrocarbons identify nest mates in two species of Formica ant. Journal of Chemical Ecology, 34, 1072-1080. Martin, S. J., Vitikainen, E., Drijfhout, F. P. & Jackson, D. 2012. Conspecific ant aggression is correlated with chemical distance, but not with genetic or spatial distance. Behavior Genetics, 42, 323-331. Martin, S. J., Vitikainen, E., Helanterä, H. & Drijfhout, F. P. 2008b. Chemical basis of nest-mate discrimination in the ant Formica exsecta. Proceedings of the Royal Society B: Biological Sciences, 275, 1271-1278. Mayr, E. 1942. Systematics and the Origin of Species. New York: Columbia University Press. McGlynn, T. P. 1999a. Non-native ants are smaller than related native ants. American Naturalist, 154, 690-699. McGlynn, T. P. 1999b. The worldwide transfer of ants: geographical distribution and ecological invasions. Journal of Biogeography, 26, 535-548. Meyling, N. V. & Eilenberg, J. 2007. Ecology of the entomopathogenic fungi Beauveria bassiana and Metarhizium anisopliae in temperate agroecosystems: potential for conservation biological control. Biological Control, 43, 145-155. Mikheyev, A. S. & Mueller, U. G. 2007. Genetic relationships between native and introduced populations of the little fire antWasmannia auropunctata. Diversity and Distributions, 13, 573-579.

34 Morel, L., Vander Meer, R. K. & Lofgren, C. S. 1990. Comparison of nestmate recognition between monogyne and polygyne populations of Solenopsis invicta (Hymenoptera, Formicidae). Annals of the Entomological Society of America, 83, 642-647. Obin, M. S. & Vander Meer, R. K. 1988. Sources of nestmate recognition cues in the imported fire ant Solenopsis invicta Buren (Hymenoptera: Formicidae). Animal Behaviour, 36, 1361-1370. Passera, L. 1994. Characteristics of tramp species. In Exotic Ants: Biology, Impact, and Control of Introduced Species. (ed. D. F. Williams), pp. 23-43. Boulder, Colorado: Westview Press. Payne, C. M., Tillberg, C. V. & Suarez, A. V. 2004. Recognition systems and biological invasions. Annales Zoologici Fennici, 41, 843-858. Peacock, A. & Baxter, A. 1949. Studies in Pharaoh’s ant, Monomorium pharaonis (L.). 1. The rearing of artificial colonies.Entomologist’s Monthly Magazine, 85, 256-260. Peacock, A., Sudd, J. & Baxter, A. 1955. Studies in Pharaoh’s Ant, Monomorium pharaonis (L.). 11. Colony foundation. Entomologist’s Monthly Magazine, 91, 125-129. Pedersen, J. S., Krieger, M. J. B., Vogel, V., Giraud, T. & Keller, L. 2006. Native supercolonies of unrelated individuals in the invasive Argentine ant. Evolution, 60, 782-791. Perdereau, E., Bagnères, A. G., Dupont, S. & Dedeine, F. 2010. High occurrence of colony fusion in a European population of the American termite Reticulitermes flavipes. Insectes Sociaux, 57, 393- 402. Perrings, C., Dehnen-Schmutz, K., Touza, J. & Williamson, M. 2005. How to manage biological invasions under globalization. Trends in Ecology & Evolution, 20, 212-215. Petersen-Braun, M. 1975. Untersuchungen zur sozialen organisation der pharaoameise Monomorium pharaonis (L.) (Hymenoptera, Formicidae) I. Der brutzyklus und seine steuerung durch populationseigene faktoren. Insectes Sociaux, 22, 269-291. Pimentel, D., Zuniga, R. & Morrison, D. 2005. Update on the environmental and economic costs associated with alien-invasive species in the United States. Ecological Economics, 52, 273-288. Provost, E., Riviere, G., Roux, M., David Morgan, E. & Bagneres, A. G. 1993. Change in the chemical signature of the ant Leptothorax lichtensteini Bondroit with time. Insect biochemistry and molecular biology, 23, 945-957. Queller, D. C. & Strassmann, J. E. 1998. Kin selection and social insects. Bioscience, 165-175. Reeve, H. K. 1989. The evolution of conspecific acceptance thresholds.American Naturalist, 133, 407-435. Reinthal, P. N. & Kling, G. W. 1994. Exotic species, trophic interactions, and ecosystem dynamics: a case study of Lake Victoria. In Theory and Application in Fish Feeding and Ecology. Ricciardi, A. & Cohen, J. 2007. The invasiveness of an introduced species does not predict its impact. Biological Invasions, 9, 309-315.

35 Richardson, D. M. & Pyšek, P. 2008. Fifty years of invasion ecology – the legacy of Charles Elton. Diversity and Distributions, 14, 161-168. Richardson, D. M., Pyšek, P., Rejmánek, M., Barbour, M. G., Panetta, F. D. & West, C. J. 2000. Naturalization and invasion of alien plants: concepts and definitions.Diversity and Distributions, 6, 93-107. Ross, K. G., Vargo, E. L. & Keller, L. 1996. Social evolution in a new environment: the case of introduced fire ants. Proceedings of the National Academy of Sciences, 93, 3021-3025. Rosset, H., Schwander, T. & Chapuisat, M. 2007. Nestmate recognition and levels of aggression are not altered by changes in genetic diversity in a socially polymorphic ant. Animal Behaviour, 74, 951-956. Roulston, T. H., Buczkowski, G. & Silverman, J. 2003. Nestmate discrimination in ants: effect of bioassay on aggressive behavior. Insectes Sociaux, 50, 151-159. Schmid-Hempel, P. 1998. Parasites in Social Insects. Princeton, New Jersey: Princeton University Press. Schmidt, A. M. 2010. The invasion biology and sociogenetics of pharaoh ants. In Department of Biology, PhD thesis. Copenhagen: University of Copenhagen. Schmidt, A. M., d’Ettorre, P. & Pedersen, J. S. 2010. Low levels of nestmate discrimination despite high genetic differentiation in the invasive pharaoh ant. Frontiers in Zoology, 7, 20. Schmidt, A. M., Linksvayer, T. A., Boomsma, J. J. & Pedersen, J. S. 2011. No benefit in diversity? The effect of genetic variation on survival and disease resistance in a polygynous social insect. Ecological Entomology, 36, 751-759. Soroker, V., Vienne, C. & Hefetz, A. 1995. Hydrocarbon dynamics within and between nestmates in Cataglyphis niger (Hymenoptera: Formicidae). Journal of Chemical Ecology, 21, 365-378. Starks, P. T. 2003. Selection for uniformity: xenophobia and invasion success. Trends in Ecology & Evolution, 18, 159-162. Stuart, R. J. 1988. Collective cues as a basis for nestmate recognition in polygynous leptothoracine ants. Proceedings of the National Academy of Sciences, 85, 4572-4575. Sturgis, S. J. & Gordon, D. M. 2012. Nestmate recognition in ants (Hymenoptera: Formicidae): a review. Myrmecological News, 16, 101-110. Suarez, A., Holway, D. & Tsutsui, N. 2008. Genetics and behavior of a colonizing species: The invasive Argentine ant. The American Naturalist, 172, S72-S84. Suarez, A. V., Holway, D. A. & Ward, P. S. 2005. The role of opportunity in the unintentional introduction of nonnative ants. Proceedings of the National Academy of Sciences of the United States of America, 102, 17032-17035.

36 Suarez, A. V., McGlynn, T. P. & Tsutsui, N. D. 2010. Biogeographic and taxonomic patterns of introduced ants. In Ant Ecology (ed. L. Lach, C. L. Parr & K. L. Abbott), pp. 233-244. Oxford: Oxford University Press. Suarez, A. V., Tsutsui, N. D., Holway, D. A. & Case, T. J. 1999. Behavioral and genetic differentiation between native and introduced populations of the Argentine ant. Biological Invasions, 1, 43-53. Sundström, L. 1997. Queen acceptance and nestmate recognition in monogyne and polygyne colonies of the ant Formica truncorum. Animal Behaviour, 53, 499-510. Thomas, M. L., Payne-Makrisa, C. M., Suarez, A. V., Tsutsui, N. D. & Holway, D. A. 2006. When supercolonies collide: territorial aggression in an invasive and unicolonial social insect. Molecular Ecology, 15, 4303-4315. Thomas, M. L., Tsutsui, N. D. & Holway, D. A. 2005. Intraspecific competition influences the symmetry and intensity of aggression in the Argentine ant. Behavioral Ecology, 16, 472-481. Torres, C., Brandt, M. & Tsutsui, N. 2007. The role of cuticular hydrocarbons as chemical cues for nestmate recognition in the invasive Argentine ant (Linepithema humile). Insectes Sociaux, 54, 363-373. Tschinkel, W. R. 1993. Sociometry and Sociogenesis of colonies of the fire ant Solenopsis invicta during one annual cycle. Ecological Monographs, 63, 425-457. Tsutsui, N. D. & Case, T. J. 2001. Population genetics and colony structure of the Argentine ant (Linepithema humile) in its native and introduced ranges. Evolution, 55, 976-985. Tsutsui, N. D., Suarez, A. V. & Grosberg, R. K. 2003. Genetic diversity, asymmetrical aggression, and recognition in a widespread invasive species. Proceedings of the National Academy of Sciences of the United States of America, 100, 1078-1083. Tsutsui, N. D., Suarez, A. V., Holway, D. A. & Case, T. J. 2000. Reduced genetic variation and the success of an invasive species. Proceedings of the National Academy of Sciences of the United States of America, 97, 5948-5953. Ugelvig, L., Drijfhout, F., Kronauer, D., Boomsma, J., Pedersen, J. & Cremer, S. 2008. The introduction history of invasive garden ants in Europe: integrating genetic, chemical and behavioural approaches. BMC Biology, 6, 11. Ugelvig, L. V. & Cremer, S. 2007. Social prophylaxis: group interaction promotes collective immunity in ant colonies. Current Biology, 17, 1967-1971. Ugelvig, L. V. & Cremer, S. 2012. Effects of social immunity and unicoloniality on host-parasite interactions in invasive insect societies. Functional Ecology, 26, 1300-1312. Van Wilgenburg, E., Torres, C. W. & Tsutsui, N. D. 2010. The global expansion of a single ant supercolony. Evolutionary Applications, 3, 136-143.

37 van Zweden, J. S., Brask, J. B., Christensen, J. H., Boomsma, J. J., Linksvayer, T. A. & d’Ettorre, P. 2010. Blending of heritable recognition cues among ant nestmates creates distinct colony gestalt odours but prevents within‐colony nepotism. Journal of Evolutionary Biology, 23, 1498-1508. van Zweden, J. S. & d’Ettorre, P. 2010. Nestmate recognition in social insects and the role of hydrocarbons. In Insect Hydrocarbons: Biology, Biochemistry and Chemical Ecology (ed. G. J. Blomquist & A. G. Bagnères), pp. 222-243. Cambridge: Cambridge University Press. Vander Meer, R. K. & Morel, L. 1998. Nestmate recognition in ants. In Pheromone Communication in Social Insects (ed. R. K. Vander Meer, M. D. Breed, K. E. Espelie & M. L. Winston), pp. 79-103. Boulder: Westview Press. Vasquez, G. M. & Silverman, J. 2008. Intraspecific aggression and colony fusion in the Argentine ant. Animal Behaviour, 75, 583-593. Vié, J.-C., Hilton-Taylor, C. & Stuart, S. N. 2009. Wildlife in a changing world: an analysis of the 2008 IUCN Red List of threatened species. Glan, Switzerland: IUCN. Vogel, V., Pedersen, J. S., D’Ettorre, P., Lehmann, L. & Keller, L. 2009. Dynamics and genetic structure of Argentine ant supercolonies in their native range. Evolution, 63, 1627-1639. Vogel, V., Pedersen, J. S., Giraud, T., Krieger, M. J. B. & Keller, L. 2010. The worldwide expansion of the Argentine ant. Diversity and Distributions, 16, 170-186. Wagner, D., Tissot, M., Cuevas, W. & Gordon, D. M. 2000. Harvester ants utilize cuticular hydrocarbons in nestmate recognition. Journal of Chemical Ecology, 26, 2245-2257. Wetterer, J. K. 2010. Worldwide spread of the pharaoh ant, Monomorium pharaonis (Hymenoptera: Formicidae). Myrmecological News, 13, 115-129. Williams, D. F., Oi, D. H., Porter, S. D., Pereira, R. M. & Briano, J. A. 2003. Biological control of imported fire ants (Hymenoptera: Formicidae).American Entomologist, 49, 150-163. Wilson, E. O. 1975. The Insect Societies. Cambridge: Harvard University Press. Wilson, E. O. & Taylor, R. W. 1967. The ants of Polynesia (Hymenoptera: Formicidiae). Pacific Insects Monographs, 14, 1-106. Zarzuela, M. F. M. d., Leite, L. G., Marcondes, J. E. & Campos, A. E. d. C. 2012. Entomopathogens isolated from invasive ants and tests of their pathogenicity. Psyche: A Journal of Entomology, 2012.

38 Genetic diversity is more important than relatedness for nestmate discrimination in the pharaoh ant

Luigi Pontieri, Raoul van Oosten, Jelle S. van Zweden, Patrizia d’Ettorre & Jes Søe Pedersen

(Manuscript in preparation) Chapter 1 39 Genetic diversity is more important than relatedness for nestmate discrimination in the pharaoh ant

Luigi Pontieri1,*, Raoul van Oosten1,2, Jelle S. van Zweden1,3, Patrizia d’Ettorre4 & Jes Søe Pedersen1

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

2 Current address: Evolutionary Ecology Group, University of Antwerp, Groenenborgerlaan 171, B-2020 Antwerp, Belgium

3 Current address: Laboratory of Socioecology & Social Evolution, KU Leuven, Naamsestraat 59, B-3000 Leuven, Belgium

4 Laboratory of Experimental and Comparative Ethology (LEEC), University of Paris 13, Sorbonne Paris Cité, Villetaneuse, France

* Correspondence Luigi Pontieri Centre for Social Evolution Department of Biology University of Copenhagen Universitetsparken 15 DK-2100 Copenhagen Denmark E-mail: [email protected]

40 Abstract

In social insect colonies the discrimination between nestmates and non-nestmates is achieved through chemical communication using cuticular hydrocarbons as the primary recognition cues. When these labels have an endogenous origin, both the degree of within-colony genetic diversity and between-colony relatedness are expected to affect the efficiency of recognition. Here we studied the effect of these two factors on nestmate recognition ability in the invasive pharaoh ant, Monomorium pharaonis, by pairing colonies at different levels of genetic diversity and relatedness in two types of behavioral assays. Despite a general ability to distinguish nestmates from non-nestmates, the degree of relatedness between colonies had no influence on the overall level of discrimination, indicating a strong colony identity mediated by its chemical profile. Low genetically diverse colonies showed better nestmate recognition abilities than high diverse ones, and workers from low genetically diverse colonies were discriminating high diverse non-nestmates better than vice versa, indicating that discrimination is asymmetrical with respect to the relative level of genetic diversity of the opponents. This asymmetry may have implications for the colonies competitive abilities and can over time increase the invasiveness of the species.

Keywords: Monomorium pharaonis, invasive ants, genetic diversity, cuticular hydrocarbons, nestmate recognition, asymmetrical aggression

41 Introduction

Recognition of group members plays a key role in the evolution, maintenance and survivorship of cooperative social units. In social insect colonies, where individuals are usually kin, the ability to discriminate between nestmates and non-nestmates is of fundamental importance because not only does it prevent colony exploitation of resources by hetero- and conspecific intruders (Wilson 1975), but it also allows group members to gain indirect fitness benefits by ensuring that altruistic acts are directed towards close relatives (Hamilton 1964; Crozier & Pamilo 1996). In many social insects, and particularly in ants, the cues for nestmate recognition are primarily encoded in the complex mixture of lipids, mainly hydrocarbons, present on the cuticle of each individual (Lahav et al. 1999; Akino et al. 2004; for review see e.g. Howard & Blomquist 2005; van Zweden & d’Ettorre 2010). These odor labels can have a strong genetic component (Greenberg 1979; Stuart 1988; Beye et al. 1998; van Zweden et al. 2010) or can be acquired from the environment (Gamboa et al. 1986; Liang & Silverman 2000; Buczkowski et al. 2005; for review see e.g. d’Ettorre & Lenoir 2010). The process of nestmate discrimination is thought to occur on the basis of the “phenotype matching model” (Lacy & Sherman 1983) and involves two categories of participants: an evaluator, which performs the act of discrimination, and a cue-bearer, which is the entity being recognized (Liebert & Starks 2004). According to the model, a social insect evaluator detects the set of odor cues (label) borne by the cue bearer on the cuticle and compares this profile to an internal template, which is a neural representation of the pattern of cues present in its own colony. The evaluator forms the template likely learning the odors carried by all its nestmates (nestmate-referent phenotype matching; Buckle & Greenberg 1981; Gamboa et al. 1986; Errard 1994; van Zweden & d’Ettorre 2010). The magnitude of the mismatch between the label carried by the cue-bearer and the one stored in the template of the evaluator triggers an appropriate behavioral response: sufficiently large differences lead to the aggressive rejection of the cue- bearer; conversely, it will be treated friendly and recognized as nestmate (cf. van Zweden & d’Ettorre 2010).

As the evaluator builds up its template on the basis of the label carried by nestmates, it has been proposed that the level of within colony genetic diversity can affect the functioning of the recognition system if labels are under strong genetic determination. When the genetic diversity of the colony increases, so does the diversity in the odor labels used for recognition, and two main effects are then expected: (1) an increase in the breadth of the neural template, as this is based on a broader array of odors; (2) an increase in the potential for overlap with labels of other colonies (i.e. nestmate and non-nestmates becoming more similar). The ultimate result of these phenomena is that individuals from high diverse colonies might be less efficient in discriminating nestmates from non-nestmates. In fact, to avoid unwanted rejection errors (i.e. rejection of nestmates), ants from genetically diverse colonies would then apply a less stringent threshold, with the drawback that more non-nestmates will be erroneously accepted as well (acceptance errors; e.g. Reeve 1989). Conversely, individuals from low diverse colonies should possess a narrower template (based on a restricted array of odors) and better discrimination abilities because the cues carried by a non-nestmate cue- 42 bearer have fewer chances to overlap with those present in the evaluator’s template.

One interesting consequence of these discrepancies in recognition abilities between high and low genetically diverse colonies might be observed when individuals from high diverse colonies encounter individuals from low diverse ones. Because of the broader template and the more permissive threshold, the former might erroneously treat the latter as nestmates. However, individuals from low diverse colonies would immediately recognize as non-nestmates individuals from high diverse ones and consequently aggressively reject them. The overall result of this interaction would be “asymmetrical aggression” driven by the relative level of genetic diversity differences of the contenders. This phenomenon has indeed been observed between introduced populations of the invasive Argentine ant (Linepithema humile) in California: ants from low diversity colonies were initiating attacks against ants from high diversity colonies more often than vice versa (Tsutsui et al. 2003).

This result seems to support the hypothesis that higher level of genetic diversity reduces discrimination ability in social insect colonies. Further support appears to come from studies that analyzed the level of aggression of workers from single queen (i.e. monogynous) and multiple queens (i.e. polygynous) colonies towards foreign conspecific workers. Polygyny is thought to increase the genetic diversity and the recognition cue variability in the nest and consequently reduce the nestmate recognition skills of these colonies relative to monogynous ones (Keller & Passera 1989). Indeed, several studies reported these expected differences (Sundström 1997; Starks et al. 1998b; Pirk et al. 2001; Adams et al. 2007). However, several other studies have failed to report lower level of aggression and higher cuticular hydrocarbon diversity in polygynous colonies relative to monogynous ones (Satoh & Hirota 2005; Rosset et al. 2007; Martin et al. 2009; Helanterä et al. 2011), indicating that higher genetic diversity might not weaken the recognition abilities. Furthermore, a study on the Argentine ant has shown that factors such as territorial status (i.e. being the owner of the territory) and colony size might be more important in determining the outcome of aggressive interactions between low and high diverse colonies than the asymmetry in genetic diversity (Buczkowski & Silverman 2005).

A possible explanation for these contrasting results might be the extent to which odors are mixed throughout colony members in the different species to create a “Gestalt” colony odor (Crozier & Dix 1979). In the “Gestalt” model, members of the colony tend to share a common chemical signature created by the admixture of individual profiles by means of grooming, trophallaxis and/or contact with nest material (Soroker et al. 1994; Lenoir et al. 1999; Bos et al. 2011). The mixing of odors would reduce overall differences in odor cues among nestmates and increase differences with non-nestmates, hence maintaining good discrimination abilities despite increased genetic diversity (van Zweden et al. 2010; van Zweden & d’Ettorre 2010). However, different species might have different rates of cue transfer (e.g. due to absence of extended trophallaxis) and the lack of odor mixing can result in individuals of the same colony having different profiles (Nehring et al. 2011; Helanterä et al. 2013). Another possibility is that environmental cues

43 (e.g. from food and nest materials) play a prominent role over heritable ones, therefore decoupling genetic diversity and recognition abilities, and the relative contribution of these two sources to the recognition label can furthermore vary across species. Moreover, the expression of nestmate recognition can vary according to the social and ecological context where the discrimination process is performed (Bos et al. 2010), and different aggression bioassays can provide different results (Reeve 1989; Starks et al. 1998a; Roulston et al. 2003; Buczkowski & Silverman 2005). Finally, aggression is only one of the possible behaviors that individuals can perform and its absence does not necessarily imply a lack of nestmate discrimination (Steiner et al. 2007; Björkman-Chiswell et al. 2008; cf. d’Ettorre & Lenoir 2010).

In the present study we investigate the influence of within-colony genetic diversity and between colony relatedness on nestmate discrimination in the highly invasive pharaoh ant (Monomorium pharaonis). This species is considered the most ubiquitous ant in the world and one of the most important pest ants (Passera 1994; Wetterer 2010). The pharaoh ant is a perfect model system to address the effect of genetic diversity and relatedness on nestmate recognition as it is possible to perform controlled crosses and rear colonies indefinitely in the laboratory (Peacock & Baxter 1949; Schmidt et al. 2011). By crossing highly genetically divergent, low diverse lineages collected around the world we were able to create several “heterogeneous” colonies characterized by a higher level of genetic diversity relative to the parental lineages. We then paired colonies at different levels of genetic diversity and relatedness in two behavioral assays. Given the low level of intraspecific aggression reported for this species (Schmidt et al. 2010) we first investigated whether colonies were generally able to distinguish between nestmates and non-nestmates and whether discrimination was based on genetically derived cues. We furthermore predicted (1) that high diverse colonies would show less efficient nestmate discrimination than low diverse ones; (2) that individuals from low diverse colonies would discriminate individuals from high diverse colonies better than vice versa, thus displaying asymmetry in discrimination; and (3) a negative association of relatedness and the efficiency of discrimination, so more related colonies display on average a lower level of rejection than less related pairs.

Materials and methods

Colony origin and maintenance

For the present study we used 13 different laboratory colonies of M. pharaonis. Five colonies were highly genetically divergent, low genetic diversity lineages obtained either from field collection or from laboratory stocks prior to being reared in captivity at the University of Copenhagen since 2004 (1L colonies: U1, U2, U4, U5, U11; see table 1 in Schmidt et al. 2010). The remaining eight colonies were part of a genetically heterogeneous stock and created through the sequential crossing of either three or eight out of eight 1L colonies (U1, U2, U3, U4, U5, U7, U11 and Gh4, Schmidt et al. 2010) which were combined following the

44 approach of Hansen and Spuhler (1984; see figure 4 in the General introduction of this thesis). Six of these colonies were created by crossing three original 1L colonies (3L colonies: 18, 31, 44, 102, 111, 158; figure 1 in Schmidt et al. 2011), whereas two were created by sequentially crossing eight original 1L colonies (8L colonies: H318, H329). Colonies in the three groups (1L, 3L and 8L) can be therefore classified according to the degree of within-colony genetic variation: “low” for the L1 colonies, “medium” for the L3 colonies, and “high” for the L8 colonies.

Colonies were maintained in Fluon™ (De Monchy, the Netherlands) coated plastic boxes (27 × 17 × 9.5 cm) with empty plastic tubes sealed with cotton serving as nesting sites, and kept in a climate room at 25 ± 2 ˚C, 35 ± 5% RH and D:L 12:12 h. Colonies were fed twice a week with almonds, boiled egg yolk, cooked liver and mealworms (Tenebrio molitor), whereas water was provided ad libitum.

Aggression bioassays

We carried out two types of aggression bioassays (cf. Roulston et al. 2003): (1) dyadic encounters where two worker ants from different colonies meet in a neutral arena (live 1–1 assay) and (2) intercolony worker introduction where a single worker is placed within a foreign colony (introduction assay). For the live 1–1 assay we used nine colonies: five 1L colonies and four 3L colonies (31, 111, 102 and 158; Fig. 1a). For the introduction assay we used six colonies: two 1L (U1 and U4), two 3L (18 and 44) and two 8L (H318 and H329, Fig.1b). The aggression bioassays were conducted by two researchers, and the observer who recorded the data did not know the identity of the interacting colonies in order to avoid confirmation bias (van Wilgenburg & Elgar 2013). Type and duration of behavior performed by ants were recorded using the software Etholog version 2.25 (Ottoni 2000).

Live 1–1 assays

Aggression tests were conducted over 10 consecutive days. Each day, ant workers were collected from each colony and marked on the abdomen with either a red or a green oil-based marker (Edding). Marked workers were then placed in Fluon-coated Petri dishes according to their colony of origin, provided with water and food (see above) and allowed to rest overnight. The following day dyadic encounters were conducted in arenas consisting of small Fluon-coated Petri dishes (Ø = 2 cm, height = 1 cm). An open-ended, Fluon- coated ring was positioned at the center of the arena prior of each trial. Workers of opposite color markings were transferred into the arena, with one ant being placed inside the ring and one outside. After 1 min of acclimatization, the ring was removed and the ants were allowed to interact for 10 min. We recorded the duration of the following behaviors performed by each ant: antennation, mandibles open, bite and prolonged fight.

Each of the four 3L colonies was tested against two 1L colonies: one used in the crossing scheme for its creation (treatment “related”; N = 4, Fig. 1a) and one that was not (treatment “not related”; N = 4, Fig.

45 1a). If genetic similarity is important for recognition we expect the level of discrimination to be higher (e.g. more aggression) in the non-related colony pairs compared to the related pairs, and both colonies in the pair should have the same response. If genetic diversity is important for recognition the prediction of the polarized aggression hypothesis is that individuals from low diverse 1L colonies should discriminate individuals from high diverse 3L colonies better than vice versa (Tsutsui et al. 2003), e.g. we expect that the paired colonies have different responses.

We also carried out all pairwise combinations between 1L colonies (N = 10) that, combined with the eight combinations described above, constituted the general “non-nestmate” treatment (N = 18). Finally, we performed intracolony controls for each colony used in the assay (treatment “control”; N = 9) and these two treatments were used to investigate whether pharaoh ants are able to discriminate between nest-mates and non-nestmates.

Introduction assays

From each colony we created two subcolonies, each consisting of one queen, 10 workers and four brood pieces (mix of 2nd and 3rd larval instars) placed in a small arena (see above, Fig. 1b). In each trial an “intruder” worker, which was previously marked as in the live 1–1 assays, was introduced with a brusher in a fluon- coated ring positioned at the center of the arena. After 1 min of acclimatization, the ring was removed. The test began when the resident workers had the first interaction with the intruder. Over 2 min, the duration of the following behavior performed by the resident workers towards the intruder, ordered by increasing level of aggressiveness, was recorded: brief contact; investigation with the antennae; antennal boxing; open mandibles; bite; prolonged fight. When more than one resident worker simultaneously interacted with the intruder, the behavior with the higher aggression level was scored.

Each 1L colony was tested against one 3L colony (treatment “more related”, N = 2) and one 8L colony (treatment “less related”, N = 2, Fig. 1b) both created with the contribution of that specific 1L lineage. For each inter-colony combination we performed 60 replicates: each colony of the pair was used 30 times as resident (15 trials per subcolony) and 30 times as source of intruder workers. We also conducted intracolony controls for all the colonies used in the assay (N = 6) replicated 15 times (seven and eight trials per subcolony).

Each resident subcolony was tested in a pseudo-randomized order to avoid consecutive trials, providing resident workers with enough time to recover prior to the subsequent test. Trials were performed over five consecutive days and subcolonies were created at the beginning of each day, meaning that resident workers, the queen and brood of each subcolony were different from those tested the previous days. Intruder workers were not tested in more than one trial.

46 Chemical analysis of cuticular hydrocarbons

For chemical analysis of cuticular hydrocarbons four samples, each consisting of five workers, were collected from each colony used in the live 1–1 assay (N = 36) one day before the beginning of the aggression tests and stored in glass vials at –20 ˚C. Cuticular extracts were obtained by washing the samples for 10 min in 50

µl of a pentane solution with n-eicosane (n-C20) as internal size standard. Samples were vortexed for the first and the last minute of the immersion period. From each extract 4 µl were injected splitless into an Agilent Technologies 7890A Gas Chromatography (GC) System connected to an Agilent Technologies 5975C mass spectrometer. The oven temperature was kept at 70 °C for 1 min, then increased at 30 °C/min to 280 °C, then at 2°C/min to 320 °C, and then held at 320 °C for 3 min. After the identification of the compounds, we measured the area under each peak using the Agilent Technologies software MSD Chemstation version E 02.00.237.

Statistical analyses

All statistical analyses were conducted in R 3.0.2 (R Core Team 2013). For the analyses that used aggression as a binary variable (i.e. aggressive or not aggressive), we defined as aggressive those trials where at least one bite or one prolonged fight could be scored. Unless stated otherwise, the overall effect of fixed factors in Generalized Linear Mixed Models (GLMMs), General Linear Mixed Models (LMMs) and General Linear Models (GLMs) was assessed using Likelihood Ratio Tests (LRTs), comparing the full model with a reduced model not including the fixed factor but with the same random effect structure. Continuous response variables in LMMs were rank transformed when needed to achieve normality and homogeneity of variance. Multiple comparison of means (post-hoc tests) were performed using the function “ghlt” in the package multcomp (Hothorn et al. 2008).

As a previous study failed to report a significant difference in aggression between nestmate (NM) and non- nestmate (nNM) combinations in M. pharaonis (Schmidt et al. 2010), we first investigated, for both types of bioassays, whether such difference could be found by performing GLMMs with binomial error structure and logit link function using the function “glmer” in the package lme4 (Bates et al. 2013). The GLMM for the live 1–1 assay included the colony membership of the interacting ants (NM or nNM) as fixed effect, colony combination as random effect and the proportion of aggressive trials per colony combination as response variable. The GLMM for the introduction assay comprised the colony membership of the interacting ants (NM or nNM) as fixed effect, the resident colony as random effect and the proportion of aggressive trials per colony combination as response variable.

For both experiments we also tested by LMMs whether nestmate and non-nestmate colony combinations differed for the time ants spent interacting. For the live 1–1 assay we used the total time of interaction after the first contact as response variable, colony membership of both ants (NM or nNM) as fixed effect, trials nested into colony combination and colony combination as random effects. Similarly, for the introduction

47 assay, we used the time resident individuals spent interacting with the intruder as response variable, colony membership of both residents and intruder (NM or nNM) as fixed effect, trials nested into host colony and host colony as random effects.

Live 1–1 assays

We tested whether the number of aggressive trials differed between the “Related” and “Not related” treatments using a binomial GLM with treatment as fixed factor and the proportion of aggressive trials per colony combination as response variable. To investigate whether ants from low diverse colonies were more prone to act as attacker in an aggressive encounter than ants from high diverse colonies, for each aggressive trial we defined the attacker as the individual spending more time performing aggressive behavior (bites and prolonged fights) than its opponent (the receiver). The overall number of aggressive trials where low diverse ants were found more aggressive over the total number of aggressive trials was then used in an exact binomial test.

As asymmetry in discrimination might involve more subtle behavior than overt aggression, for each trial in the “related” and “not related” treatments the time each individual spent interacting with the opponent from the moment of the first interaction was used to determine a “more” and a “less” interacting individual. This information was then used to investigate whether (1) ants from 1L colonies were more prone to act as more interacting individual when paired with ants from 3L colonies and (2) whether the type of opponent (related or not related) could have an influence. The first question was addressed using a GLM with binomial error structure and logit link function with the proportion of trials per colony combination where ants from 1L colonies were found as “more interacting” as response variable. The second question was tested by adding to this model the type of treatment (related and not related) as fixed factor and assessing its overall effect through a LRT.

Introduction assays

For each trial we calculated the time the resident ants spent performing agonistic behavior (investigation with antennae, antennal boxing, mandible opening, bites or prolonged fights) towards the intruder. To test whether intruders were differentially treated by the resident colony according to their degree of genetic diversity, we performed a LMM using the rank transformed time as response variable, the degree of genetic diversity of the intruder relative to the resident colony as fixed effect (with three levels: “equal”; “less” and “more”) and resident colony as random effect. Differences among the three levels of the fixed effect were assessed through a post-hoc test with Bonferroni correction.

For each of the two nNM treatments (“less related” and “more related”) we tested whether intruders being more genetic diverse than the host colony received more agonistic behavior compared to less diverse intruders introduced into more diverse resident colonies. We did so by running LMMs using the time of

48 agonistic behavior as response variable, the genetic diversity of the intruder relative to the resident colony as fixed factor and the resident colony as random effect.

Cuticular hydrocarbon data

We identified 43 different compounds, which were a mix of unbranched, mono- and di-methyl branched alkanes and unsaturated hydrocarbons. As these compounds were invariably present, we focused our analysis on quantitative differences between colonies. Briefly, peak areas were transformed following the method proposed by Aitchinson: Zij = ln[Yij/g(Yj)], where Yij is the area of peak i for the individual j, g(Yj) is the geometric mean of the areas of all peaks for individual j, and Zij is the transformed area of peak i for individual j (Aitchison 1986). We then performed a discriminant analysis of principal components (DAPC; Jombart et al. 2010) on the transformed data using the R package “adegenet” (Jombart 2008) to identify variables that differed quantitatively between colonies. This approach first transforms the variables into uncorrelated components using a principal component analysis (PCA) and then a discriminant analysis (DA) is applied to a number of principal components retained by the user in order to maximize the between colony variation and minimize the variation within colony. We retained five PCs (variance explained 82%) that were submitted to the DA. In the DA, the grouping variable was the colony of origin of the samples.

Results

Live 1−1 assay

In five trials ants did not interact during the observation period, leaving 535 trials for further analyses. When only aggressive trials were considered, i.e. trials where at least one bite or one prolonged fight could be scored, we found that nNM combinations were significantly less aggressive than NM ones (number of aggressive trials: nNM = 87/358, 24%; NM = 4/177, 2.3%; LRT: χ2 = 25.062, df = 1, P < 0.001). In addition, when considering the total interaction time of the trial, we found that nestmate individuals interacted significantly less than non-nestmates (LRT: 2χ = 19.635, df = 1, P < 0.001; Estimate ± SE for nestmate: –0.685 ± 0.13).

We did not find a significant difference between the related and not related treatments in the proportion of aggressive trials, as only 17 out of 79 trials (22%) in the “related” treatment and 13 out of 79 trials (16%) in the “not related” treatment were found to be aggressive (LRT: χ2 = 0.659, df = 1, P = 0.418). When the number of aggressive trials of the two treatments were pooled (N = 30), we did not find a significant effect of genetic diversity on the likelihood of being the attacker in the encounter: ants from 1L colonies were more aggressive than 3L opponents in 18 trials; 11 times ants from 3L colonies were more aggressive whereas in one trial individuals were equally aggressive (exact binomial test: 18 out of 29, P = 0.265).

49 When the total time of interaction was considered, ants from 1L colonies were not more likely to act as the more interacting individual in the trial when facing a 3L opponent (binomial GLM: z = 0, P > 0.999, white and black bars in Fig. 2). However, when the level of relatedness to the opponent was considered (i.e. being in the related or not related treatment), we found that 1L ants had higher probabilities of being more interacting when facing related rather than unrelated opponents, although this difference only neared statistical significance (LRT: 2χ = 3.07, df = 1, P = 0.08, Fig. 2).

Introduction assay

Overall, we scored aggression (bites and/or prolonged fight) in 14 out of 90 NM trials (16%) and in 73 out of 240 nNM trials (30%). When controlling for the host colony, we found a statistical significant effect of being nestmate or non-nestmate on the proportion of aggressive trials, with non-nestmate intruders having higher probability of receiving aggressive acts (LRT: χ2 = 7.006, df = 1, P = 0.008; estimate ± SE for nNM: 0.8547 ± 0.338). Likewise, we found a significant effect of being a nestmate or non-nestmate on the time resident workers spent interacting with the intruder ant (LRT: χ2 = 6.4401, df = 1, P = 0.011), as host colonies interacted less with nestmate intruders (estimate ± SE for NM: -0.296 ± 0.116).

When only the time spent performing agonistic behavior was considered, we found that intruders were significantly differentially treated by host workers according to their level of genetic diversity (LRT: 2χ = 14.6, df = 2, P < 0.001). Particularly, host colonies spent more time interacting with more genetically diverse intruders compared to less or equally diverse (nestmates) ones (post-hoc tests with Bonferroni correction; less versus equal: P = 0.84; more versus equal: P < 0.001; more versus less: P = 0.049; Fig. 3a). However, we found that only in the “more related” treatment intruders characterized by a higher level of genetic diversity relative to the host colony were receiving significantly more agonistic behavior than those received by less diverse intruders when introduced in high diverse host colonies (LRT: χ2 = 8.04, df = 1, P = 0.004, Fig. 3b). This association, although similar, was not statistically significant in the “less related” treatment (LRT: χ2 = 0.38, df = 1, P = 0.54, Fig. 3b).

Chemical profiles of colonies used in the live 1−1 assay

The chemical profiles of the colonies used in the 1–1 assay were significantly discriminated through the DA on the five PCs we selected, and 94% of the samples were classified to the correct colony (Fig. S1). This suggests that each pharaoh ant colony possesses a specific odor, which in principle can provide the basis for an efficient discrimination of nestmates.

Discussion

This study shows that pharaoh ants are indeed capable of discriminating between nestmates and non-

50 nestmates independently of the ecological and social context of the encounters. Discrimination must be based on endogenous cues, as each colony has retained a specific chemical signature despite being maintained under uniform laboratory condition for >6 years (Fig. S1). However, overt aggression is rare and in general ants react mildly to perceived cue differences, mostly performing subtle behavior such as antennation and prolonged inspection. The discrimination process appears to be affected by the relative level of genetic diversity of the opponents, as workers from more diverse colonies are discriminated better when introduced into less diverse colonies than less diverse workers introduced in more diverse colonies (Fig. 3a,b). The degree of relatedness between colonies has a minor influence, as more and less related colony pairs show similar levels of discrimination (Fig. 3b). Nonetheless, relatedness seems to play a role in determining whether discrimination is polarized with respect to genetic diversity, as only related colonies show polarized discrimination behavior (Figs 2, 3b). The implications of these findings are discussed below.

Nestmate discrimination is based on heritable cues

In a previous study of the pharaoh ant Schmidt et al. (2010) reported both overall low level of aggression in live 1−1 assays among genetically divergent 1L colonies, and a significant difference in aggression between intra- and inter-colony pairs was only found after one year of laboratory rearing. Here we show that significant nestmate discrimination is maintained after minimum five more years of uniform rearing conditions. The discrimination abilities of the pharaoh ants therefore appear to be enhanced under uniform laboratory conditions rather than diminished or unchanged, as it has been shown in L. humile (Buczkowski et al. 2005) and Camponotus aethiops (van Zweden et al. 2009), respectively. Together with the fact that colonies possess a distinct chemical signature (Fig. S1), this is a strong indication that heritable cues and not cues derived from the environment constitute the nestmate recognition odor, thus meeting the assumption for the polarized aggression hypothesis.

Importantly, the overall level of overt aggression among colonies was similarly low in the two types of assays (see the Results section). This is remarkable because assays using different number of individuals and mimicking different ecological and social scenarios often vary in the likelihood of scoring aggressive acts (Roulston et al. 2003). The invariably low aggression in the pharaoh ant makes it important to investigate not only aggressive responses, but also more subtle behaviors that might be performed at different rates between nestmates and non-nestmates. Previous studies in other ant species have shown that non-nestmates engage in prolonged antennation and throphallaxis compared to nestmates (Chapuisat et al. 2005; Steiner et al. 2007; Björkman-Chiswell et al. 2008), indicating that lack of aggression does not necessarily imply lack of discrimination abilities. Here we found in both bioassays that non-nestmates not only are more aggressive, but they do also interact more than nestmates.

51 The effect of genetic diversity and relatedness on nestmate discrimination

The use of heritable odor cues in nestmate discrimination raises the question whether colony genetic variation and/or the relatedness of interacting individuals affects the efficiency of recognition. Assuming that the recognition system is based only on genetic labels, a series of predictions can be made. First, we expect that an increase in genetic diversity is accompanied by an increase in odor cue diversity within the colony, with the consequence that high diverse colonies exhibit less efficient discrimination abilities than more homogeneous ones because they will accept a higher number of non-nestmates (i.e. acceptance errors; Keller & Passera 1989; Reeve 1989; Starks et al. 1998b). Second, discrimination will be “polarized” when low and high diverse colonies meet, so individuals from high diverse colonies accept individuals from low diverse colonies at a higher rate than the opposite (Tsutsui et al. 2003). Third, there will be a negative association of relatedness and the efficiency of discrimination, so more related colonies display on average a lower level of rejection than less related pairs.

In the introduction assay, the observation that low genetically diverse colonies discriminate high diverse, non-nestmates intruders better than high diverse colonies discriminating individuals from less diverse colonies (Fig. 3a,b) is in accordance with the first and the second prediction. An increase in genetic diversity therefore appears both to lower the discrimination abilities of a colony and increase the probability that its individuals are recognized as non-nestmates by more homogeneous colonies. On the contrary, the degree of relatedness has a minor effect on the recognition process. In fact the overall level of discrimination was similar both for related and unrelated colony pairs in the 1–1 assay, and more and less related pairs in the introduction assay (Fig. 3b). This indicates that colonies have a strong “identity” sensu Moffett (2012), effectively treating all non-colony mates as “different” and explaining the lack of association between genetic distance and aggression in 1L pharaoh ant colonies (Schmidt et al. 2010). However, whether two colonies are related or not appears to be important for polarized discrimination behavior, indicating that relatedness act in concert with the level of within colony genetic diversity to determine the efficiency of recognition. In fact, only in treatments where related colonies were paired (the “related” treatment in the 1–1 assay and both treatments in the introduction assay) we observed that low diverse individuals were discriminating high diverse ones better than vice versa (Figs 2, 3b), although a significant effect was observed only for the “more related” treatment in the introduction assay.

A proposed mechanism and the ecological implications

Altogether these results can be explained by the following scenario: the increase in colony genetic diversity also increases the diversity of heritable recognition cues and broadens both the mix of cues carried by workers and the recognition template of the colony (Vander Meer & Morel 1998). It follows that workers from high diverse colonies have an increased probability of erroneously accepting foreign individuals compared to low diverse ones because it is more likely that the cue-bearer presents a profile that fits within

52 the range of the acceptable ones. Consequently, low diverse intruders are better able to integrate into high diverse colonies (i.e. being treated as nestmates) whereas high diverse intruders, possessing on average more variable profiles, are more efficiently discriminated by low diverse colonies. This asymmetry in discrimination is similar to the suggestion by Tsutsui et al. (2003), but we further propose that relatedness is a factor, so that the more two colonies are related, the stronger is the polarization in recognition ability. This is because low diverse individuals are more likely to have a profile representing a subset of heritable cues acceptable by the high diverse colony when the colonies are closely related. Furthermore, the asymmetry in discrimination is not expected for completely unrelated colonies. This hypothesis is consistent with our findings, but conclusive testing will likely require identification of the exact cues used for discrimination.

The findings in this study can have important ecological implications for the spread of pharaoh ant if colonies compete in the field and low diverse ones will have higher chance of expanding and survival due to their greater discrimination abilities. Invasive populations of pharaoh ants consist of highly genetically divergent, low diverse colonies which coexist over small spatial scales (Schmidt et al. 2010) and might indeed be those that survive and thrive after competition with more genetically diverse introductions (Tsutsui et al. 2003; Van Wilgenburg et al. 2010; Abril & Gómez 2011). Moreover, as introductions of supercolonial ants are often accompanied by a reduction in genetic diversity at the population level (Ross et al. 1996; Tsutsui et al. 2000; Giraud et al. 2002; Vogel et al. 2010), new invasive populations spreading in the introduced range may concurrently become more genetically homogenous and more competitive and thus attain higher invasion success over time.

Acknowledgement

LP and JSP were funded by a grant from the Danish National Research Foundation (grant DNRF57).

References

Abril, S. & Gómez, C. 2011. Aggressive behaviour of the two European Argentine ant supercolonies (Hymenoptera: Formicidae) towards displaced native ant species of the northeastern Iberian Peninsula. Myrmecological News, 14, 99-106. Adams, E., Atkinson, L. & Bulmer, M. 2007. Relatedness, recognition errors, and colony fusion in the termite Nasutitermes corniger. Behavioral Ecology and Sociobiology, 61, 1195-1201. Aitchison, J. 1986. The Statistical Analysis of Compositional Data. London: Chapman & Hall.

53 Akino, T., Yamamura, K., Wakamura, S. & Yamaoka, R. 2004. Direct behavioral evidence for hydrocarbons as nestmate recognition cues in Formica japonica (Hymenoptera: Formicidae). Applied Entomology and Zoology, 39, 381-387. Bates, D., Maechler, M., Bolker, B. & Walker, S. 2013. lme4: Linear mixed-effects models using Eigen and S4 (R package version 1.0-5). Beye, M., Neumann, P., Chapuisat, M., Pamilo, P. & Moritz, R. F. A. 1998. Nestmate recognition and the genetic relatedness of nests in the ant Formica pratensis. Behavioral Ecology and Sociobiology, 43, 67-72. Björkman-Chiswell, B. T., Wilgenburg, E., Thomas, M. L., Swearer, S. E. & Elgar, M. A. 2008. Absence of aggression but not nestmate recognition in an Australian population of the Argentine ant Linepithema humile. Insectes Sociaux, 55, 207-212. Bos, N., Grinsted, L. & Holman, L. 2011. Wax on, wax off: nest soil facilitates indirect transfer of recognition cues between ant nestmates. PLoS ONE, 6, e19435. Bos, N., Guerrieri, F. J. & d’Ettorre, P. 2010. Significance of chemical recognition cues is context dependent in ants. Animal Behaviour, 80, 839-844. Buckle, G. R. & Greenberg, L. 1981. Nestmate recognition in sweat bees (Lasioglossum zephyrum): does an individual recognize its own odour or only odours of its nestmates? Animal Behaviour, 29, 802-809. Buczkowski, G., Kumar, R., Suib, S. L. & Silverman, J. 2005. Diet-related modification of cuticular hydrocarbon profiles of the Argentine ant,Linepithema humile, diminishes intercolony aggression. Journal of Chemical Ecology, 31, 829-843. Buczkowski, G. & Silverman, J. 2005. Context-dependent nestmate discrimination and the effect of action thresholds on exogenous cue recognition in the Argentine ant. Animal Behaviour, 69, 741-749. Chapuisat, M., Bernasconi, C., Hoehn, S. & Reuter, M. 2005. Nestmate recognition in the unicolonial ant Formica paralugubris. Behavioral Ecology, 16, 15-19. Crozier, R. H. & Dix, M. 1979. Analysis of two genetic models for the innate components of colony odor in social Hymenoptera. Behavioral Ecology and Sociobiology, 4, 217-224. Crozier, R. H. & Pamilo, P. 1996. Evolution of Social Insect Colonies: Sex Allocation and Kin Selection. Oxford: Oxford University Press. d’Ettorre, P. & Lenoir, A. 2010. Nestmate recognition. In Ant Ecology (ed. L. Lach, C. L. Parr & K. L. Abbott), pp. 194-209: Oxford University Press. Errard, C. 1994. Development of interspecific recognition behavior in the antsManica rubida and Formica selysi (Hymenoptera: Formicidae) reared in mixed-species groups. Journal of Insect Behavior, 7, 83-99.

54 Gamboa, G. J., Reeve, H. K., Ferguson, I. D. & Wacker, T. L. 1986. Nestmate recognition in social wasps: the origin and acquisition of recognition odours. Animal Behaviour, 34, 685-695. Giraud, T., Pedersen, J. S. & Keller, L. 2002. Evolution of supercolonies: the Argentine ants of southern Europe. Proceedings of the National Academy of Sciences of the United States of America, 99, 6075-6079. Greenberg, L. 1979. Genetic component of bee odor in kin recognition. Science, 206, 1095-1097. Hamilton, W. D. 1964. The genetical evolution of social behaviour. I. Journal of Theoretical Biology, 7, 1-16. Hansen, C. & Spuhler, K. 1984. Development of the National Institutes of Health genetically heterogeneous rat stock. Alcoholism: Clinical and Experimental Research, 8, 477-479. Helanterä, H., Aehle, O., Roux, M., Heinze, J. & d’Ettorre, P. 2013. Family-based guilds in the ant Pachycondyla inversa. Biology Letters, 9, 4. Helanterä, H., Lee, Y. R., Drijfhout, F. P. & Martin, S. J. 2011. Genetic diversity, colony chemical phenotype, and nest mate recognition in the ant Formica fusca. Behavioral Ecology, 22, 710-716. Hothorn, T., Bretz, F. & Westfall, P. 2008. Simultaneous inference in general parametric models. Biometrical Journal, 50, 346-363. Howard, R. W. & Blomquist, G. J. 2005. Ecological, behavioral, and biochemical aspects of insect hydrocarbons. Annual Review of Entomology, 50, 371-393. Jombart, T. 2008. adegenet: a R package for the multivariate analysis of genetic markers. Bioinformatics, 24, 1403-1405. Jombart, T., Devillard, S. & Balloux, F. 2010. Discriminant analysis of principal components: a new method for the analysis of genetically structured populations. BMC Genetics, 11, 94. Keller, L. & Passera, L. 1989. Influence of the number of queens on nestmate recognition and attractiveness of queens to workers in the Argentine ant, Iridomyrmex humilis (Mayr). Animal Behaviour, 37, 733-740. Lacy, R. C. & Sherman, P. W. 1983. Kin recognition by phenotype matching. American Naturalist, 489- 512. Lahav, S., Soroker, V., Hefetz, A. & Vander Meer, R. K. 1999. Direct behavioral evidence for hydrocarbons as ant recognition discriminators. Naturwissenschaften, 86, 246-249. Lenoir, A., Fresneau, D., Errard, C. & Hefetz, A. 1999. Individuality and colonial identity in ants: the emergence of the social representation concept. In Information Processing in Social Insects (ed. C. Detrain, J. Deneubourg & J. Pasteels), pp. 219-237. Basel: Birkhäuser. Liang, D. & Silverman, J. 2000. ”You are what you eat”: diet modifies cuticular hydrocarbons and nestmate recognition in the Argentine ant, Linepithema humile. Naturwissenschaften, 87, 412-416.

55 Liebert, A. E. & Starks, P. T. 2004. The action component of recognition systems: a focus on the response. Annales Zoologici Fennici, 41, 747-764. Martin, S. J., Helanterä, H., Kiss, K., Lee, Y. R. & Drijfhout, F. P. 2009. Polygyny reduces rather than increases nestmate discrimination cue diversity in Formica exsecta ants. Insectes Sociaux, 56, 375-383. Moffett, M. W. 2012. Supercolonies of billions in an invasive ant: what is a society? Behavioral Ecology, 23, 925-933. Nehring, V., Evison, S. E., Santorelli, L. A., d’Ettorre, P. & Hughes, W. O. 2011. Kin-informative recognition cues in ants. Proceedings of the Royal Society B: Biological Sciences, 278, 1942- 1948. Ottoni, E. B. 2000. EthoLog 2.2: a tool for the transcription and timing of behavior observation sessions. Behavior Research Methods Instruments & Computers, 32, 446-449. Passera, L. 1994. Characteristics of tramp species. In Exotic Ants: Biology, Impact, and Control of Introduced Species. (ed. D. F. Williams), pp. 23-43. Boulder, Colorado: Westview Press. Peacock, A. & Baxter, A. 1949. Studies in Pharaoh’s ant, Monomorium pharaonis (L.). 1. The rearing of artificial colonies.Entomologist’s Monthly Magazine, 85, 256-260. Pirk, C. W. W., Neumann, P., Moritz, R. F. A. & Pamilo, P. 2001. Intranest relatedness and nestmate recognition in the meadow ant Formica pratensis (R.). Behavioral Ecology and Sociobiology, 49, 366-374. R Core Team. 2013. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Reeve, H. K. 1989. The evolution of conspecific acceptance thresholds.American Naturalist, 133, 407-435. Ross, K. G., Vargo, E. L. & Keller, L. 1996. Social evolution in a new environment: the case of introduced fire ants. Proceedings of the National Academy of Sciences, 93, 3021-3025. Rosset, H., Schwander, T. & Chapuisat, M. 2007. Nestmate recognition and levels of aggression are not altered by changes in genetic diversity in a socially polymorphic ant. Animal Behaviour, 74, 951-956. Roulston, T. H., Buczkowski, G. & Silverman, J. 2003. Nestmate discrimination in ants: effect of bioassay on aggressive behavior. Insectes Sociaux, 50, 151-159. Satoh, T. & Hirota, T. 2005. Factors affecting internest variation in the aggressiveness of a polygynous ant, Camponotus yamaokai. Entomological Science, 8, 277-281. Schmidt, A. M., d’Ettorre, P. & Pedersen, J. S. 2010. Low levels of nestmate discrimination despite high genetic differentiation in the invasive pharaoh ant. Frontiers in Zoology, 7, 20.

56 Schmidt, A. M., Linksvayer, T. A., Boomsma, J. J. & Pedersen, J. S. 2011. No benefit in diversity? The effect of genetic variation on survival and disease resistance in a polygynous social insect. Ecological Entomology, 36, 751-759. Soroker, V., Vienne, C., Hefetz, A. & Nowbahari, E. 1994. The postpharyngeal gland as a “Gestalt” organ for nestmate recognition in the ant Cataglyphis niger. Naturwissenschaften, 81, 510-513. Starks, P. T., Fischer, D. J., Watson, R. E., Melikian, G. L. & Nath, S. D. 1998a. Context-dependent nestmate discrimination in the paper wasp, Polistes dominulus: a critical test of the optimal acceptance threshold model. Animal Behaviour, 56, 449-458. Starks, P. T., Watson, R. E., Dipaola, M. J. & Dipaola, C. P. 1998b. The effect of queen number on nestmate discrimination in the facultatively polygynous ant Pseudomyrmex pallidus (Hymenoptera: Formicidae). Ethology, 104, 573-584. Steiner, F. M., Schlick-Steiner, B. C., Moder, K., Stauffer, C., Arthofer, W., Buschinger, A., Espadaler, X., Christian, E., Einfinger, K., Lorbeer, E., Schafellner, C., Ayasse, M. & Crozier, R. H. 2007. Abandoning aggression but maintaining self-nonself discrimination as a first stage in ant supercolony formation. Current Biology, 17, 1903-1907. Stuart, R. J. 1988. Collective cues as a basis for nestmate recognition in polygynous leptothoracine ants. Proceedings of the National Academy of Sciences, 85, 4572-4575. Sundström, L. 1997. Queen acceptance and nestmate recognition in monogyne and polygyne colonies of the ant Formica truncorum. Animal Behaviour, 53, 499-510. Tsutsui, N. D., Suarez, A. V. & Grosberg, R. K. 2003. Genetic diversity, asymmetrical aggression, and recognition in a widespread invasive species. Proceedings of the National Academy of Sciences of the United States of America, 100, 1078-1083. Tsutsui, N. D., Suarez, A. V., Holway, D. A. & Case, T. J. 2000. Reduced genetic variation and the success of an invasive species. Proceedings of the National Academy of Sciences of the United States of America, 97, 5948-5953. van Wilgenburg, E. & Elgar, M. A. 2013. Confirmation bias in studies of nestmate recognition: a cautionary note for research into the behaviour of animals. PLoS ONE, 8, e53548. Van Wilgenburg, E., Torres, C. W. & Tsutsui, N. D. 2010. The global expansion of a single ant supercolony. Evolutionary Applications, 3, 136-143. van Zweden, J. S., Brask, J. B., Christensen, J. H., Boomsma, J. J., Linksvayer, T. A. & d’Ettorre, P. 2010. Blending of heritable recognition cues among ant nestmates creates distinct colony gestalt odours but prevents within‐colony nepotism. Journal of Evolutionary Biology, 23, 1498-1508. van Zweden, J. S. & d’Ettorre, P. 2010. Nestmate recognition in social insects and the role of hydrocarbons. In Insect Hydrocarbons: Biology, Biochemistry and Chemical Ecology (ed. G. J. Blomquist & A. G. Bagnères), pp. 222-243. Cambridge: Cambridge University Press.

57 van Zweden, J. S., Dreier, S. & d’Ettorre, P. 2009. Disentangling environmental and heritable nestmate recognition cues in a carpenter ant. Journal of Insect Physiology, 55, 159-164. Vander Meer, R. K. & Morel, L. 1998. Nestmate recognition in ants. In Pheromone Communication in Social Insects (ed. R. K. Vander Meer, M. D. Breed, K. E. Espelie & M. L. Winston), pp. 79-103. Boulder: Westview Press. Vogel, V., Pedersen, J. S., Giraud, T., Krieger, M. J. B. & Keller, L. 2010. The worldwide expansion of the Argentine ant. Diversity and Distributions, 16, 170-186. Wetterer, J. K. 2010. Worldwide spread of the pharaoh ant, Monomorium pharaonis (Hymenoptera: Formicidae). Myrmecological News, 13, 115-129. Wilson, E. O. 1975. The Insect Societies. Cambridge: Harvard University Press.

58 Figure Captions

Fig. 1. Experimental set-up with Monomorium pharaonis colony combinations and treatments in the (a) live 1−1 and (b) introduction assays. Each colony is represented with the “♀” symbol and different colors indicate different original lineages. 3L and 8L colonies are shown with the original 1L colonies’ contribution to the female’s genome.

(a)

All pairwise combinations (N = 10)

Treatments

Related 1L colonies Not related 111 31 102 158 Control U1

(b) U7 U11

18 H 318 Gh4 Treatments

Less related × 4 More related 44 H 329 × 1 × 10 Control Intruder worker

59 Fig. 2. Asymmetry in interaction time between low and high genetically diverse Monomorium pharaonis colonies in the live 1−1 assay. Number of trials where less (black bars) and more (white bars) genetically diverse individuals were interacting for a longer time than the opponent are shown according to the two treatments (“Related” and “Not related”). Numbers on top of the bar indicate the total number of trials in the treatment.

80 Less diverse

More diverse 60 N = 79 N = 79

20 More interacting individual interacting More

0 Related Not related Treatments

60 Fig. 3. Asymmetry in agonistic behavior in the Monomorium pharaonis introduction assay. (a) Box-plots showing the time resident workers spent performing agonistic behavior (investigation with antennae, antennal boxing, mandible opening, bites and prolonged fights) towards intruders characterized by equal (“Equal” = control treatment), lower (“Less”) or higher (“More”) level of genetic diversity. Different letters indicate significant differences (P < 0.05) between groups. (b) Box-plots showing the asymmetry in the time spent performing agonistic behavior according to the relative genetic diversity of resident (host) and intruder workers according to the two treatments (“More related” and “Less related”). ** P < 0.01; NS not significant.

(a) 100

40 b a viors (%) a 20 beh a

0 Equal Less More (Control) Relative intruder diversity (b) 100 ming agonistic o r 80 More related

60 Less related ** NS 40 Time spent per f 20

0 Host 1L 3L 1L 8L Intruder 3L 1L 8L 1L

61 Fig. S1. Chemical profiles of the nine Monomorium pharaonis colonies used in the live 1−1 assay. Discriminant analysis (DA) of 36 chemical samples (each consisting of five nestmate workers) based on five principal components retained for the analysis (insert), showing discrimination among the nine colonies. The proportion of variance explained by each of the two discriminant functions is provided on the axis legends.

15 1L colonies

U2 10 U4

5 U11

3L colonies 0 102

111

158 -5 iminant function 2 (28%) 31 Disc r -10 80 40 0 0 10 20 30 -15 PCA axis -20 -15 -10 -5 0 5 10 15 Discriminant function 1 (53%)

62 Unexpected fusion of genetically and chemically divergent colonies of the invasive pharaoh ant

Luigi Pontieri & Jes Søe Pedersen

(Manuscript in preparation) Chapter 2 63 Unexpected fusion of genetically and chemically divergent colonies of the invasive pharaoh ant

Luigi Pontieri1,* & Jes Søe Pedersen1

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

* Correspondence Luigi Pontieri Centre for Social Evolution Department of Biology University of Copenhagen Universitetsparken 15 DK-2100 Copenhagen Denmark E-mail: [email protected]

64 Abstract

Many invasive ant species form distinct supercolonies that are able to persist over large spans of time and space, and strong mutual aggression is thought to maintain the colonies’ integrity. In the pharaoh ant (Monomorium pharaonis) colonies are extremely genetically divergent, even within tropical populations where colonies are spatially close and free to move. This suggests that local mixing even of related colonies is very limited, and that budding often leads to new isolated lineages. Paradoxically, the ant shows low discrimination of nestmates raising the question whether genetic differentiation is sufficient to prevent unrelated colonies to fuse. In this study, experimental colonies were chosen from a breeding scheme and paired at three levels of genetic similarity (relatedness) in a fusion assay with queens, workers and brood in nest boxes connected by a common foraging arena. The distribution and social interactions of the paired colonies were followed for 10 days. We found that 94 % of even unrelated colonies merged despite initial fighting of workers, and 65 % of all fusion were completed within 40 h. The initial aggression was associated with both chemical and genetic distances of the colony pairs. This shows that the low level of nestmate discrimination generally found in the pharaoh ant has the potential to lead to fusion of colonies, thus restoring genetic diversity. Why such fusion appears not to occur in natural populations of spatially close colonies could be due to environmental factors that may also apply to other supercolonial ant species.

Keywords: colony fusion, ants, Monomorium pharaonis, nestmate recognition, cuticular hydrocarbons, genetic diversity.

65 Introduction

The majority of eusocial insect species are characterized by multicolonial populations, where individuals live in small, kin structured colonies that aggressively defend their territories and maintain colony integrity by preventing both heterospecific and conspecific intruders from entering the nest (Hölldobler & Wilson 1990). In contrast, some ant species display a peculiar social system, called unicoloniality (Hölldobler & Wilson 1977), where colonies can grow to vast networks of geographically separated but mutually tolerant nests (Bourke & Franks 1995). Individuals freely move between nests within the network without eliciting aggression and the energy that might have been invested in defending a territory is instead directed towards colony growth, foraging and interspecific competition (Holway et al. 1998). As the lack of intraspecific aggression within these huge cooperative units provides a competitive edge over other ant species through a more effective resource allocation, unicoloniality has been often invoked as the key factor explaining the enormous ecological success achieved by invasive ants (Holway et al. 2002), where this social organization appears to be over-represented (Tsutsui & Suarez 2003). These species, intriguingly, also share a suite of life-history traits collectively called “invasive ant syndrome” (Cremer et al. 2008) that is thought to facilitate their introduction, establishment and range expansion (Moller 1996; McGlynn 1999; Tsutsui & Suarez 2003): extreme polygyny, general nesting and food habits, colony reproduction by budding and high worker and nest density (Holway & Suarez 1999).

Although “true” unicoloniality (i.e. a single colony per population) appears to be a phenomenon mostly found in recent (<100 years) introductions (Helanterä et al. 2009) and many of the unicolonial populations described so far contains multiple, mutually aggressive “supercolonies” (Pedersen et al. 2006; Foucaud et al. 2009; Helanterä et al. 2009; Sunamura et al. 2009; Drescher et al. 2010), the evolution and maintenance of such social organization nonetheless poses a paradox for kin-selection theory because altruistic behaviour occurs among unrelated nestmates (Queller & Strassmann 1998).

Several hypotheses have been formulated to explain how unicoloniality can arise in spite of nestmates being unrelated (Helanterä et al. 2009). Many of them focus on the important role of diversity in recognition cues and discrimination behavior (Tsutsui et al. 2000; Giraud et al. 2002), as the main feature of unicoloniality is the lack of aggression between non-nestmates. The discrimination of nestmates in social insects is mainly achieved by chemical communication and, especially in ants, the recognition cues are primarly encoded in the cuticular hydrocarbon profile (van Zweden & d’Ettorre 2010). These odor labels, that can be genetically and/or environmentally derived (Gamboa et al. 1986; Stuart 1987; Liang & Silverman 2000), are mixed among all group members, thereby creating a Gestalt colony odour (Crozier & Dix 1979). By smelling the Gestalt odor, individuals form a neural representation of their own colony’s recognition cues (the template) that will be matched against the label carried by the conspecifics they encounter (Lenoir et al. 1999). When the dissimilarity between the neural template and the odor label carried by the encountered individual is

66 above a certain threshold, the individual will be rejected (Reeve 1989). This threshold is thought to be affected by several factors, including the level of polygyny (Reeve 1989), so that colonies with many queens will have reduced efficiency in rejecting non-nestmates. The assumption is that the colony odor will be less distinct due to a high within-colony genetic variation (Starks et al. 1998b; Satoh & Hirota 2005). It follows from purely genetically based odor labels that intercolony interactions (particularly aggression between workers) can be predicted by the degree of chemical and genetic dissimilarity between the colonies, and by genetic variation within colonies, independent of their environment.

The highly invasive Argentine ant (Linepithema humile) is known for its “closed societies” of mutually aggressive supercolonies based on an efficient inheritance of cues maintaining colony identity over large spans of time and space (reviewed by Moffett 2012) which is likely a key characteristic of supercolony organization (Pedersen 2012). Nevetheless, it was recently shown that introduced colonies of the Argentine ant can merge under laboratory conditions, and that colony fusion is modulated by levels of aggressive interactions, as well as by chemical and genetic similarity of the interacting pairs (Vasquez & Silverman 2008; Vasquez et al. 2009). Furthermore, the same authors have showed that even mutually aggressive Argentine ant colonies can occasionally fuse and experience considerable benefits in terms of increased colony growth (Vasquez et al. 2012). These findings have led the authors to speculate that fusion of incipient, unrelated Argentine ant colonies might be a possible mechanism to explain supercolony formation in this species. Although this hypothesis has been challenged (Helanterä et al. 2009; Vogel et al. 2010), it is relevant to test whether, in absence of ecological constraints, unrelated colonies of other invasive ant species can fuse.

The pharaoh ant (Monomorium pharaonis) is a unicolonial “tramp” species (Berndt & Eichler 1987; Passera 1994) that may have the potential to form supercolonies by fusion of unrelated, incipient colonies. Distributed worldwide (Wetterer 2010), in the temperate regions this ant is commonly found in buildings, depends from human for food and nesting sites and their nests are quickly relocated when subjected to physical disturbance (Buczkowski & Bennett 2009). However, in tropical areas colonies are also found outdoors, are spatially close and free to move. Unlike other invasive ants that form vast supercolonies (Giraud et al. 2002), pharaoh ant supercolonies are smaller and highly genetically differentiated even on a small spatial scale (Schmidt et al. 2010). Intriguingly, this ant appears to be tolerant of unrelated conspecifics, therefore behaving quite differently when compared to the mutually aggressive supercolonies of the Argentine ant. This scenario raises the question of why spatially close colonies of M. pharaonis do not seem to merge under natural conditions.

Here we tested whether, in absence of ecological constraints, genetic differentiation between colonies of Monomorium pharaonis prevents colonies from fusing. To assess this hypothesis, we conducted a laboratory fusion assay where experimental M. pharaonis colonies with increasing genetic distance were paired. We also determined the dissimilarity of the colonies’ cuticular hydrocarbon profiles to investigate its influence on possible colony fusion. Furthermore, we investigated whether within-colony genetic variation correlates

67 with cuticular hydrocarbon profile diversity.

Materials and methods

Colony origin and maintenance

For the present study, we used 16 different source colonies of Monomorium pharaonis being part of a larger laboratory stock reared in captivity at the University of Copenhagen. Twelve colonies were part of a genetically heterogeneous stock created through the crossing of either three or seven out of eight highly genetically divergent, low genetic diversity colony lineages (1L colonies; U1, U2, U3, U4, U5, U7, U11 and Gh4; Schmidt et al. 2010) obtained either from laboratory stocks or from field collection. Eight colonies were created by the sequential crossing of three original lineages (3L colonies), which were combined following the approach of Hansen and Spuhler (Hansen & Spuhler 1984; figure 1 in Schmidt et al. 2011), whereas four colonies were created sequentially crossing seven lineages (7L colonies). The breeding protocol was based on 10–20 gynes and 4–15 males at each cross (see figure 4 in the General introduction of this thesis). The remaining four colonies used were collected in Thailand (1L colonies; T1, T2, T3 and T4) and not used in the crossing scheme described above (Schmidt et al. 2010). We classified the colonies in three groups according to the degree of within-colony genetic variation: “low” for the 1L colonies, “medium” for the 3L colonies, and “high” for the 7L colonies (Schmidt et al. 2011).

All source colonies were maintained in Fluon™ (De Monchy, the Netherlands) coated plastic boxes (27 × 17 × 9.5 cm) in a climate room at 25 ± 2 ˚C, 35 ± 5% RH and D:L 12:12 h. Colonies were fed twice a week with almonds, boiled egg yolk, cooked liver and mealworms (Tenebrio molitor), whereas water was provided ad libitum. Colonies were also provided with empty plastic tubes sealed with cotton, serving as nesting sites.

Experimental treatments

Four of the eight 3L colonies were used as “reference” colonies (111, 44, 64 and 92; Fig. 1). Each of these colonies was paired in four treatments with colonies of increasing genetic distance: (1) “control”, where each colony was paired with itself; (2) “high relatedness”, where the reference colony was paired with one of the remaining four 3L colonies so that the pair shared two out of their three original lineages; (3) “low relatedness”, where the reference colony was paired with one of the four 7L colonies so that all lineages of the reference colony were represented in the other colony; and (4) “unrelated”, where the reference colony was paired with one of the four 1L colonies from Thailand. We obtained three replicates for each control and six replicates for each pair of the other three treatments.

68 Experimental setup

Each source colony used was first split in two subcolonies, each consisting of an approximately equal number of queens, workers and brood at different stages. Then, each subcolony was marked either blue or red by providing a glass tube containing a sugar solution in which a water-soluble food dye was diluted (Natural Red Food Colour or Blue Food Colour, Dr. Oetker). This non-invasive method allowed us to color both workers and pre-pupae stages, as workers passed part of the ingested colored solution to brood (Fig. 2a–c). Queens were marked on the abdomen with either a blue or a red oil-based marker (Edding). Colonies parts were kept in Fluon-coated Petri dishes for one week to allow an effective acquisition of the marking. From each colony part, several experimental colonies were created, each consisting of one marked queen, 30 colored brood pieces and 50 colored workers. Then, each experimental colony was paired with a differently colored counterpart according to the four different treatments described above.

Experimental colonies were maintained in individual Fluon-coated plastic boxes (7.5 × 11.5 × 4.8 cm) and provided with folded paper serving as nest. Nest boxes were connected by a vinyl tube to an identical box serving as foraging arena, with rubber plugs initially inserted at each end to prevent contact between colonies (Fig. 2d). During a 24-h acclimation period, colonies were also provided with the sugar solution containing the water-soluble dye in order to enhance the marking. During the fusion assays, cooked liver and a glass tube containing a sugar solution were provided ad libitum only in the foraging arena to promote continuous encounters between ants. All experimental colonies were kept at the same laboratory condition as the source colonies.

Colony fusion assay

The colony fusion experiment began when food provided during the acclimation period and the rubber plugs blocking the vinyl tube were removed, allowing interactions between the paired colonies. Over 10 days of observation, we recorded the total number of worker pairs fighting (only biting, gaster flexing and dragging were considered) during 2 min time checks performed after 3 h from the beginning of the assay, then twice per day (early morning and late afternoon) for the first four days and then once every day. The presence of sexual larvae, as well as the location and distribution patterns of brood piles and queens was assessed during the same time periods. To allow a correct identification of mixing of workers and the distribution of queens and brood, we also took pictures of each replicate by placing the experimental set-up on a glass table with a digital camera (Canon EOS 600D) positioned underneath. We were able to identify seven types of outcome: (1) Complete fusion: both queens and an equal amount of mixed brood located in the same chamber, and no aggression recorded between the colonies throughout the observation period; (2) Complete fusion with aggression: both queens and an equal amount of mixed brood in the same chamber, but aggression between workers recorded at least in one of the checks; (3) Brood fusion with separated queens: queens located in different chambers, both surrounded by at least one pile of mixed brood; (4) Queen fusion with killed brood:

69 both queens and the brood of only colony found together, whereas brood of the other colony has been killed; (5) Queen fusion with separated brood: both queens and the brood of only one colony found together, with brood of the other colony still present in another chamber; (6) Brood fusion with killed/isolated queen: one queen with mixed brood in one chamber, with the other queen isolated or dead; (7) No fusion: all cases where we did not find mixing of queens and/or brood. When assigning fusion type to replicates, we always chose the highest level of fusion observed.

We considered all these categories because pharaoh ants are polydomous and known to quickly relocate their nest in response to physical disturbance (Buczkowski & Bennett 2009). Therefore defining as fusion only those events where queens and brood from different colonies are found together in the same nest (Vasquez & Silverman 2008) might lead to overlook temporal and spatial distribution patterns that could change the genetic composition of the colony (i.e. brood adoption, fusion and subsequent split due to disturbance) and that should be instead considered as fusion events. The six categories of fusion fall into two main categories according to what extent the original colonies contribute to the fused colony. Types 1–3 are regarded as “fair” fusions as both original colonies have similar chances to be represented in the new colony’s future production of sexual brood. In contrast, types 4–6 are “unfair” fusions as the original colonies do not have the same chances of contributing to future reproduction.

Chemical analysis of cuticular hydrocarbons

For chemical analysis of cuticular hydrocarbons two samples, each consisting of five workers, were collected from each of the split source colonies (marked red and blue, respectively; N = 4 per colony) four days before the beginning of the fusion assays and stored in glass vials at –20 ˚C. Cuticular extracts were obtained by washing the samples for 10 min in 50 µl of a pentane solution with n-eicosane (n-C20) as internal size standard. Samples were vortexed for the first and the last minute of the immersion period. From each extract 4 µl were injected splitless into an Agilent Technologies 7890A Gas Chromatography (GC) System connected to an Agilent Technologies 5975C mass spectrometer. The oven temperature was kept at 70 °C for 1 min, then increased at 30 °C/min–1 to 280 °C, then at 2 °C/min–1 to 320 °C, and then held at 320 °C for 3 min. After the identification of the compounds, we measured the area under each peak using the Agilent Technologies software MSD Chemstation version E 02.00.237. Three samples of colony 44 and two samples of colony 23 were discarded because the signal-to-noise ratio was too low, leaving one and two samples respectively for these two colonies available for statistical analysis.

Statistical analysis

All statistical analyses were carried out using R 3.0.1 (R Core Team 2013). To investigate whether genetic distance of colony pairs had an effect on the probability of fusion, we ran a Generalized Linear Mixed Model (GLMM) with binomial error structure and logit link function using the function glmer in the package

70 “lme4” (Bates et al. 2013). We included treatment as fixed effect, colony pairs nested within treatment as random effect, and the fusion event as binary dependent variable. As an overall test of the effect of treatment we used a likelihood ratio test (LRT) comparing this full model with a null model comprising only the intercept and the same random effects structure. Mean separation between groups with Tukey post-hoc test was carried out using the function glht in the package “multcomp” (Hothorn et al. 2008).

To determine whether genetic distance of colony pairs had an effect on the number of aggressive interactions recorded at the beginning of the fusion assays, we ran a GLMM with a poisson error structure and log link function including treatment as fixed effect, colony pairs nested within treatment as random effect and the number of worker pairs fighting during the first observation as response variable. A similar model was run to assess whether genetic distance between pairs influenced the aggression level over time using the number of worker pairs fighting in the checks following the first one as response variable. Since no aggression was recorded in the colony combinations nested in the “control” group at any check, we replaced one of the “zero” values with “one” to allow both models to compute the estimates for this group. The overall effect of genetic distance was assessed in both models using a LRT, comparing these models with their respective reduced models comprising only the intercept and the same random effect. Group means were then separated as described above.

To test whether the initial aggression was a predictor of the probability of fusion, we ran a GLM with binomial error structure and logit link function including colony fusion as binary dependent variable and the number of worker pairs fighting during the first check as independent variable. A similar analysis was performed using the total number of worker pairs observed fighting in the other checks as independent variable and colony fusion as dependent variable, to assess whether continuous aggressive interactions over time could influence the probability of fusion.

In the chemical analysis we identified 43 different compounds, which were a mix of unbranched, mono- and di-methyl branched alkanes and unsaturated hydrocarbons. As these compounds were invariably present, we focused our analysis on quantitative differences between colonies. Peak areas were transformed following the method proposed by Aitchinson (1986): Zij = ln[Yij/g(Yj)], where Yij is the area of peak i for the individual j, g(Yj) is the geometric mean of the areas of all peaks for individual j, and Zij is the transformed area of peak i for individual j. We then performed a discriminant analysis of principal components (DAPC) (Jombart et al. 2010) on the transformed data using the R package “adegenet “ (Jombart 2008) to identify variables that differed quantitatively between colonies. This approach first transforms the variables into uncorrelated components using a principal component analysis (PCA). Then a discriminant analysis (DA) is applied to a number of principal components retained by the user in order to maximize the between colony variation and minimize the variation within colony. We retained seven PCs (82.3 % of variance explained) that were submitted to the DA (supplementary Fig. S1). In the DA, the grouping variable was the colony of origin of the samples (the single sample from colony 44 was excluded from the analysis). The 21 hydrocarbons that

71 were found to contribute the most to separate the different colonies (i.e. hydrocarbons with high between- to-within colony variation ratio) were then used to calculate the Euclidean distance of the colony pairs used in the fusion assays.

We used the “global centroid” method (see chapter 3) to quantify the chemical distance between colonies. For each colony combination, the arithmetic mean of each of the 21 most contributing compounds over all included individual samples was calculated (i.e. the global centroid). Subsequently the average difference between individual samples and this global centroid was determined. This implies that for colonies with very different values for a particular hydrocarbon, the average difference between individual samples and the global average will be large, whereas this difference will be small when values are very similar. For nestmate controls only samples from a single colony are included, so the global centroid will correspond to the colony centroid (see Fig. 2 in chapter 3). The chemical distance was then calculated by taking the square root of the sum of squared differences over all variables (Euclidean distance).

To assess whether the cuticular hydrocarbon dissimilarity of colony pairs could explain the aggressive interactions of workers during the fusion assays either at the beginning of the experiment or over the remaining observation period, we ran two GLMs with quasi binomial error structure to account for overdispersion of the data and logit link function. In these we included the proportion of aggressive replicates during the 1st check and in the following checks as dependent variables and the Euclidean distance between colony pairs as independent variable.

To test whether the genetic distance of colony pairs could explain their cuticular hydrocarbon dissimilarity, we ran an ANOVA using treatment and colony pairs nested into treatment as fixed effect and the Euclidean distance between pairs, calculated as previously described, as response variable. A similar test was used to assess the relationship between the level of genetic variation of each colony and its chemical distance from the origin of the Euclidean space, using treatment as fixed effect (“high”, “medium” and “low”, see Materials and Methods) and the chemical distance of each sample from the origin of the Euclidean space as response variable. In this last model, the distance from the origin was calculated using all the 43 identified hydrocarbons. Group means were separated as described above.

Results

Even though the type of fusion varied, the vast majority of Monomorium pharaonis colony pairs effective fused, with the observations spanning from all (N = 46) controls of same-colony origin to 94 % (68/72) of all pairs of unrelated colonies fusing in the 10 days observation period (Fig. 3a). Moreover, 65 % (75/114) of these fusions were completed within 40 h. We found an overall significantly negative effect of genetic distance between the paired colonies on the probability of fair colony fusion (LRT: χ2 = 9.56, df = 3, P =

72 0.023; Fig. 3b). However, none of the four treatment means were significantly different from each other in a multiple comparison test. On the other hand, colony pair relatedness significantly affected the number of aggressive interactions in the 1st check (LRT: χ2 = 49.85, df = 3, P < 0.001; Fig. 3c), with all experimental treatments showing significantly higher numbers of worker pairs fighting inst the1 check compared to the control, and with the unrelated treatment showing more aggressive interactions compared to the other experimental treatments. Genetic distance of colony pairs was also a significant factor for the number of aggressive interactions recorded in the other checks (LRT: χ2 = 11.429, df = 3, P = 0.009; Fig. 3c). However, we observed a drastic reduction in the number of aggressive interactions after the first check in all the experimental treatments, with only the “unrelated” treatment still showing a significantly higher number of worker pairs fighting compared to the nestmate control.

When considered separately, both the number of aggressive interactions recorded in the 1st check and in the other checks were significant predictors of the probability of fair fusion of the colonies (aggression 1st check: z = –2.624, df = 116, P = 0.009; Fig. 4a; total aggression other checks: z = –2.456, df = 116, P = 0.014; Fig. 4b). However, when both factors were included in the model, only the number of aggressive interactions recorded after the 1st check remained significant (aggression st1 check: z = –0.172, df = 115, P = 0.863; total aggression other checks: z = –2.069, df = 115, P = 0.038), indicating that it was a better predictor of the probability of fair fusion than the aggressive interactions at the 1st check.

The chemical profiles of the colonies were significantly discriminated through the DA on the seven PCs retained, and 79% of the samples were correctly classified (Fig. S1).

The chemical distance of colony pairs was a good predictor of the proportion of replicates being aggressive at the 1st check (quasi binomial GLM: t = 3.263, df = 25, P = 0.003; Fig. 4c) but not for the following checks (quasi binomial GLM: t = 0.331, df = 25, P = 0.74; Fig. 4d).

We found an overall significantly positive effect of genetic distance between the paired colonies on their chemical distance (F3,23 = 4.61, P = 0.011; Fig. 3d), with the chemical distance of “unrelated” colonies being significantly higher than the nestmate controls. We also found a strong and statistically significant effect of the within-colony genetic variation on the Euclidean distance of each chemical sample from the origin of the

Euclidean space (F2,55 = 7.925, P < 0.001; Fig. 5), with samples collected from colonies with “low” genetic variation being on average more distant from the origin, i.e. the population mean, compared to samples collected from colonies with “high” genetic variation.

Discussion

The most remarkable finding of this study is that the vast majority (94 %) of even unrelated pharaoh ant colonies merged within few days, comparable with fusion rates for more closely related colony pairs (Fig.

73 3a). This was not due to universal acceptance of conspecifics (Steiner et al. 2007) or lack of discrimination ability (Tsutsui et al. 2000; Giraud et al. 2002), as workers of the same colonies typically fought during the first hours of encounters (Figs 3c, 4a). The chemical cues used for discrimination had a clear genetic basis (Figs 3d, 4c) as assumed for supercolonies to maintain their long term identity (Moffett 2012). However, even continued aggression did often not prevent colony fusion, and in about 1/3 of the cases these fusions were “unfair” (Fig. 3b) in the sense that one of the original colonies had a better chance for future reproduction in the merged colony. These results raise questions about the efficiency of discrimination behaviour, the adaptive significance of colony fusion, and the association of genes and cues for recognition, which will be discussed in the following.

Intercolonial aggression and its effect on colony fusion

We found overall high levels of aggressive interactions between unrelated colonies at the beginning of the experiment. Furthermore, the probability of observing an initial aggression between two interacting colonies appears to be determined by the degree of genetic and cuticular hydrocarbon dissimilarities. These results are unexpected as they contrast what was previously found in this species, reporting both an overall low level of aggression between distantly related colonies and a lack of correlations between aggression and chemical and genetic distances (Schmidt et al. 2010). These diverging results of the two studies could be explained by the type of bioassay used to assess the aggressive behaviors of the interacting ants. While Schmidt et al. (2010) performed dyadic encounters in a neutral arena, we recorded the aggressive interactions between workers directly in the fusion assay. These two types of bioassays, which drastically differ both for the number of ants involved and for the ecological contexts they mimic, have been showed to reveal significant differences in the probability of scoring aggressive acts (Roulston et al. 2003). This is probably due to discrimination behaviour to be dependent on context and allowing individuals to adjust their acceptance threshold according to the ecological and social environment (Reeve 1989; Buczkowski & Silverman 2005). Defending the nest from intruders or competing for food likely enhanced the probability that workers engage in aggressive behaviour in our study (Holway 1999), as individuals can gain better fitness payoffs than those that might be obtained by fighting a foreign opponent during a casual encounter in a neutral area (Starks et al. 1998a). This hypothesis is further supported by the fact that most of the aggressive worker pairs we recorded were observed in the close proximity of the nest and around the food source in the foraging area. The observed association between initial aggression and chemical dissimilarity between colonies (Fig. 4c) clearly suggests that pharaoh ants are able to discriminate nestmates from non- nestmates. However, whether this ability is expressed in overt aggression is likely dependent on the social and ecological context of the actual encounters.

The strong aggression recorded at the beginning of the fusion assay declined in less than 24 h in almost all the unrelated colony pairs, remaining low or disappearing over the rest of the observation period. Conversely, we

74 observed an increasing number of colony pairs fusing as soon as aggression subsided. This situation might explain why the probability of final fusion was much better predicted by the number of aggressive worker pairs recorded after the first observation rather than those observed at the first check. Our findings agree with what has been previously found in Argentine ant, where fused colony pairs experienced a significant reduction in aggression levels over time (Vasquez & Silverman 2008). Fusion after reduction of worker- worker aggression might be explained by the initial elimination of the most aggressive individuals, or those possessing the least common recognition cues, which in turn might facilitate the merging of more similar individuals. This phenomenon has been observed in an arboreal nesting termites (Leponce et al. 1996) and might explain colony fusion after high levels of workers aggression and mortality in L. humile (Vasquez & Silverman 2008). However, unlike Argentine ants, pharaoh ants are barely able to kill opponents in one-on- one encounters. Indeed, we observed that workers engaged in fights were simply releasing each other after some time and walking away without visible damages. This sort of “transient” aggression between workers has been observed also in the clonal ant Pristomyrmex punctatus when unrelated colonies are forced to compete for limited overwintering nesting sites and the cost of fighting might be too high (Satoh & Hirota 2013). These observations suggest that reduction in aggressive interactions between workers after initial hostility and subsequent colony fusion might be due to different mechanisms rather than elimination of the least phenotypically and /or genetically similar workers.

In our study the reduced aggressiveness over time could result from a learning process, whereby worker ants may habituate to odor cues provided by repeatedly encountered strangers and therefore displaying a reduced responsiveness towards such cues (Stroeymeyt et al. 2010). The habituation mechanism has been invoked as possible explanation of why neighbors are treated less aggressively than more distant strangers (i.e. “dear enemy phenomenon”, Temeles 1994), although the presence of such effect in social insects is debated and also the opposite pattern can be found (Langen et al. 2000; Boulay et al. 2007; Saar et al. 2014). Possibilities of repeated encounters between workers, as well as the drastic reduction in aggressiveness observed, suggest that a habituation mechanism towards foreign odor labels might be at work in our fusion assay. However, this process alone it is not sufficient to explain why two colonies would eventually merge.

Reduced aggressiveness between workers and the decision of two unrelated colonies to merge rather than staying separate may be instead an adaptive strategy that could turn to be successful in particular ecological and social contexts. Nest sites and food availability, colony size and queen numbers are known to be factors that can shape intercolonial interactions in social insects (Hölldobler & Wilson 1977; Hölldobler & Wilson 1990; Foitzik & Heinze 1998). In our experiment, we used fairly small colonies each containing only one queen. Given the polygynous nature of M. pharaonis, the presence of a single reproductive individual might not ensure an efficient brood production over time, consequently reducing competitive ability and survival odds of the colony. The fusion of unrelated groups that are either small or have become queenless might therefore provide benefits in terms of colony productivity, defence against predators and resource

75 exploitation abilities that could supersede the inclusive fitness costs of reduced relatedness, as it has already been shown in the Argentine ant (Vasquez et al. 2012), African army ants (Kronauer et al. 2010), and have been suggested as a general mechanism for supercolony formation and maintenance (Steiner et al. 2007). Cohesion of the newly formed colony might then be created and maintained by homogenization of the original colony recognition labels to form a new Gestalt odour (Vasquez et al. 2009), likely through trophallaxis, preventing nepotistic behaviors and intracolony conflicts (Chapuisat et al. 2005).

Colony fusion as survival strategy for incipient colonies

The scenario we propose above can explain how spatially close, unrelated colonies of pharaoh ants might fuse despite being highly genetically distinct. However, the low intracolonial genetic diversity and the high intercolonial genetic differentiation found in contiguous field colonies suggests that colony fusion is, at best, a rare phenomenon in this species (Schmidt et al. 2010). One possible explanation is that, although found in close proximity, field colonies of pharaoh ants do not meet. Unlike our experimental setup, the presence of multiple food sources might lead colonies to avoid intercolonial battles and to possess non-overlapping foraging territories, therefore reducing the future probability of their encounter.

Another possibility is that colony size and queen number might prevent unrelated colonies from merging, and our experimental colonies do not reflect the size of the well-established field colonies. Nonetheless, these experimental units may still represent, to some extent, a natural situation. In M. pharaonis colonies are continuously expanded through budding, where small groups of workers, brood and one or more queens leave the main nest to start a new nest nearby (Peacock et al. 1955). Our fusion assay might thus mimic the intercolonial interactions occurring between unrelated, newly founded nests at the periphery of their respective supercolonies. Fusion might therefore be an occasional event that happens mainly at the border between supercolonies, which can not only increase the chance of survival or the competitive abilities of these groups, but also represent a useful mechanism to restore within-colony genetic diversity. This possible function is supported by the observation that field colonies of pharaoh ants, although forming small and isolated breeding units, still posess some genetic diversity (Schmidt et al. 2010).

Fusion between incipient, unrelated colonies could also be an important process at the beginning of the invasion process in M. pharaonis, where only small colony fragments are accidentally spread by human mediated jump dispersal in new places and likely struggle for survival. Whether fusion between unrelated propagules can happen in the field remains to be tested, but our results suggest that this is indeed possible.

Genetic background and its effect on the cuticular hydrocarbon profile

The analysis of the cuticular hydrocarbon spectra of the colonies used in this study provides a series of clues indicating that the relative proportion of each odor label that constitutes the complex chemical signature of M. pharaonis might have a genetically derived origin. A first argument to support this hypothesis is given by

76 the observation that colonies possess a specific and distinguishable gestalt odor, as the discriminant analysis correctly assigned individuals to their colony of origin based on their cuticular hydrocarbon profile (Fig. S1). Although this result might not seem surprising for a social insect field colony, it is quite remarkable for colonies kept under identical diet and laboratory condition for several years. If the relative proportions of odor cues were mostly depending by environmental factors such as type and amount of food, nest material, temperature and relative humidity, then the homogeneous rearing conditions should have led colonies’ profiles to converge (van Zweden et al. 2009). The persistence of a specific chemical signature over time despite uniform environmental conditions also matches what was found before in this species (Schmidt et al. 2010) and suggest that genetic factors play a stronger role in the determination of the chemical profile of a colony than environmental factors. Further support is also provided by the finding that the chemical distance of colony pairs are associated with both the degree of genetic similarity and the initial level of intercolonial aggression.

A final support to the genetic determination of nestmate recognition cues in M. pharaonis is the inverse relationship that appears to exist between the level of within-colony genetic variation and the chemical distance to the population mean. The analysis of the cuticular hydrocarbon spectra indeed reveals that the average profile distance from the population mean is significantly higher for samples of highly inbred, low diverse colonies than for more diverse ones. This difference indicates that genetically homogeneous colonies tend to have a broader range of variation in the relative proportion of discrimination cues compared to colonies with higher level of heterozygosity. Assuming recognition cues are genetically determined, a different level of variation in odor labels has been often associated with polygyny (Keller & Passera 1989; Starks et al. 1998b; Satoh & Hirota 2005). The presence of multiple reproducing queens usually increases the genetic diversity of the colony, which in turn might increase the degree of odor cue diversity compared to monogynous nests. However, when odor labels are shared among nestmates to create a Gestalt colony odor, differences in the cuticular profile across group members fade. The homogenization of recognition cues in the nest reduces the variation across nestmates and likely decrease the likelihood of nepotistic behaviors (van Zweden et al. 2010). On the other hand, this mechanism can lead polygynous colonies to possess Gestalt odors that are more similar to each other compared to the similarity that might exist across monogynous nests. As consequence, polygynous nests might present lower level of cue diversity and profiles that are much closer to the species mean than monogynous ones (Martin et al. 2009).

Pharaoh ant colonies, including those used in this experiment, all possess multiple reproducing queens and the sole level of gyny would thus not explain the differences in odor cue variability that we observed across group of colonies. However, we need to consider that polygyny can increase the within-colony genetic variation only when queens are genetically different from each other. Pharaoh ants colonies can usually be considered as independent breeding unit because mating takes place within the nest. Exclusive intranidal mating will over time reduce the amount of genetic variation in the colony, with the consequence that all

77 the reproducing queens after several generation of inbreeding might share the same genes. In this scenario, a polygynous colony could be considered as “functionally monogynous” because group members would be much more related to each other than individuals in a stereotypical polygynous colony. When odor cues are genetically determined, the lack of intercolonial exchanges might lead these colonies to possess divergent Gestalt odors if the level of genetic differentiation between colonies is high, which also result in an overall high level of variation in the proportion of odor labels across colonies. In our experiment the “1L” colonies collected in Thailand seems to match this scenario, as they are highly genetically differentiated but extremely inbred. On the other hand, the other colonies we used in the experiment are crosses of original lineages and therefore possess different degree of within-colony genetic variation. In these colonies the odour homogenization leading to the formation of the gestalt colony profile might rapidly dilute the existing differences between original lineages. As consequence, their chemical profiles could present high level of overlapping and might also be closer to the species mean compared to the profile showed by more isolated lineages. The proposed scenario indeed explain the pattern of cue diversity observed in the three groups of colonies in our experiment, with the level of within-colony genetic variation determining the distance of the gestalt odor from the species mean. All together, these results show that the genetic component might drastically affect both the chemical profile and the recognition system ofM. pharaonis.

Acknowledgments

We thank Patrizia d’Ettorre for help with the chemical analysis of cuticular hydrocarbons, Mette Frimodt Hansen for technical assistance in the fusion assays, Jelle van Zweden for discussions on the statistical analysis of the colony chemical profiles, and Jelle van Zweden for helpful comments on previous versions of the manuscript. This study was supported by the Danish National Research Foundation (grant DNRF57) to the Centre for Social Evolution.

References

Aitchison, J. 1986. The Statistical Analysis of Compositional Data. London: Chapman & Hall. Bates, D., Maechler, M., Bolker, B. & Walker, S. 2013. lme4: Linear mixed-effects models using Eigen and S4 (R package version 1.0-5). Berndt, K. P. & Eichler, W. 1987. Die Pharaoameise, Monomorium pharaonis (L.) (Hym., Myrmicidae). Mitteilungen aus dem Museum für Naturkunde in Berlin. Zoologisches Museum und Institut für Spezielle Zoologie (Berlin), 63, 3-186.

78 Boulay, R., Cerdá, X., Simon, T., Roldan, M. & Hefetz, A. 2007. Intraspecific competition in the ant Camponotus cruentatus: should we expect the ‘dear enemy’ effect? Animal Behaviour, 74, 985- 993. Bourke, A. F. & Franks, N. R. 1995. Social evolution in ants. Princeton: Princeton University Press. Buczkowski, G. & Bennett, G. 2009. Colony budding and its effects on food allocation in the highly polygynous ant, Monomorium pharaonis. Ethology, 115, 1091-1099. Buczkowski, G. & Silverman, J. 2005. Context-dependent nestmate discrimination and the effect of action thresholds on exogenous cue recognition in the Argentine ant. Animal Behaviour, 69, 741-749. Chapuisat, M., Bernasconi, C., Hoehn, S. & Reuter, M. 2005. Nestmate recognition in the unicolonial ant Formica paralugubris. Behavioral Ecology, 16, 15-19. Cremer, S., Ugelvig, L. V., Drijfhout, F. P., Schlick-Steiner, B. C., Steiner, F. M., Seifert, B., Hughes, D. P., Schulz, A., Petersen, K. S., Konrad, H., Stauffer, C., Kiran, K., Espadaler, X., d’Ettorre, P., Aktaç, N., Eilenberg, J., Jones, G. R., Nash, D. R., Pedersen, J. S. & Boomsma, J. J. 2008. The evolution of invasiveness in garden ants. PLoS ONE, 3, e3838. Crozier, R. H. & Dix, M. 1979. Analysis of two genetic models for the innate components of colony odor in social Hymenoptera. Behavioral Ecology and Sociobiology, 4, 217-224. Drescher, J., Blüthgen, N., Schmitt, T., Bühler, J. & Feldhaar, H. 2010. Societies drifting apart? Behavioural, genetic and chemical differentiation between supercolonies in the yellow crazy ant Anoplolepis gracilipes. PLoS ONE, 5, e13581. Foitzik, S. & Heinze, J. 1998. Nest site limitation and colony takeover in the ant Leptothorax nylanderi. Behavioral Ecology, 9, 367-375. Foucaud, J., Orivel, J., Fournier, D., Delabie, J. H. C., Loiseau, A., Le Breton, J., Cerdan, P. & Estoup, A. 2009. Reproductive system, social organization, human disturbance and ecological dominance in native populations of the little fire ant, Wasmannia auropunctata. Molecular Ecology, 18, 5059-5073. Gamboa, G. J., Reeve, H. K., Ferguson, I. D. & Wacker, T. L. 1986. Nestmate recognition in social wasps: the origin and acquisition of recognition odours. Animal Behaviour, 34, 685-695. Giraud, T., Pedersen, J. S. & Keller, L. 2002. Evolution of supercolonies: the Argentine ants of southern Europe. Proceedings of the National Academy of Sciences of the United States of America, 99, 6075-6079. Hansen, C. & Spuhler, K. 1984. Development of the National Institutes of Health genetically heterogeneous rat stock. Alcoholism: Clinical and Experimental Research, 8, 477-479. Helanterä, H., Strassmann, J. E., Carrillo, J. & Queller, D. C. 2009. Unicolonial ants: where do they come from, what are they and where are they going? Trends in Ecology & Evolution, 24, 341-349.

79 Hölldobler, B. & Wilson, E. O. 1977. The number of queens: an important trait in ant evolution. Naturwissenschaften, 64, 8-15. Hölldobler, B. & Wilson, E. O. 1990. The Ants. Cambridge: Harvard University Press. Holway, D. A. 1999. Competitive mechanisms underlying the displacement of native ants by the invasive Argentine ant. Ecology, 80, 238-251. Holway, D. A., Lach, L., Suarez, A. V., Tsutsui, N. D. & Case, T. J. 2002. The causes and consequences of ant invasions. Annual Review of Ecology and Systematics, 33, 181-233. Holway, D. A. & Suarez, A. V. 1999. Animal behavior: an essential component of invasion biology. Trends in Ecology & Evolution, 14, 328-330. Holway, D. A., Suarez, A. V. & Case, T. J. 1998. Loss of intraspecific aggression in the success of a widespread invasive social insect. Science, 282, 949-952. Hothorn, T., Bretz, F. & Westfall, P. 2008. Simultaneous inference in general parametric models. Biometrical Journal, 50, 346-363. Jombart, T. 2008. adegenet: a R package for the multivariate analysis of genetic markers. Bioinformatics, 24, 1403-1405. Jombart, T., Devillard, S. & Balloux, F. 2010. Discriminant analysis of principal components: a new method for the analysis of genetically structured populations. BMC Genetics, 11, 94. Keller, L. & Passera, L. 1989. Influence of the number of queens on nestmate recognition and attractiveness of queens to workers in the Argentine ant, Iridomyrmex humilis (Mayr). Animal Behaviour, 37, 733-740. Kronauer, D. J. C., Schöning, C., d’Ettorre, P. & Boomsma, J. J. 2010. Colony fusion and worker reproduction after queen loss in army ants. Proceedings of the Royal Society B: Biological Sciences, 277, 755-763. Langen, T. A., Tripet, F. & Nonacs, P. 2000. The red and the black: habituation and the dear-enemy phenomenon in two desert Pheidole ants. Behavioral Ecology and Sociobiology, 48, 285-292. Lenoir, A., Fresneau, D., Errard, C. & Hefetz, A. 1999. Individuality and colonial identity in ants: the emergence of the social representation concept. In Information Processing in Social Insects (ed. C. Detrain, J. Deneubourg & J. Pasteels), pp. 219-237. Basel: Birkhäuser. Leponce, M., Roisin, Y. & Pasteels, J. M. 1996. Intraspecific interactions in a community of arboreal nesting termites (Isoptera: Termitidae). Journal of Insect Behavior, 9, 799-817. Liang, D. & Silverman, J. 2000. ”You are what you eat”: Diet modifies cuticular hydrocarbons and nestmate recognition in the Argentine ant, Linepithema humile. Naturwissenschaften, 87, 412-416. Martin, S. J., Helanterä, H., Kiss, K., Lee, Y. R. & Drijfhout, F. P. 2009. Polygyny reduces rather than increases nestmate discrimination cue diversity in Formica exsecta ants. Insectes Sociaux, 56, 375-383.

80 McGlynn, T. P. 1999. The worldwide transfer of ants: geographical distribution and ecological invasions. Journal of Biogeography, 26, 535-548. Moffett, M. W. 2012. Supercolonies of billions in an invasive ant: what is a society? Behavioral Ecology, 23, 925-933. Moller, H. 1996. Lessons for invasion theory from social insects. Biological Conservation, 78, 125-142. Passera, L. 1994. Characteristics of tramp species. In Exotic Ants: Biology, Impact, and Control of Introduced Species. (ed. D. F. Williams), pp. 23-43. Boulder, Colorado: Westview Press. Peacock, A., Sudd, J. & Baxter, A. 1955. Studies in Pharaoh’s Ant, Monomorium pharaonis (L.). 11. Colony foundation. Entomologist’s Monthly Magazine, 91, 125-129. Pedersen, J. S. 2012. The logic of hypersocial colonies. Behavioral Ecology, 23, 934-935. Pedersen, J. S., Krieger, M. J. B., Vogel, V., Giraud, T. & Keller, L. 2006. Native supercolonies of unrelated individuals in the invasive Argentine ant. Evolution, 60, 782-791. Queller, D. C. & Strassmann, J. E. 1998. Kin selection and social insects. Bioscience, 165-175. R Core Team. 2013. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Reeve, H. K. 1989. The evolution of conspecific acceptance thresholds.American Naturalist, 133, 407-435. Roulston, T. H., Buczkowski, G. & Silverman, J. 2003. Nestmate discrimination in ants: effect of bioassay on aggressive behavior. Insectes Sociaux, 50, 151-159. Saar, M., Leniaud, L., Aron, S. & Hefetz, A. 2014. At the brink of supercoloniality: genetic, behavioral and chemical assessments of population structure of the desert ant Cataglyphis niger. Frontiers in Ecology and Evolution, 2. Satoh, T. & Hirota, T. 2005. Factors affecting internest variation in the aggressiveness of a polygynous ant, Camponotus yamaokai. Entomological Science, 8, 277-281. Satoh, T. & Hirota, T. 2013. Colony fusion in a parthenogenetic ant, Pristomyrmex punctatus. Journal of Insect Science, 13. Schmidt, A. M., d’Ettorre, P. & Pedersen, J. S. 2010. Low levels of nestmate discrimination despite high genetic differentiation in the invasive pharaoh ant. Frontiers in Zoology, 7, 20. Schmidt, A. M., Linksvayer, T. A., Boomsma, J. J. & Pedersen, J. S. 2011. No benefit in diversity? The effect of genetic variation on survival and disease resistance in a polygynous social insect. Ecological Entomology, 36, 751-759. Starks, P. T., Fischer, D. J., Watson, R. E., Melikian, G. L. & Nath, S. D. 1998a. Context-dependent nestmate discrimination in the paper wasp, Polistes dominulus: a critical test of the optimal acceptance threshold model. Animal Behaviour, 56, 449-458.

81 Starks, P. T., Watson, R. E., Dipaola, M. J. & Dipaola, C. P. 1998b. The effect of queen number on nestmate discrimination in the facultatively polygynous ant Pseudomyrmex pallidus (Hymenoptera: Formicidae). Ethology, 104, 573-584. Steiner, F. M., Schlick-Steiner, B. C., Moder, K., Stauffer, C., Arthofer, W., Buschinger, A., Espadaler, X., Christian, E., Einfinger, K., Lorbeer, E., Schafellner, C., Ayasse, M. & Crozier, R. H. 2007. Abandoning aggression but maintaining self-nonself discrimination as a first stage in ant supercolony formation. Current Biology, 17, 1903-1907. Stroeymeyt, N., Guerrieri, F. J., van Zweden, J. S. & d’Ettorre, P. 2010. Rapid decision-making with side-specific perceptual discrimination in ants.PLoS ONE, 5, e12377. Stuart, R. 1987. Transient nestmate recognition cues contribute to a multicolonial population structure in the ant, Leptothorax curvispinosus. Behavioral Ecology and Sociobiology, 21, 229-235. Sunamura, E., Hatsumi, S., Karino, S., Nishisue, K., Terayama, M., Kitade, O. & Tatsuki, S. 2009. Four mutually incompatible Argentine ant supercolonies in Japan: inferring invasion history of introduced Argentine ants from their social structure. Biological Invasions, 11, 2329-2339. Temeles, E. J. 1994. The role of neighbours in territorial systems: when are they ’dear enemies’? Animal Behaviour, 47, 339-350. Tsutsui, N. D. & Suarez, A. V. 2003. The colony structure and population biology of invasive ants. Conservation Biology, 17, 48-58. Tsutsui, N. D., Suarez, A. V., Holway, D. A. & Case, T. J. 2000. Reduced genetic variation and the success of an invasive species. Proceedings of the National Academy of Sciences of the United States of America, 97, 5948-5953. van Zweden, J. S., Brask, J. B., Christensen, J. H., Boomsma, J. J., Linksvayer, T. A. & d’Ettorre, P. 2010. Blending of heritable recognition cues among ant nestmates creates distinct colony gestalt odours but prevents within‐colony nepotism. Journal of Evolutionary Biology, 23, 1498-1508. van Zweden, J. S. & d’Ettorre, P. 2010. Nestmate recognition in social insects and the role of hydrocarbons. In Insect Hydrocarbons: Biology, Biochemistry and Chemical Ecology (ed. G. J. Blomquist & A. G. Bagnères), pp. 222-243. Cambridge: Cambridge University Press. van Zweden, J. S., Dreier, S. & d’Ettorre, P. 2009. Disentangling environmental and heritable nestmate recognition cues in a carpenter ant. Journal of Insect Physiology, 55, 159-164. Vasquez, G. M., Schal, C. & Silverman, J. 2009. Colony fusion in Argentine ants is guided by worker and queen cuticular hydrocarbon profile similarity. Journal of Chemical Ecology, 35, 922-932. Vasquez, G. M. & Silverman, J. 2008. Intraspecific aggression and colony fusion in the Argentine ant. Animal Behaviour, 75, 583-593.

82 Vasquez, G. M., Vargo, E. L. & Silverman, J. 2012. Fusion between southeastern United States Argentine ant colonies and its effect on colony size and productivity. Annals of the Entomological Society of America, 105, 268-274. Vogel, V., Pedersen, J. S., Giraud, T., Krieger, M. J. B. & Keller, L. 2010. The worldwide expansion of the Argentine ant. Diversity and Distributions, 16, 170-186. Wetterer, J. K. 2010. Worldwide spread of the pharaoh ant, Monomorium pharaonis (Hymenoptera: Formicidae). Myrmecological News, 13, 115-129.

83 Figure Captions

Fig. 1. Colonies of Monomorium pharaonis used in the fusion assays and the proportion of the females’ genome contributed by the original lineages. Experimental treatments are indicated by arrows (for simplicity shown only for the 1st series). See text for details.

Original lineages 1st series U1

U2 111 12 5021 T1 U3

U4 2nd series U5

U7 44 23 4009 T2 U11

Gh4 3rd series

Treatments 64 13 4013 T3 control high relatedness low relatedness 4th series unrelated

92 18 4112 T4

84 Fig. 2. Color markings and set up of the Monomorium pharaonis fusion assays with (a) a blue worker (b), a red worker, and (c) marked brood surrounded by workers of both colors during the experiment. (d) The experimental set up with two nest boxes connected by a common box for foraging.

(a) (c)

(b)

(d)

85 Fig. 3. Results of the Monomorium pharaonis fusion assays according to the four treatment groups (n = replicate sample size) having different relatedness of the colony pair: (a) proportions of the seven types of outcome, (b) proportion of replicates with fair fusion (type 1–3), (c) number of worker pairs (mean ± SE) displaying overt aggression during the first check and in the other checks, respectively, and (d) the chemical distance (mean ± SE) between colony pairs. In (c) different letters before the comma indicate significant differences (P < 0.05) across groups during the first check, whereas different letters after the comma represent significant differences across groups in the rest of the observation period. In (d), different letters show significant differences between groups in the average chemical distance between colony pairs.

n = 46 n = 24 n = 24 n = 24 (a) 100

Fusion Type 1 80 Fusion Type 2 Fusion Type 3 60 Fusion Type 4

40 Fusion Type 5 Fusion Type 6

20 No Fusion Proportion of replicates (%) of replicates Proportion

0 (b) 100

80 Fair fusion Unfair/No fusion 60

20 Proportion of replicates (%) of replicates Proportion

9 1st check

Other checks 6 (cumulative)

B , A

3 B , A

A , A

0 Aggression (no. of worker pairs ghting)

(d) 2 B

AB AB 1.5

1.0 A

0.5 Chemical distance between pairs

0 High Low Control Unrelated Relatedness Relatedness

86 Fig. 4. Relationship between the occurrence of fair fusion of Monomorium pharaonis colonies and the number of worker pairs displaying overt aggression during (a) the first check and (b) over the rest of the observation period. Relationship between chemical distance of colony pairs and the proportion of replicates where pairs of worker pairs fighting during (a) the first check and (b) over the rest of the observation period, respectively. Jitter is added on the y-axes in panels (a) and (b) to allow a better visibility of data markers.

(a) (c) 100 YES check) st 80

Fair colony fusion events NO 0

0 5 10 15 20 % Aggressive replicates (1 0.5 1.0 1.5 2.0 Aggression (1st check) (b) (d) 100 YES 80

Fair colony fusion events NO 0 0 5 10 15 20 25 30 35 0.5 1.0 1.5 2.0 % Aggressive replicates (other checks) Aggression (other checks) Euclidean distance

87 Fig. 5. Mean distance of the chemical samples from the origin (0, 0) of the Euclidean space. Each sample is assigned to one of the three groups (“Low”, “Medium”, or “High”) according to the level of within-colony genetic variation. Different letters indicate significant differences P( < 0.05) across the groups.

8 A

6 AB

B 4

Euclidean distance from the origin 0 Low Medium High

Colony genetic variation

88 Fig. S1. Discriminant analysis of 56 chemical samples (each consisting of five nestmate workers) based on seven PCs retained for the analysis (insert), showing discrimination among the 15 Monomorium pharaonis colonies. The proportion of variance explained by each of the two discriminant functions is provided on the axis legends.

Colony 111 12 13 18 23 64 92 4009 4013 4112 5021 T1 T2 T3 T4 Discriminant function 2 (28%)

Discriminant function 1 (40%)

The statistical approach to identify nestmate recognition cues

Jelle S. van Zweden, Luigi Pontieri & Jes Søe Pedersen

(Manuscript in preparation) Chapter 3 91 The statistical approach to identify nestmate recognition cues

Jelle S. van Zweden1,*, Luigi Pontieri2 & Jes Søe Pedersen2

1 Laboratory of Socioecology & Social Evolution, KU Leuven, Naamsestraat 59, B-3000 Leuven, Belgium

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

* Correspondence Jelle S. van Zweden Laboratory of Socioecology & Social Evolution KU Leuven Naamsestraat 59 B-3000 Leuven Belgium E-mail: [email protected]

92 Abstract

The ability of social insects to discriminate nestmates from non-nestmates is mainly achieved through chemical communication. To ultimately understand this recognition and its decision rules, identification of the recognition cues is essential. Although recognition cues are most likely cuticular hydrocarbons, identifying the exact cues for specific species has remained a daunting task, partly due to the sheer number of odour compounds. Perhaps unsurprisingly, one of the few species where the recognition cues have been identified, Formica exsecta, has only around ten hydrocarbons on its cuticle. In this study we use previous results on this species to search for likely nestmate recognition cues in two other species of ants, Camponotus aethiops and Monomorium pharaonis. Employing chemical distances and observed aggression between colonies, we first ask which type of data normalization, centroid, and distance calculation is most diagnostic to discriminate between nestmate recognition cues and other compounds. We find that using a “global centroid” instead of a “colony centroid” significantly improves the analysis. One reason may be that this approach, unlike previous ones, provides a biologically meaningful way to quantify the chemical distances between nestmates, allowing for within-colony variation in recognition cues. Next, we ask which subset of hydrocarbons most likely represents the cues that the ants use for nestmate recognition, which shows less clear results for C. aethiops and M. pharaonis than for F. exsecta, possibly due to the number of compounds in their respective chemical profiles. Nonetheless, some compound sets performed better than others, showing that this approach can be used to identify candidate compounds to be tested in bio-assays, and eventually crack the sophisticated code that governs nestmate recognition.

Keywords: cuticular hydrocarbons, nestmate recognition, statistical analysis.

93 Introduction

Kin recognition is a fundamental ability that allows organisms both to avoid inbreeding and to direct cooperative behaviour towards related individuals (Hepper 1991). In the eusocial insects (termites, ants, some bees and wasps), nestmate recognition – the ability to discriminate nestmates from non-nestmates – is the primary form of kin recognition, since colonies usually consist of closely related family groups. Nestmate recognition is primarily chemical in nature and based on colony-specific cuticular hydrocarbon profiles (Holldobler & Michener 1980; Clément & Bagnères 1998; Singer 1998; van Zweden & d’Ettorre 2010). These cuticular lipids, synthesized by the animals themselves (Howard & Blomquist 2005; van Zweden et al. 2010) and partly obtained from environmental sources (Obin & Vander Meer 1988; Woodrow et al. 2000; Buczkowski et al. 2005), are typically mixed throughout the colony by means of liquid food transfer, grooming, and exchange through nest material (Soroker et al. 1995; d’Ettorre et al. 2006; Couvillon et al. 2007; van Zweden et al. 2010; Bos et al. 2011). Direct evidence for the use of hydrocarbons in nestmate recognition has been obtained in several ant and bee species, by testing the level of aggression towards nestmates supplemented with synthetic hydrocarbons (Lahav et al. 1999; Dani et al. 2005; Ozaki et al. 2005; Martin et al. 2008b; Guerrieri et al. 2009), or to inert materials treated with either the hydrocarbon profile of fellow workers or synthetic mixtures of hydrocarbons (Wagner et al. 2000; Akino et al. 2004; Greene & Gordon 2007; Martin et al. 2008b). Further knowledge of the evolutionary, ecological, and physiological features of nestmate recognition relies on such direct tests to identify the exact recognition cues. However, these manipulative experiments are somewhat hampered by the fact that social insect species often have many different compounds on their cuticle (Blomquist & Bagnères 2010). It is both time-consuming and difficult to synthesise every compound, and if undertaken, it would still be near impossible totestall combinations and interactions of compounds. Hence, a statistical approach is needed to highlight candidate compounds.

Since nestmate recognition cues are expected to be uniform within colonies and variable between colonies, a relatively simple statistical analysis may give us a first clue as to which compounds follow this pattern. Martin et al. (2008a; 2008b) used within-colony correlation and species-level variation of the relative amounts of cuticular hydrocarbons to show that in Formica exsecta the group of (Z)-9-alkenes, and in F. fusca the group of dimethyl pentacosanes, were the most likely nestmate recognition cues. In a similar fashion, van Zweden & d’Ettorre (van Zweden & d’Ettorre 2010) suggested ranking compounds according to their ‘diagnostic power’ – the ratio of between- and within-colony standard deviation – to measure how likely they are to be nestmate recognition cues. In the carpenter ant Camponotus aethiops, this method highlighted predominantly 5-(di)methyl alkanes as likely candidates (van Zweden et al. 2009). These approaches, however, cannot stand alone as they are not taking the actual recognition behaviour into account. The pattern that we expect nestmate recognition cues to follow is that the more two colonies differ in these cues, the easier it will be for the insects to recognize that they belong to different colonies, and the 94 higher the level of, or more frequent, aggression should be. This may not be a linear relationship, rather a step or sigmoid function (van Zweden & d’Ettorre 2010), but the positive relationship between chemical distance and aggression is expected nonetheless. Indeed, this was found in F. exsecta for the (Z)-9-alkenes, but not for linear alkanes, again giving credit to the former as nestmate recognition cues (Martin et al. 2012).

The discovery of the exact cues used for nestmate discrimination in F. exsecta, knowing their pattern of within- and between-colony variation, and their relationship with the behavioural response of the individuals, will help researchers to search for similar compounds in other social insects. Furthermore, we can use this information to compare the efficiency of commonly used statistical procedures to highlight candidate compounds, and to develop new statistical methods with improved power and accuracy. The aim of this study was therefore to use chemical and behavioural data from F. exsecta to: (1) identify specific sets of compounds that might constitute the nestmate recognition signal in two other ant species, C. aethiops and Monomorium pharaonis; (2) investigate which of three methods of compound selection (functional group, diagnostic power and variable clustering) lead to better identification of relevant sets of compounds; (3) evaluate and compare the power of different combinations of data transformation and chemical distance calculation in differentiating between true nestmate recognition cues and other compounds and, as result of this procedure, (4) suggest the use of a new method for centroid calculation (the “global centroid” method) as alternative to the commonly used “group centroid”.

Materials and methods

Datasets analysed

We compiled several data sets of cuticular hydrocarbons (CHCs) and corresponding measurements of aggression between colonies of ants: Formica exsecta (Martin et al. 2012), Camponotus aethiops (van Zweden et al. 2009), and Monomorium pharaonis (Pontieri & Pedersen In preparation).

In addition, we created a simulated “ideal” dataset based on the F. exsecta dataset. This simulated dataset served a double purpose. First, it was used to illustrate the properties of CHC data and how different methods of data transformation can affect the degree of interdependence that compositional data naturally possess (Figure 1). Second, as we know which compounds constitute the nestmate recognition (NMR) signal in F. exsecta and how they vary both within and between colonies, we used the simulated dataset to test the efficiency of different statistical procedures to correctly discriminate NMR cues from other compounds.

Formica exsecta – The narrow-headed ant or excised wood ant, F. exsecta, is a species for which quite a lot is known about its nestmate recognition system (Martin et al. 2008b; Martin & Drijfhout 2009; Martin et al. 2012). A clear difference in aggression between nestmates and non-nestmates has been observed, even

95 when CHC profiles are quite similar (Martin et al. 2012). The CHC profile is relatively simple and consists of 10-12 (Z)-9-alkenes alkenes and linear alkanes (pairwise corresponding in chain-length), of which the composition of alkenes appears to encode the nestmate recognition signal (Martin et al. 2008). The CHC dataset (courtesy of Stephen J. Martin) consisted of 33 colonies from the same population, each containing information on ten CHCs ((Z)-9-C23:1 to n-C31) and five individual workers (Martin et al. 2012; Table S1). The aggression dataset contained 24 non-nestmate (nNM) colony combinations and five nestmate (NM) colony combinations. For each of the combinations, ten ant workers were placed on the mound of the opponent colony (and vice versa for ten of the 24 nNM combinations) and the first five interactions were recorded and classified as aggressive or non-aggressive, leading to 50–100 interactions per combination (Table S1).

Camponotus aethiops – An Italian population of this carpenter ant has been subject to several studies related to nestmate recognition (van Zweden et al. 2009; Bos et al. 2010; Stroeymeyt et al. 2010; Bos et al. 2011). The data used is from a study where colonies were kept in the laboratory for one year (van Zweden et al. 2009). Colony CHC profiles were quantified at regular intervals and their inter-colony aggression was tested by one-on-one aggression tests in a neutral arena at two time points during this year. The CHC profile consists of linear and (di)methylated alkanes (van Zweden et al. 2009). The CHC dataset consisted of six colonies, each containing information on 36 identified CHC peaks and ten individuals (five individuals at two time points; Table S1). The aggression dataset contained eight nNM and three NM colony combinations, each replicated 12 times (Table S1). For each replicate, ants were staged one-on-one in a neutral arena and when biting or abdomen flexing occurred, the interaction was classified as aggressive.

Monomorium pharaonis – The invasive pharaoh ant is known for its polygyny and low levels of nestmate recognition, albeit not being considered supercolonial (Schmidt et al. 2010). CHC profiles were found to be colony-specific, but there was no clear relationship between hydrocarbon distance and aggression, and in only one of two data sets higher aggression between non-nestmates (nNMs) than between nestmates (NMs) was found (Schmidt et al. 2010). The CHC dataset used here (Pontieri & Pedersen In preparation) consisted of 16 colonies, each containing information on 45 CHC peaks (linear and (di)methylated alkanes and alkenes) and 1–4 individual samples (each sample was an extract of five workers joined in a vial). Two of the 45 peaks were discarded as they could not be identified, thus resulting in 43 compound variables (Table S1). The aggression dataset used is of an experiment in which pairs of colonies were given the opportunity to fuse or not (Pontieri & Pedersen In preparation). Colonies were maintained in individual plastic boxes connected by a vinyl tube to a common foraging arena and allowed to interact for ten days. The number of fighting pairs of ants was recorded 12 times over the course of the assay and, when any fighting pair was found, the trial was classified as aggressive. Twelve nNM colony combinations were replicated six times and 16 NM combinations were replicated three times (Table S1).

Simulated data – A total of five colony profiles (Colony A to E) of 15 CHCs was constructed. Each colony

96 contained ten individuals. The 15 CHCs were divided into three groups of five, to mimic the simple structure observed for the CHC profile of F. exsecta. In this species the profile consists of five (Z)-9-alkenes (the supposed recognition cues), five alkanes and some additional compounds present in low concentration that are often overlooked but could carry important information nonetheless (Martin et al. 2008b). Each CHC group in our simulated dataset had a constant distribution within colonies. These distributions were specified by an arbitrary set of five numbers, each of which was used as the mean around which random numbers with a given standard deviation were created. The first set of five CHCs comprises the nestmate recognition signal, with large between-colony variation and little within-colony variation. Each colony had a distribution of means between 1 and 20, with a constant sum of 40 and a standard deviation of 2% of the mean. The variation among colony mean values was such that there was a gradual change from the first to the fifth colony, i.e. colony A and E differed the most, colony A and B differed approximately the same as D and E, and colony C had the middle colony odour (Figure 1). The second set of five CHCs comprises the task recognition signal (Greene & Gordon 2003; Martin & Drijfhout 2009), with large variation in the sum between task groups (foragers and nurses), and also larger variation around the colony means compared to the nestmate recognition signal. For foragers, the five CHCs added up to 50, whereas the sum was only 30 for nurses. Hence, foragers had a distribution of means between 2 and 20, whereas nurses had distribution of means between (2 / 50) × 30 = 1.2 and (20 / 50) × 30 = 12. The standard deviation was set to 20% of the mean for this group of CHCs. The third set of five CHCs comprises random noise of low-concentration, with the same distribution of means for all colonies. The means ranged from 0.5 to 2, with a sum of 5. The standard deviation was set to 50% of the mean. Finally, all these data (five colonies × ten individuals × 15 CHCs) were then per individual multiplied by a random number (mean = 3, standard deviation = 1, range = 1.07–5.91) to mimic concentration differences (Table S1). In addition, an artificial data file on aggression behaviour was created with ten encounters between each of the colony combinations (N = 10 combinations) and ten encounters between nestmate combinations (N = 5 combinations). Aggression between nestmates was always zero, whereas that between non-nestmate colony combinations followed a similar gradual change as the colony odour variation, i.e. there was maximum aggression in colony A vs. E, whereas in combinations A vs. B, B vs. C, C vs. D, and D vs. E only seven out of ten encounters were aggressive (Table S1).

Selection of candidate compounds

Since not all compounds function as nestmate recognition cues, it is important to select compound sets that together might make up the nestmate recognition cues. However, there are many different combinations of compound possible (e.g. with the 45 compounds of the M. pharaonis dataset there are 35,184,172,088,831 different possible combinations), so we have to rely on different strategies than simply going through each of these. There is, however, a range of methods of variable selection that might result in coherent groups which follow correlated patterns in chemical profiles. Here we test three methods that are appropriate to our 97 type of data. Two of these methods have been proposed in previous studies: functional groups (Dani et al. 2005; Martin & Drijfhout 2009) and diagnostic power (van Zweden et al. 2009). In addition, we tested the efficiency of a third approach that we term “variable clustering” where compounds are grouped according to their degree of correlation using the R package ClustOfVar (Chavent et al. 2012).

The first strategy is to subdivide the CHC variables into functional groups: linear alkanes, alkenes, 3-methyl alkanes, etc. This is based on the finding of previous studies that different structural class of hydrocarbons can have a differential importance in nestmate recognition. In the paper wasp Polistes dominulus the topical application of linear alkanes on the cuticle of anesthetized workers did not elicit an aggressive response by nestmates once the individual was reintroduced in the nest. On the other hand, the application of alkenes or methyl branched alkanes did have this effect (Dani et al. 2001). Alkenes were also found to play a major role in nestmate recognition compared to linear alkanes in the honeybees and in F. exsecta (Dani et al. 2005; Martin et al. 2012). Compounds in each of these sets often share a large part of their biosynthetic pathway (Blomquist 2010), and thus should both correlate with and experience the same genetic and physiological constraints. For example, the different methylated alkanes originate simply due to the incorporation of a propionyl-CoA instead of a malonyl-CoA during chain elongation (Morgan 2004; Blomquist 2010). On the other hand, the biosynthesis of unsaturated hydrocarbons, alkenes, likely involves an extra Δ9 desaturase step (Badouin et al. 2013), which may therefore result in a higher degree of decoupling of the relative expression of saturated and unsaturated hydrocarbons. For example, in F. exsecta the (Z)-9-alkenes and linear alkanes seem to vary relatively independently of each other (Martin & Drijfhout 2009). This subdivision into functional groups resulted in 3-6 compound sets for the different datasets (see Table S2).

The second strategy is to first determine which compounds vary most between colonies and least within colonies, as would be expected for compounds that function as nestmate recognition cues (van Zweden & d’Ettorre 2010). The ratio of between-colony standard deviation over pooled within-colony standard deviation, also termed diagnostic power (DP), can be calculated for each compound separately (after normalization to percentages or using an Aitchison-transformation, see below), after which the compounds can be ranked according to this number. For each dataset and normalization, we made four compound sets: higher than average diagnostic power (High DP), lower than average diagnostic power (Low DP), the five compounds with the highest diagnostic power (Highest 5 DP), and the five compounds with the lowest diagnostic power (Lowest 5 DP). See Table S2 for the DP value of each compound.

A third strategy is to assess the correlation between compounds (e.g. Martin et al. 2008a), so that we can group co-varying compounds directly into homogeneous clusters. Strongly co-varying variables should bring approximately the same information, so reducing the number of variables could simplify further data analysis. Principle Component Analysis (PCA) could be used to indicate such co-varying variables as well as reduce the number of variable dimensions. However, it might be difficult to find objective ways to group the variables based purely on PCA as this does not provide any statistical basis to determine a cut-off

98 point. Instead, here we use a cluster method based on squared correlations as developed in the R package ClustOfVar (Chavent et al. 2012). The advantage provided by this function relative to a PCA is that once compounds are clustered in homogeneous groups, it is possible to verify the stability of these groups by a bootstrap approach. We used the stability function to determine the best aggregation level, but set this to a maximum of five clusters for simplicity. This analysis was performed using the raw data, i.e. without prior normalization. See Table S2 for the exact compounds in each set.

Choice of calculation method

Chemical distances between colonies can be calculated in at least 12 different ways (see below). In Martin et al. (2012), the authors based their analysis on a single approach: first normalizing the raw CHC data to percentages, then calculating an average colony odour (group centroid), and then calculating the absolute distance (Manhattan distance) between these average odours. However, there are many different possible approaches for the calculation of chemical distances.

Firstly, raw CHC data can be normalized using the Aitchison-transformation: yij = log(xij/ g(Xj)), where yij is th th the transformed peak area of the i CHC component of the j individual, xij is the untransformed peak area of th th th the i component of the j individual, and g(Xj) is the geometric mean peak area of all components of the j individual (Aitchison 1986). This method of normalization, also referred to as the log-ratio transformation, may be superior to percentage-based normalization because of the reduction of the so-called ‘closure effect’. This effect can hinder the analysis of compositional data in the sense that the variation of one variable can greatly influence the variation of another variable, because of their interdependence. This effect, and the influence of normalization, are illustrated in Figure 1 using the simulated dataset.

Secondly, using a group centroid, the distance between nestmates is necessarily always zero, leading to a slight imbalance in the distribution of distances when both nestmate and non-nestmate distances are calculated. That is, even when colonies have very high variability in their CHC profiles and chemical distances between nestmates are actually very high, the calculated distances are still zero. As one way to overcome this problem we suggest using what we term a ‘global centroid’ (Figure 2). In this case, the average odour of all individuals in the colony combination is calculated, after which the average distance to this average odour is taken as the chemical distance. When two colonies are further apart in chemical space, this distance will still be higher than two colonies that are close. In addition, the distance between nestmates actually depends on their differences in individual odour profiles.

Thirdly, the actual distance calculation can be done in at least three different ways, of which the Manhattan distance is one. Another commonly used distance calculation is the usual square distance or Euclidean distance, the square root of the sum of squared differences (see also Figure 2). A third possibility is to first reduce chemical space of the data set using principal component analysis (PCA) and then take the absolute distance on the first axis (PC1). There is probably a whole array of other possibilities, but in this study we

99 limited ourselves to the three methods that are the most commonly used.

Overall, this gives 12 different combinations to calculate the chemical distance between colonies, i.e. normalization (percentages or Aitchison-transformed), centroid (group or global), and distance calculation (Manhattan distance, Euclidean distance, or PC1) give 2 × 2 × 3 = 12 different ways to calculate chemical distance. For every dataset, and for every selected set of compounds (‘compound set’), these distances were calculated using R 3.0.2 (R Core Team 2013).

Chemical distance as a predictor of observed aggression

Even though aggression was measured in different ways in the different study species (van Zweden et al. 2009; Martin et al. 2012; Pontieri & Pedersen In preparation), we could translate this in all three cases into the proportion of behaviours or encounters that were aggressive. The chemical distances based on the different compound sets (all compounds, functional groups, high/low diagnostic power, clusters of variables; see Data Selection section) and the 12 different combinations of calculation methods (see Combination section) were regressed against the proportion of aggressive behaviours/encounters and the artificial aggression data for the Simulated dataset, by fitting a general linear model (GLM) with quasi-binomial errors and a logit link function, using R 3.0.2 (R Core Team 2013). For each model run the deviance explained by chemical distance was extracted for further analysis.

To test which of the 12 combinations of calculation methods most accurately identifies nestmate recognition cues over other cues, we used the fact that we know which compounds these are in the F. exsecta and simulated datasets. We took the ratio of explained deviance from the regressions with the nestmate recognition cues (respectively CHC 1−5 and Z-9-alkenes) over the explained deviance from the regressions with task cues (respectively CHC 6−10 and n-alkanes) in both datasets and used these as the dependent variable in a GLM with Gaussian errors and ‘normalization’, ‘centroid’ and ‘distance calculation’ as the explanatory variables.

To test which compound set most likely includes the nestmate recognition cues in C. aethiops and M. pharaonis, respectively, the explained deviance of the regressions across the 12 combinations for each of the compound sets was used as the dependent variable in a GLM with Gaussian errors and ‘compound set’, ‘normalization’, ‘centroid’ and ‘distance calculation’ as the explanatory variables.

Results

The most diagnostic combination of calculation methods

Which combination of calculation methods (i.e. combination of normalization, centroid, and distance calculation) most accurately differentiates between nestmate recognition cues and other compounds?

100 In the F. exsecta dataset, the most diagnostic combination is using normalization to percentages, a global centroid and PC1 distance, as this gave the highest ratio between the explained deviance of nestmate recognition cues over the explained deviance of task cues (Z-9-alkenes/linear alkanes = 753.08/127.45 = 5.91). Overall, the global centroid and PC1 distance both gave significantly higher deviance ratios compared to group centroid and Manhattan distance (Table 1). Normalization using the Aitchison- transformation or using the Euclidean distance measure did not have such an effect (Table 1). Similarly, in the simulated dataset, centroid and distance had significant effects on the deviance ratios (Table 1). The most diagnostic combination was obtained through the combined usage of Aitchison-transformation, global centroid and PC1 distance (NMR cues/Task cues = 77.59/23.97 = 3.23), although using the same combination with percentage normalization instead gave very similar results (76.32/23.93 = 3.19). When NM control encounters were removed (i.e. similar to Martin et al. 2012) before fitting the regression models, the effect of distance calculation disappeared in both datasets (Table 1). The significant effect of centroid did, however, disappear in the simulated dataset but not in the F. exsecta dataset (Table 1). In this case, the most diagnostic combination for the F. exsecta dataset was Aitchison-transformation, global centroid, and Manhattan distance (436.66/58.60 = 7.45), and for the simulated dataset it was Aitchison-transformation, group centroid, and PC1 distance (4.85/0.13 = 38.05).

The compound set that best explains aggression

Which of the compound sets best explains the observed level of aggression in the different species, and is thus likely to be the set of nestmate recognition cues?

Based on all combinations of calculation methods, the explained deviance of the regressions using the All compound set (i.e. all compounds in the CHC profile) did not give a very good result in theF. exsecta dataset (Figure 4a). The Alkene compound set explained between-colony aggression significantly better than when using All or Linear (i.e. linear alkanes). Ranking by diagnostic power resulted in approximately the same set of compounds (see Table S2) and thus a very similar pattern in explained deviances. The cluster analysis, on the other hand, resulted in the alkenes being split into two separate groups (Table S2; Cluster 1: C23:1 + C25:1

+ C31:1; Cluster 2: linear alkanes; Cluster 3: C27:1, C29:1), both of which did not explain aggression as well as the five alkenes together (Figure 4a). This pattern hardly changed when looking only at the best combination

(stars in Figure 4), although with percentage normalization, Cluster 1 (C23:1, C25:1, and C31:1; Table S2) ranked almost as high as the Alkene compound set. The highest level of explained deviance in this case was found when using only above-average diagnostic power compounds (High DP; C23:1, C27:1, C29:1, and C31:1). Hence, the (Z)-9-alkenes can indeed be considered the most likely nestmate recognition cues, confirming earlier findings that these compounds modulate inter-colony aggression (Martin et al. 2008b; Martin et al. 2012).

In the Simulated dataset we found a very similar pattern as in the F. exsecta dataset, potentially because it was in part based on a similar number of compounds and the chemical/behavioural patterns observed in this

101 ant species. Using either All, NMR cues, High DP, High 5 DP, or Cluster 1 (each of which contained all or most of CHC 1−5; see Table S2) gave significantly higher explained deviance than the other compound sets (Figure 4b). If, however, we look only at the best combination (global centroid, PC1 distance), we find that the explained deviance of the All compound set drops somewhat, whereas using compound sets NMR cues, High DP, High 5 DP, or Cluster 1 gives similarly high levels of explained deviance. It is interesting to note here that the combinations using Manhattan or Euclidean distance and group centroid resulted in explained deviances of the Random compound set that were almost equally as high as those of the NMR cues.

In the C. aethiops dataset, using either the compound set 5-methyl, High DP, Cluster 2 or Cluster 4, resulted in the best explained deviances (Figure 4c). High DP and Cluster 4 also had an overrepresentation of 5-methyl compounds, where the High DP compound set also included 13+11+9-methyl and 3-methyl alkanes, and Cluster 4 also included 7-methyl and 3-methyl alkanes. Cluster 2 predominantly included linear alkanes and a 5-methyl alkane (see Table S2). Based on these results, the 5-methyl alkanes seem the most likely compound set to start testing in bio-assays. This is, though, still quite a limited dataset, given that we only have 11 colony combinations with 12 trials each to identify nestmate recognition cues amongst 36 compounds. In comparison, the F. exsecta dataset has 29 colony combinations with 50–100 interactions each to identify nestmate recognition cues amongst ten compounds, which likely gives a much better resolution.

In the M. pharaonis dataset, nothing explained between-colony aggression better than all compounds together (Figure 4d), but when only taking into account the most diagnostic combination, the 13+11+9-methyl alkanes appear to be the best candidate compound set. Here too, we have a more limited dataset than for F. exsecta, having 28 colony combinations, each with 3–6 trials, to differentiate amongst 45 compounds.

Discussion

Identification of likely NMR compounds in C. aethiops and M. pharaonis

Altogether, we have shown that the currently used statistical approach is capable of identifying likely nestmate recognition cues, since the known nestmate recognition cues in the F. exsecta and Simulated datasets are indeed correctly identified. On the other hand, the picture is much less clear in the two other datasets, both of which also contain many more compounds (10 and 15 versus 36 and 45). Moreover, the majority of the compounds found in C. aethiops and M. pharaonis share biosynthetic pathways, leading to high levels of multicollinearity. Hence, it is difficult to see whether the lack of a very clear pattern in the latter two datasets is statistical or behavioural in nature. It is unlikely to be due to the sheer number of compound variables, as that should be counteracted by the data selection (functional groups, ranking by diagnostic power, or variable clustering), which resulted in equally small compound sets as in the F. exsecta 102 and Simulated datasets. A more likely explanation is the shared-biosynthetic-pathway and multicollinearity issue, as this results in different compound sets having similar chemical distances between a given combination of colonies, and hence similar explained deviances. Unfortunately, we currently have no means to test if 5-methyl hydrocarbons are indeed the true nestmate recognition cues of C. aethiops, which could confirm or disprove that this statistical method filters out the proper set of compounds. Another possibility, one which we deem more likely in the case of M. pharaonis, is that any difference in odour profile leads to aggression, giving rise to a pattern where no single compound set explains aggression better than all compounds taken together. In the case of M. pharaonis, there is the additional issue that overt aggression is not readily expressed, which may result in a higher error rate in transcribing this behaviour to data that correctly describes their true hostility. The fact that the aggression data of M. pharaonis here is based on the number of trials where fighting was observed clearly does not offset this problem, since a trial with 20 couples fighting was classified as equal to a trial where only a single couple was fighting. We would have opted for this more fine-grained data, if we would also have had an estimate of the number of couples not fighting, but that was not the case.

Compound selection methods

The use of diagnostic power may prove to be quite fruitful in this type of analysis, despite the fact that this data selection method was outperformed by subdivision into functional groups in the C. aethiops and M. pharaonis datasets. The High DP compound sets contained approximately the nestmate recognition cues in the F. exsecta and Simulated datasets, and thus gave high explained deviances, so this simple ranking can be helpful when the nestmate recognition cues are not known. It is only in the M. pharaonis dataset that High DP and High 5 DP compound sets did not result in higher explained deviances than Low DP and Low 5 DP, but that might also be due to the peculiarity in nestmate recognition of this species.

Variable clustering was less successful, as can be seen in the results for F. exsecta. Here, the breakdown of the five alkene compounds into two separate clusters resulted in much lower explained deviances for both clusters. In the Simulated dataset variable clustering did identify the NMR cues as a single cluster, but in both the C. aethiops and the M. pharaonis datasets this variable selection method was outperformed by the subdivision into functional groups. In the C. aethiops dataset, Cluster 2 and Cluster 4 gave high explained deviances when the regressions were run with nestmate controls, but this effect disappeared when these controls were omitted – showing that these clusters cannot account for the differences in aggression between colonies as well as 5-methyl alkanes can.

Best combination of calculation methods to discriminate NMR from other compounds

The use of a global centroid appears to be a superior alternative to the group centroid when calculating chemical distances to detect nestmate recognition cues (Table 1). The main reason for this is not likely to

103 be that using this centroid results in an increased fit for nestmate recognition cues, but rather that it results in a much worse fit for other compounds. Using the group centroid, within-colony chemical distances are necessarily zero, whereas the global centroid allows within-colony variation in relative compound abundances to be taken into account in the regressions. And this difference in centroid usage has less of an effect on nestmate recognition cues than on other compounds because of their uniformity. The other compounds may be subject to more random variation or variation related to other functions than nestmate recognition, such as task or reproductive state. In nestmate controls, such variation translates into chemical distances above zero, while there is still low or no aggression observed, leading to a worse fit of the regression than for nestmate recognition cues (Figure 3). Because of their uniformity, these will experience lower chemical distances within colonies, which matches better with the low aggression between nestmates. However, the effect appears to be even stronger than that, because the global centroid still comes out as significant in non-nestmate encounters of theF. exsecta dataset (Table 1). This is because of approximately the same reason: The low variance of nestmate recognition cues within colonies means that using a group or a global centroid approach does not greatly alter between-colony chemical distances, so that the explained deviances resulting from the regressions are very similar. The difference between these types of centroids is more pronounced with the higher within-colony variances of other compounds. Using the global centroid, these within-colony variances impact between-colony chemical distances in a more unpredictable way and apparently decrease the explained deviance even further as compared to using the group centroid approach.

The usage of PC1 distance, rather than Manhattan or Euclidean distance, also significantly increased the deviance ratios in both datasets (Table 1). Since this ratio is mostly increased because of a decrease in the denominator, it is perhaps best to first look at what happens when linear alkanes or task cues are entered as the compound set in the regression. PCA first searches for the axis of maximum variance in the data, which then becomes the first principal axis or PC1, and variation perpendicular to this axis is filtered out only when PC1 is used in further analysis (in our case this means the distances on PC1 are taken as the chemical distances between individuals and colonies). Now, when a compound set such as linear alkanes or task cues is used, there will be some colony-specificity but the major variation in the data is related to task. Therefore, PC1 will encompass this task-related variation and filters out most of the colony-specific variation. For nestmate recognition cues, this should obviously not be the case. Comparatively, when using the Manhattan or Euclidean distance, all variation in the relative abundances of the compounds is taken into account and no such filtering is done.

Conclusions

Ultimately, nestmate recognition cues should be identified by means of bio-assays that manipulate the odour profiles and test the effect of this manipulation on the aggression of the insects, but the extension of 104 statistical methods provided in this study may be useful in identifying likely candidate compounds. The best strategy is potentially to first perform all forms of variable selection described above (functional groups, diagnostic power and variable clustering) with nestmate controls, after which the different compound sets are tested in regressions using solely non-nestmate encounters. Only when a compound set can explain the different levels of aggression observed between different combinations of colonies can it be classified as a good candidate set. The same analysis can also be done with more power: instead of using colony averages, the chemical profile of the interacting individuals in single aggression tests can be analysed and their distances regressed against the observed aggression. We also encourage the use of our new method of centroid calculation (i.e. the “global centroid” method) as alternative to the group centroid. First, the global centroid allows a better discrimination of NMR relative to other compounds, as was shown by our analysis of the F.exsecta and the simulated datasets. Second, it gives the possibility of calculating a chemical distance between nestmates.

Overall, we think that the methods reported here can lead to the identification of nestmate recognition cues in more species, so that the study of the evolutionary and physiological underpinnings of these cues can be further elucidated. We would also like to call out to anyone with a similar dataset of aggression and potential recognition cues to come forward and contact one of the authors. Perhaps we can together make sense of the nestmates recognition cues of your study species.

Acknowledgements

We thank Tabitha Mary Innocent for comments on a previous version of the manuscript. LP and JSP were funded by a grant from the Danish National Research Foundation (grant number DNRF57).

References

Aitchison, J. 1986. The Statistical Analysis of Compositional Data. London: Chapman & Hall. Akino, T., Yamamura, K., Wakamura, S. & Yamaoka, R. 2004. Direct behavioral evidence for hydrocarbons as nestmate recognition cues in Formica japonica (Hymenoptera: Formicidae). Applied Entomology and Zoology, 39, 381-387. Badouin, H., Belkhir, K., Gregson, E., Galindo, J., Sundström, L., Martin, S. J., Butlin, R. K. & Smadja, C. M. 2013. Transcriptome characterisation of the ant Formica exsecta with new insights into the evolution of desaturase genes in social Hymenoptera. PLoS ONE, 8, e68200. Blomquist, G. J. 2010. Biosynthesis of cuticular hydrocarbons. In Insect Hydrocarbons: Biology, Biochemistry, and Chemical Ecology (ed. G. J. Blomquist & A. G. Bagnères), pp. 35-52.

105 Cambridge: Cambridge University Press. Blomquist, G. J. & Bagnères, A. G. 2010. Insect Hydrocarbons: Biology, Biochemistry, and Chemical Ecology. Cambridge: Cambridge University Press. Bos, N., Grinsted, L. & Holman, L. 2011. Wax on, wax off: nest soil facilitates indirect transfer of recognition cues between ant nestmates. PLoS ONE, 6, e19435. Bos, N., Guerrieri, F. J. & d’Ettorre, P. 2010. Significance of chemical recognition cues is context dependent in ants. Animal Behaviour, 80, 839-844. Buczkowski, G., Kumar, R., Suib, S. L. & Silverman, J. 2005. Diet-related modification of cuticular hydrocarbon profiles of the Argentine ant,Linepithema humile, diminishes intercolony aggression. Journal of Chemical Ecology, 31, 829-843. Chavent, M., Simonet, V. K., Liquet, B. & Saracco, J. 2012. ClustOfVar: An R Package for the Clustering of Variables. Journal of Statistical Software, 50, 1-16. Clément, J.-L. & Bagnères, A.-G. 1998. Nestmate recognition in termites. In Pheromone Communication in Social Insects: Ants, Wasps, Bees and Termites. (ed. R. K. Vander Meer, M. D. Breed, K. E. Espelie & M. L. Winston), pp. 126-155. Boulder: Westview Press. Couvillon, M. J., Caple, J. P., Endsor, S. L., Kaercher, M. H., Russell, T. E., Storey, D. E. & Ratnieks, F. L. W. 2007. Nest-mate recognition template of guard honeybees (Apis mellifera) is modified by wax comb transfer. Biology Letters, 3, 228-230. d’Ettorre, P., Wenseleers, T., Dawson, J., Hutchinson, S., Boswell, T. & Ratnieks, F. L. W. 2006. Wax combs mediate nestmate recognition by guard honeybees. Animal Behaviour, 71, 773-779. Dani, F. R., Jones, G. R., Corsi, S., Beard, R., Pradella, D. & Turillazzi, S. 2005. Nestmate recognition cues in the honey bee: differential importance of cuticular alkanes and alkenes. Chemical Senses, 30, 477-489. Dani, F. R., Jones, G. R., Destri, S., Spencer, S. H. & Turillazzi, S. 2001. Deciphering the recognition signature within the cuticular chemical profile of paper wasps.Animal Behaviour, 62, 165-171. Greene, M. J. & Gordon, D. M. 2003. Cuticular hydrocarbons inform task decisions. Nature, 423, 32-32. Greene, M. J. & Gordon, D. M. 2007. Structural complexity of chemical recognition cues affects the perception of group membership in the ants Linephithema humile and Aphaenogaster cockerelli. Journal of Experimental Biology, 210, 897-905. Guerrieri, F. J., Nehring, V., Jørgensen, C. G., Nielsen, J., Galizia, C. G. & d’Ettorre, P. 2009. Ants recognize foes and not friends. Proceedings of the Royal Society B: Biological Sciences, 276, 2461-2468. Hepper, P. G. 1991. Kin Recognition. Cambridge: Cambridge University Press. Holldobler, B. & Michener, C. D. 1980. Mechanisms of identification and discrimination in social Hymenoptera. Life Sciences Research Report, 35-58.

106 Howard, R. W. & Blomquist, G. J. 2005. Ecological, behavioral, and biochemical aspects of insect hydrocarbons. Annual Review of Entomology, 50, 371-393. Lahav, S., Soroker, V., Hefetz, A. & Vander Meer, R. K. 1999. Direct behavioral evidence for hydrocarbons as ant recognition discriminators. Naturwissenschaften, 86, 246-249. Martin, S. & Drijfhout, F. 2009. Nestmate and task cues are influenced and encoded differently within ant cuticular hydrocarbon profiles.Journal of Chemical Ecology, 35, 368-374. Martin, S. J., Helanterä, H. & Drijfhout, F. P. 2008a. Colony-specific hydrocarbons identify nest mates in two species of Formica ant. Journal of Chemical Ecology, 34, 1072-1080. Martin, S. J., Vitikainen, E., Drijfhout, F. P. & Jackson, D. 2012. Conspecific ant aggression is correlated with chemical distance, but not with genetic or spatial distance. Behavior Genetics, 42, 323-331. Martin, S. J., Vitikainen, E., Helanterä, H. & Drijfhout, F. P. 2008b. Chemical basis of nest-mate discrimination in the ant Formica exsecta. Proceedings of the Royal Society B: Biological Sciences, 275, 1271-1278. Morgan, E. D. 2004. Biosynthesis in Insects. Cambridge: The Royal Society of Chemistry. Obin, M. S. & Vander Meer, R. K. 1988. Sources of nestmate recognition cues in the imported fire ant Solenopsis invicta Buren (Hymenoptera: Formicidae). Animal Behaviour, 36, 1361-1370. Ozaki, M., Wada-Katsumata, A., Fujikawa, K., Iwasaki, M., Yokohari, F., Satoji, Y., Nisimura, T. & Yamaoka, R. 2005. Ant nestmate and non-nestmate discrimination by a chemosensory sensillum. Science, 309, 311-314. Pontieri, L. & Pedersen, J. S. In preparation. Unexpected fusion of genetically and chemically divergent colonies of the invasive pharaoh ant. R Core Team. 2013. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Schmidt, A. M., d’Ettorre, P. & Pedersen, J. S. 2010. Low levels of nestmate discrimination despite high genetic differentiation in the invasive pharaoh ant. Frontiers in Zoology, 7, 20. Singer, T. L. 1998. Roles of hydrocarbons in the recognition systems of insects. American Zoologist, 38, 394-405. Soroker, V., Vienne, C. & Hefetz, A. 1995. Hydrocarbon dynamics within and between nestmates in Cataglyphis niger (Hymenoptera: Formicidae). Journal of Chemical Ecology, 21, 365-378. Stroeymeyt, N., Guerrieri, F. J., van Zweden, J. S. & d’Ettorre, P. 2010. Rapid decision-making with side-specific perceptual discrimination in ants.PLoS ONE, 5, e12377. van Zweden, J. S., Brask, J. B., Christensen, J. H., Boomsma, J. J., Linksvayer, T. A. & d’Ettorre, P. 2010. Blending of heritable recognition cues among ant nestmates creates distinct colony gestalt odours but prevents within‐colony nepotism. Journal of Evolutionary Biology, 23, 1498-1508. van Zweden, J. S. & d’Ettorre, P. 2010. Nestmate recognition in social insects and the role of hydrocarbons.

107 In Insect Hydrocarbons: Biology, Biochemistry and Chemical Ecology (ed. G. J. Blomquist & A. G. Bagnères), pp. 222-243. Cambridge: Cambridge University Press. van Zweden, J. S., Dreier, S. & d’Ettorre, P. 2009. Disentangling environmental and heritable nestmate recognition cues in a carpenter ant. Journal of Insect Physiology, 55, 159-164. Wagner, D., Tissot, M., Cuevas, W. & Gordon, D. M. 2000. Harvester ants utilize cuticular hydrocarbons in nestmate recognition. Journal of Chemical Ecology, 26, 2245-2257. Woodrow, R. J., Grace, J. K., Nelson, L. J. & Haverty, M. I. 2000. Modification of cuticular hydrocarbons of Cryptotermes brevis (Isoptera: Kalotermitidae) in response to temperature and relative humidity. Environmental Entomology, 29, 1100-1107.

108 Tables

Table 1. The effect of combination of calculation methods on the ratio of explained deviance of known nestmate recognition cues (Z-9-alkenes and CHC 1-5) and explained deviance of task cues (linear alkanes and CHC 6-10). Normalization (percentage or Aitchison-transformation), centroid (group or global), and distance calculation (Manhattan, Euclidean, or PC1) were set as explanatory variables.

Formica exsecta Simulated data With nestmate controls Estimate Std. Error t value Pr(>|t|) Estimate Std. Error t value Pr(>|t|) (Intercept) 0.4930 0.1614 3.0544 0.0185* 0.5312 0.0276 19.2540 <0.0001*** Normalization: Aitchison -0.1652 0.1444 -1.1445 0.2900 -0.0002 0.0247 -0.0078 0.9940 Centroid: Global 0.7267 0.1444 5.0337 0.0015** 0.5195 0.0247 21.0529 <0.0001*** Distance: Euclidean 0.0106 0.1768 0.0601 0.9538 0.0190 0.0302 0.6285 0.5496 Distance: PC1 0.4569 0.1768 2.5841 0.0363* 0.1329 0.0302 4.3965 0.0032**

Without nestmate controls Estimate Std. Error t value Pr(>|t|) Estimate Std. Error t value Pr(>|t|) (Intercept) 0.9011 0.2417 3.7286 0.0074** 2.7385 0.2152 12.7276 <0.0001*** Normalization: Aitchison -0.1746 0.2162 -0.8079 0.4457 0.0398 0.1924 0.2068 0.8420 Centroid: Global 0.7135 0.2162 3.3009 0.0131* -0.1424 0.1924 -0.7402 0.4833 Distance: Euclidean 0.0064 0.2647 0.0242 0.9814 -0.2879 0.2357 -1.2214 0.2615 Distance: PC1 0.3108 0.2647 1.1738 0.2788 0.4516 0.2357 1.9158 0.0969

109 Figure captions

Figure 1. The effect of percentage-normalization and Aitchison-transformation. The CHC data from the Simulated dataset are used to illustrate the extent to which the two types of transformations used in our analysis affect the interdependence of compositional data. Each row shows one of the three sets of 5 compounds in which the ideal profile was divided (upper row = nestmate recognition cues; middle row = task recognition cues; bottom row = random CHC variables). The average amount (± SD) of each compound in each set is showed for each of the 5 colonies present in the dataset (colony A-E). The left column (Raw data) shows the original pattern followed by CHCs in each set, whereas the middle and right column show the distribution of such compounds after Percentage and Log-ratio (Aitchison) transformation. Even though the nestmate recognition cues (upper row) are very homogeneous amongst nestmates, as is shown by the small standard deviation, they are strongly affected by other hydrocarbon variables when normalization to percentages is done (‘closure effect’). Log-ratio transformed data (Aitchison transformation) suffers less of this effect, as is shown by the relatively small standard deviations in the upper right panel.

Raw data Percentages Log-ratio

25 30 3.0

25 2.0 20 CHC 1 20 1.0 15 CHC 2 15 0.0 CHC 3 10 CHC 4 10 -1.0 CHC 5 Mean value (+/- SD) 5 5 -2.0

Nestmate recognition cues 0 0 -3.0

25 30 3.0

25 2.0 20 CHC 6 20 1.0 15 CHC 7 15 0.0 CHC 8 10 CHC 9 10 -1.0 CHC 10 Mean value (+/- SD) 5 5 -2.0 Task recognition cues Task

0 0 -3.0

25 30 3.0

25 2.0 20 CHC 11 20 1.0 15 CHC 12 15 0.0 CHC 13 10 CHC 14 10 -1.0 CHC 15 Mean value (+/- SD) 5 5 -2.0 Random CHC variables 0 0 -3.0

Colony A Colony B Colony C Colony D Colony E Colony A Colony B Colony C Colony D Colony E Colony A Colony B Colony C Colony D Colony E

110 Figure 2. Group centroid vs. global centroid. (a) Using the ‘group centroid’, a single distance is calculated between the average odours (X) of two colonies (inter-colony). The distance between nestmates (intracolony) is necessarily always zero. (b) Using a ‘global centroid’, the average odour of all individuals in a combination is first calculated, after which the average distance to this average odour is taken as the chemical distance. The distance between nestmates depends on the intra-colony variation in odour profile. For simplicity, the Euclidean distance and only two odour dimensions are depicted.

(a)

Inter-colony Intra-colony

3 4 2 1 5

10 6 8

9 7

C 1 2 3

21 22 23 4

Colony odour 2 24 25 5 26 6 28 27 29 30 8 9 13 7 14 12 11 15 10 B 20 16 18

19 17

Colony odour 1

(b) Inter-colony Intra-colony

3 4 2 1 5

10 6 8

9 7

C 1 2 21 22 3 23 24 25 4 26

Colony odour 2 5 28 6 27 29 30

13 8 14 9 12 7 11 15 B 20 16 10 18

19 17

Colony odour 1

111 Figure 3. Regressions of chemical distance against observed aggression in the Formica exsecta dataset, illustrating the effect of using group vs. global centroid. The top panel depicts chemical distances based on the (Z)-9-alkene compound set, the bottom panel uses linear alkane chemical distance. Statistics are derived from a GLM with quasi-binomial errors and a logit link function. Closed circles = nestmate control trials, open circles = non-nestmate trials. Using the global centroid, the ratio in explained deviance between (Z)-9- alkenes and linear alkanes is higher than when using the group centroid (2.74 versus 1.59).

Group centroid Global centroid s l a

i 1.0 1.0 r t

v 0.8 0.8 s i e r 0.6 0.6 g

(Z)-9-alkenes g a

explained deviance: 807.9 explained deviance: 758.7 f

o 0.4 0.4

p = 0.003 p = 0.005 n o i t

r 0.2 0.2 o p o r 0.0 0.0 P 0 20 40 60 80 100 120 10 20 30 40 50 60 s l a

i 1.0 1.0 r t

v 0.8 0.8 s i e r 0.6 0.6 Linear alkanes g g a

explained deviance: 508.1 explained deviance: 277.2 f

o 0.4 0.4

p = 0.007 p = 0.026 n o i t

r 0.2 0.2 o p o r 0.0 0.0 P 0 10 20 30 40 5 10 15 20 Chemical distance Chemical distance

112 Figure 4. Explained deviance in observed aggression by compound sets over 12 combinations of normalization, centroid, and distance calculation. For each dataset, “+NM” represents regressions with nestmate controls included, whereas “-NM” represents regressions without nestmate controls. Stars denote the most diagnostic permutation (global centroid and PC1 distance; black = percentage normalization, white = log-ratio transformed).

a Formica exsecta + NM b Simulated + NM 100 800 80 600 60 400 40 e e 200 20 a n c a n c i i 0 0 D e v D e v n e d n e d a i - NM a i - NM p l p l x 600 x E E 5 4 400 3

200 2 1 0 0 l l l l k 1 2 R 1 2 3 A A D P D P D P D P D P D P D P D P a s t e r t e r n e a r t e r t e r t e r T e n s 5 5 5 5 N M k g h g h L i l u s u s u s u s u s L o w L o w A g h g h H i H i R a n d o m C l C l C l C l C l L o w L o w H i H i

c Camponotus aethiops + NM d Monomorium pharaonis + NM

40 100 80 30 60 20 40 e e 10 a n c a n c

i i 20

D e v 0 D e v 0 n e d n e d

a i - NM a i - NM p l p l x x 20 E 15 E

15 10 10 5 5

0 0 l l l l l l l l l l l 1 2 3 4 5 l 1 2 3 4 A A D P D P D P D P D P D P D P D P t e r t e r t e r t e r t e r n e a r e t h y e t h y e t h y e t h y t e r t e r t e r t e r n e a r e n s e t h y e t h y e t h y e t h y 5 5 5 5 k g h L i g h L i l m m m m m u s u s m m m u s u s u s u s u s u s u s ------L o w A L o w g h H i g h 9 - 7 5 3 H i 9 7 5 3 C l C l C l C l C l C l C l C l C l L o w L o w H i H i 1 + 1 + 1 1 1 3 + 1 3 +

113 Supplementary materials

Table S1. Complete CHC and aggression datasets used in the study. [Chapter3-TableS1.xls]

Table S2. Division of CHCs for each of the three methods for selecting the chemical compounds (Functional groups, DP and variable cluster). [Chapter3-TableS2.xls]

114 Ant colonies prefer infected over uninfected nest sites

Luigi Pontieri, Svjetlana Vojvodic, Riley Graham, Jes Søe Pedersen & Timothy A. Linksvayer

(Manuscript submitted to PLoS ONE) Chapter 4 115 Ant colonies prefer infected over uninfected nest sites

Running title: Preference for infection

Luigi Pontieri1*, Svjetlana Vojvodic2,§, Riley Graham3, Jes Søe Pedersen1, Timothy A Linksvayer3

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

2Center for Insect Science, Department of Ecology and Evolutionary Biology, University of Arizona, 1041 E. Lowell St., Tucson, AZ 85721, USA

3Department of Biology, University of Pennsylvania, 433 South University Avenue, Philadelphia, PA 19104, USA

§Current address: Department of Biological Sciences, Rowan University, Science Hall, Glassboro, NJ 08028, USA

Fax: +45 35321250

*Corresponding author: [email protected]

116 Abstract

1. During colony relocation, the selection of a new nest involves exploration and assessment of potential sites followed by colony movement on the basis of a collective decision making process. Hygiene and pathogen load of the potential nest sites are factors worker scouts might evaluate, given the high risk of epidemics in group-living animals.

2. Choosing nest sites free of pathogens is hypothesized to be highly efficient in invasive ants as each of their introduced populations is often an open network of nests exchanging individuals (unicolonial) with frequent relocation into new nest sites and low genetic diversity, likely making these species particularly vulnerable to parasites and diseases.

3. We investigated the nest site preference of the invasive pharaoh ant, Monomorium pharaonis, through binary choice tests between three nest types: nests containing dead nestmates overgrown with sporulating mycelium of the entomopathogenic fungus Metarhizium brunneum (infected nests), nests containing nestmates killed by freezing (uninfected nests), and empty nests.

4. In contrast to the expectation pharaoh ant colonies preferentially (84 %) moved into the infected nest when presented with the choice of an infected and an uninfected nest. The ants had an intermediate preference for empty nests.

5. Pharaoh ants display an overall preference for infected nests during colony relocation. While we cannot rule out that the ants are actually manipulated by the pathogen, we propose that this preference might be an adaptive strategy by the host to “immunize” the colony against future exposure to the same pathogenic fungus.

Keywords

Social insects, Pathogen, Nest choice, Social immunity, Monomorium pharaonis, Metarhizium brunneum

117 Introduction

In social insects, the selection of a suitable nest site during colony establishment and relocation is critical for colony success. Resource availability, intra- and inter-specific competition level, as well as abiotic factors characterizing the new site can strongly affect colony fitness (Hölldobler & Wilson, 1990). The selection of a new nest site is particularly critical during colony relocation, a common phenomenon in eusocial insects (McGlynn, 2012), as migrating between nests is risky and energetically costly, and the colony usually has to select the best site among several candidates. An accurate and effective assessment of the properties of the potential nest sites is thus crucial. Studies conducted in honeybees (Seeley, 2010) and Temnothorax ants (Franks et al., 2003) have shown that individual scouts explore and assess the overall quality of potential nest sites on the basis of several factors. Among others, scouts appear able to evaluate the cleanliness of potential sites, as sites with corpses are avoided, presumably due to disease risk (Franks et al., 2005).

The presence of deadly pathogens in the selected nest might represent a particularly high cost for social insect colonies. One of the drawbacks of living in dense groups of highly related individuals is the increased chance of pathogen spread and disease outbreaks compared to non-social species (Schmid-Hempel, 1998). Social insects have evolved a battery of collective anti-parasite defences that complement individual immune resistance to reduce pathogen exposure, transmission and infection rates, also known as “social immunity” (Cremer et al., 2007). Detecting and avoiding pathogens during the nest site selection process may thus represent a first line of defence.

Pathogens may pose an especially severe problem for invasive ant species. Introduced populations often show reduced genetic diversity (Tsutsui et al., 2000; Schmidt et al., 2010), as well as a unicolonial social structure, characterized by free exchange of individuals among networks of densely occupied nests (Helanterä et al., 2009). Both factors are expected to make invasive ants particularly prone to epidemics (Ugelvig & Cremer, 2012). Avoiding pathogens in the first place is thus expected to be an important strategy for these ants.

The pharaoh ant Monomorium pharaonis is a successful “tramp” ant (Passera, 1994) which represents a good candidate to test the ability of invasive species to avoid generalist pathogens during nest emigration. Distributed worldwide (Wetterer, 2010), this unicolonial species usually thrives in human disturbed environments and nests are often established in ephemeral cavities or household items. Nests are repeatedly subjected to physical disturbance and colonies migrate very readily and frequently split to reproduce by budding. Given the high rate of nest relocation in response to physical disturbance reported in this species (Buczkowski & Bennett, 2009), as well as the particular mode of colony foundation, pharaoh ant colonies likely experience frequent encounters with a broad range of parasites, including widespread, generalist entomopathogenic fungi of the genus Metarhizium.

We therefore investigated the nest site preference of pharaoh ants during nest relocation through a series

118 of binary choice tests between three nest types: nests containing nestmates killed by the entomopathogenic generalist fungus Metarhizium brunneum (infected nests), nests containing nestmates killed by freezing (uninfected nests), and empty nests. We predicted that empty nests would be preferred, followed by nests with freeze-killed cadavers, and finally nests containing infectious cadavers.

Materials and methods

Housing and maintenance of ant colonies

Four source colonies of Monomorium pharaonis from a laboratory stock population with similar genetic background were used to each create 10 experimental colonies, consisting of 100 workers and 50 larvae, for each of the three different nest choice assays (n = 120 in total). Source colonies were reared in plastic boxes (238 × 203 mm and 175 mm high) with Fluon™-coated (BioQuip Products, Rancho Dominguez, CA, USA) walls to prevent escape. Colonies were given fresh water ad libitum in glass tubes sealed with cotton wool and fed twice a week with mealworms (Tenebrio molitor), almonds and a 1:3 proteins to carbohydrates ratio artificial diet (Dussutour & Simpson, 2008). Each colony was also provided with multiple tubes wrapped in aluminium foil, serving as nests. All colonies were maintained at constant temperature and humidity (27 ± 1 °C, 65 % RH, 12:12 L/D cycle).

Experimental arena

Nest choice assays were performed in experimental arenas consisting of 15 cm diameter Fluon™-coated Petri dishes containing a “home” nest and two “experimental” nests (details about their structure can be found in Figure S1). The three nests in the experimental arena were positioned in order to have the entrances of the two “experimental” nests facing each other and at an equal distance (≈20 mm) from the “home” nest entrance. A water tube was positioned perpendicularly to the entrance of the “home” nest, whereas food (see above) was always located on the bottom-left side of the water tube.

Pathogen infectivity test

The pathogen used in this study was Metarhizium brunneum strain ARSEF 1095, formerly named Metarhizium anisopliae (Bischoff et al., 2009), that came from the USDA-ARS Collection of Entomopathogenic Fungi Cultures in Ithaca, New York, USA and was originally isolated from Carpocapsa pomonella [Lepidoptera: Olethreutidae] from Austria. The pathogen was cultured on SDA (Sabouraud Dextrose Agar) for 4 weeks at constant 23 ˚C prior to use. Susceptibility of M. pharaonis to the fungal isolate was assessed by dipping individual ant workers (n = 42) to conidia suspensions created from sporulating culture plates in a 0.05 % solution of a surfactant (Triton™ X-100, Sigma-Aldrich®, St. Louis, USA). The suspensions were previously

119 quantified with a haemocytometer (Neubauer-improved counting chamber) and diluted to the concentration of 2 × 108 conidia ml-1. Before infecting the ants we checked the viability of the conidia by plating 100 µl of 1 × 105 per ml solution on SDA agar plates, incubating for 19 h at 24 ˚C and recording germination at ×400 magnification (germination rate was 100 %).

To confirm that the death of exposed ants was indeed caused by Metarhizium, dead ants were surface sterilized following the protocol of Lacey & Brooks (Lacey & Brooks, 1997). In short, ants were sequentially dipped in 70 % ethanol, deionized water (dH2O), 5 % sodium hypochlorite, and three subsequent changes of dH2O. Cadavers were dried on sterile filter paper, and transferred to Petri dishes containing damp filter paper and a wet cotton ball in the centre. Corpses were placed in a circle around the cotton ball, at equal distance from it and from each other. Dead ants were then inspected for fungal growth over 8 days. We also quantified the amount of conidia present on sporulating cadavers by washing them in 1 ml of a 0.05 % solution of Triton X. The conidia collected were then counted using a haemocytometer.

The exposure of ant workers to conidia of the fungal isolate resulted in the death of 86 % (n = 36) of the individuals over 8 days. Three days after the death, 28 % of the individuals (n = 10) showed mycelia growth on the cuticle. The amount of conidia present on the cuticle was 6.5 × 105 ml-1 (calculated as average of 5 sporulating cadavers).

Infected cadavers preparation

For the experiment, ant workers were collected from the foraging area of each of the four source colonies and transferred to four Fluon™-coated Petri dishes according to their colony of origin. Every Petri dish was provided with the same food used in the experiment, a water tube and several pieces of SDA agar sporulating with the entomopathogenic fungus M. brunneum to ensure effective infection.

Every day, dead workers were gently removed from the Petri plates with a soft brush, and surface sterilized as described above. Cadavers were stored at 22 ± 1 ˚C for 2 weeks. Those ants that died from M. brunneum infection (confirmed by fungal growth out of the cadaver) and were covered with conidia weregently collected with a brush and transferred into the respective experimental nests.

Uninfected cadavers preparation

Ant workers were collected from the foraging area of each of the four source colonies and killed by freezing in a –20 ˚C freezer overnight. The following day, cadavers were surface sterilized just as the infected cadavers were, in order to minimize the difference in odour between infected and uninfected cadavers. After the treatment, ant cadavers were transferred to new vials and stored at –20 ˚C. The cadavers were thawed 1 h before being transferred to the respective experimental nests.

120 Nest choice assays

Three types of experimental nests were made: nests containing 5 nestmate cadavers killed by the M. brunneum fungus with visible sporulating mycelia (infected nests); nests containing 5 nestmate cadavers killed by freezing (uninfected nests), and empty nests. In each nest choice assay, experimental colonies were presented with a choice between two of the experimental nest types. In experiment A, ants were allowed to choose between infected and uninfected nests. In experiment B, the choice was between an empty and an infected nest. In experiment C, ants could choose between an empty and an uninfected nest (Fig. 1).

Assays were prepared 24 h before the start of the experiment by introducing the experimental colony into the home nest of the test arena. The entrances of the two experimental nests were blocked by several layers of Parafilm® to avoid exploration by the ants. To avoid bias in the choice because of the asymmetric position of the food in the test arena, the position of the two experimental nests was systematically switched for each replicate. Hence, in each of the three experiments, experimental colonies obtained from the same colony source (n = 10) were presented 5 times with a particular experimental nest type on the left and the other 5 times on the right with respect to the home nest entrance.

Nest choice assays were initiated by exposing the home nest to a light source and by the simultaneous removal of the Parafilm® blocking the entrance to the experimental nests. Experimental colonies were checked for nest choice (defined as the complete migration of the experimental colony into one of the two experimental nests) at 10 min intervals within the first hour, and once every hour for the following 4 h.

Statistical analysis

All statistical analyses were conducted in R (R Core Team, 2013). Out of 120 assays, 12 were discarded because ants were able to explore the experimental nests prior to the start of the experiment, and five were discarded because colonies did not move out from the home nest within the four hour observation period (Table S1). In the remaining 103 assays used for analysis, ants were observed exploring both experimental nests and a nest choice was made within three hours, with 91 (88 %) making a choice within the first 30 min (Figure S2).

To test whether the position of the food in the test arena had an influence on the nest choice outcome, for each of the three experiments, we constructed a generalized linear model (GLM) with binomial errors and logit link function, with the type of experimental nest chosen as the dependent variable and the position (left or right) of the nest chosen as a factor, and assessed whether removing the factor “position” from the model had a significant effect based on the likelihood ratio between the two models. Similarly, to test

121 whether the colony of origin of each experimental colony had an effect on the choice outcome, for each of the experiments, we used a likelihood ratio test to compare GLMs with and without colony of origin as an explanatory factor. Since neither the position of the chosen nest nor the origin of the experimental subcolonies had an effect on the choice outcome after correcting for multiple tests, we ran an exact binomial test on the pooled nest choice data for each of the three experiments to test whether there was a preference for a particular experimental nest type. We used a G-test to test whether or not the observed responses were consistent across the three experiments.

Results

Data from replicate colonies were pooled as overall we found no significant colony effect on nest choice in the experiments (likelihood ratio tests between the full model comprising colony origin as factor and the

2 null model comprising only the intercept. Experiment A; infected vs uninfected: χ 3 = 3.267, P = 0.352; 2 2 experiment B; infected vs empty: χ 3 = 2.679, P = 0.444; experiment C; uninfected vs empty: χ 3 = 9.397, P = 0.024; cutoff P = 0.0167 for α = 0.05 when correcting for multiple tests).

In experiment A, 26 experimental colonies chose infected nests and only five chose uninfected nests, implying that ants had a significant preference for the former (exact binomial test: P < 0.001, Fig. 2a). In experiment B, 21 experimental colonies chose infected nests and 13 the empty nests (exact binomial test: P = 0.229, Fig. 2b). In experiment C, 24 experimental colonies chose the empty nests and 14 uninfected nests (exact binomial test: P = 0.143, Fig. 2c).

The position of the experimental nests in the arena did not affect nest choice (likelihood ratio test between the model comprising the position, left or right, of the experimental nest chosen as factor and the null model

2 2 comprising only the intercept. Experiment A: χ 1 = 0.169, P = 0.681; experiment B: χ 1 = 1.783, P = 0.182; 2 experiment C: χ 1 < 0.001, P = 1).

The pattern of nest choice across the experiments was similar for the four colonies from which subcolonies

2 were derived (three-way contingency table: G = 22.06, d.f. = 17, P = 0.182), suggesting that the choices are species rather than colony specific.

Discussion

We found that emigrating Monomorium pharaonis colonies were not only able to discriminate between infected and uninfected nests, but surprisingly, they also displayed a clear preference for the infected nests

122 (Fig. 2). This significant preference for infected nest sites was however not observed when ant colonies could choose between an infected nest and an empty one, although infected nest sites were chosen more frequently. Colonies were also not able to discriminate between uninfected and empty nests, although the latter type was selected more often. All together, our results show that pharaoh ant colonies have a preference for infected nest sites. Empty nests appear to rank as an intermediate choice whereas uninfected nests are the least preferred (Fig. 2). These findings contrast with our initial hypothesis and with studies reporting that insects seem to possess the ability to perceive and avoid direct physical contact with entomopathogenic fungi (Meyling & Pell, 2006; Lam et al., 2010; Ormond et al., 2011), possibly using chemicals emitted by fungal spores (Davis et al., 2013).

One possible explanation of our findings is that either the fungal isolate used or the amount of conidia present on the surface of the sporulating cadavers do not represent a strong lethal threat for the ants, resulting in only a weak repulsive effect. Recent studies showed that workers of the termite species Macrotermes michaelseni are not only able to detect the presence of virulent fungi by olfaction, but also that different fungal isolates repelled individuals in dose and virulence related manner (Mburu et al., 2009; Mburu et al., 2011). Similar dose-dependent behaviours have been also reported in the ant species Atta sexdens rubropilosa (Jaccoud et al., 1999). Although we can’t completely rule out this hypothesis, preliminary test showed that pharaoh ant workers suffer high mortality after exposure to the M. brunneum strain employed in this experiment, confirming its high virulence. Furthermore, even if the amount of conidia present on dead bodies were not representing a genuine threat for the colony, uninfected or empty nests should still represent a better or equal choice when deciding where to relocate the nest given their lower sanitary risk.

The overall preference for infected nest sites brings us to suggest instead that sporulating cadavers may exert an attractive rather than a repulsive effect. A similar effect has been recently reported in the malaria mosquito Anopheles stephensi, where female individuals appear to be highly attracted by dead caterpillars infected with the entomopathogenic fungus Beauveria bassiana (George et al., 2013). Likewise, young queens of the ant Formica selysi display an initial attraction to nest sites contaminated with entomopathogenic fungi during the dangerous stage of colony foundation (Brütsch et al., 2014) . Behavioural responses to sporulating corpses can thus be highly variable between species and the attraction towards fungal conidia does not necessarily need to be seen as non-adaptive. Sporulating cadavers might indeed represent a resource rather than a threat for the host, possibly providing direct or indirect benefits to the individuals coming into contact with. If so, our observed preference order in the choice of the experimental nests might reflect a trade-off process performed by the ants, where infected nests are highly ranked because they provide a valuable resource. Empty nests do not provide benefits or costs, thus ranking as intermediate, whereas nests containing non infected corpses might represent a potential threat and thus least preferred.

One way that infected nests may present an attractive resource is that the ants may use the fungal conidia or

123 the mycelia as a food source, therefore gaining a direct benefit that they would not obtain from the choice of a neutral area such as an empty nest or from a nest containing uninfected corpses. Insect mycophagy is well-known among insects, ranging from the simple ingestion of yeast, mycelium or sporocarps to the elegant symbiosis exhibited by fungal-farming termites and ants (Boucias et al., 2012). However, the fact that we never observed pharaoh ant workers feeding on cadavers makes this hypothesis unlikely. We argue that it is more plausible that exposure to a fungal pathogen might provide an indirect benefit to the colony.

Recently it was demonstrated that social contact with individuals exposed to fungal pathogens reduces the susceptibility of nestmates to later exposure to the same pathogen (Ugelvig & Cremer, 2007). This “social immunization” process requires the transfer of a small number of conidia from exposed to healthy group members by grooming. In turn, healthy individuals pick up low-level inoculum that, although not lethal, leads to an up-regulation of specific immune genes involved in antifungal responses (Konrad et al., 2012). While this mechanism has so far only been shown in studies involving small groups of workers, individual active immunization could represent an adaptive strategy, leading to the acquisition of colony- level “vaccination” (Konrad et al., 2012). In order to be beneficial, active immunization of the colony through emigration to infected nest sites requires that the chances of re-encountering the same pathogen are particularly high and the exposure to low pathogen levels through the immunization process confers only a low average risk of mortality to colony members. These conditions appear to be fulfilled in our study system. In fact, the scavenger life-style and the frequent opportunistic nest relocation of pharaoh ants (Buczkowski & Bennett, 2009) make them likely to have repeated encounters with a common pathogen like Metarhizium during their colony life-span. Moreover, the number of conidia present on five nestmate cadavers is unlikely to be sufficient to trigger a deadly disease outbreak in the colony.

The emigration to an infected nest would also increase the possibility that all colony members have access to the pathogen and that they have equal opportunities to prime their immune system. This early vaccination might benefit not only adult individuals, but also larval stages (Rosengaus et al., 2013). In contrast, if colonies were only exposed to conidia brought back by scouts, the limited number of conidia together with the compartmentalized social structure of ant colonies (Mersch et al., 2013) would result in limited immunization of the colony. Indeed, previous authors have argued that compartmentalization in the social network of social insects is adaptive precisely because it limits the rapid spread of pathogens from incoming foragers to the rest of the colony (Naug & Camazine, 2002). Variation in pathogen mode of transmission and virulence presumably have resulted in different evolved responses, in some cases involving controlled exposure and others complete avoidance. Thus, we suggest that our results are best explained by a social immunization process, although further research is necessary to identify the precise mechanism through which this process can occur.

The observed preference for infected nests over nests containing non infected corpses could also be

124 explained by the higher average risk represented by more virulent undetected pathogens. Since cadavers killed by undetectable pathogens like viral diseases do not readily advertise their cause of death, the choice of infected nests could be interpreted as an active decision of the colony driven by the likely higher threat posed by unknown diseases. That is, it is better to live with a known threat that can be dealt with rather than an unknown threat.

A final hypothesis that could explain our results is that the pathogen attracts its host in order to increase its transmission rate. Examples of pathogen manipulation of host behaviours have been recorded in many species (Moore, 2002). However most cases involve specialist coevolved parasites, whereas the pathogen we used is a generalist fungal pathogen that targets a broad spectrum of insect hosts.

In conclusion, we showed surprisingly that colonies of the pharaoh ant Monomorium pharaonis have an overall strong preference for moving into nest sites infected by the entomopathogenic fungus Metarhizium brunneum. Contrary to the current view, we propose that controlled exposure to a low concentration of a common pathogen can provide a long-term benefit for the colony. As already shown in other species, ants can acquire long-term protection against diseases through social contact with previously exposed individuals. This “social immunization” process can in theory lead to colony-wide “vaccination” if all members come into contact with a low level of a particular pathogen. In this respect, our study is the first to show that the need for immunity can lead an entire, functional colony to prefer an infected nest site. Alternatively, the observed preference is non-adaptive and stems from the fungus being able to manipulate the ant host by yet unknown mechanisms. In either case this behavior raises interesting perspectives for biological control, not only of this but also of the many similar ant species that are regarded as some of the World’s most problematic invasive pests.

Acknowledgements

We thank David Nash for valuable comments on previous versions of this manuscript and Bernhardt Steinwender for help with morphological determination of fungal spores. The present study was supported by the Danish National Research Foundation (grant DNRF57) to the Centre for Social Evolution (LP and JSP). SV was funded by NIH-PERT fellowship K12GM000708.

References

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

125 Boucias, D., Lietze, V.-U., & Teal, P. (2012). Chemical signals that mediate insect-fungal interactions. Biocommunication of Fungi (ed. by G. Witzany), pp. 305-336. Springer, Dordrecht, Netherlands. Brütsch, T., Felden, A., Reber, A., & Chapuisat, M. (2014). Ant queens (Hymenoptera: Formicidae) are attracted to fungal pathogens during the initial stage of colony founding. Myrmecological News, 20, 71-76. Buczkowski, G. & Bennett, G. (2009) Colony budding and its effects on food allocation in the highly polygynous ant, Monomorium pharaonis. Ethology, 115, 1091-1099. Cremer, S., Armitage, S.A.O., & Schmid-Hempel, P. (2007) Social immunity. Current Biology, 17, R693-R702. Davis, T.S., Crippen, T.L., Hofstetter, R.W., & Tomberlin, J.K. (2013) Microbial volatile emissions as insect semiochemicals. Journal of Chemical Ecology, 39, 840-859. Dussutour, A. & Simpson, S.J. (2008) Description of a simple synthetic diet for studying nutritional responses in ants. Insectes Sociaux, 55, 329-333. Franks, N.R., Hooper, J., Webb, C., & Dornhaus, A. (2005) Tomb evaders: house-hunting hygiene in ants. Biology Letters, 1, 190-192. Franks, N.R., Mallon, E.B., Bray, H.E., Hamilton, M.J., & Mischler, T.C. (2003) Strategies for choosing between alternatives with different attributes: exemplified by house-hunting ants. Animal Behaviour, 65, 215-223. George, J., Jenkins, N.E., Blanford, S., Thomas, M.B., & Baker, T.C. (2013) Malaria mosquitoes attracted by fatal fungus. PLoS ONE, 8, e62632. Helanterä, H., Strassmann, J.E., Carrillo, J., & Queller, D.C. (2009) Unicolonial ants: where do they come from, what are they and where are they going? Trends in Ecology & Evolution, 24, 341-349. Hölldobler, B. & Wilson, E.O. (1990) The Ants. Harvard University Press, Cambridge, Massachusetts. Jaccoud, D.B., Hughes, W.O.H., & Jackson, C.W. (1999) The epizootiology of a Metarhizium infection in mini-nests of the leaf-cutting ant Atta sexdens rubropilosa. Entomologia Experimentalis et Applicata, 93, 51-61. Konrad, M., Vyleta, M.L., Theis, F.J., Stock, M., Tragust, S., Klatt, M., Drescher, V., Marr, C., Ugelvig, L.V., & Cremer, S. (2012) Social transfer of pathogenic fungus promotes active immunisation in ant colonies. PLoS Biology, 10. Lacey, L.A. & Brooks, W.M. (1997). Initial handling and diagnosis of diseased insects. Manual of Techniques in Insect Pathology (ed. by A.L. Lawrence), pp. 1-16. Academic Press, London, U.K. Lam, K., Tsang, M., Labrie, A., Gries, R., & Gries, G. (2010) Semiochemical-mediated oviposition avoidance by female house flies, Musca domestica, on animal feces colonized with harmful fungi. Journal of Chemical Ecology, 36, 141-147.

126 Mburu, D.M., Ndung’u, M.W., Maniania, N.K., & Hassanali, A. (2011) Comparison of volatile blends and gene sequences of two isolates of Metarhizium anisopliae of different virulence and repellency toward the termite Macrotermes michaelseni. The Journal of Experimental Biology, 214, 956-962. Mburu, D.M., Ochola, L., Maniania, N.K., Njagi, P.G.N., Gitonga, L.M., Ndung’u, M.W., Wanjoya, A.K., & Hassanali, A. (2009) Relationship between virulence and repellency of entomopathogenic isolates of Metarhizium anisopliae and Beauveria bassiana to the termite Macrotermes michaelseni. Journal of Insect Physiology, 55, 774-780. McGlynn, T.P. (2012) The ecology of nest movement in social insects. Annual Review of Entomology, 57, 291-308. Mersch, D.P., Crespi, A., & Keller, L. (2013) Tracking individuals shows spatial fidelity is a key regulator of ant social organization. Science, 340, 1090-1093. Meyling, N.V. & Pell, J.K. (2006) Detection and avoidance of an entomopathogenic fungus by a generalist insect predator. Ecological Entomology, 31, 162-171. Moore, J. (2002) Parasites and the Behavior of Animals. Oxford University Press, Oxford, U.K. Naug, D. & Camazine, S. (2002) The role of colony organization on pathogen transmission in social insects. Journal of Theoretical Biology, 215, 427-439. Ormond, E.L., Thomas, A.P.M., Pell, J.K., Freeman, S.N., & Roy, H.E. (2011) Avoidance of a generalist entomopathogenic fungus by the ladybird, Coccinella septempunctata. Federation of European Microbiological Societies Microbiology Ecology, 77, 229-237. Passera, L. (1994). Characteristics of tramp species. Exotic Ants: Biology, Impact, and Control of Introduced Species. (ed. by D.F. Williams), pp. 23-43. Westview Press, Boulder, Colorado. R Core Team. (2013) R: a language and environment for statistical computing. R Foundation for Statistical Computing. URL http://www.R-project.org, Vienna, Austria. Rosengaus, R.B., Malak, T., & MacKintosh, C. (2013) Immune-priming in ant larvae: social immunity does not undermine individual immunity. Biology Letters, 9. Schmid-Hempel, P. (1998) Parasites in Social Insects. Princeton University Press, Princeton, New Jersey. Schmidt, A.M., d’Ettorre, P., & Pedersen, J.S. (2010) Low levels of nestmate discrimination despite high genetic differentiation in the invasive pharaoh ant. Frontiers in Zoology, 7, 20. Seeley, T.D. (2010) Honeybee Democracy. Princeton University Press, Princeton, New Jersey. Tsutsui, N.D., Suarez, A.V., Holway, D.A., & Case, T.J. (2000) Reduced genetic variation and the success of an invasive species. Proceedings of the National Academy of Sciences of the United States of America, 97, 5948-5953. Ugelvig, L.V. & Cremer, S. (2007) Social prophylaxis: group interaction promotes collective immunity in ant colonies. Current Biology, 17, 1967-1971. Ugelvig, L.V. & Cremer, S. (2012) Effects of social immunity and unicoloniality on host-parasite

127 interactions in invasive insect societies. Functional Ecology, 26, 1300-1312. Wetterer, J.K. (2010) Worldwide spread of the pharaoh ant, Monomorium pharaonis (Hymenoptera: Formicidae). Myrmecological News, 13, 115-129.

128 Fig. 1. In each assay, colonies of Monomorium pharaonis were provided with the choice between two experimental nest types: infected vs uninfected nests (experiment A); infected vs empty (experiment B) and uninfected vs empty (experiment C). The position of the experimental nest types was swapped between trials to avoid position biases in the choice. The position of the water tube (W) and food (F) is also showed.

Experiment A Experiment A Pathogen-free cadavers Infectious cadavers

× 5 H × 5

Experiment B Experiment B Empty Infectious cadavers E E

× 5 W Experiment C Experiment C Empty Pathogen-free cadavers F × 5

1 cm

129 Fig. 2. Nest choice of Monomorium pharaonis in the three experiments. (a) Infected vs uninfected; (b) Infected vs empty; (c) Uninfected vs empty.

(a) (b) (c) P = 0.143 30 P = 0.0002 25 P = 0.229 25

25 20 20 20 15 15 15 10 10 10 Number of nests Number of nests Number of nests

5 5 5

0 0 0 Infected Uninfected Infected Empty Empty Uninfected

130 Supporting information

Figure S1. - Structure of the artificial nests. Each nest consisted of four layers: two 40 × 40 mm glass plates (layer 1 and 3) sandwiching a frame of hard plastic strips (layer 2; width = 4 mm; height = 1.5 mm) fixed to the bottom plate with double-sided tape. The top glass plate was covered with a red acetate sheet of equal size (layer 4). Home and experimental nests only differed in the position of the 4 mm wide entrance hole. Floor area was 1024 mm2; nest cavity volume was 1536 mm3.

Layer Home nest Experimental nests

131 Table S1 (following three pages). – Data collected during the observation period for each of the three experiments. Colony ID is the original colony from which experimental colonies were created; Discarded indicates those assays not used in the analysis because experimental nests were explored before the behavioral observation started or colonies did not move out from the home nest during the observation period (5 h); # workers indicates the total number of workers entering experimental nests, including multiple entrances by the same ant, before a choice was made. This data was collected in 5-minute observations during the assay. Secondary relocation indicates whether the colony moved first into one experimental nest and then into the other during the observation period. If yes, the last choice was counted as the final decision. Final choice is the experimental nest where the colony was located at the end of the observation period. E: empty nest; I: infected nest; U: uninfected nest; H: home nest. Time is the period between the start of the experiment and the final choice made by the colony (NA = Not applicable).

132 Table S1 - Experiment A

Colony Assay Discarded # workers # workers Secondary Final Time ID infected uninfected relocation choice

4077 1 22 5 I 15 4077 2 69 46 I 60 4077 3 11 4 I 25 4077 4 31 28 yes U 60 4077 5 11 10 U 15 4077 6 0 9 U 15 4077 7 2 5 I 15 4077 8 53 21 I 35 4077 9 44 50 I 60 4077 10 2 0 I 15 X2 1 57 2 I 15 X2 2 11 75 U 25 X2 3 102 1 I 5 X2 4 4 4 yes I 15 X2 5 57 5 I 15 X2 6 9 10 I 15 X2 7 yes X2 8 yes X2 9 yes X2 10 yes 56 9 H NA X7 1 39 1 I 15 X7 2 yes 61 11 H NA X7 3 94 41 I 120 X7 4 117 41 I 60 X7 5 yes X7 6 yes X7 7 yes X7 8 3 0 I 15 X7 9 yes X7 10 54 1 I 15 X8 1 11 2 I 25 X8 2 1 2 I 15 X8 3 2 2 U 15 X8 4 35 1 I 15 X8 5 4 18 I 15 X8 6 35 2 I 25 X8 7 4 0 I 180 X8 8 0 2 I 25 X8 9 16 12 I 25 X8 10 21 64 I 15

133 Table S1 (continued) - Experiment B

Colony Assay Discarded # workers # workers Secondary Final Time ID infected empty relocation choice

4077 1 6 5 E 15 4077 2 6 6 E 15 4077 3 6 51 E 15 4077 4 4 2 E 15 4077 5 0 8 E 15 4077 6 61 3 I 15 4077 7 44 3 I 15 4077 8 36 11 I 25 4077 9 69 12 I 15 4077 10 60 172 yes I 120 X2 1 65 3 I 15 X2 2 1 8 I 15 X2 3 47 7 I 15 X2 4 yes 61 53 H NA X2 5 39 0 I 15 X2 6 16 1 I 15 X2 7 95 3 I 15 X2 8 84 89 yes E 25 X2 9 yes X2 10 yes X7 1 55 4 I 15 X7 2 6 3 I 15 X7 3 31 2 I 15 X7 4 2 95 E 15 X7 5 0 37 E 15 X7 6 107 6 I 15 X7 7 87 11 I 15 X7 8 yes 48 87 H NA X7 9 32 104 E 15 X7 10 yes X8 1 0 141 E 5 X8 2 0 3 E 15 X8 3 106 72 I 15 X8 4 2 0 I 15 X8 5 19 5 I 15 X8 6 5 43 E 15 X8 7 138 48 I 25 X8 8 8 82 E 25 X8 9 182 5 I 5 X8 10 yes

134 Table S1 (continued) - Experiment C

Colony Assay Discarded # workers # workers Secondary Final Time ID uninfected empty relocation choice

4077 1 18 4 U 15 4077 2 52 22 U 25 4077 3 8 6 E 15 4077 4 25 3 U 15 4077 5 40 61 E 25 4077 6 151 38 U 180 4077 7 125 15 U 25 4077 8 144 13 U 25 4077 9 32 104 E 25 4077 10 57 12 U 25 X2 1 57 49 E 60 X2 2 2 11 E 15 X2 3 5 35 E 15 X2 4 45 62 E 15 X2 5 12 92 E 5 X2 6 6 45 E 25 X2 7 8 157 E 25 X2 8 143 31 U 35 X2 9 21 181 E 15 X2 10 14 141 E 25 X7 1 26 25 E 15 X7 2 48 31 E 120 X7 3 11 97 E 5 X7 4 15 7 E 15 X7 5 32 37 E 25 X7 6 36 17 U 25 X7 7 1 52 E 15 X7 8 27 32 H NA X7 9 10 32 E 15 X7 10 76 19 U 15 X8 1 8 7 E 15 X8 2 19 51 E 25 X8 3 14 44 E 15 X8 4 6 9 E 15 X8 5 51 10 U 15 X8 6 87 14 U 15 X8 7 93 8 U 25 X8 8 158 23 U 25 X8 9 8 68 E 15 X8 10 yes

135 Figure S2. – Step plot showing the time needed for experimental colonies to make a choice between the nests offered.

110

100

60 Experiment A B 50 C Combined 40 Cumulative nest choice (number of trials)

0 5 15 25 35 45 55 120 180 Time (min)

136 Acknowledgements

137 First of all I’d like to thank my supervisor, Jes, for all his support during this almost 4 years of PhD studies and for teaching me how to be less chaotic and more organized in the work I do (I hope I have learnt something and improved…at least a bit!).

I’d like to thank Koos, the Big Boss! After the British Library, his brain is probably the place containing most of the human knowledge in whatever field you can mention or even think about! Discussing with you has been not only a great honor, but also an invaluable source of inspiration.

I want to thank David Nash for all his advices in statistics,to have helped me when I was in trouble with my projects and to have organized a super nice ant course!

How can I forget Patrizia! Thank you for providing invaluable help for analyzing those waxy, slimy layers of stuff that ants have on the cuticle and that we commonly call CHCs. But I thank you also for the hospitality in Paris, the good wine at conferences and all the laughing we had.

Sylvia, simply the best lab technician in the world and the best second-mother everyone can have  Never forget your cakes…yummy…

A special thank to my office mate Pepijn. How I’m gonna survive without our daily, pointless and never ending discussions??!! Thank you for all the help you gave me in statistics and illustrator, and for all the victories you allowed me to have in FIFA . Foremost, thanks for being a good friend more than a colleague!

To Panos, the most calm person I ever met! Despite your greekness and your love for Justin Bieber you are an amazing human being! I will never forget the good time I had with you, especially during those nights of aggression tests…

Thank you Saria. You are simply…..unique! Like your underwear! I owe you 3000 cigarettes, and many laughs as well 

To Rasmus, the best fisherman in the world! Without you I would still be living under a bridge Thank you for all, especially at the beginning of my adventure in CPH. I’m sure that the day I will defend this thesis, we will be together at “5 stars” to finish the celebrations ;-)

Special thanks to Sami, the best roommate in conferences, and Dora, who managed to survive my bad jokes

.

138 Sarah, Joanito, Marlene, Andreas, Brigitte, Rachelle, Nicoleta, Charlotte, Morten, Sanne, Bettina, Susanne,

Michael, Henning for making the CSE a lovely and enjoyable working place. Nick, Volker, Luke, Line, Tim,

Anna, Darinka, Jeroen and all those people who have left the CSE but not my memories 

To Jelle, simply because you are the best!

Kostis, David, Jon and el comandante Alex: more than friends. No words to describe you (those I have in mind are all bad and insulting ….outrageous I would say, right Jon?). No seriously…I love you guys! To

Katie, Jonas, Peter, Sal, and all the CMEC people with which I enjoyed my Fridays at the Cafeen?

Bernhardt, Louise, Milan. You made my first year in CPH so lovely and alcoholic… Thank you for being also very good friends.

Claudiuccia, Francesco, Giulia, Andrea, Sandruccio che tra un po’ si sposa, Claudio il pazzo: ma quante risate ci siamo fatti in questi anni? Grazie di essere stati sempre con me e per avermi sopportato. Specialmente

Claudia per aver avuto il coraggio di vivere con me un mese tra zucchine pelose :-D . Un saluto anche a

Nino e Franco per il cibo e a tutti gli italiani che contribuiscono a rendere la Danimarca meno seria e piu’ truffaldina.

To mamma Laura e papa’ Giuseppe, who always believed in me . To my lovely sister Mariateresa and my lovely brother-in-law Domenico. Francesco e Greta always in my heart.

To my lovely Danka, who never stopped to support me! Love you mio amore!

To Francesco e Luca, my best friends forever. Filippo, Lisa, Serafino, Peppe Federico, Eugenia, Marando,

Niccareddu, Manlio, Antoniuccio Gagliardi il porcellino, Gilduzza, Renatuccio, Lella, Giordana, Alessia,

Marcello il Cobra, Floriano/Mariano, Antonio Amodeo, Mario, Carletto, il ragionier Di Palma e l’ingegner

Tarallo....madonna quanti siete!!! So che mi sto scordando qualcuno, ma solo in queste pagine perché porto tutti dentro nel mio cuore!

Finally, the biggest thank goes to...THE PHARAOH ANT, without which I would not have had a PhD degree! Now you can freely escape the nest boxes and invade as many places as you want (no no...I’m kidding!!!)

139