ORIGINAL PAPER

Effect of group size and caste ratio on individualsurvivorship and social immunity in a subterranean termite

Qi Gao & Michael J. Bidochka & Graham J. Thompson

Received: 29 December 2010 /Revised: 20 June 2011 /Accepted: 27 June 2011 /Published online: 15 July 2011# Springer-Verlag and ISPA 2011

Abstract Individuals living within social groups maybenefit from the efficiencies of division of labour, but onthe other hand render themselves vulnerable to sociallytransmitted disease. This cost to social living shouldpromote cooperative barriers to disease transmission,especially in eusocial taxa where spatial and geneticproximity to nestmates are characteristically pronounced.Termites are eusocial yet little is known about how theirsociality is deployed to resist contagion. In this study, wemanipulate two variables that are expected to affect thenumber and nature of social interactions and measure theability of individuals within groups to resist fungalinfection. From laboratory experiments on field-collectedcolonies, we report that both group size and castecomposition directly affect the survivorship of individualswithin groups, but only caste composition moderatessurvivorship upon immune challenge. Our study thereforeprovides no statistical evidence that individual Easternsubterranean termites (Reticulitermes flavipes) have in-creased resistance to disease in crowded groups—that is,there is no evidence for a density-dependent social immuneresponse. Our results do suggest, however, that the caste-specific nature of interactions may be important forcontrolling disease in a social context.

Keywords Social insect . Social immunity . Isoptera .

Survivorship analysis . Reticulitermes flavipes . Statisticalinteraction

Introduction

The Hymenoptera (ants, bees, wasps, sawflies) and Isoptera(termites) contain many social species, some of which arefully eusocial (all termites) and thus have a permanenthelper caste and strong reproductive division of labour(Wilson 1971). Eusocial species typically live in large,densely populated colonies of closely related and frequentlyinteracting individuals. These demographic attributes arepredicted to make social insects vulnerable to the socialtransmission of disease (Alexander 1974; Hamilton 1987;Schmid-Hempel 2011). Although group living is a burdenon individual immunity, innate immune costs may be offsetby behavioural adaptations that are not available to solitaryinsects. For example, social taxa can take advantage of theirfrequent interactions for mutual grooming or exploit othergroup-enabled behaviours that may reduce pathogen load(Cremer et al. 2007; Wilson-Rich et al. 2009).

In the honey bee Apis mellifera examples of so-called‘social immunity’ include hygienic behaviour (Lapidge etal. 2002), the raising of offspring in sterile nurseries(Burgett 1997), social ‘fever’ in response to disease (Starkset al. 2000), cadaver disposal (Visscher 1983) and heightenedrisk-taking by infected individuals (Schmid-Hempel 2005).With respect to the study of termite social immunity,however, initial research has indicated that at least onedampwood (Termopsidae) species will increase mutual

Q. Gao :G. J. Thompson (*)Department of Biology, University of Western Ontario,1151 Richmond Street North,London, ON, Canada N6A 5B7e-mail: graham.thompson@uwo.ca

M. J. BidochkaDepartment of Biology, Brock University,500 Glenridge Ave,St. Catharines, ON, Canada L2S 3A1

acta ethol (2012) 15:55–63DOI 10.1007/s10211-011-0108-7

grooming in the presence of infectious spores (Rosengauset al. 1998b), communicate information about the presenceof disease to nestmates (Rosengaus et al. 1999), cannibalise-

infected nestmates (Rosengaus and Traniello 2001), produceantifungal faeces and other body exudates to controlinfection within the nest (Rosengaus et al. 1998a), and eventransfer immunity from immunised to previously unexposednestmates through a process called ‘social vaccination’(Traniello et al. 2002). These observations from Zootermopsisangusticolis suggest that termites in general may also have awell-adapted social immune system whereby behaviouralinteractions help resist the spread of disease.

The ability for termites or other insects to mount a socialdefence will likely depend on the average frequency ofinteractions within their societies. Careful studies on Z.angusticolis do support the notion that termite socialimmunity is density dependent—for example, the survivor-ship of nymphs was higher when exposed to disease insmall groups compared to those exposed in isolation(Rosengaus et al. 1998b). In addition, the effectiveness ofsocially enabled defences may also depend on the nature ofthese interactions—for example, as mediated by different taskspecialists or castes. Termite soldiers may be specialised toprotect the colony against macroorganisms like predatory ants(Prestwich 1984), but compared to workers may be poorlyadapted to mechanically defend against pathogenic micro-organisms. At any rate, behavioural repertoires of termitecastes differ (Watson et al. 1985), as they do for all socialinsects, and thus group caste composition may also affect acolony’s ability to coordinate a social defence.

In this study, we test the potential for a common pestspecies of termite, the Eastern subterranean termite Reticuli-termes flavipes (Rhinotermitidae), to mount a social defenceby quantifying the extent to which the number and nature ofsocial interactions mediate individual survivorship in the faceof disease. Specifically, we challenge groups of termites witha generalist entomopathogen and monitor their survivorshipagainst untreated controls as a function of group size andgroup caste composition. If individual immunity is affectedby the frequency of social interactions, then we expect thetreatment (immune challenge) effect on survivorship to varyas a function of group size. Moreover, if individual immunityis affected by the nature of inter-individual interactions, thenwe expect the treatment effect on survivorship to vary as afunction of caste ratio. In both cases, we do not yet know themechanisms involved, so we do not make directionalpredictions about what specific group size or caste ratioought to be important for social immunity in Reticulitermes.

Methods

Termite collection and maintenance

In the summer (June–August) of 2009, we collectedtermites from three genetically unrelated and well-

characterised field colonies (Raffoul et al. 2011) in PointPelee National Park (Essex County, ON, Canada), trans-ported them back to the laboratory at the BiotronEnvironmental Research Facility at the University ofWestern Ontario, Canada, and reared them within plasticcontainers (roughly 60×40×15 cm) within an environmentalchamber set to 26°C, 85% relative humidity (RH) and 24 hdarkness. To perpetuate each colony, we fitted each containerwith moist play sand as substrate, and provided plywood andsingle-face, two-ply cardboard as food. We checked coloniesregularly for health and vigour, and provided additionalwater via a saturated sponge and vaporizer, as necessary.Each field-collected laboratory-maintained colony initiallyconsisted between 1,000 and 2000 individuals and containedall three of the species-typical castes—that is, nymphs andfunctionally sterile workers and soldiers (Lainé and Wright2003). For northern populations of this species, the alate(imago) caste is rare (Myles 1999) and was not observed inour samples.

Fungal culture and infectivity trails

We used an entomopathogenic strain of Metarhiziumanisopliae (strain number 2575, from USDA collection,Ithaca, NY, USA) to infect groups of termites. This nativefungus is known to cause significant mortality in termites; ituses hydrolytic enzymes to penetrate the insect cuticle,subsequently the hyphae ramify the haemocoel (Bidochkaand Khachatourians 1994). First, we plated conidia (asexualspores) of M. anisopliae onto potato dextrose agar (PDA)media plates, and cultured them at 25°C in darkness (Myles2002). After approximately 14 days, we harvested theconidia by washing each plate with a sterile 0.01% Tween80 solution, and then estimated the concentration of conidiain suspension using a conventional haemocytometer(Hausser Scientific). From these washes, we made dilutionsof known concentration (103, 105, and 107 spores permilliliter), and stored these (4°C) for no more than 1 weekprior to using them in infectivity trials. We used these serialdilutions to first establish a dose for this strain that issufficient to induce mycosis.

We treated groups of termites by allowing them to walkwithin a plastic Petri dish (100×15 mm) lined with filterpaper (Whatman qualitative no. 5) impregnated with 1 mlof conidia suspension (Thompson et al. 2003). Weestablished control groups from the same procedure butusing sterile, conidia-free Tween solution. We could notprecisely control for age of termites. Because termites arehemimetabolous, however, age is correlated with instar andsize. We used only late instar (large) workers, late instar(brachypterous) nymphs, and soldiers (a terminal caste) inour analyses. Beyond this, we assumed a random distributionof ages with respect to treatment. Our sampling into Petri

56 acta ethol (2012) 15:55–63

arenas may have included individuals that would nototherwise be interacting (for example, if they were spatiallyseparated) but we assume they do not interact any more or lessas a consequence. After 1 h, we transferred treated and controlgroups of termites to new Petri dishes lined with clean filterpaper and damp cardboard. We then monitored the survivor-ship of each group (maintained at 26°C, 85% RH and 24 hdarkness, as above) daily, for a maximum of 15 days.

Fungal virulence on R. flavipes

To establish an appropriate dose for this strain that issufficient to induce mycosis and impose an immunechallenge, we tested a series of spore dilutions (spores permilliliter) for their effect on termite survivorship. For thispreliminary analysis, we simply compared the survivorshipof treated versus control groups using a Kaplan–Meiersurvivorship analysis. We used a fixed caste ratio (allworkers) and fixed group size (n=40).

Testing for social immunity

We used survivorship (time to death) as a measure of theability for individuals within groups to withstand infection.To test whether group membership affected this ability, wemeasured individual survivorship as a function of variationin two socially important traits, group size and castecomposition. Specifically, we measured differences insurvivorship between treated and control groups across arange of group sizes for a fixed caste ratio (all workers),and across a range of caste ratios for a fixed group size. Weassume that variation in group size (in effect, density) willaffect the number of social interactions within groups, whilevariation in caste composition will affect the nature ofsocial interactions within groups. Thus, if R. flavipesexpresses a type of social immunity whereby the nature ornumber of interactions affects an individual’s ability towithstand infection, then we expect a significant interactioneffect of treatment × group size or treatment × caste ratio onsurvivorship, respectively.

Group size assay

We established groups of treated and control termites acrossa range of size classes: n=5, 10, 20, and 40 workers. Thesegroup sizes are small compared to the mature colony sizesof free-living colonies, but are experimentally tractable andreflect the size range of small parties and of incipientcolonies. For each size class, we established a total of threebiological replicates whereby each replicate was from anunrelated field colony. Here, relatedness was previouslyestimated from microsatellite DNA analysis (Raffoul et al.2011). We recorded survivorship of each replicate daily and

used these data to estimate the baseline hazard function viaCox proportional hazard regression (Cox 1972), as imple-mented in PASW Statistics (version 18.0 for Mac).Significant departures in shape of the hazard function weretested via step-wise addition of the factors ‘treatment’(treated vs. control), ‘grouping factor’ (4 levels: 5, 10, 20,40), and their interaction term to the model and comparingthe fit (log likelihood) against the baseline model in whichthese parameters were not included. We tested goodness-of-fitbetween models using likelihood ratio tests.

Caste ratio assay

We established groups of treated and control termites acrossa range of caste ratios. Caste ratios vary widely in northernpopulations of R. flavipes, such that some colonies arereplete with sterile soldiers (exceeding 10%) while othersare all but void of the (potentially) reproductive nymphcaste (GJT, personal observation; Husby 1980). Thisvariation in caste composition probably reflects eachcolony’s plastic response to resources and subsequentinvestment into defence (soldiers) versus reproduction(nymphs), which may be erratic at the northernmostextreme of its range. Using a fixed group size (n=10), weexperimentally manipulated the proportion of soldier/worker/nymph. From an observed average starting ratio of2:6:2 (averaged from three laboratory colonies), weadjusted the proportion of workers to represent mild(3:4:3 and 4:2:4) or extreme (5:0:5) deficits in the workercaste. Likewise, we established groups representing a mild(1:8:1) or an extreme (0:10:0) excess of the worker caste,relative to our starting ratio. Though arbitrary, we chose tomanipulate the proportion of workers because the functionalversatility of this caste makes it a good candidate to enact asocial immune response. As above, we established a total ofthree biological replicates for each caste category andrecorded survivorship daily as input for Cox regressionanalysis.

Confirmation of mycosis

Though M. anisopliae is a confirmed entomopathogen of awide range of insects, including termites (Zoberi 1995;Rath 2000), we made an effort to verify the cause ofmortality within our own trials. As termites died daily, weremoved each cadaver and surface sterilised them individuallywith 95% alcohol. We then transferred each cadaver into a0.25-ml Eppendorf tube containing PDA medium. Weincubated each tube at 25°C for 7 days (Rosengaus andTraniello 1997). If termites died from fungal infection, thentreated termites will likely grow visible mycelia and producespores. Meanwhile, if termites died from other factors, thencadavers are not likely to show such growth.

acta ethol (2012) 15:55–63 57

Results

Fungal virulence on R. flavipes

Concentrations of 103 and 105 conidia per milliliter wereinsufficiently virulent to affect survivorship, as indicated bya basic Kaplan–Meier survivorship analysis (log rank χ2:0.45 and 0.10 respectively, P value>0.05 in both cases,df 1, n=40). A concentration of 107 conidia per milliliterwas, however, sufficient to infect groups of termites andaffect survivorship relative to controls (log rank χ2=7.76,df 1, P=0.005, n=40). At this concentration, the 2,575fungal strain is therefore sufficiently virulent to inducemycosis via topical exposure. We use this concentrationfor conducting the assays.

Group size assay

There were significant effects of treatment and socialgrouping on termite survivorship (Table 1). Both infectionwith M. anisopliae and group size directly affectedindividual termite lifespan. Challenged termites had ahazard ratio of death 8.8 times higher (95% CI 6.6–11.8,Wald statistic=226.01, df 1, P<0.001) than did untreatedtermites, after controlling for variation in group size.Likewise, the least-densely grouped termites (in group size5) had a hazard ratio of death 1.9 times higher (95% CI1.3–2.9, Wald statistic=9.55, df 1, P=0.002) than the most-densely grouped termites (in group size 40), after controllingfor treatment. This social grouping effect on survivorship was,however, not related to diseases resistance: survivorshipincreased with group size in both infected and uninfected(control) groups, providing no evidence that individualEastern subterranean termites have increased resistance in

larger groups. That is, the treatment × group size interactioneffect on survivorship was not significant (Table 1). Termitesliving within larger social groups simply live longer, but arenot necessarily better (or worse) at resisting infection as aconsequence (Fig. 1).

Caste ratio assay

There was a significant effect of treatment and socialgrouping on termite survivorship (Table 1). Both infectionwith M. anisopliae and group caste composition directlyaffected individual termite lifespan. Infected termites had ahazard ratio of death 8.5 times higher (95% CI 3.26–11.6,Wald statistic=172.76, df 1, P<0.001) than untreatedtermites, after controlling for variation in caste ratio. Thegroup caste ratio (soldier/worker/nymph) with the lowestsurvivorship (1:8:1) had a hazard of death 4.4 times (95%CI 2.9–6.8, Wald statistic=44.79, df 1, P<0.001) higherthan that of the group caste ratio with the highestsurvivorship (4:2:4), after controlling for treatment. Further,variation in caste composition does appear to be related todisease resistance, as evidenced by a significant treatment ×caste composition effect on survivorship (Table 1). Thismoderating effect of social grouping on individual diseaseresistance was pronounced for three caste configurations(4:2:4 and 2:6:2 and 5:0:5) in which treatment and casteeffects were not strictly additive. Instead, for these groupssurvivorship following exposure to pathogen was signifi-cantly higher (4:2:4 Wald statistic=13.15, df 1, P<0.001;2:6:2 Wald statistic=11.49, df 1, P=0.001, 5:0:5 Waldstatistic=10.29, df 1, P=0.001, respectively) than the mostsusceptible group (1:8:1; Fig. 2).

Confirmation of mycosis

Within social grouping factor, the effect of fungal infectionon survivorship was highly significant across all groupsizes and across all caste ratios (log-rank chi-square range18.76–195.94, df=1 in all cases, P<0.001 in all cases). Fortreated groups of termites, the confirmation rate forMetarhizium infection was high. Of 529 cadavers collectedfrom treated and control groups over the 15-day censusperiod, 312 (59%) tested positive for live mycelia, and allof these were retrieved from the treated groups. No cadavertesting positive for mycosis was retrieved from control(untreated) groups.

Discussion

In this study, we used variation in group size and castecomposition to characterise termite susceptibility to infec-tion and tested whether social context modifies this

Table 1 Cox proportional hazard regression analysis for group sizeand caste ratio assay

Factor df −2 Log likelihood Chi-square P value

Group size assay

Treatment 1 3,564.67 263.54 <0.001

Group size 3 3,551.10 13.60 0.004

Treatment × group size 3 3,548.85 2.26 0.521

Caste ratio assay

Treatment 1 2,835.14 188.07 <0.001

Caste ratio 5 2,782.52 52.62 <0.001

Treatment × caste ratio 5 2,751.91 30.61 <0.001

In the group size assay, the factors are treatment (treated vs. control)and social grouping factor (4 levels: 5, 10, 20, 40 individuals), as wellas their interaction term (treatment × group size). In the caste ratioassay, the factors are treatment (treated vs. control) and socialgrouping factor (nymph/worker/soldier ratios), as well as theirinteraction term (treatment × caste ratio)

58 acta ethol (2012) 15:55–63

susceptibility. Using the Eastern subterranean termite R.flavipes as a model, we did not find a group size effect onsusceptibility to infection (Table 1; Fig. 1). Our assaytherefore provides no evidence for a density-dependentsocial immune response, at least not to an appreciabledegree over the range of size classes tested (5–40) againstthe common fungal pathogen M. anisopliae. We did,however, find an alternative social grouping effect: varia-tion in the caste composition of groups did moderateindividual susceptibility to infection, as evidenced by astrong treatment × caste composition effect on survivorshipfollowing exposure to the fungal pathogen (Table 1; Fig. 2).The interaction means that the treatment effect size depends

on the particular caste ratio of the group. We suggest thatfor Eastern subterranean termites, it may not simply be thesheer number of interactions, as is important for socialimmunity in other termite species (Rosengaus et al. 1998b),but rather the caste-specific nature of these interactions thatcan help control disease in a social context.

Effect of group size on disease resistance

The use of a relatively small number of termites inlaboratory containers does not approach the complexity oflarge colonies and subterranean habitat (Lenz 2009).Nonetheless, several comparable laboratory studies have

Fig. 1 The association between treatment and subsequent survivor-ship (±SE) as a function of group size. A Cox proportional regressionmodel confirmed that treatment (solid lines) on mono-caste groups (allworkers) is highly significant (β=−2.181, SE=0.145) and that groupsize is positively related to survivorship (relative to group size 5,

group size 10 β=−0.192, SE=0.240, Wald=0.640, P=0.424; groupsize 20 β=−0.531, SE=0.222, Wald=5.708, P=0.017; group size 40β=−0.647, SE=0.209, Wald=9.546, P=0.002). The interactionbetween group size and treatment on survivorship is, however, notsignificant (Wald=2.265, df 3, P=0.519). Dashed line controls

acta ethol (2012) 15:55–63 59

demonstrated the potential for a group density effect ondisease resistance. Specifically, nymphs reared in isolationshow a lower resistance to infection than do nymphs reared

in a group (Rosengaus et al. 1998b; Traniello et al. 2002).Dealates (imagoes) of this species were also better able toresist fungal infection when kept in male–female pairs, than

60 acta ethol (2012) 15:55–63

individuals of either sex kept in isolation (Rosengaus et al.2000). Data from subterranean termite species has been lessforthcoming, but preliminary studies from R. speratus havealso been used to suggest a group size effect on immunity:workers reared in groups of 10 were apparently moreresistant to infection than were singletons (Shimizu andYamaji 2003), and the related subterranean pest Copto-termes formosanus showed a nearly identical patternwhereby group membership seemingly conferred a higherlevel of resistance (Meikle et al. 2005; Yanagawa andShimizu 2007). One issue with this latter group of studiesis, however, that all termites are treated, with no comparison tountreated controls. As such, the treatment effect is confoundedwith the social grouping effect. From these preliminarystudies, it is not known therefore whether termites in crowdedgroups may simply survive better, with or without treatment.To test for a social effect on resistance (not just survivorship) itis the treatment × group size interaction term that isinformative.

We do find a strong effect of social group size onsurvivorship (Table 1), but this effect was not related toresistance. Termites living in larger groups live significantlylonger, confirming that for many species of termites groupsize is positively correlated with survivorship (Lenz andBarrett 1984; Rosengaus et al. 1998b). We do not rule outthe possibility of a density effect on resistance, and futurestudies could increase power to detect it over previousstudies with larger sample sizes and increased ranges in sizeclasses, perhaps spanning three or more orders of magnitude.Moreover, one could measure the magnitude or frequency ofthe group-enabled mechanism directly, if it were known. Innatural colonies, the population size of R. flavipes ismeasured in the hundreds of thousands, if not millions ofindividuals. Thus, the potential for inter-individual interac-tion is vast, and even small effects on each other’s healthmay confer large additive, multiplicative, or even exponen-tial effects on overall colony hygiene. Future studies seekingto test for a density-dependent social immune response in R.flavipes should do so by looking for a modulatory effect of

group size on the treatment effect. As a case in point,even Z. angusticolis in which density-dependent immunityis frequently reported (Rosengaus et al. 1998b; Traniello etal. 2002) shows no such effect when infection and socialgrouping are treated as independent factors (Pie et al.2005). This and the present negative result highlight theneed to carefully test density-dependent social immunityfrom a factorial design, or even a grouped factorial designthat accommodates the population structure that is typicalof social insects. The present study does not use a groupedfactorial design and therefore does not take the potentialnon-independence of colony members into account.

Effect of group caste composition on individual immunity

How social interactions among individuals change as afunction of caste composition within groups is difficult toquantify. Indeed, few studies have tested for caste effects ongroup immunity. Rosengaus et al. (2000) report a castecomposition effect on disease resistance in two species ofNasutitermes (Termitidae) whereby worker monocastegroups outperform mixed caste or soldier monocastegroups. In their study, however, the caste groups also variedin size so it is not yet clear to what extent caste versusgroup size contributed to the observed survivorships. Rath(2000) provides contrasting survivorship data, in this caseshowing that soldier–worker mixed caste groups of Nasu-titermes exitiosis resist M. anisopliae infection better thando monocaste groups of either caste.

In our assay, we do find that individual survivorshipvaries directly with caste composition in R. flavipes and,intriguingly, this variation is related to disease resistance(Fig. 2). We speculate that pathogens are either (1)interacting differently with the natural ability of each casteto withstand infection or (2) the division in labour affordedby castes has a moderating effect on disease risk. If theformer is true, anatomical or physiological characteristics ofcertain castes (nymphs and soldiers, as the case may be)may retard the onset of disease, even in the absence ofexplicitly hygienic behaviour interactions. If the latter, thencaste diversity may sharpen behavioural divisions in labourand improve disease resistance through transfer of caste-specific physiological defences or sharing of complimentarybehavioural defences (Boomsma et al. 2005; Elliot and Hart2010). Our study cannot differentiate between these twopossibilities.

The association between social immunity and caste ratiothat we observe is not dependent on the proportion ofworkers (0–100%). Instead, disease resistance peaked withparticular combinations of workers, soldiers and nymphs(Fig. 2). We speculate that soldiers and nymphs incombination are important to the colony in pathogendefence, and that pathogen-mediated selection for particular

Fig. 2 The association between treatment and subsequent survival(±SE) as a function of group caste ratio (soldier/worker/nymph). ACox proportional regression model confirmed that treatment (solidlines) on mixed-caste groups is highly significant (β=−2.134, SE=0.162) and that caste ratio is related to survivorship (relative to 1:8:1,ratio 5:0:5 β=−0.651, SE=0.200, Wald=10.531, P=0.001; ratio 0:10:0β=−0.754, SE=0.194, Wald=15.056, P<0.001; ratio 3:4:3 β=−0.797,SE=0.192, Wald=17.128, P<0.001; ratio 2:6:2 β=−1.141, SE=0.210,Wald=29.449, P<0.001; ratio 4:2:4 β=−1.488, SE=0.222, Wald=44.791, P<0.001). Further, the interaction between caste ratio andtreatment on individual survivorship is significant (Wald=27.07, df5, P<0.001). The effect of treatment was highest for a ratio mildly inexcess of workers (1:8:1) and lowest for a ratio mildly deficient inworkers (4:2:4). Different lower case alphabetical letters denotesignificant differences in the survival distribution relative to the mostsusceptible reference (ref) population. Dashed line controls

�

acta ethol (2012) 15:55–63 61

caste ratios, or even instar or gender ratios within castes(Rosengaus and Traniello 2001; Fefferman et al. 2007),may be an important aspect of termite social evolution.Soldiers have specialised mandibles and so might beinefficient groomers, but well equipped to remove or guardcadavers away from the colony. Mankowski et al. (2005)showed that for Coptotermes spp. the presence of workerswith soldiers significantly reduced soldier mortality, and atleast one study on Z. angusticolis (Rosengaus et al. 2007)showed that the soldier caste can upregulate antimicrobialproteins that are more effective at reducing conidialviability than are antimicrobials expressed by pseudergates(a worker-like caste found in some species). Soldiers maytherefore represent an important element of communaldefence against pathogenic microorganisms.

Conclusions

By using a behavioural assay, we showed that communaldefence in termites may not simply be a matter of numberof social interactions as inferred from group density, butmay depend on the nature of these interactions as inferredfrom group composition. This finding is significant becauseit suggests that termites have greater resistance to diseasethan would be predicted from the frequency of interactionsalone, and further indicates that the social immune responsein termites may be subject to divisions in labour such thatcertain caste combinations are required for optimal function(Cremer et al. 2007), as are so many other aspects of colonygrowth, survival and reproduction. By extension, becausegroup size and group caste ratio are uniquely socialvariables that are not applicable to the demographics ofsolitary animals, our results imply that R. flavipes expressesa type of ‘social immunity’ analogous to that found in someother social insects, including the honey bee A. mellifera(Evans et al. 2006) and dampwood termites Z. angusticolis(Rosengaus et al. 1998b). While density is expected toincrease the average number of interactions among indi-viduals within a society, caste-based divisions in labour areexpected to reduce the overall number of interactions(Wilson 1971; Jeanne 1999; Ratnieks and Anderson1999). Models that describe social interaction as a functionof group size suggest that social connectivity increases withgroup density per se, but that in natural colonies an increaseto the number of individuals imposes spatial constraints thatactually make complete mixing less likely (Naug andGadagkar 1999; Pie et al. 2004). This localization ofbehavioural interactions with increases to colony sizeprobably promotes behavioural specialisation and divisionof labour (Jeanson et al. 2007). It is therefore not obvioushow variation in the number (density) and nature (caste) ofsocial interactions will interact to affect total immunity in

natural populations. Future studies should try to estimate thisinteraction effect through manipulation of both variablessimultaneously within a single, fully factorial experiment.

Acknowledgments We thank Beth MacDougall-Shackleton, BryanNeff, Yolanda Morbey, Rebeca Rosengaus and all members of theSocial Biology Group at the University of Western Ontario for usefuldiscussion and comments, and R Greg Thorn for expert advice onfungal culturing. This work is part of a Doctoral thesis project for QG,and is funded in part by a Natural Sciences and Engineering ResearchCouncil Discovery (NSERC) grant to GJT.

Conflict of interest None.

References

Alexander RD (1974) The evolution of social behavior. Annu RevEcol Syst 4:325–384

Bidochka MJ, Khachatourians GG (1994) Basic proteases ofentomopathogenic fungi differ in their adsorption properties toinsect cuticle. J Invertebr Pathol 64:26–32

Boomsma JJ, Schmid-Hempel P, Hughes WOH (2005) Life historiesand parasite pressure across the major groups of social insects. In:Fellowes MHG, Rolff J (eds) Insect evolutionary ecology. CABI,Oxon, UK, pp 139–176

Burgett DM (1997) Antibiotic systems in honey, nectar, and pollen. In:Morse RA, Flottum K (eds) Honey bee pest, predators, anddiseases. A.I. Root Co, Medina, OH, pp 455–468

Cox DR (1972) Regression models and life-tables. J R Stat Soc Ser B-Stat Methodol 34:187–220

Cremer S, Armitage SAO, Schmid-Hempel P (2007) Social immunity.Curr Biol 17:R693–R702

Elliot SL, Hart AG (2010) Density-dependent prophylactic immunityreconsidered in the light of host group living and social behavior.Ecology 91:65–72

Evans JD, Aronstein K, Chen YP, Hetru C, Imler J-L, Jiang H, KanostM, Thompson GJ, Zou Z, Hultmark D (2006) Immune pathwaysand defence mechanisms in honey bees Apis mellifera. Ins MolBiol 15:645–656

Fefferman NH, Traniello JFA, Rosengaus RB, Calleri DV (2007)Disease prevention and resistance in social insects: modeling thesurvival consequences of immunity, hygienic behavior, andcolony organization. Behav Ecol Sociobiol 61:565–577

Hamilton WD (1987) Kinship, recognition, disease, and intelligence:constraints of social evolution. In: Itô Y, Brown JL, Kikkawa J(eds) Animal societies: theories and facts. Japan Sci. Soc. Press,Tokyo, pp 81–102

Husby WD (1980) Biological studies on the Eastern subterraneantermite Reticulitermes flavipes (Kollar) (Dictyoptera, Rhinotermi-tidae) in Southern Ontario. In: Department of EnvironmentalBiology, MSc. University of Guelph, Guelph, p 200

Jeanne RL (1999) Group size, productivity, and information flow insocial wasps. In: Detrain C, Deneubourg JL, Pasteels JM (eds)Information processing in social insects. Birkhäuser, Basel, pp 3–30

Jeanson R, Fewell JH, Gorelick R, Bertram SM (2007) Emergence ofincreased division of labor as a function of group size. BehavEcol Sociobiol 62:289–298

Lainé LV, Wright DJ (2003) The life cycle of Reticulitermes spp.(Isoptera: Rhinotermitidae): what do we know? Bull EntomolRes 93:267–278

Lapidge KL, Oldroyd BP, Spivak M (2002) Seven suggestivequantitative trait loci influence hygienic behavior of honey bees.Naturwissenschaften 89:565–568

62 acta ethol (2012) 15:55–63

Lenz M (2009) Laboratory bioassays with subterranean termites(Isoptera)—the importance of termite biology. Sociobiology53:573–595

Lenz M, Barrett RA (1984) Implications for comparability oflaboratory experiments revealed in studies on the effect ofpopulation density on vigour in Coptotermes lacteus (Froggatt)and Nasutitermes exitosus (Hill) (Isoptera: Rhinotermitidae,Termitidae). Bull Entomol Res 74:477–485

Mankowski M, Kaya HK, Grace JK, Sipes BS (2005) Differentialsusceptibility of subterranean termite castes to entomopathogenicnematodes. Biocontrol Sci Technol 15:367–377

Meikle WG, Mercadier G, Rosengaus RB, Kirk AA, Derouane F,Quimby PC (2005) Evaluation of an entomopathogenic fungus,Paecilomyces fumosoroseus (Wize) Brown and Smith (Deutero-mycota: Hyphomycetes) obtained from Formosan subterraneantermites (Isop., Rhinotermitidae). J Appl Entomol 129:315–322

Myles TG (1999) Review of secondary reproduction in termites(Insecta: Isoptera) with comments on its role in termite ecologyand social evolution. Sociobiology 33:1–91

Myles TG (2002) Isolation ofMetarhizium anisopliae (Deuteromycotina:hyphomycetes) from Reticulitermes flavipes (Isoptera: Rhinotermi-tidae) with convenient methods for its culture and collection ofconidia. Sociobiology 40:257–264

Naug D, Gadagkar R (1999) Flexible division of labor mediated bysocial interactions in an insect colony—a simulation model. JTheor Biol 197:123–133

Pie MR, Rosengaus RB, Traniello JFA (2004) Nest architecture,activity pattern, worker density and the dynamics of diseasetransmission in social insects. J Theor Biol 226:45–51

Pie MR, Rosengaus RB, Calleri DV, Traniello JFA (2005) Density anddisease resistance in group-living insects: do eusocial speciesexhibit density-dependent prophylaxis? Ethol Ecol Evol 17:41–50

Prestwich GD (1984) Defense mechanisms of termites. Annu RevEntomol 29:201–232

Raffoul M, Hecnar SJ, Prezioso S, Hecnar D, Thompson GJ (2011)Trap response and genetic structure of Eastern subterraneantermites (Isoptera, Rhinotermitidae) in Point Pelee National Park,Ontario, Canada. Can Ent 143:263–271

Rath AC (2000) The use of entomopathogenic fungi for control oftermites [Review]. Biocontrol Sci Technol 10:563–581

Ratnieks FLW, Anderson C (1999) Task partitioning in insect societies.Review 46:95–108

Rosengaus RB, Traniello JFA (1997) Pathobiology and disease transmis-sion in dampwood termites [Zootermopsis angusticollis (Isoptera,Termopsidae)] infected with the fungus Metarhizium anisopliae(Deuteromycotina, Hypomycetes). Sociobiology 30:185–195

Rosengaus RB, Traniello JFA (2001) Disease susceptibility and theadaptive nature of colony demography in the dampwood termiteZootermopsis angusticollis. Behav Ecol Sociobiol 50:546–556

Rosengaus RB, Guldin MR, Traniello JFA (1998a) Inhibitory effect oftermite fecal pellets on fungal spore germination. J Chem Ecol24:1697–1706

Rosengaus RB, Maxman AB, Coates LE, Traniello JFA (1998b)Disease resistance: a benefit of sociality in the dampwood termiteZootermopsis angusticollis (Isoptera: Termopsidae). Behav EcolSociobiol 44:125–134

Rosengaus RB, Jordan C, Lefebvre ML, Traniello JFA (1999)Pathogen alarm behavior in a termite: a new form of communi-cation in social insects. Naturwissenschaften 86:544–548

Rosengaus RB, Traniello JFA, Lefebvre ML, Carlock DM (2000) Thesocial transmission of disease between adult male and femalereproductives of the dampwood termite Zootermopsis angusticollis.Ethol Ecol Evol 12:419–433

Rosengaus RB, Cornelisse T, Guschanski K, Traniello JFA (2007)Inducible immune proteins in the dampwood termite Zootermopsisangusticollis. Naturwissenschaften 94:25–33

Schmid-Hempel P (2005) Evolutionary ecology of insect immunedefenses. Annu Rev Entomol 50:529–551

Schmid-Hempel P (2011) Evolutionary parasitology: the integratedstudy of infections, immunology, ecology, and genetics. OxfordUniversity Press, Oxford

Shimizu S, Yamaji M (2003) Effect of density of the termite,Reticulitermes speratus Kolbe (Isoptera: Rhinotermitidae), onthe susceptibilities to Metarhizium anisopliae. Appl EntomolZoolog 38:125–130

Starks PT, Blackie CA, Thomas D, Seeley PT (2000) Fever inhoneybee colonies. Naturwissenschaften 87:229–231

Thompson GJ, Crozier YC, Crozier RH (2003) Isolation andcharacterization of a termite transferrin gene up-regulated oninfection. Insect Mol Biol 12:1–7

Traniello JFA, Rosengaus RB, Savoie K (2002) The development ofimmunity in a social insect: evidence for the group facilitation ofdisease resistance. Proc Natl Acad Sci USA 99:6838–6842

Visscher PK (1983) The honey bee way of death: necrophoricbehaviour in Apis mellifera colonies. Anim Behav 31:1070–1076

Watson JAL, Okot-Kotber BM, Noirot C (1985) Caste differentiationin social insects. Pergamon, Oxford, p 405

Wilson EO (1971) The insect societies. Belknap Harvard UniversityPress, Cambridge

Wilson-Rich N, Spivak M, Fefferman NH, Starks PT (2009) Genetic,individual, and group facilitation of disease resistance in insectsocieties. Annu Rev Entomol 54:405–423

Yanagawa A, Shimizu S (2007) Resistance of the termite, Coptotermesformosanus Shiraki to Metarhizium anisopliae due to grooming.Biocontrol 52:75–85

Zoberi MH (1995) Metarhizium anisopliae, a fungal pathogen ofReticulitermes flavipes (Isoptera: Rhinotermitidae). Mycologia87:354–359

acta ethol (2012) 15:55–63 63