CONCEPTS & SYNTHESISEMPHASIZING NEW IDEAS TO STIMULATE RESEARCH IN ECOLOGY

Ecology, 91(1), 2010, pp. 65–72� 2010 by the Ecological Society of America

Density-dependent prophylactic immunity reconsidered in the lightof host group living and social behavior

SIMON L. ELLIOT1,3

AND ADAM G. HART2

1Department of Animal Biology, Federal University of Vicosa, Vicosa, MG, 36571–000 Brazil2Department of Natural and Social Sciences, University of Gloucestershire, Cheltenham, Gloucestershire GL50 4AZ United Kingdom

Abstract. According to the density-dependent hypothesis (DDP), hosts living at highdensities suffer greater risk of disease and so invest more in immunity. Although there is muchempirical support for this, especially from invertebrate systems, there are many exceptions,notably in social insects. We propose that (A) density is not always the most appropriatepopulation parameter to use when considering the risks associated with disease and (B)behavioral defenses should be given a greater emphasis in considerations of a host’s repertoireof immune defenses. We propose a complementary framework stressing the connectivitybetween and within populations as a starting point and emphasizing the costs represented bydisease above the risk of disease per se. We consider the components of immune defense andpropose that behaviors may represent lower-cost defenses than their physiologicalcounterparts. As group-living and particularly social animals will have a greater behavioralrepertoire, we conclude that with group living comes a greater capacity for behavioral immunedefense, most particularly for social insects. This may escape our notice if we considerphysiological parameters alone.

Key words: disease resistance; host–pathogen interactions; immunocompetence; insects; social behavior.

INTRODUCTION

A key theme in the evolutionary ecology of disease is

the degree to which hosts invest in defense against

parasites and pathogens. Profitable paradigms in thisarea have been that (1) disease risk is a function, in the

simplest instance, of host density (Anderson and May

1979, McCallum et al. 2001) and (2) this risk of disease

leads to selection on the host to minimize the potentialcosts of disease (Hochberg 1991, Reeson et al. 1998,

Wilson and Reeson 1998). The traits employed to

minimize the costs of disease are often known as

‘‘immunocompetence,’’ although this term and its usageare not without their critics (Ryder 2003, Siva-Jothy et

al. 2005, Viney et al. 2005).

A clear prediction that arises from these paradigms is

that hosts living at high densities should invest more indefense than those living at low densities (Hochberg

1991, Reeson et al. 1998, Wilson and Reeson 1998,

Altizer et al. 2003, Wilson and Cotter 2008). This

hypothesis of ‘‘density-dependent prophylaxis’’ (DDP)

has been supported in a range of studies, particularly

with invertebrate hosts (e.g., Wilson and Reeson 1998,

Wilson et al. 2003, Cotter et al. 2004, Wilson and Cotter

2008). Some of these have adopted a comparative,

phylogenetic approach wherein physiological parame-

ters such as hemocyte counts, cuticular melanization, or

measures of hemolymph phenoloxidase titres are used as

correlates of investment in defense against disease, and

comparisons are made across species. Key papers

adopting this comparative approach have characterized

lepidopteran species according to a solitary or gregar-

ious lifestyle with the hypothesis that the latter can be

shown to invest more in defense against disease

(Hochberg 1991, Wilson et al. 2003). However, these

studies do not universally support the hypothesis,

casting some doubt on the assumption that gregarious

species will be universally exposed to a higher risk of

disease. Alternative explanations can be found, such as

the possibility that, for more aggregated populations,

between-group transmission may be more difficult

(Wilson et al. 2003) so disease risk may actually be

lower.

An alternative yet complementary approach exploits a

phenomenon displayed by several insect species, namely

Manuscript received 11 March 2009; revised 1 May 2009;accepted 5 May 2009. Corresponding Editor: S. J. Simpson.

3 E-mail: selliot@ufv.br

65

density-dependent phase polyphenism, wherein the

phenotype (or ‘‘phase state’’) of some lepidopteran and

orthopteran species varies with population density

experienced by the mother and/or during development

of immatures, and investment in immune defense is

expected to vary accordingly. Thus, locusts reared at low

densities display a suite of morphological and behavioral

traits (e.g., cryptic coloration and a tendency to avoid

conspecifics) and display the solitaria phase state. In

contrast, individuals reared at higher densities (or simply

exposed to conspecifics or associated stimuli: phase state

is highly labile) tend to have a warning coloration and

aggregate with conspecifics, part of the suite of traits

that is known as the gregaria phase state (see Simpson et

al. [1999] for a review). The expectation is that the

gregaria-phenotype individuals invest more in immunity

than do solitaria individuals. Indeed, one study of the

desert locust (Schistocerca gregaria) has shown gregaria-

phase individuals to have relatively greater hemocyte

counts and resistance to a pathogenic fungus (Meta-

rhizium anisopliae var. acridum) than their solitaria

counterparts (Wilson et al. 2002). The plasticity of this

phenomenon has allowed the comparison of genetically

highly related lepidopteran or orthopteran individuals in

high- and low-density treatments and has broadly shown

a positive correlation between gregarious behavior and

immune defense (Reeson et al. 1998, Wilson and Reeson

1998, Wilson et al. 2002, 2003, Cotter et al. 2004, Bailey

et al. 2008, Reilly and Hajek 2008, Wilson and Cotter

2008). Similarly, organisms with variable densities, but

which are not well known a priori for density-dependent

phenotypic variability (such as the coleopteran meal-

worm Tenebrio molitor), may be revealed to have

elevated investment in resistance when reared at higher

densities (Barnes and Siva-Jothy 2000). Inevitably,

however, this pattern is not always observed and

particularly not in all of the parameters measured

(Wilson and Reeson 1998, Wilson et al. 2002, Cotter

et al. 2004, Wilson and Cotter 2008).

These studies have been seminal in identifying the

relationship between host density and investment in

defense. However, they have clearly demonstrated that

there is no simple and direct positive dependence of one

upon the other. In particular, studies on eusocial insects

have provided very little support for the DDP hypoth-

esis (e.g., Pie et al. 2005). Thus, the initial assumptions

(paradigms 1 and 2, above) may be too simplistic. We

can also question the degree to which physiological

parameters such as hemocyte counts, phenoloxidase

titres, or even encapsulation responses, while being

readily quantifiable, are appropriate measures of (or

proxies for) total investment in defense (over and above

the question of whether one is measuring an induced vs.

constitutive component of defense, an issue we will not

address here). An illustration of this complexity can be

seen in a study in which a range of physiological

responses of larvae of the lepidopteran Spodoptera

littoralis are considered together (Cotter et al. 2004).

In this example, it is found that some defenses increase

with density while others do not and still others actually

decrease, synthesized by Wilson and Cotter (2008). This

highlights the importance of considering as many

relevant parameters as possible (Siva-Jothy et al. 2005)

and we emphasize that behavioral defenses are a key

aspect to include.

It has been argued (Little et al. 2005) that invertebrate

immunology has to some degree departed from a

holistic, phenomenological approach (which character-

ized the initiation of the area of vertebrate immunology)

in favor of a more reductionist approach that seeks

similarities between vertebrate and invertebrate hosts as

a starting point and/or focuses upon specific readily

quantifiable physiological traits as correlates of invest-

ment in defense. As Little et al. (2005) point out, without

considering the whole organism, and its progeny,

responses to infection that could be expressed through

life-history traits (Minchella 1985) or transgenerational

changes in phenotype (Elliot et al. 2003) will be

overlooked. The same can easily be said of behavioral

responses to infection, such as behavioral fever (Starks

et al. 2000, Elliot et al. 2002a). It stands to reason that

the whole suite of possible responses, whether physio-

logical or behavioral, should be considered together.

AN EARLIER, AND BROADER, DEFINITION

OF ‘‘IMMUNOCOMPETENCE’’

Immune defense (or ‘‘immunocompetence’’) has been

the subject of many reviews and much debate in the past

few years. As empirical investigations have dug deeper

into the mechanics of invertebrate immune systems, one

early definition seems to have been left to one side. This

definition is of particular relevance here: Owens and

Wilson (1999:171) defined immunocompetence as ‘‘a

measure of the ability of an organism to minimize the

fitness costs of an infection via any means, after

controlling for exposure to appropriate antigens’’ [our

italics].

Whether or not the term ‘‘immunocompetence’’ meets

with favor, we feel that this definition is of particular

relevance when considering the selective forces that are

expected to be operating and the possible evolutionary

responses to these forces. It distils the host-parasite

interaction into an effect on the host’s fitness and does

not predefine mechanisms an organism might use to

mitigate these effects. It is, indeed, the costs of infection

which should drive host adaptation. This has been

explored theoretically by van Baalen (1998), leading to

the thought-provoking conclusion that, at high densities,

resistance may no longer be beneficial as the force of

infection is such that a host will eventually be reinfected

and succumb to disease anyway. In such a situation

hosts should invest instead in, say, rapid reproduction to

increase fitness. (Although van Baalen’s model incorpo-

rated pathogen clearance by the hosts, the general result

makes intuitive sense for other cases.) This type of life-

history ‘‘escape attempt’’ should be considered part of

SIMON L. ELLIOT AND ADAM G. HART66 Ecology, Vol. 91, No. 1

CONCEPTS&SYNTHESIS

the organism’s immune defense as it could ameliorate

the costs of infection (Minchella 1985). Thus, the

definition given above can be interpreted to incorporate

an almost unlimited suite of traits that may allow an

organism to reduce the fitness costs of infection. This

was highlighted by Little et al. (2005) when they

emphasized that the full range of host responses to

infection should be considered. In fact, in the original

papers on density-dependent prophylaxis, behavioral

responses were explicitly cited but were generally left for

later consideration (Wilson and Reeson 1998). A

notable exception to this was an empirical study by

Reeson et al. (2000) in which host behavior was indeed

shown to be an important component of the host’s

density-dependent susceptibility. As the studies cited

above have generated a body of empirical data on

group-living (but nonsocial) insects, a number of papers

have, particularly recently, explored the defenses em-

ployed by social insects to combat disease, concluding

that individuals have increased resistance when in

groups (Rosengaus et al. 1998, Traniello et al. 2002)

and employ diverse behavioral resistance mechanisms

such as hygienic behaviors, increased brood care, waste

management, behavioral fever, exclusion of infected

individuals, and avoidance behaviors performed by

infected individuals (Starks et al. 2000, Cremer et al.

2007, Ugelvig and Cremer 2007, Cremer and Sixt 2009,

Jackson and Hart 2009, Wilson-Rich et al. 2009).

In the face of so many possible host responses

contributing to overall host defense, we are presented

with a problem in identifying the most biologically

significant responses of a given system, and in untan-

gling the individual and summative contributions of

responses that at first sight appear to have less bearing

on defense. In an effort to do this, we here consider how

group living (which we use to redefine our x-axis) can

lead to responses in terms of immune defense sensu lato

(the y-axis). In particular, we compare predictions for

nonsocial and eusocial insects.

THE GROUP-LIVING X-AXIS: POPULATION DENSITY

OR CONNECTIVITY?

Which independent variable best represents group

living when we are concerned with host investment in

defense? To take density as the independent variable (as

has commonly been done) is to assume this to be the

most important determinant of the likely costs of

disease. It may not, however, even be the key

determinant of the risk of infection, let alone the ensuing

costs. For example, it is probably more realistic to take

disease transmission as a mixture of frequency and

density dependency rather than density dependency

alone (Antonovics et al. 1995, Begon et al. 1999). For

organisms that live at more or less fixed densities, an

example being herbivorous spider mites (Oduor et al.

1997), we would expect frequency dependence to be

predominant. Similarly, we may well expect highly social

species such as eusocial insects to maintain a character-

istic density, with a constant, possibly optimal (DeSouza

et al. 2001, Altizer et al. 2003, DeSouza and Miramontes

2004), contact rate, while the colony expands to occupy

more space. Frequency-dependent transmission has

been considered to be essential in considering vertebrate

disease dynamics and immune responses (Altizer et al.

2003). Despite all of this, studies of density-dependent

prophylaxis in insects have often taken insect host

density as the key independent variable to consider in

laboratory assays. A notable example is a study by Pie et

al. (2005) in which termite density is varied experimen-

tally and over a short time period, and no effect on

resistance to a fungal pathogen is found. This result

should not come as a surprise: firstly, termites are

unlikely to experience fluctuations in density, as

explained above, so are unlikely to have evolved

responses to this; secondly, if there were an increase in

density, we would expect a colony-level response such as

the production of more resistant workers, not individual

responses over 20 days in nymphs that are already more

than a year old; thirdly, we expect behavioral defense to

be of key importance, as discussed below (see The y-

axis: . . .). Further to the above arguments, van Baalen

and Beekman (2006) point out that, given a patchy

distribution of potential hosts, contact rates (a measure

of connectivity) within and between patches can differ,

affecting epidemiological progress and therefore the

likely evolutionary outcomes, an argument similar to

that proposed by Wilson et al. (2003).

Since density-dependent transmission may not be

expected in many systems, it may seem surprising that

density-dependent prophylaxis (DDP) has found sup-

port in a number of systems. In baculovirus–caterpillar

systems, for example, pathogen transmission may

increase with host density in a mass action fashion

(Dwyer 1991), but we also expect a saturating effect seen

as a decline in per capita transmission at higher densities

(D’Amico et al. 1996, Reeson et al. 2000, Wilson and

Cotter 2008). Despite this decline in per capita

transmission, disease risk still increases with density so

we do see an adaptive increase in resistance of insects at

higher densities (i.e., DDP) (Wilson et al. 2003).

In the face of these complexities, we suggest that a

more suitable independent variable to consider (vs. host

density) is some measure of connectivity between

individuals and/or between populations, as this will

determine in large part the chances for transmission to

occur. In principle, this is much harder to quantify than

is density. Connectivity can, however, be quantified

when one considers animal behavior, in particular the

behavior of animals displaying some degree of sociality,

however limited this sociality may be. If we take

behavioral connectivity to be a key parameter, therefore,

we may be considering a better correlate of the risks of

exposure to disease (where connectivity affects trans-

mission) even though we remain a step away from the

costs of disease. This cost, again, can influence the

expected evolutionary outcomes (van Baalen and Beek-

January 2010 67GROUP LIVING AND IMMUNE DEFENSE

CONCEPTS&SYNTHESIS

man 2006). There is, in fact, recent support for the

paramount importance of connectivity: in Otterstatter

and Thomson’s (2007) study of disease transmission in

bumble bees, contact rates emerged as the only

significant measured predictor of infection risk.

Contact rates alone will not cover many systems as

transmission often does not occur through direct contact

between live individuals. For example, most insect-

pathogenic fungi kill their host and then sporulate from

the cadaver; meanwhile, insect viruses may kill the host

and liquefy it, infective particles then being ingested by

new hosts, or may be transmitted via feces. Nevertheless,

inter-individual behavior will affect transmission. If we

consider a caterpillar and a virus, then it is virions

deposited on a substrate such as leaves that will affect

transmission. Caterpillars may avoid conspecifics or

may aggregate with them. These behaviors are compo-

nents of connectivity and may have greater relevance

than density. Once adult, moths can affect transmission

through dispersal and mating behavior; again, these

behaviors make up connectivity between individuals or

populations and cannot be described by density alone.

The specific measure of connectivity that is most

applicable will vary according to the system and in

particular the modes of transmission of the important

parasites and pathogens. Thus, it may be sufficient to

consider contact rates for direct transmission, a product

of host feeding rates and local density for oral

transmission, number of mating partners for sexually

transmitted diseases, and so on.

As our focus is on the behavior of potential hosts, we

now explore a few, hopefully illustrative examples where

behavior may mitigate against density-dependent trans-

mission (before our explicit consideration of relative

roles of behavioral and physiological defenses). In these

situations, we propose that a measure of connectivity is

of far greater predictive power than density.

EXAMPLE SYSTEMS IN WHICH DENSITY DEPENDENCE MAY

BEST BE REPLACED BY A MEASURE OF CONNECTIVITY

The interaction of ants from different colonies can be

characterized by very low physical contact rates (with

notable exceptions such as slave-making ants). Ant

territories often have buffer zones, and the use of

pheromone trails within territories greatly reduces the

opportunity for inter-colony interaction. The adaptive

explanations offered for territoriality are often centered

around the consequences of competition (e.g., Adler and

Gordon 2003). At the same time, juvenile stages of

herbivorous microarthropods such as spider mites or

hemipterans (aphids, whitefly etc.) are frequently sessile.

This fact can easily be explained by their feeding

ecology.

In both instances, groups of individuals represent a

genetically homogeneous, high-density patch that can

serve as an ideal resource for an invading pathogen (but

see van Baalen and Beekman 2006, Wilson-Rich et al.

2009). The danger to the potential hosts in each instance

is that neighboring patches of conspecifics may serve as

sources of inoculum, so we should expect selection

against connectivity between patches. We should expect,

if disease risk acts as a selective force, that individuals

should minimize their contact rates with pathogens by

avoiding interactions with neighboring colonies, as is the

case with territorial ants such as attine leafcutters

(Jutsum 1979), or by adopting a sessile lifestyle where

possible, as appears to be the case with herbivorous

microarthropods (Elliot et al. 2002b). Thus, selection to

minimize disease risk is as valid an explanatory

hypothesis for these behavioral adaptations as any of

the more apparent and conventional explanations.

Indeed, the selective pressure to avoid infecting kin

may well be a driving force in structuring social-insect

colonies. Social-insect workers may take on increasingly

more dangerous tasks outside the colony as they age,

and avoid returning to the colony so as to avoid

infecting kin (Schmid-Hempel 1998). In these social

insects, we therefore expect behavioral interactions to be

of far greater importance in determining disease risk

than density per se.

Meanwhile, if we consider the migratory behavior of

locusts (marching as nymphs or swarming as adults), we

can find distal explanations for this behavior based upon

massive local resource depletion (Despland and Simpson

2000, Despland et al. 2000) or reduction of predators’

capacity to track groups (Reynolds et al. 2009), and

proximal explanations based upon avoidance of canni-

balistic conspecifics (Bazazi et al. 2008). In this situation,

it would seem at first sight that the risks of disease would

be tremendous. However, any infective propagules

produced from infective individuals will be left behind

as the band or swarm moves on. Meanwhile, infected

locusts will allocate time to increased thermoregulation

(behavioral fever) so as to combat infection (Elliot et al.

2002a) so will also be left behind. In this example, it is

entirely possible that the complexities of group living

lead to a lower risk of exposure to disease rather than the

reverse. In fact, in this system, we have little idea how

pathogens such as the fungus Metarhizium anisopliae

var. acridum) might actually be transmitted in the field.

We do know, however, that not keeping up can be lethal

(Simpson et al. 2006, Bazazi et al. 2008). Thus, an

alternative explanation for Wilson and colleagues’

(2002) observation of elevated immune responses in

gregaria-phase locusts, may be that this allows infected

animals to invest less time in behavioral fever and so

being left behind or eaten. The difference is subtle but

important: in this alternative explanation for the

published results, disease risk per se is not greater for

the high-density group livers; rather, the costs of disease

are greater to them.

These are just a few examples of the many behavioral

traits that can affect the costs incurred by disease. Our

suggestion is that, in many situations, these behaviors

are likely to be far more important than host density per

se. Accordingly, the best way to approach these systems

SIMON L. ELLIOT AND ADAM G. HART68 Ecology, Vol. 91, No. 1

CONCEPTS&SYNTHESIS

is to look at connectivity, that is, behavioral interactions

between individuals that might result in transmission or

otherwise affect the costs of disease. This is unlikely to

be simple, but Otterstatter and Thomson (2007) have

shown that it is feasible. Our prediction, then, goes

hand-in-hand with DDP (density-dependent prophylax-

is): greater connectivity between infective and susceptible

hosts will lead to greater costs of disease and so

investment in immune defenses. For the sake of argu-

ment, we can call this ‘‘connectivity-dependent prophy-

laxis’’ (CDP). In the gregarious locust example above, it

might be interesting to consider a fungal pathogen (e.g.,

Metarhizium anisopliae var. acridum), which must kill

the host in order to be transmissible; in such a system,

connectivity between infectives and susceptibles will be

far lower than straight density might indicate as the

locusts will move past the dying locust or else consume

it.

As with DDP, we can expect feedback in this system.

In the case of DDP, individuals that live at higher

densities may be expected to invest more in immune

defenses but may also be expected to respond by

reducing their tendency to aggregate. In the case of

CDP, we may expect individuals to reduce connectivity

so as to avoid infection, as has been proposed as a

response to predation (Reynolds et al. 2009). It may

then be that some of the variation we see in behaviors

(ant territoriality, microherbivore sessility, locust march-

ing) have arisen in part to reduce connectivity and

therefore vulnerability to disease. This would be a

pleiotropic benefit of the behavior and we could consider

the minimization of the costs of disease to have been a

selective force of a comparable magnitude, potentially

even the dominant force, leading to these behavioral

traits. Unfortunately, unpicking the evolutionary history

of highly stable behavioral traits is unlikely to be a very

fruitful exercise. An interesting product of this argu-

ment, though, is that it emphasizes the value in

considering behavior as a component, just as any more

apparent physiological mechanism, of an organism’s

suite of disease defenses. A recent example of this sort of

argument is one by Pontoppidan et al. (2009), where

ants restricting their foraging to the canopy is described

as a behavioral defense against contamination by fungal

pathogens in the understory.

THE y-AXIS: HOW TO DEFEND?

We have argued that the total investment in immune

defense may incorporate conventional physiological

parameters such as the production of hemocytes or a

plastic phenoloxidase response to invasion by a patho-

gen, but may equally incorporate behavioral prophylax-

es or (low cost) behavioral responses (Fig. 1). Whether

behavioral or physiological, resistance may be constitu-

tive or induced, perhaps reflecting the costs of mounting

the defenses (but see Hamilton et al. 2008), but we do not

consider this explicitly in this paper. We may, though,

expect physiological responses to be costly (Barnes and

Siva-Jothy 2000, Siva-Jothy et al. 2005, Wilson and

Cotter 2008), perhaps more so than behavioral responses

(Elliot et al. 2005, Schulenburg and Ewbank 2007).

Indeed, our prediction is that the relative costs of

behavioral vs. physiological defenses will decrease with

connectivity in the case of group livers in the first

instance (Fig. 1B) and more so with sociality (Fig. 1C).

For group-living, but nonsocial, hosts, we expect certain

behaviors (e.g., avoidance of conspecifics) to be cheaper

than physiological defenses, so we may see behavioral

FIG. 1. Graphical representations of density-dependent and connectivity-dependent prophylactic immune defense (‘‘immuno-competence’’ sensu Owens and Wilson [1999]), incorporating physiological and behavioral components and the absence or presenceof sociality in the host animals. (Note that a saturating function is chosen as transmission is unlikely to be a simple mass-actionprocess and to incorporate diminishing returns upon investment; other forms of the curves could as easily be chosen.) (A) Density-dependent prophylaxis. Here, disease risk is assumed to increase with host density and lead to a broadly proportional immune-defense response, as in the hypothesis of density-dependent prophylaxis (DDP). At higher densities, it may no longer pay off toinvest in this form of defense (dashed line; see An earlier, and broader, definition of ‘‘immunocompetence’’ for further explanation).(B) Connectivity-dependent prophylaxis in the absence of sociality. If the physiological and behavioral components of immunedefense have equal costs and are of equal efficacy, we may expect a similar investment in each of the two strategies. Group livingdoes, however, allow for comparatively cheap behavioral defenses. But since sociality is limited, a limitation is imposed upon theextent to which behavioral defenses can be employed, leaving physiological defenses in a prominent position. (C) Connectivity-dependent prophylaxis with sociality. For social animals, we expect behavioral immune defense to be cheaper than its physiologicalcounterpart. We also expect it to yield greater benefits at greater connectivities due to pleiotropy. Finally, we expect greaterpotential complexities of behaviors, so less limitations than in nonsocial animals. Thus, we expect behavioral defenses topredominate. (See The y-axis: how to defend? for a full explanation of these arguments.)

January 2010 69GROUP LIVING AND IMMUNE DEFENSE

CONCEPTS&SYNTHESIS

defenses assuming a proportionately greater role as

connectivity increases. It is even feasible that behaviors

such as cannibalism, while driven by protein deficit

(Simpson et al. 2006), have a pleiotropic benefit of

improving immune function and so reducing suscepti-

bility to pathogens (Lee et al. 2006, Povey et al. 2009). In

the case of social insects, our case for the paramountcy

of behavioral defenses is far greater and we have four

reasons to propose this. One reason is that there may be

greater pleiotropic benefits of behavioral than physio-

logical immune defense. Grooming can increase resis-

tance to pathogens (Yanagawa and Shimizu 2007), but

may have rewards beyond reductions in disease risk,

such as contributing to social cohesion, thereby render-

ing this form of investment in defense relatively cheap.

Another is that kin selection is likely to be a strong force

in social grouping, so making it more likely that

cooperative behaviors to reduce disease will evolve

(e.g., an individual employing behavioral fever in a

honey bee colony will benefit its sisters as well as itself;

Starks et al. 2000). The third is that, although

physiological responses can be inducible, we expect

behavioral responses to be highly labile (Hart et al. 2002,

Hughes and Cremer 2007), so the costs of an induced (or

otherwise tuned) response to disease risk can be held off

until this response is required, and even avoided entirely

once the risk is perceived to have passed. Finally, larger

social groupings will have the potential for more

sophisticated (thus potentially more effective and lower

cost) responses to disease through cooperative behaviors

(e.g., Hughes et al. 2002, Traniello et al. 2002, Cremer et

al. 2007, Cremer and Sixt 2009, Jackson and Hart 2009).

Thus, we expect that, while insects that live in groups

may well have higher global densities than their more

solitary counterparts, their living in social groups will

come hand-in-hand with greater behavioral complexity.

With social behavior, we expect to see less investment in

the physiological responses normally sought in experi-

mental studies and more investment in the behavioral

component of immune defense (see Fig. 1C).

CONCLUSION

To conclude, we have offered an adjustment to the

framework in which we consider how host density

(broadly speaking) may affect investment in defense

against disease. We think our connectivity-based frame-

work is broader than density-dependent prophylaxis and

may well explain some of the unusual empirical results.

In particular, we emphasize that overlooking behavioral

defenses may imply missing half of the story (or even

more than half ). We also propose that for social insects

in particular, the suite of behavioral defenses may be of

paramount importance. We propose that pleiotropy

may have acted upon some aspects of group living so

that selection for behaviors (or lack of behaviors) could

have occurred to reduce disease risk in addition to other,

more apparent benefits. This is not to exclude pleiotro-

pic benefits of physiological defenses, of course—

grasshoppers and locusts living at higher densities may

have a more melanized cuticle, which represents elevated

physiological defenses against pathogens and thismelanization may also have pleiotropic benefits as

aposomatic coloration (Sword et al. 2000).

We have emphasized that studies intended to elucidate

such mechanisms must integrate physiological and

behavioral mechanisms where possible. This promises

to be challenging as it is far from likely that suchcombined defenses will have additive effects. That said,

if it is possible to relate these combined defenses to

consequences in terms of disease progress and virulence,

then it should be quite possible. Wilson et al.’s (2002)study of locust immune defense considers physiological

defense (hemocyte levels and antibacterial activity) and,

although it does not quite incorporate behavioral fever

as an experimental variable, the study does allow thehosts to display this defense behavior. So, studies that

combine defenses as independent variables are quite

possible.

Finally, we have considered van Baalen’s (1998) idea

that, at high densities, defense may be a lost cause, as afurther potential explanatory force for lower investment

in defense at high densities. This argument can also be

used to predict low connectivities (territoriality, low

movement rates, avoidance behaviors) in animals living

at high densities (Fig. 1C).

There are several key areas around which we havedeliberately skirted in this paper. One is the distinction

between constitutive and induced defenses. This will

strongly affect both physiological and behavioral

defenses but it is too important an area to be given abrief treatment here. It is, though, worth pointing out

that we may initially expect behavioral defenses to tend

to be induced rather than constitutive, due to their

inherent lability. However, if physiological defenses are,as we have suggested, more costly, then it is these

defenses that we expect to be more inducible. Behavioral

defenses may then tend to be more constitutive, as seen

with grooming and division of labor in social insects.Another consideration, to which we have only alluded, is

host or host-offspring life-history change as a type of

defense. This can apply equally to solitary animals

(Minchella 1985), to group-living nonsocial animals

(Elliot et al. 2003, 2005), and to social insects (Schmid-Hempel 1998). While other areas are likely to be

important, such as the effects of a variable external

environment on defenses (Thomas and Blanford 2003,

Lee et al. 2008, Lazzaro and Little 2009), it seems thatthe interplay between diverse defense mechanisms over

both evolutionary time scales and the lifetime of an

individual or colony may initially provide the most

interesting insights (Siva-Jothy et al. 2005, Fefferman etal. 2007, Cotter et al. 2008).

ACKNOWLEDGMENTS

We are grateful to Og De Souza and Duncan Jackson forcritically reading versions of this manuscript, and to tworeferees (one of these Ken Wilson) for very constructive

SIMON L. ELLIOT AND ADAM G. HART70 Ecology, Vol. 91, No. 1

CONCEPTS&SYNTHESIS

comments. This work was partially funded by the ResearchFoundation for Minas Gerais (FAPEMIG).

LITERATURE CITED

Adler, F. R., and D. M. Gordon. 2003. Optimization, conflict,and nonoverlapping foraging ranges in ants. AmericanNaturalist 162:529–543.

Altizer, S., C. L. Nunn, P. H. Thrall, J. L. Gittleman, J.Antonovics, A. A. Cunningham, A. P. Dobson, V. Ezenwa,K. E. Jones, A. B. Pedersen, M. Poss, and J. R. C. Pulliam.2003. Social organization and parasite risk in mammals:integrating theory and empirical studies. Annual Review ofEcology, Evolution, and Systematics 34:517–547.

Anderson, R. M., and R. M. May. 1979. Population biology ofinfectious diseases. Part I. Nature 280:361–367.

Antonovics, J., Y. Iwasa, and M. P. Hassell. 1995. Ageneralized model of parasitoid, venereal, and vector-basedtransmission processes. American Naturalist 145:661–675.

Bailey, N. W., B. Gray, and M. Zuk. 2008. Does immunity varywith population density in wild populations of Mormoncrickets? Evolutionary Ecology Research 10:599–610.

Barnes, A. I., and M. T. Siva-Jothy. 2000. Density-dependentprophylaxis in the mealworm beetle Tenebrio molitor L.(Coleoptera: Tenebrionidae): cuticular melanization is anindicator of investment in immunity. Proceedings of theRoyal Society B 267:177–182.

Bazazi, S., J. Buhl, J. J. Hale, M. L. Anstey, G. A. Sword, S. J.Simpson, and I. D. Couzin. 2008. Collective motion andcannibalism in locust migratory bands. Current Biology 18:735–739.

Begon, M., S. M. Hazel, D. Baxby, K. Bown, R. Cavanagh, J.Chantrey, T. Jones, and M. Bennett. 1999. Transmissiondynamics of a zoonotic pathogen within and between wildlifehost species. Proceedings of the Royal Society B 266:1939–1945.

Cotter, S. C., R. S. Hails, J. S. Cory, and K. Wilson. 2004.Density-dependent prophylaxis and condition-dependentimmune function in Lepidopteran larvae: a multivariateapproach. Journal of Animal Ecology 73:283–293.

Cotter, S. C., J. P. Myatt, C. M. H. Benskin, and K. Wilson.2008. Selection for cuticular melanism reveals immunefunction and life-history trade-offs in Spodoptera littoralis.Journal of Evolutionary Biology 21:1744–1754.

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

Cremer, S., and M. Sixt. 2009. Analogies in the evolution ofindividual and social immunity. Philosophical Transactionsof the Royal Society B 364:129–142.

D’Amico, V., J. S. Elkinton, G. Dwyer, J. P. Burand, and J. P.Buonaccorsi. 1996. Virus transmission in gypsy moth is not asimple mass-action process. Ecology 77:201–206.

DeSouza, O., and O. Miramontes. 2004. Non-asymptotictrends in the social facilitated survival of termites (Isoptera).Sociobiology 44:527–538.

DeSouza, O., O. Miramontes, C. A. Santos, and D. L.Bernardo. 2001. Social facilitation affecting tolerance topoisoning in termites (Insecta, Isoptera). Insectes Sociaux 48:21–24.

Despland, E., M. Collett, and S. J. Simpson. 2000. Small-scaleprocesses in desert locust swarm formation: how vegetationpatterns influence gregarization. Oikos 88:652.

Despland, E., and S. J. Simpson. 2000. The role of fooddistribution and nutritional quality in behavioural phasechange in the desert locust. Animal Behaviour 59:643–652.

Dwyer, G. 1991. The roles of density, stage, and patchiness inthe transmission of an insect virus. Ecology 72:559–574.

Elliot, S. L., S. Blanford, C. M. Horton, and M. B. Thomas.2003. Fever and phenotype: transgenerational effect ofdisease on desert locust phase state. Ecology Letters 6:830–836.

Elliot, S. L., S. Blanford, and M. B. Thomas. 2002a. Host–pathogen interactions in a varying environment: temperature,behavioural fever and fitness. Proceedings of the RoyalSociety B 269:1599–1607.

Elliot, S. L., C. M. Horton, S. Blanford, and M. B. Thomas.2005. Impacts of fever on locust life-history traits: costs orbenefits? Biology Letters 1:181–184.

Elliot, S. L., J. D. Mumford, G. J. de Moraes, and M. W.Sabelis. 2002b. Age-dependent rates of infection of cassavagreen mites by a fungal pathogen in Brazil. Experimental andApplied Acarology 27:169–180.

Fefferman, N. H., J. F. A. Traniello, R. B. Rosengaus, andD. V. Calleri. 2007. Disease prevention and resistance insocial insects: modeling the survival consequences ofimmunity, hygienic behavior, and colony organization.Behavioral Ecology and Sociobiology 61:565–577.

Hamilton, R., M. Siva-Jothy, and M. Boots. 2008. Two armsare better than one: parasite variation leads to combinedinducible and constitutive innate immune responses. Pro-ceedings of the Royal Society B 275:937–945.

Hart, A. G., A. N. M. Bot, and M. J. F. Brown. 2002. Acolony-level response to disease control in a leaf-cutting ant.Naturwissenschaften 89:275–277.

Hochberg, M. E. 1991. Viruses as costs to gregarious feedingbehavior in the Lepidoptera. Oikos 61:291–296.

Hughes, D. P., and S. Cremer. 2007. Plasticity in antiparasitebehaviours and its suggested role in invasion biology. AnimalBehaviour 74:1593–1599.

Hughes, W. O. H., J. Eilenberg, and J. J. Boomsma. 2002.Trade-offs in group living: transmission and disease resis-tance in leaf-cutting ants. Proceedings of the Royal Society B269:1811–1819.

Jackson, D. E., and A. G. Hart. 2009. Does sanitation facilitatesociality? Animal Behaviour 77:E1–E5.

Jutsum, A. R. 1979. Interspecific aggression in leafcutting ants.Animal Behaviour 27:833–838.

Lazzaro, B. P., and T. J. Little. 2009. Immunity in a variableworld. Philosophical Transactions of the Royal Society B364:15–26.

Lee, K. P., J. S. Cory, K. Wilson, D. Raubenheimer, and S. J.Simpson. 2006. Flexible diet choice offsets protein costs ofpathogen resistance in a caterpillar. Proceedings of the RoyalSociety B 273:823–829.

Lee, K. P., S. J. Simpson, and K. Wilson. 2008. Dietary protein-quality influences melanization and immune function in aninsect. Functional Ecology 22:1052–1061.

Little, T. J., D. Hultmark, and A. F. Read. 2005. Invertebrateimmunity and the limits of mechanistic immunology. NatureImmunology 6:651–654.

McCallum, H., N. Barlow, and J. Hone. 2001. How shouldpathogen transmission be modelled? Trends in Ecology andEvolution 16:295–300.

Minchella, D. J. 1985. Host life-history variation in response toparasitism. Parasitology 90:205–216.

Oduor, G. I., M. W. Sabelis, R. Lingeman, G. J. de Moraes,and J. S. Yaninek. 1997. Modelling fungal (Neozygites cf.floridana) epizootics in local populations of cassava greenmites (Mononychellus tanajoa). Experimental and AppliedAcarology 21:485–506.

Otterstatter, M. C., and J. D. Thomson. 2007. Contactnetworks and transmission of an intestinal pathogen inbumble bee (Bombus impatiens) colonies. Oecologia 154:411–421.

Owens, I. P. F., and K. Wilson. 1999. Immunocompetence: aneglected life history trait or a conspicuous red herring?Trends in Ecology and evolution 14:170–172.

Pie, M. R., R. B. Rosengaus, D. V. Calleri, and J. F. A.Traniello. 2005. Density and disease resistance in group-living insects: Do eusocial species exhibit density-dependentprophylaxis? Ethology Ecology and Evolution 17:41–50.

January 2010 71GROUP LIVING AND IMMUNE DEFENSE

CONCEPTS&SYNTHESIS

Pontoppidan, M.-B., W. Hinaman, N. L. Hywel-Jones, J. J.Boomsma, and D. P. Hughes. 2009. Graveyards on the move:the spatio-temporal distribution of dead Ophiocordyceps-infected ants. PLoS One 4(3):e4835.

Povey, S., S. C. Cotter, S. J. Simpson, K. P. Lee, and K.Wilson. 2009. Can the protein costs of bacterial resistance beoffset by altered feeding behaviour? Journal of AnimalEcology 78:437–446.

Reeson, A. F., K. Wilson, J. S. Cory, P. Hankard, J. M. Weeks,D. Goulson, and R. S. Hails. 2000. Effects of phenotypicplasticity on pathogen transmission in the field in aLepidoptera–NPV system. Oecologia 124:373–380.

Reeson, A. F., K. Wilson, A. Gunn, R. S. Hails, and D.Goulson. 1998. Baculovirus resistance in the noctuidSpodoptera exempta is phenotypically plastic and respondsto population density. Proceedings of the Royal Society B265:1787–1791.

Reilly, J. R., and A. E. Hajek. 2008. Density-dependentresistance of the gypsy moth Lymantria dispar to itsnucleopolyhedrovirus, and the consequences for populationdynamics. Oecologia 154:691–701.

Reynolds, A. M., G. A. Sword, S. J. Simpson, and D. R.Reynolds. 2009. Predator percolation, insect outbreaks, andphase polyphenism. Current Biology 19:20–24.

Rosengaus, R. B., A. B. Maxmen, L. E. Coates, and J. F. A.Traniello. 1998. Disease resistance: a benefit of sociality inthe dampwood termite Zootermopsis angusticollis (Isoptera:Termopsidae). Behavioral Ecology and Sociobiology 44:125–134.

Ryder, J. 2003. Immunocompetence: an overstretched concept?Trends in Ecology and Evolution 18:319–320.

Schmid-Hempel, P. 1998. Parasites in social insects. PrincetonUniversity Press, Princeton, New Jersey, USA.

Schulenburg, H., and J. J. Ewbank. 2007. The genetics ofpathogen avoidance in Caenorhabditis elegans. MolecularMicrobiology 66:563–570.

Simpson, S. J., A. R. McCaffery, and B. F. Hagele. 1999. Abehavioural analysis of phase change in the desert locust.Biological Reviews of the Cambridge Philosophical Society74:461–480.

Simpson, S. J., G. A. Sword, P. D. Lorch, and I. D. Couzin.2006. Cannibal crickets on a forced march for protein andsalt. Proceedings of the National Academy of Sciences (USA)103:4152–4156.

Siva-Jothy, M. T., Y. Moret, and J. Rolff. 2005. Insectimmunity: an evolutionary ecology perspective. Pages 1–48in S. J. Simpson, editor. Advances in Insect Physiology.

Volume 32. Elsevier Academic Press, San Diego, California,USA.

Starks, P. T., C. A. Blackie, and T. D. Seeley. 2000. Fever inhoneybee colonies. Naturwissenschaften 87:229–231.

Sword, G. A., S. J. Simpson, O. T. M. El Hadi, and H. Wilps.2000. Density-dependent aposematism in the desert locust.Proceedings of the Royal Society B 267:63–68.

Thomas, M. B., and S. Blanford. 2003. Thermal biology ininsect-parasite interactions. Trends in Ecology and Evolution18:344–350.

Traniello, J. F. A., R. B. Rosengaus, and K. Savoie. 2002. Thedevelopment of immunity in a social insect: evidence for thegroup facilitation of disease resistance. Proceedings of theNational Academy of Sciences (USA) 99:6838–6842.

Ugelvig, L. V., and S. Cremer. 2007. Social prophylaxis: Groupinteraction promotes collective immunity in ant colonies.Current Biology 17:1967–1971.

van Baalen, M. 1998. Coevolution of recovery ability andvirulence. Proceedings of the Royal Society B 265:317–325.

van Baalen, M., and M. Beekman. 2006. The costs and benefitsof genetic heterogeneity in resistance against parasites insocial insects. American Naturalist 167:568–577.

Viney, M. E., E. M. Riley, and K. L. Buchanan. 2005. Optimalimmune responses: immunocompetence revisited. Trends inEcology and Evolution 20:665–669.

Wilson, K., and S. C. Cotter. 2008. Density-dependentprophylaxis in insects. Pages 137–176 in T. N. Ananthak-rishnan and D. W. Whitman, editors. Insects and phenotypicplasticity: mechanisms and consequences. Science Publishers,Plymouth, UK.

Wilson, K., R. Knell, M. Boots, and J. Koch-Osborne. 2003.Group living and investment in immune defence: aninterspecific analysis. Journal of Animal Ecology 72:133–143.

Wilson, K., and A. F. Reeson. 1998. Density-dependentprophylaxis: evidence from Lepidoptera–baculovirus inter-actions? Ecological Entomology 23:100–101.

Wilson, K., M. B. Thomas, S. Blanford, M. J. Doggett, S. J.Simpson, and S. L. Moore. 2002. Coping with crowds:density-dependent disease resistance in desert locusts. Pro-ceedings of the National Academy of Sciences (USA) 99:5471–5475.

Wilson-Rich, N., M. Spivak, N. H. Fefferman, and P. T.Starks. 2009. Genetic, individual, and group facilitation ofdisease resistance in insect societies. Annual Review ofEntomology 54:405–423.

Yanagawa, A., and S. Shimizu. 2007. Resistance of the termite,Coptotermes formosanus Shiraki to Metarhizium anisopliaedue to grooming. Biocontrol 52:75–85.

SIMON L. ELLIOT AND ADAM G. HART72 Ecology, Vol. 91, No. 1

CONCEPTS&SYNTHESIS