Colony pace: a life-history trait affecting social insect epidemiology

Abstract

Among colonies of social insects, the worker turnover rate (colony ‘pace’) typically shows considerable variation. This has epidemiological consequences for parasites, because in ‘fast-paced’ colonies, with short-lived workers, the time of parasite residence in a given host will be reduced, and further transmission may thus get less likely. Here, we test this idea and ask whether pace is a life-history strategy against infectious parasites. We infected bumblebees (Bombus terrestris) with the infectious gut parasite Crithidia bombi, and experimentally manipulated birth and death rates to mimic slow and fast pace. We found that fewer workers and, importantly, fewer last-generation workers that are responsible for rearing sexuals were infected in colonies with faster pace. This translates into increased fitness in fast-paced colonies, as daughter queens exposed to fewer infected workers in the nest are less likely to become infected themselves, and have a higher chance of founding their own colonies in the next year. High worker turnover rate can thus act as a strategy of defence against a spreading infection in social insect colonies.

1. Introduction

The evolutionary transition to a social lifestyle is considered a major reason for the spectacular ecological success of social insects [1,2]. At the same time, social living has several disadvantages. Infectious parasites, for example, may spread more easily when individuals live together at high densities, are closely related and interact frequently with one another—conditions that are characteristic for social insect colonies [3–7]. Ants and bees, however, have fewer immune genes than other insect species [8,9]. Even though this seems to be an ancestral condition in bees [8], social insects may also have additional ways to defend themselves against diseases and the transmission of pathogens, sometimes summarized as ‘social immunity’ [10]. Among these non-immunological defences are those that are based on altruistic behaviour (e.g. mutual grooming to remove parasite spores), collective decision-making (e.g. increase in nest temperature) and other collective actions that reduce parasite loads or the impact of infection (reviewed in [10]).

One aspect that has been less investigated so far is that social insects may also mitigate fitness losses caused by parasitic infection by adjusting life-history parameters such as growth, reproduction and survival [11] (see also review in [12]). Life-history theory suggests that individuals should alter the timing and investment in these traits to maximize their lifetime fitness in a given environment [13]. The presence or absence of parasites arguably is a crucial environmental factor affecting an organism's optimal pattern of resource allocation [14,15]. A life-history strategy that accelerates the host's schedule of reproduction, for instance, should be favoured when parasites shorten the remaining hosts lifespan and curtail its expected future reproductive capacity [11,12]. Evidence for a life-history response to parasitism maintained throughout the evolution of sociality, however, remains scarce (but see the classic studies in snails infected by trematodes [16]).

Here, we investigate a life-history parameter that is unique to the biology of social insects—the turnover of individuals within a colony, as determined by worker lifespan and the birth rate of new workers. This turnover defines a colony ‘pace’. In many social insects, only the queen reproduces, while the worker caste is sterile; in other species, workers also vie for some reproduction [17]. However, regardless of the level of within-colony conflict over reproduction, workers essentially are a dispensable resource for the colony's success. Thus, worker turnover is similar to the turnover of cells in a body, and an individual's life takes a different meaning from that in most other organisms. The reproductive success of an insect colony depends on the effort of its workforce—almost regardless of the individual—and hence the total number of workers present at any one time.

We studied this question in the bumblebee, Bombus terrestris L. In bumblebees, for example, daughter queens—the most valuable reproductive unit—are produced only if a critical worker number is attained [18]. Bombus terrestris colonies, moreover, show pronounced variation in their colony-specific rate of worker production and worker death. This has been observed both in the field (see electronic supplementary material) and under standardized laboratory conditions [19]. Some colonies may thus have higher worker turnover rates than others. In fast-paced colonies workers die younger, but new workers are produced more frequently than in slow-paced colonies. For an infectious parasite, an increased production of short-lived workers should have epidemiological consequences, as the expected time of residence in a given host will be reduced. In analogy to apoptosis (programmed cell death) on the within-individual level, where the death of individual cells can be an effective anti-viral defence [20]. Short host lifespan should select for rapid replication in parasites, which typically is associated with more frequent transmission events and/or higher virulence; in this context, fast growth means more infectious cells within a host and thus higher virulence according to the standard theory.

To test this idea, we manipulated worker turnover rate in experimental colonies of B. terrestris to measure its effect on the spread of the infectious parasite, Crithidia bombi. Moreover, we predict that a decrease in residence time of infection and more frequent transmissions reduce the genotypic diversity of the parasite by eliminating slowly replicating and non-suitable strains (see [21]). A reduction in parasite diversity at the time of colony reproduction is important, because high strain diversity increases the likelihood of daughter queens becoming infected [22]. Eventually, at the end of a colony cycle, fast-paced colonies may thus have fewer infected workers with fewer persisting strains, which may result in fewer daughter queens being infected and a higher colony founding success in the next generation.

Hence, can a high worker turnover rate be a life-history strategy that reduces parasite spread? In the experiment presented here, we address the following questions relevant for social insect epidemiology. (i) Does a higher worker turnover rate result in fewer infected workers over time? (ii) Does a higher worker turnover rate result in fewer last cohort workers that are responsible for rearing sexuals? (iii) Does a higher worker turnover reduce the number of persisting C. bombi strains in the colony? (iv) Are different C. bombi strains infecting colonies with high versus low worker turnover rates? (v) Does a higher worker turnover rate increase the number of infective cells in the colony?

2. Methods

(a). Parasites and insects

The bumblebee (B. terrestris) is a eusocial insect with annual life cycle. In spring, the singly mated queens, after emerging from hibernation, found colonies that grow in numbers of workers until the colony eventually reproduces in late summer; only the mated daughter queens go into hibernation to start next colonies next year. The trypanosomatid gut parasite, C. bombi, is common in most populations [23–25]. It cannot live outside the host [26] and hence experiences strong selective pressure to infect daughter queens—the only individuals that survive the winter. However, at least in our study area, at most 10–15% of spring queens are infected with C. bombi. Colonies contract the infection either from other colonies via shared use of flowers (between-colony transmission) or via faeces deposited in the nest (within-colony transmission) [27]. The parasite reproduces both clonally and sexually and is genetically highly diverse, and close to half of infections are caused by more than one strain [24,28]. The time between established infection and transmission ranges between approximately 2 days for fast developing genotypes and 10 days for slower strains [29]. Crithidia bombi increases worker mortality under stressful conditions and substantially reduces the colony founding success of daughter queens [30].

(b). Collection and culturing

Bombus terrestris queens emerging from hibernation were collected in spring 2010 from two populations in Switzerland (near Aesch BL and Neunforn TG) and allowed to initiate colonies in the laboratory. As soon as the first workers emerged, colonies were transferred to circular perlite nests [31]. Seventeen uninfected colonies were reared to reach a sufficient number of workers to serve as donor colonies for bees used in this experiment. All bees were kept under standardized laboratory conditions (28 ± 2°C, 60% RH) with constant red light illumination and pollen, and sugar water provided ad libitum.

The parasite C. bombi was collected from faeces of naturally infected queens, sampled from the same two populations in spring 2010. To be able to control the infections to single strains, we isolated single infective C. bombi cells with a fluorescence-activated cell sorter and maintained them clonally in liquid medium [32]. The six strains used in this experiment had distinct multi-locus genotypes at five polymorphic microsatellite loci and could therefore be easily distinguished by genetic markers.

(c). Infection and experimental manipulation of the worker turnover rate

To test the effect of worker turnover rate on the spread of C. bombi (prevalence, strain number, strain identity and infection intensity), we used B. terrestris workers from 17 donor colonies. From each donor colony, we created two experimental groups with 10 workers each at all times, which we exposed to either a high or low worker turnover rate, respectively, for 30 days (n = 34 experimental groups). A sketch of the experimental design is given in figure 1.

Figure 1.

Experimental design of the study. We induced different worker turnover rates in initial groups of six infected (black) and four naive (grey) workers. (a) To create a high worker turnover rate, two random workers from the group were exchanged against two naive worker (circular arrows) every 3 days. (b) To create the low worker turnover rate, workers were exchanged every 6 days. The prevalence of infected workers in the groups was measured at days 14, 30 and 38 (arrows). Six naive workers, mimicking the last-generation workers were added to the respective groups at day 30. (Online version in colour.)

The experiment started by placing four naive (uninfected) workers, and six workers each infected with a distinct C. bombi strain (i.e. six strains in total per experimental group) into a box. To infect these ‘source’ workers with a single parasite strain, and to mimic a natural colony background, four non-age-controlled workers were selected at random from each donor colony, starved for 2 h and presented with 10 µl inoculum containing 20 000 cells of out of the six single C. bombi clones in medium and sugar water (1 : 1). Workers that did not ingest the cocktail after 60 min were excluded from the experiment to ensure primary infections and viability of the infectious cells. Post-exposure, bees were kept individually to prevent cross-infection. Eight days later, their faeces were visually checked for the presence of C. bombi cells, confirming their infection status. We exposed a total of 408 workers (24 workers from each of the 17 donor colonies) to ensure successful infection of six pairs of workers per donor colony and parasite strain. One worker from each infected pair was then assigned to each experimental group (low and high turnover rate treatment), and each experimental group was completed with four naive workers from the respective donor colony to yield the total of 10 workers per experimental group (figure 1). Four donor colonies were completely resistant to either one or two parasite strains; the respective experimental groups (both low and high turnover rate) hence started the experiment with an infection proportion of 0.5 or 0.4, instead of 0.6.

To create a high turnover rate, every third day we randomly exchanged two workers from the experimental group against two randomly chosen naive workers from the same donor colony (figure 1). For the low turnover rate, two randomly chosen naive workers from the experimental group were exchanged only every sixth day. This experimental protocol mimicked the ‘birth’ and ‘death’ rate in a B. terrestris colony in the sense that workers are ‘reared’ at different rates, get exposed to infections or are removed from the pool of infection at an arbitrary time owing to uncorrelated, external events. The time intervals were chosen based on the natural variation in parasite latency after infection (2–10 days for C. bombi [29]) and the observed difference in worker lifespan among colonies of B. terrestris. The mean inherent lifespan of workers from a colony with long-lived workers was about twice the lifespan of workers from a colony with short-lived workers [19]. We exchanged workers from experimental groups following this treatment schedule for a total of 30 days and sampled all workers on day 38. This schedule allowed workers to live up to the expected four to six weeks under natural conditions [33], whereas it minimized the number of dead bees that we replaced against naive workers to keep group size constant. The proportion of infected workers (parasite prevalence) over time was monitored by visually checking the faeces of all 10 workers in the treatment groups after 14, 30 and 38 days.

On day 30, an additional six naive workers from the respective donor colonies were added to each treatment group (n = 30 groups; 15 high and 15 low turnover rate). These individually marked workers were introduced to mimic the last generation of workers reared in the colony (last cohort workers) that raises sexuals (males and daughter queens). At the same time, this last cohort of workers, and the infectious cells shed with their faeces, represents the parasite environment to which daughter queens are exposed before they leave the nest. We interpret this group as representing young daughter queens, as they represent the same genotypes, but not the same physiology, as their sister workers. Last cohort workers were exposed to the treatment groups for 8 days, which is about the time young queens spend in the nest before they leave for the mating flight (see [34]). All workers were sampled on day 38, and their gut tissue was genotyped to determine infection status (parasite prevalence) and the number and identity of C. bombi strains that successfully infected bees exposed to different worker turnover rates. For groups where all 10 workers provided faeces samples into a glass tube (n = 20 groups; 10 high and 10 low turnover rate), we additionally measured the concentration of infective cells (infection intensity) of all faeces samples pooled per treatment group with a haemocytometer (Neubauer improved counting chamber).

(d). Genotyping of infections

We dissected out the whole gut of all workers and extracted genomic DNA with a Qiagen DNAeasy blood and tissue kit (Qiagen, Hilden, Germany) following the protocol for purification of total DNA from animal tissues. Crithidia bombi infection was detected by the amplification and visualization (1.5% agarose gel) of a part of the 18S rRNA gene with Crithidia-specific primers (CB-Cytb2-F and CB-Cytb2-B, described in [35]). The C. bombi strains present in successful infections were thereafter genotyped at the C. bombi microsatellite markers Cri4, Cri2F10, Cri4G9, Cri16 and Cri1B6 [28] following the protocol in [36], and the amplification product was run on a MegaBACE sequencer (GE Healthcare, Glattbrugg, Switzerland). The microsatellite peaks were scored twice with MegaBACE Fragment Profiler software v. 1.2 by an observer blind to the treatment groups.

3. Analysis and statistics

Determinants for the number of infected workers over time were analysed using a generalized linear-mixed model (GLMM) with a binomial error structure and logit-link function. Because workers did not provide a faeces sample every time, the infection status was assessed; we treated the number of infected workers as dependent and the total number of workers per sample as independent variable [37]. In the full model, we included the experimental groups (high versus low turnover rate) as a fixed factor, and the three time points at which we assessed the number of infected workers per group (14, 30 and 38 days) as a covariate, whereas the donor colony of each experimental group was nested in experimental groups, and the individual groups were used as a random factor. Model selection was done by backwards elimination of non-significant terms [38]. If interaction terms are not presented, then they did not have a significant effect. For every time point measured, we conducted post hoc comparisons based on paired sampled t-tests on the model residuals and sequential Bonferroni correction [39] to analyse differences in the number of infected workers between groups with a high and low worker turnover rate. We used an analogous model to test the influence of a high worker turnover rate on the prevalence of infected last cohort workers. The model included the experimental groups (high versus low turnover rate) as fixed factor and the donor colony of the experimental groups as single random factor, because the infection status of the last cohort workers was assessed only at day 38. Identical model parameters were applied to test the number of persisting C. bombi strains in high versus low turnover groups, except that the number of strains after 38 days were used as dependent and the number of strains at treatment start as independent variables, because not all experimental groups started the experiment with six infected workers and thus six C. bombi strains. The distribution of the C. bombi strains that successfully infected workers exposed to a high versus low turnover rate was analysed using Fisher's exact test, and the overall infection intensity in high versus low turnover groups was tested with a Wilcoxon signed-rank test for paired samples. All statistical analyses were carried out in IBM SPSS Statistics v. 19.0 or R statistical software v. 2.15.2 [40]. Model assumptions were met for all linear models presented.

4. Results

A high turnover rate in a group of bumblebee workers significantly reduced the spread of the infectious parasite C. bombi, resulting in an average infection prevalence of Prev = 0.45 after 38 days for workers exposed to a high turnover rate, and Prev = 0.6 for workers exposed to a low turnover rate (GLMM; time: F = 0.147, p = 0.702; turnover treatment: F_1,99 = 6.941, p = 0.010; figure 2). This difference was already apparent at the first measure (figure 2) and did not change over time (GLMM; time × turnover treatment: F = 0.875, p = 0.352); the high turnover rate reduced the proportion of infected workers at all times measured (paired t-tests; 14 days: t₁₆ = −2.234, p = 0.040; 30 days: t₁₆ = −3.562, p = 0.003, 38 days: t₁₆ = −2.569, p = 0.021; figure 2).

Figure 2.

The proportion of infected workers in groups of B. terrestris workers with high and low turnover rate (n = 34), measured 14, 30 and 38 days post-infection (at time ‘0’). Shown are the estimated marginal means (±s.e.) from a model that used turnover rate as fixed factor, time as covariate and the donor colony of each experimental group nested in experimental groups as well as the individual groups as random factor (see text).

In 14 out of 17 paired experimental groups (high versus low turnover rate from the same donor colony), fewer workers were infected when exposed to a high turnover rate. In two of the remaining pairs, there were fewer infected workers in the group with the low turnover rate, and in one pair, workers were equally infected, regardless of the turnover rate they were exposed to. The infection was lost after 30 and 38 days in three groups exposed to a high turnover rate, whereas none of the experimental groups exposed to a low turnover rate lost the infection. In two groups with a low turnover rate, all workers were infected at day 38.

Importantly, a higher worker turnover rate also resulted in fewer last cohort workers contracting the infection (GLMM: F_1,28 = 6.136, p = 0.02; figure 3), and this difference in infection prevalence between groups with different worker turnover rates was unlikely to be due to different numbers of C. bombi strains infecting either treatment (GLMM: F_1,28 = 1.258, p > 0.1; figure 4)). We further did not find that different strains infected workers with a high turnover rate when compared with workers with a low turnover rate (Fisher's exact test: n = 12, p > 0.1; figure 4). Finally, the infection intensity (faeces samples pooled per group) in groups with a high turnover rate did not differ from the infection intensity of groups exposed to a low turnover rate (Wilcoxon signed-rank test: Z = −1.274, n = 20, p > 0.1).

Figure 3.

The proportion of infected B. terrestris last cohort workers after they were exposed for 8 days to worker groups with a high and low turnover rate (GLMM: F_1,28 = 6.136, n = 30, p = 0.02). Numbers in parentheses indicate the number of experimental groups measured. Bars represent the estimated marginal means (±s.e.) derived from a model with turnover rate as fixed factor, and colony as random factor (see text).

Figure 4.

Percentage of different C. bombi strains (shades of grey, letters A–F) present in groups of B. terrestris workers with a high and a low worker turnover rate, 38 days post-experimental infection. Numbers in parentheses indicate the number of experimental groups measured.

5. Discussion

In this study, we aimed to understand the consequences of different worker turnover rates for a spreading disease. We found that fewer workers and, more importantly, fewer workers that mimicked the last cohort of workers and are responsible for rearing sexuals were infected in groups with a higher worker turnover rate. Contrary to expectations, we did not find that, at the end of the experiment, fewer parasite strains infected workers exposed to a high turnover rate, nor did we find that different strains infected bumblebee workers exposed to different turnover rates. A high turnover rate did also not increase the number of C. bombi cells circulating in the colony.

In fact, the main virulence effect of C. bombi seems to be in the founding queens. C. bombi infection greatly reduces the colony founding success in daughter queens of B. terrestris after they emerge from hibernation the following spring [30]. A smaller number of infected last-generation workers, which are rearing sexuals, should result in fewer infected daughter queens per colony. Those less infected new queens should then have a higher chance of founding their own colonies. The number of C. bombi strains circulating in the colony is an important factor predicting the infection rate of daughter queens [22]. However, because a high turnover rate in our experiment did not reduce the number of circulating strains, nor the number of infective cells in the colony, the reduced number of infected last cohort workers must be due to the low number of workers with a short lifespan and thus reduced transmission period that transmit the disease within the nest with a high turnover rate. Our study thus suggests that it is the probability of encountering a parasite that is more important for its transmission within the colony than the dosage of infectious cells.

In natural colonies, a high worker turnover rate would inevitably coincide with a demographic shift towards younger workers. This may associate with increased defence levels of short-lived, ‘fast-paced’ workers as the immune system of bumblebees declines with age [41,42]. Workers sampled in the field that originate from colonies with a high turnover rate might therefore be on average younger and less infected, with a lower number of infective cells circulating in their colony. On the other hand, ‘fast-paced’ colonies with an increased death rate of workers must produce more workers over the colony cycle to achieve similar sizes when compared with colonies with a low turnover rate. For given resources, these more numerous workers reared in ‘fast-paced’ colonies might be of lower quality (e.g. low immune capacity) compared with workers exposed to lower turnover rates. On the within-individual level, associations between resource allocation into immune defence and other to important life-history characteristics are often described [43–46]. This can be explained by immunity being costly to develop and maintain [47,48], and optimal defence thus being a balance of these costs with the risk of infection. During a long lifespan, the risk of encountering and contracting a disease may be high, and long-lived hosts should be better protected (this idea is reviewed in [49]). Alternatively, short-lived hosts may trade investment into early reproduction against immunity and body maintenance—they live fast and die young [13]. For annual social insects with a narrow reproduction window at the end of the colony cycle, adopting a strategy of short lifespan and early reproduction, however, should have fitness consequences (e.g. males die before they find a mate). In this context, the choice may be to invest in either many poorly protected and sort-lived worker or few long-lived and well-defended workers to maximize colony size and thus reproductive output.

If, in bumblebees, low quality were reflected in reduced immunity, ‘fast-paced’ colonies could also be more rather than less (as affected by average age) susceptible to C. bombi infection. No evidence of this is as yet available in our system. However, a comparable pattern of immune investment can be observed within species and among castes of social insect. Short-lived male ants, for instance, have a lower immune response than the longer-lived workers and queens [50].

While shifting the timing of reproduction requires collective decision-making in infected colonies [11], worker turnover rate, or at least the egg laying rate in social insect colonies, depends on the queen. The role of the queen as the pacemaker of the colony, as an aside, is an old idea going back to Schneirla's studies on army ants [51] and was taken up repeatedly over recent decades, mainly focusing on the queen as the central force that initiates activity and coordinates work [52–55]. Whether the observed increase in worker turnover rate among certain colonies is the result of how the queen allocates her resources and might be a response to parasitism, or simply reflects the queen's high or low quality, remains to be seen. If this were the case, then we would expect that a queen produces either more short-lived workers at the expense of a reduced immune system, or few long-lived workers that have a good immune system to counter their prolonged period of transmission.

We conclude that a high worker turnover rate could be a life-history strategy that protects social insect colonies colony against the spread and transmission of an infectious parasite. This increase in the worker turnover rate may be a direct response to parasitic infections, or may indirectly be caused by factors that decrease worker longevity such as food shortage or queen quality. We also suspect that parasite-imposed selection for short-lived workers contributes towards worker lifespan variation in social insects.

Data accessibility

Genotypes of C. bombi strains and data on parasite prevalence and infection intensity will be deposited in Dryad (http://dx.doi.org/10.5061/dryad.c1483).

Authors' contributions

S.D.B. and P.S.-H. designed the study and wrote the manuscript. S.D.B. ran the experiment and analysed the data.

Competing interests

We have no competing interests.

Funding

P.S-H.: SNF (grant no. 31003AB_131076) and S.D.B.: SNF (grant no. 31003AB_131076).