Increased response to sequential infections of honeybee, Apis mellifera scutellata, colonies by socially parasitic Cape honeybee, A. m. capensis, workers

Cape honeybee, Apis mellifera capensis, workers can be social parasites and host colonies can defend themselves by rejection of such workers. Using the pseudo-clonal obligate parasitic lineage of A. m. capensis and wild-type A. m. capensis workers, which are facultative parasites, we show that host colonies significantly increase their defence behaviour towards social parasites upon secondary exposure. Most obligate and facultative social parasites were rejected before they could even produce significant amounts of the queen-like mandibular gland pheromone secretion or activate their ovaries. This suggests that other signals, like cuticular hydrocarbons, could be used by host colonies to identify potential invaders. Honeybee colonies seem to be able to utilise these potential cues, learn from their initial exposure to hive intruders and enable them to improve their defensive behaviour during subsequent infestations, resulting in increased removal rates of parasites.


Results
The survival analysis revealed that during both infections the parasitic pseudo-clones were identified and removed significantly faster than A. m. scutellata control ( Fig. 1; pseudo-clone vs. scutellata control: Infestation 1 Test statistics (TS) = 4.74 p < 0.00001; Infestation 2 TS = 7.58 p < 0.00001). Only during the second infestation were the parasitic pseudo-clones identified and removed significantly faster than A. m. capensis non-clonal social parasites ( Fig. 1; pseudo-clone vs. A. m. capensis non-clonal: Infestation 1 TS = 2.23 n.s.; Infestation 2 TS = 2.7 p < 0.0068). Furthermore, A. m. scutellata control workers were removed significantly more slowly by the host colonies than A. m. capensis non-clonal social parasites (scutellata vs. non-clonal parasites: TS > 2.32, p < 0.02 for infection 1 and 2). For each individual test group, individuals of the same group were removed at a faster rate from the test colonies during the second encounter as compared to the first infections (TS > 8.45, p < 0.00001, for all three pair wise comparisons). By comparing the increase in the average daily mortality from the first to the second round of infections it was revealed that the pseudo-clonal group increased from an average mortality of 19.7% to 39.4% (z = 4.31, p < 0.0001) during the second infestation, the A. m. scutellata control group increased from 13.8% average daily mortality to 21.3% (z = 1.66, n.s.) and the non-clonal parasitic group from 16.8% to 31.4% (z = 3.38, p < 0.001).

Discussion
The data clearly show that both, the socially parasitic pseudo-clonal and the non-clonal social parasites A. m. capensis workers were rejected significantly faster upon secondary exposure.
Pheromonal competition governs reproductive dominance hierarchies among A. m. capensis workers 18,19 . The tested clonal parasitic worker lineage has probably been selected for rapid queen-like mandibular gland secretions 20,21 as well as for its resistance against host queen pheromone signals 19,22 . The observed identification and rejection of socially parasitic workers during the first infection by the host workers is therefore most likely related to the faster reproductive development of the pseudo-clonal lineage 23 , similar to aggressive behaviour shown by non-developed workers towards workers with ongoing reproductive development in other subspecies 24 . Furthermore, internal anatomical differences like spermatheca size and numbers of ovarioles 25 could result in differential volatile profiles. Pheromones could be another set of factors, in particular the pheromonal predisposition of A. m. capensis workers seen even under queenright conditions 26 and the fact that the gland secretion of the parasitic lineage is significantly different to subordinate workers 27 . These differences in the gland and cuticular profile would allow the host workers to learn parasite specific cues, which allows them to identify and remove them in a subsequent infestation 26,27 .
Furthermore, the results show that during the second encounter the pseudo-clonal parasitic workers are rejected at a much faster rate by the A. m. scutellata host workers than all other tested groups. That pseudo-clonal parasitic workers are rejected faster could either be because of the genetically homogeneity of the lineage, which could make it is easier to learn its specific cues for the host or it is indicating a specifically directed behavioural response. More than 90% of the pseudo-clonal parasitic workers were rejected in less than 24 hours (Fig. 1) and the faster removal of the non-clonal social parasites suggests that it is rather a specifically directed behavioural response. This would mean that the faster the behavioural reaction, the stronger the resulting protection of the host colony as A. m. capensis workers need on average 6.5 days to become reproductively active 23 . Indeed, the callow parasitic workers did not remain long enough in the host colonies to activate their ovaries 23 , but signals from the tergal gland and cuticle could be used to discriminate between hosts and parasites 19,26 . Therefore, the parasitic workers could not establish themselves as reproductive parasites and can be excluded before they are able to gain reproductive dominance 9,26 and before they are able to overthrow colony defences. In any case, the observed fast behavioural reaction of the host workers upon secondary exposure provides strong support to earlier findings for www.nature.com/scientificreports www.nature.com/scientificreports/ the ejection of small hive beetle parasites by honeybee colonies 16 . Taken these observations together, it appears as if honeybee colonies may be able to learn from their initial exposure to hive intruders and are able to improve their defensive behaviour, resulting in increased removal rates of parasites. The observation that the host colonies are able to evict social parasitic workers more swiftly during the second encounter could either suggest that the host colonies get conditioned or that honeybees have evolved a potential social immune memory to fight conspecific parasites. Both are not mutually exclusive and would result in the same advantageous behaviour of repelling potential social parasites.

Material and Methods
study animals and experimental set up. Queenright, unrelated A. m. scutellata colonies (N = 4) were obtained from its endemic range (Pretoria, Gauteng Province, South Africa). All of them were set up in threeframe observation hives (∼3,000 bees). The middle frame in each hive contained brood and the top and bottom frames honey and pollen. Colonies were fed ad libitum with sugar water (1:1) and artificial beebread (honey/ icing sugar/soya flour 1:2:2). On the same day of establishing the observation hives, frames with sealed worker brood were placed in an incubator until adult emergence 28 . Twenty-four hours later, we introduced three different groups of freshly emerged individually labelled test workers (N = 50 each group, <24 h old) into the four queenright test colonies: The introduction of all 150 workers took less than five minutes and freshly emerged workers were used because they are normally readily accepted 31 as the colony odour in honeybees is strongly affected by environmental cues 32 . The experiments were conducted in 2004, all workers were labelled individually with Opalithplättchen (numbered plastic tags) and observation hives were screened for the presence of the introduced workers twice a day (13:00 day light and 18:00 red light conditions) following previously used methods 33 . procedure. All observation hives were infected at the same time and screened twice daily for the presence of labelled workers for 10 days and after the period, all test workers were carefully removed. The infections were repeated with new test workers 36 h thereafter using the same colonies and protocols. Data analyses. The data were analysed using Cox-Mantel survival analysis. Pair-wise comparisons were performed between the different groups and the Bonferroni adjustment was applied. The data are displayed as the cumulative proportion surviving. For the Cox-Mantel analysis both the test statistic (TS) and p values are given. To evaluate if any of the groups received a stronger behavioural rejection during the second round of infection we calculated the overall average daily mortality and statistically compared the proportions. All statistical analyses were performed using Statistica © .