Laboratory Assay of Brood Care for Quantitative Analyses of Individual Differences in Honey Bee (Apis mellifera) Affiliative Behavior

Care of offspring is a form of affiliative behavior that is fundamental to studies of animal social behavior. Insects do not figure prominently in this topic because Drosophila melanogaster and other traditional models show little if any paternal or maternal care. However, the eusocial honey bee exhibits cooperative brood care with larvae receiving intense and continuous care from their adult sisters, but this behavior has not been well studied because a robust quantitative assay does not exist. We present a new laboratory assay that enables quantification of group or individual honey bee brood “nursing behavior” toward a queen larva. In addition to validating the assay, we used it to examine the influence of the age of the larva and the genetic background of the adult bees on nursing performance. This new assay also can be used in the future for mechanistic analyses of eusociality and comparative analyses of affilative behavior with other animals.


Introduction
The care of offspring is a fundamental component of social behavior. The most widespread forms of this behavior involve maternal and paternal behaviors that contribute to the defense and sustenance of offspring who would otherwise perish. Offspring care has been well studied from both proximate and ultimate perspectives in vertebrates, especially birds and rodents, and is one of the best understood forms of affiliative behavior [1]. However, insects do not figure prominently in this topic because Drosophila melanogaster and other traditional models show little if any paternal or maternal care and species that do perform parental care like carrion beetles [2] and earwigs [3] still lack advanced molecular and genomic resources.
Offspring care is central to systems of eusociality, but it is usually provided cooperatively by older siblings, rather than by a parent. As in vertebrate parental care, sibling care involves high frequencies of interaction and more-or-less continuous "progressive" feeding. Cooperative rearing of individual queen honey bees. This new method can be used also for comparative studies of affilative behavior in other species.

General procedures
One-day-old adult bees were obtained from apiaries maintained by the University of Illinois Bee Research Facility, Urbana, IL, June-August 2014. Frames of honeycomb containing pupae were removed from colonies headed by naturally mated queens and placed in an incubator (34°±1°C, 45%±10% RH). Newly emerged adult bees (1-18 h old) were placed in groups of 10, each bee marked with a unique color for individual recognition. Bees were kept in vertically oriented Petri dishes (100 X 20 mm) with a beeswax foundation sheet placed on the base ("wall") of the dish to mimic in-hive conditions. Dishes were supplied with one tube of honey (~1.4 ml), 30% sucrose solution (2ml) and a mixture of fresh frozen pollen and 30% sugar solution (~10 mm diameter ball). Plate diagram can be found at S1 Fig. A group of queen larvae were raised in a queenless colony via standard commercial methods (Laidlaw and Page 1997). The queen larvae were each reared in a commercial plastic queen rearing cup (JZBZ, Location, cat #: 440). Larvae placed in such a cup are often reared as queens by workers under the right social and nutritional conditions, i.e., if they are without a queen and possess sufficient nutritional reserves. In that case they build a wax "cell" with the rearing cup as a base, and add copious amounts of royal jelly to it to promote queen development [36]. These are called "queen cells". Four-day-old queen larvae were chosen for this assay because larvae at this age are fed for only one more day prior to the start of pupation, allowing for a short-term behavioral assay whose results can be evaluated robustly by queen survival.
The groups of workers in petri dishes were held in a walk-in incubator room (34°±1°C, 45% ±10% RH) for seven days. Observations took place in the incubator room where the bees were held. On day seven, one four-day-old queen larva in a queen cell was introduced to the group through a hole in the top of the petri dish. After a queen cell was introduced to the groups, detailed observations were performed on each group for 5 min and each interaction of the bees with the cell was recorded. Preliminary observations revealed two types of behavioral interactions that the adult bees had with the larvae: short visits 2-10 sec, and long visits 11-90 sec in duration. Short visits usually included entry of only the head and the thorax into the cell, while long visits were defined by entry of the whole body and contractions of the abdomen. We interpret the short visit as "cell inspection" and long visit as "nursing," i.e., the care and feeding of the larva, in concordance with previous reports [37,38] and queen larvae [16,39]. The incubator room was kept in darkness except for during observations, which were performed under typical room fluorescent lights. The petri dishes of bees were placed on observation tables for over 30 min before the beginning of the experiment and were not interrupted except for the introduction of the queen cells (one per petri dish). A video record of the assay can be found at S1 Video.
In all experiments the same queen cells were used in more than one group. The effects of temperature and humidity was assessed by conducting some assays either at 28°±1°C, 30%±5% RH or 34°±0.5 C, 45%±10% RH which is the normal microclimate inside the beehive. The behavioral response was more robust under the latter conditions (90% vs 60% of the groups showed nursing behavior, respectively; N = 20), so all subsequent experiments were conducted under those conditions. Experiments Experiment 1: Worker response to a queen cell with a larva vs. an empty cell. This experiment was performed to determine whether the bees showed a specific response to the larva. Queen cells were introduced to groups of ten bees (N = 9) from one source colony for five minutes. We recorded the total number of times bees entered the cell ("visits"), the total time spent in the cell by all bees, and the number of bees visiting the cell in each group. As a negative control, an empty queen cup that came from the same source colony was introduced into the same groups. Cells with queen larvae or empty control cells were introduced sequentially in random order, with a 10-min interval between each test. The behavior of the bees was recorded and compared using a paired t-test. The experiment was repeated with a second colony.
Experiment 2: Worker response to a queen cell with a larva vs. royal jelly. This experiment was performed to determine whether the bees were attracted to the queen cells to scavenge the royal jelly in the cell rather than to provide care to the larva. Four-day-old queen cells were removed from a queenless colony. Half of the cells (N = 34) were treated as follows: the larva was removed from the cell, the royal jelly was removed, the cell was washed using sterile water, and the larva was placed back into the washed cell. In the other half, the larvae were removed, leaving only the royal jelly. Both types were introduced sequentially in random order to groups from two source colonies (N = 21) with a 30-min interval between the tests. The same measurements as above were taken and compared using a paired t-test. In addition, the cells were left in the dishes (one per dish) overnight to determine the proportion of each type that was "capped" (bees add wax to completely surround the developing queen) and successfully maintained to produce a mature queen. The proportion of cell capping in each group was compared using Fisher's exact test for independence. Experiment 3: Effect of group size on nursing performance. This experiment quantified the success of brood care in this laboratory assay by determining the effect of group size on brood care and the minimum number of bees needed for successful growth of the larva. Queen cells with four-day-old larvae were introduced to experimental groups (from three source colonies) of three different sizes: eight bees (N = 65), one bee (N = 38) or zero bees (N = 24, an empty dish), and were left in the dish for ten days. The honey bee queen larva goes into the pre-pupa stage after it spins a cocoon at age of five days, while the workers seal the cell with wax (Laidlaw and Page 1997). Each larva was monitored to see whether it spun a cocoon and had its cell capped with beeswax by the workers, died inside the cell, or died after falling out of the cup onto the dish floor during the first 24 h period (larvae given to zero bees could not be capped, but still could spun a cocoon). Adult queen eclosion out of cells contain cocoon or capped was monitored 7-9 days later. We compared the proportion of cells contain cocoon or capped relative to the total number introduced; the proportion of queens that successfully eclosed relative to the total number of introduced cell and relative to the number of cells contain cocoon or capped; using Fisher's exact tests. Fisher's exact test with Bonferroni correction was used for pairwise comparisons. The three source colonies used to make the groups were sampled equally.
Experiment 4: Effect of larval age on nursing behavior. The effect of larval age on nursing behavior was evaluated by giving worker groups either 3-day-old or 4-day-old queen larvae reared and removed from the a queenless colony as detailed above. Queen larvae were reared as detailed above and were removed from the queenless colony when they were 3 or 4 days old. The larvae were introduced to groups of ten 7-day-old bees from single source colony (N = 10, for each experimental group) and detailed observations were made as in Experiment 1. The results were assessed using a two-tailed student t-test. Experiment 5: Effect of adult worker genetic background on nursing behavior. This experiment determined whether the assay is capable of detecting genetic differences in brood care behavior by comparing the performance of bees from different genetic sources. Genetic effects on nursing behavior have been documented in studies performed with whole colonies in observation hives [17]. We compared nursing behavior in bees derived from three different unrelated queens (N = Colony A: 65; B: 59; C: 54), each instrumentally inseminated with semen from a single, unrelated drone (male). The latency to first visit, number of nursing visits, number of inspections, total time in the cell and number of visiting bees were compared across the colonies using one-way ANOVA and LSD post-hoc tests. The proportion of groups interacting with the queen cell from each source colony was compared using Fisher's exact tests for independence.

Statistical analyses
All statistical analyses were performed using SPSS 22.0 software (IBM). Observations for Experiments 1-4 were performed solely by the first author, and observations for Experiment 5 were performed by the first author and two additional observers, each one performing observations on groups from all three source colonies. No significant differences were found between the observers (One way ANOVA P > 0.1, for all measurements; N > 120)

Experiment 1: Worker response to a queen cell with a larva vs. an empty cell
Bees made significantly more visits and spent significantly more time in the queen cells that contained larvae compared with empty queen cells (Fig 1). The number of bees that responded to a queen cell with a larva also was significantly higher than toward an empty queen cell. Results were consistent for bees from two different unrelated colonies (Fig 1). Experiment 2: Worker response to a queen cell with a larva vs. royal jelly Bees made significantly more visits and spent more time in queen cells with a larva compared with those with royal jelly (Fig 2). The number of bees that responded to cells with queen larvae also was significantly higher than to cells with royal jelly. Bees capped 9 out of 17 queen larva cells but none of the 17 cells with only royal jelly (Fisher's exact test p = 0.001). None of the capped larvae developed into mature queens, presumably because our removal of royal jelly to set the experiment up left them with an insufficient reserve that could not be replaced by the bees in the group.

Experiment 3: Effect of group size on nursing performance
Nursing performance was assessed by quantifying the proportion of cells contain a pupae in a cocoon or capped by the workers, the proportion of adult queens that eclosed /all cells, or the proportion of queens that eclosed / cells containing a pupae in a cocoon or capped by the workers, when cared for in groups of eight bees, one bee or zero bees. The cocoon / capped counts are complementary to the percentage of larvae that died during the first 24h. There were group size effects: the proportions of all three measurements were significantly different from random distributions (Fisher's exact test, p 0.01) (Fig 3). Queen larvae fared significantly better in all three measures when placed in 8-bee groups compared with dishes with zero adult bees. Eightbee groups also had higher proportions of pupae with cocoons, pupae in cells that were capped and queens eclosing compared with one-bee-groups. Surprisingly, the proportions of queens eclosing from cells containing pupae with cocoons or of queens eclosing from cells containing pupae in cells that were capped were similar for the eight-and one-bee groups (Fisher's exact test, p = 0.78). The proportion of queens that eclosed was higher in the one-bee group than the zero-bee groups. However, one-bee groups did not show increased proportions of cells containing pupae with cocoons or pupae in cells that were capped relative to the total number of cells (p = 0.19) compared with the zero-bee experimental group. This was also the case for queens that eclosed relative to the cells containing pupae with coccoons or capped cells (p = 0.12) (Fig 3). These results maybe a result of low statistical power.  The number of visits made by worker bees to a queen larva was similar regardless of larval age (Fig 4). However, visit duration was significantly longer for 4-day-old compared to 3-day-old larvae. This is consistent with the fact that 4-day-old larvae are larger and require more feeding [40]. In addition, the number of bees performing visits was higher for 4-day-old compared to 3-day-old larvae.

Experiment 5: Effect of adult worker genetic background on nursing behavior
The percentage of groups performing nursing behavior was 97% (N = 65), 90% (N = 59) and 85% (N = 54) for colonies A, B and C, respectively (Fisher's exact test, p = 0.07). In addition to the three behavioral measures used in the previous experiments, we also measured the latency to first interaction between the workers and the larva. Also due to the large sample size, we had enough statistical power to divide the visits into inspection and nursing visits and compared each one separately. We found significant effects of genetic background in all five measurements (Fig 5).

Discussion
We present a new laboratory behavioral assay to study honey bee nursing behavior. The assay avoids the problems of working with whole colonies, is conducted under normal light conditions, and causes minimal interruption to the bees. Our results demonstrate that the assay is specific for nursing behavior and allows detailed quantification and comparison across   biologically relevant groups. This assay also will be useful for cross-species analyses of affiliative behavior, since it is highly sensitive to external and internal factors. Results presented here demonstrate the specificity of the assay. Bees showed nursing behavior in over 85% of the trials, indicating high responsiveness to the queen larva. In addition, they were much more attracted to queen cells than to empty cells. This finding suggests that the visits of the worker bees are not an exploration of a new object, but rather represent a specific response to the queen larva. This conclusion is supported by findings that bees were significantly more attracted to a queen larva than to royal jelly, even when the queen larva was in a cell without royal jelly. This result also eliminates the possibility that the bees' attraction to the queen cell was based on food scavenging rather than nursing. We also found that the bees capped only cells with larvae and not cells containing only royal jelly, which represents a specific behavior in response to signals coming from the larva and not from the royal jelly.
Another indication of assay specificity was the increased nursing behavior displayed towards 4-day-old larvae. It is known that 4-day-old larvae are larger and thus require more food. Increases in visit length and the number of visiting bees suggest that the signals produced by 4-day old larva are different and more attractive to the bees than those produced by 3-dayold larvae. This is consistent with findings that workers can assess larval age via changes in the composition of brood pheromone and adjust their behavior accordingly [40].
The results reviewed above demonstrate a specific attraction of adult worker bees to the queen larvae, but are the bees effective at rearing a queen under these artificial conditions? Based on our findings, the answer is yes. A much higher proportion of adult queens that eclosed when larvae were reared by groups of bees compared with larvae placed in dishes without adult bees. This demonstrated the effects of brood care in the dish for successful development of larvae into mature queens.
Results of other experiments reported here provide insights into the variables that are important in determining the efficiency of the assay, which should be useful to other scientists contemplating its use. First, we showed that it is possible to obtain accurate and reliable measures of nursing activity with only a five-minute, focal group observation. This means the assay can be used in a highly efficient manner to screen nursing behavior in a large number of groups. Second, we developed a set of behavioral measures to precisely quantify nursing behavior and provide the means for detailed analyses of inter-individual differences. This can be very useful in future mechanistic studies. Third, we showed that larval age has an effect on the intensity of nursing behavior, as mentioned above.
The differences detected between workers derived from three different queens suggest that this assay is capable of detecting genetic variation in nursing behavior. This also is an encouraging result because genetic variation in the intensity of nursing behavior directed toward queen or worker larvae has been reported [17,39], and inferred differences in intensity of nursing behavior have also been reported for strains of bees selected to collect high and low amounts of pollen [41]. The ability to directly measure this type of behavioral variation will be useful in elucidating the genetic and molecular bases of nursing behavior.
Observations with this assay revealed two surprising findings. First, although the presence of even one adult worker is very important in ensuring that a queen larva spins its cocoon safely without falling out of its (vertically oriented) cell; some queen larvae are able to seal themselves in cells without the assistance of adult workers. Second, single workers can nurse, cap and successfully rear a queen larva (but groups are more effective). This finding provides a direct method for measuring the behavior of individual bees, which is impossible inside the colony where brood care is performed cooperatively. This opens a wide range of opportunities for studying the mechanisms of nursing behavior to better understand the intrinsic factors that control nursing behavior including pheromones, hormones, and intracellular signaling mechanisms. The assay can also be used to explore the effects of extrinsic factors, such as parasites, pathogens, and pesticides, on nursing behavior, a topic of increasing importance given the recent large-scale losses of honey bee colonies due to Colony Collapse Disorder [42].
Another possible use of this assay is to explore the evolution of cooperative brood care in honey bees. Queens are reared on royal jelly which is placed in the queen cell while the larva is still young, and the queen cell is capped before the larva has finish eating its contents. The larva will continue to feed on royal jelly in the capped cell for about 24 hours before entering the pupal stage. By contrast, honey bee worker larvae are fed progressively by nurses and capped without any food in the cell [21]. Progressive provisioning has been recorded only in some of the social bee species and not in the solitary bee species; solitary bees, as well as some of the eusocial species like the stingless bees, rear all their offspring, worker and queen, by mass provisioning in which all the food is gathered in the cell before the egg is laid. This suggests that mass provisioning is the ancestral feeding strategy in bees, while progressive feeding evolved later in some lineages with the emergence of eusociality. Honey bee queen larva provisioning is more similar to the ancestral strategy of mass provisioning than worker provisioning, suggesting that the ancestral nursing strategy is conserved in honey bee queen rearing. Our findings that an "orphaned" larva can complete the growth cycle and that single bees engage in nursing of queen larva support this speculation.
Cooperative brood care in social insects also can be important for comparative analyses of affiliative behavior in general, especially the care of young, even though the care is provided by siblings, rather than by parents as in vertebrates. This is especially the case because the two model genetic invertebrate species, Drosophila melanogaster and Caenorhabditis elegans, do not care for their young. Honey bees can thus be employed usefully in comparative analyses because they bring the typical invertebrate advantages of short generation time and economical rearing, in addition to a growing arsenal of genomic and genetic resources [43][44][45]. Comparing honey bees with other insects that show parental care will be interesting in exploring the evolution of various forms of insect sociality. Comparing to mammals also will be interesting in terms of broad evolutionary analysis because there are intriguing similarities between honey bee "allomaternal" and mammalian maternal care. Both are progressive and occur across the circadian cycle [46][47][48], both involve glandular secretions (royal jelly and milk) rather than food collected directly from the environment, and both are performed by individuals that are not in a reproductive state (for mammals only during the period devoted to nursing; worker honey bees are almost completely non-reproductive during their entire life). The new nursing behavior assay we present here should increase the utility of honey bees in exploring the molecular foundations of this important affiliative behavior.