Final Report Summary - BEESPATNET (Mapping Spatial Interaction Networks in Honeybee Colonies)
1. Introduction
Social insects – the ants, bees, wasps and termites – have achieved ecological dominance in many terrestrial ecosystems, indeed, their total biomass equals or exceeds that of human beings. A leading explanation for this runaway success is that social insects have evolved highly effective communication mechanisms that allow a colony of many tens of thousands of individuals to coordinate their activities such that the actions and behaviours of each individuals comes to be highly dependent upon those of their nestmates. This coupling allows the group to produce adaptive behaviours, such as collective foraging and defence, which resemble the behaviour of a single, unitary organism. The social insects therefore provide a unique system for the studying the principles that underly the organisation of complex animal societies.
Social insects have a huge economic influence upon human affairs, providing valuable ecosystem services such as soil turnover, nutrient cycling, and particularly in the case of the honeybee Apis mellifera, pollination of roughly 1/3 of the fruits and vegetables consumed by humans.
However, over the last 10-15 years, honeybee populations in North America, and to a lesser extent in Europe, have suffered severe overwintering losses. The typical symptom is that when the hives are opened in spring after the winter hibernation, all the bees have died or disappeared – such losses have been termed Colony Collapse Disorder (CCD). There seems to be no single cause, rather colony collapse is triggered by a complex mixture, including habitat fragmentation & loss of pollen-diversity, sub-lethal effects of industrial agrochemicals, viral infection via ectoparasitic mite vectors.
The original project proposal was to use a novel automated tracking system (originally developed for tracking ants), to simultaneously track the movements of hundreds of individual workers as they move, work and interact with other individuals within the hive. From these tracks, we proposed to to extract the network of worker-to-worker contacts. These contact networks networks serve as the substrate for the transmission of harmful pathogens; for example, the one fungal (microsporidian) pathogen of honeybees, Nosema apis, is horizontally transferred between workers during a praticular type of contact involving liquid food exchange (trophallaxis).
Since its inception, in 2010, the project aims have shifted from questions directly concerning disease transmission and epidemiology, to questions concerning colony reproduction and so-called 'queen-signalling'. However, this shift still has a direct connection to the problem of overwintering losses, as described below.
Honeybee colonies reproduce by colony fission, in which the old queen along with roughly half the worker population split off from the original colony (in a swarm), to found a new colony, leaving behind a young queen and half the original workforce. This process naturally occurs whenever the worker population reaches a certain critical density, so over the course of a year a colony can produce many swarms. Beekeepers strive to prevent swarming, as the beekeeper loses half of the bees (swarming reduces the honey-producing capacity, and hence the profitability, of the remaining colony), and more importantly, because the workforce of a colony that has 'cast-off' several swarms is much severely reduced, swarming can increase the likelihood that the colony does not survive the winter.
Swarming is thought to be controlled by the ability of the queen to signal her presence; queens produce queen mandibular pheromone (QMP), which inhibits the workers from both raising a new queen, and laying their own unfertilized eggs. QMP is produced by the queen, and disseminated through the colony by three routes; (i) by direct contact between the queen and her attendant workers, (ii) by indirect transfer from worker to worker during physical contacts, and (iii) indirectly via deposition onto the wax substrate. It is thought the dissemination of this signal is inhibited when the worker population reaches a critical density (e.g. through physical crowding effects), and that the blocking of this signal releases the workers to raise a new queen, and hence to begin path towards swarming. Our focus has thus been to use the automatic tracking system to study the pathways by which the queen signal is disseminated, specifically:
1. Whether the movement of the queen maximises the indirect dissemination of QMP, via the wax.
2. Whether the movement of the queen maximises the rate of direct contacts with the workers.
3. Whether there movement of workers that have recently encountered the queen also serves to maximise their encounter rate with `uninformed' workers.
2. Progress so far
Objectives 1-3 require the generation of time-stamped individual trajectories, whereas objectives 2-3 additionally require the extraction of contact networks in which workers are represented by nodes, and physical interactions between bees are represented as links that connect the nodes. Furhter, as there exists strong between-colony variation, objectives 1-3 require several independent replicates.
For a single replicate we attached took newly-hatched 1000 bees (and the queen), and attached a unique barcode-like tag to the thorax of each bee. These bees were then released back into the observation colony. The simultaneous movements of these tagged 1000 bees were continuously recorded over a 31-day period. From these recordings, we extracted 31 daily time-stamped trajectories (for objectives 1-3), and daily contact-networks (for objectives 2-3). Note, the 31-day period is roughly the same length of time as the typical worker lifespan, so the trajectories and interactions additionally capture the entire ontogeny of individual spatio-temporal mobility (from the trajectories), and social position (from the contact-based interaction networks).
Progress towards objectives 1:
The initial hurdle towards all objectives was the technical challenges associated with adapting the automated ant tracking system for application to honeybees. Designing, and assembling these systems took roughly 1 year. However, we now six fully-functional honeybee tracking systems. Data collection for the first 5 replicates was completed by the end of September 2014, and analysis has been ongoing since.
As objective 1 concerns the behaviour of a single individual in each colony – the queen – rather than the remaining 999 workers, the analysis has first focused on this. Our analytical methods consist in (i) leveraging existing statistical techniques developed for studying foraging in the related field of animal movement ecology (e.g. first-passage time, time-series segmentation), and (ii) comparison of observed queen movement patterns with that seen in null-model versions of the original queen trajectories generated by constrained randomisation of the original trajectories.
The first-passage time analysis shows that although queen movement is constrained to a limited subset of the nest, within that area (their spatial fidelity zone), queens do exhibit a much greater degree of mobility than do the synthetic trajectories – statistically, their movement is classed as superdiffusive, which means that they 'get around' more than would a random-walker. This indicates that queen mobility patterns serve to maximise the dispersal of the QMP signal throughout the nest.
Furthermore, compared to the null-model trajectories, queens tend to take longer to re-encounter their previous paths, suggesting queens actively avoid areas that they have recently-visited. As queens deposit QMP onto the wax wherever they walk, and as workers can acquire QMP from the wax, we interpret this as a adaptive strategy to maximise indirect transmission of QMP to the workers, by maximising the spatial area over which QMP is deposited.
Progress towards objective 2-3:
Although the contact networks have already been built, the analysis of how QMP is transmitted both directly (queen-to-worker contacts) and indirectly (worker-to-worker contacts) over such networks has already commenced - we anticipate that work on objectives 2-3 will be complete by early 2015.
3. Results and Conclusions
Queens show active behavioural features that can be expected to maximise the dissemination of the queen signal throughout the colony. As the full dissemination of the queen signal serves not only to and inhibit the workers from raising a new queen, but also to suppress selfish worker reproduction (`cheating'), it is in the interests of the queen to signal her signal as effectively as possible. Thus, the active avoidance of previously-visited areas, and features can be interpreted as a behavioural adaptation that increases queen fitness.
4. Expected final results
The final results will allow us to fully answer the three questions outlined above. We will then be in a position to quantify the contribution of the three different transmission pathways in the dissemination of the queen signal.
The first objective (concerning queen movement) has already been partially answered. We consider that there are two means by which these results can contribute towards efforts on defeating overwintering losses. First, honeybees have been subjected to selective breeding efforts for hundreds of years – our demonstration that queens exhibit active avoidance of recently-visited areas could therefore represent a trait that might be subjected to selective breeding. These efforts could focus upon developing developing strains that have been artificially selected for increased queen mobility and area-avoidance traits; such strains should exhibit decreased tendency to swarm, and hence higher worker populations for overwintering, and so reduced overwintering mortality.
Social insects – the ants, bees, wasps and termites – have achieved ecological dominance in many terrestrial ecosystems, indeed, their total biomass equals or exceeds that of human beings. A leading explanation for this runaway success is that social insects have evolved highly effective communication mechanisms that allow a colony of many tens of thousands of individuals to coordinate their activities such that the actions and behaviours of each individuals comes to be highly dependent upon those of their nestmates. This coupling allows the group to produce adaptive behaviours, such as collective foraging and defence, which resemble the behaviour of a single, unitary organism. The social insects therefore provide a unique system for the studying the principles that underly the organisation of complex animal societies.
Social insects have a huge economic influence upon human affairs, providing valuable ecosystem services such as soil turnover, nutrient cycling, and particularly in the case of the honeybee Apis mellifera, pollination of roughly 1/3 of the fruits and vegetables consumed by humans.
However, over the last 10-15 years, honeybee populations in North America, and to a lesser extent in Europe, have suffered severe overwintering losses. The typical symptom is that when the hives are opened in spring after the winter hibernation, all the bees have died or disappeared – such losses have been termed Colony Collapse Disorder (CCD). There seems to be no single cause, rather colony collapse is triggered by a complex mixture, including habitat fragmentation & loss of pollen-diversity, sub-lethal effects of industrial agrochemicals, viral infection via ectoparasitic mite vectors.
The original project proposal was to use a novel automated tracking system (originally developed for tracking ants), to simultaneously track the movements of hundreds of individual workers as they move, work and interact with other individuals within the hive. From these tracks, we proposed to to extract the network of worker-to-worker contacts. These contact networks networks serve as the substrate for the transmission of harmful pathogens; for example, the one fungal (microsporidian) pathogen of honeybees, Nosema apis, is horizontally transferred between workers during a praticular type of contact involving liquid food exchange (trophallaxis).
Since its inception, in 2010, the project aims have shifted from questions directly concerning disease transmission and epidemiology, to questions concerning colony reproduction and so-called 'queen-signalling'. However, this shift still has a direct connection to the problem of overwintering losses, as described below.
Honeybee colonies reproduce by colony fission, in which the old queen along with roughly half the worker population split off from the original colony (in a swarm), to found a new colony, leaving behind a young queen and half the original workforce. This process naturally occurs whenever the worker population reaches a certain critical density, so over the course of a year a colony can produce many swarms. Beekeepers strive to prevent swarming, as the beekeeper loses half of the bees (swarming reduces the honey-producing capacity, and hence the profitability, of the remaining colony), and more importantly, because the workforce of a colony that has 'cast-off' several swarms is much severely reduced, swarming can increase the likelihood that the colony does not survive the winter.
Swarming is thought to be controlled by the ability of the queen to signal her presence; queens produce queen mandibular pheromone (QMP), which inhibits the workers from both raising a new queen, and laying their own unfertilized eggs. QMP is produced by the queen, and disseminated through the colony by three routes; (i) by direct contact between the queen and her attendant workers, (ii) by indirect transfer from worker to worker during physical contacts, and (iii) indirectly via deposition onto the wax substrate. It is thought the dissemination of this signal is inhibited when the worker population reaches a critical density (e.g. through physical crowding effects), and that the blocking of this signal releases the workers to raise a new queen, and hence to begin path towards swarming. Our focus has thus been to use the automatic tracking system to study the pathways by which the queen signal is disseminated, specifically:
1. Whether the movement of the queen maximises the indirect dissemination of QMP, via the wax.
2. Whether the movement of the queen maximises the rate of direct contacts with the workers.
3. Whether there movement of workers that have recently encountered the queen also serves to maximise their encounter rate with `uninformed' workers.
2. Progress so far
Objectives 1-3 require the generation of time-stamped individual trajectories, whereas objectives 2-3 additionally require the extraction of contact networks in which workers are represented by nodes, and physical interactions between bees are represented as links that connect the nodes. Furhter, as there exists strong between-colony variation, objectives 1-3 require several independent replicates.
For a single replicate we attached took newly-hatched 1000 bees (and the queen), and attached a unique barcode-like tag to the thorax of each bee. These bees were then released back into the observation colony. The simultaneous movements of these tagged 1000 bees were continuously recorded over a 31-day period. From these recordings, we extracted 31 daily time-stamped trajectories (for objectives 1-3), and daily contact-networks (for objectives 2-3). Note, the 31-day period is roughly the same length of time as the typical worker lifespan, so the trajectories and interactions additionally capture the entire ontogeny of individual spatio-temporal mobility (from the trajectories), and social position (from the contact-based interaction networks).
Progress towards objectives 1:
The initial hurdle towards all objectives was the technical challenges associated with adapting the automated ant tracking system for application to honeybees. Designing, and assembling these systems took roughly 1 year. However, we now six fully-functional honeybee tracking systems. Data collection for the first 5 replicates was completed by the end of September 2014, and analysis has been ongoing since.
As objective 1 concerns the behaviour of a single individual in each colony – the queen – rather than the remaining 999 workers, the analysis has first focused on this. Our analytical methods consist in (i) leveraging existing statistical techniques developed for studying foraging in the related field of animal movement ecology (e.g. first-passage time, time-series segmentation), and (ii) comparison of observed queen movement patterns with that seen in null-model versions of the original queen trajectories generated by constrained randomisation of the original trajectories.
The first-passage time analysis shows that although queen movement is constrained to a limited subset of the nest, within that area (their spatial fidelity zone), queens do exhibit a much greater degree of mobility than do the synthetic trajectories – statistically, their movement is classed as superdiffusive, which means that they 'get around' more than would a random-walker. This indicates that queen mobility patterns serve to maximise the dispersal of the QMP signal throughout the nest.
Furthermore, compared to the null-model trajectories, queens tend to take longer to re-encounter their previous paths, suggesting queens actively avoid areas that they have recently-visited. As queens deposit QMP onto the wax wherever they walk, and as workers can acquire QMP from the wax, we interpret this as a adaptive strategy to maximise indirect transmission of QMP to the workers, by maximising the spatial area over which QMP is deposited.
Progress towards objective 2-3:
Although the contact networks have already been built, the analysis of how QMP is transmitted both directly (queen-to-worker contacts) and indirectly (worker-to-worker contacts) over such networks has already commenced - we anticipate that work on objectives 2-3 will be complete by early 2015.
3. Results and Conclusions
Queens show active behavioural features that can be expected to maximise the dissemination of the queen signal throughout the colony. As the full dissemination of the queen signal serves not only to and inhibit the workers from raising a new queen, but also to suppress selfish worker reproduction (`cheating'), it is in the interests of the queen to signal her signal as effectively as possible. Thus, the active avoidance of previously-visited areas, and features can be interpreted as a behavioural adaptation that increases queen fitness.
4. Expected final results
The final results will allow us to fully answer the three questions outlined above. We will then be in a position to quantify the contribution of the three different transmission pathways in the dissemination of the queen signal.
The first objective (concerning queen movement) has already been partially answered. We consider that there are two means by which these results can contribute towards efforts on defeating overwintering losses. First, honeybees have been subjected to selective breeding efforts for hundreds of years – our demonstration that queens exhibit active avoidance of recently-visited areas could therefore represent a trait that might be subjected to selective breeding. These efforts could focus upon developing developing strains that have been artificially selected for increased queen mobility and area-avoidance traits; such strains should exhibit decreased tendency to swarm, and hence higher worker populations for overwintering, and so reduced overwintering mortality.