Experimental Setup and Initial Dataset
The transcriptomic data used for analysis was part of an experiment
prepared to study the molecular basis for social learning of distance in
honeybees through the waggle dance . In this experiment honeybees from 3
different colonies were trained to visit a feeder positioned at the end
of a 6m long tunnel , which was used to alter the bee’s perception of
distance as follows: vertical stripes (with respect to the direction of
flight) on the tunnel walls were used to increase the estimated flight
distance, while horizontal stripes were used to decrease it (Figure 1).
Honeybees were then marked at the feeder according to perceived distance
(similarly to , yielding two groups: “honeybees perceiving long
distance” and “honeybees perceiving short distance”.
Honeybee colonies were housed in an observation hive, which allowed
direct monitoring of the comb where honeybees normally performed the
waggle dance after returning from a foraging trip, known as the “dance
floor” . Bees were trained during the morning to visit the feeder at the
end of the tunnel – it usually took approximately 5 hours to complete
this part – and then in the afternoon (between 2pm and 4pm) foragers
that regularly visited the feeder were monitored by an observer while
flying from the dance floor to the feeder and vice-versa repeatedly.
During this 2-hour time window, the dance floor was also recorded with a
video camera, producing a recording of all waggle dance events that
occurred in the focal colony. The footage was then carefully analysed to
identify marked honeybees that performed waggle dances upon their return
from the feeder (from now on “dancers”) and separate them from those
that instead were never seen performing any dance (“non-dancers”)
despite being visible on the dancefloor upon their return from a
foraging trip. An analysis of the recorded dances confirmed that the
manipulation was successful: bees exposed to vertical stripes advertised
longer distances on average in their dances compared to bees exposed to
horizontal stripes (Manfredini et al. in prep ).
This resulted in the following 4 groups of honeybees: Dancer perceiving
Long distance (DL), Dancer perceiving Short distance (DS), Non-dancer
perceiving Long distance (NL), Non-Dancer perceiving Short distance
(NS), with 8 replicate samples in each of the 4 groups (N = 32). Brain
tissues from all these samples were processed individually for RNAseq
analysis (see Supplemental Information). RNAseq read files were aligned
to the most recent version of the A. mellifera genome (Amel_4.5)
using the intron-aware STAR aligner version 2.6.1a . Read counts were
extracted using the featureCounts function from the Bioconductor
R package ‘Subread’ version 1.8.0 . The final dataset, which represent
the starting material for this study, included the read counts for
15,314 genes, corresponding to the
whole honeybee genome across 32 bees. As we noticed some variation in
library size for some of the bee samples, we normalised read counts by
the total library sizes to correct for the effect of possible outliers.