Large-scale transcriptome changes in the process of long-term visual memory formation in the bumblebee, Bombus terrestris

Abstract

Many genes have been implicated in mechanisms of long-term memory formation, but there is still much to be learnt about how the genome dynamically responds, transcriptionally, during memory formation. In this study, we used high-throughput sequencing to examine how transcriptome profiles change during visual memory formation in the bumblebee (Bombus terrestris). Expression of fifty-five genes changed immediately after bees were trained to associate reward with a single coloured chip, and the upregulated genes were predominantly genes known to be involved in signal transduction. Changes in the expression of eighty-one genes were observed four hours after learning a new colour, and the majority of these were upregulated and related to transcription and translation, which suggests that the building of new proteins may be the predominant activity four hours after training. Several of the genes identified in this study (e.g. Rab10, Shank1 and Arhgap44) are interesting candidates for further investigation of the molecular mechanisms of long-term memory formation. Our data demonstrate the dynamic gene expression changes after associative colour learning and identify genes involved in each transcriptional wave, which will be useful for future studies of gene regulation in learning and long-term memory formation.

Figure 1

Training procedure and establishment of long-term memory with absolute conditioning. (a) Training procedure. Bees were trained individually to forage on one transparent chip, which contained 100 µl 40% sucrose solution (five trips with 10 min inter-trip intervals on each of two consecutive days). On day three, bees were trained to visit a chip containing 100 µl 40% sucrose solution five trips (with 10 min inter-trip interval) in one of three conditions: visiting a transparent chip and collected immediately after training (0-hour Control); visiting a yellow chip and collected immediately after training (0-hour Learning); visiting a yellow chip and kept in the hive for four hours without any further foraging experience prior to collection (4-hour Learning). (b) Bees form a long-term memory of a colour trained under absolute conditioning, irrespective of the colour trained. Bees discriminated the conditioned colour from a novel colour during a memory retention test (t-test, Yellow: t = 16.90, df = 5, p = 0.000; Magenta: t = 6.64, df = 5, p = 0.001, compared to chance expectation 50%). Accuracy did not differ between yellow and magenta (t-test, t = 0.65, df = 10, p = 0.532). (c) Bees’ memory performance did not differ across colonies. Bees in all three colonies formed long-term memory after training (t-test, Colony 1: t = 15.50, df = 10, p = 0.000; Colony 2: t = 8.05, df = 11, p = 0.000; Colony 3: t = 7.04, df = 9, p = 0.000, compared to chance expectation 50%) and bees performance on the memory retention test did not differ between the three colonies used for sequencing (one-way ANOVA, F_{(2, 30)} = 2.14, p = 0.136). The number within each bar indicates the number of bees tested. Vertical bars indicate standard deviation.

Figure 2

Gene expression differences associated with different learning and memory statuses. (a) The number of co-expressed and unique genes observed in the entire transcriptome among the three experimental groups. 86% of genes (16955) were shared among the three experimental groups. (b) Scatterplot of PC1 and PC2 from a principal component analysis of all samples using the gene expression values for differentially-expressed genes. The symbols represent samples from different experimental groups. PC1 and PC2 contributed 55% and 31% of the total variance, respectively. Nine samples can be separated into three experimental groups, which indicates that each learning/memory state has its own specific gene expression pattern. 0 C: 0-hour Control; 0 L: 0-hour Learning; 4 L: 4-hour Learning.

Figure 3

Hierarchical clustering of brain gene expression levels in bees with different learning and memory status. Each column represents a sequencing sample and each row represents a gene. Gene expression values are colour coded: blue indicates higher expression and yellow indicates lower expression. The normalized gene expression values of 110 differentially expressed genes were used for hierarchical clustering. It is evident that samples in each experimental group can be clustered together and 0-hour Control group and 0-hour Learning group had more similar gene expression patterns. In addition, several gene expression patterns were found and five of them stood out, as highlighted with purple rectangles. Red boxes on the right show the main GO terms for each cluster and the full list of GO terms in each cluster was shown in Supplementary File S3. 0 C: 0-hour Control; 0 L: 0-hour Learning; 4 L: 4-hour Learning.