Epigenetic regulation of the honey bee transcriptome: unravelling the nature of methylated genes

Abstract

Background: Epigenetic modification of DNA via methylation is one of the key inventions in eukaryotic evolution. It provides a source for the switching of gene activities, the maintenance of stable phenotypes and the integration of environmental and genomic signals. Although this process is widespread among eukaryotes, both the patterns of methylation and their relevant biological roles not only vary noticeably in different lineages, but often are poorly understood. In addition, the evolutionary origins of DNA methylation in multicellular organisms remain enigmatic. Here we used a new 'epigenetic' model, the social honey bee Apis mellifera, to gain insights into the significance of methylated genes.

Results: We combined microarray profiling of several tissues with genome-scale bioinformatics and bisulfite sequencing of selected genes to study the honey bee methylome. We find that around 35% of the annotated honey bee genes are expected to be methylated at the CpG dinucleotides by a highly conserved DNA methylation system. We show that one unifying feature of the methylated genes in this species is their broad pattern of expression and the associated 'housekeeping' roles. In contrast, genes involved in more stringently regulated spatial or temporal functions are predicted to be un-methylated.

Conclusion: Our data suggest that honey bees use CpG methylation of intragenic regions as an epigenetic mechanism to control the levels of activity of the genes that are broadly expressed and might be needed for conserved core biological processes in virtually every type of cell. We discuss the implications of our findings for genome-scale regulatory network structures and the evolution of the role(s) of DNA methylation in eukaryotes. Our findings are particularly important in the context of the emerging evidence that environmental factors can influence the epigenetic settings of some genes and lead to serious metabolic and behavioural disorders.

Figure 1

Histogram showing the frequency of all annotated Apis genes with CpG [o/e] frequencies between 0 and 2. For comparison, a similar analysis was performed on the nematode (C. elegans) genome that is not methylated due to the lack of the genes encoding DNA methyl-transferases. The honey bee CpG-deficient genes (CpG [o/e] <0.7) expected to be methylated are labelled in black. We used the honey bee official set of 10,742 genes available at BeeBase . The y-axis depicts the number of genes with the specific CpG [o/e] values given on the x-axis.

Figure 2

Venn diagram showing the overlap of gene expression profiles between five experimental conditions: antennae, brains, thoraces, ovaries and larvae. Only five conditions were selected for this diagram (it is impossible to plot a Venn diagram in two dimensions with more than five sets using ellipses [48]). ArrayExpress accession: E-MEXP-2093.

Figure 3

Distribution of CpG bias in: i) all Apis predicted transcription units, ii) the ubiquitously expressed genes, and in iii) the condition-specific genes (HP Gland - hypopharyngeal gland).

Figure 4

Functional categorization of methylated and unmethylated genes based on Gene Ontology (GO) classification. GO terms were assigned to honey bee predicted proteins using the corresponding GO terms of their BLASTP hits in the RefSeq Drosophila protein database. If the best hit did not have any associated GO terms, the best subsequent hit with associated GO terms was used and no GO terms were assigned to honey bee proteins that did not have any GO annotated hit with an e-value smaller than 1e-5. For illustration purposes only molecular function level 3 ontology terms (where Level 0 = root = Gene_ontology) were selected and grouped into larger categories. See additional file 4 for more details.