Abstract The honeybee is a model organism for studying learning and memory formation and its underlying molecular mechanisms. While DNA methylation is well studied in caste differentiation, its role in learning and memory is not clear in honeybees. Here, we analyzed genome-wide DNA methylation changes during olfactory learning and memory process in A. mellifera using whole genome bisulfite sequencing (WGBS) method. A total of 853 significantly differentially methylated regions (DMRs) and 963 differentially methylated genes (DMGs) were identified. We discovered that 440 DMRs of 648 genes were hypermethylated and 274 DMRs of 336 genes were hypomethylated in trained group compared to untrained group. Of these DMGs, many are critical genes involved in learning and memory, such as Creb, GABA [B] R and Ip3k, indicating extensive involvement of DNA methylation in honeybee olfactory learning and memory process. Furthermore, key enzymes for histone methylation, RNA editing and miRNA processing also showed methylation changes during this process, implying that DNA methylation can affect learning and memory of honeybees by regulating other epigenetic modification processes. Introduction Honeybees (Apis mellifera) are social insects with important economic and ecological value due to their pollination services. They are able to distinguish different colors^[38]1, odors^[39]2, even the relationship between objects, such as, sequential order^[40]3, and the upper and lower^[41]4. Even more astonishing is their ability to learn abstract rules and concepts^[42]5,[43]6. Due to their remarkable abilities in learning and memory and the relative simple structure of their brains, honeybees are considered a good model for neurobiology. But the molecular mechanisms underlying the learning and memory process of honeybees are not well understood. Epigenetic modifications, such as DNA methylation^[44]7,[45]8, histone modifications^[46]9,[47]10, and miRNA processing^[48]11,[49]12, have been shown to be involved in learning and memory processes in vertebrates. DNA methylation plays a critical role in long-term memory formation in many organisms and different learning paradigms^[50]8. The change of DNA methylation levels, which is regulated mainly by the activity of DNA methyltransferase (DNMTs), directly regulates the expression level of genes involved in memory formation in the brain^[51]13,[52]14. The honeybee genome possesses a complete, functional DNA methylation system, which includes an ortholog of Dnmt3 and two orthologs of Dnmt1 (AmDnmt1a and AmDnmt1b)^[53]15. Moreover, all these three DNMTs have catalytic activity. DNA methylation is regarded as the main mechanism regulating queen-worker caste differentiation in honeybees, and many studies have surveyed DNA methylation differences between these two castes^[54]16–[55]20. DNA methylation is also reported to be closely associated with learning and memory processes in honeybees^[56]21–[57]23. Inhibiting the activity of DNMT in honeybees using zebularine revealed that DNA methylation mediates long-term memory formation in honeybees after associative learning, but not short-term memory formation^[58]21,[59]22. By measuring the methylation of 30 memory associated genes in honeybees, Biergans et al.^[60]23 found that, during memory formation process, memory associated genes are regulated by a temporally complex epigenetic mechanism. Even though it is known DNA methylation is involved in long-term memory formation in A. mellifera, the actual pattern of DNA methylation after learning and memory has not been studied in honeybees. In this process, how many genes are methylated? To what extent does DNA methylation affect gene expression? Here, we analyzed the genome-wide DNA methylation changes following olfactory learning and memory in A. mellifera through WGBS method, and identified many DMRs and associated genes. Our results suggest extensive involvement of DNA methylation in honeybee olfactory learning and memory. Results DNA methylation patterns After sequencing, about 72,434,814 and 76,497,724 raw reads with a Q20 value of more than 92% were generated for the two samples, respectively (Table [61]1). After removing low-quality reads, 58,669,954 (7.34 Gb) and 64,500,902 (8.06 Gb) clean reads which were more than 25 × coverage of the 285 Mb A. mellifera reference genome remained, and were mapped to the genome. Finally, 56.95% and 61.10% of the reads were uniquely mapped to the honeybee genome and more than 77% of the total clean reads have a coverage of >5 × on the genome (Table [62]1). The BS conversion rates are more than 99.83% in each sample, indicating high T-C conversion during bisulfite treatment. Table 1. Data generated by whole genome bisulfite sequencing. Trained Untrained Raw reads 72434814 76497724 Raw bases (G) 9.04 9.56 Clean reads 58669954 64500902 Clean bases (G) 7.34 8.06 Q20 (%) 92.91% 93.16% BS Conversion Rate (%) 99.85% 99.83% Total mapped reads 33414826 39408700 Mapping rate (%) 56.95% 61.10% Duplication rate (%) 3.4% 3.4% 5× coverage (%) 78.50% 77.30% [63]Open in a new tab A total of 139,430 and 140,997 methylated cytosines (mCs) were detected in trained and untrained groups, both occupying about 0.19% of all the cytosine sites in the honeybee genome (Table [64]2). Of these mCs, the percentages of mCs in CG, CHG (H represents A, C or T) and CHH contexts in the trained group were 99.47%, 0.06%, 0.47% (Fig. [65]1A), occupying 0.6829%, 0.0010%, 0.0014% of the genome-wide CG, CHG and CHH sites respectively (Table [66]2); the corresponding numbers in the untrained group were 99.50%, 0.06%, 0.44% (Fig. [67]1B), occupying 0.6908%, 0.0010%, 0.0014% of the genome-wide CG, CHG and CHH sites respectively (Table [68]2). Table 2. The genome-wide percentages of methylated CG, CHG and CHH in the trained and untrained groups. Trained Untrained mC percent (%) 0.1861% 0.1882% mCG percent (%) 0.6829% 0.6908% mCHG percent (%) 0.0010% 0.0010% mCHH percent (%) 0.0014% 0.0014% [69]Open in a new tab Figure 1. [70]Figure 1 [71]Open in a new tab The distribution (%) of mCs in CG, CHG and CHH contexts. Sequence preferences flanking the methylated C sites