One unifying theme of the honey bee methylome is the broad pattern of expression of methylated genes indicating that gene activities required for the core cellular functions might be controlled, at least partly, by epigenetic means. Although these ubiquitously expressed genes may not represent the nominal size of the 'housekeeping' transcriptome in this organism, it seems likely that they are constitutively expressed in time and space. Such permanently activated genes providing 'maintenance' functions required by virtually all cells have been typically described in the past as unregulated. However, it has been suggested that in spite of their permanent activation the 'housekeeping' genes might not be required at the same level throughout development [26
], or under changing environmental conditions. Indeed, evidence suggests that even most stable transcripts are sensitive to both biotic and abiotic external influences [27
]. Our data add more weight to the notion that the activities of 'housekeeping' transcripts and their products might be modulated by epigenetic means. Such a mechanism may also exist in other organisms [9
] suggesting that a direct relationship between gene methylation and transcription is a widely spread phenomenon in both the animal and plant kingdoms.
In mammals, the majority of promoters driving the 'housekeeping' genes are associated with CpG islands [29
]. These genomic regions containing a high frequency of CG nucleotides are typically not methylated with the exception of CpG islands on the inactive X chromosome and in disease situations. In contrast to mammals, the broadly expressed genes in Apis
do not have CpG islands, whereas two out of six unmethylated genes with restricted patterns of expression selected for our detailed analyses (GOX and Impl3) are associated with classic CpG islands (table ). GOX is stringently regulated and its expression is exceptionally high in the HP gland of nurse bees, whereas Impl3 is predominantly a larval gene, and its differential expression in worker and queen larvae is part of a network that determines the reproductive fate of female bees [19
]. Although Impl3 is not directly methylated (table ), its expression is reduced in Dnmt3-silenced larvae or by feeding royal jelly [19
], suggesting that both unmethylated and methylated genes might be influenced by epigenetic controls in highly interconnected regulatory network structures. In honey bees, diet-induced changes in methylation levels lead to metabolic acceleration and increased growth driven by global, but relatively subtle changes in the expressional levels of a large number of genes [19
]. These initial changes are later followed by the activation of more specific pathways to lay down caste-specific structures, such as pollen collecting combs on workers' legs that are built during pupal stages. Thus, instead of inventing two separate developmental blueprints, the bees differentially use one common plan to produce two distinct organismal outputs [17
]. Here the entire network rather than its individual components evolved to create an alternative developmental trajectory. This might occur if a given phenotype is biologically regulated by large numbers of subtle gene expression differences that act additively, in cascade leading to a major change in the topology of a global network of interacting genes ([31
] and references therein). A recent in silico
analysis confirms that queen-worker transcriptional differences are associated with genes showing distinct CpGo/e ratios [35
]. The epigenetic regulation of phenotypic polymorphism in honey bees is an example of the adaptive value of phenotypic plasticity that was the driving force in generating the reproductive division of labour in social insects.
Like in other invertebrates [10
] the global level of genome methylation in Apis
is low and appears to be restricted to CpGs residing in coding exons [16
]. It has been argued [14
] that global methylation, a hallmark of vertebrate genomes, arose within the phylum Chordata at the time when vertebrates originated, and was a major source of innovation at the genomic level. However, Regev et al [11
] concluded that methylation, originally used as a general repressor of genomic parasites, was recruited to perform gene regulatory functions well before the transition from invertebrates to vertebrates. One possibility is that transcriptional regulation by DNA methylation is an ancient mechanism of gene control that was adequate for primordial metazoan species with limited cell type and tissue repertoires. As animal evolution progressed, novel regulatory mechanisms operating via promoters and sequence-specific transcription factors (TFs) were invented to generate both the developmental sophistication and cellular diversity that characterise modern animals. As a result, organismal complexity is largely instantiated at the level of differential gene expression that evolved by combining the specific TFs, differential splicing, non-coding RNAs, chromatin remodeling and epigenetic modification of genomic DNA by methylation [1
]. In this context, the lack of an obvious correlation between gene number and apparent morphological and behavioral complexities of diverse organisms in different phyla [37
] is not surprising. While the combinatorial interactions of TFs and their targets are now well understood [38
], the role(s) of epigenetic modifications in gene regulation are only beginning to be unraveled.
The results presented in this paper have important implications for the field of evolutionary developmental biology (evo-devo). A prominent view in this field is that morphological diversity is caused primarily by mutations in the cis-regulatory regions of genes [40
], rather than by changes in protein coding sequences as suggested by other authors (eg [41
]). A compromising proposal [42
] predicts that the relative importance of both cis-regulatory and protein coding changes will vary depending on factors such as the position or rank of a gene in a regulatory network, the population dynamics and the evolutionary time span. In this model, highly interconnected genes are preferentially subjected to cis-regulatory evolution, while mutations in protein coding sequences are more prevalent in genes residing in less densely clustered parts of the network. Our results suggest that intragenic methylation might be an additional constituent of the cis-regulatory machinery regulating the components of densely connected metabolic and information processing networks constitutively expressed in most cells. In contrast, effector genes responsible for cell differentiation and specialization might not require these rich and complex regulatory inputs, and would not be methylated.
To understand the relevance of epigenetic influences on regulatory networks to developmental and evolutionary transitions, studies of the same genes and their interacting partners are required in different phyla. By comparing epigenomes, with their developmental end-points from different phyla we should be able to reveal what is functionally common and what is different. The emerging field of insect epigenomics will undoubtedly accelerate these efforts by providing novel and exciting data on genome-wide analyses of TF-binding sites, histone modifications, DNA methylation and context-dependent gene expression.