"Epigenetic" implies changes in regulatory states of genes or genomic DNA without changes in DNA sequence. The archetypical epigenetic modification in eukaryotic genomes is the addition of a methyl group to the fifth carbon of cytosine to produce 5-methylcytosine (5meC) [1, 2], reviewed in . Cytosine DNA methylation is an epigenetic modification that is shared by many eukaryotic organisms. Along with various other epigenetic modifications such as methylation, phosphorylation and acetylation of histone amino acids, cytosine methylation is an important regulator of biological processes including transposon silencing, heterochromatin organization, genomic imprinting, and gene expression.
The distribution of cytosine methylation is highly variable within plant genomes . This overall methylation pattern, which is conserved among diverse plant taxa, is often described as "mosaic," as it consists of interspersed methylated and unmethylated regions [5, 6]. The patterns of 5meC, mechanisms for de novo and maintenance methylation and the requirement for specific proteins for cytosine methylation have been best studied in Arabidopsis thaliana, where roughly 20% of the genome is methylated in whole seedlings [7, 8]. Cytosine methylation is strongly enriched in heterochromatin at pericentromeric and subtelomeric repeats, and at rDNA clusters [7, 9]. Repetitive sequences, which consist largely of transposons, retrotransposons, and tandem or inverted repeats, are highly methylated [5, 10, 11]. A novel and unexpected finding from genome-wide surveys was that a third of A. thaliana genes are methylated within their transcribed regions ("gene body methylation") [7, 8], while perhaps 16% of rice (Oryza sativa) genes are enriched for 5meC . The relationship between gene body methylation and transcription is currently not well understood. While promoter methylation is generally associated with lower transcription in A. thaliana , the relationship of gene body methylation to expression is complex, with methylation tending to occur most often in genes transcribed at moderate to high, but not very high, levels [4, 7, 8].
In plants, 5meC can occur in all sequence contexts (CG, CHG and CHH, where H refers to A, C or T) [13, 14]. The mechanisms responsible for establishment and maintenance of 5meC are best studied in A. thaliana where the maintenance methyltransferase MET1 targets hemimethylated CG sites, and the de novo methyltransferases DRM2 and CMT3 target CHG and CHH sites. Disruption of maintenance methylation results in abnormal developmental phenotypes including stunting, malformed leaves, decreased apical dominance, lower fertility, disrupted heterochrony, delayed flowering time and abnormal flower morphology [15, 16], while DRM2 and CMT3 mutants display defects in RNA mediated silencing , as well as dwarfing and abnormal leaf phenotypes . The activity of methyltransferases appears synergistic, at least in some cases, so that deletion of DRM1/2/CMT3) affects CG methylation maintenance by MET1 . Together, these results suggest that 5meC in all contexts can affect several aspects of chromatin regulation, with consequences for plant development and differentiation.
Tissue-level variation in methylation has been noted in several plant species. For example, in Arabidopsis, about six percent of cytosines were found to be methylated in immature floral , while 24 percent of CG, six point seven percent of CHG, and one point seven percent of CHH were methylated in young plants . Few studies have compared high-resolution methylation profiles among tissues within a plant species. In rice, whole genome methylation patterns were found to be similar among mature leaves, embryos, seedling shoots and roots, but hypomethylation was correlated with preferential expression in endosperm . Patterns of 5meC in LTR transposable elements differed between rice leaves and roots and affected transcription of neighboring genes  a phenomenon common to the SINE containing FWA promoter of A. thaliana [23, 24].
In addition to well-established roles in transposable element silencing and genomic imprinting, DNA methylation may be involved in plant adaptation to stress [25, 26]. In A. thaliana, genome-wide methylation increased in the progeny of plants exposed to temperature extremes, ultraviolet light , flood, and salt but decreased in progeny of drought-stressed plants [28, 29]. In hybrid poplars (P. deltoides × P. nigra), shoot apices from drought-stressed juvenile trees exhibited genotype-dependent 5meC variation . Differential DNA methylation patterns in poplar clones that have acquired differential transcriptome responses to drought stress have been observed .
While much has been learned from work on annual plants, in-depth investigations of cytosine methylation patterns in long-lived plants have been sparse. Because of their long term tissue differentiation and perennial exposure to environmental stresses, DNA methylation may play a greater role in both tree development and homeostasis. Studies of gross cellular DNA methylation indicate that it may vary substantially during tree development, whether assessed in vivo or in vitro. In apical buds of chestnut trees, Castanea sativa, 5meC increased during bud set and decreased during bud burst . In Monterey pine, Pinus radiata, 5meC levels in needles of reproductively mature trees were double that of juvenile needles . In shoots of chestnut and Monterey pine, a gradual increase in DNA methylation accompanied aging over 5-8 years [34, 35]. Increased methylation in mature vs. juvenile leaves was associated with loss of capacity for in vitro organogenesis in P. radiata . In micropropagated Acacia, shoots with juvenile leaves exhibited higher DNA methylation levels than shoots with mature leaves . Transient DNA methylation of ovules accompanied embryogenesis in chestnut . As noted above, in poplar drought stress induced changes in total cellular DNA methylation  and was associated with transcriptome changes within separately propagated clones .
A variety of experimental techniques can be applied to study genome-wide DNA methylation (reviewed in ). On a gross scale, the proportion of 5meC can be estimated by HPLC or HPCE, as has been done to show differences in 5meC among tissue types or treatments [33, 35]. The drawback of these methods is the lack of sequence specific information. Immunoprecipitation with an antibody raised against 5-methylcytidine (MeDIP), followed by genome tiling array hybridization or high-throughput sequencing of the precipitated DNA (MeDIP-seq), has been used to enumerate and compare methylated regions in Homo sapiens [40, 41], Mus musculus , Neurospora  and A. thaliana [7, 8]. The most detailed, single-base resolution maps are generated by sequencing of genomic DNA treated with sodium bisulfite, which converts unmethylated cytosines to uracils but leaves 5meC unconverted . However, this technique requires very high sequencing depth and is not suitable for mapping to repetitive genomic regions where uniqueness can be confounded by the presence of C to T SNPs. Genome-wide bisulfite sequencing was first used in Arabidopsis , but has now also been used to assess genome methylation in Oryza sativa and P. trichocarpa [6, 21], as well as mammals including. H. sapiens  and M. musculus . For the present work, we chose MeDIP-seq of many different tissue types because it provides comprehensive methylome coverage at a lower cost than genome-wide bisulfite sequencing.
The black cottonwood, Populus trichocarpa, is widely recognized as a reference species for tree biology. It has been studied in great detail over the past 30 years, and many resources are readily available, including a draft genome sequence http://www.phytozome.net/poplar, custom microarrays, and extensive transcriptome data . For our studies we used genome assembly version 2.2 in combination with published expression microarray data from multiple tissue types [48, 49]. While mature leaves from P. trichocarpa have recently been subjected to genome-wide bisulfite sequencing , high-resolution epigenomic methods have not yet been applied to discern tissue-level variation. We investigated variation in genome-level cytosine methylation among all of the major types of differentiated poplar tissues. To this end, we sequenced methylated DNA obtained by MeDIP from seven P. trichocarpa tissues on an Illumina GAIIx. We found overall patterns of cytosine methylation that are consistent with those seen in Arabidopsis, but observed differences in methylation patterns among tissue types not previously studied. We also found a different pattern of association of gene body methylation to gene expression.