GENOME-WIDE SEQUENCING APPROACHES
Rollins (2005) directly sequenced 2565 methylated domains from human brain DNA by combining enzymatic digestion followed genome-wide sequencing. This is the first study using sequencing approach to study DNA methylation in a large-scale structure of neuronal genome. Merging with studies from other fields, the majority of neuroscientists were thinking DNA methylation in brain is a stable DNA modification, until Guo et al. (2011a, 2014) revealed methylome in postnatal brains is dynamicly regulated by neuronal activation by using genome-wide bisulfite sequencing. Then, Lister et al. reported a comprehensive profile of 5mC in human and mouse frontal cortex throughout their life span. Even more interestingly, the data from genome-wide sequencing reveals that non-CpG cytosine methylation is much more common in the brain than in any other adult tissue type (Guo et al., 2014; Lister et al., 2013). These findings have led the field of neuroscience to the realization that DNA methylation is far more dynamic and complex than initially expected and that a balance between DNA methylation and DNA demethylation is constantly fine- tuned to maintain gene expression networks in the brain.
Although indispensable for profiling whole-genome methylation at the level of the entire adult brain, there is an increasing understanding that bisulfite sequencing is not an appropriate approach to detect DNA methylation in the brain. Findings relating 5hmC, Tet proteins, and active DNA demethylation to neuronal function have made it clear that a detection method that cannot identify 5fC and 5caC will lose a great deal of information. Also, the requirement of large amount of input DNA makes it impossible to study tissues that are too small to yield the necessary DNA amount, such as small regions of the rodent brain, or even single neurons, in addition to other limitations we mentioned previously. DNA immunoprecipitation is becoming a more common approach to detect dynamic DNA modifications, such as 5mC, 5hmC, 5fC, and 5caC, under neuronal activities. Li et al. (2014b) developed a novel strategy for genome-wide sequencing based on DNA immu- noprecipitation. A quantity of input DNA as small as 50 ng was individually barcoded and then mice were pooled. Through this technique, only 50 ng of input DNA can provide a reliable, comprehensive analysis of different DNA modification patterns across the entire neuronal genome. This would theoretically permit genome-wide DNA modification profiling from as low as 10,000 cells. To further extend this new sequencing approach, fluorescence-activated cell sorting has been used to separate neurons (NeuN+) from glial cells and other cell types (NeuN-) of the ventromedial prefrontal cortex of individual adult C57BL/6 mice (Li, Baker-Andresen, Zhao, Marshall, & Bredy, 2014a). These studies described distinct differences on methylation pattern between neurons and nonneuronal cells. It is thus important to perform experiments using isolated neurons, rather than heterogenous brain tissue, to accurately profile the role of DNA modifications in information storage within the neuronal genome. This approach has been used to successfully provide a tissue- and cell-specific DNA methylation map during cocaine self-administration (Baker-Andresen et al., 2015).