Understanding the chromatin remodeling code

Plant Science 211 (2013) 137–145

Contents lists available at ScienceDirect

Plant Science journal homepage: www.elsevier.com/locate/plantsci

Review

Understanding the chromatin remodeling code Misook Ha ∗ Samsung Advanced Institute of Technology, Samsung Electronics Corporation, Yongin-Si, Gyeonggi-Do 446-712, South Korea

a r t i c l e

i n f o

Article history: Received 16 April 2013 Received in revised form 15 July 2013 Accepted 17 July 2013 Available online 24 July 2013 Keywords: Chromatin remodeling Chromatin modifications Epigenomics Computational analysis

a b s t r a c t Remodeling a chromatin structure enables the genetic elements stored in a genome to function in a condition-specific manner and predisposes the interactions between cis-regulatory elements and transacting factors. A chromatin signature can be an indicator of the activity of the underlying genetic elements. This paper reviews recent studies showing that the combination and arrangements of chromatin remodeling marks play roles as chromatin code affecting the activity of genetic elements. This paper also reviews recent studies inferring the primary DNA sequence contexts associated with chromatin remodeling that suggest interactions between genetic and epigenetic factors. We conclude that chromatin remodeling, which provides accurate models of gene expression and morphological variations, may help to find the biological marks that cannot be detected by genome-wide association study or genetic study. © 2013 Elsevier Ireland Ltd. All rights reserved.

Contents 1. 2.

3.

4.

5.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chromatin remodeling mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Chromatin remodeling as a main epigenetic mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Dynamics of chromatin remodeling in response to environmental stimuli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Encryption of parental environmental signals in descendant’s genomic DNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Analytics for inference of chromatin remodeling codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Integration of ChIP-seq and ChIP-chip data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Examination of relationships among chromatin remodeling marks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1. Nucleosome and protein bindings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2. Histone modifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3. DNA methylations mediated by small RNAs and histone modifcations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.4. Chromatin dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Clustering and classification analyses of chromatin remodeling map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Probabilistic modeling of chromatin remodeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chromatin remodeling signature in a primary genome sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. DNA sequence features of chromatin remodeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Predicting the chromatin remodeling map from a genome sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions and perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction A genome stores all of the genetic information of an organism. Eukaryotic genome is packaged in a nucleus as a chromatin

Abbreviations: H3K4me3, tri-methylations at histone H3 Lysine 4; ChIP-seq, chromatin immunoprecipitation and sequencing; siRNA, small interference RNA. ∗ Tel.: +82 31 280 9715. E-mail addresses: [email protected], [email protected] 0168-9452/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.plantsci.2013.07.006

137 138 138 138 138 139 139 140 141 141 141 141 141 142 142 142 143 144 144 144

structure, which is a complex of DNA with proteins and RNAs. Chromatin structure can tightly pack genomes and restricts the interaction of DNA segments with protein factors. On the other hand, remodeling chromatin structure can allow proteins to interact with specific genomic DNA segments. Recent studies show that chromatin structure is dynamically remodeled. Specific chromatin structures are recognized by protein factors involved in transcription and DNA replication. In this way, remodeling chromatin structure controls DNA replication and transcription in a condition-specific manner. Therefore, understanding the specific

138

M. Ha / Plant Science 211 (2013) 137–145

pattern of chromatin remodeling affecting the activity of genetic elements may enable us to predict cellular status. Chromatin’s basic structural unit is a nucleosome consisting of a ∼147 bp DNA segment and histone octamers which consist of pairs of H2A, H2B, H3 and H4. The eukaryotic chromatin structure can be remodeled by at least five mechanisms (reviewed in [1]): (1) the nucleosome formation, entailing tendency of nucleosome formation, nuclesome occupation levels, and arrangement of nucleosomes; (2) adding covalent modifications to histones; (3) replacing histones with histone variants [2,3]; (4) methylations at DNA cytosine; and (5) small and long non-coding RNAs. All of these chromatin remodeling processes change constituent, condensation, accessibility, and interacting proteins of the chromatin structures. Therefore, chromatin remodeling marks reflecting chromatin structural modification status include DNA methylations, nucleosome formation, proportion of nucleosomes containing histone variants, and histone modifications such as histone methylations, acetylations, ubiquitylations, and phosphorylations. By the action of ATP-dependent chromatin remodelers, the combination of specific chromatin modification marks are arranged and interact with specific proteins involved in transcription, DNA replications, and DNA repair [4,5]. Therefore, distinct arrangement of chromatin remodeling marks may play roles as chromatin remodeling code. This paper focuses on the data analytics for understanding features of chromatin remodeling processes at the primary organization level–nucleosome, histone variants, histone modifications, DNA methylation and RNA generations–and their coordination. We first review the significance of chromatin remodeling in epigenetics and gene regulation in response to development and environmental cues. We extend the ideas to the chromatin remodeling code hypothesis; arrangement of chromatin remodeling marks can partly regulate the function of genetic code. We then discuss several strategies to decipher chromatin signatures for prediction and controlling the gene expression and plant traits. Our objective is to apply chromatin codes to the field of genome engineering for crop improvement and human health. There are excellent reviews on chromatin remodeling and epigenetics [1,2,6,7]. Rather than reviewing detailed molecular mechanisms of chromatin remodeling, we focus on the general methods for analyses of genome-wide chromatin remodeling measurements in eukaryotes including plants. We then focus on the current literature inferring the primary DNA sequence contexts associated with chromatin remodeling that suggest interactions between genetic and epigenetic factors. 2. Chromatin remodeling mechanisms 2.1. Chromatin remodeling as a main epigenetic mechanism Various definitions of epigenetics originated from “study how genes interact with environment and result in development and phenotypes” by Waddington in 1942 [8]. Currently, epigenetics generally refers to the study of reversible and heritable morphological variations not caused by DNA sequence changes. Chromatin remodeling may be a major molecular mechanism of epigenetics. Sometimes, epigenetic trait is defined as “a stably heritable phenotype resulting from changes in a chromosome without alterations in the DNA sequence [9].” First, chromatin remodeling does not change the underlying genome sequence, although it is reversible due to the activity of chromatin modifiers and environmental changes. Second, chromatin remodeling is a reversible process by the action of chromatin modifiers. Eukaryotic genomes contain multiple copies of chromatin modifiers. Three families of histone acetyltransferases and 3 families of histone deacetylases are conserved in Arabidopsis thaliana, Saccharomyces cerevisiae, Schizosaccharomyces pombe, Caenorhabditis elegans, and Drosophila melanogaster [10,11]. Third, the cell-type specific

chromatin remodeling pattern is maintained during development and involved in gene expression regulation in developmental processes [12]. Chromatin remodeling processes are reversible and therefore play an important role in the transcriptional regulation associated with developmental programs and environmental conditions. In conclusion, chromatin remodeling may be a fundamental principle of epigenetics. Gene expression is highly regulated by the interactions among cis-regulatory elements and trans-acting factors. Debate is ongoing whether cis- or trans-regulation is the primary source of morphological variations within and between species in genetics [13,14]. In cis-regulation, the cis-regulatory elements encoded in the same DNA strand direct the specific expression of the target genes. Sequence variations in cis-regulatory elements change the affinity of the genetic elements with protein factors and differentiate the regulation of target gene expression. In trans-regulation, proteins or RNAs encoded in other DNA strands interact with the target loci and regulate the expression of the target genes. Sequence variation in trans-regulatory loci can change target specificities or expression levels of trans-acting factors which then changes the expression levels of the target loci. Remodeling the chromatin structure on top of the genes and genetic elements in part can affect the interactions among cis- and trans-regulatory factors without changing the DNA. 2.2. Dynamics of chromatin remodeling in response to environmental stimuli Chromatin is remarkably dynamically modulated in response to pathogen infection and environmental changes. The signaling networks induce chromatin modifiers that facilitate the fast gene expression changes involved in defensive systems. For example, virus-derived double-strand RNAs in plants will activate small RNA generation and RNA-dependent DNA methylations, and eventually silence the virus-derived RNA transcription [6]. In Arabidopsis and rice, pathogen infection and salicylic acid (SA) hormone can induce DNA methylation changes around genes involved in biotic stress responses [15]. These examples show that DNA methylation can change as fast as gene expression changes in response to pathogen infection. However, the kinetics of chromatin remodeling has not been well investigated. In Arabidopsis, the regulation of flowering in response to environmental and developmental cues is accompanied by chromatin remodeling (reviewed in [16]). In Arabidopsis, cold temperature in winter induces expression of VIN3, a histone methyltransferase resulting in tri-methylations at Lysine 27 of H3 histone (H3K27me3) and methylations at Lysine 9 of H3 histone (H3K9me). VIN3 induction by cold increases H3K27me3 levels at the 5 end of the FLC gene, a repressor of flowering [17]. FLC is repressed by H3K27me3, and this accelerates flowering after cold treatment [18]. Even after the plants are removed from the cold state, the chromatin remodeling states repressing FLC expression are stably maintained. 2.3. Encryption of parental environmental signals in descendant’s genomic DNA Parental environmental factors can affect the embryo chromatin structure and the morphology of progeny during development. In mice, high exposure to the mother’s xenobiotic chemical bisphenol A (BPA) leads to the fetus’s hypo-methylation of the cytosines at Agouti gene [19]. This hypo-methylation increases the risk of chronic adult diseases during development of the progeny [20]. In Arabidopsis seeds, embryo-nourishing endosperm genome is demethylated by DEMETER (DME) but embryo genome is hyper methylated [21,22]. This leads to silencing of transposable elements

M. Ha / Plant Science 211 (2013) 137–145

and maternal imprinting in embryo development [23]. These examples suggest that chromatin remodeling mediates the encryption of parental environmental signals in the descendant’s genomic DNA without changing the primary DNA sequence. In other words, chromatin remodeling may be an efficient way to reversibly control gene expression of the progeny in order to prepare for local or temporal environmental changes. Chromatin remodeling is transmitted to the progeny by DNA replication processes and the activation of chromatin modifiers at the memorized conditions. In Arabidopsis, methylations of CGs and CHGs are maintained through DNA replication by MET1, a homologue of the mammalian DNA methyltransferase DNMT1, and the plant-specific CMT3 methylatransferase, respectively [24–26]. In plants, hybrid formation by a combination of divergent genomes gives rise to phenotypic variations and hybrid vigor. However, McClintock predicted that the presence of divergent genomes in the same nucleus can result in genomic shock and infertility by rampant activation of TEs [27]. Indeed, DNA demethylations at reactivated TEs and abnormal imprinting sites have been observed in interspecific mammalian hybrids which are infertile [28]. In Arabidopsis, allopolyploids stably maintaining divergent sets of genomes inherently generate siRNAs repressing activities of TEs from both parental genomes [29]. These results suggest that chromatin remodeling is an essential mechanism for proper inheritance of the genome. The parental chromatin remodeling can be non-additively inherited to the progeny. Preferential expressions of chromatin remodeling machineries from one parent to the other induce biased chromatin remodeling. In Arabidopsis, predominant expression of RNA polymerase IV (PolIV) from maternal gametophyte leads to the generation of siRNAs from maternal chromosomes during endosperm development [30].

3. Analytics for inference of chromatin remodeling codes The specific arrangement of chromatin remodeling marks facilitates interaction with distinct protein factors and affects transcription and DNA replication. Using the energy from hydrolysis of ATP, ATP-dependent chromatin remodelers enforce ejection or formation of nucleosomes or chromatin modifications [4]. The chromatin remodelers contain DNA-dependent ATPase domains conserved among yeasts, plants and animals. However, their distinct compositions of protein-interacting domains specialize in their functions [5]. The chromatin remodelers induce distinct arrangement of nucleosomes and interact with proteins inducing incorporation of histone variants, histone modifications, or DNA methylations. In Arabidopsis and C. elegans, DNA methylations at 5 end of transposable elements and genes are recognized by MORC family ATPases [31]. MORC family ATPases induces heterochromatin condensation and silencing of the transposable elements and genes. However, DNA methylations in the middle of coding regions are preferentially observed in transcribed genes, suggesting that DNA methylations at the two ends of coding regions are deleterious to transcription in Arabidopsis [32]. Origins of DNA replication and DNA repair sites are associated with distinct arrangements of chromatin remodeling marks [33]. Enrichment of H3K27me1 at heterochromatin by the action of ATXR5 and ATXR6 prevents re-replication of DNA [34]. The transcription levels and their expression specificities can be predicted from combination and specific localizations of histone modifications, histone variants, and DNA methylation levels in coding regions [35,36]. Therefore, the distinct coordination of chromatin remodeling marks may serve as a chromatin remodeling code indicating gene expression patterns and function of genetic elements including DNA recombination

139

sites, origin of DNA replications, and transcription regulatory elements. The chromatin remodeling code is hypothesized to be a specific arrangement of chromatin remodeling marks affecting the function of the underlying genetic elements. The chromatin remodeling code hypothesis is not a new concept. The basic idea was published in a series of classic papers as histone code hypothesis by Allis and the colleagues [37]. The histone code hypothesis states that specific histone modifications recruit proteins to the genomic sites. DNA methylations and RNAs are also important processes of chromatin remodeling. Therefore, considering DNA methylations, RNAs, histone modifications, histone variants, and nuclesome, the chromatin remodeling code can be comprehensively analyzed. Furthermore, the advancement of ChIP-seq and ChIP-chip technologies allows an unbiased investigation of the chromatin remodeling landscape. The chromatin remodeling codes may be inferred from integrative analyses of whole genome measurement of the chromatin remodeling marks. Here, the chromatin remodeling code entails arrangement and combination of chromatin remodeling marks regulating, in part, the function of the genetic elements. The genetic elements can be genes, cis-regulatory elements, DNA recombination sites, or origin of DNA replication. Advances in next-generation technology, ChIP-seq, and ChIP-chip experiments allow us to comprehensively analyze the relationship among chromatin remodeling processes and the features associated with the activity of the genetic elements underlying the chromatin structure. Due to the increasing volume and complexity of data, e.g., for ChIP-seq and RNA-seq, computational analysis is essential for extracting biologically meaningful information (Fig. 1). Computational analyses address four major questions. First, they integrate heterogeneous genomic and epigenetic data generated from different methods and platforms. Second, by integrating various ChIP data, they aim to identify the correlative, antagonistic, or sequential relationships among chromatin modification processes. Third, they aim to find the features of genetic elements associated with distinct chromatin remodeling architectures. Fourth, they help to construct predictive models of genome regulation. On the other hand, the availability of thorough measurements of chromatin remodeling in the whole genome allow us to examine the biological principles we previously inferred from only a handful of exemplary loci. 3.1. Integration of ChIP-seq and ChIP-chip data ChIP-seq and ChIP-chip are commonly used to measure the genome-wide levels of chromatin modification marks. In the ChIP-chip experiment, the DNA extracted from chromatin immunoprecipitation with antibodies binding to a specific chromatin remodeling mark is labeled with dye. The DNA labeled with dye is then hybridized to a microarray of complementary DNA probes. After washing DNAs not hybridized, the dye signal from hybridized DNA is measured. The dye intensity represents the enrichment of the chromatin remodeling marks. In the ChIP-seq experiment, the DNA extracted from the chromatin immunoprecipitation is directly sequenced by next-generation sequencing technology. The number of sequencing reads or sequencing frequency represents the enrichment of the chromatin remodeling marks at a locus. A normalization procedure is required to integrate the heterogeneous data of the genome-wide measurements of histone modifications, nucleosomes, and histone variants from independent ChIP-seq and ChIP-chip experiments. The raw intensities of chromatin remodeling marks observed from ChIP-seq data are variable by sequencing depth of an experiment and technology, whereas the ChIP-chip signals are variable by dye effect and hybridization efficiency. To standardize heterogeneous ChIP-seq and ChIP-chip data and make the two different platforms comparable, the distribution of occupation levels in an experiment should be adjusted to the

140

M. Ha / Plant Science 211 (2013) 137–145

Fig. 1. Computational analyses of chromatin remodeling. Using computational analyses, genome-wide measurements of chromatin remodeling marks by ChIP-seq and ChIPchip can be integrated. From the integrated data, correlative analyses can identify relationships among the chromatin remodeling processes. Clustering and classification analyses can be used to identify distinct features of chromatin remodeling profiles associated with the activities of genetic elements.

similar distribution. The commonly used normalization method is z-score normalization, which standardizes the whole genome measurement of a chromatin remodeling mark so that the mean intensity level in a genome is 0 and the standard deviation is 1. Another common method calculates enrichment values compared to the average signal of the whole genome. In this method, the loci enriched with a chromatin mark are assigned with values greater than 1 and the depleted regions with less than 1. The number of ChIP-seq mapped on a genomic region shows Poisson distribution, in which most of the genomic region matches to at least one ChIP-seq read. Genomic regions matched with many ChIP-seq reads greater than the average of the whole genome are rare. Poisson distribution has been successfully applied for detecting significantly enriched regions [38]. 3.2. Examination of relationships among chromatin remodeling marks To understand the cooperative, antagonistic, or sequential relationships among chromatin remodeling marks, the marks themselves can be compared in a pair-wise manner. The correlative analyses can provide general and reproducible dependency relationships among chromatin remodeling processes. The significance of correlative relationships between two populations can be measured using correlation coefficients which calculate

the deviation from the independence of two random variables. The most familiar measure of dependency between two populations is the Pearson product-moment correlation coefficient which assumes that two populations are Gaussian distribution with the frequency centered on their means. In probability theory, the central limit theorem states that a mean of a sufficiently large number of independent random variables, each with a well-defined mean and well-defined variance, will be approximately normally distributed. In ChIP-seq and ChIP-chip data, the chromatin remodeling signal is measured in whole genome from a large number of cell populations. Since the ChIP-seq signal at the locus would be normally distributed according to the central limit theorem, the Pearson correlation coefficient is a reasonable measure of the dependencies between two chromatin marks. Note that Spearman’s and Kendall’s tau correlation coefficients generally can be used to measure dependency among chromatin remodeling marks, because both of these non-parametric statistics do not assume specific distribution of populations. The correlation coefficient ranges from −1 to 1, i.e., −1 for negative correlation, 0 for no correlative relationship and 1 for positive correlation. The correlation should be reproducible from many observations, because the significance of the correlation coefficient depends on both the correlation coefficient values and the number of pair-wise measurements.

M. Ha / Plant Science 211 (2013) 137–145

Correlative analyses remain popular, because these methods effectively find the general relationships among various chromatin remodeling processes. Moreover, the findings from correlative analyses complement the findings of genetic and biochemical studies. 3.2.1. Nucleosome and protein bindings Genome processing including transcription and DNA replication requires separating DNA strands from the nucleosomes for the interaction of DNA with protein factors [39,40]. Observed in cytology, heterochromatin stains intensely due to the condensed package of DNA, whereas euchromatin stains less intensely due to forming a loose and open chromatin structure. Euchromatin is enriched with coding region and actively transcribed, whereas heterochromatin is repeat-rich, non-coding-rich, and rarely transcribed. Genome-wide measurements of nucleosome occupation show that nucleosome is relatively evenly distributed across genome and in euchromatin and heterochromatin [37]. This suggests that one function of nucleosome is packaging genome and forming beads on a string chromatin structure. On the other hand, within the coding region nucleosome distribution is locally depleted by the binding of protein factors. Nucleosome is depleted at transcription start sites (TSSs), whereas RNA Pol II binding is enriched at TSS. The fact that transcription factor binding sites are hyper-sensitive to DNase I suggests that protein binding and nucleosome formation are anti-correlated [41]. 3.2.2. Histone modifications According to histone code hypothesis, the composition of various histone modifications determines euchromatin or heterochromatin structure [37]. The histone code hypothesis also proposes that coordinations of histone modifications determine gene expression and cellular states. In Arabidopsis, euchromatin is enriched with H3K4me1, H3K4me2, H3K4me3, H3K9me3, H3K27me3, H3K36me3, H2BUb, and histone acetylations [42] and the heterochromatin is enriched with DNA methylations, H3K9me2, and H3K27me1 in Arabidopsis [34,42,43]. Also in Arabidopsis, the acetylation of histones H3 and H4 and di and tri-methylations of H3 lysine 4 (H3K4me2 and H3K4me3), but not H3K4me1 [44], di- and tri-, but not mono-methylation of H3K36 [45], and H2B mono-ubiquitylation [42] are positively correlated with transcription levels, whereas H3K27me3, H2A mono-ubiquitylation, and DNA methylations are negatively correlated with transcription levels, implying association with gene repression [42,46,47]. Comprehensive genome-wide analyses of chromatin remodeling marks show that H3K4me3, H3K9ac, and H2A.Z are correlatively arranged [48]. H2A.Z and H3K4me3 precede H3K9ac deposition. H3K4me3 is recognized by histone acetylases and histone deacetylases complexes [49]. In particular, H3K4me3 is enriched at TSS. The transcription initiation complex includes histone acetylases which bind to H3K4me3 and attach H3K9ac. Therefore, H3K4me3 and H3K9ac intensity at coding regions indicates the transcription levels of the genes. 3.2.3. DNA methylations mediated by small RNAs and histone modifcations In eukaryotic genomes, DNA methylation is deposited at CG, CHG, and CHH sequences and regulates transposon silencing, genomic imprinting and stable gene silencing. It is not well understood how DNA methylations and histone modifications are regulated in a locus-specific way. The 21 to 24nt length small interference RNAs (siRNAs) have been shown to guide DNA methylation machineries to target sites via the RNA-directed DNA methylation (RdDM) pathway [47]. In the Arabidopsis RdDM pathway, the siRNA generation requires the action of the putative DNA-directed

141

RNA polymerase Pol IV, RDR2 (RNA-dependent RNA polymerase 2), and DCL3 (Dicer-like 3) [50]. The siRNAs are loaded to Argonaute proteins (AGO4/AGO6) and direct DNA methylations at the complementary genomic DNA sequences most likely through base-pairing between siRNAs and the RNAs transcribed by Pol V [51,52]. The DRM2 DNA methyltransferase execute de novo DNA methylations in the Arabidopsis genome within all sequence contexts via the RNAdirected DNA methylation (RdDM pathway). The proportion of genomes carrying DNA methylation at a cytosine can be measured using Bisulfite sequencing or methyl cytosine binding ChIP-seq. The genome-wide analyses of DNA methylation and small RNA generation suggest significant correlation among DNA methylation, small RNA generation, and RNA Pol V binding at the same locus. However, the determinants locating PolV to the specific repeat elements are still unknown. Cytosine methylation tendency is correlated with nucleosome occupation levels. The 10 bp periodicity of cytosine methylations in nucleosomal DNAs [53] suggests that nucleosomal architecture is recognized by DNA methylase complexes. According to mutant studies and genome-wide analyses, DNA methylation and H2A.Z are deposited to be mutually exclusive [54,55]. DNA methylation deposition is also mutually exclusive to H3K4me3. In mice, antagonistic localization of DNA methylations and H3K4me3 are regulated by DNA sequence composition at the locus. All of these observations suggest that chromatin modifiers recognize the chromatin remodeling marks and interact together. Co-regulation of DNA methylations and histone modifications is an integral mechanism of chromatin remodeling. Genome-wide analyses of DNA methylations and histone modifications show correlations of DNA methylations with H3K27me3, H3K9me3, and H2B monoubiquitination, but anti-correlations with H3K4me3, H3 acetylations and H2A.Z [55]. Moreover, both histone variants H3.3 and H2A.Z occupy nucleosome-free regions [56]. In Arabidopsis, the SUVR4 histone lysine methyltransferase recognizes DNA methylation and ubiquitin at H2B and converts H3K9me1 to H3K9me3 to repress transposon and peudogenes [57]. In yeasts, histone ubiquitylation at H2B lysine 34 recruits histone methyltransferases and increases methylations at H3K4 and H3K79 and acetylations at H3K16 [58,59]. Again in Arabidopsis, H3K4me3 and H3K9ac show correlative distribution across the genome. The H2BUb, H3K4me3, and H3K9ac mark actively transcribed genes. 3.2.4. Chromatin dynamics Occupation of H3.3 and H2A.Z and depletion of nucleosomes is an oxymoron. H2A.Z and H3.3 preferentially carry the histone modifications H3K4me3 and H3K9ac associated with a nucleosome-free region. The histone variants and some histone modifications may provide dynamic nucleosome exchange. However, the kinetics of nucleosomes and the mechanisms controlling the dynamics of nucleosome formation require further examination. To date, only the incorporation kinetics of H2A.Z and H3.3 into genome has been measured in yeasts [60] and flies [61]. However, the dynamics of incorporation and displacement of nucleosomes simultaneously has not been shown. In histone modifications, reversible histone acetylations and deacetylations are important for variable gene regulation, suggesting that nucleosome dynamics may accompany reversible occupations of specific histone variants and histone modifications. 3.3. Clustering and classification analyses of chromatin remodeling map The distinct arrangements of chromatin remodeling marks can affect specific functions of genetic elements. Typically, clustering and classification analyses are used to understand the combinatory depositions.

142

M. Ha / Plant Science 211 (2013) 137–145

Clustering algorithms partition the chromatin modification measurements into groups to maximize similarity within a group and maximize difference among groups. The cluster analyses can address if the chromatin remodeling profile consists of a set of distinct classes each representing substantially different properties. The distinct classes can be extracted via cluster analyses of chromatin modifications in the whole genome. Often, each class can be assigned to a genetic element or a genetic function by enrichment analyses. The annotation of activity of genetic elements by chromatin remodeling profiles has been shown to be more accurate than sequence-based analyses. For example, to characterize chromatin remodeling status in whole genome including coding and non-coding region, Ernst et al. perform genome-wide clustering analyses of 9 chromatin marks in 9 different cell types [62]. The resulting 15 chromatin states can be matched to distinct activities of genetic elements: active, weak and poised promoters, strong and weak enhancers, coding regions with high and low transcription levels, and heterochromatic and silenced states. In Arabidopsis, the chromatin remodeling profile in the coding region can be categorized into several classes [35]. In leaves, the sharp and high peaks of H3K4me3 and H3K9ac, but the depletion of H3K27me3, H3K9me2, and DNA methylations, are associated with constitutive high expression of genes, whereas the broad and high peaks of H3K4me3 and H3K9ac are associated with leafspecific high expression. Those genes marked by broad peaks of H3K4me3 and H3K9ac are preferentially involved in leaf-specific biological processes such as photosynthesis. Again, in leaves, depletion of H3K4me1, 2 and 3, H3K9ac, but enrichment of H3K27me3 in the coding region, is associated with leaf-specific repressed genes. In mice, broad distribution of H3K4me3 and H3K79me2 at TSS is associated with cell-type specific up-regulation, whereas dense narrow H3K4me3 deposition is associated with constitutive high expression in hair follicle stem cell differentiation [63]. Altogether, interpreting chromatin modification marks by transcription regulatory machineries is highly conserved among plants and mammals. In classification analysis, the property of a locus can be assigned to a class by its chromatin remodeling status. In particular, cisregulatory elements interacting with transcription factors function independently of orientation and distance from target genes. Therefore, it is difficult to identify cis-regulatory elements by DNA sequence analysis. Often, clustered and separated chromatin profile identified from clustering analysis can serve as efficient classifiers of genetic elements or activity of regulatory elements. Therefore, the distinction blurs between classification and clustering analyses. Classification analyses of chromatin remodeling marks also are efficient ways to systemically identify features of genetic elements. Heintzman et al. find that distant-acting enhancer elements, which are the cis-regulatory elements controlling condition-specific transcription of target genes, are preferentially occupied by H3K4me1, whereas the promoter is enriched with H3K4me3 occupation [64]. Using the chromatin remodeling feature of the enhancer element, they identify novel enhancer elements from H3K4me1 occupation level in the human genome. Condition-specifically active enhancer elements show significant enrichment of H3K4me2 before binding of transcription factors, and the H3K4me2 dissociates when the loci are bound to transcription factors [65]. Using the dynamics of H3K4me2 occupation levels in the transcription factor activation as features of enhancer elements, the latter authors also identify novel enhancer elements. Creyghton et al. [66], who apply classification analyses to identify cell-type specifically active enhancer elements, find the presence of both H3K27ac and H3K4me1 marks in the active enhancer element in the cell, whereas enhancers poised for activation in other conditions are marked by only H3K4me1 but are depleted of H3K27ac. Using the defined feature, the cis-regulatory

elements active only at specific cell types have been identified and validated by experiments. In Arabidopsis, the origin of DNA replication is marked by enrichment of H2A.Z, H3K4me2, H3K4me3, and H3K5ac, and depletion of H3K4me1 and H3K9me2 [33]. The recombination hot spots are modulated by DNA methylations and histone modifications [67]. In mammals, recombination host spots are preferentially enriched with H3K4me3. PRDM9 (PR domain zinc finger protein 9), a meiosis-specific histone mehtyltransferase in mice and humans, specifically trimethylates H3K4, but does not have the ability to generate H3K4me1 or H3K4me2. PRDM9 specifically adds H3K4me3 tags to meiotic homologous recombination and the process is essential for meiotic homologous recombination [68]. In mammals, the differential DNA methylations and H3K4me3 at exon-intron junction induce alternative splicing variations [69]. Alternative splicing is the differential inclusion of exons when processing messenger RNAs. Alternative splicing increases diversity of proteins from a gene. The dynamic regulation of histone modifications, especially histone acetylation and deacetylation, is a main feature of genes differentially expressed within and between species. Among species, variation of histone modifications can lead to expression change and species specific morphology. Among Arabidopsis species, differentiation of H3K4me3 and H3K4me2 levels at FLC loci vary transcription levels which leads to variegation of flower timing [70]. In hybrid species, variegation of H3K4me2 levels at circadian clock responsive transcription factor genes is correlated with differential expression of their target genes involved in photosynthesis which leads to differential energy process efficiency [71]. 3.4. Probabilistic modeling of chromatin remodeling The most probable chromatin remodeling profile can be predicted based on the correlative and antagonistic relationships among chromatin remodeling marks. A Hidden Markov Model (HMM) can be applied to understand the combinatory occupation of various histone modification and chromatin binding proteins from a subset of histone modifications. The probabilistic model calculates the probabilities of the possible states of an object based on the observations. The relationship linking observations and possible hidden states can be estimated from the datasets and optimized to maximize the likelihood. For example, HMM is used to predict the combinatory occupation of various histone modifications and chromatin binding proteins from a subset of histone modifications obtained from ChIP experiments [62]. Their hidden states are the occupation of various histone modifications and the observations are a subset of the occupancies of histone modifications. Associations between two histone modifications can be estimated by pair-wise correlation analyses among the chromatin modification marks. Based on the genome-wide dependency of the two chromatin remodeling marks, the probability of the unobserved chromatin remodeling mark can be predicted from the measurement of the other marks. Integration of more chromatin remodeling marks and analyses of the comprehensive data will lead to finding novel interactions and more details about the coordinated regulation of chromatin structure [72]. 4. Chromatin remodeling signature in a primary genome sequence 4.1. DNA sequence features of chromatin remodeling Constructing specific arrangements of histone modifications and DNA methylations is achieved by coordination among chromatin remodelers, interacting transcription factors, and targeting genome sequence composition and DNA methylation status.

M. Ha / Plant Science 211 (2013) 137–145

143

Fig. 2. Genetic and epigenetic factors in chromatin remodeling. Genomic DNA sequence context and conditional activities of chromatin remodelers act together on chromatin remodeling.

Even though nucleosome is relatively evenly distributed in the genome, the bending property of DNA sequences will affect the nucleosome formation. The energy cost for DNA to wrap around histone octamers varies by their compositions of base-pair [73]. The intrinsic flexibility of a DNA segment has been shown to facilitate its wrapping around histone octamers, whereas consecutive A/Ts such as AAAAAAA and TTTTTT form rigid DNA segments. The flexible DNA base-pair compositions are preferentially found in nucleosome, whereas the rigid DNA segments are in nucleosome-free regions [74,75]. The specific localization of chromatin remodeling marks on a genome implies distinct relationships of DNA sequence with chromatin modifications from nucleosomes. Histone methylations and acetylations such as H3K4m1, H3K4me2, H3K4me3, and H3K9ac are anti-correlated with nucleosome formation levels. Chromatin modifiers responsible for histone methylations and acetylations interact with sequence-specific protein factors and tag H3K4me3, H3K27me2, or histone acetylations at specific loci. Eukaryotic transcription initiation complex, which contains TATA-binding proteins and histone acetylases, leads to histone acetylation at the promoter. In mice, inserting non-methylated CpG-rich sequences leads to a high density of H3K4me3 occupation. In mice and humans, Cfp1 specifically binds to non-methylated CpG-rich sequences and recruits the Setd1, H3K4me3 histone methylase complex [76]. Chromatin remodelers can recognize target sites by their own sequence specificities. For example, PRDM9, a zinc finger protein causing H3K4me3, specifically binds to a GC-rich 13mer motif and marks recombination hotspots [77–79]. These results suggest that chromatin remodeling is affected by primary DNA sequence composition and DNA methylation status. Opinions differ about whether H3 acetylation neutralizes the positive charge of histone lysine residues and weakens the chargedependent interactions between histones and DNA [80]. According to nucleosome crystal structure, H3 tail does not directly interact with DNA. However, the biochemical effect of H3 histone tail modification on interaction with DNA remains to be verified. 4.2. Predicting the chromatin remodeling map from a genome sequence The use of ChIP-seq data in various cell types allows the primary DNA sequence features of chromatin remodeling to be investigated. Although the localization tendency of H3K4me3 is differentially

regulated among cell types, the 6mer primary DNA sequence preference over background in H3K4me3 nucleosomes is consistently maintained [81]. This suggests that primary DNA sequence signatures may be associated with the histone modification landscape. The sequence specificities of histone modifications are differentiated from other nucleosomes. Based on the distinct and consistent sequence specificities of H3K4me3 and other nucleosomes, a probabilistic map of H3K4me3 and other nucleosome occupation has been calculated from a reference genome sequence [81]. The formation of a H3 or H3K4me3 nucleosome at a bp can be affected by adjacent sequences and adjacent nucleosomes. Therefore, a probabilistic (HMM) model calculates the probabilities of H3 and H3K4me3 nucleosome formations at a bp considering the whole genome sequence. Here, reference genome sequence and the probability of modified nucleosome occupation at a given sequence segment are measured using ChIP-seq analyses. Using the sequence specificities and a genome sequence, the probabilities of hidden states, and the occupation of H3, H3K4me3, and sites not carrying nucleosomes can be estimated. The probability that a bp forms a modified nucleosome can be calculated considering all of the possible arrangements of modified nucleosomes from the first bp to the target bp (forward procedure) combined with the sequence composition at the target site, and then calculating all of the possible arrangements from the last bp to the target bp (backward procedure). A comparison of the predicted maps and experimental measurements shows high accuracy of the prediction model from a primary genome sequence, suggesting a significant association of the primary genome sequence context with chromatin remodeling. The sequence context affecting chromatin remodeling may be localized near the target loci or distant from the target sites carrying the chromatin remodeling. For example, since the TATAbox and consecutive A-tracts at the promoter are disfavored by both H3K4me3 and other nucleosomes, the promoter or upstream of TSS tends to be nucleosome-free and the 5 end of genes can start with H3K4me3 nucleosomes. Since the location of highly favored sequences of H3K4me3 may attract histone methylases for H3K4me3 deposition, the less-favored sequences in some proximity cannot form H3K4me3 nucleosomes due to spatial hindrances among the adjacent nucleosomes. The upstream regions of differentially expressed genes among species show both high sequence divergence and different histone modification patterns [70]. Examining the association of the histone modifications and the DNA sequence profile may help to

144

M. Ha / Plant Science 211 (2013) 137–145

translate the DNA sequence information directing the specific chromatin remodeling and regulatory elements which are difficult to predict only from DNA sequence profiles. The predictive model of chromatin remodeling from a genome sequence context which does not fully explain the dynamic modulation of chromatin structure implies the need to improve the current predictive models. On the other hand, the chromatin remodeling process is regulated by both sequencedependent factors and sequence-independent, epigenetic factors (Fig. 2). 5. Conclusions and perspectives This review discussed systematic approaches to decipher chromatin codes from the coordination of chromatin marks and primary DNA sequences. The sequence feature and feasibility of predicting a chromatin remodeling map from a primary DNA sequence suggest that the epigenetic and genetic factors cooperate for chromatin structural modulation and genome regulation. Genome-wide association studies (GWASs) have been intensively performed to identify the genetic markers causing disease and traits in humans and crops. However, genetic variations affecting trait are not easily detected [82]. The possible factors obscuring the effects of important genetic factors are penetrance and epistasis. Penetrance is the ability to express genetic variants as phenotypes. In epistasis, the modifier of the causal genes leads to the morphology. As chromatin remodeling processes are important barriers of activity of genetic elements, analyzing chromatin remodeling may resolve the epigenetic factors to model phenotypic variations. Moreover, chromatin remodeling, which provides more accurate models of gene expression and morphological variations, may help to find the biological marks that cannot be detected by GWAS or genetic study. In human health and crop technology, the following research is likely to be explored in the near future: (1) Diagnosis and prognosis of traits using chromatin remodeling profiles; the current and the time-course activity of disease/trait-causing genetic elements can be identified by characterizing chromatin remodeling. (2) The primary genome sequence-encoded chromatin remodeling code can allow more rapid diagnosing of the intrinsic defects of chromatin. (3) Identifying the DNA sequence context regulating gene through chromatin remodeling can provide efficient methods to control traits. Acknowledgments The author thanks colleagues in the Well Aging Research Center, for their valuable discussion and comments. This work is supported by a grant from the Samsung Advanced Institute of Technology to the author. References [1] S. Feng, S.E. Jacobsen, W. Reik, Epigenetic reprogramming in plant and animal development, Science 330 (2010) 622–627. [2] P.B. Talbert, S. Henikoff, Histone variants–ancient wrap artists of the epigenome, Nature reviews, Mol. Cell Biol. 11 (2010) 264–275. [3] R.B. Deal, S. Henikoff, Histone variants and modifications in plant gene regulation, Curr. Opin. Plant Biol. 14 (2011) 116–122. [4] C.R. Clapier, B.R. Cairns, The biology of chromatin remodeling complexes, Annu. Rev. Biochem 78 (2009) 273–304. [5] L. Knizewski, K. Ginalski, A. Jerzmanowski, Snf2 proteins in plants: gene silencing and beyond, Trends Plant Sci. 13 (2008) 557–565. [6] D. Baulcombe, RNA silencing in plants, Nature 431 (2004) 356–363.

[7] I.R. Henderson, S.E. Jacobsen, Epigenetic inheritance in plants, Nature 447 (2007) 418–424. [8] C.H. Waddington, The epigenotype, Endeavour 1 (1942) 18–20. [9] S.L. Berger, T. Kouzarides, R. Shiekhattar, A. Shilatifard, An operational definition of epigenetics, Genes & Development 23 (2009) 781–783. [10] S.Y. Roth, J.M. Denu, C.D. Allis, Histone acetyltransferases, Annu. Rev. Biochem. 70 (2001) 81–120. [11] R. Pandey, et al., Analysis of histone acetyltransferase and histone deacetylase families of Arabidopsis thaliana suggests functional diversification of chromatin modification among multicellular eukaryotes, Nucleic Acids Res. 30 (2002) 5036–5055. [12] J.E. Brownell, C.D. Allis, Special HATs for special occasions: linking histone acetylation to chromatin assembly and gene activation, Current opinion in genetics & development 6 (1996) 176–184. [13] P.J. Wittkopp, B.K. Haerum, A.G. Clark, Evolutionary changes in cis and trans gene regulation, Nature 430 (2004) 85–88. [14] X. Shi, et al., Cis- and trans-regulatory divergence between progenitor species determines gene-expression novelty in Arabidopsis allopolyploids, Nat. Commun. 3 (2012) 950. [15] R.H. Dowen, et al., Widespread dynamic DNA methylation in response to biotic stress, Proc. Natl. Acad. Sci. U. S. A. 109 (2012) E2183–E2191. [16] E.S. Dennis, W.J. Peacock, Epigenetic regulation of flowering, Curr. Opin. Plant Biol. 10 (2007) 520–527. [17] S. Sung, R.M. Amasino, Vernalization in Arabidopsis thaliana is mediated by the PHD finger protein VIN3, Nature 427 (2004) 159–164. [18] R. Bastow, et al., Vernalization requires epigenetic silencing of FLC by histone methylation, Nature 427 (2004) 164–167. [19] D.C. Dolinoy, D. Huang, R.L. Jirtle, Maternal nutrient supplementation counteracts bisphenol A-induced DNA hypomethylation in early development, Proc. Natl. Acad. Sci. U. S. A. 104 (2007) 13056–13061. [20] R.L. Jirtle, M.K. Skinner, Environmental epigenomics and disease susceptibility, Nat. Rev. Genet. 8 (2007) 253–262. [21] T.F. Hsieh, et al., Genome-wide demethylation of Arabidopsis endosperm, Science 324 (2009) 1451–1454. [22] M. Gehring, K.L. Bubb, S. Henikoff, Extensive demethylation of repetitive elements during seed development underlies gene imprinting, Science 324 (2009) 1447–1451. [23] J.H. Huh, M.J. Bauer, T.F. Hsieh, R.L. Fischer, Cellular programming of plant gene imprinting, Cell 132 (2008) 735–744. [24] M.W. Kankel, et al., Arabidopsis MET1 cytosine methyltransferase mutants, Genetics 163 (2003) 1109–1122. [25] A.M. Lindroth, et al., Requirement of CHROMOMETHYLASE3 for maintenance of CpXpG methylation, Science 292 (2001) 2077–2080. [26] H. Saze, O. Mittelsten Scheid, J. Paszkowski, Maintenance of CpG methylation is essential for epigenetic inheritance during plant gametogenesis, Nat. Genet. 34 (2003) 65–69. [27] B. McClintock, The significance of responses of the genome to challenge, Science 226 (1984) 792–801. [28] P.B. Vrana, et al., Genetic and epigenetic incompatibilities underlie hybrid dysgenesis in Peromyscus, Nat. Genet. 25 (2000) 120–124. [29] M. Ha, et al., Small RNAs serve as a genetic buffer against genomic shock in Arabidopsis interspecific hybrids and allopolyploids, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 17835–17840. [30] R.A. Mosher, et al., Uniparental expression of PolIV-dependent siRNAs in developing endosperm of Arabidopsis, Nature 460 (2009) 283–286. [31] G. Moissiard, et al., MORC family ATPases required for heterochromatin condensation and gene silencing, Science 336 (2012) 1448–1451. [32] D. Zilberman, M. Gehring, R.K. Tran, T. Ballinger, S. Henikoff, Genome-wide analysis of Arabidopsis thaliana DNA methylation uncovers an interdependence between methylation and transcription, Nat. Genet. 39 (2007) 61–69. [33] C. Costas, et al., Genome-wide mapping of Arabidopsis thaliana origins of DNA replication and their associated epigenetic marks, Nat. Struct. Mol. Biol. 18 (2011) 395–400. [34] Y. Jacob, et al., ATXR5 and ATXR6 are H3K27 monomethyltransferases required for chromatin structure and gene silencing, Nat. Struct. Mol. Biol. 16 (2009) 763–768. [35] M. Ha, D.W. Ng, W.H. Li, Z.J. Chen, Coordinated histone modifications are associated with gene expression variation within and between species, Genome Res. 21 (2011) 590–598. [36] B.D. Strahl, C.D. Allis, The language of covalent histone modifications, Nature 403 (2000) 41–45. [37] T. Jenuwein, C.D. Allis, Translating the histone code, Science 293 (2001) 1074–1080. [38] C. Zang, et al., clustering approach for identification of enriched domains from histone modification ChIP-Seq data, Bioinformatics 25 (2009) 1952–1958. [39] L. Bintu, The elongation rate of RNA polymerase determines the fate of transcribed nucleosomes, Nat. Struct. Mol. Biol. (2011). [40] B. Li, M. Carey, J.L. Workman, The role of chromatin during transcription, Cell 128 (2007) 707–719. [41] W. Zhang, T. Zhang, Y. Wu, J. Jiang, Genome-Wide identification of regulatory DNA elements and protein-binding footprints using signatures of open chromatin in Arabidopsis, Plant Cell 24 (2012) 2719–2731. [42] F. Roudier, et al., Integrative epigenomic mapping defines four main chromatin states in Arabidopsis, EMBO J. 30 (2011) 1928–1938.

M. Ha / Plant Science 211 (2013) 137–145 [43] Y. Jacob, et al., Regulation of heterochromatic DNA replication by histone H3 lysine 27 methyltransferases, Nature 466 (2010) 987–991. [44] X. Zhang, Y.V. Bernatavichute, S. Cokus, M. Pellegrini, S.E. Jacobsen, Genomewide analysis of mono-, di- and trimethylation of histone H3 lysine 4 in Arabidopsis thaliana, Genome Biol. 10 (2009) R62. [45] L. Xu, et al., Di- and Tri- but not monomethylation on Histone H3 Lysine 36 marks active transcription of genes involved in flowering time regulation and other processes in Arabidopsis thaliana, Mol. Cell. Biol. 28 (2008) 1348–1360. [46] F. Bratzel, G. Lopez-Torrejon, M. Koch, J.C. Del Pozo, M. Calonje, Keeping cell identity in Arabidopsis requires PRC1 RING-finger homologs that catalyze H2A monoubiquitination, Curr. Biol. 20 (2010) 1853–1859. [47] A. Molnar, et al., Small silencing RNAs in plants are mobile and direct epigenetic modification in recipient cells, Science 328 (2010) 872–875. [48] Z. Wang, et al., Combinatorial patterns of histone acetylations and methylations in the human genome, Nat. Genet. 40 (2008) 897–903. [49] Z. Wang, et al., Genome-wide mapping of HATs and HDACs reveals distinct functions in active and inactive genes, Cell 138 (2009) 1019–1031. [50] M. Matzke, T. Kanno, L. Daxinger, B. Huettel, A.J. Matzke, RNA-mediated chromatin-based silencing in plants, Curr. Opin. Cell Biol. 21 (2009) 367–376. [51] X.J. He, et al., An effector of RNA-directed DNA methylation in arabidopsis is an ARGONAUTE 4- and RNA-binding protein, Cell 137 (2009) 498–508. [52] A.T. Wierzbicki, T.S. Ream, J.R. Haag, C.S. Pikaard, RNA polymerase V transcription guides ARGONAUTE4 to chromatin, Nat. Genet. 41 (2009) 630–634. [53] R.K. Chodavarapu, et al., Relationship between nucleosome positioning and DNA methylation, Nature 466 (2010) 388–392. [54] M.L. Conerly, et al., Changes in H2A.Z occupancy and DNA methylation during B-cell lymphomagenesis, Genome Res. 20 (2010) 1383–1390. [55] D. Zilberman, D. Coleman-Derr, T. Ballinger, S. Henikoff, Histone H2A.Z and DNA methylation are mutually antagonistic chromatin marks, Nature 456 (2008) 125–129. [56] C. Jin, et al., H3.3/H2A.Z double variant-containing nucleosomes mark ‘nucleosome-free regions’ of active promoters and other regulatory regions, Nat. Genet. 41 (2009) 941–945. [57] S.V. Veiseth, et al., The SUVR4 histone lysine methyltransferase binds ubiquitin and converts H3K9me1 to H3K9me3 on transposon chromatin in Arabidopsis, PLoS. Genet 7 (2011) e1001325. [58] M.D. Shahbazian, K. Zhang, M. Grunstein, Histone H2B ubiquitylation controls processive methylation but not monomethylation by Dot1 and Set1, Mol. Cell 19 (2005) 271–277. [59] A. Shilatifard, Chromatin modifications by methylation and ubiquitination: implications in the regulation of gene expression, Annu. Rev. Biochem. 75 (2006) 243–269. [60] M.F. Dion, T. Kaplan, M. Kim, S. Buratowski, N. Friedman, O.J. Rando, Dynamics of replication-independent histone turnover in budding yeast, Science 315 (2007) 1405–1408. [61] R.B. Deal, J.G. Henikoff, S. Henikoff, Genome-wide kinetics of nucleosome turnover determined by metabolic labeling of histones, Science 328 (2010) 1161–1164.

145

[62] J. Ernst, et al., Mapping and analysis of chromatin state dynamics in nine human cell types, Nature 473 (2011) 43–49. [63] W.-H. Lien, et al., Genome-wide maps of histone modifications unwind in vivo chromatin states of the hair follicle lineage, Cell Stem Cell 9 (2011) 219–232. [64] N.D. Heintzman, et al., Histone modifications at human enhancers reflect global cell-type-specific gene expression, Nature 459 (2009) 108–112. [65] H.H. He, et al., Nucleosome dynamics define transcriptional enhancers, Nat. Genet. 42 (2010) 343–347. [66] M.P. Creyghton, et al., Histone H3K27ac separates active from poised enhancers and predicts developmental state, Proc. Natl. Acad. Sci 107 (2010) 21931–21936. [67] N.E. Yelina, et al., Epigenetic remodeling of meiotic crossover frequency in Arabidopsis thaliana DNA methyltransferase mutants, PLoS. Genet. 8 (2012) e1002844. [68] F. Smagulova, et al., Genome-wide analysis reveals novel molecular features of mouse recombination hotspots, Nature 472 (2011) 375–378. [69] A.K. Maunakea, et al., Conserved role of intragenic DNA methylation in regulating alternative promoters, Nature 466 (2010) 253–257. [70] J. Wang, L. Tian, H.S. Lee, Z.J. Chen, Nonadditive regulation of FRI and FLC loci mediates flowering-time variation in Arabidopsis allopolyploids, Genetics 173 (2006) 965–974. [71] Z. Ni, et al., Altered circadian rhythms regulate growth vigour in hybrids and allopolyploids, Nature 457 (2009) 327–331. [72] M. Tan, et al., Identification of 67 histone marks and histone lysine crotonylation as a new type of histone modification, Cell 146 (2011) 1016–1028. [73] R. Vafabakhsh, T. Ha, Extreme bendability of DNA less than 100 base pairs long revealed by single-molecule cyclization, Science 337 (2012) 1097–1101. [74] E. Segal, et al., A genomic code for nucleosome positioning, Nature 442 (2006) 772–778. [75] K. Struhl, Naturally occurring poly(dA-dT) sequences are upstream promoter elements for constitutive transcription in yeast, Proc. Natl. Acad. Sci. U.S.A. 82 (1985) 8419–8423. [76] J.P. Thomson, et al., CpG islands influence chromatin structure via the CpGbinding protein Cfp1, Nature 464 (2010) 1082–1086. [77] F. Baudat, et al., PRDM9 is a major determinant of meiotic recombination hotspots in humans and mice, Science 327 (2010) 836–840. [78] S. Myers, L. Bottolo, C. Freeman, G. McVean, P. Donnelly, A fine-scale map of recombination rates and hotspots across the human genome, Science 310 (2005) 321–324. [79] E.D. Parvanov, P.M. Petkov, K. Paigen, Prdm9 controls activation of mammalian recombination hotspots, Science 327 (2010) 835. [80] G.E. Zentner, S. Henikoff, Regulation of nucleosome dynamics by histone modifications, Nat. Struct. Mol. Biol. 20 (2013) 259–266. [81] M. Ha, S. Hong, W.H. Li, Predicting the probability of H3K4me3 occupation at a base pair from the genome sequence context, Bioinformatics (2013). [82] E.E. Eichler, et al., Missing heritability and strategies for finding the underlying causes of complex disease, Nat. Rev. Genet. 11 (2010) 446–450.