ARTICLES NATURE Vol 447 14 June 2007 replicated in multiple intervals in S phase. The remaining regions into open'and'closed chromatin territories that represent higher were sub-classified into early-, mid-or late-replicating based on the order functional domains. We explored how different chromatin average TR50 of the temporally specific probes within a 10-kb win- features, particularly histone modifications, correlate with chro- dow? For regions of each class of replication timing, we determined matin structure, both over short and long distances the relative enrichment of various histone modification peaks in Chromatin accessibility and histone modifications. We used his HeLa cells(Fig. 7b; Supplementary Information section 4.4). The tone modification studies and DNasel sensitivity data sets(intro- orrelations of activating and repressing histone modification peaks duced above) to examine general chromatin accessibility without ith TR50 are confirmed by this analysis(Fig. 7b). Intriguingly, the focusing on the specific DHS sites(see Supplementary Informa t th the o ique in being enriched for both activating tion sections 3. 1, 3.3 and 3.4). A fundamental difficulty in analysing Pan-S segments are u (H3K4me2, H3ac and H4ac)and repressing(H3K27me3)histones, continuous data across large genomic regions is determining the consistent with the suggestion that the Pan-S replication pattern appropriate scale for analysis( for example, 2 kb, 5kb, 20 kb, and so arises from interallelic variation in chromatin structure and time of on). To address this problem, we developed an approach based on replication. This observation is also consistent with the Pan-S rep- wavelet analysis, a mathematical tool pioneered in the field of signal lication pattern seen for the H19/IGF2 locus, a known imprinted processing that has recently been applied to continuous-value geno region with differential epigenetic modifications across the two mic analyses. Wavelet analysis provides a means for consistently ransforming continuous signals into different scales, enabling the The extensive rearrangements in the genome of HeLa cells led us to correlation of different phenomena independently at differing scales sk whether the detected correlations between TR50 and chromatin in a consistent manner state are seen with other cell lines. The histone modification data with Global correlations of chromatin accessibility and histone modi- GM06990 cells allowed us to test whether the time of replication of fications. We computed the regional correlation between dNasel genomic segments in HeLa cells correlated with the chromatin state sensitivity and each histone modification at multiple scales using a in GM06990 cells. Early-and late-replicating segments in HeLa cells wavelet approach( Fig. 8 and Supplementary Information section are enriched and depleted, respectively, for activating marks in 4.2). To make quantitatie mish histo of correlation values be- s between different histone rangements(see Supplementary Information section 2.12), the TR50 tween DNasel sensitivity and isograms of correlation values be- 106990 cells(Fig. 7b). Thus, despite the presence of genomic rear- modifications, we computed hi each histone modification at several and chromatin state in HeLa cells are not far from a constitutive scales and then tested these for significance at specific scales. Figure baseline also seen with a cell line from a different lineage. The enrich- 8c shows the distribution of correlation values at a 16-kb scale, which ment of multiple activating histone modifications and the depletion is considerably larger than individual cis-acting regulatory elements. f a repressive modification from segments that replicate early in s At this scale, H3K4me2, H3K4me3 and H3ac show similarly high phase extends previous work in the field at a level of detail and scale correlation. However, they are significantly distinguished from not attempted before in mammalian cells. The duality of histone H3K4mel and H4ac modifications(P<1.5X 10>; see Supple- odification patterns in Pan-S areas of the HeLa genome, and the mentary Information section 4.5), which show lower correlation with concordance of chromatin marks and replication time across two DNasel sensitivity. These results suggest that larger-scale relation- disparate cell lines(HeLa and GM06990)confirm the coordination ships between chromatin accessibility and histone modifications of histone modifications with replication in the human genom are dominated by sub-regions in which higher average DNasel sens- itivity is accompanied by high levels of H3K4me2, H3K4me3 and Chromatin architecture and genomic domains H3ac modifications Overview. The packaging of genomic DNA into chromatin is inti- Local correlations of chromatin accessibility and histone mately connected with the control of gene expression and other cations Narrowing to a scale of -2 kb revealed a more chromosomal processes. We next examined chromatin structure situation, in which H3K4me2 is the histone modification that is over a larger scale to ascertain its relation to transcription and other best correlated with DNasel sensitivity. However, there is no clear processes. Large domains(50 to >200 kb) of generalized DNasel combination of marks that correlate with DNasel sensitivity in a sensitivity have been detected around developmentally regulated way that is analogous to that seen at a larger scale(see Supplemen gene clusters", prompting speculation that the genome is organized tary Information section 4.3). One explanation for the increased (ENm013) Genomic position (kb) H3k4me2: DNasel correlation by scale Correlation value Figure 8 Wavelet correlations of histone marks and DNasel sensitivity. ffering scales decomposed by the wavelet analysis from ple, correlations between DNasel sensitivity and H3K4me2(both (in kb); the colour at each point in the heatmap represents in the GM06990 cell line)over a 1. 1-Mb region on chromosome 7(ENCODE tion at the given scale, measured in a 20 kb window region ENm013)are shown. a, The relationship between histone position. c, Distribution of correlation values at the modification H3K4me2 (upper plot) and DNasel sensitivity(lower plot)is the indicated histone marks. The yaxis is the density of shown for ENCODE region ENm013. The curves are coloured with the ralues across ENCODE; all modifications show a peak at a str. e. the s te data cos ie a ire aet resented cslewtavp let chelation The positive-correlation value E2007 Nature Publishing Groupreplicated in multiple intervals in S phase. The remaining regions were sub-classified into early-, mid- or late-replicating based on the average TR50 of the temporally specific probes within a 10-kb win￾dow75. For regions of each class of replication timing, we determined the relative enrichment of various histone modification peaks in HeLa cells (Fig. 7b; Supplementary Information section 4.4). The correlations of activating and repressing histone modification peaks with TR50 are confirmed by this analysis (Fig. 7b). Intriguingly, the Pan-S segments are unique in being enriched for both activating (H3K4me2, H3ac and H4ac) and repressing (H3K27me3) histones, consistent with the suggestion that the Pan-S replication pattern arises from interallelic variation in chromatin structure and time of replication75. This observation is also consistent with the Pan-S rep￾lication pattern seen for the H19/IGF2 locus, a known imprinted region with differential epigenetic modifications across the two alleles76. The extensive rearrangements in the genome of HeLa cells led us to ask whether the detected correlations between TR50 and chromatin state are seen with other cell lines. The histone modification data with GM06990 cells allowed us to test whether the time of replication of genomic segments in HeLa cells correlated with the chromatin state in GM06990 cells. Early- and late-replicating segments in HeLa cells are enriched and depleted, respectively, for activating marks in GM06990 cells (Fig. 7b). Thus, despite the presence of genomic rear￾rangements (see Supplementary Information section 2.12), the TR50 and chromatin state in HeLa cells are not far from a constitutive baseline also seen with a cell line from a different lineage. The enrich￾ment of multiple activating histone modifications and the depletion of a repressive modification from segments that replicate early in S phase extends previous work in the field at a level of detail and scale not attempted before in mammalian cells. The duality of histone modification patterns in Pan-S areas of the HeLa genome, and the concordance of chromatin marks and replication time across two disparate cell lines (HeLa and GM06990) confirm the coordination of histone modifications with replication in the human genome. Chromatin architecture and genomic domains Overview. The packaging of genomic DNA into chromatin is inti￾mately connected with the control of gene expression and other chromosomal processes. We next examined chromatin structure over a larger scale to ascertain its relation to transcription and other processes. Large domains (50 to .200 kb) of generalized DNaseI sensitivity have been detected around developmentally regulated gene clusters77, prompting speculation that the genome is organized into ‘open’ and ‘closed’ chromatin territories that represent higher￾order functional domains. We explored how different chromatin features, particularly histone modifications, correlate with chro￾matin structure, both over short and long distances. Chromatin accessibility and histone modifications. We used his￾tone modification studies and DNaseI sensitivity data sets (intro￾duced above) to examine general chromatin accessibility without focusing on the specific DHS sites (see Supplementary Informa￾tion sections 3.1, 3.3 and 3.4). A fundamental difficulty in analysing continuous data across large genomic regions is determining the appropriate scale for analysis (for example, 2 kb, 5 kb, 20 kb, and so on). To address this problem, we developed an approach based on wavelet analysis, a mathematical tool pioneered in the field of signal processing that has recently been applied to continuous-value geno￾mic analyses. Wavelet analysis provides a means for consistently transforming continuous signals into different scales, enabling the correlation of different phenomena independently at differing scales in a consistent manner. Global correlations of chromatin accessibility and histone modi￾fications. We computed the regional correlation between DNaseI sensitivity and each histone modification at multiple scales using a wavelet approach (Fig. 8 and Supplementary Information section 4.2). To make quantitative comparisons between different histone modifications, we computed histograms of correlation values be￾tween DNaseI sensitivity and each histone modification at several scales and then tested these for significance at specific scales. Figure 8c shows the distribution of correlation values at a 16-kb scale, which is considerably larger than individual cis-acting regulatory elements. At this scale, H3K4me2, H3K4me3 and H3ac show similarly high correlation. However, they are significantly distinguished from H3K4me1 and H4ac modifications (P , 1.5 3 10233; see Supple￾mentary Information section 4.5), which show lower correlation with DNaseI sensitivity. These results suggest that larger-scale relation￾ships between chromatin accessibility and histone modifications are dominated by sub-regions in which higher average DNaseI sens￾itivity is accompanied by high levels of H3K4me2, H3K4me3 and H3ac modifications. Local correlations of chromatin accessibility and histone modifi￾cations. Narrowing to a scale of ,2 kb revealed a more complex situation, in which H3K4me2 is the histone modification that is best correlated with DNaseI sensitivity. However, there is no clear combination of marks that correlate with DNaseI sensitivity in a way that is analogous to that seen at a larger scale (see Supplemen￾tary Information section 4.3). One explanation for the increased 1.11 Mb (ENm013) 25 15 16 8 4 2 0 0 4 8 H3k4me2 DNaseI sensitivity 89,600 89,800 90,000 90,200 90,400 H3k4me2 : DNaseI correlation by scale Genomic position (kb) Negative Positive Genomic position (kb) Correlation a c bSignal/control Scale (kb) H3k4me2 H3k4me3 H3Ac H3k4me1 H4Ac –1.0 –0.5 0 0.5 1.0 Correlation value Density 16-kb scale 1.2 1.0 0.8 0.6 0.4 0.2 0 Figure 8 | Wavelet correlations of histone marks and DNaseI sensitivity. As an example, correlations between DNaseI sensitivity and H3K4me2 (both in the GM06990 cell line) over a 1.1-Mb region on chromosome 7 (ENCODE region ENm013) are shown. a, The relationship between histone modification H3K4me2 (upper plot) and DNaseI sensitivity (lower plot) is shown for ENCODE region ENm013. The curves are coloured with the strength of the local correlation at the 4-kb scale (top dashed line in panel b). b, The same data as in a are represented as a wavelet correlation. The y axis shows the differing scales decomposed by the wavelet analysis from large to small scale (in kb); the colour at each point in the heatmap represents the level of correlation at the given scale, measured in a 20 kb window centred at the given position. c, Distribution of correlation values at the 16 kb scale between the indicated histone marks. The y axis is the density of these correlation values across ENCODE; all modifications show a peak at a positive-correlation value. ARTICLES NATURE|Vol 447| 14 June 2007 808 ©2007 NaturePublishingGroup