tered at +120 bp, the following ones at +300 and +480, respectively. RNA POL II sits right on the TSS. In the upstream region, we observe no periodic signal indicative of regular nucleosome positioning (phasing) across promoters. The closest nucleosomes are located around position −180 leaving a minimal nucleosome-free promoter region of 150 bp. As the respective peaks of the H3K4me3 and H2A.Z signals coincide, it seems likely that these two histone variants partly co-occur in the same nucleosomes. However, their relative abundance varies according to position, with H2A.Z and H3K4me3 preferring the upstream and downstream promoter regions, respectively. It is interesting to compare these results with the nucleosome architecture of yeast promoters as determined by Albert et al. (2007) for the H2A.Z histone variant using a similar data analysis method. They too observe a strong phasing of nucleosomes downstream of the TSS with a periodicity of about 170 bp. Between the promoter flanking −1 and +1 nucleosomes there appears to be a 120 bp nucleosomefree region, slightly shorter than observed in human cells. The average
TSS position was found to be located ~13 bp inside the upstream border of the +1 nucleosome in contrast to the TSS position in human cells, which is located ~45 bp upstream of the +1 nucleosome boundary. However, this difference is readily explained by the scanning model of transcription initiation in yeast (Kuehner and Brow, 2006), which suggests that RNA POL II first forms a preinitiation complex at the promoter site and then slides downstream along the DNA until it finds an appropriate sequence for initiating transcription. Thus, we demonstrate that separation of ChIP-Seq tags from the + and − DNA strands is essential for high-resolution mapping of DNAprotein complexes. Applying such an approach to the data presented in the Barski et al. (2007) paper allowed us to establish the consensus chromatin architecture of human promoters for two histone variants. With the high tag coverage obtained with the Solexa platform (about 70 tags per nucleosome for H3K4me3) we expect that individual nucleosomes can be mapped with near base-pair resolution for a majority of active promoters.
Response:
Mapping Nucleosome Positions Using ChIP-Seq Data We have recently reported genomewide maps of 20 histone lysine and arginine methylation marks as well as the histone variant H2A.Z, the insulator-binding protein CTCF, and RNA polymerase II (RNA POL II) in human CD4+ T cells. We obtained these data by sequencing the DNA samples obtained from chromatin immunoprecipitation (ChIP) using the Solexa high-throughput
sequencing technique, a method termed ChIP-Seq (Barski et al., 2007). We found that trimethylation of lysine 4 of histone H3 (H3K4me3) is greatly increased around the transcription start site (TSS) of genes and that the level of this mark correlates with transcriptional activity. Furthermore, we observed that the H3K4me3 signals downstream of the TSS were divided into several
832 Cell 131, November 30, 2007 ©2007 Elsevier Inc.
Our results further suggest that ChIPSeq may be an effective alternative technique for genome-wide promoter mapping, complementing methods based on the 5′ tag sequencing of full-length cDNAs generated with RACE (Rapid Amplification of cDNA Ends) or similar protocols. References Albert, I., Mavrich, T.N., Tomsho, L.P., Qi, J., Zanton, S.J., Schuster, S.C., and Pugh, B.F. (2007). Nature 446, 572–576. Barski, A., Cuddapah, S., Cui, K., Roh, T.Y., Schones, D.E., Wang, Z., Wei, G., Chepelev, I., and Zhao, K. (2007). Cell 129, 823–837. Kuehner, J.N., and Brow, D.A. (2006). J. Biol. Chem. 182, 14119–14128.
Christoph D. Schmid1,* and Philipp Bucher1,2,* Swiss Institute of Bioinformatics Swiss Institute for Experimental Cancer Research École Polytechnique Fédérale de Lausanne (EPFL), Ch. des boveresses 155, CH-1066, Switzerland *Correspondence:
[email protected] (C.S.),
[email protected] (P.B.) DOI 10.1016/j.cell.2007.11.017
1 2
subpeaks at +50, +210, and +360. Considering that we used mononucleosomes as our chromatin inputs for the ChIP experiments and that the short sequence reads obtained from the sequencing analysis represent the ends of the DNA templates, we suggested that these subpeaks may represent the presence of similarly positioned nucleosomes relative to the TSS in actively transcribed genes. Indeed, in their Correspondence, Schmid and Bucher have further analyzed our ChIP-Seq data by separating the sequence reads according to the + or − DNA strand. They found that the reads from the + strand of DNA define the 5′ ends and that reads from the − strand define the 3′ ends of nucleosomes. They point out that the short sequence
tags can be used to precisely map nucleosome positions near the TSS of genes. We completely agree with the observation of Schmid and Bucher that ChIP-Seq data for histone modifications and histone variants can be used to define nucleosome positions. However, this technique is limited by a number of factors. First and foremost, this technique can only be used to map nucleosomes that are associated with certain modifications, limiting the genomic regions for which nucleosomes can be mapped. Second, specific histone modifications can have an effect on the positioning of nucleosomes. For example, it has been shown in the budding yeast Saccharomyces cerevisiae that nucleosomes are first hyperacetylated before being removed from the activated promoter of the PHO5 gene (Boeger et al., 2003; Reinke and Horz, 2003). Third, the nucleosomes mapped by this method cannot be
used to determine the nucleosome level at specific regions because the number of sequence reads for a nucleosome is dependent on both the nucleosome level and the modification level of each nucleosome. Schmid and Bucher compare the nucleosome positions determined using ChIP-Seq data for the histone variant H2A.Z with data obtained by Albert et al. (2007) using similar techniques in yeast. They propose that nucleosome-free regions in human cells are slightly shorter than in yeast. However, what is really being observed here is the absence of H2A.Z-containing nucleosomes in these regions. It is possible that unmodified canonical nucleosomes exist in these regions. To precisely map the genome-wide distribution of nucleosome positions and nucleosome levels in an unbiased manner, the DNA from the mononucleosome fractions without the ChIP step should be directly sequenced.
References Albert, I., Mavrich, T.N., Tomsho, L.P., Qi, J., Zanton, S.J., Schuster, S.C., and Pugh, B.F. (2007). Nature 446, 572–576. Barski, A., Cuddapah, S., Cui, K., Roh, T.Y., Schones, D.E., Wang, Z., Wei, G., Chepelev, I., and Zhao, K. (2007). Cell 129, 823–837. Boeger, H., Griesenbeck, J., Strattan, J.S., and Kornberg, R.D. (2003). Mol. Cell 11, 1587– 1598. Reinke, H., and Horz, W. (2003). Mol. Cell 11, 1599–1607.
Artem Barski,1 Suresh Cuddapah,1 Kairong Cui,1 Tae-Young Roh,1 Dustin E. Schones,1 Zhibin Wang,1 Gang Wei,1 Iouri Chepelev,2 and Keji Zhao1,*
Laboratory of Molecular Immunology, National Heart, Lung, and Blood Institute, NIH, Bethesda, MD 20892, USA 2 Department of Human Genetics, Gonda Neuroscience and Genetics Research Center, University of California, Los Angeles, Los Angeles, CA 90095, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2007.11.018 1
Cell 131, November 30, 2007 ©2007 Elsevier Inc. 833
