Chromatin plays a crucial role in gene regulation, and chromatin immunoprecipitation followed by sequencing (ChIP–seq) has been the standard technique for examining protein–DNA interactions across the whole genome. However, it is difficult to obtain epigenomic information from limited numbers of cells by ChIP–seq because of sample loss during chromatin preparation and inefficient immunoprecipitation. In this study, we established an immunoprecipitation-free epigenomic profiling method named chromatin integration labelling (ChIL), which enables the amplification of genomic sequences closely associated with the target molecules before cell lysis. Using ChIL followed by sequencing (ChIL–seq), we reliably detected the distributions of histone modifications and DNA-binding factors in 100–1,000 cells. In addition, ChIL–seq successfully detected genomic regions associated with histone marks at the single-cell level. Thus, ChIL–seq offers an alternative method to ChIP–seq for epigenomic profiling using small numbers of cells, in particular, those attached to culture plates and after immunofluorescence.
Access optionsAccess options
Subscribe to Journal
Get full journal access for 1 year
only $18.75 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Deep-sequencing (ChIP–seq and ChIL–seq) data in this study have been deposited in the Gene Expression Omnibus (GEO) under the accession code GSE115047. Previously published ChIP–seq, ATAC-seq and mRNA-seq data reanalysed in this work are available under accession codes GSE36023 (ChIP–seq C2C12 cells; H3K4me3 and H3K27me3), GSE65493 (ChIP–seq C2C12 cells; H3K9me3), GSE36024 (ChIP–seq C2C12 cells; CTCF and MyoD), GSE37525 (ChIP–seq C2C12 cells; H3K27ac), GSE89977 (ChIP–seq C2C12 cells; H3K4me3, H3K27ac and H3K27me3), GSE75169 (ChIP–seq MCF-7 cells; H3K4me3; ChIP–seq MDA-MB-231cells; H3K4me3), GSE63523 (ULI-NChIP TT2 cells; H3K4me3 for 5K, 10K and 100K cells)7, GSE74359 (Mint-ChIP K562 cells; H3K4me3 for 500, 1K, 10K and 100K cells)14, GSE72784 (Micro-ChIP E14 cells; H3K4me3 for 500, 1K, 10K and 100K cells)13, GSE71434 (STAR-ChIP R1 cells; H3K4me3 for 200 and 5K cells)15 and GSE104389 (ATAC-seq and mRNA-seq for C2C12 growth cells). Source data for Figs. 1e, 2a,b, 3c, 4d–f and 6c and Supplementary Figs. 1g, 3a, 4a,c,e, 5b and 6a,c have been provided as Supplementary Table 7. All other data supporting the findings of this work are available from the corresponding authors upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Mardis, E. R. ChIP–seq: welcome to the new frontier. Nat. Methods 4, 613–614 (2007).
Park, P. J. ChIP–seq: advantages and challenges of a maturing technology. Nat. Rev. Genet. 10, 669–680 (2009).
Hitchler, M. J. & Rice, J. C. Genome-wide epigenetic analysis of human pluripotent stem cells by ChIP and ChIP–seq. Methods Mol. Biol. 767, 253–267 (2011).
Gilfillan, G. D. et al. Limitations and possibilities of low cell number ChIP–seq. BMC Genomics 13, 645 (2012).
Shankaranarayanan, P. et al. Single-tube linear DNA amplification (LinDA) for robust ChIP–seq. Nat. Methods 8, 565–567 (2011).
Zwart, W. et al. A carrier-assisted ChIP–seq method for estrogen receptor-chromatin interactions from breast cancer core needle biopsy samples. BMC Genomics 14, 232 (2013).
Brind’Amour, J. et al. An ultra-low-input native ChIP–seq protocol for genome-wide profiling of rare cell populations. Nat. Commun. 6, 6033 (2015).
Cao, Z. N., Chen, C. Y., He, B., Tan, K. & Lu, C. A microfluidic device for epigenomic profiling using 100 cells. Nat. Methods 12, 959–962 (2015).
Jakobsen, J. S. et al. Amplification of pico-scale DNA mediated by bacterial carrier DNA for small-cell-number transcription factor ChIP–seq. BMC Genomics 16, 46 (2015).
Rotem, A. et al. Single-cell ChIP–seq reveals cell subpopulations defined by chromatin state. Nat. Biotechnol. 33, 1165–1172 (2015).
Schmidt, C., Rendeiro, A. F., Sheffield, N. C. & Bock, C. ChIPmentation: fast, robust, low-input ChIP–seq for histones and transcription factors. Nat. Methods 12, 963–965 (2015).
Zheng, X. B. et al. Low-cell-number epigenome profiling aids the study of lens aging and hematopoiesis. Cell Rep. 13, 1505–1518 (2015).
Dahl, J. A. et al. Broad histone H3K4me3 domains in mouse oocytes modulate maternal-to-zygotic transition. Nature 537, 548–552 (2016).
van Galen, P. et al. A multiplexed system for quantitative comparisons of chromatin landscapes. Mol. Cell 61, 170–180 (2016).
Zhang, B. J. et al. Allelic reprogramming of the histone modification H3K4me3 in early mammalian development. Nature 537, 553–557 (2016).
Skene, P. J. & Henikoff, S. An efficient targeted nuclease strategy for high-resolution mapping of DNA binding sites. eLife 6, e21856 (2017).
Skene, P. J., Henikoff, J. G. & Henikoff, S. Targeted in situ genome-wide profiling with high efficiency for low cell numbers. Nat. Protoc. 13, 1006–1019 (2018).
Zarnegar, M. A., Reinitz, F., Newman, A. M. & Clarke, M. F. Targeted chromatin ligation, a robust epigenetic profiling technique for small cell numbers. Nucleic Acids Res. 45, e153 (2017).
Schmid, M., Durussel, T. & Laemmli, U. K. ChIC and ChEC; genomic mapping of chromatin proteins. Mol. Cell 16, 147–157 (2004).
Kazane, S. A. et al. Site-specific DNA-antibody conjugates for specific and sensitive immuno-PCR. Proc. Natl Acad. Sci. USA 109, 3731–3736 (2012).
Sos, B. C. et al. Characterization of chromatin accessibility with a transposome hypersensitive sites sequencing (THS-seq) assay. Genome Biol. 17, 20 (2016).
Buenrostro, J. D., Giresi, P. G., Zaba, L. C., Chang, H. Y. & Greenleaf, W. J. Transposition of native chromatin for fast and sensitive epigenomic profiling of open chromatin, DNA-binding proteins and nucleosome position. Nat. Methods 10, 1213–1218 (2013).
Mikkelsen, T. S. et al. Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature 448, 553–560 (2007).
Daley, T. & Smith, A. D. Predicting the molecular complexity of sequencing libraries. Nat. Methods 10, 325–327 (2013).
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
McLean, C. Y. et al. GREAT improves functional interpretation of cis-regulatory regions. Nat. Biotechnol. 28, 495–501 (2010).
Holliday, D. L. & Speirs, V. Choosing the right cell line for breast cancer research. Breast Cancer Res. 13, 215 (2011).
Lara-Astiaso, D. et al. Chromatin state dynamics during blood formation. Science 345, 943–949 (2014).
Chen, X. Q. et al. ATAC-see reveals the accessible genome by transposase-mediated imaging and sequencing. Nat. Methods 13, 1013–1020 (2016).
Stahl, P. L. et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science 353, 78–82 (2016).
Stoeckius, M. et al. Simultaneous epitope and transcriptome measurement in single cells. Nat. Methods 14, 865–868 (2017).
Harada, A. et al. Incorporation of histone H3.1 suppresses the lineage potential of skeletal muscle. Nucleic Acids Res. 43, 775–786 (2015).
Kimura, H., Hayashi-Takanaka, Y., Goto, Y., Takizawa, N. & Nozaki, N. The organization of histone H3 modifications as revealed by a panel of specific monoclonal antibodies. Cell Struct. Funct. 33, 61–73 (2008).
Hayashi-Takanaka, Y. et al. Tracking epigenetic histone modifications in single cells using Fab-based live endogenous modification labeling. Nucleic Acids Res. 39, 6475–6488 (2011).
Picelli, S. et al. Tn5 transposase and tagmentation procedures for massively scaled sequencing projects. Genome Res. 24, 2033–2040 (2014).
Handa, T., Harada, A., Maehara, K., Ohkawa, Y. & Kimura, H. Detailed protocol—chromatin integration labeling. Protoc. Exch. https://doi.org/10.1038/protex.2018.122 (2018).
Kim, D., Landmead, B. & Salzberg, S. L. HISAT: a fast spliced aligner with low memory requirements. Nat. Methods 12, 357–360 (2015).
Ramirez, F., Dundar, F., Diehl, S., Gruning, B. A. & Manke, T. deepTools: a flexible platform for exploring deep-sequencing data. Nucleic Acids Res. 42, W187–W191 (2014).
Maehara, K. & Ohkawa, Y. agplus: A rapid and flexible tool for aggregation plots. Bioinformatics 31, 3046–3047 (2015).
Zuo, C. & Keles, S. A statistical framework for power calculations in ChIP–seq experiments. Bioinformatics 30, 753–760 (2014).
We are grateful to the staff of Ohkawa Lab and S. Sekine (Waseda University, Tokyo, Japan) for technical supports, and Y. Sato (Tokyo Tech, Yokohama, Japan) for discussion and drawing the ChIL scheme illustrations. We also thank the Advanced Computational Scientific Program of the Research Institute for Information Technology, Kyushu University and the National Institute of Genetics, for providing high-performance computing resources. This work was in part supported by MEXT/JSPS KAKENHI (JP25116010, JP17H03608, JP17K19356, JP18H04802 and JP18H05527 to Y.O., JP16H01219, JP15K18457 and JP18K19432 to A.H., JP16K18479, JP16H01577, JP16H01550 and JP18H04904 to K.M., JP25116002, JP17H01408 and JP18H05534 to H.Kurumizaka and JP25116005, JP26291071, JP17H01417 and JP18H05527 to H.Kimura), JST CREST (JPMJCR16G1 to Y.O., H.Kurumizaka, and H.Kimura), the Platform for Drug Discovery, Informatics, and Structural Life Science, and the Platform Project for Supporting Drug Discovery and Life Science Research from the Japan Agency for Medical Research and Development (to K.S., H.Kimura and H.Kurumizaka). H.Kurumizaka was also supported by the Waseda Research Institute for Science and Engineering and by programs of Waseda University.
Integrated supplementary information
(a) Scheme of ChIL probe preparation. Antibodies were first conjugated with amine-reactive DBCO-(PEG)5-NHS ester for the subsequent Cu2+-free Click reaction with the 5’-azide of pre-annealed double-stranded ChIL DNA. (b-f) Analysis of ChIL probes by 15% polyacrylamide gel electrophoresis (PAGE) followed by DNA staining with GelRed (b-e) and by 7.5% non-reduced SDS-PAGE followed by protein staining with Coomassie Brilliant Blue (f). (b) Conjugation kinetics of azide-DNA. The reaction mixtures (IgG:DNA = 3:1) were incubated at 4 °C for the indicated period. Most DNA was free after 1 day of incubation but became conjugated by 7 days. (c) Purification of ChIL probe. DNA-conjugated antibody was purified through multiple rounds of filtration using DNA Fast Flow Membrane filters (Merck Millipore, Germany). Free DNA was mostly removed by four rounds of filtration and the recovery efficiency of ChIL probes fell during the fifth round of filtration. (d) IgG:DNA ratio. Under the reaction conditions of IgG:DNA at a 1:2 molar ratio, one to four DNA molecules were conjugated per IgG. For data shown in b-d, similar results were obtained in two independent experiments. (e and f) Reproducibility of ChIL probe preparation. ChIL probes prepared from two independent experiments (#1 and #2) migrated similarly by PAGE (e) and SDS-PAGE (f), indicating that ChIL probe preparation was well controlled. The unprocessed blots (b-f) are shown in Supplementary Fig. 7. (g and h) Immunofluorescence with C2C12 cells using Alexa Fluor 488-conjugated mouse anti-H3K4me3 antibody and TAMRA-conjugated anti-mouse ChIL probe. DNA was counterstained with Hoechst 33342. Scale bar, 10 μm. Two independent immunofluorescence experiments showed similar results. (g) Comparison between purified and unpurified (conjugated) ChIL probes. Intensity plots of the primary antibodies (x-axis) and ChIL probes (y-axis) are shown with Pearson’s correlations (averages with standard deviations from 20 cells; 10 cells from each experiment). The lower correlation observed for unpurified ChIL probe suggests the non-specific binding of unconjugated ChIL DNA. (f) Incubation of ChIL probes in phosphate-buffered saline (low salt) resulted in higher background staining in both the cytoplasm and nucleus compared with buffer containing 0.5 M NaCl (high salt).
(a) Typical ChIL-seq library fragment profiles. ChIL-seq H3K4me3 (left) and H3K27me3 (right) libraries were analyzed using a bioanalyzer (Agilent Technologies, CA, USA) with a DNA high-sensitivity kit. The fragment sizes were ~300 bp (~200–400 bp). Fragment sizes were analyzed in all libraries prepared, and similar results were seen. (b and c) Comparison of library complexity among ChIL-seq and ChIP-seq data by extrapolation of distinct uni-reads (b) and total mapped reads (c). Labels (#1 and #2) represent two biologically independent experiments. (b) The dotted lines indicate the lower limits of the ENCODE target non-redundancy fractions at 1 × 107 reads for the active markers and at 2 × 107 reads for the repressive H3K27me3. (c) Comparison of library complexities between H3K4me3 ChIL-seq with low-input H3K4me3 ChIP-seq data, that is, Micro-ChIP (E14 cells), ULI-NChIP (TT2 cells), STAR-ChIP-seq (R1 cells), and Mint-ChIP-seq (K562 cells) for 100–1K, 5–10K, and 100K cells.
(a-c) Optimization of bin sizes. Data set from a single experiment for each modification (#1) was used for the analysis. (a) Coefficients of the logistic regression model represent the contribution of ChIL-seq signal in each 100-bp bin around TSSs to predicting the presence of ChIP-seq peak. The optimal bin-sizes (ranges) for ChIL-seq were determined to be the ranges of statistically significant coefficients. Significant (P < 0.01 at n = 27,051 TSSs in the two-sided Wald tests without multiple correction) and non-significant coefficients are indicated in red and gray, respectively. The shapes of the corresponding ChIL-seq signal densities (solid line) are shown as reference. (b) Evaluation of the bin sizes to predict ChIP-seq peaks by ROC (receiver operating characteristics) curves showing false positive rates (FPRs) vs. true positive rates (TPRs) of prediction. Shown are ROC curves for logistic regression and read counts in TSS ± 1 kb, ± 2 kb, ± 5 kb and in 2–5 kb from the TSS. The ChIL-seq signals within 2 as well as 1 kb from the TSS were sufficient to predict the H3K4me3 and H3K27ac peaks. For H3K27me3, while 2–5 kb from the TSS (excluding TSS ± 1 kb) was found to be the optimal range, TSS ± 5 kb (including TSS ± 1 kb) was also sufficient. (c) Comparison of ChIL-seq and ChIP-seq signals with respect to transcript expression ranking. ChIP-seq data are normalized to the input data. RNA expression levels in C2C12 cells were quantified by RNA-seq. The x-axis shows the ranking of transcripts and the y-axis shows log2-transformed ChIL-seq read counts in the regions within 2 kbp (top) and 5 kbp (bottom) from the corresponding TSS. If an expression level was shared by multiple transcripts, the mean of ChIL-seq and ChIP read counts for the regions around the corresponding TSSs was used. The smooth curves are local regression (LOESS) curves and the shaded gray region represents the 95% confidence interval.
(a) The required Benjamini–Hochberg FDR thresholds corresponded to the maximal Jaccard coefficients (JC) in Figure 4. Higher cell numbers required higher thresholds. Total 8 points (2 replicates x 4 different cell numbers) were used to draw the curves. (b) Precision (TP/(TP+FP)-recall (TP/(TP+FN)) curves of ChIL-seq for predicting ChIP-seq peaks. Each circle represents the optimal threshold in terms of JC. ChIP-seq data were normalized to the input. Two individual experiments are labeled #1 and #2 (also for e). (c) Prediction performance of low-input ChIP-seq. Maximal JC are plotted as in Figure 4. H3K4me3 assays of ES (micro [GSE72784], STAR [GSE71434], ULI-N [GSE63523]), and K562 (ChIL-seq and Mint-ChIP [GSE74359]) were evaluated (4, 3, 3, 4 and 4 points for the curve fitting, respectively). K562 cells were analyzed to test whether ChIL-seq can be applied to suspension cells. As cells were gradually detached from the plate, the JC (0.28 and 0.38 for 100 and 1000 cells in average of replicates) were not as high as those of C2C12 cells. (d) Estimated performance of ChIL-seq at given total reads. The lines indicate the median values estimated by bootstrap sampling (n = 20 for each of 41 intervals in 104–108) based on the assumption that read counts in genome-wide bins follow a multinomial distribution. The vertical lines indicate recommended read numbers at which no more than 5% improvement on the current maximal JC was expected when the total reads were doubled. The number of distinct mapped reads was estimated by Preseq under the assumption that 70% of the total reads were uniquely mapped. The widths of the 95% confidence intervals were narrower than the line thickness. (e) Inspecting ChIL-seq conditions. ChIL-seq for H3K4me3 and H3K27me3 were performed using 100 C2C12 cells in various conditions (Supplementary Table 4). In some cases, control experiments using ChIL DNA were also performed. The maximal JC are plotted, as in Figure 4. The labels of ‘100’ and ‘Tn5_0.1’ indicate the standard conditions of ChIL-seq. Each condition had two replicates, represented as points.
(a) Fluorescence microscope images of entire cells used for ChIL-seq in Figure 6b. A tiling image was generated by aligning 20 × 20 images to cover a single well of a 96 well plate (left panel). Magnified views of the indicated area (red square) are shown (right panel). Scale bar, 100 μm. The experiments were repeated three times with similar results. (b-d) Analysis of CTCF and MyoD ChIL-seq. Two independent ChIL-seq experiments (#1 and #2; Fig. 6) were performed for each protein and the control (GSE115047). Data set of each single experiment (CTCF #1 and MyoD #2) were used for the analysis. (b) Coefficients of the logistic regression model represent the contribution of the ChIL-seq signal in each 100-bp bin around transcription factor binding sites (TFBSs) to predicting the presence of the ChIP-seq peak, as explained in Supplementary Figure 3a. For CTCF and MyoD, ChIL-seq reads in TFBSs ± 2 kb evidently contributed to the prediction of the presence of ChIP-seq peaks. PWMScan (http://ccg.vital-it.ch/pwmtools/pwmscan.php) was used to select the potential TFBSs with the parameters of P-value cutoffs of 10−7 and 10−6 for CTCF and MyoD, respectively. (c) Evaluation of the bin sizes for predicting ChIP-seq peaks at TFBSs by ROC curves. ROC curves for logistic regression and read counts in TFBSs ± 1 kb, ± 2 kb, ± 5 kb, and 2–5 kb from the TFBSs are shown. ChIL-seq signals within 2 kb from the TFBSs were sufficient to predict the CTCF and MyoD peaks. (d) Determining the optimal thresholds for genome-wide ChIL-seq peak detection. The sensitivity (TPR), specificity (1 − FPR), and Jaccard coefficient at different thresholds (read counts ≥ x) were plotted. The results for CTCF (#1) were: threshold, 5 (P = 0.001, Benjamini–Hochberg-adjusted P = 0.034); sensitivity, 44.2%; specificity, 98.1%; Jaccard, 23.0%, and the results for MyoD (MyoD #2) were: threshold, 8 (P = 0.001, adjusted P = 0.034); sensitivity, 33.3%; specificity, 99.0%; Jaccard, 18.8%. The optimal sizes of the bins for CTCF and MyoD were set at 2 kbp, respectively, based on the analysis presented in c.
(a) Colocalization of ChIL probes with the H3K9me3 antibody. Mouse C2C12 cells were stained with Alexa Fluor 488-conjugated mouse anti-H3K9me3 and TAMRA-labeled ChIL probe and counterstained with Hoechst 33342. Similar results were obtained by two independent experiments. Intensity plots of the primary antibodies (x-axis) and ChIL probes (y-axis) are shown with Pearson’s correlations (averages and standard deviations from 20 cells; 10 cells from each experiment). Scale bar, 10 μm. (b) H3K9me3 profiles from ChIL-seq with various numbers of C2C12 cells by IGV. H3K9me3 ChIP-seq data (GSE65493), H3K4me3 and H3K27me3 100-cell ChIL-seq, in-house ATAC-seq and mRNA-seq data for C2C12 cells are shown for comparison. The y-axis shows the average normalized signal intensity of each genome locus. ChIL-seq signal patterns of H3K9me3 were, regardless of the cell number, analogous to that of the corresponding ChIP-seq data and differed markedly from the H3K4me3 and H3K27me3 signal patterns. Some broadly enriched regions by ChIP-seq (like Aap1 region in Supplementary Fig. 6b), however, appears less highlighted by ChIL-seq. The labels ChIL-seq #1, #2, ChIP-seq and ATAC-seq represent two biologically independent experiments. (c) Enrichment of H3K9me3 ChIL-seq signals in repeat elements. Percentages of ChIL-seq signals in GSAT MM (major satellite), IAPEZI, L1Md A/F/G/T, and MERVL repeat elements plotted against total mapped reads. ChIP-seq signal accumulation in the repeat elements are indicated as log2 fold change against input for comparison (left and middle panels). H3K9me3 was enriched in GSAT MM and IAPEZI in both ChIL-seq and ChIP-seq. The consensus sequences of repetitive elements from the RepBase database (https://www.girinst.org/repbase) were used. About 30% of H3K9me3 ChIL-seq reads with 100, 1K, and 10K cells intersected with GSAT MM sequences. ChIP-seq signal accumulations on repetitive elements are represented by the log2-transformed fold changes of the ChIP-seq reads over the input reads (right panel), using GSE36023 (H3K4me3), GSE37525 (H3K27ac), GSE36023 (H3K27me3), and GSE65493 (H3K9me3) for C2C12 cells. GSAT MM sequences was also enriched in H3K9me3 ChIP-seq.
Unprocessed images of key blots/gels in whole study.
About this article
Nature Communications (2019)