DNA contains all the information and instructions needed to build an organism and enables its cells to fulfil their role. There are many different cell types in animals, and each type only needs a small portion of the information found in the DNA to do its job. Hence, only the sections – also known as genes – specific to a particular role need to be active or ‘expressed’ in any given cell type. The sets of genes that are active in a cell determine what that cell can do and make it different from other cell types. Short regions that surround the genes act as ‘switches’ to control their activity.

To study how these switches work, researchers have developed techniques that can measure when specific ones are on or off in different cell types. A technique called NOMe-seq can simultaneously identify active and inactive regions in the genome by measuring markers that are specific for active and inactive regions. However, this approach was created for large samples containing thousands of cells, so-called bulk samples. Since the pattern of gene expression can vary between individual cells even if they are of the same cell type, it is important to analyse each individual cell.

Sebastian Pott has now adapted the NOMe-seq technique for use in single cells. The modified protocol was used to study human cells that have been well studied and can be grown in the laboratory. The results of this proof-of-principle study showed that the adapted version of the technique obtained similar results as when used on bulk samples. This demonstrates that the method can reliably measure active and inactive regions in single cells across the genome.

The next step will be to use this tool to study how genes are affected in diseases like cancer, where gene expression can be variable between individual cells. For example, understanding the differences between individual cells within a cancer sample might help to understand why some cancers react to treatments better than others.

Extensive transcriptional variation between individual cells has been observed using single cell RNA-seq. These data facilitate identification of functional subpopulations in seemingly homogeneous cell populations (Shalek et al., 2014), or characterization of the cellular composition of complex tissues (Jaitin et al., 2014; Treutlein et al., 2014; Macosko et al., 2015). To gain mechanistic insights into regulatory features that underlie cellular heterogeneity it is essential to measure chromatin organization in individual cells. A number of methods that map chromatin organization in populations of cells have been adapted for single cells, including ATAC-seq (Cusanovich et al., 2015; Buenrostro et al., 2015b), DNase-seq (Jin et al., 2015), methylome sequencing (Smallwood et al., 2014; Farlik et al., 2015), and ChIP-seq (Rotem et al., 2015). Interpretation of these data in single cells is complicated because of the near binary and extremely sparse signal (Cusanovich et al., 2015; Buenrostro et al., 2015b; Maurano and Stamatoyannopoulos, 2015). Nucleosome Occupancy and Methylome-sequencing (NOMe-seq) (Kelly et al., 2012) employs the GpC methyltransferase (MTase) from M.CviPI to probe chromatin accessibility (Kelly et al., 2012; Kilgore et al., 2007). The GpC MTase methylates cytosines in GpC dinucleotides in non-nucleosomal DNA in vitro. Combined with high-throughput bisulfite sequencing this approach has been used to characterize nucleosome positioning and endogenous methylation in human cell lines (Kelly et al., 2012; Taberlay et al., 2014) and in selected promoters of single yeast cells (Small et al., 2014). NOMe-seq data have several unique features that are advantageous considering the challenges associated with single cell measurements (Figure 1a). First, NOMe-seq simultaneously measures chromatin accessibility (through GpC methylation) and endogenous CpG methylation. Chromatin accessibility indicates whether a putative regulatory region might be utilized in a given cell (ENCODE Project Consortium, 2012), while endogenous DNA methylation in regulatory regions has been connected to a variety of regulatory processes often associated with repression (Schübeler, 2015). The ability to combine complementary assays within single cells is essential for a comprehensive genomic characterization of individual cells since each cell represents a unique biological sample which is almost inevitably destroyed in the process of the measurement. Second, each sequenced read might contain several GpCs which independently report the accessibility status along the length of that read. NOMe-seq therefore captures additional information compared to purely count-based methods, such as ATAC-seq and DNase-seq, which increases the confidence associated with the measurements and allows detection of footprints of individual transcription factor (TF) binding events in single cells. Third, the DNA is recovered and sequenced independently of its methylation status, which is a pre-requisite to distinguish between true negatives (i.e. closed chromatin) and false negatives (i.e. loss of DNA) when assessing accessibility at specified locations in single cells. This is especially important in single cells where allelic drop-out is pervasive. In single cells, NOMe-seq can therefore measure the fraction of accessible regions among a set of covered, pre-defined genomic locations. In this proof- of-principle study, I showed that NOMe-seq, which previously had only been performed on bulk samples (Kelly et al., 2012; Taberlay et al., 2014), can be performed on single cells. In addition to endogenous methylation at CpG dinucleotides, single cell NOMe-seq (scNOMe-seq) measured chromatin accessibility at DHSs and TF binding sites in individual cells, and detected footprints of CTCF binding at individual loci. Finally, the average phasing distance between nucleosomes within individual cells can also be estimated from scNOMe-seq data.

Figure 1 with 3 supplements see all

Download asset Open asset

To adapt the NOMe-seq protocol (Kelly et al., 2012; Miranda et al., 2010) to single cells, individual nuclei were first incubated with GpC MTase and then sorted into wells of a 96-well plate using fluorescence-activated cell sorting (FACS) (Figure 1b and Figure 1—figure supplement 1). DNA from isolated nuclei was subjected to bisulfite conversion and sequencing libraries were prepared using a commercial kit for amplification of low amounts of bisulfite-converted DNA (Materials and methods). To assess the feasibility and performance of NOMe-seq in single cells, I used the well-characterized cell lines GM12878 and K562. The scNOMe-seq datasets in this study represent 19 individual GM12878 cells and 11 individual K562 cells. The set of GM12878 cells included seven control cells that were not treated with GpC MTase (Figure 1—figure supplement 2). Each GpC MTase-treated library was sequenced to at least 16 M individual reads (Materials and methods). Reads were aligned to the human genome using the aligner Bismark (Krueger et al., 2012) and, after removal of duplicate reads, between 2.5M and 5M reads were retained per library (Supplementary file 1). On average 6,679,864 (2.9%) of all cytosines in GpCs and 1,291,180 (3.6%) of all cytosines in CpGs were covered per cell (Figure 2—figure supplement 1 and Supplementary file 1).

scNOMe-seq accurately detected accessible chromatin at DNaseI hypersensitive sites

To test whether the GpC methylation observed in GpC MTase treated samples (Figure 2—figure supplement 1) captured known chromatin accessibility patterns, I focused on DNaseI hypersensitive sites (DHSs) that were previously identified in GM12878 and K562 cell lines (ENCODE Project Consortium, 2012). DHSs were associated with strong enrichment of GpC methylation, both in data from pooled and individual GM12878 (Figure 2a,b, Figure 2—figure supplement 2) and K562 cells (Figure 2—figure supplements 3 and 4). Conversely, endogenous CpG methylation decreased around the center of the DHSs in agreement with previous reports (Stadler et al., 2011; Ziller et al., 2015) (Figure 2a and Figure 2—figure supplement 3). These data show that scNOMe-seq detected chromatin accessibility at DHSs. To assess how many of the known DHSs regions were recovered in a single cell, I first filtered DHSs that contained GpC dinucleotides within their primary sequence and thus could be theoretically detected by NOMe-seq. The frequent occurrence of GpC di-nucleotides renders the majority (>85%) of DHSs detectable by NOMe-seq (Figure 2—figure supplements 5 and 6). Of the theoretically detectable DHSs, 10.6% (20388/191566) and 17.3% (33182/191598) had one or more GpCs covered and, using a more stringent criterion, 5.2% (9083/174896) and 9.5% (16608/174828) were covered at four or more GpCs in individual GM12878 cells and K562 cells, respectively (Figure 2c). Chromatin accessibility signal can vary along the length of a given DHSs due to binding of transcription factors (Neph et al., 2012) and the specific position of a GpC within a DHS will thus affect its chance of being methylated. To account for this variability and to obtain more robust estimates of GpC methylation only DHSs with at least four covered GpC were used for the subsequent analyses and referred to as ‘covered DHSs’.

Figure 2 with 15 supplements see all

Download asset Open asset

In single cells, the average GpC methylation at covered DHSs was strongly correlated with the observed DNaseI accessibility at these sites in bulk populations (Figure 2d, Figure 2—figure supplements 7 and 8). The opposite trend was observed for endogenous CpG methylation which was lowest for DHSs with the highest DNaseI accessibility (Figure 2—figure supplement 7). The correlation between GpC methylation and DNaseI accessibility was lower for scNOMe-seq data compared to bulk NOMe-seq data in the same cell line (Figure 2—figure supplement 8). At the level of individual sites the distribution of GpC methylation suggested that around 50% of the covered DHS showed less than 25% GpC methylation in individual cells (Figure 2—figure supplement 9). To estimate the proportion of covered DHSs that were concurrently accessible in a single cell I applied a fixed threshold of 40% GpC methylation above which sites were considered accessible (Materials and methods). At this GpC methylation threshold 32–44% and 26–37% of all covered DHSs were determined to be accessible in single GM12878 and K562 cells, respectively. As expected these results depended to some degree on the cutoffs used for GpC methylation and the number of required GpCs per DHS. However, even under the most lenient conditions less than 50% of DHSs were accessible in individual cells (Figure 2—figure supplement 10). Grouping the DHSs based on DNaseI accessibility in bulk samples, confirmed that the degree of DNaseI accessibility related closely to the frequency of DHS accessibility in single cells (Figure 2e). This analysis leveraged the NOMe-seq-specific property that the DNA sequence is recovered independently of its accessibility status. It provided direct evidence for the notion that the degree of DNaseI accessibility observed in DNase-seq of bulk samples reflects the frequency with which a region is accessible in individual cells. Consequently, chromatin accessibility between cells is less variable at regions with high DNaseI accessibility in bulk samples (Figure 2—figure supplement 11). Correspondingly, correlation of GpC methylation between individual cells is stronger at DHS loci compared to randomized locations (Figure 2—figure supplement 12).

scNOMe-seq captured characteristic chromatin organization associated with transcription

Chromatin accessibility and endogenous methylation show characteristic patterns at gene promoters and within gene bodies (Schübeler, 2015; ENCODE Project Consortium, 2012). To test whether these features can be observed in scNOMe-seq data, I first plotted the average GpC and CpG methylation around transcription start sites (TSS). The average GpC methylation showed the expected increase of chromatin accessibility directly upstream of the TSS (Figure 3a, Figure 3—figure supplement 1). In contrast, and as expected, the endogenous CpG methylation decreased towards the TSS (Figure 3b). To visualize the distribution of CpG methylation throughout entire gene loci, I plotted the aggregated CpG methylation across regions containing the entire gene body and 50 kb upstream and 50 kb downstream of each gene (Figure 3c, Figure 3—figure supplement 1). Endogenous methylation was specifically reduced at the narrow promoter region and gradually increased throughout the gene body. Downstream of the transcription end site (TES) the average level CpG methylation level fell back to the non-genic background level. Endogenous CpG methylation is typically increased within highly expressed genes (Schübeler, 2015). This trend was clearly apparent in the single cell data where gene body methylation was highest in highly expressed genes (Figure 3d, Figure 3—figure supplement 1). Correspondingly, in promoter regions (−500 bp to +150 bp) chromatin accessibility (GpC methylation) increased with the transcript level of the adjacent gene (Figure 3e, Figure 3—figure supplement 2). In contrast to chromatin accessibility, endogenous methylation was lowest in promoters of genes with high transcript levels (Figure 3f). These data show that scNOMe-seq recapitulated known characteristics of chromatin accessibility and endogenous methylation at gene promoters and within gene bodies.

Figure 3 with 2 supplements see all

Download asset Open asset

GpC methylation and endogenous CpG methylation data separated individual GM12878 and K562 cells

A potentially powerful application for single cell genomic approaches is the label-free classification of single cells from heterogeneous mixtures of cells solely based on the measured feature (Cusanovich et al., 2015; Buenrostro et al., 2015a; Jaitin et al., 2014; Macosko et al., 2015). Of note, using a union set of DHSs from both cell types was sufficient to classify individual GM12878 and K562 cells into their respective cell types based on GpC methylation (Figure 4a, Figure 4—figure supplement 1). While this assessment might have been influenced in part by the separate processing of the cell types, both cell types showed preferential enrichment of GpC methylation at their respective DHSs compared to DHSs identified in the other cell type (Figure 4b). Similar to GpC methylation, endogenous CpG methylation at multiple sets of genomic features was sufficient to separate the cells into the respective cell types (Figure 4c, Figure 4—figure supplement 1).

Figure 4 with 1 supplement see all

Download asset Open asset

Detection of footprints of CTCF binding at individual loci in single cells

To examine in detail whether scNOMe-seq captures features of chromatin accessibility that are specifically associated with transcription factor binding, I analyzed scNOMe-seq data at transcription factor binding sites (TFBS). The average GpC methylation around CTCF ChIP-seq peaks (ENCODE Project Consortium, 2012) in single cells recapitulated the accessibility previously observed in NOMe-seq bulk samples (Kelly et al., 2012): Accessibility increased strongly towards the CTCF binding sites while the location of the CTCF motif at the center of the region showed low accessibility suggesting that CTCF binding protected from GpC MTase activity and thus creating a footprint of a CTCF binding event, both when averaged across data from all single cells (Figure 5a and Figure 5—figure supplement 1) and in individual cells (Figure 5b and Figure 5—figure supplement 2). In contrast, endogenous CpG methylation was generally depleted around the center of CTCF binding sites (Figure 5a and Figure 5—figure supplement 1). Similar accessibility profiles, albeit less pronounced compared to CTCF, were observed for additional transcription factors, for example EBF1 and PU.1 (Figure 5—figure supplement 3). These analyses provided evidence that, in aggregate, scNOMe-seq detected chromatin accessibility characteristic of CTCF binding in single cells. To test whether scNOME-seq data detected CTCF footprints at individual motifs loci, GpC methylation at motifs within CTCF ChIP-seq peaks was compared to the GpC methylation level in the regions flanking each motif (Figure 5c). On average, two-thirds of CTCF motif instances within these accessible regions showed no GpC methylation, suggesting that CTCF binding prevented the GpC MTase from methylating the cytosines within the binding motif and thus creating a footprint (Figure 5d and f). Of note, motifs associated with a footprint had significantly higher scores than motifs without a footprint suggesting that the motif score is a strong determinant of CTCF binding within these accessible regions (p= 5.429e-12, paired t-test) (Figure 5e,g and Figure 5—figure supplement 4). Of note, the CTCF footprints could be observed at individual loci within individual cells and were shared across cells (Figure 5h and Figure 5—figure supplement 5).

Figure 5 with 5 supplements see all

Download asset Open asset

Estimating nucleosome phasing in single cells

The pattern of GpC methylation adjacent to CTCF sites suggested that scNOMe-seq also detected the well-positioned nucleosomes flanking these regions (Figure 5a) (Kelly et al., 2012). This observation was confirmed by the oscillatory distribution of the average GpC and CpG methylation around locations of well-positioned nucleosomes identified from MNase-seq data (ENCODE Project Consortium, 2012) (Figure 6a). While nucleosome core particles are invariably associated with DNA fragments of 147 bp, nucleosomes are separated by linker DNA of varying lengths, resulting in different packaging densities between cell types and between genomic regions within a cell (Valouev et al., 2011; Schones et al., 2008). To determine whether scNOMe-seq data can be used to measure the average linker length, average distances between nucleosome midpoints in single cells (phasing distances) were estimated by correlating the methylation status between pairs of cytosines in GpC di-nucleotides at offset distances from 3 bp to 400 bp (Figure 6c,d and Figure 6—figure supplements 1 and 2). The estimated phases fell between 187 bp and 196 bp (mean = 196.7 bp) in GM12878 cells, and between 188 bp and 200 bp (mean = 194.2 bp) in K562 cells (Figure 6e). These estimates are in general agreement with phase estimates derived from MNase-seq data in human cells (Valouev et al., 2011). In addition, estimated phasing distances varied within individual cells depending on the chromatin context, similar to observation from bulk MNase-seq data (Valouev et al., 2011) (Figure 6f).

Figure 6 with 2 supplements see all

Download asset Open asset

In this study, I demonstrated that scNOMe-seq simultaneously measures chromatin accessibility by GpC methylation as well as endogenous CpG and DNA methylation in single cells. scNOMe-seq detected chromatin accessibility at DHSs and TFBS and, in aggregate, these data recapitulated NOMe-seq data obtained from bulk cells (Kelly et al., 2012). scNOMe-seq data also detected footprints of CTCF binding, and was used to estimate nucleosome phasing distances.

Similar to other single cell genomic methods, scNOMe-seq relies on annotations obtained from bulk measurements (Cusanovich et al., 2015; Buenrostro et al., 2015b; Smallwood et al., 2014; Farlik et al., 2015). A limitation of single cell genomic methods is their sparse coverage which leads to high allelic drop-out. For methods in which the signal is based on counting the sequenced fragments, such as ATAC-seq and DNase-seq, this poses a challenge since true negatives at a specific location cannot be distinguished from false negatives that are a consequence of read loss. Compared to these methods, scNOMe-seq has the unique advantage, that reads are recovered independently of the signal and allelic drop-out events therefore can be distinguished from closed or inaccessible chromatin configurations. The frequency of accessible sites in the population of DHSs can be estimated. Using this approach only about 30–50% of DHSs detected in the population were found accessible in a single cell, depending on the thresholds chosen to call a site accessible. While this assessment would have been possible using bulk NOMe-seq data, scNOMe-seq offers important possibilities for future applications. For example, to compare accessibility across multiple loci within a single cell and the use of heterogeneous cellular mixtures as input material.

As expected, the chance of a covered DHS to being open or closed is not equally distributed across all DHSs from the population. Instead, DHSs with strong DNaseI accessibility showed a higher frequency of accessibility in single cells compared to those sites with low DNaseI accessibility in the population (Figure 2e) suggesting that the peak height is indeed directly related to the frequency with which a site is accessible in individual cells. In agreement with this observation a large proportion of variability observed between cells was attributable to DHSs with low DNaseI accessibility in bulk samples (Figure 2—figure supplement 11). In principle, variation between cells could be due to differential GpC MTase enzyme activity. However, the genome-wide levels of GpC methylation reached comparable levels in all cells and the variability between cells was not equally distributed across all DHS (Figure 2d, Figure 2—figure supplement 1)

Measuring similarity of chromatin accessibility between cells was sufficient to group GM12878 and K562 cells based on their cell type of origin (Figure 3a). In this particular case, the separation is confounded with experimental batches. However, higher average GpC methylation in DHSs for the respective cell type compared to the DHSs of the other cell type indicated that scNOMe-seq can differentiate the two cell types (Figure 2—figure supplement 14). Similarly, endogenous CpG methylation at different genomic features (DHS, 10 kb windows, gene bodies) was sufficient to distinguish between the two cell types. This approach should be extendable to scNOMe-seq data from samples containing mixtures of cell types.

scNOMe-seq measures chromatin accessibility at GpC di-nucleotides along the entire length of a sequencing read. Since most features that bind DNA are smaller than the length of 100 bp (200 bp within 200–500 bp regions in the case of paired end reads), the regions covered by sequence-specific transcription factors and nucleosomes can be captured within a single fragment. This allows one to directly detect binding of TFs provided that their sequencing motif contains at least one GpC di-nucleotide. I demonstrated the feasibility of this approach using CTCF binding sites. Of note, most motifs within regions of CTCF ChIP-seq peaks were protected from GpC methylation (‘footprint’) (Figure 5). In agreement with an inferred binding event as the cause for this protection, scores for CTCF motifs that were associated with a footprint were significantly higher than for motifs without a footprint. Depending on the motif specificity of a given TF and provided that their motifs contain a GpC dinucleotide, similar measurements should be feasible for many TFs and could be used to infer the activity of a range of transcription factors in single cells or to measure combinatorial binding of two or more TFs.

Estimation of the average nucleosome phasing distances allows one to study chromatin compaction and complements the measurements of chromatin accessibility at regulatory regions and DNA methylation. The estimates from individual cells fit very well with measurements made from MNase-seq data in bulk samples (Valouev et al., 2011). It remains to be established whether the variation in phasing distances between individual cells is of biological or technical nature (Figure 6e).

These proof-of-principle experiments have been performed using commercial kits for bisulfite conversion and library amplification, additional optimization or alternative amplification approaches (Smallwood et al., 2014) are likely to increase the yield substantially. Compared to other single cell methods, for example ATAC-seq, scNOMe-seq does not enrich for accessible chromatin regions and thus requires significantly more sequencing coverage. Ultimately, it should be possible to integrate the GpC MTase treatment into microfluidic workflows and combine this method with scRNA-seq, similar to recently published methods that combine scRNA-seq and methylome- sequencing (Angermueller et al., 2016). This study was primarily designed to test the feasibility of NOMe-seq in single cells and only a small number of nuclei where sequenced for each cell line. As a consequence, this set up could not be used to study cell-to-cell variation in detail. scNOMe-seq will be particularly useful for studies that aim to simultaneously measure chromatin accessibility and DNA methylation. This approach will be especially powerful for the characterization of chromatin organization in single cells from heterogeneous mixtures or complex tissues, for example to samples of brain tissues or primary cancer cells.

1. 1. Pott S

   (2016) Simultaneous measurement of chromatin accessibility, DNA methylation, and nucleosome phasing in single cells

   Publicly available at the NCBIGene Expression Omnibus (accession no: GSE83882).

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE83882

1. 1. The ENCODE Project Consortium

   (2012) DNaseI HS peaks; GM12878 cells

   Publicly available under link provided (file name: wgEncodeRegDnaseUwGm12878Peak).

   http://hgdownload.cse.ucsc.edu/goldenPath/hg38/database/
2. 1. The ENCODE Project Consortium

   (2012) DNaseI HS peaks; K562 cells

   Publicly available under link provided (file name: wgEncodeRegDnaseUwK562Peak).

   http://hgdownload.cse.ucsc.edu/goldenPath/hg38/database/
3. 1. The ENCODE Project Consortium

   (2012) CTCF ChIP-seq; GM12878 cells

   Publicly available under link provided (file name: wgEncodeAwgTfbsBroadGm12878CtcfUniPk.narrowPeak.gz).

   http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeAwgTfbsUniform/
4. 1. The ENCODE Project Consortium

   (2012) CTCF ChIP-seq; K562 cells

   Publicly available under link provided (file name: wgEncodeAwgTfbsBroadK562CtcfUniPk.narrowPeak.gz).

   http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeAwgTfbsUniform/
5. 1. The ENCODE Project Consortium

   (2012) Pu.1 ChIP-seq; GM12878 cells

   Publicly available under link provided (file name: wgEncodeAwgTfbsHaibGm12878Pu1Pcr1xUniPk.narrowPeak.gz).

   http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeAwgTfbsUniform/
6. 1. The ENCODE Project Consortium

   (2012) Pu.1 ChIP-seq; K562 cells

   Publicly available under link provided (file name: wgEncodeAwgTfbsHaibK562Pu1Pcr1xUniPk.narrowPeak.gz).

   http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeAwgTfbsUniform/
7. 1. The ENCODE Project Consortium

   (2012) EBF1 ChIP-seq; GM12878 cells

   Publicly available under link provided (file name: wgEncodeAwgTfbsHaibGm12878Ebf1sc137065Pcr1xUniPk.narrowPeak.gz).

   http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeAwgTfbsUniform/
8. 1. The ENCODE Project Consortium

   (2012) MNase-seq; GM12878

   aligned Bam files publicly available under link provided.

   http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeSydhNsome/
9. 1. Lay FD

   (2015) NOMe-seq; K562

   Publicly available at the NCBIGene Expression Omnibus (accession no: GSM1583563).

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM1583563
10. 1. Lay FD

    (2015) NOMe-seq; K562

    Publicly available at the NCBIGene Expression Omnibus (accession no: GSM1583567).

    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM1583567

1. 1. Akalin A
   2. Franke V
   3. Vlahoviček K
   4. Mason CE
   5. Schübeler D

   (2015) Genomation: a toolkit to summarize, annotate and visualize genomic intervals

   Bioinformatics 31:1127–1129.

   https://doi.org/10.1093/bioinformatics/btu775

   • PubMed
   • Google Scholar
2. 1. Angermueller C
   2. Clark SJ
   3. Lee HJ
   4. Macaulay IC
   5. Teng MJ
   6. Hu TX
   7. Krueger F
   8. Smallwood SA
   9. Ponting CP
   10. Voet T
   11. Kelsey G
   12. Stegle O
   13. Reik W

   (2016) Parallel single-cell sequencing links transcriptional and epigenetic heterogeneity

   Nature Methods 13:229–232.

   https://doi.org/10.1038/nmeth.3728

   • PubMed
   • Google Scholar
3. 1. Buenrostro JD
   2. Wu B
   3. Chang HY
   4. Greenleaf WJ

   (2015a) ATAC-seq: a method for assaying chromatin accessibility Genome-Wide

   Current Protocols in Molecular Biology 109:21.29.1–21.29.9.

   https://doi.org/10.1002/0471142727.mb2129s109

   • PubMed
   • Google Scholar
4. 1. Buenrostro JD
   2. Wu B
   3. Litzenburger UM
   4. Ruff D
   5. Gonzales ML
   6. Snyder MP
   7. Chang HY
   8. Greenleaf WJ

   (2015b) Single-cell chromatin accessibility reveals principles of regulatory variation

   Nature 523:486–490.

   https://doi.org/10.1038/nature14590

   • PubMed
   • Google Scholar
5. 1. Chen K
   2. Xi Y
   3. Pan X
   4. Li Z
   5. Kaestner K
   6. Tyler J
   7. Dent S
   8. He X
   9. Li W

   (2013) DANPOS: dynamic analysis of nucleosome position and occupancy by sequencing

   Genome Research 23:341–351.

   https://doi.org/10.1101/gr.142067.112

   • PubMed
   • Google Scholar
6. 1. Cusanovich DA
   2. Daza R
   3. Adey A
   4. Pliner HA
   5. Christiansen L
   6. Gunderson KL
   7. Steemers FJ
   8. Trapnell C
   9. Shendure J

   (2015) Multiplex single cell profiling of chromatin accessibility by combinatorial cellular indexing

   Science 348:910–914.

   https://doi.org/10.1126/science.aab1601

   • PubMed
   • Google Scholar
7. 1. ENCODE Project Consortium

   (2012) An integrated encyclopedia of DNA elements in the human genome

   Nature 489:57–74.

   https://doi.org/10.1038/nature11247

   • PubMed
   • Google Scholar
8. 1. Ernst J
   2. Kheradpour P
   3. Mikkelsen TS
   4. Shoresh N
   5. Ward LD
   6. Epstein CB
   7. Zhang X
   8. Wang L
   9. Issner R
   10. Coyne M
   11. Ku M
   12. Durham T
   13. Kellis M
   14. Bernstein BE

   (2011) Mapping and analysis of chromatin state dynamics in nine human cell types

   Nature 473:43–49.

   https://doi.org/10.1038/nature09906

   • PubMed
   • Google Scholar
9. 1. Farlik M
   2. Sheffield NC
   3. Nuzzo A
   4. Datlinger P
   5. Schönegger A
   6. Klughammer J
   7. Bock C

   (2015) Single-cell DNA methylome sequencing and bioinformatic inference of epigenomic cell-state dynamics

   Cell Reports 10:1386–1397.

   https://doi.org/10.1016/j.celrep.2015.02.001

   • PubMed
   • Google Scholar
10. 1. Jaitin DA
    2. Kenigsberg E
    3. Keren-Shaul H
    4. Elefant N
    5. Paul F
    6. Zaretsky I
    7. Mildner A
    8. Cohen N
    9. Jung S
    10. Tanay A
    11. Amit I

    (2014) Massively parallel single-cell RNA-seq for marker-free decomposition of tissues into cell types

    Science 343:776–779.

    https://doi.org/10.1126/science.1247651

    • PubMed
    • Google Scholar
11. 1. Jin W
    2. Tang Q
    3. Wan M
    4. Cui K
    5. Zhang Y
    6. Ren G
    7. Ni B
    8. Sklar J
    9. Przytycka TM
    10. Childs R
    11. Levens D
    12. Zhao K

    (2015) Genome-wide detection of DNase I hypersensitive sites in single cells and FFPE tissue samples

    Nature 528:1–17.

    https://doi.org/10.1038/nature15740

    • PubMed
    • Google Scholar
12. 1. Karolchik D
    2. Barber GP
    3. Casper J
    4. Clawson H
    5. Cline MS
    6. Diekhans M
    7. Dreszer TR
    8. Fujita PA
    9. Guruvadoo L
    10. Haeussler M
    11. Harte RA
    12. Heitner S
    13. Hinrichs AS
    14. Learned K
    15. Lee BT
    16. Li CH
    17. Raney BJ
    18. Rhead B
    19. Rosenbloom KR
    20. Sloan CA
    21. Speir ML
    22. Zweig AS
    23. Haussler D
    24. Kuhn RM
    25. Kent WJ

    (2014) The UCSC Genome browser database: 2014 update

    Nucleic Acids Research 42:D764–D770.

    https://doi.org/10.1093/nar/gkt1168

    • PubMed
    • Google Scholar
13. 1. Kelly TK
    2. Liu Y
    3. Lay FD
    4. Liang G
    5. Berman BP
    6. Jones PA

    (2012) Genome-wide mapping of nucleosome positioning and DNA methylation within individual DNA molecules

    Genome Research 22:2497–2506.

    https://doi.org/10.1101/gr.143008.112

    • PubMed
    • Google Scholar
14. 1. Kent WJ
    2. Sugnet CW
    3. Furey TS
    4. Roskin KM
    5. Pringle TH
    6. Zahler AM
    7. Haussler D

    (2002) The human genome browser at UCSC

    Genome Research 12:996–1006.

    https://doi.org/10.1101/gr.229102

    • PubMed
    • Google Scholar
15. 1. Kilgore JA
    2. Hoose SA
    3. Gustafson TL
    4. Porter W
    5. Kladde MP

    (2007) Single-molecule and population probing of chromatin structure using DNA methyltransferases

    Methods 41:320–332.

    https://doi.org/10.1016/j.ymeth.2006.08.008

    • PubMed
    • Google Scholar
16. 1. Krueger F
    2. Kreck B
    3. Franke A
    4. Andrews SR

    (2012) DNA methylome analysis using short bisulfite sequencing data

    Nature Methods 9:145–151.

    https://doi.org/10.1038/nmeth.1828

    • PubMed
    • Google Scholar
17. 1. Langmead B
    2. Salzberg SL

    (2012) Fast gapped-read alignment with bowtie 2

    Nature Methods 9:357–359.

    https://doi.org/10.1038/nmeth.1923

    • PubMed
    • Google Scholar
18. 1. Li H
    2. Handsaker B
    3. Wysoker A
    4. Fennell T
    5. Ruan J
    6. Homer N
    7. Marth G
    8. Abecasis G
    9. Durbin R
    10. 1000 Genome Project Data Processing Subgroup

    (2009) The sequence alignment/Map format and SAMtools

    Bioinformatics 25:2078–2079.

    https://doi.org/10.1093/bioinformatics/btp352

    • PubMed
    • Google Scholar
19. 1. Macosko EZ
    2. Basu A
    3. Satija R
    4. Nemesh J
    5. Shekhar K
    6. Goldman M
    7. Tirosh I
    8. Bialas AR
    9. Kamitaki N
    10. Martersteck EM
    11. Trombetta JJ
    12. Weitz DA
    13. Sanes JR
    14. Shalek AK
    15. Regev A
    16. McCarroll SA

    (2015) Highly parallel Genome-wide expression profiling of individual cells using Nanoliter Droplets

    Cell 161:1202–1214.

    https://doi.org/10.1016/j.cell.2015.05.002

    • PubMed
    • Google Scholar
20. 1. Maurano MT
    2. Stamatoyannopoulos JA

    (2015) Taking stock of regulatory variation

    Cell Systems 1:18–21.

    https://doi.org/10.1016/j.cels.2015.07.011

    • PubMed
    • Google Scholar
21. 1. Miranda TB
    2. Kelly TK
    3. Bouazoune K
    4. Jones PA

    (2010) Methylation-sensitive single-molecule analysis of chromatin structure

    Current Protocols in Molecular Biology Chapter 21:Unit 21.17.1–21.1716.

    https://doi.org/10.1002/0471142727.mb2117s89

    • PubMed
    • Google Scholar
22. 1. Neph S
    2. Vierstra J
    3. Stergachis AB
    4. Reynolds AP
    5. Haugen E
    6. Vernot B
    7. Thurman RE
    8. John S
    9. Sandstrom R
    10. Johnson AK
    11. Maurano MT
    12. Humbert R
    13. Rynes E
    14. Wang H
    15. Vong S
    16. Lee K
    17. Bates D
    18. Diegel M
    19. Roach V
    20. Dunn D
    21. Neri J
    22. Schafer A
    23. Hansen RS
    24. Kutyavin T
    25. Giste E
    26. Weaver M
    27. Canfield T
    28. Sabo P
    29. Zhang M
    30. Balasundaram G
    31. Byron R
    32. MacCoss MJ
    33. Akey JM
    34. Bender MA
    35. Groudine M
    36. Kaul R
    37. Stamatoyannopoulos JA

    (2012) An expansive human regulatory lexicon encoded in transcription factor footprints

    Nature 489:83–90.

    https://doi.org/10.1038/nature11212

    • PubMed
    • Google Scholar
23. 1. Pages H
    2. Aboyoun P
    3. Gentleman RC
    4. DebRoy S

    (2016)

    Biostrings: String Objects Representing Biological Sequences, and Matching Algorithms

    Biostrings: String Objects Representing Biological Sequences, and Matching Algorithms.

    • Google Scholar
24. 1. Quinlan AR
    2. Hall IM

    (2010) BEDTools: a flexible suite of utilities for comparing genomic features

    Bioinformatics 26:841–842.

    https://doi.org/10.1093/bioinformatics/btq033

    • PubMed
    • Google Scholar
25. Software

    1. R Core Team

    (2015) R: A Language and Environment for Statistical Computing

    R: A Language and Environment for Statistical Computing.

    https://www.R-project.org/
26. 1. Rotem A
    2. Ram O
    3. Shoresh N
    4. Sperling RA
    5. Goren A
    6. Weitz DA
    7. Bernstein BE

    (2015) Single-cell ChIP-seq reveals cell subpopulations defined by chromatin state

    Nature Biotechnology 33:1165–1172.

    https://doi.org/10.1038/nbt.3383

    • PubMed
    • Google Scholar
27. 1. Schones DE
    2. Cui K
    3. Cuddapah S
    4. Roh TY
    5. Barski A
    6. Wang Z
    7. Wei G
    8. Zhao K

    (2008) Dynamic regulation of nucleosome positioning in the human genome

    Cell 132:887–898.

    https://doi.org/10.1016/j.cell.2008.02.022

    • PubMed
    • Google Scholar
28. 1. Schübeler D

    (2015) Function and information content of DNA methylation

    Nature 517:321–326.

    https://doi.org/10.1038/nature14192

    • PubMed
    • Google Scholar
29. 1. Shalek AK
    2. Satija R
    3. Shuga J
    4. Trombetta JJ
    5. Gennert D
    6. Lu D
    7. Chen P
    8. Gertner RS
    9. Gaublomme JT
    10. Yosef N
    11. Schwartz S
    12. Fowler B
    13. Weaver S
    14. Wang J
    15. Wang X
    16. Ding R
    17. Raychowdhury R
    18. Friedman N
    19. Hacohen N
    20. Park H
    21. May AP
    22. Regev A

    (2014) Single-cell RNA-seq reveals dynamic paracrine control of cellular variation

    Nature 510:263–269.

    https://doi.org/10.1038/nature13437

    • PubMed
    • Google Scholar
30. 1. Small EC
    2. Xi L
    3. Wang JP
    4. Widom J
    5. Licht JD

    (2014) Single-cell nucleosome mapping reveals the molecular basis of gene expression heterogeneity

    PNAS 111:E2462–E2471.

    https://doi.org/10.1073/pnas.1400517111

    • PubMed
    • Google Scholar
31. 1. Smallwood SA
    2. Lee HJ
    3. Angermueller C
    4. Krueger F
    5. Saadeh H
    6. Peat J
    7. Andrews SR
    8. Stegle O
    9. Reik W
    10. Kelsey G

    (2014) Single-cell genome-wide bisulfite sequencing for assessing epigenetic heterogeneity

    Nature Methods 11:817–820.

    https://doi.org/10.1038/nmeth.3035

    • PubMed
    • Google Scholar
32. 1. Stadler MB
    2. Murr R
    3. Burger L
    4. Ivanek R
    5. Lienert F
    6. Schöler A
    7. van Nimwegen E
    8. Wirbelauer C
    9. Oakeley EJ
    10. Gaidatzis D
    11. Tiwari VK
    12. Schübeler D

    (2011) DNA-binding factors shape the mouse methylome at distal regulatory regions

    Nature 480:1–7.

    https://doi.org/10.1038/nature10716

    • PubMed
    • Google Scholar
33. 1. Taberlay PC
    2. Statham AL
    3. Kelly TK
    4. Clark SJ
    5. Jones PA

    (2014) Reconfiguration of nucleosome-depleted regions at distal regulatory elements accompanies DNA methylation of enhancers and insulators in Cancer

    Genome Research 24:1421–1432.

    https://doi.org/10.1101/gr.163485.113

    • PubMed
    • Google Scholar
34. Software

    1. Tan G

    (2014) Data package for JASPAR

    JASPAR.

    http://jaspar.genereg.net/
35. 1. Treutlein B
    2. Brownfield DG
    3. Wu AR
    4. Neff NF
    5. Mantalas GL
    6. Espinoza FH
    7. Desai TJ
    8. Krasnow MA
    9. Quake SR

    (2014) Reconstructing lineage hierarchies of the distal lung epithelium using single-cell RNA-seq

    Nature 509:371–375.

    https://doi.org/10.1038/nature13173

    • PubMed
    • Google Scholar
36. 1. Valouev A
    2. Johnson SM
    3. Boyd SD
    4. Smith CL
    5. Fire AZ
    6. Sidow A

    (2011) Determinants of nucleosome organization in primary human cells

    Nature 474:516–520.

    https://doi.org/10.1038/nature10002

    • PubMed
    • Google Scholar
37. Software

    1. Warnes GR
    2. Bolker B
    3. Bonebakker L
    4. Gentleman R

    (2016) gplots: Various R programming tools for plotting data

    R Package Version.

    https://CRAN.R-project.org/package=gplots
38. Software

    1. Wickham H

    (2009)

    Ggplot2

    Springer Science & Business Media, New York.
39. 1. Ziller MJ
    2. Edri R
    3. Yaffe Y
    4. Donaghey J
    5. Pop R
    6. Mallard W
    7. Issner R
    8. Gifford CA
    9. Goren A
    10. Xing J
    11. Gu H
    12. Cacchiarelli D
    13. Tsankov AM
    14. Epstein C
    15. Rinn JL
    16. Mikkelsen TS
    17. Kohlbacher O
    18. Gnirke A
    19. Bernstein BE
    20. Elkabetz Y
    21. Meissner A

    (2015) Dissecting neural differentiation regulatory networks through epigenetic footprinting

    Nature 518:355–359.

    https://doi.org/10.1038/nature13990

    • PubMed
    • Google Scholar

Author details

1. Sebastian Pott

   Department of Human Genetics, University of Chicago, Chicago, United States

   Contribution

   SP, Conceptualization, Investigation, Visualization, Writing—original draft, Writing—review and editing

   For correspondence

   spott@uchicago.edu

   Competing interests

   The author declares that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-4118-6150

• 11,426

  views

• 1,400

  downloads

• 191

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.