CUT&Tag for efficient epigenomic profiling of small samples and single cells — TEI-tagged view

Hatice S Kaya-OkurpersName, Steven J. WupersName, Christine A. CodomopersName, Erica S. PledgerpersName, Terri D. BrysonpersName, Jorja G. HenikoffpersName, Kami AhmadpersName, Steven HenikoffpersName (2019date). “CUT&Tag for efficient epigenomic profiling of small samples and single cells.” Nature Communicationstitle·venue.

Object of study — where a given chromatin proteinterm sits on the genome - a quantity with NO GROUND TRUTH, so that every method for measuring it can only be checked against another method for measuring it

Research question — whether tethering a Tn5 transposase to an antibody in intact cells can profile chromatin at high resolution from very few cells - and how you would know it had

Contribution — CUT&Tagterm: antibody-tethered pA-Tn5term tagmentation in situ, giving sequencing-ready libraries in one tube in one day, down to 60 cells and into single cellsterm

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Abstract. Abstract Many chromatin features play critical roles in regulating gene expression. A complete understanding of gene regulation will require the mapping of specific chromatin features in small samples of cells at high resolution. Here we describe Cleavage Under Targets and Tagmentationterm (CUT&Tagterm), an enzyme-tethering strategyterm that provides efficient high-resolution sequencing librariesterm for profiling diverse chromatin componentsterm. In CUT&Tagterm, a chromatin proteinterm is bound in situ by a specific antibodyterm, which then tethers a protein A-Tn5 transposase fusion proteinterm. Activation of the transposaseterm efficiently generates fragment librariesterm with high resolution and exceptionally low background. All steps from live cells to sequencing-ready libraries can be performed in a single tubeterm on the benchtop or a microwellterm in a high-throughput pipelineterm, and the entire procedure can be performed in one day. We demonstrate the utility of CUT&Tagterm by profiling histone modificationsterm, RNA Polymerase IIterm and transcription factorsterm on low cell numbersterm and single cellsterm. Subject terms: Biological techniques, Epigenetics analysis, Chromatin analysis, Biotechnology, Genomics Understanding gene regulation will require mapping specific chromain features in a small number of cells at high resolution. Here the authors describe CUT&Tagterm, which uses antibody-mediated tethering of Tn5 transposase to a chromatin proteinterm to generate high resolution libraries. Introduction The advent of massively parallel sequencing and the dramatic reduction in cost per base has fueled a genomics revolution, however, the full promise of epigenomic profiling has lagged owing to limitations in methodologies used for mapping chromatin fragments to the genome 1 . Chromatin immunoprecipitation with sequencing (ChIP-seqterm) and its variations 2 – 5 suffer from low signalsterm, high backgroundsterm and epitope maskingterm due to cross-linkingterm, and low yieldsterm require large numbers of cells 2 , 6 . Alternatives to ChIP include enzyme-tethering methodsterm for unfixed cells, such as DamIDterm 7 , ChEC-seqterm 8 , and CUT&RUNterm 9 , 10 , where a specific protein of interest is targeted in situ and then profiled genome-wide. For example, CUT&RUNterm, which is based on Laemmli’s Chromatin ImmunoCleavage (ChIC) strategy 11 , maps a chromatin proteinterm by successive binding of a specific antibodyterm, and then tethering a Protein A/Micrococcal Nuclease (pA-MNase) fusion proteinterm in permeabilized cellsterm without cross-linkingterm 9 . MNase is activated by addition of calcium, and fragments are released into the supernatant for extraction of DNA, library preparation and paired-end sequencing. CUT&RUNterm provides base-pair resolution of specific chromatin components with background levels that are much lower than with ChIP-seqterm, dramatically reducing the cost of genome-wide profiling. Although CUT&RUNterm can generate high-quality data from as few as 100–1000 cells, it must be followed by DNA end polishing and adapter ligationterm to prepare sequencing libraries, which increases the time, cost and effortterm of the overall procedure. Moreover, the release of MNase-cleaved fragments into the supernatant with CUT&RUNterm is not well-suited for application to single-cell platformsterm 12 , 13 . Here we overcome the limitations of ChIP-seqterm and CUT&RUNterm using a transposometerm that consists of a hyperactive Tn5 transposaseterm 14 , 15 —Protein A (pA-Tn5term) fusion protein loaded with sequencing adaptersterm. Tethering in situ followed by activation of pA-Tn5term results in factor-targeted tagmentationterm, generating fragments ready for PCR enrichmentterm and DNA sequencing. Beginning with live cells, Cleavage Under Targets and Tagmentationterm (CUT&Tagterm) provides amplified sequence-ready libraries in a day on the bench top or in a high-throughput format. We show that a variety of chromatin components can be profiled with exceptionally low backgrounds using low cell numbersterm and even single cellsterm. This easy, low-cost method will empower epigenetic studies in diverse areas of biological research. Results Efficient profiling of nucleosomes and RNAPII with CUT&Tagterm To implement chromatin profiling by tagmentationterm (Fig. 1a ), we incubated intact permeabilized human K562 cells with an antibody to lysine-27-trimethylation of the histone H3 tail (H3K27me3term), an abundant histone modificationterm that marks silenced chromatin regions. We incubated cells with a secondary anti-rabbit antibodyterm to increase the local concentration of antibodyterm bound on chromatin sites, then incubated cells with an excess of pA-Tn5term fusion protein pre-loaded with sequencing adaptersterm to tether the enzyme at antibody-bound sites in the nucleus. The transposometerm has inherent affinity for exposed DNA 16 , 17 , and so we washed cells under stringent conditionsterm to remove un-tethered pA-Tn5term. We then activated the transposometerm by addition of Mg ++term , integrating adapters spanning sites of H3K27me3-containing nucleosomes. Finally, fragment librariesterm were enriched from purified DNA and pooled for multiplex paired-end sequencing on an Illumina HiSeq flow-cell. The entire protocol manipulates all steps in a single reaction tube (Fig. 1b ), where permeabilized cellsterm are first mixed with an antibody, and then immobilized on Concanavalin A-coated paramagnetic beadsterm, allowing magnetic handling of the cellsterm in all successive wash and reagent incubation steps. For standardization between experiments, we used the small amount of tracer genomic DNAterm derived from the E. coli during transposase protein production to normalize sample read countsterm in lieu of the heterologous spike-in DNA that is recommended for CUT&RUNterm 9 (see Methods section and Supplementary Fig. 1a ). Fig. 1 In situ tethering for CUT&Tagterm chromatin profiling. a The steps in CUT&Tagterm. Added antibody (green) binds to the target chromatin proteinterm (blue) between nucleosomes (gray ovals) in the genome, and the excess is washed away. A second antibody (orange) is added and enhances tethering of pA-Tn5term transposometerm (gray boxes) at antibody-bound sites. After washing away excess transposometerm, addition of Mg ++term activates the transposometerm and integrates adapters (red) at chromatin proteinterm binding sites. After DNA purification genomic fragments with adapters at both ends are enriched by PCR. b CUT&Tagterm is performed on a solid support. Unfixed cells or nuclei (blue) are permeabilized and mixed with antibody to a target chromatin proteinterm. After addition and binding of cells to Concanavilin A-coated magnetic beads (M), all further steps are performed in the same reaction tube with magnetic capture between washes and incubations, including pA-Tn5term tethering, integration, and DNA purification Display of ~8 million reads mapped to the human genome assembly shows a clear pattern of large chromatin domains marked by H3K27me3term (Fig. 2a ). We also obtained profiles for H3K4me1 and H3K4me2 histone modificationsterm, which mark active chromatin sites. In contrast, incubation of cells with a non-specific IgG antibody, which measures untethered integration of adapters, produced very sparse landscapes (Fig. 2a ). To assess the signal-to-noiseterm of CUT&Tagterm relative to other methods we compared it with profiling generated by CUT&RUNterm 18 and by ChIP-seqterm 19 for the same H3K27me3term rabbit monoclonal antibody in K562 cells. To directly compare the three techniques, we set the read depthterm of each dataset to 8 million reads each. Landscapes for each of the three methods are similar, but background noiseterm dominates in ChIP-seqterm datasets (Fig. 2a ), and it is thus appears that ChIP-seqterm will require substantially greater read depthterm to distinguish chromatin features from background. In contrast, both CUT&RUNterm and CUT&Tagterm profiles have extremely low background noiseterm levels. As expected, very different profiles were seen in the same region for a different human cell type, H1 embryonic stem (H1 ES) cells (Fig. 2b ). To more quantitatively compare signal and noise levels in each method, we generated heatmaps around genomic sites called from H3K4me1 modification profiling for each method, where the same antibody had been used. After sampling each dataset to 8 million reads for comparison, we found that CUT&Tagterm for this histone modificationterm shows moderately higher signals compared to CUT&RUNterm throughout the list of sites (Fig. 2c ). Both methods have low backgrounds around the sites. In contrast, ChIP-seqterm signal has a very narrow dynamic rangeterm that is ~1/20 of the CUT&Tagterm signal range, and much weaker signals across the majority of sites. To quantitatively compare methods, we displayed the average read counts for CUT&Tagterm, CUT&RUNterm and ChIP-seqterm datasets for the H3K4me1 histone mark around the top 10,000 peaks defined by MACS2 on an H3K4me1 ChIP-seqterm dataset (Fig. 2g ). We found that CUT&Tagterm profiling gives substantially more signal accumulationterm at these sites, implying that CUT&Tagterm will be most effective at distinguishing chromatin features with fewest reads. Fig. 2 CUT&Tagterm for histone modificationterm profiling and RNAPII. a Representative chromatin landscapes across a 3 Mb segment of the human genome generated by the indicated method. For H3K27me3term, we downsampled ChIP-seqterm and CUT&RUNterm datasets to the same total mapped read counts as CUT&Tagterm for direct comparison. The high backgroundterm in downsampled ChIP-seqterm is from singleton reads distributed across the genome. b Same as a except for H1 ES cells. c Comparison of profiling methods for the H3K4me1 histone modificationterm in K562 cells. The same antibody was used in all experiments. Peaks were called and ordered for each dataset using MACS2. Each dataset was downsampled to the same read depthterm for comparison and plotted on their called peaks. Color intensities are scaled to the maximum read count at peaks in each dataset. d Detection of gene activity by RNAPII CUT&Tagterm. Gene promoters were ordered by associated RNA-seq counts (gray wedge) and read counts from RNAPII S2/5p CUT&Tagterm were plotted on these sites. e Active RNAPII is enriched at RNAPII CUT&Tagterm peaks. Peaks were called from RNAPII S2/5p CUT&Tagterm and ordered using MACS2 (gray wedge). PRO-seqterm reads were displayed onto these positions for (+) strand reads (yellow) and (–) strand reads (blue). f Comparison of ATAC-seqterm and H3K4me2 CUT&Tagterm profiling in K562 cells. Peaks were called on ATAC-seqterm data and heat maps were produced as in c . The top and bottom 2.5% of peaks were discarded to remove outliers. g Metaplot comparison of H3K4me1 histone modificationterm signal in CUT&RUNterm, CUT&Tagterm, and ChIP-seqterm in K562 cells, averaged at the top 10,000 peaks detected by MACS2 in ChIP-seqterm data. Profiling with the same antibody was compared at the downsampled read depthsterm of 8 million mapped reads for all three methods (blue, cyan, and green), and for 40 million mapped reads (orange) from ChIP-seqterm. h Metaplot comparison of ATAC-seqterm and H3K4me2 CUT&Tagterm profiling in K562 cells. 53,805 peaks were called on ATAC-seqterm data using MACS2, and read counts from each method were averaged across the intervals. The top and bottom 2.5% of peaks were discarded to remove outliers. Read counts at 17,000 randomly-chosen intervals for each dataset are displayed as dotted lines. Source data are available in the Source Data file The transcriptional status of genes and regulatory elements can be inferred from histone modificationterm patterns, but gene expression is directly read out by profiling chromatin-bound RNA polymerase IIterm (RNAPII). We used an antibody to the S2/S5-phosphorylation (S2/5p) forms of RNAPII, which distinguish engaged polymerase 20 . Landscapes show enrichment of RNAPII CUT&Tagterm reads at many genes (Fig. 2a , Supplementary Fig. 2a ), and a promoter heatmap reveals that this enrichment is predominantly at the 5′ ends of active genes 21 (Fig. 2d ). These results were confirmed by the observation of similar CUT&Tagterm patterns using antibodies to S2p, S5p and S7p forms of RNAPII (Supplementary Fig. 2a and Supplementary Fig. 3a, b ). To validate RNAPII CUT&Tagterm without relying upon annotations, which are typically based on mapping of processed transcripts, we chose transcriptional run-on data obtained with the base-pair-resolution PRO-seqterm technique, which provides direct mapping of RNAPII using a method that is unrelated to chromatin profiling 22 . PRO-seqterm maps the position of the 5′ end of engaged RNAPII as it is activated in situ, and is used to identify paused RNAPII just downstream of the transcriptional start site. Peaks were called from RNAPII S2/5p CUT&Tagterm and ordered using MACS2, and processed datasets from PRO-seqterm run-on for human K562 cells (SRA GSM1480327 ) were aligned to the peak calls. When ordered by RNAPII CUT&Tagterm MACS2 score, a close correspondence between PRO-seqterm occupancy and RNAPII-Ser2/5p CUT&Tagterm occupancy is seen (Fig. 2e ). Similar heat maps were obtained using antibodies to S2p, S5p, and S7p phosphorylation of the RNAPII C-terminal domain (Supplementary Fig. 3c ). CUT&Tagterm sensitively maps active sites in chromatin Replicates for profiling of H3K4me1 modification by CUT&Tagterm are highly similar, demonstrating the reproducibility of the method (Fig. 3a ). We obtained similar reproducibility when we compared H3K27me3term CUT&Tagterm replicates (Supplementary Fig. 2c ). In previous experiments with CUT&RUNterm profiling, we found that H3K4me2 histone modificationterm landscapes, which are associated with active promoters and enhancers, resemble ATAC-seqterm profiles 18 . We therefore performed CUT&Tagterm using an antibody to H3K4me2. An example of H3K4me2 CUT&Tagterm profiling to published ATAC-seqterm in K562 cells 23 is shown (Fig. 2a ). We found high occupancies for H3K4me2 at strong ATAC-seqterm peaks (Fig. 2f ), with much higher read counts (Fig. 2h ), implying that H3K4me2 profiling captures the most prominent accessible chromatin sites in the genome with greater sensitivity. Fig. 3 Reproducibility and efficiency of CUT&Tagterm. a Hierarchically clustered correlation matrix of CUT&Tagterm replicates (R1 and R2) and with CUT&RUNterm and ChIP-seqterm profiling for the H3K4me1 histone modificationterm. The same antibody was used in all experiments. Pearsonterm·stat correlations were calculated using the log 2 -transformed values of read counts split into 500 bp bins across the genome. b Efficiency of peak-calling between methods. Mitochondrial reads were removed from datasets from each method profiling the H3K4me2 histone modificationterm. The remaining read counts were downsampled to varying depths, and then used to call peaks using MACS2. The summed number of reads falling within called peaks in each dataset was plotted. Source data are available in the Source Data file To quantify the sensitivity of H3K4me2 CUT&Tagterm relative to H3K4me2 CUT&RUNterm 18 , H3K4me2 ChIP-seqterm 19 , and ATAC-seqterm 23 , we downsampled reads from each method, and used MACS2 with default parameters to call peaks on each dataset. We then estimated the fraction of reads falling within the called peaksterm. We found that both CUT&RUNterm and CUT&Tagterm populate peaks more deeply than ChIP-seqterm or ATAC-seqterm, demonstrating that they have exceptionally low signal-to-noiseterm (Fig. 3b ). In addition, CUT&Tagterm more rapidly populates peaks at low sequencing depthsterm, where ~2 million reads are equivalent to 8 million for CUT&RUNterm (or 20 million for ChIP-seqterm), demonstrating the exceptionally high efficiency of CUT&Tagterm. Of all the methods, only CUT&Tagterm reaches a fraction of 0.6 within peaks. Thus, with two histone modificationsterm (H3K4me2 and H3K27me3term), we segment the chromatin landscape into both active and silenced regionsterm, even with relatively low sequencing depthsterm. CUT&Tagterm simultaneously maps factor binding and accessible DNA To determine if we could use CUT&Tagterm for mapping transcription factorterm binding, we tested if pA-Tn5term tethered at transcription factorsterm can be distinguished from accessible DNA sitesterm in the genome. We used an antibody to the NPAT nuclear factor, a transcriptional coactivator of the replication-dependent histone genes, in CUT&Tagterm reactions. NPAT binds only ~80 accessible sitesterm in the histone clusters on chromosome 1 and chromosome 6 24 , thus we can compare true binding sitesterm with accessible sitesterm. In NPAT CUT&Tagterm profiles, ~99% of read counts accumulate at the promoters of the histone genes (Fig. 4a ). By scoring sites for correspondence to published ATAC-seqterm data 23 , we found that a smaller number of counts are distributed across accessible sitesterm in the K562 genome (Fig. 4b ). This probably results from some un-tethered pA-Tn5term binding to exposed DNA in situ, but it is straightforward to distinguish antibody-tethered sites from accessible sitesterm by the vast difference in read coverage (Fig. 4c ). Indeed, calling peaks by standard algorithms on NPAT CUT&Tagterm data generates a list of ~9000 sites that includes histone gene promoters and 10% of ATAC-defined accessible sitesterm (Supplementary Fig. 4 ). While this is only a fraction of the ~54,000 accessible sitesterm defined in K562 cells, adjusting the threshold and stringency of NPAT peak calling may improve detection. Fig. 4 CUT&Tagterm profiling of the NPAT chromatin factor and chromatin accessibility. a Ideograms of chromosome 1 and chromosome 6 with the clusters of replication-dependent histone genes (6 genes at 1q21.2 and 55 genes at 6p22.2) indicated. An NPAT CUT&Tagterm profiling track is displayed over each chromosome. The major NPAT peaks fall at the histone genes. b Distribution of read counts in CUT&Tagterm profiling. Called accessible sitesterm from ATAC-seqterm data were segregated into those at histone genes and other ATAC sites. Read counts from NPAT CUT&Tagterm were plotted for each category. c Boxplots of NPAT CUT&Tagterm signal at the promoters of replication-dependent histone genes ( His ), at other accessible sitesterm called from ATAC-seqterm data, and at a random selection of genomic background sites (bkgd). Source data are available in the Source Data file To test if CUT&Tagterm is tractable for profiling more abundant transcription factorterm binding sites, we profiled the CCCTC-binding factor (CTCF) DNA-binding protein. For these experiments, we varied the stringency of wash buffers to assess displacement of transcription factorsterm from chromatin. Under low-salt and medium-salt concentration conditions we observed read counts at CTCF sites detected by CUT&RUNterm and by ChIP-seqterm (Fig. 5a ), but with additional minor peaks (Supplementary Fig. 2a ). These additional peaks suggest that un-tethered pA-Tn5term contributes to coverage in these experiments. To determine if true CTCF binding sitesterm could be distinguished from accessible features by read depthterm, we compared the CUT&Tagterm read count at high-confidence CTCF sitesterm·def (defined by peak-calling on CUT&RUNterm data 18 ) to the CUT&Tagterm read count at accessible sitesterm (defined by peak-calling on ATAC-seqterm data 23 ). We found that these two distributions of read counts overlap, but that of accessible sitesterm is lower than that of CTCF sites (Fig. 5b ). Based solely on read depthterm, we discriminate ~5600 CTCF bound sites with a 1% false discovery rateterm. Comparing motif enrichment in these two classes demonstrates that the high signals correspond to CTCF motifs ( E -value = 2.1 × 10 −69 ), and the low signalsterm do not. Fig. 5 CUT&Tagterm profiling of the CTCF DNA-binding protein. a Comparison of methods for CTCF mapping. CTCF motifs in the genome were ranked by e -value, datasets from each method were downsampled to the same read depthterm, and then read counts were plotted on the fixed order of sites. b Distribution of read counts in CTCF CUT&Tagterm profiling. Sites were called from CUT&RUNterm profiling (blue) and at non-overlapping accessible sitesterm (red, called from ATAC-seqterm). Read counts from CTCF CUT&Tagterm were plotted for each category. c Resolution of CUT&Tagterm. Mean plots of fragment end positions from CTCF CUT&Tagterm centered over CTCF motifs in called peaks. Three different NaCl concentrations were used in wash buffers during and after pA-Tn5term tethering. Data are represented as a fraction of the maximum signal within the interval. d Resolved structure of a CTCF binding siteterm. The promoter of the SLC39A6 gene on chromosome 18 shows the chromatin features around a CTCF-bound site. Source data are available in the Source Data file We assessed the resolution of the CUT&Tagterm procedure by plotting the ends of reads centered on CTCF binding sitesterm. This shows that CUT&Tagterm protects a “footprint” spanning 80 bp directly over the CTCF motif (Fig. 5c ). While the segment protected from Tn5 integration is larger than the ~45 bp protected from MNase in CUT&RUNterm 9 , this indicates that the tethered transposase produces high resolution maps of factor binding sites. Similar footprints were obtained using different salt concentration washes, although 300–500 mM salt concentrations resulted in somewhat reduced signal-to-noiseterm (Fig. 5c ). The high resolution of CUT&Tagterm provides structural details of individual sites. For example, superposition of CTCF, H3K4me1, H3K4me2, H3K4me3, and ATAC mapping at a representative site reveals the relationship between accessible DNA, CTCF binding, and modified neighboring nucleosomes (Fig. 5d ). CUT&Tagterm profiles low cell numberterm samples and single cellsterm ChIP requires substantial cellular material, limiting its application for experimental and clinical samples. However, we and others have previously demonstrated that tethered profiling strategies like CUT&RUNterm have sufficient sensitivity that profiling small cell numbers routinely becomes feasible 9 , 25 . Signal improvements in CUT&Tagterm suggest that this method may work even more efficiently with limited samples. We first tested CUT&Tagterm for the H3K27me3term modification across a ~1500× range of material, from 100,000 down to 60 cells. We observed very similar high-quality chromatin profiles from all experiments (Supplementary Fig. 1b ), demonstrating that high data quality is still maintained with low input material. Analyzing sample and tracer DNA in these CUT&Tagterm series revealed that sequencing yield is proportional to the number of cells (Supplementary Fig. 1a ). CUT&Tagterm has the advantage that the entire reaction from antibody binding to adapter integration occurs within intact cells. The transposase and chromatin fragments remain bound together 15 , 26 , and thus fragmented DNA is retained within each nucleus. We developed a simple strategy to generate chromatin profiles of individual cells, which we term single-cell CUT&Tagterm (scCUT&Tag) (Fig. 6a ). We performed scCUT&Tag to the H3K27me3term modification on a bulk population of K562 cells, but with gentle centrifugation between steps instead of Concanavalin A magnetic beads. After integration, we used a Takara ICELL8 nano-dispensing system to aliquot single cellsterm into nanowells of a 5184 well chip, identifying the nanowells that contained one and only one cell by imaging the chip. We then performed PCR enrichmentterm of libraries in each passing nanowell using two indexed primers, and finally pooled all enriched libraries from the chip for Illumina deep sequencing to high redundancy to assess the sampling and coverage in each cell (Supplementary Fig. 6a ). Libraries from each well are distinguished by unique combinations of the two indices. Fig. 6 Chromatin profiling of individual cells. a Single cellterm CUT&Tagterm (scCUT&Tag) processing. All steps from antibody incubations through adapter tagmentation are done on a population of permeabilized unfixed cells. Individual cells are then dispensed into nanowells of a Takara ICELL8 chip. After verifying nanowells with single cellsterm by microscopy, combinations of two indexed barcoded primers are added to each well and fragment librariesterm are enriched by PCR. Libraries from the chip are pooled for multiplex sequencing. b A chromatin landscape across a 500 kb segment of the human genome is shown for H3K27me3term CUT&Tagterm on K562 cells. Tracks from bulk CUT&Tagterm, aggregated scCUT&Tag, and for 956 single cellsterm are shown. Single cellsterm are ordered by total read counts in each cell. c A chromatin landscape across a 500 kb segment of the human genome for H3K4me2 CUT&Tagterm on K562 cells. Tracks from bulk CUT&Tagterm, aggregated scCUT&Tag, and for 808 single cellsterm are shown. Single cellsterm are ordered by total read counts in each cell. d Fraction of reads in single K562 cells falling within called active peaks for the H3K4me2 histone modificationterm using stringent criteria. Narrow peaks were called using MACS2 on bulk profiling data, and reads from scCUT&Tag were assigned to those peaks. e Fraction of reads in single K562 and H1 cells falling within called silenced domains for the H3K27me3term histone modificationterm. Domains were called using SEACR on bulk profiling data for each cell type, and reads from scCUT&Tag were assigned to those domains. f Comparison of chromatin landscapes in H1 and K562 single cellsterm across a 2 Mb segment including the HoxB locus. Four hundred and seventy-nine single cellsterm of each type were ordered by total read counts. Source data are available in the Source Data file The aggregate of single cellterm chromatin profiling closely matched profiles generated in bulk samples (Fig. 6b ), with high correlations (Pearsonterm·stat’s r = 0.89, Supplementary Fig. 6b ). Individual cells were ranked by the genome-wide number of reads, and the unique fragments are displayed in tracks for each cell. Strikingly, the majority of reads from individual cells fall within H3K27me3term blocks defined in bulk profiling, indicating high recovery in single cellterm chromatin profiling (Fig. 6b ). A second replicate of H3K27me3term scCUT&Tag demonstrated the reproducibility of single cellterm profiling. Similarly, single cellterm profiling of the H3K4me2 modification recapitulates genomic landscapes of accessible and active chromatin (Fig. 6c ). A significant fraction of reads in single cellsterm fall within defined active and silenced chromatin features (Fig. 6d, e ). The breadth of chromatin features—from ~5 nucleosomes for H3K4me2 to hundreds in H3K27me3term domains—assists the detection of chromatin features even with sparse sampling from individual cells. To assess if chromatin features in individual cells could be used to distinguish cell types, we performed scCUT&Tag to the H3K27me3term modification in H1 cells. Again, we found that a high fraction of reads fell within domains defined by bulk profiling (Fig. 6e ), with high correlations between bulk and aggregated single cellterm data (Pearsonterm·stat’s r = 0.85, Supplementary Fig. 6b ). Comparing a 2 Mb region encompassing the HoxB domain reveals clear histone methylation in single cellterm tracks specifically in H1 cells, while this region is depleted in K562 cells (Fig. 6f ). These genome-wide patterns are sufficient to discriminate single H1 cells from K562 cells with high efficiency (Supplementary Fig. 6c, d ). The small fraction of K562 cells that are mis-called have the sparsest read coverage. Thus, chromatin profiling provides a method to discriminate single cellterm types. Discussion Chromatin profiling by CUT&Tagterm efficiently reveals regulatory information in genomes. In contrast to RNA-seq 27 , which only measures expressed genes, chromatin profiling has the unique advantage of identifying silenced regions, which is a key aspect of establishing cell fates in development. Although methods like ATAC-seqterm map accessible and factor-bound sites 17 , the specific chromatin proteinsterm bound at these sites must be inferred from motif or chromatin profiling data. While ChIP-based methods have been extensively used in model cell line systems, the vagaries of crosslinkingterm and fragmenting chromatin have limited chromatin profiling by ChIP-seqterm to an artisan technique where each experiment requires optimization. Likewise, a recently described alternative cross-linked chromatin profiling method, ChIL-seq 28 , requires many more steps than CUT&Tagterm and requires 3–4 days to perform all of the steps. In contrast, the CUT&Tagterm procedure, like CUT&RUNterm, is an unfixed in situ method, and is easily implemented in a standardized approach. This, combined with the cost-effectiveness of CUT&Tagterm, makes it appropriate for a high-throughput pipelineterm that can be implemented in a core facility 18 . It is conceivable that diverse users may provide their mixture of cells and antibody and receive processed deep sequencing files in just days. Since the first step in high-throughput CUT&Tagterm is antibody incubation at 4 °C, samples can be accumulated overnight in a facility and then loaded together onto a 96-well plate for robotic handling, as we previously demonstrated for AutoCUT&RUN 18 . With efficient use of reagents and better signal-to-noiseterm, CUT&Tagterm requires even fewer reads per sample than AutoCUT&RUN, which is already much cheaper than commercial exome sequencing. While the ease and low cost of this pipeline is appealing, the primary virtue of automated chromatin profiling is the minimization of batch and handling effects, and thus maximum reproducibility. Such aspects are critical for clinical assays and testing for chromatin-targeting drugs. We have shown that CUT&Tagterm provides high-quality single-cell profiles using the ICELL8 nano-dispensation system 12 , which allows for imaging prior to reagent addition and PCR. Likewise, CUT&Tagterm should be suitable for the 10× Genomics encapsulation systemterm 13 by adaptation of their recently announced ATAC-seqterm single-cell protocol 29 . Adaptability to high-throughput single-cell platformsterm is possible for CUT&Tagterm because adapters are added in bulkterm, whereas previous single-cell adaptations of antibody-based profiling methods, including ChIP-seqterm 30 , ChIL-seq 28 , and CUT&RUNterm 25 require reactions to be performed after cells are separated. The distinct distributions of low-level untargeted accessible DNA sitesterm and high-level CTCF-bound sites in CUT&Tagterm datasets suggests that by modeling the two expected underlying distributionsterm, true binding sitesterm can be distinguished from accessible DNA sitesterm without using other data. An advantage of this strategy is that the statistical distinction between true binding sitesterm and accessible features allows characterization of two chromatin features in the same experiment, where accessible DNA sitesterm can be annotated as well as binding sites for the targeted factor. This parsing out of the low-level ATAC-seqterm background from the strong targeted CUT&Tagterm signal makes possible de novo “multi-OMIC” CUT&Tagterm 31 . In the future, we expect that barcoding of adaptersterm 26 will allow for multiple epitopesterm to be simultaneously profiled in single cellsterm in large numbers, maximizing the utility of single-cell epigenomic profiling for studies of development and disease. Methods Biological materials Human K562 cells were purchased from ATCC (Manassas, VA, Catalog #CCL-243) and cultured following the supplier’s protocol. H1 ES cells were obtained from WiCell (Cat#WA01-lot#WB35186). We used the following antibodies: Guinea Pig anti-Rabbit IgG (Heavy & Light Chain) antibody (Antibodies-Online ABIN101961). H3K27me3term (Cell Signaling Technology, 9733, Lot 14), H3K27ac (Millipore, MABE647), H3K4me1 (Abcam, ab8895), H3K4me2 (Upstate 07–030, Lot 26335), H3K4me3 (Active Motif, 39159), PolSer2P, PolSer5P, PolSer2+5P, PolSer7P (Cell Signaling Technology, Rpb1 CTD Antibody Sampler Kit, 54020), CTCF (Millipore 07–729), NPAT (Thermo Fisher Scientific, PA5–66839 ALX-215-065-1), and Sox2 (Abcam, ab92494). Transposometerm preparation Using the pTXB1-Tn5 15 expression vector, sequences downstream of lac operator were replaced with an efficient ribosome binding site, three tandem FLAG epitope tags and two IgG binding domains of staphylococcal protein A, which were PCR amplified from the pK19pA-MN vector 11 . The C-terminus of Protein A was separated from the transposase by a 26 residue flexible linker peptide composed of DDDKEF(GGGGS) 4 . The pTXB1-Tn5 plasmid was a gift from Rickard Sandberg (Addgene plasmid # 60240) and the pK19pA-MN plasmid was a gift from Ulrich Laemmli (available through Addgene, plasmid # 86973). The 3XFlag-pA-Tn5-Fl plasmid (Addgene plasmid # 124601) was transformed into C3013 cells (NEB) following the manufacturer’s protocol. Each colony tested was inoculated into 3 mL LB medium and growth was continued at 37 °C for 4 h. That culture was used to start a 400 mL culture in 100 µg/mL carbenicillin containing LB medium (as it is more stable than ampicillin) and incubated on a shaker until it reached O.D. ~0.6, whereupon it was chilled on ice for 30 min. Fresh IPTG was added to 0.25 mM to induce expression, and the culture was incubated at 18 °C on a shaker overnight. The culture was collected by centrifugation at 10,000 rpm, 4 °C for 30 min. The bacterial pellet was frozen in a dry ice-ethanol bath and stored at −70 °C. Protein purification was performed as previously described 15 with minor modifications. Briefly, a frozen pellet was resuspended in 40 mL chilled HEGX Buffer (20 mM HEPES-KOH at pH 7.2, 0.8 M NaCl, 1 mM EDTA, 10% glycerol, 0.2% Triton X-100) including 1× Roche Complete EDTA-free protease inhibitor tablets. The lysate was sonicated 10 times for 45 s at a 50% duty cycle with output level 7 while keeping the sample chilled and holding on ice between cycles. The sonicated lysate was centrifuged at 10,000 rpm in a Fiberlite rotor at 4 °C for 30 min. A 2.5 mL aliquot of chitin slurry resin (NEB, S6651S) was packed into each of two disposable columns (Bio-rad 7321010). Columns were washed with 20 mL of HEGX Buffer. The soluble fraction was added to the chitin resin slowly, then incubated on a rotator at 4 °C overnight. The unbound soluble fraction was drained and the columns were rinsed with 20 mL HEGX and washed thoroughly with 20 mL HEGX containing Roche Complete EDTA-free protease inhibitor tablets. The chitin slurry was transferred to a 15 mL conical tube and resuspended in elution buffer (10 mL HEGX with 100 mM DTT). The tube was placed on rotator at 4 °C for ~48 h. The eluate was collected and dialyzed twice in 800 mL 2X Tn5 Dialysis Buffer (100 mM HEPES-KOH pH 7.2, 0.2 M NaCl, 0.2 mM EDTA, 2 mM DTT, 0.2% Triton X-100, 20% Glycerol). The dialyzed protein solution was concentrated using an Amicon Ultra-4 Centrifugal Filter Units 30 K (Millipore UFC803024), and sterile glycerol was added to make a final 50% glycerol stock of the purified protein. To generate the pA-Tn5term adapter transposometerm, 16 µL of a 100 µM equimolar mixture of preannealed Tn5MEDS-A and Tn5MEDS-B oligonucleotides 15 were mixed with 100 µL of 5.5 µM pA-Tn5term fusion protein. The mixture was incubated on a rotating platform for 1 h at room temperature and then stored at −20 °C. The complex is stable at room temperature, with no detectable loss of potency after 10 days on the benchtop (Supplementary Fig. 5A ), and without noticeable loss of data quality (Supplementary Fig. 5B ). Unexpectedly, this extended room temperature incubation resulted in a 1–2 order-of-magnitude increase in the number of tagmented E. coli fragments (Supplementary Fig. 5C ), which can be used as a calibration standard within a series using a constant amount of pA-Tn5term. This observation suggests that the E. coli DNA that co-purifies with pA-Tn5term is subject to tagmentation both during room temperature incubation and during tagmentation in situ, where the dramatic increase seen with pre-incubation results from subsequent trapping of tagmented pA-Tn5-bound DNA within the cell. In support of this interpretation, we note that E. coli carry-over DNA suitable for calibration is also released by pA-MNase in CUT&RUNterm reactions during digestion 32 . A likely explanation for the trapping of these different fusion protein-bound DNAs within cells is that the protein-protein interaction domains of Protein A that are specific for IgG bind non-specifically to cellular proteins, whereupon addition of divalent cation results in MNase digestion and release (pA-MNase) or tagmentation (pA-Tn5term). CUT&Tagterm for bench-top application Cells were harvested, counted and centrifuged for 3 min at 600× g at room temperature. Aliquots of cells (60–500,000 cells), were washed twice in 1.5 mL Wash Buffer (20 mM HEPES pH 7.5; 150 mM NaCl; 0.5 mM Spermidine; 1× Protease inhibitor cocktail) by gentle pipetting. Concanavalin A coated magnetic beads (Bangs Laboratories) were prepared as described 9 and 10 µL of activated beads were added per sample and incubated at RT for 15 min. We observed that binding cells to beads at this step increases binding efficiency. The unbound supernatant was removed and bead-bound cells were resuspended in 50–100 µL Dig-wash Buffer (20 mM HEPES pH 7.5; 150 mM NaCl; 0.5 mM Spermidine; 1× Protease inhibitor cocktail; 0.05% Digitonin) containing 2 mM EDTA and a 1:50 dilution of the appropriate primary antibody. Primary antibody incubation was performed on a rotating platform for 2 h at room temperature (RT) or overnight at 4 °C. The primary antibody was removed by placing the tube on the magnet stand to clear and pulling off all of the liquid. To increase the number of Protein A binding sites for each bound antibody, an appropriate secondary antibody (such as Guinea Pig anti-Rabbit IgG antibody for a rabbit primary antibody) was diluted 1:50 in 50–100 µL of Dig-Wash buffer and cells were incubated at RT for 30 min. Cells were washed using the magnet stand 2–3× for 5 min in 0.2–1 mL Dig-Wash buffer to remove unbound antibodies. A 1:200 dilution of pA-Tn5term adapter complex (~0.04 µM) was prepared in Dig-med Buffer (0.05% Digitonin, 20 mM HEPES, pH 7.5, 300 mM NaCl, 0.5 mM Spermidine, 1× Protease inhibitor cocktail). After removing the liquid on the magnet stand, 50–100 µL was added to the cells with gentle vortexing, which was incubated with pA-Tn5term at RT for 1 h. Cells were washed 2–3× for 5 min in 0.2–1 mL Dig-med Buffer to remove unbound pA-Tn5term protein. Next, cells were resuspended in 50–100 µL Tagmentation buffer (10 mM MgCl 2 in Dig-med Buffer) and incubated at 37 °C for 1 h. To stop tagmentation, 2.25 µL of 0.5 M EDTA, 2.75 µL of 10% SDS and 0.5 µL of 20 mg/mL Proteinase K was added to 50 µL of sample, which was incubated at 55 °C for 30 min or overnight at 37 °C, and then at 70 °C for 20 min to inactivate Proteinase K. To extract the DNA, 122 µL Ampure XP beads were added to each tube with vortexing, quickly spun and held 5 min. Tubes were placed on a magnet stand to clear, then the liquid was carefully withdrawn. Without disturbing the beads, beads were washed twice in 1 mL 80% ethanol. After allowing to dry ~5 min, 30–40 µL of 10 mM Tris pH 8 was added, the tubes were vortexed, quickly spun and allowed to sit for 5 min. Tubes were placed on a magnet stand and the liquid was withdrawn to a fresh tube. To amplify libraries, 21 µL DNA was mixed with 2 µL of a universal i5 and a uniquely barcoded i7 primer 33 , using a different barcode for each sample. A volume of 25 µL NEBNext HiFi 2× PCR Master mix was added and mixed. The sample was placed in a Thermocycler with a heated lid using the following cycling conditions: 72 °C for 5 min (gap filling); 98 °C for 30 s; 14 cycles of 98 °C for 10 s and 63 °C for 30 s; final extension at 72 °C for 1 min and hold at 8 °C. Post-PCR clean-up was performed by adding 1.1× volume of Ampure XP beads (Beckman Counter), and libraries were incubated with beads for 15 min at RT, washed twice gently in 80% ethanol, and eluted in 30 µL 10 mM Tris pH 8.0. A detailed, step-by-step protocol can be found at https://www.protocols.io/view/bench-top-cut-amp-tag-wnufdew/abstract High-throughput CUT&Tagterm For high-throughput 96-well microplate application, cells were first permeabilized and incubated with the primary antibodies before binding to beads. Two biological replicates of human K562 and H1 ES cells were washed twice with Wash Buffer, resuspended in Dig-wash buffer with 2 mM EDTA and arrayed in a 96-well plate. Permeabilization before antibody incubation varied from 1 to 5 h. Then, dilutions of appropriate antibodies were added (making final antibody concentrations 1:50) as duplicates. Cells were incubated with primary antibodies overnight. The next day, 10 µL of activated Concanavalin A coated magnetic beads were added to each sample, mixed gently and incubated at room temperature for 10 min. The plate was placed on a magnetic plate holder and supernatants were discarded. Appropriate secondary antibodies were prepared as 1:50 dilutions in Dig-wash and added to each well. Cells were washed three times with Dig-wash and then incubated with 1:200 dilution of pA-Tn5term adapter complex in Dig-med buffer at RT for 1 h. Cells were washed 3× for 5 min in Dig-med Buffer and resuspended in 50 µL Tagmentation buffer and incubated at 37 °C for 1 h. To stop tagmentation, 2.25 µL of 0.5 M EDTA, 2.75 µL of 10% SDS and 0.5 µL of 20 mg/mL Proteinase K was added to the sample, which was incubated at 55 °C for 30 min and then at 70 °C for 20 min to inactivate Proteinase K. Samples were held at 4 °C overnight until ready to continue. A 1.1× volume of AMPure XP beads was added to each well, vortexed and incubated at room temperature for 10–15 min. The plate was placed on a magnet and unbound liquid was removed. Beads were gently rinsed twice with 80% ethanol, and DNA was eluted with 35 µL of 10 mM Tris-HCl pH 8. 30 µL of eluted DNA was amplified by PCR as described above. DNA sequencing and data processing The size distribution of libraries was determined by Agilent 4200 TapeStation analysis, and libraries were mixed to achieve equal representation as desired aiming for a final concentration as recommended by the manufacturer. Paired-end Illumina sequencing was performed on the barcoded libraries following the manufacturer’s instructions. Paired-end reads were aligned using Bowtie2 version 2.2.5 with options: –local–very-sensitive-local–no-unal–no-mixed–no-discordant–phred33 -I 10 -X 700. Because of the very low background with CUT&Tagterm, typically 3 million paired-end reads suffice for nucleosome modifications, even for the human genome. For maximum economy, up to 96 barcoded samples per 2-lane flow cell can be pooled for 25 × 25 bp sequencing. For peak calling, parameters used were macs2 callpeak—t input_file –p 1e-5 –f BEDPE/BED(Paired End vs. Single End sequencing data) –keep-dup all –n out_name. Single-cell CUT&Tagterm Approximately 50,000 exponentially growing K562 cells were processed by centrifugation between buffer and reagent exchanges in low-retention tubes throughout. Centrifugations were performed at 600× g for 3 min in a swinging bucket rotor for the initial wash and incubation steps, and then at 300× g for 3 min after pA-Tn5term binding. Cells were collected and washed twice with 1 mL Wash Buffer (20 mM HEPES, pH 7.5; 150 mM NaCl; 0.5 mM Spermidine, 1× Protease inhibitor cocktail) at room temperature. Nuclei were isolated by permeabilizing cells in NP40-Digitonin Wash Buffer (0.01% NP40, 0.01% Digitonin in wash buffer) and resuspended in 1 mL of NP40-Digitonin Wash buffer with 2 mM EDTA. Antibody was added at a 1:50 dilution and incubated on a rotator at 4 °C overnight. Permeabilized cellsterm were then rinsed once with NP40-Digitonin Wash buffer and incubated with anti-Rabbit IgG antibody (1:50 dilution) in 1 mL of NP40-Digitonin Wash buffer on a rotator at room temperature for 30 min. Nuclei were then washed 3× for 5 min in 1 mL NP40-Digitonin Wash buffer to remove unbound antibodies. For pA-Tn5term binding, a 1:100 dilution of pA-Tn5term adapter complex was prepared in 1 mL NP40-Dig-med-buffer (0.01% NP40, 0.01% Digitonin, 20 mM HEPES, pH 7.5, 300 mM NaCl, 0.5 mM Spermidine, 1× Protease inhibitor cocktail), and permeabilized cellsterm were incubated with the pA-Tn5term adapter complex on a rotator at RT for 1 h. Cells were washed 3× for 5 min in 1 mL NP40-Dig-med-buffer to remove excess pA-Tn5term protein. Cells were resuspended in 150 µL Tagmentation buffer (10 mM MgCl 2 in NP40-Dig-med-buffer) and incubated at 37 °C for 1 h. Tagmentation was stopped by adding 50 µL of 4× Stop Buffer (40.4 mM EDTA and 2 mg/mL DAPI) and the sample was held on ice for 30 min. The SMARTer ICELL8 single-cell system (Takara Bio USA, Cat. #640000) was used to array single cellsterm as described for scATAC-seq 12 . DAPI-stained nuclei were visualized under the microscope and if there were clumps, they were strained through 10 micron cell strainers. Cells were counted using a hematocytometer and diluted at 28 cells/µL in 0.5× PBS and 1× Second Diluent (Takara Bio USA, Cat. # 640196). Cells were loaded to a source loading plate. Control wells containing 0.5× PBS (25 µL) and fiducial mix (25 µL) (Takara Bio USA, Cat. #640196) were also included in the source loading plate. Using the ICELL8 MultiSample NanoDispenser (MSND) FLA program, cells were dispensed into a SMARTer ICELL8 350 v chip (Takara Bio USA, Cat. # 640019) at 35 nanoliter per well. After cell dispense was complete, chips were sealed with the imaging film (Takara Bio USA, Cat. #640109) and centrifuged at 400× g for 5 min at room temperature and imaged using the ICELL8 imaging station (Takara Bio USA). Images were analyzed using automated microscopy image analysis software (CellSelect, Takara Bio USA). Since cells were stained only with DAPI, they were propidium iodide negative, so that permeabilized cellsterm would not be excluded by default software settings. Additional single cellsterm were manually selected for dispensing using a manual triaging procedure. Immediately following imaging, the filter file, which notes single-cell containing wells and control wells, was generated. We typically obtained ~1000 single cellsterm per chip. All of the following reagents were added to the selected set of wells which contained single cellsterm. To index the whole chip, 72 i5 and 72 i7 unique indices 33 were dispensed at 7.32 µM using ICELL8 MSND FLA program using the index 1 and index 2 filtered dispense tool respectively at 35 nanoliter per well. NEBNext High-Fidelity 2X PCR Master Mix (NEB, M0541L) was dispensed twice using the ICELL8 MSND Single Cellterm/TCR program for the filtered dispense tool at 50 nL per well. The chip was sealed and centrifuged at 2250× g at 4 °C for at least 5 min after each dispense. The chip was sealed with a TE Sealing film (Takara Bio USA, Cat. #640109) and on-chip PCR was performed using a SMARTer ICELL8 Thermal Cycler (Takara Bio USA) as follows: 5 min at 72 °C and 2 min at 98 °C followed by 15 cycles of 10 s at 98 °C, 30 s at 60 °C, and 5 s at 72 °C, with a final extension at 72 °C for 1 min. PCR products were collected by centrifugation at ~2250× g for 20 min using the supplied SMARTer ICELL8 Collection Kit (Takara Bio USA, Cat.#640048). Pooled libraries were purified using Ampure XP beads (Beckman Counter) in a 1:1.1 ratio. Briefly, libraries were incubated with beads for 15 min at RT, washed twice in 80% ethanol, and eluted in 10 mM Tris pH 8.0. Paired-end 25 × 8 × 8 × 25 bp Illumina sequencing was performed on the pooled barcoded libraries following the manufacturer’s instructions. Paired-end reads were aligned using Bowtie2 version 2.2.5 with options:–local–very-sensitive-local–no-unal–no-mixed–no-discordant–phred33 -I 10 -X 700. Reporting summary Further information on research design is available in the Nature Research Reporting Summary linked to this article. Supplementary information Supplementary Information Peer Review File Reporting Summary Source Data Acknowledgements We thank Tayler Hentges and Aaron Hernandez for technical support, and Matt Fitzgibbon and Michael Meers for helpful discussions on data analysis. We also thank all members of the HenikoffpersName lab for valuable discussions. This work was supported by the Howard Hughes Medical Institute, grants from the National Institutes of Health (R01 GM 108699 and 4DN TCPA A093) and the Fred Hutch Summer Undergraduate Research Program. Author contributions H.S.K. and S.H. designed and performed all experiments. C.A.C and E.S.P. and T.D.B. assisted with the experiments. H.S.K., S.J.W., J.G.H., K.A., and S.H. developed algorithms and analyzed the data. H.S.K, K.A., and S.H. wrote the manuscript. Data availability Publically available datasets analyzed in this work are available in Supplementary Note 1. All sequencing data generated in this study have been deposited in GEO under accession GSE124557 . The 3XFlag-pA-Tn5-Fl plasmid has been deposited with Addgene (#124601). Source data for the figures can be found in the Source Data file. All other data are available from the authors upon reasonable request. Competing interests The authors declare no competing interests. Footnotes Journal peer review information: Nature Communications thanks Sebastian Pott and Gabriella Ficz for their contribution to the peer review of this work. Peer reviewer reports are available. Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Kami AhmadpersName, Email: kahmad@fredhutch.org. StevenpersName HenikoffpersName, Email: steveh@fhcrc.org. Supplementary information Supplementary Information accompanies this paper at 10.1038/s41467-019-09982-5.