Abstract
Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Appendix 1: Native ChIP protocol Appendix 2: CUT&RUN protocol for yeast Appendix 3: CUT&RUN protocol for mammalian cells Data availability References Decision letter Author response Article and author information Metrics Abstract We describe Cleavage Under Targets and Release Using Nuclease (CUT&RUN), a chromatin profiling strategy in which antibody-targeted controlled cleavage by micrococcal nuclease releases specific protein-DNA complexes into the supernatant for paired-end DNA sequencing. Unlike Chromatin Immunoprecipitation (ChIP), which fragments and solubilizes total chromatin, CUT&RUN is performed in situ, allowing for both quantitative high-resolution chromatin mapping and probing of the local chromatin environment. When applied to yeast and human nuclei, CUT&RUN yielded precise transcription factor profiles while avoiding crosslinking and solubilization issues. CUT&RUN is simple to perform and is inherently robust, with extremely low backgrounds requiring only ~1/10th the sequencing depth as ChIP, making CUT&RUN especially cost-effective for transcription factor and chromatin profiling. When used in conjunction with native ChIP-seq and applied to human CTCF, CUT&RUN mapped directional long range contact sites at high resolution. We conclude that in situ mapping of protein-DNA interactions by CUT&RUN is an attractive alternative to ChIP-seq. https://doi.org/10.7554/eLife.21856.001 eLife digest The DNA in a person’s skin cell will contain the same genes as the DNA in their muscle or brain cells. However, these cells have different identities because different genes are active in skin, muscle and brain cells. Proteins called transcription factors dictate the patterns of gene activation in the different kinds of cells by binding to DNA and switching nearby genes on or off. Transcription factors interact with other proteins such as histones that help to package DNA into a structure known as chromatin. Together, transcription factors, histones and other chromatin-associated proteins determine whether or not nearby genes are active. Sometimes transcription factors and other chromatin-associated proteins bind to the wrong sites on DNA; this situation can lead to diseases in humans, such as cancer. This is one of the many reasons why researchers are interested in working out where specific DNA-binding proteins are located in different situations. A technique called chromatin immunoprecipitation (or ChIP for short) can be used to achieve this goal, yet despite being one of the most widely used techniques in molecular biology, ChIP is hampered by numerous problems. As such, many researchers are keen to find alternative approaches. Skene and Henikoff have now developed a new method, called CUT&RUN (which is short for “Cleavage Under Targets & Release Using Nuclease”) to map specific interactions between protein and DNA in a way that overcomes some of the problems with ChIP. In CUT&RUN, unlike in ChIP, the DNA in the starting cells does not need to be broken up first; this means that protein-DNA interactions are more likely to be maintained in their natural state. With CUT&RUN, as in ChIP, a specific antibody identifies the protein of interest. But in CUT&RUN, this antibody binds to the target protein in intact cells and cuts out the DNA that the protein is bound to, releasing the DNA fragment from the cell. This new strategy allows the DNA fragments to be sequenced and identified more efficiently than is currently possible with ChIP. Skene and Henikoff showed that their new method could more accurately identify where transcription factors bind to DNA from yeast and human cells. CUT&RUN also identified a specific histone that is rarely found in yeast chromatin and the technique can be used with a small number of starting cells. Given the advantages that CUT&RUN offers over ChIP, Skene and Henikoff anticipate that the method will be viewed as a cost-effective and versatile alternative to ChIP. In future, the method could be automated so that multiple analyses can be performed at once. https://doi.org/10.7554/eLife.21856.002 Introduction The action of transcription factors (TFs) at their binding sites on DNA drives gene expression patterns, and so genome-wide TF mapping has become a central goal both for individual researchers and large-scale infrastructural projects. TF profiling is most commonly carried out using chromatin immunoprecipitation (ChIP), a protocol that has changed little since it was first introduced over 30 years ago (Solomon and Varshavsky, 1985). Cells are crosslinked with formaldehyde, chromatin is fragmented and solubilized, an antibody is added, and the antibody-bound chromatin is recovered for DNA extraction. Successive advances in DNA mapping technologies have revolutionized the use of X-ChIP (formaldehyde crosslinking ChIP), and with ChIP-seq, base-pair resolution mapping of TFs became feasible (Rhee and Pugh, 2011; Skene and Henikoff, 2015; He et al., 2015). Improvements to ChIP-seq retain the crosslinking step to preserve the in vivo pattern while the entire genome is fragmented to create a soluble extract for immunoprecipitation. However, crosslinking can promote epitope masking and can generate false positive binding sites (Teytelman et al., 2013; Park et al., 2013; Jain et al., 2015; Baranello et al., 2016; Meyer and Liu, 2014). ChIP can also be performed without crosslinking, using ionic conditions that do not disrupt electrostatic contacts (Kasinathan et al., 2014). ‘Native’ ChIP provides a map of direct protein-DNA interactions with sensitivity and specificity trade-offs that compare favorably with X-ChIP methods. Native ChIP also minimizes problems with epitope masking and improves efficiency relative to X-ChIP, making it more amenable to low starting numbers of cells (O’Neill et al., 2006; Brind’Amour et al., 2015). But problems remain with incomplete extraction efficiency of protein-DNA complexes and the potential loss of binding. The uncertainties caused by systematic biases and artifacts in ChIP emphasize the need for methods based on different principles. An important class of non-ChIP mapping methods involves tethering of an enzyme to a DNA-binding protein by a chimeric fusion and action of the enzyme on DNA in the local vicinity. For example, in DamID (van Steensel et al., 2001) and related methods (Southall et al., 2013; Hass et al., 2015), Escherichia coli Dam methyltransferase is tethered to the TF and catalyzes N6-methylation of adenine at GATC sites in vivo. Sites can be mapped genome-wide using an N6-methyl-directed restriction enzyme. However, as the resolution of DamID is limited by the distribution of GATC sites, DamID cannot obtain the high resolution that is potentially attainable using a sequencing read-out (Aughey and Southall, 2016). An alternative enzyme tethering method, chromatin endogenous cleavage (ChEC) tethers the endo-exonuclease Micrococcal Nuclease (MNase) to the TF (Schmid et al., 2004). In ChEC, MNase is activated by permeabilizing cells and adding calcium for controlled cleavage. We recently applied an Illumina sequencing read-out to ChEC (ChEC-seq), achieving near base-pair resolution (Zentner et al., 2015). Enzyme tethering methods fundamentally differ from ChIP because they are carried out in vivo (DamID) or in situ (ChEC), with extraction of DNA directly from live or permeabilized cells, thus eliminating the need to solubilize and recover chromatin. Both DamID and ChEC require that a different chimeric fusion construct be produced for each TF to be mapped, limiting their transferability, for example to animal models, patient biopsies and post-translational modifications. In the original chromatin immunocleavage (ChIC) method, crude nuclei from crosslinked cells are first treated with a TF-specific antibody, followed by addition of a chimeric fusion between Protein A and MNase (pA-MN) and activation by calcium (Schmid et al., 2004). Protein A binds specifically to Immunoglobulin G, which obviates the need for a fusion protein. Here we report a major development of ChIC that retains the advantages of enzyme tethering methods, while extending its applicability and ease-of-use to a point that it replaces other existing methodologies. A key feature of our protocol is that in the absence of crosslinking, seconds after calcium-induced MNase cleavage on both sides of the TF, the TF-DNA complex is released into solution, allowing for recovery of pure TF-bound DNA fragments for sequencing simply by centrifugation and DNA extraction. By carrying out the procedure on magnetic beads, our ‘Cleavage Under Targets and Release Using Nuclease’ (CUT&RUN) technique is simpler than ChIP-seq while retaining the advantages of in situ methods. Targeted digestion by CUT&RUN greatly reduces background relative to complete genomic fragmentation for ChIP, requiring only ~1/10th the sequencing depth of standard ChIP methods. Simple spike-in controls allow accurate quantification of protein binding not possible by other methods. Furthermore, the method allows low starting cell numbers, and robotic automation is possible by performing the reaction on magnetic beads. Results Overview of the strategy Chromatin Immuno-Cleavage (ChIC) has the advantage of using TF-specific antibodies to tether MNase and cleave only at binding sites. To adapt ChIC for deep sequencing, it was necessary to reduce the representation of background breaks in DNA that otherwise dominate deep sequencing libraries. The CUT&RUN modifications of ChIC were motivated by the observation that light MNase treatment of nuclei liberates mononucleosomes and TF-DNA complexes, leaving behind oligonucleosomes (Sanders, 1978; Teves and Henikoff, 2012). We expected that targeted cleavage on both sides of a TF would release the TF-DNA complex into the supernatant, leaving the remainder of the genome in the pelleted nuclei. By performing brief digestion reactions on ice, we would recover TF-DNA complexes in the supernatant before TF-bound MNase diffuses around the genome and cleaves accessible chromatin. Based on this rationale, we developed a simple CUT&RUN protocol (Figure 1A). Unfixed nuclei are (1) immobilized on lectin-coated magnetic beads, (2) successively incubated with antibodies and protein A-MNase (pA-MN) followed by minimal washing steps, (3) mixed with Ca++ on ice to initiate the cleavage reaction then stopped seconds-to-minutes later by chelation, and (4) centrifuged to recover the supernatant containing the released TF-DNA complexes. DNA is then extracted from the supernatant and used directly for sequencing library preparation. Figure 1 with 4 supplements see all Download asset Open asset CUT&RUN produces limit digestion TF-DNA complexes. (A) Schematic diagram of the CUT&RUN strategy. Nuclei attached to magnetic beads can be treated successively with an antibody (or optionally with a primary and secondary antibody) and Protein A-MNase (pA-MN), which diffuse in through the nuclear pores. After Ca++ addition to activate MNase cleavage, fragments are released and diffuse out of the nucleus. DNA extracted from the supernatant is used to prepare libraries for paired-end sequencing. (B) CUT&RUN cleaves and releases chromatin particles into the supernatant. S. cerevisiae nuclei in which the endogenous H2A genes were replaced with H2A-3XFLAG were subjected to CUT&RUN and incubated at 0°C in Ca++ for the indicated times. DNA extracted from both the insoluble (ins) and soluble (sol) fractions was electrophoresed on a 1% agarose gel. The No 1o Ab control was digested for 10 min in parallel but without having added the primary mouse anti-FLAG antibody. (C) Size distributions of mapped paired-end reads from sequencing of indicated TF samples. An H2A size distribution is included for comparison. Data are normalized such that the sum of all points at each length step in base pairs equals 1. (D) Time-course profiles for Abf1 and Reb1 samples (~2–3 million mapped paired-end reads per track) showing ≤120 bp (blue) and ≥150 bp (brown) fragment length classes, compared to ORGANIC ChIP-seq (~20–30 million mapped paired-end reads) and standard ChIP-seq (Paul et al., 2015) (~5 million Abf1 and ~126 million Reb1 mapped single-end 50 bp reads). A negative control track shows the result of leaving out the primary antibody (No 1o Ab). Within each TF and fragment size group, the Y-axis scale is autoscaled by IGV showing normalized counts and the fragment size classes are superimposed. Ticks mark the location of significant Abf1 (upper) and Reb1 (lower) motifs. This region was chosen as having the largest cluster of Abf1 motifs on Chromosome 3. https://doi.org/10.7554/eLife.21856.003 CUT&RUN produces limit digestion of chromatin complexes We first performed CUT&RUN using crude yeast nuclei. To rigorously compare CUT&RUN and ChIP-seq, we used the same FLAG-tagged TF strains, the same nuclear preparation protocol, the same mouse anti-FLAG monoclonal antibody and the same procedure for Illumina library preparation and paired-end sequencing (Kasinathan et al., 2014). As mouse Protein A binds only weakly to mouse IgG, we used a rabbit anti-mouse secondary antibody for CUT&RUN. To test the efficiency of CUT&RUN, we used a Saccharomyces cerevisiae strain expressing 3XFLAG-tagged histone H2A, which would be expected to release nucleosomal fragments genome-wide. Indeed, over a 100-fold digestion time course at 0°C, we observed gradual cleavage and release of fragments down to mononucleosomal size, completely dependent on the presence of the primary antibody (Figure 1B). We next applied CUT&RUN to two structurally distinct S. cerevisiae TFs, ARS binding factor 1 (Abf1), and rDNA enhancer binding protein 1 (Reb1), obtaining ~2–3 million mapped paired-end reads per sample. We found that the size distributions of mapped fragments were virtually superimposable below ~150 bp for time points between 4 s and 128 s (Figure 1C). This close concordance between time points over a 32-fold range suggests that limit digestion of TF-bound fragments occurs rapidly upon addition of Ca++, and demonstrates that digestion time is not a critical parameter. Mapped TF fragment sizes peaked at ~100 bp, in contrast to H2A fragments, which peaked at ~150 bp. As we expect TF complexes to be smaller than ~100 bp, and nucleosomes are ~150 bp, we mapped ≤120 bp and ≥150 bp fragments separately. Time-point profiles show crisp CUT&RUN peaks within the ≤120 bp size class for each TF motif in each region (Figure 1D and Figure 1—figure supplement 1). Except for a slow monotonic increase in peak occupancy when normalized to the spike-in control (Figure 1—figure supplement 2), no consistent differences between time points were observed within the 1 s to 128 s interval, confirming that gradual release of TF-DNA complexes yields limit digestion reactions. Total DNA extraction and purification of small fragments produced nearly identical results (Figure 1—figure supplement 3), which demonstrates that extraction of DNA from the supernatant quantitatively recovers TF-bound fragments. CUT&RUN robustly maps yeast TF binding sites in situ at high resolution To verify that the ≤120 bp fragments represent cleavages around TF binding sites, we first identified all significant Abf1 and Reb1 motifs in the genome and found that motifs based on CUT&RUN data and motifs based on ORGANIC data were nearly identical (Figure 1—figure supplement 4A–D). We use ORGANIC-derived motifs to scan the yeast genome, which provided us with a comprehensive list of 1899 Abf1 and 1413 Reb1 motifs determined completely independently of CUT&RUN. We confirmed that the majority of peak calls overlapped the motif for each dataset, with somewhat better performance for CUT&RUN than ORGANIC for Abf1 and vice versa for Reb1 (Figure 1—figure supplement 4E). We then aligned the ≤120 bp and ≥150 bp profiles centered over these motifs and constructed heat maps. When rank-ordered by occupancy over the 2 kb interval centered over each Abf1 and Reb1 motif, we observed that >90% of the TF sites were occupied by fragments over the corresponding motif relative to flanking regions (Figure 2 and Figure 2—figure supplement 1, upper panels), representing likely true positives. CUT&RUN occupancies over Abf1 and Reb1 motifs showed high dynamic range relative to nuclease accessibility (Figure 2—figure supplement 1, lower panels), seen in heat maps as higher contrast above background for CUT&RUN. In contrast, Abf1 fragments showed negligible occupancy at non-overlapping Reb1 sites and vice-versa for Reb1 fragments at non-overlapping Abf1 sites (Figure 2 and Figure 2—figure supplement 1, middle panels). The almost complete correspondence between the presence of a TF motif and occupancy of the TF, and the general absence at sites of a different TF, imply that CUT&RUN is both highly sensitive and specific for TF binding. Figure 2 with 1 supplement see all Download asset Open asset CUT&RUN accuracy and robustness compares favorably with ChIP-seq. Abf1 (A) and Reb1 (B) heat maps of CUT&RUN datasets from a single experiment (20160630), pooling 1" to 32" time-course samples, and separating into ≤120 bp and ≥150 bp size classes (left). Also shown is the ORGANIC ChIP-seq ≤120 bp size class (middle) and standard ChIP-seq datasets (right). Abf1 has two DNA-binding domains spaced ~10 bp apart (Cho et al., 1995), whereas Reb1 has a single Myb-like DNA-binding domain (Morrow et al., 1990). Solubilization of Abf1 chromatin after MNase digestion required 600 mM NaCl to obtain the best trade-off between specificity and sensitivity, whereas for Reb1, 80 mM gave the best results (Kasinathan et al., 2014), and these are the datasets used for comparison. As in our previous comparison of ORGANIC to ChIP-exo and ChIP-chip (Kasinathan et al., 2014), we consider the set of all statistically significant Abf1 and Reb1 motifs as the ‘gold standard’ for judging sensitivity (occupancy of sites by the correct TF) and specificity (exclusion from sites of an incorrect TF). Aligned profiling data were centered and oriented over the motif for the same TF (top) and for the other TF (bottom) for display (removing 81 sites where Abf1 and Reb1 sites were within 50 bp of one another) and were ordered by average pixel density over the −1 kb to +1 kb span of the ≤120 bp datasets using Java Treeview with log2 scaling and contrast = 5. Ordering was performed independently for CUT&RUN (based on ≤120 bp fragments) and ChIP-seq, in which case the approximate fraction of sites occupied relative to flanking regions becomes evident, and comparison of the top panel (correct TF) to the bottom panel (incorrect TF) reflects the sensitivity/specificity tradeoff for a dataset. Sites were determined by MAST of the S. cerevisiae genome using the based on ChIP-seq data (Figure 1—figure supplement but results were using MAST with based on CUT&RUN data To directly compare CUT&RUN to high-resolution ChIP-seq, we up ChIP-seq data over Abf1 and Reb1 motifs. As (Kasinathan et al., 2014), ORGANIC ChIP-seq the majority of Abf1 true positive motifs and nearly all Reb1 motifs the genome (Figure upper middle panels). The best Reb1 data were with 80 mM NaCl and the best Abf1 data were with 600 mM the dynamic range for Reb1 was better than that for Abf1 with false positive occupancy (Figure lower middle panels). In contrast, CUT&RUN showed the same dynamic range for both TFs over the same range of digestion time points with ~10 paired-end that CUT&RUN is more than ORGANIC ChIP-seq. to these high-resolution methods (Kasinathan et al., 2014), standard ChIP-seq using crosslinking and showed sensitivity and specificity (Figure panels). CUT&RUN provides TF occupancy maps with sensitivity/specificity trade-offs relative to ChIP-seq. To the resolution of CUT&RUN, we the of each TF as the average density of fragment around the motif For both Abf1 and Reb1, we observed bp that these transcription factors bp centered over the motif with near base-pair resolution (Figure and in the cleavage maps show a pattern on of both Abf1 and Reb1 with between ~10 bp apart over bp, and confirmed by to be of base (Figure 10 bp cleavage the 10 of which suggests that the DNA on of these bound TFs is oriented such that tethered MNase has to one of the DNA of MNase to a TF it to cleave nearby DNA on the of a of the chromatin (Figure the that we at 0°C are to immobilized MNase for cleavage near the tethering Figure Download asset Open asset CUT&RUN maps TF binding sites at high resolution. (A) of fragment a deep and for Abf1 and Reb1 CUT&RUN datasets at their oriented and aligned motifs all normalized base-pair counts from datasets (Figure patterns with an ~10 bp on the and are confirmed by (B) of the between the 1 bp resolution shown in (A) and the same with an bp which also shows that is no corresponding in average (C) as but to with an bp and at The that the around Reb1 show at and likely reflects the presence of shown below Y-axis based on the ≥150 bp size class from ORGANIC data (Kasinathan et al., 2014). CUT&RUN maps chromatin-associated complexes mapping of of the chromatin can be for methods. For example, the complex a that it et al., et al., 2015), and the DNA protein from sites of binding (Zentner and Henikoff, 2013; et al., X-ChIP complexes to their native ChIP yeast chromatin at their sites of both in regions and on nucleosomes (Zentner et al., For CUT&RUN to such chromatin-associated complexes we found it necessary to extract total DNA than chromatin by CUT&RUN in situ, which be to diffuse through nuclear pores. we extracted all DNA and DNA fragments with beads. When this protocol was applied to over a digestion we observed chromatin profiles that were to using ORGANIC but with only the number of paired-end reads (Figure As peaks on the of binding sites are seen for both CUT&RUN and ORGANIC confirming that from the in vivo (Zentner and Henikoff, as it does in et al., map and average analyses show that the ≤120 bp fragments track with sites, whereas the ≥150 bp fragments are in the local representing (Figure supplement 1). Figure 4 with 2 supplements see all Download asset Open asset CUT&RUN maps chromatin complexes. (A) showing a CUT&RUN time-course experiment million paired-end reads per a no primary antibody (No 1o negative aligned with ORGANIC data for two MNase digestion time points and average million reads per (Zentner and Henikoff, sites shown as that peaks are of peak (B) profiles for CUT&RUN digestion over a spike-in showing (C) ORGANIC profiles million reads) show concordance with the CUT&RUN s million reads). that the same CUT&RUN s ≤120 bp is shown in both (B) and but at different We also applied CUT&RUN to the of the to nucleosomes at and so we aligned yeast genes at the of the +1 of the et al., 2015). We observed digestion over a s to 30 min time course (Figure supplement and of the of directly over the (Figure supplement et al., peaks are most in the where CUT&RUN profiles show a gradual increase in with digestion between s and 10 min (Figure that quantitative limit are using our we observed a nearly for our negative control from nuclei treated in parallel for the digestion but where the primary anti-FLAG antibody was results for CUT&RUN are to results for ORGANIC profiling et al., 2015), but with higher (Figure We conclude that CUT&RUN provides high-resolution mapping of chromatin-associated complexes, that are and CUT&RUN insoluble DNA-binding protein complexes Abf1 and Reb1 are TFs, but many DNA-binding proteins of are and so can be to by ChIP. In is only only one per which is only the of Abf1 or An to the which the histone in of is that it is of the complex the cell et al., which it highly insoluble et al., 2012). To the by CUT&RUN, we the samples after the supernatant from one and total DNA from the In this way we could compare the recovery of the soluble and the insoluble In we histone By the between total and soluble chromatin, we can the occupancy of each histone in the insoluble As expected for the insoluble the occupancy on the is seen at the (Figure occupancy of insoluble H2A, which is in the genome, is also at the Indeed, at all yeast we of and H2A to the bp over the digestion with resolution that is better than that of standard X-ChIP (Figure We also extracted total DNA from the chromatin from cells that crosslinked to CUT&RUN, with results (Figure crosslinking results in a more distinct and the of nucleosomes on which we as a in chromatin with crosslinking, while that the strategy can be applied to crosslinked cells. Figure with 1 supplement see all Download asset Open asset CUT&RUN maps the highly insoluble S. cerevisiae (A) After digestion for the indicated samples were in and both the soluble fraction and total DNA were fragments were from total DNA with beads before library preparation. counts are shown for S. cerevisiae 1, where and H2A are on the same Y-axis over was also seen genome-wide. (B) as (A) but in over the kb interval at the (C) of insoluble and H2A, where we = = for the of all S. cerevisiae aligned around their A et al., is shown on the same scale for comparison (left). over two (D)