Chromatin, the complex of DNA and its bound proteins, is a very dynamic structure that constantly changes shape responding to the variety of cellular stimuli. Such rearrangements allow for the selective gene expression, robust DNA replication as well as for DNA repair. As DNA is wrapped around nucleosomes and as regulatory proteins, such as transcription factors (TF), exert their function on chromatin, the chromatin landscape constantly changes. Such changes can be either the primary cause or the consequence of a pathological condition. Thus, it is imperative to have appropriate methods to study such changes in a comprehensible, robust, quantitative and reproducible manner.

In this article, we describe the four major non-immunological assays for high-throughput chromatin studies: ATAC-seq, MNase-seq, DNase-seq, which are enzymatic methods, and FAIRE-seq, a biochemical method. Other methods have been used as well; for example, Liu J et al used deoxyribonuclease (DNase) I–treated terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay to assess the chromatin accessibility [2].

The assay for transposase-accessible chromatin using sequencing (ATAC-seq) method was developed by the Greenleaf lab [3] and has gained popularity [4, 5]. For example, B de Laval et al evaluated epigenetic memory in hematopoietic stem cells in response to immune challenges [6]. Calvanese V et al investigated the localization of MLLT3 at active transcription start sites in hematopoietic stem or progenitor cells using ATAC-seq [7]. Nott A et al identified open regions of chromatin in microglial, neuronal, oligodendrocyte and astrocyte nuclei through ATAC-seq [8]. Barisic D et al investigated the role of SNF2H in chromosomal structure with ATAC-seq [9]. Xia W et al performed a mini-version of ATAC-seq on human germinal vesicle oocytes to investigate histone modifications during human parental-to-zygotic transition [10]. Lee J et al investigate the effect of haploinsufficiency of lamin A/C, which interacts with heterochromatin-rich genomic regions at the nuclear envelope, on open chromatin and gene transcription [11]. ATAC-seq employs the enzyme Tn5 transposase in complex with short oligonucleotides of known sequence (adaptors). The transposase binds to and cuts accessible DNA and, at the same time, adds these oligos at the beginning/end of the chromatin fragments. ATAC-seq has shown almost identical results with DNase-seq in terms of DHSs while it offers additional important advantages. We will discuss these further below.

Briefly, the protocol includes the following steps [12] : nuclear isolation from freshly collected cells and digestion with the Tn5 transposase-adaptor complex, for example, Tn5 Tagment DNA Enzyme 1 [13]. This initiates chromatin digestion and, simultaneously, adds the adaptors at the ends of each fragment. This step is termed “tagmentation” as it “tags” the DNA fragments while it fragments the chromatin to create a library, which can be barcoded at will. After removal of proteins by column purification, a small number of PCR reactions follow in order to amplify the created library, with care so that saturation does not become an issue. The ideal result of the DNA fragment analysis is, first, a pool of 60-120 bp-long fragments representing the nucleosome-free regions (NFR), and secondly, a pool of fragments of about 150 bp increments, representing the nucleosome-containing chromatin. Updated protocols exist, for example, Omni-ATAC [14].

One of the greatest advantages of ATAC-seq is the simultaneous recovery of most chromatin structures, from NFR to mono- to di- to poly-nucleosomal regions. Whereas ATAC-seq NFR regions overlap almost completely with DHSs signal from DNase-seq, only the former allows the additional chromatin structures (nucleosome positioning and packing, TF occupancy etc) to be analysed simultaneously (Fig. 1). This is achieved by size separation of the ATAC signal bioinformatically. Secondly, ATAC-seq provides high-resolution analysis with only a few thousands, or even hundreds of cells; whereas low-input DNase-seq or scDNA-seq protocols have been developed, pre-existing DHS maps are required [15, 16]. Single cell analysis is also possible (scATAC-seq) with specialised equipment such as Chromium Single Cell ATAC Solution from 10x Genomics [17]. Nevertheless, the “tagmentation” step reduces dramatically the number of steps required minimising sample loss. Thus, the ability to use ATAC-seq with rare biological or clinically relevant samples is, possibly, the greatest gain (Table 1).

Indeed, ATAC-seq analysis in lymphocytes of chronic (CLL) patients have revealed the accessibility maps that differ between disease subtypes [18]. Others have done the same in human pancreatic islets revealing chromatin accessibility in type-2 diabetes [19], in primary tissues of both male and female mice [20], in mouse nephron progenitor cells [21] and many other tissues. Interestingly, ATAC-seq in alpha and beta pancreatic cells have identified novel, distinct gene signatures [22]. A comparison between normal and Ras-induced cancer tumour cells have revealed a more open chromatin landscape in the latter case, and it identified a region in a p53 intron where chromatin was gradually opening associated with tumour progression [23].

Lastly, ATAC-seq has been used in combination with other methods allowing simultaneous analysis of additional chromatin characteristics. For example, the methyl-ATAC-seq (mATAC-seq) method combines ATAC-seq with bisulfite sequencing reveals the methylation status of ATAC-generated fragments in a single protocol [24]. By applying ATAC-seq directly on bead-bound, chromatin immunoprecipited (ChIP) nucleosomes, the time and material loss in classical ChIP-seq approaches could be reduced [25].

The main drawback of ATAC-seq is similar to the other enzymatic methods (Fig. 1). For a nucleosomal region to be mapped, it is required to be in close proximity (1-2 kbp) to or surrounded by two accessible regions, one upstream and one downstream.

The purification of micrococcal nuclease (MNase) from Staphylococcus aureus by Cunningham et al. in the late 1950s, offered the first tool to elucidate chromatin structure. Indeed, Cunningham observed that, in the presence of calcium, MNase catalysed the hydrolysis of the 5’-phosphodiester bond of deoxyribonucleotides. Before the first electron microscopy images of chromatin revealing the nucleosomal structures (beads-on-a-string), the ability of MNase to cleave the linker DNA between nucleosomes offered the first clues that chromatin is consisted of a repeating array of subunits.

[enlarge]

Figure 1. Scheme of chromatin structures analyzed by each method shown by the corresponding signal obtained. Adapted from [1].

As expected, the condition in which the digestion takes place is very important. MNase has slight preference towards a consensus sequence (TATA(A/T)A(A/T)) [26], regions of low helix stability, single-stranded regions and unpaired bases. MNase initially cleaves one DNA strand but then achieves a second cut at the other strand in proximity to the first cut. As soon as the double-strand break occurs, MNase possesses a slower exonucleolytic activity that can degrade the newly generated termini [27]. Thus, temperatures lower than optimal may introduce bias toward these regions producing repeating subunits of 198 bp, instead of 146 bp. MNase progressively digest DNA until it meets an obstacle, e.g. a nucleosome. Thus, longer incubation times may over-digest chromatin and displace nucleosomes from chromatin resulting in fragments smaller than the nucleosomal DNA [28, 29]. Nevertheless, with standardized, optimised protocols none of the above has been shown to introduce significant bias to nucleosome mapping [26].

Initially, MNase digestion was used in low-throughput studies of chromatin structure. In the early 2000s, MNase has been used in combination with microarrays while a few years later with next generation sequencing (NGS), providing quantitative and qualitative genome wide information on nucleosome occupancy and positioning (reviewed in [1] ). Extensive MNase digestion of chromatin (either native or cross-linked) generates a pool of mononucleosomes (Table 1). Thus, MNase digestion followed by deep sequencing (MNase-seq) of mono-nucleosomal DNA has yielded nucleosomal maps in various cell types by precisely determining the positions of nucleosomes (Fig. 1) [30-32].

By altering the conditions, the Shore lab has demonstrated the presence of unstable, easily digestible nucleosomes that are displaced by MNase hence they are missing from classical MNase-seq analyses. These fragile nucleosomes are present at yeast promoters, which in turn were considered nucleosome-free [33].

The mononucleosomal DNA resulting from MNase digestion can be directly analysed by PCR methods, microarrays, or NGS. However, MNase digestion in combination with chromatin immunoprecipitation (ChIP) has been used to study the positions of nucleosomes bearing specific post-translational modifications (PTMs) or the association of non-nucleosomal proteins, such as transcription factors, with DNA. Briefly, upon chromatin digestion, PTM-bearing nucleosomes or TFs of interest are immunoprecipitated and the recovered DNA is the analysed further.

Overall, MNase-seq is an indirect method that studies chromatin structure in a plethora of organisms. As it digests the free (linker) but not the nucleosome-protected DNA, it reveals the position of nucleosomes and, possibly, of other chromatin-bound proteins. For example, Barisic D et al used MNase-seq to study the effect of SNF2H to nucleosome patterns [9].

Deoxyribonuclease I (DNase I) is a double-strand, non-specific endonuclease which binds to and cuts accessible, non-nucleosomal DNA [34]. It cleaves preferentially at the 5’ of pyrimidines [35] with some sequence bias [36]. Traditionally, the cleavage sites of DNase I have been termed DNase I hypersensitivity sites (DHSs) [37] and are characterised as hallmarks of chromatin regulation. DHSs comprise about 2% of the total genome and are dynamic as they are created by nucleosome or transcription factor (TF) turnover during chromatin rearrangements. DHSs include all cis-regulatory elements such as enhancers, promoters, insulators, silencers, locus control regions, transcription start sites (TSS) or gene bodies of active genes.

The advances in NGS have rendered DNase-seq the golden standard method for studying open chromatin and mapping of regulatory DNA elements replacing the low-throughput methods which relied on Southern blotting [38] (Fig. 1).

Two protocols, with minor differences have been established by the Crawford [39] and Stamatoyannopoulos [40] labs, with the second being extensively used in the ENCODE project [41]. One important parameter of both protocols is the time of digestion as DNase I cleaves first the most accessible sites, while with prolonged digestions, all chromatin will be eventually digested into ~300 bp fragments [42, 43]. Briefly, the protocol entails purification of intact nuclei so that the DNase I can enter the nucleus to reach the chromatin. The following, chromatin digestion, is the most crucial step as many factors can influence the outcome, such as enzyme purity, concentration, lot activity, and buffer composition. Thus, titration of DNase I and identification of the optimal enzyme-to-number of cells ratio is a crucial parameter. The digested chromatin is then subjected to protein degradation and size-selected DNA fragments are recovered with gel electrophoresis or column purification. Then, another crucial step is the library preparation when oligonucleotides of known sequence are attached to the 5’ and/or 3’ ends of the fragments. These oligos, termed adaptors, are used in subsequent PCR reactions that amplify the digested DNA fragments before being sequenced.

Several studies have identified dynamic cis-regulatory sites all over the genome upon different conditions in a cell type-specific manner. The feasibility of the protocols has been very crucial as 14% of all DHSs are cell type-specific [44]. Others, using quantitative comparison of DHSs and the fact that TF occupancy protects from DNase I cleavage, have predicted TF occupancy associated with functional and biological states [36, 45].

DNase-seq has been used with, but is not restricted to, almost all cell lines, formalin-fixed paraffin-embedded (FFPE) human tissue samples [16], various animal embryos and others. In a study using DNase-seq in mouse pre-implantation embryos [46], the authors compared the DHS in oocytes, sperm and fertilized embryos and they have revealed the epigenetic reprogramming steps that take place upon fertilization. The DHS are gradually increasing in number and in intensity upon embryo development while the differences between genetically imprinted genes could be easily identified. In a comparative study, only 1.28% of DHSs were found common between multipotent cells, normal differentiated primary cells, immortalised primary cells, and malignant cancer cell lines [15, 47]. Similarly, as differentiation progresses, DHSs overlap is decreasing. Besides, DHSs are disorganised in cancer. Interestingly, mutations in intergenic DHSs have been found associated with the expression of neighbouring genes in breast cancer [48] or thyroid carcinoma [16] suggesting that cancer driver mutations can occur in regulatory elements.

Formaldehyde-Assisted Isolation of Regulatory Elements (FAIRE) is an alternative method used to separate, collect and study the regulatory elements of chromatin. The chromatin environment of a plethora of eukaryotic cells and tissues has been studied using FAIRE. Briefly, the FAIRE protocol contains the following steps: short fixation with formaldehyde followed by mechanical or biochemical cell lysis and sonication. The lysate is then subjected to phenol/chloroform extraction. The method is based on the fact that when in phenol/chloroform solution, the NFR of the DNA are found in the aqueous phase while the protein-bound chromatin is aggregated between the two phases. The difference of formaldehyde crosslinking efficiency between the different forms of chromatin, i.e. tight crosslinking in the presence of proteins, seems to be the rationale behind FAIRE [49]. The NFR fraction is collected from the aqueous phase, precipitated with ethanol and, after crosslinking reversal, the DNA is prepared for subsequent analyses, such as qPCR, microarrays or deep sequencing (FAIRE-seq).

As with the other methods, FAIRE signal is peaks with heights analogous to the activity of promoter or gene expression, active chromatin. Conversely, FAIRE signal is negatively correlated with nucleosome dense regions. Similar to the previous methods, FAIRE-seq has been instrumental in creating open chromatin maps in numerous cell types, comparing chromatin environments in healthy vs disease states, TF binding etc [1].

FAIRE-seq is a fairly simple method that avoids any sequence bias that may be introduced by endonucleases (DNase I, Tn5 transposase, MNase). Similarly, sample preparation is straightforward thus reducing the number of steps that may lead to irreproducibility. Importantly, enzyme titrations are not required. However, all these do not mean that FAIRE-seq does not have its own limitations. For example, the fixation time needs to be adjusted for each cell type, even though a 5-min fixation time seems to be a good starting point. More importantly, the signal-to-noise ratio is much smaller compared to the other methods, and is the main reason why FAIRE-seq is less often utilized than the other methods. More frequently, FAIRE-seq is a secondary method that accompanies ATAC-seq or DNase-seq [1].