The Current State of Chromatin Immunoprecipitation
The biological significance of interactions of nuclear proteins with DNA in the context of gene expression, cell differentiation, or disease has immensely been enhanced by the advent of chromatin immunoprecipitation (ChIP). ChIP is a technique whereby a protein of interest is selectively immunoprecipitated from a chromatin preparation to determine the DNA sequences associated with it. ChIP has been widely used to map the localization of post-translationally modified histones, histone variants, transcription factors, or chromatin modifying enzymes on the genome or on a given locus. In spite of its power, ChIP has for a long time remained a cumbersome procedure requiring large numbers of cells. These limitations have sparked the development of modifications to shorten the procedure, simplify sample handling and make ChIP amenable to small numbers of cells. In addition, the combination of ChIP with DNA microarray and high-throughput sequencing technologies has in recent years enabled the profiling of histone modification, histone variants, and transcription factor occupancy on a genome-wide scale. This review highlights the variations on the theme of the ChIP assay, the various detection methods applied downstream of ChIP, and examples of their application.
KeywordsChromatin immunoprecipitation (ChIP)MeDIPHistoneAcetylationMethylationDNA bindingEpigenetics
Reversible Modifications of DNA and Histones
Interaction between proteins and DNA is essential for many cellular functions such as DNA replication and DNA repair, maintenance of genomic stability, chromosome segregation at mitosis, and regulation of gene expression. Transcription is controlled by the dynamic association of transcription factors and chromatin modifiers with target DNA sequences. These associations take place not only within regulatory regions of genes (promoters and enhancers) but also within coding sequences. They are modulated by modifications of DNA, such as the methylation of CpG dinucleotides , by post-translational modifications of histones , and by incorporation of histone variants [3–7]. These alterations are commonly referred to as epigenetic modifications: they modify the composition of DNA and chromatin without altering genome sequence, and they are passed onto daughter cells (they are heritable).
DNA methylation is generally seen as a hallmark of long-term gene silencing [8, 9]. Methyl groups on the cytosine in CpG dinucleotides create target sites for methyl-binding proteins which induce transcriptional repression by recruiting transcriptional repressors such as histone deacetylases or histone methyltransferases . DNA methylation largely contributes to gene repression and as such it is essential for development [10–12], X chromosome inactivation , and genomic imprinting [14, 15]. The relationship between DNA methylation and gene expression is intricate and recent evidence based on genome-wide CpG methylation profiling has highlighted CpG content and density of promoters as one component of this complexity [16, 17].
In addition to DNA methylation, post-translational modifications of histone proteins regulate gene expression. The core element of chromatin is the nucleosome, which consists of DNA wrapped around two subunits of histone H2A, H2B, H3, and H4. Nucleosomes are spaced by the linker histone H1. The amino-terminal tails of histones are post-translationally modified to confer physical properties that affect their interactions with DNA. Histone modifications not only influence chromatin packaging but are also read by adaptor molecules, chromatin modifying enzymes, transcription factors, and transcriptional repressors, and thereby contribute to the regulation of transcription [2, 18–20].
Analysis of DNA-Bound Proteins by Chromatin Immunoprecipitation
Chromatin immunoprecipitation (ChIP) has become the technique of choice to investigate protein–DNA interactions inside the cell [36, 37]. ChIP has been used for mapping the localization of post-translationally modified histones and histone variants in the genome, and for mapping DNA target sites for transcription factors and other chromosome-associated proteins.
Development of techniques leading to the ChIP assay as we know it since the mid 1990s has occurred over many years . The use of formaldehyde to cross-link proteins with proteins or proteins with DNA, however, was first reported in the 1960s and its application to study histone–DNA interactions within the nucleosome goes back to the mid-late 1970s. The development of anti-histone antibodies 20 years ago, to investigate the association of histones with DNA in relation to transcription, led the path to the ChIP assay . Pioneering studies showed that during heat shock, histone H4 remained associated with the HSP70 gene . Subsequent improvements in the procedure enabled the demonstration that the interaction of histone H1 with DNA was altered during changes in transcriptional activity in Tetrahymena . The availability of antibodies to post-translationally modified histones, in combination with ChIP, has been instrumental in the understanding of transcription regulation in the early 1990s. For instance, antibodies to acetylated histones have been used to show that, using the β-globin locus as a target genomic sequence, core histone acetylation is associated with chromatin that is active or poised for transcription [49–52]. The ChIP assays has since been extended to non-histone proteins, including less abundant protein complexes, and to a wide range of organisms such as protozoa, yeast, sea urchin, flies, fish and avian and mammalian cells .
For well over a decade, ChIP has remained a cumbersome protocol, requiring 3–4 days and large numbers of cells—in the multi-million range per immunoprecipitation. These limitations have limited the application of ChIP to large cell samples. Classical ChIP assays also involve extensive sample handling [37, 53], which is a source of loss of material, creates opportunities for technical errors and enhances inconsistency between replicates. To remedy to these limitations, modifications have been made to make ChIP protocols shorter, simpler and allow analysis of small cell samples [39, 54–57].
This introductory review addresses modifications of conventional ChIP assays which have recently been introduced to simplify and accelerate the procedure and enable the analysis of DNA-bound proteins in small cell samples. Analytical tools that can be combined with ChIP to address the landscape of DNA–protein interactions are also presented.
ChIP Assays Designed for Small Cell Numbers
A major drawback of ChIP has for a long time been the requirement for large cell numbers. This has been necessary to compensate for the loss of cells upon recovery after cross-linking, for the overall inefficiency of ChIP, and for the relative insensitivity of detection of ChIP-enriched DNA. The need for elevated cell numbers has hampered the application of ChIP to rare cell samples, such as cells from small tissue biopsies, rare stem cell populations, or cells from embryos. Several recent publications have addressed this issue and report alterations of conventional ChIP protocols to make the technique applicable to smaller numbers of cells.
Carrier ChIP: CChIP
The rationale behind carrier ChIP, or CChIP, is that the immunoprecipitation of a small amount of chromatin prepared from few mammalian cells (100–1,000) is facilitated by the addition of carrier chromatin from Drosophila—or any species sufficiently evolutionarily distant from the species investigated . CChIP involves the mixing of cultured Drosophila cells with a small number of mammalian cells. Native chromatin fragments are prepared from purified nuclei by partial MNase digestion and immunoprecipitated using antibodies to modified histones. To compensate for the small amount of target DNA precipitated, the ChIP DNA is detected by radioactive PCR and phosphorimaging. Specificity of amplification is monitored for each ChIP by determination of the size of the DNA fragment produced .
CChIP has proven to be suitable for the analysis of 100-cell samples. A limitation, however, is that analysis of multiple histone modifications require multiple aliquots of 100 cells which may or may not be identical. Furthermore, in its published form, CChIP is based on the NChIP procedure  and as such is not suited for precipitation of transcription factors. Nonetheless, there is no reason to believe that CChIP is not compatible with cross-linking, and thereby becomes more versatile. Despite these limitations, however, the benefit of CChIP for analyzing small cell samples is already clear.
Using CChIP, O’Neill et al.  have reported an analysis of active and repressive histone modifications on a handful of target loci in mouse inner cell mass and trophectoderm cells, the two cell types of the blastocyst. Application of CChIP to embryonic transcription factors in embryos and embryonic stem (ES) cells to unravel common and distinct target genes should enhance our understanding of the molecular basis of pluripotency.
A Quick and Quantitative ChIP assay: Q2ChIP
As alternative to CChIP, we have developed a quick and quantitative (Q2)ChIP protocol suitable for up to 1,000 histone ChIPs or up to 100 transcription factor ChIPs from as few as 100,000 cells . Q2ChIP involves a chromatin preparation from a larger number of cells than CChIP, but includes chromatin dilution and aliquoting steps which enable storage of many identical chromatin samples from a single preparation. Because Q2ChIP involves a cross-linking step, chromatin samples are also suitable for immunoprecipitation of transcription factors or other non-histone DNA-bound proteins. DNA protein cross-linking in suspension in the presence of a histone deacetylase inhibitor, elimination of essentially all non-specific background chromatin through a tube-shift after washes of the ChIP material, and combination of cross-linking reversal, protein digestion and DNA elution into a single 2-h step, shorten the procedure and enhance ChIP efficiency . Suitability of Q2ChIP to small amounts of chromatin has been attributed to the reduction of the number of steps in the procedure, increased ratio of antibody-to-target epitope, resulting in an enhanced signal-to-noise ratio. Q2ChIP has been validated against the conventional ChIP assay from which it was derived . It has been used to illustrate changes in histone H3K4, K9 and K27 acetylation and methylation associated with differentiation of embryonal carcinoma cells on developmentally regulated promoters .
With the aim of further reducing the number of cells used, we subsequently devised a micro (μ)ChIP protocol suitable for up to nine parallel ChIPs of modified histones and/or RNA polymerase II (RNAPII) from a single batch of 1,000 cells without carrier chromatin [57, 58]. The assay can also be downscaled for monitoring the association of one protein with multiple genomic sites in as few as 100 cells, and has been adapted for small (~1 mm3) tissue biopsies. Modifications of μChIP for analysis of tumor biopsies have been reported recently . The assay was validated by assessing several post-translational modifications of histone H3 and binding of RNAPII in embryonal carcinoma cells and in human osteosarcoma biopsies, on developmentally regulated and tissue-specific genes .
In μChIP, chromatin is prepared from 1,000 cells and divided into nine aliquots (‘100-cell ChIP’), of which eight can be dedicated to parallel ChIPs, including a negative control, and one serves as an input reference sample. When starting from 100 cells, only one ChIP is possible using the current protocol. Regardless of the starting cell number, the 100-cell ChIP enables the analysis of 3–4 genomic sites by duplicate qPCR without amplification of the ChIP DNA . We have since successfully amplified μChIP DNA using whole-genome DNA amplification kits and have been able to apply μChIP to microarrays .
At the time our μChIP assay was being evaluated , a miniaturized ChIP protocol for 10,000 cells also coincidentally called microChIP, was published . From batches of 10,000 cells, the assay allows analysis of histone or RNAPII binding throughout the human genome using a ChIP-on-chip approach with high-density oligonucleotide arrays. This microChIP assay takes approximately 4 days, but it presents the main advantage of being applicable to genome-wide studies rather than being restricted to a few genomic regions . Of note, a ChIP assay followed by high-throughput sequencing (ChIP-seq) has recently been reported for cell numbers significantly smaller than what has commonly been used in ChIP-seq . This assay is presented below.
Fast ChIP Protocols
Using the ChIP material directly as template in the PCR (on-bead PCR) has also been reported in yeast, with results comparable to PCR using purified DNA . The possibility of performing the PCR reaction directly on the immunoprecipitated material indicates that the formaldehyde cross-linking reversion step may be omitted, likely because the initial PCR heating step suffices to partially reverse the cross-link. Direct PCR, therefore, holds promises for speeding up the analysis of ChIP products.
Whether end-point or quantitative on-bead PCR can be performed seems, however, to depend on the nature of the carrier beads used in ChIP. Direct on-bead PCR is successful with magnetic Protein G beads  and with agarose-conjugated Protein A beads (J.A. Dahl and P. Collas, unpublished data). Furthermore, we have shown that ChIP products precipitated by agarose beads can be directly analyzed by qPCR using SYBR® Green (J.A. Dahl and P. Collas, unpublished data). This is in contrast to magnetic beads which due to their opacity, interfere with quantification of the SYBR® Green signal during the real-time PCR (Fig. 3). These observations argue, then, that while direct qPCR is possible with ChIP templates bound to agarose, and most likely Sepharose, beads, magnetic beads are currently incompatible with qPCR.
An alternative to Chelex-100 and on-bead PCR has recently been reported in the context of a higher throughput ChIP assay than those reported to date . To enable rapid access of the ChIP DNA for PCR with minimal sample handling, the authors have replaced Chelex-100 with a high-pH Tris buffer containing EDTA. PCR-ready DNA recovery is identical to that of Chelex-100, with the advantage that it can be performed in a single tube or in wells without a need for centrifugation .
Thus the past 2 years have seen the emergence of creative and attractive variations on the classical ChIP assay, which have enabled a considerable reduction in time, greatly simplified the procedure and made ChIP compatible with the analysis of small cell numbers. Notably, the Q2ChIP and μChIP assays also fit into the 1-day ChIP protocol category.
A Microplate-Based Assay to Enhance Throughput: Matrix ChIP
To increase the throughput of ChIP and simplify the assay, a microplate-based ChIP assay, Matrix ChIP, was recently reported . Matrix ChIP takes advantage of antibodies immobilized with Protein A coated into each well of a 96-well plate. Besides simplification of sample handling, one rationale for immobilizing antibodies is that they can be maintained in the correct orientation. Such specific orientation can enhance binding capacity to up to tenfold compared to random-oriented antibodies . All steps, from immunoprecipitation to DNA purification, are done in the wells without sample transfers, enabling a potential for automation. As mentioned above, recovery of PCR-ready ChIP DNA from the surface-bound antibodies is permitted by the use of simple buffer that facilitates DNA extraction. In its current format, matrix ChIP enables 96 ChIP assays for histone and DNA-bound proteins, including transiently bound protein kinases, in a single day .
Cleaning Up Nucleosomes to Enhance Histone ChIP Efficacy: HAP-ChIP
Many modified residues on histone tails serve as docking sites for transcription factors or chromatin modifying enzymes. In a ChIP assay, binding of these proteins may sterically hinder access of antibodies to a fraction of histone epitopes, with as a result an underestimation of the amount of a given modified histone enriched at a specific locus. To overcome this limitation, a variation on ChIP has been introduced to remove chromatin-bound non-histone proteins prior to immunoprecipitation of nucleosomes . This assay takes advantage of high-affinity interaction of DNA with hydroxyapatite (HAP) to wash out chromatin-associated proteins before ChIP under native conditions (HAP-ChIP) .
HAP-ChIP consists primarily of five steps, including purification of nuclei, fragmentation of chromatin with MNase, purification of nucleosomes by HAP chromatography, immunoprecipitation of the nucleosomes, and qPCR analysis of the precipitated DNA. Lysis of nuclei takes place in high concentration of NaCl and is immediately followed by chromatin fragmentation. High-salt lysis is believed to produce an even representation of both euchromatin and heterochromatin, which other NChIP protocols do not necessarily provide (regions of tightly packed heterochromatin are insensitive to MNase under lower salt concentrations). Additionally, elution of nucleosomes from HAP occurs with up to 500 mM NaPO4 at pH 7.2 under low-salt conditions. This preserves the interaction of DNA with core histones (histone octamers are eluted from DNA with 2 M NaCl). These procedures result in a preparation of polynucleosomes (1–3 nucleosomes per chromatin fragment), stripped of non-histone proteins . HAP-ChIP has been used in combination with qPCR; however, with a few modifications , it is speculated to be adaptable to ChIP-on-chip or ChIP-seq.
Flow Cytometry Analysis of ChIP DNA: ChIP-on-Beads
Quantitative determination of the amount of DNA associated with an immunoprecipitated protein is commonly done by qPCR [46, 56, 67]. A recent protocol, however, calls for the capture of conventional PCR products on microbeads and flow cytometry analysis . A standard ChIP is performed, and the ChIP DNA is used as template for end-point PCR in which primers are tagged in their 5′ end with Fam (forward primer) and biotin (reverse primer). The Fam/biotin PCR products are captured and analyzed by flow cytometry. Importantly, labeling must occur in the linear phase of the PCR to ensure reliable quantification. The similarity of the data obtained by qPCR and flow cytometry has been shown for the enrichment of H4 and H3 epitopes on a specific locus in Jurkat cells .
The ChIP-on-beads assay has been proposed to be useful for quantitative assessments of ChIP products in a high-throughput manner . However, the complexity of the procedure makes it at present difficult to foresee the advantage of ChIP-on-beads over ChIP-qPCR or ChIP-on-chip approaches, especially as long as a qPCR analysis of ChIP products is necessary for evaluation of the linear phase of the PCR-mediated labeling step. Simplification of the ChIP DNA fragment labeling procedure would, however, make ChIP-on-beads amenable for assessing large numbers of samples for a limited number of genes.
Analysis of Proteins Co-Enriched on Single Chromatin Fragments: Sequential ChIP
An important issue in deciphering the epigenetic code is whether two given histone modifications, transcription factors or chromatin modifiers are co-enriched on the same locus. Notably, trimethylated H3K4 and H3K27 have been suggested to constitute a ‘bivalent mark’ on genes encoding transcriptional regulators in ES cells [26, 27, 35], because both modifications could be co-precipitated from the same genomic fragment . Indeed, genome-wide approaches in cell types such as ES cells, fibroblasts, and T cells support a view of chromatin domains co-enriched in H3K4me3 and H3K27me3, albeit with distinct profiles and peaks [27, 33, 35, 45, 69]. Based on these observations, one may conclude that H3K4me3 and K27me3 may be found on distinct genomic fragments (e.g., two alleles), on the same promoter but on distinct nucleosomes, or may co-exist in a subpopulation of nucleosomes. Similar questions apply to the co-occupancy of two transcription factors on a single locus.
To resolve these issues, a sequential ChIP assay has been developed, wherein one protein is immunoprecipitated from a chromatin sample, and a second protein, presumed to be co-enriched on the same genomic fragment, is subsequently immunoprecipitated from chromatin eluted from the first ChIP [70, 71]. Sequential ChIP has been used to demonstrate the existence of bivalent histone marks on a single genomic fragment . In that study, ES cell chromatin was first immunoprecipitated with antibodies against H3K27me3 and the ChIP chromatin was used for a second immunoprecipitation using antibodies against H3K4me3. Sequential immunoprecipitation, then, retains only chromatin which concomitantly carries both histone modifications. Sequential ChIP has also been used to show the co-occupancy of two or more transcription factors on a genomic site [43, 72–76]. The sequential ChIP approach has been detailed and reviewed elsewhere [77, 78].
Of note, in a recent publication, Felsenfeld and co-workers have assessed the genome-wide distribution of nucleosomes co-enriched in the histone variants H3.3 and H2A.Z using a sequential ChIP approach coupled to high-throughput sequencing (see below) . Their findings led to the discovery that unstable nucleosomes containing both histone variants are enriched in what was previously referred to as ‘nucleosome-free regions’ on active promoters, enhancers, and insulator regions. Interestingly, these authors also show that common chromatin preparation methods result in the loss of such unstable double-variant containing nucleosomes . The level of analysis of co-occupancy of two proteins on a locus can potentially be further refined using purified mono-nucleosomes as chromatin templates.
Methods for Genome-Wide Mapping Protein Binding Sites on DNA
ChIP has for several years been limited to the analysis of pre-determined candidate target sequences analyzed by PCR using specific primers. Recently, several strategies have been developed to enable application of ChIP to the discovery of novel target sites for transcriptional regulators and to map the positioning of post-translationally modified histones throughout the genome. These genome-wide approaches have immensely contributed to characterizing the chromatin landscape primarily in the context of pluripotency, differentiation, and disease.
The advent of oligonucleotides microarrays has revolutionized analysis of gene expression and our understanding of transcription profiles. Subsequent development of genomic DNA microarrays (chips) has, when combined with ChIP assays, enabled the mapping of transcription factor binding sites [80, 81] and of histone modifications [82, 83] on large areas in the genome through an approach known as ChIP-on-chip. Despite its relatively recent introduction, ChIP-on-chip has been largely exploited to, for example, map c-Myc binding sites in the genome [84, 85], elaborate Oct4, Nanog, and Sox2 transcriptional networks in ES cells , identify polycomb target genes [87, 88] or provide a histone modification landscape in T cells . Several reviews dedicated to ChIP-on-chip, its variations and limitations have been published [89–91], thus we only provide here a brief account of the principle.
ChIP-on-chip is only as informative as the oligonucleotide microarrays onto which the ChIP-enriched DNA is hybridized. This limitation has stimulated the development of methods for unbiased determination of genomic sequences associated with a given protein. Novel transcription factor binding sites can be identified by cloning and sequencing DNA from the ChIP material [92, 93]. However, the overwhelming excess of non-specifically precipitated DNA fragments makes ChIP-cloning unpractical. A ChIP-display strategy has been designed and applied to the identification of target genes occupied by the transcription factor Runx2 . ChIP-display concentrates DNA fragments containing each target sequence and scatters the remaining, non-specific, DNA. Target sequences are concentrated by restriction digestion and electrophoresis, as fragments harboring the same target site acquire the same size. To scatter non-specific fragments, the total pool of restriction fragments is divided into families on the basis of identity of nucleotides at the ends of these fragments. Because all restriction fragments displaying each given target harbor the same nucleotide ends, they remain in the same family and the family detection signal on gel is not altered. Non-specific background fragments, however, are scattered into many families so that each family detection signal is markedly lower .
ChIP-display can unravel transcription factor targets in ChIPs that are enriched for targets by as little as 10- to 20-fold over bulk chromatin , and as such shows reasonable sensitivity. Gel electrophoresis display of ChIP DNA products allows a direct comparison of patterns (i.e., targets) obtained from different cell types . ChIP-display is also relatively insensitive to background which characterizes ChIP-PCR or ChIP-on-chip approaches. However, ChIP-display is not well suited for a comprehensive analysis of target sequences for proteins with a large number of genomic targets, such as SP1, GATA proteins, histone deacetylases, polycomb proteins, or RNAPII , or for the mapping of histone modifications. It is better suited for transcription factors with a more limited number of targets; however, it lacks the quantification of the relative abundance of a transcription factor associated with a given locus, which is enabled by qPCR.
ChIP-Paired End Ditag: ChIP-PET
A second strategy developed in response to the limitations of the ChIP-on-chip assay is based on sequencing portions of the precipitated DNA. Indeed, with a limited survey of the cloned ChIP DNA fragment library, distinguishing between genuine binding sites and noise without additional molecular validation is challenging. In contrast, with a wide sampling of the ChIP DNA pool, sequencing approaches can identify DNA fragments enriched by ChIP.
ChIP-paired end ditag (PET) exploits the efficiency of sequencing short tags, rather than entire inserts, to enhance information content and increase accuracy of genome mapping . ChIP-PET relies on the recently reported gene identification signature strategy in which 5′ and 3′ signatures of full-length cDNAs are extracted into PETs that are concatenated [95, 96]. The sequences are subsequently mapped to the genomic sequences to delineate the transcription boundaries of every gene. As in the gene identification signature strategy, a pair of signature sequences (tags) is extracted from the 5′ and 3′ ends of each ChIP DNA fragment, concatenated and mapped to the genome.
The PET approach has recently been exploited to characterize ChIP DNA fragments in order to achieve unbiased, genome-wide mapping of transcription factor binding sites [43, 44]. From a saturated sampling of over 500,000 PET sequences, Wei and colleagues characterized over 65,000 unique p53 ChIP DNA fragments and established overlapping PET clusters to define p53 target sequences with high specificity. The analysis also enabled a refinement of the consensus p53 binding motif and unraveled nearly 100 previously unidentified p53 target genes implicated in p53 function and tumorigenesis . In addition, a ChIP-PET analysis of binding sites for Oct4 and Nanog in mouse ES cells has laid out a transcription network regulated by these proteins in these cells .
ChIP with DNA Ligation and Selection: ChIP-DSL
With the aim of detecting DNA target motifs with higher sensitivity and specificity than through conventional ChIP-on-chip, a multiplex assay coined as ChIP-DSL was introduced. ChIP-DSL combines ChIP with a DNA ligation and selection (DSL) step . The assay involves the pre-determined use, or construction, of a microarray of 40-mer probes onto which the ChIP DNA fragments are to be hybridized. The reason for this is that a pair of 20-mer ‘assay oligonucleotides’ is synthesized corresponding to each half of each 40-mer. These 20-mer oligonucleotides are flanked on both sides by a universal primer binding site. These oligonucleotides are mixed into a ‘DSL oligo pool’. Following conventional ChIP, the purified ChIP DNA is randomly biotinylated and annealed to the DSL oligo pool. The annealed fragments are captured on streptavidin-conjugated magnetic beads, allowing elimination of the non-annealed 20-mers (the noise). All selected DNA fragments are immobilized onto the beads and those paired by a specific DNA target motif are ligated. Thus, the correctly targeted oligonucleotides are specifically turned into templates for PCR amplification. One of the PCR primers is fluorescently labeled to enable detection after hybridization on the 40-mer probe microarray. The DSL procedure is also carried out for input DNA using PCR primers labeled with a different fluorophore.
ChIP-DSL is claimed to present advantages over ChIP-on-chip  on several grounds. Only unique signature motifs are targeted, alleviating potential interference with repetitive and related sequences upon hybridization. Sensitivity of the assay is increased due to the PCR amplification step. Amplification is presumably unbiased because DNA fragments bear the same pair of specific primer binding sites and have the same length.
ChIP-DSL has been used to identify a large number of novel binding sites for the estrogen receptor alpha in breast cancer-derived MCF7 cells . ChIP-DSL has also been used to show widespread recruitment of the histone demethylase LSD1 on active promoters, including most estrogen receptor alpha gene targets .
Perhaps the most powerful strategy to date for identifying protein binding sites across the genome consists of directly and quantitatively sequencing ChIP products. In an ultra high-throughput sequencing approach [35, 45, 99], DNA molecules are arrayed across a surface, locally amplified, subjected to successive cycles of single-base extension (using fluorescently labeled reversible terminators) and imaged after each cycle to determine the inserted base. The length of the reads is short (25–50 nucleotides using the Illumina/Solexa platform), however, millions of DNA fragments can be read simultaneously.
ChIP-seq has been used to generate ‘chromatin-state maps’ for ES and lineage-committed cells . The data corroborate ChIP-on-chip data on the same cell types reported earlier by the same group , as well as results reported independently by ChIP-PET . Using the Illumina/Solexa 1G platform, binding sites for the transcription factor STAT1 in HeLa cells  and a profiling of histone methylation, histone variant H2A.Z binding, RNAPII targeting, and CTCF binding throughout the genome  have also been reported. All results claim robust overlap between ChIP-seq, ChIP-on-chip, and ChIP-PCR data. Interestingly, the ChIP-seq data illustrate the potential for using ChIP for genome-wide annotation of novel promoters and primary transcripts, active transposable elements, imprinting control regions, and allele-specific transcription . As outlined above, ChIP-seq can be applied downstream of a sequential ChIP approach to identify genomic regions co-enriched in histone variants . Insights into the analysis of large data sets related to array and sequencing data have recently been published .
ChIPseq for Small Cell Numbers
Current state-of-the-art technologies for genome-wide mapping of protein localization by ChIPseq entail many steps that introduce bias and inefficacy. One limiting step in that regard is the amplification of the immunoprecipitated DNA and construction of the library. Bernstein and colleagues have recently applied a single-molecule approach to directly sequence ChIPed chromatin with minimal sample manipulation . The approach is compatible with as little as 50 pg of ChIP DNA and is expected to facilitate the establishment of chromatin-state maps from small cell numbers.
Controls in ChIP Assays
In spite of improvements in the ChIP assays to reduce or eliminate background chromatin , background does exist and needs to be accounted for using appropriate negative controls. A survey of the ChIP literature reveals the use of various controls, the nature of which seems to mainly depend on the investigator. One classical negative control is the use of no antibodies (also often referred to as a ‘bead-only’ control). Bead-only controls for unspecific binding of chromatin fragments to the beads used to precipitate the complex of interest. Although it is useful, this control is not as stringent as using an irrelevant antibody, preferably of the same isotype as the experimental antibody, in a parallel chromatin preparation. Enhanced stringency of the control also implies the use or an irrelevant antibody against a nuclear protein. A third negative control consists of comparing, in the same ChIP, protein enrichment on a target sequence relative to enrichment on another, irrelevant, region. This control was performed in our laboratory to demonstrate the specificity of occupancy of Oct4 on the NANOG promoter in pluripotent carcinoma cells, whereas it was virtually absent from the GAPDH promoter . In ChIP-PCR experiments, the negative control may generate a PCR signal that can be used as a reference to express a ChIP-specific enrichment. In ChIP-on-chip or ChIP-cloning-sequencing (such as ChIP-PET) assays, the negative control IP is used in a subtractive approach at the level of array analysis. In addition to a negative control, some investigators use a positive control, such as a high-quality antibody against a well-characterized ubiquitous transcription factor . Positive control antibodies are particularly important when setting up new methodologies.
Variations on a Theme: Analysis of DNA Methylation Using Precipitation Strategies
In addition to the ChIP techniques reviewed here, various approaches have also been developed and applied to investigate other aspects of chromatin organization—notably DNA methylation states.
ChIP Coupled to Bisulfite Genomic Sequencing: ChIP-BA
Profound understanding of the interplay between histone modifications, DNA methylation, transcription factor binding, and transcription requires the combination of multiple analyses from a single chromatin or DNA sample. The CG content of a transcription factor binding site, thus its methylation state, is likely to affect binding . In an attempt to relate transcription factor binding to DNA methylation, ChIP has been combined with bisulfite genomic sequencing analysis in a ChIP-BA approach . ChIP DNA fragments are processed for PCR analysis (or array hybridization) and for bisulfite conversion to determine the CpG methylation pattern. ChIP-BA has been used to determine the DNA methylation requirements for binding of a methyl-CpG binding protein . The method can also potentially be useful to unravel methylation patterns that are compatible, or incompatible, with the targeting of a specific protein to a genomic region . A potential problem with ChIP-BA, however, is noise that is directly turned into a sequence which may be irrelevant. Subtractive strategies may conceivably be utilized provided appropriate controls are performed.
An alternative to ChIPing a protein is to label DNA close to the target site of the protein of interest . Labeling consists of a methylation tag put on by a DNA adenine methyltransferase (Dam) fused a DNA binding protein (the protein of interest) (DamID approach) . Binding of the transcription factor-Dam protein to DNA elicits adenine methylation in the vicinity of the protein target site. The methylated sites are detected by digestion with a methyl-specific restriction enzyme. The digestion products are purified, amplified using a methylation-specific PCR assay, labeled, and hybridized onto a microarray. DamID has been used to uncover binding sites for transcription factors, DNA methyltransferases, and heterochromatin proteins in Drosophila, Arabidopsis, and mammalian cells [105–109], and more recently, nuclear lamin B1 . Of interest, a comparison of the DamID and ChIP-on-chip approaches has been reported .
Methylated DNA Immunoprecipitation: MeDIP
A variation of the ChIP assay has been introduced to determine genome-wide profiles of DNA methylation. Methyl-DNA immunoprecipitation (MeDIP) consists of the immunoprecipitation of methylated DNA fragments using an antibody to 5-methyl cytosine [111, 112]. Detection of a gene of interest in the methylated DNA fraction can be done by PCR, hybridization to genomic (promoter or comparative genomic hybridization) arrays [112, 113] or high-throughout sequencing . Although MeDIP proves to be a potent method, a constraint of the assay is its limitation to regions with a CpG density of at least 2–3% . Below this density, even methylated CpGs will be regarded as unmethylated relative to genome average. MeDIP is being increasingly used to map methylation profiles (the ‘methylome’) of promoters in a variety of organisms and cell types [16, 112] including mesenchymal stem cells (A.L. Sørensen, A.H. Reiner, B.M. Jacobsen, and P. Collas, unpublished data). Protocols and reviews on the MeDIP approach have been recently published [114–118].
Conclusions and Prospects
ChIP has become the technique of choice for mapping DNA–protein interactions in the cell, identify novel binding sites for transcription factors or other chromatin-associated proteins, map the localization of post-translationally modified histones and map the localization of histone variants. These studies unravel an increasingly complex epigenetic landscape in the context of gene expression, definition of gene boundaries, development, differentiation, and disease. The advent of ChIP assays for small cell samples has moved ChIP forward into the fields of early embryo development and small cancer biopsies. The combination of small-scale ChIP assays with increasingly robust DNA amplification methods enables genome-wide analyses of histone modifications or RNAPII binding in small cell samples. Approaches to directly sequence small amounts of ChIPed DNA also open the door to eliminating bias unavoidably generated by amplifying ChIP DNA. This is expected to move the field towards true whole-genome analyses of embryos.
ChIP assays have also in recent years become significantly more user-friendly, with fewer steps, reduced sample handling and faster assays. Efforts have been put into simplifying isolation of ChIP DNA, for a quicker analysis and minimizing sample loss. Some of the new developments such as matrix ChIP also seem to be suited for automation. In an era which promotes the concept of personalized medicine in a context where epigenetics is increasingly linked to disease, automated whole-genome epigenetic analyses of individual patient material is likely to become reality. To set the path in this area, a proof-of-concept of an automated ChIP assay has recently been reported using a lab-on-a-chip microfluidic device . The current device is suited for four parallel automated ChIPs each for ~2,000 cells. Increased parallelism should truly enhance the throughput of epigenetic analyses.
Our work is supported by grants from the Research Council of Norway, the Norwegian Cancer Society and the University of Oslo.