Chromatin immunoprecipitation: optimization, quantitative analysis and data normalization
1Swammerdam Institute for Life Sciences, Universiteit van Amsterdam, Kruislaan 318, 1098 SM Amsterdam, The Netherlands
2Institute for Biology I, Aachen University, Worringer Weg 1, 52074 Aachen, Germany
author email corresponding author email
Plant Methods 2007, 3:11doi:10.1186/1746-4811-3-11
The electronic version of this article is the complete one and can be found online at: http://www.plantmethods.com/content/3/1/11
© 2007 Haring et al; licensee BioMed Central Ltd.
Chromatin remodeling, histone modifications and other chromatin-related processes play a crucial role in gene regulation. A very useful technique to study these processes is chromatin immunoprecipitation (ChIP). ChIP is widely used for a few model systems, including Arabidopsis, but establishment of the technique for other organisms is still remarkably challenging. Furthermore, quantitative analysis of the precipitated material and normalization of the data is often underestimated, negatively affecting data quality.
We developed a robust ChIP protocol, using maize (Zea mays) as a model system, and present a general strategy to systematically optimize this protocol for any type of tissue. We propose endogenous controls for active and for repressed chromatin, and discuss various other controls that are essential for successful ChIP experiments. We experienced that the use of quantitative PCR (QPCR) is crucial for obtaining high quality ChIP data and we explain why. The method of data normalization has a major impact on the quality of ChIP analyses. Therefore, we analyzed different normalization strategies, resulting in a thorough discussion of the advantages and drawbacks of the various approaches.
Here we provide a robust ChIP protocol and strategy to optimize the protocol for any type of tissue; we argue that quantitative real-time PCR (QPCR) is the best method to analyze the precipitates, and present comprehensive insights into data normalization.
Epigenetic regulation of gene expression is crucial for cell differentiation, and thus essential for normal growth and development of higher eukaryotes. Epigenetic control is an intricate interplay between various molecular mechanisms, e.g. DNA methylation and histone modifications (reviewed in [1-4]). Whereas DNA methylation has been studied in great detail for several decades, the role of histone modifications has only been fully appreciated for about 10 years . Since then the number of papers on new histone modifications and their possible functions has exploded.
The most widely used procedure to examine histone modifications is Chromatin Immunoprecipitation (ChIP), a technique first established for cultured Drosophila cells . In short, ChIP relies on antibodies to identify the presence of specific histone modifications at DNA regions of interest. Chromatin is extracted from cells or tissue, fragmented and incubated with antibodies against specific histone modifications (Figure 1). The chromatin fragments bound to the antibodies are captured using protein A/G beads, and DNA is isolated from the precipitate. This DNA is usually analyzed by (quantitative) PCR to determine the abundance of a region of interest in the precipitated material. ChIP has proven to yield very valuable information on chromatin-associated processes in eukaryotes, including plants and humans. Despite the fact that ChIP on plant material is established for the widely used model system Arabidopsis, to implement and optimize the technique for other model plant species, such as maize, is still very challenging and time consuming.
Figure 1. Outline of the ChIP-QPCR procedure. This outline represents the ChIP procedure as described in the text. IP, ChIP sample; NoAb, No-antibody control.
In current literature, conventional PCR is mostly used to analyze plant ChIP precipitates. In this method, the intensity of a DNA band on an agarose gel is assumed to reflect the initial abundance of a specific DNA fragment in the precipitate. Instead, the intensity of the band reflects the endpoint of a non-linear PCR reaction. Since a careful quantification of ChIP signals is important for a correct interpretation of the data, we discuss the application of quantitative real-time PCR (QPCR).
Once the QPCR data are obtained, they can be normalized and presented in different ways. All normalization methods have their own advantages and drawbacks. Therefore, making a well-informed choice is important for correct interpretation of the data. For example, the most commonly used methods, '% of input' (%IP) and 'fold enrichment', may obscure the biological meaning of the ChIP signal by relating the signal intensity to an arbitrary amount of chromatin or to background levels, respectively. Besides the inherent disadvantages of the various methods, the wide range of normalization methods that is currently used can hamper the comparison of published data sets .
In this paper we present a robust, optimized ChIP protocol, and in addition a strategy to optimize the protocol when dealing with different experimental systems or conditions. Quantitative real-time PCR (QPCR) is presented as best practice to analyze the precipitated material. A quantitative interpretation of ChIP-QPCR data requires normalization, an often under-illuminated aspect of the ChIP-QPCR procedure. Therefore, this paper provides a discussion on the use of various normalization methods, enabling a well-informed choice for a specific normalization method.
Results and Discussion
Chromatin immunoprecipitation experiments can roughly be divided into two categories. One uses crosslinked chromatin sheared by sonication (X-ChIP), and the other native chromatin digested by nucleases (N-ChIP). Both methods have their own advantages and disadvantages . Our paper provides and discusses a protocol for X-ChIP that involves formaldehyde crosslinking of the chromatin in intact tissue, ensuring the rapid fixation of the existing chromatin structure. Without crosslinking, we systematically failed to obtain significant amounts of precipitate. In this paper, we discuss the whole procedure in four sections: (1) The isolation of good-quality chromatin from plant material, (2) the chromatin immunoprecipitation itself, (3) the analysis of precipitated material by QPCR and (4) data normalization. An outline of the various steps in the whole procedure is shown in Figure 1.
(1) Isolation of chromatin
The isolation of plant chromatin needs a plant-specific approach; plant cells are surrounded by a cell wall, and generally whole tissues, rather than uniform cell cultures, are used to isolate chromatin. Plant cells contain large vacuoles, resulting in a relatively low number of nuclei per gram of tissue. In addition, vacuoles are a source of proteolytic activities . We advise the use of healthy, unfrozen plant tissue as starting material for the isolation of chromatin. If possible, use tissue enriched in unexpanded cells. Such tissue will provide the best yield and purity of the isolated chromatin. In case there are doubts if the tissue of interest will yield good-quality chromatin, non-crosslinked nuclei can be isolated followed by micrococcal nuclease digestion. Good-quality chromatin gives rise to a distinct nucleosome ladder [8,10].
Crosslinking of the starting material by formaldehyde is used to ensure that the chromatin structure is preserved during the isolation and ChIP procedure [6,11]. Crosslinking of chromatin within the plant tissue requires the fixative to penetrate the cells, which is hampered by the plant's waxy cuticle and spongy air-filled mesophyll. In our protocol, efficient penetration of the fixative is achieved by vacuum infiltration of buffer that contains formaldehyde. It is important that the buffer penetrates the plant material completely; after vacuum infiltration the plant material should appear translucent or 'water-soaked'. The buffer volumes used in the presented protocol are suitable for efficiently crosslinking up to 5 grams of plant material. Increasing the amount of material impedes efficient crosslinking.
The crosslinking step should be optimized, since too little crosslinking will not sufficiently preserve the chromatin structure, and too much crosslinking will hamper the ChIP procedure . A method to determine the optimal crosslinking conditions is illustrated in Figure 2. Essentially, when the crosslinking is optimal for ChIP, decrosslinking is required to efficiently isolate DNA from the nuclei by phenol-chloroform extraction. The chromatin is over-crosslinked when it is impossible to recover a substantial amount of DNA from the nuclei by decrosslinking. The nuclei are under-crosslinked when most of the DNA can be recovered without decrosslinking. Crosslinked material can be stored for several months at -80°C after freezing in liquid nitrogen.
Figure 2. Crosslinking efficiency analysis. Leaf material was crosslinked in buffer containing increasing amounts of formaldehyde. Samples were decrosslinked (+DC) or not (-DC), and DNA isolated using phenol/chloroform extraction followed by ethanol precipitation. While DNA is efficiently isolated from samples that were not crosslinked (lanes indicated with 0×), decrosslinking is required for the isolation of DNA from crosslinked samples (lanes indicated with 1× and 3×). Over-crosslinking strongly decreases DNA isolation efficiency (lanes indicated with 3×). DC, decrosslinking; 0×, 1×, 3×, relative concentration of formaldehyde; M, lambda DNA cut with PstI.
Shearing of chromatin
The resolution obtained by the ChIP procedure is determined by the size of the chromatin fragments used as input material. Two methods are commonly used to fragment chromatin, sonication (hydrodynamic shearing) and micrococcal nuclease (MNase) digestion. Both methods can show preferential fragmentation of certain chromosomal regions [13,14]. When using formaldehyde crosslinking, sonication is the preferred method, as crosslinking restricts the access of MNase to chromatin .
Optimal fragmentation can be achieved by testing various sonication conditions on chromatin, followed by DNA isolation and estimating the sonication efficiency by gel electrophoresis. Ideally, the bulk of the chromatin is sonicated to a length between 250 and 750 bp. For efficient fragmentation, sonication at low power, in combination with several pulses, is preferred over sonication at high power and few pulses, but conditions vary with the sonication device used. It is important to keep the chromatin sample cooled on ice during sonication, as heat released by the sonication probe can reverse the crosslinks. The presence of detergent (SDS) in the sonication buffer improves sonication efficiency considerably, but can induce foaming during sonication. Foam makes the chromatin sample unsuitable for ChIP, probably as a result of the surface tension imposed by the foam, which can disrupt protein conformation . Foaming can be prevented by decreasing the sonication power. Sonicated chromatin can be stored at -80°C for at least a few months, but repeated freeze/thaw cycles should be avoided.
(2) The chromatin immunoprecipitation procedure
Antibodies (Ab) are the most important factor for a successful ChIP experiment. It is crucial to choose an antibody carefully, especially when antibodies are raised against non-plant proteins. Papers reporting the use of a specific antibody for ChIP are a good indication of the suitability of that antibody, but one should realize that quality can differ between antibody batches. The successful use of a specific antibody in experiments other than ChIP (i.e. Western blotting, immunocytochemistry) does not automatically mean the antibody is suitable for ChIP; that has to be tested.
Antibodies can be available as polyclonal or monoclonal preparations. Monoclonal antibodies have a high specificity compared to polyclonal sera, but the polyclonal sera may recognize several epitopes of the target, increasing signal levels of low-abundance templates. Be aware that polyclonal serum against multiple modifications (i.e. against hyperacetylated H4) may have an undocumented preference for one modification over the others , impeding the biological interpretation of ChIP data.
Different antibody preparations have distinct properties, which can affect the ChIP results. The affinity for epitopes differs between antibodies, affecting the resulting signal levels. For example, antibodies can differ in their sensitivity towards crosslinks or adjacent modifications (discussed in ). In addition, the relation between the availability of epitopes and antibody binding may not be linear. Some antibodies are sensitive to inhibitory factors present in the input chromatin sample, resulting in a decrease in binding efficiency of the antibody when increasing the amount of input. This is exemplified in Figure 3. We tested the effect of changing the amount of input chromatin on ChIP efficiency (the recovery of DNA relative to the amount of input) for two different antibodies. With an antibody against hyperacetylated H4 (Upstate #06–946), the ChIP efficiency is constant over a broad range of chromatin concentrations. When this experiment is performed with an antibody recognizing an invariant domain of H3 (H3core; Abcam #AB1791), dilution of the input chromatin improves the ChIP efficiency. These data show that the relation between ChIP signal and chromatin input depends on the characteristics of the antibody used. To ensure the comparability of results obtained with different input samples, we recommend determining the optimal Ab: chromatin ratio by titration of the amount of input chromatin, and the use of similar amounts of input chromatin when ChIP results are to be compared.
Figure 3. Titration of chromatin relative to a fixed amount of antibody. (a) ChIP titration experiment using an antibody against acetylated H4. Dilution of the chromatin has no effect on the precipitation efficiency. (b) ChIP titration experiment using an antibody against an invariant domain of H3 (H3core). Dilution of the chromatin improves the precipitation efficiency, suggesting that the antibody recognizing the H3core is sensitive to inhibitory factors present in the chromatin sample. ChIP-QPCR was performed as described in this paper. Chromatin was isolated from leaves of 2 week-old plants, sonicated and used in 1:1, 1:2 and 1:4 dilutions for ChIP. The input samples are diluted by the same factors as the chromatin samples. Results are represented as %IP, the error bars indicate the standard error. ChIP samples are represented by closed bars, while open bars indicate the signals from the 'no-antibody' serum controls. Each data point represents the average of two different chromatin samples, each analyzed in duplo.
When establishing the ChIP procedure we recommend using an antibody against an invariant domain of histones as a starting point [19,20]. In our hands, such an antibody usually results in a good signal-to-noise ratio. When setting up ChIP to detect histone acetylation, the initial detection can be facilitated by treating plant material with butyrate or TSA before the isolation of chromatin [7,20]. Such treatments inhibit histone deacetylation, resulting in increased global acetylation levels.
Once the ChIP protocol has been established, it has to be determined which histone modifications will be analyzed. For this purpose, data obtained from immunocytochemistry experiments are a valuable source of information [21,22]. Such experiments should however, be considered as informative but not conclusive for the decision on which histone modification to analyze. Immunocytochemistry experiments provide a global view of the localization of histone modifications at eu- or heterochromatin, while ChIP experiments can have a single-nucleosome resolution . At such high resolution, histone modifications can be observed where they are not expected based on immunolocalization experiments ([20,22]; Haring et al, in preparation). Another factor to keep in mind is that specific chromatin states can be associated with different histone modifications in distinct species. For example, immunolocalization experiments showed the presence of H3K9me2, but not H3K9me3, in heterochromatin of Arabidopsis, while H3K9me3, but not H3K9me2, is observed in heterochromatin of mice [21,24,25].
Controls for ChIP
To ensure the reliability of ChIP data, two control samples specific for the ChIP experiment should be included: the input sample and the 'no-antibody' (NoAb) control sample (Table 1). Both the input sample and the NoAb control sample provide essential information about the ChIP experiment, and should be analyzed with every primer set used. Additional controls are needed for the QPCR procedure; these will be discussed in the QPCR section.
Table 1. Controls for ChIP and QPCR
The input sample will be indicative for the presence and amount of chromatin used in the ChIP reaction. It is an aliquot taken from the chromatin before preclearing (step 9 in the protocol). The chromatin aliquot is decrosslinked and DNA is isolated. This DNA sample should yield a PCR product with all primer sets used. Besides serving as a positive control, the data derived from the input sample can be used for normalization by the '%IP' method discussed in the section on data normalization.
The NoAb control is a chromatin sample to which non-specific control serum is added instead of a specific antibody (see materials and methods). The NoAb sample is treated the same way as the ChIP samples. The QPCR signals resulting from the NoAb samples indicate the amount of background signal generated by the chromatin preparations and ChIP procedure. Ideally, the washing steps remove non-specifically bound chromatin, resulting in an absence of QPCR signals for the NoAb samples. In reality however, it is not uncommon to find a PCR product for the NoAb control sample. It is very important that the DNA isolated from the NoAb samples is amplified with every primer pair used, as the level of background signal can differ for each primer pair.
Optimizing the signal to noise ratio
When setting up ChIP, one of the main problems is a high level of background signal (NoAb control) relative to the level of the signal of interest (ChIP sample). This hampers distinguishing a 'true signal' from the background signal. Therefore, we recommend setting up ChIP using an antibody against an invariant domain of histone H3 [19,20]. In our experiments, such an antibody yields a very clear difference in signal between the ChIP samples and NoAb control. Another helpful tool to discriminate between background and ChIP signal is a positive control sequence for ChIP, as discussed in the 'controls for specific chromatin states' section.
Several options are available to optimize the signal-to-noise ratio. To prevent non-specific binding of chromatin to the protein A/G agarose beads, the beads are blocked with BSA and non-specific blocking DNA before they are used in ChIP assays. To further reduce background, BSA, non-specific serum and unrelated DNA can be added to the pre-incubation step of chromatin with beads (step 9). Options for preventing chromatin from binding to plastic tubes are pre-incubation of the tubes with BSA and unrelated DNA (e.g. salmon sperm DNA), or the use of siliconized tubes. The background can also be reduced by lowering the amount of input chromatin, the amount of protein A/G agarose beads or the amount of antibody, but this may lower signal levels as well.
In case ChIP experiments with an antibody against an invariant domain of histone H3 would result in a low ChIP-QPCR signal relative to the background signal, factors inhibiting the PCR reaction may be involved. We recommend purifying the DNA after the reverse crosslinking (step 17) with a commercial 'PCR purification spin column kit'. In our hands, DNA purification by phenol-chloroform extraction does not sufficiently remove detergents present in the ChIP elution buffer, inhibiting subsequent QPCR reactions.
(3) Analysis of precipitated material by QPCR
DNA isolated from the precipitated chromatin has to be analyzed to determine which DNA fragments are present in the precipitate (Figure 1). For the detection of specific DNA fragments, various methods are available, and the chosen method determines to what extent the data can be analyzed quantitatively. Commonly used analysis methods are conventional PCR and quantitative PCR. Alternative methods are microarray analysis and slot blotting. Microarray analysis is useful when studying the genome-wide distribution of histone modifications . Analysis by slot blotting is feasible when dealing with highly repetitive sequences [27,28], but appears not sensitive enough to detect single copy sequences in ChIP samples (M. Haring and M. Stam, unpublished). This paper focuses on quantitative PCR for the detection of immunoprecipitated DNA (see Figure 4 for an outline how to set up QPCR).
Figure 4. Outline for setting up QPCR. This outline represents setting up the QPCR analysis, as described in the text. Abbreviations: TF: transcription factor.
Until recently, ChIP precipitates have primarily been analyzed by conventional PCR (e.g. bands on a gel). Such an approach requires lots of testing per individual primer set to ensure that measurements are taken in the linear range of amplification. If this condition is not fulfilled, the resulting data cannot be considered quantitative, impeding data interpretation. For this reason, analyzing ChIP precipitates by quantitative real-time PCR (QPCR) has several advantages over conventional PCR. The QPCR technique does not quantify the amount of PCR product at the end of the PCR reaction, as with conventional PCR band densitometry. Instead, the initial amount of template DNA is calculated from the kinetics of the PCR reaction. During a QPCR run, the accumulation of PCR product is measured every cycle. The number of cycles needed to reach a certain amount of PCR product – 'Cycle threshold' or Ct value -, and a calibration line (see below), are used to calculate the initial amount of DNA template.
Quantitative PCR analysis for ChIP can be performed using DNA-dye based or probe-based PCR product detection chemistries. The most widely used DNA-dye based QPCR chemistry employs the fluorescent dye SYBRgreen for detection of the amplicon. SYBRgreen is only fluorescent when bound to double-stranded DNA (dsDNA), and the amount of fluorescence is proportional to the amount of dsDNA. SYBRgreen detects a PCR product independent of its DNA sequence, so this chemistry can be used for all primer sets. At the same time, this sequence-independent binding of SYBRgreen requires primer sets to be thoroughly optimized; amplification of non-target DNA fragments and formation of primer dimers will also yield fluorescent signal, hampering measurements of the ChIP-QPCR signal.
When using hybridization probe-based chemistries, one or more fluorescently labeled oligonucleotide probes are designed to anneal to a specific DNA sequence . This adds specificity to the detection of PCR fragments, as both the primers and the probe will have to anneal to template DNA to enable detection of the product.
QPCR primer sets
In order to obtain high quality QPCR data, the primer sets used need to meet specific criteria [29,30]. If primer sets are not optimized, this may result in amplification artifacts and/or an inaccurate quantification. Primer sets can be tested by performing QPCR on serial dilutions of template DNA isolated from crosslinked, sonicated chromatin. The efficiency of the amplification should be as close as possible to two (a 2× dilution of the template should result in a 1 Ct increase), and should remain constant over a wide range of template concentrations. The amplification efficiency can be improved considerably by designing the length of the amplified fragment between 80 and 150 bp. Importantly, the ChIP-QPCR results should be within the linear range of amplification. To verify that the primer sets amplify the correct fragment, sequencing the obtained PCR product is recommended. When using SYBRgreen for detection of the QPCR signal, the amplification of non-target DNA and the formation of primer dimers should be avoided, as the dye does not discriminate between these and target DNA fragments (discussed above).
The sequence regions to be amplified have to be chosen very carefully. Different combinations of histone modifications mark distinct DNA sequence elements, such as regulatory sequences, promoter- or coding sequences. The histone modifications present at such specific DNA elements can be different from those at the flanking chromatin . We therefore advise to design primer sets within a single DNA element, such as a promoter, intron or other regulatory sequence. When the amplified region spans more than one DNA element, the data obtained by ChIP may consist of a mixture of data, hampering the biological interpretation. In conclusion, knowing as many features of the amplified sequences as possible is useful for the design of useful primer sets.
Controls for QPCR
A good-quality QPCR analysis can only be performed when taking along the proper controls. These controls are in addition to the earlier discussed input and NoAb controls, which are specific for the ChIP experiment. This section will discuss the required QPCR specific controls: a calibration line, melting curve analysis and the no-template control (Table 1B).
A QPCR calibration line is required for the quantification of ChIP signals, and should be included for every primer set used. A calibration line consists of QPCR reactions performed with multiple dilutions of template DNA that is isolated from sonicated crosslinked chromatin. The calibration line corrects for differences in PCR efficiency between various primer sets, as well as for differences in signal level between multicopy and single copy sequences. Besides enabling quantification, the calibration line functions as a positive control for the QPCR assay, independent of the ChIP samples. The dilution series for the calibration line must be made fresh from a DNA stock every time, as the quality of diluted DNA rapidly decreases. The latter is probably due to binding of DNA to the surface of the reaction tube . Addition of BSA or unrelated DNA may reduce this effect.
When performing QPCR with SYBRgreen, a melting curve analysis can be performed to determine if the correct fragment is amplified. In this extension of the PCR program, the temperature is gradually increased while continuously measuring SYBRgreen fluorescence. PCR products will denature at a temperature specific for their size and sequence, and this will be measured as a loss of fluorescence. Multiple PCR fragments in one reaction, or the presence of primer-artifacts, will result in a step-wise decrease of fluorescence. Optimal primer sets should yield a single denaturation event. Even so, multiple fragments of similar size and/or sequence may yield identical melting curves. The PCR products should therefore be verified by sequencing before using the primers for ChIP-QPCR analysis. The melting curve analysis is not possible when using QPCR chemistries based on fluorescent oligonucleotide probes.
A no-template control, which is commonly included in conventional PCR, is useful for the detection of master mix contamination and primer dimers. The no-template control is distinct from the NoAb control, as the latter usually results in PCR products.
(4) Controls for specific chromatin states
Genomic sequences which are known to be associated with specific histone modifications are essential controls for the interpretation of ChIP data (Table 1C). When, in a ChIP experiment, the regions of interest do not show any enrichment of a specific histone modification, the control sequences can indicate if this is due to a failed experiment or if the outcome reflects the actual situation. Ideally, the controls function as a positive and negative control for several antibodies in various tissues.
The development of positive ChIP controls is challenging as chromatin states can not simply be classified as "active" or "silent", many different types of active and silent chromatin exist. These chromatin states are the result of several different molecular networks that can act together at a particular genomic location, each resulting in a distinct set of histone modifications and other chromatin features [33,34]. Identifying a positive control for a specific chromatin state is facilitated by knowledge of the regulatory processes that underlie the presence of the histone modification of interest.
Here, we will discuss controls for two different chromatin states: transcriptional active and repressed chromatin. In our example, the positive control for active chromatin is a negative control for repressed chromatin and vice versa.
Importantly, if antibodies are used against histone modifications associated with active chromatin, the control sequence for active chromatin must result in a positive signal in all tissues of interest. A good candidate is a housekeeping gene that is constitutively expressed, such as the actin gene, which is frequently used as a positive control for active chromatin [35-39]. Even if the expression level of a control gene varies in different tissues, it may still function as a good control; quantitative changes in expression level do not necessarily reflect changes in histone modifications.
We have used maize actin1 coding sequences to develop positive controls for active chromatin. Cloning and sequencing of actin1 DNA from our maize lines indicated it is a single copy gene and well suited for QPCR analysis. RT-PCR analyses showed it is active in the tissues we have examined (husk and young leaves; data not shown). ChIP was performed as described in this paper, and input chromatin was isolated from leaves of one month old plants (young leaves) and from husks of three months old plants (husk leaves). The precipitates were analyzed with primers specific for two positions on the actin gene: the untranslated leader (UTR) and the second exon (exon 2; Figure 5a). We observed positive signals for both the