1. High-quality genomic DNA is isolated from cells/tissues using the Qiagen DNeasy kit.
2. Genomic DNA is digested with Mse I (T▼TAA) to produce small fragments (200 - 1,000bp) while keeping CpG islands intact.
3. Fragmented DNA is heat denatured to produce single-stranded DNA, improving the efficiency of immunoprecipitation.
4. Antibodies against 5-methyl cytidine (available from Eurogentec, Abcam, and Diagenode) are used to immunoprecipitate methylated DNA fragments.
   Note: The enrichment of methylated DNA fragments increases as the number of methylated CpG per fragment increases.
5. The immune complexes are captured with agarose beads.
6. Complexes are washed to remove nonspecifically bound material.
7. Following elution of bound complexes, ethanol precipitation, and resuspension of MeDIP DNA, a small aliquot of DNA and PCR primers can be used to amplify a known methylated DNA region (positive) and a known unmethylated DNA region (negative).
8. After validating the enrichment of MeDIP DNA, fragments are amplified by either whole-genome or random priming methods (if necessary). Experimental and total DNA samples are denatured and labeled by Klenow fragment-catalyzed primer extension using Cy dye coupled 9mer primers.
9. The labeled experimental IP and total DNAs are co-hybridized to the array for 16 - 20 hours, washed, and scanned.
10. Array images are used for data extraction as pair files. Genomic feature format (GFF) files are then produced for visualization of scaled log₂-ratio data. The intensity ratio of immunoprecipitated to total DNA (not taken through immunoprecipitation steps) is plotted versus genomic position to identify regions where increased signal (i.e. methylated DNA fragment enrichment) is observed relative to the control sample.
11. P-value files (.gff) files are generated from the scaled log₂-ratio data, where each probe is tested for positive enrichment of DNA methylation against all probes on the array.
12. Peak files (.gff) identifying regions of DNA methylation are generated from the p-value files, and peaks can be mapped to the transcription start site of each gene.

Before embarking on a MeDIP experiment to analyze DNA methylation on a genome-wide scale, you need to optimize several experimental conditions:

• Enzymatic digestion of DNA – The size distribution of digested DNA fragments needs to be within a 200 - 1,000bp range for optimal methylation detection. Underdigestion of DNA with Mse I will result in inefficient immunoprecipitation.
• Antibody titration – The anti 5-methyl cytidine antibody will need to be titrated empirically to determine the optimal ratio of antibody to genomic DNA. Roche NimbleGen recommends using a 1:1 mass ratio of antibody to genomic DNA as a starting point for any titration experiments.
• MeDIP experiment validation (before and after array analysis) – To be truly confident that the anti 5-methyl cytidine antibody successfully enriched methylated DNA fragments, several different methods can be employed to validate your results. The table below compares validation methods. Click on the method to view a publication that utilizes that technology.

Using the proper experimental controls is also an important aspect of any genome-wide DNA methylation experiment. Listed below are common controls that researchers are using in their experiments:

• Nonspecific IgG antibodies – The most common type of negative control involves adding antibodies that do not recognize a specific epitope, for example pre-immune serum or IgG. A potential pitfall is that since the antibodies do not immunoprecipitate effectively, the nonspecific DNA yield is often extremely low. Hence, the hybridization tends to be much noisier and can result in many false positives due to amplification of trace amounts of nonspecific DNA. Alternatively, the primary antibody can be omitted (i.e. no antibody control). Be aware that testing this negative control requires a separate array of IgG versus input and that the hybridization of MeDIP versus IgG is not recommended.
• DNA methyltransferase (DNMT) deficient cell line – Several upstream applications, such as target deletion or siRNA, can be used to perturb the expression of the DNMT1 gene (maintains methylation after DNA synthesis). This effectively decreases DNA methylation and hence the amount of methylated DNA fragments that are immunoprecipitated. Alternatively, a cell line that deficient in DNMT could be used as a negative control.
• 5-aza-2’-deoxycytidine (5-aza-dC) treatment – The DNMT inhibitor 5-aza-dC can be used to treat cells to inhibit global DNA methylation throughout successive cell divisions. 5-aza-dC treatment is generally in the micromolar range for 3 - 7 days, where freshly prepared drug (in DMSO) is added each day.

A MeDIP experiment in which 40µg is used as the amount of starting material typically yields approximately 5 - 10%, or 2 - 4µg, MeDIP DNA, which is adequate for subsequent labeling reactions and hybridization to the array. However, in cases where DNA quantity is limited (e.g. tumor biopsy sample), MeDIP DNA (and input DNA) must be amplified using one of three methods described below to obtain adequate yields. Multiple cycles can be performed to achieve the desired amount of amplification product. Regardless of the method selected, Roche NimbleGen recommends the QIAquick PCR Purification Kit (Qiagen #28104) for purification of amplified DNA samples.

• Whole Genome Amplification (WGA) – Genomic MeDIP and input fragments are converted to PCR-amplifiable OmniPlex™ Library (Rubicon Genomics, Inc.) molecules flanked by universal priming sites. The library is then amplified by PCR using universal primers and a limited number of cycles. Roche NimbleGen recommends the WGA Kit (Sigma #WGA2-50RXN). The WGA protocol used by the Farnham lab at University of California - Davis can be accessed at genomecenter.ucdavis.edu/farnham/pdf/8-18-06WGA.pdf Roche NimbleGen recommends this method if amplification of your MeDIP sample is required. Comparisons of MeDIP amplified versus MeDIP unamplified samples showed that roughly 60 - 70% of peaks were present in amplified versus unamplified samples.
• Ligation Mediated-Polymerase Chain Reaction (LM-PCR) – WARNING: Linkers must be added prior to performing MeDIP reaction. Unidirectional linkers are ligated to blunt-ended MeDIP and input fragments and PCR amplified. Refer to www.chiponchip.org/protocol_itm3.html for a detailed LM-PCR protocol.
• T7 Amplification – MeDIP and input DNA fragments are tailed on the 3' end to generate a 20 - 40bp polyT tail with a terminal dideoxycytidine base. A T7 promoter -(A)18B anchored primer adaptor is annealed to the polyT tail of each template strand. During second-strand synthesis, Klenow removes the excess bases from the tail overhang (via its 3' - 5' exonuclease activity) and extends from the primer to produce the second strand. This process results in two double-stranded DNAs identical to the original template, except that each has a T7 promoter at the opposite end. The product of the second-strand synthesis reaction is used as template in an in vitro transcription reaction. These RNA samples are reverse transcribed to generate DNA probes for labeling and hybridization. Refer to www.biomedcentral.com/1471-2164/4/19 to view a publication that describes T7-based linear amplification of genomic DNA.

Quality Control of Amplification Methods

After amplifying MeDIP and input samples, DNA should be run on an Agilent Bioanalyzer or agarose gel to analyze the size and quality of DNA fragments. The size of MeDIP DNA fragments should be between 100 - 2,000bp with the majority of fragments between 200 - 1,000bp. Also, the DNA should appear as a smear, as opposed to several discrete bands. Samples exhibiting discrete banding will typically produce noisy DNA methylation data that are difficult to interpret.