Microarrays and DNA methylation profiling

 

Dr. Axel Schumacher © 2004-2007

 

 

 

The method described in detail below is based on the analysis of the enriched unmethylated DNA fraction, using a series of treatments with methylation-sensitive restriction enzymes, adaptor ligation, PCR amplification, and quantitative mapping of unmethylated DNA sequences using microarrays. The key advantages of this approach are the ability to investigate DNA methylation patterns using very small DNA amounts and also relatively high informativeness in comparison to the other restriction enzyme- based strategies for DNA methylation profiling (1).

 

 

1. Introduction

 

The key principles and some technical details of this method are described in our recent article (1). Briefly, genomic DNA (gDNA) is digested with the methylation-sensitive restriction enzymes such as HpaII and Hin6I. Whereas methylated restriction sites remain unaltered, the sites containing unmethylated CpGs are cleaved by the enzymes, and DNA fragments with 5’-CpG protruding ends are generated. In the next step, the double-stranded adapter CG-1 is ligated to the CpG-overhangs. At this point, it is expected that most of the relatively short (<1.5 kb) and amplifiable DNA fragments derive from the unmethylated DNA regions. Some ligation fragments, however, may still contain methylated cytosines. A large proportion of these fragments are eliminated by treatment with McrBC, thereby increasing the specificity of the enrichment of the unmethylated DNA fraction. McrBC cleaves DNA containing methylcytosine on one or both strands, recognizing two half-sites of the form (G/A)mC; these half-sites can be separated by up to 3 kb, but the optimal separation is 55–103 base pairs. The remaining pool of unmethylated DNA fragments is then enriched by aminoallyl-PCR amplification that uses primers complementary to the adapter CG-1. An important advantage of using protruding ends in the adapter–ligation step is that degraded gDNA fragments will not be ligated and amplified and, therefore, will not interfere with DNA methylation analysis (which is especially useful when analyzing tissues with relatively long post-mortem interval or paraffin-embedded samples). The enriched unmethylated DNA fractions are then labelled with fluorescent dyes and hybridized to microarrays. Several different types of microarrays can be used for epigenetic analysis, for example oligonucleotide arrays of individual genes or microarrays containing relatively larger DNA fragments of gene regulatory regions, such as CpG islands (1).

 

 

Fig 1: Schematic outline of the microarray-based method for identification of DNA methylation differences in genomic DNA. Samples are cleaved by methylation-sensitive restriction endonucleases, such as HpaII and Hin6I, ligated to the CpG-overhang specific adapter, and then cut by McrBC to eliminate residual methylated DNA fragments. The resulting unmethylated DNA fragments are then selectively enriched by adapter-specific aminoallyl-PCR, labelled, and hybridized to microarrays.

 

2. Materials

 

2.1 Adapter design

  1. Adapter storage buffer ST: Prepare 100 ml of 10 mM Tris, pH 8.5; 50 mM NaCl.
  2. Primer for preparation of universal-adapter CG-1: primer CG-1a: 5’-CGTGGAGACTGACTACCAGAT-3’ and primer CG-1b: 5’-AGTTACATCTGGTAGTCAGTCTCCA-3’.

 

2.2 Methylation-sensitive cleavage of DNA

 

1.        Restriction enzymes: 10 U/µl HpaII (Fermentas, Lithuania) and 10 U/µl Hin6I (Fermentas). (Note: HhaI is an isoschizomer of Hin6I, however, the HhaI restriction enzyme is not suitable for adapter-based methylation analyses, since this enzyme does not produce a CpG-overhang that can be ligated to the adapter).

2.   Spike DNA (optional; see Methods)

 

2.3 Ligation

  1. Prepare a 10 µM ATP stock-solution. Make aliquots and store at -20°C.
  2. T4 DNA Ligase (Fermentas, Lithuania)

 

2.4 McrBC digestion

  1. McrBC enzyme from New England BioLabs (NEB).
  2. Guanosine triphosphate (GTP; is delivered with the enzyme): make 10x aliquots, since GTP is highly unstable.

 

2.5 CpG-specific adapter amplification

  1. Labelling nucleotides [20 mM]: Thaw one vial aminoallyl (aa)-dUTP from Ambion, Austin, TX (50 µl solution 5-(3-aminoallyl)-2'-deoxyuridine 5'-triphosphate). Add 16.5 µl H2O and from 100 mM nucleotide stock solutions (Fermentas) add 16.6 µl dTTP, 41.6 µl dGTP, 41.6 µl dATP, and 41.6 µl dCTP. Mix and store at -20°C.
  2. Taq polymerase [5 U/µl] (NEB).
  3. 200 pmol/µl CG-1b-primer (see 2.1).

 

2.6 Purification of amplification product

  1. Microcon YM-50 columns from Millipore.

 

2.7 Labelling

  1. Sodium bicarbonate buffer (0.1 M, pH 9.0): Prepare 10 ml of a 0.1 M Na2CO3 solution. Add 0.42 g of NaHCO3 to another flask. Add 50 ml of water, and mix with a magnetic stirrer until the NaHCO3 dissolves. The pH should be around 8.5. Adjust the pH to 9.0 by adding ~ 2.5-3.0 ml of the 0.1 M Na2CO3 solution. Monitor the pH change carefully and do not exceed pH 9.3. Filter-sterilize the buffer, which should be prepared as fresh as possible. Aliquot and store at –20°C. Do not re-freeze the buffer. Use every aliquot only once.
  2. Prior to coupling, prepare the dyes (FluoroLink monofunctional dye, Cy3 and Cy5, Amersham Biosciences). The Cy-dyes come in packages, which contain 5 vials. Dissolve 1 vial of dye in 72 µl DMSO. Aliquot 4.8 µl in 15 light protected amber reaction tubes, dry immediately in a speedvac and store at –20°C protected from light. (H2O is also used to dissolve Cy-dyes, however the mono reactive ester is labile in water. Do not use Dioxane, which was suggested elsewhere as an alternative to DMSO. It appears that Cy5 is not fully soluble in Dioxane. DMSO on the other hand is hygroscopic and will absorb moisture from the air, which will react with the NHS ester of the dye and significantly reduces the coupling reaction efficiency. Therefore, keep the DMSO supplied in an amber screw-capped vial at -20°C, and let the vial warm to room temperature before opening to prevent condensation. While the dyes are usually good when first opened, they are sensitive to moisture with a short half-life in aqueous environments. Thus take care to store aliquots of them desiccated, if possible under vacuum).
  3. Quenching buffer: Prepare a 10 ml solution of 4 M hydroxylamine (Sigma). Make aliquots and store at -20°C. Hydroxylamine has top be handled with care, since it may explode when heated. It is an irritant to the respiratory tract, skin, eyes, and other mucous membranes. It may be absorbed through the skin, is harmful if swallowed, and is a possible mutagen.
  4. Purification of dyes: MiniElute PCR-purification kit, Qiagen, Hilden/GER.
  5. Prepare a 3 M NaOAc solution (pH 5.2).

 

2.9 Hybridization

  1. Prehybridization buffer: To 2 ml SlideHyb Glass Array Hybridization Buffer #2 from Ambion add 80 µl of 20 µg/µl yeast tRNA (Sigma) and 200 µg bovine serum albumin (BSA).
  2. Hybridization buffer: To 2 ml SlideHyb Buffer #2 from Ambion (other hybridization buffers may be used. The optimal buffer depends on several experimental variables such as type of glass slide used, the length and GC content of the attached nucleic acid targets, and the manufacturing protocol employed.) Add 50 µl of 20 µg/µl yeast tRNA (Sigma) and 200 µl Cot-1 DNA [1 µg/µl] from Roche Diagnostics. The COT fraction of human placental gDNA consists mainly of repetitive DNA elements. COT DNA is expensive and hence may be omitted if the microarray used does not contain larger amounts of repetitive sequences. If using cDNA arrays, 10-20 µg poly(dA)-poly(dT), which blocks hybridization to polyA tails of cDNA, may be added.
  3. Wash solution I (low stringency): prepare 1 litre of 2x SSC, 0.5 % SDS. Filter-sterilize the buffer. (Note: Wash buffer impurities can cause grainy background; SDS will fluoresce in the Cy5 channel while salt will fluoresce in the Cy3 channel. To eliminate this problem use high quality SDS and SSC for preparation of wash buffers and filter sterilize. Insufficient washing may also contribute to grainy background, therefore, increasing the wash times and/or slight agitation may help to decrease the background).
  4. Wash solution II (high stringency): prepare 1 litre 0.5x SSC, 0.5 % SDS. Filter-sterilize the buffer.
  5. Wash solution III: 1 litre Isopropanol (2-propanol). Use only high quality isopropanol (>99.8%). When it is less concentrated or exposed for a prolonged time to air (it absorbs water quickly), isopropanol causes streaks or residue on the array surface.
  6. Cover slips: Hybri-Slips (Sigma). Hybri-slips have hydrophobic surfaces, which guarantee a constant probe concentration and, unlike glass cover slips, do not adsorb the probes.

 

 

3. Methods

 

Here we present a complete protocol of DNA methylation profiling using the unmethylated fraction of the genome. This protocol works with most microarray types, as exemplified here with CpG island- and oligonucleotide microarrays. A detailed description of how specific ‘epigenetic’ microarrays are designed and how the slides are processed can be found in Schumacher et al., 2006 (1). DNA samples can either be interrogated as pairs (tester and control) on one array, as described in this protocol, or hybridized independently with only one fluorescent dye (e.g. for Affymetrix tiling arrays).

 

3.1 Adapter design

 

In the CpG-specific adapter design, the following aspects must be taken into account:

 

  1. It must contain a CpG-overhang, which fits to the restriction site of the enzymes used (Fig. 2a).
  2. The two nucleotides next to the CpG-overhang have to be different from the nucleotides within the recognition sequence of all the enzymes used in the restriction-digests. This ensures that, after ligation of the adapters, the old restriction-site is disrupted and the sequence cannot be re-cut by the enzymes.
  3. The melting-temperature (Tm) should be about equal throughout the adapter and decreasing at the 3’-end opposite to the CpG-overhang.
  4. The core-sequence of the adapter should be non-homologous to any sequence in the gDNA. Additionally, it is advisable to avoid palindromic sequences, which can lead to mispriming.
  5. If you want to cut the adapter from the target-fragments, include a very specific, new recognition site within the adapters core-sequence (close to CpG-overhang), which does not cut frequently in the genome you analyze. Use a rare 8-mer (e.g. SdaI (5’-CC↑TGCA↓GG) for sticky ends or MssI (GTTT↓AAAC) for blunt ends.
  6. Use a specific, long antisense-overhang, which prevents the forming of tandem-repeats and random blunt-end ligation to the genome (e.g. through degraded DNA).
  7. To avoid improper annealing between the two primers, choose a non-5’-complementary end (Fig. 2a).

 

 

Fig. 2: Preparation of the CpG-specific adapter. a: CG1-adapter components. b: Annealing of the CG-primers produces a double-stranded adapter that can be identified by electrophoresis as a strong band (200 pmol loaded; 2.5 % agarose gel).

 

3.1.1 Adapter preparation

1.        Dissolve the non-phosphorylated ssDNA oligonucleotides in ST buffer with a concentration of approximately 800 pmol. The presence of some salt is necessary for the oligos to hybridize into a double stranded (ds)-adapter. Keep some of the CG-1b oligo, as it will be later used as a primer to amplify the unmethylated DNA fraction. Measurethe concentration in a spectrophotometer and adjust the concentration to 400 pmol. (Note: Quantify in spectrophotometer. Re member the OD260 = 1 of single stranded DNA is equivalent to 33 µg/ml not 50 µg/ml! This relationship, however, can be inaccurate for short fragments of DNA, such as oligonucleotides. Base composition and even linear sequence will affect optical absorbance, hence the precise value of the OD to mass relationship is unique for each oligo. For example, 1.0 OD260 of CCCCCCCCCCCC (homopolymeric deoxycytidine) equals 39 micrograms while 1.0 OD260 of AAAAAAAAAAAA (homopolymeric deoxyadenosine) equals only 20 micrograms. Do not believe the amount of primer as indicated by the manufacqturer. Perform the appropriate calculation (800 pmol primer = 0.0008 x MW of oligo)).

  1. Mix together equal molar amounts of each complementary oligonucleotide to a final concentration of 200 pmol, and heat 5 min in a thermalcycler at 80°C. Cool down with 1°C/min until the mixture reaches room temperature (RT). The longer the cooling period, the lower is the risk of hairpin structures.
  2. Check a small amount of the ds-Adapter on a 2.5% Agarose-gel (see Fig. 2b), which should generate a strong band. Run the single stranded primers next to the adapter as a control.
  3. Store the adapters at –20°C.

 

3.2 Methylation-sensitive cleavage of DNA

 

Over 260 different methylation-sensitive restriction enzymes (MSRE’s; incl. isoschizomers) are now available, however, not all enzymes are useful and informative for DNA methylation profiling. Informative MSREs are defined by the number of cleavage fragments that can be ligated to adapters and efficiently amplified, and are not lost during column-purification steps (3). Some enzymes, although they cut frequently in the genome, produce fewer informative fragments compared to enzymes which do not cut as frequently. For example, the non-palindromic AciI (5’-CCGC-3’) recognizes more than twice as many CpG sites in CpG island regions when compared to HpaII, but on the other hand, produces fewer fragments in the size-range that can be detected by PCR or amplified fragment length polymorphism (AFLP) methods (Table 1).

 

MSRE

Cut site 5’-3’

~% of CpGs in CpG islands*

~% of CpGs in non-CpG islands*

Fragments/kb in CpG islands*

Fragments/kb in non-CpG islands*

AciI (SsiI)

CCGC, GCGG

30.60

17.36

3.23

1.79

Hin6IΔ (HinP1I)

GCGC

14.40

5.05

3.98

0.61

HpaII (BsiSI)

CCGG

11.70

9.33

3.98

1.18

Hin1I (BsaHI)

GRCGYC

2.6

0.9

1.92

0.11

HpyCH4IV (TaiI)

ACGT

1.66

6.73

1.24

0.97

Bsp119I (AsuII)

TTCGAA

0.11

0.13

0.11

<0.02

Bsu15I (ClaII)

ATCGAT

<0.05

0.39

<0.02

0.02

 

Table 1: Methylation-sensitive restriction enzymes. All the above MSRE’s produce sticky ends that can be ligated to the CG-1 adapter for high throughput microarray-based DNA methylation profiling. Other MSRE’s that fit to the CG-1 adapter are Psp1406I, XmiI, BstBI or NarI, however these enzymes have very few restriction sites in the human genome or are too expensive to be used in the required amount. Asterisk (*) indicates the number of 75 bp - 2 kb long, i.e. ‘informative’, fragments, derived from CpG island and non-CpG island sequences in the human genome (3). R = A/G; Y = C/T.

 

Different requirements for the enzyme reduces the list of potentially useful and informative MSRE’s to about 17, which would cover up to 85% of all CpG island CpG dinucleotides, but less than 50% of all CpG dinucleotides in other genomic regions (3). Here, we demonstrate the technology with a double-digestion using HpaII and Hin6I enzymes (Note: To gain the most out of restriction analyses, it is crucial to choose the right enzyme combination for the targets to be interrogated. For example, some MSRE’s cut relative frequently in CpG islands, but rarely recognize a sequence outside of a CpG island region, as is the case for Hin6I (5’-GCGC-3’) or Bsp143II (5’-PuGCGCPy-3’). In contrast, enzymes such as HpyCH4IV (5’-ACGT-3’) cut predominantly outside of CpG island sequences and are less useful in the interrogation of CpG islands, for instance in CpG island-microarray based studies (3). Several other methods rely on the specific methylation-sensitive cleavage of the rare cutter NotI (5’-GCGGCCGC-3’), for example RLGS and AFLP methods and a couple of microarray approaches (6, 7). However, NotI-sites are not well represented in the genome and will only provide a very rough overview of methylation patterns. Hence, it is not advisable to include NotI in genome-wide analyses of e.g. complex diseases).

Before the test DNA can be processed, it is advantageous to add “spiking” DNA to the gDNA samples. The spiking DNA consists of ‘alien’ (exogenous) DNA, which allows monitoring of each step of the experiment (Note: Additionally, spike DNA can be used to test the methylation-sensitive cleavage reaction, ligation efficiency, McrBC digestion and PCR reaction. Spiking DNA is added to the reactions with concentrations equivalent to the template concentration (1 genome equivalent, GE, assuming 3x109 bp in the human genome). For example, 16.2 pg of λ plasmid is added to 1 µg of template DNA. Make 12x GE stock solutions in 1 mM EDTA; aliquot into small volumes and store at -20°C. To prepare a McrBC digestion control, digest pUC57 with HpaII, ligate the CG-adapter and premethylate with SssI methylase. McrBC controls are added to the McrBC cleavage step as premethylated plasmid. Other suitable controls are ФX174 to test for PCR bias or pBR322 to test the ligation efficiency (1). ФX174 DNA has to be digested separately with HpaII/Hin6I and then ligated to the CG-adapter. Complementary spiking oligonucleotides have to be representative sequences of HpaII and Hin6I cleavage fragments).

 

As with any microarray experiment, controls are crucial for monitoring experimental variability. Suitable spiking DNA consists of sequences that are frequently spotted on commercially available microarrays, such as Arabidopsis, RNA spikes or artificial sequences. (Note: SNPs within the recognition sites for HpaII and Hin6I may simulate epigenetic differences (8). In order to exclude the impact of DNA sequence variation it is suggested to check the available SNP databases and identify the DNA sequence variation within the restriction sites of the used enzymes. From CpG island microarray studies, the estimate is that 10 to 30% of methylation variation detected in between individuals could, in fact, be due to DNA sequence variation (1). For comparison, in a study for the Human Epigenome Project (HEP), interrogation of 3,273 unique CpG sites on chromosome 6 revealed that 101 CpGs overlapped with known SNPs (3 %) (9). Another way to differentiate DNA sequence effects from genuine epigenetic differences consists of performing an identical microarray experiment on the same DNA sample that has been stripped of all methylated cytosines