Professional Documents
Culture Documents
Methods
Abstract
DNA methylation contributes to the control of gene expression and plays an essential role in cellular physiology.
Well-defined patterns of DNA methylation are established and fixed during embryonic development, and changes
in these patterns may be a contributing factor in developmental disorders, cancer and aging. Not least the possibility
of using DNA methylation as a marker for disease has created a strong need for techniques to detect and measure
DNA methylation. Different techniques provide information on DNA methylation at different levels, spanning
from genome-wide methylation content to methylation of single residues in specific genes. The limitations of
individual techniques strongly affect interpretation of data. In this review, we discuss some general themes in DNA
methylation analysis and outline the basic principles of current key techniques. We discuss the advantages and
disadvantages of these techniques, including potential artifacts and pitfalls, and suggest some overall guidelines
that may be instructive for a rational choice of methodology.
Abbreviations: COBRA – combined bisulfite restriction analysis; DHPLC – denaturing high-performance liquid
chromatography; DMH – differential methylation hybridization; HPCE – high-performance capillary electro-
phoresis; HPLC – high-performance liquid chromatography; MCA-RDA – methylated CpG island amplification-
representational difference analysis; MS-AP-PCR – methylation-sensitive arbitrarily primed PCR; MS-DGGE
– methylation-specific denaturing gradient gel electrophoresis; MS-DHPLC – methylation-specific denaturing
high-performance liquid chromatography; MS-MCA – methylation-specific melting curve analysis; Ms-SnuPE
– methylation-sensitive single nucleotide primer extension; MS-SSCA – methylation-specific single-strand
conformation analysis; MSO – methylation-specific microarray; MSRE – methylation-sensitive restriction
endonuclease; MSRF – methylation-sensitive restriction fingerprinting; PCR – polymerase chain reaction; RLGS
– restriction landmark genomic scanning; SAM – S-adenosylmethionine; TLC – thin-layer chromatography
Methylation is the predominant ‘epigenetic’ found in ICF syndrome, a rare disorder characterized
modification of DNA in mammalian genomes. The by immunodeficiency, facial anomalies, and hypo-
term epigenetic is used to denote that methylation methylation of the satellite DNA in juxtacentromeric
can modify the information content of DNA without heterochromatin of chromosomes 1 and 16 (Okano et
changing the primary nucleotide sequence. Hence, al. 1999).
DNA methylation does not alter the structure or In normal, fully differentiated cells, DNA methyla-
function of a gene, but provides information as to tion patterns are replicated with high fidelity during
where and when the gene should be expressed. There mitosis. However, a large body of data has shown that
is now substantial evidence that, in general, the these patterns can become altered and cause cellular
level of methylation within a gene’s promoter CpG functional abnormalities in the course of human
island correlates with its transcriptional silencing. disease and aging. The pathologic consequences of
Altering the binding of some transcription factors is altered DNA methylation have been most extensively
one possible mechanism by which 5-methylcytosine studied in cancer. Compared with normal cells, cancer
controls gene expression. An alternative model for cells frequently show genome-wide hypomethylation,
methylation-based gene silencing implicates a group hypermethylation of tumor suppressor genes, and
of proteins and protein complexes that specifically loss of genomic imprinting (Baylin et al. 1998).
bind to methyl-CpG and recruit histone deacetylates Recent studies have demonstrated increased rates of
to the sites of methylation (Wade 2001). The removal spontaneous mutations, deletions and rearrangements
of acetyl groups from the histones of the nucleosomal in cells lacking DNMT1, the DNA methyltransferase
core converts the open, transcriptionally competent responsible for maintaining DNA methylation patterns
chromatin structure into a closed structure that is after DNA replication (Chen et al. 1998). Accord-
inaccessible to the transcriptional machinery. A third ingly, loss of genomic methylation may represent an
possible mechanism of methylation-dependent gene early event in multistep carcinogenesis by providing
silencing involves DNA methyltransferases, which the premalignant cell with a mutator phenotype. In
have transcriptional repressor function and can recruit contrast, hypermethylation of CpG islands may cause
histone deacetylates to methylated DNA (Robertson gene silencing and can be functionally equivalent to
2002). mutation and deletion in activating tumor suppressor
In humans, methylation-dependent gene silencing genes (Jones and Laird 1999). Intriguingly, evidence is
has been implicated in inactivation of the X chromo- emerging that many of the changes observed in cancer
some in females, inactivation of the silent allele at cells are also present in aging cells in normal tissues. It
imprinted loci, and silencing of intragenomic para- has been proposed that these age-related methylation
sitic DNA elements (Costello and Plass 2001). Several changes are precursors for the aberrant methylation
lines of research have shown that the establishment patterns in cancer cells and thus may contribute to
and maintenance of well-defined methylation patterns the age-related increase in cancer risk (Ahuja and Issa
are required for normal embryonic development. First, 2000).
elimination of any of the three known DNA methyl- The central role of DNA methylation in main-
transferases, Dnmt1, Dnmt3a and Dnmt3b, is lethal taining cellular function and the recognition that
in mice (Okano et al. 1999; Li et al. 1992). Second, changes in DNA methylation patterns may have broad
a number of human disorders are associated with implications for human health have created a strong
genetic abnormalities involving chromosomal regions need for techniques to reliably detect and measure
that are subject to genomic imprinting, includ- methylation in DNA from diverse sources. In this
ing Beckwith-Wiedemann syndrome, Prader-Willi review, we describe different ways of analyzing DNA
syndrome, and Angelman syndrome (Paulsen and methylation and outline the basic principles of current
Ferguson-Smith 2001). Third, two human syndromes techniques (listed in Table 1), with emphasis on tech-
have been identified that are caused by inherited niques for gene-specific analysis of DNA methylation.
defects in genes affecting the DNA methylation Each of these techniques is limited in the information
machinery. Rett syndrome, an X-linked, neuropsychi- content of the data it generates, which strongly affects
atric disorder with onset in early childhood, is caused interpretation of data.
by loss-of-function mutations in the gene encoding
methyl-CpG binding protein MeCP2 (Amir et al.
1999), and mutations in the DNMT3B gene have been
235
Table 1. Techniques described in this review, grouped according to the dimensions in which they can resolve DNA methylation.
Quantitative analysis Ms-SnuPE Bisulfite treatment Yes Gonzalgo and Jones 1997
of single CpGs MethyLight Bisulfite treatment Yes Eads et al. 2000
COBRA Bisulfite treatment Yes Xiong and Laird 1997
Direct bisulfite genomic sequencing Bisulfite treatment Yes Frommer et al. 1992
Analysis of allelic Direct bisulfite genomic sequencing Bisulfite treatment Yes Frommer et al. 1992
methylation Cloned bisulfite genomic sequencing Bisulfite treatment Yes Frommer et al. 1992
heterogeneity MethyLight Bisulfite treatment Yes Eads et al. 2000
MS-MCA Bisulfite treatment Yes Worm et al. 2001
MS-DGGE Bisulfite treatment Yes Aggerholm et al. 1999
MS-SSCA Bisulfite treatment Yes Maekawa et al. 1999
MS-DHPLC Bisulfite treatment Yes Baumer et al. 2001
MSO Bisulfite treatment Yes Gitan et al. 2002;
Adorjan et al. 2002
Cytosine and 5-methylcytosine can be identified and Despite its simplicity, the SssI acceptance assay is
quantified by including external standards of bases prone to a number of technical problems. First of all,
and monitoring UV absorbance at 254 nm. To avoid the day-to-day variation is considerable, primarily due
interference from RNA contamination, it is usually to the instability of both SAM and SssI methyltrans-
necessary to remove RNA from the nucleic acid ferase. Another problem may be associated with the
preparations by enzymatic hydrolysis using a combi- difficulty of dissolving genomic DNA to homogeneity,
nation of ribonuclease A and ribonuclease T1. which may lead to inaccurate DNA concentration
Fraga et al. (2002) used high-performance capil- measurements. A useful way of obtaining a more
lary electrophoresis (HPCE) to resolve DNA hydroly- homogeneous solution is to digest the DNA with a
sates and determine the content of 5-methylcytosine. restriction enzyme (e.g., EcoRI) that does not contain
Compared with HPLC, HPCE is more simple, rapid a CpG sequence in its recognition site (Oakeley 1999).
and cost-effective. The sensitivity of HPLC- or HPCE-
based determination of genome-wide DNA methyla- Chloroacetaldehyde assay
tion is relatively high (<5 µg), but may still be too low
for routine processing of tissue specimens. Signifi- Oakeley et al. (1999) described a method for studying
changes in genome-wide levels of DNA methylation
cant improvements in sensitivity have been achieved
by fluorescent labeling using chloroacetaldehyde. The
by combining HPLC separation of DNA hydrolysates
with mass spectrometry detection (del Gaudio et al. DNA is first treated with sodium bisulfite, which
converts cytosine into uracil, but leaves 5-methyl-
1997), which also may be useful for HPCE-based
cytosine unchanged (Frommer et al. 1992; described
approaches.
In laboratories where equipment for high-resolu- in detail below). The bisulfite-treated DNA is then
depurinated under acidic conditions, and the purines
tion separation of hydrolyzed DNA is not available,
thin-layer chromatography (TLC) may be a useful are removed by silver nitrate precipitation or by
alternative (Schmitt et al. 1997). In brief, DNA is cut column chromatography. Subsequent incubation of the
sample with chloroacetaldehyde yields the intensely
at all CCGG sites using the restriction enzyme MspI,
and the internal cytosine in these sites is then radioac- fluorescent ethenocytosine derivative of 5-methyl-
cytosine, which can be quantified using a fluorimeter
tively labeled using γ -labeled ATP and polynucleotide
and used as a direct measure of the level of 5-methyl-
kinase. Following hydrolysis of the DNA to mono-
nucleotides using nuclease P1, radioactively labeled cytosine in the genome. The chloroacetaldehyde assay
is an attractive alternative to the SssI acceptance assay.
cytosine monophosphate and 5-methylcytosine mono-
phosphate are separated on TLC plates and identified Not least the use of stable and inexpensive chem-
and quantified as individual spots using a phosphor- icals and the avoidance of radioactivity are major
advantages.
imager. Because only cytosines in CCGG sites are
probed, this approach only provides a rough estimate
Immunochemical analysis
of the ratio between 5-methylcytosine and cytosine.
Monoclonal antibodies raised against 5-methylcyto-
SssI acceptance assay sine provide a powerful tool for studying global DNA
methylation. To determine genome-wide methylation
The bacterial enzyme SssI DNA methyltransferase, content, genomic DNA is denatured and immobil-
which uses S-adenosylmethionine (SAM) as methyl ized on a nitrocellulose membrane and then incubated
donor, effectively methylates unmethylated cytosine in with an anti-5-methylcytosine antibody. The intensity
CpG sites. This enzyme can be used in a simple assay of the staining, which is proportional to the degree
to determine global CpG methylation: genomic DNA of methylation, may be determined by subsequent
is incubated with SssI methyltransferase in the pres- incubation with a fluorescein-conjugated secondary
ence of tritium-labeled SAM and then immobilized on antibody followed by fluorescence scanning (Oakeley
nitrocellulose paper (Wu et al. 1993). The amount of et al. 1997).
radioactive label incorporated can be determined using Immunofluorescence microscopy has been used to
a scintillation counter and is a direct measure of the map patterns of DNA methylation in cellular chromo-
number of unmethylated CpGs in the total pool of somes in individual cells. Chromosomal regions
DNA. with particularly high densities of 5-methylcytosine,
such as juxtacentromeric regions, emit the strongest
237
Figure 1. Principles of methylation analysis using MSREs (A) or bisulfite treatment (B). A: Genomic DNA is digested with a restriction
endonuclease that will not cut if the cytosine within the cleavage site is methylated. B: Treatment of genomic DNA with sodium bisulfite
deaminates unmethylated cytosine residues to uracil under conditions where 5-methylcytosine remains largely unaffected.
the bisulfite reaction (Paulin et al. 1998). Several the standard overnight incubation. This unintended
independent groups have confirmed the beneficial conversion is supposed to occur randomly and may
effect of bisulfite treatment in agarose beads, whereas only become evident when individual DNA molecules
the possible effects of the two latter modifications are analyzed. Second, treatment with bisulfite may
remain controversial. Some genomic regions seem lead to DNA degradation due to partial acid-catalyzed
to be particularly resistant to bisulfite conversion. depurination. Long incubations can result in extensive
Warnecke et al. (2002) showed that such artifacts damage to template sequences, which subsequently
might be less significant when DNA has been isolated may be difficult to amplify (Raizis et al. 1995). Several
using a protocol involving extended proteinase K groups have reported that a 4-h incubation with
treatment to remove residual protein. bisulfite may be sufficient for complete conversion,
Other potential artifacts of the sodium bisulfite provided that the DNA is completely denatured. Third,
reaction should be considered when designing exper- when PCR is used to amplify bisulfite-treated DNA,
iments and interpreting the results. First, 2–5% of the same primer pair may not amplify the methylated
5-methylcytosines may be converted to thymine in and unmethylated sequences with equal efficiency. At
239
present, there is no protocol that can be used to avoid Several precautions must be taken to exclude false
such biased PCR amplification, which should there- positives resulting from an incomplete digest, in
fore be excluded experimentally by analyzing samples particular when the digestion product is subjected to
containing both methylated and unmethylated DNA in PCR amplification. Because reduced accessibility of
known ratios (Warnecke et al. 1997). DNA to cleavage is usually caused by impurities in the
DNA, the best way to ensure complete DNA cleavage
is careful purification of the DNA using proteinase
Techniques for gene-specific methylation analysis K. Although parallel cleavage reactions using the
two isoschizomers may serve as a good control for
A wide variety of methods have been developed to complete cleavage, it cannot be excluded that the
assess the methylation status of specific genes in different enzymes are affected differently by DNA
a broad range of applications. While MSREs are impurities. Even the inclusion of a control DNA frag-
still widely used in combination with either Southern ment with the same cleavage site as the target sequence
analysis or PCR, most current methods rely on in the same reaction is not always sufficient to exclude
treatment of genomic DNA with bisulfite prior to incomplete digestion. In general, because methylated
PCR amplification. Several of these methods were sites may not be distinguished from incomplete diges-
originally developed for detection of single-nucleotide tion, MSRE methods should not be used for detection
polymorphisms (SNPs) and disease-causing mutations of low-abundant methylated molecules.
in genomic DNA, but later have been adapted for
detection of bisulfite-induced sequence differences
between methylated and unmethylated alleles. Methylation-specific PCR
Figure 2. Schematic outline of methylation-specific PCR. Genomic DNA is treated with bisulfite and used as template in a PCR containing
primers designed to discriminate between methylated and unmethylated alleles.
resulting PCR product is purified following separa- analysis of a number of individual clones. While
tion in agarose gels and is then hybridized with a the experimental procedure is laborious and time-
primer flanking the CpG site(s) being investigated. consuming, interpretation of data is simple: the pres-
To prevent unwanted bias, oligonucleotides used in ence of a band in the C-track indicates the presence of
the primer extension reactions should not anneal to a methylated cytosine in the corresponding genomic
sequences that originally contained one or more CpG DNA (Figure 4A). Alternatively, the PCR product may
sites, which may not always be possible in CpG- be sequenced directly to provide an average across
dense regions. After the primer has hybridized to its all molecules in the sample. The presence of bands
target sequence, a single-nucleotide extension reac- in both the C- and T-tracks at a specific position in
tion is performed in the presence of Taq polymerase the sequence indicates partial methylation. However,
and either [α-32 P]dCTP or [α-32 P]dTTP. Quantifica- as pointed out by Oakeley (1999), care should be
tion of the ratio of methylated versus unmethylated taken when using the relative intensities of these
cytosine (C vs. T) at the original CpG sites can be bands to estimate the proportion of 5-methylcytosine
determined by denaturing polyacrylamide gel electro- to cytosine in the genome. The major disadvantages
phoresis and phosphorimager analysis. Alternatively, of cloning and sequencing are that a high number
Ms-SnuPE reaction products can be transferred to of clones have to be sequenced for reliable results,
nylon membranes for quantification. and that artifacts relating to PCR infidelity, incom-
Although fraught with the technical problems of plete bisulfite conversion, or bisulfite conversion of 5-
the bisulfite method (incomplete conversion, PCR methylcytosine to thymine become significant. Direct
bias, etc.) and some experimental pitfalls (Oakeley sequence analysis is less prone to artifacts but does not
1999), Ms-SnuPE has several advantages over other provide information about the methylation patterns of
methods for determining methylation levels: it is individual alleles.
quantitative; it avoids the use of restriction enzymes;
and small amounts of DNA can be analyzed. Further- MethyLight
more, the methylation status of several CpG sites
can be assessed simultaneously by including multiple MethyLight (Eads et al. 2000) is a high-throughput
oligonucleotides in a single primer extension reaction. DNA methylation analysis technique that relies on
El Maarri et al. (2002) described a non-radioactive fluorescence-based real-time PCR. After bisulfite
variant of Ms-SnuPE, in which ion-pair reverse phase conversion, the DNA is amplified by PCR in a reaction
HPLC is used to discriminate and quantify the containing three oligonucleotides; a probe flanked by
extended primer products on the basis of different a forward primer and a reverse primer, all of which
masses and hydrophobicities. After annealing to are specific for the target. The probe is dual-labeled
bisulfite-treated DNA, the primers are extended in with a 5 fluorescent reporter dye and a 3 quencher
the presence of both ddCTP and ddTTP. In order dye. When the probe is intact, the quencher dye
to quantify the incorporation of either ddTTP or absorbs the fluorescence of the reporter dye due to
ddCTP, the extension products are loaded directly their proximity. After annealing to the PCR product,
onto a denaturing high performance liquid chroma- the probe is cleaved by the 5 to 3 exonuclease
tography (DHPLC) column that allows separation of activity of Taq DNA polymerase. This cleavage will
non-extended, ddCTP-extended, and ddTTP-extended release the reporter from the quencher, resulting in an
primers. increased fluorescence signal, which is proportional to
the amount of PCR product generated. The fluores-
Bisulfite genomic sequencing cent signal is measured at every cycle, and the initial
template methylation level can be derived from the
Sequence analysis of PCR products generated from cycle number at which the fluorescent signal crosses
bisulfite treated DNA (Frommer et al. 1992) is consid- the detection threshold in the exponential phase of the
ered the gold standard for gene-specific methylation PCR reaction.
analysis, first of all because it provides information The MethyLight technique has several important
about the methylation status of every cytosine residue advantages. First, it is a highly sensitive assay, capable
within the target sequence. For highest resolution of of detecting methylated alleles in the presence of a
methylation patterns, the PCR products are cloned into 10,000-fold excess of unmethylated alleles, which
an appropriate plasmid vector followed by sequence allows identification of aberrant DNA methylation
242
Figure 3. Schematic outline of Ms-SnuPE. Genomic DNA is treated with bisulfite followed by PCR amplification using primers that do
not discriminate between methylated and unmethylated alleles. The PCR product is then incubated with a primer that anneals to sequences
upstream and terminates immediately 5 to the site being monitored. After primer annealing, a single-nucleotide extension reaction is performed
in the presence of Taq polymerase and [α-32 P]dTTP or [α-32 P]dCTP. The radiolabeled reaction products are typically electrophoresed in
polyacrylamide gels and quantified by phosphorimage analysis.
Figure 4. Methods to resolve sequence differences in PCR products generated from bisulfite-treated DNA using primers that do not discriminate
between methylated (M) and unmethylated (U) alleles.
is treated with sodium bisulfite, and the region of alized using high-sensitivity fluorescent gel stains, e.g.
interest is amplified using primers that do not discrimi- SYBR Green II and SYBR Gold. Ethidium bromide
nate between methylated and unmethylated alleles. binds poorly to single-stranded DNA and is usually
The bisulfite-induced sequence differences between not recommended for staining SSCP gels, in particular
methylated and unmethylated alleles can then be when denaturation of the PCR product is incomplete
resolved by denaturation of the PCR product into and/or re-annealing of the strands is extensive. Effi-
single strands followed by electrophoresis in a poly- cient denaturation and prevention of re-annealing both
acrylamide gel under non-denaturing conditions may be achieved by denaturing the PCR product in
(Figure 4B). After electrophoresis, bands can be visu- a loading buffer containing 95% formamide and 10
244
mM sodium hydroxide at 95 ◦ C for 5 min, and then staining. For semi-quantitative analysis, the digested
immediately placing the sample on ice. PCR products may be denatured, separated by
One of the great advantages of SSCA is that denaturing polyacrylamide gel electrophoresis, and
alleles with different methylation patterns are physi- then transferred to a membrane by electroblotting
cally separated in the gel and may be extracted and hybridized with radioactively labeled oligonucle-
for subsequent sequence analysis. This provides a otides. Quantification can be performed using a
simple means for identifying individual epigeno- phosphorimager, and the percentage of fully methyl-
types in a pool of DNA without time-consuming ated sites in a genomic DNA sample can be calculated
cloning analysis. The limitation of MS-SSCA is that as the ratio between the cleaved PCR product and the
probably not all methylation changes are resolved. total amount of PCR product. For optimal restriction
First, conventional SSCA analysis has a specificity enzyme digestion, purification of the PCR products
of <70%. Second, in unmethylated bisulfite-treated may be necessary.
DNA, the number of bases is reduced from 4 to 3, COBRA analysis allows the determination of DNA
which significantly reduces the ability of the single methylation levels at specific loci in small amounts
stranded molecules to form secondary structures. of genomic DNA, provided that the DNA is not too
Conversion of non-CpG cytosines may also reduce extensively degraded during bisulfite treatment. Even
or entirely abolish internal base pairing of methyl- though COBRA is based on the use of restriction
ated sequences. Until more systematic studies have enzymes, it is not nearly as prone to false positives
clarified this issue, MS-SSCA may not be considered as MSRE-PCR because it does not rely on the absence
a generally applicable method for resolving gene- of cleavage to detect methylated sites. As with most
specific methylation changes. other bisulfite methods, incomplete bisulfite conver-
sion of the DNA and PCR bias will severely affect
Combined bisulfite restriction analysis (COBRA) the reliability of the COBRA technique. An additional
limitation is that it is limited to restriction sites and
COBRA (Xiong and Laird 1997) is a powerful tech- thus cannot be used to analyze the methylation status
nique that exploits differences in restriction enzyme of all possible CpG sites.
recognition sites between methylated and unmethyl-
ated DNA after treatment with bisulfite (Figure 4C). Methylation-specific denaturing gradient gel
Bisulfite conversion can lead to both methylation- electrophoresis (MS-DGGE)
dependent creation of new restriction enzyme sites and
methylation-dependent retention of pre-existing sites. Denaturing gradient gel electrophoresis (DGGE) is
After bisulfite treatment, the sequence of interest is widely used in combination with PCR as a tech-
amplified by PCR using primers that do not discrimi- nique to detect single-base mutations in genomic
nate between methylated and unmethylated alleles. DNA (Abrams and Stanton 1992). Separation of DNA
The resulting PCR products will represent a mixed molecules by DGGE is based on the electrophoretic
population of methylated and unmethylated alleles. mobility of a partially denatured (melted) molecule in
The fraction of molecules that has a newly created polyacrylamide gels, which is decreased significantly
or retained restriction site containing a CpG directly compared with that of the corresponding helical form.
reflects the methylation level at that site in the original In a gel containing an increasing gradient of chem-
pool of genomic DNA. The PCR products are then ical denaturants, the migration of the molecules will
digested with a restriction enzyme that has cytosines almost stop at the position at which the transition from
only within CpG sites in its recognition sequence. For the helical to the partially melted form occurs. Single-
example, TaqI cleaves the sequence TCGA, which will base substitutions and other sequence alterations can
be retained in bisulfite-treated DNA if the cytosine is significantly alter the melting temperature, causing
methylated, but will be transformed to TTGA if the the variant molecules to stop migrating at different
cytosine is unmethylated. By appropriate choice of positions in the denaturing gradient gel.
restriction site and enzyme, a positive result (cleavage) MS-DGGE (Aggerholm et al. 1999) resolves dif-
can be varied to indicate either non-methylation or ferentially methylated DNA molecules on the basis of
methylation. differences in melting temperature after bisulfite treat-
The digestion products may be analyzed by ment (Figure 4D). After non-discriminatory amplific-
agarose gel electrophoresis and ethidium bromide ation of methylated and unmethylated sequences, the
245
PCR product contains a mixture of molecules with to double-stranded DNA, e.g. SYBR Green I. The
different contents of G:C base pairs, reflecting the melting properties of the PCR product can be
distribution of methylated CpGs in the genomic DNA examined immediately after amplification by slowly
template. The PCR products are then resolved by elevating the temperature under continuous fluores-
electrophoresis in a polyacrylamide gel containing a cence monitoring. An abrupt decrease in fluorescence
gradient of urea and formamide, which is kept at a reflects the cooperative melting of a single domain in
constant temperature of 54–60 ◦ C. Because a G:C base the PCR product and can be identified if the melting
pair contains three hydrogen bonds and is more stable curves are converted into melting peaks by plotting the
than an A:T base pair with only two hydrogen bonds, negative derivative of fluorescence over temperature
the migration distance of the PCR product increases versus temperature. The 5-methylcytosine content of
with an increase in methylation density. After electro- the genomic DNA template determines the absolute
phoresis, the gel is stained with ethidium bromide or a position of the PCR product’s melting peak (Figure
high-resolution fluorescent gel stain. 4E).
An attractive feature of MS-DGGE is that The major advantages of MS-MCA are that it
sequences with different degrees of methylation will requires only standard, inexpensive PCR reagents; it
be retarded at different positions in the gel. This resolves heterogeneous methylation both in pools of
provides a detailed visual display of the methylation DNA and in cloned PCR products; and it detects
status of the total pool of DNA molecules. This methylated and unmethylated alleles in the same
is particularly useful for resolving heterogeneous reaction. PCR amplification and subsequent melting
methylation (Aggerholm et al. 1999). Furthermore, analysis may even be integrated using a thermal
like with MS-SSCA, different epigenetic clonotypes cycler coupled with a fluorometer (LightCycler or
will be revealed as distinct bands in the gel, which may equivalent). The major advantage of this integrated
be recovered for subsequent sequence analysis. The procedure, compared with existing gel-based methods,
drawback of this method is the need for careful design is the avoidance of post-PCR experimental steps,
of PCR primers to ensure optimal melting behavior of which also reduces the risk of PCR contamination.
the amplified product. In general, it is recommended The major limitation of this method is that no informa-
that the primers be designed to generate a PCR tion is provided on the methylation status of individual
product that, for both unmethylated and methylated cytosines, and that the sensitivity is relatively low,
sequences, contains two distinct melting domains: making it unsuitable for detection of low-abundant
a high-melting domain and a low-melting domain methylated alleles. In addition, like in MS-DGGE,
containing the entire sequence of interest. Introduc- manipulation of the PCR product’s melting properties
ing a new high-melting domain and converting the may be required to ensure that the entire region of
sequence of interest into a single low-melting domain interest melts as a single peak.
may both be accomplished by PCR-mediated attach-
ment of a GC-rich sequence, generally known as Methylation-specific denaturing high-performance
a GC-clamp (Sheffield et al. 1989). The impact of liquid chromatography (MS-DHPLC)
adding a GC-clamp to one of the ends of a PCR
product can be accurately predicted by using the DHPLC (Xiao and Oefner 2001) was developed as a
computer algorithm MELT87 (Lerman and Silverstein mutation-detection technique and has later been
1987). adapted for detection of changes in DNA methyla-
tion patterns. In the original protocol for analysis of
Methylation-specific melting curve analysis genomic DNA, mutant and wild-type amplicons are
(MS-MCA) subjected to analysis on reversed-phase chromatog-
raphy supports under partially denaturing conditions
MS-MCA (Worm et al. 2001) is an in-tube assay and detected by on-line UV or fluorescence monitor-
that, like MS-DGGE, exploits differences in melting ing. The stationary phase, which has become commer-
temperature between methylated and unmethylated cially available, is made of alkylated nonporous
alleles after bisulfite treatment. Essentially, non- poly(styrene-divinylbenzene) particles. Separation of
discriminatory amplification of the bisulfite-treated DNA molecules is achieved by means of a hydro-
template is performed in the presence of a PCR- organic eluent containing triethylammonium and
compatible, fluorescent dye that specifically binds acetate ions. At elevated temperatures, retention of
246
DNA molecules is sequence dependent, forming the of techniques for gene-specific methylation analysis.
basis for resolution of sequence differences (Xiao and Within recent years, however, new high-throughput
Oefner 2001). methods have made it possible to simultaneously
Deng et al. (2000) and Baumer et al. (2001) analyze the methylation status of thousands of CpG
showed that DHPLC is also useful for resolution of islands in a pool of DNA.
differences in DNA methylation. After bisulfite treat-
ment of DNA, the region of interest is amplified using Restriction landmark genomic scanning (RLGS)
non-discriminatory primers, and the amplicon is then
applied directly onto the DHPLC column. Ampli- RLGS is a method that provides a bird’s-eye view of
cons generated from methylated sequences have a the methylation status of thousands of unselected CpG
higher G:C-content and, therefore, melt less and elute islands in the genome within a single gel (Costello
later than the corresponding unmethylated sequences. et al. 2000). Genomic DNA is initially digested
DHPLC is a rapid and cost-effective technique for with methylation-sensitive restriction enzymes, such
high-throughput analysis of DNA methylation. The as NotI, which recognize large CpG-rich sequences.
disadvantages include the need for expensive, special- These sequences usually occur in CpG islands and are
ized equipment and the use of the toxic substance, cleaved only if the CpG dinucleotides are unmethyl-
acetonitrile. At present, little is known about the ated. The digested DNA is radioactively labeled at
sensitivity of MS-DHPLC. the restriction sites using [α-32 P]dCTP, [α-32 P]dGTP
and DNA polymerase, digested with a second restric-
Methylation-specific microarray (MSO) tion enzyme, such as EcoRV, and then subjected to
electrophoresis in an agarose tube gel (first-dimension
MSO uses bisulfite-treated DNA as template for separation). The agarose gel is then equilibrated in
non-discriminatory PCR amplification, followed by HinfI digestion buffer and the DNA digested in the gel
hybridization of the PCR product to glass slides with HinfI. The agarose gel is then placed horizon-
carrying oligonucleotides that discriminate between tally across the top of a nondenaturing polyacrylamide
methylated and unmethylated cytosines at specific gel, and the DNA is separated by electrophoresis in
CpG positions (Gitan et al. 2002; Adorjan et al. the second dimension. The gels are then dried and
2002). To allow detection of hybridization signals exposed to X-ray film or inspected using a phosphor-
by fluorescence analysis, the amplification primers imager.
carry a fluorescent label. Each of the detection oligo- The above approach produces a complex pattern
nucleotides is designed to hybridize to the bisulfite- of spots, in which a spot will be missing if a particular
modified sequence around one CpG site, which is NotI site is methylated and has not, therefore, been
either originally unmethylated (TG) or methylated recognized and cleaved. This spot may be stabbed
(CG). Hybridization conditions must be selected to from the corresponding control gel for further identifi-
allow detection of the single nucleotide differences cation by sequence analysis. Alternatively, software
between the TG and CG oligonucleotides. One of the systems have been developed for automated analysis
great potentials of MSO is that multiple genes can be of RLGS profiles. Differences between RLGS profiles
analyzed on the same array. For example, Adorjan et have been used to identify imprinted genes and genes
al. (2002) assessed the methylation status of 232 indi- involved in disease states such as cancer.
vidual CpG sites representing 56 different genes as a One of the most critical parameters in RLGS is
means to distinguish different solid tumors. A poten- DNA quality; even small amounts of degraded DNA
tial limitation of this method is that closely spaced may lead to a significant reduction in the quality of
CpGs may not be amenable to analysis if the gene in the RLGS profiles. Other disadvantages include the
question is heterogeneously methylated. labor-intensive and time-consuming procedure, the
need for large amounts of starting DNA, and the diffi-
cult interpretation of the two-dimensional gel images.
Global CpG island methylation analysis Furthermore, although the technique is useful for
simultaneously assessing the methylation status of a
Previous attempts to identify new genes that are differ- large number of CpG sites, it does not provide infor-
entially methylated in human disease have primarily mation about whether or not a given CpG island
been candidate gene approaches relying on the use resides within the promoter region of a gene. Accord-
247
ingly, only a proportion of the aberrantly methylated for methylation changes and to isolate specific DNA
CpG islands are associated with transcriptional gene fragments associated with these changes. This method
silencing. relies on digesting genomic DNA with methylation-
sensitive and insensitive restriction enzymes followed
Methylated CpG island by amplification with random primers that preferen-
amplification-representational difference analysis tially bind to sequences associated with CpG islands.
(MCA-RDA) A similar approach was described by Huang et al.
(1997), who called it methylation-sensitive restriction
MCA (Toyota et al. 1999) is a PCR-based method
fingerprinting (MSRF). According to the protocol of
that allows preferential amplification of closely spaced Gonzalgo et al. (1997), genomic DNA is digested
methylated SmaI sites, which occur predominantly in
in three independent reactions, containing RsaI,
CpG islands. Unmethylated SmaI sites in genomic RsaI + HpaII, and RsaI + MspI, respectively, and
DNA are first eliminated by digestion with SmaI, subsequently amplified and radioactively labeled with
which cuts only when CpG sites within the target
GC-rich primers under low-stringency PCR conditions
sequence are not methylated, and leaves the fragment in the presence of [α-33 P]dATP. The resulting PCR
blunt ended. DNA is then digested with the SmaI
products are resolved on high-resolution polyacryla-
isoschizomer, XmaI, which digests at methylated sites
mide gels under denaturing conditions, followed by
and leaves an overhang. Adaptors are then ligated to exposure of the dried gel to autoradiographic film.
these overhangs, and PCR is performed using primers
Fragments that show differential methylation can be
complementary to these adaptors. The MCA ampli-
stabbed from the gel and identified by sequence
cons represent a pool of methylated CpG islands that analysis. Separate digestions of DNA using the
can be used directly in a dot blot analysis to study the
methylation-sensitive enzyme, HpaII, and its methyla-
methylation status of any gene for which a probe is tion-insensitive isoschizomer, MspI, are important for
available. Alternatively, to identify CpG islands whose controlling for the presence of PCR products unrelated
methylation status differs between normal DNA and
to methylation changes. Double digestion with either
DNA from pathological tissues, MCA can be coupled of these enzymes in combination with RsaI is used to
with representational difference analysis (RDA), a
generate smaller fragments. MS-AP-PCR can detect
subtractive hybridization technique designed to isolate
methylation changes in as little as 200 ng of genomic
genomic differences between two complex genomes DNA, but like RLGS and MCA-RDA it does not
(Lisitsyn et al. 1993). The subtractive and kinetic
provide information about the possible associations
enrichment of differentially methylated sequence by between methylation changes and changes in gene
RDA has the potential to identify sequences methyl- expression.
ated only in pathological conditions.
Like RLGS, the MCA-RDA method can be useful Differential methylation hybridization (DMH)
for simultaneously determining the methylation status
of multiple CpG-dense loci. It should be noted, Future studies on global changes in DNA methylation
however, that only those SmaI sites that are <1 kb will probably be speeded up by the use of microarray-
apart can be amplified using MCA. On the one hand, based technologies. The principle of one such method,
this will ensure that the most CpG-rich sequences are DMH, was first described by Huang et al. (1999)
represented in the MCA amplicons; on the other hand, and has since been further developed for large-scale,
∼25% of all possible CpG islands do not contain global analysis using a microarray format (Yan et al.
two closely spaced SmaI sites. Other disadvantages 2001). Genomic DNA is pre-cut with a methylation-
of MCA-RDA are that it examines only a limited insensitive enzyme, such as MseI. Linkers are then
number of CpG sites within a given CpG island, and ligated to the digested DNA before it is incubated with
that ∼60% of the recovered clones are Alu repetitive the methylation-sensitive enzymes, BstUI and HpaII.
sequences. The resulting digests are amplified by PCR and the
products hybridized to an array of immobilized CpG
Methylation-sensitive arbitrarily primed PCR islands. Yan et al. (2001) used a microarray containing
(MS-AP-PCR) ∼8,000 GC-rich tags to identify epigenetic alterations
MS-AP-PCR (Gonzalgo et al. 1997) is a PCR-based in breast cancer.
fingerprinting method that can be used to screen
248
So many techniques – how to choose? to provide the necessary information. For example,
neither extensive analysis of the methylation pattern
Mapping of 5-methylcytosine in genomic DNA has nor sophisticated quantitative analysis are required to
become an important tool for studying the molecular make a diagnosis of an imprinting disorder, in which
basis of human disease and for understanding tissue- either methylated or unmethylated alleles are entirely
specific gene expression. However, to generate reli- missing in all cells. In this case, it may be fully
able data on DNA methylation is by no means a trivial sufficient to rely on a technique with relatively low
task. First, the cellular heterogeneity of most tissues sensitivity, such as MSRE-Southern analysis. On the
makes it difficult to assign a particular methylation other hand, highly sensitive techniques are required
signal to a specific cell type. Second – as revealed for detection of low-abundant, aberrantly methylated
by the plethora of available techniques – no single alleles in aging or cancer tissues, or in DNA released
technique fulfills all criteria for generating unambigu- from neoplastic and preneoplastic lesions into serum,
ous data on DNA methylation. The purpose of this urine or sputum. Usually, the most rational approach is
review has been to provide an overview of current to use methylation-specific PCR to determine whether
key techniques and describe the most important advan- or not aberrantly methylated alleles are present in
tages and disadvantages, which should be carefully the sample. Once a positive signal has been obtained
considered before deciding on which approach would using methylation-specific PCR, further quantitative
best provide a useful set of data. Some of these or semi-quantitative analysis is necessary to exclude
techniques are technically challenging, and choice false positives. In-depth resolution of methylation
of methodology will often depend on the available patterns at the nucleotide level requires sequence
equipment and expertise. analysis of PCR products.
The quality and quantity of the starting DNA Unfortunately, bisulfite-based procedures for
are important limiting factors in DNA methylation detection of aberrant DNA methylation are neither
analyses; several of the techniques described in this trivial nor foolproof. As described in detail above, a
review require >1 µg of high-quality DNA from number of potential artifacts and pitfalls should be
cultured cells or fresh tissue samples for reliable carefully considered, including PCR bias, incomplete
results. Accordingly, due to the limited quantity and conversion of cytosine, and unintended conversion
extensive degradation of DNA, only sparse infor- of 5-methylcytosine. Furthermore, PCR assays based
mation on DNA methylation may be derived from on bisulfite treatment of DNA may not be easily
archived pathology samples, such as formalin-fixed, established. Li and Dahiya (2002) have developed
paraffin-embedded tissue sections. Degradation of a program, MethPrimer, that may be useful for
DNA during bisulfite treatment may further reduce designing primers to specifically amplify bisulfite-
the number of intact DNA molecules amenable to treated DNA. In its present version, this program
PCR analysis, and two rounds of amplification using is useful for designing primers for methylation-
nested or semi-nested primers are usually necessary to specific PCR, but further development is undergoing
generate sufficient amounts of the product. The low to make it supportive for COBRA and Ms-SnuPE as
number of starting molecules may result in stochastic well. MethPrimer has been made freely accessible
amplification, leading to the generation of a PCR (http://itsa.ucsf.edu/˜urolab/methprimer/).
product that does not accurately reflect the distribu- A particularly powerful tool for identifying an
tion of methylated and unmethylated molecules in the aberrantly methylated cell type in a complex tissue
original sample. Millar et al. (2002) have described a is bisulfite analysis of microdissected cells. Indeed,
number of bisulfite protocols that may help optimizing bisulfite treatment of DNA in agarose beads has
the analysis of archived samples. Nevertheless, data been used to generate detailed methylation patterns
on DNA methylation obtained from limited material of single cells representing different stages of male
must always be interpreted with great caution, unless germ cell differentiation (Kerjean et al. 2000). The
the results can be verified in several independent combination of microdissection and bisulfite analysis
experiments. is likely to significantly increase our knowledge about
For gene-specific methylation analysis, where the the role of DNA methylation in development, disease
target under investigation usually is an aberrantly and aging.
methylated CpG island, it is important to consider how
detailed the resolution of 5-methylcytosine must be
249
Li E, Bestor TH and Jaenisch R (1992) Targeted mutation of the Rein T, DePamphilis ML and Zorbas H (1998) Identifying 5-
DNA methyltransferase gene results in embryonic lethality. Cell methylcytosine and related modifications in DNA genomes.
69: 915–926 Nucleic Acids Res 26: 2255–2264
Li LC and Dahiya R (2002) MethPrimer: designing primers for Robertson KD (2002) DNA methylation and chromatin – unraveling
methylation PCRs. Bioinformatics 18: 1427–1431 the tangled web. Oncogene 21: 5361–5379
Lisitsyn N, Lisitsyn N and Wigler M (1993) Cloning the differences Schmitt F, Oakeley EJ and Jost JP (1997) Antibiotics induce
between two complex genomes. Science 259: 946–951 genome-wide hypermethylation in cultured Nicotiana tabacum
Maekawa M, Sugano K, Kashiwabara H, Ushiama M, Fujita S, plants. J Biol Chem 272: 1534–1540
Yoshimori M and Kakizoe T (1999) DNA methylation analysis Sheffield VC, Cox DR, Lerman LS and Myers RM (1989) Attach-
using bisulfite treatment and PCR-single-strand conforma- ment of a 40-base-pair G+C-rich sequence (GC-clamp) to
tion polymorphism in colorectal cancer showing microsatellite genomic DNA fragments by the polymerase chain reaction in
instability. Biochem Biophys Res Commun 262: 671–676 improved detection of single-base changes. Proc Natl Acad Sci
Millar DS, Warnecke PM, Melki JR and Clark SJ (2002) Methyl- USA 86: 232–236
ation sequencing from limiting DNA: embryonic, fixed, and Singer-Sam J, LeBon JM, Tanguay RL and Riggs AD (1990) A
microdissected cells. Methods 27: 108–113 quantitative HpaII-PCR assay to measure methylation of DNA
Oakeley EJ (1999) DNA methylation analysis: a review of current from a small number of cells. Nucleic Acids Res 18: 687
methodologies. Pharmacol Ther 84: 389–400 Southern EM (1975) Detection of specific sequences among DNA
Oakeley EJ, Podesta A and Jost JP (1997) Developmental changes fragments separated by gel electrophoresis. J Mol Biol 98: 503–
in DNA methylation of the two tobacco pollen nuclei during 517
maturation. Proc Natl Acad Sci USA 94: 11721–11725 Toyota M, Ho C, Ahuja N, Jair KW, Li Q, Ohe-Toyota M, Baylin
Oakeley EJ, Schmitt F and Jost JP (1999) Quantification of 5- SB and Issa JP (1999) Identification of differentially methyl-
methylcytosine in DNA by the chloroacetaldehyde reaction. ated sequences in colorectal cancer by methylated CpG island
Biotechniques 27: 744–50, 752 amplification. Cancer Res 59: 2307–2312
Okano M, Bell DW, Haber DA and Li E (1999) DNA methyl- Wade PA (2001) Methyl CpG binding proteins: coupling chromatin
transferases Dnmt3a and Dnmt3b are essential for de novo architecture to gene regulation. Oncogene 20: 3166–3173
methylation and mammalian development. Cell 99: 247–257 Warnecke PM, Stirzaker C, Melki JR, Millar DS, Paul CL and Clark
Olek A, Oswald J and Walter J (1996) A modified and improved SJ (1997) Detection and measurement of PCR bias in quantita-
method for bisulphite based cytosine methylation analysis. tive methylation analysis of bisulphite-treated DNA. Nucleic
Nucleic Acids Res 24: 5064–5066 Acids Res 25: 4422–4426
Orita M, Iwahana H, Kanazawa H, Hayashi K and Sekiya T (1989) Warnecke PM, Stirzaker C, Song J, Grunau C, Melki JR and Clark
Detection of polymorphisms of human DNA by gel as single- SJ (2002) Identification and resolution of artifacts in bisulfite
strand conformation polymorphisms. Proc Natl Acad Sci USA sequencing. Methods 27: 101–107
86: 2766–2770 Worm J, Aggerholm A and Guldberg P (2001) In-tube DNA methyl-
Paulin R, Grigg GW, Davey MW and Piper AA (1998) Urea ation profiling by fluorescence melting curve analysis. Clin
improves efficiency of bisulphite-mediated sequencing of 5 - Chem 47: 1183–1189
methylcytosine in genomic DNA. Nucleic Acids Res 26: 5009– Wu J, Issa JP, Herman J, Bassett DE, Jr., Nelkin BD and Baylin
5010 SB (1993) Expression of an exogenous eukaryotic DNA methyl-
Paulsen M and Ferguson-Smith AC (2001) DNA methylation in transferase gene induces transformation of NIH 3T3 cells. Proc
genomic imprinting, development, and disease. J Pathol 195: Natl Acad Sci USA 90: 8891–8895
97–110 Xiao W and Oefner PJ (2001) Denaturing high-performance liquid
Raizis AM, Schmitt F and Jost JP (1995) A bisulfite method of chromatography: A review. Hum Mutat 17: 439–474
5-methylcytosine mapping that minimizes template degradation. Xiong Z and Laird PW (1997) COBRA: a sensitive and quantitative
Anal Biochem 226: 161–166 DNA methylation assay. Nucleic Acids Res 25: 2532–2534
Rein T, Zorbas H and DePamphilis ML (1997) Active mammalian Yan PS, Chen CM, Shi H, Rahmatpanah F, Wei SH, Caldwell CW
replication origins are associated with a high-density cluster of and Huang TH (2001) Dissecting complex epigenetic alterations
mCpG dinucleotides. Mol Cell Biol 17: 416–426 in breast cancer using CpG island microarrays. Cancer Res 61:
8375–8380