Introduction to epigenetics
The mechanisms governing phenotypic plasticity,
or the capacity of a cell to change its state in response to stimuli,
have long been the subject of research (Phenotypic plasticity 1). The
traditional central dogma of biology states that the DNA of a cell is transcribed to RNA, which is translated to proteins, which perform cellular processes and functions.
A paradox exists, however, in that cells exhibit diverse responses to
varying stimuli and that cells sharing identical sets of DNA such as in
multicellular organisms can have a variety of distinct functions and
phenotypes. Classical views have attributed phenotypic variation to differences in primary DNA structure, be it through aberrant mutation or an inherited sequence allele.
However, while this did explain some aspects of variation, it does not
explain how tightly coordinated and regulated cellular responses, such
as differentiation, are carried out.
A more likely source of cellular plasticity is through the Regulation of gene expression,
such that while two cells may have near identical DNA, the differential
expression of certain genes results in variation. Research has shown
that cells are capable of regulating gene expression at several stages:
mRNA transcription, processing and transportation as well as in protein
translation, post-translational processing and degradation.
Regulatory proteins that bind to DNA, RNA, and/or proteins are key
effectors in these processes and function by positively or negatively
regulating specific protein level and function in a cell.
And, while DNA binding transcription factors provide a mechanism for
specific control of cellular responses, a model where DNA binding transcription factors are the sole regulators of gene activity is also unlikely. For example, in a study of Somatic-cell nuclear transfer, it was demonstrated that stable features of differentiation remain after the nucleus
is transferred to a new cellular environment, suggesting that a stable
and heritable mechanism of gene regulation was involved in the
maintenance of the differentiated state in the absence of the DNA
binding transcription factors.
With the finding that DNA methylation and histone modifications
are stable, heritable, and also reversible processes that influence gene
expression without altering DNA primary structure, a mechanism for the
observed variability in cell gene expression was provided.
These modifications were termed epigenetic, from epi “on top of” the
genetic material “DNA” (Epigenetics 1). The mechanisms governing
epigenetic modifications are complex, but through the advent of high-throughput sequencing technology they are now becoming better understood.
Epigenetics
Genomic
modifications that alter gene expression that cannot be attributed to
modification of the primary DNA sequence and that are heritable mitotically and meiotically
are classified as epigenetic modifications. DNA methylation and
histone modification are among the best characterized epigenetic
processes.
DNA methylation
The
first epigenetic modification to be characterized in depth was DNA
methylation. As its name implies, DNA methylation is the process by
which a methyl group is added to DNA. The enzymes responsible for catalyzing this reaction are the DNA methyltransferases (DNMTs).
While DNA methylation is stable and heritable, it can be reversed by
an antagonistic group of enzymes known as DNA de-methylases. In
eukaryotes, methylation is most commonly found on the carbon 5 position
of cytosine residues (5mC) adjacent to guanine, termed CpG dinucleotides.
DNA methylation patterns vary greatly between species and even
within the same organism. The usage of methylation among animals is
quite different; with vertebrates exhibiting the highest levels of 5mC and invertebrates more moderate levels of 5mC. Some organisms like Caenorhabditis elegans
have not been demonstrated to have 5mC nor a conventional DNA
methyltransferase; this would suggest that other mechanisms other than
DNA methylation are also involved.
Within an organism, DNA methylation levels can also vary
throughout development and by region. For example, in mouse primordial germ cells, a genome wide de-methylation even occurs; by implantation stage, methylation levels return to their prior somatic levels. When DNA methylation occurs at promoter regions,
the sites of transcription initiation, it has the effect of repressing
gene expression. This is in contrast to unmethylated promoter regions
which are associated with actively expressed genes.
The mechanism by which DNA methylation represses gene expression
is a multi-step process. The distinction between methylated and
unmethylated cytosine residues is carried out by specific DNA-binding
proteins. Binding of these proteins recruit histone deacetylases (HDACs) enzyme which initiate chromatin remodeling such that the DNA becoming less accessible to transcriptional machinery, such as RNA polymerase, effectively repressing gene expression.
Histone modification
In eukaryotes, genomic DNA is coiled into protein-DNA complexes called chromatin. Histones,
which are the most prevalent type of protein found in chromatin,
function to condense the DNA; the net positive charge on histones
facilitates their bonding with DNA, which is negatively charged. The
basic and repeating units of chromatin, nucleosomes, consist of an octamer of histone proteins
(H2A, H2B, H3 and H4) and a 146 bp length of DNA wrapped around it.
Nucleosomes and the DNA connecting form a 10 nm diameter chromatin
fiber, which can be further condensed.
Chromatin packaging of DNA varies depending on the cell cycle stage and by local DNA region.
The degree to which chromatin is condensed is associated with a certain
transcriptional state. Unpackaged or loose chromatin is more
transcriptionally active than tightly packaged chromatin because it is
more accessible to transcriptional machinery. By remodeling chromatin
structure and changing the density of DNA packaging, gene expression can
thus be modulated.
Chromatin remodeling occurs via post-translational modifications of the N-terminal tails of core histone proteins. The collective set of histone modifications in a given cell is known as the histone code. Many different types of histone modification are known, including: acetylation, methylation, phosphorylation, ubiquitination, SUMOylation, ADP-ribosylation, deamination and proline isomerization;
acetylation, methylation, phosphorylation and ubiquitination have been
implicated in gene activation whereas methylation, ubiquitination,
SUMOylation, deamination and proline isomerization have been implicated
in gene repression. Note that several modification types including
methylation, phosphorylation and ubiquitination can be associated with
different transcriptional states depending on the specific amino acid on
the histone being modified. Furthermore, the DNA region where histone
modification occurs can also elicit different effects; an example being
methylation of the 3rd core histone at lysine residue 36 (H3K36). When
H3K36 occurs in the coding sections of a gene, it is associated with
gene activation but the opposite is found when it is within the promoter
region.
Histone modifications regulate gene expression by two mechanisms: by disruption of the contact between nucleosomes and by recruiting chromatin remodeling ATPases. An example of the first mechanism occurs during the acetylation of lysine terminal tail amino acids, which is catalyzed by histone acetyltransferases (HATs).
HATs are part of a multiprotein complex that is recruited to chromatin
when activators bind to DNA binding sites. Acetylation effectively
neutralizes the basic charge on lysine, which was involved in
stabilizing chromatin through its affinity for negatively charged DNA.
Acetylated histones therefore favor the dissociation of nucleosomes and
thus unwinding of chromatin can occur. Under a loose chromatin state,
DNA is more accessible to transcriptional machinery and thus expression
is activated. The process can be reversed through removal of histone
acetyl groups by deacetylases.
The second process involves the recruitment of chromatin
remodeling complexes by the binding of activator molecules to
corresponding enhancer regions. The nucleosome remodeling complexes
reposition nucleosomes by several mechanisms, enabling or disabling
accessibility of transcriptional machinery to DNA. The SWI/SNF protein complex
in yeast is one example of a chromatin remodeling complex that
regulates the expression of many genes through chromatin remodeling.
Relation to other genomic fields
Epigenomics
shares many commonalities with other genomics fields, in both
methodology and in its abstract purpose. Epigenomics seeks to identify
and characterize epigenetic modifications on a global level, similar to
the study of the complete set of DNA in genomics or the complete set of
proteins in a cell in proteomics.
The logic behind performing epigenetic analysis on a global level is
that inferences can be made about epigenetic modifications, which might
not otherwise be possible through analysis of specific loci. As in the other genomics fields, epigenomics relies heavily on bioinformatics, which combines the disciplines of biology, mathematics and computer science.
However while epigenetic modifications had been known and studied for
decades, it is through these advancements in bioinformatics technology
that have allowed analyses on a global scale. Many current techniques
still draw on older methods, often adapting them to genomic assays as is
described in the next section.
Methods
Histone modification assays
The cellular processes of transcription, DNA replication and DNA repair
involve the interaction between genomic DNA and nuclear proteins. It
had been known that certain regions within chromatin were extremely
susceptible to DNAse I digestion, which cleaves DNA in a low sequence specificity manner. Such hypersensitive sites were thought to be transcriptionally active regions, as evidenced by their association with RNA polymerase and topoisomerases I and II.
It is now known that sensitivity to DNAse I regions correspond to
regions of chromatin with loose DNA-histone association.
Hypersensitive sites most often represent promoters regions, which
require for DNA to be accessible for DNA binding transcriptional
machinery to function.
ChIP-Chip and ChIP-Seq
Histone modification was first detected on a genome wide level through the coupling of chromatin immunoprecipitation (ChIP) technology with DNA microarrays, termed ChIP-Chip.
However instead of isolating a DNA-binding transcription factor or
enhancer protein through chromatin immunoprecipitation, the proteins of
interest are the modified histones themselves. First, histones are cross-linked to DNA in vivo through light chemical treatment (e.g., formaldehyde). The cells are next lysed, allowing for the chromatin to be extracted and fragmented, either by sonication or treatment with a non-specific restriction enzyme (e.g., micrococcal nuclease). Modification-specific antibodies in turn, are used to immunoprecipitate the DNA-histone complexes. Following immunoprecipitation, the DNA is purified from the histones, amplified via PCR and labeled with a fluorescent tag (e.g., Cy5, Cy3).
The final step involves hybridization of labeled DNA, both
immunoprecipitated DNA and non-immunoprecipitated onto a microarray
containing immobilized gDNA. Analysis of the relative signal intensity
allows the sites of histone modification to be determined.
ChIP-chip was used extensively to characterize the global histone modification patterns of yeast.
From these studies, inferences on the function of histone
modifications were made; that transcriptional activation or repression
was associated with certain histone modifications and by region. While
this method was effective providing near full coverage of the yeast
epigenome, its use in larger genomes such as humans is limited.
In order to study histone modifications on a truly genome level,
other high-throughput methods were coupled with the chromatin
immunoprecipitation, namely: SAGE: serial analysis of gene expression (ChIP-SAGE), PET: paired end ditag sequencing (ChIP-PET) and more recently, next-generation sequencing (ChIP-Seq).
ChIP-seq follows the same protocol for chromatin immunoprecipitation
but instead of amplification of purified DNA and hybridization to a
microarray, the DNA fragments are directly sequenced using next
generation parallel re-sequencing. It has proven to be an effective
method for analyzing the global histone modification patterns and
protein target sites, providing higher resolution than previous methods.
DNA methylation assays
Techniques
for characterizing primary DNA sequences could not be directly applied
to methylation assays. For example, when DNA was amplified in PCR or bacterial cloning techniques, the methylation pattern was not copied and thus the information lost. The DNA hybridization technique
used in DNA assays, in which radioactive probes were used to map and
identify DNA sequences, could not be used to distinguish between
methylated and non-methylated DNA.
Restriction endonuclease based methods
Non genome-wide approaches
The earliest methylation detection assays used methylation modification sensitive restriction endonucleases.
Genomic DNA was digested with both methylation-sensitive and
insensitive restriction enzymes recognizing the same restriction site.
The idea being that whenever the site was methylated, only the
methylation insensitive enzyme could cleave at that position. By
comparing restriction fragment
sizes generated from the methylation-sensitive enzyme to those of the
methylation-insensitive enzyme, it was possible to determine the
methylation pattern of the region. This analysis step was done by
amplifying the restriction fragments via PCR, separating them through gel electrophoresis and analyzing them via southern blot with probes for the region of interest.
This technique was used to compare the DNA methylation modification patterns in the human adult and hemoglobin gene loci. Different regions of the gene (gamma delta beta globin) were known to be expressed at different stages of development.
Consistent with a role of DNA methylation in gene repression, regions
that were associated with high levels of DNA methylation were not
actively expressed.
This method was limited not suitable for studies on the global
methylation pattern, or ‘methylome’. Even within specific loci it was
not fully representative of the true methylation pattern as only those
restriction sites with corresponding methylation sensitive and
insensitive restriction assays could provide useful information. Further
complications could arise when incomplete digestion of DNA by
restriction enzymes generated false negative results.
Genome wide approaches
DNA methylation profiling on a large scale was first made possible through the Restriction Landmark Genome Scanning (RLGS)
technique. Like the locus-specific DNA methylation assay, the
technique identified methylated DNA via its digestion methylation
sensitive enzymes. However it was the use of two-dimensional gel electrophoresis that allowed be characterized on a broader scale.
However it was not until the advent of microarray and next
generation sequencing technology when truly high resolution and
genome-wide DNA methylation became possible.
As with RLGS, the endonuclease component is retained in the method but
it is coupled to new technologies. One such approach is the
differential methylation hybridization (DMH), in which one set of
genomic DNA is digested with methylation-sensitive restriction enzymes
and a parallel set of DNA is not digested. Both sets of DNA are
subsequently amplified and each labelled with fluorescent dyes and used
in two-colour array hybridization. The level of DNA methylation at a
given loci is determined by the relative intensity ratios of the two
dyes. Adaptation of next generation sequencing to DNA methylation
assay provides several advantages over array hybridization.
Sequence-based technology provides higher resolution to allele specific
DNA methylation, can be performed on larger genomes, and does not
require creation of DNA microarrays which require adjustments based on
CpG density to properly function.
Bisulfite sequencing
Bisulfite sequencing
relies on chemical conversion of unmethylated cytosines exclusively,
such that they can be identified through standard DNA sequencing
techniques. Sodium bisulfate and alkaline treatment does this by converting unmethylated cytosine residues into uracil
while leaving methylated cytosine unaltered. Subsequent amplification
and sequencing of untreated DNA and sodium bisulphite treated DNA allows
for methylated sites to be identified. Bisulfite sequencing, like the
traditional restriction based methods, was historically limited to
methylation patterns of specific gene loci, until whole genome
sequencing technologies became available. However, unlike traditional
restriction based methods, bisulfite sequencing provided resolution on a
nucleotide level.
Limitations of the bisulfite technique
include the incomplete conversion of cytosine to uracil, which is a
source of false positives. Further, bisulfite treatment also causes DNA
degradation and requires an additional purification step to remove the
sodium bisulfite.
Next-generation sequencing is well suited in complementing bisulfite sequencing in genome-wide methylation analysis.
While this now allows for methylation pattern to be determined on the
highest resolution possible, on the single nucleotide level, challenges
still remain in the assembly step because of reduced sequence complexity
in bisulphite treated DNA. Increases in read length seek to address
this challenge, allowing for whole genome shotgun bisulphite sequencing
(WGBS) to be performed. The WGBS approach using an Illumina Genome
Analyzer platform and has already been implemented in Arabidopsis thaliana.
Chromatin accessibility assays
Chromatin
accessibility is the measure of how "accessible" or "open" a region of
genome is to transcription or binding of transcription factors. The
regions which are inaccessible (i.e. because they're bound by nucleosomes) are not actively transcribed by the cell while open and accessible regions are actively transcribed.
Changes in chromatin accessibility are important epigenetic regulatory
processes that govern cell- or context-specific expression of genes.
Assays such as MNase-seq, DNase-seq, ATAC-seq or FAIRE-seq are
routinely used to understand the accessible chromatin landscape of
cells. The main feature of all these methods is that they're able to
selectively isolate either the DNA sequences that are bounded to the histones,
or those that are not. These sequences are then compared to a reference
genome that allows to identify their relative position.
MNase-seq and DNase-seq both follow the same principles, as they
employ lytic enzymes that target nucleic acids to cut the DNA strands
unbounded by nucleosomes or other proteic factors, while the bounded
pieces are sheltered, and can be retrieved and analyzed. Since active,
unbound regions are destroyed, their detection can only be indirect, by
sequencing with a Next Generation Sequencing
technique and comparison with a reference. MNase-seq utilizes a
micrococcal nuclease that produces a single strand cleavage on the
opposite strand of the target sequence. DNase-seq employs DNase I, a non-specific double strand-cleaving endonuclease.
This technique has been used to such an extent that nucleosome-free
regions have been labelled as DHSs, DNase I hypersensitive sites, and has been ENCODE consortium's election method for genome wide chromatin accessibility analyses. The main issue of this technique is that the cleavage distribution can be biased, lowering the quality of the results.
FAIRE-seq (Formaldehyde-Assisted Isolation of Regulatory
Elements) requires as its first step crosslinking of the DNA with
nucleosomes, then DNA shearing by sonication.
The free and linked fragments are separated with a traditional
phenol-chloroform extraction, since the proteic fraction is stuck in the
interphase while the unlinked DNA shifts to the aqueous phase and can
be analyzed with various methods.
Sonication produces random breaks, and therefore is not subject to any
kind of bias, and is also the bigger length of the fragments (200-700
nt) makes this technique suitable for wider regions, while it's unable
to resolve the single nucleosome.
Unlike the nuclease-based methods, FAIRE-seq allows the direct
identification of the transcriptionally active sites, and a less
laborious sample preparation.
ATAC-seq is based on the activity of Tn5 transposase. The
transposase is used to insert tags in the genome, with higher frequency
on regions not covered by proteic factors. The tags are then used as
adapters for PRC or other analytical tools.
Direct detection
Polymerase sensitivity in single-molecule real-time sequencing
made it possible for scientists to directly detect epigenetic marks
such as methylation as the polymerase moves along the DNA molecule being
sequenced. Several projects have demonstrated the ability to collect genome-wide epigenetic data in bacteria.
Nanopore sequencing is based on changes of electrolytic current signals according to base modifications (e.g. Methylation). A polymerase mediates the entrance of ssDNA
in the pore: the ion-current variation is modulated by a section of the
pore and the consequently generated difference is recorded revealing
the position of CpG. Discrimination between hydroxymethylation and methylation is possible thanks to solid-state nanopores even if the current while passing through the high-field region of the pore may be slightly influenced in it. As a reference amplified DNA is used which will not present copied methylated sites after the PCR process. The Oxford Nanopore Technologies MinION sequencer is a technology where, according to a hidden Markov
model, it is possible to distinguish unmethylated cytosine from the
methylated one even without chemical treatment that acts to enhance the
signal of that modification. The data are registered commonly in
picoamperes during established time. Other devices are the Nanopolish
and the SignaAlign: the former expresses the frequency of a methylation
in a read while the latter gives a probability of it derived from the
sum of all the reads.
Single-molecule real-time sequencing (SMRT) is a single-molecule DNA sequencing method. Single-molecule real-time sequencing utilizes a zero-mode waveguide (ZMW).
A single DNA polymerase enzyme is bound to the bottom of a ZMW with a single molecule of DNA as a template.
Each of the four DNA bases is attached to one of four different fluorescent dyes.
When a nucleotide is incorporated by the DNA polymerase, the
fluorescent tag is cleaved off and the detector detects the fluorescent
signal of the nucleotide incorporation.
As the sequencing occurs, the polymerase enzyme kinetics shift when it
encounters a region of methylation or any other base modification. When
the enzyme encounters chemically modified bases, it will slow down or
speed up in a uniquely identifiable way.
Fluorescence pulses in SMRT sequencing are characterized not only by
their emission spectra but also by their duration and by the interval
between successive pulses. These metrics, defined as pulse width
and interpulse duration (IPD), add valuable information about DNA
polymerase kinetics. Pulse width is a function of all kinetic steps
after nucleotide binding and up to fluorophore release, and IPD is
determined by the kinetics of nucleotide binding and polymerase
translocation.
In 2010 a team of scientists demonstrated the use of
single-molecule real-time sequencing for direct detection of modified
nucleotide in the DNA template including N6-methyladenosine, 5-methylcytosine and 5-hydroxylcytosine. These various modifications affect polymerase kinetics differently, allowing discrimination between them.
In 2017, another team proposed a combined bisulfite conversion
with third-generation single-molecule real-time sequencing, it is called
single-molecule real-time bisulfite sequencing (SMRT-BS), which is an
accurate targeted CpG methylation analysis method capable of a high
degree of multiplying and long read lengths (1.5 kb) without the need
for PCR amplicon sub-cloning.
Theoretical modeling approaches
First
mathematical models for different nucleosome states affecting gene
expression were introduced in 1980s [ref]. Later, this idea was almost
forgotten, until the experimental evidence has indicated a possible role
of covalent histone modifications as an epigenetic code.
In the next several years, high-throughput data have indeed uncovered
the abundance of epigenetic modifications and their relation to
chromatin functioning which has motivated new theoretical models for the
appearance, maintaining and changing these patterns. These models are usually formulated in the frame of one-dimensional lattice approaches.
