Science Wiki

DNA Methylation is a conserved biochemical and epigenetic process in which DNA methyltransferase (DNMT) enzymes catalyze the covalent addition of a methyl group (–CH₃) to specific nucleobases within a DNA molecule—most prominently at the carbon-5 position of cytosine to yield 5-methylcytosine (5mC)—without altering the primary Watson–Crick base-pairing sequence. Often referred to as the "fifth base" of the genome, 5mC projects its hydrophobic methyl group directly into the major groove of the B-DNA double helix, where it modulates transcription factor binding, recruits repressive chromatin-remodeling complexes, silences transposable elements, maintains chromosomal stability, and orchestrates mammalian genomic imprinting and X-chromosome inactivation. Across prokaryotes (bacteria and archaea), DNA methylation at both cytosine and adenine (N⁶-methyladenine, 6mA) additionally regulates DNA replication timing, mismatch DNA repair, and restriction-modification self/non-self immunity.

In mammals and most vertebrates, DNA methylation occurs almost exclusively on cytosine bases followed immediately by a guanine base in the 5' → 3' direction, a palindromic dinucleotide motif denoted 5'-CpG-3' (where the "p" represents the intervening phosphodiester bond). Because a 5'-CpG-3' dinucleotide on one strand pairs with a complementary 3'-GpC-5' (which reads 5'-CpG-3' on the antiparallel strand), every CpG site contains two symmetrically positioned cytosines on opposite strands, providing the structural basis for semiconservative mitotic inheritance of methylation patterns.

S-Adenosylmethionine (SAM) Catalysis and Base Flipping

All eukaryotic DNA methyltransferases share a conserved C-terminal catalytic domain that uses S-adenosyl-L-methionine (SAM or AdoMet) as the universal electrophilic methyl donor. Because C5 of the pyrimidine ring is tightly stacked inside the DNA double helix and chemically unreactive toward direct nucleophilic substitution, DNMT enzymes execute a remarkable base-flipping and covalent catalysis mechanism:

1. The DNMT enzyme binds the DNA minor and major grooves and rotates ("flips") the target cytosine nucleotide 180° completely out of the helical stack into a buried catalytic pocket.
2. A conserved catalytic cysteine thiolate side chain (–S⁻) in the enzyme's PCQ motif performs a nucleophilic Michael addition onto the C6 carbon of the cytosine ring, forming a covalent enzyme–DNA intermediate and generating high carbanion (enamine) electron density at the C5 carbon.
3. The activated C5 carbanion attacks the electrophilic methyl group attached to the sulfonium ion of SAM, transferring the –CH₃ group to C5 and releasing *S*-adenosyl-L-homocysteine (SAH).
4. Abstraction of the proton at C5 followed by β-elimination breaks the covalent C6–cysteine bond, releasing the enzyme and returning the newly formed 5-methylcytosine (5mC) back into the double-stranded helix.

CpG Depletion and CpG Islands in Vertebrate Genomes

Whereas bulk cytosine (C) deaminates spontaneously by hydrolysis into uracil (U)—an unnatural base in DNA that is rapidly recognized and excised by uracil-DNA glycosylase (UDG) during base excision repair—5-methylcytosine (5mC) spontaneously deaminates into thymine (T), a canonical DNA base:

Because T:G mismatches are repaired less efficiently than U:G mismatches, methylated CpG dinucleotides act as mutational hotspots that undergo C → T transition mutations at rates 10- to 50-fold higher than unmethylated cytosines. Over hundreds of millions of years of vertebrate evolution, this hypermutability has depleted CpG dinucleotides across the human genome to only ~21% of their statistically expected frequency (~28 million CpG sites in the haploid human genome, of which 70%–80% are constitutively methylated in somatic tissues).

The major exception to this global depletion is the presence of roughly 30,000 CpG islands (CGIs)—stretches of DNA 300 to 3,000 base pairs long with a G+C content >50% and an observed-to-expected CpG ratio >0.6. Over 70% of annotated human gene promoters (including virtually all housekeeping genes and developmental master regulators) contain CpG islands that remain unmethylated in the germline and healthy somatic cells, protecting their CpGs from deamination and keeping chromatin transcriptionally permissive.

Mammalian cells coordinate DNA methylation dynamics through a division of labor among three catalytically active DNA methyltransferases ("writers") and three Ten-Eleven Translocation (TET) methylcytosine dioxygenases ("erasers"):

Replication-Coupled Maintenance by DNMT1 and UHRF1

When a symmetrically methylated 5'-^mCpG-3' / 3'-GpC^m-5' duplex undergoes semiconservative DNA replication, both daughter duplexes emerge transiently hemimethylated (methylated on the parental template strand, unmethylated on the newly synthesized strand). The multi-domain protein UHRF1 specifically recognizes hemimethylated CpG dinucleotides via its SRA (SET and RING-associated) domain, flips out the parental 5mC, and ubiquitylates histone H3 tails (H3K18ub/H3K23ub). This recruits DNMT1—which is also tethered to the sliding clamp PCNA at the replication fork—to methylate the opposing cytosine on the nascent daughter strand within minutes of replication, preserving epigenetic memory across dozens of mitotic cell divisions with >99% fidelity per site.

Active Demethylation via the TET–TDG Oxidation Cascade

Because the carbon–carbon bond connecting the C5 ring atom of cytosine to the methyl group (C5–CH₃) is thermodynamically extremely stable, no human enzyme can cleave it directly in a single step. DNA demethylation occurs either passively (by inhibiting or excluding DNMT1/UHRF1 during replication so 5mC is diluted by 50% per cell division) or actively through enzymatic oxidation catalyzed by TET1, TET2, and TET3 dioxygenases.

Using molecular oxygen (O₂), ferrous iron (Fe²⁺), ascorbate (vitamin C), and the Krebs-cycle metabolite α-ketoglutarate (2-oxoglutarate), TET enzymes iteratively oxidize 5-methylcytosine through three distinct oxidized cytosine intermediates:

Once 5fC or 5caC is formed, the base-excision repair enzyme Thymine DNA Glycosylase (TDG) recognizes and excises the oxidized base, leaving an abasic (AP) site that is cleaved by APE1 endonuclease and filled in with a canonical unmethylated cytosine by DNA Polymerase β and DNA Ligase III. Notably, 5hmC (the "sixth base") accumulates stably to high levels (~0.5%–1% of all cytosines) in post-mitotic Purkinje and cortical neurons, where it functions as an independent epigenetic mark associated with active transcription.

When a promoter CpG island becomes densely methylated, gene transcription is potently silenced through two synergistic mechanisms:

1. Direct Steric Blockade: The bulky C5 methyl group protruding into the major groove physically prevents methylation-sensitive transcription factors (such as CTCF, MYC, E2F, and NRF1) from binding their cognate DNA recognition motifs.
2. Methyl-CpG-Binding Domain (MBD) Recruitment: Specialized "reader" proteins containing a methyl-CpG-binding domain—including MeCP2 (mutated in Rett syndrome), MBD1, MBD2, and Kaiso—bind symmetrically methylated CpGs and recruit corepressor complexes (such as NuRD and Sin3A) containing histone deacetylases (HDACs) and H3K9 methyltransferases. These remodel the surrounding nucleosomes into compacted, inaccessible heterochromatin.

Paradoxically, while promoter CpG island methylation silences transcription initiation, gene body methylation (within exons and introns of actively transcribed genes, guided by H3K36me3 recruiting DNMT3B) correlates positively with active elongation by RNA Polymerase II by suppressing spurious intragenic initiation and regulating alternative splicing.

In mammals, two waves of global epigenetic reprogramming—first in the fertilized zygote (where the paternal pronucleus is rapidly demethylated by TET3) and second in developing primordial germ cells (PGCs)—reset the methylome. Specific loci called Imprinting Control Regions (ICRs) escape zygotic reprogramming via ZFP57/KAP1 protection, enforcing genomic imprinting where ~150 mammalian genes (such as *IGF2*/*H19* and the *SNRPN* Prader–Willi/Angelman locus) are expressed exclusively from either the maternal or the paternal allele. In human oncology, tumors characteristically exhibit a dual methylome catastrophe: global hypomethylation of repetitive elements (triggering retrotransposon reactivation and chromosomal instability) coupled with focal CpG island hypermethylation that epigenetically silences critical tumor suppressor genes (*BRCA1*, *MLH1*, *CDKN2A/p16*, *RB1*).

How does bisulfite sequencing distinguish 5-methylcytosine from unmethylated cytosine?

Standard polymerase chain reaction (PCR) and DNA polymerases cannot distinguish 5mC from C because both base-pair identically with guanine (G). In sodium bisulfite sequencing, single-stranded DNA is treated with bisulfite (HSO₃⁻), which sulfonates and hydrolytically deaminates unmethylated cytosine into uracil (U) (read as thymine T after PCR amplification), whereas 5-methylcytosine (5mC) is chemically resistant to bisulfite deamination and remains read as C at single-base resolution.

Why do mammals die during embryonic development if DNA methylation is removed, whereas yeast and *C. elegans* lack 5mC entirely?

Model organisms such as budding yeast (*Saccharomyces cerevisiae*), the nematode *Caenorhabditis elegans*, and *Drosophila melanogaster* lost canonical DNMT1/DNMT3 enzymes during evolution and rely exclusively on histone modifications (H3K9me3, H3K27me3) and RNA interference for gene silencing. In vertebrates, however, genomes are vastly larger (~3 Gb) and >45% composed of transposable elements (LINEs, SINEs, ERVs); without DNA methylation, transposons reactivate catastrophically and genomic imprinting collapses.

How does the Horvath epigenetic clock measure biological aging?

As humans age, reproducible stochastic drift and regulated shifts occur at specific CpG dinucleotides across all tissues—some CpGs progressively gain methylation while others lose it. Developed by Steve Horvath in 2013, the pan-tissue epigenetic clock uses regularized machine-learning regression across 353 specific CpG sites to predict chronological and biological age within ±3.6 years, reflecting cumulative mitotic history and TET/DNMT maintenance fidelity.