Our Research Interests

DNA methylation

In mammals, DNA methylation is most often observed at cytosine residues in the context of a CG dinucleotide. The transfer of a methyl group (-CH3) to the position 5 of the cytosine ring is carried out by DNA methyltransferases using S-adenosyl-methionine (SAM) as a methyl group donor.

DNA methylation, particularly when applied to CG-rich promoter sequences, has been shown to silence gene expression in a heritable manner. DNA methylation is therefore a form of cellular memory. Because DNA methylation is not encoded in the DNA sequence itself, it is called an epigenetic modification (“epi”, Greek origin: “above” or “upon”). The transcriptional silencing associated with 5-methylcytosine is required for fundamental biological processes such as embryonic development, protection against intragenomic parasites, X-inactivation, genomic imprinting and cognitive functions. In addition, aberrant promoter methylation and inappropriate silencing of tumor suppressor genes has recently emerged as a major cause leading to cancer.

Did you know?

About 60% of all human genes are transcribed from a CG-rich promoter sequences (called CpG islands). Most of these islands are unmethylated, which correlates with an active status for the gene.

The bulk of DNA methylation in mammalian genomes is found at the many repetitive sequences that litter our genomes. It is critical to maintain this “junk DNA” in a transcriptionally silent stage, and DNA methylation is a key process that allows the cells to do just that. Evolutionary studies have revealed that silencing of repetitive “junk DNA” is probably the primary function of epigenetics.

The CG dinucleotide is the most under-represented dinucleotide in the human genome, due to the high rate of spontaneous deamination of 5-methylcytosine to thymine. CpG sites are clustered in CpG islands, which serve as promoter regions for many genes.

In Bacteria, DNA methylation is also prevalent, but is mostly used as part of site-specific “restriction/modification” systems that help the bacteria to protect their genome from invading foreign DNA. The catalytic domains of DNA methyltransferases are conserved from Bacteria to humans.

DNA methyltransferases

Four active DNA methyltransferases have been identified and studied in mammals. They are named DNMT (for DNA MethylTransferase)1, 2, 3A and 3B. Although these enzymes all carry highly conserved signature DNA methyltransferase motifs in their catalytic domains (black bars), genetic and biochemical studies indicate functional differences between these family members. DNMT3A and DNMT3B are considered to be de novo DNA methyltransferases, since they show equal activity on unmethylated and hemi-methylated DNA in vitro and are responsible for de novo methylation during early embryonic development. DNMT1, on the other hand, shows preference for hemi-methylated DNA substrates, is associated with replication foci, and is therefore considered a maintenance DNA methyltransferase. Together, these enzymes can generate new DNA methylation patterns and maintain them through cell divisions.

The DNMT2 enzyme was recently shown to possess a weak DNA methyltransferase activity, although its biological role remains unclear.

The DNMT3-like protein, DNMT3L

The DNMT3L protein was the last member assigned to the DNMT3 family on the basis of the high conservation of its N-terminal PHD-like zinc finger domain with the corresponding domains of DNMT3A and DNMT3B. The C-terminus of DNMT3L also shows partial homology to the C-terminal catalytic regions of DNMT3A and DNMT3B, although key catalytic residues involved in the transfer of the methyl groups are not conserved in DNMT3L. This raised the possibility that DNMT3L might not function as a DNA methyltransferase proper. Indeed, we showed that the human DNMT3L is inactive on its own ( Chédin et al, 2002).

Mouse knockouts for Dnmt3Lindicated however, that DNMT3L plays a key role in allowing DNA methylation during the maturation of germ cells. In the absence of DNMT3L, methylation imprints, which are normally established during oocyte growth in females or during spermatogonial differentiation in males, are lost. Embryos derived from Dnmt3L^-/-females die during gestation, show loss of maternal methylation at imprinting centers and aberrant expression of imprinted genes. Dnmt3L^-/-males are sterile, but recent data shows that spermatogonial cells from DNMT3L^-/- males also show loss of methylation at imprinting centers. DNMT3L is therefore required for maternal and paternal methylation imprints.

The paradox was therefore to understand how an inactive methyltransferase can nonetheless be critical for DNA methylation? In theory, DNMT3L could “regulate” other active DNA methyltransferases or could target DNA methylation to certain areas, such as imprinting centers. One clue came from the observation that DNMT3L co-localizes and co-immunoprecipitates with DNMT3A and DNMT3B. This suggested that DNMT3L might function together with these two de novo DNA methyltransferases.

Using an episomal DNA methyltransferase assay, we showed for the first time that DNMT3L was a general stimulatory factor for de novo methylation by DNMT3A (Chédin et al, 2002). DNMT3L is the first stimulatory factor for DNA methylation to be described.

In the presence of DNMT3L, DNMT3A’s DNA methyltransferase activity was increased on average about 4-fold at each CpG site. This means that the likelihood of, say, 10 CpG sites, to be all methylated is 4¹⁰=1,048,576 times higher in the presence of DNMT3L than in its absence. Not surprisingly, DNMT3L favored the formation of long, uninterrupted, methylation patterns. Interestingly, DNMT3L did not confer any sequence specificity to the methylation reaction, therefore indicating that it is not a targeting factor.

More recently, we have shown that the human DNMT3A, DNMT3A2, DNMT3B1, DNMT3B2 are all catalytically active DNA methyltransferases in vivo(the numbers refer to different enzymatic isoforms that are expressed with different time- and tissue-specificity). We also showed that DNMT3L stimulates all of these active isoforms, albeit to different levels. Finally, we showed that DNMT3L physically interacts with DNMT3A and DNMT3B through their C-terminal regions.

Did you know?

The bacterial HhaI and HaeIII DNA methyltransferases flip their target cytosine residue by 180 degrees outside of the DNA helix. Base-flipping is thought to be a conserved feature of all DNA methyltransferases.
Just after fusion of the sperm and the egg, the paternal pro-nucleus is actively de-methylated. This de-methylation probably contributes to the reprogramming of the paternal genome. However, the enzyme(s) responsible for this activity is currently unknown.
5-aza-cytidine and its derivatives are cytosine analogs which, once incorporated in the DNA, will trap DNA methyltransferases in a covalent complex, thus depleting them from the cells. Treatment with 5-aza-C or its derivatives results in artificial demethylation of the genome and reactivation of silenced genes. It is currently being tested as an anti-cancer drug in clinical trials.

Genomic imprinting

Genomic imprinting is both an amazing epigenetic phenomenon and somewhat of an oddity in mammalian genetics.

Imprinted genes, unlike most other genes, are expressed only from one allele in a strict parent-of-origin manner. Hence, some genes are only expressed from the allele inherited from the mother, while the others are solely expressed from the allele inherited from the father. This strange mode of inheritance explains why the maternal and paternal genomes, while genetically almost identical, are epigenetically very different. This also explains some classic experiments in which genomes from different parental origins were introduced in enucleated recipient oocytes and then transferred to pseudo-pregnant females in order to assess their developmental potential (see below).

Modified drawing from Janine LaSalle

Oocytes in which a maternal and a paternal genome were combined gave rise to normal embryos. Oocytes in which two paternal genomes were introduced failed to develop; the resulting embryos never grew properly and showed an abundance of extra-embryonic tissues (placenta). Maternally derived embryos developed further than their paternal counterparts, but had very reduced extra-embryonic tissues and ultimately died. We now understand that imprinted genes are the main stumbling block preventing such asexual reproduction (is that true? Check out recent news of the birth of the first mammal ever to be born without a father. What’s the trick?).

Imprinted genes / evolution of imprinting

There are about 70 imprinted genes known so far (Imprinted Gene Catalogue). ). It is estimated that imprinted genes might account for ~1% of the total genes, or about 300 genes. These genes play key roles in controlling cell proliferation, a particularly important aspect of embryonic development, and cognitive functions (many imprinted genes are expressed only in the brain). Many paternally expressed genes tend to favor cellular proliferation or to increase the growth of the placenta or the rate of transfer of nutrients from the mother to the embryo through the placenta. Many maternally expressed genes tend to do just the opposite. These observations are well explained in the context of the “conflict theory”, which explains the evolution of imprinting in mammals with in utero development as a tug of war between both sexes. The premise of the theory is that each sex is trying to maximize its reproductive potential. The trick, however, is that females are the sole provider of nutrients for the growth of the fetus during gestation. Taking advantage of this fact, males try to favor the transfer of nutrients from the mother to their offspring to ensure the birth of “healthy babies”, even if this comes at the expense of the mother’s future reproductive fitness. Females, on the other hand, try to tip the balance back to relieve some of the burden placed on them and increase their future reproductive fitness while providing just the right amount of nutrients to the fetus. Today, establishing this delicate balance is the only way to achieve a successful pregnancy. A similar battle between both sexes over how much care should be given to newborns is also apparent and determined, at least in part, by imprinted genes. Paternally expressed genes tend to favor increased maternal care (read more about this fascinating issue of behavior and imprinting here).

Imprinting centers

Other characteristics of imprinted genes include the fact that they tend to cluster in large regions of the genome. The mouse chromosome 7 is a haven for imprinted genes and carries at least three independent clusters of imprinted genes. Chromosomes 11p15 or 15q11-13 in humans carry extensive clusters as well (see figure below). Genetic studies have indicated that each cluster is controlled by one cis-acting imprinting center (IC), which coordinates the imprinted expression of all genes in the cluster.

15q11-q13
Modified from Nicholls and Knepper (2001)

In the case of the 15q11-q13 cluster represented above, an approximately 2 megabases segment is imprinted and is under the control of a single IC located at the 5’-end of the SNRPN gene. All genes in deep purple are paternally expressed. UBE3A (red) is the only maternally expressed gene within this cluster. Deletions in the IC can cause Prader-Willi syndrome (PWS), characterized by aberrant silencing of all normally paternally expressed genes, or in Angelman syndrome (AS), characterized by aberrant reactivation of all the normally maternally-silent genes. Mutations in UBE3A can also cause AS. PWS and AS patients carrying a set of overlapping IC deletions have allowed researchers to narrow down the IC to a few kilobases in length. Understanding the function of ICs is a major goal in the field of genomic imprinting and a current goal of the Chedin Laboratory.

Epigenetic regulation of imprinting

Genomic imprinting is an amazing example of epigenetic regulation through DNA methylation and chromatin modifications. It was indeed established that the regulation of imprinted gene functions solely through epigenetics. All known ICs and many imprinted genes show differential, parent-specific, DNA methylation marks. For example, the SNRPN IC is maternally methylated, which correlates with silencing (the H19 gene is an example for the opposite situation). Our current understanding of genomic imprinting indicates that ICs are differentially marked by DNA methylation (and probably, chromatin modifications) during gametogenesis. This phase of establishment of epigenetic settings is preceded by an erasure phase, where the settings inherited from the previous generation are erased in primordial germ cells (see figure below). Once the ICs have been marked in their respective germ lines, a new zygote can be formed that inherits alleles equipped with the proper epigenetic settings. The last phase of genomic imprinting corresponds to the maintenance of these settings through the billions of cell divisions that are going to occur during development so that each imprinted gene maintains its proper expression status.

Modified from Falls et al., 1999

Clearly, genomic imprinting provides a unique system where parent-specific, epigenetic modifications are targeted to specific loci to ensure the control of gene expression. It also clearly illustrates the power of epigenetics as both a heritable and reversible manner to control the expression of our genome.

You can read about this in more detail by checking out our publications.