News & Views | Published:


Reversing the 'irreversible'

Nature volume 450, pages 357359 (15 November 2007) | Download Citation

“Do not speak — unless it improves on silence” is generally wise advice, and is even vital for a subset of essential genes. New studies describe how, when appropriate, the silence of these genes is broken.

In cellular tissues, it is essential that certain genes are turned on in appropriate cells but remain silent in others. For years, the dogma has been that some forms of gene silencing are irreversible, or at least extremely stable. One such form of silencing is mediated by the Polycomb group of proteins, which repress gene expression partly by altering the structure of chromatin — complexes of DNA and histone proteins. Such repression can be achieved by, for example, adding a specific molecular tag to a histone. Five papers1,2,3,4,5, including one by Jepsen et al.1 on page 415 of this issue, now describe how enzymatic removal of one molecular tag reverses the silencing of gene expression that is mediated by the Polycomb group.

It has long been known that the way genes are packaged inside the nucleus of a cell can determine their expression (transcription). Molecular tags on histones have crucial roles in controlling the activity of associated genes; such tags include acetyl, methyl and phosphate groups. Each of these tags may be added to or removed from specific amino acids in histones. The particular array of tags on histones at a specific genomic location may either facilitate or hinder the binding of regulatory proteins to DNA, resulting in the activation or inhibition of transcription6.

Of the various types of histone modification, methylation is the most stable. For example, precise patterns of trimethylated lysine residues at positions 9 or 27 of the amino-acid sequence of histone H3 can be maintained through many cell cycles in specific cell lineages, preserving the silenced state of the associated genes. However, in recent years, several enzyme groups that remove methyl tags from histones have been identified. The largest group of these histone demethylase enzymes is characterized by a catalytic structural domain called JmjC (refs 7, 8). Each of these JmjC-containing enzymes removes a methyl group from a specific lysine residue of histone H3.

The only enzymes that methylate lysine 27 of histone H3 are members of the Polycomb group of transcriptional repressors9. The mammalian methylase enzyme for this reaction is EZH2, which functions as part of a protein complex called PRC2, adding a trimethyl group to lysine 27. This leads to the recruitment of other Polycomb-group proteins and silencing of target genes (Fig. 1a). These include genes that are essential for regulating developmental pathways, such as HOX genes.

Figure 1: Reversal of Polycomb-group silencing by demethylases.
Figure 1

DNA (blue) wraps around histone proteins (brown spheres) to form nucleosomes, the building-blocks of chromatin. a, In the silenced state, the PRC2 complex, which contains Polycomb-group proteins, binds to target genes and adds three methyl groups (Me) to histone H3 at lysine 27 (K27). A second Polycomb-group-containing complex, PRC1, is recruited, partly because a component of this complex has a specific affinity for histone H3 trimethylated at K27. Consequently, PRC1, and possibly other Polycomb-group proteins, prevent transcription of the genes associated with this lysine. b, Five studies1,2,3,4,5 suggest that such gene silencing might be rapidly reversed by the MLL class of protein complexes, which contain either of two histone demethylase enzymes, JMJD3 and UTX. The binding of MLL complexes to regions containing histone H3 trimethylated at K27 is associated with displacement of PRC2, allowing JMJD3 or UTX to actively demethylate this particular lysine. Consequently, the association between PRC1 and the target gene is destabilized, permitting gene transcription and additional modifications that also promote transcription.

The reversible nature of Polycomb-group-mediated gene silencing and the significance of this regulatory plasticity for development and other physiological processes have become apparent only recently9. But because no demethylase enzyme had been identified for trimethylated lysine 27 of histone H3, there was a growing sense that Polycomb-group-mediated gene silencing might be reversed only by the slow loss of the trimethyl group from this lysine residue during cell proliferation, following removal of the PRC2 protein complex from target genes.

The new studies1,2,3,4,5 change this view, demonstrating that the trimethylated lysine is specifically demethylated by two similar JmjC-containing enzymes — JMJD3 and UTX. Both of these demethylases are components of MLL protein complexes, which antagonize Polycomb-group-mediated gene silencing (Fig. 1b).

The SMRT gene is known to be associated with neuronal development, and Jepsen et al.1 studied the brains of mouse embryos in which SMRT was deleted and also cultured cells derived from them. They found that the product of this gene is needed to maintain neural stem cells in their undifferentiated state. They also found that SMRT-deficient progenitor cells have a strong tendency to differentiate either into neurons or into non-neuronal brain support cells called glia. The authors explain these observations in terms of a direct, inhibitory effect of the SMRT protein on the expression of JMJD3.

The SMRT protein is a co-repressor that works with a nuclear receptor known as the retinoic-acid receptor to prevent expression of retinoic-acid-responsive genes. Treatment of normal cells with retinoic acid results in the dissociation of SMRT from the retinoic-acid receptor, triggering JMJD3 expression. Jepsen et al. suggest that, in turn, JMJD3 removes the trimethyl group from lysine 27 of histone H3, thereby contributing to the de-repression of several genes that trigger neurogenesis. This seems to happen while the levels of the EZH2 methylase enzyme remain constant.

Through this model, Jepsen and colleagues propose a mechanism by which histone tags added by Polycomb-group proteins are rapidly removed in response to retinoic-acid signalling. Consequently, neural stem cells are released to take on more specialized roles.

Two other studies2,3 report that retinoic-acid signalling leads to the recruitment of another enzyme that demethylates lysine 27 — UTX. This enzyme acts on multiple HOX genes in cultured human embryonic teratocarcinoma cells, causing displacement of EZH2, demethylation of lysine 27 and de-repression of the HOX genes. When mouse embryonic stem cells are treated with retinoic acid, EZH2 is rapidly displaced from retinoic-acid-responsive genes and the histone H3 lysine 27 residues associated with these genes are swiftly demethylated. Intriguingly, these changes are much slower in genes that are not retinoic-acid responsive10. This suggests a dual mode of 'reawakening' Polycomb-group-silenced genes in stem cells: a fast mode of de-repression, possibly involving demethylation of lysine 27 by JMJD3 or UTX; and a slow mode resulting from passive diminution of trimethylated lysine 27 in the absence of PRC2.

A further study4 finds that the JMJD3 demethylase also participates in regulating the differentiation of bone-marrow cells and the activation of immune-system cells called macrophages. When macrophages are exposed to bacterial products or other activators, JMJD3 expression is induced by a transcription factor called NF-κB, facilitating the activation of a Polycomb-group-silenced gene, BMP-2. So both the retinoic-acid receptor and NF-κB can trigger JMJD3 expression, suggesting that JMJD3 is regulated by different external stimuli in different cell types. In both cases, however, the targets of JMJD3 are discrete sets of genes that were initially silenced by Polycomb-group proteins.

Clearly, the identification of two demethylases for lysine 27 of histone H3 is only the first step in understanding their antagonistic relationship with Polycomb-group silencing proteins. Many questions remain concerning targeting of these enzymes to specific genes, their roles in the initial establishment of patterns of trimethylated lysine 27, and other factors with which they might interact to initiate transcription. Moreover, genetic analysis of JMJD3, UTX and their homologues in other species will probably reveal other developmental and physiological processes that depend on the reversal of Polycomb-group-mediated gene silencing.


  1. 1.

    et al. Nature 450, 415–419 (2007).

  2. 2.

    et al. Nature 449, 731–734 (2007).

  3. 3.

    et al. Science 318, 447–450 (2007).

  4. 4.

    et al. Cell 130, 1083–1094 (2007).

  5. 5.

    et al. Nature 449, 689–694 (2007).

  6. 6.

    , & Curr. Opin. Genet. Dev. 15, 163–176 (2005).

  7. 7.

    & Nature Rev. Mol. Cell Biol. 8, 307–318 (2007).

  8. 8.

    & Mol. Cell 25, 1–14 (2007).

  9. 9.

    & Nature Rev. Cancer 6, 846–856 (2006).

  10. 10.

    , & Stem Cells 25, 2191–2199 (2007).

Download references

Author information


  1. Richard S. Jones is in the Department of Biological Sciences, Southern Methodist University, Dallas, Texas 75275-0376, USA.

    • Richard S. Jones


  1. Search for Richard S. Jones in:

About this article

Publication history



Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing