Is It a Code: The Debate
Yes: An epigenetic histone code may allow for unprecedented predictive power
The nuclear signaling networks that operate through chromatin define the mechanisms by which genomes interact with the environment. Posttranslational modifications to histone tails act and interact in many different ways, and discoveries of complex and "cross-talking" phenomena appear regularly. Given that we seem to be making perfectly good progress with no more than sound protein biochemistry, what is to be gained by postulating a histone or epigenetic code?
The short answer is that a code takes us into a new dimension by addressing the possibility that environmental effects may be inherited from one cell generation to the next and possibly through the germ line. A histone-based epigenetic code would act in parallel with the genetic code and should offer comparable predictive power. The search for such a code is surely worth undertaking, but first we must establish what we are looking for, and where to look.
As a start, we might borrow from semiotics. As 40 million plus readers of a recent best seller are likely aware, semiotics is the study of signs and symbols and their use or meaning. A semiotic system consists of a sign, its meaning (i.e., the outcome), and the code by which the sign brings that meaning into effect. In practice, one or more adaptors will be required to link sign and outcome to "read" the code. This simple system has two important properties. First, the code is arbitrary. For example, a red traffic light will generally stop traffic because we have established a code that says red means stop, but we could just as well stop for green lights if everyone was in agreement. Second, the sign is independent of the outcome. The red light does not appear because the traffic has stopped (as a red brake light would) nor does the red light have any direct role in stopping traffic (as a pile of rocks in the road might).
The genetic code meets the requirements of a semiotic system. DNA triplets (signs) give rise to specified amino acids (meanings) according to rules specified by a code and interpreted by tRNAs (adaptors). There is no physical link between base triplet and amino acid, and the code is arbitrary in that its meaning can be altered by introducing a mutant tRNA.
Asking whether histone modifications and their functional effects constitute valid semiotic systems is informative. Results linking patterns of histone acetylation to transcriptional activity fail because the sign (acetylation) is itself directly involved in the meaning/outcome (transcription). If the histone code is to be just a generic, vaguely defined term encompassing all the histone modifications linked to chromatin function, this need not concern us. But if we are searching for a truly predictive epigenetic code, transcriptionally active chromatin is not the place to look.
Transcriptionally silent heterochromatin regions may be more promising. H3s di- and trimethylated at lysine 9 (H3K9me2/3) are good markers of silent heterochromatin, both in metaphase chromosomes and in interphase. They act as receptors for the heterochromatin protein HP1, whose binding triggers major changes in chromatin structure. Unfortunately, H3K9me2/3 molecules are present in the final outcome (heterochromatin), though their involvement beyond the initial HP1 binding stage is uncertain. The H3K9me2/3 sign is arbitrary (as required of a code) as it can also occur in transcriptionally active regions, where it is read by a different HP1 isoform.1 These multiple outcomes also demonstrate that, as expected, H3K9me2/3 alone does not constitute a code, and additional modifications to adjacent residues are needed to generate a meaningful sign. More worrying for H3K9me2/3, however, gene silencing can sometimes precede H3K9 methylation,2 suggesting that it can be the consequence and not the cause of silencing, a brake light rather than a traffic signal.
In truth we do not yet have the finding that will prove or disprove the epigenetic code hypothesis. But we do have candidates, and by clearly defining what we expect our code to do, we can at least start to look in the right places.
Bryan M. Turner is professor of experimental genetics at the University of Birmingham's Institute of Biomedical Research.
1. C.R. Vakoc et al., "Histone H3 lysine 9 methylation and HP1gamma are associated with transcription elongation through mammalian chromatin," Mol Cell, 19:381-91, 2005.
2. V. Mutskov, G. Felsenfeld, "Silencing of transgene transcription precedes methylation of promoter DNA and histone H3 lysine 9," EMBO J, 23:138-49, 2004.
Is It a Code: The Debate
No: Histone modifications, although diverse, do not constitute a complex code of chromatin states
Some of the best-understood regulatory processes in biology involve posttranslational modifications of proteins. The PDGF-b receptor (PDGFbR) is regulated via phosphorylation of seven tyrosines on its cytoplasmic domain. Upon extracellular binding of the PDGF ligand, PDGFbR dimerizes and transphosphorylates these tyrosines. The different phosphotyrosines are then bound by SH2-containing proteins, resulting in transduction of extracellular PDGF into a cytoplasmic signal.
Histones are subject to an extraordinary number of diverse posttranslational modifications. Acetylations, methylations, and phosphorylations have evidently constrained the unstructured tails of histones H3 and H4 to be essentially invariant in eukaryotes. Furthermore, functional studies have implicated many of these modifications in maintaining the active state of genes during development. Therefore, like phosphorylation in signal transduction, modification of histones in cellular memory has emerged as an important theme in biological regulation.1
The large number of different histone modifications and their potential for gene regulation has been interpreted as a "histone code," in which combinations of modifications on the same or different histones result in a variety of chromatin states. Histone-modifying enzymes are said to "write" the code, and modification-specific binding proteins are said to "read" it. How is this a code, compared to modification and independent binding as with PDGF signal transduction?
Central to the concept of a code is translation of information. The genetic code translates base triplets into amino acids. The Morse code translates a series of dots and dashes into letters and digits. Significantly, a single Morse code dot translates as "e" while two dots code for "i," not "ee." The latter would entail a trivial representation of each dot by an e. By the same token, if two histone modifications are independently bound with strictly cumulative effects, this is straightforward binding chemistry, not the readout of an informational code.
An alternative to a combinatorial histone code is that each modification directly changes a nucleosomal property with only cumulative effects. For example, acetylation of lysine neutralizes its positive charge and thus reduces its interaction with negatively charged DNA. This charge-neutralization model predicts that the precise location of an acetylated residue on a histone tail is not important, but two acetylated lysines will have a stronger effect in the same direction as one. Indeed, this is the case: For the yeast genome as a whole, nearly every affected gene shows essentially the same effect of multiple lysine acetylations on the histone H4 tail, regardless of which lysines are acetylated.2
Methylation of lysine is different from acetylation in that it is not thought to directly change the properties of chromatin, but rather to provide a binding platform for specific chromatin-associated proteins. Different methyl-bound proteins in proximity might interact to produce levels of chromatin activity that are not simply the cumulative effects of independent action. In this way, different combinations of modifications would result in the complex differentiation of chromatin. However, most modifications associated with transcriptional activity are strongly correlated. Even important modifications on different histone molecules, such as di- and trimethylation of H3K4 and acetylation of H3 and H4 tails closely correspond with one another along the chromosome in yeast, flies, and humans.3 This leaves very little room for differentiation of active chromatin attributable to a histone code.
In summary, histone modifications do not appear to differ in any fundamental way from other important protein modifications involved in biological regulation, nor do they appear to constitute a combinatorial code for gene expression. The combinatorial histone code has not fulfilled its promise as a paradigm for cellular memory. Rather, current evidence favors a binary alternative, in which the cumulative effects of histone modifications favor active or silent chromatin. Active chromatin is inherited through the action of Trithorax Group proteins, which counteract silencing by Polycomb Group proteins. Both groups include diverse chromatin-modifying enzymes. The many correlated histone modifications that result from their actions would assure the robustness of this binary system of epigenetic inheritance.1
1. S.L. Schreiber, B.E. Bernstein, "Signaling network model of chromatin," Cell, 111:771-8, 2002.
2. M.F. Dion et al., "Genomic characterization reveals a simple histone H4 acetylation code," Proc Natl Acad Sci, 102:5501-6, 2005.
3. B. van Steensel, "Mapping of genetic and epigenetic regulatory networks using microarrays," Nat Genet, 37(suppl):S18-24, 2005.
Steven Henikoff is a Member of the Basic Sciences Division at the Fred Hutchinson Cancer Research Center, Seattle, and an Investigator of the Howard Hughes Medical Institute.
The pirate site plans to ignore the lawsuits from Elsevier and the American Chemical Society.