Histones serve as slates to a dizzying array of modifications, but researchers are confident they can decipher the epigenetic puzzle.
Roughly two meters of DNA gets packed into every cell nucleus in the human body. In addition to stuffing all that information into a sphere 3 to 10 microns across, the proteins that perform this task must also ensure that in each cell certain genes are constantly transcribed, while others lie ready, and other regions remain dormant, practically inaccessible. Within this cramped, chaotic space, an army of proteins must manage cellular information, decide cell fate with a moment's notice and maintain it, often passing that fate to daughter cells.
This regulation takes place in the context of the histone proteins. Two each of the four standard histones - H3, H4, H2A, and H2B - join together to form an octameric nucleosome, a spool around which roughly 146 nucleotides wind in a near double loop. By no means inert packing material, the nucleosome serves as a slate for a rich variety of modifications or "marks" that appear to play a role in managing the genome. Acetylation, methylation, phosphorylation, ubiquitination, sumoylation - the list of modifications grows monthly. Researchers have mapped as many as 50 different marks to specific amino acid residues on the histones' highly conserved N-terminal tails and elsewhere on the molecules, says Thomas Jenuwein from the Research Institute of Molecular Pathology in Vienna. "One of the challenges, I think, is to identify all of the modifications that exist," says Tony Kouzarides, of the Wellcome Trust/Cancer Research UK Gurdon Institute.
When a methyltransferase like G9a methylates H3K9, heterochromatin protein 1 (HP1) can bind, generally resulting in a silent state. Phosphorylation to the adjacent S10 residue by Aurora kinase B may loosen the association. Further modification of the H3 tail, such as acetylation at H3K14 through a Histone acetyltransferase (HAT) prevents HP1 binding. Additionally, the histone methytransferase Suv39 associates with HP1 (inset) providing a potential mechanism for spread of the methyl signal.
Constant additions to the cast of histone marks (see image below), and the regulatory proteins that make them, remove them, or act upon their presence, put the number of potential combinations in the astronomical range. Say there were 30 different residues marked in a given nucleosome, 230 gives a billion possible combinations. A human cell has approximately only 40 million incorporated nucleosomes. Nevertheless, many appear to act together. Take one oft-described scenario on histone H3: Methylated lysine 4 (H3K4Me) and an acetylated lysine 14 (H3K14Ac) together produce an active gene state similar to a phosphorylated serine 10 in combination with acetylated lysines 14 and 9. Methylated lysine 9 is antagonistic to the serine 10 phosphorylation acting like a switch to shut down expression (see image above left). And, adjacent serine-lysine pairs appear elsewhere in the histone tails.
These "cross-talking" units and other signaling cascades (see image below) have an alluring quality to many researchers. A layer of information is definitely encoded in these signals, says C. David Allis, a Howard Hughes Medical Institute investigator at Rockefeller University. "As you develop from a totipotent or pluriopotent cell to a developmentally differentiated cell ... the DNA is remaining constant. It sort of tells us [that] something non-DNA is going on," says Allis, but just how much information is stored in these signals has been a matter of debate.
HDACs AND BEYOND
As histones potentially set and maintain the fate not only for adult cells, but also for cancers, and possibly even for offspring, an explosion of research on their epigenetic roles has ensued in the 10 years since the first histone acetyltransferase (HAT)1 and the first histone deacetylase (HDAC)2 were described by Allis, and Harvard's Stuart Schreiber, also an HHMI investigator, respectively. Acetylated lysines on the histone tails had been known about for decades, but the identification of the proteins that specifically make and remove these marks opened the field. Now efforts are concentrated on uncovering all the possible marking sites and finding out just what they do.
Jeff Boeke's group at Johns Hopkins University School of Medicine has been attacking histones by mutating them. The group has targeted modifiable lysines in yeast in an effort to dissect biological pathways and networks, one of the goals of the NIH Roadmap for Medical Research. Although modifiable lysines are not restricted to histones, they appear there "to a ridiculous extent," says Boeke. By changing them to glutamine or arginine, they can mimic acetylated or unacetylatable lysines, respectively. Then by examining the effects of these mutations on silencing phenotypes, researchers can predict which residues might be modified in a way that affects expression status of silenced regions. Boeke's group has been able to make accurate predictions about H3K56, a modified lysine in the core of histone H3.3 "There's a couple of other residues where we've made a similar prediction. We haven't found evidence by mass spectrometry, but we're making antibodies against the potential modified forms," Boeke says.
Unfortunately, such genetic analyses simply can't mimic histone methylation, a mark that has attracted much attention in the community. Methyl marks are more stable than acetyl marks and once thought to be practically permanent. But a lysine demethylase, LSD1, was described by Yang Shi's group at Harvard Medical School in late 2004.4 Dimethyl lysines can be stripped of their signal, too, by a protein called JHDM1 as Yi Zhang's group at the University of North Carolina showed in February,5 revealing a function for a domain known as JmjC. And in March, Shi's group reported on an enzyme that can reverse lysine trimethylation.6 JMJD2A, which also contains a JmjC domain, changes H3K9 and K36 trimethylation to dimethylation. They report that they've also found JMJD2 family members that revert trimethyl marks to monomethyl marks directly, potentially adding to the fine tuning of this system.
Some of the known posttranslational modifications including acetylation (Ac), methylation (Me), phosphorylation (P), and ubiquitination (Ub) made to histone tails are shown in this diagram showing half of a nucleosome. Additional marks (not shown here) have been mapped to residues within the cores or on the C-terminal tails of H3, H4, and H2B, as well as to the H1 linker histone.
Kouzarides' group has been focusing on both lysine and arginine methylation. No arginine demethylases have yet been discovered, but his group among others has identified a deimination process that changes a histone's methylated arginine into a citrulline residue.7 More recently he's been presenting data on the cis/trans isomerization of proline 38 on histone H3. Isomerization flips the conformation of proline 38 which in turn regulates methylation of the nearby lysine 36 and affects transcription, he says.
CONDUCTING THE ORCHESTRA
Such work is likely to uncover many more modifications but won't always explain how these work in concert across large segments of the genome or even over whole genomes. Oliver Rando at the Bauer Center for Genomic Research at Harvard Medical School mutated individual lysine residues on the tail of yeast histone H4. By changing H4's lysines 5, 8, 12, and 16 to arginine individually and in combination, his group surmised that expression goes up or down for a few hundred genes in a roughly cumulative way for lysines 5, 8, and 12 (see image below).8 Here, it appears that acetylation acts largely as a function of charge associated with the individual modifications. H4K16 acetylation has an all-or-nothing "digital" property for a subset of genes but retains some of the "analog" qualities of its neighbor lysines, Rando says. "The interesting thing is that if you want to predict the expression of a triple mutant, say K5, 8, and 16, you would do pretty well to just add [the] individual expression profiles together," says Rando. Although Rando calls the effect a "code" in the manuscript, he says it's a very simple code, "There's no fancy combinatorial stuff going on."
Unlike yeast, metazoans have dozens of copies of each "standard" histone protein in addition to several so-called variants, which makes genetic analysis impossible. But observing histone modification patterns in organisms with a variety of cell types has yielded interesting clues. This past month, Bradley Bernstein of Massachusetts General Hospital and a group including Schreiber and Eric Lander reported that in mouse embryonic stem cells, many crucial developmental genes carry histone marks for active and silent chromatin simultaneously.9 "You've got both repressive K27 and activating K4 methylation marks," Bernstein says, creating what he calls a bivalent state. "The model is that in ES cells you're silencing these critical developmental genes, but keeping them poised so that they can be activated when the cells differentiate."
Strikingly, says Bernstein, in embryonic stem cells, the underlying DNA sequence itself is the best predictor for the patterns of methylation at these developmental loci. H3K27 methylation appears across broad swaths of so-called transposon exclusion zones, which are areas free of transposons. H3K4 methylation happens at CpG islands. "Bivalent domains are where you have a CpG island in the middle of a transposon exclusion zone," says Bernstein. During differentiation, one or the other methylation marks disappears presumably by demethylation or histone replacement, although precisely when the chromatin changes occur is hard to pin down.
The chicken and egg question is an important one. An army of regulatory proteins sets and maintains the histone marks, often replacing histones with variants that are slightly different. DNA sequence has to be controlling this epigenetic state at some point. "If you just keep peeling back ... these writers have to know where they go in chromatin," says Allis. Another predictor for nucleosome state appears to be simply gene transcription.
Experiments in which one, two, three, or four histone H4 lysine residues were changed to arginine show that for many genes, incremental decreases or increases in expression correlate roughly to the number of mutations. K16 mutations resulted in significant expression changes for a subset of genes (*all combinations of 2 and 3 mutations were tested).
Steven Henikoff, HHMI investigator at Fred Hutchinson Cancer Research Center in Seattle, recently characterized the incorporation of a histone variant known as H3.3 across much of the Drosophila melanogaster genome.10 While H3 is predominantly available for incorporation during DNA replication, H3.3 is available and gets incorporated during transcription. The more a gene transcribes, the more H3.3 is incorporated, providing a possible means for cells to "remember" which genes are active. Other variants appear to be broadly associated with other chromatin states. H2AX, for example, which accounts for 10% to 15% of all H2A nuclear content, appears to be required to mend double strand breaks and manage B-cell DNA breaks during class switch recombination. Steven Jackson's group at the Wellcome Trust/Cancer Research UK Gurdon Institute recently showed that the phosphorylated C-terminus of H2AX is specifically bound by the DNA damage-response factor MDC1/NFBD1.11
Hypersensitivity of H2AX mutants to DNA damage has made it a promising clinical marker for cancer, says Jackson, and other epigenetic marks are increasingly being linked to cancer states, including DNA and histone methylation as well as histone acetylation. Michael Grunstein of the University of California, Los Angeles, notes that many published hints suggest epigenetic state might serve a prognostic purpose in cancer. He points to the work of his former postdoc Siavash Kurdistani, now at the David Geffen School of Medicine in Los Angeles. Kurdistani's group used highly specific antibodies to detect levels of acetylation at H3K9, K18, and H4K12, as well as dimethylation at H4R3 and H3K4, in tumor samples of prostate cancer patients.12 Differences between patients in the bulk modifications at these five spots could successfully place patients into two prognostic groups, and two marks - H3K4 dimethylation and H3K18 acetylation - were very successful for grouping patients on their own.
Much of the wide-scale genomic work suggests that these marks do appear in regular patterns but don't act with any high degree of combinatorial complexity. "What's going on is complex, but it's hardly a code," Grunstein says. Allis famously subscribed to the "histone code hypothesis" when he wrote about it in 2000,13 positing that combinations of histone modifications produce specific outcomes through proteins that "read" the marks and modify chromatin accordingly. Several researchers contend that this combinatorial complexity simply isn't there and that such regulation is no different from other signaling pathways. Histone marks simply don't carry the easy readout of a genetic code, in which each combination of three nucleotides specifies something.
Sequential modifications regulate processes in chromatin. The paf1 complex which associates with elongating RNA PolII is required for ubiquitination of H2BK123 by Rad6. This mark enables subsequent methylation of H3K79 or H3K4. Trimethylation at these locations has different functional consequences due to the recruitment of different effector proteins.
Allis remains undeterred. Recent work, he says, comes closer than ever to fulfilling some of the requirements for a code. His latest reason for optimism is a PHD finger motif that appears to specifically recognize trimethylated H3K4. In work he presented at a Keystone Symposium in January, he talked about the protein BPTF, which contains a PHD domain as well as a bromodomain, a well-established acetyl recognition motif. The two recognition motifs in the protein are separated by an unstructured alpha helix, what he calls a spacer or "ruler" that might allow the protein to read different marks on the same or adjacent histones. Recalling a figure from his 2000 article that helped popularize the histone code hypothesis, he says, "We explicitly said, there should be docking molecules that have the right patches. I think the new work on that PhD, bromodomain, and the linker comes real close." The dimethyl eraser, JHDM1, and the trimethyl eraser, JMJD2, also contain PHD domains.
Both Allis and Jenuwein, who have written together on the subject, say they're a bit reluctant to bring up "the code" at conferences and talks, but they still think about it. On one level, the disagreement is a semantic argument about the definition of a code (see debate). Nevertheless, contention remains as to whether histone marks act with any real combinatorial complexity. Bryan Turner, considered the first to propose a code,14 continues to engage in debate (in these pages and in a friendly face-off with Rando at the American Association for Cancer Research this past month).
Allis hasn't stopped thinking bigger. He's been collaborating with Alexander Tarakhovsky, another lab head at Rockefeller who described a histone methyltransferase working outside the nucleus.15 Although phosphorylation, ubiquitination, acetylation, and other modifications occur regularly in the cytosol, methylation has largely been restricted to the nucleus. Allis posits, "That gets to the bigger problem of, is it a histone code or is it a protein code?"
Semantics aside, the question remains: How do the marks work? "Our view of how modifications function is very two-dimensional.... They recruit a protein, but really what effect that protein or the complex of proteins has on chromatin is really a mystery," says Kouzarides. This mystery has drawn him and a slew of other researchers into charting the rules for this intricate regulatory system. At the center of it all, the nucleosome's influence is only likely to grow. "It may well be the case that the kinds of rules identified for histones are going to be operating in other parts of the cell," says Jackson. "Maybe this is a good paradigm for other systems."
1. J. Taunton, "A mammalian histone deacetylase related to the yeast transcriptional regulator Rpd3p," Science, 272:408-11, 1996.
2. J.E. Brownell, "Tetrahymena histone acetyltransferase A: a homolog to yeast Gcn5p linking histone acetylation to gene activation," Cell, 84:843-51, 1996.
3. E.M. Hyland et al., "Insights into the role of histone H3 and histone H4 core modifiable residues in Saccharomyces cerevisiae," Mol Cell Biol, 25:10060-70, 2005.
4. Y. Shi et al., "Histone demethylation mediated by the nuclear amine oxidase homolog LSD1," Cell, 119:941-53, 2004.
5. Y.-I. Tsukada et al., "Histone demethylation by a family of JmjC domain-containing proteins," Nature, 439:811-6, Feb. 16, 2006.
6. J.R. Whetstine et al., "Reversal of histone lysine trimethylation by the JMJD2 family of histone demethylases," Cell, doi:10.1016/j.cell.2006.03.028.
7. G.L. Cuthbert et al., "Histone deimination antagonizes arginine methylation," Cell, 118:545-553, 2004.
8. M.F. Dion et al., "Genomic characterization reveals a simple histone H4 acetylation code," Proc Natl Acad Sci, 102:5501-6, 2005.
9. B.E. Bernstein, "A bivalent chromatin structure marks key developmental genes in embryonic stem cells," Cell, April 21, 2006.
10. Y. Mito et al., "Genome-scale profiling of histone H3.3 replacement patterns," Nat Genet, 37:1090-7, October 2005.
11. M. Stucki et al., "MDC1 directly binds phosphorylated histone H2AX to regulate cellular responses to DNA double strand breaks," Cell, 123:1213-26, Dec. 29, 2005.
12. D.B. Seligson et al., "Global histone modification patterns predict risk of prostate cancer recurrence," Nature, 435:1262-6, 2005.
13. B.D. Strahl, C.D. Allis, "The language of covalent histone modifications," Nature, 403:41-5, 2000.
14. B.M. Turner, "Decoding the nucleosome," Cell, 75:5-8, 1993.
15. I.-H. Su, A. Tarakhovsky, "Lysine methylation and 'signaling memory,'" Curr Opin Immunol, 18:152-7, April 2006.