DOI: 10.1126/science.1066726, 2113 (2001);294 Science

Adrian BirdMethylation Talk Between Histones and DNA

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

  here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

  ): July 7, 2014 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/294/5549/2113.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/294/5549/2113.full.html#ref-list-1, 4 of which can be accessed free:cites 12 articlesThis article

66 article(s) on the ISI Web of Sciencecited by This article has been

http://www.sciencemag.org/content/294/5549/2113.full.html#related-urls8 articles hosted by HighWire Press; see:cited by This article has been

http://www.sciencemag.org/cgi/collection/molec_biolMolecular Biology

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2001 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

July

7, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

July

7, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

July

7, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

July

7, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

6564636261605958575655545352515049484746454443424140393837363534333231302928272625242322212019181716151413121110987654321

www.sciencemag.org SCIENCE VOL 294 7 DECEMBER 2001 2113

ErbB-4 receptor is important for expres-sion of genes controlling growth inhibition.Ni et al. also report that the carboxyl-ter-minal region of ErbB-4 has weak transcrip-tional activity when fused to the DNAbinding domain of GAL4. Interestingly,Lin et al. (8) recently reported that theEGF receptor, a relative of ErbB-4, may al-so act directly in the nucleus and may af-fect gene expression. They found that afraction of the EGF receptor moved to thenucleus after activation of the receptor byligand binding. A strong transcriptional ac-tivity was observed when the carboxyl-ter-minal region of the EGF receptor wasfused to the DNA binding domain of Gal4.Furthermore, the EGF receptor complexbound to an AT-rich DNA sequence motif.The promoter for the gene encoding cyclinD1 contains two copies of this AT-rich mo-tif, which are required for activation of thisgene by EGF. Together, the results of Ni etal. (1) and Lin et al. (8) suggest that cyto-plasmic fragments of members of the EGFreceptor family are important regulators ofgene expression.

A crucial question is how receptorcleavage is itself regulated. It appears that,for the cases studied so far, the triggeringevent is the removal of most of the extracy-toplasmic part of the protein. This appearsto be a prerequisite for intramembranecleavage by γ-secretase or related proteas-es. In the case of ErbB-4, the metallopro-tease TACE that removes its ectodomain isactivated by protein kinase C (PKC). Thephysiological activator of PKC is phospho-lipase, but experimentally, as used by Ni et

al. (1), the phorbol ester TPA also does thejob. Interestingly, phospholipase C–γ is oneof the SH2 domain proteins that are acti-vated by several tyrosine kinase receptors.Thus, intramembrane proteolysis of ErbB-4 may be induced by its activation of SH2domain proteins (see the figure).

It is still unclear how the liberated intra-cellular fragments of the receptors aretranslocated to the nucleus. Because only asmall portion of ErbB-4 undergoes RIP, itis likely that nuclear transport does not oc-cur by random diffusion but rather is care-fully regulated. Ni et al. report that ErbB-4has putative nuclear localization and nucle-ar export sequences. At the very least, thenuclear export sequence is involved in nu-clear shuttling of the receptor fragment, be-cause accumulation of the ErbB-4 receptorfragment in the nucleus is enhanced by lep-tomycin B, which blocks nuclear export ofproteins (1). Lin et al. (8) also identified aputative nuclear localization sequence inthe juxtamembrane portion of the EGF re-ceptor. This sequence is conserved in allmembers of the EGF receptor family.

In both ErbB-4 and the EGF receptor, itis the extreme carboxyl-terminal region ofthe receptor that is the strongest activatorof transcription. However, the entire cyto-plasmic domain of both receptors appearsto be translocated to the nucleus. It is pos-sible that additional specific proteolyticevents take place in the nucleus such thatthe juxtamembrane domain (containingthe nuclear localization signal) and the ki-nase domain are cleaved off, liberating thetranscriptionally active carboxyl-terminal

tail. Another intriguing possibility is thatthe cytoplasmic receptor domains in thenucleus also regulate transcription throughphosphorylation events.

The discovery that RIP is another wayin which the cell conveys signals from theplasma membrane (and endoplasmic retic-ulum membrane) to the nucleus provides anew example of the multifaceted nature ofintracellular signal transduction. The Ni etal. work shows that tyrosine kinase recep-tors are also subject to RIP. One avenuefor future investigation will be to elucidatehow important RIP is compared with otherpathways in ErbB-4 signal transduction,and to see whether RIP is required for sig-naling through other tyrosine kinase recep-tors. Different members of this receptorfamily exhibit substantial overlap in theirabilities to activate various SH2-domainsignaling molecules. RIP offers one way toinduce specific signals from different re-ceptors (see the figure). Finally, it will beimperative to identify the nuclear partnersof RIP-derived receptor fragments andtheir target genes.

References1. C.-Y. Ni, M. P. Murphy, T. E. Golde, G. Carpenter, Sci-

ence 294, 2179 (2001).2. M. S. Brown, J. Ye, R. B. Rawson, J. L. Goldstein, Cell

100, 391 (2000).3. H. Steiner, C. Haass, Nature Rev. Mol. Cell Biol. 1, 217

(2000).4. F. Oswald et al., Mol. Cell. Biol. 21, 7761 (2001).5. P. A. Edwards, J. Ericsson, Annu. Rev. Biochem. 68, 157

(1999).6. X. Cao, T. C. Südhof, Science 293, 115 (2001).7. W. T. Kimberly, J. B. Zheng, S. Y. Guenette, D. J. Selkoe,

J. Biol. Chem. 276, 40288 (2001).8. S.-Y. Lin et al., Nature Cell Biol. 3, 802 (2001).

S C I E N C E ’ S C O M P A S S

Biological phenomena are complex,but biologists, being human, cravesimplicity. Hence the frisson of ex-

citement, mixed with relief, with theunion of two hitherto separate domains ofstudy—in this case, the methylation ofDNA and the methylation of histone pro-teins. DNA methylation is a mark on ge-nomic DNA made by addition of methylgroups to cytosine bases, whereas histonemethylation marks proteins that coat theDNA by addition of methyl groups to cer-

tain lysine residues. In their recent Naturepaper, Tamaru and Selker (1) report thatDNA methylation and histone methylationshare a common pathway in the filamen-tous fungus Neurospora crassa. Their dis-covery sets the stage for an experimentalattack on one of the abiding mysteries ofthe genome: How do patterns of DNAmethylation originate?

The problems inherent in managinglarge eukaryotic genomes can be eased bymarking regions of DNA. Such markedDNA becomes structurally adapted so thatit can perform certain activities. The mostdirect mark is one applied to the DNA it-self, but this addition presumably mustavoid adverse effects on the genome’s sta-

bility and coding properties. The onlymarking system that is widespread amongeukaryotes is methylation of DNA cyto-sine rings at position 5 to give the modi-fied base, 5-methylcytosine (m5C). Moreelaborate marking can be achieved bycoating the genome with immobile pro-teins that are specifically designed to car-ry covalent messages. Core histones—theproteins assembled into the beadlike nu-cleosomes around which the DNA iswrapped—can be thought of as informa-tion modules that acquire coded informa-tion based on addition or removal ofchemical groups. Their amino-terminaltails protrude from the nucleosome beadsand can be modified through the attach-ment (or removal) of acetyl, phosphate, ormethyl groups (2).

The resulting “histone code” affects theaccessibility of the packaged DNA to tran-scriptional activator proteins, and hencethe ability to switch on gene expression.For example, the lysine residue K9 in his-tone H3—the ninth amino acid from the

P E R S P E C T I V E S : M O L E C U L A R B I O L O G Y

Methylation Talk Between

Histones and DNAAdrian Bird

The author is at the Wellcome Trust Centre for CellBiology, Institute of Cell and Molecular Biology, Uni-versity of Edinburgh, Edinburgh EH9 3JR, UK. E-mail:a.bird@ed.ac.uk

6564636261605958575655545352515049484746454443424140393837363534333231302928272625242322212019181716151413121110987654321

7 DECEMBER 2001 VOL 294 SCIENCE www.sciencemag.org2114

amino-terminal end of H3—is acetylatedin active chromatin, but methylated in re-gions of the genome that are silenced.When methylated, this lysine (calledH3K9) attracts HP1, a protein associatedwith condensed (silenced) “heterochromat-ic” regions of the genome. Thus, methyla-tion of H3K9 appears to be read by HP1 asa signal to condense that region of theDNA (3–5). Like H3K9 methylation, DNAmethylation too is associated with tran-scriptional silencing and can lead in ani-mals to the recruitment of m5C bindingproteins (6). The resemblances have ledsome to wonder if the two methyltransferreactions might turn out to be connected.There would be a precedent of sorts, be-cause mammalian proteins that bind to ge-nomic m5C sites recruit proteins thatdeacetylate histone tails, leading to tran-scriptional shutdown (7, 8). In that case,though, the flow of “epigenetic” informa-tion is from DNA to histones. The Tamaruand Selker report is exciting because itsuggests the reverse: that histone modifica-tion can influence the methylation of DNA.

In Neurospora, DNA methylation affectsduplicated DNA sequences that were detect-ed early in the organism’s sexual cycle andsubjected to active mutation (9, 10). As a re-sult of this process, known as repeat-in-duced point mutation (RIP), the wild-typeNeurospora genome contains a small frac-tion of methylated DNA, the majority of theDNA remaining nonmethylated. The victimsof RIP are predominantly mobile genetic el-ements that normally move from place toplace within the genome, unless eliminatedby a defense mechanism of this kind. Ourstory begins with a search for mutant formsof Neurospora that lack DNA methylation.Several dim (defective in methylation) genesthat encode DNA methyltransferase en-zymes have already been identified. Tamaruand Selker (1) now show that a mutation ina newly discovered Neurospora gene, dim-5,abolishes methylation of all tested DNA se-quences. Through genetic mapping andDNA sequence analysis, these investigatorsshow that this gene encodes a protein con-taining a so-called SET domain. Intriguing-ly, the SET domain of chromosomal pro-teins found in mammals, insects, and yeastcan be an enzyme that transfers methylgroups to H3K9 (11). By providing the Neu-rospora mutant with the wild-type dim-5gene, the authors were able to correct themutation and to restore methylation.

To look for a connection between dim-5, DNA methylation, and gene silencing,Tamaru and Selker exposed differentstrains of Neurospora to the toxic drug hy-gromycin. Strains engineered to carry anonmethylated hygromycin-resistance gene(hph) grow in the presence of the drug, but

strains with a methylated hph gene cannotgrow because the gene is silenced. Interfer-ing with dim-5 gene expression in strainswith a methylated hph gene causeddemethylation and reactivation of hph,thereby allowing the fungal hyphae to growin the presence of hygromycin. Methyla-tion and silencing clearly were undermined

in the absence of the dim-5 gene, but stillthe question remained: Was methylation ofhistones in the region of dim-5 responsiblefor this effect? Early results already fa-vored the involvement of histone methyla-tion, because the original dim-5 mutationwas located at a highly conserved residueof the catalytic SET domain. Following upon this lead, Tamaru and Selker expressedthe DIM-5 protein and exposed it to corehistones in the presence of labeled S-adenosyl methionine. Only histone H3 be-came methylated, although on whichresidue was not clear. It is important toknow the particular residue, becausemethylation of H3K9 correlates with genesilencing, but methylation of a different ly-sine H3K4 in the same histone signifiestranscriptionally active chromatin (12). Toanswer this question, the investigators cre-ated strains of Neurospora in which K9 inH3 was replaced with amino acids (argi-nine and leucine) that cannot be methylat-ed. Although these strains retained theirnatural H3 genes, they showed marked re-activation and demethylation of the hphgene in the presence of mutant histone H3.Therefore, the presence of H3 lacking a K9residue that could be methylated mimicsthe effect of losing the H3 methyltrans-ferase DIM-5 through mutation. The obvi-ous deduction is that methylation of cyto-sine in Neurospora DNA depends uponprior methylation of K9 in histone H3.

Eukaryotic DNA methylation does notcover the genome like a featureless blanket,but conforms to a pattern of methylatedand nonmethylated DNA sequence do-mains. An attractive feature of the new datais its potential to shed light on the origin ofthese DNA-methylation patterns, a subjectabout which we currently know very little.It is possible that DNA methyltransferasescan only productively access DNA that iswrapped around nucleosomal histones car-rying an H3K9 methylation signal (see thefigure). This scenario is compatible withother known features of H3K9 behavior: itsacetylation in transcriptionally active chro-matin, its deacetylation under the influenceof many gene-expression silencers (tran-scriptional repressors), and its methylationby SET domain proteins such asSU(VAR)3-9 and DIM-5. The notional“descent” from active to profoundly inac-tive chromatin culminates in DNA methy-lation. This would also agree with the pre-vailing evidence (at least in animals) thatcytosine methylation is not a primary agentof gene silencing, but affects genes thathave already been shut down in other ways.

Linear progressions of this kind (seethe figure) are seductive, but should beviewed with suspicion. There is as yet noevidence in Neurospora that the sequences

Nucleosome

DNA

DNMT

AcetylationH3K9

DeacetylationH3K9

MethylationH3K9

MethylationDNA cytosine

Active

Transcription

Inactive

Silenced

Stablysilenced

Methyl groups stick together. Modifications

of the lysine residue K9 in histone protein H3.

Depicted are the effects of deacetylation, his-

tone methylation, and DNA methylation of

H3K9 on gene transcription. The H3 amino-ter-

minal tail protrudes from a nucleosome

(green). H3K9 can be acetylated (pink) in tran-

scribed chromatin, but is deacetylated under

the influence of repressors of transcription. This

permits methylation (orange) of H3K9 by a

SET domain protein. Methylated H3K9 may di-

rectly or indirectly affect enzymes called DNA

methyltransferases (DNMT), leading to

methylation of cytosine on nearby DNA (blue).

DNA methylation prompted by methylation of

H3K9 on a local nucleosome has not yet been

demonstrated. Interaction of methylated H3K9

with chromodomain proteins such as HP1 and

interactions of methylated cytosine residues

with methylated DNA binding proteins are not

shown.

S C I E N C E ’ S C O M P A S S

6564636261605958575655545352515049484746454443424140393837363534333231302928272625242322212019181716151413121110987654321

www.sciencemag.org SCIENCE VOL 294 7 DECEMBER 2001 2115

that become methylated under the influ-ence of methyl-H3K9 are in the same ge-nomic location as those harboring themethylated histone. The effect on DNAmodification might be remote, rather thanlocalized to the methylated nucleosome.Extending the relationship between H3K9methylation and DNA methylation to non-fungal eukaryotes is also premature. DNAmethylation is not a fact of life for manyeukaryotes, including other fungi (yeasts)

and certain animals (the nematode wormCaenorhabditis elegans), so rigorous con-servation of its function cannot be as-sumed. Cautionary notes aside, however,there is little doubt that the search for SETdomain proteins that influence mammalianDNA methylation will now proceed at afrantic pace. Thanks to the awesome powerof Neurospora genetics, there has neverbeen a better time to probe the mysteriousorigins of DNA methylation.

References1. H. Tamaru, E. Selker, Nature 414, 277 (2001).2. T. Jenuwein, C. D. Allis, Science 293, 1074 (2001).3. A. J. Bannister et al., Nature 410, 120 (2001).4. J. Nakayama et al., Science 292, 110 (2001).5. M. Lachner et al., Nature 410, 116 (2000).6. A. Bird, A. P. Wolffe, Cell 99, 451 (1999).7. X. Nan et al., Nature 393, 386 (1998).8. P. L. Jones et al., Nature Genet. 19, 187 (1998).9. J. T. Irelan, E. U. Selker, Genetics 146, 509 (1997).

10. E. U. Selker, B. C. Jensen, G. A. Richardson, Science238, 48 (1987).

11. S. Rea et al., Nature 406, 593 (2000).

12. M. D. Litt et al., Science 293, 2453 (2001).

S C I E N C E ’ S C O M P A S S

The gastrointestinal tract endodermdifferentiates into specialized ep-ithelial cells that perform digestive,

absorptive, protective, and endocrinetasks. These cells have a relatively shortlife-span and must be continuously re-placed from a pool of progenitor cells. Itis still unclear why a progenitor cellmakes a “choice” to leave the progenitorpool and to adopt a specific cell fate. Aninitial decision in epithelial lineage deter-mination involves the Notch signalingpathway, yet subsequent choices and thedownstream targets of this pathway haveyet to be completely identified. On page2155 of this issue, Yang et al. (1) provideevidence that Math1, a basic helix-loop-helix transcription factor, is a “pro-choice” determinant of epithelial cellcommitment, and is a downstream targetof the Notch pathway.

The intestinal epithelium is thrown intofingerlike folds called villi, which are sepa-rated from each other by troughs calledcrypts (see the figure). The differentiatedepithelial cells at the tips of the villi are re-placed every few days by progenitor stemcells that dwell in the crypts and move upthe villi as they differentiate (2). In thecrypts, the progenitor stem cells succumb tolineage determination and eventually giverise to four types of differentiated gut ep-ithelial cells—enterocytes, Paneth cells,goblet cells, and enteroendocrine cells.These initial differentiation events are af-fected by the position of the epithelial cellalong the crypt-to-villus axis and by its in-teractions with neighboring cells (3). It is

not clear how progenitor stem cells in thecrypt become lineage restricted. But, justbefore they leave the crypt, these lineage-restricted undifferentiated cells undergo aswitch, withdraw from the cell cycle, andbegin to express specialized proteins. Thedifferentiated cells then migrate up the villi

and, after reaching the top, undergoapoptosis, thereby maintaininghomeostasis of the intestinal ep-ithelium. Coordination betweenproliferation, differentiation, andapoptosis requires the well-timedinterplay of different signalingpathways.

The Notch pathway specifiescell fate through feedback amplifi-cation of relative differences incellular levels of Notch and its li-gand Delta (4). This results in sub-sets of cells that produce largeamounts of Notch. These cells in-duce expression of transcriptionfactors, such as Hes1 (5). Hes1 is atranscriptional repressor, andtherefore cells that express theHes1 gene remain precursor cells(6). Yang et al. further dissect thispathway by showing that Math1 isa downstream target of Hes1 andcontrols the initial choice of fatemade by crypt progenitor stemcells (see the figure).

Math1 is expressed in devel-oping and mature mouse intesti-nal epithelium. Yang and col-leagues used reporter constructsto show that Math1-def icientmice have increased expression ofthe reporter gene in crypt cells.However, they lack goblet, Paneth,and enteroendocrine cells, andshow no increase in the pro-grammed cell death of cells at thetips of the villi. These findingssuggest that Math1 is involved inepithelial cell fate decisions. Theauthors propose that Math1 ex-pression is needed for cells tomake the first lineage-specifyingchoice, that is, to adopt one of thefollowing three fates: Paneth,

P E R S P E C T I V E S : D E V E L O P M E N T

Epithelial Cell Differentiation—

a Mather of ChoiceGijs R. van den Brink, Pascal de Santa Barbara, Drucilla J. Roberts

G. R. van den Brink is at the Academic Medical Cen-ter, Department of Experimental Internal Medicine inthe Netherlands. D. J. Roberts and P. de Santa Bar-bara are in the Department of Pathology, Mas-sachusetts General Hospital, Boston, MA 02114,USA. E-mail: djroberts@partners.org

Vi

Cr

Notch high

No choice Choice

Delta high

Hes-1

ProsecretoryMath1

Hes-1

ProsecretoryMath1

?

A

B

X

A gut instinct about cell fate. (A) Low-power 4-µm sec-

tion of adult murine small intestine. Precursor cells (brown)

are stained for cyclin PCNA (proliferating cell nuclear anti-

gen); enterocytes (red) express intestinal alkaline phos-

phatase (IAP); goblet cells (blue) secrete mucins. Inset

shows high-power image of small intestine enterocytes and

goblet cells. (B) Math1, a component of the Notch signaling

pathway, influences intestinal epithelial cell fate decisions.

In crypt progenitor stem cells that express high levels of

Notch, the Hes1 transcription factor is switched on and the

expression of Math1 and of other “prosecretory” genes is

blocked. The result is that the precursor cells become ente-

rocytes. In cells expressing low amounts of Notch, levels of

Delta are high, production of Hes1 is blocked, and Math1

expression is induced. Production of the Math1 helix-loop-

helix transcription factor allows precursor cells to make a

choice: whether to become goblet cells, Paneth cells, or en-

teroendocrine cells (Vi, villus; Cr, crypt).