Zhang Meaney 10 AnnRPsychol Epigenetics Review

8/11/2019 Zhang Meaney 10 AnnRPsychol Epigenetics Review

1/34

Epigenetics andthe EnvironmentalRegulation of the Genomeand Its Function

Tie-Yuan Zhang and Michael J. Meaney

Sackler Program for Epigenetics and Psychobiology of McGill University, Douglas Ment

Health University Institute and the Singapore Institute for Clinical Sciences, Montreal,Quebec, H4H 1R3 Canada; email: [email protected]

Annu. Rev. Psychol. 2010. 61:43966

TheAnnual Review of Psychologyis online atpsych.annualreviews.org

This articles doi:10.1146/annurev.psych.60.110707.163625

Copyright c2010 by Annual Reviews.All rights reserved

0066-4308/10/0110-0439$20.00

Key Words

maternal care, stress responses, DNA methylation, gene x

environment interactions, glucocorticoid receptor

Abstract

There are numerous examples in psychology and other disciplines o

the enduring effects of early experience on neural function. In this article, we review the emerging evidence for epigenetics as a candidate

mechanism for these effects. Epigenetics refers to functionally relevanmodifications to the genome that do not involve a change in nucleotid

sequence. Such modifications include chemical marks that regulate th

transcription of the genome. There is now evidence that environmental events can directly modify the epigenetic state of the genome. Thu

studies with rodent models suggest that during both early develop

ment and in adult life, environmental signals can activate intracellulapathways that directly remodel the epigenome, leading to changes ingene expression and neural function. These studies define a biologica

basis for the interplay between environmental signals and the genomin the regulation of individual differences in behavior, cognition, an

physiology.

439


2/34

Phenotype: any

observablecharacteristic or traitof an organism, such asits morphology,development,biochemical orphysiologicalproperties, or behavior

Contents

INTRODUCTION .. . . . . . . . . . . . . . . . . 440GENE TRANSCRIPTION .. . . . . . . . . 441

Chromatin Modifications. . . . . . . . . . . 442Regulation of Glucocorticoid

Receptor Expression . . . . . . . . . . . . 443

ENVIRONMENTALPROGRAMMING OF

GENE EXPRESSION..... . . . . . . . . 444EPIGENETIC REGULATION

OF THE GENOME . . . . . . . . . . . . . . 446Epigenetics and the

Social Environment . . . . . . . . . . . . . 449THE FUNCTIONAL

IMPORTANCE OF THESOCIAL IMPRINT . . . . . . . . . . . . . . . 451

ACTIVITY-DEPENDENT

REGULATION OFTHE EPIGENOME . . . . . . . . . . . . . . 452Summary (and Perhaps

Some Constraints) . . . . . . . . . . . . . . 455

EPIGENETICS ANDMENTAL HEALTH . . . . . . . . . . . . . . 456

CONCLUSIONS . . . . . . . . . . . . . . . . . . . . 458

INTRODUCTION

Psychology has seen a major transition in

personality theory. Personality traits were oncethought to emerge under the dominion of

influences associated with nurture. The post-natal family environment was considered as the

primary candidate force in the developmentof individual differences in personality. This

perspective changed dramatically in responseto integration of the biological sciences into

personality psychology. First, evolutionary ap-proaches established the idea that the brain and

its development, like any other organ, are sub-

ject to evolutionary forces. Second, behavioralgenetics (Ebstein 2006, Kendler 2001, Plomin

& Rutter 1998) provided evidence for a relationbetween variation at the level of the genome

and that in personality and mental health.Although efforts to quantify the independent

contribution of genetic and environmental

influences are fraught with complications (e

geneenvironment interactions, nongenommechanisms of inheritance), measures of co

cordance in specific traitsbetween monozygo

and dizygotic twins, among other approachsuggest a pervasive influence of genetic var

tion. Indeed, it is impossible to imagine thatfunction of brain cells could occur independ

of variations in the genes that encode proteins that regulate neuronal functions.

Genomic variation at the level of nucleotsequence is associated with individual diff

ences in personality and thus with vulnerabity and resistance to a wide range of chro

illness (Ebstein 2006, Meyer-LindenbergWeinberger 2006, Rutter 2007). Such var

tions can take multiple forms, including vaation at the level of (a) a single nucleotide (i

single-nucleotide polymorphisms or SNP

(b) variation in the number of nucleotide peats (i.e., variable number of tandem repeat

VNTRs), or (c) chromosomal reorganizatiEach form of variation can potentially alter

nomic function and thus phenotype. The chlenge for psychology is that of conceptually

tegrating findings from genetics into the stuof personality and our understanding of t

pathophysiology of mental illness. Howand uder what conditions does genomic variation

fluence brain development and function? Hmight relevant findings from the field of g

netics influence the design of public policy a

therapies in psychology?It is important to note the simple f

that genes encode for protein, not functiThus, as described below, the effects of gene

variation are contextually determined and bconsidered as probabilistic. Cellular functi

can only be understood in terms of the constdialogue that occurs between the genome a

its environment. The environment regulathe cellular signals that control the operat

of the genome. The objective of this reviis to describe recent advances in molecu

biology, notably in the field of epigenetics, a

to suggest that epigenetic mechanisms areideal candidate mechanism for the effects

environmental signals, including events su

440 Zhang Meaney


3/34

as social interactions, on the structure and

function of the genome (Harper 2005). Theintent is to first consider the processes by which

cellular signals, referred to as transcription

factors, regulate the activity (or expression)of a gene. The biological primacy of gene

environment interactions is apparent from thesimple realization that the levels and the activ-

ity of these transcription factors is controlledby environmental signals. Thus the operation

of the genome is dependent upon context. Thequestion concerns the mechanisms responsible

for such contextual influences. We suggestthat epigenetics is one such process and can

account, in part, for instances in which envi-ronmental events occurring at any time over

the lifespan exert a sustained effect on genomic

function and phenotype.Epigenetics signals refer to a series of chem-

ical modifications to the DNA or to regionssurrounding the DNA. The transcriptional

activity of the genome is regulated by signals,transcription factors, that physically bind to

specific DNA sites. The importance of epige-netics mechanisms lies in the ability to regulate

the ease with which transcription factors canaccess the DNA. Epigenetic signals can thus

determine the capacity for environmentalregulation of the genome. There is emerging

evidence for the idea that epigenetic marks are

directly altered in early life by environmentalevents and thus influence the development

of individual differences in specific neuralfunctions that underlie cognition and emotion.

More recent studies suggest that dynamicalterations in these same epigenetic signals are

crucial for the synaptic remodeling that medi-ates learning and memory. Thus, epigenetics

provides a remarkable insight into the biologythat governs the function of the genome in

response to environmental signals.

GENE TRANSCRIPTION

The most compelling evidence for the pre-

dominance of geneenvironment interactionsin cellular function emerges from the study of

gene transcription [Gilbert (2006) provides a

Gene transcriptionthe synthesis of RNAunder the direction DNA. RNA synthesis the process oftranscribing DNA

nucleotide sequenceinformation into RNsequence informatio

Receptor: a proteinembedded in eitherthe plasma membranor cytoplasm of a ceto which a mobilesignaling molecule(ligand) may bind. Tsignaling molecule cbe a peptide, ahormone, a

pharmaceutical drugor a toxin, and whensuch binding occursthe receptor goes inta conformationalchange that ordinariinitiates a cellularresponse

Chromatin: thecombination of DNARNA, and protein thmakes upchromosomes. The

major components ochromatin are DNAand histone proteins

very clear and well-illustrated description]. The

transcription of the genome is a highly regu-lated event. At the heart of this process lies a

class of proteins referred to as transcription fac-

tors. As the name implies, these proteins havethe ability to bind to regulatory elements of the

gene and to activate or repress gene transcrip-tion. Importantly, the expression and activation

of the transcription factors themselves are dy-namically regulated by environmental signals.

Many of the earliest cellular responses to envi-ronmental stimuli involve either the activation

of pre-existing transcriptional signals throughchemical modifications such as phosphoryla-

tion (i.e., the addition of a phosphate) of spe-cific amino acids of the protein, or an increase

in gene expression that results in the rapid syn-

thesis of proteins (e.g., immediate early geneproducts) that then serve to regulate the ac-

tivity of other genes. This includes genes thatare involved in synaptic plasticity. The bind-

ing of transcription factors to DNA sites is thebiological machinery for the dynamic gene

environment interactions that result in alteredrates of gene transcription.

Figure 1 portrays the organization of theglucocorticoid receptor gene as an example of

genomic organization and a target for discus-sion below. The schema is actually somewhat

misleading. For reasons of graphic simplicity,

we often describe the organization of a geneor the interactions between transcription

factors and DNA as if the DNA were a linearmolecule to which transcription factors gain

unimpeded access. The reality of proteinDNA interactions is very different. Figure 2

presents the classic crystallographic analysis ofthe organization of DNA (Luger et al. 1997).

DNA is organized into units referred to asnucleosomes, each of which contains about

145150 base pairs wrapped around a coreregion of histone proteins (Turner 2001). The

histones and DNA together are referred to as

chromatin; the nucleosome is the organizationof chromatin. Under normal conditions there

is a tight physical relation between the histoneproteins and its accompanying DNA, resulting

in a rather closed nucleosome configuration.

www.annualreviews.org Epigenetics and the Environment 441


4/34

..tggg16cggggg17cgggag.

11 14 15 1617 18 19 110 111 2

NGFI-A

GR Promoter 17 Sequence

9

Regulatory (non-coding) region Coding region

Figure 1

A schema describing the organization of the rat glucocorticoid receptor gene including 9 exon regions.Exons 29 participate in the coding for the glucocorticoid receptor protein. Exon 1 is composed of multipregulatory regions, each of which is capable of activating gene transcription (i.e., promoter sequences). Tactivity of the various exon 1 promoters is tissue-specific, with evidence suggesting that certain promotersare more active in areas such as liver or thymus, and others are more active in brain (e.g., exon 1 7; based oMcCormick et al. 2000; see Turner & Muller 2005 for comparable data in humans). The use of multiplepromoters permits regulation in one tissue independently from other regions (i.e., increased glucocorticoreceptor in pulmonary tissues prior to birth that is necessary for respiratory competency at parturition, whmaintaining reduced glucocorticoid receptor levels in brain, where glucocorticoid effects might inhibitneurogenesis). The consensus binding site for nerve growth factorinducible factor A (NGFI-A) lying witthe exon 17promoter is highlighted. The reader should note that this organization is not necessarily typicRegulatory elements (promoters or enhancers) can exist between exons (i.e., within introns) or at sites thaare either 5 or 3 to the coding region, sometimes at considerable distances.

Enzyme: a molecule,usually protein, thatcatalyzes (i.e., increasesthe rate of ) a specificchemical reaction

Histone deacetylases(HDACs): a class ofenzymes that removeacetyl groups from an-N-acetyl lysineamino acid on a

histone. The action ofHDACs is thus theopposite to that ofhistoneacetyltransferases, andHDACs are associatedwith transcriptionalsilencing

This restrictive configuration is maintained,

in part, by electrostatic bonds between thepositively charged histones and the negatively

charged DNA. The closed configurationimpedes transcription factor binding and

is associated with a reduced level of geneexpression. An increase in transcription factor

binding to DNA and the subsequent activationof gene expression commonly requires chem-

ical modification of the chromatin that occurson thehistone proteins. The primary targets for

such events are the amino acids that form thehistone tails extending from the nucleosome

(Figure 2). These modifications alter chro-matin in a manner that either increases or

decreases the ability of transcription factorsto access regulatory sites on the DNA thatcontrol gene transcription.

Chromatin Modifications

The dynamic alteration of chromatin structure

is achieved through modifications to thehistone

proteins at the amino acids that form the

stone protein tails that extend out from tnucleosome (Figure 2). These modificatio

are achieved through a series of enzymes tbind to the histone tails and modify the

cal chemical properties of specific amino ac(Grunstein 1997, Hake & Allis 2006, Jenuw

& Allis 2001). For example, the enzyme histoacetyltransferase transfers an acetyl group on

specific lysines on the histone tails. The adtion of the acetyl group diminishes the posit

charge, loosening the relation between the stones and DNA, opening the chromatin a

improving the ability of transcription factorsaccess DNA sites. Thus, histone acetylation

specific lysine sites is commonly associatedwactive gene transcription.

The functional antagonists of the histo

acetyltransferases are a class of enzymes knoas histone deacetylases (HDACs). These e

zymes remove acetyl groups and prevent fther acetylation, thus maintaining a clo

442 Zhang Meaney


5/34

chromatin structure, decreasing transcription

factor and gene expression. Both the acety-lation and deacetylation of histones are dy-

namic processes that are regulated by environ-

mental signals. Indeed, a number of proteinsthat were known to be associated with tran-

scriptional activation (e.g., transcriptional co-factors) have been identified as histone acetyl-

transferases. These factors enhance the efficacyof transcription factors by opening chromatin

and thus increasing the binding of the factor tothe regulatory regions of the gene.

The reader should note that there are ac-tually multiple modifications to histone tails,

including methylation (in this case on the hi-stones), phosphorylation, and ubiquitination.

For the sake of simplicity, the discussion is lim-

ited to histone acetylation/deacetylation.

Regulation of GlucocorticoidReceptor Expression

The neurotransmitter serotonin (5-

hydroxytryptamine; 5-HT) regulates glu-cocorticoid receptor gene transcription in

hippocampal neurons (Figure 3; Mitchell et al.1990, 1992; Weaver et al. 2007). This effect

is dependent upon the binding of the tran-

scription factor nerve growth factorinduciblefactor A (NGFI-A) to a specific binding site

on the exon 17 glucocorticoid (GR) promoter(Figure 1). The importance of this interaction

can be precisely defined. For example, mutatinga single nucleotide, in this case simply exchang-

ing a cytosine for an adenine, in the regionof the promoter that normally binds NGFI-A

abolishes the ability of NGFI-A to associatewith the exon 17 promoter and eliminates

the effect of NGFI-A on gene transcription(Weaver et al. 2007). However, the ability of

NGFI-A to bind to the exon 17 promoter is

regulated by another protein, a transcriptionalcofactor, the CREB-binding protein, that is

activated by the same 5-HT-cyclic adenosinemonophosphate (cAMP)/cyclic nucleotide

dependent kinases (PKA)-signaling cascadethat results in the increased levels of NGFI-A

(Figure 3). The CREB-binding protein is a

Histoneacetyltransferases:enzymes that acetylalysine amino acids ohistone proteins bytransferring an acety

group from acetyl Cto form -N-acetyllysine. Histoneacetylation isassociated with theactivation of genetranscription

Promoter: a regionDNA that facilitatesthe transcription of particular gene.Promoters aretypically located nea

the genes theyregulate, on the samstrand and upstreamfrom the coding regi

histone acetyltransferase. The association of

the CREB-binding protein with the exon 17promoter is accompanied by an increase in

the acetylation of a specific lysine on the tail

of histone 3 of the exon 17 promoter (Weaveret al. 2004, 2007). Thus, 5-HT activates both

NGFI-A and the CREB-binding protein.Interestingly, NGFI-A and the CREBbinding

protein physically associate with one anotherprior to DNA binding. The CREBbinding

protein acetylates histones associated withthe exon 17 promoter, enhancing the abil-

ity of NGFI-A to bind and activate genetranscription.

Environmental signals alter 5-HT activity.Indeed, the effect of 5-HT on glucocorticoid

receptor expression reflects the dependency

of gene transcription on signals that derivefrom environmental events (note that the

relevant environmental event may be internalor external to the organism; e.g., a change

in the availability of glucose, an electricalimpulse, or a social interaction). Such effects

underlie the dynamic interdependence of geneand environment. However, psychologists, and

in particular developmental psychologists, arefamiliar with more enduring environmental

influences; instances where experience inearly life has shaped neural development and

function in a manner that is sustained into

adulthood. Such effects are considered as thebasis for environmental influences over the de-

velopment of individual differences. In certaincases, the sustained effects of early experience

have been associated with structural alterationsin neural circuits that mediate specific func-

tions. The process of sexual differentiationamong vertebrates provides excellent examples

where environmental signals lead to differencesin morphology and thus to gender. However,

more recent studies suggest another form ofenvironmentally regulated plasticity that exists

at the level of genome itself. Such effects ap-

pear to involve the modification of epigeneticmarks on the DNA. These studies suggest

that environmental events alter the activity ofspecific intracellular signals that modify the

nature of the epigenetic marks at specific sites



6/34

in the genome, leading to sustained effects on

gene expression and thus neural function.

ENVIRONMENTALPROGRAMMING OFGENE EXPRESSION

Studies in developmental psychobiology and

physiology are replete with examples of theenvironmental programming of gene expres-

sion. Such studies commonly report that a vari-ation in the early environment associates with

changes in gene expression and biological func-tion that persists into adulthood and thus well

beyond the duration of the relevant environ-mental event. In the rat, for example, prena-

tal nutrient deprivation or enhanced exposure

to hormonal signals associated with stress sta-bly alter, or program, the activity of genes in

the liver and other sites that are associatedwith glucose and fat metabolism, including the

gene for the glucocorticoid receptor (Batesonet al. 2004; Gluckman & Hanson 2004, 2007;

Jirtle & Skinner 2007; Meaney et al. 2007;Seckl & Holmes 2007). These findings are as-

sumed to represent instances in which the op-eration of a genomic region in adulthood varies

as a function of early environmental influences.The results of recent studies suggest that such

programming effects can derive from gene

environment interactions in early life that leadto a structural alteration of the DNA, which

in turn mediates the effects on gene expressionas well as more complex levels of phenotype

( Jirtle & Skinner 2007, Meaney 2007, Meaney& Szyf 2005). These studies were performed

in rodents but were inspired by the vast lit-erature reporting the pervasive effects of fam-

ily environment on health outcomes in humans(Repetti et al. 2002). No less compelling are the

results of studies on maternaleffects in plants,insects, reptiles, and birds showing that varia-

tions in nongenomic signals of maternal ori-

gin associate with sustained effects on the phe-notype of the offspring (Cameron et al. 2005,

Mousseau & Fox 1998, Rossiter 1998).The objective of these studies is to examine

the biological mechanisms whereby variations

in motherinfant interactions might directly

fluence gene expression and behavior (Mean2001). Such studies focus on variations

maternal behavior that lie within the normrange for the species, in this case the Norw

rat, and that occur in the absence of a

experimental manipulations (i.e., naturally curring variations in motherpup interaction

Variations on maternal care in the rat are stuied with simple, albeit very time-consumi

observations on animals in their home ca(Champagne 2008, Champagne et al. 200

One behavior, pup licking/grooming (Lemerges as highly variable across mothers. P

LG is a major source of tactile stimulatifor the neonatal rat that regulates endocr

and cardiovascular function in the pup (Ho2005, Levine 1994, Schanberg et al. 1984). T

question then was whether such variations

pup LG might directly alter the developmof individual differences in behavior a

physiology. For the studies reviewed hethe focus is on the development of individ

differences in defensive responses.Subsequent findings revealed considera

evidence for the effect of maternal care the behavioral and endocrine responses

stress in the offspring. The male or femadult offspring of mothers that natur

exhibit increased levels of pup LG (i.e., toffspring of high-LG mothers) show m

modest behavioral and endocrine responses

stress compared to animals reared by low-Lmothers (Caldji et al. 1998, Francis et al. 19

Liu et al. 1997, Menard et al. 2004, Toki et2007, Weaver et al. 2004). Specifically,

offspring of high-LG mothers show reducfearfulness and more modest hypothalam

pituitary-adrenal (HPA) responses to strCross-fostering studies, where pups born

high-LG mothers are fostered at birth low-LG mothers (and vice versa), revea

direct relationship between maternal cand the postnatal development of individ

differences in behavioral and HPA respon

to stress (Caldji et al. 2000, 2003; Francis et1999; Weaver et al. 2004). In these studies,

rearing mother determined the phenotype

444 Zhang Meaney


7/34

Hypothalamus

Pituitary

Adrenals

Hippocampus Hippocampus

Glucocorticoids

() ()

()

()

()

()

()

CRF

ACTH

()

Hypothalamus

Pituitary

AdrenalsGlucocorticoids

()

CRF

ACTH

()

Figure 4

A schema outlining the function of the hypothalamic-pituitary-adrenal axis, the nexus of which are the

corticotropin-releasing factor (CRF) neurons of the paraventricular nucleus of the hypothalamus. CRF isreleased into the portal system of the anterior pituitary, stimulating the synthesis and release ofadrenocorticotropin (ACTH), which then stimulates adrenal glucocorticoid release. Glucocorticoids act onglucocorticoid receptors in multiple brain regions, including the hippocampus, to inhibit the synthesis andrelease of CRF (i.e., glucocorticoid negative feedback). The adult offspring of high-LG mothers, bycomparison to those of low-LG dams, show (a) increased glucocorticoid receptor expression, (b) enhancednegative-feedback sensitivity to glucocorticoids, (c) reduced CRF expression in the hypothalamus, and(d) more modest pituitary-adrenal responses to stress.

the offspring. Thus variations within a normal

range of parental care can dramatically alterphenotypic development in the rat.

The effects of maternal care on the devel-opment of defensive responses to stress in the

rat involve alterations in the function of thecorticotrophin-releasing factor (CRF) systems

in selected brain regions (Figure 4). The CRF

system furnishes the critical signal for the acti-vation of behavioral, emotional, autonomic, and

endocrine responses to stressors (Bale & Vale2004, Koob et al. 1994, Plotsky et al. 1989). As

adults, the offspring of high-LG mothers showdecreased CRF expression in the hypothalamus

as well as reduced plasma ACTH and gluco-

corticoid responses to acute stress by compar-ison to the adult offspring of low-LG mothers(Francis et al. 1999; Liu et al. 1997; Weaveret al.

2004, 2005). Circulating glucocorticoids act atglucocorticoid receptor sites in corticolimbic

structures, such as thehippocampus, to regulate

HPA activity (Figure 4). Such feedback effects

commonly inhibit hypothalamic CRF expres-

sion. The high-LG offspring showed signif-icantly increased hippocampal glucocorticoid

receptor expression, enhanced glucocorticoidnegative feedback sensitivity, and decreased hy-

pothalamic CRF levels. Indeed, the magnitudeof the glucocorticoid response to acute stress

is significantly correlated with the frequency of

pup LG during the first week of life, as is thelevel of both hippocampal glucocorticoid re-

ceptor andhypothalamicCRF expression (all rs>0.70; Liu et al. 1997). Importantly, pharmaco-

logicalmanipulationsthatblocktheeffectoftheglucocorticoid receptor eliminate the maternal

effect on the HPA response to stress, suggesting

that the differences in hippocampal glucocorti-coid receptor expression are directly related tothose at the level of HPA function.

Pup LG is a major source of tactile stimula-tion for the neonate. Experimental models that

directly apply tactile stimulation, through the

stroking of the pup with a brush, provide direct



8/34

evidence for the importance of tactile stimu-

lation derived from pup LG. Thus, strokingpups over the first week of life increases hip-

pocampal glucocorticoid receptor expression

( Jutapakdeegul et al. 2003) and dampens be-havioral and HPA responses to stress (Bur-

ton et al. 2007, Gonzalez et al. 2001). Like-wise, manipulations of lactating mothers that

directly increase the frequency of pup LGalso increase hippocampal glucocorticoid re-

ceptor expression and decrease HPA responsesto stress (Francis et al. 1999, Toki et al. 2007).

Manipulations, notably stressors imposed onthemother, that decrease pupLG areassociated

with increased behavioral and HPA responsesto stress and are associated with decreased hip-

pocampal glucocorticoid receptor expression

and increased hypothalamic expression of CRF(Champagne & Meaney 2006, Fenoglio et al.

2005).The offspring of the high-LG and low-LG

mothers also differ in behavioral responses tonovelty (Caldji et al. 1998, Francis et al. 1999,

Zhang et al. 2004). As adults, the offspringof the high-LG mothers show decreased star-

tle responses, increased open-field exploration,and shorter latencies to eat food provided in

a novel environment. There are also behav-ioral differences in response to more precise

forms of threat. Thus, the offspring of low-

LG mothers show greater burying of an elec-trified probe in the defensive burying paradigm

(Menard et al. 2004), which involves an activeresponse to a threat. These differences in be-

havioral responses to stress are associated withaltered activity in the CRF system that links

the amygdala (and bed nucleus of the stria ter-minalis) to the noradrenergic cells of the locus

coeruleus (Caldji et al. 1998, Zhang et al. 2004).The results of these studies suggest that the

behavior of the mother toward her offspringcan program stable changes in gene expres-

sion that then serve as the basis for individual

differences in behavioral and neuroendocrineresponses to stress in adulthood. The maternal

effects on phenotype are associated with sus-tained changes in the expression of genes in

brain regions that mediate responses to stress

and form the basis for stable individual diff

ences in stress reactivity. These findings pvide a potential mechanism for the influen

of parental care on vulnerability/resistancestress-induced illness over the lifespan. But

critical issue is simply that of how mater

care might stably affect gene expression. Hare variations in the social interactions betw

the mother and her offspring biologically ebedded so as to stably alter the activity of s

cific regions of the genome? The answersthese questions appear to involve the abil

of social interactions in early developmentstructurally modify relevant genomic regio

For the sake of this review, we focus on the mternal effect on the regulation of hippocam

glucocorticoid receptor expression.

EPIGENETIC REGULATIONOF THE GENOME

The molecular processes that lead to the inittion of gene transcription involve modificatio

to the histone proteins that form the core ofnucleosome (Figure 2). Such modificati

open chromatin, permitting transcriptfactor binding and the activation of g

transcription. A second level of regulat

occurs not on the histone proteins, but rathdirectly on the DNA. Indeed, the classic e

genetic alteration is that of DNA methylatiwhich involves the addition of a methyl gro

(CH3) onto cytosines in the DNA (Bird 19Holliday 1989, Razin & Riggs 1980). DN

methylation is associated with the silencof gene transcription. This effect appears

be mediated in one of two ways (Bird 200First, wide swaths of densely methylated DN

preclude transcription factor binding to DNsites, thus silencing gene expression. The s

ond manner is subtler and probably far mo

prevalent in regions with more dynamic vaations in gene transcription, such as the bra

In this case, selected cytosines are methylatand the presence of the methyl group attra

a class of proteins know as methylated-DNbinding proteins (Klose & Bird 2007). Th

proteins, in turn, attract an entire cluster

446 Zhang Meaney


9/34

proteins that form a repressor complex, which

includes active mediators of gene silencing.The HDACs are a critical component of the

repressor complex. HDACs prevent histone

acetylation and favor a closed chromatin statethat constrains transcription factor binding

and gene expression (Figure 2and see above).Compounds that inhibit HDACs can thus

increase transcription from methylated DNA.When we think of genomic influences, we

most commonly imagine effects associated withvariation in nucleotide sequence. Yet this is only

one form of information contained within thegenome. Despite the reverence afforded DNA,

a gene is basically like any other molecule inthe cell; it is subject to physical modifications.

As described above, these modifications alter

the structure and chemical properties of theDNA and thus gene expression. Collectively,

themodifications to the DNA andits chromatinenvironment can be considered as an additional

layer of information that is contained within thegenome. This information is thus epigenetic in

nature (the name derives from the Greek epimeaning upon and genetics). The acetylation

of histone proteins or the methylation of DNAare examples of epigenetic modifications. Epi-

genetic modifications do not alter the sequencecomposition of the genome. Instead, these epi-

genetic marks on the DNA and the histone pro-

teins of the chromatin regulate the operation ofthe genome. Thus, epigenetics has been de-

fined as a functional modification to the DNAthat does not involve an alteration of sequence

(Waddington 1957). Although this definitionhas been subjected to revision (Bird 2007, Hake

& Allis 2006), the essential features of epige-netic mechanisms are (a) structural modifica-

tions to chromatin either at the level of the hi-stone proteins (Figure 2) or the DNA, (b) the

associated regulation of the structure and func-tion of chromatin, (c) the downstream effects

on gene expression, and (d) the fact that these

effects occur in the absence of any change innucleotide sequence.

The methylation of DNA in mammals is anactive biochemical modification that selectively

targets cytosines and is achieved through the

actions of enzymes, DNA methyltransferases,

that transfer the methyl groups from methyldonors. There are two critical features to DNA

methylation: First, it is a stable chemical modi-fication, and second, it is associated with the si-

lencing of gene transcription (Bestor 1998, Bird

2002, Bird & Wolffe 1999, Razin 1998).Until very recently, it wasthought that DNA

methylation patterns on the genome were over-laid upon the genome only during early peri-

ods in embryonic development. Indeed, DNAmethylation is considered as a fundamental

feature of cell differentiation. It is importantto consider a simple feature of cell biology:

All cells in the body generally share the sameDNA. Thus, the processes of cell specializa-

tion, whereby liver cells specialize in functionsrelated to energy metabolism and brain cells

establish the capacity for learning and memory,

involve silencing certain regions of the genomein a manner that is specific for each cell type.

Genes associated with gluconeogenesis are si-lenced in brain cells but remain active in liver

cells. Such processes define the function of thecell type (e.g., Fan et al. 2005). DNA methyla-

tion is considered as a mechanism for the ge-nomic silencing that underlies cell specializa-

tion. Such events occur early in developmentand are considered to be highly stable,such that

dedifferentiation (whereby a cell loses its spe-cialization) is rare and often is associated with

organ dysfunction.Thus DNA methylation was considered

both unique to early periods in develop-

ment and irreversible. Experimental modelscommonly used to study DNA methyla-

tion further reinforced this view. DNAmethylation-induced gene silencing mediates

two of the most commonly studied examplesof the epigenetic silencing of genes, namely

X-chromosome inactivation and gene im-printing. Mammalian females bear two copies

of the X-chromosome. The inactivation ofone copy of the X-chromosome occurs in

all mammalian females and is essential fornormal function (i.e., maintaining a constant

gene dosage in males and females). The

silencing of the X-chromosome is associated



10/34

with DNA methylation (Mohandas et al. 1981,

Riggs & Pfieffer 1992; but also see Hellman& Chess 2007 for a more current update).

The second example of epigenetic-mediated

gene silencing is that of gene imprinting (daRocha & Ferguson-Smith 2004, Reik 2001),

a remarkable subject in its own right and onewith considerable implications for growth and

development (Charalambous et al. 2007). Forhumans and other mammals, the expression of

specific genes is determined by the parent oforigin. For certaingenes, the copy derived from

the mother is active while that emanating fromthe father is silenceda maternally imprinted

gene. In other cases, it is the reverse: The copyof the gene inherited from the father is active

while that from the mother is silenceda

paternally imprinted gene. The silent copyis methylated in DNA regions that regulate

gene expression and thus is inactive. Again,the epigenetic marks associated with gene

imprinting are established very early in life.These marks, as well as those associated with

X-chromosome inactivation, are largely stable.Collectively, these models have left biolo-

gists with the impression that under normalconditions, DNA methylation occurs early in

embryonic life and is irreversible. DNA methy-lation was considered to be an actively dy-

namic process only during periods of cell di-

vision and differentiation (see above) such thatin mature postmitotic cells, further alteration

of methylation patterns was improbable. More-over, the extensive loss of cytosine methylation

in the models described above is associated withpathology. This perspective was further rein-

forced by findings showing that an alterationof DNA methylation at critical genomic tar-

gets (i.e., tumor suppressors) is associated withcancer (Eden et al. 2003, Feinberg 2007, Laird

2005).At this point, dynamic changes in DNA

methylation were of considerable interest for

developmental biologists but somewhat less sofor psychologists, who study the aftermath of

more subtle variations in neuronal differentia-tion that occur in later periods of development

or even in the fully mature brain. The issue for

developmental psychologists concerns less

process by which cells specialize as neuroand more the issues related to why neurons

one individual function differently from thof another, or how neurons might dynamica

later alter functional properties in relation

experience (i.e., activity-dependent neuroplasticity). The studies reviewed below prov

an importantrevision to this perspective. This now considerable evidence in neuroscien

and other fields, including immunology aendocrinology/metabolism, that the state

DNA methylation at specific genomic siteindeed dynamic even in adult animals (B

2007, Jirtle & Skinner 2007, Meaney & S2005). Moreover, alterations in DNA methy

tion are emerging as a candidate mechanismthe effects of early experience in individual d

ferences in neural function as well as in learn

and memory. Thus, although the assumptioconcerning DNA methylation appear valid

the examples cited above, recent studies revthat DNA methylation patterns are activ

modified in mature (i.e., fully differentiatcells including, and perhaps especially, n

rons, and that such modifications can occuranimals in response to cellular signals driv

by environmental events ( Jirtle & Skin2007, Meaney & Szyf 2005, Sweatt 2009). F

example, variations in the diet of mice durgestation or later in development, such as

early postweaning period, can stably alter methylation status of the DNA (Cooney et

2002, Waterland & Jirtle 2003, Waterla

et al. 2006, Whitelaw & Whitelaw 200Likewise, both mature lymphocytes (Bruniq

& Schwartz 2003, Murayama et al. 2006) aneurons (e.g., Champagne 2008, Champag

et al. 2006, Lubin et al. 2008, Martinowet al. 2003, Sweatt 2009) show changes in t

DNA methylation patterns at critical genomregions in response to environmental stim

that stably alter cellular function. The abiof environmental signals to actively remo

epigenetic marks that regulate gene expressis a rather radical change in our understand

of the environmental regulationof gene expr

sion. Such epigenetic modifications are th

448 Zhang Meaney


11/34

a candidate mechanism for the environmental

programming of gene expression.

Epigenetics and theSocial Environment

The section below describes studies of themolecular basis for the effects of maternal care

on the development of individual differencesin gene expression and stress responses. The

mechanism for this interaction is epigenetic, in-volving alterations in DNA methylation at spe-

cific sites in the genome. In summary, variationsin motherinfant interactions in theratalter the

extra- and intracellular environment of neuronsin selected brain regions. Such alterations di-

rectly modify the epigenetic marks on regions

of the DNA that regulate the transcription ofthe glucocorticoid receptor, which in turn reg-

ulates the HPA response to stress. These epige-neticmarksarestable,enduringwellbeyondthe

period of maternal care, and provide a molec-ular basis for a stable maternal effect on the

phenotype of the offspring. Thus, the behaviorof the mother directly alters cellular signals that

then actively sculpt the epigenetic landscape ofthe offspring, influencing the activity of specific

regions of the genome and the phenotype of theoffspring.

The critical feature of the maternal effects

described above is that of persistence. The dif-ferences in the frequency of pup LG between

high- and low-LG mothers are limited to thefirst week of postnatal life. And yet the differ-

encesingeneexpressionandneuralfunctionareapparent well into adulthood. How might the

effects of an essentially social interaction stablyalter the expression of the genes that regulate

the activity of neural systems that mediate en-docrine and behavioral responses to stress? To

address this question, we focused on the sus-tained effect of maternal care on glucocorticoid

receptor gene transcription in the hippocampus

as a model system for the environmental pro-gramming of gene expression.

The focus of the epigenetic studies isthe NGFI-A consensus sequence in the

exon 17 promoter (Figure 1) that activates

glucocorticoid receptor expression in hip-

pocampal neurons. The tactile stimulationassociated with pup LG increases 5-HT activ-

ity in the hippocampus. In vitro studies with

cultured hippocampal neurons show that 5-HTacts on 5-HT7 receptors to initiate a series of

intracellular signals that culminate with an in-crease in the expression of NGFI-A as well as in

the CREB-binding protein (Figure 3). Com-parable effects occur in vivo. Manipulations

that increase pup LG by lactating rats result inan increased level of cAMP as well as NGFI-A

(Meaney et al. 2000). Pups reared by high-LGmothers show increased NGFI-A expression

in hippocampal neurons as well as an increasedbinding of NGFI-A to the exon 17promoter se-quence (Weaver et al. 2007, Zhang et al. 2009).

Moreover, the binding of NGFI-A to the exon17 promoter sequence is actively regulated by

motherpup interactions, such that there isincreased NGFI-A bound to the exon 17 pro-

moter immediately following a nursing bout,but not at a period that follows 25 minuteswith-

out motherpup contact (Zhang et al. 2009).NGFI-A and the CREB-binding protein

form a complex that binds directly to the exon17 promoter sequence and actively redesigns

the methylation pattern at this region of the

genome (Weaver et al. 2004, 2007). Thus, asadults, the offspring reared by high-LG moth-

ers show very modest levels of methylationat the 5 CpG of the NGFI-A consensus se-

quence (Figure 5). This effect on methylationis very precise. Located only a few nucleotides

removed from this site is the 3 CpG site(Figures 1 and 5), the methylation status of

which is unaffected by maternal care.A rather novel aspect of the effect of ma-

ternal care on DNA methylation was apparentin the results of a simple developmental study

examining the methylation status of the 5 and

3 CpG sites from late in fetal life to adulthood(Weaver et al. 2004). Neither the 5 nor the

3 CpG sites within the NGFI-A bindingregion is methylated in hippocampal neurons

from fetal rats, whereas both sites are heavilymethylated on the day following birth, with

no difference as a function of maternal care.



12/34

These findings reflect what is referred to as

de novo methylation, whereby a methyl groupis applied to previously unmethylated sites.

However, between the day following birth and

the end of the first week of life, the 5 CpGis demethylated in pups reared by high-LG,

but not low-LG, mothers. This differencethen persists into adulthood. Importantly, the

period over which the demethylation occursfalls precisely within that time when high- and

low-LG mothers differ in the frequency of pupLG; the difference in pup LG between high-

and low-LG mothers is not apparent in thesecond week of postnatal life (Caldji et al. 1998,

Champagne 2008, Champagne et al. 2003).The demethylation of the 5 CpG siteoccurs

as a function of the same 5-HT-activated sig-

nals that regulate glucocorticoid receptor geneexpression in cultured hippocampal neurons

(Weaver et al. 2007). Thus, when hippocam-pal neurons of embryonic origin are placed in

culture and treated with 5-HT, which mimicsthe extracellularsignal associated with maternal

LG,the5 CpG site is demethylated; there is noeffectatthe3 CpG site. The binding of NGFI-

A to the exon 17 site is critical. Hippocampalneurons that are rendered incapable of increas-

ing NGFI-A expression through antisense or

siRNA treatment show neither the demethyla-tion of the 5 CpG site nor the increase in glu-

cocorticoid receptor expression (Weaver et al.2007). Likewise, a mutation of the NGFI-A

site (exchanging a C for an A at the 3 CpGsite) that completely abolishes the binding of

NGFI-A to the exon 17 promoter also pre-vents the demethylation of the 5 CpG. Finally,

the infection of hippocampal neurons with avirus containing a nucleotide construct engi-

neered to express high levels of NGFI-A pro-duces demethylation of the 5 CpG of the exon

17 promoter sequence and increases glucocor-

ticoid receptor expression.These findings suggest that maternal lick-

ing of pups increases NGFI-A levels in thehippocampal neurons of the offspring, thus al-

tering DNA methylation. But there is a compli-cation. If DNA methylation blocks transcrip-

tion factor binding and the 5 CpG site of

the exon 17 promoter is heavily methylated

neonates, then how might maternally activaNGFI-A bind to and remodel the exon 17

gion? And why is the effect apparent at the

but not the 3 CpG? The answer to these qutions appears to involve other transcriptio

signals that are affected by maternal care. Lels of the transcription factorspecific protein

(SP-1) and the CREB-binding protein are aincreased in the hippocampus of pups reared

high-LG mothers (Weaver et al. 2007, Zhaet al. 2009). The exon 17 promoter contain

DNA sequence that binds SP-1, and this gion overlaps with that for NGFI-A. SP-1 c

actively target both methylation and demethlation of CpG sites (Brandeis et al. 1994). T

5CpG site is the region of overlap in the bin

ing sites. The CREB-binding protein, on tother hand, acts as a histone acetyltransfera

an enzyme capable of acetylating histone taincluding the exon 17 region, opening ch

matin and permitting the binding of transcrtion factors such as NGFI-A and SP-1. Incre

ing histone acetylation can lead to transcriptfactor binding at previously methylated si

and the subsequent demethylation of these gions (Fan et al. 2005, Szyf et al. 2005). Th

we suggest that the binding of this complex

proteins, NGFI-A, the CREB-binding proteand SP-1 is critical in activating the process

demethylation. The results to date are certaiconsistent with this model, but we should n

that we have yet to firmly establish the idetity of the enzyme that is responsible for t

demethylation of the 5 CpG site.These findings suggest that materna

induced increases in hippocampal NGFlevels can initiate the remodeling of DN

methylation at the regions of the DNA thregulate glucocorticoid receptor expressi

The NGFI-A transcription factor binds

multiple sites across the genome. If NGFI-related complexes target demethylation, th

one might assume that other NGFI-A-sensitregions should show a maternal effect on DN

methylation and gene expression comparato that observed with the glucocorticoid rec

tor. Zhang and colleagues (2009) showed t

450 Zhang Meaney


13/34

the hippocampal expression of theGAD1gene

that encodes for glutamic acid decarboxylase,an enzyme in the production of the neuro-

transmitter GABA, is increased in the adult

offspring of high-LG mothers. This effect isassociated with altered DNA methylation of

an NGFI-A response element in a mannercomparable to that for the glucocorticoid

receptor gene. Moreover, as with the effecton the glucocorticoid receptor, an in vitro

increase in NGFI-A expression mimics theeffects of increased pup LG. The function of

GABAergic neurons in the limbic system is alsoregulated by maternal care (Caldji et al. 1998,

2000, 2003) and is a major target for anxiolytic

agents. These findings are therefore likelyrelevant for the decreased fearfulness observed

in the adult offspring of high-LG mothers.In summary, the maternallyinduced changes

in specific intracellular signals in hippocampalneurons can physically remodel the genome.

The increased binding of NGFI-A that derivesfrom pup LG appears critical for the demethy-

lation of the exon 17promoter. We suggest thatthis process involves accompanying increases in

SP-1 and the CREB-binding protein, and thatthe combination of these factors results in the

active demethylation of the exon 17 promoter.

It should be noted that there are important fea-tures of this model that remain to be clearly de-

fined, includingthe identificationof theenzymethat is directly responsible for the demethyla-

tion. Nevertheless, the events described to daterepresent a model by which the biological path-

ways activated by a social event may become im-printed onto the genome. This imprint is then

physically apparent in the adult genome, result-ing in stable alterations (or programming) of

gene expression.

THE FUNCTIONALIMPORTANCE OF THESOCIAL IMPRINT

A critical issue is that of relating the epige-

netic modifications at specific DNA regionsto function. The presence of a methyl group

on the 5 CpG of the NGFI-A binding site is

functionally related to glucocorticoid receptor

gene expression in adult animals. In vitro stud-ies reveal that the methylation of the 5 CpG

site reduces the ability of NGFI-A to bind

to the exon 17 promoter and activate gluco-corticoid receptor transcription (Weaver et al.

2007). These findings are consistent with themodel described above, whereby DNA methy-

lation impedes transcription factor binding andthus the activation of gene expression. The

next question concerns the in vivo situationand function at a level beyond that of gene

expression.In contrast to the situation with neonates,

there is no difference in NGFI-A expressionas a function of maternal care among adult

animals: Hippocampal levels of NGFI-A are

comparable in the adult offspring of high-and low-LG mothers. However, the altered

methylation of the exon 17 promoter wouldsuggest differences in the ability of NGFI-A to

access its binding site on the exon 17promoter.Chromatin-immunoprecipitation assays, which

permit measurement of the interaction be-tween a specific protein and a defined region of

the DNA, reveal increased NGFI-A associationwith the exon 17promoter in hippocampi from

adult offspring of high- compared to low-LG

mothers (Weaver et al. 2004, 2005). This dif-ference occurs despite the comparable levels of

NGFI-A. These findings show that in the livinganimal, under normal conditions, there is more

NGFI-A associated with the exon 17promoterin hippocampal neurons of adult animals reared

by high- compared with low-LG mothers.There is also evidence that directly links

the maternal effect on the epigenetic state ofthe exon 17 promoter to the changes in glu-

cocorticoid receptor expression and HPA re-sponses to stress. Recall that the methylation

of specific CpG sites can diminish transcrip-

tion factor binding through the recruitmentof repressor complexes that include HDACs.

The HDACs deacetylate histone tails, thusfavoring a closed chromatin configuration. In-

deed, the exon 17 promoter is more promi-nently acetylated in hippocampi from adult off-

springof high- compared with low-LGmothers



14/34

(Weaver et al. 2004, 2005). This finding is con-

sistent with the increased transcription of theglucocorticoid receptor gene in animals reared

by high- versus low-LG mothers. A subsequent

study (Weaver et al. 2004) examined the ef-fects of directly blocking the actions of the

HDACs in the adult offspring of high- and low-LG mothers by directly infusing an HDAC in-

hibitor into the hippocampus daily for four con-secutive days. The treatment with the HDAC

inhibitor produces a series of predictable re-sults that reflect a cause-effect relation between

DNA methylation andgene expression. First, asexpected, HDAC blockade eliminates the dif-

ferences in the acetylation of the histone tails(open chromatin) of the exon 17 promoter inhippocampal samples from high- and low-LG

mothers. Second, the increased histone acety-lation of the exon 17promoter in the offspring

of low-LG mothers is associated with an in-crease in the binding of NGFI- A to the exon

17promoter in the offspring of low-LG moth-ers, eliminating the maternal effect on NGFI-

A binding to the exon 17 promoter. Compa-rable levels of NGFI-A binding to the exon

17 promoter then eliminate the maternal ef-fect on hippocampal glucocorticoid receptor

expression, such that glucocorticoid receptor

levels in the adult offspring of low-LG motherstreated with the HDAC inhibitor are compa-

rable to those in animals reared by high-LGmothers. And most importantly, the infusion of

the HDAC inhibitor reversed the differences inthe HPA response to stress.

HDAC inhibition increases NGFI-A bind-ing to the exon 17 promoter in the offspring

of low-LG mothers. The studies with neonatesreveal that increased NGFI-A binding results

in the demethylation of the 5 CpG. In vitro,the introduction of a viral tool that leads to the

increased expression of NGFI-A is sufficient

to demethylate the exon 17 promoter. Weaveret al. (2007) argue that the binding of NGFI-A

is critical for the demethylation of the 5CpGsite. The same effect is apparent in vivo and

even with the adult animals used in the stud-ies described above. HDAC infusion into the

hippocampus increases NGFI-A binding to the

exon 17promoter in the adult offspring of lo

LG mothers and decreases the level of methytionof the 5CpG site onthe exon17promo

Another study (Weaver et al. 2005) show

that the reverse pattern of results could be otained in response to the infusion of meth

nine into the hippocampus. The methioninfusion produced greater methylation of

5CpG in the offspring of high-LG mothedecreasedNGFI-Abinding andGR expressi

and increased HPA responses to stress (Weaet al. 2005).

Although these studies employ rather crupharmacological manipulations, the results

criticalastheysuggestthatfullymatureneurin an adult animal express the necessary en

matic machinery to demethylate or remeth

late DNA. The importance of this plasticat the level of DNA methylation is revea

in subsequent studies of cognition (see belowhich suggest that dynamic modification

DNA methylation in critical neuronal poplations in adult animals is involved in spec

forms of learning and memory.

ACTIVITY-DEPENDENTREGULATION OF

THE EPIGENOME

The maternal effect on the epigenetic state

the glucocorticoid receptor exon 17 promoand glucocorticoid receptor gene expression

apparent over the first week of life and occin response to an increased NGFI-A signa

hippocampal neurons. The increased exprsion of NGFI-A and its binding to the exon

GR promoter over the first week of life are avated by maternal behavior (Weaver et al. 20

Zhang et al.2009). An increase in theexpressof NGFI-A is associated with synaptic plast

ity and with learning and memory (Dragun

1996, Jones et al. 2001, Knapska & Kaczma2004, Li et al. 2005, ODonovan et al. 199

Thus it is not surprising that the offspringhigh-LG mothers show increased synaptic d

sity both in early life (Liu et al. 2000) ain adulthood (Bagot et al. 2009, Bredy et

2003, Champagne et al. 2008, Liu et al. 200

452 Zhang Meaney


15/34

Such events occur as a function of a series of

activity-dependent changes in neuronal activ-ity triggered by the action of glutamate at the

NMDA receptor site (Ali & Salter 2001, Bear &

Malenka 1994, Malenka & Nicoll 1993, Morris& Frey 1997). Thus, it is possible that envi-

ronmentally driven changes in neuronal tran-scriptional signals could potentially remodel

the methylation state of specific regions of theDNA (Meaney & Szyf 2005, Sng & Meaney

2009). These effects could, in turn, prove es-sential for sustained alterations in synaptic

function.Learning and long-term memory com-

monly require changes in gene expression andprotein synthesis (Alberini et al. 1995, Kandel

2001, Lynch 2004). As described above, gene

transcription is associated with chromatinremodeling engineered by enzymes that

modify the histone proteins within chromatincomplexes. A number of the intracellular

signals that are crucial for learning and mem-ory are in fact enzymes that modify histone

proteins. One example is that of the CREB-binding protein, which functions as a histone

acetyltransferase and is strongly implicated incognitive function (e.g., Alarcon et al. 2004).

Thus, contextual fear conditioning, which isa hippocampus-dependent learning paradigm

whereby an animal associates a novel context

with an aversive stimulus, is accompanied byincreased acetylation of histone H3 (Levenson

et al. 2004). Likewise, there is evidence for theimportance of epigenetic modifications of his-

tones in the amygdala during fear conditioning(Yeh et al. 2004). Interestingly, extinction of

the conditioned fear response is associatedwith increased histone acetylation in the pre-

frontal cortex, which mediates the inhibitionof conditioned fear responses (Bredy et al.

2007). The CREB-binding protein is probablyinvolved in the relevant histone modifications.

Mice that are heterozygous for a dysfunction

form of the CREB-binding protein showsignificant impairments in multiple forms of

hippocampal-dependent, long-term memory(Bourtchouladze et al. 2003, Korzus et al. 2004,

Wood et al. 2006; also see Guan et al. 2002,

Vecsey et al. 2007). Importantly, the cognitive

impairments are reversed with HDAC ad-ministration, suggesting that CREB-binding

protein-induced histone acetylation mediateseffects on learning and memory.

There is also evidence for the importance of

dynamic changes in DNA methylation at spe-cific sites during learning and memory. Fear

conditioning results in the rapid methylationand transcriptional silencing of the gene for

protein phosphatase 1 (PP1), which suppresseslearning. The same training results in the

demethylation and transcriptional activation ofthe synaptic plasticity gene reelin. These find-

ings imply that both DNA methylation anddemethylation might be involved in long-term

memory consolidation.BDNF has been implicated in adult neu-

ral plasticity, including learning and memory

(West 2001). Thegenomic structure of the Bdnfgene contains multiple promoters that gener-

ate mRNAs containing different noncoding ex-ons spliced upstream of a common coding exon

(Timmusk et al. 1993). This organization issomewhat like that described above for the glu-

cocorticoid receptor (Figure 1). In the case ofBDNF, the exon IV promoter in rat is activated

upon membrane depolarization in cultured cor-tical and hippocampal neurons by means of

KCl, which leads to calcium influx, activatingsignaling cascades and inducing the expression

of an array of genes that are involved in neuralplasticity (West 2001).

Importantly, the activity-dependent Bdnf

gene is also regulated through epigenetic modi-fications that involve dynamic changes in DNA

methylation and the association of methylated-DNA binding proteins to the relevant sites

on the bdnf promoter. Thus, increased DNAmethylation of the exon IV promoter at sites

that bind to transcriptional activators is associ-atedwiththepresenceofthemethylated-DNA-

binding protein, MeCP2, and a decreased levelof bdnf expression. This transcriptionally

quiescent state prior to depolarization is alsoassociated with the presence of histone deacety-

lases (i.e., HDAC1) and mSIN3A, which form

a common repressor complex. Membrane



16/34

Proteinphosphorylation:a modification ofproteins in which aspecific amino acid(serine, a threonine or

a tyrosine) isphosphorylated by aprotein kinase by theaddition of a covalentlybound phosphategroup. In such cases ofphosphoregulation,the protein switchesbetween aphosphorylated and anunphosphorylatedstate, with one in anactive form and the

other inactive

depolarization of the neuron leads to a de-

crease in CpG methylation and a dissociationof MeCP2-related repressor complex from the

exon IV promoter. As described above for the

glucocorticoid receptor, the decrease in CpGmethylation is then associated with an increase

in histone acetylation and the binding of thetranscription factor, CREB. CREB is known

to activatebdnfexpression. These data suggestthat DNA methylation at a particular site can

suppress activity-dependent transcription of

Bdnf. These findings also indicate that DNA

methylation patterns in postmitotic neuronscan undergo dynamic changes in response to

neuronal activation, and a lower level of DNAmethylation correlates with a higher level of

Bdnfgene transcription in neurons.

Interestingly, MeCP2 levels increase as neu-rons mature (Zoghbi 2003). The high level of

MeCP2 protein in mature neurons is consis-tent with a possible role for MeCP2 in synaptic

remodeling associated with learning and mem-ory (Zhou et al. 2006). Further supporting a

roleforMeCP2inmaturesynapticfunctionandplasticity, Mecp2-null mice exhibit abnormali-

ties in dendritic arborization (Chen et al. 2003,Kishi & Macklis 2004), basal synaptic transmis-

sion (Moretti et al. 2006), presynaptic function(Asaka et al. 2006, Moretti et al. 2006, Nelson

et al. 2006), excitatory synaptic plasticity (Asaka

et al. 2006, Moretti et al. 2006), and hippocam-pal and amygdalar learning (Moretti et al. 2006,

Pelka et al. 2006). Zhou et al. (2006) found thatneuronal activity (membrane depolarization) is

associated with a phosphorylation of MeCP2 atSerine421 that led to its dissociation from the

bdnfexon IV promoter and an increase in bdnfexpression (also see Chen et al. 2003). Impor-

tantly, activity-dependent increases in BDNFlevels are blocked in cells bearing a mutant

version of MeCP2 that is unable to undergophosphorylation. Glutamate is a primary neural

signalforsynapticplasticity,andbothglutamate

as well as the direct activation of its NMDAreceptor produced MeCP2 phosphorylation in

neurons. Glutamate activates NMDA recep-tors, resulting in a neuronal calcium influx and

the activation of calcium-modulated kinase II

(CaMKII), which regulates synaptic plastic

(Lisman et al. 2002). Zhou et al. (2006) fouthat CaMKII actively phosphorylates MeCP

The protein phosphorylation occurringMeCP2 in response to neuronal activation

a transient event. The results described abo

(Martinowich et al. 2003) suggest that nronal activation can lead to changes in DN

methylation, which is a potentially more stabepigenetic alteration that could conceiva

result in a long-term change inbdnfexpressiThus far, this review has considered

relation between DNA methylation, histoacetylation/deacetylation, transcription fac

binding, and gene expression. However, this evidence that the chromatin alteratio

can alter DNA methylation. Thus, HDinhibitors result in an increase in histo

acetylation, enhanced transcription fac

binding, and decreased DNA methylatiSuch effects were described above in relation

DNA methylation and glucocorticoid recepexpression (Weaver et al. 2004, 2005). Thus

is possible that (a) neuronal activation leadsthe transient phosphorylation of MeCP2 a

its dissociation from the exon IVbdnfpromoand (b) an increase in histone acetylation a

CREB binding, producing increased bexpression; and that (c) the histone acetylat

and CREB binding are also associated wDNA demethylation, as described above

the case of the glucocorticoid receptor histone acetylation and NGFI-A bindi

Such events could underlie a common proc

of activity-dependent modification of DNmethylation (Meaney & Szyf 2005).

Studies by Sweatt and colleagues suggthat the changes in DNA methylation at t

exon IVbdnfpromoter are involved in specforms of learning and memory (Sweatt 200

Bdnf gene expression increases in the hpocampus with contextual and spatial learn

and appears essential for the synaptic remoeling that accompanies such forms of learni

and memory (Hall et al. 2000, Linnarsson et1997). NMDA receptor activation is criti

for both contextual (Maren & Quirk 20

and spatial (Morris et al. 2003) learning

454 Zhang Meaney


17/34

well as for the increase in bdnf expression that

accompanies such events. Lubin et al. (2008)found that contextual fear conditioning was

associated with a demethylation of the exon IV

bdnfpromoter and an increase in bdnfexpres-sion: Both effects were blocked with a gluta-

mate receptor antagonist.Taken together, thesefindings suggest that the activity-dependent

changes in neuronal activity that associatewith learning and memory induce a dynamic

alteration in DNA methylation that, in turn,subserves the sustained changes in gene expres-

sion critical for long-term memory. Althoughthis remains a working hypothesis, the findings

discussed above further emphasize the degree

to which neuronal activation can structurallyremodel the genome and alters its operation.

Interestingly, there is also evidence that en-vironmental influences prevailing during early

development may determine the capacity forsuch activity-driven, epigenetic modifications.

Disruptions to motherinfant interactions dur-ing early development are associated with alter-

ations in hippocampalbdnfexpression (Branchiet al. 2006; Lippman et al. 2007; Roceri et al.

2002, 2004; but also see Griesen et al. 2005)and increased DNA methylation at the exon IV

bdnfpromoter (Roth et al. 2009). Rearing micein a communal nest, with three mothers and

their litters, increases maternal care toward the

offspring, which in turn is associated with in-creased BDNF expression (Branchi et al. 2006).

And in the rat, the offspring of high-LG moth-ers show decreased MeCP2 association with the

exon IVbdnfpromoter (Weaver et al. 2007) andincreased bdnfexpression (Liu et al. 2000). Such

maternal effects might bias in favor of reducedcapacity for epigenetic remodeling at this crit-

ical site and restrain synaptic plasticity associ-ated with learning and memory.

Summary (and PerhapsSome Constraints)

Studies over the past five years have created

considerable enthusiasm for epigenetic mod-els of the effects of early experience, synaptic

plasticity, and neural function. The hypothesis

underlying this approach considers epigenetic

effectsongeneexpressionasacandidatemecha-nism for the effects of environmental signals on

the future behavior of the organism. This hy-

pothesis is particularly attractive for those ex-amining the sustained effects of early experi-

ence or of chronic, biologically relevant eventsin adulthood (e.g., environmental enrichment,

chronic stress) on gene expression and neuralfunction. Mature neurons undergo consider-

able changes in phenotype and are thereforean ideal cell population for epigenetic regu-

lation. Nevertheless, there are constraints onthe influence of epigenetic marks. For exam-

ple, the effects of DNA methylation on geneexpression are influenced by the organization

of the relevant genomic region. DNA methy-

lation appears to have a reduced effect ongene expression in regions that have a very

high density of cytosine-guanine paired sites(Weber et al. 2007). Moreover, much of the

DNA within a cell is packed tightly in het-erochromatin (Fraser & Bickmore 2007) and

is probably inaccessible to environmentally in-duced chromatin remodeling signals. Thus, the

infusion of an HDAC inhibitor directly intothe adult hippocampus alters the expression of

only about 2% of all the genes normally ex-pressed in the rat hippocampus (Weaver et al.

2006). Were the entire genome subject to dy-

namic epigenetic regulation such as describedabove for thebdnfgene, then we could expect

this percentage to be substantially higher. Itis likely that there is a pool of genes that re-

tains the capacity for dynamic environmentalregulation through epigenetic mechanisms. Of

course this begs questions concerning the fac-tors that determine the nature and contents of

suchpools.These considerations notwithstand-ing, it appears that with neurons, a number of

genes are closely related to synaptic plasticityand neural function and are subject to dynamic

regulation through epigenetic mechanisms, in-

cluding DNA methylation.Epigenetics refers to a collection of chemi-

cal modifications that occur to histones or di-rectly on the DNA. These modifications, in

turn, alter gene transcription. One might argue



18/34

that in defining such mechanisms we have, in

effect, simply better defined the processes thatregulate gene transcription. Although the value

of such findings is obvious for molecular biol-

ogy, how might such processes revise our think-ing at the level of the systems sciences? We

suggest that these findings provide researcherswith a renewed appreciation of the environ-

mental regulation of cellular activity. We nowunderstand the physical basis for the Hebbian

synapse (Hebb 1958), whereby environmen-tal signals activate intracellular pathways that

result in the remodeling of synaptic connec-tions in a manner that influences subsequent

activity at relevant sites. There is a physicalreference for the process of neuroplasticity.

Epigenetic modifications provide the mecha-

nism for a comparable level of plasticity at thelevelof the genome. We once thought of synap-

tic connections as being fixed, immutable tofurther changes beyond some critical period in

development. Studies of synaptic plasticity re-vised our appreciation of the brain, revealing

instead a dynamic tissue, subject to constant re-modeling through the environmental activation

of activity-dependent synaptic plasticity. Thestudy of epigenetics suggests a comparable pro-

cess at the level of the genome, also once con-sidered a constant, static source of influence.

Indeed, we must emphasize that epigenetic

modifications do not alter DNA sequence. Theproduct of the glucocorticoid receptor gene is

unaffected by epigenetic marks. However, it ap-pears that the operation of thegenomeis indeed

subject to environmental regulation in a man-ner that may be no less dynamic than that of

synaptic connections.Recent studies from Nestler and colleagues

reveal considerable epigenetic modification atspecific genomic sites associated with chronic

stress or repeated exposure to psychostimulantdrugs, both of which produce sustained influ-

ences on behavior (Nestler 2009, Renthal et al.

2009). Although such effects have yet to bereported for DNA methylation, modifications

of histone proteins are associated with expo-sure to drugs of abuse and stressors in rodent

models (Renthal et al. 2009, Renthal & Nestler

2008). These findings suggest that epigene

states, including DNA methylation, are alteby a wide range of biologically relevant eve

(Meaney & Szyf 2005, Renthal & Nestler 20Szyf et al. 2005). Such epigenetic modificatio

might therefore underlie a wide range of s

ble changes in neural function following expsure to highlysalient events (e.g.,chronic stre

drugs of abuse, reproductive phases such as penting) and are thus logical mechanisms

environmentally induced alterations in menhealth (Akbarian & Huang 2009, Jiang et

2008, Tsankova et al. 2007).

EPIGENETICS ANDMENTAL HEALTH

Emerging evidence links the alterations in gexpression associated with DNA methylati

to psychiatric illness. Cortical dysfunctionschizophrenia is associated with changes

GABAergic circuitry (Benes & Berretta 200This effect is associated with a decrease

the expression of the GAD1 gene that ecodes for a specific form of glutamic acid

carboxylase (GAD67), one to two key enzymfor GABA synthesis in cortical interneuro

There is compelling evidence for the decrea

expression of GAD67 in cortical tissues frschizophrenic patients (Akbarian & Hua

2006, Costa et al. 2004). The dysregulaGAD67 expression in the chandelier GA

neurons is thought to result in disruptionsynchronized cortical activity and impairm

of executive functions in schizophrenia subje(Lewis et al. 2005). Likewise, allelic variation

GAD1is associated with schizophrenia (Straet al. 2007).

In addition to GAD67, there is alsodecrease in cortical expression of reelin

schizophrenic brains (Eastwood & Harris

2003); reelin is closely associated with syntic plasticity. The same GABAergic neurons

the schizophrenic brain that express reelin aGAD67 exhibit an increase in DNA meth

transferases 1 (DNMT1; Veldic et al. 200DNMT1 is a member of a family of

zymes that transfers a methyl group from t

456 Zhang Meaney


19/34

methyl donor S-adenosyl-methionine (SAM)

onto cytosines, thus producing DNA methy-lation. The promoter for thereelingene shows

increased methylation in the brains of patients

with schizophrenia compared with control sub-jects (Abdolmaleky et al. 2005, Grayson et al.

2005). Kundakovic et al. (2007) showed that theinhibition of DNMT1 in neuronal cell lines re-

sulted in the increased expression of both reelinand GAD67. The increase in gene expression

was associated with a decreased association ofMeCP2, further suggesting that these differ-

ences are associated with alteration in DNAmethylation. Recall that maternal care directly

alters DNA methylation of the GAD67 pro-

moter in the rat (Zhang et al. 2009). This effectis associated with a decrease in DNMT1 ex-

pression and reduced MeCP2 association withtheGAD1promoter.

An important question is that of the devel-opmental origins of such differences in DNA

methylation. A set of recent studies (McGowanet al. 2009) suggests that epigenetic modifi-

cations might occur in humans in responseto variations in parentoffspring interactions.

DNA was extracted from hippocampal samplesobtained from victims of suicide or from

individuals who had died suddenly from othercauses (auto accidents, heart attacks, etc.).

The samples were obtained from the Quebec

Suicide Brain Bank, which conducts forensicphenotyping that includes a validated assess-

ment of psychiatric status and developmentalhistory (e.g., McGirr et al. 2008). The studies

examined the methylation status of the exon 1Fpromoter of the glucocorticoid receptor, which

corresponds to the exon 17promoter in the rat(Turner & Muller 2005). The results showed

increased DNA methylation of the exon 1Fpromoter in hippocampal samples from suicide

victims compared with controls, but only if

suicide was accompanied with a developmentalhistory of child maltreatment. Child maltreat-

ment, independent of psychiatric state, pre-dicted the DNA methylation status of the exon

1Fpromoter. As in the previous rodent studies,the methylation state of the exon 1F promoter

also determined the ability of NGFI-A to bind

to the promoter andactivate gene transcription.

Although such studies are obviously correla-tional and limited by postmortem approaches,

the results are nevertheless consistent with

the hypothesis that variations in parental carecan modify the epigenetic state of selected

sites of the human genome. Moreover, thefindings are also consistent with studies that

link childhood abuse to individual differencesin stress responses (Heim et al. 2000). Child-

hood abuse is associated with an increase inpituitary ACTH responses to stress among

individuals with or without concurrent majordepression. These findings are particularly

relevant, since pituitary ACTH directly reflectscentral activation of the HPA stress response,

and hippocampal glucocorticoid receptor

activation dampens HPA activity. The findingsin humans are consistent with the rodent

studies cited above investigating epigeneticregulation of the glucocorticoid receptor

gene and with the hypothesis that early lifeevents can alter the epigenetic state of relevant

genomic regions, the expression of which maycontribute to individual differences in the risk

for psychopathology (Holsboer 2000, Neigh& Nemeroff 2006, Schatzberg et al. 1985).

Certain limitations need to be considered aswe integrate epigenetics into the study of psy-

chopathology. The study of epigenetic mech-

anisms in humans is complicated by the factthat epigenetic marks are often tissue-specific.

For example, the brain contains some neu-rons that synthesize and release dopamine as

a neurotransmitter and others that rely onacetylcholine. We might assume that among

dopaminergic neurons, the genes associatedwith the capacity for acetylcholine production

are silenced, likely through some level of epi-genetic regulation. Such processes are inherent

in the specialization of brain cells, as with allother differentiated cells in the body. This pro-

cess of specialization involves epigenetic reg-

ulation and implies that the epigenetic marksvary from cell type to cell type. Indeed, there is

considerable variation in epigeneticmarks fromone brain region to another, perhaps even more

so than variation within the same brain region



20/34

across individuals (Ladd-Acosta et al. 2007).

Brain samples are for the most part beyonddirect examination in the living individualat the

level of molecular analysis. This often leaves us

with measures of DNA extracted from bloodor saliva and with the question of whether the

epigenetic marks within such samples actuallyreflect those within the relevant neuronal pop-

ulation. Thus, for the time being advances inthe study of neuroepigenetics will rely heavily

on relevant models with nonhuman species aswell as complementary studies of samples from

postmortem human brains.

CONCLUSIONS

It is now evident that genomic variation at

the level of nucleotide sequence is associatedwith individual differences in personality and

thus with vulnerability and resistance to a widerange of chronic illness (Ebstein 2006, Meyer-

Lindenberg & Weinberger 2006, Rutter 2007).The challenge is how to conceptually integrate

thefindings from genetics into psychology. Theoperation of the genome is regulated by cellu-

lar signals that are responsive to environmentalconditions. Thus, the effects of genetic varia-

tion are contextually determined and therefore

best considered as probabilistic. Genetic vari-ations influence cellular activity and, depend-

ing upon current and past environmental con-ditions, will bias toward particular functional

outcomes. The molecular events that mediategene transcription reveal the interdependence

of gene and environment (Sokolowski 2001,Sokolowski & Wahlsten 2001). Oddly, what is

perhaps the most profound comment on thisissue dates back several years. In response to

a question from a journalist considering therelative importance of nature versus nurture

in defining individual differences in personal-ity, Hebb responded that such comparisons are

akin to asking what contributes more to the

area of rectangle, the length or the width? Therecent flush of studies examining gene x en-

vironment effects on personality and vulnera-bility/resistance to mental illness (Caspi et al.

2003, Meaney 2009, Rutter 2007, Suomi 2006)

reflects the interdependence of genetic and

vironmental influences, such that the effectone level can only be understood within t

context of the other. Indeed, developmenprocesses are best considered as the outco

of a constant dialog between the genome a

its environment (Bateson 1994; Gottelieb 191998; Lewontin 1974).

The gene x environment perspective is cical in the establishment of an understand

of the development of individual differencesneural function and personality. Until recen

most experimental approaches were limitedidentifying factors that could influence neu

development. Our own studies of maternal cin the rat are a case in point. This research

amines the effects of variation in maternal cin animals that are housed from weaning o

ward under identical conditions. We systema

cally minimize variation from weaning onwaThis approach permits conclusions as to the p

tential effects of variations in maternal care bcannot estimate the importance of such effe

for individual differences in adult function uder natural conditions. Indeed, environmen

enrichment in the postweaning period can verse effects associated with the variations

maternal care (Bredy et al. 2004, Champag& Meaney 2006, Zhang et al. 2006). Likew

studies of monozygotic-dizygotic twins exaine what are, in effect, differences in paren

gene dosage while minimizing variation in

early environment. Such approaches have pvided convincing evidence that genetic fact

caninfluence thedevelopment of individualdferences, but do not identify how. Indeed,

challenge is to define how, when, and unwhich conditions specific genetic or enviro

ment factors operate to regulate developmeHerein lies the enormous contribution of t

gene x environment perspective, particulawhen integrated into longitudinal studies

development.The excitement concerning the findin

in the area of epigenetics derives from

realization that such mechanisms could fothe biological basis for the interplay betwe

environmental signals and the genome. T

458 Zhang Meaney


21/34

studies reviewed here suggest that (a) epi-

genetic remodeling occurs in response tothe environmental activation of the classic

activity-dependent cellular signaling path-

ways that are associated with synaptic plasticity,(b) epigenetic marks, particularly DNA methy-

lation, are actively remodeled over early devel-opment in response to environmental events

that regulate neural development and function,and (c) epigenetic marks at histone proteins and

the DNA are subject to remodeling in responseto environmental influences even at later stages

in development. We have highlighted examplesof environmental influences that are of obvious

relevance for psychologists. However, increas-

ing evidence from animal studies indicates thatprenatal and early postnatal environmental

factors, including nutritional supplements,xenobiotic chemicals, and reproductive tech-

nologies, can alter the epigenetic state ofspecific genomic regions ( Jirtle & Skinner

2007).These findings suggest that epigenetic re-

modeling might serve as an ideal mechanismfor phenotypic plasticitythe process whereby

the environment interacts with the genome toproduce individual differences in the expression

of specific traits. One could easily imagine that

such processes mediate observed discordancesbetween m

Documents

Zhang Meaney 10 AnnRPsychol Epigenetics Review