Available online at www.sciencedirect.com

www.elsevier.com/locate/brainres

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 8

http://dx.doi.org/100006-8993/& 2015 Pu

Abbreviations: A

regulated kinase; EV

KCl, potassium ch

5; mGluR5pS, mGl

interacting 1; PRP,nCorresponding aE-mail addresse

Review

Homer 1a and mGluR5 phosphorylation inreward-sensitive metaplasticity: A hypothesis ofneuronal selection and bidirectionalsynaptic plasticity

Tanya M. Marton, Marshall G. Hussain Shuler, Paul F. Worleyn

Department of Neuroscience, Johns Hopkins University School of Medicine, 725 N. Wolfe Street, Baltimore, MD 21205,USA

a r t i c l e i n f o

Article history:

Accepted 23 June 2015

Drug addiction and reward learning both involve mechanisms in which reinforcing

neuromodulators participate in changing synaptic strength. For example, dopamine

Available online 14 July 2015

Keywords:

Homer

mGluR5

PIN1

Dopamine

NMDA receptor

Immediate early gene

Addiction

Reward learning

Reinforcement learning

Synaptic plasticity

Synaptic tag

Eligibility trace

Protoweight

Provisional weight

Neuronal selection

.1016/j.brainres.2015.06.03blished by Elsevier B.V.

rc, activity-regulated cy

H1, Ena/VASP Homolo

loride; LFS, low-frequen

uR5 with phosphorylated

plasticity-related proteinuthor.s: tanya.marton@gmail.c

a b s t r a c t

receptor activation modulates corticostriatal plasticity through a mechanism involving

the induction of the immediate early gene Homer 1a, the phosphorylation of metabotropic

glutamate receptor 5 (mGluR5)0s Homer ligand, and the enhancement of an NMDA

receptor-dependent current. Inspired by hypotheses that Homer 1a functions selectively

in recently-active synapses, we propose that Homer 1a is recruited by a synaptic tag to

functionally discriminate between synapses that predict reward and those that do not. The

involvement of Homer 1a in this mechanism further suggests that decaminutes-old firing

patterns can define which synapses encode new information.

This article is part of a Special Issue entitled SI:Addiction circuits.

& 2015 Published by Elsevier B.V.

7

toskeleton-associated protein; D1R, dopamine receptor 1; ERK, extracellular signal-

gy 1; H1a, Homer 1a; HFS, high-frequency stimulation; IEG, immediate early gene;

cy stimulation; LTP, long-term potentiation; mGluR5, metabotropic glutamate receptor

S1126; NMDAR, NMDA receptor; PIN1, peptidyl-prolyl cis-trans isomerase NIMA-

; SIC, slow inward current

om (T.M. Marton), shuler@jhmi.edu (M.G. Hussain Shuler), pworley@jhmi.edu (P.F. Worley).

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 818

Contents

1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182. Biochemical model describes contingencies for neuromodulator-dependent synaptic plasticity. . . . . . . . . . . . . . . . . . 18

2.1. Metabotropic glutamate receptors, Homer, and cocaine addiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.2. D1R activation induces mGluR5 phosphorylation: a de novo binding site for PIN1. . . . . . . . . . . . . . . . . . . . . . . . . 192.3. H1a/mGluR5pS/PIN1 contributes to the in vivo effects of cocaine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202.4. mGluR5-PIN1 mechanism in synaptic plasticity; requirement for H1a induction creates a temporal contingency. . . . . 212.5. PIN1 isomerization of mGluR5 enhances coupling to NMDA type glutamate receptors . . . . . . . . . . . . . . . . . . . . 22

3. The H1a/mGR5pS/PIN1 mechanism describes an overlap between reward learning and synaptic tagging,and predicts novel features of neuronal selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.1. The H1a/mGR5pS/PIN1 mechanism detects the co-occurrence of glutamatergic, neuromodulatory, and

neuronal events and depends on reactivation over decaminutes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.2. The “synaptic eligibility trace” model describes glutamatergic and neuromodulatory contingencies;

“synaptic tags” further relate neuronal events to synaptic plasticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.3. Synaptic tags recruit H1a to punish unreinforced active synapses but enhance reactivated, reinforced

active synapses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.4. H1a-dependent metaplasticity might direct neuronal selection to neurons with decaminutes-old activity . . . . . 25

4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

1. Introduction

Animals experience many events in the course of a day, yetpreferentially encode those that have special salience. The basisfor this selectivity can be related generally to “reinforcementpathways” that release neuromodulators including dopamine,noradrenaline, or serotonin in specific brain areas in responseto an experience. Drug addiction can be conceptualized as aform of learning that usurps these neuromodulatory rewardpathways (Hyman et al., 2006; Luscher and Malenka, 2011). Howdo neuromodulators influence memory formation and howdoes their misappropriation by drugs of abuse lead to addic-tion? As we will describe here, neuromodulators evoke intra-cellular signaling events that create conditional states ofenhanced synaptic plasticity, which can alter synaptic driveand create long-lasting associations between neurons. Theseassociations underlie reward learning and memory. Wedescribe newly discovered biochemical events in this pathwaythat are time-dependent and conditional in ways that suggestparallels with theoretical models of learning. We begin with abiochemical model that is built upon the function of a class ofproteins encoded by cellular immediate early genes (IEGs) andhighlight the events that create such conditionality. We thenspeculate on how these molecular events might relate tophysiological and theoretical models of learning.

2. Biochemical model describes contingenciesfor neuromodulator-dependent synaptic plasticity

2.1. Metabotropic glutamate receptors, Homer, andcocaine addiction

Dr. Conquet's lab at GlaxoSmithKline first reported that metabo-tropic glutamate receptor type 5 (mGluR5) plays a role in thebehavioral responses to cocaine (Chiamulera et al., 2001). Sincethen, pharmacological manipulation of mGluR5 has similarly

been shown to influence self-administration of cocaine andother drugs of abuse (Kenny et al., 2003; Paterson and Markou,2005). mGluR5 is a group 1 metabotropic glutamate receptor thatactivates phospholipase C to generate the intracellular secondmessengers diacyl glycerol (which activates protein kinase C)and inositol trisphosphate (which releases Ca2þ from intracel-lular stores). In addition to these canonical signaling outputs,group 1 mGluRs can regulate signaling pathways that controlactivation of PI3 kinase, tyrosine kinase and phosphatase, andelongation factor 2 kinase, generate endocannabinoids and D-serine, and activate membrane ion channels (Nicoletti et al.,2011). Given this diversity of function, it is perhaps surprisingthat mGluR5 deletion mice (mGluR5�/�) are viable and fertile,and possess many baseline behavioral responses that allowthem to survive in a lab setting. However, mGluR5�/� miceshow several behavioral endophenotypes relevant to addictionincluding reduced locomotion in response to cocaine, reducedcocaine motor sensitization, and reduced cocaine self-administration (Bird and Lawrence, 2009). The finding thatmGluR5 is involved in the behavioral responses to cocaine wasunexpected, as cocaine was known to influence dopamine levelsand receptors, but not directly affect glutamate receptors. Howthen is mGluR5 involved in responses to cocaine?

One lead toward answering this question came from

studies of genes whose mRNAs are rapidly up-regulated inresponse to brain activity. Classical studies reported morethan 50 years ago that de novo RNA and protein synthesis arerequired for animals to establish long-term memory (Davisand Squire, 1984). Importantly, protein and RNA synthesis arerequired only during a brief time window �3 h immediatelyfollowing a salient experience (Montarolo et al., 1988). Thissuggested that it may be possible to discover critical eventsthat underlie memory by identifying mRNAs and encodedproteins that are up-regulated during this time window and

examining their synaptic function.Homer 1a (H1a) is an example of a gene that is rapidly

transcribed in brain neurons in response to neuronal

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 8 19

activating stimuli, including cocaine (Brakeman et al., 1997).H1a is a member of a family of gene products generated bythree different genes, termed Homer 1, 2, and 3 (Xiao et al.,1998; Tu et al., 1998). All of the Homer gene products share aconserved protein domain at their N-terminus called theEVH1 (Ena Vasp Homology 1) domain. H1a is generated fromthe Homer 1 gene by a mechanism that involves activity-dependent transcript termination within an intron (Bottaiet al., 2002). This mechanism prevents splicing of the intron,and the intron becomes an exon that encodes the shorter C-terminus of the H1a protein. Homer 1 is additionally notablein that intron 1 is large (�40 kb). Because of the finiteprocessivity rate of transcription by Pol II, it takes �20 minfrom the time of a stimulus for H1a mRNA to appear in theneuronal soma (Vazdarjanova et al., 2002). This contrastswith 2–3 min for the de novo transcription of Arc mRNA(Guzowski et al., 1999). Despite the complexity of H1a'stranscriptional induction, H1a and Arc are both closely tiedto physiological synaptic activity; these two gene productscan be monitored in coordination to identify neurons of thehippocampus and cortex that are engaged in experience-dependent information processing (Vazdarjanova et al., 2002).

Homer's EVH1 domain binds the consensus sequencePPXXF (Tu et al., 1998). mGluR50s C-terminus contains aHomer binding site (TPPSPF), as do regions of the IP3 receptor,the ryanodine receptor, Shank, and dynamin 3 (Tu et al.,1999). These proteins are likely co-functional within thesynapse. H1a is unique within the Homer gene family; itlacks the C-terminal coiled-coil and leucine zipper region thatmediates dimerization or tetramerization of all other Homermolecules (Hayashi et al., 2009; Xiao et al., 1998). These longerforms of Homer (in forebrain Homer 1c and Homer 2 are themajor forms; Homer 3 predominates in cerebellum) can formcrosslinked complexes, connecting mGluR5 with downstreamsignaling partners. H1a lacks the dimerization domain andtherefore acts as a “dominant negative” that can disruptcrosslinking, freeing mGluR5 from crosslinked complexes.mGluR5 signaling might be modulated in response to neuro-nal activity through the dynamic expression of H1a andconsequent disruption of these complexes (Xiao et al., 2000).

Evidence that Homer proteins contribute to behavioralresponses to cocaine came with studies that observedchanges in Homer expression following cocaine administra-tion (Swanson et al., 2001) and examined the effect ofgenetically deleting Homer 1 or Homer 2 (Szumlinski et al.,2004). Interestingly, naïve Homer1�/� or Homer2�/� miceexhibit neurochemical changes in the striatum that aretypically seen in wildtype mice upon repeated administrationof cocaine (Szumlinski et al., 2004). An emergent modelsuggested Homer crosslinking restricts mGluR5 function toinhibit an mGluR5-dependent cocaine-induced plasticity(Szumlinski et al., 2006).

2.2. D1R activation induces mGluR5 phosphorylation: a denovo binding site for PIN1

To test the hypothesis that interruption of Homer binding tomGluR5 creates a condition that mimics or enhances cocainesensitization, we created a knock-in mouse that incorporatesa single amino acid mutation of mGluR5 [TPPSPR; mGluR5R/R]

that markedly reduces Homer EVH1 binding (Tu et al., 1998).By disrupting Homer binding to mGluR5, we hoped to mimicHomer 1�/� and Homer 2�/� animals' responses to cocaine.However, mGluR5R/R mice showed normal neurochemicalparameters, normal locomotor responses to cocaine, andnormal motor sensitization (Park et al., 2013). Thus, interrup-tion of Homer-dependent crosslinking by mutation ofmGluR5 did not produce the same effect as deleting cross-linking Homer. A different model was needed.

An alternate model for the role of Homer binding tomGluR5 emerged with the discovery that the Homer bindingsite in mGluR5 is phosphorylated by kinases that targetproline-adjacent serine or threonine (proline-directedkinases) (Orlando et al., 2009). Since proline-directed ERKkinases are activated by dopamine receptor (D1R) stimulation(Gerfen, 2003; Gerfen et al., 2008; Kim et al., 2006; Valjentet al., 2004), we reasoned that mGluR5 might be phosphory-lated by ERK following D1R stimulation by cocaine (Koob andBloom, 1988; Koob, 1992). Indeed, cocaine administration ordirect D1R stimulation induced mGluR5 phosphorylation inan ERK-dependent manner (Fig. 1, steps 2 and 3) (Park et al.,2013). These findings suggest a model whereby addictivedrugs and natural rewards enhance dopamine levels (Carelliand Wightman, 2004; Luo et al., 2011), stimulate D1Rs, andactivate ERKs to phosphorylate mGluR5; this phosphorylationcould mediate some of the behavioral effects of cocaine.

Phosphorylation of the Homer binding site in mGluR5 hastwo important effects: it increases the affinity of Homer EVH1binding and creates a binding site for a prolyl isomerase(Fig. 1, steps 3 and 5) (Park et al., 2013). These effects arerevealed in dynamic structural studies conducted by Dr.Kerns' lab at Brandeis that used NMR to determine theconformation and binding affinity of synthetic phosphopep-tides and Homer 1 EVH1. A prior crystal structure of the EVH1domain bound to the mGluR5 sequence (TPPSPF) showed thatthe alignment of phenylalanine's aromatic side chain alongthe aromatic surface of EVH1 is energetically important to themGluR5-Homer interaction (Beneken et al., 2000). The cisconformation of the S–P bond is required to create a tighthairpin turn between phenylalanine and the trigonal prism ofthe peptide's N-terminus. Phosphorylated forms of themGluR5 peptide enable an increase in the cis conformationof the S–P bond, allowing a near 40-fold increase in EVH1binding affinity (Park et al., 2013).

A second consequence of mGluR50s phosphorylation isthat it creates a binding site for the prolyl isomerase PIN1(Fig. 1, step 5) (Park et al., 2013). Proline's side chain ischemically bonded to the amino acid backbone, whichreduces rotation around this bond. Rotation is important inprotein folding during synthesis and for functional propertiesof certain proteins. Phosphorylation of the S–P bond in theHomer binding site of mGluR5 further slows spontaneousisomerization such that prolyl isomerases become essentialto allow isomerization on a time scale of sec�1 (Fig. 1, step 6)(Lu et al., 2007; Lu and Zhou, 2007). PIN1 is a member of theparvulin family of prolyl isomerases and is unique in that itrequires phosphorylation of its substrate for its binding andcatalytic activity. We therefore hypothesized that PIN1 couldbe part of a mechanism involving cocaine-induced phosphor-ylation of mGluR50s Homer ligand and subsequent

Fig. 1 – H1a/mGR5pS/PIN1 interactions connect neuronal activity and dopaminergic input to enhance NMDAR current. Theevents enabling mGluR5 Homer ligand's interaction with pERK, Homer 1a, and PIN1 are depicted as steps 1–7: 1. Sufficientneuronal firing induces the transcription and translation of the immediate early gene (IEG) Homer 1a (H1a), 2. Stimulation ofthe D1 dopamine receptor activates pERK, 3. causing phosphorylation of mGluR50s Homer ligand (mGR5pS), 4. H1a competesoff crosslinking long-form Homer, 5. allowing the prolyl isomerase PIN1 to bind, and 6. isomerize mGluR50s phosphorylatedHomer ligand, 7. PIN1 isomerization of mGluR50s Homer ligand enhances an NMDAR-dependent current.

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 820

interactions. NMR studies confirmed that PIN1 binds phos-phorylated mGluR5 peptide sequence and accelerates iso-merization of the pS–P bond by41000-fold (Park et al., 2013).We tested the hypothesis that phospho mGluR5-PIN1 inter-actions are enhanced by cocaine administration and foundthat in mice lacking all forms of Homer, cocaine induces arapid increase of mGluR5 phosphorylation and a parallelincrease of PIN1 binding to mGluR5. However, in wildtypemice, cocaine induced phosphorylation of mGluR5 but did notresult in a detectable increase in PIN1 binding. This suggestedthat cross-linking Homer might compete with PIN1 for inter-action with mGluR50s Homer ligand, thereby restricting PIN1binding even when mGluR5 is phosphorylated. Biochemicalstudies confirmed this hypothesis; they also showed that theaddition of H1a greatly enhances PIN1-mGluR5 interactionsdespite the presence of crosslinking Homer. In this model,H1a competes effectively with crosslinking forms of Homer(Fig. 1, step 4) and permits PIN1 to bind in the presence ofcross-linking Homer (Fig. 1, step 5; compare Fig. 2A (H1apresent and PIN1 binding) to Fig. 2C (long-form Homer, noPIN1 binding). NMR and biochemical studies indicated thatenhanced PIN1 binding only occurs when H1a reaches stoi-chiometric equivalence to crosslinking forms of Homer. Sincecrosslinking forms of Homer are very abundant at excitatorysynapses, and total cellular cross-linking Homer is in stoi-chiometric excess to H1a even at the peak of its inducedexpression (Park et al., 2013), this model requires that newlysynthesized H1a be able to accumulate at specific synapses.This property of H1a to accumulate at synapses has beendemonstrated (Okada et al., 2009). The consequence of H1aaccumulation is the facilitated binding of PIN1 to phosphory-lated mGluR5. Operationally, this means that PIN1 bindingdepends on both mGluR5 phosphorylation and the induction

of H1a. Thus, cocaine administration enhances PIN1 bindingand facilitates isomerization selectively for those mGluR5complexes that, by virtue of their subcellular localization andpresence within recently active neurons, are associated withH1a. We therefore inferred that only the small fraction oftotal brain mGluR5 that interacts with synaptic H1a can bindPIN1 following cocaine-induced phosphorylation, suggestingthat the increased mGluR5-PIN1 interaction might be unde-tectable in total brain lysate. We specifically tested thehypothesis that the joint action of mGluR5 phosphorylationand H1a are required for cocaine's behavioral effects, sup-porting a model in which PIN1-mediated isomerization iscentral (Park et al., 2013).

2.3. H1a/mGluR5pS/PIN1 contributes to the in vivo effectsof cocaine

To assess the relevance of mGluR5 phosphorylation to thebehavioral effects of cocaine, we generated another mGluR5genetic substitution model that expresses an mGluR5 mutantthat cannot be phosphorylated at the Homer binding site(APPAPF; mGluR5AA/AA) and is consequently unable to bindPIN1. We compared the effect of repeated cocaine adminis-tration in different genetic models that included H1a�/�,mGluR5R/R, mGluR5AA/AA, and PIN1þ/� mGluR5þ/A. We mon-itored motor sensitization, the process whereby repeatedcocaine exposure results in enhanced motor responses tothe drug. Cocaine motor sensitization was reduced or absentin H1a�/�, mGluR5AA/AA, and PIN1þ/� mGluR5þ/AA, but wasintact in mGluR5R/R (Park et al., 2013). Thus, the ability ofcocaine to sensitize motor response appears to depend jointlyon mGluR5 phosphorylation and H1a, which together enablePIN1 interactions. Accordingly, cocaine motor sensitization

Fig. 2 – H1a potentiates synapses with mGluR5 phosphorylation, depresses synapses without. (A) H1a and mGluR5phosphorylation together enable PIN1 binding, leading to an enhanced NMDAR current. This enhanced current correspondsto a bias towards synaptic potentiation, which may be expressed as an increase in surface expression of AMPARs. (B) H1a,when alone, agonizes mGluR5, causing a downregulation of AMPARs and a depression of synaptic strength. (C) In synapseswithout H1a, neither form of mGluR5-mediated plasticity is expected to occur.

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 8 21

depends on mGluR5-Homer interactions only in as much asthey enable or inhibit the mGluR5-PIN1 interaction. Ongoingstudies will assess the role of mGluR5-PIN1 in cocaine self-administration and cocaine seeking behaviors.

2.4. mGluR5-PIN1 mechanism in synaptic plasticity;requirement for H1a induction creates a temporal contingency

How might the mGluR5-PIN1 interactions influence synapsesto enable cocaine motor sensitization? Motor sensitizationinvolves an enhancement of synaptic input to the striatum ofcocaine-exposed mice (Pascoli et al., 2012). Slice electrophy-siology experiments suggest how cocaine exposure mightinduce this enhancement through D1R and mGluR5 signaling(Centonze et al., 2006). Long-term potentiation (LTP) of stria-tal responsivity can be induced by standard high-frequency

stimulation (HFS) of cortical afferents and subsequentlyreversed by low-frequency synaptic stimulation (LFS). Boththe potentiation and depotentiation of these responsesdepend on NMDA receptors (NMDAR). Interestingly, depoten-tiation can be blocked by treatment of striatal slices with D1Ragonists prior to induction of LTP or by prior repeated cocaineadministration in vivo (Centonze et al., 2006). In fact, thedetails of the necessary conditions for motor sensitizationcorrespond closely with those that prevent depotentiation(Park et al., 2013). The enhancement of total synaptic inputonto the striatum could therefore occur as a consequence ofthe prevention of depotentiation following D1R stimulationby cocaine. Nevertheless, it is worth noting that the mechan-isms mediating potentiation following cocaine exposuremight independently prevent depotentiation. Either way,D1R stimulation can be thought to produce a metaplastic

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 822

state that enhances synaptic strength and consequentlyenhances motor responses to cocaine.

Given that locomotor sensitization depends on mGluR5phosphorylation, H1a induction, and PIN1, we investigatedthe possibility that these events are also involved in D1R'sability to block depotentiation. To determine if the poten-tiating metaplastic effect of D1R is mediated by phos-phorylated mGluR5 and H1a, we examined acute striatalslices from mGluR5AA/AA, mGluR5R/R, and H1a�/� mice.LTP and depotentiation were intact in all three mouselines, but D1R activation failed to block depotentiationonly in the mGluR5AA/AA and H1a�/� mice. Thus, inductionof H1a and phosphorylation of mGluR5 are both requiredfor D1R to exert its metaplastic effect. Moreover, theresponse to D1R agonist was different in mGluR5R/R micecompared to littermate wildtype mice in an interestingway. In wildtype slices, D1R agonist must be applied�30 min prior to the depotentiation stimulus for it toblock depotentiation. However, in mGluR5R/R mice, D1Ragonist could be applied immediately before the depoten-tiation stimulus and still block depotentiation. In wildtypebut not mGluR5R/R mice, H1a is required to compete withcross-linking forms of Homer to enable PIN1 binding tophosphorylated mGluR5. However, because mGluR5F1128Rdoes not bind Homer, PIN1 interactions are possibleimmediately upon phosphorylation without the inductionof H1a. This finding implies that D1R stimulation caninduce both mGluR5 phosphorylation and H1a synthesis(the two events necessary for PIN1 binding) but at differenttimes; pretreatment is needed in wildtype mice to allowsufficient time for H1a to be synthesized. Thus, thepotentiating metaplastic effect of D1R stimulationdepends not only on the phosphorylation of mGluR5 buton the presence of H1a, induced by events that are several10 s of minutes (decaminutes) old. Restated, time con-straints indicate how D1R stimulation may metaplasti-cally alter synaptic strength: the mGluR5-PIN1 mechanismwill not have an effect in synapses that do not havedecaminutes-old histories of activity.

2.5. PIN1 isomerization of mGluR5 enhances coupling toNMDA type glutamate receptors

By what mechanism might mGluR5–PIN1 interactions enablepotentiating metaplasticity? Stimulation of the mGluR5–PIN1interaction in striatal cultures enhanced the slow inwardcurrent (SIC) that normally follows mGluR5 activation(mGluR5-SIC) (Park et al., 2013). In conditions that enablemGluR5–PIN1 interactions, treatment of cells with D1R ago-nist resulted in a rapid ERK-dependent increase in theamplitude of the mGluR5-SIC with D1R-expressing neurons.In particular, striatal cultures lacking Homer (Homer1�/�, 2�/

�, 3�/�), lacking mGluR5–Homer interactions (cultures pre-pared from mGluR5R/R), or in which H1a was introduced byvirus or induced through KCl application, all exhibited anenhanced mGluR5-SIC following ERK stimulation. By con-trast, cultures in which H1a was not induced (wildtype,untreated) or could not be induced (H1a�/�, w/wo KCltreatment) exhibited normal baseline mGluR5-SICs that were

not enhanced by ERK stimulation. Similarly, cultures inwhich mGluR5 phosphorylation could not be induced(mGluR5AA/AA mutants, w/wo H1a virally introduced) alsoexhibited normal baseline mGluR5-SICs that were unaffectedby ERK stimulation. Thus, as with cocaine locomotor sensi-tization and the potentiating metaplastic effect of D1R sti-mulation, mGluR5-SIC potentiation also requires bothmGluR5 phosphorylation and H1a. Furthermore, amplifica-tion of mGluR5-SIC could be inhibited by two different classesof PIN1 inhibitors and was absent in cultures generated fromPIN1�/� mice. It is therefore reasonable to conclude that themGluR5–PIN1 interactions enabled by the phosphorylation ofmGluR5 and the induction of H1a are critical to all three ofthese phenomena.

There is even further evidence that the mGluR5-SIC isrelated to potentiating metaplastic effects of D1R stimulationand induction of locomotor sensitization following cocaineexposure. Across the many listed conditions of striatalculture, pharmacological investigations revealed that boththe normal and the enhanced mGluR5-SIC are mediated byan NMDAR current (Park et al., 2013). All of these findingspoint to one integrated model of cocaine's behavioral effects.D1R stimulation causes mGluR5 phosphorylation, which incombination with induced H1a, enables mGluR5–PIN1 inter-actions to create the potentiating metaplastic condition ofenhanced NMDAR currents (Fig. 1). When these NMDARs arestimulated, their enhanced current facilitates an overallpotentiation of synaptic input onto the striatum. Theenhanced synaptic input into the striatum mediates theanimal's sensitized response to cocaine.

While it is plausible to envision mechanisms connectingan enhanced NMDAR current to the behavioral effects ofcocaine, we do not yet understand how mGluR5 isomeriza-tion might enable this enhanced NMDAR current. A biochem-ical explanation of the effect of mGluR5–PIN1 interactions onNMDARs awaits future investigations.

3. The H1a/mGR5pS/PIN1 mechanismdescribes an overlap between reward learning andsynaptic tagging, and predicts novel features ofneuronal selection

The H1a/mGR5pS/PIN1 mechanism therefore describes a setof molecular events that cause synaptic plasticity. Impor-tantly, potentiation occurs through this mechanism as aconsequence of the coordinated activity of the neuron, itssynaptic input, and reward. The temporal relationshipbetween combinations of these events is the topic of theore-tical models of learning and plasticity. The H1a/mGR5pS/PIN1mechanism therefore elaborates and merges these theoreti-cal models. It suggests that reward learning depends uponthe synaptic recruitment of plasticity related proteins (PRPs)whose global synthesis is induced by decaminute-old pat-terns of neuronal firing. It also suggests that reward learningoccurs selectively on synapses experiencing coincident inputand reward in neurons that are sufficiently active in theirnetworks.

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 8 23

3.1. The H1a/mGR5pS/PIN1 mechanism detects the co-occurrence of glutamatergic, neuromodulatory, and neuronalevents and depends on reactivation over decaminutes

Stated specifically, H1a/mGR5pS/PIN1 potentiating metaplas-ticity can be realized as a consequence of four events: (1)neuronal firing induces H1a (Fig. 3A-α), after which (2)synaptic activity, when closely incident with (3) D1R stimula-tion, elevates pERK levels (Fig. 3A-β). High pERK causesphosphorylation of mGluR5 that endures until (4) synapticreactivation occurs decaminutes after (1) (Fig. 3A-γ). Synapticreactivation in the context of mGR5pS and H1a enables orpreserves potentiation (Fig. 3A-δ and C). Notably, event (1)happens at the neuron-level, but (2), (3), and (4) co-occur inselect synapses (Figs. 3A-S1 and C). Upon reactivation ofprevious potentiated synapses by LFS, the H1a/mGR5pS/PIN1 potentiating metaplasticity mechanism protectssynapses meeting these conditions from depotentiation(Centonze et al., 2006; Otmakhova and Lisman, 1998; Park

Fig. 3 – H1a/mGR5pS/PIN1 potentiating metaplasticity requires dSynapses S1, S2, and S3 on two neurons (A and B) are comparedBetween A and B, each of these synapses has identical synapticvary in their relative time of neuronal firing. (α) Neuron A fires endepicted, whereas Neuron B fires similarly only seconds beforedecaminutes before Neuron B. In both neurons, (β) pERK is activa(S3). (γ) S1 and S3 are subsequently active. (C) The combination oA enhances synaptic strength. (D) In synapses without pERK thatS3 of Neuron A, H1a downregulates synaptic strength. (E) In syninducing activity, no H1a-mediated plasticity occurs.

et al., 2013). While necessary for blocking depotentiation, thismechanism might also operate to enhance synaptic strengthin other contexts. Indeed, stimulation of D1Rs or otherneuromodulator receptors can cause potentiation after nor-mally benign or depressing synaptic activity (Shen et al., 2008;Pawlak et al., 2010). These phenomena all serve as examplesof potentiating metaplasticity in which neuromodulators biassome synapses towards potentiation.

Models of synaptic plasticity that reference subsets ofthese four neuronal and synaptic events have been pre-viously proposed, including those that invoke a “synapticeligibility trace,” a “synaptic tag,” or a process of neuronalselection influenced by patterns of reactivation. The H1a/mGR5pS/PIN1 mechanism suggests how these distinct mod-els might be integrated and extended. It provides an exampleof how neuronal firing, synaptic glutamate, neuromodula-tion, and reactivation can all interact to allow a network toencode reward-related experiences. In particular, the H1a/mGR5pS/PIN1 mechanism integrates the “synaptic eligibility

ecaminute-old neuronal activity prior to synaptic activation.to illustrate the effect of neuronal firing patterns on plasticity.activity and pERK patterns, but the neurons they belong toough to induce H1a decaminutes before the synaptic events

. Neuron A can therefore exhibit synaptically-localized H1ated in synapse 1 (S1) and synapse 2 (S2), but not in synapse 3f H1a, pERK, and subsequent NMDAR activity in S1 of Neuronalso belong to neurons with decaminute-old activity, such asapses belonging to neurons without decaminute-old H1a-

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 824

trace” and “synaptic tag” models. While synaptic tags recruitH1a to all active synapses, those synapses whose activity isuncorrelated with reward are downregulated through H1a'sagonism of mGluR5 (Hu et al., 2010), whereas those whoseactivity correlates with reward can exhibit potentiationthrough enhancement of NMDAR current (Park et al., 2013).Further, a requirement for H1a in metaplasticity provides amechanistic explanation for neuronal selection favoring neu-rons with recent activity (Rogerson et al., 2014) and suggeststhat decaminutes-old activity may be of particularimportance.

3.2. The “synaptic eligibility trace” model describesglutamatergic and neuromodulatory contingencies; “synaptictags” further relate neuronal events to synaptic plasticity

The earliest models of learning and plasticity highlighted thehypothesis that potentiation should occur when postsynapticneuronal firing is coincident with presynaptic input (Mageeand Johnston, 1997). Although these models have clearexperimental support and identifiable molecular mechan-isms (Cotman and Monaghan, 1988), they do not explainhow the association between more temporally distant events—like a cue followed by a reward a second later—might occur.Further, because an outcome and its predictor may beseparated in time, it can be a challenge to identify which ofthe many preceding events are predictors. To solve this“distal reward problem,” and the related “credit assignmentproblem,” reinforcement learning theoreticians proposed thatan “eligibilty trace” is initiated by the first event (cue) anddecays until the second (reinforcing) event occurs (Fosteret al., 2000; Izhikevich, 2007). The cue's value is augmented inproportion to the remaining trace, or “protoweight” (Gavorniket al., 2009). In this way, value is readily assigned to cues thatpredict reward at temporal delays. Over many iterations, thisprocess can ultimately separate random cue-reward coinci-dences from robust predictions (Izhikevich, 2007). A cellularversion of this general conception imagines that cue-relatedglutamatergic input initiates a decaying synaptic state orsignal (an eligibility trace) in the activated synapse (Shouvaland Gavornik, 2011). Upon encountering reward-communicating neuromodulators, the remaining trace isconverted into a new synaptic weight. Through this mechan-ism, synapses that reliably receive input before reward arestrengthened until more distal reward relationships arelearned. Computational models of synaptic eligibility tracehave replicated the neural findings in cue-reward learningparadigms (Gavornik et al., 2009; Shuler and Bear, 2006).

However, the synaptic eligibility trace model presumeschanges in synaptic strength occur instantaneously andtherefore does not address the biological mechanisms ofsynaptic plasticity or their limitations. In particular, severalforms of synaptic plasticity depend on de novo synthesis ofPRPs in the hours following their induction by sufficientneuronal firing (Barondes, 1965; Agranoff et al., 1965;Flexner et al., 1965). But how can a neuron-wide event liketranscription enable synapse-specific plasticity? In otherwords, how do nascent proteins exert their effects on onlythe right synapses? To explain how proteins that are tran-scribed in the nucleus might act selectively on synapses

undergoing plasticity, a “synaptic tag” was imagined (Freyand Morris, 1997). In this model, glutamatergic input tagsrecently active synpases; this tag then selectively recruitsnascent PRPs.

The eligibility trace model can be extended to include thisprocess. Synaptic activity initiates a decaying protoweight,which upon reward, is converted into a provisional weight.Such a provisional weight is only realized as an actual changein synaptic strength upon interaction with PRPs. We proposethat pERK and mGluR5 phosphorylation represent suchprovisional weights; they interact with the PRP H1a to enablepotentiation of rewarded synapses. Notably, high pERK activ-ity results from the coactivation of NMDARs and D1Rs(Valjent et al., 2004; Sarantis et al., 2009) or mGluR and D1R(Voulalas et al., 2005), detecting the coincidence of glutama-tergic input and reinforcement by dopamine (Cahill et al.,2014). cAMP is coordinately activated with ERK in manymodels, and as expected of such a provisional weight model,the cAMP-dependent effects of dopamine on synapsesdepend on the relative timing of glutamate and dopaminerelease. The optimal response occurs when dopamine isdelivered within the first second of glutamate release; how-ever, no effect is observed when dopamine is delivered twoseconds after glutamate release (Yagishita et al., 2014).

3.3. Synaptic tags recruit H1a to punish unreinforcedactive synapses but enhance reactivated, reinforced activesynapses

Like H1a/mGluR5pS/PIN1 potentiating metaplasticity, synap-tic tag-related plasticity describes a synaptic event thatdepends on IEG induction and PRP synthesis. In fact, becauseit was known to be synthesized de novo following neuronalactivity and to interact with mGluR5 at the synapse, the IEGH1a was one of the first candidate PRPs (Kato et al., 2001).Consistent with its hypothesized interaction with a theore-tical synaptic tag, H1a was more recently shown to beselectively recruited to recently active synapses (Okadaet al., 2009). The emerging knowledge of H1a's role in synapticplasticity therefore provides new insight into the interfacebetween synaptic activity, synaptic tagging, and rewardlearning.

In conditions that do not induce pERK or mGR5pS, H1adownregulates synaptic strength (Hu et al., 2010). Three typesof synapses can therefore be described: those without suffi-cient synaptic activity to recruit H1a, those with synapticactivity that is not coincident with reward, and those withsynaptic activity coincident with reward. Because it is notrecruited there, H1a will not affect the synaptic strength ofinactive synapses (Fig. 2B). In contrast, H1a downregulatesthe synaptic strength of active, but unrewarded synapses inwhich pERK and mGR5pS are not abundant (Fig. 2C). Finally,H1a enables potentiation of synapses whose activity is coin-cident with reward, in which pERK and mGR5pS levels arehigh (Fig. 2A). This model modifies an earlier view that IEGshomeostatically downregulate all synapses (Hu et al., 2010;Shepherd et al., 2006). If H1a's synaptic recruitment anddownregulatory function is proportional to the level of recentsynaptic input (given little pERK activity), this might helpexplain why activity-driven downregulation has been

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 8 25

observed to scale with initial synaptic strength (Turrigianoand Nelson, 2004). In other words, scaling of synapticstrength could be explained by the observation that H1a isinduced in active neurons and recruited to synapses inproportion to their synaptic input.

Interestingly, another IEG, Arc, has also been shown toexert a synapse-specific effect: it accumulates preferentiallyin inactive synapses (Okuno et al., 2012). In contrast to H1a,Arc's function has not been shown to depend on a synapse'sreward history and has only been shown to decrease synapticstrength (Shepherd and Bear, 2011). Additionally, there is noevidence that Arc transcription is necessary decaminutesbefore the induction of the forms of plasticity for which it isrequired (Park et al., 2008). In fact, Arc-dependent mGluR-LTDcan be expressed without any specific protocol that inducesArc transcription decaminutes beforehand. As discussed inSection 3.4, the requirement for H1a induction decaminutesbefore the time of plasticity induction might inform neuronalselection, restricting plasticity to neurons with sufficientdecaminutes-old firing rates to have induced H1a. It iscurrently unknown if Arc might similarly contribute toneuronal selection.

In summary, H1a is a PRP induced by neuronal activity;pERK and mGR5pS are provisional weights, induced byreinforced synaptic activity. In neurons with sufficient activ-ity to induce H1a, unreinforced active synapses might exhibitlow pERK and low mGR5pS and undergo reductions in

Fig. 4 – The requirement for decaminute-old neuronal activity fi

propose that the H1a/mGR5pS/PIN1 mechanism may mean that tsignificant decaminute-old activity. (1) Many synapses experiensome of these synapses have previously experienced input-rew(3) Only some of these synapses belong to neurons with sufficieactivity sufficient for H1a to be present in the synapse at the timactivity of these neurons will influence their downstream targe

synaptic strength. Reinforced active synapses will exhibithigh mGR5pS and enable potentiation upon reactivation.

3.4. H1a-dependent metaplasticity might direct neuronalselection to neurons with decaminutes-old activity

Like all PRPs hypothesized to be recruited by synaptic tags,H1a is only induced by sufficient neuronal activity. Thismeans synaptic input's ability to modify synaptic strengthdepends on the recent firing history of its parent neuron.Notably, H1a synthesis takes more than 15 min following itsinduction (Vazdarjanova et al., 2002).

While the coordinated activity of a neuron and itssynapses has been repeatedly hypothesized to influencesynaptic plasticity, the H1a/mGR5pS/PIN1 mechanism furtherentails a novel requirement for its realization: sufficientneuronal activity decaminutes prior to synaptic activity(Figs. 3 and 4). For a synapse with a synaptic tag and mGR5pSprovisional weight to realize synaptic change, it must belongto a neuron that exhibited sufficient activity decaminutesearlier.

This requirement is further entailed by the H1a/mGR5pS/PIN1 potentiating metaplasticity mechanism's dependenceon reactivation: an enhanced NMDAR current is convertedinto a change in synaptic strength only by glutamatergicactivation of NMDARs in synapses that already have H1a andmGR5pS from previous events—H1a from sufficient

lters for synapses on select neurons within networks. Wehe synapses that learn an event only belong to neurons withce NMDAR activation, depicted by green asterisks. (2) Onlyard pairings sufficient to activate pERK, depicted in pink.nt recent firing, of which (4) fewer have decaminutes-olde of NMDAR stimulation. (5) Subsequent changes in the

ts.

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 826

decaminute-old neuronal activity and mGR5pS from morerecent input-reward pairing.

Consider the following illustration: imagine equivalentcorticostriatal synapses that receive D1R stimulation milli-seconds after HFS (causing pERK and mGluR5 phosphoryla-tion) and several seconds before LFS (causing the synapticreactivation on which metaplasticity operates). A synapsebelonging to a neuron that fired frequently more than 30 minago (sufficient time for H1a synthesis) would remain poten-tiated, whereas a synapse belonging to a neuron that had notfired or had only fired 5 min prior (insufficient time for H1asynthesis), would become depotentiated. In the second case,H1a would not yet be present near synapses. Potentiatingmetaplasticity mediated by the H1a/mGR5pS/PIN1 mechan-ism would therefore only be realized in neurons reactivatedover decaminute-long timeframes. The H1a/mGR5pS/PIN1model therefore suggests a novel constraint on how rewardinfluences synaptic strength: the metaplastic effect of neuro-modulators like dopamine might only matter in neuronsrepeatedly activated over decaminutes. Such a constrainthas not previously been applied to models of reward-relatedlearning and could produce novel results.

Interestingly, the prior activity of neurons within a net-work has been shown to influence synaptic plasticity and thenetwork's subsequent encoding of its experiences (Rogersonet al., 2014; Yiu et al., 2014). Although many neurons withinthe lateral amygdala show similar initial responses to a cueprior to its pairing with an aversive shock, only some of theseresponses will become potentiated by cue-shock pairings.The process by which some of many possible neurons cometo encode an experience has been termed “neuronal alloca-tion“ or “neuronal selection.” Presumably, which neuronslearn this cue-shock pairing is relevant for how the networkencodes its experiences more generally (Ghosh and Chattarji,2015). Although many neurons in the network report the cue,only some of them express augmented responses. The fearresponse will become more likely in conditions that activatethe selected neurons and comparatively less likely in condi-tions activating cue-responsive neurons that did not undergoplasticity. Thus, the cue-shock pairing might be better under-stood as a more complex condition-cue-shock pairing inwhich the conditions are defined by the subset of neuronsthat undergo plasticity due to their prior activity. The net-work might therefore partially encode experience through theenvironmental features that are repeated over decaminutesand not those that are more uniquely observed at the time ofreward. The dependence of reward plasticity on IEG PRPswithin synapses might therefore entail a mechanism bywhich those contextual features that are robust enough tobe repeated over decaminute timeframes are selectivelyincorporated into a network's memory.

Consistently, artificial induction of neuronal activity byexperimenters can bias the neuronal selection processtowards the activated neurons when activity precedes cue-shock pairing, but not when it occurs afterwards (Rogersonet al., 2014; Yiu et al., 2014). Although the hypothesis thatdecaminute-old activity and H1a (or other PRP IEGs) contri-bute to neuronal selection is novel and has theoreticalappeal, it has yet to be demonstrated. The precise temporalfeatures of the activity required for neuronal selection have

not been investigated in detail, and the effect ofdecaminutes-old activity compared to more recent neuronalfiring has not been demonstrated. Nevertheless, such a modelpredicts that prior induction of H1a would be a particularlygood predictor of which neurons come to encode an experi-ence. The coordinated imaging of H1a induction and changesin response of a neural network during a behavioral learningparadigm could provide confirmation of the broader role H1amight serve in neuronal selection and the subsequent dimen-sionality of a neural network. We note that the mGluR5R/R

mouse does not require H1a for neuromodulator-dependentpotentiating metaplasticity and may be useful to test the roleof mGR5pS/PIN1 in neuronal selection.

Interestingly, H1a might not be required to enableneuromodulator-dependent potentiating metaplasticity indiseases where mGluR5-Homer interactions are abnormal.For example, in mouse models of Fragile X syndrome,mGluR5 exhibits reduced interactions with long-form Homer(Ronesi, et al., 2012) and may be less dependent upon H1a forneuromodulator-induced PIN1 binding. Similarly, age-relatedmemory impairment involves a decrease in mGluR5-long-form Homer interactions (Menard and Quirion, 2012). Disrup-tions in long-form Homer function have also been proposedto contribute to schizophrenia (Szumlinski et al., 2005;Spellman, et al., 2011). H1a's proposed role in neuronalselection suggests that in these cases, reward experiencemight be encoded much more promiscuously and much lessspecifically, perhaps altering many of the basic ways in whichthese nervous systems represent the world. A disruption ofthe proposed neuronal selection mechanism might explaincognitive dysfunction in these cases.

4. Conclusions

In conclusion, pERK and mGluR5 phosphorylation may serveas provisional weights, detecting the coincidence betweenglutamatergic synaptic input and reinforcement by dopa-mine. mGR5pS may be realized as potentiating metaplasticityin neurons in which H1a is induced by sufficient neuronalactivity. H1a, recruited to active synapses, might depressunrewarded active synapses and potentiate reactivatedrewarded ones. This process will selectively occur withinneurons based on their decaminutes-old activity histories,filtering the encoding of new experiences into those featuresthat persist over this timeframe. The H1a/mGR5pS/PIN1mechanism therefore enables a novel synthesis of systems-level and cellular models of learning.

Acknowledgments

This work was supported by NIDA 010309. We thank JenniferFairman for her artwork in the figures.

r e f e r e n c e s

Agranoff, B.W., Davis, R.E., Brink, J.J., 1965. Memory fixation in thegoldfish. Proc. Natl. Acad. Sci. USA 54, 788–793.

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 8 27

Barondes, S.H., 1965. Relationship of biological regulatorymechanisms to learning and memory. Nature 205, 18–21.

Beneken, J., et al., 2000. Structure of the Homer EVH1 domain-peptide complex reveals a new twist in polyprolinerecognition. Neuron 26, 143–154.

Bird, M.K., Lawrence, A.J., 2009. The promiscuous mGluR5receptor–a range of partners for therapeutic possibilities?.Trends Pharmacol. Sci. 30, 617–623.

Bottai, D., et al., 2002. Synaptic activity-induced conversion ofintronic to exonic sequence in Homer 1 immediate early geneexpression. J. Neurosci. 22, 167–175.

Brakeman, P.R., et al., 1997. Homer: a protein that selectivelybinds metabotropic glutamate receptors. Nature 386, 284–288.

Cahill, E., et al., 2014. Convergence of dopamine and glutamatesignaling onto striatal ERK activation in response to drugs ofabuse. Front. Pharmacol. 4, 172.

Carelli, R.M., Wightman, R.M., 2004. Functional microcircuitry inthe accumbens underlying drug addiction: insights from real-time signaling during behavior. Curr. Opin. Neurobiol. 14,763–768.

Centonze, D., et al., 2006. Chronic cocaine preventsdepotentiation at corticostriatal synapses. Biol. Psychiatry 60,436–443.

Chiamulera, C., et al., 2001. Reinforcing and locomotor stimulanteffects of cocaine are absent in mGluR5 null mutant mice.Nat. Neurosci. 4, 873–874.

Cotman, C.W., Monaghan, D.T., 1988. Excitatoy amino acidneurotransmission: NMDA receptors and Hebb-type synapticplasticity. Ann. Rev. Neurosci. 11, 61–80.

Davis, H.P., Squire, L.R., 1984. Protein synthesis and memory: areview. Psychol. Bull. 96, 518–559.

Flexner, L.B., et al., 1965. Loss of memory as related to inhibitionof cerebral protein synthesis. J. Neurochem. 12, 535–541.

Foster, D., Morris, R.G., Dayan, P., 2000. A model of hippocampallydependent navigation, using the temporal difference learningrule. Hippocampus 10, 1–16.

Frey, U., Morris, R.G., 1997. Synaptic tagging and long-termpotentiation. Nature 385, 533–536.

Gavornik, J.P., et al., 2009. Learning reward timing in cortexthrough reward dependent expression of synaptic plasticity.Proc. Natl. Acad. Sci. USA 106, 6826–6831.

Gerfen, C.R., 2003. D1 dopamine receptor supersensitivity in thedopamine-depleted striatum animal model of Parkinson’sdisease. Neuroscientist 9, 455–462.

Gerfen, C.R., Paletzki, R., Worley, P., 2008. Differences betweendorsal and ventral striatum in Drd1a dopamine receptorcoupling of dopamine- and cAMP-regulated phosphoprotein-32 to activation of extracellular signal-regulated kinase. J.Neurosci. 28, 7113–7120.

Ghosh, S., Chattarji, S., 2015. Neuronal encoding of the switchfrom specific to generalized fear. Nat. Neurosci. 18, 112–120.

Guzowski, J.F., et al., 1999. Environment-specific expression of theimmediate-early gene Arc in hippocampal neuronalensembles. Nat. Neurosci. 2, 1120–1124.

Hayashi, M.K., et al., 2009. The postsynaptic density proteinsHomer and Shank form a polymeric network structure. Cell137, 159–171.

Hu, J.H., et al., 2010. Homeostatic scaling requires group I mGluRactivation mediated by Homer1a. Neuron 68, 1128–1142.

Hyman, S.E., Malenka, R.C., Nestler, E.J., 2006. Neuralmechanisms of addiction: the role of reward-related learningand memory. Annu. Rev. Neurosci. 29, 565–598.

Izhikevich, E.M., 2007. Solving the distal reward problem throughlinkage of STDP and dopamine signaling. Cereb. Cortex. 17,2443–2452.

Kato, A., et al., 2001. Phorbol esters promote postsynapticaccumulation of Vesl-1S/Homer-1a protein. Eur. J. Neurosci.13, 1292–1302.

Kenny, P.J., et al., 2003. Metabotropic glutamate 5 receptor

antagonist MPEP decreased nicotine and cocaine self-

administration but not nicotine and cocaine-induced

facilitation of brain reward function in rats. Ann. N.Y. Acad.

Sci. 1003, 415–418.Kim, D.S., et al., 2006. Reversal of supersensitive striatal

dopamine D1 receptor signaling and extracellular signal-

regulated kinase activity in dopamine-deficient mice.

Neuroscience 137, 1381–1388.Koob, G.F., Bloom, F.E., 1988. Cellular and molecular mechanisms

of drug dependence. Science 242, 715–723.Koob, G.F., 1992. Drugs of abuse: anatomy, pharmacology and

function of reward pathways. Trends Pharmacol. Sci. 13,

177–184.Lu, K.P., et al., 2007. Prolyl cis-trans isomerization as a molecular

timer. Nat. Chem. Biol. 3, 619–629.Lu, K.P., Zhou, X.Z., 2007. The prolyl isomerase PIN1: a pivotal

new twist in phosphorylation signalling and disease. Nat. Rev.

Mol. Cell Biol. 8, 904–916.Luo, A.H., et al., 2011. Linking context with reward: a functional

circuit from hippocampal CA3 to ventral tegmental area.

Science 333, 353–357.Luscher, C., Malenka, R.C., 2011. Drug-evoked synaptic plasticity

in addiction: from molecular changes to circuit remodeling.

Neuron 69, 650–663.Magee, J.C., Johnston, D., 1997. A synaptically controlled,

associative signal for Hebbian plasticity in hippocampal

neurons. Science 275, 209–213.Menard, C., Quirion, R., 2012. Successful cognitive aging in rats: a

role for mGluR5 glutamate receptors, Homer1 proteins and

downstream signaling pathways. PLoS One 7, e28666.Montarolo, P.G., Kandel, E.R., Schacher, S., 1988. Long-term

heterosynaptic inhibition in Aplysia. Nature 333, 171–174.Nicoletti, F., et al., 2011. Metabotropic glutamate receptors: from

the workbench to the bedside. Neuropharmacology 60,

1017–1041.Okada, D., Ozawa, F., Inokuchi, K., 2009. Input-specific spine entry

of soma-derived Vesl-1S protein conforms to synaptic tagging.

Science 324, 904–909.Okuno, H., et al., 2012. Inverse synaptic tagging of inactive

synapses via dynamic interaction of Arc/Arg3.1 with CaMKIIβ.Cell 149, 886–898.

Orlando, L.R., et al., 2009. Phosphorylation of the homer-binding

domain of group I metabotropic glutamate receptors by

cyclin-dependent kinase 5. J. Neurochem. 110, 557–569.Otmakhova, N.A., Lisman, J.E., 1998. D1/D5 dopamine receptors

inhibit depotentiation at CA1 synapses via cAMP-dependent

mechanism. J. Neurosci. 18, 1270–1279.Park, S., et al., 2008. Elongation factor 2 and fragile X mental

retardation protein control the dynamic translation of Arc/

Arg3.1 essential for mGluR5-LTD. Neuron 59, 70–83.Park, J.M., et al., 2013. A prolyl-isomerase mediates dopamine-

dependent plasticity and cocaine motor sensitization. Cell

154, 637–650.Pascoli, V., Turiault, M., Luscher, C., 2012. Reversal of cocaine-

evoked synaptic potentiation resets drug-induced adaptive

behaviour. Nature 481, 71–75.Paterson, N.E., Markou, A., 2005. The metabotropic glutamate

receptor 5 antagonist MPEP decreased break points for

nicotine, cocaine and food in rats. Psychopharmacology 179,

255–261.Pawlak, V., et al., 2010. Timing is not everything:

neuromodulation opens the STDP gate. Front. Synaptic

Neurosci. 2, 146.Rogerson, T., et al., 2014. Synaptic tagging during memory

allocation. Nat. Rev. Neurosci. 15, 157–169.

b r a i n r e s e a r c h 1 6 2 8 ( 2 0 1 5 ) 1 7 – 2 828

Ronesi, J.A., et al., 2012. Disrupted Homer scaffolds mediateabnormal mGluR5 function in a mouse model of fragile Xsyndrome. Nat. Neurosci. 15, 431–440.

Sarantis, K., Matsokis, N., Angelatou, F., 2009. Synergisticinteractions of dopamine D1 and glutamate NMDA receptorsin rat hippocampus and prefrontal cortex: involvement ofERK1/2 signaling. Neuroscience 163, 1135–1145.

Shen, W., et al., 2008. Dichotomous dopaminergic control ofstriatal synaptic plasticity. Science 321, 848–851.

Shepherd, J.D., et al., 2006. Arc/Arg3.1 mediates homeostaticsynaptic scaling of AMPA receptors. Neuron 52, 475–484.

Shepherd, J.D., Bear, M.F., 2011. New views of Arc, a masterregulator of synaptic plasticity. Nat. Neurosci. 14, 279–284.

Shouval, H.Z., Gavornik, J.P., 2011. A single spiking neuron thatcan represent interval timing: analysis, plasticity and multi-stability. J. Comput. Neurosci. 30, 489–499.

Shuler, M.G., Bear, M.F., 2006. Reward timing in the primary visualcortex. Science 311, 1606–1609.

Spellman, I., et al., 2011. Homer-1 polymorphisms are associatedwith psychopathology and response to treatment inschizophrenic patients. J. Psychiatr. Res. 45, 234–241.

Swanson, C.J., et al., 2001. Repeated cocaine administrationattenuates group I metabotropic glutamate receptor-mediatedglutamate release and behavioral activation: a potential rolefor Homer. J. Neurosci. 21, 9043–9052.

Szumlinski, K.K., et al., 2004. Homer proteins regulate sensitivityto cocaine. Neuron 43, 401–413.

Szumlinski, K.K., et al., 2005. Behavioral and neurochemicalphenotyping of Homer1 mutant mice: possible relevance toschizophrenia. Genes Brain Behav. 4, 273–288.

Szumlinski, K.K., Kalivas, P.W., Worley, P.F., 2006. Homer proteins:implications for neuropsychiatric disorders. Curr. Opin.Neurobiol. 16, 251–257.

Tu, J.C., et al., 1998. Homer binds a novel proline-rich motif and

links group 1 metabotropic glutamate receptors with IP3

receptors. Neuron 21, 717–726.Tu, J.C., et al., 1999. Coupling of mGluR/Homer and PSD-95

complexes by the Shank family of postsynaptic density

proteins. Neuron 23, 583–592.Turrigiano, G.G., Nelson, S.B., 2004. Homeostatic plasticity in the

developing nervous system. Nat. Rev. Neurosci. 5, 97–107.Valjent, E., et al., 2004. Addictive and non-addictive drugs induce

distinct and specific patterns of ERK activation in mouse

brain. Eur. J. Neurosci. 19, 1826–1836.Vazdarjanova, A., et al., 2002. Experience-dependent coincident

expression of the effector immediate-early genes arc and H1a

in hippocampal and neocortical neuronal networks. J.

Neurosci. 22, 10067–10071.Voulalas, P.J., et al., 2005. Metabotropic glutamate receptors and

dopamine receptors cooperate to enhance extracellular

signal-regulated kinase phosphorylation in striatal neurons. J.

Neurosci. 25, 3763–3773.Xiao, B., et al., 1998. Homer regulates the association of group 1

metabotropic glutamate receptors with multivalent

complexes of homer-related, synaptic proteins. Neuron 21,

707–716.Xiao, B., Tu, J.C., Worley, P.F., 2000. Homer: a link between neural

activity and glutamate receptor function. Curr. Opin.

Neurobiol. 10, 370–374.Yagishita, S., et al., 2014. A critical time window for dopamine

actions on the structural plasticity of dendritic spines. Science

345, 1616–1620.Yiu, A.P., et al., 2014. Neurons are recruited to a memory trace

based on relative neuronal excitability immediately before

training. Neuron 83, 722–735.

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具