MicroRNA regulation of homeostatic synaptic plasticityJonathan E. Cohena, Philip R. Leea, Shan Chenb, Wei Lib, and R. Douglas Fieldsa,1

aSection on Nervous System Development and Plasticity, The Eunice Kennedy Shriver National Institute of Child and Human Development, National Institutesof Health, Bethesda, MD 20892-3714; and bUnit on Retinal Neurophysiology, National Eye Institute, National Institutes of Health, Bethesda, MD 20892-2510

Edited by Michael Eldon Greenberg, Children’s Hospital Boston, Boston, MA, and approved May 31, 2011 (received for review January 12, 2011)

Homeostatic mechanisms are required to control formation andmaintenanceof synaptic connections tomaintain thegeneral level ofneural impulse activity within normal limits. How genes controllingthese processes are co-coordinately regulated during homeostaticsynapticplasticity is unknown.MicroRNAs (miRNAs)exert regulatorycontrol over mRNA stability and translation and may contribute tolocal and activity-dependent posttranscriptional control of synapse-associated mRNAs. However, identifying miRNAs that functionthroughposttranscriptional gene silencingat synapses has remainedelusive. Using a bioinformatics screen to identify sequence motifsenriched in the 3′UTR of rapidly destabilized mRNAs, we identifieda developmentally and activity-regulatedmiRNA (miR-485) that con-trols dendritic spine number and synapse formation in an activity-dependent homeostatic manner. We find that many plasticity-associatedgenes containpredictedmiR-485binding sitesand furtheridentify thepresynapticproteinSV2Aasa targetofmiR-485.miR-485negatively regulateddendritic spinedensity, postsynaptic density95(PSD-95) clustering, and surface expression of GluR2. Furthermore,miR-485 overexpression reduced spontaneous synaptic responsesand transmitter release, as measured by miniature excitatory post-synaptic current (EPSC) analysis and FM 1–43 staining. SV2A knock-down mimicked the effects of miR-485, and these effects werereversed by SV2A overexpression. Moreover, 5 d of increased syn-aptic activity induced homeostatic changes in synaptic specializa-tions that were blocked by a miR-485 inhibitor. Our findings reveala role for this previously uncharacterizedmiRNAand thepresynapticprotein SV2A in homeostatic plasticity and nervous system develop-ment, with possible implications in neurological disorders (e.g., Hun-tington and Alzheimer’s disease), where miR-485 has been found tobe dysregulated.

activity-dependent development | posttranscriptional gene regulation |presynaptic terminal

Homeostatic synaptic plasticity is the process of regulatingsynaptic connections to compensate for changes in levels of

impulse activity in neural networks over a time course of days,thus maintaining circuit function within an optimal range (1).This homeostasis limits changes in synaptic strength induced byHebbian synaptic plasticity (2), compensates for nervous systemdisorders causing hyper- or hypoexcitability, and adjusts excitabilityof neural circuits during development. How genes involved in thisprocess are coordinately regulated is unknown.Increasing excitability of hippocampal or cortical neurons in

culture using bicuculline (BiC) reduces synaptic strength throughboth pre- and postsynaptic mechanisms, and decreasing excit-ability with TTX produces the opposite response (1, 3, 4). Severalmolecules have been identified that contribute to homeostaticsynaptic plasticity over different time scales, including alterationsin the α/β Ca2+/CaM-dependent kinase II (CaMKII) ratio (5),astrocytic TNF-α release (6), regulation of cell surface AMPAreceptors (7), Arc/Arg3.1 expression (4), and polo-like kinase 2(8). Structural changes in synaptic connectivity that follow physi-ological changes in synaptic strength must involve gene regulatorynetworks controlling synaptic development, maturation, andmaintenance. MicroRNAs (miRNAs) rapidly and coordinatelyregulate stability and translation of sets of mRNAs mediatingspecific processes (9, 10), suggesting that miRNAs could have animportant role in homeostatic synaptic plasticity. In response to

increased activity, miRNAs could rapidly destabilize mRNAs andsuppress translation of proteins required for synaptic functionand development, thereby reducing connectivity homeostatically.There is evidence that translation of mRNAs regulating synapticstrength can be regulated locally at the synapse (11).Hundreds of miRNAs have been characterized, many of which

are enriched in the brain and synapses (12–14), but identifyingspecific miRNAs contributing to activity-dependent developmentand plasticity has proven difficult for two reasons (15–18): miR-NAs regulating synaptic plasticity are assumed to be relativelyrare in comparison to miRNAs regulating more general cellularprocesses, and plasticity is localized to individual synapses. Thus,potential miRNAs that may regulate mRNA transcript stability inactivity-dependent synaptic plasticity may not emerge from globalmiRNA expression analysis.To overcome this problem, we used a bioinformatics approach.

Microarray analysis showed that the abundance of several mRNAtranscripts was rapidly decreased in hippocampal neurons in cellculture after increasing activity. Using motif-based sequenceanalysis (19), we then determined whether these transcripts sharedmotifs in the 3′UTRs corresponding to putative binding sites formiRNAs or RNA binding proteins. This approach revealed thatmany of these rapidly down-regulated transcripts shared sequen-ces in their 3′UTR for several miRNAs that could, in theory,increase degradation of mRNAs that promote synaptic develop-ment. One of these motifs was identified as a putative binding sitefor a rare brain-enriched miRNA (miR-485) for which little isknown. Our analysis indicated that several of the genes with pu-tative miR-485 binding sites control synaptic function or de-velopment. The following studies were undertaken to determinewhether miR-485 regulates homeostatic synaptic plasticity inhippocampal neurons through destabilization of specific genetranscripts regulating synaptic development and function. Theresults support the hypothesis, and the findings provide a newmolecular mechanism regulating homeostatic synaptic plasticity inhippocampal neurons, which is likely to be important in regulatingsynaptic connectivity during neural development and disease.

ResultsSynaptic activity in hippocampal cultures was increased with 50μM BiC and 500 μM 4-aminopyridine (4-AP) (20); Ca2+ imagingconfirmed the increase in repetitive firing in these hippocampalcultures (Fig. S1A). Chronic elevation of synaptic activity withBiC/4-AP for 5 d reduced dendritic spine density by 22 ± 3% (n=36; P < 0.001) (Fig. S1B), consistent with previously describedactivity-dependent structural changes (21, 22).To identify mRNAs that are rapidly reduced by posttran-

scriptional mechanisms, hippocampal neurons in culture weretreated with BiC/4-AP for 5 min in the presence of the tran-

Author contributions: J.E.C., P.R.L., and R.D.F. designed research; J.E.C., P.R.L., S.C., andW.L. performed research; J.E.C. contributed new reagents/analytic tools; J.E.C., P.R.L., S.C.,W.L., and R.D.F. analyzed data; and J.E.C., P.R.L., and R.D.F. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: fieldsd@mail.nih.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1017576108/-/DCSupplemental.

11650–11655 | PNAS | July 12, 2011 | vol. 108 | no. 28 www.pnas.org/cgi/doi/10.1073/pnas.1017576108

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 5,

202

1

scriptional blocker actinomycin D (25 μM). Microarray profilingidentified many rapidly regulated mRNAs under these condi-tions, and many were synapse- and morphogenetic-associated. Abioinformatics approach was used to examine the 3′UTR of thisset of destabilized mRNAs to identify enriched motifs comparedwith transcripts that were up-regulated or unregulated by in-creased synaptic activity. If these transcripts are regulated bymiRNAs, they would be expected to share miRNA binding se-quences that would suppress translation or increase degradationof transcripts. Using motif prediction and analysis tools, MultipleEm for Motif Elicitation (MEME) (19), many of these rapidlydestabilized mRNAs were found to share a motif in the 3′UTRcorresponding to the binding sequence for miR-485 (Fig. S1C).We then tested whether this miRNA could participate in thehomeostatic activity-dependent reduction in synaptic connectivitycaused by increased activity.

miR-485 Is Regulated in a Developmental and Activity-DependentManner in the Hippocampus.miR-485 was detected in hippocampalneuronal cell bodies and dendrites in culture by FISH (Fig. 1A).These findings were confirmed by localization of miR-485 inneuronal processes by RT-PCR analysis of processes extendingthrough 3-μm pores in transwell cultures (Fig. 1B). Expressionlevels for miR-485 were low in developing rat hippocampuscompared with other brain-enriched miRNAs at embryonic day18.5 (23, 24) but increased 12.2± 2.5-fold (n=3) by postnatal day14 (Fig. 1C); in culture, both miR-485 and the precursor miRNAwere regulated in an activity-dependent manner, increasing 1.3 ±0.1-fold (n = 3; P < 0.05) after a 1-h treatment with BiC/4-AP.

miR-485 Regulates the Presynaptic Protein SV2A. Gene ontologyanalysis of miR-485–predicted targets (Table S1) showed en-richment in genes functioning in CNS development and disease(e.g., Huntington disease). One of these predicted targets, the

synaptic vesicle protein SV2A (Fig. S2), is enriched in presynapticterminals, where it controls neurotransmitter release (25).Function of predicted miRNA responsive elements (MREs) wastested by luciferase assays on constructs containing either the full-length 3′UTR or the 3′-end of SV2A 3′UTR. Overexpression ofa miR-485 mimic (miR-M; a small chemically modified dsRNAdesigned to mimic the endogenous miRNA) significantly reducedluciferase activity of both constructs (n = 6; P < 0.001) (Fig. 2A),and this was reversed when both MREs were mutated within the5′-seed region (Fig. S2B). The effects of miR-485 were specific,because cotransfection with a miR-485 inhibitor (miR-I; an an-tisense RNA oligonucleotide decoy that specifically and com-petitively binds to and inhibits the function of endogenous miR-485) reduced the suppression of a miR-485 luciferase construct bythe miR-M (n = 3; P < 0.01). Furthermore, transcript levels forSV2A 3′UTR constructs were reduced by miR-485 (Fig. S2C),consistent with a degradative mechanism for the miRNA.We tested if miR-485 controlled SV2A expression in hippo-

campal neurons by miR-485 overexpression and inhibition. miR-485 overexpression decreased both transcript abundance (n = 3;P < 0.05) (Fig. 2B) and protein expression (Fig. 2C) of SV2Arelative to both an untreated control and negative controls for themiR-M and miR-I containing random sequences validated to notaffect miR function. Furthermore, the effect on SV2A was spe-

A

U6

snR

NA

miR

-485

scr-

miR

LNA ProbeMAP2LNA Probe

C

0 5 10 15 2002468

10121416

Time (postnatal day)

miR

-485

abu

ndan

ce(n

orm

. to

E18

.5)

- 10

- 100- 200

P = Process

U6 g-acti

nb-a

ctin

miR-48

5

P S

S = Soma + Process

P S P S P S

B

Fig. 1. miR-485 is expressed in hippocampal neurons and developmentallyregulated. (A and B) Expression of miR-485 was detected by FISH and RT-PCR.U6 snRNA and scrambled probe are shown as positive and negative controls,respectively. scr-miR, scrambled miR probe. Dendrites (red) were identifiedby MAP2 immunoreactivity, and nuclei (blue) were stained with Hoechst33342. (Scale bar = 25 μm.) (B) RT-PCR was performed on neuronal processesgrowing through 3-μm pores in transwell inserts to isolate from cell soma inthe upper compartment. P, process; S, soma + process. Mature miR-485(based on product size) was present in both fractions. U6 (U6 snRNA) andγ-actin served as reference genes. β-actin was approximately twofold higherin processes, indicating enrichment of neuronal processes. (C) miR-485transcript is developmentally regulated in the hippocampus. The increase inmature miR-485 (n = 3) at each time point was first normalized to the ref-erence gene U6 snRNA and the abundance expressed relative to embryonicday 18.5. E, embryonic day.

B

SY

P38

PS

D-9

5

RN

A a

bund

ance

(nor

m to

con

trol)

-0 .4

-0.2

0.0

0.2

0.4miR-MmiR-I

*

SV

2A

A

0

0.2

0.4

0.6

0.8

1.0

1.2

Control

miR

-485

SV2A

SV2A MUT

SV2A 3'U

TR

******

***

n.s.

Tran

slat

iona

l sup

pres

sion

(rel

ativ

e lu

cife

rase

act

ivity

)C

E

0.0

0.2

0.4

0.6

0.8

***S

YP

38+ -

punc

ta e

xpre

ssin

gS

V2A

(col

ocal

izat

ion

inde

x)

cont

rol

miR

-Mm

iR-I

D

contr

ol

miR-M

- co.

miR-M

miR-I

- co.

miR-I

SV2A

GAPDHSYP38

GAPDHPSD-95

control miR-M miR-I

SV

2A S

YP

38S

V2A

SY

P38

Fig. 2. miR-485 decreases SV2A expression in hippocampal neurons. (A)SV2A is a putative target of miR-485. Luciferase assays to validate SV2A asa target of miR-485 were performed on constructs containing either the full-length 3′UTR (SV2A-3′UTR), distal 3′UTR containing two miR-485 bindingsites (SV2A), or mutated sites in the 5′-seed of MRE1 and MRE2 (SV2A MUT)as controls (Fig. S2B). Translational suppression by miR-485 was normalizedto empty psiCHECK-2 plasmid. Treatment with the miR-M decreased SV2Atranscript abundance (B) and protein levels (C) significantly without affect-ing SYP38 or PSD-95. Treatment with negative controls for either the mimicor inhibitor did not alter transcript or protein levels for SV2A, SYP38, or PSD-95. -co. control. (D) miR-485 overexpression reduced SV2A expression in hip-pocampal neuron synapses. Cultures were cotransfected with DsRed andmiRNAs at 7 days in vitro (DIV) and analyzed by immunocytochemistry at 12DIV for SV2A and SYP38 expression. Representative examples of untreated(control) and neurons transfected with the miR-M or miR-I showing that SV2Aexpression is reduced in individual presynaptic terminals. (Lower) Highermagnification is shown. Treatment with negative controls did not alter SV2Aor SYP38 expression. (Scale bars: Upper, 10 μm; Lower, 2.5 μm.) (E) Quanti-tative analysis of randomly chosen dendritic fields showing that the miRNAspecifically reduced colocalization of SV2A and SYP38.

Cohen et al. PNAS | July 12, 2011 | vol. 108 | no. 28 | 11651

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 5,

202

1

cific, because neither transcript nor protein expression of thesynaptic markers synaptophysin (SYP38) and postsynaptic density95 (PSD-95) was altered (Fig. 2 B and C). Neither of thesetranscripts contains conserved miR-485 binding sites within their3′UTRs, consistent with a specific effect of the miRNA on genescontaining the MRE.SV2A immunoreactivity at synapses (Fig. 2 D and E and Fig.

S3) was specifically reduced by the miR-M. To identify pre-synaptic terminals, cultures were double-labeled for the pre-synaptic markers SV2A (green) and SYP38 (red). Colocalizationanalysis of SV2A+ and SYP38+ puncta showed a reduction ofSV2A in cultures overexpressing miR-485 (n = 21; F4,102 =20.57, P < 0.001 by one-way ANOVA). Colocalization indices forneurons transfected with either control (n = 20, 0.73 ± 0.01) andnegative controls for the mimic (n = 25, 0.71 ± 0.01) or inhibitor(n= 20, 0.73 ± 0.01) were not affected. Treatment with the miR-M or miR-I did not affect the number of presynaptic terminals(SYP38+ puncta).The finding that miR-485 regulated the expression of SV2A, a

ubiquitous presynaptic protein that regulates neurotransmitterrelease (26), suggested that synapses may be altered structurallyas a consequence of changes in synaptic function. We thereforetested the function of miR-485 on several morphological prop-erties of hippocampal synapses.

miR-485 Regulates Spine Density and Synapse Morphology. Hippo-campal dendritic spine density increases over several weeks inculture and in vivo, and immature spine protrusions graduallybecome stubby with a distinctive neck and head structure as theymature (27, 28). miR-485 overexpression reduced spine densityby 14 ± 5% (n = 19; P < 0.001) and increased the appearance oflong and thin immature spines by 35 ± 9% (n = 6; P < 0.01) (Fig.3 A and E). Conversely, the miR-I increased spine density by 15 ±4% (n = 19; P < 0.005) and increased the number of short andstubby spines by 46 ± 19% (n = 6; P < 0.05) (Fig. 3E, comparearrowheads in untreated samples vs. miR-M– and miR-I–treatedsamples). Effects of miR-485 on spine density were highly sig-nificant by one-way ANOVA (F4,90 = 9.46, P < 0.001). Themagnitude of these changes is consistent with structural changesassociated with modulating synaptic activity in culture (22, 29).Furthermore, the miR-I fully reversed the effects of the miR-M,restoring spine density to control levels (Fig. S4). These resultsshow that miR-485 suppresses spine formation and maturation.The effects of miR-485 are consistent with the homeostaticdecrease in synaptic connectivity seen after elevating impulseactivity and are consistent with the parallel activity-dependentincrease in miR-485 levels.The effects of miR-485 on presynaptic SV2A expression, den-

dritic spine number, and maturation may have postsynapticeffects because of the loss of mature spines containing post-synaptic density components, such as PSD-95 and AMPA recep-tors. Consistent with this, miR-485 overexpression reduced PSD-95 clustering and colocalization with SYP38 (Fig. 3D). Blockingendogenous miR-485 increased the number of PSD-95 puncta.Effects of miR-485 on PSD-95 clustering were further examinedwith a GFP-tagged PSD-95 construct to visualize PSD-95 punctadirectly (30). PSD-95–GFP+ puncta density was reduced 34 ± 3%(n = 35; P < 0.001) (Fig. 3 B and E) by miR-485 overexpression.Conversely, inhibition of endogenous miR-485 increased thedensity of PSD-95–GFP+ puncta by 24 ± 4% (n= 35; P < 0.001).Effects of the miR-M and miR-I were highly significant (F4,170 =48.73, P < 0.001 by one-way ANOVA). Moreover, effects of themiR-M on PSD-95–GFP+ puncta density were specific becausethey were reversed by the miR-I (Fig. S4).AMPA receptor trafficking into the postsynaptic membrane is

regulated by synaptic activity to control synaptic strength in ahomeostatic manner (7). The role of miR-485 in AMPA receptortrafficking was investigated in hippocampal neurons transfected

with pCl–super-ecliptic pH-sensitive (SEP)–GluR2(R) to visual-ize cell-surface GluR2 receptor clustering (31). miR-485 over-expression significantly reduced SEP-GluR2 density by 17 ± 5%(n = 16; P < 0.01) (Fig. 3 C and E). Conversely, inhibition ofendogenousmiR-485 significantly increasedSEP-GluR2density by19 ± 4% (n= 16; P < 0.05). The effects of functional miR-485 onSEP-GluR2were highly significant (F4,76 = 8.55, P < 0.001 by one-way ANOVA). The morphological changes in synapse numberwere observable 3 d following miR-485 overexpression. Thesefindings show that miR-485 acts to regulate spine density andmaturation, PSD-95 clustering, and surface GluR2 expressionnegatively in the postsynaptic membrane (Fig. 3).

miR-485 Overexpression Alters Synapse Function. The homeostaticchanges in spine density andGluR2 receptor expression regulatedby miR-485 were accompanied by corresponding changes insynaptic function. Treatment with the miRNA inhibitor for 6 dincreased spontaneous miniature excitatory synaptic current(mEPSC) frequency relative to controls, whereas mEPSC am-plitude was not changed [9.1 pA for the negative control com-pared with 10 pA for the miR-I; χ2(3,313) = 191.984, P < 0.001](Fig. 4 A and B).SV2A functions to facilitate synaptic vesicle fusion and neu-

rotransmitter release through interaction with synaptotagmin (32,33). Therefore, miR-485 should decrease the number of synapticvesicles released following a stimulus. This was tested by imagingFM 1–43 uptake and release following depolarization. Consistentwith this hypothesis, treatment with either the miRNA mimic(30.8 ± 3.0 s, n= 9) or inhibitor (33.2 ± 1.8 s, n= 10) did not alterdestaining rates significantly from control (Fig. 4C). However,

D SYP38PSD-95 MERGE

Spi

ne d

ensi

ty(%

cha

nge

from

con

trol)

***

***

-20

-10

0

10

20

cont

rol

miR

-Mm

iR-I

***

0

-40

-20

20

40

cont

rol

miR

-Mm

iR-I

***

Pos

tsyn

aptic

den

sitie

s(%

cha

nge

from

con

trol)

-30

-20

-10

0

10

20

30

Sur

face

Glu

R2

dens

ity(%

cha

nge

from

con

trol)

cont

rol

miR

-Im

iR-M

***

***

A B C

miR-M

miR-I

control

SEP-GFP-GluR2

14

9

11▲ 24

▲

20

▲28

DsRed PSD-95-GFP

9

17

13

control

miR-MmiR-I

E

Fig. 3. miR-485 negatively regulates spine density, PSD-95 clustering, andsurface GluR2 expression. miR-485 reduces spine density (A), PSD-95 clus-tering (B), and surface GluR2 expression (C) along dendrites. Spine density,PSD-95, and SEP-GluR2 were 12.7 ± 0.4 spines per 20 μm (n = 18), 9.4 ± 0.3PSD-95–GFP+ puncta per 20 μm (n = 34), and 6.9 ± 0.3 SEP-GluR2 puncta per20 μm (n = 16) in untreated controls, respectively. These values were notaffected by transfection with negative controls for the mimic [12.4 ± 0.3spines per 20 μm (n = 20), 9.1 ± 0.2 PSD-95–GFP+ puncta per 20 μm (n = 36),and 7.3 ± 0.3 SEP-GluR2 puncta per 20 μm (n = 16)] or inhibitor [12.6 ± 0.4spines per 20 μm (n = 19), 9.3 ± 0.2 PSD-95–GFP+ puncta per 20 μm (n = 35),and 7.3 ± 0.2 SEP-GluR2 puncta per 20 μm (n = 17)]. (D) Endogenous PSD-95clustering (green) in hippocampal cultures was decreased by transfectionwith the miR-M and increased by the miR-I. (Scale bar = 10 μm.) (E) Repre-sentative examples of neurons transfected with DsRed, PSD-95 (PSD-95–GFP),or surface GluR2 (SEP-GluR2). (Scale bar = 2.5 μm.)

11652 | www.pnas.org/cgi/doi/10.1073/pnas.1017576108 Cohen et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 5,

202

1

total dye loss was decreased by treatment with the miRNA mimicand increased by the inhibitor (0.28 ± 0.01 for the miR-I and0.36 ± 0.02 for the miR-M; P < 0.05) (Fig. 4A). The finding thatmEPSC frequency was decreased by the miRNA mimic withouteffects on mEPSC amplitude or kinetics of FM 1-43 destaining(presynaptic release) supports the morphological findings thatmiR-485 decreases synapse number.

SV2A Knockdown Mimics the Effects of miR-485 Overexpression.These results suggest that miR-485 regulation of the presynapticprotein SV2A contributes to the activity-dependent decrease insynaptic connectivity induced by increased neuronal activity.Consistent with this, siRNA-mediated knockdown of SV2A inhippocampal cultures reduced expression of SV2A protein with-out affecting SYP38 or PSD-95 expression levels (Fig. S5 A andB), and the reduction in SV2A by miR-485 was rescued to levelsseen for untreated cultures by SV2A overexpression (pSV2A).Effects of SV2A knockdown and overexpression on dendritic

spines (Fig. 5A), PSD-95 (Fig. 5B), andGluR2 (Fig. 5C) were thentested. Spine density was reduced by 29 ± 4% (n= 20; P < 0.001)(Fig. 5 A and D). A second siRNA targeting a distinct site withinthe SV2A transcript also reduced spine density by 28± 3% (SV2AsiRNA-2: n=19; P< 0.001). SV2A knockdown reduced PSD-95+

GFP puncta by 25 ± 2% (SV2A siRNA-1: n= 30; P < 0.001) (Fig.5B) and 28± 2% (SV2A siRNA-2: n=34; P< 0.001). SEP-GluR2density was similarly reduced by 22± 6% (SV2A siRNA-1: n=16;P < 0.001) and 17 ± 3% (SV2A siRNA-2: n= 16; P < 0.001) (Fig.5C). SV2A overexpression (pSV2A) increased PSD-95+ GFPpuncta density by 8 ± 3% (n = 16; P < 0.01) and SEP-GluR2 by17 ± 3% (n = 17; P < 0.05). Moreover, SV2A overexpressionreversed the effects of miR-485 on spine density, PSD-95, andGluR2 (compare the miR-M with miR-M + pSV2A; Fig. 5D).Effects on spines, PSD-95–GFP+ puncta, and SEP-GluR2 were

highly significant by one-way ANOVA (F7,178 = 21.96, P < 0.001for spines; F7,198 = 33.86, P< 0.001 for PSD-95; and F7,120 = 8.98,P < 0.001 for SEP-GluR2). These results demonstrate a newrole for SV2A in controlling dendritic spines and postsynapticspecialization.

miR-485 Controls Homeostatic Plasticity of Dendritic Spines andPostsynaptic Specialization. The effects of miR-485 on synapse-associated transcripts, including SV2A (Table S1); spine density;and postsynaptic specialization are consistent with a possible roleof this miRNA in homeostatic synaptic plasticity acting to reducetranscripts rapidly and/or suppress translation to decrease syn-aptic connectivity under conditions of persistent hyperactivity.Consistent with this, chronic BiC/4-AP treatment for 5 d reducedSV2A transcript levels by 20 ± 6% (n = 4; P < 0.05). This hy-pothesis was tested by increasing synaptic activity with BiC/4-APfor 5 d and blocking endogenous miR-485 to determine whetherthe homeostatic changes in spines, PSD-95 clustering, and surfaceGluR2 expression observed following chronic activity are im-paired when miR-485 function is blocked. The results supportthis conclusion. Spine density (−20 ± 3%, n= 32; P < 0.001) (Fig.6A), PSD-95–GFP+ puncta (−20 ± 3%, n = 35; P < 0.001)(Fig. 6B), and SEP-GluR2 puncta (−17 ± 5%, n= 16; P < 0.001)(Fig. 6C) were significantly reduced after increasing synaptic ac-tivity with BiC/4-AP for 5 d. Consistent with the hypothesis,inhibiting endogenous miR-485 blocked the homeostatic re-duction in spine density (Fig. 6A), PSD-95–GFP+ puncta density(Fig. 6B), and SEP-GluR2 expression (Fig. 6C) induced by BiC/4-AP. Effects of BiC/4-AP and miR-485 inhibition on dendriticspine (n= 32), PSD-95 clustering (n= 35), and SEP-GluR2 (n=16) were highly significant by one-way ANOVA (F4,166 = 31.19,P < 0.001 for spines; F4,157 = 24.66, P < 0.001 for PSD-95–GFP+;

10 pA200 msec

b

d

c

a

A B

abcd

mE

PS

C fr

eque

ncy

(eve

nts

per

3 m

in)

0

30

60

90

120

150

180

a b dc

0 50 100 150 200 250 3000.2

0.4

0.6

0.8

1.0

60 mM K+-saline

Nor

mal

ized

fluo

resc

ence

C

Time (sec)

p < 0.05

Fig. 4. miR-485 reduces spontaneous synaptic activity in hippocampalneurons. (A) Representative traces of spontaneous mEPSCs recorded fromcultured hippocampal neurons transfected with negative control for themiR-M (a), miR-M (b), negative control for miR-I (c), and miR-I (d) showinga reduction in spontaneous mEPSC frequency after transfection with themimic (b) and an increase following transfection with the inhibitor (d). (B)Quantitative analysis showing the transfection significantly changed mEPSPfrequency [χ2(3,313) = 191.984, P < 0.001]. (C) Synaptic vesicle release (FM1-43 destaining kinetics) was not affected by miR-485, but the number ofrecycled vesicles (amount of destaining) was reduced by miR-485 (yellow)and increased by the inhibitor (red) (P < 0.05). Destaining curves for negativecontrols (n = 8 coverslips with at least 60 boutons each) and for culturestreated with the miR-M (n = 8) and miR-I (n = 11) are shown.

D

-30

-20

-10

0

10

20

30

miR

-M +

pSV

2A

cont

rol

siR

NA

miR

-MpS

V2A

Sur

face

Glu

R2

dens

ity(%

cha

nge

from

con

trol)

****** **

*

-40

-30

-20

-10

0

10

20

*** ***

miR

-M +

pSV

2A

cont

rol

siR

NA

miR

-MpS

V2A

***Pos

tsyn

aptic

den

sitie

s(%

cha

nge

from

con

trol) **

-40

-30

-20

-10

0

10

Spi

ne d

ensi

ty(%

cha

nge

from

con

trol)

miR

-M +

pSV

2A

cont

rol

siR

NA

miR

-MpS

V2A

***

n.s.

*** ***

A B C

cont

rol

siR

NA

pSV

2Am

iR-M

pSV

2A +

miR

-M

DsRed

▲

▲

▲

▲

▲

19

21

13

15

19

SEP-GFP-GluR2

9

14

8

11

11PSD-95-GFP

15

11

17

12

14

Fig. 5. SV2A siRNA-mediated knockdown mimics the effects of miR-485overexpression on reducing dendritic spine density, PSD-95 clustering, andsurface GluR2 expression. SV2A siRNA reduces spine density (A), PSD-95 clus-tering (B), and surface GluR2 expression (C) along dendrites. Spine density,PSD-95, and SEP-GluR2were 10.4± 0.3 spines per 20 μm(n=33), 10.3± 0.4 PSD-95–GFP+ punctaper 20 μm(n=20), and 6.9± 0.3 SEP-GluR2punctaper 20 μm(n= 16) in untreated controls, respectively. These values were not affected bytransfectionwithnegative controls for the siRNA [10.5±0.3 spinesper20μm(n=23), 10.3±0.3 PSD-95–GFP+punctaper 20μm(n=27), and7.3±0.3 SEP-GluR2puncta per 20 μm(n= 16)] ormiRNAmimic [10.5± 0.3 spines per 20 μm(n = 23),9.8±0.3PSD-95–GFP+punctaper20μm(n=20), and7.3±0.3SEP-GluR2punctaper 20 μm (n = 16)]. (D) Representative examples for A–C. (Scale bar = 2.5 μm.)

Cohen et al. PNAS | July 12, 2011 | vol. 108 | no. 28 | 11653

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 5,

202

1

and F4,75 = 9.99, P < 0.001 for SEP-GluR2) and Kruskal–Wallispost hoc tests. These findings indicate a role for miR-485 in ho-meostatic plasticity in hippocampal neurons acting throughmorphological changes in dendritic spines and postsynaptic spe-cialization to decrease synaptic connectivity and function.

DiscussionThe results identify an activity-regulated miRNA, miR-485, thatfunctions in a homeostatic manner to control synaptic develop-ment and plasticity. Individual miRNAs can target up to severalhundred genes (10, 34) to destabilize sets of mRNA transcriptsthat mediate complex cellular processes, such as development andplasticity, in a coordinated manner. This provides a rapid post-transcriptional mechanism regulating cellular function that can belocalized to specific subcellular domains, thus making this mech-anism well suited to function at synaptic terminals undergoingactivity-dependent homeostasis (9). miRNA-485 is expressed inthe hippocampus and cerebral cortex (23, 24), and the results ofthis study show that this miRNA potently suppresses dendriticspine development, structure, and function. We found that miR-485 did not alter release properties at individual presynaptic ter-minals but, instead, altered the number of functional synapses(Fig. 4). A gene ontology analysis shows that many predicted tar-gets of miR-485 are implicated in neural development and mor-phogenesis (Table S1), including the target, SV2A, which weconfirmed as a target regulated by both translational suppression(Fig. 2) and mRNA degradation (Fig. S2). Furthermore, SV2Aexpression affects spine maturation, PSD-95 clustering, and sur-faceGluR2 expression. Chronic synaptic activity induced by BiC/4-AP treatment reduced the density of spines and PSD-95 andGluR2 puncta, and this activity-dependent regulationwas reversedby the miR-I.

Activity-dependent control of miR-485 and its targets couldfunction as amechanism to rapidly down-regulate other transcriptsenriched at the pre- and postsynaptic terminals that mediatea decrease in synaptic connectivity to adapt to chronically elevatedexcitation. The experiments blocking these homeostatic changeswith the miR-I support this conclusion. Mature miR-485 levelswere elevated by increased excitatory activity in hippocampalneurons, and other studies show that the miRNA processing en-zyme Dicer is regulated in a Ca2+-dependent manner (35), sug-gesting that pre-miRNAs targeted to synapses may undergoprocessing into mature miRNAs in an activity-dependent manner.SV2A regulation by miR-485 and synaptic activity in the context

of homeostasis are important for several reasons. SV2A is highlyexpressed in all presynaptic terminals of the hippocampus, where itfacilitates neurotransmitter release through the readily releasablepool (25, 26, 32). Mice lacking SV2A develop seizures by postnatalday 7 and die (26), demonstrating the importance of SV2A inpostnatal brain development and synaptic function. Moreover,SV2A is down-regulated following seizures (36, 37), suggesting thatmiR-485 may participate in a homeostatic response to hyper-excitation during seizure through posttranscriptional regulation ofSV2A to reduce neurotransmitter release. More importantly, wefound that SV2A knockdown mimicked the morphological effectsof the miR-M and SV2A overexpression reversed the effects of themiR-M, supporting a downstream role of this protein in homeo-static synaptic plasticity.These results differ in some respects from previous studies.

Studies in microisland cultures (32) and hippocampal slices (26)did not find differences in minifrequency and synapse number inSV2A KO mice. Other targets of miR-485 (Table S1) may con-tribute to the decrease in minifrequency observed in response toincreasing miR-485 levels (Fig. 4).We have altered activity for longer than most other studies

investigating homeostatic mechanisms to study the morphologicalconsequences of increased synaptic activity on development ofsynaptic connections. The results show changes in postsynapticspecialization attributed to miR-485 acting through SV2A atpresynaptic terminals. Pre- and postsynaptic sites undergo ho-meostatic changes through both the size of the readily releasablepool of synaptic vesicles and scaling of synaptic strength (38–40). Furthermore, our findings are in agreement with transcrip-tional requirements for homeostatic plasticity occurring pre-synaptically as well as postsynaptic morphological changes insynaptic density (3, 39).Homeostatic synaptic plasticity is also important in adjusting

neural circuit connectivity and excitation in neurological diseaseand in response to brain trauma. A survey of public databases frommiRNA expression studies reveals that miR-485 is regulated ina number of neurological diseases and after brain trauma; miR-485 is down-regulated in Huntington disease (41), Alzheimer’sdisease (42), and traumatic brain injury (43) and is up-regulated instroke (44). These disease studies, together with our findings onthe effects of miR-485 on spines and synaptic development, sug-gest that miR-485may contribute to pathophysiology and adaptiveresponses to several neurodegenerative diseases disrupting normallevels of neural network activity. Down-regulating miR-485 func-tion increased spine density and accompanying postsynapticchanges, suggesting the feasibility of using a miR-I therapeuticallyto promote regeneration and synaptogenesis.In conclusion, these results identify miR-485 as an important

regulator of synaptic development in an activity-dependentmanner,and they extend the presently identified mechanisms controllinghomeostatic synaptic plasticity to include a new posttranscriptionalmechanism involving miRNAs. The ability of miRNAs to regulatelarge sets of mRNAs controlling complex cellular processes ina coordinatedmanner rapidly and locally within different regions ofa neuron makes the activity-dependent effects of miR-485 wellsuited to participate in homeostatic regulation of synaptic circuits

-30

-20

-10

0

10

20

30

40

******

miR

-I

miR

-I +

BiC

/4-A

P

cont

rol

BiC

/4-A

P

***

Spi

ne d

ensi

ty(%

cha

nge

from

con

trol)

DsRed

21

26

17

23

cont

rol

miR

-I +

BiC

/4A

PB

iC/4

AP

miR

-I

13

16

11

14

PSD-95GFP

0

-30

-20

-10

10

20

30

miR

-I +

BiC

/4-A

P

cont

rol

Pos

tsyn

aptic

den

sitie

s(%

cha

nge

from

con

trol)

miR

-I

***

BiC

/4-A

P

***

***

A B

D-30

-20

-10

0

10

20

30

miR

-I +

BiC

/4-A

P

cont

rol

miR

-IB

iC/4

-AP

Sur

face

Glu

R2

dens

ity(%

cha

nge

from

con

trol) ***

***

***

C

SEP-GFP-GluR2

11

14

8

12

Fig. 6. BiC/4-AP for 5 days in vitro (DIV) induces a homeostatic reduction insynaptic connectivity that is reversed by blocking endogenous miR-485.Hippocampal neurons transfected with a miR-I and constructs to visualizechanges in synaptic properties were treated with 50 μm BiC plus 500 μm 4-APfor 5 d. BiC/4-AP reduces spine density (A), PSD-95 clustering (B), and surfaceGluR2 expression (C) along dendrites that are reversed by the miR-I. Spinedensity, PSD-95, and SEP-GluR2 were 10.8 ± 0.4 spines per 20 μm (n = 35), 9.3± 0.3 PSD-95–GFP+ puncta per 20 μm (n = 29), and 6.9 ± 0.3 SEP-GluR2 punctaper 20 μm (n = 16) in untreated controls. These values were not affected bytransfection with negative controls for the miRNA inhibitor [11.2 ± 0.4 spinesper 20 μm (n = 30), 9.1 ± 0.2 PSD-95–GFP+ puncta per 20 μm (n = 32), and7.3 ± 0.2 SEP-GluR2 puncta per 20 μm (n = 17)]. (D) Representative examplesfor A–C. (Scale bar = 2.5 μm.)

11654 | www.pnas.org/cgi/doi/10.1073/pnas.1017576108 Cohen et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 5,

202

1

under many different conditions in pathology and in the normaldevelopment and function of neural circuits.

Materials and MethodsAnimals and Cell Culture. Timed-pregnant albino Sprague–Dawley rats wereused throughout this study. Dissociated hippocampal cultures were pre-pared as previously described (45). HEK293 cells were maintained in DMEMplus 10% (vol/vol) FBS at 5% (vol/vol) CO2. All procedures conformed to Na-tional Institutes of Health animal welfare guidelines and approved animalstudy protocols. Details are provided in SI Materials and Methods.

Drug Treatments and Reagents. Cultures were treated by replacingmediumwithfresh medium containing drugs at the following final concentrations: 50 μM BiCmethiodide, 1 μM TTX Na+ citrate (Sigma), 500 μM 4-AP, 50 μM D-(−)-2-amino-5-phosphonopentanoic acid (D-APV), 40 μM 6-cyano-7-nitroquinoxaline-2,3-dione(CNQX), 50 μM picrotoxin (Tocris Biosciences), and 25 μM actinomycin D (Invi-trogen). Short-term treatments of cultures were in sterile-filtered saline con-taining 145mMNaCL, 4.5 mMKCl, 0.8 mMMgSO4, 1.8 mM CaCl2, 10 mMHepes,and 10 mM glucose, adjusted to 320 mOsm with sucrose at pH 7.3.

Transfections.Neurons were cotransfected at 7 days in vitro (DIV) with 2 μg ofDsRed-C1 monomer (Clontech); PSD-95–GFP (30) or pCl-SEP-GluR2(R) (31);and 25 pmol of the miR-M, miR-I, miR-M negative control, miR-I negativecontrol, or distilled H2O (Ambion) with Lipofectamine 2000 (Life Technolo-gies). siRNA-mediated knockdown of SV2A was performed using 25 pmol ofSilencer Select siRNA targeting SV2A (s138846 and s138848) (Invitrogen) andsiRNA negative controls (Invitrogen). Transfected cultures were analyzed byelectrophysiological recordings, immunocytochemistry, live imaging, lucif-erase assays, RT-PCR, and Western blotting as described in SI Materialsand Methods.

Data and Statistical Analysis. All values are reported as the mean ± SE of themean of at least three biological replicates per experiment. In all figures,statistical significance is presented as *P < 0.05, **P < 0.01, and/or ***P <0.001. Details are provided in SI Materials and Methods.

ACKNOWLEDGMENTS. We are grateful to Drs. David Bredt and RobertJ. Wenthold for PSD-95–GFP and Robert Malinow for pCI–SEP-GluR2(R) usedin this study. This work was supported by the National Institute of ChildHealth and Human Development and the National Eye Institute IntramuralResearch Program.

1. Turrigiano GG, Leslie KR, Desai NS, Rutherford LC, Nelson SB (1998) Activity-dependent scaling of quantal amplitude in neocortical neurons. Nature 391:892–896.

2. Hebb DO (1949) The Organization of Behavior: A Neuropsychological Theory (Wiley,New York).

3. Thiagarajan TC, Lindskog M, Tsien RW (2005) Adaptation to synaptic inactivity inhippocampal neurons. Neuron 47:725–737.

4. Shepherd JD, et al. (2006) Arc/Arg3.1 mediates homeostatic synaptic scaling of AMPAreceptors. Neuron 52:475–484.

5. Thiagarajan TC, Piedras-Renteria ES, Tsien RW (2002) alpha- and betaCaMKII. Inverseregulation by neuronal activity and opposing effects on synaptic strength. Neuron 36:1103–1114.

6. Stellwagen D, Malenka RC (2006) Synaptic scaling mediated by glial TNF-alpha.Nature 440:1054–1059.

7. Hou Q, Zhang D, Jarzylo L, Huganir RL, Man HY (2008) Homeostatic regulation ofAMPA receptor expression at single hippocampal synapses. Proc Natl Acad Sci USA105:775–780.

8. Seeburg DP, Feliu-Mojer M, Gaiottino J, Pak DT, Sheng M (2008) Critical role of CDK5and Polo-like kinase 2 in homeostatic synaptic plasticity during elevated activity.Neuron 58:571–583.

9. Kosik KS (2006) The neuronal microRNA system. Nat Rev Neurosci 7:911–920.10. Guo H, Ingolia NT, Weissman JS, Bartel DP (2010) Mammalian microRNAs predominantly

act to decrease target mRNA levels. Nature 466:835–840.11. Sutton MA, et al. (2006) Miniature neurotransmission stabilizes synaptic function via

tonic suppression of local dendritic protein synthesis. Cell 125:785–799.12. Kim J, et al. (2004) Identification of manymicroRNAs that copurify with polyribosomes

in mammalian neurons. Proc Natl Acad Sci USA 101:360–365.13. Kye MJ, et al. (2007) Somatodendritic microRNAs identified by laser capture and

multiplex RT-PCR. RNA 13:1224–1234.14. Lugli G, Torvik VI, Larson J, Smalheiser NR (2008) Expression of microRNAs and their

precursors in synaptic fractions of adult mouse forebrain. J Neurochem 106:650–661.15. Siegel G, et al. (2009) A functional screen implicates microRNA-138-dependent

regulation of the depalmitoylation enzyme APT1 in dendritic spine morphogenesis.Nat Cell Biol 11:705–716.

16. Fiore R, et al. (2009) Mef2-mediated transcription of the miR379-410 cluster regulatesactivity-dependent dendritogenesis by fine-tuning Pumilio2 protein levels. EMBO J28:697–710.

17. Schratt GM, et al. (2006) A brain-specific microRNA regulates dendritic spinedevelopment. Nature 439:283–289.

18. Wayman GA, et al. (2008) An activity-regulated microRNA controls dendritic plasticityby down-regulating p250GAP. Proc Natl Acad Sci USA 105:9093–9098.

19. Bailey TL, Elkan C (1994) Fitting a mixture model by expectation maximization todiscover motifs in biopolymers. Proc Int Conf Intell Syst Mol Biol 2:28–36.

20. Hardingham GE, Fukunaga Y, Bading H (2002) Extrasynaptic NMDARs opposesynaptic NMDARs by triggering CREB shut-off and cell death pathways. Nat Neurosci5:405–414.

21. Piccoli G, et al. (2007) Proteomic analysis of activity-dependent synaptic plasticity inhippocampal neurons. J Proteome Res 6:3203–3215.

22. Verpelli C, et al. (2010) Synaptic activity controls dendritic spine morphology bymodulating eEF2-dependent BDNF synthesis. J Neurosci 30:5830–5842.

23. Liang Y, Ridzon D, Wong L, Chen C (2007) Characterization of microRNA expressionprofiles in normal human tissues. BMC Genomics 8:166.

24. Pena JT, et al. (2009) miRNA in situ hybridization in formaldehyde and EDC-fixedtissues. Nat Methods 6:139–141.

25. Bajjalieh SM, Frantz GD, Weimann JM, McConnell SK, Scheller RH (1994) Differentialexpression of synaptic vesicle protein 2 (SV2) isoforms. J Neurosci 14:5223–5235.

26. Crowder KM, et al. (1999) Abnormal neurotransmission in mice lacking synapticvesicle protein 2A (SV2A). Proc Natl Acad Sci USA 96:15268–15273.

27. Harris KM, Jensen FE, Tsao B (1992) Three-dimensional structure of dendritic spines andsynapses in rathippocampus (CA1)atpostnatalday15andadultages: Implications for thematuration of synaptic physiology and long-termpotentiation. J Neurosci 12:2685–2705.

28. Papa M, Bundman MC, Greenberger V, Segal M (1995) Morphological analysis ofdendritic spine development in primary cultures of hippocampal neurons. J Neurosci15:1–11.

29. Segal M, Kreher U, Greenberger V, Braun K (2003) Is fragile X mental retardationprotein involved in activity-induced plasticity of dendritic spines? Brain Res 972:9–15.

30. El-Husseini AE, Schnell E, Chetkovich DM, Nicoll RA, Bredt DS (2000) PSD-95involvement in maturation of excitatory synapses. Science 290:1364–1368.

31. Kopec CD, Li B, Wei W, Boehm J, Malinow R (2006) Glutamate receptor exocytosis andspine enlargement during chemically induced long-term potentiation. J Neurosci 26:2000–2009.

32. Custer KL, Austin NS, Sullivan JM, Bajjalieh SM (2006) Synaptic vesicle protein 2enhances release probability at quiescent synapses. J Neurosci 26:1303–1313.

33. Yao J, Nowack A, Kensel-Hammes P, Gardner RG, Bajjalieh SM (2010) Cotrafficking ofSV2 and synaptotagmin at the synapse. J Neurosci 30:5569–5578.

34. Baek D, et al. (2008) The impact of microRNAs on protein output. Nature 455:64–71.35. Lugli G, Larson J, Martone ME, Jones Y, Smalheiser NR (2005) Dicer and eIF2c are

enriched at postsynaptic densities in adult mouse brain and are modified by neuronalactivity in a calpain-dependent manner. J Neurochem 94:896–905.

36. Gorter JA, et al. (2006) Potential new antiepileptogenic targets indicated by microarrayanalysis in a rat model for temporal lobe epilepsy. J Neurosci 26:11083–11110.

37. van Vliet EA, Aronica E, Redeker S, Boer K, Gorter JA (2009) Decreased expression ofsynaptic vesicle protein 2A, the binding site for levetiracetam, during epileptogenesisand chronic epilepsy. Epilepsia 50:422–433.

38. Murthy VN, Schikorski T, Stevens CF, Zhu Y (2001) Inactivity produces increases inneurotransmitter release and synapse size. Neuron 32:673–682.

39. Han EB, Stevens CF (2009) Development regulates a switch between post- andpresynaptic strengthening in response to activity deprivation. Proc Natl Acad Sci USA106:10817–10822.

40. Wierenga CJ, Walsh MF, Turrigiano GG (2006) Temporal regulation of the expressionlocus of homeostatic plasticity. J Neurophysiol 96:2127–2133.

41. Packer AN, Xing Y, Harper SQ, Jones L, Davidson BL (2008) The bifunctional microRNAmiR-9/miR-9* regulates REST and CoREST and is downregulated in Huntington’s disease.J Neurosci 28:14341–14346.

42. Cogswell JP, et al. (2008) Identification of miRNA changes in Alzheimer’s disease brainand CSF yields putative biomarkers and insights into disease pathways. J AlzheimersDis 14:27–41.

43. Redell JB, Liu Y, Dash PK (2009) Traumatic brain injury alters expression ofhippocampal microRNAs: potential regulators of multiple pathophysiologicalprocesses. J Neurosci Res 87:1435–1448.

44. Jeyaseelan K, Lim KY, Armugam A (2008) MicroRNA expression in the blood and brainof rats subjected to transient focal ischemia by middle cerebral artery occlusion.Stroke 39:959–966.

45. Cohen JE, Fields RD (2008) Activity-dependent neuron-glial signaling by ATP andleukemia-inhibitory factor promotes hippocampal glial cell development. NeuronGlia Biol 4:43–55.

Cohen et al. PNAS | July 12, 2011 | vol. 108 | no. 28 | 11655

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 5,

202

1