Neuron

Article

Nuclear Calcium-VEGFD Signaling ControlsMaintenance of Dendrite ArborizationNecessary for Memory FormationDaniela Mauceri,1 H. Eckehard Freitag,1 Ana M.M. Oliveira,1 C. Peter Bengtson,1 and Hilmar Bading1,*1Department of Neurobiology, Interdisciplinary Centre for Neurosciences (IZN), University of Heidelberg, INF 364, 69120 Heidelberg,

Germany

*Correspondence: hilmar.bading@uni-hd.deDOI 10.1016/j.neuron.2011.04.022

SUMMARY

The role of neuronal dendrites is to receive and pro-cess synaptic inputs. The geometry of the dendriticarbor can undergo neuronal activity-dependentchanges that may impact the cognitive abilities ofthe organism. Here we show that vascular endothe-lial growth factor D (VEGFD), commonly known asan angiogenic mitogen, controls the total lengthand complexity of dendrites both in cultured hippo-campal neurons and in the adult mouse hippocam-pus. VEGFD expression is dependent upon basalneuronal activity and requires nuclear calcium-calmodulin-dependent protein kinase IV (CaMKIV)signaling. Suppression of VEGFD expression in themouse hippocampus by RNA interference causesmemory impairments. Thus, nuclear calcium-VEGFDsignaling mediates the effect of neuronal activity onthe maintenance of dendritic arbors in the adulthippocampus and is required for cognitive func-tioning. These results suggest that caution be em-ployed in the clinical use of blockers of VEGFDsignaling for antiangiogenic cancer therapy.

INTRODUCTION

Many brain functions, including memory formation and acquired

neuroprotection, are controlled by transient increases in the

intracellular calcium concentration induced by synaptic activity

(Hardingham and Bading, 2010; Silva et al., 1998). Calcium can

act locally near the site of entry to switch on signaling mecha-

nisms that modulate several biochemical processes that in turn

lead to changes in neuronal excitability and/or the efficacy of

synaptic transmission (Bliss et al., 2007). The long-term mainte-

nance of such activity-induced, functional adaptations requires

that calcium transients invade the cell nucleus and activate or

repress gene expression (Hardingham and Bading, 2010; Greer

and Greenberg, 2008). Nuclear calcium is one of the most potent

signals in neuronal gene expression and represents a key player

in the dialog between synapse and nucleus (Zhang et al., 2009). It

controls cAMP response element binding (CREB)- and CREB-

binding protein (CBP)-mediated transcription (Hardingham

et al., 1997, 1999, 2001; Chawla et al., 1998; Hu et al., 1999)

and is critical for the acquisition of memories and the build-up

of neuroprotective activity in synaptically activated neurons

(Limback-Stokin et al., 2004; Papadia et al., 2005; Zhang et al.,

2007, 2009). A picture of how genomic events induced by

nuclear calcium signaling regulate persistent neuroprotection is

emerging (Zhang et al., 2009, 2011; Lau and Bading, 2009). In

contrast, nuclear calcium-regulated processes required for

memory formation are unknown. Here we considered the possi-

bility that nuclear calcium signaling modulates structural

features of neurons, in particular the complexity of the dendritic

arbor, that determine their ability to receive and process inputs

(Cline and Haas, 2008). The calcium/calmodulin-dependent

protein kinase IV (CaMKIV), a target of calcium in the nucleus

(Chow et al., 2005), has been implicated in the regulation of

dendritic growth and spine remodeling (Redmond et al., 2002;

Marie et al., 2005), suggesting that nuclear calcium may repre-

sent an important signal in these processes. In this study we

identify vascular endothelial growth factor D (VEGFD), a mitogen

for endothelial cells and regulator of angiogenesis and lymphatic

vasculature (Lohela et al., 2009), as a target of nuclear calcium-

CaMKIV signaling in hippocampal neurons. We also show that

VEGFD is required for the maintenance of a complex dendritic

arbor and provides the molecular link between neuronal activity,

the regulation of dendritic geometry, and cognitive functioning.

RESULTS

Nuclear Calcium Controls Dendrite Geometryand Spine DensityTo investigate the role of nuclear calcium signaling in the regula-

tion of dendrite architecture, we expressedCaMBP4 in the nuclei

of hippocampal neurons. This protein contains four repeats of

the M13 calmodulin (CaM) binding peptide derived from the

rabbit skeletal muscle myosin light chain kinase (Wang et al.,

1995). CaMBP4 effectively inactivates the nuclear calcium/

CaM complex and blocks genomic responses induced by

nuclear calcium signaling (Zhang et al., 2007, 2009; Papadia

et al., 2005). Morphometric analyses revealed that, compared

to control, hippocampal neurons expressing CaMBP4 along

with humanized Renilla reniformis green fluorescent protein

(hrGFP) to visualize the cells showed a significant decrease

Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc. 117

Figure 1. Nuclear Calcium Signaling Regu-

lates Neuronal Morphology

(A) Representative micrographs of hippocampal

neurons transfected with an expression vector for

hrGFP or cotransfected with expression vectors

for hrGFP and CaMBP4 or for hrGFP and CaM-

KIVK75E. Scale bar = 20 mm.

(B and C) Quantification of total dendritic length

and Sholl analysis in hippocampal neurons trans-

fected with the indicated constructs.

(D) Representative micrographs of dendritic

spines of hippocampal neurons transfected as

indicated. Scale bar = 5 mm.

(E) Quantification of spine density of neurons

transfected as in (D).

(F and G) Cumulative frequency plots of spine

length and spine width in neurons transfected as

indicated. More than 1000 spines and 12 neurons

from a minimum of 3 independent preparations

were examined for each construct.

Statistically significant differences are indicated

with asterisks, **p < 0.005, ***p < 0.0005. Bars

represent mean ± SEM.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

both in the total dendritic length and in the complexity of the

dendrites assessed by Sholl analysis (Figures 1A–1C). Expres-

sion of CaMBP4 also caused a significant decrease in dendritic

spine density (Figures 1D and 1E) and a considerable shortening

and thinning of the remaining spines (Figures 1F and 1G). A

similar reduction in total dendritic length, dendritic complexity,

spine size, and spine density was observed in hippocampal

neurons expressing CaMKIVK75E, a dominant negative mutant

of CaMKIV (Anderson et al., 1997) (Figure 1). These results indi-

cate that nuclear calcium is an important signal in the control of

dendritic geometry and spine density.

Nuclear Calcium-CaMKIV Signaling RegulatesVEGFD ExpressionWe next aimed at identifying nuclear calcium/CaMKIV-regulated

genes that mediate the observed structural changes. Examina-

tion of transcriptome data obtained from hippocampal neurons

expressing CaMBP4 (Zhang et al., 2009) drew our attention to

VEGFD (also known as Fos-induced growth factor) (Lohela

et al., 2009). VEGFD is well known for its role in angiogenesis

118 Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc.

and lymphangiogenesis in healthy tissues

and in several types of cancer (Achen and

Stacker, 2008). VEGFD is detectable in

the nervous system (http://www.brain-

map.org/; Lein et al., 2007) but a function

for this secreted factor in neurons has not

been described, although two other

VEGF family members, VEGF (also

known as VEGFA) and VEGFC, have

been implicated in neurogenesis and

the maturation of newly born neurons

(Cao et al., 2004; Le Bras et al., 2006;

Licht et al., 2010). Quantitative reverse

transcriptase PCR (QRT-PCR) analysis

revealed that VEGFD is expressed in

cultured hippocampal neurons and reaches peak levels of

expression after a culturing period of 10–13 days (Figure 2A

and Figure S1A, available online). VEGFD mRNA and protein

are detectable in cultured neurons and in vivo at different devel-

opmental stages in the mouse hippocampus and cortex (Figures

S1C–S1F). VEGFD expression is significantly lower in hippo-

campal neurons infected with a recombinant adenoassociated

virus (rAAV) containing an expression cassette for either

CaMBP4 (rAAV-CaMBP4) or CaMKIVK75E (rAAV-CaMKIVK75E)

than in uninfected neurons or in neurons infected with an

rAAV-expressing LacZ (rAAV-LacZ) (Figure 2B and Figure S1A).

The expression levels of many other genes, including other

members of the VEGF family, were not affected by CaMBP4 or

CaMKIVK75E (Zhang et al., 2009) (Figure S1B). The neurotrop-

ism of our rAAVs specifically targets neurons over glia (Xu

et al., 2001) indicating that the modulation of VEGFD expression

is restricted to neurons.

To investigate a possible contribution of neuronal activity to

VEGFD expression, we treated cultured hippocampal neurons

with TTX for 5 days (from day in vitro [DIV] 5 until the time point

Figure 2. VEGFD Expression Is Nuclear

Calcium Signaling/CBP Dependent

(A) QRT-PCR analysis of VEGFD expression in

hippocampal cultures at the indicated DIV (n = 5).

(B) QRT-PCR analysis of VEGFD expression in

uninfected hippocampal neurons and in hippo-

campal neurons infected with rAAVs giving rise to

the indicated proteins (n = 5).

(C) QRT-PCR analysis of VEGFD, VEGFC, cFos,

gapdh, b tubulin, and b microglobulin expression

in cultured hippocampal neurons on DIV 10 with or

without treatment from DIV 5 to DIV 10 with 1 mM

TTX (n = 3).

(D) QRT-PCR analysis of VEGFD, VEGFC, and

cFos expression in cultured hippocampal neurons

with or without treatment for 2 hr with 50 mM

bicuculline. VEGFD and VEGFC (y) axis scale is

shown on the left; cFos (y) axis scale is shown on

the right (n = 3).

(E) QRT-PCR analysis of VEGFD, VEGFC, and

cFos expression in cultured hippocampal neurons

with or without treatment from DIV 5 to DIV 10 with

10 mM MK801, 10 mM nifedipine, or both (n = 3).

(F) QRT-PCR analysis of expression of cFos and

VEGFD in uninfected hippocampal neurons and

in hippocampal neurons infected with rAAV-

CaMBP4 with or without treatment with actino-

mycin D (10 mg/ml) for 30 min, 1 hr, 2 hr, 4 hr, and

24 hr (n = 3).

(G) QRT-PCR analysis of expression of cFos and

VEGFD in uninfected hippocampal neurons and in

hippocampal neurons infected with rAAV-E1A or

rAAV-E1ADCR1 with or without treatment for 2 hr

with 50 mM bicuculline (bic). VEGFD (y) axis scale

is shown on the right; cFos (y) axis scale is shown

on the left (n = 3).

(H) QRT-PCR analysis of expression of GAPDH,

CBP, and VEGFD in hippocampal neurons with or

without treatment from DIV 8 to DIV 13 with the

indicated siRNAs (n = 3).

(I and J) Analysis of total dendritic length and Sholl

analysis of neurons infected with rAAV-E1A or

rAAV-E1ADCR1. Twelve neurons from a minimum

of three independent preparations were examined

for each construct.

Statistically significant differences are indicated

with asterisks (*p < 0.05, **p < 0.005, ***p <

0.0005). Bars represent mean ± SEM. See also

Figure S1.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

of gene expression analysis on DIV 10). We found that compared

to controls, this treatment reduced expression of VEGFD and

also that of cFos, a well-characterized neuronal activity marker

(Figure 2C). For several other genes analyzed in parallel we found

no differences in expression levels after TTX treatment (Fig-

ure 2C). Given the prolonged duration of the TTX treatment, we

cannot rule out the possibility that the observed reduction in

VEGFD expression may be caused indirectly through secondary

effects. To determine if VEGFD mRNA levels are affected by an

increase in synaptic activity, we exposed hippocampal neurons

to the GABAA receptor blocker bicuculline. This treatment

relieves tonic, GABAA receptor-mediated inhibition of synaptic

transmission from the hippocampal network and induces bursts

of action potentials (APs) (Hardingham et al., 2001; Arnold et al.,

2005). Bicuculline treatment caused a robust induction of cFos

mRNA but did not alter VEGFC or VEGFD mRNA levels (Fig-

ure 2D). To investigate the involvement of NMDA receptors

and voltage-gated calcium channels in VEGFD expression, we

treated hippocampal neurons with the NMDA receptor blocker

MK801 and/or nifedipine, a blocker of L-type voltage-gated

calcium channels. Both MK801 and nifedipine significantly

reduced the expression of VEGFD mRNA; treatment with a

combination of both channel blockers yielded the largest reduc-

tion in VEGFD mRNA levels (Figure 2E). These results indicate

that VEGFD expression is controlled by basal neuronal activity

through a mechanism that is initiated by NMDA receptors and

Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc. 119

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

L-type voltage-gated calcium channels and requires nuclear

calcium-CaMKIV signaling.

Because the observed reduction of VEGFDmRNA levels after

inhibition of nuclear calcium signaling could be due to a change

in VEGFD mRNA stability, we next determined the half-life of

VEGFDmRNA. The levels of VEGFDmRNA and in parallel those

of cFos were measured in both uninfected and rAAV-CaMBP4-

infected hippocampal neurons before and at various time points

(0.5–24 hr) after treatment of the cells with actinomycin D, an

inhibitor of gene transcription. We found that VEGFD mRNA

has a half-life of more than 24 hr in uninfected hippocampal

neurons; a virtually identical decay rate for VEGFDwas observed

in rAAV-CaMBP4-infected neurons (Figure 2F), although com-

pared to uninfected controls, the absolute amounts of VEGFD

mRNA in these neurons were lower (see also Figure 2B and Fig-

ure S1A). Analysis of cFosmRNA revealed a half-life of less than

1 hr (Figure 2F), which is in agreement with other studies (Schiavi

et al., 1992). These results indicate that the regulation of VEGFD

expression by nuclear calcium signaling takes place at the level

of gene transcription rather than at the posttranscriptional level.

CBP Regulates VEGFD ExpressionIn silico analysis with the Transcription Element Search System

(TESS; http://www.cbil.upenn.edu/cgi-bin/tess/tess) of a 2000

base pairs-long upstream regulatory region of themurine VEGFD

gene revealed a large number of possible binding sites for

several transcription factors including the AP-1 complex,

NF-AT, MEF-2, HiNF, NF-kB, POU2-Oct, and HNF4. However,

a cAMP response element (CRE) appears to be lacking, suggest-

ing that nuclear calcium-CaMKIV-mediated regulation of VEGFD

takes place by transcription factors other than CREB, the proto-

typical target of this signaling pathway (Hardingham et al., 1997,

2001). Because the activity of the transcriptional coactivator

CBP is controlled by nuclear calcium and CaMKIV (Chawla

et al., 1998), we next tested the role of CBP in VEGFD regulation.

CBP interacts with a variety of transcription factors (Bedford

et al., 2010), which includes some of those for which putative

binding sites have been identified in the VEGFD gene (see

above). Moreover, a contribution of CBP to the regulation of

the human VEGFD promoter in cancer cells has been suggested

(Schafer et al., 2008). To directly investigate a possible role of

CBP in the regulation of the endogenous VEGFD gene in hippo-

campal neurons, we infected the neurons with an rAAV express-

ing the adenovirus protein E1A. E1A binds to CBP via its amino

terminal-conserved region 1 (CR1) and disrupts CBP function

(Arany et al., 1995; Lundblad et al., 1995; Bannister and Kouzar-

ides, 1995). As expected, rAAV-mediated expression of E1A

blocked the AP bursting-induced increase in the expression of

cFos (Figure 2G), a known target of the CREB/CBP transcription

factor complex (Chawla et al., 1998; Hardingham et al., 1999;

Greer andGreenberg, 2008). Expression of E1A also significantly

reduced VEGFD mRNA levels (Figure 2G). Infection of hippo-

campal neurons with an rAAV expressing a mutant version

of E1A (E1ADCR1) that lacks CR1 and fails to interact with

CBP (Arany et al., 1995; Lundblad et al., 1995; Bannister and

Kouzarides, 1995) had no effect on cFos regulation or VEGFD

transcription (Figure 2G). We also used RNA interference

(RNAi) to specifically decrease CBP mRNA levels in hippo-

120 Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc.

campal neurons (Figure 2H). This caused a significant reduction

of VEGFD mRNA levels (Figure 2H) confirming the role of CBP

in modulating VEGFD transcription. A morphometric analysis re-

vealed that hippocampal neurons expressing E1A have shorter

and simplified dendritic trees compared to neurons expressing

E1ADCR1 (Figures 2I and 2J). These results indicate that CBP

acts downstream of nuclear calcium-CaMKIV signaling to regu-

late VEGFD expression in hippocampal neurons.

VEGFD Restores Dendrite Complexity but Not SpineDensity in Nuclear Calcium Signaling-Depleted NeuronsTo investigate whether VEGFD is involved in mediating the

effects of nuclear calcium-CaMKIV signaling on neuronal struc-

ture, we either transfected or infected hippocampal neurons,

respectively, with an rAAV plasmid (pAAV-VEGFD) or an rAAV

(rAAV-VEGFD) containing an expression cassette for HA-tagged

VEGFD or we treated the neurons with recombinant VEGFD

(rVEGFD). Expression of HA-tagged VEGFD was detected im-

munocytochemically and by immunoblotting in rAAV-VEGFD-in-

fected hippocampal neurons and in the culture media (Figures

S1G and S1H). Although VEGFD-HA expression or exogenously

applied rVEGFD had no detectable effect on neuronal

morphology, both treatments rescued the reduction in dendrite

length and complexity caused by expression of CaMBP4 or

CaMKIVK75E (Figures 3A–3C, 3E, and 3F). In contrast,

VEGFD-HA and rVEGFD failed to restore normal spine density

in CaMBP4 or CaMKIVK75E-expressing neurons (Figures 3D

and 3G), indicating that the mechanisms through which nuclear

calcium-CaMKIV signaling regulate dendrite geometry and spine

density are distinct. Because VEGFD belongs to a family of

closely related factors that in part share the receptors (Achen

and Stacker, 2008), we tested whether VEGF or VEGFC also

affect dendrite arborization. However, neither recombinant

VEGF (rVEGF) nor recombinant VEGFC (rVEGFC) was able to

rescue the reduction in dendrite length and complexity caused

by CaMBP4 or CaMKIVK75E expression (Figure S2), indicating

a specific role for VEGFD in the control of dendrite arborization

by nuclear calcium-CaMKIV signaling.

VEGFD Autocrine Signaling Is Essential for ComplexDendritic ArborizationTo determine whether the observed reduction in VEGFD expres-

sion that followed blockade of nuclear calcium-CaMKIV

signaling is sufficient to alter dendritic architecture, we used

RNAi to lower VEGFD expression in hippocampal neurons.

DNA sequences encoding short hairpin RNAs (shRNAs) de-

signed to target the mouse VEGFD mRNA were inserted down-

stream of the U6 promoter of an rAAV vector. The resulting rAAV,

rAAV-shVEGFD, also harbors a calcium/calmodulin-dependent

protein kinase II (CaMKII) promoter-containing expression

cassette for the red fluorescent protein, mCherry (Figure 4A).

Control rAAVs were identical to rAAV-shVEGFD except that

they either lacked DNA sequences encoding shRNAs (rAAV-

emptymC) or contained DNA sequences encoding a scrambled

version of the VEGFD-specific shRNA (rAAV-shSCR). Infection

rates of 80%–95% of the neuron population were obtained for

all three rAAVs (Figures S3A and S3B). QRT-PCR and immuno-

blot analysis revealed that rAAV-shVEGFD, but not rAAV-shSCR

Figure 3. VEGFD Regulates Dendrite Archi-

tecture

(A) Representative micrographs of neurons trans-

fected with expression vectors for the indicated

proteins with or without treatment for 3 days with

rVEGFD (100 ng/ml). Scale bar = 20 mm.

(B–D) Sholl analysis and analysis of total dendritic

length and spine density of neurons transfected,

as indicated, with expression vectors for hrGFP,

CaMBP4, CaMKIVK75E, and VEGFD-HA.

(E–G) Sholl analysis and analysis of total dendritic

length and spine density of neurons transfected

with expression vectors for hrGFP, CaMBP4, and

CaMKIVK75E with or without treatment for 3 days

with rVEGFD. Twelve neurons from a minimum of

three independent preparations were examined

for each construct.

Statistically significant differences are indicated

with asterisks (*p < 0.05, **p < 0.005, ***p <

0.0005). Bars represent mean ± SEM. See also

Figure S2.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

or rAAV-emptymC, reduced VEGFD mRNA levels and blocked

VEGFD protein expression (Figure 4B and Figure S3C). Expres-

sion of VEGFC was not affected by rAAV-shVEGFD or by the

two control rAAVs (Figure 4B). It has been reported that expres-

sion of certain shRNAs can have an effect on neuronal mor-

phology because of the induction of an interferon response

(Alvarez et al., 2006). However, by using an interferon-responsive

reporter gene system we found no evidence for an interferon

response induced by rAAV-shVEGFD (Figure S3D). In addition,

we observed no increase in cell death in hippocampal neurons

infected with rAAV-shVEGFD (Figure S3E).

Morphological analyses revealed that compared to hippo-

campal neurons transfected with pAAV-shSCR or pAAV-emp-

tymC, neurons transfected with pAAV-shVEGFD showed a less

complex dendritic arbor and a reduction in total dendritic length

Neuron 71, 117–

(Figures 4C–4E). In contrast, RNAi-medi-

ated knockdown of VEGFD did not

change spine density (number of spines/

20 mm: 7.1 ± 0.36, pAAV-emptymC;

6.17 ± 0.56, pAAV-shSCR; 6.52 ± 0.51,

pAAV-shVEGFD). Similar results were ob-

tained with different shRNA sequences

directed against VEGFD (Figure S3F and

Supplemental Experimental Procedures).

The effect of pAAV-shVEGFD transfec-

tion on the dendritic tree could be re-

versed by treatment with rVEGFD. In

contrast, rVEGFD did not affect dendrite

length or complexity of hippocampal

neurons transfected with pAAV-shSCR

and pAAV-emptymC (Figures 4C–4E).

These results identify a role for VEGFD

in the regulation of dendritic architecture

and further support the above-mentioned

concept (see Figure 3) that dendrite

arborization and spine morphogenesis

are controlled by distinct nuclear calcium/CaMKIV-regulated

processes.

The observation that the dendrite structure is altered in

shVEGFD-expressing neurons even if the surrounding untrans-

fected cells have a normal VEGFD expression level suggests

a possible autocrine mechanism of action of VEGFD. To inves-

tigate this further, we transfected hippocampal neurons with

pAAV-VEGFD-HA or with a plasmid containing an expression

cassette for HA-tagged VEGFD resistant to shVEGFD (pAAV-

resiVEGFD-HA) together with pAAV-shVEGFD in order to over-

express VEGFD in the same neurons expressing shVEGFD.

Expression of resiVEGFD-HA rescued the reduction of dendrite

length and complexity caused by expression of shVEGFD

(Figures 4F–4H), indicating that VEGFD acts in an autocrine

manner. This conclusion is further supported by an experiment

130, July 14, 2011 ª2011 Elsevier Inc. 121

Figure 4. RNAi-Mediated Suppression of

VEGFD Expression Is Sufficient to Alter

Dendritic Morphology

(A) Schematic representation of the rAAV vector

used for shRNA expression.

(B) QRT-PCR analysis of expression of VEGFD

and VEGFC in uninfected hippocampal neurons

and in hippocampal neurons infected with rAAV-

emptymC, rAAV-shVEGFD, or with rAAV-shSCR

(n = 5).

(C) Representative micrographs of hippocampal

neurons transfected with pAAV-emptymC, pAAV-

shVEGFD, or pAAV-shSCR with or without treat-

ment for 3 days with rVEGFD (100 ng/ml). The

mCherry signal was used to visualize the neurons.

Scale bar = 20 mm.

(D and E) Sholl analysis and analysis of total

dendritic length of neurons transfected as in (C)

and, where indicated, treated with rVEGFD.

Twelve neurons from a minimum of three inde-

pendent preparations were examined for each

construct.

(F) Analysis of total dendritic length of neurons

transfected as in (G). Twelve neurons from a

minimum of three independent preparations were

examined for each construct.

(G) Representative micrographs of hippocampal

neurons transfected with pAAV-hrGFP (vector),

pAAV-VEGFD-HA, or pAAV-resiVEGFD-HA with

or without cotransfection with pAAV-shVEGFD.

Scale bar = 20 mm.

(H) Sholl analysis of neurons transfected as in (G).

Twelve neurons from a minimum of three inde-

pendent preparations were examined for each

construct.

Statistically significant differences are indicated

with asterisks (*p < 0.05, **p < 0.005, ***p <

0.0005). Bars represent mean ± SEM. See also

Figure S3.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

in which hippocampal neurons were first infected with rAAV-

VEGFD and subsequently transfected with pAAV-shVEGFD.

Because infection rates are very high but transfection rates

are very low, this creates a situation in which a small number

of transfected cells with low VEGFD expression levels and

reduced dendrite length and arborization are surrounded by in-

fected cells overexpressing VEGFD. We found that even under

these conditions the impairment in dendrite morphology caused

by shVEGFD cannot be overcome by the VEGFD overex-

pressed in the infected neurons (Figures S3G–S3I). Thus,

although we cannot fully exclude paracrine action of VEGFD,

all available evidence strongly suggests that VEGFD regulates

total dendrite length and complexity through an autocrine

mechanism.

122 Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc.

VEGFD Regulates DendriticArborization via VEGFR3Human VEGFD and its close relative

VEGFC can bind and activate both VEGF

receptors 2 and 3 (VEGFR2 andVEGFR3);

however,murineVEGFDcanonly activate

VEGFR3 (Baldwin et al., 2001). To investi-

gate whether dendritic architecture is specifically controlled by

VEGFD acting via VEGFR3, we generated rAAVs expressing

shRNAs specific for VEGF (rAAV-shVEGF), VEGFC (rAAV-

shVEGFC), and VEGFR3 (rAAV-shVEGFR3) (Wong et al., 2005;

Kleinman et al., 2008). By using QRT-PCR we showed that

rAAV-shVEGF, rAAV-shVEGFC, rAAV-shVEGFR3, and rAAV-

shVEGFD reduced mRNA levels of their respective targets

leaving unaltered the expression of the other VEGF family

members (Figure 5A). Morphological analyses revealed that

transfection of hippocampal neurons with pAAV-shVEGF or

pAAV-shVEGFC, similar to transfection with the control plas-

mids, pAAV-shSCRorpAAV-emptymC, hadnoeffect ondendrite

length or complexity (Figures 5B–5D). In contrast, knockdown of

VEGFR3 by transfecting neurons with pAVV-shVEGFR3 led to

Figure 5. RNAi-Mediated Suppression of

VEGFR3 Causes the Same Simplified

Dendritic Arborization Phenotype Observed

by Silencing VEGFD

(A) QRT-PCR analysis of expression normalized to

uninfected cells of VEGFD, VEGF, VEGFC, and

VEGFR3 in uninfected hippocampal neurons

and in hippocampal neurons infected with rAAV-

emptymC, rAAV-shVEGFD, rAAV-shVEGF, rAAV-

shVEGFC, or rAAV-shVEGFR3 (n = 3).

(B) Representative micrographs of hippocampal

neurons transfected with pAAV-emptymC, pAAV-

shSCR, pAAV-shVEGFD, pAAV-shVEGF, pAAV-

shVEGFC, or pAAV-shVEGFR3. The mCherry

signal was used to visualize the neurons. Scale

bar = 20 mm.

(C and D) Sholl analysis (C) and analysis of total

dendritic length (D) of neurons transfected as in

(B). Twelve neurons from a minimum of three

independent preparations were examined for

each construct. Statistically significant differences

are indicated with asterisks (*p < 0.05, **p < 0.005).

Bars represent mean ± SEM.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

changes in the dendritic structure that were virtually identical

to those obtained in hippocampal neurons transfected with

pAAV-shVEGFD (Figures 5B–5D; see also Figures 4C–4H and

Figures S3G–S3I for the effects of pAAV-shVEGFD on dendrite

morphology). These results indicate that among VEGF family

members, VEGFD, acting through VEGFR3, plays a specific

role in the regulation of dendrite arborization.

VEGFD Regulates Multiple Signaling Pathwaysin Hippocampal Neurons and Shapes DendriteMorphology via p38 MAP Kinase ActivationWe next determined the signaling mechanisms through which

VEGFD controls dendrite architecture. Cell lysates from hippo-

Neuron 71, 117–

campal neurons treated with rVEGFD

for various lengths of time were sub-

jected to immunoblot analysis by using

a large panel of antibodies that are

specific for the phosphorylated (i.e., acti-

vated) forms of signaling molecules (Fig-

ure 6). We found that rVEGFD activates

ERK1/2, p38 MAP kinase (MAPK), and

CREB (Figures 6A and 6B). The increase

in ERK1/2 phosphorylation and CREB

phosphorylation (which takes place in

neurons and not in glial cells as shown

by double immunostaining with the

neuronal marker NeuN; Figure 6C) was

significant but moderate (Figures 6A

and 6B). In contrast, the activation of

p38 MAPK was very robust (Figures 6A

and 6B), indicating that it may be a major

transducer of VEGFD signaling in hippo-

campal neurons. We therefore deter-

mined whether p38 MAPK mediates the

effects of VEGFD on dendrite geometry.

Indeed, selective blockade of p38 MAPK by using the p38

MAPK inhibitor SB203580 severely compromised the ability of

rVEGFD to rescue the impairment of dendrite length and com-

plexity in hippocampal neurons expressing CaMBP4 and CaM-

KIVK75E (Figures 6D and 6E; see also Figures 3A, 3E, and 3F).

Because SB203580 inhibits the alpha and beta isoforms of

p38 MAPK (Bain et al., 2007), we further investigated the role

of p38 MAPK in VEGFD-mediated dendritic arborization by

RNAi. We generated two pAAVs, shp38a and shp38b, that

contain expression cassettes for shRNAs specific for the alpha

and the beta isoform, respectively, of p38 MAPK. We found

that the reduction of p38 alpha MAPK expression prevented

the rVEGFD-induced rescue of the dendrite phenotypes of

130, July 14, 2011 ª2011 Elsevier Inc. 123

Figure 6. VEGFD-Induced Signaling Events

(A) Immunoblot analysis by using phosphospecific

antibodies of hippocampal neurons with or with-

out rVEGFD treatment for the indicated times.

Tubulin was used as loading control.

(B) Quantification of the experiment shown in (A).

rVEGFD treatment causes a significant increase of

CREB, ERK1/2, and p38 MAP kinase phosphory-

lation. All values are expressed as percentage of

untreated controls (n = 5).

(C) Immunocytochemical analysis by using phos-

pho-CREB-specific antibodies of hippocampal

neurons with or without rVEGFD treatment for the

indicated times. Nuclei of cells were counter-

stained with Hoechst 33258; NeuN was used as a

neuronal marker. Representative images are

shown. Scale bar = 20 mm.

(D) Analysis of total dendritic length of neurons

transfected with expression vectors for hrGFP,

CaMBP4, and CaMKIVK75E with or without

treatment for 3 days with rVEGFD and/or

SB203580. Twelve neurons from a minimum of

three independent preparations were examined

for each experimental condition.

(E) Sholl analysis of neurons transfected and

treated as in (D). Sholl analysis data of neurons

expressing hrGFP, CaMBP4, or CaMKIVK75E

treated with SB203580 showed no difference

compared to the respective untreated controls

and were omitted from the graph for clarity.

(F) Analysis of total dendritic length of neurons

transfected with expression vectors for hrGFP,

CaMBP4, shSCR, shp38a or shp38b (as indicated)

with or without treatment for 3 days with rVEGFD.

Twelve neurons from a minimum of three inde-

pendent preparations were examined for each

experimental condition. Statistically significant

differences are indicated with asterisks (*p < 0.05,

**p < 0.005, ***p < 0.0005). Bars represent

mean ± SEM.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

hippocampal neurons expressing CaMBP4 (Figure 6F). These

results indicate that p38 alpha MAPK is required for VEGFD

regulation of dendrite architecture.

VEGFD Modulates Network ActivityTo investigate whether VEGFD-regulated changes in the struc-

ture of dendrites are associated with changes in neuronal

network activity, we used microelectrode array (MEA) record-

ings. Indeed, the spike frequencies of hippocampal cultures in-

fected with rAAV-shVEGFD were reduced compared to cultures

124 Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc.

infected with rAAV-shSCR or rAAV-

emptymC. This decrease could be partly

rescued by the addition of rVEGFD to

the media (Figure 7A). The decrease in

network activity caused by infection with

rAAV-shVEGFD was first observed at

DIV 10 (Figure 7A), coinciding with the

onset of robust VEGFDmRNA expression in vitro (see Figure 2A

and Figure S1A).

Patch-Clamp Recordings Verify Reduced Surface Areaand Excitability after Silencing VEGFD ExpressionThe effects of silencing VEGFD expression on the electrical prop-

erties of neurons were investigated with whole-cell patch clamp

recordings (Table S1 and Figure 7B). Neurons either transfected

with pAAV-shVEGFD or infected with rAAV-shVEGFD showed, in

comparison to their respective control group, a markedly smaller

Figure 7. MEA and Patch-Clamp Analysis

Reveal Reduced Activity, Surface Area,

and Synaptic Transmission in shVEGFD-

Expressing Neurons

(A) MEA analysis of absolute spike frequencies

of uninfected hippocampal neurons and hippo-

campal neurons infected with rAAV-emptymC,

rAAV-shVEGFD, and rAAV-shSCR. Where indi-

cated, cultureswere treatedatDIV6with100ng/ml

rVEGFD. Statistically significant differences are

indicated with asterisks (*p < 0.05). Bars represent

mean ± SEM.

(B) Scatter plot showing the distribution and mean

(horizontal bars) membrane capacitance (Cm) of

transfected and infected neurons expressing

shVEGFD or shSCR. Statistically significant differ-

ences are indicated with asterisks (****p < 0.0001).

(C) Representative mEPSCs averaged from 160–

220 events in a pAAV-shVEGFD and apAAV-shSCR

transfected cell showing the rawaverage above and

the scaled average below.

(D) Representative responses to bath application of

AMPA (10 mM, black or red line) in the presence of

cyclothiazide (20 mM), TTX (1 mM), and gabazine

(5 mM) recorded from pAAV-shVEGFD and pAAV-

shSCR transfected cells.

(E–G) Scatter plots showing the distribution and

mean (horizontal bars) mEPSC IEI (E), mEPSC

amplitude (F), and AMPA response (G). Note the log

scale of the ordinate axis in (E) and (F). Significant

differences between shVEGFD-expressing cells

and their respective transfected or infected shSCR-

expressing controls are indicated (*p < 0.05, **p <

0.01, with Kolmogorov-Smirnov two-sample tests;

n = 26 to 31 cells per group). See also Table S1.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

membranecapacitance indicativeofa reducedplasmamembrane

surface area, a finding consistent with the observed reduction in

dendritic arborization (Figures 4 and 5). Despite this difference,

shVEGFD-expressing neurons did not show an altered resting

membrane potential or threshold membrane potential for action

potential initiation (Table S1). This reflects the healthy integrity of

these neurons despite their altered morphology. Slightly more

current injection was necessary to elicit an action potential in

shVEGFD-expressingcells,although this trendwasonlysignificant

in thegroupofhippocampalneurons inwhich infectionwasused to

express shVEGFD (Table S1). Moreover, we found stronger

accommodation in spike patterns induced by squarewave current

injections in shVEGFD-expressing neurons (data not shown). This

suggests a mildly reduced excitability in shVEGFD-expressing

neurons, consistent with the reduced absolute spike frequency

identified with MEA recordings (see Figure 7A). Increased accom-

modation may be due to a reduced contribution of dendritic

sodium channels, h-channels, or calcium-activated potassium

channels, which in pyramidal neurons can drive slow repetitive

firing and influence burst waveforms.

Neuron 71, 117–

Lowering VEGFD ExpressionReduces the Number of FunctionalAMPA ReceptorsThe influence of VEGFD expression on

synaptic transmission in hippocampal

neurons in culture was directly assessed by recording miniature

excitatory postsynaptic currents (mEPSCs) in the presence of

TTX and the GABAA receptor blocker, gabazine. Neurons trans-

fected with pAAV-shVEGFD or infected with rAAV-shVEGFD

showed longer mEPSC interevent intervals (IEIs, 1/frequency)

and smaller mEPSC amplitudes than their respective shSCR-

expressing controls (Figures 7E and 7F). The reduced mEPSC

frequency in transfected hippocampal neurons suggests that

the effect was not mediated by a reduced release probability

presynaptically because the low transfection rate ensures that

the majority of synaptic input to shRNA-expressing cells comes

from non-shRNA-expressing cells. The reduced mEPSC fre-

quency is thus most likely indicative of fewer AMPA receptor-

containing synapses per cell. The 21%–24% reduction in

mEPSC amplitude also suggests a lower density of AMPA recep-

tors at synapses in shVEGFD-expressing cells. mEPSCs of

hippocampal neurons expressing shVEGFD also showed faster

rise and decay time constants than their respective shSCR-ex-

pressing controls (Figure 7C and Table S1), most likely due to

reduced filtering of mEPSCs in their more compact dendritic

130, July 14, 2011 ª2011 Elsevier Inc. 125

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

trees. Alternatively, a synaptic NMDA receptor-mediated slow

component of the mEPSC may have been reduced in

shVEGFD-expressing neurons, although significant NMDA

currents are unlikely in our recording conditions (�71mV holding

potential, 1.3 mM Mg2+). Responses were also recorded to

bath-applied AMPA, which produced a peak within 30 s whose

amplitude was used as an indication of the total number of func-

tional AMPA receptors per cell (Figure 7D). AMPA response

amplitudes were smaller in hippocampal neurons expressing

shVEGFD (Figures 7D and 7G), indicative of a reduced total

number of surface-expressed AMPA receptors per cell. Taken

together, our patch-clamp analysis has identified a reduced

plasma membrane surface area, as well as a reduced number

of AMPA receptor-containing synapses, a reduced number of

AMPA receptors per synapse, and a reduced total number of

AMPA receptors in shVEGFD-expressing cells. These results

are consistent with the reduced dendritic morphology identified

by morphometric analyses.

VEGFD Shapes Dendritic Arborization In VivoWe next investigated the role of VEGFD in vivo. rAAV-shVEGFD

or the appropriate control rAAVs were stereotaxically delivered

to the dorsal hippocampus of 2-month-old C57BL/6 male

mice. Infected neurons were readily identified by analysis of

the mCherry fluorescence (Figure S4). The morphology of

neurons in the CA1 area of the hippocampus was assessed by

manually tracing the basal dendrites of Golgi-stained brain slices

obtained from animals 2.5 weeks after viral gene delivery. As in

cultured neurons, infection of hippocampal neurons in vivo

with rAAV-shVEGFD, but not with rAAV-shSCR or rAAV-emp-

tymC, reduced both the total length of dendrites and their

complexity (Figures 8A–8C).

VEGFD Is Required for Memory FormationWe finally investigated whether the changes in neuronal struc-

ture induced by RNAi-mediated knockdown of VEGFD causes

cognitive deficits. Mice stereotaxically injected with rAAV-

shVEGFD or rAAV-shSCR into the hippocampus were tested in

two well-characterized hippocampus-dependent memory tests:

Morris water maze and contextual fear conditioning. In the

hidden-platform version of the Morris water maze, mice learn

the location of the platform by using distal visual cues (Morris

et al., 1982). We found that mice stereotaxically injected with

rAAV-shSCR or rAAV-shVEGFD needed significantly less time

to find the hidden platform across training trials (main effect of

training session: F[7,91] = 11.30, p < 0.0001); however, no effect

of treatment was observed (main effect of treatment: F[1,13] =

1.34, p = NS) (Figure 8D). This suggests that both groups have

developed a learning strategy to find the hidden platform.

Multiple behavioral strategies may be employed by the mice to

obtain the reward (escape from water) and some of these strat-

egiesmay be comparably efficient but distinct in the requirement

for hippocampal function (Lipp and Wolfer, 1998; Gerlai, 2001).

To assess spatial memory, we performed a probe trial during

which the platform was removed from the water maze and the

mice were given 60 s to search for it. The search pattern during

the probe trial can reveal a spatial preference that is believed

to represent spatial memory. We observed that rAAV-shSCR-

126 Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc.

injected mice show a spatial preference for the target quadrant,

whereas rAAV-shVEGFD-injected mice did not (Figure 8E). This

was expressed as a significantly longer time spent in the target

quadrant by the rAAV-shSCR-injected mice in comparison

to the other quadrants (one-way analysis of variance [ANOVA]:

F[3,28] = 9.84, p < 0.001; multiple post hoc comparisons: time

in target quadrant versus time in adjacent or opposite quadrants;

p < 0.05). In contrast, rAAV-shVEGFD-injected mice spent

similar amounts of time in the different quadrants (F[3,24] =

1.2, p = NS). This suggests that the rAAV-shVEGFD-injected

mice did not develop a spatial searching strategy, pointing to

spatial memory impairment in these mice. Swimming speed

was not different between the two groups (Figure 8F, main effect

of treatment: F[1,13] = 3.94, p = NS).

To determine whether abnormalities in motivation, motor

coordination, or vision could account for the deficit in spatial

memory, we also trained mice on a visible-platform version of

the water maze, a hippocampus-independent task. In this task,

the mice use a proximal, visual cue to locate the platform (Fig-

ure 8G). rAAV-shSCR- and rAAV-shVEGFD-injected mice

showed similar escape latencies to find the visible platform

(main effect of training session: F[2,26] = 7.12, p < 0.01; main

effect of treatment: F[1,13] = 0.36, p = NS), demonstrating that

both groups acquired the task (Figure 8G). Overall, these results

suggest that rAAV-shVEGFD-injected mice have impaired

spatial memory.

In the contextual fear conditioning, mice learn the association

between an aversive stimulus, a mild foot shock, and the context

in which it was delivered. Inmice that have formed an associative

memory, a second exposure to the same context induces a fear-

ful response expressed as freezing or immobility, parameters

used to quantify the formation of memory (Maren, 2001). We

found that mice stereotaxically injected with rAAV-shVEGFD

showed significantly lower levels of freezing during the 24 hr

test session than didmice injected with rAAV-shSCR (Figure 8H).

The reduction in freezing levels was not due to decreased loco-

motor activity or pain sensitivity because the basal exploratory

activity and reaction to shock during the training session were

not different between the two groups (Figures 8I and 8J). These

findings together with the results obtained with the Morris water

maze indicate that VEGFD is important for memory formation.

DISCUSSION

In this study, we identify VEGFD as a regulator of neuronal

dendrite geometry. VEGFD mediates the effects of synaptic

activity and nuclear calcium-CaMKIV signaling on the mainte-

nance of complex dendrite arborization, which is necessary for

memory formation.

Nuclear Calcium, Dendrite Geometry, and CognitionNeurons, even once fully developed, remain plastic and undergo

activity-dependent functional or structural alterations. Changes

in gene expression – induced by synaptic activity and calcium

transients propagating toward and into the nucleus (Chawla

et al., 1998; Hardingham et al., 1997, 2001; Zhang et al., 2009) –

are often essential for the long-term maintenance of adaptive re-

sponses (Hardingham and Bading, 2010; Greer and Greenberg,

Figure 8. RNAi-Mediated Suppression of

VEGFD Expression In Vivo Affects Den-

dritic Morphology and Impairs Memory

Formation

(A) Golgi-stained CA1 pyramidal neurons from the

ipsilateral and contralateral hemispheres of mice

stereotaxically injected into the hippocampus with

rAAV-shVEGFD, rAAV-emptymC, or rAAV-shSCR.

Representative tracings of the basal dendritic tree

are shown. Scale bar = 10 mm.

(B and C) Sholl analysis and analysis of total

dendritic length of CA1 Golgi-stained pyramidal

neurons from the ipsilateral and contralateral

hippocampus of mice stereotaxically injected into

the hippocampus with the indicated constructs.

Twenty neurons for each condition obtained from

four injected animals per construct were analyzed.

(D) Mean escape latency during acquisition of the

hidden-platform version of the Morris water maze

(rAAV-shVEGFD, n = 8; rAAV-shSCR, n = 7).

(E) Mean percent time spent in each quadrant

during the probe trial test of the Morris water

maze. T, target quadrant; CW, clockwise quadrant

in relation to target; CCW, counterclockwise

quadrant in relation to target; O, opposite quad-

rant in relation to target (rAAV-shVEGFD, n = 8;

rAAV-shSCR, n = 7).

(F) Mean swimming speed during acquisition of

the hidden-platform version of the Morris water

maze (rAAV-shVEGFD, n = 8; rAAV-shSCR, n = 7).

(G) Mean escape latency during acquisition of the

visible-platform version of the Morris water maze

(rAAV-shVEGFD, n = 8; rAAV-shSCR, n = 7).

(H) Long-term contextual fear memory. Mice were

injected with rAAV-shVEGFD or rAAV-shSCR;

results are expressed as the percentage of time

spent immobile during the contextual memory test

24 hr after training (rAAV-shVEGFD, n = 10; rAAV-

shSCR, n = 11).

(I) Speed of movement during the 2 s electric foot

shock delivered in training session of animals

represented in (H) (rAAV-shVEGFD, n = 10; rAAV-

shSCR n = 11).

(J) Distance traveled during the initial 2.5 min

exposure to the training chamber of animals

represented in (H) (rAAV-shVEGFD, n = 10; rAAV-

shSCR, n = 11).

Statistically significant differences (ANOVA fol-

lowed by Tukey’s post hoc test for morphometric,

repeated-measures ANOVA and one-way ANOVA followed by Tukey’s post hoc test for Morris water maze and Student’s t test for behavioral analysis) are

indicated with asterisks (*p < 0.05, **p < 0.005, ***p < 0.0005). Bars represent mean ± SEM. See also Figure S4.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

2008). Dendritic trees, the branched projections of the input-

receiving endsof neurons, are prime targets for activity-regulated

structural alterations. The geometry of dendrites specifies the

connectivity of neurons and strongly influences how signals are

integrated and transmitted to the cell soma and therefore also

which output is produced. Changes in the lengths and branching

patterns of dendrites would be expected to alter not only the

performance of a neuron but also the computational power of

the network the neuron is part of, ultimately causing changes in

the organism’s behavior. Support for such a link between

dendritic architecture and cognitive abilities comes from theoret-

ical considerations and mathematical modeling (Hausser et al.,

2000; Segev and London, 2000) aswell as frombrainmorphology

studies of neurological diseases. In particular, shortening and

simplification of dendrites have been observed in a variety of

disorders that are associatedwithmental retardation or cognitive

deficits, including genetic abnormalities, such as Down syn-

drome or Rett syndrome (Kaufmann andMoser, 2000), neurode-

generative conditions, including Alzheimer’s disease and aging

(Dickstein et al., 2007), metabolic dysfunctions (DeLong, 1993;

Patil et al., 2007), and infection with human immunodeficiency

virus (Masliah et al., 1997). Intrinsic, predetermined genetic

programs and cell-autonomous mechanisms have been shown

to determine dendrite morphogenesis in developing neurons

Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc. 127

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

(Kim et al., 2009; Goodman, 1978; Jan and Jan, 2003; Goldberg,

2004; Corty et al., 2009). However, there is an increasing appreci-

ation of the influence of the electrical activity of neurons on

dendrite arborization (Zhang and Poo, 2001; Chen and Ghosh,

2005). We show here that VEGFD plays a central role in this

process; as a target of nuclear calcium-CaMKIV signaling, it links

basal neuronal activity to the control of total dendrite length and

branching patterns, thereby providing neurons of the adult

nervous system with the structural features needed for proper

cognitive performance of the organism. These findings explain

why interference either with nuclear calcium signaling or with

CaMKIV activity compromises the ability of mice to form long-

term memories (Limback-Stokin et al., 2004; Kang et al., 2001).

Theyalsosuggestagenerallyapplicableconcept, inwhich impair-

ments of synaptic transmission, e.g., because of synapse loss in

aging or Alzheimer’s disease (LaFerla, 2002; Shankar et al.,

2007; Kuchibhotla et al., 2008) and/or malfunctioning of activity-

induced calcium signaling toward and within the cell nucleus

(‘‘nuclear calciopathy,’’ Zhanget al., 2011)may lead to a decrease

in VEGFD expression, followed by a reduction in dendrite

complexity, and finally, an emergence of cognitive deficits. Strat-

egies aimed at maintaining or restoring appropriate dendrite

lengths and branching patterns—either through supplementation

of VEGFD or enhancement of nuclear calcium signaling—may

therefore represent avenues for the development of effective

therapies for age- and disease-related cognitive dysfunction.

Anti-VEGFD Cancer Treatment and CompromisedBrain FunctionSeveral members of the VEGF family, including VEGFD, are well-

known angiogenic and lymphangiogenic mitogens that drive the

formation of blood vessels and lymphatic vasculature in healthy

tissues. VEGFD can also enhance the formation of tumor

lymphatics, thereby promoting tumor growth and the lymphatic

spreadof cancercells (Stackeret al., 2001). Indeed,VEGFD levels

in cancer patients correlate with parameters that indicate a poor

clinical prognosis (Achen and Stacker, 2008). Consequently,

tremendous efforts are presently being directed toward the

development and use of VEGF-targeted drugs as antiangiogenic

and antilymphangiogenic treatments. These drugs are intended

to inhibit the growth of tumors by cutting off their blood supply

and blocking themetastatic spread of tumor cells to lymph nodes

via the lymphatic vasculature (Heath and Bicknell, 2009). The

unexpected role we have uncovered for VEGFD in the control of

dendritic architecture and cognitive function thus calls for caution

in the use of blockers of VEGFD signaling as a cancer therapy.

EXPERIMENTAL PROCEDURES

Hippocampal Cultures and Transfection

Hippocampal neurons from newborn C57BL/6 mice were cultured as

described (Bading and Greenberg, 1991; Zhang et al., 2011). Experiments

were done after a culturing period of 10–14 DIV. DNA transfection was done

on DIV 8 by using Lipofectamine 2000 (Invitrogen, San Diego, CA) as described

(Wiegert et al., 2007). For the studies on dendrite morphology, neurons were

analyzed 4–5 days after transfection.

Stereotaxic Delivery of rAAVs

rAAVs were delivered by stereotaxic injection into the right dorsal hippo-

campus of 2-month-old male C57BL/6 mice. For the Golgi staining-based

128 Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc.

morphometric analysis, viral particles were unilaterally injected over a period

of 20 min at the following coordinates relative to Bregma: anteroposterior,

�2.1 mm; mediolateral, �1.4 mm; and dorsoventral, �1.4 to �1.8 mm from

the skull surface. For behavior experiments mice were injected bilaterally.

Morphometric Analyses

For three-dimensional Sholl analysis (Sholl, 1953), total dendritic length and

spine morphology were calculated by using Object-Image freeware software

with a specific set of macros (written by Dr. E. Ruthazer, McGill University,

Quebec, Canada). Briefly, a z-stack acquisition was imported, calibrated in

Object-Image, and manually traced. Total dendritic length was then com-

puted. For Sholl analysis, the shell interval was set at 5 mm. All analyses

were performed blind. In all in vitro experiments, for each condition, aminimum

of 12 neurons from 3 independent preparations was analyzed. For the Golgi

staining, quantifications of 20 neurons for each condition from 4 different

injected animals per viral construct were traced and analyzed.

MEA Recordings

MEA recordings were done as described (Arnold et al., 2005). From DIV 7 to

DIV 13, recordings of spontaneous network activity were acquired for 5 min

once per day.

Patch-Clamp Recordings

Whole-cell patch-clamp recordings were made at room temperature from

cultured hippocampal neurons plated on coverslips secured with a platinum

ring in a recording chamber (Open Access Chamber-1, Science Products

GmbH, Hofheim, Germany) mounted on a fixed-stage upright microscope

(BX51WI, Olympus, Hamburg, Germany). Differential interference contrast

optics, infrared illumination, and a CCD camera (PCO, Visitron Systems, Puch-

heim, Germany) connected to a contrast enhancement unit (Argus, Hama-

matsu, Herrsching am Ammersee, Germany) were used to view neurons on

a videomonitor. Recordingsweremadewith aMulticlamp 700A amplifier, digi-

tized through a Digidata 1322A A/D converter, and acquired by using pClamp

software (Molecular Devices, Sunnyvale, CA). All membrane potentials were

corrected for the calculated junction potential of �11 mV (JPCalc program

by Dr. Peter H. Barry). mEPSCs and whole-cell AMPA (10 mM, 6 ml/min, Bio-

trend) responses were recorded as previously described (Wiegert et al.,

2009) except for the solutions (see Supplemental Information) and except

that a holding potential of –71 mV was used and both TTX (1 mM, Biotrend,

Cologne, Germany) and gabazine (5 mM, Biotrend) were included in the

ACSF. TTX was applied before gabazine to prevent the induction of action

potential bursting. Access (7–18 MU) and membrane resistance (see Table

S1) were monitored before and after mEPSC recordings and data were

rejected if changes greater than 20% occurred.

Behavioral Experiments

The behavioral experiments started 3 weeks after stereotaxic delivery of

rAAVs. Mice were habituated to the experimental room and handled once a

day for 3 consecutive days before testing started. Two different sets of mice

were used for Morris water maze and contextual fear conditioning experi-

ments. At the end of the experiments, virus infection was assessed. For details

on the behavioral experiments see Supplemental Information.

Data Analysis

All plotted data representmean ± SEM. ANOVAwith Tukey’s post hoc test was

used for statistical analyses except where stated otherwise. In those experi-

ments where only two conditions are tested, comparisons were made by using

a Student’s t test for independent samples. In the Morris water maze experi-

ment, repeated-measures ANOVA and one-way ANOVA were used to analyze

acquisition curves and probe trials, respectively.

SUPPLEMENTAL INFORMATION

Supplemental Information includes four figures, one table, and Supplemental

Experimental Procedures and can be found with this article online at

doi:10.1016/j.neuron.2011.04.022.

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

ACKNOWLEDGMENTS

We thank I. Bunzli-Ehret for her help with the preparation of hippocampal

cultures; U. Weiss for western blot analysis; Drs. S. Oliviero, B.L. Sabatini,

and A. Asano for providing plasmids; Drs. S.J. Zhang, D. Lau, and S. Romorini

for discussions; and Dr. A.M. Hagenston for comments on the manuscript. All

images of Golgi-impregnated tissues were taken at the Nikon Imaging Center

at Heidelberg University. This work was supported by the Alexander von Hum-

boldt Foundation (Wolfgang Paul Prize to H.B.), a European Research Council

(ERC) Advanced Grant (to H.B.), the Sonderforschungsbereich (SFB) 488 and

SFB 636 of the Deutsche Forschungsgemeinschaft (DFG), the European

Network of Excellence NeuroNE, and the European Union Project Glutamate

Receptor Interacting Proteins as Novel Neuroprotective Targets (EU Project

GRIPANNT). H.B. is a member of the Excellence Cluster ‘‘CellNetworks’’ at

Heidelberg University. D.M. is recipient of a EuropeanMolecular Biology Orga-

nization (EMBO) Long-Term Fellowship; A.M.M.O. is recipient of a Postdoc-

toral Fellowship from the Foundation for Science and Technology, Portugal.

Accepted: April 9, 2011

Published: July 13, 2011

REFERENCES

Achen, M.G., and Stacker, S.A. (2008). Molecular control of lymphatic metas-

tasis. Ann. N Y Acad. Sci. 1131, 225–234.

Alvarez, V.A., Ridenour, D.A., and Sabatini, B.L. (2006). Retraction of synapses

and dendritic spines induced by off-target effects of RNA interference.

J. Neurosci. 26, 7820–7825.

Anderson, K.A., Ribar, T.J., Illario, M., and Means, A.R. (1997). Defective

survival and activation of thymocytes in transgenic mice expressing a catalyt-

ically inactive form of Ca2+/calmodulin-dependent protein kinase IV. Mol.

Endocrinol. 11, 725–737.

Arany, Z., Newsome, D., Oldread, E., Livingston, D.M., and Eckner, R. (1995). A

family of transcriptional adaptor proteins targeted by the E1A oncoprotein.

Nature 374, 81–84.

Arnold, F.J., Hofmann, F., Bengtson, C.P., Wittmann, M., Vanhoutte, P., and

Bading, H. (2005). Microelectrode array recordings of cultured hippocampal

networks reveal a simple model for transcription and protein synthesis-depen-

dent plasticity. J. Physiol. 564, 3–19.

Bading, H., and Greenberg, M.E. (1991). Stimulation of protein tyrosine phos-

phorylation by NMDA receptor activation. Science 253, 912–914.

Baldwin, M.E., Catimel, B., Nice, E.C., Roufail, S., Hall, N.E., Stenvers, K.L.,

Karkkainen, M.J., Alitalo, K., Stacker, S.A., and Achen, M.G. (2001). The spec-

ificity of receptor binding by vascular endothelial growth factor-d is different in

mouse and man. J. Biol. Chem. 276, 19166–19171.

Bain, J., Plater, L., Elliott, M., Shpiro, N., Hastie, C.J., McLauchlan, H.,

Klevernic, I., Arthur, J.S., Alessi, D.R., and Cohen, P. (2007). The selectivity

of protein kinase inhibitors: a further update. Biochem. J. 408, 297–315.

Bannister, A.J., and Kouzarides, T. (1995). CBP-induced stimulation of c-Fos

activity is abrogated by E1A. EMBO J. 14, 4758–4762.

Bedford, D.C., Kasper, L.H., Fukuyama, T., and Brindle, P.K. (2010). Target

gene context influences the transcriptional requirement for the KAT3 family

of CBP and p300 histone acetyltransferases. Epigenetics 5, 9–15.

Bliss, T.V., Collingridge, G.L., andMorris, R.G. (2007). Synaptic Plasticity in the

Hippocampus. In The Hippocampus Book, P. Andersen, R. Morris, D. Amaral,

T. Bliss, and J. O’Keefe, eds. (New York: Oxford Univ. Press), pp. 343–473.

Cao, L., Jiao, X., Zuzga, D.S., Liu, Y., Fong, D.M., Young, D., and During, M.J.

(2004). VEGF links hippocampal activity with neurogenesis, learning and

memory. Nat. Genet. 36, 827–835.

Chawla, S., Hardingham, G.E., Quinn, D.R., and Bading, H. (1998). CBP:

a signal-regulated transcriptional coactivator controlled by nuclear calcium

and CaM kinase IV. Science 281, 1505–1509.

Chen, Y., and Ghosh, A. (2005). Regulation of dendritic development by

neuronal activity. J. Neurobiol. 64, 4–10.

Chow, F.A., Anderson, K.A., Noeldner, P.K., and Means, A.R. (2005). The

autonomous activity of calcium/calmodulin-dependent protein kinase IV is

required for its role in transcription. J. Biol. Chem. 280, 20530–20538.

Cline, H., and Haas, K. (2008). The regulation of dendritic arbor development

and plasticity by glutamatergic synaptic input: a review of the synaptotrophic

hypothesis. J. Physiol. 586, 1509–1517.

Corty, M.M., Matthews, B.J., and Grueber, W.B. (2009). Molecules and

mechanisms of dendrite development in Drosophila. Development 136,

1049–1061.

DeLong, G.R. (1993). Effects of nutrition on brain development in humans. Am.

J. Clin. Nutr. 57 (2, Suppl), 286S–290S.

Dickstein, D.L., Kabaso, D., Rocher, A.B., Luebke, J.I., Wearne, S.L., and Hof,

P.R. (2007). Changes in the structural complexity of the aged brain. Aging Cell

6, 275–284.

Gerlai, R. (2001). Behavioral tests of hippocampal function: simple paradigms

complex problems. Behav. Brain Res. 125, 269–277.

Goldberg, J.L. (2004). Intrinsic neuronal regulation of axon and dendrite

growth. Curr. Opin. Neurobiol. 14, 551–557.

Goodman, C.S. (1978). Isogenic grasshoppers: genetic variability in the

morphology of identified neurons. J. Comp. Neurol. 182, 681–705.

Greer, P.L., and Greenberg, M.E. (2008). From synapse to nucleus: calcium-

dependent gene transcription in the control of synapse development and

function. Neuron 59, 846–860.

Hausser, M., Spruston, N., and Stuart, G.J. (2000). Diversity and dynamics of

dendritic signaling. Science 290, 739–744.

Hardingham, G.E., Chawla, S., Johnson, C.M., and Bading, H. (1997). Distinct

functions of nuclear and cytoplasmic calcium in the control of gene expres-

sion. Nature 385, 260–265.

Hardingham, G.E., Chawla, S., Cruzalegui, F.H., and Bading, H. (1999). Control

of recruitment and transcription-activating function of CBP determines gene

regulation by NMDA receptors and L-type calcium channels. Neuron 22,

789–798.

Hardingham, G.E., Arnold, F.J., and Bading, H. (2001). Nuclear calcium

signaling controls CREB-mediated gene expression triggered by synaptic

activity. Nat. Neurosci. 4, 261–267.

Hardingham, G.E., and Bading, H. (2010). Synaptic versus extrasynaptic

NMDA receptor signalling: implications for neurodegenerative disorders.

Nat. Rev. Neurosci. 11, 682–696.

Heath, V.L., and Bicknell, R. (2009). Anticancer strategies involving the vascu-

lature. Nat Rev Clin Oncol 6, 395–404.

Hu, S.C., Chrivia, J., and Ghosh, A. (1999). Regulation of CBP-mediated

transcription by neuronal calcium signaling. Neuron 22, 799–808.

Jan, Y.N., and Jan, L.Y. (2003). The control of dendrite development. Neuron

40, 229–242.

Kang, H., Sun, L.D., Atkins, C.M., Soderling, T.R., Wilson, M.A., and

Tonegawa, S. (2001). An important role of neural activity-dependent CaMKIV

signaling in the consolidation of long-term memory. Cell 106, 771–783.

Kaufmann, W.E., and Moser, H.W. (2000). Dendritic anomalies in disorders

associated with mental retardation. Cereb. Cortex 10, 981–991.

Kim, A.H., Puram, S.V., Bilimoria, P.M., Ikeuchi, Y., Keough, S., Wong, M.,

Rowitch, D., and Bonni, A. (2009). A centrosomal Cdc20-APC pathway

controls dendrite morphogenesis in postmitotic neurons. Cell 136, 322–336.

Kleinman, M.E., Yamada, K., Takeda, A., Chandrasekaran, V., Nozaki, M.,

Baffi, J.Z., Albuquerque, R.J., Yamasaki, S., Itaya, M., Pan, Y., et al. (2008).

Sequence- and target-independent angiogenesis suppression by siRNA via

TLR3. Nature 452, 591–597.

Kuchibhotla, K.V., Goldman, S.T., Lattarulo, C.R., Wu, H.Y., Hyman, B.T., and

Bacskai, B.J. (2008). Abeta plaques lead to aberrant regulation of calcium

homeostasis in vivo resulting in structural and functional disruption of neuronal

networks. Neuron 59, 214–225.

LaFerla, F.M. (2002). Calcium dyshomeostasis and intracellular signalling in

Alzheimer’s disease. Nat. Rev. Neurosci. 3, 862–872.

Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc. 129

Neuron

Nuclear Calcium and VEGFD in Dendrite Geometry

Lau, D., and Bading, H. (2009). Synaptic activity-mediated suppression of p53

and induction of nuclear calcium-regulated neuroprotective genes promote

survival through inhibition of mitochondrial permeability transition.

J. Neurosci. 29, 4420–4429.

Le Bras, B., Barallobre, M.J., Homman-Ludiye, J., Ny, A., Wyns, S., Tammela,

T., Haiko, P., Karkkainen, M.J., Yuan, L., Muriel, M.P., et al. (2006). VEGF-C is

a trophic factor for neural progenitors in the vertebrate embryonic brain. Nat.

Neurosci. 9, 340–348.

Lein, E.S., Hawrylycz, M.J., Ao, N., Ayres, M., Bensinger, A., Bernard, A., Boe,

A.F., Boguski, M.S., Brockway, K.S., Byrnes, E.J., et al. (2007). Genome-wide

atlas of gene expression in the adult mouse brain. Nature 445, 168–176.

Licht, T., Eavri, R., Goshen, I., Shlomai, Y., Mizrahi, A., and Keshet, E. (2010).

VEGF is required for dendritogenesis of newly born olfactory bulb interneu-

rons. Development 137, 261–271.

Limback-Stokin, K., Korzus, E., Nagaoka-Yasuda, R., and Mayford, M. (2004).

Nuclear calcium/calmodulin regulates memory consolidation. J. Neurosci. 24,

10858–10867.

Lipp, H.P., and Wolfer, D.P. (1998). Genetically modified mice and cognition.

Curr. Opin. Neurobiol. 8, 272–280.

Lohela, M., Bry, M., Tammela, T., and Alitalo, K. (2009). VEGFs and receptors

involved in angiogenesis versus lymphangiogenesis. Curr. Opin. Cell Biol. 21,

154–165.

Lundblad, J.R., Kwok, R.P., Laurance, M.E., Harter, M.L., and Goodman, R.H.

(1995). Adenoviral E1A-associated protein p300 as a functional homologue of

the transcriptional co-activator CBP. Nature 374, 85–88.

Maren, S. (2001). Neurobiology of Pavlovian fear conditioning. Annu. Rev.

Neurosci. 24, 897–931.

Marie, H., Morishita, W., Yu, X., Calakos, N., and Malenka, R.C. (2005).

Generation of silent synapses by acute in vivo expression of CaMKIV and

CREB. Neuron 45, 741–752.

Masliah, E., Heaton, R.K., Marcotte, T.D., Ellis, R.J., Wiley, C.A., Mallory, M.,

Achim, C.L., McCutchan, J.A., Nelson, J.A., Atkinson, J.H., and Grant, I.;

HNRC Group. The HIV Neurobehavioral Research Center. (1997). Dendritic

injury is a pathological substrate for human immunodeficiency virus-related

cognitive disorders. Ann. Neurol. 42, 963–972.

Morris, R.G., Garrud, P., Rawlins, J.N., and O’Keefe, J. (1982). Place naviga-

tion impaired in rats with hippocampal lesions. Nature 297, 681–683.

Papadia, S., Stevenson, P., Hardingham, N.R., Bading, H., and Hardingham,

G.E. (2005). Nuclear Ca2+ and the cAMP response element-binding protein

family mediate a late phase of activity-dependent neuroprotection.

J. Neurosci. 25, 4279–4287.

Patil, C.G., Lad, S.P., Katznelson, L., and Laws, E.R., Jr. (2007). Brain atrophy

and cognitive deficits in Cushing’s disease. Neurosurg. Focus 23, E11.

Redmond, L., Kashani, A.H., and Ghosh, A. (2002). Calcium regulation of

dendritic growth via CaM kinase IV and CREB-mediated transcription.

Neuron 34, 999–1010.

Schafer, G., Wissmann, C., Hertel, J., Lunyak, V., and Hocker, M. (2008).

Regulation of vascular endothelial growth factor D by orphan receptors

hepatocyte nuclear factor-4 alpha and chicken ovalbumin upstream promoter

transcription factors 1 and 2. Cancer Res. 68, 457–466.

130 Neuron 71, 117–130, July 14, 2011 ª2011 Elsevier Inc.

Schiavi, S.C., Belasco, J.G., and Greenberg, M.E. (1992). Regulation of proto-

oncogene mRNA stability. Biochim. Biophys. Acta 1114, 95–106.

Segev, I., and London, M. (2000). Untangling dendrites with quantitative

models. Science 290, 744–750.

Shankar, G.M., Bloodgood, B.L., Townsend, M., Walsh, D.M., Selkoe, D.J.,

and Sabatini, B.L. (2007). Natural oligomers of the Alzheimer amyloid-beta

protein induce reversible synapse loss by modulating an NMDA-type gluta-

mate receptor-dependent signaling pathway. J. Neurosci. 27, 2866–2875.

Sholl, D.A. (1953). Dendritic organization in the neurons of the visual andmotor

cortices of the cat. J. Anat. 87, 387–406.

Silva, A.J., Kogan, J.H., Frankland, P.W., and Kida, S. (1998). CREB and

memory. Annu. Rev. Neurosci. 21, 127–148.

Stacker, S.A., Caesar, C., Baldwin, M.E., Thornton, G.E., Williams, R.A., Prevo,

R., Jackson, D.G., Nishikawa, S., Kubo, H., and Achen, M.G. (2001). VEGF-D

promotes the metastatic spread of tumor cells via the lymphatics. Nat. Med. 7,

186–191.

Wang, J., Campos, B., Jamieson, G.A., Jr., Kaetzel, M.A., and Dedman, J.R.

(1995). Functional elimination of calmodulin within the nucleus by targeted

expression of an inhibitor peptide. J. Biol. Chem. 270, 30245–30248.

Wiegert, J.S., Bengtson, C.P., and Bading, H. (2007). Diffusion and not active

transport underlies and limits ERK1/2 synapse-to-nucleus signaling in hippo-

campal neurons. J. Biol. Chem. 282, 29621–29633.

Wiegert, J.S., Hofmann, F., Bading, H., and Bengtson, C.P. (2009). A transcrip-

tion-dependent increase in miniature EPSC frequency accompanies late-

phase plasticity in cultured hippocampal neurons. BMC Neurosci. 10, 124.

Wong, S.Y., Haack, H., Crowley, D., Barry, M., Bronson, R.T., and Hynes, R.O.

(2005). Tumor-secreted vascular endothelial growth factor-C is necessary for

prostate cancer lymphangiogenesis, but lymphangiogenesis is unnecessary

for lymph node metastasis. Cancer Res. 65, 9789–9798.

Xu, R., Janson, C.G., Mastakov, M., Lawlor, P., Young, D., Mouravlev, A.,

Fitzsimons, H., Choi, K.L., Ma, H., Dragunow, M., et al. (2001). Quantitative

comparison of expression with adeno-associated virus (AAV-2) brain-specific

gene cassettes. Gene Ther. 8, 1323–1332.

Zhang, L.I., and Poo,M.M. (2001). Electrical activity and development of neural

circuits. Nat. Neurosci. Suppl. 4, 1207–1214.

Zhang, S.J., Steijaert, M.N., Lau, D., Schutz, G., Delucinge-Vivier, C.,

Descombes, P., and Bading, H. (2007). Decoding NMDA receptor signaling:

identification of genomic programs specifying neuronal survival and death.

Neuron 53, 549–562.

Zhang, S.J., Zou, M., Lu, L., Lau, D., Ditzel, D.A., Delucinge-Vivier, C., Aso, Y.,

Descombes, P., and Bading, H. (2009). Nuclear calcium signaling controls

expression of a large gene pool: identification of a gene program for acquired

neuroprotection induced by synaptic activity. PLoS Genet. 5, e1000604.

Zhang, S.J., Buchthal, B., Lau, D., Hayer, S., Dick, O., Schwaninger, M.,

Veltkamp, R., Zou, M., Weiss, U., and Bading, H. (2011). A signaling cascade

of nuclear calcium-CREB-ATF3 activated by synaptic NMDA receptors

defines a gene repression module that protects against extrasynaptic NMDA

receptor-induced neuronal cell death and ischemic brain damage.

J. Neurosci. 31, 4978–4990.