Exercise Reverses Behavioral and Synaptic Abnormalities ...neuronet.jp/pdf/O_220.pdf · Cell Reports Report Exercise Reverses Behavioral and Synaptic Abnormalities after Maternal

Report

Exercise Reverses Behavi

Graphical Abstract

Highlights

d MIA causes behavioral and synaptic abnormalities in the

offspring

d Microglia-dependent synaptic engulfment is impaired byMIA

d Voluntary running in adulthood ameliorates MIA-induced

abnormalities

d Voluntary running stimulates microglia-mediated engulfment

of surplus synapses

Andoh et al., 2019, Cell Reports 27, 2817–2825June 4, 2019 ª 2019 The Author(s).https://doi.org/10.1016/j.celrep.2019.05.015

Authors

Megumi Andoh, Kazuki Shibata,

Kazuki Okamoto, ..., Yuki Miura,

Yuji Ikegaya, Ryuta Koyama

[email protected]

In Brief

Andoh et al. find that maternal immune

activation (MIA) causes autism spectrum

disorder (ASD)-like behaviors and

synaptic surplus in the offspring in mice.

Voluntary running normalizes synaptic

density and ameliorates abnormal

behaviors even after the onset of ASD-like

behaviors, probably by boosting synaptic

engulfment by microglia.

Cell Reports

Report

Exercise Reverses Behavioral and SynapticAbnormalities after Maternal InflammationMegumi Andoh,1,3 Kazuki Shibata,1,3 Kazuki Okamoto,1 Junya Onodera,1 Kohei Morishita,1 Yuki Miura,1 Yuji Ikegaya,1,2

and Ryuta Koyama1,4,*1Laboratory of Chemical Pharmacology, Graduate School of Pharmaceutical Sciences, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku,

Tokyo 113-0033, Japan2Center for Information and Neural Networks, 1-4 Yamadaoka, Suita City, Osaka 565-0871, Japan3These authors contributed equally4Lead Contact

*Correspondence: [email protected]

https://doi.org/10.1016/j.celrep.2019.05.015

SUMMARY

Abnormal behaviors in individuals with neurodeve-lopmental disorders are generally believed to be irre-versible. Here, we show that voluntary wheel runningameliorates the abnormalities in sociability, repeti-tiveness, and anxiety observed in a mouse model ofa neurodevelopmental disorder induced by maternalimmune activation (MIA). Exercise activates a portionof dentate granule cells, normalizing the density ofhippocampal CA3 synapses, which is excessive inthe MIA-affected offspring. The synaptic surplus inthe MIA offspring is induced by deficits in synapseengulfment by microglia, which is normalized byexercise through microglial activation. Finally, che-mogenetically induced activation of granule cellspromotes the engulfment of CA3 synapses. Thus,our study proposes a role of voluntary exercise inthe modulation of behavioral and synaptic abnormal-ities in neurodevelopmental disorders.

INTRODUCTION

Physical exercise has attracted attention as a potential thera-

peutic strategy for abnormal behaviors in neurodevelopmental

disorders such as autism spectrum disorders (ASDs) (Pan,

2010; Anderson-Hanley et al., 2011), but the cellular and molec-

ular mechanisms for how exercise ameliorates the symptoms

remain unrevealed. Previous studies reported that exercise

may help improve cognitive performance in humans and rodents

(Aberg et al., 2009; Gomes da Silva et al., 2012), probably by

enhancing the production of neurotrophic factors and neurogen-

esis (de Almeida et al., 2013; Pereira et al., 2007). However, the

question of whether exercise modifies synaptic connections,

which are fundamental structures for brain function, remains

unanswered.

The frequency of neurodevelopmental disorders is positively

correlated with maternal immune activation (MIA), which is

induced by viral infection during pregnancy (Knuesel et al.,

2014), and several studies have reported that deficits in synap-

CelThis is an open access article und

tic function and structure underlie the pathogenesis of neurode-

velopmental disorders (Koyama and Ikegaya, 2015). Specif-

ically, the disruption of synapse excitatory versus inhibitory

(E/I) balance (Rubenstein and Merzenich, 2003; Yizhar et al.,

2011; Tyzio et al., 2014) and the increased spine or synapse

density (Tang et al., 2014; Jawaid et al., 2018; Andoh et al.,

2016) are shared pathological features between ASD patients

and ASD animal models. These findings motivated us to

examine whether voluntary exercise in adulthood reverses the

behavioral abnormalities and affects synaptic properties in

mouse offspring prenatally subjected to MIA (MIA offspring).

To test this idea, we focused on the hippocampus, because

physical exercise has been suggested to activate neurons in

the hippocampal dentate gyrus (Clark et al., 2011) and sociabil-

ity, impairment of which is a major symptom of ASD, is partly

mediated by the hippocampus (Hitti and Siegelbaum, 2014;

Okuyama et al., 2016).

In the present study, we examined the role of microglia, the

resident immune cells in the brain, in synaptic deficits in MIA

offspring. Microglia continuously survey the brain environment,

monitoring and modulating synaptic structure and function; mi-

croglia prune synapses in the lateral geniculate nucleus (LGN)

and hippocampal CA1 (Schafer et al., 2012; Paolicelli et al.,

2011) during development. We examined whether MIA induces

synaptic deficits by affecting synaptic pruning by microglia,

because MIA has been suggested to affect microglial properties

such as number, morphology, and cytokine expression (Zhu

et al., 2014; Hui et al., 2018; Mattei et al., 2017), possibly leading

to abnormalities in microglial functions, including their phago-

cytic capacities (Giovanoli et al., 2013; Fernandez de Cossıo

et al., 2017).

RESULTS

Exercise in Adulthood Ameliorates MIA-AssociatedBehaviorsWe used offspring from pregnant mice intraperitoneally injected

with the synthetic double-stranded RNA polyinosinic:polycyti-

dylic acid (poly(I:C)), on embryonic days 12.5 and 17.5 to model

MIA induced by viral infection during pregnancy (Malkova et al.,

2012; Naviaux et al., 2013). Consistent with previous reports, we

found that MIA offspring at postnatal day (P) 60 exhibited deficits

l Reports 27, 2817–2825, June 4, 2019 ª 2019 The Author(s). 2817er the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Figure 1. Adult Exercise Ameliorates MIA-Associated Behaviors

(A) Social interaction behavior assessed by the three-chamber test and graphed as a social preference index (percentage of total investigation time spent

interacting with a stranger). n = 40–45 mice.

(B) Repetitive behavior assessed by self-grooming time. n = 31–33 mice.

(C) Anxiety behavior assessed by the novelty-suppressed feeding test. Latency to feed was used as an index of anxiety and is graphed as the fraction of mice that

did not eat over a period of 5 min. n = 12–24 mice.

(D) Representative confocal images of the dentate gyrus and hippocampal CA3 immunostained for NeuN and c-Fos at P60. Mice underwent exercise from P30 to

P60. DG, dentate gyrus.

(E) Density (percentage of control) of c-Fos(+) cells in several brain regions in P60 mice that underwent exercise. n = 4 mice (2–6 regions/mouse). S1, primary

somatosensory cortex; PRh, perirhinal cortex.

Mean ± SEM (A, B, and E). *p < 0.05 and **p < 0.01 (A–C), and **p < 0.01 versus control (E); Dunnett’s test (A and B), Gehan-Breslow test (C), and Student’s t test

(E). Mice from 7 litters in (A), 6 litters in (B), and 4 litters in (C) were used for each condition (control, MIA, and MIA + exercise).

in social interaction using the three-chamber assay (Figure 1A)

and repetitive behaviors by analyzing self-grooming time (Fig-

ure 1B). BecauseMIA is likely associatedwith anxiety in offspring

(Ulmer-Yaniv et al., 2018), we also performed the novelty-sup-

pressed feeding test and found that the MIA offspring displayed

enhanced anxiety (Figure 1C). Next, we examined whether the

MIA-associated behaviors could be reversed by adult voluntary

exercise. For this purpose, we placed a runningwheel in the cage

of MIA offspring from P30 to P60. After a month of voluntary

running exercise, all observed MIA-associated behaviors were

ameliorated (Figures 1A–1C). We confirmed that neither MIA

treatment nor exercise affected basal motility (total distance:

control, 3,621.29 ± 143.69 cm; MIA, 3,472.24 ± 115.58 cm;

MIA + exercise, 3,225.90 ± 163.16 cm; n = 15–20 mice, mean

± SEM, p > 0.05, Dunnett’s test) or appetite (home cage latency

to feed: control, 58.50 ± 7.53 s; MIA, 62.42 ± 6.73 s; MIA + exer-

cise, 59.67 ± 6.56 s; n = 12–24 mice, mean ± SEM, p > 0.05,

Dunnett’s test) of the mice. Altogether, these results suggest

2818 Cell Reports 27, 2817–2825, June 4, 2019

that exercise can attenuate MIA-associated behaviors even in

adulthood.

Exercise in Adulthood Ameliorates Synaptic Defects inMIA OffspringImmunostaining of the immediate-early gene c-Fos demon-

strated that the running exercise preferentially activated the den-

tate granule cells compared to its effect on other brain regions

(Figures 1D and 1E), in agreement with a previous report (Clark

et al., 2011). Therefore, we focused on the properties of the

synapses between the granule cell axons, i.e., the mossy fibers,

and the CA3 pyramidal cells. First, we measured the develop-

mental changes in the density of mossy fiber synapses, which

were defined as puncta that colocalized presynaptic elements

(synaptoporin [SPO], a mossy fiber-specific presynaptic marker)

(Williams et al., 2011) and postsynaptic elements (postsynaptic

density 95 [PSD95]) in the CA3 (Figures 2A and 2B; Figure S1).

The density of mossy fiber synapses decreased from P15 to

Figure 2. Adult Exercise Ameliorates Synaptic Defects in MIA Offspring

(A) P30 hippocampus immunostained for SPO and PSD95.

(B) Upper: representative images of the mossy fiber pathway stratum lucidum (SL) and CA3 stratum pyramidale (SP) at P30. Lower: colocalized signals of SPO

and PSD95 in the boxed areas in the upper images.

(C) Developmental changes in the mossy fiber synapse density. n = 4–9 mice (4–12 regions/mouse).

(D) Mossy fiber synapse density at P60. n = 4–9 mice (10–12 regions/mouse).

(E) mEPSC traces taken from CA3 pyramidal cells from P30–P34 control or MIA offspring.

(legend continued on next page)

Cell Reports 27, 2817–2825, June 4, 2019 2819

P30, probably via developmental synapse elimination, and the

decreased synapse density was maintained until P60 in control

mice. However, the developmental decrease in synapses was

not observed in MIA offspring (Figure 2C), and the synapse

density at P15 was maintained until P60 (Figure 2D). The hippo-

campal neuron density was comparable between control and

MIA offspring (granule cell: control, 14.78 ± 0.34 3 105/mm3;

MIA, 14.75 ± 0.39 3 105/mm3; pyramidal cell: control, 5.63 ±

0.08 3 105/mm3; MIA, 5.77 ± 0.06 3 105/mm3; mean ± SEM,

n = 4 mice per condition, p > 0.05, Student’s t test) and thus

was unlikely to contribute to the increased synapse density in

MIA offspring. The density of inhibitory synapses was not

affected in MIA offspring (Figure S2). Furthermore, we examined

the functional properties of the mossy fiber synapses using

whole-cell patch-clamp recordings from CA3 pyramidal neu-

rons. We measured the miniature excitatory postsynaptic cur-

rents (mEPSCs) derived from the mossy fiber synapses, defined

as the difference in the mEPSCs before and those after applica-

tion of 10 mM DCG-IV, a mGluR2 agonist that specifically blocks

mossy fiber transmission (Mellor andNicoll, 2001).We found that

the mEPSC frequency, but not the amplitude, was higher in MIA

offspring than in control mice (Figures 2E–2I). Finally, we investi-

gated the structural properties of the large mossy fiber boutons,

which include multiple synaptic sites (Rollenhagen et al., 2007),

using Thy1-mGFP mice in which the membrane structure of a

portion of granule cells was labeled with GFP (Figure 2K) (Tao

et al., 2016). We found that the area of each mossy fiber bouton

was greater in the MIA offspring than in the controls (Figure 2L),

but the overall density of mossy fiber boutons was not changed

(Figure 2M). Because the area of mossy fiber boutons and that of

postsynapses in the boutons were positively correlated (Fig-

ure 2N), these results indicate that the density of functional

mossy fiber synapses was increased in adult MIA offspring.

Exercise in Adulthood Stimulates Microglia to PruneSynapses in MIA OffspringNext, we examined whether and how adult exercise affected the

synaptic properties ofMIA offspring. First, we found that theP30–

P60 exercise normalized the increased density of mossy fiber

synapses in MIA offspring to the control level (Figure 2C). More-

over, the elevated frequency ofmEPSCwas reversed by exercise

to the control level (Figure 2J). These results suggested that

exercise attenuates structural and functional abnormalities of

mossy fiber synapses. Next, we examined whether microglia,

the brain immune cells that prune synapses during development

(Schafer et al., 2012; Paolicelli et al., 2011), were involved in the

elimination of mossy fiber synapses. For this purpose, we per-

(F and G) mEPSC frequency (F) and delta frequency (G), i.e., the mossy fiber transm

(1–2 cells/mouse).

(H and I) mEPSC amplitude (H) and delta amplitude (I). n = 12 cells from 8–10 mi

(J) Delta frequency of mEPSC taken from CA3 pyramidal cells from P60–P67 mic

(K) Left: P30 hippocampus of a Thy1-mGFP mouse. Upper right: images of SL a

(L and M) Area (L) and density (M) of mossy fiber boutons. n = 5–6 mice (25–110

(N) Relationship between the area of mossy fiber boutons and the area of Homer

p = 0.029.

Mean±SEM (C, D, F–J, L, andM). *p < 0.05 and **p < 0.01; Tukey’s test after ANOV

correlation coefficient (N). Mice from 3 litters were used for each condition (contro

2820 Cell Reports 27, 2817–2825, June 4, 2019

formed a synapse engulfment assay (Schafer et al., 2012) in the

mossy fiber pathway to detect the volume of synaptic elements

engulfed by microglia in mice, in which microglia are selectively

labeled with GFP (Zhan et al., 2014) (Figure 3A). We confirmed

that no GFP(+) cell in the hippocampal CA3 was immunopositive

for CCR2 (Prinz and Priller., 2010), an infiltrating cell-specific

marker (Figure S3), while GFP(+) cells were stained with an anti-

body for Siglec-H (Zhang et al., 2006; Konishi et al., 2017) and

P2Y12R (Sasaki et al., 2003; Haynes et al., 2006; Mildner et al.,

2017), which were generally used as microglia-specific markers

(Figure S3). Furthermore, GFP(+) cells expressed interleukin-1b

(IL-1b), one of the typical cytokines released from microglia (Fig-

ure S3). We also confirmed that the expression patterns of these

markers were unchanged, even in MIA or minocycline-injected

animals (data not shown). Daily injection of minocycline from

P15 (75 mg/kg subcutaneously [s.c.]), which is frequently used

to suppress microglial activity (Kobayashi et al., 2013),

decreased the volume of CD68-positive lysosomes in microglia

at P18 (Figures S4A andS4B) and suppressed the developmental

decrease in synapse density at P20 (Figure 3B). Furthermore,

PSD95 was engulfed by microglia in the mossy fiber pathway

(Figure 3A; Figures S4C and S4D), and the engulfment of SPO

and the density of microglia were not affected by minocycline

treatment (Figures S4E–S4G). These results suggest that micro-

glia selectively engulfed the postsynaptic sites of mossy fiber

synapses during development. Whether presynapses or postsy-

napses are engulfedmay depend on the developmental changes

of synaptic structure in mossy fiber boutons that occur in the

course of synaptic formation (Figure S8A). In P18 MIA offspring,

the volume of CD68 (Figure 3C) and that of engulfed PSD95

puncta (Figure 3D), but not the density of microglia (Figure 3E),

were lower than in control mice, suggesting that the impaired

synaptic engulfment by microglia underlies the increased density

of mossy fiber synapses in MIA offspring. Finally, we examined

whether microglia were involved in the exercise-induced

enhancement of synapse elimination inMIA offspring (Figure 2C).

To test this idea, we gave MIA offspring daily injections of mino-

cycline throughout the exercise period (P30 to P60). The engulf-

ment of the presynaptic protein VGLUT1 by microglia was lower

in the mossy fiber pathway of the MIA offspring than in that of the

control mice (Figure 3F). In contrast, exercise promoted synaptic

engulfment in MIA offspring to a level comparable to that in con-

trol (Figure 3F) without affecting the density of microglia or the

CD68 volume (microglial density: control, 1.40 3 104 ± 0.17 3

104/mm3; MIA, 1.39 3 104 ± 0.15 3 104/mm3; MIA + exercise,

1.59 3 104 ± 0.15 3 104/mm3; MIA + exercise + minocycline,

1.423 104 ± 0.173 104/mm3; CD68 ofmicroglia volume: control,

ission (frequency without DCG-IV�with DCG-IV). n = 12 cells from 8–10mice

ce (1–2 cells/mouse).

e. n = 4 cells from 3 mice (1–2 cells/mouse).

nd SP. Lower right: representative mossy fiber bouton.

boutons/mouse).

1 in the boutons. n = 31 boutons from 3 mice (8–12 boutons/mouse). r = 0.39,

A (C, F, andH), Dunnett’s test (D), Student’s t test (G, I, L, andM), andPearson’s

l, MIA, and MIA + exercise) at each postnatal day. See also Figures S1 and S2.

Figure 3. Adult Exercise Stimulates Microglia to Prune Synapses in MIA Offspring

(A) (i) Representative image of microglia with CD68 and PSD95 immunostaining in the stratum lucidum (SL) of a CX3CR1GFP/+ mouse at P18. (ii) Themicroglia with

signals outside of the image field were removed. (iii) Merged CD68 (iv) and PSD95 (v) signals within the microglia. Arrows indicate PSD95 in microglial lysosomes.

(B) Mossy fiber synapse density at P15 and P20. Minocycline was injected daily from P15 to P19. n = 8 mice (74–93 fields/mouse).

(C) CD68 volume (percentage of microglial volume) at P18. n = 9 mice (4 fields/mouse).

(D) Engulfed PSD95 volume in microglia (percentage of control average) at P18. n = 8 mice (4–6 fields/mouse).

(E) Microglial density in the SL at P18. n = 4 mice (4 fields/mouse).

(F) Engulfed VGLUT1 volume in microglia (percentage of control average) at P60. n = 7–12 mice (6 fields/mouse).

(G) Mossy fiber synapse density at P60. Minocycline was injected daily from P30 to P60. n = 5–10 mice (10–12 fields/mouse).

Mean ± SEM (B–G). *p < 0.05 and **p < 0.01; Tukey’s test after ANOVA (B), Dunnett’s test (F and G), and Student’s t test (C–E). Mice from 3 litters were used for

each condition (control, MIA, MIA + exercise, and MIA + exercise + minocycline). See also Figures S3–S5.

1.62% ± 0.27%; MIA, 2.37% ± 0.19%; MIA + exercise, 2.41% ±

0.60%; MIA + exercise + minocycline, 1.51% ± 0.36%; mean ±

SEM, n = 7–8 mice, p > 0.05, Dunnett’s test). In addition, the

effect of exercise on synaptic engulfment (Figure 3F) was atten-

uated by minocycline. In agreement with these results, minocy-

cline treatment resulted in an increase in synapse density (Fig-

ure 3G). Because minocycline has various biological actions,

we examined whether minocycline directly affects exercise-

induced activity in the granule cells. We found that minocycline

did not affect the granule cell activity (Figure S5), supporting a

more specific action of minocycline on microglia downstream

of neural activity, though other indirect effects of minocycline

cannot be excluded. For example, minocycline given systemi-

cally might affect other cells beyond microglia, including reduc-

tion of infiltrating cells. In the current experimental conditions,

however, we did not detect evidence of contamination of the

Cell Reports 27, 2817–2825, June 4, 2019 2821

Figure 4. Neuronal Activation Enhances Microglial Synapse Engulfment

(A) AAV8-CaMKIIa-eGFP or AAV8-CaMKIIa-hM3Dq-mCherry was injected into the bilateral dentate gyrus at P10.

(B) Mossy fiber pathway stratum lucidum (SL) and CA3 stratum pyramidale (SP) immunostained for GFP and SPO at P31.

(C) Mossy fiber pathway SL immunostained for mCherry and Iba1 at P31.

(D) Images of microglial processes touching or not touching the mossy fiber boutons (arrow).

(E) Rate of the microglial processes touching the mossy fiber boutons. n = 3 mice (89–180 boutons/mouse).

(F) Relationship between the engulfment of VGLUT1 bymicroglia and the volume ofmCherry(+)mossy fiber boutons in the SLwith or without CNO treatment. n = 3

mice (5–6 fields/mouse). Saline: r = 0.071, p = 0.79; CNO: r = 0.74, p = 0.0007.

Mean ± SEM (E). *p < 0.05; Student’s t test (E) and Pearson’s correlation coefficient (F). See also Figures S6 and S7.

infiltrating cells (Figure S3) in both MIA and minocycline-injected

conditions. In addition, the density of Iba1(+) cells was not

changed in MIA and minocycline-injected conditions (control,

1.40 3 104 ± 0.17 3 104/mm3; MIA, 1.39 3 104 ± 0.15 3

104/mm3; MIA + exercise, 1.59 3 104 ± 0.15 3 104/mm3; MIA +

exercise + minocycline, 1.42 3 104 ± 0.17 3 104/mm3; mean ±

SEM, n = 7–8mice, p > 0.05, Dunnett’s test). From these findings,

we speculate that it is likely that minocycline did not affect the

number of infiltrating cells under the conditions and regions of in-

terest used in the current study. Altogether, these results suggest

that exercise activates microglia in adult MIA offspring to engulf

surplus synapses in the mossy fiber pathway.

Neuronal Activation Enhances Microglial SynapseEngulfmentFinally, we examined whether increased activity of granule cells

is sufficient to induce engulfment of the mossy fiber synapses

by microglia, because activity-dependent synapse competition

is a key factor that primes microglia to eat weaker synapses

(Schafer et al., 2012). For this purpose, we used the chemoge-

2822 Cell Reports 27, 2817–2825, June 4, 2019

netic technology designer receptors exclusively activated by

designer drugs (DREADD), a method for remote and transient

manipulation of the activity of cells that express the designer re-

ceptors, namely, mutated human muscarinic receptors (hM3Dq

and hM4Di) that are exclusively activated by the designer drug

clozapine N-oxide (CNO) (Alexander et al., 2009). We first

confirmed that the local injection of the adeno-associated virus

(AAV) into the dentate gyrus (Figure 4A) successfully labeled the

SPO-positive mossy fiber synapses using AAV8-CaMKIIa-eGFP

(Figure 4B). Then, we transfected the granule cells with hM3Dq

in combination with mCherry (27.0% ± 3.7% of granule cells

expressed mCherry, mean ± SEM, n = 6 mice) to induce neural

activity using AAV8-CaMKIIa-hM3Dq-mCherry and confirmed

with c-Fos immunoreactivity that hM3D-expressing granule

cells increased activity in response to CNO injection (Figure S6).

To examine the interaction between the mossy fiber bouton and

the microglia (Figure 4C), we assessed the rate of microglial pro-

cesses touching hM3Dq-expressing mossy fiber boutons (Fig-

ure 4D) and found that the intraperitoneal injection of CNO at

P30 decreased the touching events in 24 h (Figure 4E). We

further found that the microglia engulfed more VGLUT1 puncta

when the total volume of hM3Dq-expressing boutons was

higher in CNO-injected mice (Figure 4F). However, we found

no correlation between the engulfment of VGLUT1 puncta

and the total volume of hM3Dq-expressing boutons in saline-

injected mice (Figure 4F). These results suggest that CNO

injection induced activity-dependent synapse competition be-

tween hM3Dq-expressing and hM3Dq-non-expressing mossy

fiber boutons, resulting in the engulfment of weaker synapses,

i.e., hM3Dq-non-expressing synapses, by microglia. In addition,

we pharmacologically assessed the role of neuronal activity

in the pruning of mossy fiber synapses by using hippocampal

slice cultures (Figure S7) (Kasahara et al., 2016). We confirmed

that neuronal activity was necessary in the developmental elim-

ination of synapses by microglia and sufficient to induce the

elimination of synapses by microglia in slice cultures prepared

from MIA offspring (Figures S7C–S7E). Overall, these results

suggest that induced competition between the granule cells re-

sults in the engulfment of mossy fiber synapses by microglia

(Figure S8B).

DISCUSSION

The present study unveiled the role of microglia in the synapse

surplus and ASD-related behaviors observed in MIA offspring

and its modification by adult voluntary exercise. The modifica-

tion of dentate gyrus-hippocampal CA3 connections may affect

the functional connectivity of the hippocampus with other brain

regions, which is associated with autistic-like behaviors in mice

(Zhan et al., 2014). In addition, the increased density of functional

mossy fiber synapses in adult MIA offspring is consistent with

previous studies that demonstrated increased excitability and

spine density of hippocampal neurons in rodent models of

ASD (Tyzio et al., 2014; Jawaid et al., 2018).

Thoughwe have not assessed themolecular link thatmediates

neuronal activity and microglial activation in the present study,

possible candidates aremolecules involved in the classical com-

plement pathway such as C1q and C3. C1q and C3 have been

suggested to tag relatively weak or inactive synapses to be en-

gulfed bymicroglia in the retino-geniculate connections (Schafer

et al., 2012; Stevens et al., 2007). Thus, it is possible that comple-

ment molecules also contribute to exercise-induced pruning of

mossy fiber synapses, because the phenomenon likely depends

on synaptic competition. The competition between mossy fiber

synapses might be modulated by other molecules, such as

brain-derived neurotrophic factor (BDNF), which is enriched in

mossy fiber boutons (Koyama et al., 2004) and upregulated by

exercise (Voss et al., 2013). Furthermore, BDNF is released

from mossy fiber boutons in response to neuronal activity and

induces the maturation of synapses (Yoshii and Constantine-Pa-

ton, 2010). Therefore, it is possible that exercise-induced granule

cell activation promotes the release of BDNF from mossy fiber

boutons and strengthen some synapses, leading to the synapse

competition. However, changes in hippocampal BDNF levels

have not been reported in mouse models of MIA (Han et al.,

2016, 2017) except for a decrease in aged (22-month-old) MIA

mice (Giovanoli et al., 2015). In addition to these molecules,

newly generated granule cells in adulthood would be a key to

induce synaptic competition. Adult neurogenesis is enhanced

by voluntary exercise (Voss et al., 2013), and adult-born granule

cells exhibit higher excitability than existing mature granule cells

(Danielson et al., 2016). It has been revealed that MIA inhibits

neurogenesis in dentate gyrus during both postnatal days and

adulthood, leading to impaired maturation of newborn granule

cells (Zhang and van Praag, 2015). Thus, enhanced neurogene-

sis by exercise may also contribute to competition between syn-

apses derived from adult-born granule cells and existing mature

granule cells.

Our results revealed that exercise reversed the abnormal be-

haviors and synaptic surplus in adult MIA offspring. We deter-

mined that exercise activated a portion of neurons, resulting in

the engulfment of excess synapses by microglia, likely primed

by synaptic competition. However, whether the CA3 synaptic

surplus is the underlying cause of the behavioral abnormalities

in MIA offspring remains undetermined.

The abnormal behaviors in patients with neurodevelopmental

disorders such as ASDs typically appear by 8 to 10 months

of age and have been recognized to be irreversible after that

period. Reports have shown that autistic-like behaviors may be

ameliorated, but the effect was transient or necessitated genetic

modification (Yizhar et al., 2011; Mei et al., 2016). In contrast, our

findings propose clinically applicable methods to ameliorate

ASD symptoms. Further investigations are needed to better

understand the cellular and molecular mechanisms that are

involved in not only the development but also the reversible as-

pects of ASD symptoms.

STAR+METHODS

Detailed methods are provided in the online version of this paper

and include the following:

d KEY RESOURCES TABLE

d CONTACT FOR REAGENT AND RESOURCE SHARING

d EXPERIMENTAL MODEL AND SUBJECT DETAILS

B Animals

B Organotypic culture of hippocampal slices

d METHOD DETAILS

B Gestational exposure to poly(I:C)

B Running

B Sample preparation and immunohistochemistry

B Stereotaxic surgery for virus infection

B Three-chamber social interaction test

B Grooming test

B Novelty-suppressed feeding test

B Open field test

B Electrophysiology

d QUANTIFICATION AND STATISTICAL ANALYSIS

B Quantification

B Statistical Analysis

d DATA AND SOFTWARE AVAILABILITY

SUPPLEMENTAL INFORMATION

Supplemental Information can be found online at https://doi.org/10.1016/j.

celrep.2019.05.015.

Cell Reports 27, 2817–2825, June 4, 2019 2823

ACKNOWLEDGMENTS

We thank Dr. Crocker, Dr. Kiyama, and Dr. Konishi for providing us with anti-

Siglec-H antibodies and the method for immunohistochemistry with the anti-

body. We thank the members of JCAR (The Japanese Consortium for Autism

Research) for discussion. This research was supported in part by a Grant-in-

Aid for Scientific Research (C) (26460094 to R.K.) and by a Grant-in-Aid for

Scientific Research on Innovation Area ‘‘Glia Assembly’’ (26117504 and

16H01329 to R.K.) from JSPS and by the Brain Science Foundation, Japan

to R.K. and by ERATO (JPMJER1801 to Y.I.) from JST.

AUTHOR CONTRIBUTIONS

M.A. conducted and analyzed the experimental data and wrote the manu-

script. K.S., J.O., and K.M. conducted and analyzed the experiments. K.O.

and Y.M. helped with electrophysiological experiments. R.K. designed and

planned the project and wrote the manuscript. Y.I. discussed the results and

commented on the manuscript.

DECLARATION OF INTERESTS

The authors declare no competing interests.

Received: April 27, 2018

Revised: March 11, 2019

Accepted: May 1, 2019

Published: June 4, 2019

SUPPORTING CITATIONS

The following reference appears in the Supplemental Information: Wilke et al.

(2013).

REFERENCES

Aberg, M.A., Pedersen, N.L., Toren, K., Svartengren, M., Backstrand, B.,

Johnsson, T., Cooper-Kuhn, C.M., Aberg, N.D., Nilsson, M., and Kuhn, H.G.

(2009). Cardiovascular fitness is associated with cognition in young adulthood.

Proc. Natl. Acad. Sci. USA 106, 20906–20911.

Alexander, G.M., Rogan, S.C., Abbas, A.I., Armbruster, B.N., Pei, Y., Allen,

J.A., Nonneman, R.J., Hartmann, J., Moy, S.S., Nicolelis, M.A., et al. (2009).

Remote control of neuronal activity in transgenic mice expressing evolved G

protein-coupled receptors. Neuron 63, 27–39.

Anderson-Hanley, C., Tureck, K., and Schneiderman, R.L. (2011). Autism and

exergaming: effects on repetitive behaviors and cognition. Psychol. Res. Be-

hav. Manag. 4, 129–137.

Andoh, M., Koyama, R., and Ikegaya, Y. (2016). Severity of Kainic Acid-

Induced Seizures is not Aggravated in the Maternal Immune Activation Mouse

Model of Gestational Poly (I:C) Exposure. J Autism Epilepsy 1, 1003.

Clark, P.J., Bhattacharya, T.K., Miller, D.S., and Rhodes, J.S. (2011). Induction

of c-Fos, Zif268, and Arc from acute bouts of voluntary wheel running in new

and pre-existing adult mouse hippocampal granule neurons. Neuroscience

184, 16–27.

Danielson, N.B., Kaifosh, P., Zaremba, J.D., Lovett-Barron, M., Tsai, J., Denny,

C.A., Balough, E.M., Goldberg, A.R., Drew, L.J., Hen, R., et al. (2016). Distinct

Contribution of Adult-Born Hippocampal Granule Cells to Context Encoding.

Neuron 90, 101–112.

de Almeida, A.A., Gomes da Silva, S., Fernandes, J., Peixinho-Pena, L.F.,

Scorz, F.A., Cavalheiro, E.A., and Arida, R.M. (2013). Differential effects of

exercise intensities in hippocampal BDNF, inflammatory cytokines and cell

proliferation in rats during the postnatal brain development. Neurosci. Lett.

553, 1–6.

Fernandez de Cossıo, L., Guzman, A., van der Veldt, S., and Luheshi, G.N.

(2017). Prenatal infection leads to ASD-like behavior and altered synaptic

pruning in the mouse offspring. Brain Behav. Immun. 63, 88–98.

2824 Cell Reports 27, 2817–2825, June 4, 2019

Giovanoli, S., Engler, H., Engler, A., Richetto, J., Voget, M., Willi, R., Winter, C.,

Riva, M.A., Mortensen, P.B., Feldon, J., et al. (2013). Stress in puberty un-

masks latent neuropathological consequences of prenatal immune activation

in mice. Science 339, 1095–1099.

Giovanoli, S., Notter, T., Richetto, J., Labouesse, M.A., Vuillermot, S., Riva,

M.A., and Meyer, U. (2015). Late prenatal immune activation causes hippo-

campal deficits in the absence of persistent inflammation across aging.

J. Neuroinflammation 12, 221.

Gomes da Silva, S., Unsain, N., Masco, D.H., Toscano-Silva, M., de Amorim,

H.A., Silva Araujo, B.H., Simoes, P.S., Naffah-Mazzacoratti, Mda.G., Mortara,

R.A., Scorza, F.A., et al. (2012). Early exercise promotes positive hippocampal

plasticity and improves spatial memory in the adult life of rats. Hippocampus

22, 347–358.

Han, M., Zhang, J.C., Yao, W., Yang, C., Ishima, T., Ren, Q., Ma, M., Dong, C.,

Huang, X.F., and Hashimoto, K. (2016). Intake of 7,8-Dihydroxyflavone During

Juvenile and Adolescent Stages Prevents Onset of Psychosis in Adult

Offspring After Maternal Immune Activation. Sci. Rep. 6, 36087.

Han, M., Zhang, J.C., Huang, X.F., and Hashimoto, K. (2017). Intake of 7,8-di-

hydroxyflavone from pregnancy to weaning prevents cognitive deficits in adult

offspring after maternal immune activation. Eur. Arch. Psychiatry Clin. Neuro-

sci. 267, 479–483.

Haynes, S.E., Hollopeter, G., Yang, G., Kurpius, D., Dailey, M.E., Gan, W.B.,

and Julius, D. (2006). The P2Y12 receptor regulates microglial activation by

extracellular nucleotides. Nat. Neurosci. 9, 1512–1519.

Hitti, F.L., and Siegelbaum, S.A. (2014). The hippocampal CA2 region is essen-

tial for social memory. Nature 508, 88–92.

Hui, C.W., St-Pierre, A., El Hajj, H., Remy, Y., Hebert, S.S., Luheshi, G.N., Sri-

vastava, L.K., and Tremblay, M.E. (2018). Prenatal Immune Challenge in Mice

Leads to Partly Sex-Dependent Behavioral, Microglial, and Molecular Abnor-

malities Associated with Schizophrenia. Front. Mol. Neurosci. 11, 13.

Jawaid, S., Kidd, G.J., Wang, J., Swetlik, C., Dutta, R., and Trapp, B.D. (2018).

Alterations in CA1 hippocampal synapses in a mouse model of fragile X syn-

drome. Glia 66, 789–800.

Kasahara, Y., Koyama, R., and Ikegaya, Y. (2016). Depth and time-dependent

heterogeneity of microglia in mouse hippocampal slice cultures. Neurosci.

Res. 111, 64–69.

Knuesel, I., Chicha, L., Britschgi, M., Schobel, S.A., Bodmer, M., Hellings, J.A.,

Toovey, S., and Prinssen, E.P. (2014). Maternal immune activation and

abnormal brain development across CNS disorders. Nat. Rev. Neurol. 10,

643–660.

Kobayashi, K., Imagama, S., Ohgomori, T., Hirano, K., Uchimura, K., Saka-

moto, K., Hirakawa, A., Takeuchi, H., Suzumura, A., Ishiguro, N., and Kado-

matsu, K. (2013). Minocycline selectively inhibits M1 polarization of microglia.

Cell Death Dis. 4, e525.

Konishi, H., Kobayashi, M., Kunisawa, T., Imai, K., Sayo, A., Malissen, B.,

Crocker, P.R., Sato, K., and Kiyama, H. (2017). Siglec-H is a microglia-specific

marker that discriminates microglia from CNS-associated macrophages and

CNS-infiltrating monocytes. Glia 65, 1927–1943.

Koyama, R., and Ikegaya, Y. (2015). Microglia in the pathogenesis of autism

spectrum disorders. Neurosci. Res. 100, 1–5.

Koyama, R., Yamada, M.K., Fujisawa, S., Katoh-Semba, R., Matsuki, N., and

Ikegaya, Y. (2004). Brain-derived neurotrophic factor induces hyperexcitable

reentrant circuits in the dentate gyrus. J. Neurosci. 24, 7215–7224.

Malkova, N.V., Yu, C.Z., Hsiao, E.Y., Moore, M.J., and Patterson, P.H. (2012).

Maternal immune activation yields offspring displaying mouse versions of the

three core symptoms of autism. Brain Behav. Immun. 26, 607–616.

Mattei, D., Ivanov, A., Ferrai, C., Jordan, P., Guneykaya, D., Buonfiglioli, A.,

Schaafsma, W., Przanowski, P., Deuther-Conrad, W., Brust, P., et al. (2017).

Maternal immune activation results in complexmicroglial transcriptome signa-

ture in the adult offspring that is reversed by minocycline treatment. Transl.

Psychiatry 7, e1120.

Mei, Y., Monteiro, P., Zhou, Y., Kim, J.A., Gao, X., Fu, Z., and Feng, G. (2016).

Adult restoration of Shank3 expression rescues selective autistic-like pheno-

types. Nature 530, 481–484.

Mellor, J., and Nicoll, R.A. (2001). Hippocampal mossy fiber LTP is indepen-

dent of postsynaptic calcium. Nat. Neurosci. 4, 125–126.

Mildner, A., Huang, H., Radke, J., Stenzel, W., and Priller, J. (2017). P2Y12

receptor is expressed on human microglia under physiological conditions

throughout development and is sensitive to neuroinflammatory diseases.

Glia 65, 375–387.

Naviaux, R.K., Zolkipli, Z., Wang, L., Nakayama, T., Naviaux, J.C., Le, T.P.,

Schuchbauer, M.A., Rogac, M., Tang, Q., Dugan, L.L., and Powell, S.B.

(2013). Antipurinergic therapy corrects the autism-like features in the poly(IC)

mouse model. PLoS ONE 8, e57380.

Okuyama, T., Kitamura, T., Roy, D.S., Itohara, S., and Tonegawa, S. (2016).

Ventral CA1 neurons store social memory. Science 353, 1536–1541.

Pan, C.Y. (2010). Effects of water exercise swimming program on aquatic skills

and social behaviors in children with autism spectrum disorders. Autism 14,

9–28.

Paolicelli, R.C., Bolasco, G., Pagani, F., Maggi, L., Scianni, M., Panzanelli, P.,

Giustetto, M., Ferreira, T.A., Guiducci, E., Dumas, L., et al. (2011). Synaptic

pruning by microglia is necessary for normal brain development. Science

333, 1456–1458.

Pereira, A.C., Huddleston, D.E., Brickman, A.M., Sosunov, A.A., Hen, R.,

McKhann, G.M., Sloan, R., Gage, F.H., Brown, T.R., and Small, S.A. (2007).

An in vivo correlate of exercise-induced neurogenesis in the adult dentate

gyrus. Proc. Natl. Acad. Sci. USA 104, 5638–5643.

Prinz, M., and Priller, J. (2010). Tickets to the brain: role of CCR2 and CX3CR1

in myeloid cell entry in the CNS. J. Neuroimmunol. 224, 80–84.

Rollenhagen, A., Satzler, K., Rodrıguez, E.P., Jonas, P., Frotscher, M., and

L€ubke, J.H. (2007). Structural determinants of transmission at large hippocam-

pal mossy fiber synapses. J. Neurosci. 27, 10434–10444.

Rubenstein, J.L., and Merzenich, M.M. (2003). Model of autism: increased

ratio of excitation/inhibition in key neural systems. Genes Brain Behav. 2,

255–267.

Sasaki, Y., Hoshi, M., Akazawa, C., Nakamura, Y., Tsuzuki, H., Inoue, K., and

Kohsaka, S. (2003). Selective expression of Gi/o-coupled ATP receptor P2Y12

in microglia in rat brain. Glia 44, 242–250.

Schafer, D.P., Lehrman, E.K., Kautzman, A.G., Koyama, R., Mardinly, A.R., Ya-

masaki, R., Ransohoff, R.M., Greenberg, M.E., Barres, B.A., and Stevens, B.

(2012). Microglia sculpt postnatal neural circuits in an activity and comple-

ment-dependent manner. Neuron 74, 691–705.

Stevens, B., Allen, N.J., Vazquez, L.E., Howell, G.R., Christopherson, K.S.,

Nouri, N., Micheva, K.D., Mehalow, A.K., Huberman, A.D., Stafford, B., et al.

(2007). The classical complement cascade mediates CNS synapse elimina-

tion. Cell 131, 1164–1178.

Tang, G., Gudsnuk, K., Kuo, S.H., Cotrina, M.L., Rosoklija, G., Sosunov, A.,

Sonders, M.S., Kanter, E., Castagna, C., Yamamoto, A., et al. (2014). Loss

of mTOR-dependent macroautophagy causes autistic-like synaptic pruning

deficits. Neuron 83, 1131–1143.

Tao, K., Ichikawa, J., Matsuki, N., Ikegaya, Y., and Koyama, R. (2016). Exper-

imental febrile seizures induce age-dependent structural plasticity and

improve memory in mice. Neuroscience 318, 34–44.

Tyzio, R., Nardou, R., Ferrari, D.C., Tsintsadze, T., Shahrokhi, A., Eftekhari, S.,

Khalilov, I., Tsintsadze, V., Brouchoud, C., Chazal, G., et al. (2014). Oxytocin-

mediated GABA inhibition during delivery attenuates autism pathogenesis in

rodent offspring. Science 343, 675–679.

Ulmer-Yaniv, A., Djalovski, A., Yirmiya, K., Halevi, G., Zagoory-Sharon, O., and

Feldman, R. (2018). Maternal immune and affiliative biomarkers and sensitive

parentingmediate the effects of chronic early trauma on child anxiety. Psychol.

Med. 48, 1020–1033.

Voss, M.W., Vivar, C., Kramer, A.F., and van Praag, H. (2013). Bridging animal

and human models of exercise-induced brain plasticity. Trends Cogn. Sci. 17,

525–544.

Wilke, S.A., Antonios, J.K., Bushong, E.A., Badkoobehi, A., Malek, E., Hwang,

M., Terada, M., Ellisman, M.H., and Ghosh, A. (2013). Deconstructing

complexity: serial block-face electron microscopic analysis of the hippocam-

pal mossy fiber synapse. J. Neurosci. 33, 507–522.

Williams, M.E., Wilke, S.A., Daggett, A., Davis, E., Otto, S., Ravi, D., Ripley, B.,

Bushong, E.A., Ellisman, M.H., Klein, G., and Ghosh, A. (2011). Cadherin-9

regulates synapse-specific differentiation in the developing hippocampus.

Neuron 71, 640–655.

Yizhar, O., Fenno, L.E., Prigge, M., Schneider, F., Davidson, T.J., O’Shea, D.J.,

Sohal, V.S., Goshen, I., Finkelstein, J., Paz, J.T., et al. (2011). Neocortical exci-

tation/inhibition balance in information processing and social dysfunction. Na-

ture 477, 171–178.

Yoshii, A., and Constantine-Paton, M. (2010). Postsynaptic BDNF-TrkB

signaling in synapse maturation, plasticity, and disease. Dev. Neurobiol. 70,

304–322.

Zhan, Y., Paolicelli, R.C., Sforazzini, F., Weinhard, L., Bolasco, G., Pagani, F.,

Vyssotski, A.L., Bifone, A., Gozzi, A., Ragozzino, D., and Gross, C.T. (2014).

Deficient neuron-microglia signaling results in impaired functional brain con-

nectivity and social behavior. Nat. Neurosci. 17, 400–406.

Zhang, Z., and van Praag, H. (2015). Maternal immune activation differentially

impacts mature and adult-born hippocampal neurons in male mice. Brain Be-

hav. Immun. 45, 60–70.

Zhang, J., Raper, A., Sugita, N., Hingorani, R., Salio, M., Palmowski, M.J., Cer-

undolo, V., and Crocker, P.R. (2006). Characterization of Siglec-H as a novel

endocytic receptor expressed on murine plasmacytoid dendritic cell precur-

sors. Blood 107, 3600–3608.

Zhu, F., Zheng, Y., Liu, Y., Zhang, X., and Zhao, J. (2014). Minocycline allevi-

ates behavioral deficits and inhibits microglial activation in the offspring of

pregnant mice after administration of polyriboinosinic-polyribocytidilic acid.

Psychiatry Res. 219, 680–686.

Cell Reports 27, 2817–2825, June 4, 2019 2825

STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER

Antibodies

rabbit anti-Synaptoporin - 1:1000 Synaptic Systems Cat# 102 002; RRID:AB_261974

mouse anti-PSD95 – 1:500 Thermo Fisher Scientific Cat# MA1-046; RRID:AB_2092361

guinea pig anti-VGLUT1 – 1:1000 Synaptic Systems Cat# 135304; RRID:AB_887878

mouse anti-VGAT – 1:1000 Synaptic Systems Cat# 131 011; RRID:AB_887872

rabbit anti-Gephyrin – 1: 1000 Synaptic Systems Cat# 147 002; RRID:AB_2619838

rabbit anti-Iba1 – 1:400 Wako Cat# 019-19741; RRID:AB_839504

rat anti-CD68 – 1:200 Bio-Rad Cat# MCA1957GA; RRID:AB_324217

chicken anti-GFP – 1:1000 Abcam Cat# ab13970; RRID:AB_300798

mouse anti-mCherry – 1:1000 Abcam Cat# ab125096; RRID:AB_11133266

guinea pig anti-Homer1 – 1:500 Synaptic Systems Cat# 160 004; RRID:AB_10549720

rabbit anti-DsRed – 1:1000 Takara Bio Cat# 632496; RRID:AB_10013483

rabbit anti-CCR2 – 1:100 Abcam Cat# ab203128

rabbit anti-P2Y12 – 1:500 AnaSpec Cat# AS-55043A; RRID:AB_2298886

mouse anti-IL-1b – 1:100 Cell Signaling Technology Cat# 12242; RRID:AB_2715503

rabbit anti-c-Fos – 1:1000 Santa Cruz Biotechnology Cat# sc-52-G; RRID:AB_2106783

goat anti-c-Fos – 1:100 Santa Cruz Biotechnology Cat# sc-52; RRID:AB_2629503

mouse anti-NeuN – 1:1000 Millipore Cat# MAB377; RRID:AB_2298772

sheep anti-Siglec-H – 1:1000 Dr. Konishi N/A

secondary antibodies: conjugated to AlexaFluor dyes – 1:500 Thermo Fisher Scientific N/A

NeuroTrace 435/455 Blue Fluorescent Nissl Stain – 1:200 Thermo Fisher Scientific N21479

CY3 – 1:1000 Perkin Elmer Cat# NEL704A001KT; RRID:AB_2572409

Bacterial and Virus Strains

AAV8-CaMKIIa-eGFP Addgene 50469

AAV8-CaMKIIa-hM3D(Gq)-mCherry Addgene 50476

Chemicals, Peptides, and Recombinant Proteins

polyinosinic:polycytidylic acid Sigma-Aldrich P9582

minocycline Sigma-Aldrich M9511

clozapine-N-oxide Abcam ab141704

tetrodotoxin Tocris Bioscience 04330-31

picrotoxin Sigma-Aldrich P1675

DCG-IV Tocris Bioscience 0975

Experimental Models: Organisms/Strains

B6.129P2(Cg)-Cx3cr1tm1Litt/J The Jackson Laboratory Stock No. 005582

Thy1-mGFP Dr. Pico Caroni N/A

Software and Algorithms

ImageJ N/A

MATLAB N/A

CONTACT FOR REAGENT AND RESOURCE SHARING

Requests for resources, reagents, or questions about methods should be directed to and will be fulfilled by the Lead Contact, Ryuta

Koyama ([email protected]).

e1 Cell Reports 27, 2817–2825.e1–e5, June 4, 2019

EXPERIMENTAL MODEL AND SUBJECT DETAILS

AnimalsC57BL/6J mice (SLC, Shizuoka, Japan) and mice from the reporter line Thy1-mGFP (Lsi1; a generous gift from Dr. Pico Caroni) and

CX3CR1-GFP (Stock No: 005582, The Jackson Laboratory) were maintained under conditions of controlled temperature and a light

schedule and provided with unlimited food and water. Pregnant mice were used from E12.5 to make the maternal immune activation

(MIA) model (please see Gestational exposure to poly(I:C) in Method Details). All the born pups were used and sacrificed by P67. All

experimental procedures conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and to the

guidelines provided by the University of Tokyo (approval number: P29-15). Unless otherwise noted, males were used.

Organotypic culture of hippocampal slicesHippocampal slice cultures were prepared from P10 mice (males and females) as previously described with minor modifications

(Kasahara et al., 2016). Briefly, the posterior part of the mouse brain was cut into 400-mm-thick transverse slices using a DTK-

1500 vibratome (Dosaka, Kyoto, Japan) in aerated, ice-cold GBSS (consisting of the following (g/l): 7.00 NaCl, 0.212 MgCl2･6H2O, 2.27 NaHCO3, 0.14 MgSO4･7H2O, 0.286 Na2HPO4･12H2O, 0.37 KCl, 0.22 CaCl2･2H2O, 0.33 KH2PO4) containing 25 mM

glucose. The slices were incubated for 30-90 minutes at 4�C with cold incubation medium containing minimum essential medium

(MEM) and Hanks’ balanced salt solution (HBSS) (consisting of the following (g/l): 8.00 NaCl, 0.10 MgCl2･6H2O, 0.35 NaHCO3,

0.10 MgSO4･7H2O, 0.12 Na2HPO4･12H2O, 0.40 KCl, 0.185 CaCl2･2H2O, 0.06 KH2PO4) at a ratio of 2:1, 10 mM Tris and 25 mM

HEPES. The slices were placed on Omnipore membrane filters (JHWP02500; Millipore) in a solution containing 50% MEM, 25%

horse serum (Cell Culture Lab, Cleveland, OH, USA) and 25%HBSS, 10 mM Tris, 25 mMHEPES and 5 mMNaHCO3, supplemented

with 33mMglucose. The slices were then incubated at 37�C in a humidified incubator with 5%CO2 and 95% air. Hippocampal slices

were cultured with medium containing 1 mM tetrodotoxin (TTX; Tocris Bioscience, Bristol, UK) or 50 mMpicrotoxin (PIC; Sigma) from

5 days in vitro (DIV) to 10 DIV.

METHOD DETAILS

Gestational exposure to poly(I:C)Thematernal immune activation (MIA) model wasmade by injecting polyinosinic:polycytidylic acid (poly(I:C)) (Potassium salt; Sigma)

as previously described (Malkova et al., 2012; Naviaux et al., 2013). Pregnant dams received intraperitoneal (i.p.) injection of poly(I:C)

at two doses (3 mg/kg on E12.5 and 1.5 mg/kg on E17.5). The same volume of saline was injected into pregnant dams at the same

time to prepare the control mice. Poly(I:C)-injected offspring were used as MIA offspring and weaned at P21. All the born pups were

alive and normally developed, though some pregnant mice had abortion (the ratio was 11% for Saline group and 33% for MIA group).

No mice were housed alone. Male mice were used for the experiments.

RunningMice in the exercise group were housed in a cage (3–4 mice/cage) with a freely accessible running wheel (130 mm in diameter). The

running wheel was placed in the corner of the cage for 30 days from P30. Minocycline (Sigma, 30 mg/kg) was intraperitoneally

injected once a day beginning on the first day that the running wheel was placed in the cage. The same volume of saline was injected

as a control.

Sample preparation and immunohistochemistryExperimental mice were deeply anesthetized with isoflurane and perfused transcardially with cold phosphate-buffered saline (PBS)

followed by 4% paraformaldehyde (PFA). The brain samples were postfixed with 4% PFA for 2–4 h on ice and subsequently

immersed in 20% and 30% sucrose in PBS for 24 h and 48 h, respectively, at 4�C. Coronal hippocampal sections (40 mm thick)

were preparedwith a cryostat (HM520; Thermo Fisher Scientific,Waltham,MA, USA) at�24�C. For slice cultures, sampleswere fixed

overnight in 4% PFA at 4�C. Floating sections were used for PSD95 staining.

The fixed sampleswere rinsed three timeswith 0.1Mphosphate buffer (PB). Slices were then permeabilized and blocked for 30min

at room temperature in 0.1 M PB with 0.3% Triton X-100 and 10% goat serum. The samples were subsequently incubated with

primary antibodies in 0.1 M PB with 0.3% Triton X-100 and 10% goat serum overnight at 4�C. After the samples were rinsed three

times with 0.1 M PB, they were incubated with secondary antibodies in 0.1 M PBwith 0.3% Triton X-100 and 10% goat serum for 4 h

at room temperature. For slice cultures, the samples were incubated with primary antibodies 2 overnights at 4�C under agitation and

secondary antibodies and NeuroTrace (435/455 blue fluorescent Nissl stain, 1:200; Thermo Scientific, MA, USA) 1 overnight at 4�Cunder agitation. Finally, the samples were rinsed three times with 0.1 M PB and embedded in VECTASHIELD with DAPI (Vector Lab-

oratories, Burlingame, CA, USA) or MOUNTANT (Thermo Scientific). The primary antibodies used in this study were as follows: rabbit

anti-SPO (1:1000; Synaptic Systems, Gottingen, Germany), mouse anti-PSD95 (1:500; Thermo Scientific), guinea pig anti-VGLUT1

(1:1000; Synaptic Systems), mouse anti-VGAT (1:1000; Synaptic Systems), rabbit anti-gephyrin (1:1000; Synaptic Systems), rabbit

anti-Iba1 (1:400;Wako, Osaka, Japan), rat anti-CD68 (1:200; Bio-Rad, CA, USA), chicken anti-GFP (1:1000; Abcam, Cambridge, UK),

mouse anti-mCherry (1:1000; Abcam), guinea pig anti-Homer1 (1:500, Synaptic Systems), rabbit anti-DsRed (1:1000, Takara Bio,

Cell Reports 27, 2817–2825.e1–e5, June 4, 2019 e2

Shiga, Japan), rabbit anti-CCR2 (1:100, Abcam), rabbit anti-P2Y12 (1:500, AnaSpec, CA, USA) andmouse IL-1b (1:100, Cell Signaling

Technology, MA, USA). Secondary antibodies conjugated to Alexa fluor dyes (1:500; Invitrogen, MD, USA) were used.

c-Fos immunostaining was performed on free-floating sections as described below. The samples were incubated for 1 h in 0.2%

Triton X-100 in PBS containing 5% goat serum and with rabbit anti-c-Fos (1:1000; Santa Cruz Biotechnology, CA, USA) and mouse

anti-NeuN (1:1000; Millipore, Bedford, MA, USA) overnight at 4�C. The sections were then incubated with the goat anti-rabbit

biotinylated secondary antibody (1:500; Vector Laboratories) and Alexa fluor dye-conjugated secondary antibodies (1:500; Invitro-

gen) for 4 h at room temperature, with the avidin-biotin complex (1:100; Vector Laboratories) for 1.5 h at room temperature, and

with CY3 (1: 1000; PerkinElmer, MA, USA) for 1 h at room temperature. After three rinses, the samples were embedded. Or the

samples were incubated for 1 h in 0.3% Triton X-100 in PBS containing 5% bovine serum albumin and with goat anti-c-Fos

(1:100; Santa Cruz Biotechnology, CA, USA) and mouse anti-NeuN (1:1000; Millipore, MA, USA) overnight at 4�C. The sections

were then incubated with Alexa fluor dye-conjugated secondary antibodies (1:500; Invitrogen) for 1.5 h at room temperature. After

three rinses, the samples were embedded.

For Siglec-H immunostaining, CX3CR1GFP/+ mice were perfused with cold phosphate-buffered saline (PBS) followed by 2%

paraformaldehyde (PFA). The brain samples were immersed in 30% sucrose in PBS for 24 h at 4�C. Coronal hippocampal

sections (30 mm thick) were prepared with a cryostat and mounted on glass slides. The samples were incubated for 1 h in 0.3%

Triton X-100 in PBS containing 5% bovine serum albumin and with goat anti-GFP, rabbit anti-Iba1 and sheep anti-Siglec-H

(1:1000, provided by Dr. Konishi, assistant professor at Nagoya University) overnight at 4�C. The sections were then incubated

with Alexa fluor dye-conjugated secondary antibodies (1:1000; Invitrogen) for 1 h at room temperature. After three rinses, the samples

were encapsulated.

Stereotaxic surgery for virus infectionMIA offspring were used at P10. Mice were anesthetized using pentobarbital (25 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.) and placed

in a stereotaxic apparatus. The virus (0.5 mL per side) was injected into the bilateral dentate gyrus (AP: �1.9 mm, LM: ± 0.9 mm,

DV: �1.8 mm) at a rate of 0.25 mL/min using glass pipettes. For the expression of DREADD, we bilaterally injected AAV8-

CaMKIIa-eGFP or AAV8-CaMKIIa-hM3Dq-mCherry (purchased from Addgene) using glass pipettes, which were then left in

place for a few minutes before they were slowly removed. Then, the mice were intraperitoneally injected with the selective ligand

clozapine-N-oxide (CNO; Abcam, 5 mg/ kg) or saline at P30 and sacrificed at P31 for further analysis.

Three-chamber social interaction testA three-chamber test was carried out to investigate the social interaction of mice. The test was performed in a white three-

chamber box (25 3 51 3 25 cm). Age- and gender-matched C57BL/6J mice that had never been exposed to the test mice

were used as stranger mice and individually placed in a chamber. In the other side of the chamber, a cage-mate mouse was

used as a familiar mouse. Both the stranger and familiar mice were placed in wire cups to allow the test mice to freely access

and sniff the mice in the cups. The three-chamber box and cups were cleaned with 70% ethanol and wiped with paper towels

between each trial. In the first 10-min session, a test mouse was placed in the three-chamber box, where two empty cups were

located in both sides of the chamber, and allowed to freely explore to habituate the test mouse. In the second 10-min session, a

stranger mouse and a familiar mouse were placed in the cage on each side. Then, the test mouse was placed in the center

chamber of the three-chamber box and allowed to freely explore the box 10 min. The brightness of the room was 10 lux

throughout the test. The movement of the mice was recorded by a USB webcam and PC-based video capture software. The

recorded video file was further analyzed by ImageJ, and the time spent in each chamber was measured. A preference index

to stranger mice was calculated to assess the social interaction using the following formula: 100 3 time in stranger chamber/

total time in stranger and familiar chamber.

Grooming testA grooming test was performed the day after the three-chamber test to investigate repetitive behaviors. A test mouse was placed in a

normal cage without wood bedding for 20 minutes, in which the first 10 minutes was the habituation session and the last 10 minutes

was the test session. The movement of the test mouse was recorded by a USB webcam and PC-based video capture software. The

recorded video file was further analyzed to manually measure the grooming time during the test session.

Novelty-suppressed feeding testThe novelty-suppressed feeding test was carried out for 5 min in a white polystyrene box (403 403 30 cm) whose floor was covered

with wood bedding. Twenty-four hours before testing, all food was removed from the home cage, but water was freely accessible

during this period. During the test period, a single pellet was placed on a white paper platform located in the center of the test

box at which the light intensity was approximately 1000 lux. The test mouse was placed in a corner of the box, and the latency

from the starting time until the time when the mouse bit the food pellet in the center of the field was measured to examine the level

e3 Cell Reports 27, 2817–2825.e1–e5, June 4, 2019

of anxiety. Immediately after this test, the mice were transferred to their home cage, and the latency to feed in their home cage was

measured to examine their basal appetite.

Open field testThe open field test was conducted using a large, square, white polystyrene box (40 3 40 3 30 cm) with an open top and a floor

covered with clear acrylic sheeting and lasted for 10 min. The arena of the open field included the center zone (27 3 27 cm). The

movement of the test mouse was recorded by a USB webcam and PC-based video capture software, and the recorded video file

was analyzed using ImageJ to measure the total distance traveled.

ElectrophysiologyP30–34 (before exercise) or P60-67 (after exercise) mice were deeply anesthetized with isoflurane, and the brains were quickly

removed and placed in ice-cold, oxygenated (95% O2/5% CO2) artificial cerebrospinal fluid (ACSF) containing 127 mM NaCl,

1.6 mM KCl, 1.24 mM KH2PO4, 1.3 mM MgSO4, 2.4 mM CaCl2, 26 mM NaHCO3, and 10 mM glucose. Transverse slices containing

the CA3 area were cut at a thickness of 400 mm with a vibratome in ice-cold, oxygenated modified ACSF containing 222.1 mM

sucrose, 27 mM NaHCO3, 1.4 mM NaH2PO4, 2.5 mM KCl, 0.5 mM ascorbic acid, 1 mM CaCl2, and 7 mM MgSO4. Slices were

maintained for 30 min at 37�C and then incubated for at least 30 min at room temperature before use.

Slices were transferred to a recording chamber and superfused with oxygenated ACSF containing 0.1 mM picrotoxin (30–33�C,1–3 mL/min). Whole-cell recordings were made from visually identified, pyramidal neurons located in the CA3 region using infrared

differential interference contrast (IR/DIC) techniques. Patch pipettes (3-6MU) were fabricated from borosilicate glass and filled with a

solution containing 127 mM CsMeSO4, 8 mM CsCl, 10 mM HEPES, 1 mM MgCl2, 10 mM phosphocreatine-Na2, 4 mM MgATP,

0.3 mM NaGTP, and 0.2 mM EGTA (pH 7.2–7.3, 280–295 mOsm). mEPSCs were recorded at a holding potential of �70 mV in the

presence of tetrodotoxin (1 mM). mEPSCs were detected using an in-house MATLAB program and were defined as inward currents

with amplitudes > 7 pA. DCG-IV (Tocris Bioscience, 10 mM for slices from P30-34 mice and 1 mM for slices from P60-67 mice) was

applied to block synaptic transmission from mossy fiber synapses. The series resistance was monitored, and if it exceeded 40 MU,

the data were discarded. Data were sampled at 20 kHz and filtered at 2 kHz using an Axopatch 200B, 700B amplifier (Molecular De-

vices), DIGIDATA 1320A, 1440A (Molecular Devices), and pClamp 10.2 (Molecular Devices).

QUANTIFICATION AND STATISTICAL ANALYSIS

QuantificationThe immunostained samples were analyzed with a TCS SP8 (Leica Microsystems, Wetzlar, Germany) or FV1200 (Olympus, Tokyo,

Japan) confocal system under 10 3, 20 3, 60 3 and 100 3 objectives. Z series images were collected with 0.33-mm steps, and 4 Z

sections (1 mm thick) were stacked using ImageJ (NIH) to quantify the synapse density. 11 Z series sections (5 mm thick) for the quan-

tification of synapse engulfment or 21 Z series sections (10 mm thick) for CD68 expression were collected at 0.50 mmsteps. 25 Z series

sections (8 mm thick) for the quantification of post-synaptic volume in mossy fiber bouton were collected at 0.33 mm steps. ImageJ

was used to quantify the synapse density. After the stacked images in the stratum lucidum were cropped, pre- and postsynaptic

signals were detected by determining the threshold of the fluorescent intensities. Then, the number of colocalized signals of

pre- and postsynaptic puncta were counted.

The quantification of synapse engulfment and CD68 expression was carried out as previously described with minor modifications

(Schafer et al., 2012). Microglial, synaptic, and CD68 immunofluorescent signals were detected by determining the threshold of the

fluorescent intensities in ImageJ. Then, colocalized images of microglia and synapses or CD68 were prepared, and the volume of

microglia and the colocalization were measured in the stratum lucidum. To determine the synapse engulfment by microglia, the

following calculation was used: Volume of internalized synaptic puncta (mm3)/ Volume of microglia (mm3). To determine the % of

CD68 expression, the following calculation was used: Volume of internalized CD68 (mm3)/ Volume of microglia (mm3).

For assays of microglia-dependent synaptic pruning in hippocampal slices, microglial, synaptic, and CD68 immunofluorescent

signals were detected by determining the threshold of the fluorescent intensities in ImageJ. Then, colocalized images of microglia,

synapses and CD68 were prepared, and the volume of microglia and the colocalization were measured in the stratum lucidum.

To determine the synaptic engulfment by microglia, the following calculation was used: Volume of internalized synaptic puncta

(mm3)/ Volume of microglia (mm3).

The quantification of post-synaptic volume in mossy fiber bouton of Thy-1 mGFP mice was carried out as follows; GFP-labeled

bouton and postsynaptic immunofluorescent signals were detected by determining the threshold of the fluorescent intensities in

ImageJ. Then, colocalized images of boutons and synapses were prepared, and the area of boutons and the colocalization were

measured in the stratum lucidum. All analyses were carried out in a blinded manner.

Statistical AnalysisThe data are represented as the mean ± standard error of the mean (SEM) or as box and whisker plots showing the distribution and

median (solid line) of the data (box edges indicate 25th and 75th percentiles; whiskers, min and max) and were pooled from at least 3

Cell Reports 27, 2817–2825.e1–e5, June 4, 2019 e4

independent experiments. The data were collected and statistically analyzed independently by 2 researchers in a blinded manner.

Student’s t test or Dunnet’s test, or one-way or two-way analysis of variance (ANOVA) followed by the Tukey test was performed for

statistical analysis unless otherwise described. The Gehan-Breslow test (Figure 1C) was also used.

DATA AND SOFTWARE AVAILABILITY

Reasonable requests for data will be fulfilled by the Lead Contact.

e5 Cell Reports 27, 2817–2825.e1–e5, June 4, 2019