Full Text

Turn on search term navigation

ARTICLE

Received 25 Apr 2014 | Accepted 3 Jun 2014 | Published 11 Jul 2014

Giulia E. Tyzack1, Sergey Sitnikov2,*, Daniel Barson1,*, Kerala L. Adams-Carr1, Nike K. Lau1, Jessica C. Kwok1, Chao Zhao1,2, Robin J.M. Franklin1,2, Ragnhildur T. Karadottir1,2, James W. Fawcett1 & Andrs Lakatos1

The role of remote astrocyte (AC) reaction to central or peripheral axonal insult is not clearly understood. Here we use a transgenic approach to compare the direct inuence of normal with diminished AC reactivity on neuronal integrity and synapse recovery following extra-cranial facial nerve transection in mice. Our model allows straightforward interpretations of ACneuron signalling by reducing confounding effects imposed by inammatory cells. We show direct evidence that perineuronal reactive ACs play a major role in maintaining neuronal circuitry following distant axotomy. We reveal a novel function of astrocytic signal transducer and activator of transcription-3 (STAT3). STAT3 regulates perineuronal astrocytic process formation and re-expression of a synaptogenic molecule, thrombospondin-1 (TSP-1), apart from supporting neuronal integrity. We demonstrate that, through this new pathway, TSP-1 is responsible for the remote AC-mediated recovery of excitatory synapses onto axotomized motor neurons in adult mice. These data provide new targets for neuroprotective therapies via optimizing AC-driven plasticity.

DOI: 10.1038/ncomms5294 OPEN

Astrocyte response to motor neuron injury promotes structural synaptic plasticity via STAT3-regulated TSP-1 expression

1 John van Geest Centre for Brain Repair, Department of Clinical Neurosciences, University of Cambridge, E.D. Adrian Building, Forvie Site, Robinson Way, Cambridge CB2 0PY, UK. 2 Department of Veterinary Medicine & Wellcome Trust-MRC Stem Cell Institute, University of Cambridge, Madingley Road, Cambridge CB3 0ES, UK. * These authors contributed equally to this work. Correspondence and requests for materials should be addressed to A.L. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294

Plasticity of synaptic input around cell bodies of neurons occurs rapidly following injury to their central or peripheral axon projections. Proximal to the lesion, this adaptive

remodelling mechanism may both switch functioning neurons into the circuitry1,2 and enhance neuronal viability3. A critical but unresolved issue is the role of perineuronal AC reactivity in this process. Could remotely activated grey matter glial cells orchestrate dynamic structural changes in neuronal connections? Synapse elimination and rearrangement around the neuronal soma and dendrites occur concomitantly with glial cell activation46. Similar to microglia, perineuronal ACs are known to respond to distant insults by extending hypertrophic processes around synapses. However, the extent to which this type of reactive transformation represents a detrimental or a protective response for neuronal function and integrity in the adult central nervous system (CNS) has never been directly addressed. This is highly pertinent, as targeting reactive AC neuronal interactions proximal to damage is a plausible alternative approach for facilitating neuronal survival and repair. This is due to the relative lack of negative inuence from other invading cell types6,7. Thus, understanding the precise mechanisms and consequences of remote AC activation may hold the key to optimizing the efciency of functional recovery.

Most reports on the astrocytic behaviour have derived from studies on ACs residing in the vicinity of a lesion, but often led to conicting views. Although reactive ACs have been shown to express a number of molecules inhibitory for plasticity8,9, recent studies argue that they can also be benecial for repair. Within traumatic spinal cord lesions, signal transducer and activator of transcription-3 (STAT3) signalling was found to regulate AC activation, which was also shown to be necessary for wound healing and functional recovery1012. The exact mechanism underlying synapse recovery in the vicinity of these lesions is unknown. AC-derived thrombospondin-1/2 (TSP-1/2) attracted interest as they have been found to increase excitatory synapse density in the developing CNS13,14. Recent studies using experimental ischaemic CNS damage reported that thrombospondin-1/2 can be re-expressed locally at the lesion site15,16, but their regulation in reactive ACs also remains unresolved.

The complexity of these lesions may cloud direct interpretations of default AC behaviour, and the above ndings may not be relevant to situations when ACs are remotely activated in a less inammatory environment1719. The inux of other cells around the lesion and the release of cytokines that may reach distant areas by diffusion and via the cerebrospinal uid can inuence both AC response and synaptic recovery.

We directly addressed the effect of remote AC reactivity on neuronal integrity and recovery of their synaptic input while reducing potential confounding factors. We used unilateral extracranial facial nerve transection where the AC response is distant and predominantly triggered by neuronal insults in a less inammatory environment. By selectively impeding AC activation in an established transgenic system12, we could reliably examine its effect on neuronal function. We show that remotely activated grey matter ACs directly promote structural synaptic plasticity and support network integrity. We also provide the underlying mechanism, showing that STAT3 activation upregulates astrocytic TSP-1 re-expression and release, which is required to facilitate the recovery of synaptic input onto surviving motor neurons after their distant axonal insult.

ResultsSTAT3 induces AC process formation after axotomy. To explore whether STAT3 signalling plays a key role in the AC response to remote axonal injury, we used glial brillary acidic

protein (GFAP)-Cre/STAT3-loxP conditional knockout (CKO) mice in comparison with wild-type (WT) controls. Astrocytic STAT3 activation was veried by nuclear translocation of immunoreactivity (IR) in the facial nuclei (FN) 14 days after facial axotomy (Fig. 1af). To identify all ACs, tissue sections were immunostained for ALDH1L1, a pan-astrocytic marker that is less dependent on STAT3 activity20 (Fig. 1ad). In WT mice (n 3) the proportion of ACs with STAT3 activation was

signicantly higher at 49.445.26% compared with the contralateral (CL) side (4.60.54%) or for unlesioned controls(7.511.9%; n 3, Po0.001; Fig. 1a,c,g). In CKO mice (n 5),

STAT3 activation was effectively abolished in ACs (Fig. 1b,d,g).

We then addressed whether STAT3 activation is associated with characteristic features of reactive AC transformation, such as GFAP expression. First, we examined whether the STAT3-activated population overlaps with GFAP-expressing ACs (Fig. 1e,f). In WT mice, 43.775.84% of GFAP-positive cells displayed nuclear STAT3 (nSTAT3) IR at day 5, which rose to60.91.54% at day 14 when compared with the CL side at each time point (n 3, Po0.001; Fig. 1e,h). In CKO mice, this

response was negligible or absent (Fig. 1f,h). Quantications of AC reaction assessed by intensity (optical density) measurements of GFAP immunoreactivity showed that intensity values were equally (P 0.49) increased 7-fold and 5.8-fold in the affected FN

of WT C57BL/6 mice and WT littermate controls compared to the CL side (n 3; Fig. 1k,l). By contrast, in the CKO group,

GFAP density values only rose 2.5-fold (n 6, Po0.01). AC

death was not the cause of decreased GFAP immunoreactivity, as pyknosis in this population was equally marginal in the two groups (P 0.634; Fig. 1j,k). Furthermore, AC numbers were also

comparable as assessed by labelling for ALDH1L1 (n 4, P40.05

for all; Fig. 1i).

We also tested whether STAT3 phosphorylation, another aspect of STAT3 activation, is associated with increased GFAP expression by immunoblotting whole FN tissue samples. In the ipsilateral (IL) FN of WT mice, the density of pSTAT3 IR bands was increased by 9.6-fold (P 0.001) when compared with the CL

side. While this response was much lower in CKO mice, overall STAT3 phosphorylation was not entirely abolished (n 3,

Fig. 1m,n). This mirrors the diminished extent of nSTAT3 IR observed in CKO mice, and it also indicates that other cells, such as neurons can efciently activate STAT3 after axotomy (Fig. 1e,f,o). We also found a positive correlation (R 0.9,

P 0.004) between the density of pSTAT3 and GFAP IR bands in

the axotomized FN. The latter was greater (Po0.001) by 2.9-fold in WT mice than in CKO mice (n 3; Fig. 1m,n).

By electron microscopy we then determined whether STAT3 activation can inuence AC process availability, a sensitive measure of AC reaction. The abundance of AC processes was conrmed by ultrastructural studies in which the proportion of neuronal membrane contacted by AC endfeet was quantied21. We detected a signicantly greater (Po0.001) neuronal coverage by WT AC processes in the axotomized FN compared with that seen for the CKO mice (WT IL: 44.72.94%, CKO IL:31.22.79%, n 14 cells; Fig. 2a,b). The astrocytic coverage in

the non-affected FN was comparable between the groups (Fig. 2a,b). Having conrmed that the abundance of perineuronal processes depends on astrocytic STAT3 activity, we examined whether it also correlates with changes in cytoskeletal lament content, such as GFAP, in individual ACs. The number of AC processes that were positively labelled for GFAP was quantied 15 mm away from their soma (6390 cells per group analysed, n 3; Fig. 2c). We found a 1.74-fold increase

in process numbers in the WT group where ACs displayed an average of 10.020.46 processes (P 0.009) compared with

5.750.15 processes seen for the CKO mice. To assess whether

2 NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294 ARTICLE

WT GFAP-STAT3-CKO

a b

IL FN

STAT3/Aldh1l1

/DAPI

c d

CL FN

STAT3/Aldh1l1

/DAPI

e f

IL FN

STAT3/GFAP

/DAPI

g h i j

nSTAT3/Aldh1L1 IR ACs

nSTAT3 / GFAP IR ACs

AC density

Density of non-neuronal nuclei (IL FN)

Day 5 Day 14

***

*** *** ***

900

2,500

nSTAT3 IR AC of total

Aldh1l1 IR cells (%)

nSTAT3 IR AC of total

GFAP IR cells (%)

Aldh1L1 IR AC mm2

CL IL

***

CL IL

750

CL IL

2,000

Nuclei mm2

600

CKO

1,500

450

Pyknotic ACs

300

1,000

150

500

2.1%

2.3%

CKO

0 WT CKO

0 WT

WT GFAP-STAT3-CKO

IL FN - GFAP/DAPIFN -GFAP

Intensity of GFAP immunoreactivity

Fold increase

(normalized to CL FN)

10 NS

WT (littermates)

WT (C57BL/6) CKO

m n o

Immunoblots from FN tissue samples

pSTAT3 band density

Relative band density

to -actin

Proportion of nSTAT3 IR neurons

GFAP band density

***

WT contralateral

WT ipsilateral

CKO contralateral

CKO ipsilateral

CL IL

0.7

***

9.0

1 3

2 1 3

1 3

2 1 3

NS NS

86 kDa

42 kDa

50 kDa

pSTAT3

STAT3

GFAP

-actin

nSTAT3/NeuN IR cells

of all neurones (%)

0.6

***

7.5

CL IL

0.5

6.0

0.4

4.5

0.3

0.2

3.0

***

0.1

1.5

0.0

CKO

WT CKO

0 WT CKO

Figure 1 | STAT3 activation is necessary for reactive transformation of perineuronal ACs after remote injury. (af,k) Grey matter AC response in the facial nucleus (FN) 14 days post-facial nerve transection. (ad) Nuclear localization of STAT3 in Aldh1L1 IR ACs (uorescent microscopy), (e,f) in GFAP IR reactive ACs (thin arrows; confocal images) and in motor neurons (arrowheads) in the IL FN of WTmice (a,e) and its relative absence (empty arrows) in the CL side in WT mice (c) and in CKO mice (b,d,f). (g) Proportion of Aldh1L1 IR ACs and (h) GFAP IR ACs displaying nSTAT3 labelling at day 5 or 14 post axotomy in WT mice (n 3) and in CKO mice (n 5) (g, ***Po0.001, one-way analysis of variance (ANOVA); h, n 3, ***Po0.001, one-way ANOVA Po0.0001). (i) AC

density is expressed as mean of Aldh1L1 IR cell count (mm 2), which is comparable in WT and CKO groups and remains unchanged after axotomy (n 4,

NSP 0.645, one-way ANOVA). (j) Numbers of intact non-neuronal nuclei (n 3, NSP 0.39, t-test ) and the proportion of pyknotic ACs in both groups

(n 3, NSP 0.634, t-test ). (k) GFAP immunoreactivity in the axotomized and CL FN in WT mice and in CKO mice. GFAP IR ACs (arrowheads) surround

axotomized neurons (full arrows) and display marginal pyknosis (thin arrows) in the two groups. (l) Mean intensity (optical density) values for IL GFAP immunoreactivity normalized to the CL side within the same section (WT n 3, CKO n 6, NSP 0.49, t-test; **Po0.01, one-way ANOVA P 0.0013).

(m) Means of band densities for pSTAT3 and GFAP immunoblots (n, Supplementary Fig. 1) of IL and CL FN tissue samples, demonstrating overall STAT3 phosphorylation induced by axotomy in WTmice and to a lesser extent in CKO mice (n 3, ***Po0.001, *Po0.05, one-way ANOVA). (o) Mean percentages

of neurons displaying nSTAT3 labeling. Scale bar, 30 mm for af; 250 mm and 20 mm for k; n mice per group. Data represent means.e.m.

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294

the increase in the number of GFAP IR processes correspond with STAT3-induced dynamic cytoskeletal changes, we quantied the number of cells displaying cortical F-actin assembly with connections to radial laments in cultured ACs activated by interleukin-6 (IL-6). At 24 h, signicantly more WT ACs retained their reorganized cytoskeleton (P 0.043), while CKO cells did

not respond and showed at morphology with parallel laments (n 3 experiments per replicate cultures; Fig. 2d). We conclude

that STAT3 activation transforms grey matter ACs into a reactive phenotype, thereby increasing their perineuronal presence following axotomy.

Finally, we asked whether there are differences in the relative availability of reactive ACs to microglia to dene the degree of specic astrocytic inuence on neurons. Thus, we assessed the microglial response in WT and CKO mice after axotomy. We quantied both the density of CD11b-positive areas and the presence of activated proinammatory microglia/macrophages labelled for IBA1/Arg1 or for IBA1/Lamp2, respectively. We found comparable density values for CD11b immunoreactivity in the axotomized FN in the two groups at day 14 (n 3, P 0.299;

Fig. 2e). Although there was a more than 5.6-fold increase (P 0.001) in density of total IBA1-positive activated microglia

after axotomy, their numbers were comparable (P 0.82) in the

presence of normal and diminished AC reactivity (n 4;

Fig. 2f,g). Moreover, the proportions of Arg1-labelled cells among IBA1-positive microglia was equal between the two groups as early as day 5 (n 3, P 0.098) and also at day 14

when cell densities signicantly declined (Po0.001) by more than3.5-fold (n 4, P 0.307; Fig. 2f,h). We did not nd differences

in the fraction of IBA1/Lamp2-positive microglia/macrophages up to day 14 (n 3, P 0.218; Fig. 2f,i). These results indicate

that microglia/macrophage numbers were not affected signicantly by the reduction in astrocytic STAT3 activation.

ACs support neuronal integrity via STAT3 signalling. We addressed whether ACs maintain neuronal viability through STAT3 signalling. First, we compared neuronal integrity in the FN of WT mice and CKO mice with diminished AC reactivity and STAT3 activation (Fig. 3a). Viable, non-atrophic neurons were dened by strong NeuN IR and intact nuclei at days 1, 7, 14 (Fig. 3a,b) and 28 post-facial nerve transection. From day 7, the preservation of neuronal integrity was nearly twofold greater (Po0.0001) in the vicinity of activated WT ACs (Fig. 3a,b). In

CKO mice, the proportion of pyknotic nuclei with weak NeuN immunoreactivity was 2.63-times higher (P 0.041) than in WT

mice at day 14 (n 7; Fig. 3a,c). Neuronal death was also con

rmed by activated caspase-3/NeuN co-labelling. 30.141.93% of neurons displayed activated caspase-3 immunoreactivity in the IL FN of CKO mice, which was 1.7-fold greater than that seen for WT mice (n 3, P 0.025; Fig. 3df).

We also addressed whether Schwann cells (SCs) could be affected by GFAP-Cre-dependent reduction in STAT3 activation in CKO mice, which could potentially contribute to neuronal death by the loss in their trophic inuence at the nerve stump.We found a reduction in the density of nSTAT3-positive and S100 IR myelinating or non-myelinating SCs in axotomized nerves sampled 1 mm proximal to the stump (P 0.002).

However, the overall density of SCs raised to equal values in WT and CKO mice following axotomy (n 3, P 0.857;

Fig. 3g,h).

We then sought to determine in vitro whether ACs via STAT3 can exert a direct neuroprotective effect independently of SCs. We set up a puried ACneuron co-culture system in which ACs were initially ablated by AraC and then replaced either by WT, CKO or no ACs for comparison (n 3). Cortical neurons derived

from neonatal mice are known to be vulnerable in culture, especially without the presence of astrocytic factors, providing a simple survival assay22. Viable neurons were dened by the criteria used in our in vivo experiments (Fig. 4af). Consistent with our in vivo ndings, WT ACs were nearly 1.73 times more efcient (Po0.01) than CKO ACs in supporting neuronal integrity (Fig. 4g). This was also mirrored by a signicantly reduced proportion of activated caspase-3-positive neurons in the presence of WT ACs (10.11.4%; n 3, P 0.001) when

compared with the CKO group (33.82.49%) or to cultures with no added ACs (38.14.04%; n 4, Po0.01; Fig. 4h), while

the number of re-plated ACs in the WT and CKO groups was equal (Fig. 4i). Together, these results suggest that ACs exert an independent STAT3-mediated protective effect.

AC reaction is associated with synaptic recovery. We also asked whether the AC reaction enhances the maintenance or recovery of synaptic input on neurons to help restore function. To address this question, initially we compared synapse densities in WT and CKO mice at day 14 post axotomy. This time point correlates with accelerated synapse formation, which follows the initial phase of synapse stripping6. At the ultrastructural level axosomatic synapse density was dened by the number of synapses along a 1 mm length of the neuronal membrane (Fig. 5a).

The values were equal in the CL sides in the two groups, suggesting normal developmental synaptogenesis in the CKO group. However, synapse density in the injured side was 1.54-fold higher in WT mice (n 12 cells) compared with CKO mice

(n 17 cells, Po0.05; Fig. 5b). The density of axodendritic

synapses reached the values of the CL side in WT mice (n 13

dendrites), whereas this was still reduced by 1.76-fold in the CKO group (n 12 dendrites, Po0.001; Fig. 5c). These results suggest

that AC reactivity promotes structural synaptic rearrangements.

To assess whether these observed changes represent functional synapses, we measured frequencies of excitatory postsynaptic

Figure 2 | STAT3 mediates reactive AC process formation. (a,b) Differential increase in perineuronal AC process availability 14 days post facial nerve transection in WT and CKO mice. (a) Electron micrographs of AC endfeet (green) along the neuronal membrane (blue) in the IL and CL FN. (b) Ultrastructural analysis: percentage of neuronal perimeter contacted by AC endfeet (CL n 12, IL n 14 cells, ***Po0.001, **Po0.01, one-way analysis of

variance (ANOVA) Po0.0001). (c) Mean number of AC processes that are GFAP IR 15 mm beyond the soma (n 3 mice, 6390 cells in total, **P 0.009,

t-test). (d) In vitro examples of activated ACs displaying cortical F-actin bres (phalloidin, empty arrow) with connections to radial laments (small arrow) and phosphorylated focal adhesion kinase (pFAK) 24 h after IL-6 treatment. Non-activated ACs retain their at morphology with mostly parallel laments (full arrow). Mean numbers of WTor CKO ACs that retained their activated phenotype 24 h after activation by IL-6 or in serum-free medium (Sato) (n 3

experiments per replicate cultures, *P 0.043, t-test). (e) CD11b immunolabelling in the CL and IL FN in WT and CKO mice. Lower panel represents

confocal images of the axotomized FN. In WT mice, CD11b IR structures (full arrows) appear to intermingle with processes that are strongly positive for GFAP (empty arrows). Mean density values for CD11b IR areas normalized to the CL FN in WT and CKO mice at day 14 post axotomy. (f) Illustrations of IBA1 IR-activated microglia in IL FN (green) and of its subtypes co-labelled (arrows) for markers, Arg1 or Lamp2 (red) in WT and CKO mice at day 14 post axotomy. (g) Means of cell densities for IBA1 IR-activated microglia at day 14. (h,i) Means of percentages of Arg1/IBA1 (h) or Lamp2/IBA1 (i) double-labelled microglia/macrophages over the total number of IBA1-positive cells in the IL FN (day 5 n 4 and day 14 n 3, ***Po0.001, **Po0.01, one-way

ANOVA). Scale bar, 0.25 mm for a; 20 mm for d; 100 and 20 mm for e; 200 and 25 mm for f. Data represent means.e.m.

4 NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294 ARTICLE

currents in surviving motor neurons 1014 days after axotomy. Whole-cell patch clamp recording was performed on acute brainstem slice preparations taken from axotomized postnatal mice (P20P24). In CKO mice, there was a clear reduction in synaptic

activity (3.671.25 Hz, n 6, P 0.036) in surviving neurons

when compared with that seen in WT mice (9.471.87 Hz, n 5;

Fig. 5d). This suggests a more abundant excitatory synaptic input post axotomy in WT animals over CKO mice, which also correlates

15 m

GFAP IR filaments in AC processes in vivo

WT GFAP-STAT3-CKO

GFAP-STAT3-CKO

CKO

IL FN

GFAP IR processes / AC

Cytoskeletal changes in vitro

Astrocytes with strong

corticalF-actin (%)

GFAPSTAT3-CKO

CKO

CL FN

Phal / pFAK / DAPI

WT-Sato

WT-IL6

CKO-IL6

0 CKO-Sato

Coverage of neurons by astrocyte processes (TEM)

IL FN -GFAP/ CD11bCD11b

CL FN IL FN IL FN

CL FN

Density of CD11b IR area

NS ***

***

5 CKO

Contact length

(% of perimeter)

Normalized integrated

density values

0 WT CL

WT IL

CKO CL

CKO IL

WT GFAP-STAT3-CKO

IBA1/ Arg1 IBA1

CL FN IL FN CL FN IL FN

IL FN

IBA1/ Lamp2

Microglia density

WT CKO

800

CL IL

Arg1 IR microglia Lamp2 IR microglia

*** NS

WT CKO

IBA1 IR

microglia mm2

Arg1 IR cells in

the IBA1 IR

population (%)

Lamp2 IR cells in

the IBA1 IR

600

400

population (%)

200

0 WT Day 5 Day 14 Day 5 Day 14

CKO

D14 post axotomy

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294

GFAP-STAT3-CKO

CL FN - NeuN

IL FN - NeuN

IL FN - NeuN/DAPI

Neuronal survival following in vivo axotomy

1.0

neurons in vivo

(normalized to CL FN)

CKO

NeuN IR intact

Proportion of caspase-3/NeuN IR

neurons (%)

** *

Proportion of

NeuN IR pyknotic

neurons (%)

0.5

0.0 D1 D7 D14 D28

0 WT CKO WT CKO

Days post axotomy

Day 14 Day 14

WT GFAP-STAT3-CKO

NeuN/Casp3

GFAP-STAT3-CKO

S100 IR SC density

CL IL

25 * *

STAT3/S100/DAPI

IL FN CL FN

Cells per area

(3 x 104 m2 )

NS**

Cells per area

(3 x 104 m2 )

nSTAT3 IR SC density

0 WT CKO

Figure 3 | Neuronal integrity is maintained in areas of full AC activation in vivo. (a) Fluorescent microscopic images of NeuN IR neurons in the CL and IL FN in brain sections from WT and CKO mice 14 days after facial nerve transection. DAPI-stained pyknotic neuronal nuclei are present in the IL FN (full arrows) among intact neurons (empty arrows). (b) Relative number of intact and strong NeuN IR neurons at days 1, 7, 14 and 28 post axotomy (n 3, 3; 4,

4; 7, 7; 4, 3 mice for WT, CKO groups, two-way analysis of variance (ANOVA) Po0.0001, F(1/24) 23.72 for the genotype effect, **P 0.0026,

**P 0.0029, *P 0.044, respectively, t-test for pairs). (c) Mean percentages represent the proportion of neurons undergoing pyknosis (n 3 mice,

*P 0.041, t-test) and (d) the proportion of NeuN IR neurons that are also positive for activated caspase-3 labelling (n 3 mice, *P 0.025, t-test).

(e,f) Examples of activated caspase-3/NeuN IR neurons in the IL FN of WT and CKO mice. (g) Axotomized and also CL facial nerves sampled close to the stylomastoid foramen 1 mm away from the nerve stump. Images show nuclear localization of STAT3 (nSTAT3, arrows) in S100-positive SCs in both WTand CKO mice. Insets represent confocal microscopic images. (h) Means of density values (cells per area) for the total number of S100-positive cells and for nSTAT3/S100 double-labelled cells in the intact CL and axotomized nerves (n 3 mice, *Po0.05, **P 0.002, one-way ANOVA P 0.0016). Scale bar,

100 mm and 25 mm for a; 30 mm for e,f; 10 mm for g and for insets. Data represent means.e.m.

6 NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294 ARTICLE

NeuN/Casp3/DAPI

AraC + No AC

AraC + CKO AC

AraC + WT AC

NeuN

a b

c d

e f

Neuronal survival in N-AC Co-cultures at 21 DIV

** ** **

** **

NeuN IR cells per

field (105 m2)

Proportion of

caspase-3/NeuN

IR neurons (%)

Astrocyte per field

(105 m2)

No ACs

added

CKO ACs

added

WT ACs

added

No ACs

added

CKO ACs

added

WT ACs

added

No ACs

added

CKO ACs

added

WT ACs

added

Figure 4 | Astrocytic STAT3 supports neuronal integrity independently of other cell types in vitro. (af) Fluorescent images of NeuN/DAPI co-labelled cells in puried neuronal cultures in the near absence of ACs (a,b) and in the presence of CKO ACs (c,d) or WT ACs (e,f). Examples of pyknotic cells (full arrows, b,d). Insets illustrate the same areas with activated caspase-3/NeuN co-labelling. Activated caspase-3-positive neurons with reduced NeuN immunoreactivity (empty arrows) are prevalent in the no added AC (b) and CKO AC (d) conditions. Some activated caspase-3-positive cells are already pyknotic (small arrows) and/or disintegrated into debri (empty arrows). (g) Mean number of surviving NeuN IR neurons per experiment at 21 DIV (n 3

cultures per 9 elds each, **Po0.01, one-way analysis of variance (ANOVA) P 0.0016). (h) Mean percentage values of NeuN IR neurons that are also co-

labelled for activated caspase-3. (i) The numbers of WTand CKO ACs surviving at 21 DIV were comparable. Only a small number of WT ACs remained in the no added ACs group after AraC treatment (for h and i: n 3 cultures per 9 elds each, **Po0.01, one-way ANOVA P 0.001). Scale bar, 30 mm for a,c,e;

10 mm for b,d,f and for insets. Data represent means.e.m.

with a 1.45-fold greater density of PSD-95-labelled postsynaptic puncta in WT mice (P 0.036), while the density of dendrites has

not changed (Fig. 5eg).

Next, we addressed whether the greater excitatory input in WT mice represents synapse recovery rather than maintenance of connections in adult animals. We identied synapses by synaptotagmin-1/PSD-95 co-labelling during synapse stripping (day 5) and the re-organization phase (day 14) in both IL (Fig. 6a) and CL FN (Fig. 6e). Synapse densities were dened by the number of IR puncta per 0.01 mm2 area, and were normalized to the CL measurements to demonstrate the efcacy of synapse recovery. Initially synapse densities declined equally (P 0.125)

to values of 0.680.03 and 0.610.02 for WT and CKO mice (n 3), respectively. However, signicant recovery was only

observed in WT mice to a ratio of 0.840.02 (n 6, P 0.001;

Fig. 6b). In contrast, in areas of STAT3-decient AC populations, this ratio remained as low as 0.540.05 (n 4), 30% lower when

compared with that seen in the WT mice (Fig. 6b). Despite greater neuronal degeneration in the CKO mice, the IL/CL ratios of MAP-2-positive dendritic areas (Fig. 6a) did not differ signicantly in the two groups (n 4; Fig. 6c). This may suggest

dendritic compensatory mechanisms23 in surviving neurons in CKO mice; however, synapse recovery was still abolished. Our results indicate that surviving neurons in the CKO mice receive signicantly less excitatory input per cell than their WT counterparts.

TSP-1 is required for synapse recovery following axotomy. We then asked what factors are directly responsible for the observed synaptic changes mediated by reactive ACs. TSP-1, an AC-

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294

a b

GFAP-STAT3-CKO Synapse density on the soma

Synapse density on dendrites

IL FN-neuronal soma IL FN-dendritic area

0.8

0.0

***

Synapses per mSynapses per m

0.6

0.4

0.2

0.0

WT IL

CKO CL

CKO IL

0.4

***

0.3

0.2

0.1

WT IL

CKO CL

CKO IL

WT GFAP-STAT3-CKO

CL FN

PSD-95/MAP-2

CL FN

IL FN

PSD-95/MAP-2

d f g

Synaptic activity (EPSC)

PSD-95 IR puncta

MAP-2 IR area

CKO

WT WTCKO

CKO

0.8

Frequency (Hz)

Number of puncta

(normalized to CL)

* CKO

0.6

1.0

CKO

0.4

Area

(normalized to CL)

0.5

10 pA

0.2

5 pA

10 ms

0.0

500 ms

Figure 5 | STAT3-mediated AC response corresponds with greater synaptic density and input onto neurons. (a) Electron micrographs of axosomatic and axodendritic synapses within a 50 mm radius from the soma in the IL and CL FN 14 days post axotomy. WT mice, in contrast to CKO mice, have more presynaptic terminals (green) opposing the postsynaptic density (arrowheads). Insets demonstrate postsynaptic densities. (b) Number of synaptic contacts (mm 1) along the surface of the soma (WT n 12 cells, CKO n 17 cells) and (c) within the dendritic area (WT n 13 dendrites, CKO n 12

dendrites, * Po0.05, ***Po0.001, one-way analysis of variance Po0.0001). (d) Frequency of excitatory postsynaptic currents from whole-cell recordings in the IL FN 1214 days post axotomy (WT n 5, CKO n 6, *P 0.036, t-test), and representative trace recordings with single EPSCs. (e) Confocal images

of immunostained acute brainstem slices of axotomized postnatal (P2123) mice processed after electrophysiological studies. Examples demonstrating PSD-95 IR synapses (red) and MAP-2 IR dendrites (insets, white) in the CL and IL FN at 1214 days post axotomy. (f) Means of number of PSD-95 IR puncta (n 3 mice, *P 0.041, t-test) and (g) MAP-2 IR areas (n 3 mice, NSP 0.176, t-test) normalized by CL values in the same section. Scale bar,

0.5 mm for a; 0.1 mm for insets; 10 mm for e and 25 mm for insets. Data represent means.e.m.

derived molecule responsible for excitatory synaptogenesis during development, was recently shown to be re-expressed locally in CNS injuries16,24. Thus, we hypothesized that TSP-1 plays a role in the plasticity of synaptic input on axotomized motor neurons in adult mice. To explore this possibility, we compared synaptic densities in the FN of TSP-1 KO mice with those found in WT and CKO mice at days 5 and 14 (Fig. 6a,b). Successful TSP-1 protein depletion in ACs derived from TSP-1 KO mice was veried by immunoblots (Fig. 7m). Synapse recovery was signicantly impaired in TSP-1 KO mice, but to a smaller degree than that found in the CKO group (see above). Between days 5 and 14, the normalized synapse density values had only returned to 0.740.03 from 0.690.01 (n 4, Po0.05; Fig. 6b).

The IL/CL ratio of MAP-2 IR dendritic areas was also comparable with the values seen in all other groups (Fig. 6c). In addition, synapse density and MAP-2 area fraction values measured in the CL FN of all groups were equal, which reects no differences in developmental excitatory synaptogenesis or dendritic areas between WT and CKO or TSP-1 mice (Fig. 6ei). In summary, our ndings provide evidence that TSP-1 plays an important role in adult plasticity following axonal insults.

Astrocytic STAT3 induces TSP-1 expression and release. We asked whether TSP-1 could be responsible for reactive AC-driven synaptic recovery via STAT3 signalling. TSP-1 (Thbs1) messenger

8 NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294 ARTICLE

Ipsilateral facial nucleus

Contralateral facial nucleus

WT GFAP-STAT3-CKO TSP-1 KO

Synapse density

***

1.0

0.8

0.6

0.4

0.2

0.0

Day 5

Day 14

WT GFAP-STAT3-CKO TSP-1 KO

NS NS

SYT-1/PSD-95IR puncta

(IL/CL ratio)

CL FN-MAP-2CL FN-SYT-1/PSD-95IL FN-SYT-1/PSD-95

IL FN-MAP-2

WT CKO TSP1 KO

0.8

MAP-2 IR area

(IL/CL ratio)

0.6

0.4

0.2

0.0

3D view of MAP-2

/PSD-95IR dendrites

CKO

TSP1 KO

GFAP-STAT3-CKO

TSP-1 KO

PSD-95 density

SYT-1 density

NS NS

PSD-95IR puncta

per area (0.01 mm2 )

500

400

300

200

100

SYT-1 IRpuncta

per area (0.01 mm2 )

800

600

400

200

Synapse density MAP-2 area

SYT-1/PSD-95 IR puncta

per area (0.01 mm2 )

MAP-2 IR area fraction

per 0.01 mm2

0.25

0.20

0.15

0.10

150

100

0.05

0.00

Figure 6 | Recovery of excitatory synaptic density is associated with astrocytic STAT3 activation and is TSP-1 dependent. (a) Confocal images of synapses (arrowheads) demonstrated by colocalization of synaptotagmin-1 (SYT-1) and PSD-95 IR puncta and the MAP-2 IR dendritic tree in the IL FN in WT, CKO and TSP-1 KO mice 14 days post axotomy. (b) Means of ratios of synapses in the IL FN relative to the CL FN at day 5 (n 3 mice, one-way

analysis of variance (ANOVA) P 0.299) and at day 14 post axotomy (WT n 6, CKO and TSP-1 KO n 4 mice, ***Po0.001, *Po0.05, one-way ANOVA

P 0.0002 and t-test between day 5 and day 14 values: **P 0.001). (c) Means of ratios of MAP-2 IR areas (n 4 mice, NSP 0.272, one-way ANOVA).

(d) Three-dimensional reconstruction of confocal images showing a MAP-2 IR dendrite (grey) with areas of PSD-95 immunoreactivity at shaft and spinal synapses (green) in the WT IL FN. (e) Images showing synapses illustrated by SYT-1 and PSD-95 immunoreactivity and MAP-2 labeling (white) in the CL FN in adult WT, CKO and TSP-1 KO mice. Quantication of (f) PSD-95, (g) SYT-1, (h) SYT-1/PSD-95 IR puncta and (i) MAP-2 area fraction in the CL FN in the three groups (n 4, NSP40.05 for all, one-way ANOVA). Data represent the mean of number of synaptic puncta or MAP-2 IR area per 0.01 mm2 eld.

Scale bar, 10 mm and 25 mm for a; 1 mm for d; 25 mm for e. Data represent means.e.m.

RNA (mRNA) expression was examined in the FN of WT and CKO mice at day 14 post axotomy. By in situ hybridization, we detected a 6.73-fold increase in the number of Thbs1 mRNA/ GFAP-expressing ACs (Fig. 7ac) in the WT mice when compared with the CL side (n 3, P 0.039). This increase was

reduced to 2.21-fold in reactive ACs with impaired STAT3 activation in the CKO mice. Motor neurons also displayed Thbs1 mRNA expression, however, their numbers were only marginally increased by 1.71-fold on the axotomized side and was equal in the two groups (n 3; P 0.407; Fig. 7b,c). The number of

Thbs1 mRNA-positive cells excluding ACs and neurons were negligible.

To conrm the link between astrocytic STAT3 activation and Thbs1 expression, we also quantied mRNA levels by quantitative PCR (qPCR) from FN tissue samples. In WT mice, Thbs1 mRNA levels were B3-fold greater in the axotomized FN when compared with the non-affected FN or to the FN of CKO mice (n 3, P 0.035; Fig. 7d). This response to axotomy was also

reected by increased TSP-1 protein expression indicated by a1.9-fold higher band density value in immunoblots from IL FN

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294

DiscussionWe provide denitive evidence that, despite the histological resemblance to ACs found in inhibitory glial scars in direct CNS injuries, remotely activated grey matter ACs promote structural plasticity and circuit integrity in the adult CNS. We describe novel mechanisms underlying this effect by showing that: (1) remote activation of ACs and their process formation in the vicinity of axotomized facial neuronal cell bodies is STAT3 dependent, despite the limited inammatory response; (2)STAT3-dependent AC activation is required for the recovery of functional synaptic input onto surviving motor neurons, and not only for supporting neuronal preservation; (3) a direct regulatory link exists between astrocytic STAT3 activation and the re-expression/release of the synaptogenic molecule TSP-1; (4) TSP-1 re-expression, in part, is directly responsible for the remote AC-mediated recovery of excitatory input onto facial motor neurons in adult plasticity.

ACs responding to remote cues extend processes and tend to remain in their original domains without proliferation6,19. We provide both ultrastructural and immunohistochemical evidence that the formation of perineuronal astrocytic processes is mediated via STAT3 phosphorylation despite the lack of a signicant inammatory response. In our in vivo experiments, the majority of reactive ACs displayed nSTAT3 labelling, leading to a cascade of transcriptional events26. However, cytoplasmic pSTAT3 may also directly inuence cell migration and process outgrowth, as found in broblasts, by binding to Rac1 activator b-PIX, modulating cytoskeletal re-organization27. Our ndings show that STAT3 signalling is also a key element in remote reactive transformation of ACs, explaining the role of STAT3 activation seen in previous studies18,28. This could be potentially triggered by injured neurons via release of cytokines, such as IL-6 (refs 2931).

What are the effects that remotely activated ACs exert through this pathway then? ACs with a reactive phenotype have often been regarded as cells inhibitory for repair9. However, recent evidence suggests that the outcome of glial reaction to CNS damage may vary depending on regional differences32,33 and on the severity of injury, and may contribute to tissue preservation in cases of moderate insults to neurons34,35. Our ndings are in line with these observations. It is possible that the limited inammatory cell response in the facial nucleus36 allows ACs to display a reactive but neuroprotective phenotype, indicating a default protective role of AC activation.

We provide evidence that astrocytic STAT3 signalling plays a direct and important role in supporting the integrity of the neuronal network, albeit they are likely to do so in conjunction with other cells. We demonstrate that ACs can independently support neuronal integrity via STAT3 signalling, using puried ACneuron co-cultures. Indeed, microarray studies indicate that

Figure 7 | STAT3 activation directly regulates TSP-1 expression in ACs. (a,b) In situ hybridization for Thbs1 in the FN 7 days post axotomy in WTand CKO mice. (b) Triple-labelling shows co-localization of Thbs1 mRNA with GFAP (empty arrows) or NeuN immunoreactivity (arrows). (c) Quantication of ACs and neurons expressing Thbs1 mRNA (n 3 mice, *P 0.039, NSP 0.407, t-test). (d) Thbs1 mRNA measured by qPCR in IL and CL FN tissue samples in

WT and CKO mice at day 14 (n 3, *Po0.05, one-way analysis of variance (ANOVA) P 0.035). (e) Immunoblots of samples taken from FN tissues at

day 14, representing TSP-1 content against pSTAT3, STAT3 and b-actin immunoreactivity. (f) Mean density values for TSP-1 immunoblots of IL and CL FN (n 3, **Po0.01, ***P 0.0002, one-way ANOVA). (g) ChIP demonstrates specic STAT3 binding to the TSP-1 promoter by showing PCR product

enrichment in the STAT3 pull-down sample. Positive control: sonicated/non-precipitated input sample. Negative controls: normal rabbit immunoglobulin G and No DNA samples. (h) Thbs1 mRNA levels measured by qPCR in puried WT (n 3) and CKO AC cultures (n 6) with or without IL-6 treatment

(***Po0.001, one-way ANOVA Po0.001). (i) Aldh1L1/GFAP/DAPI- and pSTAT3 or TSP-1 quadruple labelling in WT and CKO ACs activated by IL-6 or in serum-free conditions (Sato). Proportion of GFAP/Aldh1L1 IR ACs displaying (j) pSTAT3 (empty arrows) or (k) TSP-1 immunoreactivity in Sato and IL-6-activated conditions (n 3 experiments per replicate cultures, ***Po0.001, *Po0.05, one-way ANOVA P 0.0012). (l) Confocal images illustrating

intracellular or extracellular distribution of TSP-1 immunoreactivity in non-activated or IL-6-treated WT ACs, respectively. (m) TSP-1, pSTAT3 and STAT3 immunoblots for WT, CKO and TSP-1 KO ACs treated with Sato or IL-6. (n) ELISA: mean concentrations of TSP-1 (pg ml 1) released into supernatants by

WT and CKO ACs after treatment with Sato or IL-6. Dots represent chemiluminescence (n 3 experiments per replicate samples, **Po0.01, one-way

ANOVA P 0.002). Scale bar, 200 mm for a; 50 mm for b; 40 mm for i; 20 mm for l. Data represent means.e.m. Supplementary Figs 24.

tissue samples of WT mice (n 3, P 0.002; Fig. 7e,f). Selective

reduction of AC reactivity in CKO mice resulted in no response in Thbs1 mRNA (P 0.589) and almost completely abolished

TSP-1 protein expression when compared with that seen in WT mice (P 0.001).

We then asked whether STAT3 directly regulates Thbs1 expression in ACs. We tested this by a series of molecular studies using puried ACs cultured with or without the STAT3 activator IL-6. First, chromatin immunoprecipitation (ChIP) assay was performed using DNA samples from cultured ACs activated by IL-6. The TSP-1 promoter PCR product was enriched in the STAT3 pull-down sample, indicating that STAT3 specically binds to the promoter sequence (Fig. 7g). To conrm that STAT3 binding also initiates transcription, we then measured Thbs1 mRNA levels from these samples by qPCR. We found a 2.2-fold increase in WT ACs cultures in the presence of IL-6 when compared with controls (n 3, Po0.001; Fig. 7h). Inhibition of

STAT3 activation in CKO ACs abolished the increase in Thbs1 mRNA (n 3, P40.05), suggesting a tight link between STAT3

activation and Thbs1 expression (Fig. 7h).

Next, we examined the TSP-1 protein content in non-activated and IL-6-activated astrocyte cultures. ACs were identied by GFAP/Aldh1L1 co-immunolabelling, as grey matter ACs may express no or low levels of GFAP25. IL-6 induced a signicant rise (P 0.001) in the proportion of ACs with nSTAT3 labelling in

WT cultures to 82.11.66% from 8.672.59% (n 3; Fig. 7i,j),

whereas this was negligible or absent in all CKO ACs regardless the presence of IL-6 (P 0.521). In WT cultures, STAT3

activation resulted in signicantly greater (P 0.025) propor

tion of ACs displaying TSP-1 immunoreactivity (42.423.4%) over non-treated cells (27.822.41%) or when compared with CKO cultures (Po0.001). In CKO ACs, there was no response to

IL-6 (n 3, experiments/replicate cultures, P 0.292; Fig. 7ik).

These results indicate that TSP-1 protein expression in ACs is STAT3 driven.

Since TSP-1 can be secreted, we addressed whether it is released from ACs. Along with the increase of TSP-1 mRNA levels in activated ACs, immunoblotting and confocal imaging revealed that the intracellular TSP-1 content was reduced while TSP-1 was deposited on AC surfaces or in the extracellular matrix (Fig. 7l,m). In contrast, CKO ACs already had weaker TSP-1 bands, and this was not reduced upon activation to the extent seen for the WT cultures. We then measured the amount of TSP-1 released from ACs by enzyme-linked immunosorbent assay (ELISA). We observed signicantly higher levels of TSP-1 concentration(2.080.10 pg ml 1) in the supernatant of IL-6-treated ACs in the WT group over the CKO cultures (1.530.09 pg ml 1) or controls (n 3, Po0.01; Fig. 7n). In summary, our experiments

provide novel evidence that TSP-1 is re-expressed via STAT3 signalling, and it is also concomitantly released from the cells.

10 NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294 ARTICLE

astrocytic STAT3 activation may induce neuronal survival by expression of molecules responsible for antioxidant defence37. This includes glutathione synthase, whose deciency leads to less efcient protection against glutamate excitotoxicity38. Although, SCs can also support neuronal viability via c-Jun39, we cannot be entirely certain whether reduction in STAT3 activation in SCs could have contributed to an impaired trophic effect in CKO mice. The relevance of this possibility is yet to be addressed.

Furthermore, it is plausible that ACs exert their protective effects in combination with microglia, too. Their interaction through purinergic gliotransmission has been shown to be protective40. We also found that the relative increase of the availability of reactive ACs to microglia corresponds with enhanced neuronal preservation. Other factors, such as limited lymphocyte inux, have shown to be less relevant in inuencing survival in this model36.

WT GFAP-STAT3-CKO

Thbs1 mRNA Thbs1 mRNA

TSP-1 band density

NeuN/GFAP/DAPI

Thbs1 mRNA positive cells

Thbs1 qPCR-FN tissue

WB-FN tissue

10 *

Astrocytes Neurons

CL IL

WT CKO

0.5

*** **

Fold increasein Thbs1

mRNA-positive cells

(IL versus CL)

Expression fold change

CL IL

IL CL

0.4

pSTAT3

STAT3

TSP1

-actin

86 kDa

Relative density

to -actin

0.3

0.2

170 kDa

0.1

42 kDa

0.0

WT CKO WT CKO

WT CKO

WT CKO WT CKO

WT CKO

WT GFAP-STAT3-CKO

Chromatin immunoprecipitation

Thbs1 qPCR-cultured ACs

Ladder

TSP-1 promoter

Input STAT3 IgG No DNA

Sato

IL-6IL-6

300 bp

50 bp

pSTAT3

200 bp 100 bp

Expression fold

change

1.5

GFAP/Aldh1L1/DAPI/TSP-1

2.5 Sato IL-6

***

2.0

1.0

0.5

0.0

Immunoblots from cultured AC

TSP-1

TSP-1 ELISA AC supernatants

Sato

GFAP/Aldh1L/DAPI

IL-6

WT CKO TSP-1 KO

Sato

IL-6

TSP-1 IR AC

Sato Sato Sato

IL-6

IL-6 IL-6

TSP-1 concentration (pg ml1 )

*** ***

***

2.0

Proportion of pSTAT3/

Aldh1L1 IR AC (%)

pSTAT3 IR AC

Proportion of TSP-1

IR AC (%)

170 kDa

86 kDa

1.5

TSP-1

pSTAT3

STAT3

-actin

1.0

0.5

42 kDa

0.0

WT+IL6

CKO

CKO+IL6

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294

Apart from protecting against the harmful environment, ACs can directly boost neuronal viability. Several reactive AC-derived trophic factors and cytokines have also been implicated in contributing to neuron survival41. There is also evidence for AC-induced long-term potentiation4245, which may increase synaptic activity-induced neuronal survival46,47. This may occur when glutamatergic synapses are restored on neurons4852. Whether ACs can also mediate survival by re-establishing balanced excitatory synaptic input needs to be examined in more detail.

Do perineuronal reactive ACs orchestrate structural plasticity to support circuit integrity? The involvement of reactive processes in reorganizing connections is suggested by descriptive studies showing enhanced glial process extension in areas with active synapse formation17,19,53,54. Initially, reactive microglial and perhaps AC processes may remove synapses from injured motor neurons during the rst stages of synaptic rearrangements55,56. Our ndings suggest the possibility that during the early phases of synapse stripping, microglial cells play a more signicant role than do reactive ACs as the initial reduction in synapses was independent from the extent of early astrocytic response. However, the eventual structural recovery of synaptic input on motor neurons corresponded well with the relative availability of reactive ACs and their processes over microglial presence. Indeed, the numbers of activated microglia/macrophages were equal in the WT and CKO groups, yet synapse densities were reduced in the latter. These results suggest that reactive ACs promote the recovery of neuronal connections. Specically, our electrophysiological ndings and synapse density analysis indicate the restoration of functional afferents on motor neurons.

Thus, it was pertinent to ask whether AC activation can directly and independently facilitate excitatory synaptic rearrangements. AC-derived synaptogenic molecules were shown to stabilize excitatory synapses in vitro14 and in development13,57,58. Although the joint action of TSP-1 and TSP-2 is required for postnatal plasticity58, using TSP-1 knockout mice, we show that TSP-1 alone has a signicant role in synaptic rearrangements in the adult brainstem. We therefore asked whether reactive ACs can supply TSP-1 to neurons in the recovery phase. We show that reactive ACs are the major populations that signicantly upregulate TSP-1 expression proximal to the site of axotomy. The contribution of cells other than ACs in TSP-1-mediated adult plasticity is less likely, although in development, TSP-1 was seen in phagocytic cells and thrombospondin immunoreactivity could be also detected in microglia around dying neurons in adult mice24. Our study suggests that it is possibly a result of phagocytic internalization of neuronal debris or extracellular TSP-1 content, as Thbs1 mRNA was almost exclusively detected in ACs and neurons. We propose that contribution by neurons to the increase of TSP-1 expression is modest. The number of Thbs1 mRNA-labelled neurons was only marginally raised upon injury, and individual neurons did not seem to increase TSP-1 production. Indeed, in CKO mice in which neuronal STAT3 function is not affected, the increase in both TSP-1 mRNA and protein levels was almost abolished in tissue samples. These results indicate that the main source of TSP-1 is reactive ACs following axotomy, and our experiments conducted using puried AC cultures strengthen our conclusion. Using IL-6-stimulated ACs in culture, we demonstrated direct STAT3 binding to the TSP-1 promoter, increased mRNA levels and rapid loss of intracellular TSP-1 content concomitant with raised TSP-1 protein levels in supernatants upon astrocytic STAT3 activation. Our work provides direct evidence for STAT3-regulated TSP-1 expression and release.

Other mechanisms or factors may also be responsible for synaptic remodelling via activated ACs. Our study indicated a

greater reduction in synapse recovery in CKO mice with impaired AC activation compared with that seen in TSP-1 KO mice. Recent evidence suggests that other AC-derived factors, such as glypican 4 and 6, may also play a role in neonatal plasticity59. In vitro PY2 purinergic signalling and type IV collagen/a1b1 integrin-mediated cascades also lead to an increase in TSP-1 expression60,61. We provide a novel regulatory mechanism for TSP-1 and show direct evidence for its involvement in reactive AC-mediated structural synaptic plasticity in adulthood.

Finally, what is the relevance to novel therapies in neurology? A common characteristic feature of CNS trauma or advanced neurodegenerative disease is widespread AC activation. It is essential to understand to what degree these astrocytic phenotypic changes represent pathology rather than an adaptive protective response. Emerging evidence suggests that the astrocytic effect may vary, even if the primary pathogenic trigger is derived from glia62. Microarray studies based on selective sampling of ACs in various neurodegenerative conditions have revealed both degeneration-and regeneration-related features63,64. In our reductionist paradigm of selective motor neuron injury, in which ACs are activated primarily by neuronal injury, reactivity leads to a restorative outcome. This approach provided an opportunity to dissect out the benecial aspects of AC reaction that may also be relevant in disease. We present a novel regulatory mechanism mediated via astrocytic STAT3 signalling, which plays a key role in adult plasticity and network integrity in combination with other cells. Remotely activated ACs promote the recovery of excitatory input on surviving motor neurons by upregulation of TSP-1 expression. Further genomic and proteomic approaches are necessary to better understand the complexity of the AC response in neurodegeneration and trauma. This may open up opportunities to develop new neuroprotective strategies.

Methods

Animals and transgenic models. For modelling diminished AC activation, 810-week-old adult male GFAP-STAT3-CKO mice (CKO) were used (received from Prof. M. Sofroniew). Briey, the transgenic mice were generated by crossing the73.12 GFAP-Cre mouse line65 with STAT3-loxP mice using C57BL/6 mice as background. The loxP sites anked exon 22 of the STAT3 gene encoding a tyrosine residue (tyr705) that is critical for its activation, resulting in the inhibition of STAT3 activation specically in GFAP-expressing cells in crossed CKO mice. These mice show comparable development and phenotypes of ACs to that seen for adult WT mice12. For WT control, 810 week-old adult male GFAP-Cre / /

STAT3-loxP littermates and/or age-matched background C57BL/6 mice were used. For synapse recovery analysis, in addition to the above, 810-week-old adult male Thbs1tm1Hyn mice (TSP-1 KO, C57BL/6J background) were also used. These mice are homozygous for Thbs1 deletion (Jackson Labs).

Surgical facial axotomy. All experimental procedures were carried out in accordance with the UK Scientic Procedures Act (1986) and guidelines set out by the International Association for the Study of Pain guidelines for the care and use of animals. The right facial nerve of 810-week-old male mice or postnatal P1112 male mice was transected at its extracranial course near the stylomastoid foramen under uothane (2%) anaesthesia with oxygen (1.5 l h 1). Before surgery, a subcutaneous injection of buprenorphine (Vetergesic; 0.1 mg kg 1) was administered to minimize pain and discomfort together with antibiotics (penicillin and streptomycin) to minimize potential infection. The adult animals were killed by a lethal injection of phenobarbital (300 mg kg 1) in humane conditions following brief anaesthesia by uothane (2%) at different time points post axotomy (1, 5, 7, 14, 28 days).

AC cultures. To generate ACs for both monocultures and neuronal co-cultures, cerebral cortices of P1 transgenic GFAP-STAT3-CKO mice, WT littermates, C57BL/6 WT mice (Harlan Olac & Charles River) and TSP-1 KO mice (Jackson Lab) were prepared as described in previous studies66. In general, this method establishes a purity of over 95% for ACs with less than 1% microglia. For ourin vitro experiments, 98% pure AC cultures were used. In some cases, further purication steps were necessary to achieve this. To do so, immunopanning or complement-mediated cell lysis methods were applied to remove broblasts according to published protocols (Supplementary Methods). ACs were cultured for 6 weeks (mature ACs) in 10% fetal bovine serum, 1 mM GlutaMAX in Dulbeccos modied Eagle medium (Life Technologies) to provide cells for in vitro assays. To evaluate the effects of AC reactivity, mature ACs were plated onto PDL coated

12 NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294 ARTICLE

coverslips (5 103cells per well) in 24 well plates or in T25 asks. After 23 days,

the medium was replaced with Satos serum-free medium before treatment with IL-6 (50 ng ml 1), after which the cells were either immunostained or lysed for protein (complete lysis-M, EDTA-free buffer, Roche, UK) or RNA extraction. The supernatants were kept for analysing TSP-1 content by ELISA.

ACneuron co-culture assays. To produce puried cortical ACneuron co-cultures in which ACs were replaced by their GFAP-STAT3-CKO or littermate counterparts, initially a mixed glialneuron co-culture was produced using WT E18 embryos, similarly to previously published protocols22. To ablate contaminating and proliferating cell populations such as ACs, progenitor cells, broblasts and microglia, cytosine arabinoside (1 mM AraC, Sigma-Aldrich) was added to the cultures between days 57, while the medium was supplemented with AC-conditioned medium. This signicantly reduced the number of ACs in the cultures (o6%). For the survival assays, mature ACs derived from either GFAP-STAT3-

CKO mice or their WT littermates were re-added to the puried neuronal cultures. Cells were cultured for 21 days in total. See also Supplementary Methods.

ChIP. To detect physical binding of STAT3 to the TSP-1 promoter region in ACs activated by IL-6, ChIP was performed using the Magna ChIP A/G Kit (Millipore) according to the manufacturers instructions. Briey, 107 mature ACs were xed in 1% formaldehyde for 10 minutes followed by quenching with glycine for 5 min. ACs were washed in phosphate-buffered saline (PBS) before cell and nuclear lysis, which was performed in the presence of protease and phosphatase inhibitors. The isolated chromatin was sonicated to obtain DNA fragments of 2001,000 base pairs. Then, immunoprecipitation was performed overnight at 4 C, using a rabbit anti-STAT3 antibody (Cell Signalling, clone 79D7) bound to A/G magnetic beads. After isolation and elution, reverse cross-linking was performed according to standard protocols. DNA was then puried and analysed by PCR using a primer pair designed to detect a predicted STAT3 binding site67 along the TSP-1 promoter. The size of the PCR product was 176 bp. Forward primer: 50-TGGCTTC

CTCTGTGGTCTCT-30. Reverse primer: 50-GTCAAGGTCATGGGATGGTC-30.

Synthesis of digoxigenin-labelled RNA probes. To synthesize the probe for TSP-1 in situ hybridization, RNA from the brain of a P8 mouse was extracted using ISOLATE RNA Mini Kit (Bioline, UK), following the manufacturers instructions for RNA isolation. The isolated RNA was reverse transcribed into complementary DNA (cDNA) using the cDNA Synthesis Kit (Bioline) using a standard protocol for the Oligo dT primer. The resultant cDNA was amplied using standard PCR protocols. Murine TSP-1-specic primer sequences used were 50-GTTCGTCG

GAAGGATTGTTA-30 for the forward primer and 50-TCTATTCCAATGGCA ACGAG-30 for the reverse primer68. The size of the PCR product was 733 bp. Adaptor sequences were added to the original pair of primers to create attachment sites for the SP6 and T7 polymerases for the synthesis of RNA probes. cDNA from the rst round of PCR was further amplied to produce the template for RNA probe synthesis, using the above forward primer with T7 phage promoter sequence (50-TAATACGACTCACTATAGG-30) added to its 50 end and the above reverse primer with SP6 sequence (50-ATTTAGGTGACACTATAGA-30) added to its 50 end. The SP6 and T7 promoter sequence allowed the antisense and sense RNA probe to be synthesized respectively. After conrming the identity of the PCR product by sequencing, digoxigenin (DIG)-labelled RNA probes were synthesized using the DIG RNA Labelling Kit (SP6/T7) (Roche) according to the manufacturers instructions. The yield of labelling (31.25 ng ml 1) was estimated using standardized dot blot protocol.

In situ hybridization. TSP-1-specic in situ hybridization was performed following instructions described previously69, with minor modications. In brief, frozen coronal mice brain sections (10 mm) were treated with proteinase K(200 ng ml 1 in PBS; Roche) for 20 s before xing in 4% paraformaldehyde (PFA).

After permeabilization, acetylation and prehybridization treatments, hybridization with prepared DIG-labelled probes (1:400 in hybridization buffer) was carried out at 60 C overnight. Following stringency washing, the hybridized probe was recognized with alkaline phosphatase-conjugated DIG-specic antibody (1:5,000; Roche) for 1 h at 25 C and the signal was then visualized with NBT/BCIP (Roche).

Reverse transcription and qPCR. For measurements of Thbs1 mRNA levels, mouse brainstems were snap-frozen and 300-mm thick coronal sections were cut, including the FN, using a cryostat (Leica). CL and IL FN were isolated using a purpose-made 0.8 mm punch needle; then, the tissue was triturated using a pestle. For in vitro samples, RNA was extracted from 5 105 ACs.

Total RNA was isolated using the ISOLATE II RNA Mini Kit (Bioline) according to the manufacturers instructions. Reverse transcription was performed using 0.11 mg of total RNA, with the Tetro cDNA Synthesis Kit (Bioline)

according to the dened protocol. qPCR was performed using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) in the Bio-Rad CFX96 thermocycler. Primers were chosen from the primePCR library (Bio-Rad) with a preferenceof intron-spanning primers when available. The Thbs1 assay ID was qMmuCID0014108 and the GAPDH (glyceraldehyde 3-phosphate dehydrogenase)

assay ID was qMmuCED0027497. Gene expression data were analysed using the relative quantication (ddCt) method using GAPDH as housekeeping gene to evaluate quantitative variation.

Tissue processing. For immunohistochemistry, mice were perfused with 4% PFA solution in PBS under phenobarbital anaesthesia. Frozen blocks were cut at 10-mm thickness to gain a series of coronal brainstem sections, using a cryostat (Leica). For electron microscopy, after the same anaesthesia, mice were preperfused with 20 ml of 0.1 M HEPES buffer containing nitrates followed by perfusion with 6% formaldehyde in 0.1 M HEPES buffer at pH 7.4 containing 2 mmol l 1 calcium chloride 4 weeks post axotomy. The brains were kept in 0.1 M HEPES buffer for 1 h before trimming. Following inltration by Lowicryl HM20 (Taab Laboratories) selected resin blocks were trimmed for ultrathin sectioning (Leica Ultracut UCT, Leica), stained with lead citrate/uranyl acetate, mounted on nickel grids and viewed by electron microscopy.

Immunolabelling and imaging. For immunohistochemistry, following standard protocols, frozen sections were blocked in 10% normal goat serum (NGS) and permeabilized in 0.3% Triton X-100 (Sigma-Aldrich; in PBS) at RT for 30 min. They were then stained with primary antibodies in NGS (3%) and Triton X-100(0.1% in PBS) at 4 C overnight followed by species-specic secondary antibodies in PBS for 1 h and DAPI/Hoechst (100 ng ml 1) for 510 min at RT. For immunocytochemistry, on day 21, the coverslips were washed in PBS four times before xing in 4% PFA (Sigma-Aldrich), then were blocked in 5% NGS in PBS. The cells were permeabilized in conjunction with the application of primary antibodies at RT for 45 min (2% NGS, 0.1% Triton X-100 in PBS, Sigma-Aldrich). Primary antibodies were used as follows: rabbit anti-Aldh1l1, 1:200 (Abcam, ab87117), goat anti-Arginase1, 1:100 (Abcam, ab60176), rabbit anti-activated caspase-3, 1:300 (Cell Signalling, P175), rat monoclonal anti-CD11b, 1:1,000 (Serotec, clone 5C6), rabbit anti-p-FAK, 1:1,000 (Life Technologies, 44624G), rabbit anti-GFAP, 1:500 (DAKO, Z0334), mouse monoclonal anti-GFAP, Cy3-conjugated, 1:500 (Sigma-Aldrich, clone FN-15), rabbit anti-IBA1, 1:500 (Wako, 019-19741), rat anti-Lamp2, 1:500 (Abcam, clone GL2A7), mouse anti-MAP-2, 1:100 (Sigma-Aldrich, clone HM-2), mouse monoclonal anti-NeuN, 1:200 (Milli-pore, clone A60), rabbit anti-PSD-95, 1:200 (Thermo Scientic, clone 6G6-1C9), mouse anti-S100, 1:500 (BD Biosciences, clone 19/S100B), rabbit monoclonal and mouse monoclonal anti-STAT3, 1:200 (Cell Signalling, clones 79D7 and 126H6, respectively), rabbit monoclonal anti-pSTAT3 (Tyr705), 1:100 (Cell Signalling, clone D3A7), mouse anti-SYT-1, 1:250 (Synaptic Systems, clone 41.1), mouse monoclonal anti-TSP-1, 1:50 (Santa Cruz, clone A6.1). This was followed by incubation with species-specic secondary antibodies for 30 min at RT. Images were taken either by light, uorescent or confocal microscopes (Leica). The specicity for all antibodies were pretested by the companies. At rst applications of primary antibodies, secondary antibodies alone were also used as negative controls. For analysis, either the Leica Application Suite software (Leica), Fiji or ImageJ v1.26 software was used.

Western blotting and ELISA. Western blots were performed according to standard protocols (Life Technologies). Electrophoresis was run in 412% gradient NuPAGE Novex Bis-Tris Pre-Cast Gels (Life Technologies) and followed by protein transfer to a PDVF membrane (Life Technologies) according to standard protocols. Membranes were then washed three times in PBS 1 with 0.2% Tween

at pH 7.4. Blocking was performed in PBS 1 , 0.2% Tween, 5% dry milk powder

(Marvel). Membranes were stained with the primary and secondary antibodies diluted in the same blocking solution. Primary antibodies used were as follows: mouse monoclonal anti-b-actin, 1:20,000 (Abcam), mouse monoclonal anti-GFAP, 1:500 (Sigma-Aldrich), rabbit monoclonal anti-STAT3, 1:1,000 (Cell Signalling), rabbit monoclonal anti-pSTAT3 (Tyr705), 1:1,000 (Cell Signalling), mouse monoclonal anti-TSP-1, 1:200 (Santa Cruz). Detection was performed by exposing the membrane to the Amersham ECL Prime Western Blotting Detection Reagents (GE Healthcare). Chemiluminescence on membranes was detected and imaged with the Alliance 4.7 CCD Image System (UVITEC).

To detect TSP-1 release by ACs, a sensitive chemiluminescent ELISA was used according to standard protocols provided by Thermo Scientic. Briey, the supernatants were ultracentrifuged rst to concentrate the content of molecules above 50 kDa, using Amicon Ultra tubes (Millipore). Ninety-six well plates were coated with either standard dilutions of TSP-1 (0.520 pg ml 1 range) or with the concentrated samples (1:5 volume) in the coating buffer (0.2 M Na carbonate/

bicarbonate, pH 9.4) at 4 C overnight. TSP-1 dilution served as positive control and coating with buffer alone represented negative control. Non-specic binding sites were blocked in 2% BSA in the standard diluent (0.1 M phosphate, 0.15 M Nachloride, 0.05% Tween 20, pH7.2; Sigma-Aldrich). For specic TSP-1 detection, a mouse TSP-1-specic antibody (1:1,000; Santa Cruz, clone A6.1) was applied, followed by a HRP-conjugated mouse-specic secondary antibody (1:50,000; Vector). The SuperSignal ELISA Femto chemiluminescent substrate (Thermo Scientic) was used for chemiluminescent detection, which was analysed by the GloMax plate-reader (Promega) at the wavelength of 420 nm. The antibodies were diluted in 1:5 strength of the standard diluent (50 ml total volume). The experiments included technical duplicates with biological triplicates.

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 13

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294

Image analysis for immunohistochemistry and western blots. In general at least 34 sections were taken for analysis 150 mm apart along the whole rostrocaudal axis of the FN. Cell counts, measurements of MAP-2 IR area fraction, intensity (optical density) of GFAP labelling and integrated density measurements for CD11b immunoreactivity were carried out in both axotomized and CL sides. LASAF (Leica) and ImageJ/Fiji applications were used to analyse areas of the entire FN in each tissue section according to stereotaxic parameters. The values were normalized to the non-axotomized CL side/background in the same section for accurate comparison between the groups. The methods for cell identication and counts for ACs, microglia, SCs and neurons have been described in the results. Western blot densitometry was performed by standard protocols using ImageJ/Fiji applications. Quantitative comparisons were performed based on normalization to density values of b-actin within the same blot.

Analysis of AC processes and synapse density. We used classical ultrastructural criteria21 for quantifying AC processes and synapses. Astrocytic processes were identied by their irregular, concave shape, glycogen granules and intermediate lament bundles in a relatively electron-lucent cytoplasm. Electron microscopic images (86112) were analysed per group, and the proportion of neuronal perimeter contacted by astrocytic endfeet was determined for 1214 neurons. To examine STAT3-induced cytoskeletal changes, GFAP immunoreactivity was analysed in vivo in AC processes and F-actin was visualized by phalloidin in cultured ACs. In axotomized FN, AC processes showing GFAP immunoreactivity were counted 15 mm away from the soma19. For in vitro analysis, we counted ACs with strong cortical F-actin bres, which were rounding up but kept connections with radial bres extending to lopodia or processes. This represented the activated phenotype in contrast to resting cells that displayed at morphology with mostly parallel laments.

Axosomal and axodendritic synapses were identied by the presence of presynaptic vesicles and electron-dense postsynaptic densities. Synapse density was expressed as number of synapses per 1 micrometre length along the cell body or dendritic membrane. For immunohistochemistry-based synapse quantication, we followed a previously developed method58. Briey, for quantication of synapse densities in the FN of mouse brains, four coronal sections per animal were immunostained for pre- and postsynaptic markers, synaptotagmin-1 and PSD-95, respectively. Then sections were scanned by confocal laser microscopy in 5 mm depth (optical section width 0.5 mm, 10 optical sections each). The parameters for scanning were set up for WT brain sections, and the same imaging parameters were used for transgenic animals. Merged single optical section images at 1 mm intervals were analysed using ImageJ (v1.0, NIH) for which the ImageJ puncta analyzer plug-in application was provided by Dr C. Eroglu70 (developed by B. Wark in ProfB. Barres laboratory). Threshold values were set to the CL side in each section. The number of colocalized pre- and postsynaptic puncta (ve optical planes per section; 20 images per brain) was counted. Average synaptic density per imaged area(0.01 mm2) was calculated for each condition, and was normalized to the non-affected CL side.

Electrophysiology. Whole-cell patch clamp recording of facial motor neurons was performed on coronal brainstem slices (225-mm thick) prepared from WT and CKO P2124 old male mice 2 weeks following axotomy. Mice were killed by decapitation, in accordance with UK animal usage legislation. Tissue samples for genotyping were collected during tissue preparation. Synaptic currents were analysed and tted using Electrophysiology Data Recorder V3.2.1 and Whole Cell Analysis Program V4.2.7 (Strathclyde Electrophysiology Software, Strathclyde University, UK), and were dened to occur if their amplitude was 41.5 times the s.d. of the current noise and their 1090% decay time was longer than their rise time. Solutions used in electrophysiological studies were obtained from Sigma-Aldrich and described as follows. Slices were prepared in bicarbonate-buffered solution containing (mM) 126 NaCl, 24 NaHCO3, 1 NaH2PO4, 2.5 KCl, 1 MgCl2,2 CaCl2, 1 Na-kynurenate to block glutamate receptors bubbled with 95% O2/5% CO2, pH 7.4. Electrodes contained internal solution comprising (mM) 130 Csgluconate, 4 NaCl, 0.5 CaCl2, 10 HEPES, 10 BAPTA, 4 MgATP, 0.5 Na2GTP, 2

K-Lucifer yellow, pH set to 7.3 with CsOH (ECl 88 mV). Pipette series resis

tance was 59 MO and electrode junction potentials were compensated. Slices were superfused at 241 C with HEPES-buffered solution containing (mM) 144 NaCl,2.5 KCl, 10 HEPES, 1 NaH2PO4, 2.5 CaCl2, 10 glucose, 0.1 glycine (to co-activate NMDA receptors), 0.005 strychnine (to block glycine receptors), pH set to 7.4 with NaOH, bubbled with 100% O2.

Statistical analysis. All experiments included both biological and technical replicates, and were repeated three times at least. For in vivo experiments, the subjects were not randomized and two animals have been excluded from analysis due to an unsuccessful surgical procedure. The assessor was blinded by another investigator to the allocation of assays or animals during outcome assessment. The sample size was estimated from pilot experiments. Data and graphs are presented as means.e.m, and n values refer to the number of cells, cultures, tissue samples or animals analysed per group. GraphPad Prism 5 (GraphPad Software) was used to generate graphs and to perform tests for distribution and statistical signicance. Data were analysed using two-tailed, unpaired t-test for comparison of two groups,

which was referred to as t-test in the text. One-way analysis of variance was applied for comparison of multiple groups and two-way analysis of variance was used to analyse independent variables in different groups. Bonferroni post hoc test was applied to analyse data pairs unless stated otherwise. Statistical signicance was accepted at P-values of o0.05. *P, **P, ***P indicate o0.05, o0.01, o0.001, respectively. Non-signicant P-values were labelled as NSP in the text.

References

1. Raisman, G. Neuronal plasticity in the septal nuclei of the adult rat. Brain Res. 14, 2548 (1969).

2. Raineteau, O. & Schwab, M. E. Plasticity of motor systems after incomplete spinal cord injury. Nat. Rev. Neurosci. 2, 263273 (2001).

3. Hardingham, G. E. Coupling of the NMDA receptor to neuroprotective and neurodestructive events. Biochem. Soc. Trans. 37, 11471160 (2009).

4. Tetzlaff, W., Graeber, M. B., Bisby, M. A. & Kreutzberg, G. W. Increased glial brillary acidic protein synthesis in astrocytes during retrograde reaction of the rat facial nucleus. Glia 1, 9095 (1988).

5. Aldskogius, H. & Kozlova, E. N. Central neuron-glial and glial-glial interactions following axon injury. Progr. Neurobiol. 55, 126 (1998).

6. Moran, L. B. & Graeber, M. B. The facial nerve axotomy model. Brain Res. Rev. 44, 154178 (2004).

7. Fawcett, J. W. & Asher, R. A. The glial scar and central nervous system repair. Brain Res. Bull. 49, 377391 (1999).

8. Bradbury, E. J. et al. Chondroitinase ABC promotes functional recovery after spinal cord injury. Nature 416, 636640 (2002).

9. Silver, J. & Miller, J. H. Regeneration beyond the glial scar. Nat. Rev. Neurosci. 5, 146156 (2004).

10. Okada, S. et al. Conditional ablation of Stat3 or Socs3 discloses a dual role for reactive astrocytes after spinal cord injury. Nat. Med. 12, 829834 (2006).

11. Faulkner, J. R. et al. Reactive astrocytes protect tissue and preserve function after spinal cord injury. J. Neurosci. 24, 21432155 (2004).

12. Herrmann, J. E. et al. STAT3 is a critical regulator of astrogliosis and scar formation after spinal cord injury. J. Neurosci. 28, 72317243 (2008).

13. Eroglu, C. et al. Gabapentin receptor alpha2delta-1 is a neuronal thrombospondin receptor responsible for excitatory CNS synaptogenesis. Cell 139, 380392 (2009).

14. Xu, J., Xiao, N. & Xia, J. Thrombospondin 1 accelerates synaptogenesis in hippocampal neurons through neuroligin 1. Nat. Neurosci. 13, 2224 (2010).

15. Lin, T. et al. Differential regulation of thrombspondin-1 and thrombospondin-

2 after focal cerebral ischemia/reperfusion. Stroke 34, 177186 (2003).

16. Liauw, J. et al. Thrombospondins 1 and 2 are necessary for synaptic plasticity and functional recovery after stroke. J. Cereb. Blood Flow Metab. 28, 17221732 (2008).

17. Graeber, M. B. & Kreutzberg, G. W. Astrocytes increase in glial brillary acidic protein during retrograde changes of facial motor neurons. J. Neurocyt. 15, 363373 (1986).

18. Schubert, K. O. et al. Activation of STAT3 signaling in axotomized neurons and reactive astrocytes after mbria-fornix transection. Exp. Brain Res. 165, 520531 (2005).

19. Wilhelmsson, U. et al. Redening the concept of reactive astrocytes as cells that remain within their unique domains upon reaction to injury. Proc. Natl Acad. Sci. USA 103, 1751317518 (2006).

20. Emery, B. & Barres, B. A. Unlocking CNS cell type heterogeneity. Cell 135, 596598 (2008).

21. Ventura, R. & Harris, K. M. Three-dimensional relationships between hippocampal synapses and astrocytes. J. Neurosci. 19, 68976906 (1999).

22. Boehler, M. D., Wheeler, B. C. & Brewer, G. J. Added astroglia promote greater synapse density and higher activity in neuronal networks. Neuron Glia Biol. 3, 127140 (2007).

23. Brown, C. E., Boyd, J. D. & Murphy, T. H. Longitudinal in vivo imaging reveals balanced and branch-specic remodeling of mature cortical pyramidal dendritic arbors after stroke. J. Cereb. Blood Flow Metab. 30, 783791 (2010).

24. Mller, J. C. et al. Regulation of thrombospondin in the regenerating mouse facial motor nucleus. Glia 17, 121132 (1996).

25. Cahoy, J. D. et al. A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J. Neurosci. 28, 264278 (2008).

26. Hamby, M. E. et al. Inammatory mediators alter the astrocyte transcriptome and calcium signaling elicited by multiple G-protein-coupled receptors.J. Neurosci. 32, 1448914510 (2012).27. Teng, T. S., Lin, B., Manser, E., Ng, D. C. H. & Cao, X. Stat3 promotes directional cell migration by regulating Rac1 activity via its activator betaPIX.J. Cell Sci. 122, 41504159 (2009).28. Xia, X. G., Hans-Dieter, H., Deller, T. & Kirsch, M. Induction of STAT3 signaling in activated astrocytes and sprouting septal neurons following entorhinal cortex lesion in adult rats. Mol. Cell Neurosci. 21, 379392 (2002).

14 NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5294 ARTICLE

29. Klein, M. A. et al. Impaired neuroglial activation in interleukin-6 decient mice. Glia 19, 227233 (1997).

30. Streit, W. J., Hurley, S. D., McGraw, T. S. & Semple-Rowland, S. L. Comparative evaluation of cytokine proles and reactive gliosis supports a critical role for interleukin-6 in neuron-glia signaling during regeneration. J. Neurosci. Res. 61, 1020 (2000).

31. Zigmond, R. E. Gp130 cytokines are positive signals triggering changes in gene expression and axon outgrowth in peripheral neurons following injury. Front. Mol. Neurosci. 4, 62 (2011).

32. Zhang, Y. & Barres, B. A. Astrocyte heterogeneity: an underappreciated topic in neurobiology. Curr. Opin. Neurobiol. 20, 588594 (2010).

33. Tsai, H.-H. et al. Regional astrocyte allocation regulates CNS synaptogenesis and repair. Science 337, 358362 (2012).

34. Pekny, M. & Nilsson, M. Astrocyte activation and reactive gliosis. Glia 50,

427434 (2005).

35. Myer, D. J., Gurkoff, G. G., Lee, S. M., Hovda, D. A. & Sofroniew, M. V. Essential protective roles of reactive astrocytes in traumatic brain injury. Brain 129, 27612772 (2006).

36. Kalla, R. et al. Microglia and the early phase of immune surveillance in the axotomized facial motor nucleus: impaired microglial activation and lymphocyte recruitment but no effect on neuronal survival or axonal regeneration in macrophage-colony stimulating factor-deci. J. Comp. Neurol. 436, 182201 (2001).

37. Fontana, X. et al. c-Jun in Schwann cells promotes axonal regeneration and motoneuron survival via paracrine signaling. J. Cell. Biol. 198, 127141 (2012).

38. Saraan, T. A. et al. Disruption of astrocyte STAT3 signaling decreases mitochondrial function and increases oxidative stress in vitro. PLoS ONE 5, e9532 (2010).

39. Shih, A. Y. et al. Cystine/glutamate exchange modulates glutathione supply for neuroprotection from oxidative stress and cell proliferation. J. Neurosci. 26, 1051410523 (2006).

40. Shinozaki, Y. et al. Microglia trigger astrocyte-mediated neuroprotection via purinergic gliotransmission. Sci. Rep. 4, 111 (2014).

41. Sofroniew, M. V. Molecular dissection of reactive astrogliosis and glial scar formation. Trends Neurosci. 32, 638647 (2009).

42. Jourdain, P. et al. Glutamate exocytosis from astrocytes controls synaptic strength. Nat. Neurosci. 10, 331339 (2007).

43. Filosa, A. et al. Neuron-glia communication via EphA4/ephrin-A3 modulates LTP through glial glutamate transport. Nat. Neurosci. 12, 12851292 (2009).

44. Henneberger, C., Papouin, T., Oliet, S. H. R. & Rusakov, D. A. Long-term potentiation depends on release of D-serine from astrocytes. Nature 463, 232236 (2010).

45. Pascual, O. et al. Astrocytic purinergic signaling coordinates synaptic networks. Science 310, 113116 (2005).

46. Lee, B., Butcher, G. Q., Hoyt, K. R., Impey, S. & Obrietan, K. Activity-dependent neuroprotection and cAMP response element-binding protein (CREB): kinase coupling, stimulus intensity, and temporal regulation of CREB phosphorylation at serine 133. J. Neurosci. 25, 11371148 (2005).

47. Lveill, F. et al. Suppression of the intrinsic apoptosis pathway by synaptic activity. J. Neurosci. 30, 26232635 (2010).

48. Genoud, C. et al. Plasticity of astrocytic coverage and glutamate transporter expression in adult mouse cortex. PLoS Biol. 4, e343 (2006).

49. Yang, Y. et al. Presynaptic regulation of astroglial excitatory neurotransmitter transporter GLT1. Neuron 61, 880894 (2009).

50. Eroglu, C. & Barres, B. A. Regulation of synaptic connectivity by glia. Nature 468, 223231 (2010).

51. Eleore, L., Vassias, I., Vidal, P.-P. & De Waele, C. Modulation of the glutamatergic receptors (AMPA and NMDA) and of glutamate vesicular transporter in the rat facial nucleus after axotomy. Neurosci. 136, 147160 (2005).

52. Van Damme, P. et al. Astrocytes regulate GluR2 expression in motor neurons and their vulnerability to excitotoxicity. Proc. Natl Acad. Sci. USA 104, 1482514830 (2007).

53. Emirandetti, A., Graciele Zanon, R., Sabha, M. & De Oliveira, A. L. R. Astrocyte reactivity inuences the number of presynaptic terminals apposed to spinal motoneurons after axotomy. Brain Res. 1095, 3542 (2006).

54. Theodosis, D. T., Poulain, D. A. & Oliet, S. H. R. Activity-dependent structural and functional plasticity of astrocyte-neuron interactions. Physiol. Rev. 88, 9831008 (2008).

55. Wake, H., Moorhouse, A. J., Miyamoto, A. & Nabekura, J. Microglia: actively surveying and shaping neuronal circuit structure and function. Trends Neurosci. 36, 209217 (2013).

56. Stephan, A. H., Barres, B. A. & Stevens, B. The complement system: an unexpected role in synaptic pruning during development and disease. Ann. Rev. Neurosci. 35, 369389 (2012).

57. Lo, F.-S., Zhao, S. & Erzurumlu, R. S. Astrocytes promote peripheral nerve injury-induced reactive synaptogenesis in the neonatal CNS. J. Neurophysiol. 106, 28762887 (2011).

58. Christopherson, K. S. et al. Thrombospondins are astrocyte-secreted proteins that promote CNS synaptogenesis. Cell 120, 421433 (2005).

59. Allen, N. J. et al. Astrocyte glypicans 4 and 6 promote formation of excitatory synapses via GluA1 AMPA receptors. Nature 486, 410414 (2012).

60. Tran, M. D. & Neary, J. T. Purinergic signaling induces thrombospondin-1 expression in astrocytes. Proc. Natl Acad. Sci. USA 103, 93219326 (2006).

61. Yonezawa, T. et al. Type IV collagen induces expression of thrombospondin-1 that is mediated by integrin alpha1beta1 in astrocytes. Glia 58, 755767 (2010).

62. Hall, E. D., Oostveen, J. A. & Gurney, M. E. Relationship of microglial and astrocytic activation to disease onset and progression in a transgenic model of familial ALS. Glia 23, 249256 (1998).

63. Simpson, J. E. et al. Microarray analysis of the astrocyte transcriptome in the aging brain: relationship to Alzheimers pathology and APOE genotype. Neurobiol. Aging 32, 17951807 (2011).

64. Ferraiuolo, L. et al. Dysregulation of astrocyte-motoneuron cross-talk in mutant superoxide dismutase 1-related amyotrophic lateral sclerosis. Brain 134, 26272641 (2011).

65. Garcia, A. D. R., Doan, N. B., Imura, T., Bush, T. G. & Sofroniew, M. V. GFAP-expressing progenitors are the principal source of constitutive neurogenesis in adult mouse forebrain. Nat. Neurosci. 7, 12331241 (2004).

66. Noble, M. & Murray, K. Puried astrocytes promote the in vitro division of a bipotential glial progenitor cell. EMBO J. 3, 22432247 (1984).

67. Hutchins, A. P., Poulain, S. & Miranda-Saavedra, D. Genome-wide analysis of STAT3 binding in vivo predicts effectors of the anti-inammatory response in macrophages. Blood 119, e110e119 (2012).

68. Cursiefen, C. et al. Roles of thrombospondin-1 and -2 in regulating corneal and iris angiogenesis. Invest. Ophthalmol. Vis. Sci. 45, 11171124 (2004).

69. Carulli, D. et al. Composition of perineuronal nets in the adult rat cerebellum and the cellular origin of their components. J. Comp. Neurol. 494, 559577 (2006).

70. Ippolito, D. M. & Eroglu, C. Quantifying synapses: an immunocytochemistry-based assay to quantify synapse number. J. Vis. Exp. 45, 18 (2010).

Acknowledgements

This work has been mainly supported by the Walker Fellowship (A.L.) at the University of Cambridge and by the Van Geest Fund (G.E.T.). We also received personal support by the Gates Scholarship (D.B., S.S.) and the Wellcome Trust Biomedical Vacation Scholarship (N.K.L., A.L.; RG60754). The authors are also grateful to Dr A. Tolkovsky for reading the manuscript and for her useful comments. We are grateful to Professor M. Sofroniew (UCLA, USA) for his advice and providing the STAT3-GFAP-CKO mice and to ProfessorG. Raivich for commenting on our work. We thank Dr C. Eroglu for the availability of the puncta analyzer software for synapse analysis (developed by B. Wark in Professor B. Barres laboratory) and to Dr J. Skepper for the help with electron microscopy.

Authors contributions

G.E.T. carried out part of the cell counting and analysis, western blots, qPCR, and performed the ChIP. S.S. conducted and analysed electrophysiological experiments guided by R.T.K. D.B. performed analysis on astrocyte and dendrite densities, and helped with the gures. K.L.A.-C. carried out the in vitro neuronal survival assay. N.K.L. and J.C.K. completed the in situ hybridization. C.Z. provided some of the transgenic mice and gave advice. R.J.M.F., R.T.K., J.W.F. provided feedback on the manuscript. A.L. developed the concept, directed the research, conducted and analysed experiments and wrote the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Tyzack, G. E. et al. Astrocyte response to motor neuron injury promotes structural synaptic plasticity via STAT3-regulated TSP-1 expression. Nat. Commun. 5:4294 doi: 10.1038/ncomms5294 (2014).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 5:4294 | DOI: 10.1038/ncomms5294 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 15

Word count: 13645

Show less

Abstract

Translate

The role of remote astrocyte (AC) reaction to central or peripheral axonal insult is not clearly understood. Here we use a transgenic approach to compare the direct influence of normal with diminished AC reactivity on neuronal integrity and synapse recovery following extracranial facial nerve transection in mice. Our model allows straightforward interpretations of AC-neuron signalling by reducing confounding effects imposed by inflammatory cells. We show direct evidence that perineuronal reactive ACs play a major role in maintaining neuronal circuitry following distant axotomy. We reveal a novel function of astrocytic signal transducer and activator of transcription-3 (STAT3). STAT3 regulates perineuronal astrocytic process formation and re-expression of a synaptogenic molecule, thrombospondin-1 (TSP-1), apart from supporting neuronal integrity. We demonstrate that, through this new pathway, TSP-1 is responsible for the remote AC-mediated recovery of excitatory synapses onto axotomized motor neurons in adult mice. These data provide new targets for neuroprotective therapies via optimizing AC-driven plasticity.

Details

Title

Astrocyte response to motor neuron injury promotes structural synaptic plasticity via STAT3-regulated TSP-1 expression

Author

Tyzack, Giulia E; Sitnikov, Sergey; Barson, Daniel; Adams-carr, Kerala L; Lau, Nike K; Kwok, Jessica C; Zhao, Chao; Franklin, Robin J M; Karadottir, Ragnhildur T; Fawcett, James W; Lakatos, András

Pages

4294

Publication year

2014

Publication date

Jul 2014

Publisher

Nature Publishing Group

e-ISSN

20411723

Source type

Scholarly Journal

Language of publication

English

DOI

https://doi.org/10.1038/ncomms5294

ProQuest document ID

1544367852

Astrocyte response to motor neuron injury promotes structural synaptic plasticity via STAT3-regulated TSP-1 expression

Jump to:

Full Text

Abstract

Details

Suggested sources