Regulation of RhoA by STAT3 coordinates glial scar formation

Scarring is a general tissue response after injury, which separates injured areas from healthy tissue and promotes wound healing. In response to various central nervous system (CNS) insults, lesion-proximal, reactive astrocytes form a glial scar, working in concert with secreted molecules found in the lesion (Burda et al., 2016). Seclusion of the injured area limits tissue damage, restricts inflammation, and preserves function in the subacute phase of CNS injuries (Okada et al., 2006; Sofroniew, 2015). The long-standing view that glial scars hinder further axonal regeneration after CNS injury was recently challenged by a study showing that glial scars can support axon outgrowth, even in chronic spinal cord injuries (Anderson et al., 2016). These studies underscore the clinical relevance of a better understanding of the cellular and molecular mechanisms involved in glial scar formation.

Signal transducer and activator of transcription–3 (STAT3) has been identified as a key factor in astrogliosis (Ceyzériat et al., 2016). Specific ablation of STAT3 in reactive astrocytes is associated with defective glial scars, deficient lesion seclusion, enhanced demyelination, and increased neuronal losses after spinal cord injury (SCI) in mice (Okada et al., 2006; Herrmann et al., 2008). However, although the seclusion of the various nonneural cells that invade lesion centers has been shown to rely on STAT3-dependent reorientation of astrocytic processes (Wanner et al., 2013), the molecular mechanisms underlying the effect of STAT3 remains unknown. In the present study, we focused on the molecular mechanisms by which reactive astrocytes form glial scars and the molecular effectors of STAT3 in reactive astrocytes dynamics.

STAT3 signaling promotes astrocyte migration in vitro (Okada et al., 2006) and promotes the seclusion of contusive spinal cord lesions by reactive astrocytes (Okada et al., 2006; Herrmann et al., 2008), but the underlying molecular mechanisms remain unclear. We sought to elucidate the role of STAT3 in controlling reactive astrocytes dynamics during glial scar formation by a classical migration study approach.

We first sought to determine whether matrix metallopeptidase-2 (MMP2), a major protease involved in tissue remodeling and a direct transcriptional target of STAT3 (Xie et al., 2004), is controlled by STAT3 in astrocytes. Real-time quantitative PCR analyses showed that the expression level of MMP2 was significantly reduced in astrocytes prepared from conditional knockout (KO) mice Nestin-Stat3^−/− (STAT3-CKO astrocytes) compared with WT mice, grown in the usual culture conditions (Fig. 1 A). Expression of MMP9, another critical protease in ECM remodeling activated by STAT3, was unchanged in STAT3-CKO astrocytes. Expression of MMP2 was also significantly less in STAT3-CKO astrocytes versus WT astrocytes after challenge with lysophosphatidic acid (LPA), or in vitro injury (wound scratch; Fig. 1 B), consistent with the role of STAT3 in astrocytes after injury. To confirm the role of STAT3 in regulating MMP2, we used gelatin zymography to assess levels of MMP2 in medium collected from WT astrocytes versus STAT3-CKO astrocytes in culture. This assay confirmed the significantly reduced secretion of MMP2 in STAT3-CKO astrocytes, compared with WT astrocytes (Fig. 1 C). To test whether reduced secretion of MMP2 impairs astrocyte migration, we examined the ability of STAT3-CKO astrocytes to migrate through Matrigel (BD). As a positive control, we also tested astrocytes from Mmp2^−/− mice. STAT3-CKO astrocytes exhibited defective invasion through Matrigel, and reduced invasion of MMP2-KO astrocytes indicated that MMP2 is necessary for invasion through Matrigel (Fig. 1 D). MMP2-KO astrocytes did not show any significant reduction in migration in a trans-well assay compared with WT astrocytes, indicating that MMP2 does not affect nonproteolytic migration in this assay (Fig. 1 E). Although the reduction of MMP2 secretion by STAT3-CKO astrocytes is a relevant observation in the context of SCI, where MMP2 promotes scar formation (Hsu et al., 2006), the noninvolvement of MMP2 in trans-well migration indicates that other effectors of STAT3 are involved in the proteolysis-independent movements of astrocytes.

STAT3 is known to stabilize the microtubules in various cells (Ng et al., 2006; Verma et al., 2009). Ablation of STAT3 in astrocytes did not result in the expected destabilization of microtubules (Fig. 2 A). Quantitative analysis showed that α-tubulin immunoreactivity (IR) was even moderately increased in STAT3-CKO astrocytes, for an unknown reason. The formation of glial scars involves the reorientation of astrocytic processes, a process which is defective in STAT3-ablated reactive astrocytes (Wanner et al., 2013). We therefore evaluated the response of astrocytes to an in vitro wound-scratch, an established model in studies of polarization (Etienne-Manneville, 2006). Surprisingly, centrosomes (means ± SEM; 65.1% ± 7.1 and 61.7% ± 5.1 for WT and STAT3-CKO, respectively) and Golgi apparatus (64.8 ± 6.2% and 60.2 ± 1.8%, respectively) showed normal polarization in STAT3-CKO astrocytes (Fig. 2 B), suggesting that the defective scarring by STAT3-ablated astrocytes is not a result of defective polarization. However, the protrusions formed in response to wounding were significantly shorter in STAT3-CKO astrocytes than they were in the WT counterparts (Fig. 2 C). In addition, the focal adhesions (FAs) of STAT3-CKO astrocytes were significantly longer than those of WT cells (Fig. 2 D). To evaluate FA functionality, we examined the migration of STAT3-CKO astrocytes on inhibition of FAK. Inhibition of FAK reduced the migration of WT astrocytes in a dose-dependent manner (ultimately down to the level of the reduced migration observed in STAT3-CKO), but the migration of STAT3-CKO cells was not significantly affected by PF573228, which suggests that FAK-dependent migration is constitutively defective in STAT3-CKO astrocytes (Fig. 2 E). However, the activation level of FAK, as assessed by the phosphorylation levels of tyrosines 397 and 576, was not reduced in STAT3-CKO astrocytes (Fig. 2 F), indicating that the defect in FAK-dependent migration is not due to a defect in FAK enzymatic activity. This led us to evaluate whether the effect of STAT3 on migration depends on the strength of cell-substratum adhesion (net adhesion). At any applied concentration of laminin, migration of STAT3-CKO astrocytes was significantly less than that it was for WT astrocytes (Fig. 2 G), which excludes the involvement of cell-adhesion strength, the effect of which on migration varies with the concentration of substrate ECM (DiMilla et al., 1991).

We then examined FA dynamics using time-lapse imaging of RFP-CrkI (Nagashima et al., 2002) in migrating astrocytes (Fig. 2 H and Video 1). Coordination between forward motion of the cells and FA dynamics appeared disrupted in STAT3-CKO astrocytes. In STAT3-CKO cells, FAs that formed at the tips of lamellipodia persisted beyond the lamella, whereas the FA in WT cells disassembled once they were located within a zone ∼20–25 µm from the leading edge. Quantitative analysis confirmed that the lifetime of the FA was significantly longer in migrating STAT3-CKO astrocytes (Fig. 2 I).

To investigate the possibility that FA disassembly is defective in STAT3-CKO astrocytes, we controlled the induction of FAs by treating serum-starved astrocytes with LPA (Ridley and Hall, 1992). LPA-induced FAs were longer in STAT3-CKO astrocytes than they were in WT astrocytes (Fig. 3, A and B). Colocalization of phospho-Y418-Src with paxillin and F-actin indicated that activated Src family kinases were concentrated in mature FAs in astrocytes (Fig. 3 C), which is consistent with their role in FA disassembly (Lee et al., 2010). These astrocytes were subsequently treated for an additional 45 min with an inhibitor of Src kinases, SU6656, to inhibit FA disassembly (Lee et al., 2010). As expected, Src inhibition resulted in significant elongation of FAs in WT astrocytes. In STAT3-CKO astrocytes, inhibition of Src did not result in FA elongation (Fig. 3, A and B), indicating that Src-dependent disassembly of FA is deficient in these cells. In a trans-well assay, inhibition of Src kinases consistently reduced the migration of WT astrocytes in a dose-dependent fashion but did not affect STAT3-CKO astrocytes (Fig. 3 D). Phospho-Src levels by immunoblotting showed that Src signaling is not reduced in STAT3-CKO astrocytes, compared with that in WT astrocytes (Fig. 3 E).

Actomyosin-generated tensions regulate FA dynamics (Ridley and Hall, 1992; Wolfenson et al., 2011), so we examined F-actin and actomyosin contractility in STAT3-CKO astrocytes. Astrocytes exhibit a high degree of morphological diversity under usual culture conditions, which makes analysis of stress fibers difficult. However, we found that astrocytes undergo stereotypical changes during the spreading that follows replating. In this context, the central stress fibers that formed during the spreading of STAT3-CKO astrocytes on laminin were much stronger than those observed in WT astrocytes (Fig. 4 A). Unlike the resistant, peripheral actin ring, these central stress fibers were sensitive to inhibition of RhoA effector kinase (ROCK) by H1152 (Fig. 4 B). Evaluation of actomyosin contractility in a collagen-gel assay showed that WT and STAT3-CKO astrocytes contract to the same extent on stimulation by TGF-α (Fig. 4 C). In both cell types, inhibition of ROCK by H1152 prevented such contraction, and even induced relaxation, which confirmed the specificity of the test. In the absence of contractile stimulation, STAT3-CKO astrocytes exhibited more basal actomyosin tonus than WT astrocytes did.

Given the role of the small GTPase RhoA in actin dynamics (Ridley and Hall, 1992), we next examined the activation of RhoA in STAT3-CKO astrocytes. We found that although RhoA expression level was significantly reduced in STAT3-CKO astrocytes (Fig. 4 E), the activation of RhoA was significantly increased in STAT3-CKO astrocytes under usual culture conditions (Fig. 4 D). Evaluation of downstream signaling indicated that this constitutive activation of RhoA is potent in STAT3-CKO astrocytes. The phosphorylation of myosin light chain-2 (MLC2) was increased during the usual culture conditions (Fig. 4 F). In addition, the phosphorylation of MLC2 on stimulation by leukemia inhibitory factor (LIF) was stronger and was delayed in STAT3-CKO astrocytes (Fig. 4 G), compared with WT astrocytes. Suppressor of cytokine signaling-3 (SOCS3), a negative regulator of LIF receptor signaling (Takahashi et al., 2003), was not induced in response to LIF in STAT3-CKO astrocytes, in contrast to the transient induction observed in WT ones (Fig. 4 G), which suggests that it may enhance the deregulation of the RhoA–ROCK pathway in STAT3-CKO astrocytes.

We next examined the role of RhoA deregulation in the defective migration of STAT3-CKO astrocytes. Although the inhibition of RhoA-GTPase by a cell-permeable C3 exoenzyme significantly reduced the trans-well migration of WT astrocytes, STAT3-CKO astrocytes appeared insensitive to that inhibition (Fig. 4 H). Inhibition of ROCK by either of two drugs, Y27632 or H1152, produced similar results; unlike WT cells, the migration of STAT3-CKO astrocytes was insensitive to ROCK inhibition (Fig. 4, I and J). Immunoblotting showed that strong inhibition of ROCK in astrocytes, as confirmed by the markedly reduced phosphorylation of MLC2, did not affect the phosphorylation level of Y705 in STAT3 (Fig. 4 K). This excludes the possibility that the reduction of WT astrocyte migration is due to the inhibition of STAT3 itself by ROCK inhibitors, as suggested in a previous study (Sanz-Moreno et al., 2011). Downstream targeting of the Rho/ROCK axis through inhibition of MLC kinase with ML-7 or myosin II ATPase with blebbistatin also indicated that the decreased migration of STAT3-CKO astrocytes is caused by deregulation of the RhoA–ROCK pathway (Fig. 4, L and M). These in vitro data suggest that constitutive activation of RhoA in STAT3-CKO astrocytes impairs actomyosin tonus, FA dynamics, and migration.

Next, we sought to identify the mechanisms by which RhoA is constitutively activated in STAT3-CKO astrocytes. Using quantitative real-time PCR, we first examined the expression levels of obvious candidates, i.e., guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs), that might contribute directly to the activation of RhoA in astrocytes. None of the GEFs examined were up-regulated (Fig. S1 A), and none of the GAPs examined were down-regulated (Fig. S1 B), in STAT3-CKO astrocytes.

We next examined ezrin, radixin, and moesin (ERM) proteins, which are involved in multiple aspects of cell migration (Clucas and Valderrama, 2014) and antagonize the activity of RhoA (Speck et al., 2003). Expression levels of these three proteins were not significantly modified in STAT3-CKO astrocytes compared with WT astrocytes. In contrast, immunoblotting with an antibody recognizing threonine-phosphorylated residues (Thr567 of ezrin, Thr564 of radixin, and Thr558 of moesin), which are known to have key roles in regulating ERM protein conformation and function, revealed a significant reduction in the phosphorylation of ezrin in STAT3-CKO astrocytes under usual culture conditions, i.e., without any specific experimental stimulus (Fig. 4 N). This suggests that the constitutive activation of RhoA in STAT3-CKO astrocytes is attributable to reduced phosphorylation of ezrin. GRK2, a kinase that phosphorylates ezrin (Cant and Pitcher, 2005), has been shown to positively regulate cell migration (Penela et al., 2009); we thus examined its expression level in astrocytes. Expression of GRK2 is not significantly changed in STAT3-CKO astrocytes versus WT astrocytes (Fig. 4 O). The reason for the reduced phosphorylation of ezrin in STAT3-CKO astrocytes is unknown.

Ablation of STAT3 in reactive astrocytes results in the defective seclusion of leukocytes and meningeal fibroblasts that infiltrate spinal cord lesions (Okada et al., 2006; Herrmann et al., 2008). As RhoA mediates the signaling of various repulsive pathways (Fujita and Yamashita, 2014), we wondered whether RhoA is necessary for the corralling of inflammatory cells by reactive astrocytes (Wanner et al., 2013). We observed that in vitro corralling of microglia and bone marrow–derived macrophages by astrocytes was lost on inhibition of the RhoA/ROCK pathway (Fig. 5 and Fig. S2), which suggests that deregulation of the RhoA/ROCK pathway in STAT3-CKO astrocytes affects not only the dynamics of reactive astrocytes themselves but also their ability to repel infiltrating cells within CNS lesions.

We next searched for in vivo evidence to support these in vitro findings.

We first speculated that a subset of elongated, reactive astrocytes located in the immediate vicinity of CNS traumatic lesions, which are known to express the intermediate filament protein Nestin in the acute phase after focal brain injury (Suzuki et al., 2012), might be more likely to contribute directly to glial scar formation than the wider population of GFAP-positive, reactive astrocytes. Examination of Nestin IR in a contusive SCI model confirmed that up-regulation of Nestin selectively identifies lesion-proximal, reactive astrocytes in the injured spinal cord (Fig. S3). To target this subpopulation of reactive astrocytes, we compared three lines of Nestin-EGFP reporter mice: Nestin-EGFP(A) (Kawaguchi et al., 2001), Nestin-EGFP(B) (Yamaguchi et al., 2000), and Nestin-CCE (Okada et al., 2006; Fig. 6). The conditional mouse line Nestin-CCE was particularly interesting. To generate Nestin-CCE mice, we crossed Nestin-Cre mice (Betz et al., 1996) with another transgenic line (Kawamoto et al., 2000) carrying a CAG promoter-loxP-CAT-loxP-EGFP (CAG-CAT^loxP/loxP-EGFP) reporter gene construct, which directs expression of EGFP on Cre-mediated recombination. Remarkably, in Nestin-CCE, there was no detectable fluorescence in the spinal cord of sham-operated animals, even in the area surrounding the central canal in which stem/progenitor cells are typically found (Fig. 6 A), in contrast with the two other Nestin mouse lines examined. Moreover, GFAP⁺ and Nestin⁺, which were reactive to astrocytes surrounding the lesion, were readily identified by strong EGFP fluorescence in injured Nestin-CCE animals (Fig. 6 B), suggesting that specific genetic targeting of lesion-proximal, reactive astrocytes occurred in Nestin-CCE mice.

We used this Nestin-Cre driver line to examine the effect of STAT3 deletion on the morphology of lesion-proximal, reactive astrocytes in the spinal cord. Using GFAP-Cre driver mice undergoing severe crush SCI, Sofroniew and colleagues have shown that STAT3-ablated, reactive astrocytes fail to hypertrophy (Herrmann et al., 2008) and that astrocytic processes fail to assemble in bundles (Wanner et al., 2013). In relation to the reduced length of α-tubulin protrusions that we observed after wound scratch in vitro (Fig. 2 C), we performed a wound stab in the dorsal spinal cord of adult Nestin-Stat3^−/− and WT mice to examine the appearance of reactive astrocytes located proximal to the lesion. Simultaneous immunostaining for both GFAP and AQP4 allowed clear visualization of the astrocyte processes in the injured area located in the dorsal white matter (Fig. 7). At 7 d postinjury (dpi), reactive astrocytes at the lesion site were clearly elongated with a few, thick processes extending toward the lesion site in WT mice (Fig. 7). In contrast, STAT3-ablated, reactive astrocytes appeared very different, with no obvious elongation and, therefore, no preferential orientation toward the lesion (Fig. 7). This in vivo observation is consistent with the morphological changes observed in vitro and with a possible modification of RhoA in the injured spinal cord of Nestin-Stat3^−/− mice.

This led us to examine the expression and activation of RhoA in vivo. We observed that RhoA was significantly down-regulated in the spinal cords of Nestin-Stat3^−/− mice at 5 d after contusive injury (Fig. 8 A), in agreement with our observation in primary astrocytes (Fig. 4 E). Furthermore, the quantification of the pool of active RhoA (RhoA-GTP) by G-Lisa assay showed an augmentation of activated RhoA in the injured spinal cords of Nestin-Stat3^−/− mice at 5 dpi but which did not reach statistical significance (Fig. 8 B). Immunostaining for phosphorylated-ERM proteins in injured spinal cords showed indistinct expression of p-ERM IR in both reactive astrocytes and nonastrocyte cells (i.e., based on their morphology, very likely leukocytes) within the lesion in WT mice (Fig. 8 C). In contrast, p-ERM IR appeared specifically reduced in the reactive astrocytes of Nestin-Stat3^−/− mice (Fig. 8 C). Z-stack analysis at high magnification by confocal microscopy confirmed that p-ERM IR was localized within both nonastrocytes cells and reactive astrocytes bordering the lesion in WT mice (Fig. 8 D and Video 2). In contrast, p-ERM IR was specifically reduced in the reactive astrocytes of Nestin-Stat3^−/− mice, whereas the signal was not affected in surrounding nonastrocyte cells (Fig. 8 D and Video 3). These observations suggest that the regulation of RhoA by STAT3 observed in primary astrocytes also occurs within reactive astrocytes in the context of SCI.

Phosphatase and tensin homologue (PTEN) controls astrocyte migration and inhibits the activation of Rac1 (Dey et al., 2008), an antagonist of RhoA. In various cell types, STAT3 represses PTEN through the expression of miR-21 (Iliopoulos et al., 2010), so we investigated whether PTEN is involved in the regulation of astrocytes dynamics by STAT3. Real-time quantitative PCR showed a significant reduction of miR-21 in STAT3-CKO astrocytes in comparison to WT astrocytes (Fig. 9 A). Consistent with derepression from the down-regulation of miR-21, PTEN levels were significantly increased in STAT3-CKO astrocytes compared with those of WT mice (Fig. 9 B). We then examined the activation of the protein kinase Akt, an indirect marker of PTEN phosphatase activity. The phosphorylation of Akt on Ser473 was consistently and significantly reduced in STAT3-CKO astrocytes. Immunocytochemistry indicated that PTEN is unambiguously distributed in the cytoplasm and nuclei of STAT3-CKO astrocytes in vitro (Fig. 9 C). Next, we examined regulation of PTEN by STAT3 in vivo. miR-21 is transiently up-regulated after SCI (Strickland et al., 2011), so we examined its expression in the injured spinal cords of WT and Nestin-Stat3^−/− mice at 5 dpi; however, we observed only a nonsignificant reduction of miR-21 in the contused spinal cords of Nestin-Stat3^−/− mice (Fig. 9 D). Although PTEN IR is very weak and sparse in the sham spinal cord, contusive SCI induced strong up-regulation of PTEN IR (Fig. 9 E). Confocal microscopy showed elevated neuronal expression of PTEN in both WT and Nestin-Stat3^−/− mice at 5 dpi (Fig. 9 F). In contrast, the up-regulation of PTEN in reactive astrocytes appeared stronger in Nestin-Stat3^−/− mice than it did in WT mice (Fig. 9 F). These data indicate that STAT3 represses PTEN expression, likely via miR-21, in reactive astrocytes in vitro and in vivo after SCI.

To examine whether PTEN repression is involved in the control of astrocyte migration by STAT3, we generated mice harboring conditional, heterozygous deletions of both STAT3 and PTEN in reactive astrocytes (Nestin-Stat3^+/−/Pten^+/−). Reduction of PTEN rescued the migration of STAT3^+/−/PTEN^+/− astrocytes in a trans-well migration assay (Fig. 9 G). Inhibition of PI3K by LY294002 resulted in a similar reduction of the migration of WT and STAT3-CKO astrocytes, suggesting that the lipid phosphatase activity of PTEN is not involved in the STAT3-dependent migration of astrocytes (Fig. 9 H). Analysis of LPA-induced FAs showed that PTEN reduction rescued the lengths of FAs in STAT3^+/−/PTEN^+/− astrocytes to levels resembling those in WT astrocytes (Fig. 9, I and J). Moreover, STAT3^+/−/PTEN^+/− astrocytes recovered their sensitivity to SU6656-induced FA elongation (Fig. 9, I and K). These data indicate that the regulation of FA turnover via PTEN is a major mechanism in the control of astrocytes dynamics by STAT3.

Finally, we examined whether PTEN had an influence on defective glial scar formation in Nestin-Stat3^+/− mice. In agreement with previous studies (Okada et al., 2006; Herrmann et al., 2008), we observed that the scar borders at 21 dpi are poorly defined in Nestin-Stat3^+/− mice (Fig. 10, A and B) compared with WT mice. In contrast, Nestin-Stat3^+/−/Pten^+/− mice harbor scar borders similar to those in WT mice, where the segregation between reactive astrocytes and lesion center is easily defined. Quantitative analysis confirmed that reduction of PTEN in Nestin-Stat3^+/− background rescues potently glial scar formation (Fig. 10 C). Examples of injury sites in other animals are presented in Fig. S4. Immunostainings for CD11b, a marker of inflammatory cells, and for fibronectin, a fibroblast marker, supported these observations because the interface between the lesion center and the reactive astrocytes was readily delineated in WT and Nestin-Stat3^+/−/Pten^+/− mice, compared with the diffuse border in Nestin-Stat3^+/− mice (Fig. 10 D).

Recent studies have dramatically revised our understanding of the role of reactive astrocytes. Lesion scarring by reactive astrocytes has been shown to provide benefits in the subacute phase (Bush et al., 1999; Faulkner et al., 2004) and, to some extent, in the chronic phase of CNS injury (Anderson et al., 2016), as well as in models of autoimmune disease (Voskuhl et al., 2009). Although the role of astrocytic STAT3 in glial scar formation has strong support (Okada et al., 2006; Herrmann et al., 2008; Wanner et al., 2013), the precise molecular mechanisms have remained elusive.

Although MMP2 is known to be a direct transcriptional target of STAT3 (Xie et al., 2004), to our knowledge, it had never been shown that STAT3 actually controls MMP2 expression in astrocytes. Our observation that MMP2 levels are reduced in the conditioned medium of STAT3-CKO astrocytes is relevant, given that MMP-2 promotes functional recovery after injury by regulating the formation of the glial scar and white matter sparing (Hsu et al., 2006). The unaltered trans-well migration of MMP2-KO astrocytes indicates that the regulation of astrocyte dynamics by STAT3 involves another mechanism.

In the present study, for the first time, to our knowledge, we report that STAT3 regulates the morphology and dynamics of FAs. FAs are considered to be important signaling hubs (Kuo et al., 2011), which suggests that regulation of FAs by STAT3, an important factor in several pathological contexts (e.g., inflammation and cancer), may have consequences in important biological processes, such as cell proliferation, differentiation, survival, and the regulation of gene expression. In STAT3-CKO astrocytes, the slower FA dynamics appear to rely on the impaired disassembly of FAs, which become insensitive to inhibition by Src. While Src and FAK are known to be involved in FA disassembly (Jones et al., 2000), they both exhibit normal levels of expression and phosphorylation in STAT3-CKO astrocytes. Even if FA disassembly does occur in cells that express constitutively active RhoA (Ezratty et al., 2005), it has previously been shown that actomyosin-generated tension controls the molecular kinetics of FAs (Wolfenson et al., 2011). This led us to examine stress fibers and actomyosin contractility in astrocytes. We found both to be elevated on ablation of STAT3.

We show that STAT3 controls the activation level of RhoA and that this is crucial to astrocyte dynamics and migration. This finding may account for the previous observation that inhibition of STAT3 potently reduces the migration rate only in cancer cells, in which invasion is dependent on actomyosin (Mierke et al., 2011). However, it appears that this scheme cannot be generalized to all cell types because loss of STAT3 expression in mouse embryonic fibroblasts, in contrast, leads to increased Rac1 activity (Teng et al., 2009), which antagonizes RhoA (Sander et al., 1999).

Beyond reactive astrocytes dynamics, the regulation of RhoA by STAT3 may take on additional significance in the context of SCI. After SCI, the activation of RhoA in both neuronal and glial cells is indeed rapid and sustained (Dubreuil et al., 2003). The activation of STAT3 observed after injury (Okada et al., 2006) may help to constrain the activation of RhoA. Because the inactivation of Rho–ROCK is required for the generation of Il-1β–induced reactive astrogliosis (John et al., 2004), the constitutive activation of RhoA when STAT3 is ablated is consistent with the observed, attenuated up-regulation of GFAP, failure of astrocyte hypertrophy, and pronounced disruption of astroglial scar formation after SCI in GFAP–STAT3-CKO mice (Herrmann et al., 2008). These findings were recently corroborated by the observation that the loss of STAT3 prevents or attenuates most genome-wide changes in astrocytes associated with astrogliosis and scar formation in WT mice (Anderson et al., 2016). Interestingly, John et al. (2004) reported that Il-1β–induced astrogliosis is associated with the disruption of FAs, deactivation of the contractile apparatus of the cell, and activation of ERM adapter proteins. We found that phosphorylation of ezrin is reduced in STAT3-ablated astrocytes, which is consistent with a previous study showing that ezrin regulates proper FA turnover (Hoskin et al., 2015). Our results suggest that ezrin, which is known to inhibit the activation of RhoA (Matsumoto et al., 2014), mediates the inhibition of RhoA by STAT3 in astrocytes.

Our observation that the RhoA–ROCK axis is required for the corralling of leukocytes by reactive astrocytes in vitro is also important because the neuroprotective and myelin-sparing effects of reactive astrocytes during the subacute phase after injury is thought to rely on this walling-off mechanism (Bush et al., 1999; Faulkner et al., 2004; Khakh and Sofroniew, 2015).

Additionally, we were able to identify lesion-proximal, scar-forming, reactive astrocytes by the expression of Nestin in the acute phase after SCI. We also extended the characterization using Nestin-Cre mice, in which recombination does not occur in stem/progenitor cells in the spinal cord. Recombination in the adult injured spinal cord of the Nestin-Cre mouse does not occur in Hu (ELAVL)-positive neurons and only exceptionally in some GST-π–positive cells (Okada et al., 2006), which may be a subpopulation of astrocytes, as suggested by Herrmann et al. (2008). These findings show that the Nestin-CCE mouse enables specific labeling of scar-forming reactive astrocytes in the spinal cord, and thus constitutes a unique tool for their analysis.

In various cell types, STAT3 represses PTEN expression, through miR-21 (Iliopoulos et al., 2010), and PTEN itself regulates astrocyte migration through Rac1 (Dey et al., 2008), an antagonist of RhoA (Sander et al., 1999). While the regulation of PTEN output by RhoA–ROCK has been shown previously (Lima-Fernandes et al., 2011), to our knowledge, regulation of RhoA by PTEN has not been reported. This led us to examine whether PTEN is involved in the STAT3-dependent signaling that regulates astrocytes dynamics. We observed that the reduction of PTEN could rescue FA dynamics and migration of STAT3^+/− astrocytes, as well as glial scar formation after SCI. This is consistent with previous studies showing that PTEN prevents glial scar formation. (Goncalves et al., 2015; Chen et al., 2016).

Together with the recent finding that glial scars may, to some extent, promote, rather than exclusively prevent, axon regeneration (Anderson et al., 2016), the deleterious action of astrocytic PTEN in glial scar formation suggests that the therapeutic value of targeting PTEN may extend beyond its direct neuronal action (Liu et al., 2010; Zukor et al., 2013).

Key functions of normal astrocytes, such as glutamate clearance in the synaptic cleft (Pannasch et al., 2014), or their ezrin-dependent contributions to structural, synaptic plasticity (Lavialle et al., 2011), are tightly coupled with changes in their morphology (Heller and Rusakov, 2015), suggesting that the present findings, which relate to the regulation of reactive astrocytes in the context of SCI, may also be relevant to the roles of astrocytes in subnormal CNS.

Control solutions (vehicle) used in the study included a 0.1% final concentration of DMSO equal to the amount used to dissolve the test reagents, when necessary. Aphidicolin, blebbistatin, H89, human LIF, LPA, PF573228, and TGF-α were obtained from Sigma-Aldrich. The Rho inhibitor CT04 (Cytoskeleton, Inc.), LY294002 (Promega), H1152 (Tocris Bioscience), Y27632 (Wako Pure Chemical Industries), and ML-7 and SU6656 were obtained from EMD Millipore. The following antibodies were used: mouse β-actin (A1978; Sigma-Aldrich), rat CD11b (MCA711G; Bio-Rad Laboratories), rabbit polyclonal β-COP (PC175; EMD Millipore), mouse fibronectin (610077; BD), rat GFAP (130300; Invitrogen), chicken nestin (NES; supplied by Hiroaki Kanki, AVES Labs, Inc., Tigard, OR), mouse paxillin (610568; BD), rabbit polyclonal pericentrin (PRB-432C; Covance), mouse monoclonal PTEN (559600; BD) for immunoblotting, rabbit monoclonal PTEN (9559; Cell Signaling Technology) for immunofluorescence, mouse monoclonal RhoA (Cytoskeleton, Inc.), rabbit polyclonal SOCS3 (a gift from A. Yoshimura, Keio University School of Medicine, Tokyo, Japan; 18395; IBL-America), mouse monoclonal nonphospho-Y416-Src (2101; Cell Signaling Technology) for blotting and rabbit polyclonal p-Y418-Src (44660G; Invitrogen) for immunostaining, and mouse α-tubulin (T9026; Sigma-Aldrich). Rabbit polyclonal Akt (9271), p-S473-Akt (9277), ERM (3142), p-ERM (3276), FAK (3285), p-Y576-FAK (3281), p-Y397-FAK (3283), GAPDH (2118), myosin light chain-2 (3672), p-S19-MLC (3671), STAT3 (9132), and p-Y705-STAT3 (9131) were all obtained from Cell Signaling Technology. Secondary antibodies used for immunostaining were conjugated to Alexa488, 555, or 647 (Molecular Probes). Peroxidase-conjugated secondary antibodies for Western blotting were obtained from Jackson ImmunoResearch Laboratories, Inc. Texas red–conjugated Tomato lectin was from Vector Laboratories. Cell nuclei were stained using Hoechst 33258 (Thermo Fisher Scientific).

The MMP2-KO mice (Itoh et al., 1997) were a gift from Shigeyoshi Itohara (Institute of Physical and Chemical Research BioResource Center [RIKEN BRC], Ibaraki, Japan). The PTEN^LoxP mice (Suzuki et al., 2001) were a gift from Tak W. Mak (University Health Network, Toronto, Canada) and were provided by the RIKEN BRC through the National BioResource Project of the Ministry of Education, Culture, Sports, Science and Technology (Tokyo, Japan). Nestin-EGFP(A) (Kawaguchi et al., 2001), Nestin-EGFP(B) (Yamaguchi et al., 2000), Nestin-CCE (Okada et al., 2006), and STAT3^loxP/loxP mice were previously reported (Okada et al., 2006). In brief, in Nestin-Cre mice (Meiji Institute of Health Science, Odawara, Kanagawa, Japan) Cre-recombinase is expressed under the control of the 5.8 Kb Nes gene promoter and the 1.8-kb rat second-intronic enhancer (Johansson et al., 2002). To generate Nestin-CCE mice, the Nestin-Cre mice were crossed with another transgenic line that carried the CAG promoter-loxP-CAT-loxP-EGFP (CAG-CAT^loxP/loxP-EGFP) reporter gene construct (a gift from J. Miyazaki, Osaka University Graduate School of Medicine, Osaka, Japan; Kawamoto et al., 2000), which directs expression of EGFP upon Cre-mediated recombination.

Primary astrocytes were prepared as previously described (Renault-Mihara et al., 2011) from cortices and striata of 1- or 2-d-old mice. Nestin-Cre; STAT3^+/−/PTEN^+/− astrocytes were obtained by mating Nestin-Cre⁺PTEN^flox/WT mice with STAT3^flox/flox mice. All mice used in this study had the C57BL/6J genetic background.

Isolation of microglial cells was adapted from a previous protocol (Calvo et al., 2005). In brief, cerebral hemispheres from newborn C57BL/6J mice were dissociated by trituration after removal of the meninges. Cells were seeded in DMEM/F12 supplemented with 33 mM glucose, 2 mM l-glutamine, 20 mM NaHCO₃, 5 mM Hepes, pH 7.4, 50 IU/ml penicillin, and 5 µg/ml streptomycin and containing 10% heat-inactivated FBS. The culture medium was changed on days 1 and 3. On day 8, fresh medium was supplemented with 20% medium conditioned by L929 cells. Microglial cells were harvested by gently flushing the medium with a fire-polished Pasteur pipette onto the surface of the cultures to detach cells adhering to the astrocyte monolayer. Immunostaining with anti-Iba1 antibody confirmed the purity of microglial preparations.

Bone marrow–derived macrophages were obtained as follows: femurs and tibias from adult C57BL/6J mice were dissected and crushed with a pestle. The crushed bones were washed five times in HBSS⁺ (supplemented with 2% FBS, 10 mM Hepes, and 1% penicillin/streptomycin), and the cell suspensions were filtered through a 70-µm cell strainer (Falcon 352350; Corning) to remove debris and bone fragments and were collected by centrifugation at 280 g for 7 min at 4°C. The pellet was immersed in 1 ml water for 5–10 s to burst the RBCs; after which, 1 ml of 2× PBS containing 4% FBS was added, and the suspension was filtered again through a cell strainer. Pelleted cells were resuspended in DMEM/F12 supplemented with 33 mM glucose, 2 mM l-glutamine, 20 mM NaHCO₃, 5 mM Hepes, pH 7.4, 50 IU/ml penicillin, and 5 µg/ml streptomycin and containing 10% heat-inactivated FBS. On day 4, nonadherent cells were collected and, after centrifugation, were resuspended in fresh medium supplemented with 20% medium conditioned by L929 cells to induce their differentiation into bone marrow–macrophages. 3 d later, immunostaining for CD11b confirmed the purity of the macrophages prepared.

Total RNA was isolated from primary astrocytes using TRIzol (Invitrogen), according to the manufacturer’s protocol, and was further purified with the RNeasy Mini kit (QIAGEN). Total RNA was used to synthesize cDNA with 500 ng oligo-dT primers. The cDNA synthesis was performed at 42°C for 50 min in a final volume of 20 µl using Superscript III reverse transcription (Invitrogen) followed by RNase H (Invitrogen). Real-time RT-PCR was performed using a ViiA 7 real-time PCR system (Thermo Fisher Scientific) with SYBR green (Takara Bio Inc.). In brief, the parameters were 40 cycles at annealing temperature 60°C with a hold time of 1 min, and a ramp rate of 1°C/s. The primers used are listed in Table S1. The specificity of the primers, which had been validated by the primer bank project (https://pga.mgh.harvard.edu/primerbank/), was further confirmed by examining amplification and melt curves of reactions from the astrocyte cDNA template. Analysis was performed with ViiA 7 analysis software using the automatic thresholding function, with GAPDH as an internal control. Five biological replicates per phenotype (i.e., WT versus STAT3-CKO) were run in triplicates, in at least two independent experiments.

Gelatin zymography was performed as described previously (Xie et al., 2004). In brief, astrocytes were replated at the same density after 12 d in vitro and were then treated with 50 ng/ml TGF-α for 24 h in serum-free medium. The collected medium was further concentrated using a Centricon filter (EMD Millipore) and was then loaded into a gelatin gel that was further digested for 45 h at 37°C and was ultimately revealed using Coomassie staining.

For the Matrigel invasion assay, 100 µl of Matrigel matrix (BD) at 1 mg/ml was gelled for 5 h at 37°C in trans-well inserts with 8-µm pores polyethylene terephthalate filters (BioCoat; BD). Astrocytes were trypsinized in vitro on day 13, stained with trypan blue, and counted using a hemocytometer. A total of 10⁵ cells resuspended in MEM–F12 complete medium containing 1% FBS, 10 ng/ml TGF-α, and 10 µg/ml aphidicolin (Sigma-Aldrich) were allowed to migrate toward the same culture medium supplemented with 50 ng/ml TGF-α and 5 µg/ml fibronectin (Sigma-Aldrich), in the lower chamber. After 48 h at 37°C, the filters were fixed for 20 min at RT with PFA (4% in PBS, pH 7.5), and the nonmigrated cells were removed by wiping the upper side of the membranes with cotton swabs. Cell nuclei were stained using Hoechst 33258, and the filters were mounted on slides in Fluoromount medium (Diagnostic BioSystems). Using AxioVision software (ZEISS) connected to an epifluorescence microscope (Axioplan 2; ZEISS), 12 fields per membrane were captured, and the cells were counted manually. Invasion index was calculated by normalizing the counts to the WT values that were obtained within the same experiment.

Polyethylene terephthalate filters with 8-µm pores (BioCoat; BD) were coated with 5 µg/ml mouse laminin (Thermo Fisher Scientific) for 3 h at 37°C. Astrocytes were trypsinized on day 13 in vitro, stained with trypan blue, and counted using a hemocytometer. Astrocytes were resuspended in MEM–F12 complete medium containing 1% FBS and 10 µg/ml aphidicolin and were preincubated for 20 min in an Eppendorf tube with inhibitors before seeding 3 × 10⁴ cells into the trans-well. The cells were allowed to migrate toward the same culture medium supplemented with 10% FBS in the lower chamber. After 15 h at 37°C, the filters were fixed for 20 min at RT with PFA (4% in PBS, pH 7.5), and the nonmigrated cells were removed by wiping the upper side of the membranes with cotton swabs. Cell nuclei were stained using Hoechst 33258, and the filters were mounted on slides in Fluoromount medium. Using AxioVision software connected to an epifluorescence microscope (Axioplan 2), 12 fields per membrane were captured, and the cells were counted manually. The migration index was calculated by normalizing the counts to control values that were obtained within the same experiment.

In vitro astrocytes were replated at 14 d onto culture plates coated with 5 µg/ml laminin; 15 h later, cells were fixed with 4% PFA, and then immunostained for α-tubulin. Pictures were acquired at RT using 20× objective (Plan-Neofluar, NA 0.5) on the Axioplan-2 epifluorescence microscope (ZEISS). Cell outlines were manually traced using ImageJ software, and the mean densitometry was determined in each cell (n = 221 and 241 WT and STAT3-CKO, respectively, from 31 and 37 fields of view from three independent experiments).

Scratch-induced polarization assay was performed as previously described in detail (Etienne-Manneville, 2006) using astrocyte monolayers prepared from WT and Nestin-Stat3^−/− P1 pups.

Primary astrocytes at 13 d in vitro were replated in normal culture medium (i.e., with 10% FBS) at low density into 8-well chambers (glass bottom; Iwaki Co.) that were previously coated with 5 µg/ml of laminin; 36 h later, cells were fixed and then immunostained for paxillin. Pictures were acquired at RT using an LSM 700 confocal microscope (ZEISS) with 63× objective (ZEISS; C-Apochromat, water immersion, 1.2 W). FA length was measured using ImageJ software (version 1.48) after manual calibration.

FA elongation was induced as previously described (Lee et al., 2010). In brief, astrocytes at 13 d in vitro were replated in normal culture medium (i.e., with 10% FBS) at low density into 8-well chambers (glass bottom; Iwaki) coated with 5 µg/ml of laminin. About 24 h after replating, the cells were serum starved for 14 h. Astrocytes were then treated with 10 µM LPA for 1 h in serum-free medium. In the indicated groups, SU6656 and LPA were added for the last 45 min, with final concentrations of 10 µM for both reagents. Cells were then fixed, stained, and processed as indicated in the previous section.

Primary astrocytes were transfected with RFP-CrkI (a gift from N. Mochizuki, National Cerebral and Cardiovascular Center, Osaka, Japan; Nagashima et al., 2002) on day 11 using Amaxa nucleofector (Lonza) and replated the next day into an 8-well, glass-bottom chamber (Iwaki) that was coated by 5 µg/ml laminin; 2 d later, and 2 h after a wound was made in the astrocyte monolayer using a P10 pipette tip, the fluorescent cells at the edge of the wound were recorded using a confocal microscope (LSM5, Pascal Exciter; ZEISS) that was equipped with a heat- (37°C) and gas-controlled incubation chamber (5% CO₂; Tokai Hit), which was coupled to a heated-motor stage. The objective lens (EC Plan-Neofluar, 10×, NA 0.3) was maintained at 37°C and was used to acquire a Z-stack, time-lapse series with the following parameters: zoom 4× every 15 min, seven Z-stacks (every 5 µm). Z-projections were performed with the Image Browser software (ZEISS). The formation of FA at the leading edge was tracked until disappearance using the Manual Tracking plugin of the ImageJ software (1.48). The data were collected in six different recording sessions using three independent, primary cultures of astrocytes.

After treatment, cells were rinsed twice with ice-cold PBS, scraped into MAPK lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% NP-40, 2 mM EDTA, 2 mM EGTA, and 2 mM sodium orthovanadate, supplemented with phosphatase inhibitor cocktails 1 and 2 (Sigma-Aldrich) and Mini Complete protease inhibitor cocktail without EDTA (Roche) on ice, passed three times through a 26-gauge needle, and then centrifuged at 13,000 g for 20 min at 4°C. Protein concentration was measured using a Micro BCA protein assay (Thermo Fisher Scientific), and the supernatants were mixed with sample buffer (Laemmli), boiled, and separated on polyacrylamide gels. After protein transfer to a polyvinylidene fluoride membrane (Immobilon-P; EMD Millipore), the blots were probed with the indicated primary antibodies, visualized with HRP-conjugated secondary antibodies, and developed using an enhanced chemiluminescence kit (ECL; GE Healthcare). Band intensity was quantified using an LAS-3000 imaging system (Fujifilm) and Multi Gauge software (Fujifilm).

Protein lysates from spinal cords were prepared as follows: mice were deeply anesthetized and perfused with ice-cold heparinized saline, and the spinal cords were quickly dissected and removed; 4-mm-long portions of the spinal cord from the lesion epicenter were snap-frozen in liquid nitrogen. Spinal cords were then dissociated into QIAzol lysis reagent (QIAGEN) using a sterile pilar and homogenization with P1000 micropipette (Gilson, Inc.). The proteins were then extracted using the Isogen procedure. In brief, proteins were precipitated from the organic phase using isopropanol, followed by treatment with guanidine hydrochloride and ethanol. Ultimately, the pellets were resuspended for 10 min at 55°C in SDS 1% lysis buffer containing proteases and phosphatases inhibitors. The collected supernatants were then separated by electrophoresis as described for the cellular lysates.

Gel-contraction assay was performed as previously described (Iwabu et al., 2004). In brief, astrocytes at 14 DIV were trypsinized and, after serial washes, were ultimately resuspended at 2 × 10⁶ cells/ml in ice-cold, serum-free medium containing 1 mg/ml BSA. Astrocytes were then incubated with 20 µM ML-7 for 30 min on ice. Cell suspensions were then diluted 1:1 with ice-cold, neutralized collagen solution (final concentration of cell matrix collagen-type IA of 1 mg/ml), and 50 ng/ml TGF-α was added. 500 µl of this mixture (corresponding to 500,000 astrocytes) were dispensed in multiple 24-well plates and left to polymerize for 60 min at 37°C. Each matrix was then covered with 1 ml of serum-free medium containing 1 mg/ml BSA and eventually 50 ng/ml TGF-α and/or 20 µM ML-7. The compaction of the matrices was evaluated by weighting them after a 24-h incubation period at 37°C.

The activation of RhoA was assessed using two different techniques. The classical RhoA–GTP pull-down assay was performed with a kit from Thermo Fisher Scientific, with 900 µg of protein lysates for each replicate. RhoA constitutive activation in STAT3-CKO astrocytes was confirmed with an ELISA kit, according to manufacturer indications (G-Lisa assay, Cytoskeleton, Inc.); 35 µg of total proteins were used per assay. In vivo assessment of RhoA activation was performed as follows: mice were deeply anesthetized and perfused with ice-cold heparinized saline, and the spinal cords were dissected. A 4-mm-long portion of the spinal cord from the lesion epicenter was crushed using a BioMasher (Wako Pure Chemical Industries) in G-Lisa lysis buffer, according to the manufacturer’s specification; sonicated twice in ice; and spun at 15,000 rpm for 5 min. The collected supernatants were distributed in vials and snap-frozen in liquid nitrogen. RhoA–GTP G-Lisa assay was performed with 50 µg of total proteins per assay. RhoA–GTP was normalized to the total RhoA expression level, which was determined by Western blotting.

Mouse astrocytes grown in 10% FBS were replated onto laminin on day 13 and were then gradually switched to 10% horse serum, as previously described (Wanner et al., 2013). On day 21, astrocytes and leukocytes (either microglia or bone marrow–derived macrophages. The protocol for their obtention is described in the Cell culture section) were co-cultured in 10% FBS-containing medium in the presence of the vehicle, 2 µg/ml CT04 or 2 µM H1152. On day 24, the co-cultures were fixed and stained with GFAP antibody, Alexa Fluor 488 dye–conjugated phalloidin and Texas red–conjugated-Tomato lectin (lectin from Lycopersicon esculentum). Images were acquired with 10× objective (Nikon Plan Apochromat, NA 0.45) using the Z-stack and tile functions of KEYENCE BZ-9000 microscope operated with a monochrome camera. BZ-II (version 1.31) analyzer software (KEYENCE) was used to obtain full-focus, tiled images.

The protocol for contusive SCI was previously described (Renault-Mihara et al., 2011). All surgical and animal-care procedures were performed in accordance with the Laboratory Animal Welfare Act and the Guide for the Care and Use of Laboratory Animals (U.S. National Institutes of Health) and were approved by the animal experimentation committee of Keio University School of Medicine (approval 13020-1). In brief, adult mice (8 wk old) were anesthetized with an i.p. injection of 100 mg/kg ketamine and 10 mg/kg xylazine. The dorsal surface of the dura mater was exposed through a laminectomy at the tenth thoracic vertebra, and SCI was induced using an Infinite Horizon Impactor (60 kilodynes; Precision Systems and Instrumentation), as previously described (Scheff et al., 2003); 1 ml of saline was s.c. injected for rehydration at the end of the procedure, and mice were kept on a 37°C warming table until awakening from anesthesia. The mice were then returned to their cages and given free access to water and food.

Adult mice (8 wk old) were anesthetized with an i.p. injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). The dorsal surface of the dura mater was exposed through a laminectomy at the tenth thoracic vertebra. A 2–3-mm-long, superficial, midline incision was made along the dorsal column of the spinal cord using a surgical blade (Feather blade 11). The injured spinal cords were gently washed with saline before closing the lesion. 1 ml of saline was s.c. injected for rehydration at the end of the procedure. The mice were sacrificed 7 d after injury. Three WT and three Nestin-Stat3^−/− mice were injured; one WT mouse died shortly after surgery, likely from hemorrhage before the wound stab.

At the indicated times after SCI, the mice were deeply anesthetized and transcardially perfused with ice-cold heparinized saline containing 50 mM NaF (Nacalai), followed by 4% PFA containing 50 mM NaF. The spinal cords were removed and immersed overnight in 4% PFA at 4°C, followed by 24 h in 10% sucrose and 24 h in 30% sucrose at 4°C. The spinal cords were then embedded in optimal cutting temperature (OCT) compound and sectioned with a cryostat (CM3050S; Leica Biosystems). 20-µm-thick, frozen sections were washed in PBS and then blocked for 30 min with a solution of PBS containing 0.4% Triton X-100 and 1.5% FBS. Primary antibodies diluted within this solution were applied overnight at 4°C. Solutions of secondary antibodies were applied for 45 min at RT. Slides were mounted using PermaFluor mounting medium (Thermo Fisher Scientific). Negative controls for immunostaining (no primary antibodies) were systematically performed and were used for setting the microscope parameters.

Except for live imaging, all images were acquired at RT. Fig. 7 A shows a Z-stack tile series of pictures acquired with the KEYENCE BZ-9000 microscope, processed with full focus function of the Bz-II Analyzer software. Fig. 7 B shows ZEISS LSM700 confocal microscope images (objective 40×, water immersion, C-Apochromat, 1.2 W). Fig. 8 C shows confocal imaging from a ZEISS LSM 700 microscope with a 10× objective (Plan-Apochromat, NA 0.45). Fig.8 D shows images that correspond to a single plane extracted from 1-µm step, Z-stack series, acquired using the ZEISS LSM 700 confocal microscope with a 40 × objective (water immersion, C-Apochromat, 1.2 W). Fig. 9 E shows a Z-stack tile series of images acquired with a KEYENCE BZ-9000 microscope, which were processed with the full-focus function of the Bz-II Analyzer software. Fig. 9 F shows images acquired with the ZEISS LSM700 (objective 20×, Plan-Apochromat, NA 0.8). Fig. 10 (A and D) provides images acquired with a KEYENCE BZ-9000 microscope (objective 20×, Nikon, Plan-Apochromat, NA 0.75).

microRNAs were purified with the miRNeasy Mini kit (QIAGEN) from QIAzol lysates. Quantitative analysis was performed using a TaqMan assay 000397 for miR-21 and 001232 for normalization to snoRNA202 (Invitrogen) and either a 7900HT real-time PCR cycler (analysis in primary astrocytes) or a ViiA 7 real-time PCR system (Thermo Fisher Scientific; for spinal cord samples).

For each spinal cord analyzed, five 20-µm-thick sagittal sections were immunostained with anti-GFAP and AQP4 antibodies. Tile pictures were acquired with a KEYENCE BZ700 microscope, with 10× objective (CFI Plan Apochromat, NA 0.45). Fluorescences for GFAP (Alexa Fluor 488) and AQP4 (Alexa Fluor 555) were summed using Adobe Photoshop CC2015, as previously described (Wanner et al., 2013), to compensate for the down-regulation of GFAP in Nestin-STAT3^+/− mice. For each section analyzed, 15 radial lines 300 µm long were evenly distributed along the entire perimeter of the lesion and centered on the interface (150 µm inside the lesion, 150 µm inside the tissue) and were manually traced using ImageJ software. Blind scoring of the glial scar borders at the intersection with those lines (whether a neat border or not) generated the mean score per each section (e.g., 10 neat borders of 15; i.e., 66.6%). Five mice per group (= 25 sections) were analyzed.

All analyses were performed using GraphPad Prism software version 5.00 (GraphPad Software), and all tests were two-tailed; significance is indicated as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001. Unless otherwise specified, values represent the means ± SEM of at least three independent experiments that were performed at least in triplicate. Replicates correspond to biological replicates, except notably, for the gel contraction assay and the trans-well migration assays, where technical replicates were made from a common cell suspension, which was obtained by pooling the cells from several culture dishes originating from multiple animals. Migration experiments were analyzed using two-way ANOVA and Bonferroni multiple comparison test. Where n < 10 or the data were nonnormally distributed, we used nonparametric Mann-Whitney U tests. Where n ≥ 10 for normally distributed data, an unpaired t test or ANOVA (for more than two variables) were used for analysis. Normality of the distribution of the data was tested with d'Agostino and Pearson omnibus, Shapiro-Wilk, and Kolmogorov–Smirnov normality tests using the column statistics function of GraphPad Software.

Fig. S1 shows the expression levels of candidate molecules for the regulation of RhoA activation in STAT3-CKO astrocytes. Fig. S2 shows the corralling of bone marrow–derived macrophages by astrocytes, which relies on RhoA–ROCK pathway. Fig. S3 shows up-regulated Nestin, which identifies the lesion-proximal, reactive astrocytes after SCI. Fig. S4 describes the genetic reduction of PTEN in Nestin-Stat3^+/− mice, which rescues the glial scar border formation after contusive injury. Table S1 lists the sequences of the primers used for gene expression analysis. Video 1 provides time-lapse, confocal imaging of RFP-CrkI–expressing, migrating astrocytes. Video 2 shows a confocal, 1-µm step, Z-stack series of the immunostaining for p-ERM and GFAP in the injured spinal cord of WT mice (Fig. 8 D), which is animated at 2 frames/s. Video 3 provides a confocal, 1-µm step, Z-stack series of the immunostaining for p-ERM and GFAP in the injured spinal cord of Nestin-Stat3^−/− mice (Fig. 8 D), animated at 2 frames/s.

We are grateful to Drs. Tak W. Mak, Shigeyoshi Itohara, and Junichi Miyazaki for providing PTEN^loxP, MMP2-KO, and CAG-CAT^loxP/loxP-EGFP mice, respectively. We also thank Hiroaki Kanki, Akihiko Yoshimura, and Naoki Mochizuki for the gifts of Nestin antibody, SOCS3 antibody, and the RFP-CrkI construct, respectively. We are grateful to Tokuko Harada, Fumiko Ozawa, Shiho Nakamura, and Satoe Banno for mouse care. We thank Okano Laboratory (Shinjuku-ku, Tokyo, Japan) members, Professor Yoshiaki Toyama, and members of the SCI research group. We thank Douglas Sipp (Keio University, Shinjuku-ku, Tokyo, Japan) for invaluable comments on the manuscript.

This work was supported by fellowships from the Japan Society for Promotion of Science (P07451), as well as grants-in-aid for scientific research from Japan Society for Promotion of Science (Kakenhi grant WakateB 22700342) and Ministry of Education, Culture, Sports, Science and Technology (Kakenhi grant number WakateB 16K21360), Keio University (Sakaguchi fund), the Naito Foundation, and a medical research grant on traffic accidents from the General Insurance Association of Japan (to F. Renault-Mihara).

H. Okano is a paid scientific advisor of SanBio, Co., Ltd., and K Pharma, Inc., and M. Nakamura is a paid scientific advisor of K Pharma, Inc.

The authors declare no further competing financial interests.

Author contributions: F. Renault-Mihara conceived the study, designed and carried out the experiments, analyzed the data, and wrote the manuscript. H. Kumamaru and S. Okada characterized the different strains of Nestin reporter mice. F. Renault-Mihara, M. Mukaino, M. Shinozaki, M. Baudoux, T. Ishibashi, S. Kawabata, Y. Nishiyama, K. Sugai, and H. Kumamaru performed spinal cord injuries. S. Kawase participated in plasmid preparation. K. Yasutake performed cryostat sectioning. M. Nakamura and H. Okano supervised the study and edited the manuscript.