|Home | About | Journals | Submit | Contact Us | Français|
Pathological findings in neonatal brain injury associated with preterm birth include focal and/or diffuse white matter injury (WMI). Despite the heterogeneous nature of this condition, reactive astrogliosis and microgliosis are frequently observed. Thus, molecular mechanisms by which glia activation contribute to WMI were investigated.
Postmortem brains of neonatal brain injury were investigated in order to identify molecular features of reactive astrocytes. The contribution of astrogliosis to WMI was further tested in a mouse model in genetically-engineered mice.
Activated STAT3 signaling in reactive astrocytes was found to be a common feature in postmortem brains of neonatal brain injury. In a mouse model of neonatal WMI, conditional deletion of STAT3 in astrocytes resulted in exacerbated WMI, which was associated with delayed maturation of oligodendrocytes. Mechanistically, the delay occurred in association with over-expression of TGFβ-1 in microglia, which in healthy controls, decreased with myelin maturation in age-dependent manner. TGFβ-1 directly and dose-dependently inhibited the maturation of purified oligodendrocyte progenitors, and pharmacological inhibition of TGFβ-1 signaling in vivo reversed the delay in myelin development. Factors secreted from STAT3-deficient astrocytes promoted elevated TGFβ-1 production in cultured microglia compared to wild type astrocytes.
These results suggest that myelin development is regulated by a mechanism involving cross-talk between microglia and oligodendrocyte progenitors. Reactive astrocytes may modify this signaling in STAT3-dependent manner, preventing the pathological expression of TGFβ-1 in microglia and the impairment of oligodendrocyte maturation.
Oligodendrocytes (OL) are the myelinating cells of the central nervous system, facilitating fast conduction of action potentials along the axon. In humans, myelin development begins prenatally and continues into young adulthood to achieve complex neurological functions. Failure of proper myelin development is a common pathology in neonatal brain injury associated with preterm and term births associated with hypoxic-ischemia or other insults1–7. The pathology, collectively called neonatal white matter injury (WMI), ranges from diffuse non-necrotic changes affecting mainly subcortical white mater (hypoxic-ischemic encephalitis, HIE) to focal lesions affecting periventricular white matter (periventricular leukomalacia, PVL), and is characterized by underdeveloped myelin and wide-spread reactive gliosis, along with limited neuronal loss in some cases. Afflicted children frequently present with diminished cognitive, motor, and psychiatric functions that persist beyond infancy8–10. Although etiology of WMI is likely to be multifactorial, neuroinflammation appears to be a major contributing factor, as suggested by consistent positive correlation with infections during pregnancy11, 12 and postnatal period13, elevation of immune activity in the fetal blood14, amniotic fluid15, and umbilical cord16. Despite these findings, limited knowledge with respect to the causal link and molecular mechanisms by which such insults lead to actual damages to oligodendrocytes has hindered the development of effective therapies.
Similar to many adult disorders of the central nervous system (CNS), reactive gliosis (astrocyte and microglia activation) is a hallmark of neonatal WMI17–20. In adult CNS, astrocytes are known to play homeostatic roles in active maintenance and surveillance of CNS environment, and in activated state, play putative roles in both destruction and promotion of repair (reviewed in 21). However, it is unclear if developing glial cells possess similar phenotypes and functions in the unique environment of the neonatal brain, and in particular, whether they have direct contribution to the pathology observed in WMI. In fact, recent evidence suggests that an arrest in OL maturation, rather than cellular loss, underlies some cases of WMI22, 23, raising the possibility that the pathological brain environment may be inhibitory for OL differentiation, leading to growth failure and improper damage repair. The purpose of this study was to identify cellular/molecular mechanisms of reactive gliosis underlying WMI in the neonatal brain, and to determine the consequences to myelin development. We report suporting evidence that STAT3-mediated astrogliosis plays a protective role in WMI, by restricting the aberrant expression of microglial TGFβ-1, a factor we found to inhibit OL maturation. These studies highlight not only a possible involvement of microglial signals in normal development of myelin, but also a protective role of astrogliosis in WMI involving microglial TGFβ-1.
All human tissues were collected as part of University of California Multicampus Research Programs and Initiatives. They were collected in accordance with guidelines established by the University of California San Francisco and University of California Los Angeles Committees on Human Research. Four cases of WMI (case 1–4) and three cases of controls (case 5–7) were analyzed (see Table 1). Cases 1 and 2 were 35 and 37 and 4/7 weeks of gestation respectively, with diagnoses of hypoxic-ischemic encephalopathy (HIE). Cases 3 and 4 were 31 weeks and 38 weeks of gestation, respectively, with diagnoses of periventricular leukomalacia (PVL). Control case 5 was 36 and 4/7 weeks, and case 6 was 37 and 3/7 weeks of gestations, and the causes of death were diaphragmatic hernia. Control case 7 was 38 weeks of gestation, suggestive but not diagnosis of acute anoxia. Brain tissues were harvested within 48 hrs of patient death, and were fixed in 4% paraformaldehyde, cryoprotected in sucrose solution (cases 1,2,5,6), embedded in OCT, sectioned, and stored at −72°C or fixed in neutral buffered formalin (cases 3,4,7), paraffin embedded, and sectioned, and stored at 4°C. Subcortical white matter below cingulate gyrus or periventricular white matter outside necrotic lesion was used for immunohistochemistry. Please see supplementary materials for additional information on diagnoses.
Experiments were performed according to protocols approved by the Chancellor’s Animal Research Committee of the Office for Protection of Research Subjects at the University of California, Los Angeles. LPS (Escherichia coli strain O55:B5, lot number 20316A1) was purchased from List Laboratories. We determined that the concentration of 10 mg/kg in this lot was adequate to induce the white matter injury in C57BL/6 mice as described in the results. LPS was diluted in sterile saline to a concentration of 0.5 mg/ml and injected subcutaneously between the shoulders. Please see supplementary materials for detailed information on injection.
The generation of mice deficient in STAT3 selectively in astrocytes was described previously24–26. The breeding strategy used in the experiments was GFAPCre+/STAT3loxploxp × GFAPCre−/STAT3loxploxp. The offspring were approximately 50% Cre+/STAT3loxploxp and 50% Cre−/STAT3loxploxp. All lines were backcrossed to C57BL/6 for at least eight generations, and littermates served as genotype controls. GFAPCre+ mice were also bred with a reporter line expressing β-galactosidase 27. C57Bl/6 WT mice were obtained from our ongoing breeding colony.
A selective TGFβ receptor kinase inhibitor SB431542 (Cayman chemical) was dissolved in Dimethyl Sulfoxide and sterile saline (1:1 vol/vol) to a concentration of 0.5 mg/ml. An intraperitoneal injection at 5 mg/kg was administered at P2.
Mouse forebrain at specified ages was homogenized in RIPA buffer containing protease inhibitor cocktail (mini complete, Roche). 25–35 ug of protein was resolved on a 4–20% gradient polyacrylamide gel (invitrogen) and transferred to PVDF membranes (Biorad). Membranes were then incubated with primary antibodies, appropriate secondary antibody conjugated to horseradish peroxidase, and visualized using ECL or ECL plus (GE healthcare). Quantification of relative amounts of proteins was made in ImageJ from developed films. Primary antibodies used are: mouse MBP (1:1000, Covance), rabbit phosphorylated SMAD2 (1:500; Cell Signaling), mouse phosphorylated STAT3 (1:500, BD Biosciences), rat GFAP (1:2500, Invitrogen), and mouse β-actin (1:3000; Sigma).
All human postmortem WMI samples were subjected first to antigen retrieval protocol which consisted of a 10-min incubation in 10 mM sodium citrate at 90°C followed by 20-min cool-down at room temperature. After three washes with PBS, the same protocol for mouse immunohistochemistry was followed.
Animals were anesthetized and perfusedintracardially with 4% PFA in 0.1 M phosphate buffer, pH 7.4. Brains were postfixed in the same solution, cryopreserved in 20% sucrose in 0.1 M phosphatebuffer. Frozen sections were sectioned, mounted for slide imunohistochemistry (15 um) or stored for free-float immunohistochemistry (30 um). After blocking, brain sections were incubated with various primary antibodies, followed by appropriate secondary antibodies (Invitrogen, or Jackson ImmunoResearch). 4′,6-Diamidino-2-phenylindole (DAPI, Vector Laboratories) was used as a nuclear counterstain. For TGFβ-1 and pSMAD2 staining, fluorescence was amplified using Tyramide Signal Amplification kit (Perkin Elmer). Please see supplementary materials for list of antibodies and method for quantification.
Enriched oligodendrocyte progenitors were prepared as described28. First, a mixed glial culture was prepared by mechanically dissociating the cerebral hemispheres from 1 day old C57BL/6 WT mice and plating on poly-D-lysine-coated flasks in Dulbecco’s modified Eagle’s medium and Ham’s F12 (1:1 vol/vol, hereafter referred to DMEM/F12 medium) (Invitrogen), containing 100 μg/ml gentamicin and supplemented with 4 mg/ml dextrose anhydrous, 3.75 mg/ml HEPES buffer, 2.4 mg/ml sodium bicarbonate and 10% fetal bovine serum (FBS) (Omega Scientific). After 24 hrs the medium was changed and the cells were grown in DMEM/F12 supplemented with insulin (5 μg/ml), human transferrin (50 μg/ml), sodium selenite (30 nM), d-Biotin (10 mM), 0.1% BSA (Sigma Aldrich), 1% horse serum and 1% FBS (Omega Scientific). After 9 days, OL progenitors were purified from the mixed glial culture by the differential shaking and adhesion procedure29 and allowed to grow on polylysine-coated coverslips in defined culture medium30. The purity of culture was 95%, as confirmed by NG2 as well as PDGFRα staining. Please see supplementary materials for list of antibodies and method for assessment.
Primary astrocyte cultures were prepared as described31. As STAT3-cKO and WT pups were born in the same litter and experimenters were blind to the genotype at the time of tissue collection, individual cultures were obtained from each mouse, resulting in one astrocyte conditioned medium obtained from each mouse pup. Cerebral hemispheres from 0 to 1 day old mice were mechanically dissociated, passed through a 310 um nylon mesh, and plated on poly-D-lysine-coated 100 mm dish containing Eagle’s minimum essential medium (MEM) supplemented with 10% FBS, glucose (2 mg/ml), insulin (5 ug/ml), and gentamycin (100 ug/ml). After 3 days, the medium was changed to MEM supplemented with G5 consisted of glucose (4.5 mg/ml), hydrocortisone (10 nM), sodium selenite (30 nM), insulin (5 ug/ml), transferrin (50 ug/ml), biotin (10 ng/ml), bFGF (5 ng/ml) and EGF (10 ng/ml) (Peprotech)32. The purity of culture was 98%, as confirmed by GFAP staining as well as exclusion criteria for Olig2 (oligodendrocyte) and CD11b (microglia). Medium was changed to match the microglia medium DMEM/F12 containing 10% FBS at 11 days in culture, and astrocyte conditioned medium was collected after 72 hrs. For the treatment of microglia, 25% astrocyte conditioned medium supplemented with 75% DMEM/F12 containing 10% FBS was added as described below.
Primary microglia culture was obtained from the mixed glial culture explained above. At 9 days in culture, microglia were purified by shaking at 200RPM for 1 hr and plated on 24-well plates in DMEM/F12 containing 10% FBS. The purity of culture was at least 92%, confirmed by CD11b and Iba1 staining. Treatment with astrocyte conditioned medium started the following day, in presence of 25 ng/ml LPS to activate microglia. The medium was collected 48 hours later for determination of secreted TGFβ-1 by ELISA.
Culture medium was collected from purified astrocyte and microglia cultures before and at the end of treatments specified above and used in sandwich ELISA for TGFβ-1 following manufacturer’s protocols (eBioscience).
Data are presented as mean ± SEM. Statistical analysis was performed using GraphPad Prism 4.01 software with ANOVA followed by Bonferroni’s multiple comparison test, or Friedman test followed by Bonferroni’s multiple comparison test. A two-tailed unpaired Student’s t test was used when two groups were compared. All statistics were considered significant when P < 0.05.
The presence of reactive gliosis in neonatal WMI has been reported in several histological studies19, 20, 33. We confirmed the existence of widespread reactive astrogliosis as well as microgliosis in the white matter in four cases of postmortem brain samples of WMI using the astrocyte marker GFAP and microglia marker Iba1. In all WMI, but not control cases, we observed an intense staining of GFAP+ hypertrophic cells with elongated processes, a hallmark morphology indicative of reactive astrogliosis (Figure 1A, a–b). An abundant Iba1+ microglia with amoeboid morphology were observed in the same areas (Figure 1A, c–d), implying a reactive response to an insult. More detailed examination of reactive astrogliosis revealed that a subset of astrocytes in the white matter showed nuclear localization of phosphorylated STAT3 (pSTAT3), indicating that STAT3 pathway was activated in these cells (Figure 1B). STAT3 activation in astrocytes has been implicated as a principal inducer of the hypertrophic morphology observed in reactive astrogliosis 24, and appears to regulate numerous target genes 34, probably affecting the cellular phenotype. pSTAT3 containing astrocytes were observed in all four cases of WMI regardless of the type of diagnoses (HIE or PVL, see Methods) but none of three age-matched control cases, implicating a molecular signature of reactive astrocytes shared among WMI cases (Table 1).
In order to test the specific contribution of glia activation in neonatal WMI, we optimized a mouse model in which astrogliosis and microgliosis along with WMI could be consistently induced in the absence of complicating factors such as gross ischemia or necrosis. To specifically address the effect of gliosis on white matter development, we felt it was important to consider inter-species differences between humans and mice with respect to the time course of myelin development. The highest risk of WMI in humans occurs at 23–32 gestational weeks, when majority of the white matter is composed of immature, pre-myelinating OL, which are known to be vulnerable to exogenous insults35, 36. In mice, this period of white matter development corresponds to the early postnatal period37, when the great majority of OL-lineage cells express immature markers such as PDGFRα and O4, but not mature markers such as MBP and GSTπ38. Indeed, we found that a single subcutaneous injection of lipopolysaccharide (LPS), 10 mg/kg, in the early postnatal period resulted in sustained reactive micro- and astro-gliosis, (Figure 2A and Supplementary Figure 1A) and a subsequent deficit in myelin development (Figure 2C–E), similar to that previously reported inflammation-induced WMI models40, 41. Moreover, like in human postmortem tissues, we observed nuclear localization of pSTAT3 in GFAP positive astrocytes starting six hours after peripheral LPS injection (Figure 2B), and lasting up to 5d (later time points shown in Supplementary Figure 1A). Despite activation of STAT3 pathway and the enhanced expression of GFAP, an intermediate filament protein known to increase expression in gliosis (Figure 2A and Supplementary Figure 1A), the expression of another astrocyte marker, S100, a calcium binding protein, was not affected, indicating that LPS had no gross effects on astrocyte development (Supplementary Figure 1B).
In agreement with the notion for time-sensitive risk for white matter injury after various stressors in humans42, we identified a narrow time window for LPS injection between P1 and P2 that resulted in a significant reduction in MBP (30% – 54% reduction compared to control), whereas beyond P2 there were only marginal non-significant reductions (Figure 2C). These data highlighted the importance of the early and restricted postnatal period P1–2 as a critical window in which the induction of reactive gliosis can disrupt healthy myelination in the mouse forebrain.
Further examination of the immunoreactivity to mature myelin marker MBP revealed that in LPS injected animals, the corpus callosum was less stained and thinner, striatal myelin bundles were smaller, and cortical radiation from the corpus callosum was limited to deep cortical layers (Figure 2D and 2E, e–f) in broad areas spanning from frontal to temporal cortex (Supplementary Figure 1C). Notably, many of LPS injected animals showed enlarged lateral ventricles, a common finding in human patients with WMI (Figure 2D and Supplementary Figure 1C, b–c f–g)43–46.
In order to determine if the mechanism underlying the hypo-myelinated phenotype was an arrest of OL maturation as implied in studies of human pathological samples22, 23, we investigated the appearance of maturation-dependent OL lineage markers in the mouse model. Quantification of temporal changes revealed that by 5 days post injection, immunoreactivity for O4, was significantly reduced in large areas of forebrain white matter including corpus callosum and striatum (Figure 2E, c–d and Figure 2F). At this developmental stage, O4 almost exclusively marks OL lineage cells at the preOL stage (beyond progenitor, but pre-differentiated stage) because immunoreactivityies for mature OL markers MBP and GSTπ are barely detectable at this stage. The reduction of O4 at 5 days after LPS was associated with a reduction in mature OL markers MBP and GSTπ in LPS-treated mice at 10 and 15 days post injection, respectively (Figure 2E, e–h and Figure 2F). We did not find evidence for an increased apoptotic death in Olig2+ pan-OL, as assessed by either activated caspase-3 antibody or terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay at any of seven time points examined between 2 hr – 20 days (Supplementary Figure 2A shows 3 time points). Delayed appearance of differentiated OL together with the lack of OL death is consistent with a recent clinical study in which myelin staining was altered without apparent OL loss22, 23, which has led to the proposal that arrested differentiation underlies the observed pathology. Interestingly, the delay in the appearance of mature markers was preceded by a transient increase in the number of PDGFRα+ OL progenitors (Figure 2E, a–b and Figure 2F), suggesting an accumulation of cells in the white matter at this very early stage. The increased number of progenitors did not appear to be due to enhanced proliferation; in fact, there was a trend for a decrease, which was significant at the 48 hr time point (Supplementary Figure 2B). Finally, numbers of cells labeled with the pan-OL lineage marker Olig2 was not changed by LPS (Fig. 2F), suggesting that the observed changes reflected a shift in cells numbers towards an immature state rather than an overall loss of cells in the OL lineage. In summary, the data suggest that LPS induces an arrest of OL development, without major effects on proliferation or survival.
Over the 20 day time course examined, no evidence of increased neuronal cell death was observed, as assessed by detection of TUNEL staining in NeuN+ cells. Moreover, immunostaining for the pan axonal marker NF160 did not differ between control and LPS-treated mice in callosal bundles where myelin marker was reduced (Supplementary Figure 2C–D), suggesting that the myelin defect was not likely secondary to neuronal or axonal loss. Collectively, these results suggested that a neuroinflammatory insult at this critical period of myelination is sufficient to induce a deficit in myelin development, through a mechanism of impaired production and/or arrested differentiation of OL progenitors.
As discussed above, nuclear localization of pSTAT3 was detected in GFAP positive cells in LPS-treated mice (Figure 2B and Supplementary Figure 1A), similar to that observed in clinical WMI cases (Figure 1B). To investigate the role of STAT3-mediated induction of reactive astrogliosis in neonatal WMI, we used mice harboring an astrocyte-specific deletion of STAT3 gene (STAT3-cKO mice). The transgenic Cre driver mouse line used a 15 kb mouse GFAP promoter cassette, and targets astrocytes that express high levels of GFAP, postnatal GFAP+ neural stems cells, but not embryonic radial glia (described in47). By immunohistochemistry, the expression of a reporter protein β-galactosidase driven by a floxed ROSA26 promoter27 in mice harboring the same GFAP-Cre cassette was found to be almost undetectable in the forebrain white matter of non-LPS injected animals at P2 (Supplementary Figure 3A, a–c), confirming the limited Cre-mediated excision activity at this time point in the absence of LPS. On the other hand, when astrocytes become hypertrophic and upregulate GFAP expression in response to LPS, β-gal was specifically detected in GFAP+ astrocytes (Supplementary Figure 3A, d–f). These data suggested that the GFAP-Cre line spared endogenous astrocytes at the time of LPS injection, and induced Cre-mediated recombination in response to the inflammatory insult.
Untreated STAT3-cKO mice showed a normal time course of induction of OL maturation markers, tested at P2 (PDGFRα for OL progenitors), P7 (O4 for immature OL), and at P12 (MBP for mature OL) (Supplementary Figure 3B, a–f). The nuclear localization of pSTAT3 observed in numerous astrocytes of STAT3-WT mice after LPS injection was abolished in STAT3-cKO mice (Figure 3A), confirming the integrity of the STAT3 excision in reactive astrocytes. The functional inactivation of the STAT3 pathway was demonstrated by the fact that STAT3-cKO mice failed to exhibit the hypertrophic reactive astrocyte morphology observed in STAT3-WT littermates (Figure 3A–B) even in the presence of an inflammatory insult. This was expected because the STAT3 pathway is reported to be a critical inducer of astrocyte hypertrophy24.
In order to test the effects of attenuated STAT3 signaling in reactive astrocytes on myelin development, MBP, a myelin protein expressed primarily in mature OL was quantified. The decreased MBP expression induced by LPS in STAT3-WT mice was more pronounced in STAT3-cKO mice at 10 days post LPS, as determined by Western blot and immunohistochemistry (Figure 3C–D). Moreover, there was no evidence for increased TUNEL-dependent cell death in OL (Supplementary Figure 4A), suggesting that the exacerbated myelin defect in STAT3-cKO was not associated with appreciable OL apoptosis. Because GFAP is also strongly expressed in postnatal stem cells that potentially generate OL48, we addressed the possibility that STAT3-deletion in GFAP+ neural stem cells diminished the production of OL. To do this we followed the fate of cells undergoing Cre-mediate recombination and their progeny using the aforementioned β-gal reporter driven by the same GFAP-Cre driver. Only 1–2% of mature (GSTπ+) OL appearing at this stage expressed β-galactosidase in the basal state, and the LPS injection did not alter this percentage (Supplementary Figure 4B) at P20. These results indicated that the contribution of postnatal GFAP+ stem cells to the OL lineage is small at this stage, and not significantly altered by LPS injection. Thus, a diminished production of OL from GFAP+ neural stem cells is unlikely to play a role in the exacerbated myelin injury.
In order to investigate the mechanism by which astrocytic STAT3 pathway imposed a non-cell autonomous effect on the process of myelination, we screened for candidate factors with aberrant temporal production in STAT3-cKO brain after LPS injection. TGFβ-1 was found to fulfill the criteria as a factor whose over-expression might delay OL maturation. TGFβ-1, a BMP family cytokine, signals through transcription factors SMAD2 and SMAD3, distinct from those induced by BMPs (SMAD1, 5, and 8). TGFβ-1 has been shown to affect differentiation of a broad range of cell types outside the CNS49–52, but its functions on oligodendrocyte development are unknown. Interestingly, TGFβ-1 expression in the postnatal brain of untreated wild type mice was strongly developmentally regulated, being the highest at the neonatal stage (P2), then decreasing and becoming nearly undetectable by immunohistochemistry by P12 (Figure 4A, a–c). A similar age-dependent decrease was observed in gene expression levels (Figure 4A, right panel). Importantly, essentially all detectable TGFβ-1 immunoreactivity was colocalized to CD68+ microglia at each of the time points examined, whereas no detectable colocalization was found in GFAP+ astrocytes (Figure 4A, d–i). After LPS injection, TGFβ-1 expression was significantly increased in LPS-injected mice compared to control-injected mice in broad regions of forebrain, and increased even further in mice with the conditional STAT3 deletion (Figure 4B). Double immunostaining revealed that regardless of genotype and time point, TGFβ-1 expression was exclusively colocalized to microglia (Figure 4B), suggesting microglia as the primary source of TGFβ-1.
If TGFβ-1 indeed acted on nearby OL precursors to regulate their development, it might be expected that the downstream TGFβ-1 signal transducer, phosphorylated SMAD2 (pSMAD2) would be detectable in cells in the OL lineage at early postnatal stages (when TGFβ-1 expression is high). Moreover, pSMAD2 levels are expected to be increased after LPS administration, due to its apparent induction of TGFβ-1 expression. We confirmed the expression of pSMAD2 in Olig2+ cells in non-injected WT mice at P3, which decreased as animals aged to P12 (data not shown). Consistent with the hypothesis, we found that the proportion of pSMAD2+/Olig2+ double positive cells was significantly increased by LPS injection, and further increased in LPS-treated STAT3-cKO mice beyond observed in similarly treated STAT3-WT mice (Figure 4C–D).
The observation that LPS-induced TGFβ-1 expression in microglia was differentially affected in STAT3-cKO vs. WT mice raised the question of how astrocyte-specific differences in STAT3 activity could mediate this altered microglial phenotype. In order to test the ability of astrocytes to directly control microglial TGFβ-1 expression, we purified astrocytes from STAT3-cKO and STAT3-WT littermates (therefore astrocytes were either KO or WT in STAT3) and collected their conditioned media. We confirmed constitutive basal expression of pSTAT3 in cultured WT astrocytes (Supplementary figure 5) as suggested in literature53. Media from these cells, containing secreted factors from astrocytes, were added to purified primary microglia obtained from C57BL/6 WT mice, which had not been affected by genetic manipulation during their development, and the secretion of TGFβ-1 was measured by ELISA. Microglia cultured alone were found to secrete 0.761±0.042 ng/ml TGFβ-1 into their medium. However, when they were cultured in presence of astrocyte conditioned medium obtained from STAT3-WT astrocytes, TGFβ-1 in the medium was significantly reduced (Figure 4E), suggesting that astrocytes secrete factor(s) with the capacity to limit TGFβ-1 production by microglia. On the other hand, when microglia were treated with medium from STAT3-KO astrocytes, levels of TGFβ-1 were not significantly reduced, and remained at a similar level to microglia cultured alone (Figure 4E). The source of TGFβ-1 is most likely microglia, because astrocyte conditioned media did not contain live cells and the purity of microglia cultures was >92%. In addition, the contribution of TGFβ-1 secreted by astrocytes before addition to microglia was found to be low and similar between STAT3-cKO and STAT3-WT (0.072±0.047 ng/ml in KO, 0.065±0.072 ng/ml in WT). These results suggested that STAT3 signaling in astrocytes regulates the secretion of factors to limit the production of TGFβ-1 from microglia in vitro. Moreover, the results raised the possibility that astrocytic STAT3 signaling might serve as a cell non-autonomous mechanism to inhibit the microglial production of TGFβ-1 during an inflammatory insult, when STAT3 pathway becomes activated. On the other hand, exogenously applied TGFβ-1 did not induce STAT3 pathway activity in astrocytes, suggesting that activation of STAT3 signaling in astrocytes did not occur as a result of overexpressed TGFβ-1 (Supplementary Figure 7).
The fact that that the disappearance of TGFβ-1 expression in untreated wild type mice coincided with the appearance of mature OL marker MBP, and that the disappearance was delayed under conditions of excess TGFβ-1 in association with myelin injury (see Figure 4B) suggested that microglial TGFβ-1 might function as a molecular “brake” in normal OL differentiation, and might further delay OL differentiation after induction of astrogliosis. In this respect, the disappearance of TGFβ-1 expression could be important trigger for the commencement of myelin development, and that its disappearance needs to be carefully timed to result in properly-timed myelination. To test whether TGFβ-1 has the capacity to act directly on OL to regulate their development, we prepared cultures of purified OL progenitors from WT mice, and investigated their development in presence and absence of TGFβ-1. As assessed by Western blot, there was an upregulation of pSMAD2 expression within 2 hrs of TGFβ-1 application, confirming their ability to respond to exogenous TGFβ-1 (Figure 5A). The development of the OL progenitors proceeds with proliferation and differentiation, so to study these processes independently, we examined the effects of TGFβ-1 under separate conditions. In conditions under which the proliferation of OL progenitors is maintained (i.e., in media containing mitogens PDGF-A and bFGF for 48 hours), proliferation, as determined by bromo-deoxyuridine (BrdU) incorporation after a 24 hr pulse was decreased by TGFβ-1 in a dose dependent manner. Similar results were obtained using another mitotic marker phosphorylated histone H3 (Figure 5B). Under conditions where OL progenitors were allowed to differentiate for 48 hr, TGFβ-1 induced a dose-dependent decrease in the proportion of cells that expressed mature markers CC1, O1, and MBP, and an increase in cells that retained immature OL markers PDGFRα, Sox9 and Sox2 (Figure 5C). Additionally, delayed differentiation of OL progenitors was confirmed by assessing the morphological complexity, which yielded supporting results that TGFβ-1 treatment dose-dependently limited the extension of processes, maintaining the cells in immature morphology (Supplementary Figure 6).
Next, to test the involvement of TGFβ-1 signaling on myelination in vivo, a selective inhibitor of the signaling pathway, SB431542 (Cayman Chemical), was administered during normal myeilin development as well as during LPS-induced WMI paradigm. SB431542 inhibits the kinase activity of TGFβ specific receptor ALK5 thereby blocking SMAD2 signaling but spares SMAD1/5/8 signaling used by BMPs54. Its efficacy has been shown in vivo in adult mouse models of various diseases55–57, and in neonatal mice, was found to be effective in reducing the phorphosrylation of SMAD2 in the brain when administered intraperitoneally at P2 (Figure 5D). The administration of SB431542 at P2 during normal development induced a small, non-significant increase in mature myelin protein MBP. Importantly, upon LPS injection, SB431542 was found to partially rescue the reduction of MBP by about 50% (Figure 5E), suggesting a critical involvement of sustained SMAD2 activity in the delayed myelin development observed in LPS injected mice. Collectively, these in vitro and in vivo results demonstrated that prolonged activation of TGFβ-1 signaling may be a key pathological modulator in neonatal brain development. The observation that the reduction of TGFβ-1 expression was delayed after LPS injection and that microglia were the only detectable source of TGFβ-1 at all time points (see Figure 4B), together with the finding that expression levels of TGFβ-1 were affected by astrocyte-specific STAT3 signaling, suggests a non-cell autonomous mechanism involving at least three different cell types, whereby reactive astrocytes alters the microglial secretion of TGFβ-1, which in turn acts on OL progenitors to alter their development.
To determine if overt TGFβ-1 expression in microglia is also a characteristic of human cases of WMI, we stained samples of the human postmortem cases in Table 1 with antibodies against human TGFβ-1 and the microglia marker Iba1. In all cases of WMI, there was abundant expression of TGFβ-1 in the white matter. Moreover, a majority of the TGFβ-1 staining localized to Iba1 positive cells, demonstrating the microglial expression of TGFβ-1 (Figure 6). Notably, microglia containing TGFβ-1 showed hypertrophic, amoeboid morphology, indicative of an activated phenotype. These studies indicate that TGFβ-1 signaling might represent a therapeutic target to limit WMI in babies at risk for this pathology.
Reactive astrogliosis is a hallmark of neonatal brain injury and many other pathologies of the CNS. Despite this commonality, its beneficial vs. harmful consequences in various contexts are not well understood, and cell non-autonomous effects of astrogliosis on neonatal brain development have never been demonstrated. Although inflammatory insults and hypoxic-ischemic events are considered the main causal factors for neonatal WMI, the multifactorial nature of the pathology has limited the advances of our mechanistic understanding. In this regard, pathological studies on hypoxic-ischemia cases generally show concurrent micro- and astro-gliosis, making it difficult to dissociate the sequence of key factors leading to the white matter pathology. In the current model, we attempted to study the reactive glia component of neonatal WMI in relative isolation, and therefore optimized an inflammation to consistently induce gliosis in mice at a time corresponding to the heightened vulnerability to WMI in humans. We found that a single peripherally-administered dose of LPS was sufficient to induce micro-and astro-gliosis and WMI injury in mice in a pattern consistent with that observed in human postmortem cases. This protocol, along with genetically-engineered mice were then used to probe the complex nature of neonatal brain injury. Importantly, the studies revealed a protective role of astrocytes, and led to the identification of a candidate pathway for disease protection involving modulation of the TGFβ-1 signaling pathway.
Our data in human postmortem tissues of neonatal WMI samples demonstrated that the STAT3 pathway was commonly active in hypertrophic astrocytes, so the potential consequences of STAT3-mediated astrogliosis were investigated in our model using conditional knockout mice. The STAT3 pathway has been demonstrated to be a mediator of reactive astrogliosis, particularly with respect to GFAP upregulation, hypertrophic morphology, and scar formation 24. By manipulating the STAT3 gene in astrocytes, we demonstrated a protective role of reactive astrocytes in limiting myelin injury. An investigation of potential mechanisms revealed that aberrant TGFβ-1 in microglia might be a key factor in this pathology. In this regard, TGFβ-1 was found to be tightly regulated during development in healthy mice, and was detected exclusively in microglia. Moreover, it was suggested that excess TGFβ-1 had a direct action on developing OL, as evidenced by the fact that higher levels of the signal transducer pSMAD2 were observed in OL lineage cells in vivo under conditions of high TGFβ-1, and that TGFβ-1 inhibited the maturation of purified OL progenitors in culture. The that fact that pharmacological inhibition of pSMAD2 pathway partially reversed the inflammation-induced delayed myelin development in vivo suggest that TGFβ-1 signaling might be a therapeutic target for limiting myelin defects in infants at risk for WMI. However, it must be cautioned that the ability of TGFβ-1 blockade to diminish perinatal white matter injury may be model dependent. For example, other factors, including cytokines IL-6, IL-10, ciliary neurotrophic factor (CNTF), and leukemia inhibitory factor (LIF) in addition to LPS can activate STAT3 pathway in astrocytes, and it is uncertain if they would trigger a similar STAT3-dependent induction of TGFβ-1. In any case, the current study highlights a new role of TGFβ-1 in neonatal brain development, whose dysregulation may mediate pathology associated with a type of WMI.
Interestingly, whereas we showed TGFβ-1 was expressed primarily in microglia of postmortem WMI samples, and exclusively in microglia in the normal and inflamed neonatal mouse brain, TGFβ-1 immunoreactivity has been demonstrated in both macrophages and reactive astrocytes in multiple sclerosis (MS) lesions58. Zhang et al. recently proposed an alternative mechanism in which astrocytes expressing TGFβ-1 in MS lesions was hypothesized to induce Jagged1/Notch1 signaling in OL, thereby inhibiting their differentiation59. Being a secreted molecule, TGFβ-1 can be secreted by a variety of cellular sources and may potentially act on multiple cell types with appropriate receptors. However, as discussed, we did not observe expression of TGFβ-1 in astrocytes in the mouse model, and the great majority of cells expressing TGFβ-1 in human postmortem tissues were Iba+ microglia. On the other hand, in the developing mouse brain, Jagged1 expression was found exclusively in callosal axons in discrete locations and rapidly became undetectable during the examined postnatal period (data not shown). Based on these observations, it is unlikely that the Jagged1 expression by reactive astrocytes contributed significantly to the broadly affected pathology we reported in the neonatal mouse model used here.
Another potential mechanism is suggested by the fact that astrocytes are known to secrete cytokines including LIF60 and CNTF61 which promote survival and maturation of OL. In addition, reactive astrocytes are known to express factors such as FGF262 and hyaluronan63 that are inhibitory for OL differentiation. Of those, we examined expression levels of LIF in our model and found that it was mildly upregulated by LPS, but was not regulated in astrocytic STAT3-dependent manner (data not shown). How other molecules are regulated in the context of inflammatory insult in the neonatal brain requires further examination.
Why should an inhibitory factor for myelination be highly expressed in newborn brain, at the time when myelin is developing? One potential explanation is that OL progenitors must be maintained at an immature stage until neuronal connections are fully established and/or until other developmental processes are complete. The disruption of such a carefully-timed process by astrogliosis could presumably have pathological effects on subsequent CNS function. Notably, microglial expression of TGFβ-1, as well as nuclear pSTAT3 in reactive astrocytes was observed in the white matter of human neonatal WMI examined here. Using cell culture system, we demonstrated that astrocyte-derived factors (obtained from conditioned media from WT vs. STAT3-deficient astrocytes) had the capacity to differentially regulate TGFβ-1 secretion by cultured microglia. Although the identification of that factor remains unknown, Sarafian et al. recently reported microarray data on STAT3 WT and deficient astrocytes, which may suggest potential candidate genes53.
In summary, the clinical evidence presented along with the findings reported here in mouse and tissue culture models suggest that STAT3-mediated reactive astrocytes attempt to protect myelin development against neuroinflammation by constraining microglial TGFβ-1 expression in cell non-autonomous manner. These observations imply a potential benefit of promoting STAT3 pathway activation during inflammatory insult. Collectively, the results presented here support the concept that process of healthy myelination involves intricate and timed communications among multiple glial cell types, and that the temporally-regulated communication can be dysregulated by reactive astrogliosis during a critical time, leading to impairment in myelin development.
(A) Left panel: Time course changes in pSTAT3 expression in astrocytes. In the white matter (striatum shown), pSTAT3 expression in astrocytes peaks at 24 hrs after LPS injection and becomes near undetectable at 5 days after injection. Scale bar = 100 um. Right panel: Time course changes in GFAP expression in the forebrain after LPS injection determined by Western Blot. (B) Immunohistochemistry for another astrocyte marker S100 shows unchanged staining pattern and morphology throughout the examined time course after LPS injection, unlike GFAP expression shown in (A). Scale bar = 200 um. (C) Immunohistochemistry for myelin protein MBP in coronal sections at various levels revealed that LPS caused a reduction in MBP in widespread regions of the forebrain at 10 days post injection (P12). Notable reduction in MBP immunofluorescence was observed in white matter regions such as corpus callosum (arrows) and striatum as well as in gray matter regions (cortex, thalamus). Enlarged lateral ventricles (asterisks) was a common characteristic in LPS injected mice. Scale bar = 2 mm.
(A) Photomicrographs show Olig2+, TUNEL+, and TUNEL+/Olig2+ cells at 24 hr post LPS injection. Scale bars = 200 um. Histograms show the respective numbers of total TUNEL+ cells and double positive (TUNEL+/Olig2+) cells at 24 hr, 5 d and 10d after LPS injection. Although a significant increase in TUNEL+ cells was observed in white matter after LPS, the numbers of cells double positive for TUNEL and pan-OL marker Olig2 was extremely low and did not differ between LPS and control groups (see quantification on right), suggesting that reduced OL cell survival was not the mechanism for hypomyelination. N=3–4 in each group. P-values represent ***<0.001. Experiments were repeated using alternative cell death marker caspase-3, which gave similar results. (B) PDGFRα+ OL progenitors that are positive for proliferative marker Ki67 were counted in corpus callosum. Although proportion of proliferating PDGFRα cells were small (1~18 cells of >300 cells, also see Figure 2F), there was a significant reduction at 2d after LPS, and strong trend for reduction in other time points, suggesting negative effects of LPS for the proliferation of OL progenitors. N=3–5 in each group. P-values represent * <0.05. (C) Lack of neuronal cell death was confirmed by quantifying TUNEL+/NeuN+ cells in cortex and striatum (shown) after LPS injection. Scale bar = 200 um.
(A) In non-injected P2 mice, the number of β-gal+ cells is minimal in the cortex (a–c) and white matter (not shown), whereas β-gal+ appears in GFAP+ astrocytes 24 hr post LPS, suggesting Cre activity at baseline is minimal but increased in GFAP+ cells in response to LPS injection (d–f). Scale bar = 100 um. (B) Baseline expression of OL progenitors at P2 (PDGFRα, a–b), immature OL at P7 (O4, c–d), and mature OL at P12 (MBP, e–f) show comparable patterns of expression in cKO and WT littermates. GFAP immunoreactivity was also normal at P2, but was somewhat reduced at P7 and P12 in untreated STAT3-cKO mice (Fig S3, g–l). This may have been due to age-dependent increase in astrocyte maturation and subsequent enhancement of GFAP-Cre promoter-mediated STAT3 deletion, as the GFAP gene is a known target of STAT3 signaling 64. On the other hand, the staining pattern, fluorescent intensity, and overall appearance of another astrocyte marker S100 was indistinguishable between genotypes at all time points (Fig S3, m–r). Scale bar = 200 um.
(A) There was no evidence for increased OL death associated with LPS injection in STAT3 cKO mice. Colocalization of a pan-OL marker Olig2 and TUNEL showed similar survival of OL in cKO and WT mice. N=3 in each group. N=3 in each group. Scale bar = 200 um. (B) At P20, many β-gal+ cells were found in control as well as LPS injected mice, but colocalization within mature OL marker GSTπ was rare, and remained unchanged in both treatment groups (1.40% in control, 1.44% in LPS; a total of 1889 cells in control, 1772 cells in LPS counted). Because β-gal serves as a lineage tracing reporter protein arising from Cre mediated gene excision, these results suggest that the LPS treatment did not skew the contribution of GFAP+ postnatal stem cells away from mature OL phenotype. N=3 in each group. Scale bar = 100 um.
Primary astrocyte cultures were obtained from P1 WT and STAT3-cKO mice and stained for astrocyte marker S100 and pSTAT3 after obtaining conditioned media. Unlike astrocytes in vivo, cultured WT astrocytes were postivie for pSTAT3.
OL progenitors were kept in defined culture media for 24h and then treated with 5 ng and 50 ng of TGFβ-1 for 48h. O4+ cells were scored for morphological complexity (see method), and the percentage of cells with processes greater or equal to 4 times the cell body diameter was quantified. Scale bar = 35 um. P-values represent * <0.05, **<0.01.
Cultured astrocytes were treated with TGFβ-1 (10 ng or 50 ng) for 48 hours and expression of pSTAT3 was measured. Note TGFβ-1 did not directly increase pSTAT3 in astrocytes, suggesting STAT3 activation in astrocyets is independent of available TGFβ-1 in the environment.
This work was supported by the grants from the US National Institute of Health to J.A.W. (NS070580, HD057557), to J.d.V. (HD04612), and to M.V.S. (NS057624), and by the University of California Multicampus Research Programs and Initiatives (E.J.H., H.V.V.)