STAT3 Drives GFAP Accumulation and Astrocyte Pathology in a Mouse Model of Alexander Disease

Alexander disease (AxD) is caused by mutations in the gene for glial fibrillary acidic protein (GFAP), an intermediate filament expressed by astrocytes in the central nervous system. AxD-associated mutations cause GFAP aggregation and astrogliosis, and GFAP is elevated with the astrocyte stress response, exacerbating mutant protein toxicity. Studies in mouse models suggest disease severity is tied to Gfap expression levels, and signal transducer and activator of transcription (STAT)-3 regulates Gfap during astrocyte development and in response to injury and is activated in astrocytes in rodent models of AxD. In this report, we show that STAT3 is also activated in the human disease. To determine whether STAT3 contributes to GFAP elevation, we used a combination of genetic approaches to knockout or reduce STAT3 activation in AxD mouse models. Conditional knockout of Stat3 in cells expressing Gfap reduced Gfap transactivation and prevented protein accumulation. Astrocyte-specific Stat3 knockout in adult mice with existing pathology reversed GFAP accumulation and aggregation. Preventing STAT3 activation reduced markers of reactive astrocytes, stress-related transcripts, and microglial activation, regardless of disease stage or genetic knockout approach. These results suggest that pharmacological inhibition of STAT3 could potentially reduce GFAP toxicity and provide a therapeutic benefit in patients with AxD.


Introduction
Alexander disease (AxD) is a progressive and generally fatal disorder of the central nervous system (CNS), with a range of clinical phenotypes including cognitive and motor impairment and white matter deficits that are distributed throughout the neuroaxis [1]. AxD has been subclassified into two groups, with type I demonstrating frontal predominance and early onset, and type II presenting with lesions concentrated in the hindbrain and cervical spinal cord, less pronounced leukodystrophy and cognitive deficits, and onset at any age [2]. The hallmark pathological feature in both groups is the presence of Rosenthal fibers, cytoplasmic protein aggregates of astrocytes, particularly in perivascular, subpial, and subependymal locations.
Nearly all cases of AxD result from heterozygous missense mutations in the astrocyte intermediate filament, GFAP [3,4]. To better understand how mutations in GFAP cause disease, we generated mouse models with knock-in mutations causing single amino acid changes orthologous to common human mutations at the endogenous Gfap locus [5]. These models replicate many of the pathological characteristics found in the human disease. AxD mice spontaneously increase expression of Gfap above normal levels, form the characteristic Rosenthal fibers within the cytoplasm of astrocytes, induce stress responses including Nrf2/ARE and small heat shock protein pathways, and activate microglia [6].
A major factor in the pathology induced by mutant GFAP in AxD is accumulation of GFAP protein to toxic levels. Even overexpression of wild-type GFAP causes protein aggregation, and at high enough levels, is lethal [7]. The cascade of stress response signaling pathways associated with protein misfolding leads to reactive gliosis with increased GFAP expression, which is in part due to transactivation of the Gfap promoter [8]. The transcription factor STAT3 regulates astrocyte development and astrogliosis and directly controls transcription of Gfap [9][10][11][12][13]. STAT3 binds to the Gfap promoter and is necessary to induce expression [13,14], and STAT3 activation has been shown to drive reactive gliosis in several models of acute injury [11,12,15] and neurodegeneration [16,17].
Although the disease process in AxD is initiated by expression of mutant protein, the resulting astrogliosis and transactivation of the Gfap promoter further promote protein accumulation and toxicity [8]. We have shown that STAT3 is phosphorylated in both mouse and rat models of AxD [18,19] and hypothesize that activation of STAT3 contributes to a positive feedback loop that drives increased GFAP expression, protein accumulation, and ultimately astrocyte dysfunction. Here, we used a combination of genetic approaches to manipulate STAT3 activation and clarify its involvement in chronic gliosis and to test whether blocking phosphorylation reduces astrocyte pathology in mouse models of the disease. Therapeutic approaches to target activation of the JAK/STAT pathway have been approved to treat inflammatory disorders, hematopoietic malignancies, and solid tumors [20,21], and inhibitors that cross the blood-brain barrier could potentially be repurposed to benefit patients with AxD.

Mice
All animal studies were approved by the College of Letters and Sciences and Vice Chancellor Office for Research and Graduate Education Animal Care and Use Committee within the University of Wisconsin-Madison. Gfap +/R236H mice (Gfap tm2Mes ) were maintained as heterozygotes congenic in either FVB/N-Tac or C57BL/6J genetic backgrounds [5]. Stat3 fl/fl mice (Stat3 tm2Aki ) [22], engineered with loxP sites flanking the exon coding for the Tyr-705 phosphorylation site, were congenic in C57BL/6J and maintained as homozygotes. Tg(Gfap-cre)73.12Mvs (C57BL/6J, JAX stock #012886) [23] and Tg(Aldh1l1cre/ERT2)1Khakh (C57BL/6NJ, JAX stock #031008) [24] lines were maintained as hemizygotes. Ai14 Cre reporter mice (Gt(ROSA)26Sor tm14(CAG-tdTomato)Hze /J, C57BL/6J, JAX stock #007914) [25] were maintained as heterozygotes. Heterozygous constitutive Stat3 +/− knockout mice were generated by crossing a nestin-promotor-driven Cre line (Tg(Nescre)1Kln, C57BL/6J, JAX stock #003771) [26] with Stat3 fl/fl mice for germline excision of the floxed allele. Cre excision removes the exon containing the Tyr-705 phosphorylation site while leaving the remaining coding sequence in frame and the translated protein intact. The genes for STAT3 and GFAP are located within a 2 megabase interval on mouse chromosome 11 and required homologous recombination between the two genes to combine the floxed Stat3 allele with the R236H Gfap mutation. These mice were maintained by crossing Gfap +/R236H ; Stat3 fl/fl mice with Stat3 fl/fl mice. Genotyping was performed by Transnetyx using real time PCR to analyze tail biopsy DNA.

Human Brain Tissue
Frozen frontal cortex from a 1-year-old AxD patient was obtained from the NICHD Brain and Tissue Bank for Developmental Disorders at the University of Maryland, Baltimore, MD, USA. Tissue was thawed and fixed in 4% paraformaldehyde overnight before paraffin embedding for immunolabeling.

Tamoxifen Induction
For inducible Cre activation in mice expressing the Aldh1l1-CreER T2 transgene, all mice were given intraperitoneal injections of 80 mg/kg tamoxifen (Sigma T5648, dissolved in corn oil at 20 mg/mL) daily for 5 days starting at approximately P56. Personnel administering the injections were blinded to genotypes. Tissues were collected for molecular and histological analysis at 3 months of age (~P90) as described below.

Tissue Collection
For molecular analysis of crosses including Gfap +/R236H ; Stat3 +/− or Gfap +/R236H ; Stat3 −/fl ; Gfap-Cre Tg73.12 mice, animals were euthanized by CO 2 asphyxiation, and tissues were collected on ice and frozen immediately with dry ice and stored at −80 • C until processed. For experiments performed with crosses to generate Gfap +/R236H ; Stat3 fl/fl ; Gfap-Cre Tg73.12 mice, animals were first anesthetized with isoflurane and transcardially perfused with phosphate buffered saline (PBS, pH 7.4). Brains were collected and bisected at the midline: half was placed in 4% paraformaldehyde for histological analysis, and half was microdissected and frozen for molecular analysis.
For experiments performed with mice from the Gfap +/R236H ; Stat3 fl/fl ; Aldh1l1-CreER T2 ; Ai14 cross, animals that were positive for Ai14 were anesthetized with isoflurane and transcardially perfused with phosphate buffer, followed by 4% paraformaldehyde for histological analysis. Animals that were Ai14 negative were euthanized by CO 2 asphyxiation, and tissues were collected on ice and frozen immediately with dry ice and stored at −80 • C until processed.

RNA Extraction and Quantitative PCR
Tissues were homogenized in Trizol reagent (Thermo Fisher Scientific, Invitrogen, Waltham, MA, USA) for RNA analysis per the manufacturer's protocol. For quantitative PCR, 1 µg RNA was transcribed as cDNA using 1 µm random hexamers, 2 µm poly-(dT)15 primers and 500 nm (each) dNTP in a 20 µL reaction with Superscript III (Thermo Fisher Scientific, Invitrogen protocol). Reverse transcriptase reactions were diluted and incorporated into quantitative PCR at a ratio of 16 nl cDNA per µL reaction using SYBR Green Master Mix with 150 nM primers (Table 1) on an ABI ViiA7 Real-Time PCR System (Thermo Fisher Scientific, Applied Biosystems Inc.). Standards generated from purified amplicons were used to determine relative concentrations, and values were normalized to Rn18s.  Il6  GAC TTC CAT CCA GTT GCC TTC T  AAG TAG GGA AGG CCG TGG TT  Lcn2  AGA CTT CCG GAG CGA TCA GT  TCT GAT CCA GTA GCG ACA GC  Nqo1  CGG TAT TAC GAT CCT CCC TCA ACA  AGC CTC TAC AGC AGC CTC CTT CAT  Socs3  AAG GCC GGA GAT TTC GCT TC  GGG AAA CTT GCT GTG GGT GA  Stat3  ACG AAA GTC AGG TTG CTG GT  GCT GCC GTT GTT AGA CTC CT  Thbs1  TGC CCT GCC CAC CAC GAT TCA  GCA TAG CCG GGC TTG CAC TCA CA  Rn18s  CGC CGC TAG AGG TGA AAT TCT  CGA ACC TCC GAC TTT CGT TCT

Protein Extraction and Western Analysis
Total protein lysates were prepared from frozen brain regions by homogenizing with a bead mill mixer (GenoGrinder, SPEX SamplePrep) with 50 mg tissue per ml 2% SDS, 50 mM, Tris-HCl (pH 7.4), 5 mM EDTA, 1 mM Pefabloc SC (Sigma-Aldrich, St. Louis, MO, USA) and Complete Protease Inhibitor Cocktail (Sigma-Aldrich, Roche). Samples were boiled for 20 min, and protein concentration was determined by bicinchoninic acid assay using the BCA Protein Assay kit (Thermo Fisher Scientific) with bovine serum albumin (BSA) as a standard. For western analysis, samples were diluted with lysis buffer to equivalent concentrations and 20-30 µg electrophoresed (125V at room temperature until bromophenol blue reached the bottom of the gel) on 26-well 10% Criterion Tris-HCl or TGX gels (BioRad, Hercules, CA, USA) before transferring to an Immobilon-FL membrane using the BioRad Criterion system. For most experiments, immunoblots were stained with REVERT Total Protein Stain (LI-COR Biosciences, Lincoln, NE, USA) and imaged with a LI-COR Odyssey for subsequent normalization (GAPDH was used for normalization otherwise). Membranes were blocked in SEA Blocking Buffer (Thermo Fisher Scientific, Pierce) before incubating with primary antibodies overnight at 4 • C. Immunoblots were washed before adding secondary antibodies and incubating for 2 h at room temperature, washing again, and imaging with a LI-COR Odyssey. Tris buffered saline (TBS, pH 7.5) with 0.05% Tween 20 was used for all antibody dilutions and washes. A final wash in PBS was performed before imaging. Antibodies are listed in Table 2.

GFAP ELISA
GFAP was quantified from total protein extracted with 2% SDS (described above) by a sandwich enzyme linked immunosorbent assay (ELISA) as described previously [8]. Briefly, microtiter plates (Nunc MaxiSorp, Thermo Fisher Scientific, Waltham, MA, USA) were coated with a cocktail of monoclonal GFAP antibodies diluted in PBS (SMI-26, see Table 2 for antibodies) and blocked with BLOTTO (5% milk in PBS) before adding protein samples and standards diluted in PBS with 1% BSA and 0.05% Tween 20 (Tw20). Plates were incubated at room temperature for 2 h and washed with PBS/0.05% Tw20 before adding rabbit anti-GFAP detection antibody (Dako Z0334) diluted in BLOTTO and incubated overnight at 4 • C. Plates were washed with PBS/0.05% Tw20 followed by incubation with secondary HRP conjugated goat anti-rabbit IgG diluted in BLOTTO at room temperature for 2 h. Plates were washed with PBS/0.05%Tw20 followed by PBS before adding SuperSignal ELISA Femto Substrate (Thermo Fisher Scientific, Pierce, Waltham, MA, USA) and analyzing with a GloRunner luminometer (Turner Biosystems).

Histological Analysis
Mouse brains were collected as described above, fixed in 4% paraformaldehyde overnight before cryoprotecting in graded concentrations of sucrose (10-30%), and sectioning on a sliding microtome at 40 µm. Sections were stored at −20 • C in 25% glycerol/25% ethylene glycol/0.1 M phosphate buffer (pH 7.4) as cryoprotectant. For labeling, tissue sections were rinsed with PBS, blocked and permeabilized with 5% normal goat or donkey serum, 0.5% Triton-X-100 (TX100) in PBS, and incubated in primary antibody ( Table 2) diluted in 1% BSA, 0.3% TX100 in PBS for approximately 72 h at 4 • C. Sections were washed in PBS with 0.05% TX100 before incubating in secondary antibody ( Table 2) diluted in 1% BSA, 0.3% TX100 in PBS overnight at 4 • C. Finally, sections were washed in PBS/0.05% TX100 and mounted with ProLong Gold mounting media with DAPI (Thermo Fisher Scientific, Invitrogen). For phophoY705-STAT3 immunofluorescence, sections were further permeabilized with 100% methanol for 10 min at −20 • C prior to adding block. Fluorescence images were collected with a Nikon A1R-HD confocal microscope system with equivalent settings for comparisons between genotypes.

Image Analysis
All fluorescent images were collected with identical settings at 1024 × 1024 pixel resolution for comparisons and quantification, and analysis performed on grayscale images from individual fluorescent channels. For signal intensity, the entire field of view was selected, and the mean pixel intensity measured with ImageJ/Fiji (NIH). Values are expressed as arbitrary units (a.u.) of fluorescence normalized to the control group. To quantify p62 as a representation of protein aggregation and Rosenthal fibers, gray-scale images were converted to RGB, and p62 positive puncta highlighted using the color threshold adjustment in Fiji (default thresholding method, dark background on) by setting the upper limit for brightness to separate high intensity signals from background in the histogram. The highlighted puncta were selected, and the measured area converted to a percentage of the total image area for comparisons.

STAT3 Activation in AxD
Gfap +/R236H mouse models of AxD demonstrate GFAP accumulation and a marked astrocyte stress response. In addition to perivascular, periventricular, and subpial astrocytes, protein aggregation and increased GFAP expression occur predominantly in forebrain regions including hippocampus, corpus callosum and olfactory bulb [5,8]. As a marker of astrogliosis, changes in the Gfap transcript show a similar pattern with greater elevation in forebrain compared to hindbrain regions and spinal cord, as previously reported [8] ( Figure 1A). STAT3 is known to transactivate Gfap expression in reactive gliosis [12,15], and transcriptome analysis of a mouse model of AxD overexpressing human GFAP also suggests JAK/STAT pathway activation [27]. To determine whether STAT3 activation correlates with GFAP elevation in mouse models of AxD, we performed western analysis of STAT3 and phosphorylated STAT3 (Tyr705) in different regions of the CNS and found that STAT3 expression and activation were more pronounced in forebrain regions compared to hindbrain regions and spinal cord, similarly to Gfap transcript ( Figure 1B,C). Although STAT3 protein was elevated approximately 2-fold, expression levels did not account for the 15-to 40-fold increase in phospho-STAT3. Further analysis showed other genes regulated by STAT3, including Il6, Socs3, Cxcl1, Thbs1, Ccnd1, as well as Stat3, were also elevated in Gfap +/R236H mice ( Figure 1D). Immunofluorescence labeling demonstrated nuclear localization of activated pSTAT3 specifically in Gfap +/R236H mouse astrocytes in both forebrain and hindbrain regions ( Figure 1E). Importantly, activation of STAT3 was also apparent in brain tissue from a patient with AxD, with nuclear localization of pSTAT3 in astrocytes ( Figure 1F). These results suggest that STAT3 contributes to astrocyte reactivity in AxD, but whether the transcription factor is necessary for GFAP accumulation is not clear.
analysis of STAT3 and phosphorylated STAT3 (Tyr705) in different regions of the CN and found that STAT3 expression and activation were more pronounced in forebrain re gions compared to hindbrain regions and spinal cord, similarly to Gfap transcript (Figur 1B,C). Although STAT3 protein was elevated approximately 2-fold, expression levels did not account for the 15-to 40-fold increase in phospho-STAT3. Further analysis showed other genes regulated by STAT3, including Il6, Socs3, Cxcl1, Thbs1, Ccnd1, as well as Stat3 were also elevated in Gfap +/R236H mice ( Figure 1D). Immunofluorescence labeling demon strated nuclear localization of activated pSTAT3 specifically in Gfap +/R236H mouse astro cytes in both forebrain and hindbrain regions ( Figure 1E). Importantly, activation o STAT3 was also apparent in brain tissue from a patient with AxD, with nuclear localiza tion of pSTAT3 in astrocytes ( Figure 1F). These results suggest that STAT3 contributes t astrocyte reactivity in AxD, but whether the transcription factor is necessary for GFAP accumulation is not clear. (C) Graphs represent fold change comparisons of R236H over WT for immunoblots as shown in (B) (males at 3 months, ANOVA with Sidak's post-tests) (D) Quantification of transcripts regulated by STAT3 or related to neuroinflammation in hippocampus (Hip) and corpus callosum (CC) in Gfap +/R236H mice. Relative values are normalized to Rn18s and expressed as a fold change compared to the wild-type average (males and females, Hip at 8 weeks, CC at 12 weeks, two-tailed t-test).

Genetic Knockdown of STAT3 in AxD Model Mice
STAT3 is a pleiotropic transcription factor, and complete knockout by deletion of both alleles is lethal [28]. To further explore whether reduced STAT3 activation affects GFAP expression, we generated heterozygous constitutive knockout mice from the Stat3 tm2Aki strain [22] (see methods), and this line was crossed with Gfap +/R236H mice (Figure 2A). At 12 weeks of age, Gfap +/R236H ; Stat3 +/− mice exhibited reduced pSTAT3 protein in hippocampus and olfactory bulb ( Figure 2B), two regions known to have elevated GFAP in AxD mouse models. Quantitation of Gfap transcript and GFAP protein demonstrated only subtle and inconsistent differences ( Figure 2C), suggesting that even at reduced levels, pSTAT3 activation is still sufficient to elevate Gfap expression, or that other pathways are involved in Gfap transactivation in AxD.

Genetic Knockdown of STAT3 in AxD Model Mice
STAT3 is a pleiotropic transcription factor, and complete knockout by deletion of both alleles is lethal [28]. To further explore whether reduced STAT3 activation affects GFAP expression, we generated heterozygous constitutive knockout mice from the Stat3 tm2Aki strain [22] (see methods), and this line was crossed with Gfap +/R236H mice ( Figure  2A). At 12 weeks of age, Gfap +/R236H ; Stat3 +/− mice exhibited reduced pSTAT3 protein in hippocampus and olfactory bulb ( Figure 2B), two regions known to have elevated GFAP in AxD mouse models. Quantitation of Gfap transcript and GFAP protein demonstrated only subtle and inconsistent differences ( Figure 2C), suggesting that even at reduced levels, pSTAT3 activation is still sufficient to elevate Gfap expression, or that other pathways are involved in Gfap transactivation in AxD.

Conditional Knockout of STAT3 in Gfap Expressing Cells Prevents GFAP Elevation and Stress Response in AxD Model Mice
To specifically assess the role of STAT3 in Gfap expressing cells, we next used a transgenic line (Tg73.12) expressing Cre recombinase under the control of the mouse Gfap promoter [23], in combination with the floxed (fl) Stat3 mice. The genes for GFAP and STAT3 are located within a 2 megabase interval on mouse chromosome 11. However, we were successful in combining the floxed Stat3 allele with the R236H Gfap mutation, which required a homologous recombination event to generate a Gfap +/R236H ; Stat3 fl/fl line. To facilitate complete STAT3 knockout in astrocytes, the Gfap-Cre-Tg73.12 line was crossed with heterozygous Stat3 +/− knockout mice to produce Tg73.12; Stat3 +/− mice. This line was crossed with the Gfap +/R236H ; Stat3 fl/fl line, and animals carrying both knockout and floxed Stat3 alleles (-/fl) were analyzed ( Figure 3A, all mice congenic in C57BL/6J). At 8 weeks of age Gfap +/R236H ; Stat3 −/fl mice negative for the Tg73.12 Cre transgene showed the typical increase in pSTAT3 activation, while those that were positive for Cre demonstrated pSTAT3 levels similar to mice with wild-type Gfap in both hippocampus and olfactory bulb ( Figure 3B). Analysis of Gfap mRNA in the same mice demonstrated that conditional Stat3 knockout in astrocytes reduced transcription to levels at or below the amount observed in Gfap +/+ mice ( Figure 3C). Quantification of Lcn2 (lipocalin 2) and Aif1 (Iba1) as markers of reactive astrocytes and microglia, respectively, also showed normalization in Gfap +/R236H mice with Stat3 knockout ( Figure 3C). Finally, GFAP protein measures showed reduction with Stat3 knockout in Gfap +/R236H mice ( Figure 3D), suggesting that protein accumulation is tied to elevated expression.  To assess whether GFAP mutation might eventually cause protein accumulation with age, we performed similar experiments in older mice (12 weeks and 6 months), when pathology was increased in the model. For these experiments, we used mice carrying the floxed Stat3 gene on both alleles (fl/fl) for a more exact comparison between full STAT3 expression and the targeted knockout condition ( Figure 4A). At 12 weeks of age, we observed a reduction in phosphorylated STAT3 similar to that observed above at 8 weeks, and total STAT3 expression, which was also elevated in Gfap +/R236H mice, was normalized by blocking STAT3 activation ( Figure 4B). GFAP protein was reduced by STAT3 inactivation in all brain regions analyzed and spinal cord ( Figure 4C), with no significant difference between Cre positive Gfap +/R236H and Gfap +/+ mice. Complete knockout of Stat3 resulted in GFAP levels at or below those measured in wild-type animals. difference between Cre positive Gfap +/R236H and Gfap +/+ mice. Complete knockout of Stat3 resulted in GFAP levels at or below those measured in wild-type animals.
To assess disrupted proteostasis and protein aggregation, we analyzed the autophagy cargo protein p62/SQSTM1 (sequestosome), which facilitates degradation of ubiquitinated proteins and colocalizes with Rosenthal fibers [29,30], and found that blocking STAT3 activation reduced p62 accumulation in Gfap +/R236H mice as shown by western analysis in Figure 5A. These results were confirmed by immunofluorescence in animals at 6 months of age, which exhibited no GFAP aggregation or p62 accumulation when STAT3 phosphorylation was prevented in astrocytes ( Figure 5B,C). Finally, Iba1 immunolabeling showed reduced microglial expression, indicating the prevention of neuroinflammation ( Figure 6).  (B,C) Quantitation of pY705-STAT3 and total STAT3 protein in olfactory bulb and hippocampus by immunoblotting (B), and GFAP protein in multiple brain regions and cervical spinal cord by ELISA (C). Values in (B) are normalized to total protein and represent fold change relative to the average for Gfap +/+ ; Stat3 fl/fl ; Cre (-) mice. C57BL/6J males and females, two-way ANOVA with Sidak's post-tests, * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns = not significant.
To assess disrupted proteostasis and protein aggregation, we analyzed the autophagy cargo protein p62/SQSTM1 (sequestosome), which facilitates degradation of ubiquitinated proteins and colocalizes with Rosenthal fibers [29,30], and found that blocking STAT3 activation reduced p62 accumulation in Gfap +/R236H mice as shown by western analysis in Figure 5A. These results were confirmed by immunofluorescence in animals at 6 months of age, which exhibited no GFAP aggregation or p62 accumulation when STAT3 phosphorylation was prevented in astrocytes ( Figure 5B,C). Finally, Iba1 immunolabeling showed reduced microglial expression, indicating the prevention of neuroinflammation ( Figure 6).

Inducible STAT3 Knockout in Adult AxD Mouse Model Astrocytes Reverses GFAP Pathology and Normalizes Reactive Astrocytes and Microglia
To investigate whether genetic manipulation of STAT3 phosphorylation can affect GFAP accumulation in adult AxD model mice with established pathology [5,6], we took advantage of a tamoxifen-inducible astrocyte-specific CreER T2 line using the Aldh1l1 promoter [24]. We first crossed the Aldh1l1-CreER T2 line to the Cre-reporter line Ai14 [25] and subsequently bred those mice with the Stat3 fl/fl line. Gfap +/R236H ; Stat3 fl/fl mice were then crossed to the CreER T2 ; Ai14; Stat3 fl/fl line. Gfap +/R236H and Gfap +/+ animals that were either positive or negative for Cre were treated with tamoxifen for 5 days at 2 months of age, and tissues collected at 3 months of age ( Figure 7A).
An initial assessment of efficacy for tamoxifen induction of Cre recombination in astrocytes using the Ai14 tdTomato reporter showed that 83% of GFAP positive astrocytes in hippocampus and 87% in olfactory bulb were also positive for tdTomato (364+/436 and 748+/858 astrocytes, respectively). We determined whether STAT3 was activated and found that Gfap +/R236H ; Stat3 fl/fl mice (without Cre) showed the same increase in pSTAT3, whereas mice expressing Cre (Stat3 ∆/∆ ) demonstrated pSTAT3 levels similar to those of Gfap +/+ mice in both hippocampus and olfactory bulb ( Figure 7B). Total STAT3 protein showed similar changes, although not as pronounced.
We next examined whether blocking STAT3 activation in adult AxD model mice could reverse GFAP accumulation. Transcript analysis demonstrated a significant reduction in Gfap expression in hippocampus and olfactory bulb ( Figure 7C). Analysis of GFAP protein showed reductions in all brain regions examined except for cerebellum ( Figure 7D). Spinal cord, which had decreased amounts of GFAP in Gfap +/R236H mice, also showed no further reduction in protein. Lcn2 transcript was measured as a surrogate marker of astrogliosis, and expression was normalized with STAT3 inactivation ( Figure 7E). Other transcripts regulated by STAT3, including Socs3 and Ccnd1, were also decreased in hippocampus, and Socs3 was reduced in olfactory bulb. Transcripts for small chemokines Cxcl1 and Ccl2 were reduced, suggesting a normalization of the neuroinflammatory response, and Nqo1 (NAD(P)H quinone oxidoreductase), a target of the Nrf2/ARE antioxidant stress response pathway, was reduced in hippocampus but not olfactory bulb ( Figure 7E).
To determine whether reduced GFAP expression led to reduced aggregation and improved proteostasis, we immunolabeled p62 positive aggregates as a measure of Rosenthal fibers and found that protein aggregation was reversed in Gfap +/R236H mice by blocking STAT3 activation ( Figure 8A,B). Quantification of p62 by western analysis showed a similar reduction in hippocampus and olfactory bulb ( Figure 8C). Cells 2023, 12, x FOR PEER REVIEW 11 of 21  Figure 4A. (B,C) Immunofluorescence labeling of p62 and GFAP in hippocampus (B), and quantification of percent area for p62 (C) in mice from the same cross at 6 months of age. Values in (A) are normalized to total protein and represent fold change relative to the average for Gfap +/+ ; Stat3 fl/fl ; Cre (-) mice. C57BL/6J males and females, two-way ANOVA with Sidak's post-tests, * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns = not significant. Scale bars = 50 µ m.  Figure 4A. (B,C) Immunofluorescence labeling of p62 and GFAP in hippocampus (B), and quantification of percent area for p62 (C) in mice from the same cross at 6 months of age. Values in (A) are normalized to total protein and represent fold change relative to the average for Gfap +/+ ; Stat3 fl/fl ; Cre (-) mice. C57BL/6J males and females, two-way ANOVA with Sidak's post-tests, * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns = not significant. Scale bars = 50 µm.  GFAP accumulation in adult AxD model mice with established pathology [5,6], we took advantage of a tamoxifen-inducible astrocyte-specific CreER T2 line using the Aldh1l1 promoter [24]. We first crossed the Aldh1l1-CreER T2 line to the Cre-reporter line Ai14 [25] and subsequently bred those mice with the Stat3 fl/fl line. Gfap +/R236H ; Stat3 fl/fl mice were then crossed to the CreER T2 ; Ai14; Stat3 fl/fl line. Gfap +/R236H and Gfap +/+ animals that were either positive or negative for Cre were treated with tamoxifen for 5 days at 2 months of age, and tissues collected at 3 months of age ( Figure 7A).   Finally, to assess the effects of STAT3 inhibition in astrocytes on neuroinflammation in general, we examined microglia activation by labeling with Iba1 and found that microglia appeared to return to a normal state with reduced Iba1 expression ( Figure 9A,B). Transcript analysis of the gene for Iba1 (Aif1) also showed a reduction in hippocampus and olfactory bulb after blocking STAT3 activation in Gfap +/R236H mice ( Figure 9C).  Figure 7A. Stat3 alleles in animals expressing the Aldh1l1-CreER T2 transgene are indicated as ∆ (Stat3 ∆/∆ ). (C) Quantification of p62/SQSTM1 protein in olfactory bulb and hippocampus by western analysis of a separate group of mice from the same genetic cross and treatment described in (A,B). C57BL/6J males and females, two-way ANOVA with Sidak's post-tests, * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns = not significant. Scale bars = 50 µm.
Finally, to assess the effects of STAT3 inhibition in astrocytes on neuroinflammation in general, we examined microglia activation by labeling with Iba1 and found that microglia appeared to return to a normal state with reduced Iba1 expression ( Figure 9A,B). Transcript analysis of the gene for Iba1 (Aif1) also showed a reduction in hippocampus and olfactory bulb after blocking STAT3 activation in Gfap +/R236H mice ( Figure 9C).

Discussion
GFAP expression and accumulation have been shown to correlate with AxD severity [8], and STAT3 is known to regulate GFAP transcription in development and in response to injury [14]. In mouse models of AxD, regions of the CNS with the greatest increase in GFAP expression also demonstrate significant STAT3 phosphorylation. In addition, we showed that STAT3 was activated in gray matter astrocytes in a patient with the disease. Here, we investigated whether blocking STAT3 phosphorylation could be used to manipulate Gfap expression and prevent or reverse protein accumulation in Gfap +R236H mice. Initial experiments with heterozygous Stat3 +/− knockout mice achieved significant reduction in phosphorylated STAT3-Y705, but these differences were not sufficient to consistently reduce Gfap expression. In addition to STAT3, several other transcription factors and their binding sites in the GFAP promoter have been tested for their influence on basal expression and the injury response [14]. AP-1 in particular has been suggested to regulate Gfap elevation in different injury models [31], including a transgenic mouse model of AxD that overexpresses wild-type human GFAP [7], but whether AP-1 or other factors contribute to Gfap elevation in Gfap +R236H mice has not yet been tested.
To further investigate the role of STAT3 in reactive AxD astrocytes, we used a conditional knockout approach to target Stat3 in Gfap expressing cells and found that Stat3 knockout reduced GFAP levels and prevented the reactive astrocyte response, similarly to other studies in acute injury models [12,15] and models of neurodegeneration [16,32]. STAT3 is an essential regulator of astrocyte differentiation and maturation, and by prohibiting STAT3 activation during astrocyte development, GFAP transcript and protein were reduced even in Gfap +/+ mice. AxD severity is tied to GFAP expression, and in these experiments, further transactivation of the Gfap promoter was prevented. Reduced GFAP suggests that mutant protein does not reach a toxic threshold and that astrocytes are able to maintain proteostasis and avoid a reactive stress response.
GFAP is also expressed in progenitor cells, and early activation of the Gfap-promoterdriven Cre transgene (Tg73.12) as described above leads to excision of floxed genes in both astrocytes and other differentiated cell types, including neurons [23]. To specifically evaluate the role of STAT3 activation in reactive astrocytes with existing GFAP accumulation and protein aggregation [5,6], we used an inducible CreER T2 line controlled by the Aldh1l1 promoter [24]. Knocking down STAT3 in adult mice with established GFAP pathology reduced Gfap expression and the concomitant stress and neuroinflammatory response and reversed GFAP accumulation and aggregation. We cannot exclude the formal possibility that Cre itself has an effect independent of STAT3 [33]; however, we consider this unlikely given the lack of change in markers for astro-and microgliosis (e.g., Lcn2, Aif1) in the presence of Cre in Gfap +/+ mice ( Figures 3C, 7E and 9C). These results demonstrate that STAT3 is a key factor in Gfap elevation and GFAP accumulation, and that blocking STAT3 activation may provide a therapeutic target for AxD. STAT3 activation is not necessarily detrimental, and as has been shown in other models of CNS injury or neurodegenerative disease, the consequences are context-dependent [34][35][36][37][38][39][40]. In Huntington's disease, for example, activation of JAK/STAT3 signaling improves proteostasis by increasing gene expression in proteasome and lysosomal pathways [39]. However, given the role of STAT3 in Gfap transactivation, inhibiting STAT3 in AxD would have the direct benefit of reducing levels of mutant GFAP and preventing protein misfolding and the resulting cellular toxicity. More extensive analyses of AxD tissues and controls will be needed to firmly establish the role of STAT3 in the human disorder.
Although we do not know the initiating events leading to STAT3 activation, our results support our hypothesis that activation of STAT3 contributes to a positive feedback loop that drives increased GFAP expression, protein accumulation, and ultimately astrocyte dysfunction. GFAP is one of the most highly expressed proteins in astrocytes, and a misfolded mutant form could disrupt a myriad of cellular processes, including vesicular transport and mitochondrial function, which could lead to oxidative stress and activation of MAP kinase pathways. Activation of JNK and p38 has been demonstrated in cellular models of AxD [41,42], and MAP kinases could activate AP-1 and NFκB, leading to increased IL-6 to activate the IL-6/JAK/STAT3 pathway.
In this study, we took advantage of the wealth of available genetic tools to manipulate STAT3 activation in our mouse models of AxD. Gfap +R236H mice recapitulate many of the pathological features of human disease and have been instrumental in investigating mechanisms and modifiers of the disease [43]; however, they have only subtle behavioral deficits and limited functional outcome measures [5,44]. For future studies to test whether inhibiting STAT3 pharmacologically has therapeutic value, we recently developed a rat model of AxD [19], which, in addition to STAT3 activation, exhibits motor deficits and other clinically relevant phenotypes for improved functional assessment. Inhibitors of the JAK/STAT pathway have been approved to treat inflammatory disorders, hematopoietic malignancies, and solid tumors [20,21], and JAK inhibitors have been shown to be effective in suppressing STAT3 activation in glioblastoma, experimental autoimmune encephalomyelitis (EAE) models of multiple sclerosis, and a model of Parkinson's disease [45][46][47][48]. The JAK inhibitor baricitinib was recently evaluated in clinical trials for Aicardi-Goutieres syndrome, an interferonopathy that also leads to white matter deficits, with results suggesting improved neurological function [49]. Although JAK inhibition can have adverse side effects [50,51], strategies to specifically target STAT3 with small molecules have been effective in preclinical models of Alzheimer's disease [32,52] and may provide alternatives for reducing GFAP pathology in AxD.