|Home | About | Journals | Submit | Contact Us | Français|
Molecular markers associated with CNS injury are of diagnostic interest. Mechanical trauma generates cellular deformation associated with membrane permeability with unknown molecular consequences. We used an in vitro model of stretch-injury and proteomic analyses to determine protein changes in murine astrocytes and their surrounding fluids. Abrupt pressure-pulse stretching resulted in the rapid release of 59 astrocytic proteins with profiles reflecting cell injury and cell death, i.e. mechanoporation and cell lysis. This acute trauma-release proteome was overrepresented with metabolic proteins compared to the uninjured cellular proteome, bearing relevance for post-traumatic metabolic depression. Astrocyte-specific deletion of signal transducer and activator of transcription 3 (STAT3-CKO) resulted in reduced stretch-injury tolerance, elevated necrosis and increased protein release. Consistent with more lysed cells, more protein complexes, nuclear and transport proteins were released from STAT3-CKO versus non-transgenic astrocytes. STAT3-CKO astrocytes had reduced basal expression of GFAP, lactate dehydrogenase B (LDHB), aldolase C (ALDOC) and astrocytic phosphoprotein 15 (PEA15), and elevated levels of tropomyosin (TPM4) and α actinin 4 (ACTN4). Stretching caused STAT3 dependent cellular depletion of PEA15 and GFAP, and its filament disassembly in subpopulations of injured astrocytes. PEA15 and ALDOC signals were low in injured astrocytes acutely after mouse spinal cord crush injury and robustly expressed in reactive astrocytes one day post-injury. In contrast, α crystallin (CRYAB) was present in acutely injured astrocytes, and absent from uninjured and reactive astrocytes, demonstrating novel marker differences among post-injury astrocytes. These findings reveal a proteomic signature of traumatically-injured astrocytes reflecting STAT3-dependent cellular survival with potential diagnostic value.
After trauma to the CNS, mechanical impact from a pressure wave or physical deformation force acutely inflicts cellular injuries to neurons, glia and endothelial cells. This primary trauma impact causes acute plasmalemmal permeability, mechanoporation, resulting in “membrane deformation or failure” (Barbee 2005) that is demonstrated by cellular dye uptake acutely after injuries to the brain or spine (Choo et al. 2007; Farkas et al. 2006; Whalen et al. 2008). Diffuse axonal injury is a well-recognized pathophysiology due to neuronal membrane damage (Buki and Povlishock 2006). Acute astrocyte injury and demise after trauma are reported, but these early events in glia are poorly understood and have been overshadowed by studies examining delayed reactive astrogliosis (Colombo et al. 1997; Li et al. 2012; Sofroniew and Vinters 2010; Zhao et al. 2003).
Changes in protein composition of cells after mechanical trauma inflicting membrane integrity compromise are not known. Despite proteomic studies and advances in high fidelity neurotrauma models (Kwon et al. 2011; Lubieniecka et al. 2011), insight into cellular and molecular pathophysiology early after mechanical injury remains elusive. Analyses are confounded by diverse responses including trauma-inflicted cell damage, cell death, proliferation, inflammation and reactivity that occur over various times in different cell types (Burda and Sofroniew 2014). Changing cell composition due to infiltrating cells prevents standardized tissue sampling, making it difficult to isolate cellular and molecular injury mechanisms (Kazanis 2005; Shen et al. 2014; Wanner et al. 2013). Further, tissue homogenization mixes cells, interstitial fluids and blood, precluding the measurement of protein shifts between cellular and extracellular compartments. Our rationale is to identify protein changes specific to acute cellular traumatic injury and to determine an astroglial signature of traumatic protein release into surrounding fluids. We have chosen a simple, well-characterized and controlled in vitro injury model that produces widespread and reproducible mechanical strain injury to cells grown on deformable membranes by abrupt pressure-pulse stretching producing traumatic deformation forces (Wanner 2012; Wanner et al. 2008; Ellis et al. 1995; Shreiber et al. 1999).
In this study, we determined the changes in protein composition in astrocytes and their surrounding fluids acutely after mechanical trauma (Sondej et al. 2011). The resulting protein release profiles of traumatized mouse astrocytes were related to inflicted transient mechanoporation, cell injury, or cell death. Mice with astrocyte-selective deletion of STAT3 were used to capture acute trauma-induced differences in astrocyte injury responses dependent on an acute survival role of STAT3 in traumatized astrocytes. While STAT3 activation has been shown to govern reactive gliosis after spinal cord injury (Herrmann et al. 2008; Wanner et al. 2013), we focus on its acute role in astroglial survival after traumatic injury. Our studies using genetic deletion and acute pharmacological inhibition of STAT3 show its critical role in survival of acutely traumatized astrocytes, and we identify various new STAT3-regulated astrocytic proteins.
This work provides the first fluid-derived proteomic signature of traumatized astrocytes, the mouse astrocyte trauma-release proteome. Trauma-released proteins reflect the extent of mechanoporation and cell death. Cell population fate analyses document cytosolic protein depletion, induction and posttranslational modification after mechanical trauma in injured and reactive astrocytes. This study is translationally significant because it provides a fluid-based astrocyte signature of trauma that carries potential diagnostic marker candidates.
All mice used were derived from in-house breeding colonies backcrossed onto C57/BL6 backgrounds. Mice with astrocytic STAT3 (conditional knockout, STAT3-CKO) were generated by mice having loxP sites flanking the exon 22 of the STAT3 gene, encoding a tyrosine residue (Y705) essential for STAT3 activation (Takeda et al. 1998). STAT3 loxP mice were crossed with the mGFAP-Cre mouse line 73.12. We characterized the resulting mGFAP-STAT3-CKO mice in the context of SCI (Herrmann et al. 2008; Wanner et al. 2013). Control mice (CON) included (i) no transgene, (ii) were heterozygous for mGFAP-Cre, or (iii) were heterozygous or homozygous for loxP without carrying mGFAP-Cre (Herrmann et al. 2008). All experiments were approved by the Animal Research Committee of the Office for Protection of Research Subjects at UCLA. Mice were genotyped by PCR (Hermann et al., 2008) and Western blotting.
All surgical procedures were performed under general anesthesia with isoflurane in oxygen-enriched air using an operating microscope (Zeiss), and rodent stereotaxic apparatus (David Kopf). Crush injury was made after laminectomy of a single vertebra at the level of L1/L2 using No. 5 forceps (Dumont, Fine Science Tools, tip width of 0.5 mm, fitted with a 0.25 mm spacer) compressing the cord for 5 seconds (Faulkner et al. 2004). This inflicted a moderate crush injury for CON mice but exacerbated severity for STAT3-CKO mice (Herrmann et al. 2008). Animals were evaluated blind to genotype and experimental condition. Spinal cords were analyzed 6 hours (h), 1 day (D) and 3 D post-injury.
Cortical tissue from 3–4 D old mice was dissected, cut into < 0.5 mm pieces in cold Ca2+ and Mg2+-free HBSS (Wanner 2012). Cells were isolated by trituation in medium (DMEM/F12, 10% FBS, 1:1000 gentamicin), filtered through nylon meshes (70 μm, then 10 μm) and collected by centrifuging (6 min at 380 x g). Cells of each animal were resuspended and cultured in individual T25 flask in same medium. Upon reaching confluence, proliferating cells were removed by shaking (de Vellis et al. 2009).
Astrocytes were passaged onto deformable membranes after a brief Versene rinse (137 mM NaCl, 1.5 mM KH2PO4, 2.7 mM KCl, 8.1 mM Na2HPO4, 685 μM EDTA) followed by 6 min digestion with 0.25% trypsin in 0.53 mM EDTA at 37°C that was stopped by adding medium and centrifugation at 400 x g. After resuspension in the same medium, 150–200,000 cortical cells per 2.5 ml medium were seeded onto each elastic culture well (962 mm2). Upon confluence medium with heat inactivated 10% adult horse serum was used, which was stepwise withdrawn. Cultures were serum free prior to stretching and contained postmitotic cells expressing mature markers (Wanner 2012; Wanner et al. 2008).
Cells were mechanically stretch-injured using a controlled pneumatic pressure device (2nd generation cell injury controller, CIC II, Custom Design and Fabrication Inc.). A 50 ms nitrogen pressure-pulse deformed the elastic culture bottom, causing widespread astrocyte stretching and shearing. Injury pressures between 2.1–2.7 psi lead to 11.2–12.2 mm deflection of the BioFlex II plates using the CIC II (CIC Operation Manual). Cells and media were analyzed at 5, 30 min, at 5 and 48 h post-injury.
Astrocytes from 9 CON and 8 STAT3-CKO mice were monitored on a time-lapse microscopy stage. In addition, 5 CON cultures were pretreated overnight with the STAT3 inhibitor indirubin, 10 μM E804, which was removed prior to imaging (IRE, Nam et al. 2005). Plasmamembrane permeability was assayed using 10 kD dextrans conjugated with fluoro-rhodamine or Alexa Fluor 555 that was (0.05 mM, Molecular Probes) replaced 3 min post-stretching. A 25 mm diameter circle was cut out of the silastic culture bottom using a leather puncher (arch puncher) and was placed upside down into a 35 mm glass bottom culture dish with a recessed space (Matek Corp.) fixed by a messing ring. Pre-warmed Alexa 488 conjugated caspase 3 substrate was added (10 μM, NucView 488 caspase 3 substrate, Biotium). Within 15 min the dish was transferred onto a stage top chamber with 5% CO2 at 37°C (Wheather station, Precision Control) and 6 areas per culture were monitored using a spinning disc confocal inverted microscope (Olympus IX81). Focus corrected images were obtained at 6 min intervals, using a 10X phase contrast objective (UPlan FLN, numerical aperture, NA 0.3) with appropriate filters for red (excitation λ: 535–555 nm, emission λ: 570–625 nm) and green (excitation λ: 457–487 nm, emission λ: 502–538 nm) and a CCD camera (ORCA-R2, Hamamatsu). Images were analyzed between 30 min, as caspase substrate conversion became visible and 4.5 h after stretching (SlideBookTM 4.1, Intelligent Imaging Innovations). Intact and dead cells were determined on phase-contrast images by counting nuclei of unlysed (gray) and lysed cells (phase-dark nuclei with white halo). Permeable, resealed injured cells were determined by counting fluoro-rhodamine-dextran positive cells. Apoptotic cells were determined by counting Alexa 488 caspase substrate positive nuclei. Cell fractions were determined from 7 stretched cultures of CON mice, 6 STAT3-CKO mice and 5 IRE treated cultures and normalized to their respective genotype unstretched cultures (n=5).
A 10 min propidium iodide (PI, 60 μM in pre-warmed Leibovitz’ medium, Sigma) uptake in unfixed cells determined percentages of cell death and permeability (molecular weight: 668 Da) at 5, 30 min or 48 h after stretching. PI stain was crosslinked by UV light for 5 min after rinses in same medium. Cultures were fixed for 30 min in cold, freshly depolymerized 4% PFA in TBS (pH 7.4, filtered).
A detailed proteomic protein preparation protocol is provided (Sondej et al. 2011). Conditioned medium (CM, “fluid”, 24 ml, 12 cultures per condition) at 30 min and 5 h after injury was supplemented with DTT (5 mM) and protease inhibitors (Roche, 2.5 tablets/24 ml, including 1 mM EDTA). Debris was removed by centrifuging media for 5 min at 600 x g (J-6B, Beckman). Fluid samples were concentrated to one twentieth of the volume (3 kD molecular weight cutoff centrifugal concentrators, Vivaspin, GE Healthcare), yielding a final protein amount between 1–1.6 mg. Samples were desalted using Sephadex G-25 desalting columns (PD10 spin columns, GE Healthcare) and proteins were precipitated in cold 10% TCA and then centrifuged for 10 min at 17,400 x g (RC-5C Plus, Sorvall). Pellets were rinsed in H2O followed by 10x volume of cold acetone, and stored overnight at −20°C. Samples were centrifuged (10 min, 17,400 x g, 4°C), supernatants removed and pellets were air dried. Proteins were resuspended in 500–900 μl rehydration solution (7 M urea, 2 M thiourea, 4% CHAPS, 0.2% Bio-Lytes pH 3–10, 0.0002% bromophenol blue and 0.4% DTT) and the protein concentration was assayed using a colorimetric assay read at 660 nm (Pierce).
For the first dimension isoelectric focusing gel electrophoresis, samples were centrifuged for 5 min at 16,000 x g, 4 °C, diluted to 0.5 μg protein/μl rehydration solution and 200 μl (100 μg) was loaded onto each of triplicate immobilized pH gradient (IPG) strips (11cm, pH 3–11 nonlinear ReadyStrip, Biorad). After incubating for 1 h at room temperature (RT), strips were overlaid with mineral oil and actively rehydrated overnight at 50 V, 20 °C in an isoelectric focusing cell (Protean). Isoelectric focusing proceeded by applying a maximum current of 30 μA/strip using 4 h at 100 V (linear) followed by 2 h at 250 V (linear), 5 h at 4,000 V (linear) and then for 85,000 volt-h at 4,000 V (rapid) at 20 °C.
For the second dimension, the proteins in the IPG strips were reduced using 2% DTT in 50 mM Tris-HCl (pH 8.8), 7 M urea, 20% glycerol, 2% SDS for 20 min with gentle shaking at RT and then alkylated using 2.5% iodacetamide in the same buffer. IPG strips were placed on top of precast 10.5–14% gradient SDS-PAGE gels (Criterion, Biorad), overlaid with pre-warmed low melting agarose (0.5%, 0.008% bromophenol blue in 1X SDS-PAGE running buffer), and run at 19 °C starting for 10 min at 40 V, followed by 2–2.5 h at 120 V in 25 mM Tris (pH 8.3), 192 mM glycine, 0.1% SDS along with unstained molecular weight standards (Benchmark, Invitrogen, Dodeca Criterion, Biorad). Gels were rinsed in water, fixed for 1 h in fixer solution (10% methanol, 7% acetic acid) and stained overnight at RT with Sypro Ruby (Biorad). Gels were rinsed, destained for 4 h in fixer solution and fluorescence was scanned (532 nm excitation, 555 nm LP emission filter; FX Pro Plus, Biorad).
Matching spot intensities were quantified in unstretched and stretched samples using triplicate gels per condition (SameSpots software, Progenesis). Cultures from 11 CON mice and 8 STAT3-CKO mice were analyzed. Stretched cultures of 6 CON and 4 STAT3-CKO mice were compared with unstretched cultures (5 CON, 4 STAT3-CKO) in triplicate for each sample. Spot changes with at least 1.8-fold difference and a probability of P<0.05 were defined as significant (multiple comparison ANOVA). Proteins with levels unchanged by stretching were also found (Fig. 2 A1).
Selected spots were excised from the 2D gels using a robotic spot cutter (ProPic II, Digilab). Proteins were trypsin digested (Shevchenko et al. 1996), peptides extracted and vacuum concentrated (SpeedVac, Thermo). Peptides were reconstituted with 20 μl of 2% acetonitrile, 0.1% formic acid and 10 μl (≈100–200 fmol peptides) were loaded on a C18 Nano-Trap pre-column (Dionex) using a flow rate of 5 μl/min. Peptides were separated by a C18 analytical HPLC column at a flow rate of 300 nl/min with an increasing acetonitrile gradient. Peptides were measured by LC-tandem mass spectrometry (LC-MS/MS) using a hybrid linear ion trap mass spectrometer (LTQ-Orbitrap) with a pico/silica tip ESI emitter (New Objective) and helium as collision gas. Data dependent non-redundant MS/MS spectra were acquired using: R = 100,000; full range m/z 300–2000 scan followed by 7 MS/MS scans for most intense peptide ions. Sequences of mass spectra were searched (Mascot, Matrix Science) and identified (Swiss-Prot mouse database). Criteria for protein identification for a greater than 95% confidence required a minimum of two unique peptides with an ion score greater than 34 and were manually validated.
LC-MS/MS was used to determine the proteome of mouse astrocytes and their released proteins. Unstretched astrocytes were grown in T75 flasks. Medium (11 ml/flask) was collected supplemented with EDTA (5 mM), bestatin (40 μM), pepstatin (10 μM) and phosphoramidon (10 μM), concentrated and desalted. Cells were lifted by shaking in Versene pelleted (centrifugation for 5 min at 100 x g), and then lysed in 0.5% NP40, 150 mM NaCl, 25 mM Tris, including same protease inhibitor additives. Upon homogenization with a G25 then G27½ needle, whole cell lysates (WCL) were diluted to 0.1% NP40. Part of the samples was depleted from abundant proteins using either immunoaffinity (Sigma ProteoPrep) or silicon carbide ion exchange based enrichment (Norgen ProteoSpin). Protein concentration was determined and samples were vacuum-concentrated and adjusted to pH 8 using freshly prepared 100 mM NH4HCO3 buffer. Proteins were reduced in the same buffer with 5 mM DTT for 20 min at 50°C followed by 5 min at 96°C, and then alkylated with 10 mM iodacetamide in 100 mM NH4HCO3 for 45 min at 37°C. Fluid samples of 10–50 μg protein per 50 μl medium and cell lysates of 20–90 μg protein per 100 μl WCL were trypsin digested overnight at 37°C using 1–2 μg trypsin. Digested protein (1.0–1.5 μg) was loaded on the C18 Nano-Trap pre-column at 5 μl/min for 5 min and separated by a C18 HPLC column at a flow rate of 300 nl/min with an increasing acetonitrile gradient. Peptides were measured by LC-MS/MS using the same instrument (LTQ-Orbitrap).
Fluid (CM) proteins were prepared as described omitting desalting and precipitation. WCL proteins were prepared by adding 150 μl/well of ice cold lysis buffer (150 mM NaCl, 20 mM Tris, 5 mM EDTA, 10 mM sodium pyrophosphate, 30 mM phosphatase substrate, 5 mM DTT, 1% NP40 and 1 protease inhibitor tablet per 10 ml buffer). Cells were collected and homogenized by trituation, passaged through a 27½ gauge needle avoiding bubbles, and protein concentration was determined.
One fifth of Laemmli sample buffer (6x) was added to samples (125 mM Tris-HCl, pH 6.8, 10% glycerol, 2% (w/v) SDS, 0.1% (w/v) bromophenol blue), samples were denatured by boiling for 5 min at 100°C, and placed on ice. Protein (30 μg) was loaded on 10% or 15% Tris-HCl polyacrylamide handcast gels (Mini-PROTEAN, Bio-Rad) along with a molecular weight standard (Precision Plus Protein™ Kaleidoscope, Bio-Rad). Gels were run in SDS running buffer (200 mM glycine, 25 mM Tris Base, 0.1% (w/v) SDS) and blotted onto nitrocellulose membranes (Amersham Hybond-ECL, GE Healthcare) in 200 mM Glycine, 25 mM Tris (pH 8.3), 20% methanol. Blots were stained for 10 min with 0.1% Ponceau S in 5% acetic acid and gels were stained overnight at 4°C using 0.4% Brilliant Blue R, 45% methanol, 10% acetic acid. Membranes were washed with Tween (0.05%) supplemented Tris buffered saline (TBST) and incubated in 10% dry milk in TBST for 1 h at RT. Washed blots were incubated in primary antibody diluted in 5% BSA in TBST overnight at 4°C (Table 1). After rinsing, blots were incubated for 1 h at RT in secondary antibodies coupled to horseradish peroxidase diluted in 10% milk in TBST (Table 2). Proteins were visualized using enhanced chemiluminescent substrate (Supersignal West Pico, or Femto; Thermo Scientific) and autoradiography film (Hyblot CL, Denville Scientific).
Western blots of cultures from 11 CON and 9 STAT3-CKO animals were analyzed. A ranked intensity score was assigned to each of the 13 proteins (0, no signal; 4, maximal signal within a blot) of stretched and unstretched, CON and STAT3-CKO samples processed together. Protein release was implied when an intracellular protein, based on its gene ontology (GO) assignment, was elevated (≥0.5 difference from base level) in the fluid in at least 4 different animals on Western blots and was significantly increased on 2-D gels (Table 3). Statistical comparisons of injury and genotype mean intensity differences were made using a non-parametric repeated mixed measure ANOVA model and resampling, since ordinal intensity data did not follow the normal distribution (Efron and Tibshirani 1991). Means that followed the normal distribution were compared using t-tests or parametric one way ANOVA. Significance was assessed using the Tukey-Kramer criterion to adjust for multiple comparisons. Trauma-induced changes were compared to baseline. Differences between CON and STAT3-CKO astrocytes were determined in unstretched WCL samples from 5–8 genotype culture pairs. Injury differences between genotypes were compared in stretched CON and STAT3-CKO cultures correcting for variability due to different penetration of the STAT3 deletion. Spearman’s correlations were computed between protein release amounts and cell death. Factor analysis based on Spearman correlations determined commonality among multiple proteins based on genotype. Statistical analyses were carried out using SAS9.4 (SAS Inc.) and Instat (GraphPad Inc.).
After rinsing the cells, non-specific binding was blocked by 5% normal donkey serum (NDS). Rabbit anti-PEA15 and chicken anti-GFAP were applied overnight in blocking solution at 4 °C (Table 1). Donkey anti-rabbit Alexa Fluor 488 and anti-chicken Alexa Fluor 647 were diluted in the same blocking solution and centrifuged for 30 min at 13,000 x g, 4 °C and incubated for 1 h in the dark at RT on rinsed cultures (Table 2). After rinsing, cultures were stained for 5 min with 10 μg/ml bisbenzimide in H2O (Hoechst, Sigma), rinsed, dried and coverslipped using a 25 mm circular cover glass (491 mm2) and solidifying anti-fade mounting medium (Fluoro-gel, Electron Microscopy Sciences). Dried membranes were cut and mounted onto glass slides with 8% gelatin solution. Slides were fitted with tape strips (3M) for imaging on an inverted fluorescence microscope (Olympus IX70) equipped with a CCD camera and capturing program (Axiocam, Axiovision, Zeiss).
Cultures from 4–6 CON and 3–4 STAT3-CKO animals for each condition (unstretched, 5, 30 min and 48 h post-stretching) were stained. Twelve pictures per culture were analyzed using a 20x objective (UPLAN Apo, NA 0.7, PH2) and 600 to 1140 cells per culture were scored for incidences of PI-, PEA15- and GFAP-positive cells. GFAP cytoskeletal filament assembly, brightness subcellular localization were scored. Filament disassembled GFAP signals appeared homogeneous and blurred adjacent to filament assembled GFAP displaying typical cytoskeletal lines, supporting a structural change in selected cells after stretching. The proportion of PI-positive cells among the PEA15 and the various GFAP populations was determined. Astrocytes with granular GFAP were excluded. “Intact cells” were PI-negative with oval nuclei with pale Hoechst staining. “Dead cells” had shrunken, pyknotic, raisin-shaped nuclei with condensed, bright Hoechst-stained nuclei. The “acute” post-injury dataset combined 5 and 30 min post-stretching. Significant differences in numbers of PEA15- and GFAP-positive cell populations and PI-positive cells among them were determined using post-hoc multiple comparison ANOVA (Tukey, Instat).
After terminal anesthesia by barbiturate overdose, mice were perfused transcardially with buffered 4% PFA. Spinal cords were removed, post-fixed for 2 h and cryoprotected in buffered 30% sucrose overnight. Horizontal, 40 μm frozen sections were cut using a cryostat microtome (Leica).
For enzyme-enhanced immunohistochemistry, endogenous peroxidase was quenched for 15 min using 1% H2O2 in 100% methanol, then sections were permeabilized for 45 min in 0.5% Triton X100, rinsed and incubated in 5% normal goat serum in TBS. Rabbit anti-aquaporin 4 and rabbit anti-GFAP were combined in the same blocking solution and incubated overnight at 4 °C (Table 1). After rinsing, biotinylated anti-rabbit antibody was incubated in blocking solution for 1 h at RT (Table 2). After incubation and rinsing an avidin-biotin-peroxidase complex (ABC kit, Vector) was prepared and applied to the sections. Rinsed sections were equilibrated in 100 mM sodium acetate buffer and the freshly prepared, sonicated and filtered modified DAB substrate solution (0.5 mg DAB/ml, 2 mg β-D-glucose /ml, 6 units glucose oxidase /ml and 25 mg nickel ammonium sulfate /ml in 100 mM sodium acetate buffer) was applied at RT for 5–10 min and stopped by rinses in acetate buffer (Shu et al. 1988). Sections were rinsed, thaw-mounted onto glass slides in 0.5% gelatin solution, dried and dehydrated in an increasing ethanol series, rinsed in xylene and mounted (Permount) with weights for 20 min at 70 °C on a slide warmer. DAB-stained sections were photographed on an upright brightfield microscope (AX70, Olympus) and images of various focus levels were overlaid in Photoshop (Adobe).
For immunofluorescence (IF) staining endogenous tissue autofluorescence was blocked by incubating water-equilibrated sections in a 0.5% Sudan black solution in 70% ethanol for 5 min followed by rinses. Sections were permeabilized in 0.5% Triton X100 in TBS for 20 min, rinsed and incubated for 2–4 h in 5% NDS in TBS. Goat anti-ALDOC and mouse-anti-CRYAB were each combined with rabbit anti-GFAP. Rabbit anti-PEA15 was combined with chicken anti-GFAP in 5% NDS in TBS and sections were incubated overnight at 4 °C. After rinsing, anti-goat Alexa 488 combined with anti-rabbit-Cy3, anti-mouse Cy3 combined with anti-rabbit Alexa 488 and anti-rabbit-Alexa 488 combined with anti-chicken Cy3 were prepared in blocking solution, centrifuged for 30 min at 13,000 x g at 4 °C and incubated on the sections for 1 h at RT (Table 2). Rinsed sections were stained for 5 min with 10 μg/ml bisbenzimide in H2O. After final rinses, sections were dried and coverslipped as above. Sections were analyzed using a confocal laser scanning microscope (LSM 510, Zeiss) with a 63x oil immersion objective (PLAN Apochromat, NA 1.4), a 488 nm argon laser and a 543 nm helium-neon laser with a 505–530 nm bandpass filter for Alexa 488 fluorescence, and a 560–615 nm bandpass filter for Cy3 fluorescence. The same pinhole (1 μm) and photomultiplier settings were used for all images and scans over a depth of 4 μm were overlaid.
Cell death rates rose with increasing pressure intensities 48 h post-injury with exacerbated dose-response relationship in STAT3 deficient astrocytes (STAT3-CKO) compared to controls (CONs, Fig. 1A). Stretch pressures of 2.4 ± 0.3 psi caused cell death in one fifth of stretched CON and nearly half of STAT3-CKO astrocytes, a 2.4 fold difference (Fig. 1B). Post-injury membrane permeability was monitored by a 10 kDa rhodamine-conjugated dextran and apoptosis using a green fluorescent caspase 3 substrate. Time-lapse video analysis shows that 22% of astrocytes stained with rhodamine-dextran during stretching retained the dye following washing (Fig. 1C). These “resealed injured” astrocytes were membrane porated during stretching and regained cell integrity afterwards, which is a new finding. Acute cell lysis occurred in 23% of CON astrocytes, without significant additional subsequent necrosis, while rates were higher in STAT3-CKO astrocytes, and necrosis continued over time post-injury (Fig. 1D). Apoptotic CON cells rose from 10% at 30 min to 25% by 4.5 h post-injury (Fig. 1D), with some resealed injured cells becoming apoptotic (Fig. 1C4). There were hardly any apoptotic cells among STAT3-CKO stretched astrocytes anytime post-injury (Fig. 1C8, 1D). Most leaky STAT3-CKO astrocytes underwent lysis without activating caspase 3 by 4.5 h post-injury (Fig. 1D). Thus, abrupt pressure-stretching caused transient membrane permeability followed by apoptosis and necrosis that was elevated in STAT3-deficient astrocytes, showing reduced survival and apoptosis; hence they were less stretch-resilient than controls.
We next tested whether STAT3-CKO astrocytes were more vulnerable due to a pre-existing deficiency of STAT3 regulated proteins (Sriram et al. 2004). We inhibited STAT3 prior to stretch-injury in non-transgenic astrocytes using the indirubin derivative E804 (IRE) that inhibits STAT3 activation without being toxic (Nam et al. 2005). All astrocytes had initially similar membrane permeability rates after stretching, yet dye-retaining rates were poor in IRE-treated (not shown) and STAT3-CKO astrocytes compared to controls by 4.5 h post-injury. IRE reduced apoptosis by 22% after stretching compared to untreated CON astrocytes (at 4.5 h, SEM is 6 in CON, 1 in IRE treated cultures, n=5 cultures, P<0.01, Tukey ANOVA). Stretch-induced necrosis rates of IRE-treated were similar to those of STAT3-CKO astrocytes and significantly higher than those of untreated CON astrocytes (by 23%, at 4.5 h, SEM is 2 in CON, 4 in IRE treated cultures; P<0.01, Tukey, ANOVA). Thus, we show for the first time that both, pharmacological inhibition and genetic deletion of STAT3 elevated necrotic cell death and suppressed apoptosis, making astrocytes without STAT3 activation more vulnerable to mechanical trauma and unable to overcome integrity compromise.
We next determined the release profile of traumatized astrocytes using a proteomic approach chosen for its robust and unbiased detection of protein differences. We reproducibly measured ~ 700 protein spots derived from fluid, i.e., conditioned medium (CM) from unstretched and stretched cultures of both genotypes at 30 min and 5 h post-injury. Spots with reproducibly different intensities (≥ 1.8x increase or ≤ 0.6x decrease) after stretching in either genotype were selected, excised and digested. We identified 59 trauma-influenced proteins by mass spectrometry (Fig. 2A), 53 proteins increased in CM (Fig. 2A2; Table 3) while 6 proteins decreased. Three change profiles were found at 30 min (Figure 2B) and similarly, at 5 hr after stretching (not shown). Example protein spots for each of these profiles with corresponding protein immunoblots confirmed the observed profiles (Figures 2C and D). 1: Similar stretch-induced protein increase in fluid of CON and STAT3-CKO cultures, as seen in Ywhae 14-3-3ε (1433E) and aldolase C (ALDOC, Fig. 2B1, C1, D1). 2: Larger protein increase in STAT3-CKO versus CON culture fluids post-injury, e.g. Cu-Zn superoxide dismutase 1 (SODC, Fig. 2B2, C2, D2). 3: Only CON fluids showed stretch-induced protein increase not -STAT3-CKO fluids, e.g. apolipoprotein E (APOE, Fig. 2B3, C3, D3). Thus, mechanical trauma caused acute and substantial release of astrocytic proteins with STAT3 modulated release profiles.
Gene ontology (GO) analysis shows that released proteins originated from all cellular compartments and across all functional categories (Fig. 3B, E). The largest group of released proteins was organelle-derived (37%), followed by cytoplasmic ones (23%, Fig. 3C for 30 min post-injury, 5hr data were similar, not shown). STAT3-CKO cultures had overall 1.7 fold more trauma-released proteins than CON cultures (Fig. 3A, 30 min post-injury, 5hr data were similar, not graphed, see Table 3). Proportionally more nuclear proteins and protein complexes were released from STAT3-CKO versus CON stretched astrocytes, including filamin-A, an ATPase (OLA1) and β actin (ACTB, Fig. 3B).
To determine whether trauma caused selective or unselective protein release from the cellular pool, we identified all proteins of CON mouse astrocytes by proteomic mass spectrometry and (LC-MS/MS) compared their cellular compartments and functional assignments with that of trauma-released proteins (Fig. 3C, F). Trauma-released proteins originated from all subcellular compartments and fluids similarly (with the exception of membrane-bound organelles, which were nearly doubled (Fig. 3C1+2). This correlation was comparable for STAT3-CKO cells (not shown). Consistent with finding organelle-derived proteins, including peroxiredoxin 2, phosphatidyl ethanolamine binding protein (PEBP-1) and S-formylglutathione hydrolase/esterase D (ESTD), we frequently found dye-incorporated vesicles during lysis of membrane-permeable astrocytes after stretching Cytosolic dye uptake by membrane-wounded cells appeared in vesicles just prior to disappearing as cells lysed. (Fig. 3D). This reflects an organelle- mediated process, like vesicle-shedding, which was associated with trauma induced secondary cell death and might contribute to protein loss.
Extracellular space and plasma membrane proteins were common in CM of unstretched cultures (25% of the assignments, Greco et al. 2010), yet they were 2–3 fold under-represented in the trauma-release proteome (Fig. 3C2). Among these were classically secreted proteins, like APOE, gelsolin, haptoglobin, and inter-alpha-trypsin inhibitor that were decreased in fluids acutely after stretching (Fig. 2B3, C3, D3; Table 3). This finding suggests reduced active protein secretion as a new aspect of astroglial traumatic injury.
Considering astroglia’s major roles in brain metabolism, inflammation and reactive gliosis, we analyzed the biological processes of the trauma-released proteins (Fig. 3E). Metabolic proteins were the largest acute trauma-release group of CON astrocytes after stretching at both times post-injury (Figs. 3E, F). Together with catabolic proteins, they contributed to one third of the trauma-release proteome. STAT3-CKO astrocytes had similar metabolic and catabolic proteins released. For example, GAPDH had the largest release amount from injured astrocytes in both genotypes (Table 3). Proteins of cell component organization, communication and transport were released more selectively by STAT3-CKO than CON injured astrocytes (Fig. 3E, shown are 30 min post-injury, 5hr had similar release profiles, not shown). Together, over-represented release of nuclear proteins and protein complexes from stretched STAT3-CKO astrocytes (Fig. 3B), suggests that cell lysis contributed more to the STAT3-CKO than to the CON release profile, which is consistent with finding elevated necrotic cell death in STAT3-CKO versus CON traumatized astrocytes (Fig. 1D).
We next compared biological processes of the trauma-released proteins with those of the cellular proteome of CON astrocytes (Fig. 3F). Again, most functional categories (9 out of 13) were released proportionally to their presence inside uninjured cells, except metabolic plus catabolic proteins, which were nearly doubled in the trauma-release proteome relative to their representation in cells. Release of metabolic proteins contained enzymes processing carbohydrates, such as aldolases, enolase, GAPDH, cytoplasmic isocitrate, malate dehydrogenases and LDHB. Amino acid processing enzymes like glutamine synthetase and nitrogen compound related proteins including SODC, nucleoside diphosphate kinase B, and PEBP1 were also abundantly released. In contrast, proteins of cell communication, transport and cell component organization were underrepresented compared with the LC-MS/MS cell proteome (Fig. 3F, 30 min trauma-release profile, similar data at 5hr post-injury, not shown). Biological process assignment of STAT3-CKO released profiles was not compared to their whole cell lysate profile, as substantially higher cell death rates were observed (Fig. 1D). In summary, metabolic pathways including glycolysis, TCA cycle, ATP biosynthesis, nitrosylation, and response to oxidative stress were most affected by acute trauma-inflicted protein release. This new data may contribute to a reported energetic and metabolic depression in mechanically traumatized astrocytes in the early hours after TBI (Bartnik et al. 2007; Scafidi et al. 2009).
Proteomic changes were validated using Western blotting for 13 proteins, selected based on their enrichment in astrocytes (Cahoy et al. 2008) or due to genotype differences (Fig. 2B). CON cells, not STAT3-CKO cells, rapidly increased cellular levels of GFAP, PEA15 and vinculin (VINC) by 30 min after stretching compared to unstretched cells (Fig. 4B, see also Fig. 7A, B). GFAP increase reflects fast post-translational modification rather than de novo synthesis (see Discussion). PEA15 and VINC increases are new findings. Trauma-release (“fluid”) and cellular reduction (“cells”) were determined at 30 min and 5 h post-stretching for both genotypes (Fig. 4B). Repeated immuno-blots confirmed robust fluid elevation for all 13 proteins at both time points post-injury, except SODC in CON fluid, which was elevated only 5 h after stretching (Fig. 4B, Fig. 2D2). Representative immunoblots for 9 proteins are shown, either at 30 min (Figs. 4A, 7A, B), with similar release by 5 h after stretching (not shown), or, at 5 h (Figs. 2D, ,6A),6A), with similar release already at 30 min post-injury (not shown).
Cellular protein levels did not decrease markedly at 30 min after stretching in CON cell lysates collectively (Fig. 4B); while protein changes in individual cells are addressed below. By 5 h post-injury, CON astrocytes had reduced cellular levels for transgelin (TAGL), CRYAB and 1433E (Fig. 4B, blots not shown). In contrast, stretched STAT3-CKO cells had already by 30 min reduced cellular levels of SODC, VINC, 1433E and GAPDH and additional proteins were reduced by 5 h post-injury including PEA15 (Figs. 4A, B, ,7A).7A). Most proteins were reduced in STAT3-CKO astrocytes compared to unstretched levels, except GFAP and ACTB (Fig. 4B, blots shown in Figs. 6A1, ,7B).7B). Thus, our approach documents for the first time a substantial protein flux from cells to fluids acutely after traumatic astrocyte injury, which was greater in STAT3-CKO versus CON cultures.
Immunoblots show that STAT3-CKO cultures released larger amounts of GAPDH than CON cultures and had greater depletion of cellular GAPDH levels post-injury (Figs. 4A1, 4B). ACTN4 release was also larger in stretched STAT3-CKO versus CON cultures, with cellular post-injury levels unchanged at first in both genotypes, but reduced by 5 h in STAT3-CKO stretched cells (Fig. 4A2, B). LDHB had lower signal in STAT3-CKO versus CON cells by 5 h post-injury, yet fluid signals were similar in both genotypes but unstretched LDH base expression was reduced in STAT3-CKO versus CON astrocytes (Fig. 4A3, ,5B).5B). This prompted us to compare base protein expression in the two genotypes (Fig. 5). Proteins with consistently lower expression in STAT3-CKO compared to CON astrocytes included GFAP, ALDOC, LDHB and PEA15 (Figs. 4A3, 5A1, B and 7A, B). GFAP expression is reported to be regulated by STAT3 activation (Sriram et al. 2004; Herrmann et al., 2008), whereas ALDOC, PEA15 and LDHB are novel STAT3-dependent astrocytic proteins. Expression of GAPDH, TPM4 and ACTN4 was higher in STAT3-CKO versus CON astrocytes, suggesting that these targets were suppressed by STAT3 (Figs. 5A2, B, 4A1–2). Alternatively, they could have been increased to compensate for downregulated proteins in STAT3-CKO cells (Di Domenico et al. 2010; Herrmann et al. 2008, Fig. 5B). The covariation factor plot segregated CON and STAT3-CKO by two underlying factors. PEA15, LDHB and ALDOC made Factor 1, while TPM4, GAPDH, ACTN4 and GFAP made a second factor. Together these two factors segregated astrocytes by their genotype with CON astrocytes showing higher values for factor 1 and STAT3-CKO showing higher values for Factor 2, supporting a role for STAT3 in regulating these proteins.
To correlate genotype and trauma induced protein release amounts, we chose 5 proteins with similar base expression in the two genotypes (Fig. 5B). SODC, ACTB, VINC, CRYAB and TAGL release was larger in stretched STAT3-CKO versus CON cultures at either or both post-injury times (Figs. 2C2, D2, 6A, B). Cellular levels for SODC, CRYAB, TAGL and VINC decreased further after injury in STAT3-CKO than in CON cells, except those of ACTB that were saturated (Figs. 4B, 6A, B). Factor analysis grouped VINC, SODC and CRYAB into Factor 1, yielding higher values for STAT3-CKO than CON cells, thereby segregating the genotypes in the factor plot. This demonstrates trauma-inflicted protein covariance based on genotype. We next correlated acute post-injury release of SODC with delayed cell death rates in CON and STAT3-CKO astrocytes (Fig. 6D). Less SODC release acutely post-injury associates with lower cell death rates in CON cells, whereas larger acute SODC release associates with higher delayed cell death rates in STAT3-CKO astrocytes. Together, these data support the idea that acute fluid levels of these markers can signify the magnitude of astrocyte loss, an outcome measure of this injury model.
CON cells increased PEA15 and GFAP signals 30 min after stretching but not STAT3-CKO cells (Figs. 4B, 7A, B). PEA15 fluid levels were similar in both genotypes, whereas those of GFAP were lower in STAT3-CKO versus CON stretched cultures (Fig. 7A, B). In addition to the 50 kDa full size GFAP, stretched CON cells showed a 47 kD GFAP and fluids showed a 37 kD sized GFAP specific signal (Fig. 7B). These proteolytically cleaved GFAP fragments are reported after calcium dependent GFAP proteolysis of in situ degraded spinal cord tissue and cultured cortical astrocytes (Ciesielski-Treska et al. 1984; DeArmond et al. 1983).
To better understand these patterns, we analyzed membrane leakage using PI-uptake and monitored it with PEA15 and GFAP staining in individual cells. Uniform PEA15 expression in all unstretched astrocytes changed into a patchwork of PEA15-positive and -negative cells acutely after stretching, documenting striking diversity of PEA15 absence and presence among injured astrocytes (Fig. 7D). For GFAP, we found 84% of CON unstretched astrocytes expressed GFAP as reported (Fig. 9A, Chiu et al. 1981). Stretching caused GFAP loss and staining changes that were consistent with filament disassembly (see Methods) and associated with permeable cells (Table 5), suggesting post-translational modifications in response to cell wounding (Fig. 7C2, E3, 4). Bright GFAP signals after stretching resemble those seen rapidly after elevating intracellular calcium levels (Lee et al. 2000). Stretched astrocytes with intact membranes retained PEA15 and GFAP was filamentous. Some permeable cells lacking PEA15 still showed GFAP, but with disassembled filaments (Fig. 7D3,4, E3,4). By 48 h post-injury, CON cultures displayed reactive astrocytes with elongated thick processes and elevated PEA15 and GFAP expression compared to signals in unstretched astrocytes and PI-stained nuclei indicated dead astrocytes (see Methods, Fig. 7D5,6, E5,6).
Quantifying cell integrity, fates, and marker pattern showed that stretching caused PEA15 loss in 20% of CON and 40% of STAT3-CKO astrocytes (Figs. 1, ,8A,8A, Table 4). Most stretched astrocytes retaining PEA15 had intact membranes while the majority of PEA15 depleted astrocytes were permeable in both genotypes except 22% of CON astrocytes that displayed intact membranes (Fig. 8B1,2, Table 4). These cells might have been leaky during stretch, released PEA15, and resealed prior to the permeability assay. Time-lapse imaging showed fast resealing after stretching in a similar fraction of injured astrocytes that endure after rhodamine-dextran uptake as “resealed injured” cells (Fig. 1). Together, immunoblot and cell population analyses conclude that PEA15 was released from twice as many STAT3-CKO than CON astrocytes, resulting in similar PEA15 fluid signals in STAT3-CKO and CON stretched cultures despite lower PEA15 ‘base’ expression in STAT3-CKO astrocytes (Fig. 7A).
There were significantly fewer GFAP-positive cells in STAT3-CKO versus CON cultures (Table 5, Fig. 9A). A notable group of STAT3-CKO astrocytes displayed mislocalized GFAP that did not form filaments, but appeared as granular aggregates similar to those found in inclusion bodies (Franke et al. 1982). Resulting Filamentous GFAP-positive population of CON astrocytes was 83% and 48% of STAT3-CKO unstretched astrocytes, excluding cells with granular GFAP distribution (Figs. 7C1, 9 A, Table 5). Stretching reduced the GFAP-positive CON population to nearly half and lowered it by one-third in STAT3-CKO cultures (Fig. 9A, Table 5). Thus, lower GFAP fluid levels in STAT3-CKO versus CON cultures were due to release from fewer cells of a smaller GFAP-positive population with lower GFAP expression (Figs. 7B, ,9A).9A). Surprisingly, stretch-induced GFAP depletion occurred in twice as many cells as PEA15 depletion (Tables 4, ,5,5, Figs. 8A, ,9A,9A, P<0.01, Tukey). Permeabilizing astrocytes is reported to result in varying depletion ratios of different astrocytic proteins, the cause of which is unknown (Karl et al. 2000, see Discussion).
GFAP loss and filament disassembly resulted in a residual 25% of cells with regular, filament assembled GFAP in CON and only 8% in STAT3-CKO cultures acutely after stretching (Table 5, Fig. 9A). In STAT3-CKO cultures, cell with disassembled GFAP were present in similar fractions acute and two days post-injury, while their fraction decreased in CON cultures by 2 days after stretching (P<0.05, Tukey, Table 5). Over the same time period CON cultures showed a 27% increase in cells with assembled GFAP filaments (P<0.05, Tukey, Table 5). To what extent new GFAP expression in reactive astrocytes and filament reconstruction in injured cells contribute to this ‘recovery’ remains to be determined.
We next correlated membrane integrity with GFAP filament assembly after injury. Almost a quarter of CON and nearly half of STAT3-CKO GFAP expressing cells were permeable acutely after injury (Fig. 9B, Table 5). Most cells with intact membranes had filamentous GFAP whereas the largest fraction of permeable cells was among astrocytes bearing disassembled GFAP in both genotypes acutely after stretching (Fig. 9B). A small cell fraction with bright GFAP signals had highest permeability rates (Table 5, P< 0.05, Tukey). Thus, astrocytes with disassembled GFAP filaments and altered GFAP antigenicity associated with membrane wounding acutely post-injury. GFAP was more frequently retained in permeable cells than PEA15, but cellular GFAP underwent these structural changes after injury (Tables 4, ,5,5, Figs. 8B, ,9B,9B, P<0.01, Tukey). In sum, mechanical trauma inflicted membrane wounding in a subpopulation of homogeneous astrocytes that affected astroglial markers PEA15 and GFAP differently in release, breakdown and cellular organization.
Studies on acutely injured astrocytes are sparse compared to reports on reactive gliosis. To begin investigating astroglial cell injury and distinguish injured from reactive astrocytes, we chose a well-characterized moderate crush spinal cord injury (SCI) model (Faulkner et al. 2004; Herrmann et al. 2008; Wanner et al. 2013). This model has the advantage that reactive gliosis and scar formation are structurally, temporally and spatially well-defined (Wanner et al., 2013). To this end, we first identified structural features that distinguished acutely injured from reactive astrocytes; we then analyzed selected trauma release markers in pathologically defined astrocytes after crush SCI (see below).
Damaged tissue with injured and dead astrocyte profiles was identified 3 days after SCI when it was easily distinguished from the surrounding penumbra populated by reactive astrocytes (Fig. 10A). Reactive astrocytes typically had hypertrophied cell bodies and thickened processes (Fig. 10D), whereas damaged gray matter tissue near the lesion had astroglial debris and astrocytes with truncated processes (Fig. 10C). Injured white matter tissue contained more astrocytes with tortuous, beaded or disintegrated processes (Fig. 10C), reminiscent of diffuse axonal injury after mechanical shear forces that have been rarely described in traumatically injured astrocytes (Colombo et al. 1997; Farkas et al. 2006).
Both genotypes had pathologically swollen astrocytes with disproportionally enlarged cell bodies and thin or shortened processes in damaged gray matter tissues 6 h after SCI, but they were larger in STAT3-CKO than in CON damaged gray matter (Fig. 11A). By 1 day post-injury most STAT3-CKO astrocytes were absent in injured gray matter and few remaining astrocytes were degenerating while CON astrocytes endured (Fig. 11A). Injured white matter contained reactive and damaged astrocytes with tortuous and beaded processes in CON cords, while reactive astrocytes were missing in STAT3-CKO cords (Fig. 11B). STAT3-CKO white matter damage zones extended 1.7 fold further from the lesion center into adjacent neural tissues than CON damage zones (Fig. 11C). Poor recovery of STAT3-CKO mice after crush SCI is reported (Herrmann et al. 2008; Wanner et al. 2013). We now document larger damage zones and more prominent acute astroglial pathology and demise in STAT3-CKO versus CON injured cords, providing evidence for a role of STAT3 in astrocyte survival.
After quantifying injured cell populations and their marker release into fluids in vitro, we next addressed whether the identified release markers are present in the crushed spinal cord, providing polarity of a defined injury site. Presence of four astrocyte enriched trauma-release markers was determined in uninjured, injured and reactive astrocytes at the cellular level, rather than by immunoblotting of tissue homogenates to avoid inconsistent tissue compositions after SCI between the two genotypes (Wanner et al. 2013) and to investigate different cell fates and markers coexisting in neighboring areas. Acutely after SCI, injured astrocytes were located near the lesion, while reactive astrocytes in CON cords were more distally around the lesion (see Fig. 10). In uninjured spinal cords ALDOC was robustly expressed in CON astrocytes, but signals were weak in STAT3-CKO astrocytes, supporting the relevance of in vitro quantified STAT3-dependent ALDOC expression (Figs. 12A, ,5).5). Cytosolic ALDOC colocalized with GFAP, but ALDOC was also present in cell bodies and fine process endings, making it a useful astrocyte-specific marker. PEA15 showed expression in some uninjured CON astrocytes, but signals were weak in STAT3-CKO astrocytes (Fig. 12B). In uninjured mouse spinal cords CRYAB was not colocalized with GFAP expressing astrocytes of either genotype (Fig. 12C). CRYAB was chosen based on its role after SCI and its interaction with GFAP (see below, Klopstein et al. 2012; Koyama and Goldman 1999).
After SCI, ALDOC signal was depleted in injured astrocytes near the lesion at 6 hr post-SCI, but was present in enlarged cell bodies of injured astrocytes one day later (Fig. 13A and insert). Reactive astrocytes in the penumbra had strong ALDOC and GFAP expression (Fig. 13A). Injured STAT3-CKO astrocytes extended farther in lesioned cords and had only marginal ALDOC signals (Figs. 11C, 13A). PEA15 was depleted 6 h after injury (Fig. 13B, insert). By one day post-SCI, few injured CON astrocytes displayed marginal PEA15 signal, while it was expressed in reactive astrocytes distal from the lesion (Fig. 13B). STAT3-CKO, injured astrocytes had marginal PEA15 signals (Fig. 13B). Thus, ALDOC and PEA15 are newly in vivo confirmed STAT3 regulated astrocyte markers in addition to GFAP (Herrmann et al. 2008; Wanner et al. 2013). They are acutely lost from injured astrocytes, but robustly expressed in territorial, reactive astrocytes after SCI, serving as new markers for reactive gliosis.
At 6 h after SCI near the lesion, injured astrocytes clearly expressed CRYAB in both genotypes (Fig. 13C). In contrast, no reactive astrocytes expressed detectable CRYAB (Fig. 13C). CRYAB was colocalized with GFAP in pathologically swollen astrocytes with pruned processes (Fig. 13B and insert). In these astrocytes reduced GFAP signals appeared faint and fragmented resembling stains in injured astrocytes after stretching (Figs. 13C, 7C, E). CRYAB associates with GFAP in pathological aggregates called Rosenthal fibers and it has a function in regulating GFAP assembly (Hagemann et al. 2009; 2006). CRYAB expression remained in injured astrocytes one day post-SCI whereas adjacent reactive astrocytes were CRYAB-negative (Fig. 13C). CRYAB colocalized with decreased GFAP signals in wounded astrocytes supports its acute function as a heat shock protein and cytoskeletal organizer after mechanical trauma. Decreased, filament disrupted GFAP appears as a feature of swollen astrocytes near SCI lesions and mechanoporated astrocytes after stretching (Figs. 7C, E, ,9).9). These mouse SCI data document astroglial marker differences that demonstrate a new biochemical diversity that is related to specific astrocyte states of cell injury and reactivity, which were both affected by the deletion of cell survival-supporting transcription factor STAT3.
We used a simple mechanical astrocyte trauma model with and without STAT3 deletion, measured proteome profiles, and identified proteins released after pressure-pulse stretching. Cell fate after trauma was related to release profiles that reflect cell injury, mechanoporation, and cellular outcomes of survival, reactivity and cell death. Our studies are the first to reveal subpopulations of traumatized astrocytes based on proteome profiles and specific astrocytic marker elevation, depletion and post-injury modification that were validated in vivo. This work will help elucidate a signature of astrocyte injury and demise in biofluids after CNS trauma. Our results showed that astrocytes responded with mixed vulnerability to mechanical trauma, which we have begun to investigate in vivo.
Mechanical trauma causes temporary and different degrees of plasma membrane wounding (Barbee 2005) that can explain the large variability observed in culture trauma models (Cullen et al. 2011; Geddes et al. 2003;). Various injury responses and cell fates are systematically documented as a homogeneous astrocyte population that turns into a mosaic of intact, permeable, resealed and dead cells after a simple mechanical trauma pulse, a finding relevant to acute in vivo traumatic brain and spinal cord injuries. Dye-uptake reveals perforated membranes of rat cortical neurons after a traumatic impact with some regaining integrity while others endured plasmalemma wounding or underwent delayed integrity compromise (Farkas et al. 2006). After central cortical fluid percussion populations of rat neurons are acutely porated, reseal, endure or re-open associating with different ultrastructural morphologies including mitochondrial irregularity, vacuolization and neurofilament disassembly (Lafrenaye et al. 2012; Tang-Schomer et al. 2010). Diffuse axonal injury causes axolemmal permeability with swelling and bead formation (Buki and Povlishock 2006; Kilinc et al. 2009). Mechanoporated cells accumulate around the lesion acutely after mouse SCI (Choo et al. 2007). When following porated cells using dye pulse-labeling over time after controlled cortical impact, Whalen and colleagues found dye overlap with pyknotic cells and label disappearance, suggesting wounded mouse cortical cells undergo protracted cell death (Whalen et al. 2008). Thus, acute plasma membrane permeability is a fundamental feature of traumatic CNS injury of unknown mechanism. Acutely after SCI, we show structural astroglial pathology including swelling and process fragmentation and astrocyte membrane wounding after in vitro trauma that associated with significant protein release into fluids, leaving an injury “trace”. Importantly, this release process reflects the population of injured and dying cells as shown in the proteomic profiles and individual marker levels. Thus, acute biofluid marker elevation captures ensuing astroglial demise with the possibility for future clinical diagnostic use.
Our data suggests an acute role of STAT3 in astroglial resilience to mechanical trauma because its deletion increased necrosis. Failure to undergo apoptosis, an ATP dependent process, is consistent with exacerbated energy deficiency in STAT3 CKO cells compared to non-transgenic astrocytes after trauma. Energy control is critical for cell survival after traumatic injury. Stretched rat cortical astrocytes had vacuolated mitochondria, decreased mitochondrial potential and reduced ATP levels that recovered by 24 h (Ahmed et al. 2000; Ellis et al. 1995). A transient ATP deficit occurs after mechanical injury in rat neurons (Geddes et al. 2003). Mouse cortical neurons show decreased mitochondrial potential after moderate stretching and a failure to recover after a secondary insult (Arundine et al. 2004).
STAT3 has a key function in oxidative metabolism regulating a switch between mitochondrial respiration and cytosolic glycolysis (Reich 2009). Phosphorylated STAT3 is present in mitochondria of cancer and astrocytes (Sarafian et al. 2010; Wegrzyn et al. 2009). STAT3-deficient astrocytes have reduced ATP production, lower mitochondrial membrane potential and glutathione levels, and are vulnerable to glucose restriction and free oxygen radicals (Sarafian et al., 2010). These findings document an essential role of STAT3 for cell survival and metabolic capacity, not addressing the molecular players. Our work extends previous proteomic data on STAT3-deficient mouse neurons reporting changes in metabolic proteins after ischemic injury and oxidative modification of mitochondrial proteins after rat controlled cortical impact (Di Domenico et al. 2012; Opii et al. 2007). Our new data shows lower expression of glucose controlling and metabolizing enzymes, including LDHB, ALDOC and PEA15 in STAT3-deficient astrocytes. In addition, stretching caused greater trauma-release of metabolic enzymes including alpha enolase, transitional ER ATPase and ESTD among others, amounting to an unfavorable energy status and cell death of traumatized STAT3-CKO astrocytes. Thus, the STAT3-goverend metabolic capacity is important for cellular recovery from trauma. Hence, mechanoporation inflicted metabolic enzyme decrease adds to changes in their substrates making traumatized astrocytes vulnerable. A transient ATP deficit and elevated free oxygen radicals further exacerbate a post-traumatic energy deficit (Arundine et al. 2004;). Impaired glucose utilization via oxidative metabolism and a shift to glycolysis occur acutely after trauma and during post-traumatic metabolic depression in rat cortical models of controlled contusion and moderate lateral fluid percussion (Bartnik et al. 2007; Hovda et al. 1991; Scafidi et al. 2009). Thus, understanding acute injury mechanisms and survival strategies of transiently compromised astrocytes in metabolic distress is essential for overcoming a state of metabolic crisis that is found even after mild TBI (Giza and Hovda 2014).
Heterogeneous responses to trauma and astrocyte demise after neurotrauma are understudied. Post-injury induction of GFAP expression occurs between 12–24 h after mild cortical contusion in the mouse (Hinkle et al. 1997). Astrocyte injury and death are acute and are followed by reactive gliosis, producing prominent, enlarged and numerous astrocytes around the lesion, masking their preceding astrocyte demise. Significant rat hippocampal astrocyte loss is reported acutely after fluid percussion injury (Zhao et al. 2003). Mouse fibrous astrocyte populations are nearly halved in white matter after contusion SCI (Lytle and Wrathall 2007), while new astrocytes are added by proliferation, a capacity also shared by cortical and hippocampal astrocytes after traumatic injuries (Lytle and Wrathall 2007; Wanner et al. 2013; Topp et al. 1989). Diffuse astrocyte loss is quantitated 6 h onwards after human TBI in postmortem cortical and hippocampal white matter biopsies (Li et al. 2012). Ultrastructure of brain biopsies show extensive astroglial swelling 3 hours to 3 days after cerebral contusion reflecting cytotoxic edema followed by cell death (Bullock et al. 1991). Thus, histopathology documents astroglial demise after neurotrauma, yet without molecular markers. Our work adds molecular properties of traumatized astrocytes including ALDOC and PEA15 depletion as well as CRYAB induction and association with filament-degraded GFAP.
SCI inflicted astrocyte swelling and process beading are pathological features of trauma and of ischemia that has been more extensively studied (Fern et al. 2014; Salter and Fern 2008). The structural and proteomic changes likely reflect a combination of mechanical trauma and a subsequent ischemic episode, both share similar pathology including metabolic depression. Astrocytes are particularly vulnerable to ischemia resulting in cytotoxic edema and energy demise (Cao et al. 2010; Hertz 2008). Degenerating astrocytic processes with swollen endfeet, abnormal mitochondria and GFAP reduction are part of astroglial necrosis after stroke (Ito et al. 2009; Lukaszevicz et al. 2002). Ischemia after middle cerebral artery occlusion induces STAT3 mainly in astrocytes with acute activation in ischemic areas followed by maintained activity over days in reactive astrocytes of neighboring cortical and hippocampal regions (Choi et al. 2003; Justicia et al. 2000). SCI, STAT3-CKO cords had exacerbated astrocytes swelling and necrosis and broader tissue damage and neuronal death than CON cords, thus, STAT3-mediated astrocyte survival is neuroprotective (Wanner et al. 2013). This adds to the previously described neuroprotection of STAT3 activation in rat neurons acutely following focal cerebral ischemia (Dziennis et al. 2007).
In addition to increased protein depletion and metabolic deficit, changes in structural proteins could contribute to the vulnerability of STAT3-deficient, traumatized astrocytes, as they expressed less GFAP (Herrmann et al. 2008; Sarafian et al. 2010). TPM4 and ACTN4 are two actin-regulating proteins that are elevated in STAT3-CKO compared to CON astrocytes and could increase cell rigidity, making the cells more vulnerable. A function of tropomyosin in cell stiffness is consistent with its reduction during shape changes of astrocytes becoming reactive (Canonne-Hergaux et al. 1994; Ferrier et al. 1994). Further studies on cytoskeletal proteins and their regulators are needed to address astrocyte structural resilience to trauma.
What influences the extent of individual proteins released after trauma? Contributing to a protein’s unique “release pool” are its connection to the cytoskeleton, protein complex assembly, cell compartmental residence, molecular weight and varying base expression levels. Differences between PEA15 and GFAP release could be caused by PEA15 association with astrocyte microtubules, thereby diminishing its releasable pool (Estelles et al. 1996). GFAP filaments were disassembled and degraded after stretching; part was released while part remained cellular. Possible causes include elevated intracellular calcium after trauma (Mills et al. 2004; Rzigalinski et al. 1998; Floyd et al. 2001), which is known to mediate GFAP filament degradation (Ciesielski-Treska et al. 1984; DeArmond et al. 1983). GFAP filament assembly is supported by S100β, 14-3-3E and CRYAB (Bianchi et al. 1994; Li et al. 2006; Nicholl and Quinlan 1994; Ellis et al. 2007). Their trauma-release and cellular loss may attenuate GFAP filament assembly. GFAP proteolytic cleavage in human white matter lesions and in degenerated frontal lobes supports its clinical relevance, yet associated pathophysiology needs to be further investigated (Martinez et al. 2008; Newcombe et al. 1986).
Astrocytes are known for shedding vesicles enriched in intracellular proteins, including carbohydrate metabolic enzymes like ALDOC (Sandoval et al. 2013), here identified as an astrocyte trauma-release protein. Proteins released after trauma and via exosomes bypass classical secretory pathways. Exosomes carry trauma-released proteins, and carry altered proteomic profiles in brains of patients with neurodegenerative disease than those of healthy brains (Harrington et al. 2009). Leak of protein-sized dyes occurs also from membrane-wounded Chinese hamster ovary cells (Mellgren et al. 2007). Vesicle-associated protein release occurs during repair of wounded membranes or, as our time-lapse findings indicate, can precede terminal cell lysis (Godell et al. 1997; Lusardi et al. 2003). Thus, fluid-derived astrocytic proteins are released from transiently plasmalemmal-compromised cells via membrane poration, exosome shedding or during necrotic cell lysis. Clinical relevance of astroglial protein release into biofluids is provided by reporting GFAP elevation in CSF and blood of TBI patients (Mondello et al. 2011; Okonkwo et al. 2013; Pelinka et al. 2004). A biofluid signature of astrocyte injury and death is supported by proteomic analyses of individual TBI patient’s CSF showing metabolic and astrocytic proteins present in our trauma-release proteome (Hanrieder et al. 2009; Sjodin et al. 2010). We began to explore astrocyte-specific proteins in biofluids of TBI patients revealing a substantial overlap with our mouse astrocyte trauma-release proteome (manuscript in preparation). This work provides a proteomic reference list that captures astroglial injury and cell death after acute trauma and offers new neurotrauma biomarker candidates with potential diagnostic and predictive value for TBI patients.
This work was supported by grants from NIH (1R21NS072606 to IBW; R01NS057624 to MVS; R01GM104610 to JAL), the Neilsen Foundation (82776 to IBW), and the Dr. Miriam and Sheldon Adelson Medical Research Foundation (to MVS).
We acknowledge statistical analysis done by Dr. Jeff Gornbein, Statistical Biomathematical Consulting Clinic, Dept. of Biomathematics, UCLA. Further acknowledgement for assistance goes to Anji Li, Nathalie Focha, Melania Apostolidou, Jumana Aljumainy and Julia Halford from the Wanner lab; Donna Crandall at the Semel Institute, Melissa Sondej and Philip Doran of the UCLA Molecular Instrumentation Center, and Yan Ao and Rose Korsak of the Neurobiology Department, Dr. Sofroniew’s lab.
The authors have no financial conflicts or interests.
Author contributions: IBW designed the research with contributions from MVS, PP and JAL. IBW, JL, EK as well as students and assistants acknowledged below performed experiments. IBW, JL, EK and WY analyzed the data with advice from PP, JAL, MVS and statistical input by Jeff Gornbein, PhD (see acknowledgment). IBW, JL, MVS and JAL wrote and edited the paper.