Expression of the ALS-causing variant hSOD1^G93A leads to an impaired integrity and altered regulation of claudin-5 expression in an in vitro blood–spinal cord barrier model

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by progressive paralysis due to the loss of primary and secondary motor neurons. Mutations in the Cu/Zn-superoxide dismutase (SOD1) gene are associated with familial ALS and to date numerous hypotheses for ALS pathology exist including impairment of the blood–spinal cord barrier. In transgenic mice carrying mutated SOD1 genes, a disrupted blood–spinal cord barrier as well as decreased levels of tight junction (TJ) proteins ZO-1, occludin, and claudin-5 were detected. Here, we examined TJ protein levels and barrier function of primary blood–spinal cord barrier endothelial cells of presymptomatic hSOD1^G93A mice and bEnd.3 cells stably expressing hSOD1^G93A. In both cellular systems, we observed reduced claudin-5 levels and a decreased transendothelial resistance (TER) as well as an increased apparent permeability. Analysis of the β-catenin/AKT/forkhead box protein O1 (FoxO1) pathway and the FoxO1-regulated activity of the claudin-5 promoter revealed a repression of the claudin-5 gene expression in hSOD1^G93A cells, which was depended on the phosphorylation status of FoxO1. These results strongly indicate that mutated SOD1 affects the expression and localization of TJ proteins leading to impaired integrity and breakdown of the blood–spinal cord barrier.

Introduction

Amyotrophic lateral sclerosis (ALS) is a severe progressive motor neuron disease that affects both lower motor neurons in the brainstem and spinal cord and the upper motor neurons in the motor cortex. Degeneration of these neurons leads to muscle atrophy, paralysis, fasciculation, and spasticity. Most patients with ALS die within 3 to 5 years after symptom onset caused by respiratory failure, but the clinical disease duration is very variable and ranges from death within months to more than 20 years after onset.¹ To date, only one drug, riluzole, is approved for ALS treatment. Approximately 90% of all patients have sporadic ALS (SALS), which is clinically indistinguishable to patients with a familial history (FALS). The first causative mutations for familial forms of ALS were found in 1993 within the Cu/Zn superoxide dismutase (SOD1) gene.² The gene encodes a 153 amino acid ubiquitously expressed metalloenzyme, which forms a functional homodimer that converts the superoxide anion (O₂^.−), a by-product of the oxidative phosphorylation in the mitochondria, to hydrogen peroxide and oxygen by the cyclical reduction and oxidation of the copper atom.³ Mutations in the SOD1 gene cause about 20% of all FALS cases and, to date, more than 150 mutations have been reported to be pathogenic. Transgenic mice expressing mutant human SOD1, e.g. SOD1^G93A, have been generated to elucidate SOD1-mediated motor neuron degeneration, but the pathologic events leading to ALS are still unknown.⁴ Numerous hypotheses for the pathogenesis of ALS have been proposed, which include glutamate toxicity, oxidative stress, mitochondrial dysfunction, autoimmune mechanisms, protein aggregation, and glial involvement.^{5, 6} Recently, it was postulated that ALS is a neurovascular disease and that the impairment of the blood–spinal cord barrier is one of the first pathologic events.⁷

The blood–spinal cord barrier separates the circulating blood from the spinal cord parenchyma and comprises endothelial cells, astrocytes, and pericytes. Tight junctions (TJs) between the endothelial cells form an intricate complex of transmembrane proteins and regulate the movements of ion, molecules, and cells by closing the paracellular space. The barrier integrity is furthermore facilitated by surrounding pericytes and astrocytes, which contribute to the maintaining of the barrier function by chemokine secretion.⁸ Since the blood–spinal cord barrier protects the spinal cord from pathogens and toxic agents, its disruption has a crucial role in the pathogenesis of several disorders.⁹ Ultrastructural changes at the blood–spinal cord barrier and its leakage to IgG and albumin through disease progression as well as endothelial cell damage have been shown in hSOD1^G93A mice.^{10, 11} In this context, Zhong et al have shown a disrupted blood–spinal cord barrier in different SOD1 mutants by the reduction of TJ proteins ZO-1, occludin, and claudin-5. Furthermore, they observed that this early breakdown occurred before inflammatory changes and motor neuron degeneration and suggested that barrier damage may be initiating the disease. Analyses of post-mortem tissues of ALS patients (both SALS and FALS) also indicated the evidence of blood–spinal cord barrier impairment including endothelial cell damage, pericyte degeneration, and reduced TJ protein expression.^{12, 13, 14, 15} However, it still has to be elucidated if the barrier damage is an initial disease factor and which molecular mechanisms trigger the barrier disruption.

In this study, the expression of TJ proteins in hSOD1^G93A endothelial cells was examined in vivo and in vitro. We observed a reduction of TJ proteins in the lumbar spinal cord in hSOD1^G93A mice and in cultures of primary mouse spinal cord endothelial cells (pMSCECs) from presymptomatic mice indicating a direct link between mutant SOD expression and the expression of TJ proteins. By stably transfecting bEnd.3 cells with hSOD1^G93A, we generated endothelial cells exhibiting comparable barrier integrity characteristics and TJ protein levels as pMSCECs from hSOD1^G93A mice. Therefore, they represent a valuable tool to examine pathways and biochemical processes that may lead to impaired barrier integrity. We examined the β-catenin (β-cat)/AKT/forkhead box protein O1 (FoxO1) pathway as well as the transcriptional activity of FoxO1 in hSOD1^G93A bEnd.3 cells and could show that binding of FoxO1 to the claudin-5 promoter is increased due to an altered β-cat/AKT/FoxO1 pathway in hSOD1^G93A cells compared with control. This subsequently results in decreased TJ protein levels at the cell surface eventually leading to an impaired integrity in hSOD1^G93A endothelial cells.

Materials and methods

Antibodies

The monoclonal antibody 9E10 recognizing the myc-epitope was generated by a mouse hybridoma cell line and used to detect all overexpressed hSOD1 constructs in stable bEnd.3 cell lines. Furthermore, the following antibodies were used: rat-anti-CD31 (BD Biosciences, Heidelberg, Germany), rabbit-anti-claudin-5 and mouse-anti-occludin (both from Zymed, Invitrogen, Darmstadt, Germany), rabbit-anti-phospho-FoxO1 (Ser256), rabbit-anti-FoxO1, rabbit-anti-phospho-β-cat (Ser675), mouse-anti-phospho-AKT (Ser473), and rabbit-anti-AKT were purchased from Cell Signaling (Frankfurt, Germany), rabbit-anti-β-actin and mouse-anti-β-tubulin (both from Sigma-Aldrich, Darmstadt, Germany), mouse-anti-β-cat (BD Biosciences), rabbit-anti-SOD1 from (Epitomics, Burlingame, CA, USA). The secondary HRP-conjugated antibodies against mouse and rabbit were purchased from Jackson Lab, Bar Harbor, ME (USA).

Cell Culture

bEnd.3 cells (ATCC, Manassas, VA, USA) were cultured in Dulbecco's Modified Eagle Medium (DMEM; Gibco) high glucose medium containing 10% fetal bovine serum and 100 U/ml penicillin/streptomycin (Gibco). For the generation of stably expressing cells, the packaging cell line GP2-293 (Clontech, Mountain View, CA, USA) was double transfected with the pLBCX contructs and pVSV-G at a ratio of 1:1 using the calcium phosphate transfection method to induce virus production. The medium was changed after 4 hours and viruses were collected for another 24 hours. After infection with recombinant viruses in the presence of 20 μg/mL polybrene (Sigma-Aldrich, Schnelldorf, Germany) for 24 hours, bEnd.3 cells were selected with 5 μg/mL blasticidin (Invitrogen, Carlsbad, CA, USA). Stable pools with comparable expression levels were analyzed in further experiments. In all experiments, cells were seeded with 5 × 10⁴ cells per cm² and experiments were performed 3 days after seeding when cells had reached postconfluency.

Cloning

The plasmids pEGFP-N1-hSOD1^WT and pEGFP-N1-hSOD1^G93A were used as a template to subclone hSOD1 sequences into the retroviral expression vector pLBCX with a 5′-HindIII and 3′-NotI restriction site containing an additional 3′-myc-epitope using the following fwd primer: 5′-CCC AAG CTT ATG GCG ACG AAG GCC GTG TGC GTG CTG AA-3′ and rev primer: 5′-ATT TGC GGC CGC TTA CAG ATC CTC TTC TGA GAT GAG TTT TTG TTC TTG GGC GAT CCC AAT TAC ACC ACA A-3′.¹⁶

Animals

hSOD1-transgenic mice were kept in the Central Facility for Animal Research of the University Medical Center (ZVTE) of the Johannes Gutenberg-University of Mainz under a 12- hour light cycle with food and drinking water ad libitum. All animal procedures were approved by the state of Rheinland-Pfalz and were performed in accordance with European guidelines and German law for the care and use of laboratory animals (amendment in effect July 2013). hSOD1^WT (B6SJL-Tg(SOD1)2Gur/J) and hSOD1^G93A (B6.Cg-Tg(SOD1-G93A)dl1Gur/J)⁴ mouse lines originate from the Jackson Laboratories (Bar Harbor, ME, USA). Primary cultures from brains and spinal cords were established before mice developed phenotypic signs of disease. The time of onset was defined when mice lost 10% of their maximal weight. hSOD1^G93A mice in our colony develop disease with about 270 days of age. Disease endstage was defined as when mice were not able to right themselves up within 10 seconds when laid on the side (about 300 days of age).

Isolation and Cultivation of Primary Mouse Endothelial Cells

Primary mouse spinal cord endothelial cells (pMSCECs) or primary mouse brain capillary endothelial cells (pMBCECs) were isolated from 7-month-old hSOD1^G93A mice and their respective littermates. In brief, spinal cords or cortices were pooled and tissue was mechanically dissociated, followed by a digest with a mixture of 0.75 mg/mL collagenase CLS2 (Worthington, Lakewood, NJ, USA) and 10 U/mL DNaseI (Sigma-Aldrich, Schnelldorf, Germany) in DMEM performed at 37°C on a shaker set at 1000 g for 1 hour. The pellet was resuspended in 20% BSA-DMEM (w/v) and centrifuged at 1,000 g for 20 minutes to remove myelin. The pellet was further digested with 1 mg/mL collagenase-dispase (Roche, Mannheim, Germany) and 10 U/mL DNaseI in DMEM at 37°C on a shaker for 1 hour. Endothelial capillaries were separated on a 33% continuous Percoll (GE Healthcare, Munich, Germany) gradient, collected, and seeded in dishes coated with 0.4 mg/mL collagen IV and 0.1 mg/mL fibronectin (all from Sigma-Aldrich, Schnelldorf, Germany). Cultures were maintained in DMEM supplemented with 20% plasma-derived bovine serum (First Link, Birmingham, UK), 100 U/mL penicillin/streptomycin, 2 mmol/L L-glutamine (all from Gibco), 4 μg/mL puromycin (Alexis, Loerrach, Germany) and 1 ng/mL basic fibroblast growth factor (R&D Systems, Wiesbaden, Germany) at 37°C and 5% CO₂.

Cell Surface Biotinylation

To examine surface levels of the TJ proteins claudin-5 and occludin, bEnd.3 cells were grown on 60-mm dishes to postconfluency. Cells were washed three times with ice-cold PBS and cell-surface proteins were biotinylated with 0.25 mg/mL sulfo-NHS-biotin (Pierce Chemicals, Rockford, IL, USA) in ice-cold PBS for 40 minutes at 4°C. After 20 minutes, the biotin solution was changed. Cells were washed four times with ice-cold PBS containing 50 mmol/L NH₄Cl to quench nonconjugated biotin and lysed in NP40 lysis buffer. Equal amounts of proteins were incubated with NeutrAvidin Agarose resin (Pierce Chemicals) at 4°C overnight. Biotinylated proteins were recovered by boiling in RotiLoad (Carl Roth, Karlsruhe, Germany) for 10 minutes and separated on 12% bis-tris gels.

Cell Extraction and Western Blot Analysis

Cells were washed with cold PBS, scraped off culture dishes and lysed in NP40 lysis buffer plus complete protease inhibitors (Roche) for 20 minutes. Extracts were centrifuged at 18,000 g for 20 minutes at 4°C in a microcentrifuge. Protein concentrations were determined by BCA protein assay (Pierce Chemicals). Equal amounts of total proteins were incubated with RotiLoad (Carl Roth) and heat-denaturated for 10 minutes. Proteins were separated on 10% or 12% bis-tris gels and transferred onto nitrocellulose membranes (Millipore, Bedford, MA, USA). Membranes were blocked for 1 hour with 5% nonfat dry milk in TBS containing 0.01% Tween-20 (Carl Roth), before incubation with appropriate primary and secondary antibodies. Proteins were detected using enhanced chemiluminescence (Millipore, Schwalbach, Germany) by using the LAS-3000mini (Fujifilm, Duesseldorf, Germany).

Immunofluorescense of Tight Junction Proteins

Cells were grown on glass coverslips (Marienfeld, Lauda-Königshofen, Germany) or chamber slides (Nunc, Langenselbold, Germany) and fixed with 4% paraformaldehyde and 0.12 mol/L sucrose in PBS for 30 minutes at room temperature (RT). Permeabilization was performed by incubation with 0.1% (v/v) Triton X-100 in PBS for another 30 minutes at RT. The cells were blocked with 5% (v/v) goat serum and 1% (w/v) BSA for 90 minutes at RT and the primary antibodies were incubated for 1 hour at 37°C. The secondary antibodies were incubated for 90 minutes at RT. Cell nuclei were stained with 5 μmol/L DRAQ5 (Biostatus Limited, Leicestershire, UK) for 10 minutes at RT and cells were embedded in Prolong Gold antifade reagent (Invitrogen, Darmstadt, Germany). Stainings were documented with a LSM710 (Zeiss, Jena, Germany) equipped with a ZEN 2011 SP2 software.

Immunohistochemistry

For immunodetection of claudin-5 and CD31 in spinal cords, 10-month-old transgenic mice expressing hSOD1^WT or hSOD1^G93A were anesthetized and spinal cords were prepared and snap-frozen in liquid nitrogen. Unfixed tissues were embedded in tissue freezing medium (Jung, Leica, Wetzlar, Germany) and cryosections were placed on slides. Sections were blocked and permeabilized (1% BSA, 0.2% powdered skim milk, 0.3% Triton-100 in TBS, pH 7.4) and stained with antibodies against claudin-5 and CD31 followed by appropriate secondary antibodies conjugated to Cy5 and Alexa488 (Jackson Laboratories, West Grove, PA, USA), respectively, and DAPI (Calbiochem, Darmstadt, Germany) to detect nuclei. Stainings were documented with a LSM710 (Zeiss) equipped with a ZEN 2011 SP2 software.

Measurement of the Transendothelial Electrical Resistance and the Permeability to [¹⁴C]-Inulin and FITC-Dextran

Transendothelial electrical resistance (TER) and permeability of high molecular paracellular marker [¹⁴C]-inulin were used to analyze the integrity of pMSCECs or stable bEnd.3 cell monolayer. Therefore, cells were seeded on 24-transwell cell culture inserts (ThinCerts, Greiner Bio-One, Frickenhausen, Germany) and placed into the cellZscope device (nanoAnalytics, Münster, Germany). Transendothelial electrical resistance and capacitance (C_Cl) of cells were measured automatically every hour under physiologic conditions by impedance spectroscopy. When capacitance values were between 1.0 and 0.8 μF/cm², indicating a confluent monolayer of cells, cells were used for the permeability assay. Culture media was changed to serum-free media containing 40 mmol/L HEPES and 1 μCi/mL [¹⁴C]-inulin (Perkin-Elmer, Waltham, MA, USA) was added to the luminal compartment of the ThinCerts. At each point in time, 10 μL samples were taken from the abluminal compartment and counted on a Tri-Carb 2800 TR Liquid Scintillation Analyzer (Perkin-Elmer). The permeability was evaluated by calculating the apparent permeability coefficient (P_app): P_app[cm/s]=dQ/(dt × A × c₀), where dQ is the amount of permeated [¹⁴C]-inulin in the incubation time, A is the surface area, c₀ is the initial concentration in the luminal compartment and dt is the incubation time. Furthermore, for the selective size opening of the endothelial cells, the permeability of 4 kDa and 40 kDa FITC-dextran was analyzed. Therefore, bEnd.3 cells were seeded on 24-transwell cell culture inserts in phenol red-free DMEM. When cells were postconfluent, 1 mg/mL of either 4 kDa or 40 kDa FITC-dextran (Sigma-Aldrich, Karlsruhe, Germany) was added to the luminal compartment of the ThinCerts. After 18 hours, 150 μL samples were taken from the abluminal compartment and measured on a Wallac 1420 multilabel counter (Perkin-Elmer). The P_app was calculated as previously described.

Chromatin Immunoprecipitation

The chromatin immunoprecipitation assay was performed using the SimpleChIP Plus Enzymatic Chromatin IP Kit from Cell Signaling (New England Biolabs GmbH, Frankfurt, Germany) according to the manufacturers' instructions. In brief, bEnd.3 cells were grown on 150-mm dishes to postconfluency and crosslinked with 1.5% formaldehyde for 20 minutes at RT. Addition of 125 mmol/L glycine inactivated unreacted formaldehyde. Cells were then washed and scraped off culture dishes and resuspended in lysis buffer for nuclei preparation. Nuclei were incubated with 1,000 gel units Micrococcal Nuclease for 10 minutes at 37°C to digest DNA to length of approximately 150 to 900 bp. Nuclear membrane was broken by sonification for 5 × 10 seconds at a power of 100% using a Sonoplus UW2070 sonificator (Bandelin Electronics, Berlin, Germany). For the immunoprecipitation, 10 μg of chromatin was incubated with rabbit-anti-FoxO1 at 4°C overnight. DNA was recovered and amplified by standard PCR techniques. Primers used in this experiment were as follows: 5′-CCG CTC GAG CTG CTG AAC TTG GGG AAG AC-3′ and 5′-CCC AAG CTT AAG GGA GTG AGG GAA GGA AA-3′. As a negative control, primers for the claudin-5 coding sequence (NM_013805) were used to exclude nonspecific precipitated DNA: 5′-GGC ACT CTT TGT TAC CTT GAC C-3′ and 5′-CAG CTC GTA CTT CTG TGA CAC C-3′. For qualitative PCR analyses, we used the following PCR conditions: 94°C for 15 minutes; (94°C for 30 seconds; 54°C (claudin-5 promoter) and 64°C (claudin-5 CDS) for 30 seconds; 72°C for 1 minute) for 35 cycles; 72°C for 7 minutes.

Statistical Analysis

All graphs and statistical analyses were performed using GraphPad Prism 4 software (GraphPad, La Jolla, CA, USA). Data were analyzed by t-test or one-way analysis of variance coupled to Tukey post-test for multiple comparisons. P<0.05 was considered as statistical significance.

Results

The Expression of Tight Junction Proteins Is Decreased in hSOD1^G93A Endothelial Cells

Previously, decreased levels of TJ proteins claudin-5, occludin, and ZO-1 were detected in freshly isolated cells of dissected microvessels of presymptomatic SOD1 transgenic mice.¹¹ Here, we extended these analyses by examining the in vivo and in vitro TJ protein levels of claudin-5 (Figure 1). In hSOD1^G93A endstage mice, we observed reduced levels of claudin-5 in the spinal cord vasculature compared with hSOD1^WT controls (Figures 1A and 1F). We furthermore found claudin-5 present in the surrounding tissue in hSOD1^G93A mice. However, in isolated and cultivated pMSCECs of presymptomatic hSOD1^G93A mice, we also observed reduced levels of the claudin-5 protein in comparison with their respective littermates (Figures 1G and 1H). These data demonstrated that the expression of the ALS-causing hSOD1^G93A variant results in the reduction of TJ proteins in endothelial cells. To further understand the underlying mechanism of reduced claudin-5 protein levels in ALS, we stably transfected the immortalized mouse brain endothelial cell line bEnd.3 with hSOD1^WT and hSOD1^G93A (Figures 2A and 2B) and stable pools with comparable transgene levels were used for further biochemical experiments. The expression of hSOD1^G93A led to decreased claudin-5 protein levels in the whole lysate and at the cell surface (Figures 2C and 2D). Compared with hSOD1^WT cells, the claudin-5 levels are reduced in the lysate by approximately 35% (Figure 2E) and by approximately 32% at the cell surface (Figure 2F). Furthermore, we investigated the protein levels of occludin, another TJ protein (Supplementary Figure S1A). Likewise, the levels of occludin were reduced in the lysate of hSOD1^G93A cells compared with hSOD1^WT cells by approximately 18% (Supplementary Figure S1B) and by approximately 21% (Supplementary Figure S1C) at the cell surface. Taken together, these data strongly indicate that the expression of hSOD1^G93A results in decreased levels of the TJ proteins claudin-5 and occludin already at a presymptomatic stage of disease. This effect persists after cultivation of primary endothelial cells and can be imitated by stably overexpressing hSOD1^G93A in an immortalized endothelial cell line.

Figure 1.

Distribution of claudin-5 in spinal cords of superoxide dismutase (SOD1) transgenic mice. (A–F) Native, unfixed cryo sections from lumbar spinal cords from mice overexpression hSOD1^G93A at endstage (10 months of age) or from age-matched hSOD1^WT mice were stained with polyclonal antibodies against claudin-5 (red) and rat monoclonal antibodies against CD31 (green). Nuclei were detected with DAPI (blue). (A'–F') An enlargement of the boxed areas of the corresponding panels is shown. The arrow in (D–F) points to a blood vessel weakly stained for claudin-5 in a mutant animal. Scale bar is 20 μm. (G, H) Primary mouse spinal cord endothelial cells (pMSCECs) of 7-month-old nonsymptomatic hSOD1^G93A transgenic mice and their respective littermates were fixed with paraformaldehyde and immunostained for the tight junction (TJ) protein claudin-5, followed by incubation with AlexaFlour 546 secondary antibody. Cell nuclei were stained with DRAQ5. Scale bar is 5 μm.

Figure 2.

Expression of hSOD1^G93A results in decreased claudin-5 levels in endothelial cells. (A, B) bEnd.3 cells were stably transfected via retroviral transfection. The expression of hSOD1 was analyzed by SDS-PAGE and western blot. The hSOD1 variants were fused to a C-terminal myc-tag. Blots were incubated with either mouse monoclonal 9E10 antibody recognizing the myc-tag (A) or rabbit monoclonal SOD1 antibody (B), which detects mouse and human SOD1. (C) bEnd.3 cells overexpressing hSOD1^WT or hSOD1^G93A were fixed with paraformaldehyde and immunostained for tight junction (TJ) protein claudin-5, followed by incubation with AlexaFlour 546 secondary antibody. Cell nuclei were stained with DRAQ5. Scale bar is 5 μm. (D) Surface proteins of postconfluent hSOD1^WT and hSOD1^G93A bEnd.3 cells were biotinylated using sulfo-NHS-biotin. Biotinylated samples were precipitated with NeutrAvidin agarose beads and analyzed by SDS-PAGE and western blot. All samples were analyzed on the same western blot but in a different order (PD: NeutrAvidin; left panel). As input controls, 20 μg of cell lysates was used (input; left panel). Polyclonal β-actin antibody served as a loading control to verify the absence of endomembrane contaminants in the biotinylation. (E, F) Claudin-5 levels in the lysate (E) or at the cell surface (F) were quantified by densitometric analysis of western blots. Intensities of hSOD1^WT cells were set as 100%. The data represent mean±s.e.m. of six independent experiments. *Statistically significant difference (P<0.05, t-test) between claudin-5 levels of stable hSOD1^WT and hSOD1^G93A cells is indicated. hSOD1, human SOD; mSOD1, mouse SOD1. SOD1, superoxide dismutase.

Expression of hSOD1^G93A Leads to an Impaired Integrity in Endothelial Cells

As we found that expression of hSOD1^G93A results in decreased levels of TJ proteins claudin-5 and occludin, we investigated whether these decreased TJ levels result in changes in barrier integrity. Therefore, hSOD1 bEnd.3 cells were seeded on cell culture inserts and TER and capacitance (C_Cl) were measured by impedance spectroscopy using a cellZscope device. Transendothelial electrical resistance of both hSOD1^G93A and hSOD1^WT cells increased over time because the expression of TJ proteins is induced when cells build a confluent monolayer (Figure 3A). However, when C_Cl values of both cells were comparable (Figure 3B), which showed that the cells had built a confluent monolayer, the TER of hSOD1^G93A bEnd.3 cells was decreased by approximately 20% compared with the hSOD1^WT cells. Furthermore, hSOD1^G93A cells exhibited an approximately 40% higher permeability to [¹⁴C]-inulin (Figure 3C) compared with hSOD1^WT cells. Furthermore, we examined the integrity of hSOD1^G93A pMSCECs and compared it with the stable bEnd.3 cell lines. Transendothelial electrical resistance of hSOD1^G93A pMSCECs from presymptomatic mice was decreased by approximately 50% compared with their respective littermates (Figure 3D), even when both pMSCECs cultures had built a confluent monolayer (Figure 3E). In addition, hSOD1^G93A pMSCECs exhibited an approximately 40% higher permeability to [¹⁴C]-inulin compared with littermate pMSCECs (Figure 3F). These data showed that the expression of hSOD1^G93A not only results in decreased levels of TJ proteins, but also furthermore leads to an impaired integrity in these cells. We furthermore examined whether increased permeability of hSOD1^G93A endothelial cells exhibits size selectivity, since claudin-5-deficient mice have a size-selective opening of the blood–brain barrier. Therefore, the permeability of 4 kDa or 40 kDa FITC-dextran was investigated in postconfluent hSOD1^WT and hSOD1^G93A bEnd.3 cells (Supplementary Figure S2). In both cell type monolayers, we detected an increased P_app to 4 kDa FITC-dextran than to 40 kDa FITC-dextran demonstrating that small molecules leak through the cell layer, whereas paracellular diffusion of big molecules is restricted. However, we observed higher P_app values in hSOD1^G93A endothelial cells, whereas the P_app of 4 kDa FITC-dextran exhibits a tendency of higher leakage and the P_app of 40 kDa FITC-dextran is significantly increased in hSOD1^G93A cells. Thus, the observed impaired integrity in hSOD1^G93A cells exhibits a size-selective opening of the endothelial barrier. As claudin-5-deficient mice have a size-selective opening of the blood–brain barrier,¹⁷ we checked whether the blood–brain barrier of hSOD1^G93A mice is affected as well. Therefore, pMBCECs were isolated in parallel with the pMSCECs and the TER and the P_app were compared between pMSCECs and pMBCECs of hSOD1^G93A mice and their respective littermates (Supplementary Figure S3). The TER and the P_app of hSOD1^G93A pMSCECs were decreased compared with their respective littermates, which confirmed our previous observations. We did not observe any differences in the TER (Supplementary Figure S3A) and P_app (Supplementary Figure S3B) of hSOD1^G93A pMBCECs compared with their respective littermates. Taken together, the expression of hSOD1^G93A results in decreased levels of TJ proteins that eventually leads to an impaired integrity in these cells, and as newly generated bEnd.3 cell lines stably expressing SOD1^G93A constructs showed the same functional response as primary endothelial cells from SOD1^G93A mice, it represents a relevant in vitro blood–spinal cord barrier model for ALS.

Figure 3.

Impaired integrity of hSOD1^G93A endothelial cells. (A–C) Stable hSOD1^WT and hSOD1^G93A bEnd.3 cells were cultivated on cell culture inserts in the cellZscope device. (A) Representative transendothelial electrical resistance (TER) measurement of stable bEnd.3 cells of at least three independent experiments. (B) Representative C_Cl measurement of stable bEnd.3 cells of at least three independent experiments. (C) Apparent permeability (P_app) of [¹⁴C]-inulin of hSOD1^G93A bEnd.3 cells compared with hSOD1^WT overexpressing cells. The data represent mean±s.e.m. with n⩾20 of 3 independent experiments. Statistical significance (*P<0.05, t-test) between hSOD1^G93A and hSOD1^WT overexpressing cells. (D–F) Primary mouse spinal cord endothelial cells (pMSCECs) were isolated from 7-month-old hSOD1^G93A mice and their nontransgenic littermates. Cells were cultivated on cell culture insert in the cellZscope device, which measured online the TER and C_Cl of the cells. (D) Representative TER and C_Cl measurement (E) of cultured pMSCECs of at least three independent experiments. (F) Apparent permeability (P_app) of [¹⁴C]-inulin of pMSCECs hSOD1^G93A compared with their littermates. The data represent mean±s.e.m. with n⩾9 of 3 independent experiments. Statistical significance (*P<0.05, t-test) between pMSCECs from hSOD1^G93A mice and their littermates is indicated. SOD1, superoxide dismutase; hSOD1, human SOD1.

The β-cat/AKT/FoxO1 Pathway and the Gene Regulation of Claudin-5 Is Altered in hSOD1^G93A Cells

As we demonstrated the suitability of hSOD1 overexpressing bEnd.3 cells as an in vitro model for ALS endothelial cells, this cell line was used to examine the molecular processes triggering barrier disruption. Recently, it was shown that the β-cat/AKT/FoxO1 pathway regulates the claudin-5 expression in endothelial cells and that the phosphorylation states of β-cat, AKT, and FoxO1 are crucial for the claudin-5 expression. The phosphorylation of FoxO1 through AKT activation (pAKT) prevents the nuclear accumulation of β-cat and FoxO1, which consequently leads to the expression of claudin-5. Thus, unphosphorylated FoxO1 is translocated to the nucleus, forming a complex with pβ-cat and T-cell factor and thereby repressing the claudin-5 expression.^{18, 19, 20} To assess whether the expression of mutant SOD1 affects this pathway, we analyzed the phosphorylation states of β-cat, AKT, and FoxO1 by western blot (Figure 4). We observed increased levels of pβ-cat and decreased levels of pAKT in mutant bEnd.3 cells compared with SOD1^WT overexpressing cells (Figures 4A and 4D). In line with these data, we observed decreased levels of pFoxO1 (Figures 4E and 4F) suggesting an increased activity of FoxO1 in the nucleus and thereby inhibiting transcription. To analyze a direct effect on claudin-5 transcription, we examined the binding of FoxO1 to the promoter region of claudin-5 by a chromatin immunoprecipitation assay. Therefore, proteins and DNA of postconfluent hSOD1^WT and hSOD1^G93A bEnd.3 cells were crosslinked and a FoxO1 antibody was used to precipitate protein bound to the claudin-5 promoter. The amounts of precipitated DNA in hSOD1^WT and hSOD1^G93A cells were analyzed using a standard PCR techniques followed by a qualitative agarose gel (Figure 5A). We observed higher levels of precipitated claudin-5 promoter in hSOD1^G93A cells demonstrating an increased interaction between FoxO1 and the claudin-5 promoter, which results in a repression of the claudin-5 expression in these cells (Figure 5B). To confirm the repression of the claudin-5 promoter, we analyzed the claudin-5 mRNA levels as well as the production of claudin-5 protein in hSOD1^G93A cells (Supplementary Figure S4). By quantitative real-time PCR analysis, we detected higher ΔC_T values for claudin-5 in hSOD1^G93A cells compared with hSOD1^WT cells representing a 93-fold amount of claudin-5 mRNA in hSOD1^WT cells than in hSOD1^G93A cells (Supplementary Figure S4A). Subsequently, these low amounts of claudin-5 mRNA result in a lower production of the claudin-5 protein, which we observed via metabolic labeling (Supplementary Figure S4B). Taken together, these data strongly indicate that the expression of hSOD1^G93A alters the β-cat/AKT/FoxO1 pathway resulting in a repression of the claudin-5 promoter which subsequently leads to reduced TJ levels and the impaired integrity observed in transgenic endothelial cells.

Figure 4.

Decreased levels of p[S675]β-cat, p[S473]AKT, and p[S256]FoxO1 in stable hSOD1^G93A bEnd.3 cells. The phosphorylation states of p[S675]β-cat (A, B), p[S473]AKT (C, D), and p[S256]FoxO1 (E, F) were analyzed in postconfluent hSOD1^WT and hSOD1^G93A bEnd.3 cells by SDS-PAGE and western blot (A, C, E). Levels of phosphorylated and total proteins were quantified by densitometric analysis of western blots (B, D, F). Intensities were normalized to β-tubulin and the ratio of phosphorylated to total protein was calculated. Ratios of hSOD1^WT cells were set as 100%. The data represent mean±s.e.m. of at least five independent experiments. *Statistically significant difference (P<0.05, t-test) between ratios of stable hSOD1^WT and hSOD1^G93A cells are indicated. SOD1, superoxide dismutase; hSOD1, human SOD1.

Figure 5.

Binding of FoxO1 to the promoter region of claudin-5 in stable hSOD1^G93A bEnd.3 cells. The binding of FoxO1 to the promoter region of claudin-5 was analyzed by chromatin immunoprecipitation in postconfluent hSOD1^WT and hSOD1^G93A bEnd.3 cells. (A) Crosslinked chromatin was precipitated with FoxO1 antibody. As a negative control, primers for the claudin-5 CDS were used to exclude nonspecific precipitated DNA. Levels of precipitated DNA were quantified by densitometric analysis of agarose gels (B). Intensities were normalized to input and the ratio of hSOD1^WT cells was set as 100%. The data represent mean±s.e.m. of three independent experiments. *Statistically significant difference (P<0.05, t-test) between ratios of stable hSOD1^WT and hSOD1^G93A cells is indicated. SOD1, superoxide dismutase; hSOD1, human SOD1.

Discussion

Amyotrophic lateral sclerosis is a neurodegenerative disease hallmarked by progressive motor neuron degeneration in the brain and the spinal cord leading to muscle atrophy, paralysis, and death typically within 3 to 5 years from diagnosis.¹ Mutations in the gene of SOD1 are associated with approximately 20% of familial ALS cases and since the clinical and pathologic profiles of SALS and FALS are similar, it is believed that insights from studies of ALS-causing gene mutations can be applied to SALS cases.^{2, 21} The molecular mechanisms resulting in ALS are still unknown, but there is convincing evidence that the disruption of the blood–spinal cord barrier is one of the primary pathologic events.⁷ In both FALS and SALS patients, endothelial cell damage, pericyte degeneration, and loss of endothelial integrity were observed, and studies with SOD1 transgenic mice showed that the blood–spinal cord barrier disruption was found before motor neuron degeneration and neurovascular inflammation.^{7, 8, 10, 11, 14, 15} This disruption was accompanied by a reduction of the TJ proteins claudin-5, occludin, and ZO-1 in freshly isolated and dissected cells of the blood–spinal cord barrier. In our study, we extended these analyses by examining the in vivo and in vitro levels of the TJ protein claudin-5 in presymptomatic and endstage hSOD1^G93A mice and observed a reduction of TJ protein levels during the disease. Despite this, we observed that the claudin-5 staining of vessels was strongly reduced in hSOD1^G93A endstage mice and that the protein was distributed in the surrounding tissue where it partially colocalized with CD31. This observation will be of interest for further investigations. Furthermore, we observed that a stable integration of hSOD1^G93A in the immortalized endothelial cell line bEnd.3 leads to a significant reduction of the TJ proteins claudin-5 and occludin at the cell surface. Zhong et al¹¹ already showed that the levels of TJ proteins are downregulated in isolated capillaries of presymptomatic mutant hSOD1 transgenic animals. However, the authors examined the TJ levels in whole lysates of primary tissue and not the decisive and functional surface expression. Thus, we propose that mutant SOD1^G93A expression have a direct effect on the expression level of TJ proteins in ALS.

Previous studies have shown that the integrity characteristics of endothelial cells are dependent on the localization and the protein levels of TJ proteins. Using bEnd.3 cells, it was shown that the localization of claudin-5 at the cell surface directly influenced the development of the TER in a cellular monolayer. Furthermore, the integrity of an endothelial monolayer could be disturbed by silencing the translation of claudin-5.²² In addition, reduced levels of occludin resulted in decreased TER values and higher permeability of fluorescently labeled glucane.^{23, 24} An even stronger phenotype was detected when both claudin-5 and occludin protein levels were decreased in matrix metalloproteinase-9 expressing cells. In matrix metalloproteinase-9 expressing bEnd.3 cells, claudin-5 and occludin are disrupted by matrix metalloproteinase-9 that resulted in an even stronger increase in barrier permeability.²⁵ On the basis of these observations, we examined the integrity of the hSOD1^WT and hSOD1^G93A expressing bEnd.3 cells and pMSCECs of presymptomatic hSOD1^G93A mice. For both bEnd.3 cells and pMSCECs, we detected decreased TER values, an increased P_app as well as a size-selective opening of the barrier in hSOD1^G93A expressing cells representing an impaired integrity compared with the control cells. Thus, expression of hSOD1^G93A in endothelial cells leads to reduced cell surface levels of TJ proteins and subsequent impaired integrity of the endothelial barrier. These results are in line with previous studies demonstrating the selective gene excision of the ALS-causing mutant hSOD1^G37R within the endothelia attenuated the blood–spinal cord barrier disruption by 70% compared with age-matched hSOD1^G37R mice.²⁶ Double transgenic animals still had about 50% greater IgG leakages compared with wild-type mice most likely because hSOD1^G37R expression in cells surrounding the endothelium contribute to the pathology in this multifactorial disease. Nevertheless, as in our study primary endothelial cells were selected in the presence of puromycin, our results strongly indicate that the expression of hSOD1^G93A in endothelial cells results in the impaired integrity.

We furthermore checked whether the blood–brain barrier is disrupted in hSOD1^G93A mice as well. Therefore, the TER and the P_app of pMBCECs and pMSCECs from the same animals were compared. We did not observe any differences in the TER and P_app in pMBCECs from hSOD1^G93A, indicating that the blood–brain barrier is not affected at a presymptomatic stage. Previous studies have shown that the endothelial cells of the blood–brain barrier in the brain stem are degenerated and damaged in hSOD1^G93A mice at an early stage of the disease.¹⁰ However, we isolated endothelial cells from the cortex from presymptomatic animals and could not detect any changes in the integrity of pMBCECs compared with their respective littermates. Thus, we suggest that at a presymptomatic stage, the blood–brain barrier is not disrupted that changes during the disease progression. Taken together, our data clearly provide strong evidence for a direct role of mutant SOD in the expression pattern of TJ proteins, since stably transfected bEnd.3 cells as well as pMSCECs expressing mutant hSOD1^G93A showed decreased TJ protein levels and impaired integrity of the blood–spinal cord barrier.

As we observed that the protein levels of claudin-5 are generally reduced in hSOD1^G93A cells, we were interested whether the transcriptional regulation of the claudin-5 expression was altered in these cells. The claudin-5 expression is regulated by several transcription factors that can act either as activators (e.g., SOX18) or as repressors (e.g., FoxO1, β-cat).^{18, 27, 28} We concentrated on the β-cat/AKT/FoxO1 pathway, which was described by Taddei et al (Figure 6A).¹⁸ In their study, Taddei et al postulated that the phosphorylation of FoxO1 through pAKT prevents the nuclear accumulation of pβ-cat and FoxO1, which consequently leads to the expression of claudin-5. The phosphorylation of β-cat at Ser675 was previously shown to induce its nuclear translocation. In the nucleus, pβ-cat forms a complex with FoxO1 and T-cell factor and the subsequent interaction of this complex with the claudin-5 promoter was shown to result in reduced levels of claudin-5.^{20, 29, 30} In our study, we observed increased levels of p[S675]β-cat in the hSOD1^G93A cells compared with the wild-type cells indicating a mechanistic explanation for the reduced TJ protein levels. Furthermore, it was recently shown that the nuclear levels of β-cat are increased in whole lysates of ALS spinal cords of hSOD1^G93A mice,³¹ pointing again to a direct involvement of mutant SOD expression in impaired barrier integrity due to an altered pathway responsible for TJ protein expression.

Figure 6.

Suggested model for the regulation of the claudin-5 expression in hSOD1^G93A endothelial cells. (A) In hSOD1^WT endothelial cells, the adhesion of VE-cadherin leads downstream to an activation of the PI3K and AKT pathway. Phosphorylation of FoxO1 through AKT activation (pAKT) prevents the nuclear accumulation of β-cat and FoxO1 and consequently leads to the expression of claudin-5. (B) Binding of mutant hSOD1^G93A to Rac1 may result in a local overproduction of reactive oxygen species (ROS).⁴² Increased oxidative stress results in increased pβ-cat levels and a reduced phosphorylation of FoxO1 through the interaction of AKT. This may cause an accumulation of β-cat and FoxO1 in the nucleus leading to a complex formation of β-cat and FoxO1 with the transcription factor TCF, and this complex may bind to the claudin-5 promoter and may inhibit its transcription. This subsequently may result in the loss of claudin-5 in tight junctions (TJs), in the disruption of the endothelial barrier function and in enhanced permeability (modified from Gavard and Gutkind¹⁹ and Taddei et al¹⁸). TCF, T-cell factor. SOD1, superoxide dismutase; hSOD1, human SOD1.

To further analyze the molecular regulation of mutant hSOD1-induced TJ protein expression, we focused the AKT kinases that have been shown to influence the β-cat transcriptional pathway. Since pAKT indirectly influences the β-cat transcriptional activity, we also examined the protein level of p[S473]AKT and detected decreased phosphorylation of AKT in hSOD1^G93A cells compared with controls. The role of AKT in ALS has been studied in various systems, but so far the results are contradictory.³² For instance, decreased levels of phosphorylated AKT were observed in motor neurons of both FALS and SALS patients and in hSOD1^G93A mice before symptom onset,³³ as well as in atrophic muscles of symptomatic hSOD1^G93A mice and ALS patients,^{34, 35} whereas no changes on phosphorylated AKT were examined in an alternate study of hSOD1^G93A mouse spinal cords.³⁶ Since AKT is a multifactorial protein that has a crucial role in several cellular processes including survival, apoptosis, cell-cycle progression, and cell growth and furthermore regulates glycogen metabolism and the expression of TJ proteins, it is complicated to generalize its function in different tissues and cell types.³⁷ Here, we examined the phosphorylation state of AKT in endothelial cells, and, according to previous studies in endothelial cells, an inactivation of AKT has been shown to lead to a decreased TJ protein expression.¹⁸ A downstream target of the pAKT is FoxO1, which was shown to directly regulate the expression of claudin-5.³⁸ Unphosphorylated FoxO1 is translocated to the nucleus to repress the expression of its target genes, and the phosphorylation of FoxO1 prevents its transfer to the nucleus. Analyzing endothelial cells expressing mutant SOD revealed decreased levels of pFoxO1 in hSOD1^G93A cells compared with the hSOD1^WT cells. So far, the role of FoxO1 in ALS was studied with a focus on muscle atrophy. In muscle biopsies, no difference in the nuclear protein contents of FoxO1 between ALS patients and control patients was detected.³⁵ However, in a cell culture model for muscle atrophy, reduction of pFoxO1 could be observed demonstrating that observations depend on the chosen cellular model.³⁹ In the hSOD1^G93A endothelial cells, we detected reduced levels of pFoxO1 indicating the cytosolic accumulation of FoxO1 explaining its decreased transcriptional activity. To further underline this hypothesis, we performed a chromatin immunoprecipitation assay to analyze a direct effect of FoxO1 on the claudin-5 transcription. We could clearly show an increased interaction of FoxO1 with the claudin-5 promoter in hSOD1^G93A cells resulting subsequently in a repression of the claudin-5 promoter. Furthermore, we confirmed that the repression of the claudin-5 promoter eventually results in decreased amounts of mRNA as well in a decreased production of claudin-5. Taken together, our data show a direct role of mutant SOD on claudin-5 expression that leads to decreased levels of the TJ protein and subsequently impaired integrity of hSOD1^G93A endothelial cells.

Previous studies have shown that an increase in the vascular permeability can be caused by the local formation of reactive oxygen species (ROS) through a Rac-dependent mechanism, which is associated with the functional interaction of FoxO1 and β-cat and the loss of endothelial cell–cell junctions.^{40, 41} A direct link between the production of ROS and ALS-causing mutants of SOD1 was recently postulated by Harraz et al.⁴² The authors proposed a redox sensor model by which SOD1 can regulate NADPH oxidase-dependent (Nox-dependent) production of O₂^.− through its ROS-sensitive control of Rac1-GTP hydrolysis. Under reducing conditions, SOD1 binds to Rac1 and inhibits its GTPase activity, which consequently leads to an increased production of O₂^.−. When the local concentrations of H₂O₂ rise (spontaneous or by SOD1), SOD1 dissociates from Rac1, Rac1 then is inactivated by GTP hydrolysis, and this leads to an inactivation of the Nox complex and a reduction in ROS production. The authors observed that this redox-sensitive uncoupling of SOD1 was defective in hSOD1^G93A cells, leading in an overproduction of ROS.^{42, 43} We hypothesize that this altered production of ROS may subsequently alter the expression of the TJ proteins via the β-cat/AKT/FoxO1 pathway, which lead to the impaired integrity in SOD1 mutants causing ALS (Figure 6B). Since previous data as well as ours provide strong evidence that the endothelial damage and accordingly the breakdown of the blood–spinal cord barrier is one of the primary events in ALS, the stabilization of the blood–spinal cord barrier represents a promising approach as a viable therapeutic target. In hSOD1^G93A mice, the peripherally administration of activated protein C (APC) or APC analogs with reduced anticoagulant activity slowed the disease progression and extended the survival after disease onset.²⁶ Moreover, the authors observed an APC-mediated downregulation of mutant SOD1 in endothelial cells, astrocytes, and microglia resulting in the delay of the inflammatory response as well as in the maintenance of the barrier integrity. Treatment of hSOD1^G93A mice at an early disease stage furthermore restored the blood–spinal cord barrier integrity, which delayed the onset of motor-neuron impairment and degeneration and clearly demonstrated the therapeutic potential of APC for ALS.⁴⁴ Since this protective potential of APC is mediated by a Rac1-dependent signaling in association with an early reduction of ROS, and considering that Rac1 signaling and the production of ROS is altered in hSOD1^G93A cells, further studies are needed to elucidate the role of Rac1 signaling in endothelial cells of the blood–spinal cord barrier as well as the molecular mechanism leading to the barrier breakdown in ALS. Taken together, previous data from independent groups as well as ours provide strong evidence that the endothelial damage and accordingly the breakdown of the blood–spinal cord barrier is one of the primary events in ALS and that the regeneration of the barrier integrity may be a future promising target.

The authors declare no conflict of interest.