|Home | About | Journals | Submit | Contact Us | Français|
Brain injury may result in the development of epilepsy, one of the most common neurological disorders. We previously demonstrated that albumin is critical in the generation of epilepsy following blood-brain barrier (BBB) compromise. Here we identify TGF-β pathway activation as the underlying mechanism. We demonstrate that direct activation of the TGF-β pathway by TGF-β1 results in epileptiform activity similar to that following exposure to albumin. Co-immunoprecipitation revealed binding of albumin to TGF-β receptor II and Smad2 phosphorylation confirmed downstream activation of this pathway. Transcriptome profiling demonstrated similar expression patterns following BBB breakdown, albumin and TGF-β1 exposure, including modulation of genes associated with the TGF-β pathway, early astrocytic activation, inflammation, and reduced inhibitory transmission. Importantly, TGF-β pathway blockers suppressed most albumin-induced transcriptional changes and prevented the generation of epileptiform activity. Our present data identifies the TGF-β pathway as a novel putative epileptogenic signaling cascade and therapeutic target for the prevention of injury-induced epilepsy.
Epilepsy, affecting 0.5–2% of the population worldwide, is one of the most common neurological disorders. Focal neocortical epilepsy often develops following traumatic, ischemic or infectious brain injury. While the characteristic electrical activity in the epileptic cortex has been extensively studied, the mechanisms underlying the latent period preceding the occurrence of spontaneous epileptic seizures (epileptogenesis) are poorly understood. Following injury, local compromise of blood-brain barrier (BBB) integrity is common (Tomkins et al., 2001; Neuwelt, 2004; Abbott et al., 2006; Oby and Janigro, 2006), as revealed by ultrastructural studies of animal and human epileptic tissue in multiple forms of epilepsy (Kasantikul et al., 1983; Cornford and Oldendorf, 1986; Cornford, 1999; Marchi et al., 2007; van Vliet et al., 2007), raising the possibility that vascular damage, and specifically BBB opening, may serve as a trigger event leading to epilepsy. This hypothesis has been confirmed by animal studies, in which opening of the BBB was sufficient to induce delayed epileptiform activity (Seiffert et al., 2004). Subsequent studies have shown that albumin, the most common serum protein, is sufficient to recapitulate the epileptiform activity induced by BBB disruption.
Furthermore, uptake of serum components such as albumin and IgGs, associated with BBB disruption, has been demonstrated in various cell populations. Albumin is taken up by astrocytes (Ivens et al., 2007; van Vliet et al., 2007), neurons (Marchi et al., 2007; van Vliet et al., 2007), and microglia although to a lesser extent (van Vliet et al., 2007), while IgG uptake has been found in neurons (Rigau et al., 2007). In rat lung endothelial cells, albumin endocytosis is mediated by transforming growth factor beta receptors (TGF-βRs), leading to phosphorylation of the proximate effector of the canonical TGF-β signaling pathway, Smad2, and translocation of the activated Smad2/Smad4 complex to the nucleus (Siddiqui et al., 2004). TGF-βRs are also implicated in albumin uptake by astrocytes, as blocking TGF-βRs prevents albumin uptake and suppresses albumin-induced epileptiform activity (Ivens et al., 2007). This raises the possibility that albumin activation of the TGF-β signaling pathway serves as the underlying mechanism, however, this hypothesis remains unconfirmed.
Here we show that activation of the TGF-β signaling pathway is sufficient to induce epileptiform activity. Furthermore, we show that global transcriptional cascades induced by TGF-β1 or albumin exposure before the development of epileptiform activity (during the epileptogenesis window) are nearly identical and can be blocked by application of TGF-βR blockers. Given the pleiotropic effects of the TGF-β signaling pathway, these findings provide a plausible mechanism for epileptogenesis following brain injury, and advocate a specific therapeutic target.
All experimental procedures were approved by the animal care and use ethical committees at Charité University Medicine, Berlin and Ben-Gurion University of the Negev, Beer-Sheva. The in vivo experiments were performed as previously described (Seiffert et al., 2004; Ivens et al., 2007). In short, adult male Wistar rats (120–250 gr) were anesthetized and placed in a stereotactic cage, a 4 mm diameter bone window was drilled over the somatosensory cortex, and the dura was opened. The underlying cortex was then perfused with artificial cerebrospinal fluid (aCSF, composition in mM: 129 NaCl, 21 NaHCO3, 1.25 NaH2PO4, 1.8 MgSO4, 1.6 CaCl2, 3 KCl, 10 glucose) supplemented with either deoxycholic acid (DOC, 2 mM, Sigma-Aldrich, Steinheim, Germany), bovine serum albumin (BSA, 0.1 mM, Merck, Darmstadt, Germany) corresponding to 25% of serum albumin concentration, or with TGF-β1 (10 ng/ml, Peprotech, Rocky Hill, NJ) for 30 min. Sham-operated animals (perfused with aCSF) served as controls. Only rats with no apparent injury to the cortical surface or bleeding from cortical vessels (as seen under the surgical microscope) at the end of the procedure were used.
In order to investigate transcriptional changes occurring during the epileptogenesis time window (before the development of epileptiform activity), animals were sacrificed 7/8, 24, or 48 hours following treatment. RNA isolated from these animals was used for microarray and qRT-PCR analyses described below. As robust changes in gene expression were observed after the 24 hour treatments, a second set of animals including sham-operated controls and animals treated with BSA (0.1 mM) or BSA plus TGF-βR blockers (TGF-βRII antibody, 50 μg/ml, Santa Cruz Biotechnology, Santa Cruz, CA; SB431542, 100 μM, TGF-βRI kinase activity inhibitor, Tocris, Bristol, UK) were sacrificed 24 hours following treatment. RNA isolated from these animals was also used for microarray and qRT-PCR analyses. The last set of animals was treated with 0.1–0.2 mM BSA and sacrificed 24 or 46–50 hours following treatment for Smad2-P immunodetection.
Brain slices for the in vitro experiments were prepared using standard techniques (Pavlovsky et al., 2003; Seiffert et al., 2004; Ivens et al., 2007). Slices were transferred to a recording chamber where they were incubated in aCSF containing BSA (0.1 mM), TGF-β1 (10 ng/ml) or artificial serum (aSERUM, composition based on aCSF with the following changes, composition in mM: 0.8 MgSO4, 1.3 CaCl2, 5.7 KCl, 1 L-glutamine, 0.1 BSA). To block the activity of TGF-β1, slices were incubated in aCSF containing SB431542 (10 μM) before the addition of TGF-β1 (10ng/ml). To block TGF-βRs, slices were incubated in aCSF containing SB431542 (10 μM) and TGF-βRII antibody (10 μg/ml) for 30 min followed by incubation in BSA in the presence of TGF-βR blockers. For detection of epileptiform activity, field potentials were recorded 4–12 h following incubation in cortical layer IV using extracellular glass microelectrodes (~3MΩ) in response to bipolar stimulation at the border of white and grey matter. The time of recording was chosen based on the occurrence of epileptiform activity 4–8 hours following albumin exposure in the slice preparation (Ivens et al., 2007).
To prepare cortical lysates, brains were isolated from naïve adult Wistar rats, dissected in cold saline solution and lysed in RIPA buffer. BSA (3 μg) was added to lysates to approximately match the amount of precipitating anti-albumin antibodies. Immunoprecipitation was performed using the Catch and Release® v2.0 Reversible Immunoprecipitation System (Upstate, Charlottesville, Virginia) with the following modifications to the standard protocol: the starting amount of protein was increased to 1,500 μg and the incubation time with precipitating antibodies was increased to 90 minutes. Lysate samples (positive or negative for albumin) were immunoprecipitated with an anti-TGF-βRII antibody (Upstate) or an anti-albumin antibody (Biogenesis, Poole, UK).
The immunoprecipitated samples were separated with SDS-PAGE and transferred onto a nitrocellulose membrane. The membrane was stained with Ponceau S stain to confirm that the IP procedure was successful. It was then destained, and blocked with 5% BSA in standard TBS-T buffer overnight at 4°C. TGF-βRII was detected with a rabbit anti-TGF-βRII antibody (Upstate) and an AP-conjugated donkey anti-rabbit IgG secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA). Chemiluminescent detection was done using Lumi-Phos WB Chemiluminescent Substrate (Pierce, Rockfort, IL) and standard X-ray film according to the manufacturer’s instructions.
Cortical lysate samples from sham-operated controls and animals treated with BSA were separated by SDS-PAGE and transferred onto a nitrocellulose membrane. The membrane was blocked with 5% nonfat milk overnight at 4°C, incubated with a rabbit polyclonal antibody against phospho-Smad2 (Millipore Corporation, Bedford, MA) for 48 hr at 4°C, and incubated with a peroxidase-conjugated goat anti-rabbit IgG secondary antibody (Jackson ImmunoResearch Laboratories) for 2 hr at room temperature.
Total RNA was isolated using the TRIzol reagent (Invitrogen, Carlsbad, CA) and prepared using the Affymetrix GeneChip one-cycle target labeling kit (Affymetrix, Santa Clara, CA). Biotinylated cRNA was fragmented and hybridized to the GeneChip Rat Genome 230 2.0 Array according to company protocols (Affymetrix Technical Manual). Normalization of the array data was done using GCRMA (GC Robust Multi-Array Average) analysis, which takes into account the GC content of the probe sequences (Wu et al., 2004). Functional annotation analysis was performed with the program Database for Annotation, Visualization, and Integrated Discovery (DAVID) 2008 (Dennis et al., 2003) (http://david.abcc.ncifcrf.gov). Unspecific (e.g., cellular process) and redundant terms (e.g., death and cell death) were removed and the full lists are provided in Supplementary Tables 1–8. The GenMAPP 2.0 program (Salomonis et al., 2007) (http://www.genmapp.org/) was used to visualize genes involved in TGF-β signaling. For the time course analysis one array was run for each treatment (DOC, BSA, TGF-β1) for the following time points: 7/8, 24, and 48 hr. In addition a sample from a sham treated animal (24hr) was run and used to normalize the other arrays. Pairwise Pearson correlation coefficients for the three treatments were determined with Excel (Microsoft Corp., Richmod, WA). Hierarchical clustering was performed with Gene Cluster and displayed with TreeView software (Eisen et al., 1998).
Arrays were then run for the second set of animals sacrificed 24 hr following treatment (Sham, n=2; BSA, n=3; BSA + TGF-βR blockers, n=4). Significance analysis of microarrays (SAM) was performed with a false discovery rate (FDR) threshold of 9.2%. A 1.5 fold change cutoff was also used to filter this list. Genes from this filtered list which demonstrated a log2 ratio difference > 0.5 between the albumin and albumin plus blocker treatments were considered part of the attenuated response. The remaining genes were considered part of the unattenuated response. All microarray data are available at the GEO website (http://www.ncbi.nlm.nih.gov/geo) under accession number GSE12304.
mRNA expression levels were determined by quantitative reverse transcriptase-PCR by real-time kinetic analysis with an iQ5 detection system (Bio-Rad, Hercules, CA). Real - time PCR data were analyzed using the PCR Miner program (Leonoudakis et al., 2008). 18S mRNA levels were used as internal controls for variations in sample preparation. Primer sequences are provided in Supplementary Methods.
For the electrophysiological data, differences between treated and control slices were determined by the Mann-Whitney U test for two independent samples or the chi-square test using SPSS 13.0 (SPSS Inc., Chicago, IL). Linear regression analysis for the microarray data was performed with Graphpad Prism (GraphPad Software, Inc., San Diego, CA). PCR data were analyzed with an unpaired Student’s t test (p < 0.05 was considered significant) in Excel (Microsoft) or with the relative expression software tool (REST) (Pfaffl, 2001). REST determines significance of the group ratio results with a randomization test. p < 0.05 was taken as the level of statistical significance.
To assess the hypothesis that activation of the TGF-β signaling pathway is the mechanism underlying albumin-induced epileptogenesis, we activated this pathway directly by incubating neocortical slices with TGF-β1 (10 ng/mL) in artificial cerebrospinal fluid (aCSF) and performed electrophysiological recordings. These recordings were compared to those of slices treated with a solution containing serum levels of electrolytes and 0.1mM albumin (aSERUM, previously shown to induce epileptogenesis (Ivens et al., 2007), albumin in aCSF, or aCSF (control). Spontaneous, prolonged and hypersynchronous interictal-like activity was observed in slices treated with aSERUM after 6–10 hours (n = 6 out of 9 slices, 3 animals) but never in aCSF treated slices (Fig. 1B). When albumin was added to the control aCSF solution, epileptiform activity was recorded only in response to stimulation of the white matter (Fig. 1C; n = 8 out of 12 slices, 6 animals). Importantly, epileptiform activity was also recorded following incubation in TGF-β1, which is similar to that seen following treatment with aSERUM and albumin in aCSF (Fig. 1C; n = 5 out of 5 slices, 4 animals; n = 7 out of 9 slices, 3 animals; and n = 8 out of 12 slices, 6 animals, respectively). Recordings were performed 4–12 hr following treatment. While treatment with either TGF-β1 or albumin in aCSF resulted in the appearance of evoked epileptiform activity, only the altered electrolytic solution (i.e. aSERUM) resulted in spontaneous activity in the slice preparation. Importantly, when aSERUM was applied without albumin neither spontaneous or evoked epileptiform activity was recorded. In all three cases (albumin, TGF-β1, and aSERUM), the evoked epileptiform activity was all-or-none in nature, paroxysmal, prolonged and propagating along the cortical slice, similar to that seen following BBB opening with bile salts (Seiffert et al., 2004; Ivens et al., 2007) and typical to that observed in acute models of epilepsy (Gutnick et al., 1982). No epileptiform activity was seen in the control aCSF treated slices.
To further confirm that the TGF-β1 induced epileptiform activity was dependent on the TGF-βR mediated pathway, we performed additional trials of the above experiments in the presence of two TGF-βR blockers (SB431542 and anti-TGF-βRII antibody). TGF-βR blockers prevented epileptiform activity induced by TGF-β1 or albumin (Fig. 1C). The measured integral of the field potential (albumin: 117.2 ± 35.4 mV*ms; TGF-β1: 84.1 ± 20.1 mV*ms) was significantly lower in slices treated with albumin or TGF-β1 in the presence of TGF-βR blockers (albumin and blockers: 23.7 ± 6.9 mV*ms, n = 20 slices, 4 animals, p = 0.001; TGF-β1 and blockers: 12.5 ± 3.9 mV*ms, n =8 slices, 4 animals, p =0.005) (Fig. 1D).
To determine whether albumin binds to TGF-β receptors, co-immunoprecipitation using antibodies against albumin or TGF-βRII was performed on cortical lysate samples (obtained from naïve rats). An expected band corresponding to TGF-βRII was detected in samples immunoprecipitated with the TGF-βRII antibody. More importantly, this band was also detected in samples pre-incubated with albumin when immunoprecipitated with the albumin antibody and probed for TGF-βRII (Fig. 2A). These results reveal a direct interaction between albumin and TGF-βRII. In the canonical TGF-β signaling pathway, Smad2 and/or 3 are phosphorylated following TGF-β receptor activation and form a complex with Smad4, which then translocates into the nucleus and activates transcription (Beattie et al., 2002). To investigate whether albumin activates downstream components of the TGF-β pathway, Smad2 phosphorylation levels in cortical lysates were assessed by Western blot, revealing an increase in Smad2 phosphorylation during the epileptogenic time window in animals exposed to albumin as compared to sham-operated controls (Fig. 2B).
BBB opening or exposure to albumin in vivo (Ivens et al., 2007), as well as in vitro exposure of neocortical slices to albumin or TGF-β1 (Fig. 1C) all result in the gradual development of hypersynchronous neuronal epileptiform activity. The delayed appearance of abnormal activity (5–7 hours in vitro and >4 days in vivo (Ivens et al., 2007)) suggests a transcription-mediated mechanism. In search of a common pathway and transcriptional activation pattern that underlie epileptogenesis following BBB opening, we performed transcriptome analysis using Affymetrix rat microarrays. RNA was extracted from cortical regions of rats treated with sodium deoxycholate (DOC, to induce BBB opening), albumin or TGF-β1 for various durations (7/8, 24, 48hr). These time points were chosen in order to evaluate changes in transcription occurring before the appearance of epileptiform activity. Control RNA was extracted from cortical regions excised from sham-operated animals. Hierarchical clustering analysis of these arrays showed that overall the three treatments resulted in strikingly similar gene expression profiles, as arrays representing similar time points clustered together regardless of the treatment (Fig. 3A). These similarities are exemplified in figure 3B, which shows a high correlation between the expression profiles for the albumin and TGF-β1 treatments at 24 hours (r2 = 0.75, p < 0.0001).
To identify biological themes common to the three treatments, the gene list was filtered to include genes showing at least a 1.5 fold change in expression and a Pearson correlation coefficient ≥ 0.95 for pair-wise comparisons between all treatments (see methods). Hierarchical clustering was performed and the main clusters were used for gene ontology (GO) analysis with DAVID (Database for Annotation, Visualization, and Integrated Discovery)(Dennis et al., 2003). DAVID calculates the probability that particular GO annotations are overrepresented in a given gene list using a Fisher exact probability test. Molecular function and biological process GO terms with a p-value < 0.05 containing at least three genes were considered significant. This analysis revealed major gene expression trends that occur in response to all three epileptogenic treatments (Fig 3C). Early responses include genes involved in general stress-related cellular, metabolic and intracellular signaling pathways; early responses persisting to later time points include inflammatory processes as well as genes involved in induction of cell cycle, differentiation, proliferation, and apoptosis; responses at middle to late time points include repression of synaptic transmission and ion transport genes (Fig 3C; for complete GO term annotation results see Supplementary Tables 1–4).
Selected GO term groups were chosen for further analysis of individual gene expression profiles (Fig. 4). The most dramatic change observed in all treatments across all time points was the early and persistent upregulation of genes associated with immune response activation (Fig. 4A–B). Inflammatory genes included NF-kappa B pathway related genes, cytokines and chemokines (Il6, Ccl2, Ccl7), transcription factor Stat3, the pattern recognition receptor CD14 and extracellular matrix proteins (Fn1 and Spp1) (Fig. 4A). Activation of the complement pathway was also prominent (Fig. 4B) and included C1 subcomponents (C1qa, C1qb, C1qg), the associated protease C1s, Masp1, and C2. A significant neuronal response was prominent in the middle-late time points and included downregulation of genes associated with GABAergic (inhibitory) neurotransmission (including the GABA A receptor subunits, Gabra4, Gabrd, Gabrg1, and Gabrb2 as well as glutamic acid decarboxylase (Gad67) (Fig. 4C) and modulation of genes associated with glutamatergic (excitatory) neurotransmission (including upregulation of the ionotropic glutamate receptor subunits, GluRdelta2 and GluR1 and downregulation of the NMDA receptor subunits, NR2B, NR2A and NR2C and the metabotropic glutamate receptor, mGluR7, Fig. 4D). Furthermore, a variety of voltage gated ion channels including calcium, sodium, chloride, and potassium channels were affected by all three epileptogenic treatments (Fig. 4E), including a noteworthy downregulation of voltage gated (Kv7.3 and Kv8.1) and inward rectifying (Kir3.1) potassium channels. We also found significant modulation of glial-specific genes beginning at the early time point (Fig. 4F): the cytoskeletal proteins GFAP and vimentin (Vim), and several calcium binding proteins (S100a6, S100a10, s100a11) were all upregulated while gap junction connexins 30 and 43 (Cx30 and Cx43) and the inward rectifying potassium channel Kir4.1 were downregulated. Microarray-based gene expression measurements for selected genes were further verified using quantitative real-time PCR. Expression patterns were similar although the magnitude of the fold changes sometimes differed (Supplementary Fig. 1).
Given the high correlation between expression profiles following the three epileptogenic treatments, combined with the biochemical evidence that albumin binds to TGF-β receptors and the physiological evidence that TGF-β1 induces evoked epileptiform activity, we assessed the extent to which each treatment activates transcription of genes known to be associated with the TGF-β signaling pathway using GenMAPP (Salomonis et al., 2007). 43% of genes analyzed in the TGF-β signaling pathway were modulated by both treatment with albumin and TGF-β1. Genes which showed at least a 1.5 fold change in expression following albumin or TGF-β1 treatment, are highlighted in figure 5. Importantly, 86% of genes modulated by TGF-β1 treatment are also modulated following albumin treatment indicating a high degree of overlap. To check the specificity of this pathway activation, additional pathways were analyzed. Indeed, there was still a high degree of overlap but the percent of genes modulated by both albumin and TGF-β1 treatments was much lower. For example, for the androgen receptor and alpha6-beta4 integrin signaling pathways, only 21.7% and 24% of genes analyzed were modulated by both treatments.
The above evidence indicates that TGF-β signaling is a key mediator of albumin-induced epileptogenesis. To determine if the global transcriptional response seen following albumin treatment is dependent on activation of the TGF-β signaling pathway, we performed an additional set of microarray expression profiles using rats treated with albumin in the absence (n=3) or presence of TGF-βRI and II blockers (n=4, TGF-βR1 kinase activity inhibitor SB431542 and anti-TGF-βRII antibody) and sacrificed 24 hours following treatment. Although some changes in gene expression resulting from albumin treatment were still present following the blocker treatment, the majority of these changes were absent or attenuated following TGF-β pathway blocker treatment (Fig. 6A), confirming dependence of the albumin-induced transcriptional response on TGF-β signaling.
Gene ontology analysis was then used to reveal which biological processes were blocked following TGF-β pathway blocker treatment (Fig. 6A; for complete GO term annotation results see Supplementary Tables 5–8). Genes in the TGF-β signaling GO term demonstrated a dramatic suppression of the albumin-induced expression changes in the presence of TGF-β signaling blockers (Fig. 6B). In addition, TGF-β pathway blocker treatment prevented the albumin-induced modulation of genes involved in neuronal processes, immune response, and ion and cellular transport (Fig 6B). Several prominent signaling pathways including the NF-kappaB cascade, Jak-Stat cascade, and MAPKKK cascade were upregulated following albumin treatment, but did not show a similar upregulation following albumin treatment in the presence of TGF-β pathway blockers. Quantitative real-time PCR was also performed with these samples to confirm the microarray results (Fig. 6C). Indeed, TGF-β pathway blocker treatment completely blocked expression changes following albumin exposure for Stat3 and Glt-1 and partially blocked changes for Cx43 and GFAP.
BBB breakdown is a hallmark of vascular injury in the brain and is observed in numerous neurological diseases including traumatic brain injury, stroke and neurodegenerative diseases (Neuwelt, 2004; Abbott et al., 2006; Oby and Janigro, 2006; Zlokovic, 2008). Compromise of the BBB is triggered by preceding processes that in turn cause vascular injury. For example, perivascular astrocytes or perivascular microglia could be activated by an initial precipitating event (trauma or ischemia) and cause vascular injury leading to serum albumin extravasation into the brain parenchyma. We have previously shown development of cortical dysfunction – specifically, hypersynchronous neuronal activity following BBB opening or exposure to serum albumin (Seiffert et al., 2004; Ivens et al., 2007). Cortical dysfunction was followed by reduced dendritic branching and neuronal loss several weeks following either treatment (Ivens et al., 2007).
In this study, we now extend these findings to include the appearance of spontaneous, prolonged and hypersynchronous interictal-like activity following treatment with albumin in a solution containing serum levels of electrolytes (aSERUM). We did not observe spontaneous activity when slices were treated with albumin or TGF-β in aCSF, probably due to the generally low excitability and lack of spontaneous activity of the deafferented slice preparation (Connors et al., 1982; Gutnick et al., 1982). The changes in electrolyte concentrations in the aSERUM solution (e.g. higher K+, lower Mg2+ and Ca2+) are probably sufficient to increase neuronal excitability such that the epileptiform activity, which characterizes the network upon activation, appears spontaneously. We have previously shown that spontaneous activity is rarely evoked in slices treated with albumin in aCSF (~10% of slices, see also (Seiffert et al., 2004). Importantly, spontaneous recurrent seizures followed by secondary generalization were also observed in some animals treated with albumin (Ivens et al., 2007). These observations combined with the appearance of epileptiform activity following albumin treatment in vitro likely reflect abnormal network epileptic network activity in vivo.
Since serum albumin is sufficient to induce epileptic-like activity, our initial hypothesis was that albumin enters the normally inaccessible CNS environment, binds to TGF-β receptors, and causes a cascade of events culminating in epileptiform activity. Here we report similar development of epileptiform activity following exposure to TGF-β1, demonstrating the importance of TGF-β pathway activation in this injury model. We further show that albumin binds TGF-βRII and induces the phosphorylation of Smad2. Given the latent period prior to the appearance of epileptiform activity (ie epileptogenesis), we hypothesized that BBB breakdown, albumin, and TGF-β1 share a common mechanism, specifically a transcriptional mechanism involving the TGF-β pathway. Indeed, clustering and gene ontology analysis revealed striking similarities across the three treatments. The most prominent finding from our microarray results is the identification of a key role for TGF-β signaling, with activation of TGF-β related genes seen in response to TGF-β1, albumin, and BBB breakdown. Blocking TGF-β signaling prevented the majority of the albumin-induced transcriptional responses, allowing us to narrow our gene list and identify the genes that are most relevant to epileptogenesis under these conditions. Furthermore, application of TGF-βR blockers suppressed the development of epileptiform neuronal activity following albumin or TGF-β1 treatment, highlighting this pathway as a novel therapeutic target.
Interestingly, TGF-β1 is generally considered to have a protective function. Previous studies have demonstrated the ability of TGF-β1 to prevent glutamate neurotoxicity in vitro and protect against ischemic injury in vivo (Prehn et al., 1993; Henrich-Noack et al., 1996; Ruocco et al., 1999). On the other hand, studies have also suggested TGF-β1 may not always be neuroprotective. Transgenic mice overexpressing TGF-β1 specifically in astrocytes (Wyss-Coray et al., 1995) developed seizures (along with overproduction of extracellular matrix components, severe communicating hydrocephalus, motor incoordination, and early runting). Furthermore, a recent study has demonstrated that blockade of TGF-β-Smad2/3 signaling in peripheral macrophages in a mouse model for Alzheimer’s disease results in marked attenuation of cerebrovascular-amyloid deposits (Town et al., 2008). These studies along with our work illustrate the complexity and importance of TGF-β signaling in neurological diseases. Additional work is needed to elucidate the mechanisms responsible for the protective and detrimental effects of TGF-β signaling.
Other significant findings from our microarray results include the early upregulation of genes involved in inflammatory processes and the delayed downregulation of genes involved in neuronal processes including synaptic transmission and ion transport. Features of CNS inflammation, such as glial and complement activation, cytokine production and adhesion protein expression, were all present in our array data. Upregulation of genes involved in activation of the NF-kappaB pathway and complement cascade reflects a significant innate immune response. In recent years, several studies have shed light on the importance of inflammatory processes in epilepsy. Both NF-kappaB pathway activation and complement activation have been reported in various epilepsy animal models (Rozovsky et al., 1994; Rong and Baudry, 1996; Lubin et al., 2007; van Vliet et al., 2007), as well as the involvement of other immune response genes including Il6 (Balosso et al., 2008), Ccl2 (Calvo et al., 1996; Manley et al., 2007), Stat3 (Choi et al., 2003), and Fn1 (Hoffman and Johnston, 1998). A study by Rizzi et al., (2003), where status epilepticus (SE) was induced with kainic acid, found cytokines to be causally involved in the SE-induced neuronal damage, as cytokine synthesis preceded hippocampal neuronal injury and this injury only occurred when cytokines were produced (Rizzi et al., 2003).
Some immune-related genes have also been shown to play a role in neuronal functions. Cytokines such as interleukin-1β (Il1b) and tumor necrosis factor (TNF) have been shown to increase neuronal excitability. Il1b mediates increased neuronal glutamate release (Casamenti et al., 1999), induces phosphorylation of the NMDA NR2B subunit (Viviani et al., 2003; Balosso et al., 2008) and prevents glutamate uptake by astrocytes (Hu et al., 2000). TNF promotes recruitment of AMPA receptors lacking the Glur2 subunit to the neuronal membrane leading to increased calcium influx and also promotes endocytosis of GABA A receptors (Beattie et al., 2002; Stellwagen et al., 2005; Leonoudakis et al., 2008). An upregulation of TNF was observed in our arrays indicating it may be involved in increasing neuronal excitability during epileptogenesis in addition to its immune related functions. Another immune related protein with a possible non-immune function following injury is C1q, the first component of the classical complement cascade. C1q has recently been shown to play a role in synapse elimination during development and in neurodegeneration (Stevens et al., 2007). Upregulation of several C1q subcomponents in our arrays may be associated with synaptic remodeling and neuronal loss following the generation of the epileptic focus (Ivens et al., 2007).
An imbalance of excitatory and inhibitory transmission resulting in increased network excitability is a feature of most epilepsy models. In our study, several GABA-A receptor subunits were downregulated while ionotropic glutamate receptor subunits were upregulated. Genes involved in glutamate transport (Glt-1) as well as voltage gated ion channels were also downregulated. These changes, appearing in the middle-late time points predict an increase in network excitability. These results are also consistent with a previous microarray study examining gene expression changes during epileptogenesis following electrically induced SE (van Vliet et al., 2007), which found downregulation of GABA-A and NMDA receptor subunits.
Our initial findings showing albumin uptake by astrocytes (Ivens et al., 2007) raised the hypothesis that an early glial response precedes the development of abnormal neuronal activity. This hypothesis is reinforced by our microarray data. Glial cells are responsible for the homeostasis of the extracellular environment (i.e. uptake of glutamate and potassium) under normal conditions and for the initiation of inflammatory processes in the central nervous system when activated following injury. Our results support various studies that have implicated a role for these cells in epileptic activity (Tian et al., 2005; Wetherington et al., 2008). Importantly, the early astrocytic response, before epileptic activity is observed, further strengthens the putative role of these cells in the epileptogenic process itself. The downregulation of the astrocytic glutamate transporter, Glt-1, the inward rectifier potassium channels (Kir 4.1) and connexins Cx26, Cx30 and Cx43 all predict impaired buffering of glutamate and potassium which in turn will enhance local neuronal excitability. The upregulation of the transcription factor Stat3 predicts astrocytic differentiation (Morita et al., 1995) and transformation (Choi et al., 2003). In addition, upregulation of GFAP and vimentin provide evidence for reactive gliosis found in many epilepsy models, which is characterized by hypertrophy and increased intermediate filament expression (Wetherington et al., 2008). Overall, our microarray data suggest a pronounced immediate glial response followed by a later neuronal response. It remains to be determined whether our treatments induce the release of mediators from astrocytes that in turn activate transcription in neurons as well as in other cell types such as microglia and endothelial cells.
While many of the molecular changes revealed by our microarray study are shared with different epilepsy models, our study is the first to demonstrate that these changes, following direct vascular injury (i.e. BBB breakdown), are associated with brain exposure to serum albumin and are mediated via the TGF-β signaling pathway. Indeed, using TGF-β pathway blockers we were able to not only block the transcriptional response following albumin exposure in vivo, but also the development of epileptiform activity. Combined, these results present the TGF-β pathway as a novel therapeutic tool for preventing injury-related epileptogenesis.
The authors thank Prof. John Ngai and Karen Vranizan for their help with the microarray design and analysis and Aaron Friedman for comments on the manuscript. This study was supported by the CURE foundation award (DK and AF), Sonderforschungsbereich TR3(C8) (AF and UH), the Mary Elizabeth Rennie Epilepsy Foundation research grant (DK), the Israel Science Foundation (AF), and the United States-Israel Binational Science Foundation (AF and DK)