We investigated in vivo the expression of the tissue inhibitor of metalloproteinases-1 (TIMP-1) in the rat CNS after kainate (KA)-induced excitotoxic seizures. In situhybridization revealed that TIMP-1 mRNA is induced rapidly and massively in most regions of the adult forebrain after KA treatment. Neuronal activity seems to be necessary but not sufficient to trigger TIMP-1 induction, because it is not observed in seizing 10-d-old pups, unlike what is observed in 21- and 35-d-old animals after seizures. The rapid induction of TIMP-1 is not prevented by the inhibitor of protein synthesis cycloheximide, suggesting that, after seizures, TIMP-1 is induced in neurons as an immediate early gene (IEG). The initial neuronal upregulation is followed by enhanced expression in astrocytes, as assessed by double-labeling experiments. In the hippocampus rapid increases in mRNA are followed by relatively delayed (8 hr after KA) increases in TIMP-1 immunoreactivity in the perisomatic and dendro–axonic areas, suggesting secretion of the protein. At 3 d after KA treatment, strong immunoreactivity is found in astrocytes and in the cell bodies and dendro–axonic projections of resistant neurons such as the dentate granule cells. Taken together, the results suggest that TIMP-1 may be instrumental for neurons and astrocytes in coupling early cellular events triggered by seizures with the regulation of long-lasting changes involved in tissue reorganization and/or neuroprotection.
- protease inhibitor
- brain seizures
- neuronal death
- immediate early gene
- extracellular matrix
- tissue remodeling
Tissue inhibitor of metalloproteinases-1 (TIMP-1) is a 28.5 kDa secreted glycoprotein (Murphy et al., 1977; Cawston et al., 1981) that shares 40 and 28% amino acid sequence homology with its family partners TIMP-2 (Stetler-Stevenson et al., 1989) and TIMP-3 (Apte et al., 1994; Leco et al., 1994), respectively. Although considered a multifunctional molecule that possesses growth-promoting properties (Docherty et al., 1985; Gasson et al., 1985) and mitogenic activity (Bertaux et al., 1991; Hayakawa et al., 1992) in a variety of cultured cells, TIMP-1 is recognized mostly as an inhibitor of matrix metalloproteinases (MMPs) (Cawston et al., 1981). MMPs constitute a family of zinc-binding and Ca2+-dependent endopeptidases, including gelatinases, collagenases, and stromelysins (Emonard and Grimaud, 1990). The excessive cleavage of the extracellular matrix (ECM) associated with an imbalance of the MMPs/TIMPs ratio has been correlated with the invasive potential of brain tumor cells (Khokha et al., 1989; Nakagawa et al., 1994; Mohanam et al., 1995) and with the histopathogenesis of inflammatory-related diseases like rheumatoid arthritis or multiple sclerosis, as reviewed by Dayer and Burger (1994)and by Hodgson (1995).
In non-neural cells the expression of MMPs and TIMP-1 is modulated by extracellular factors, including cytokines and growth factors like TNF-α (Chua and Chua, 1990; So et al., 1992), IL-1β (Shingu et al., 1993), bFGF, EGF, and TGFβ (Edwards et al., 1987) and by phorbol esters (Murphy et al., 1985). In contrast, very little is known about the regulation of TIMP-1 expression in the CNS. Nedivi et al. (1993)first isolated TIMP-1 as a candidate plasticity gene in the dentate gyrus of kainate (KA)-treated rats and demonstrated, by using Northern blot approaches, that its expression is modulated in this region by stimuli producing long-term potentiation and during postnatal development. These observations are in keeping with the hypothesis that controlled proteolysis is crucial in the development and plasticity of the CNS (Lynch and Baudry, 1984; Monard, 1988; Frey et al., 1996), whereas excessive proteolysis contributes to various neuropathologies (Siman et al., 1989; Lee et al., 1991; Rosenberg et al., 1992). In particular, brain seizures and associated pathology recently have been linked with alterations of serine protease expression (Qian et al., 1993; Chen et al., 1995; Tsirka et al., 1995), mutations in an endogenous cysteine protease inhibitor (Pennacchio et al., 1996), and, most interestingly, with the induction of stromelysin-1 (Reeben et al., 1996), a metalloproteinases that is inhibited by TIMP-1 (Murphy et al., 1991). Taken together, these data strongly suggest that proteases are involved in seizure-associated pathology and that their inhibitors, including TIMP-1, may be instrumental in regulating their deleterious effects. To evaluate this hypothesis, we find it is important first to establish the distribution and regulation of TIMP-1 in the seizing brain.
As an extension of the findings of Nedivi et al. (1993), we report here that KA-mediated excitotoxic seizures induce TIMP-1 expression in a time-, region-, cell-, and age-dependent manner and that the initial induction in neurons is independent of novel protein synthesis. In addition, we discuss the differential and sustained expression of TIMP-1 in neurons and astrocytes in relation with pathology.
MATERIALS AND METHODS
Animals and treatments. Experiments involving animals were approved by the French Ethical Science Committee (Statement No. 04223). They were performed with male Wistar rats at various postnatal ages (P). Rat had access to food and water ad libitum and were housed under a 12 hr light–dark cycle at 22–24°C. All drugs were injected intraperitoneally (0.5 ml/100 gm). Seizures were induced, as previously described (Tremblay et al., 1984), by KA (Sigma, St. Louis, MO) dissolved in 0.9% saline and given at varying doses, depending on the postnatal age: 1–2 mg/kg at P10, 6–7 mg/kg at P21, and 9–10 mg/kg at P35–P60. After KA injection each animal was placed in a Plexiglas cylinder for behavioral observation. The animals that did not show a stereotyped seizure behavior were not included in the study. Control animals were injected with the same volume of 0.9% saline.
After various delays after KA injections (from 50 min to 14 d), 65 adult animals (P60) were anesthetized with pentobarbital and decapitated; their brains were removed rapidly onto a cold plate, dipped for 1 min in cold isopentane (−50°C), and then stored at −80°C until they were sliced into cryostat sections 15 μm thick, mounted onto gelatinized slides, and kept at −70°C until used forin situ hybridization, immunohistochemistry, or histology. Three KA-treated animals and their respective saline controls received, in addition, repeated injections of the anticonvulsive drug diazepam (25 mg/kg, Hoffmann-La Roche, Basel, Switzerland) and were killed 90 min later.
A separate group of 12 adult rats was used to study the effects of the protein synthesis inhibitor cycloheximide on the KA-mediated TIMP-1 mRNA induction. Animals were injected with (1) KA (10 mg/kg) alone; (2) 2 mg/kg cycloheximide twice, 1 hr before and 1 hr after KA injection; (3) cycloheximide plus saline twice, as in (2); and (4) saline control. In this paradigm animals were killed 2 hr after KA or saline injections.
For studies addressing the developmental regulation of TIMP-1 mRNA expression by seizures, we treated 18 rats of P10, P21, and P35 with either saline or KA. Sixteen hours later they were killed, and tissue was processed as indicated above.
For TIMP-1 immunohistochemistry and double-labeling procedures combining in situ hybridization and immunohistochemistry, 18 adult rats were treated with KA (10 mg/kg) and 6 with saline and intracardially perfused with 0.9% saline, followed by 4% paraformaldehyde in 0.1 m phosphate buffer (PPB) at various times: 1.5, 4, 8, 16, and 72 hr and 14 d after treatment. Brains were removed rapidly onto a cold plate, post-fixed for 24 hr in PPB at 4°C, cryoprotected for 24 hr in PPB containing 20% sucrose, and sectioned (24 μm thickness) in the coronal plane with a freezing microtome. Tissue sections were collected into PPB and stored at 4°C for at least 3 d before further processing.
In all experiments a minimum of three animals per age and treatment were used for statistical analysis with the nonparametric Kruskal–Wallis test, followed by the Mann–Whitney U test, for comparison between pairs of groups.
In situ hybridization. In vitrotranscription of 35S-UTP labeled TIMP-1 riboprobes was performed from linearized pBluescript plasmids with a Promega (Madison, WI) in vitro transcription kit with T7 (Pharmacia, Uppsala, Sweden) and T3 (Boehringer Mannheim, Mannheim, Germany) RNA polymerases for sense and antisense probes, respectively. The plasmid (kindly provided by Dr. Yoav Citri, Weizmann Institute of Science, Israel) carried a 700 bp rat cDNA containing part of the 5′ untranslated region and the entire coding region. This probe recognized only one band where size corresponds to the TIMP-1 mRNA on Northern blots with brain RNA (Nedivi et al., 1993). Because a 150 bp sequence of the TIMP-1 3′ coding region is highly homologous to a sequence contained in murine β-interferon (Skup et al., 1982), control experiments were performed with a 35S-dATP-labeled oligonucleotide selected in a region of the rat TIMP-1 (5′ 79–123 3′) nonhomologous to β-interferon. The in situ hybridization results obtained with such an oligoprobe were identical to those obtained with the TIMP-1 riboprobe in both control and KA-treated animals.
Frozen sections were brought to room temperature and fixed in cold 4% paraformaldehyde in 0.1 m phosphate buffer (PB) for 30 min and then rinsed in glycine-phosphate buffer, acetylated, rinsed, dehydrated, delipidated, and incubated with 1 × 106cpm/100 μl of antisense or sense riboprobes overnight at 50°C. Tissue was rinsed twice for 30 min at 60°C in 4× SSC, treated with 20 μg/ml RNase A (Boehringer Mannheim), and finally washed in increasingly stringent conditions up to 0.1× SSC at 60°C during 30 min. All tissue was processed for both film (Biomax, Kodak, Rochester, NY) and emulsion autoradiography (NTB2, Kodak), with exposure times of 3–5 d and 3–4 weeks, respectively. After development of emulsion autoradiograms, the sections were counterstained with cresyl violet and mounted with Permount.
Densitometric analysis of mRNA levels on film autoradiograms was performed with the Samba/2005 image analysis system (TITN Alcatel, France). Multiple measures were obtained from at least four tissue sections per brain from each rat.
Immunohistochemistry. For TIMP-1 immunohistochemistry PPB fixed floating sections were rinsed in PBS. Endogenous peroxidase activity was inhibited with 0.1% H2O2 (Sigma) for 20 min. Tissue was preincubated for 30 min at room temperature in PBS plus 0.1% Triton X-100 (Sigma), 0.2% gelatin (Bio-Rad, Munich, Germany), and 3% normal goat serum (Vector Laboratories, Burlingame, CA) and then incubated at 4°C overnight in a solution containing a rabbit polyclonal anti-TIMP-1 (1:500, Chemicon, Temecula, CA) plus 0.1% Triton plus 0.2% gelatin plus 1% normal goat serum. This antibody recognizes the rat protein and does not cross-react with any other protein on Western blots. Tissue then was rinsed in PBS, incubated 1 hr at room temperature with a biotinylated goat anti-rabbit antibody (1:400, Vector), rinsed in PBS, incubated for 90 min at room temperature in the avidin–biotin–peroxidase kit (ABC kit, Vector), developed in 0.05% diaminobenzidine (Sigma) plus 0.2% ammonium nickel (II) sulfate (Sigma) plus 0.1% H2O2 in 0.05m Tris buffer, dehydrated, and mounted.
To verify an efficient inhibition of protein synthesis by cycloheximide, we tested whether KA-induced expression of Fos protein was blocked by cycloheximide administered before and after KA injections. Adjacent brain sections to those used in the in situ hybridization study for TIMP-1 mRNA expression were fixed for 20 min in cold PPB and processed as described above with slight modifications: the incubation solution contained 0.1 mlysine instead of normal goat serum; the first antibody was a rabbit polyclonal anti-human Fos (PCO5, Oncogene Science, Uniondale, NY) diluted 1:1000 in PBS containing 0.1% gelatin and 0.15% Triton; streptavidine–biotin–peroxidase (1:400, Vector) was used instead of the ABC kit.
In all cases, sections incubated in the absence of the primary antibody were not immunoreactive.
Double-labeling procedures combining in situhybridization and immunohistochemistry. To identify precisely the cell types expressing TIMP-1 mRNA, we combined in situ hybridization with a TIMP-1 35S-cRNA probe and immunohistochemistry, using either a neuronal (MAP-5 antibody) or an astrocytic (GFAP antibody) marker. Using microtome-cut floating sections, we performed in situ hybridization as indicated above, except for the dehydration and delipidation steps that were performed after the immunohistochemistry procedure. After the last wash in 0.1× SSC, tissue was rinsed three times for 10 min in PBS, pH 7.4, preincubated for 1 hr in a PBS solution containing 2% gelatin, 3% normal goat serum and 0.1% Triton X-100, followed by overnight incubation at 4°C with either rabbit polyclonal anti-GFAP (1:300, Dako, High Wycombe, UK) or mouse monoclonal anti-MAP-5 (1:2000, Sigma) in PBS containing 2% gelatin and 1% normal goat serum. The next day, tissue was rinsed in PBS, incubated with a biotinylated goat anti-rabbit or goat anti-mouse antibody (1:400, Vector) for 1 hr, rinsed in PBS, incubated with the ABC kit (1:200) for 1 hr, rinsed, and developed with DAB (without nickel intensification). Sections were mounted, dehydrated, delipidated, and processed for emulsion autoradiography, as indicated above.
Histology. Cresyl violet was used to evaluate general histology and cell degeneration in cryostat sections adjacent to those used for in situ hybridization studies.
Seizures in developing and adult rats
The behavioral effects induced by systemic administration of KA varied with the postnatal age, as previously described (Ben-Ari et al., 1981; Tremblay et al., 1984). Thus, in P10 animals scratching-like movements appeared after 20–25 min of KA injection, rapidly followed by tonicoclonic seizures. Scratching-like movements still were observed in P21 rats after KA, but not in P35–P60 animals receiving the glutamate analog. Moreover, in P21, P35, and P60 animals, KA injections induced “wet-dog shakes” 35–45 min after injection, followed 25–30 min later by typical limbic motor seizures (but no tonicoclonic seizures) and status epilepticus that could last for several hours. Cycloheximide treatment did not modify the behavioral manifestations of KA-induced seizures.
TIMP-1 mRNA expression after KA-induced seizures in adult rats
In control adult animals (Fig.1 A) mRNA expression appeared to be very low in most forebrain areas with the exception of the hippocampus, which exhibited above average hybridization levels. In contrast, parenteral injections of KA dramatically induced TIMP-1 mRNA in a time- and region-specific manner (Figs. 1, 2). No changes in basal TIMP-1 hybridization signal were detected before the manifestation of behavioral seizures (50 min after KA injection). However, as early as 90 min after KA injection (between 15 and 30 min after the onset of behavioral seizures), the expression of TIMP-1 mRNA increased strongly in the stratum granulosum of the hippocampus (Fig.1 B). Peak expression in this layer occurred between 4 and 8 hr (Fig. 1 C), as shown by tenfold increases measured by densitometric analysis of the film autoradiograms (Fig. 2). During the first 8 hr statistically significant (p < 0.01) increases in mRNA levels, between two- and fourfold, also were detected in the strata pyramidale of regions CA1 and CA3, respectively. By 8 hr post-KA injection (Fig. 1 C,I), increased labeling no longer was restricted to the neuronal cell layers but also was distributed as scattered clusters of autoradiographic grains in all laminae. In addition, at this time point, hybridization also was highly elevated in amygdaloid nuclei, the entorhinal cortex, the primary olfactory cortex, and the hypothalamus and was increased slightly in superficial layers of the neocortex and in the paraventricular thalamic nuclei. In brains from animals killed at 16 (Fig.1 D,J) and 24 hr post-KA treatment, the hybridization signal was increased further in the strata radiatum, oriens lucidum, and lacunosum moleculare of the hippocampus proper and in the molecular layer and the hilus of the dentate gyrus, along with areas of the neocortex, the medial thalamic nuclei, and the hypothalamus. Hybridization signal remained highly elevated in the entorhinal and olfactory cortex. Densitometric analysis (Fig. 2) revealed that, by 24 hr post-KA injection, TIMP-1 mRNA levels peaked in the strata pyramidale of regions CA1 (fourfold increase, p < 0.01) and CA3 (sevenfold increase, p < 0.01) while dropping to values six- to sevenfold (p < 0.01) of those of paired control animals in the dentate gyrus. At 48 and 72 hr (Fig. 1 E), labeling remained elevated in areas of degeneration such as the stratum pyramidale of the hippocampus (p < 0.5), the amygdaloid nuclei, and the superficial and deep layers of the entorhinal cortex, as compared with labeling in control animals. Thereafter, hybridization signal progressively declined across all brain regions, virtually reaching control levels by 14 d post-KA treatment (Fig. 1 G). The KA-induced increase in TIMP-1 mRNA levels was blocked by systemic co-injections of the GABA/benzodiazepine receptor agonist diazepam (results not shown), which also blocked seizure activity, indicating that the induction of the mRNA expression was not attributable to a nonspecific effect of KA. Hybridizations with the cRNA sense probe gave no labeling in brain sections of saline- and KA-treated animals (Fig.1 H).
Developmental regulation of TIMP-1 mRNA expression after KA-induced seizures
To investigate whether neuronal activity was sufficient to induce TIMP-1 mRNA expression and whether it could be correlated with the developmentally linked neuronal vulnerability to KA, we treated animals 10, 21, and 35 d postnatal with convulsant doses of KA. We then evaluated the expression of TIMP-1 mRNA 16 hr later by in situ hybridization both in these animals and in their respective controls (Fig. 3). TIMP-1 mRNA was expressed poorly in the brains of control animals at all developmental stages studied. Maximum levels of TIMP-1 mRNA expression were found in the granule cell layers of the cerebellum and the main hippocampal neuronal layers of 10-d-old animals. The density of labeling was greater in the stratum pyramidale of region CA1 and in the stratum granulosum than in the stratum pyramidale of CA3. Hybridization signal in the pyramidal and granule cell layers of the hippocampus decreased throughout postnatal development and was lowest in P35 animals (Fig. 3 E). In KA-treated animals TIMP-1 mRNA was not induced in response to generalized tonicoclonic seizures in 10-d-old rats (Fig.3 B). In contrast, by P21 (Fig. 3 D) and P35 (Fig.3 F), increases of TIMP-1 mRNA levels in response to limbic seizures were ∼five- to tenfold, as compared with age-paired controls in specific brain structures. Increases in P21 brains were confined to the hippocampus, the anterior septal, hypothalamic, and rhinencephalic areas of projection. In P35 animals, these increases extended to the entire cortical area, the caudate putamen, and the thalamic and hypothalamic nuclei. No changes in TIMP-1 mRNA expression were observed in the cerebellum or the midbrain after KA treatment.
Effects of cycloheximide on KA-mediated induction of TIMP-1 mRNA
The rapid induction of TIMP-1 mRNA expression after KA-induced seizures and its propagation in the limbic system was reminiscent of the pattern of expression of various immediate early genes (IEG). To establish whether the rapid seizure-mediated induction of TIMP-1 mRNA was dependent on novel protein synthesis, we injected the protein synthesis inhibitor cycloheximide 1 hr before and 1 hr after KA or saline injections. As illustrated in Figure 4, cycloheximide treatment (Fig. 4 B,D) did not alter the expression of TIMP-1 mRNA in the granule cell and pyramidal layers of control (Fig. 4 A) and KA-treated animals (Fig.4 C). However, it blocked (Fig. 4 F) the KA-mediated induction of Fos protein (Fig. 4 E), indicating that inhibition of protein synthesis had been effective. No Fos immunostaining was detected in brains from saline- or saline/cycloheximide-injected animals (results not shown).
Colocalization of TIMP-1 mRNA with neuronal and astroglial markers
The distribution of TIMP-1 labeling after KA treatment suggested that mRNA expression predominantly increased in neurons up to 8 hr after treatment and in both neurons and glial cells thereafter. For confirmation, KA-treated animals and their respective controls were killed at 1.5, 8, and 16 hr after injections. Brain tissue sections were processed for simultaneous in situ hybridization autoradiography and immunohistochemistry with either MAP-5-specific or GFAP-specific antibodies for neuronal and astroglial identification, respectively. At 1.5 and 8 hr post-KA, the highest density of labeling overlapped MAP-5 stained areas of high neuronal body density. When they were observed at high magnification, it was obvious that dense clusters of silver grains associated primarily with MAP-5-positive dentate granule cells (Fig. 5 A), piriform cortex (Fig. 5 C) and amygdaloid neurons (data not shown), and less densely packed grains overlapped neurons in the hippocampal pyramidal cell layer, in the neocortex, in the thalamus, and in the hypothalamus (data not shown). Hippocampal interneurons appeared generally unlabeled, and GFAP-positive cells exhibiting astrocyte morphology displayed a low density of grain clustering (Fig. 5 B). At 16 post-KA injection, the density of the autoradiographic grains was increased or reduced over MAP-5-positive cells (Fig. 5 E), depending on the neuronal population, and clearly increased over GFAP-stained cells (Fig. 5 D,F) across all of the regions expressing TIMP-1 mRNA. Most GFAP-positive cells were labeled with TIMP-1 cRNA, and a particularly intense signal was observed in the non-neuronal cell body layers of the hippocampus and in the amygdala. Overall, radiolabeling was confined to neuronal and glial cell bodies, with some astrocytes showing hybridization signal in their immunostained processes only occasionally.
TIMP-1 immunoreactivity in normal brain and after KA-induced seizures
TIMP-1 immunostaining in adult brains was distributed broadly across cerebellar (data not shown) and forebrain regions. In all brain areas examined, the main locus of TIMP-1 immunoreactivity appeared to be the cell bodies and processes of cells morphologically identified as neurons. In most cases, the latter exhibited a clear perinuclear staining; occasionally, a few neurons appeared to be labeled in the nuclear compartment also. Light immunostaining also was observed in dendritic areas of different brain regions such as the stratum lucidum of region CA3 and the dentate molecular layer in the hippocampus (Fig.6 A). Within the neuronal hippocampal layers immunoreactivity decreased along the rostrocaudal axis of the brain. In the same coronal plane the intensity of staining ranked as follows: CA1 > DG > CA3 (Fig. 6 A). Outside the hippocampus robust immunostaining was observed in neurons of the neocortex, the cingulate, and the piriform cortex, whereas it was lighter in thalamic and amygdaloid neurons (data not shown). Staining in areas enriched in white matter, like the corpus callosum, was close to background levels, and only at high magnification were we able to detect the presence of immunoreactive glial cells. The pia mater and the ependymal lining of the ventricles also were immunoreactive. No other signs of glial TIMP-1 immunoreactivity were detected in control brains.
Changes in TIMP-1 immunoreactivity after seizures seemed to be region- and cell type-specific across time. At 4 hr after KA injections there was no evidence for changes in TIMP-1 immunoreactivity in the hippocampus (data not shown), as compared with saline-injected controls or in any other brain structure studied. By 8 (Fig.6 B) and 16 hr (data not shown) post-KA, we observed a consistent reduction in cell body immunostaining of pyramidal and granule cells in the hippocampus. In contrast, diffuse staining appeared to increase at these time points in the molecular layer of the dentate gyrus (in particular in the distal area) and in the strata radiatum and oriens. Similar observations were made in thalamic and cortical areas, namely a decrease in cell body staining and an increase in the surrounding diffuse immunoreactivity (data not shown). At 72 hr post-KA injection (Fig. 6 C), differences in immunostaining were increased further between treated and control hippocampi: TIMP-1 immunoreactivity in pyramidal CA1 neurons virtually was undetectable, whereas it was increased dramatically in all layers of the CA3 and dentate gyrus regions. The presence of strongly immunostained glial cells distributed across all hippocampal laminae was another important feature at 72 hr post-KA. By 14 d post-KA injection (Fig.6 D), immunostaining in the hippocampus was still high in the dendritic areas and in glial cells located mostly in the stratum radiatum of region CA1 and in the molecular layers of the dentate gyrus. Microscopic observations at higher magnification (Fig.7) revealed that TIMP-1 immunoreactivity was distributed homogeneously in the cell bodies of neurons in the strata pyramidale (Fig. 7 A,B) and granulosum (data not shown). In all cases, 8 hr after KA injection, immunostaining in the cell bodies was confined mostly to the perisomatic area and was increased as diffuse immunostaining in the dendritic areas (Fig. 7 C,D). At 72 hr post-KA treatment, TIMP-1 immunoreactivity in the CA1 region (Fig.7 E,F) was strong in cells that appeared to be, by size, morphology, and spatial distribution, mostly astrocytes. These were found primarily in vulnerable structures such as the hippocampus, the entorhinal cortex, and the amygdala and displayed a strong immunoreactivity in the nuclear compartment and also in the processes. Moreover, glial cells located within the corpus callosum were stained more darkly than in control brains.
Delayed neuronal death in the CA1 region after systemic KA injections
Incipient signs of degeneration in pyramidal neurons of CA1 were detected first by Nissl staining at 24 hr post-KA. At 48 hr, degeneration was obvious in this layer (data not shown). At 72 hr post-KA treatment, there was strong evidence for CA1 pyramidal neuron demise, as revealed by Nissl staining (Fig. 7 G,H). In keeping with previously reported results (Schowb et al., 1980; Ben-Ari et al., 1981), neuronal degeneration in the hippocampus after KA systemic injections was found mainly in the CA1 stratum pyramidale. Only occasionally did we find overt signs of degeneration in the boundary of CA3 with CA2–CA1 subfields. In any case, degeneration in region CA3 was never detected before 72 hr post-KA treatment. As previously shown (Ben-Ari et al., 1980), the CA3 area is most vulnerable after intra-amygdaloid injections of KA. Other regions that showed signs of neurodegeneration (data not shown) after systemic KA administration include the amygdaloid nuclei, the entorhinal cortex, and the dorsal thalamus, in agreement with Schowb et al. (1980) and with Ben-Ari et al. (1981).
Our results show that KA-induced seizures dramatically upregulate in a region-, time-, and age-dependent manner the mRNA of TIMP-1 in both neurons and astrocytes. We provide evidence that seizure-dependent induction of TIMP-1 mRNA in vivo is independent of de novo protein synthesis, indicating that after seizures TIMP-1 is induced as an IEG. Finally, sustained immunoreactivity of TIMP-1 in vulnerable brain structures such as the hippocampus suggests that it may be involved in tissue repair and/or cell survival after seizures in the mammalian CNS.
TIMP-1 mRNA expression is upregulated differentially in adult and young rats after seizures
In situ hybridization histochemistry reveals that basal levels of TIMP-1 mRNA are very low in the normal neonatal and adult rat brain, in agreement with previous Northern blot data (Nedivi et al., 1993; Okada et al., 1994). We further have established a developmental downregulation of TIMP-1 mRNA expression in all hippocampal neuronal layers, with the lowest levels found in the adult brain. Developmental changes in TIMP-1 gene expression may be associated with the plastic remodeling of neural tissue during maturation of the brain circuitry, as has been proposed for ECM protease inhibitors (Monard, 1988).
Low basal expression of TIMP-1 mRNA in the normal brain contrasted with its rapid and dramatic induction in the dentate gyrus after KA treatment. The subsequent propagation to other hippocampal, cortical, and subcortical areas was closely related to the electrographic and metabolic activation of limbic structures after systemic KA injections (Ben-Ari et al., 1981). Interestingly, although TIMP-1 mRNA expression globally decreased at 2 or 3 d after KA treatment, it remained elevated in areas where degeneration occurred, as confirmed by Nissl staining and in agreement with the well established pattern of pathology produced by systemic KA administration (Schowb et al., 1980; Ben-Ari et al., 1981). These data indirectly suggest that TIMP-1 may be involved in processes of tissue repair and/or cell survival in the injured CNS.
The blockade of KA-mediated TIMP-1 mRNA induction in adult rats by the anticonvulsant drug diazepam indicates that the transcriptional regulation of TIMP-1 expression is activity-dependent and not the result of nonspecific effects of KA. Nonetheless, evidence that neuronal activity per se may not be sufficient to trigger TIMP-1 expression is provided by the absence of TIMP-1 induction in 10-d-old animals, despite intense seizure activity manifested by tonicoclonic seizures. Full limbic seizures and emerging signs of pathology in response to KA are not present in rats until the third postnatal week (Nitecka et al., 1984; Tremblay et al., 1984), at which time TIMP-1 mRNA first was seen abundantly induced in the hippocampus and areas of projection. By 35 d postnatal, when both seizures and pathology are indistinguishable from those of adult rats, TIMP-1 expression extended to all limbic and limbic-related structures. The close parallel between developmental changes in seizures and neuronal vulnerability on one hand and the increased TIMP-1 expression on the other further supports the idea that TIMP-1 is associated with seizure-induced neuropathology. The necessity of a mature limbic system and/or the appropriate signal transduction pathways at a given age may account for differences in the developmental regulation of TIMP-1 after seizures. Because similar developmental patterns of mRNA induction have been documented for c-fos (Schreiber et al., 1992) and brain-derived neurotrophic factor (BDNF) (Dugich-Djordjevic et al., 1992), it is possible that these genes and TIMP-1 share similar stimulus–transcription coupling mechanisms in the brain under intense neuronal activity.
The early TIMP-1 mRNA induction in neurons is independent ofde novo protein synthesis
The TIMP-1 gene also shares with c-fos (Morgan and Curran, 1991) and BDNF (Hughes et al., 1993; Lauterborn et al., 1996) the property of being transcribed as an IEG. Indeed, cycloheximide did not block the KA-mediated induction of TIMP-1 mRNA but did block the induction of Fos protein, indicating that protein synthesis was successfully inhibited, as previously demonstrated in vivo(Jonec and Wasterlain, 1979; Goto et al., 1990; Papas et al., 1992). This raises the possibility that activity-dependent TIMP-1 expression is under the control of consitutively expressed transcription factors that are post-translationally activated. One potential candidate is the protein STAT3 (Signal Transducer and Activator of Transcription 3) that translocates to the nucleus on phosphorylation (see Schindler and Darnell, 1995) and binds to the STAT binding site of the rat TIMP-1 promoter, a major regulatory element in TIMP-1 gene expression (Bugno et al., 1995). Although it is not known whether STAT3 is phosphorylated and activated after seizures, a recent report (Rajan et al., 1995) documents the activation of STAT proteins after axotomy in vivo, suggesting a role for this class of transcription factors in neuronal injury. As an inhibitor of protease activity, an early TIMP-1 expression agrees with the necessity for neurons to develop protective strategies shortly after stress stimuli. Interestingly, the TIMP-1 promoter region also contains an AP-1 binding site (Edwards et al., 1992; Bugno et al., 1995) and may bind AP-1 proteins like Fos or Jun, both induced after seizures (Morgan and Curran, 1991). This site could account for delayed expression of TIMP-1 in both neurons and glial cells, which, in contrast, may depend on novel protein synthesis. Previous in vitro work showing that cycloheximide abrogates serum-mediated TIMP-1 mRNA induction in cultured fibroblasts (Edwards et al., 1985) supports the contention of an alternative mechanism of transcriptional regulation for TIMP-1 in the CNS.
Seizures alter the neuronal distribution and levels of TIMP-1 immunoreactivity
Ubiquitous TIMP-1 immunoreactivity was detected in neurons across the entire control rat brain. This constitutive expression may exert constant control over the activity of MMPs under physiological conditions.
The upregulation of TIMP-1 mRNA after seizures was followed by generalized increases in TIMP-1 immunoreactivity in most brain structures studied. However, theses increases were delayed somewhat after the induction of the mRNA. Indeed, dramatic and rapid induction of TIMP-1 mRNA in the dentate granule cells (90 min after KA) translated into increased immunoreactivity at 8 hr post-KA. At later time points (72 hr post-KA) immunoreactivity was elevated in the granule cells, when mRNA levels had returned to control values. Similar discrepancies between mRNA and protein expression after seizures have been reported previously for other proteins such as BDNF (Wetmore et al., 1994; Nawa et al., 1995). In similar experimental paradigms, proteases of the ECM such as tPA (Qian et al., 1993; Tsirka et al., 1995) and MMP3 (Reeben et al., 1996) are induced. The activity of other MMPs also is induced after brain traumatic lesions (Rosenberg, 1995) and focal cerebral ischemia (Rosenberg et al., 1996). The increased expression of TIMP-1, a potent inhibitor of MMP activity, in different neural cells in the stressed brain may be part of a compensatory homeostatic cell response to preserve the integrity of the ECM, challenged by an increase in proteolytic activity. Accordingly, we observed in the hippocampus an increased TIMP-1 immunoreactivity at 8–16 hr post-KA in the perisomatic compartment and in dendritic areas of pyramidal and granule cells. This strongly suggests a translocation of the protein via secretory pathways, in agreement with the secreted nature of TIMP-1 (Murphy et al., 1977; Cawston et al., 1981).
At 72 hr after treatment, TIMP-1 immunoreactivity appears dramatically increased not only in dendro–axonic fields of the hippocampus but also within the granule cell bodies, suggesting differential regulation of TIMP-1 synthesis, stability, transport, and secretion across time. The delayed accumulation of TIMP-1 in these neurons may result from a decreased secretion at a time when astrocytes may contribute significantly to the extracellular TIMP-1 pool, as suggested by their increased immunoreactivity at 72 hr post-KA. At this time point, in the degenerating CA1 area, the sustained TIMP-1 mRNA expression was not correlated with increased immunoreactivity in neurons of this subfield. This apparent paradox may result from an altered capacity to proceed with translation of TIMP-1 or from an increased proteolysis of this protein. Alternatively, glial cells infiltrating the CA1 neuronal layer may contribute to the maintenance of elevated TIMP-1 mRNA levels. Our results demonstrate a good correlation between the absence of TIMP-1 protein in the vulnerable CA1 neurons and the presence of TIMP-1 in resistant granule cells. In brain excitotoxic seizures a local alteration of the MMP/TIMP ratio in favor of proteolytic activity could contribute to the pathological outcome, as suggested in other pathologies (Apodaca et al., 1990; Stetler-Stevenson, 1990; Rosenberg et al., 1992; Shingu et al., 1993; Giraudon et al., 1995; Mohanam et al., 1995). Further biochemical and anatomical studies addressing alterations of the MMP/TIMP ratio will be necessary to elucidate their implication in neuronal plasticity and neuropathology in the seizing brain.
Delayed increase of TIMP-1 in astrocytes of lesioned areas
The expression of TIMP-1, first in neurons and, as revealed by double-labeling experiments, subsequently also in astrocytes, suggests that TIMP-1 transcriptional regulation in the seizing brain may be cell- and stimulus-dependent. Although neuronal hyperactivity may trigger early TIMP-1 upregulation in neurons, other factors, set in motion by seizures, could influence its expression in both neurons and glial cells at later time points. Among these factors proinflammatory cytokines (i.e., IL-1 β, TNF-α) and bFGF, known to be induced by seizures (Minami et al., 1990; Bugra et al., 1994; Gall et al., 1994;de Bock et al., 1996) and supposedly involved in tissue repair after lesions, are efficient modulators of the expression of TIMP-1 in non-neural cells (Edwards et al., 1987; Chua and Chua, 1990; Okada et al., 1990; So et al., 1992; Shingu et al., 1993; Dayer and Burger, 1994). In addition to its most widely recognized function as inhibitor of MMPs, TIMP-1 also exhibits trophic activity (Docherty et al., 1985;Gasson et al., 1985). Experimental evidence indicates that trophic factors can modulate their expressions mutually in the CNS (Yoshida and Gage, 1991; Rivera et al., 1994; Ferhat et al., 1997). In keeping with this notion and considering the complex cellular and temporal expression of TIMP-1 after seizures, it is possible that TIMP-1 contributes to the cascade of trophic interactions among neural cells. It is noteworthy that TIMP-1 immunoreactivity was induced strongly at 72 hr and still elevated at 14 d in cells morphologically identified as astrocytes. These were observed mainly in vulnerable structures such as the hippocampus, the amygdala, or the entorhinal cortex. Astrocytes synthesize ECM components such as neural cell adhesion molecules (Le Gal La Salle et al., 1992; Niquet et al., 1993), fibronectin (Niquet et al., 1994), or tenascin-C (Niquet et al., 1995) in response to seizures and demonstrate increased MMP activity when stimulated with proinflammatory cytokines (Gottschall and Yu, 1995). Taken together, these data suggest that the contribution of astrocytes to brain tissue remodeling after seizures (Represa et al., 1995) may be, at least in part, based on their capacity to produce essential components of the ECM as well as to regulate its turnover via the modulation of MMPs/TIMPs ratio. The idea of TIMP-1 participating in tissue remodeling after seizures is reinforced by the fact that increased immunoreactivity still was observed in the dendritic areas and in the glial cells of the hippocampus as late as 2 weeks after KA treatment.
In summary, we have described early changes in TIMP-1 mRNA expression and protein localization, along with the differential modulation of these changes across time in neurons and astrocytes. This raises the interesting possibility that TIMP-1 plays a role in coupling early cellular events triggered by environmental stimuli with the intracellular and extracellular regulation of long-lasting changes. Additionally, our results support the contention that TIMP-1 influences tissue remodeling and/or neuronal survival after seizures.
This research was supported by funding from the Institut National de la Santé et de la Recherche Médicale. S.R. was granted a fellowship from the government of Catalonia and a Biomed I fellowship from the European Union. The c-fos antibody was a generous gift from Dr. R. Wehrle. We thank Elena Prats for technical assistance and Dr. M. Esclapez and J. Levav for critical reading of this manuscript. We dedicate this article to Dr. Yoav Citri, in memoriam.
Correspondence should be addressed to Dr. Santiago Rivera, Institut National de la Santé et de la Recherche Médicale U-29, Hôpital de Port Royal, 123 Boulevard Port Royal, 75014 Paris, France.