Abstract
Protease nexin-1 (PN-1), a member of the serpin superfamily, controls the activity of extracellular serine proteases and is expressed in the brain. Mutant mice overexpressing PN-1 in brain under the control of the Thy-1 promoter (Thy 1/PN-1) or lacking PN-1 (PN-1−/−) were found to develop epileptic activity in vivo and in vitro. Theta burst-induced long-term potentiation (LTP) and NMDA receptor-mediated synaptic transmission in the CA1 field of hippocampal slices were augmented in Thy 1/PN-1 mice and reduced in PN-1−/− mice. Compensatory changes in GABA-mediated inhibition in Thy 1/PN-1 mice suggest that altered brain PN-1 levels lead to an imbalance between excitatory and inhibitory synaptic transmission.
- protease nexin-1
- knock-out and transgenic mice
- long-term potentiation
- protease modulation
- epileptiform activity
- synaptical activity
The activity of extracellular serine proteases in the brain is controlled by specific inhibitors called serpins. Unlike the coagulation/fibrinolytic pathway where several different inhibitors are known, in the CNS only protease nexin-1 (PN-1; Guenther et al., 1985; Sommer et al., 1987) is present at significant levels (Mansuy et al., 1993; Sappino et al., 1993; Reinhard et al., 1994). PN-1 is expressed in both glial and neuronal cells of the developing and adult CNS (Mansuy et al., 1993; Sappino et al., 1993; Reinhard et al., 1994) and in glomeruli of the olfactory bulbs, where synapse formation and elimination remain an active process throughout life (Reinhard et al., 1988; Mansuy et al., 1993). High PN-1 expression also is detected after injury of the peripheral nervous system (Meier et al., 1989) or the CNS (Hoffmann et al., 1992; Scotti et al., 1994).
PN-1 is a 43 kDa protein that, when bound to the cell surface or the extracellular matrix (ECM), acts potently as a thrombin inhibitor (Stone et al., 1987; Wagner et al., 1989). PN-1 also can block plasmin, tissue plasminogen activator (tPA), urokinase plasminogen activator (uPA), and trypsin (Baker et al., 1980; Guenther et al., 1985; Stone et al., 1987; Wagner et al., 1989), suggesting that in the brain PN-1 could modulate the activity of several serine proteases known to process or induce physiologically active macromolecules that control neurite extension (Krystosek and Seeds, 1981; Gurwitz and Cunningham, 1988; Jalink and Moolenaar, 1992; Suidan et al., 1992), neuronal cell viability (Smith-Swintosky et al., 1995; Tsirka et al., 1995; Vaughan et al., 1995), neuronal cell excitability (Yamada and Bilkey, 1993;Tsirka et al., 1995), and synaptic plasticity (Mansuy et al., 1993;Qian et al., 1993; Liu et al., 1994; Meiri et al., 1994; Romanic and Madri, 1994; Seeds et al., 1995; Frey et al., 1996; Huang et al., 1996).
To assess the physiological role of the equilibrium between PN-1 and its target protease(s), we disturbed the balance by generating transgenic mice that overexpress PN-1 in the CNS under the control of the Thy 1 promoter (Thy 1/PN-1) and transgenic mice that lack the PN-1 gene (PN-1−/−). The Thy 1/PN-1 mutants should provide a model to study the effect of increased inhibition of serine protease(s), and the PN-1−/− mutants should provide a model for diminished serine protease inhibition. Because PN-1 can inhibit tPA, plasmin, and trypsin, previous findings led us to investigate the incidence of epileptic events in the mutant mice: (1) tPA−/− mice are less susceptible to kainic acid (KA)-induced seizures (Tsirka et al., 1995); (2) tPA is one of the genes upregulated on seizure induction (Qian et al., 1993), and it could be involved in the control of distinct forms of late long-term potentiation (L-LTP), depending on the particular tetanization paradigm (Frey et al., 1996; Huang et al., 1996); (3) plasminogen enhances the NMDA-induced increase in intracellular Ca2+ concentration (Inoue et al., 1994); and (4) trypsin induces epileptiform activity in hippocampal slices (Yamada and Bilkey, 1993). In this study we have used Thy 1/PN-1 and PN-1−/− mice to evaluate the role of PN-1 in epileptiform activity and in hippocampal synaptic transmission and plasticity.
MATERIALS AND METHODS
Generation and analysis of Thy 1/PN-1 mice. An 8.2 kbEcoRI genomic DNA fragment (Evans et al., 1984) encompassing the murine Thy 1.2 gene was used to construct the Thy 1 expression cassette (a gift from Drs. Glen A. Evans and Shizhong Chen, Salk Institute, San Diego, CA). This cassette was generated originally by deleting a DNA fragment (from the BanI site in exon 2, upstream of the translation start codon, to a XhoI site in exon 4) and inserting an XhoI linker. A bluntedHindIII–EcoRI DNA fragment containing the rat PN-1 cDNA (Sommer et al., 1987) was inserted into the bluntedXhoI site of the Thy 1 cassette. The PN-1 cDNA encompasses the authentic PN-1 translation initiation (ATG) codon, signal peptide, and termination codon (TGA). Before DNA microinjection, the fusion genes were excised and freed from plasmid sequences by agarose gel electrophoresis and purified further with an Elutip-d column (Schleicher & Schuell, Dassel, Germany). Transgenic mice were generated and analyzed as described before (Botteri et al., 1987). Male mice in each generation were used for breeding and maintenance of the Thy 1/PN-1 lines (neither transgene is on the X chromosome). Both female and male heterozygous mice were used for analysis after the inheritance of the transgene by Southern blot analysis had been verified (Chen et al., 1987). In situ hybridization and Northern and immunoblot analyses were performed as described (Mansuy et al., 1993). Northern blots were hybridized to random-primed 32P-labeled DNA probes. The following probes were used: a 750 bpXhoI–BamHI DNA fragment from exon 4 of the mouse Thy 1.2 gene (Ingraham et al., 1986) and a 1351 bpXhoI–XbaI DNA fragment carrying the rat PN-1 cDNA (Gloor et al., 1986; Sommer et al., 1987). For the thrombin inhibition assay, a previously described method (Nick et al., 1990) was adapted to microtiter plates. Calibration was performed by using rat recombinant PN-1 (Sommer et al., 1989).
Generation and analysis of PN-1−/− mice. Electroporation and PCR analysis of E14 embryonic stem (ES) cell clones resistant to G418 and gancyclovir were completed as previously described (Stief et al., 1994). ES cells carrying the disrupted PN-1 allele were injected into C57BL/6 recipient blastocysts (Stief et al., 1994) or cocultured with denuded post-compacted eight-cell-stage mouse embryos (Wood et al., 1993). Eight-cell embryos from [(C57BL/6 × Balb/c) F1 females × C57BL/6 males] at a post-compaction stage were placed in M2 medium (Hogan et al., 1986). Batches of 20 embryos were incubated briefly in acidified Tyrode’s solution (Hogan et al., 1986) until dissolution of their zona pellucida. Meanwhile, ES cells were trypsinized to obtain a single-cell suspension. After preplating was done to remove most of the fibroblast feeder cells, the ES cells were resuspended at a concentration of 106 cell/ml in coculture medium (Wood et al., 1993). Ten zona-deprived embryos were placed in 50 μl droplets of the ES cell suspension and incubated at 37°C for 2–3 hr to allow random aggregation of ES cells with post-compaction embryos. Embryos were allowed to recover and develop overnight in M16 medium (Hogan et al., 1986); finally, they were transferred into pseudo-pregnant foster females. Male chimeras were mated with C57BL/6 females. The genotype of the mice was confirmed by Southern blot analysis.
Electrophysiology. Transverse hippocampal slices (400 μm) from 6- to 12-week-old mutant mice and wild-type mice from the same litter were prepared by standard methods and maintained in an interface chamber. In the experiments that used extracellular field recordings, the slices were perfused at 35°C with a medium containing (in mm): NaCl 124.0, KCl 2.5, MgSO4 2.0, CaCl2 2.5, KH2PO4 1.25, NaHCO3 26.0, glucose 10, and sucrose 4, bubbled with 95% O2/5% CO2, pH 7.4. The CA1 stratum radiatum usually was stimulated by a bipolar platinum–iridium electrode (100 μm in diameter; 0.05 Hz, 100 μsec pulse duration), and field EPSPs were recorded from the CA1 stratum radiatum or the stratum pyramidale by means of glass microelectrodes (2 m NaCl, 1–5 MΩ). In the experiments investigating late LTP monopolar, laquer-coated stainless steel electrodes were used for stimulation and recording. Biphasic pulses (0.1 msec per polarity) were applied for testing. In the experiments that used the whole-cell voltage-clamp technique (Blanton et al., 1990) (HEKA EPC-9 patch-clamp amplifier), the slices were perfused at 28°C with the same extracellular medium, but the MgSO4 concentration was 1.3 mm, the KCl concentration was 1.25 mm, and 50 μmpicrotoxin was added to the medium throughout the experiment. The CA3 region routinely was removed to prevent bursting. Patch electrodes (3–8 MΩ) were filled with a solution (280–290 mOsm) containing (in mm): Cs-methanesulphonate 130, NaCl 8, HEPES 10, EGTA 5, Mg-ATP 2, Na-GTP 0.2, and QX-314 5, pH-adjusted to 7.25 with CsOH. Input and series resistance were monitored constantly, and cells showing >10% change during the experiment were discarded. In the experiments that used the conventional intracellular recording technique, the slices were perfused at 30°C with the same medium as in the extracellular experiments, but the KCl concentration was 1.25 mm. Recordings were obtained in the discontinuous voltage-clamp mode (Axoclamp-2A amplifier) with sharp electrodes (60–90 MΩ) filled with 4 m potassium acetate; clamp efficiency was 80–85% in both experimental groups; stimulation intensity was adjusted to elicit maximal inhibitory currents. Statistical evaluations were performed by ANOVA and Student’st test. All experiments and analyses were performed without knowledge of the respective genotypes of the animals. Results in the text or in the figures are expressed as mean ± SEM.
RESULTS
Generation and analysis of PN-1-overexpressing mice
Elevated neuronal expression levels of rat PN-1 specifically in the CNS of transgenic mice were achieved by mouse Thy 1.2 regulatory sequences (Chen et al., 1987) to control expression of the rat PN-1 cDNA (Fig. 1A). From a total of four lines expressing the transgene, two were chosen for further studies. Line T1 expressed moderate and line T2 expressed high levels of PN-1. Individuals from each line and seven successive generations showed a stable pattern of brain-specific transgene mRNA expression (Fig.1B). In these animals the tissue-specific expression pattern of the transgene was similar, but transgene mRNA levels were higher in brain of line T2 than T1 (Fig. 1B). Mice from both lines also showed low levels of transgene mRNA expression in lung (lane Lu, Fig. 1B), but not in thymus, because of the lack of the thymocyte enhancer (Gordon et al., 1987; Vidal et al., 1990).
In both lines onset of transgene expression occurred around birth and by postnatal day 14 reached maximum levels that were retained throughout adult life (Fig. 1C). This pattern mimics the expression of the endogenous Thy 1 gene and offers the advantage of reducing the potential for triggering of compensating gene expression patterns during development. In agreement with this hypothesis, we observed no increases in the expression of tPA, uPA, or thrombin in Thy 1/PN-1 mice.
Immunoblot analysis revealed that transgene expression led to a substantial increase in CNS PN-1 protein levels overall and in defined regions, including cerebellum, cortex, and hippocampus (Fig.1D). Identification of transgene-derived rat PN-1 protein was facilitated by a slower migration of the endogenous mouse PN-1 protein, as compared with the 43 kDa rat PN-1 protein (Mansuy et al., 1993). Like CNS transgene mRNA levels, PN-1 protein levels were approximately twofold higher in line T2, as compared with line T1. Functional activity of transgene-derived PN-1 protein was assessed by testing brain homogenates for serine protease inhibitory activity (Fig.1E). Compared with nontransgenic littermates, extracts of transgenic mouse brains consistently contained two- to fourfold more thrombin inhibitory activity. Moreover, the relative activities measured in various brain regions correlated with the amounts of PN-1 protein detected by immunoblot analysis (Fig.1D) and with PN-1 immunoreactivity (Fig.1G).
Endogenous PN-1 expression occurs in both neurons and glial cells (Mansuy et al., 1993; Reinhard et al., 1994). The predominantly neuronal expression of the transgene was confirmed by in situ hybridization with a rat PN-1 cRNA probe. High levels of transgene mRNA were detected in characteristic neuronal cell layers of the transgenic mouse brain (line T2). For reference, endogenous PN-1 mRNA expression was reflected by weak and diffuse hybridization signals in nontransgenic littermate brains (Fig. 1F; see alsoMansuy et al., 1993). Transgene mRNA was particularly prominent in several layers of the neocortex, in the olfactory bulb, in the stratum pyramidale of all CA subfields in the hippocampus, in the dentate gyrus granule cells, in the cerebellar nuclei, and also in many neurons scattered throughout various brain regions, including pons, medulla, and midbrain (Fig. 1F). Using staining conditions that revealed little mouse endogenous PN-1 immunostaining in sections of control littermates (Mansuy et al., 1993; Reinhard et al., 1994), we detected a major increase in rat PN-1 protein levels in the hippocampus (Fig. 1G) and all other transgene mRNA-expressing cell populations (results not shown).
Epileptic potential of Thy 1/PN-1 mice
In wild-type mice intraperitoneal injection of the convulsive glutamate agonist kainic acid (KA; 30 mg/kg) was followed by clonic spasms of the forelegs in three of seven animals, with onset times ranging between 43 and 46 min. In contrast, in littermate Thy 1/PN-1 mice (line T2) clonic spasms of the forelegs were observed in six of seven animals and full clonic spasms and chronic seizures in four of seven animals. Onset times ranged between 14 and 52 min (exitus, 3/7). These experiments suggest a clear increased susceptibility for kainic acid in the transgenic mice. Despite the fact that we used age-matched transgenic and nontransgenic littermates of similar weight (25 gm), it is difficult to obtain statistically valid data by using the intraperitoneal injection of kainic acid. Effects caused by individual differences in kainic acid metabolism cannot be excluded. We therefore assessed whether Thy1/PN-1 mice showed increased susceptibility after stereotactic intrastriatal injection of another excitotoxin, ibotenic acid (0.3 μl/10 min; 1% solution in PBS). This procedure has usually negligible epileptic effects, and, as expected, no seizures were observed in the nontransgenic mice (n= 6). In contrast, all Thy 1/PN-1 mice (n = 6) suffered from seizures. Histological evaluation of control and transgenic brains revealed no significant differences in injection site position and/or local tissue damage or appearance. Altogether, these observations suggest an increased susceptibility of Thy 1/PN-1 mice to glutamatergic excitotoxins.
We also tested two convulsive agents that interfere mainly with GABAergic transmission. Intraperitoneal injection of isoniazid (250 mg/kg), the convulsant action of which is ascribed to a reduction of brain GABA levels via inhibition of glutamic acid decarboxylase (GAD), revealed a significant delay in seizure onset in the Thy 1/PN-1 mice (line T2, n = 14; median onset time, 56 min; Student’st test, p < 0.01) as compared with littermate control mice (n = 14; median onset time, 37 min). In contrast, pentylenetetrazole (PTZ) was equally efficient in inducing seizures in transgenic and control mice when it was used at a threshold dose of 50 mg/kg (n = 6 for each group) or at a suprathreshold dose of 60 mg/kg (n = 6 for each group), whereas no seizures were observed in either group with a subthreshold dose of 40 mg/kg (n = 15 for each group). In summary, augmented levels of PN-1 seem to increase seizure susceptibility evoked by glutamatergic excitotoxins, and the prolonged latency phase for isoniazid-induced convulsions would be compatible with an increase in GABAergic inhibition.
As an in vitro correlate of epileptic activity, we analyzed the potential of hippocampal slices to develop burst discharges, which are characterized by the appearance of multiple population spikes (polyspikes) in response to repetitive stimulation. These burst discharges are based on the activity-dependent disinhibition of excitatory synaptic transmission (Thompson and Gähwiler, 1989) and on the activation of NMDA receptors (Dingledine et al., 1986;Masukawa et al., 1991). Repetitive stimulation (1 Hz for 30 sec) of the Schaffer collateral/commissural fibers provoked an increased polyspiking of the CA1 pyramidal neurons in Thy 1/PN-1 mice, as compared with littermate controls (n = 8 animals/8 slices; ANOVA, p < 0.01; Fig.2A,B). No bursting activity resembling interictal spikes was recorded, either spontaneously or at the normal test frequency of 0.05 Hz.
In both in vivo and in vitro experiments mice expressing high levels of PN-1 thus developed epileptic activity on overactivation of the glutamatergic system.
LTP in Thy 1/PN-1 mice
Basal synaptic transmission measured by the ratio between the presynaptic fiber volley and the corresponding initial slope of the field excitatory postsynaptic potential (fEPSP) was identical in Thy 1/PN-1 mice and littermate wild-type controls (0.24 ± 0.03 vs 0.24 ± 0.02 msec at maximal fEPSP amplitude; n = 6 animals/12 slices). Paired-pulse facilitation, a short-term synaptic plasticity relying on presynaptic mechanisms (Hess et al., 1987), was unchanged in Thy 1/PN-1 mice as well. These results suggest that overexpression of PN-1 does not interfere with presynaptic release processes.
Induction of LTP at excitatory synapses facilitates the development of a seizure-prone state (for review, see McNamara, 1994). We have analyzed two different forms of LTP, theta burst stimulation (TBS)-induced LTP and induction of late LTP (L-LTP), using repeated strong tetanization. Recent observations reveal that both forms involve different cellular mechanisms (Frey et al., 1996; Huang et al., 1996). In the first set of experiments investigating TBS-induced stimulation, stimulation strength was adjusted to evoke a fEPSP of 30% of its maximum amplitude, which was identical in the two experimental groups (0.24 ± 0.02 vs 0.23 ± 0.02 mV/msec; n = 6 animals/12 slices). After stable baseline recording of fEPSPs for 20 min, LTP was induced by a TBS paradigm (Larson and Lynch, 1986). LTP was increased significantly in Thy 1/PN-1 mice when compared with their wild-type littermates (199 ± 12 vs 166 ± 7%, 50 min after TBS; n = 6 animals/12 slices; ANOVA,p < 0.05; Fig. 3A,B1). This increase in LTP occurred without modifications in post-tetanic potentiation (Fig. 3A), a result that is in agreement with the unchanged paired-pulse facilitation. The postsynaptic responses to TBS showed a significant increase in Thy 1/PN-1 mice, as compared with littermate controls (n = 6 animals/12 slices; ANOVA,p < 0.05 for area measurements; Fig. 3B2). This indicates that the enhancement of LTP is based on an increase in the postsynaptic responsiveness to the LTP-inducing stimulus. The present results were confirmed by monitoring LTP in a second transgenic line (line T1) overexpressing PN-1 and in a transgenic line overexpressing a different protein, the secreted mouse IL-1 receptor antagonist (Zahedi et al., 1991; Sauer et al., 1996), under control of the same Thy 1 expression cassette. Although LTP was increased to a similar extent in the second PN-1-overexpressing line T1 (n = 4 animals/8 slices; p < 0.05; results not shown), no differences from control values were observed in the transgenic mice overexpressing the IL-1 receptor antagonist (n = 4 animals/8 slices; p = 0.24; results not shown). Thus, the observed increase in LTP was not attributable to the use of the Thy 1 expression cassette per se, but it was attributable to the overexpression of PN-1, leading to a selective increase in TBS-induced LTP.
Because tPA, one of the potential target proteases of PN-1, has been implicated directly in the long-term maintenance of LTP (Qian et al., 1993; Frey et al., 1996), a long-lasting form of LTP (L-LTP) also was recorded with an induction protocol specific for L-LTP (Frey et al., 1996). In contrast to TBS-induced LTP, L-LTP in Thy 1/PN-1 and in littermate control mice was similar (150 ± 21 vs 138 ± 18%, 8 hr after induction; n = 5 and 10 animals), suggesting no interference of PN-1 overexpression with the mechanisms of L-LTP maintenance.
Excitatory synaptic transmission in Thy 1/PN-1 mice
To study the mechanisms underlying the increase in polyspiking and in LTP, we first analyzed excitatory synaptic transmission by performing whole-cell patch-clamp recordings of evoked EPSCs in CA1 pyramidal cells in hippocampal slices. As illustrated in Figure 3,C and D, the voltage dependence and the reversal potential of both 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX)-sensitive non-NMDA receptor-mediated EPSCs (measured as the initial slope of dual-component EPSCs) and NMDA receptor-mediated EPSCs (in the presence of CNQX, 50 μm; Tocris Cookson, Bristol, UK) were not different in Thy 1/PN-1 mice from their littermate control mice. The ratio between the NMDA and the non-NMDA receptor-mediated EPSCs measured at a holding potential of −40 mV was unchanged as well (Fig.3E). In contrast, as illustrated in Figure 3F, the decay time course of NMDA receptor-mediated EPSCs recorded at −40 mV was prolonged significantly in Thy 1/PN-1 mice, as compared with littermate control mice (time required to decay to 50% of the peak amplitude: 41.0 ± 3.0 vs 50.6 ± 3.1 msec; n= 7 cells/6 animals; Student’s t test, p < 0.05). There was no change in the rise time of the NMDA currents. The prolonged decay time course was selective for the NMDA component, because the time course of the non-NMDA receptor-mediated component recorded at −90 mV was unchanged. Paired-pulse facilitation of NMDA receptor-mediated EPSCs recorded at −60 mV was similar in both experimental groups (results not shown), confirming the results we had obtained in extracellular recordings. In conclusion, the prolonged NMDA receptor-mediated EPSCs concur with the enhancement of TBS-induced LTP and the increased epileptic potential in Thy 1/PN-1 mice.
Inhibitory synaptic transmission in Thy 1/PN-1 mice
Because Thy 1/PN-1 mice exhibited an increased onset latency to seizures induced by the GAD inhibitor isoniazid and because changes in GABA-mediated inhibition are known to influence seizure activity and/or LTP (Wigström and Gustafsson, 1983; Dingledine et al., 1986), we analyzed monosynaptic fast IPSCs in CA1 pyramidal cells, using conventional intracellular recording techniques. Recordings in the current-clamp mode revealed no difference in the resting membrane potential (−64.9 ± 2.2/−66.7 ± 2.6 mV), the input resistance (55 ± 4/50 ± 2 MΩ), the passive membrane time constant (2.0 ± 0.3/2.4 ± 0.2 msec), or the shape/amplitude of the action potential between Thy 1/PN-1 and littermate control mice. Discontinuous single-electrode voltage clamp was used to investigate GABAA receptor-mediated fast IPSCs in the presence of CNQX (50 μm) and d-2-amino-5-phosphonopentanoic acid (d-AP5, 25 μm; Tocris Cookson) to block excitatory synaptic transmission; slow IPSCs were blocked by 50 mm QX-314 (RBI, Natick, MA) in the recording electrode. Analysis of frequency-dependent paired-pulse depression (PPD) of monosynaptic fast IPSCs, under conditions that account for GABA acting on presynaptic GABAB receptors (Davies et al., 1990), revealed a small but significant reduction in paired-pulse depression (n = 7 cells from 6 animals; ANOVA, p< 0.05; Fig. 4A). The reduction of PPD was not accompanied by changes in the voltage dependence, in the reversal potential, or in the maximal amplitude of the fast IPSC (Fig.4B). There was also no change in the rise time of the IPSC and the membrane conductance during the IPSC (results not shown). However, the decay time course of the fast IPSC was significantly shorter in Thy 1/PN-1 than in control mice (time required to decay to 50% of the peak amplitude, 28.6 ± 3.7 vs 40.6 ± 4.0 msec;n = 7 cells from 6 animals; Student’s ttest, p < 0.05; Fig. 4C).
Generation and analysis of PN-1-deficient mice
So that the role of PN-1 in modulating neuronal excitability could be confirmed, mutant mice lacking a functional PN-1 gene were generated (Fig. 5). A targeting vector for disruption of the murine PN-1 gene in mouse ES cells was constructed from two contiguous 129Sv/J genomic EcoRI DNA fragments spanning a total of 10.8 kb and harboring exons 2 and 3 of the PN-1 gene (Fig. 5A). Two of ten properly targeted ES cell clones were used to generate chimeric mice. Chimeric mice derived from either ES cell clone transmitted the mutant allele to male as well as female offspring (PN-1+/−), which subsequently were used to generate homozygous PN-1−/− mice. Successful disruption of the PN-1 gene was confirmed by Northern blot (Fig. 5B) and immunoblot analyses (Fig.5C). Analysis of total brain RNA revealed an approximately twofold reduction in PN-1 mRNA levels in PN-1+/− mice, as compared with wild-type, and a complete lack of PN-1 transcripts in PN-1−/− mice (Fig. 5B). As expected, on Western blots no PN-1 protein was detected in tissues from PN-1−/− mice (Fig.5C). PN-1−/− mice were viable and showed no obvious gross abnormalities in health and behavior when compared with PN-1+/− and wild-type littermates.
Epileptic potential, synaptic transmission, and LTP in PN-1−/− mice
Similar to the Thy 1/PN-1 mice, PN-1−/− mice showed an increased susceptibility to KA-induced seizures, as compared with genetically and age-matched control littermates. Intraperitoneal KA injection (30 mg/kg) was followed by full clonic spasms and seizures in 6 of 10 PN-1−/− mice, with onset times ranging between 16 and 72 min (4/10, only forelegs involved). In contrast to the PN-1−/− mice, none of the littermate control animals exhibited full clonic spasms but showed only clonic spasms of the forelegs, with onset times ranging from 30 to 59 min (n = 10). In vitro, hippocampal slices from PN-1−/− mice exhibited an increased polyspiking activity in the CA1 field on 1 Hz of stimulation, as compared with littermate controls (n = 5 animals/10 slices; ANOVA, p < 0.05; Fig. 6A,B).
Basal hippocampal synaptic transmission measured by the ratio between the presynaptic fiber volley and the corresponding initial slope of the fEPSP at Schaffer collateral/commissural→CA1 pyramidal cell synapses was similar in PN-1−/− and littermate control mice (0.22 ± 0.03 vs 0.23 ± 0.04 msec at maximal fEPSP amplitude; n = 9 animals/13 slices). The absolute values of the fEPSP initial slope at the stimulation strength used to induce TBS–LTP (30% of maximal fEPSP amplitude) were identical (0.20 ± 0.02 vs 0.19 ± 0.02 mV/msec;n = 9 animals/13 slices). Under these conditions LTP was found to be reduced in PN-1−/− mice, as compared with littermate controls (157 ± 8 vs 199 ± 11%, 50 min after TBS;n = 9 animals/13 slices; ANOVA, p < 0.05; Fig. 7A,B1). The postsynaptic responses to TBS were diminished, as compared with control mice (n = 9 animals/13 slices; ANOVA, p < 0.05 for area measurements; Fig. 7B2), suggesting that the reduction in LTP was a consequence of a less efficient TBS. Similar to the Thy 1/PN-1 mice, no significant difference in the post-tetanic potentiation was measured. In contrast to TBS-induced LTP, the maintenance of L-LTP induced via stronger repeated tetanization was not influenced in PN-1−/− mice when compared with littermate controls (168 ± 11 vs 161 ± 11%; n = 7 animals/14 slices). Recently it was shown (Frey et al., 1996) that a different form of L-LTP can be observed in tPA-deficient mice, which simulated a “normal glutamatergic potentiation” by reduction of GABAergic inhibition. To test whether L-LTP in PN-1−/− mice is carried by similar mechanisms, we continuously applied the GABAA-receptor inhibitor picrotoxin (10 μm) before LTP induction. No differences were observed between mutant and wild-type animals (123 ± 8%, n = 4 from 4 animals vs 124 ± 8%, n = 6 from 6 animals, 2 hr after induction and during continuous application of picrotoxin).
Whole-cell recordings of evoked EPSCs from CA1 pyramidal cells in slices from PN-1−/− mice revealed no difference in the voltage dependence or the reversal potential of non-NMDA receptor-mediated (Fig. 7C) and NMDA receptor-mediated EPSCs (Fig.7D). The decay time course of the NMDA receptor-mediated EPSCs recorded at −40 mV was not changed in PN-1−/− mice, as compared with littermate controls (Fig. 7F). However, in contrast to the Thy 1/PN-1 mice, the ratio between the NMDA and the non-NMDA receptor-mediated EPSC was reduced markedly in PN-1−/− mice (15.1 ± 1.9 vs 24.3 ± 0.9%; n = 7 cells/7 and 6 animals; Student’s t test, p < 0.01; Fig. 7E). These results are in agreement with a diminished postsynaptic response to the TBS and the consequent reduction in LTP, but they seem to contradict the increased epileptic potential observed in PN-1−/− mice.
DISCUSSION
Lowered threshold for epileptic activity and enhancement of theta burst-induced LTP in PN-1-overexpressing mice
In the hippocampus, epileptiform activity is based on extensive excitatory interactions among pyramidal cells (Miles and Wong, 1987;Meier and Dudek, 1993) and on reduction of synaptic inhibition, which promotes excitatory interactions and ultimately leads to excessive NMDA receptor activation (Dingledine et al., 1986; Merlin and Wong, 1993). In the CA1 field, synaptic activation of the NMDA receptors plays a primary role in the induction of epileptiform activity (Herron et al., 1985; Dingledine et al., 1986; Anderson et al., 1987) as well as in the induction of LTP (Harris et al., 1984) (for review, see Bliss and Collingridge, 1993). The observed enhancement of postsynaptic responses to TBS in Thy 1/PN-1 mice suggested an increased activation of the NMDA receptor system. In agreement with this, we found that Thy 1/PN-1 mice exhibited a selectively prolonged decay time course of the NMDA receptor-mediated EPSCs. Interestingly, this decay time course is regulated developmentally, being slower in younger animals, which are also more prone to epilepsy (McDonald and Johnston, 1990; Carmignoto and Vicini, 1992; Hestrin, 1992). Moreover, the loss of susceptibility to LTP observed during the early development of the sensory cortex is accompanied by a shortening of the time course of NMDA receptor-mediated currents and by a decrease in the ratio between NMDA and non-NMDA receptor-mediated currents (Crair and Malenka, 1995). The decay kinetics of the NMDA receptor-mediated EPSC can vary according to the subunit composition of the NMDA receptor complex (Monyer et al., 1992). The prolonged time course of the EPSCs mediated by native NMDA receptors in neonates correlates with a higher expression of the NR2B subunit of the NMDA receptor during early development (Williams et al., 1993; Sheng et al., 1994). The subunit composition of endogenous NMDA receptors in Thy 1/PN-1 mice remains to be determined.
The enhancement of NMDA receptor-mediated excitation in CA1 pyramidal cells was accompanied by modifications in GABAergic inhibition. Thy 1/PN-1 mice exhibited a reduced sensitivity to isoniazid-induced seizures, suggesting increased levels of GABA. In addition, the fast IPSC exhibited a shorter decay time course, as compared with littermate wild-type mice. Because the uptake of GABA is shaping the decay of the fast IPSC (Dingledine and Korn, 1985), this suggests that chronically increased GABA levels might be compensated by a more rapid GABA uptake. Thy 1/PN-1 mice also exhibited a reduction in the GABABreceptor-mediated autoinhibition of GABAergic synaptic transmission. A similar reduction of the presynaptic autoinhibition at GABAergic nerve terminals leading to a frequency-dependent increase in inhibition was observed in the dentate gyrus of kindled rats, a model for the temporal lobe epilepsy (Buhl et al., 1996). Our results suggest that in Thy 1/PN-1 mice the increased neuronal excitability also has led to compensatory changes in GABAergic inhibition.
In contrast to the reduced activation of GABABreceptors on inhibitory nerve terminals, a reduction in the activation of GABAB receptors on excitatory nerve terminals could result in an increased excitability during repetitive stimulation (Isaacson et al., 1993). However, the finding that the paired-pulse facilitation of NMDA receptor-mediated EPSCs was similar in slices of Thy 1/PN-1 mice, as compared with control animals (results not shown), argues against a significant change in GABAergic transmission at the heterosynaptic level, unlike in lethargic mice, a genetic model for absence seizures (Hosford et al., 1992).
Eventually a reduced autoinhibition at GABAergic nerve terminals paradoxically may lead to a frequency-dependent hyperexcitation. During TBS the large inhibitory chloride currents may reverse into excitatory outward currents, a phenomenon attributed to compartmentalized shifts in ECl within the neurons (Alger and Nicoll, 1982; Thompson and Gähwiler, 1989). Moreover, a shortened time course of the GABA concentration in the extracellular space also should result in less spillover of GABA to the neighboring synaptic terminals (Isaacson et al., 1993) and eventually lead to a net reduction of GABAergic inhibition of the CA1 pyramidal neurons. A more detailed analysis of GABAergic inhibition, however, is needed to assess fully its contribution to the observed effects in Thy 1/PN-1 mice.
Lowered threshold for epileptic activity and diminution of theta burst-induced LTP in PN-1-deficient mice
In PN-1−/− mice the susceptibility to KA-induced seizures and the in vitro epileptiform activity was increased, but in contrast to Thy 1/PN-1 mice TBS-induced LTP was reduced and accompanied by a diminished postsynaptic response to TBS. In agreement with this, we found that the ratio between the NMDA receptor-mediated and the non-NMDA receptor-mediated components of the synaptic EPSC was reduced in PN-1−/− mice. This may reflect a reduced number of functionally active synaptic NMDA receptors or a change in the subunit composition (Sakimura et al., 1995). The enhancement and the reduction in LTP in PN-1-overexpressing and deficient mice, respectively, are in line with the observed changes in NMDA receptor-mediated synaptic transmission. In contrast to TBS-induced LTP, no changes were found of L-LTP induced via a strong tetanization. This particular stimulation protocol might involve different and/or additional events, such as the activation of voltage-dependent calcium channels and the subsequent influx of the calcium required for L-LTP. More subtle changes at the NMDA receptor could be covered by these processes and therefore are indistinguishable in these experiments. However, it is clear that PN-1−/− mice, although exhibiting a reduced NMDA component, were more susceptible to KA-induced seizures and showed an increased in vitroepileptiform activity, as compared with littermate controls. This might be attributable to a reduced excitation of the inhibitory interneurons or to nonsynaptic mechanisms, such as modifications in the extracellular space (Hochman et al., 1995), in the extracellular matrix, or in the fine wiring of the neuronal network propagating seizures (McNamara, 1994). Whatever the mechanisms, a reduction of LTP does not necessarily imply an overall reduced excitability; e.g., mice lacking the α-subunit of calcium/calmodulin kinase II exhibit a severe limbic epilepsy despite a deficient LTP (Silva et al., 1992;Butler et al., 1995), and mice lacking the prion protein also showed decreased LTP and increased epileptiform activity (Collinge et al., 1994).
Molecular mechanisms of PN-1-induced changes in neuronal excitability
PN-1 can inhibit several proteases, such as tPA and thrombin, which play a critical role in extracellular matrix remodeling (Romanic and Madri, 1994). The interaction of cells with the extracellular matrix not only regulates cell shape, motility, differentiation, and gene expression via integrin-mediated signal transduction (Schwartz et al., 1995) but also modifies transmitter release (Chen and Grinnell, 1995), hippocampal LTP (Staubli et al., 1990; Xiao et al., 1991), and kindling (Grooms and Jones, 1995) (for review, see Jones, 1996).
In the hippocampus, tPA is upregulated after seizures, kindling, or LTP (Qian et al., 1993). However, tPA-catalyzed proteolysis has been detected throughout most regions of the hippocampus except in the CA1 field because of the presence of an inhibitory activity (Sappino et al., 1993). mRNA analysis of three putative tPA inhibitors (PAI-1, PAI-2, and PN-1) indicated that PN-1 is the only potential candidate (Sappino et al., 1993). The increase in GABAergic inhibition in tPA−/− mice (Frey et al., 1996) concurs with our finding that inhibition also is enhanced in Thy 1/PN-1 mice. In addition, the decreased sensitivity of tPA−/− mice to KA-induced seizures (Tsirka et al., 1995) fits with the increased susceptibility of PN-1−/− mice but provides at the same time a counter-argument for the action of PN-1 on tPA in Thy 1/PN-1 mice. Moreover, neither an excess nor an absence of PN-1 significantly altered tPA-mediated proteolysis, as detected by zymographic overlay assays performed on cryostat hippocampal sections derived from both PN-1−/− and Thy 1/PN-1 mice (results not shown).
Alternatively, PN-1 might act on a thrombin-like protease and interfere with the activation of the thrombin receptor (Dihanich et al., 1991;Niclou et al., 1994), which has been shown to promote the secretion of thrombospondin-1 (TSP1), an extracellular matrix component also upregulated after KA-induced seizures (Chamak et al., 1994). TSP1 is a ligand for the α3/β1 integrin (DeFreitas et al., 1995) and can inhibit plasmin and, like PN-1, uPA (Hogg, 1994). NMDA receptor-mediated neuronal excitability has been shown to be enhanced by plasmin (Inoue et al., 1994), and uPA-overexpressing mice exhibit learning deficits (Meiri et al., 1994). Thus, complex protease-dependent extracellular interactions relevant to neuronal plasticity may be influenced by the level of PN-1 expression.
In conclusion, we have shown that synaptic plasticity, as measured by theta burst-induced LTP, correlates with the level of PN-1 in the brain of mutant mice and that the respective change in this form of LTP can be explained by opposite modifications of NMDA receptor-mediated synaptic transmission with no change in basal non-NMDA receptor-mediated synaptic transmission. The lowered threshold for epileptiform activity in both PN-1-overexpressing and deficient mice indicates that the equilibrium between brain serine proteases and their inhibitors, like PN-1, contributes to the balance between excitation and inhibition during sustained neuronal activity.
Footnotes
We thank Dr. A. Pavlik for dissecting brain regions; Dr. E. Reinhard for initial advice on the PN-1 immunoblot analysis; Dr. J. Moll for cloning and inserting the IL-1Ra cDNA into the Thy 1 expression cassette; Dr. A. Stief for initial advice and screening of transfected ES cells; Dr. M. Pozza for the gift of CGP57250; Drs. P. Caroni, J. Hagmann, D. Hartman, H.-R. Olpe, J. Nicholls, and A. Pavlik for critical reading of this manuscript; Mrs. M.-O. Schellinger for excellent care of the mice; Dr. S. Takeda for the wild-type neomycin gene; Dr. H. Blüthmann for the transgenic line carrying the neomycin gene; and C. Lapize and S. A. Leuenberger for valuable technical help.
Correspondence should be addressed to Dr. Denis Monard, Friedrich Miescher Institut, P.O. Box 2543, CH-4002 Basel, Switzerland.
Dr. Lüthi’s present address: Department of Anatomy, School of Medical Sciences, University of Bristol, University Walk, Bristol BS8 1TD, UK.
Dr. Mansuy’s present address: College of Physicians and Surgeons, Columbia University, 722 West 168th Street, New York, NY 10032.