|Home | About | Journals | Submit | Contact Us | Français|
Mice with inactivation of the Tuberous sclerosis complex-1 (Tsc1) gene in glia (Tsc1GFAPCKO mice) have deficient astrocyte glutamate transporters and develop seizures, suggesting that abnormal glutamate homeostasis contributes to neurological abnormalities in these mice. We examined the hypothesis that Tsc1GFAPCKO mice have elevated extracellular brain glutamate levels that may cause neuronal death, abnormal glutamatergic synaptic function, and associated impairments in behavioral learning. In vivo microdialysis documented elevated glutamate levels in hippocampi of Tsc1GFAPCKO mice and several cell death assays demonstrated neuronal death in hippocampus and neocortex. Impairment of long-term potentiation (LTP) with tetanic stimulation was observed in hippocampal slices from Tsc1GFAPCKO mice and was reversed by low concentrations of NMDA antagonist, indicating that excessive synaptic glutamate directly inhibited LTP. Finally, Tsc1GFAPCKO mice exhibited deficits in two hippocampal-dependent learning paradigms. These results suggest that abnormal glutamate homeostasis predisposes to excitotoxic cell death, impaired synaptic plasticity and learning deficits in Tsc1GFAPCKO mice.
Tuberous sclerosis complex (TSC) is a relatively common multi-system genetic disease caused by mutation of either the TSC1 or TSC2 gene (Crino et al., 2006; Kwiatkowski, 2003; Sparagana and Roach 2000). Neurological symptoms are usually the most disabling clinical problems associated with TSC and typically include seizures, autism, mental retardation, and learning disabilities. Current treatments of the neurological manifestations of TSC are purely symptom-based and are largely ineffective. Thus, understanding the pathophysiology of TSC is critical for developing novel, more effective therapies for the neurological symptoms in TSC. The neuropathological hallmark of brains from TSC patients are areas of disrupted cortical lamination, termed tubers, which likely represent the epileptogenic foci for seizures and also contribute to the cognitive deficits in TSC (Doherty et al., 2005; Goodman et al., 1997). Cortical tubers demonstrate dramatic histopathological abnormalities in glia, suggesting that defective glial function may be centrally involved in the pathogenesis of TSC (Gutmann et al., 2000; Ess et al., 2004). In this regard, we have previously demonstrated that conditional Tsc1 gene inactivation primarily in glia of mice (Tsc1GFAPCKO mice) results in progressive epilepsy, encephalopathy, and premature death (Uhlmann et al., 2002). However, the specific mechanisms causing neuronal dysfunction in Tsc1GFAPCKO mice are unknown.
Glutamate, the major excitatory neurotransmitter in mammalian brain, may have detrimental effects on neurons when present in excessive amounts and has been implicated in the pathogenesis of a number of neurological disorders. Impairment of glutamate uptake by astrocytes can lead to excessive extracellular glutamate levels, resulting in abnormal synaptic function under some conditions (Mennerick and Zorumski, 1994; Tong and Jahr, 1994), impaired long-term potentiation (Katagari et al. 2001), and excitotoxic neuronal death (Rothstein et al., 1996; Tanaka et al., 1997), which may all cause cognitive dysfunction. Furthermore, inactivation of the astrocyte glutamate transporters, GLT-1 or GLAST, results in decreased seizure threshold or spontaneous seizures in mice (Tanaka et al., 1997; Watanabe et al., 1999). We have previously shown that astrocytes from Tsc1GFAPCKO mice exhibit decreased expression and function of GLT-1 and GLAST (Wong et al., 2003), suggesting the possibility that abnormal glutamate homeostasis could also contribute to neuronal dysfunction in the Tsc1GFAPCKO mice. Thus, in the present study, we examined the hypothesis that Tsc1GFAPCKO mice have elevated extracellular glutamate levels, which may contribute to excitotoxic neuronal death, abnormal glutamatergic synaptic physiology, and impaired behavioral conditioning and learning.
Tsc1flox/flox-GFAP-Cre (Tsc1GFAPCKO) knockout mice with conditional inactivation of the Tsc1 gene in glia were produced using Cre-Lox technology as described previously (Uhlmann et al., 2002). Tsc1flox/+-GFAP-Cre or Tsc1flox/flox littermates were used as controls in these experiments and have previously been found to have no abnormal phenotype (Uhlmann et al., 2002). Care and use of animals conformed to a protocol approved by the Washington University School of Medicine Animal Studies Committee.
Four week old mice were implanted with microdialysis probes as previously described (Cirrito et al., 2003; 2005). Briefly, under isoflurane anesthesia, guide cannulae (BR-style, Bioanalytical Systems, Indianapolis, IN) were inserted stereotaxically into the left hippocampus (bregma -3.1 mm, 2.5 mm lateral to midline, and 0.6 mm below dura at a 12° angle). After the guide cannulae were cemented, 2 mm microdialysis probes (BR-2, 38 kDa MWCO membrane, Bioanalytical Systems) were inserted through the guides into the hippocampus. Mice were allowed to recover from anesthesia and were housed in a Raturn Cage system (Bioanalytical Systems), which permitted freedom of movement and ad lib food and water for the remainder of the experiment. The microdialysis probe was connected to a Univentor syring pump (SciPro) and artificial cerebrospinal fluid (ACSF, in mM: 1.3 CaCl2, 1.2 MgSO4, 3 KCl, 0.4 KH2PO4, 25 NaHCO3, and 122 NaCl, pH 7.35) was perfused through the microdialysis probe. All dialysate samples were collected with a refrigerated fraction collector into polypropylene tubes for subsequent measurement of glutamate concentration, as described below. To ensure that brain extracellular fluid (ECF) glutamate levels reached a steady-state concentration after probe insertion, six 1-hr samples were taken at a constant flow rate during an initial equilibration phase prior to starting the protocol below. After a stable baseline was obtained, an extrapolated zero flow protocol was used to calculate the in vivo concentration of glutamate within the brain ECF (Menacherry et al 1992; Cirrito et al. 2003), by measuring glutamate concentrations from dialysate samples acquired at different flow rates and extrapolating back to zero flow rate, at which point the dialysate should reach equilibrium with and equal the in vivo ECF glutamate concentration. The extrapolation was based on a second order polynomial fit: y = a*x2 + b*x + E, where y = glutamate concentration, x = flow rate, and E = extrapolated in vivo ECF concentration at zero flow rate. To assess whether the microdialysis sampling technique and other biological factors were consistent between different conditions (e.g. control versus Tsc1GFAPCKO mice), the percentage recovery of glutamate at each flow rate was determined and compared by the following equation: (Cx/E) * 100, where Cx is the measured glutamate concentration at a given flow rate and E is the in vivo concentration calculated by extrapolation. At the end of microdialysis, animals were anesthetized with isobutel and transcardially perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde in PBS pH 7.4 to allow histological confirmation of the microdialysis probe location in the hippocampus.
Dialysate glutamate concentrations were measured using an Amplex red glutamic acid/ glutamate oxidase assay kit (Molecular Probes, Eugene, OR) on the same day microdialysis was performed. For each sample, a total volume of 100 μl per microplate well was obtained by mixing 50 μl of sample with 50 μl of working solution (100 μM Amplex Red, 0.25 U/mL horseradish peroxidase (HRP), 0.08 U/mL L-glutamate oxidase, 0.5 U/mL L-glutamate-pyruvate transaminase, and 200 μM L-alanine). Samples were then incubated at 37°C for 30 min and analyzed with a FL600 microplate reader (BioTek, Winooski, VT) with 530 nm excitation and 590 nm emission wavelengths. Glutamate concentrations of samples were determined by interpolation from a standard curve derived by measurements of other samples with known, pre-measured concentrations of glutamate. Each point was corrected for background fluorescence by subtracting values derived from glutamate-free control samples.
One and three month old control or Tsc1GFAPCKO mice were anesthetized with ketamine and perfused transcardially with 0.1 M phosphate-buffered saline (PBS), pH 7.4, followed by 4% paraformaldehyde in 0.1 M PBS. Brains were removed immediately after perfusion and placed in 4% paraformaldehyde for 30 min at 4°C and then stored in 0.1 M PBS containing 30% sucrose for 72 h at 4°C. Fifty μm coronal sections were cut with a vibratome and kept in cryoprotective solution (ethylene glycol and sucrose in 0.1 M PBS, pH 7.4) at −20°C until used for immunostaining or TUNEL studies.
The TUNEL assay was performed on brain sections by standard protocol using the ApopTag-Fluorescein Kit (Serologicals Corporation, Norcross, GA). Briefly, sections were washed with 0.1 M PBS and pretreated sequentially with 0.3% Triton X for 15 min at room temperature (RT), 20 μg/ml proteinase K for 20 min at 37°C, and 0.2% H2O2 for 20 min at RT, with intervening washes in 0.1 M PBS. Tissue was then placed in the equilibration buffer for 1 min at room temperature, followed by incubation in the reaction buffer containing terminal deoxynucleotidyltransferase enzyme (TdT) for 1 hr at 37°C. The reaction was then terminated by transferring sections to stop buffer for 10 min at RT and sections were washed in PBS three times for 5 min each. The sections were then incubated with FITC-conjugated anti-digoxigenin antibody for 30 min at RT, rinsed in PBS, and mounted in Vectashield with DAPI (Vector Laboratories, Burlingame, CA). Negative controls for each TUNEL experiment omitted the TdT enzyme from the reaction mixture. Positive controls for the TUNEL procedure were utilized from the kit.
Immunostaining to caspase-3 was performed by standard techniques using a rabbit polyclonal caspase -3 antibody (Cell Signaling Techologies, Beverly, MA) and Alexa Fluor 488 anti-rabbit secondary antibody (Molecular Probes, Eugene, OR). Labeling for Fluoro-Jade B (Chemicon International, Temecula, CA) was performed using previously-published methods (Schumed and Hopkins, 2000). To identify neurons positive for TUNEL, double-labeling was performed with a mouse monoclonal anti:NeuN antibody (Chemicon International, Temecula, CA). Neocortex and hippocampus of fixed coronal sections were examined for positively-stained neurons using confocal microscopy. For quantification, the number of positive neurons per section were counted in frontal sections (~1.5 – 2.5 mm posterior to bregma) containing neocortex and hippocampus by an observer blinded to the genotype of the sections.
Horizontal entorhinal cortex-hippocampal slices (400 μm) were prepared with a vibratome from two to four week old control or Tsc1GFAPCKO mice, using standard methods as described previously (Wong et al., 2003). Slices were placed in a submerged recording chamber and perfused continuously with an ACSF containing (in mM): NaCl 124, KCl 3, CaCl2 2.5, MgCl2 1.3, NaH2PO4 1.3, NaHCO3 22, and glucose 10. Patch pipettes were filled with (in mM) CsCl 130, NaCl 4, CaCl2 0.5, HEPES 10, EGTA 5, and MgCl2 2. QX-314 (10 mM) was also added to the pipette solution to block sodium currents. Pyramidal neurons in the CA1 region were directly visualized on an upright microscope with differential interference contrast/infrared optics. Whole cell voltage clamp recordings were obtained and analyzed using an Axopatch 200B amplifier, Digidata 1322 data acquisition system, and pCLAMP 8 software (Molecular Devices, Sunnyvale, CA). Series resistance compensation of 70-85% was used. Membrane capacitance was calculated from fitting of current transient responses to small hyperpolarizing voltage steps. Excitatory postsynaptic currents were activated at a holding potential of −60 mV by electrical stimulation of Schaffer collaterals in stratum radiatum with a tungsten concentric bipolar stimulating electrode (0.1 ms, 0.1 Hz). The amplitude of the stimulation pulses was adjusted to the lowest intensity necessary to induce a maximal response (0.5 – 2 mA). Five current traces per condition were averaged and analyzed with pCLAMP software. The decay phase of synaptic currents was fitted by a two-component exponential Chebyshev method. All single cell experiments were performed at both room temperature (~25° C) and at ~33° C. For studies of non-NMDA currents, MK-801 (10 μM) and bicuculline (25 μM) were added to the ACSF. For isolation of NMDA currents, magnesium was omitted and CNQX (20 μM), bicuculline (25 μM) and glycine (10 μM) were added to the ACSF. GABAergic inhibitory postsynaptic currents were isolated by addition of MK-801 (10 μM) and CNQX (20 μM) to the standard ACSF. All drugs were purchased from Sigma (St. Louis, MO), unless otherwise specified.
Other experiments examining extracellular field EPSPs and long-term potentiation were performed in coronal (400 μm) hippocampal slices from ~8 week old control and Tsc1GFAPCKO mice. Extracellular field potentials were recorded with glass microelectrodes filled with ACSF from the stratum radiatum of CA1 region. Signals were recorded by an Axoclamp 2B amplifier in bridge mode with additional amplification from a Warner Instrument DC amplifier and saved and analyzed by pCLAMP software. Stable, baseline field EPSPs of approximately one third of maximal amplitude were activated by stimulation of the Schaffer collateral pathway with a bipolar stimulating electrode at 0.016 Hz. To induce LTP, a series of 4 tetani of 100 Hz stimuli for 1 second were applied with each tetanus separated by 30 seconds.
During the early postweaning period (P24 ± 1 day), Tsc1GFAPCKO (n = 10) mice and littermate controls (n = 9) were evaluated on a battery of tests (walking initiation, inclined and inverted screens, platform, ledge, and pole) designed to assess balance, strength, and coordination as previously described (Wozniak et al., 2004). Two trials were administered and averaged for each test with a 2-h interval intervening between trials.
Locomotor activity was quantified using a computerized, photobeam system (MotorMonitor, Hamilton-Kinder, LLC, Poway, CA) according to previously published methods (Wozniak et al., 2004). The mice were tested in transparent polystyrene enclosures (47.6 × 25.4 × 20.6 cm high) just before beginning spatial (place) learning trials in the water maze (see below) on P30 (± 1 day). The dependent variables analyzed included the number of ambulations (whole body movements), rearings, distance traveled in the periphery, distance traveled in the central area of the “field”, and time spent and entries made into the center during a 1-h test period. Analysis of measures involving behaviors emitted in the center was used to evaluate possible alterations in emotionality beyond changes in activity.
Spatial learning and memory capabilities were assessed during the early juvenile period using the Morris water navigation test utilizing procedures similar to previously published methods (Wozniak et al., 2004). The protocol included conducting cued (visible platform), place (submerged and not visible platform), and probe (platform removed) trials in a round pool (118 cm inner diameter) of opaque water. Swim paths were tracked and recorded by a computerized system (Polytrack, San Diego Instruments, San Diego, CA), which calculated escape path length and latency. Swimming speeds were also calculated for cued and place trials. Mice were first trained on the cued condition (P28 ± 1 day) to determine if nonassociative factors (e.g., sensorimotor or visual disturbances or alterations in motivation) were likely to affect acquisition performance during subsequent place trials. Mice received 4 trials per day for two consecutive days of cued training, using an intertrial interval of approximately 20 min, and allowing the mouse to remain on the platform for 15 sec. The platform was moved to a different location for each trial within a day in the presence of very few distal cues thus substantially limiting spatial learning during this time. Three days after completing the cued trials, the mice were trained on the “place” condition to learn the location of a submerged (hidden) platform. The place trials were conducted in the presence of several salient distal cues positioned around the room to facilitate association of the spatial cues with the submerged platform location. During place training the mice were given 4 trials per day for 5 days (60 s maximum for a trial) with the platform remaining in the same location for all place trials. The daily protocol involved administering 2 blocks of 2 consecutive trials with each block being separated by approximately 1 h and a mouse being released from a different quadrant for each trial. A probe trial was administered approximately 1 h after completion of the place trials on the 5th day of place trials to evaluate retention of the platform location. During the 60-s probe trial, the escape platform was removed and a mouse was placed in the quadrant diagonally opposite from the previous platform location. Times spent in the pool quadrants, and the number of crossings made over the previous platform location (platform crossings), were recorded.
Two or three days after completing water maze testing, the mice were evaluated on the conditioned fear test (P39 ± 1 day) using a protocol similar to our previously published methods (Khuchua et al., 2003) except that in the present study, freezing behavior was scored by a computerized video-based image analysis system (see below). Briefly, the mice were trained and tested in two-Plexiglas conditioning chambers (26 cm ×, 18 cm, and 18 cm high) (Med-Associates, St. Albans, VT) with each chamber differing in terms of visual, tactile, and olfactory cues present. On day 1, each mouse was placed into the conditioning chamber for 5 min. Freezing behavior was quantified during a 2-min baseline period, after which an 80 dB tone (conditioned stimulus; CS) consisting of broadband white noise was presented for 20 sec. During the last second of the pulse, the mice received a 1.0 mA continuous foot shock (unconditioned stimulus; US). This CS-US (tone-shock) pairing was repeated each minute over the next two minutes, and freezing was quantified after each of the three tone-shock pairings. The mice were removed from the testing chamber 40 sec after the third shock and returned to their home cages. Twenty-four hours later, each mouse was placed in the same conditioning chamber in which it was trained to test contextual fear, which involved quantifying freezing behavior for 8 min. Twenty-four hours later, each mouse was placed into a different chamber (with different cues present) to be evaluated on the auditory cue test. The mice were observed for 10 min in this “altered context” and freezing behavior was quantified during a 2 min baseline period and over a subsequent 8 min period, during which time the auditory cue (tone; CS) was presented. Freezing was quantified using FreezeFrame image analysis software (Actimetrics, Evanston, Ill) which allowed for simultaneous visualization of behavior while adjusting a “freezing threshold” which categorized behavior as freezing or not freezing during 0.75 sec intervals. Freezing was defined as no movement except for that associated with normal respiration and the data was presented as percent of time spent freezing.
After completing the conditioned fear testing, the mice were evaluated for their shock sensitivity to determine if fear conditioning may have been affected by this variable. The levels of shock that were necessary to elicit flinching, running, vocalizing, or jumping were determined for each mouse using our previously published procedures (Khuchua et al., 2003). The shock sensitivity measures and all other behavioral tests were conducted by observers who were unaware of the genotypic status of individual mice. In addition, mice were carefully monitored for seizure-like behaviors, and those that exhibited such behaviors were not tested on any behavioral measure until they appeared to be fully recovered.
For behavioral experiments, analysis of variance (ANOVA) models with repeated measures (including Huynh-Feldt adjustment) were used. The model typically contained one between-subjects variable (genotype) and one within-subjects variable (e.g., blocks of trials) and included Bonferroni corrections for multiple comparisons. For other experiments, two-tailed t-tests were used for all comparisons of quantitative data between two groups (e.g. control and Tsc1GFAPCKO mice). One-way ANOVA was used in the hippocampal slice LTP experiments, comparing three groups (e.g control, Tsc1GFAPCKO, Tsc1GFAPCKO + APV). All numerical data are presented as mean ± standard error of the mean (SEM). Statistical significance was defined as p<0.05.
Our previous work has demonstrated a >75% decrease in GLT-1 and GLAST expression and a corresponding decrease in astrocyte glutamate transporter functional activity in the hippocampus of Tsc1GFAPCKO mice (Wong et al., 2003). Given this impairment in astrocyte glutamate transport in Tsc1GFAPCKO mice, we performed in vivo microdialysis to determine the ECF concentration of glutamate in the hippocampus in 4-week old Tsc1GFAPCKO and control mice. The 4-week age was chosen as a time that should precede seizure onset (Uhlmann et al., 2002; Erbayat-Altay, 2007). Within three hours of microdialysis probe insertion, glutamate levels achieved a steady-state in both control and Tsc1GFAPCKO mice (data not shown), and between four and six hours following probe insertion, the relative dialysate glutamate concentration was significantly higher in Tsc1GFAPCKO mice 131.9 ± 15.4 nM versus 84.2 ± 12.2 nM in controls; n=5 mice per group, p<0.05). To calculate the absolute ECF glutamate concentration in vivo, an extrapolated zero flow protocol was then utilized, in which glutamate concentrations were measured at different flow rates and extrapolation was performed to a theoretical zero flow rate (Fig. 1A), at which point the dialysate and ECF should reach equilibrium and equal the in vivo ECF glutamate concentration (Menacherry et al. 1992). Using this method individually in each mouse, the average ECF glutamate concentration was significantly elevated in Tsc1GFAPCKO mice (Fig. 1B; 902.1 ± 88.1 nM versus 607.3 ± 52.1 for controls; p<0.01, n=6 mice per group). By comparison, the percentage recovery of glutamate (determined by dividing the measured glutamate concentration at a given flow rate by the absolute extrapolated concentration) was not different between Tsc1GFAPCKO mice and wild type mice at each flow rate (Fig. 1C), indicating that the microdialysis technique and other potentially confounding biological factors were consistent between the two groups. Evidence of histological damage from the microdialysis technique was minimal (Fig. 1D; see also Cirrito et al., 2003).
Since increased extracellular glutamate may result in neuronal excitotoxicity, we examined neuronal death in control and Tsc1GFAPCKO mice at one and three months of age using multiple methods, including TUNEL, caspase-3, and Fluoro-Jade B labeling. The one and three month time points were chosen to represent times before and after onset of seizures, based on previous video-EEG studies (Uhlmann et al., 2002; Erbayat-Altay et al., 2007; see Discussion), thus controlling for the additional effects of seizures on neuronal death. TUNEL-positive immunoreactivity was detected in the neocortex, hippocampal CA1 and CA3 pyramidal cell layers, and dentate gyrus (granular and subgranular layers) in both one and three month Tsc1GFAPCKO mice (Figs. (Figs.2B,2B, ,3A,3A, ,4).4). By comparison, no TUNEL-positive cells were seen in control mice (Fig. 2A), with the exception of a few cells in the subgranular layer of dentate gyrus (DG), consistent with previous reports and corresponding to a region of ongoing neurogenesis (Biebl et al., 2000). Although there was a trend toward increased TUNEL-positive cells in the hippocampus (but not neocortex) of Tsc1GFAPCKO mice at three months compared to one month, this increase was not statistically significant (Fig. 3A). Fluoro-Jade B (Fig. 2D) and Caspase-3 (Figs. 2E, 2F, ,3B)3B) staining revealed similar spatial and temporal patterns of positive cells.
The pattern of TUNEL, capase-3, and Fluoro-Jade B labeling in hippocampus corresponded with clearly-identifiable neurons in the CA1 or CA3 pyramidal cell layers or dentate granule cell layer. To more precisely determine the identity of the affected cells in the neocortex, double-labeling with Neu-N, a neuron-specific antigen, was performed in conjunction with TUNEL staining. Approximately 75% of the TUNEL-positive cells in the neocortex were double-labeled with Neu-N (Fig. 4), indicating that neurons accounted for the majority of cell death in both neocortex and hippocampus.
To determine whether the impaired glutamate homeostasis affects synaptic physiology in Tsc1GFAPCKO mice, we assessed glutamatergic synaptic transmission in hippocampal slices. As we have previously documented that glutamate transporter function is impaired in astrocytes in the CA1 region of hippocampal slices from Tsc1GFAPCKO mice (Wong et al., 2003), we examined glutamatergic excitatory postsynaptic currents (EPSCs) at the Schaffer collateral-CA1 pyramidal neuron synapse in hippocampal slices in situ. There were no significant differences in non-NMDA or NMDA-mediated EPSCs in CA1 neurons activated at low frequency (0.1 Hz) from hippocampal slices obtained from 2-4 week old control and Tsc1GFAPCKO mice (Fig. 5, Table 1). In addition, there were no significant differences in the effect of the glutamate transporter antagonists, DHK or TBOA, on EPSCs in control versus Tsc1GFAPCKO hippocampal slices (data not shown). Since glutamate transporters exhibit significant temperature-sensitivity (Tong and Jahr, 1994), experiments were conducted at both room temperature and ~33°C with similar results. Lastly, as inhibitory interneurons may be dependent upon astrocyte glutamate uptake and the subsequent shunting of glutamine for GABA synthesis (Patel et al., 2001), we also evaluated GABAergic IPSCs in the CA1 pyramidal neurons. There were no significant differences in GABAergic IPSCs between control and Tsc1GFAPCKO hippocampal slices (Fig. 5, Table 1).
While there were no significant abnormalities in glutamatergic synaptic transmission in Tsc1GFAPCKO mice under conditions of low frequency synaptic stimulation, we next sought to determine the effect of impaired astrocyte glutamate transport on long-term potentiation due to high frequency synaptic stimulation. We considered this possibility based on the findings in Glt1 knockout mice, which also exhibit normal glutamatergic synaptic transmission during low frequency stimulation (Tanaka et al., 1997), but display impaired LTP due to excessive synaptic glutamate exposure (Katagiri et al., 2001). Stable, extracellular field EPSPs were activated in the stratum radiatum of CA1 region of hippocampal slices. Baseline field EPSPs activated by low frequency synaptic stimulation (0.016 Hz) appeared to have a lower amplitude in Tsc1GFAPCKO mice compared to control mice (Fig. 6A; 0.87 ± 0.40 mV measured at one-third maximal amplitude in Tsc1GFAPCKO mice versus 1.23 ± 0.40 in controls; n=6 mice per group), but this difference was not statistically significant. However, LTP induced by tetanic stimulation (100 Hz) was significantly decreased in Tsc1GFAPCKO mice compared to control mice (Fig. 6B,C; 105.6% ± 9.3 of baseline EPSP slope in Tsc1GFAPCKO mice versus 156.2% ± 10.3 in controls; n = 6 mice per group, p<0.05). Furthermore, similar to Glt-1 knockout mice (Katagiri et al., 2001), this impairment of LTP was partially reversed by application of a low concentration of D-APV (0.5 μM; to partially block NMDA receptors) during the tetanic stimulation in Tsc1GFAPCKO mice (Fig. 6B,C; 137.9% ± 11.8 of baseline in the presence of D-APV versus 105.6% ± 9.3 in the absence of D-APV; n = 6-8 mice per group, p<0.05), suggesting that the impairment of LTP in Tsc1GFAPCKO mice was due to excessive activation of NMDA receptors by glutamate.
Given the impaired LTP and cell death observed in the hippocampus of Tsc1GFAPCKO mice, we next tested the mice for performance on two behavioral learning tasks that rely on intact hippocampal function (Silva et al., 1998; Maren and Holt, 2002): a Pavlovian conditioned fear task and a spatial (place) learning/Morris water navigation task.
When first introduced into the conditioning chamber (Day 1), the Tsc1GFAPCKO mice tended to freeze more often than controls (Fig. 7A) although an ANOVA of the freezing behavior over the two-minute baseline period did not yield any significant overall effects involving genotype. In contrast, analysis of the freezing data following the three tone(T)/shock(S) pairings showed that the control mice froze more often than the Tsc1GFAPCKO mice, as documented by a significant main effect of genotype, [F(1,16) 5.33, p = 0.035]. Subsequent pairwise comparisons showed that the Tsc1GFAPCKO mice froze much less often than controls after the second (p = 0.024) and third T/S pairing (p = 0.007; less than Bonferroni corrected p = 0.017), suggesting that Tsc1GFAPCKO mice may be impaired in short-term context conditioning. When re-introduced into the conditioning chamber 24 h after T/S training (Day 2), the Tsc1GFAPCKO mice exhibited virtually no freezing behavior suggesting profound impairment in terms of contextual fear conditioning (Fig 7B). In contrast, the control mice froze often during the first half of the test period indicating strong contextual fear conditioning. An ANOVA of the freezing data revealed a significant main effect of genotype, [F(1,16) = 7.87, p = 0.013], and a significant Minute by Genotype interaction, [F(7,112) = 5.43, p = 0.001]. Another striking aspect about the contextual fear data was that the control mice showed a gradual, steady decline in freezing over time after peaking at 2 min, thus demonstrating habituation (min 2 vs min 8; p = 0.001), whereas the Tsc1GFAPCKO mice did not habituate but rather showed potentiation of freezing over most of the session (min 1 vs min 7 (peak); p = 0.025). Thus, control and Tsc1GFAPCKO mice showed virtually opposite behavioral responses to the contextual cues that were paired with the shock.
To evaluate conditioning to the auditory cues (tone), the mice were placed into a new chamber (Day 3) containing different sets of visual, tactual, and olfactory cues 24 h after the contextual fear test, and both groups exhibited similar amounts of freezing over the 2-min baseline period (Fig 7C). In contrast, the onset of the “tone” elicited much more freezing in control mice compared to the Tsc1GFAPCKO mice (Fig. 7C), as evidenced by a significant Genotype by Minute interaction, [F(7,98) = 9.10, p < 0.0005]. Subsequent pairwise comparisons showed that the control mice froze significantly more often than Tsc1GFAPCKO mice during minute 1 (p < 0.0005) and minute 2 (p = 0.009). Similar to the contextual fear data, the control mice showed gradual but significant habituation of freezing over time (Block 1 vs Block 8: p = 0.006), while the Tsc1GFAPCKO mice exhibited a significant potentiation of the freezing response across most of the test session (Block 1 vs Block 7, peak freezing: p = 0.044). The mice were also tested for their sensitivity to shock to determine if this may have contributed to the abnormal conditioning, but there were no differences between control and Tsc1GFAPCKO mice in terms of the levels of shock that elicited flinching, running, vocalizing or jumping (Fig. 7D). One Tsc1GFAPCKO mouse exhibited seizure-like behaviors during conditioned fear testing which appeared to produce long-term effects on its behavior and conditioning performance and therefore was eliminated from the dataset.
In the cued trials in the water maze, no differences were observed between Tsc1GFAPCKO and control mice in escape path length (Fig. 7E) or latency (not shown), indicating that Tsc1GFAPCKO mice had no deficits in nonassociative functions (e.g., visual, sensorimotor, or motivational disturbances) that would affect their performance on subsequent place trials. In contrast, the Tsc1GFAPCKO mice were significantly impaired during acquisition in the place condition (Fig. 7F) relative to the controls. Specifically, an ANOVA of the path length data revealed a significant main effect of genotype, [F(1,7) = 6.56, p = 0.020] showing that the controls had significantly shorter path lengths to the escape platform across the blocks of trials with differences between groups being greatest during the second (p = 0.004) and eighth blocks (p = 0.049) of trials. Also, control mice showed significant improvement between the first and last block of trials (p = 0.024) suggesting spatial learning had occurred while no such learning was observed in Tsc1GFAPCKO mice. Differences between groups were even larger when escape latency was analyzed (data not shown); however, the latency data were confounded by the finding that Tsc1GFAPCKO mice swam significantly slower than controls during the place trials (see below), thus making path length to the submerged platform the more appropriate dependent variable to assess acquisition performance during the place trials. An ANOVA of the swimming speed data resulted in a significant main effect of genotype, [F(1,17) = 7.64, p = 0.013), indicating that the control mice swam significantly faster that the Tsc1GFAP CKO mice across the blocks of trials (Fig. 7G), with differences being greatest during the sixth and tenth blocks (p < 0.009). With regard to the probe trial data, one-way ANOVAs conducted within each group on the times spent in the pool quadrants showed significant differences in quadrant times for controls, [F(3,24) = 8.91, p = 0.006] but not Tsc1GFAPCKO mice. Subsequent pairwise comparisons showed that control mice, but not Tsc1GFAPCKO mice, exhibited spatial bias in spending more time in the quadrant that had contained the platform (target quadrant) compared to the left (p = 0.044), opposite (p = 0.022), and right (p < 0.0005] quadrants (Fig. 7H), again indicating an impairment in spatial learning in the Tsc1GFAPCKO mice.
To provide important control data for aiding interpretation of the conditioning and learning results, as well as to characterize the behavioral phenotype of Tsc1GFAPCKO mice in more detail, mice were evaluated on a battery of sensorimotor tests and on a 1-h locomotor activity measure. No significant differences were observed between the two groups of mice on the walking initiation, ledge, platform, 60° inclined screen, 90° inclined screen, or inverted screen tests from the sensorimotor battery when the mice were assessed during the early postweaning period (P25 ± 1; data not shown). However, Tsc1GFAPCKO mice (Fig. 7I) were significantly impaired on the pole test, which was originally developed to study bradykinesia associated with striatal dopamine depletion in mice. Tsc1GFAPCKO mice took significantly longer to climb down the pole compared to controls (t17 = 3.69, p = 0.002), a finding that is consistent with the swimming speed data from the water maze testing. With regard to the activity data, no significant differences between groups were observed on any of the activity or emotionality variables when the mice were evaluated just before undergoing place training in the water maze (P32 ± 1; data not shown). In summary, except for slow movement during a difficult task that requires complex coordination between the forelimbs and hindlimbs, the Tsc1GFAPCKO mice were not impaired in a wide variety of motor/sensorimotor functions. These results and the lack of differences in the path length and latency data during the cued trials in the water maze suggest that the impaired acquisition performance of the Tsc1GFAPCKO mice during the place trials and the contextual fear studies were not secondary to nonassociative sources, but represent primary learning and conditioning deficits in these mice.
The mechanisms underlying neuronal dysfunction in TSC are poorly understood. We have previously shown that Tsc1GFAPCKO mice have impaired astrocyte glutamate transport, suggesting a potential role of disturbed glutamate homeostasis in TSC pathophysiology. In the present study, we report that extracellular glutamate is elevated in Tsc1GFAPCKO mice in vivo. Furthermore, we show that Tsc1GFAPCKO mice exhibit increased excitotoxic neuronal death and impaired synaptic plasticity, as well as associated behavioral deficits in hippocampal-dependent learning. Remarkably, the impairment in hippocampal synaptic plasticity is partially reversed by a glutamate antagonist at low dose, indicating that in addition to cell death, excessive synaptic glutamate directly causes this defect in synaptic function. These results are novel in indicating that abnormal glutamate homeostasis may contribute to mechanisms of neuronal dysfunction and cognitive deficits in TSC.
Given the dramatic reduction in astrocyte glutamate transport in Tsc1GFAPCKO mice (Wong et al., 2003), it is not surprising that we observed elevated extracellular glutamate levels in the Tsc1GFAPCKO mice. Previous studies indicate that astroglia glutamate transporters are responsible for the majority of glutamate transport and inactivation of these transporters leads to increased extracellular glutamate levels (Tanaka et al. 1997; Rothstein et al. 1996). While other confounding factors, such as seizures, might also contribute to elevated extracellular glutamate levels in Tsc1GFAPCKO mice, the age used in the microdialysis studies likely precedes the age of onset of seizures in these mice (see below). Thus, we suggest that elevated glutamate levels most likely constitute a primary pathogenic factor promoting neuronal death and impaired synaptic function in Tsc1GFAPCKO mice.
Significant neuronal death was observed in Tsc1GFAPCKO mice in this study, potentially resulting from the elevated extracellular glutamate. Previous studies have shown that astrocytes protect neurons from glutamate-mediated excitotoxic cell death (Rosenberg and Aisenman, 1989; Ye and Sontheimer, 1998) and inactivation of glutamate transporters induces neuronal death in vivo (Rothstein et al., 1996; Tanaka et al., 1997). While impaired astrocyte glutamate transport represents a rational mechanism for the observed neuronal injury in Tsc1GFAPCKO mice, potential confounding effects of other factors, in particular seizures, on cell death again need to be considered, as seizures by themselves can cause neuronal death. To control for the effect of seizures, we assayed cell death at two time points, corresponding to times before and after the usual onset of seizures, based on previous video-EEG studies (Uhlmann et al., 2002; Erbayat-Altay et al., 2007). Tsc1GFAPCKO mice typically exhibit onset of epilepsy between one and two months of age with infrequent seizures. By two to three months of age, seizures progressively worsen in frequency and severity, culminating in death by three to four months of age. In the present study, we observed a trend towards increased neuronal death in the hippocampus of Tsc1GFAPCKO mice at three months of age compared to one month, suggesting a possible additional effect of seizures on neuronal death. However, the presence of neuronal death in both neocortex and hippocampus at one month of age and the lack of definite progression in cell death with age, especially in neocortex, indicates that neuronal death is likely an independent event preceding seizures in Tsc1GFAPCKO mice and may be related to elevated extracellular glutamate levels from impaired astrocyte glutamate transport. Thus, neuronal death triggered by excessive glutamate may contribute to the development of epileptogenesis and other neurological deficits in these mice. Of clinical relevance, markers of cell death have also been found in cortical tubers from patients with TSC (Maldonado et al., 2003).
Impaired glutamate transport and elevated extracellular glutamate levels could also have direct effects on synaptic function. However, glutamatergic synaptic transmission was normal in hippocampal slices from Tsc1GFAPCKO mice under conditions of low frequency stimulation. Although some studies have reported a modest effect of glutamate transporter antagonists on excitatory postsynaptic currents (Mennerick and Zorumski, 1994; Tong and Jahr, 1994), others have found no effect (Isaacson and Nicoll, 1993; Sarantis et al., 1993; Tanaka et al., 1997), indicating that glutamate transport has minimal contribution to rapid, post-synaptic glutamatergic responses during low-frequency synaptic transmission. In contrast, LTP induced by tetanic stimuli was impaired in Tsc1GFAPCKO mice, suggesting that abnormal glutamate transport and elevated extracellular glutamate are more likely to impact synaptic function during maximal glutamate release, such as with high-frequency tetanic synaptic signalling. The reduction in LTP could be related to the neuronal death seen in hippocampus in Tsc1GFAPCKO mice, as well as other circuit abnormalities. Interestingly, however, this impairment in LTP was reversed by the presence of low concentrations of a NMDA antagonist, suggesting that excessive synaptic glutamate, at least in part, can directly account for abnormal synaptic plasticity. Analogously, Glt-1 knockout mice also exhibit normal low frequency glutamate synaptic transmission but impaired LTP that is reversed with low concentrations of APV (Katagari et al., 2001; Tanaka et al., 1997). Other studies suggest that excessive NMDA receptor activation may also impair LTP in normal rodent hippocampus (Bashir and Collingridge, 1992; Coan et al., 1989). Collectively, our findings suggest that elevated extracellular glutamate in Tsc1GFAPCKO mice may cause pathologically excessive activation of NMDA receptors during high frequency synaptic stimulation, which may result in impaired LTP induction. Consistent with a recent report of abnormal synaptic plasticity in the Eker rat model of TSC (von der Brelie et al., 2006), impaired LTP in Tsc1GFAPCKO mice could provide a mechanistic explanation for cognitive deficits commonly seen in TSC.
In support of this idea, our behavioral studies demonstrate significant deficits in two hippocampal-dependent conditioning/learning paradigms. First, in a Pavlovian conditioned fear task, Tsc1GFAPCKO mice showed virtually no evidence of contextual fear conditioning, despite an intact sensitivity to shock compared to controls. Moreover, the responses of Tsc1GFAPCKO mice during tone-shock training suggest that short-term memory deficits pertaining to aversive stimuli may have also contributed to their contextual fear conditioning impairments, a possibility that should be evaluated in future studies. Second, the water maze data suggest that compromised cognitive functions of Tsc1GFAPCKO mice extend to spatial learning and memory, although the magnitude of this impairment was not as profound. Differences between the control and Tsc1GFAPCKO mice in spatial learning may have been even greater with an extended testing period, since controls would have likely continued to improve with additional testing relative to the Tsc1GFAPCKO mice who showed no evidence of learning by the fifth block of trials. Unfortunately, our design was limited by our need to obtain other important behavioral data (e.g conditioned fear, sensorimotor battery) within a constrained time frame, to minimize possible confounding effects of seizures.
Although other factors could have confounded interpretation of the behavioral results, the bulk of the present data indicate that Tsc1GFAPCKO mice have primary cognitive deficits in hippocampal-dependent tasks. The lack of significant performance differences between groups on all activity measures, on 6 of 7 sensorimotor tests, in shock sensitivity, and during the cued water maze trials indicates that the impaired conditioning and learning performance of Tsc1GFAPCKO mice was not secondary to compromised nonassociative functions. The one deficit found in Tsc1GFAPCKO mice relating to complex co-ordination functions likely did not have a significant impact on conditioning/learning performance, given the normal performance of Tsc1GFAPCKO mice on all activity measures and during the cued trials. As discussed for the other experiments, the presence of seizures could also affect certain behavioral indices. However, the effect of seizures was likely minimal, since most of the behavioral testing occurred at an age that was likely prior to seizure onset (see above) and we were careful to monitor for seizure activity and minimize direct confounding effects on behavior. Future behavioral studies are warranted to determine if there are other significant functional impairments in these mice.
We appreciate that additional studies are needed to more definitively prove a causal link between impaired astrocyte-regulated glutamate homeostasis and the associated neurological phenotype in Tsc1GFAPCKO mice. Although multiple lines of evidence indicate that astrocytes play a key role in the pathophysiology of Tsc1GFAPCKO mice, it is possible that Tsc1 inactivation in GFAP-positive neuroglial progenitor cells during early cortical development could lead to primary neuronal defects, which could also contribute to abnormal glutamatergic synaptic physiology and impaired learning in these mice. Future experiments in our laboratory will test the effect of pharmacological or genetic manipulations that upregulate astrocyte glutamate transporters on reversing the glutamate-related abnormalities observed in the present study (Ganel et al., 2006; Rothstein et al., 2005), thus potentially providing mechanistic proof for the importance of astrocytic regulation of glutamate homeostasis in this model. Of clinical relevance, the findings of impaired glutamate homeostasis and associated abnormalities in synaptic plasticity and neuronal death in Tsc1GFAPCKO mice represent rational mechanisms that may contribute to cognitive deficits and epileptogenesis in TSC and suggest the possibility of novel therapeutic approaches for TSC directed specifically at astrocyte glutamate transport, or glutamate homeostasis in general.
We thank Nicholas Rensing and Sara Conyers for technical support. This work was supported by grants from the National Institutes of Health (K02NS045583 and R01NS056872, MW; DA07261, JRC; Neuroscience Blueprint Core Grant NS057105), Tuberous Sclerosis Alliance (MW), and U.S. Army Medical Research and Materiel Command grant (DAMD17-03-1-0073, DHG).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.