Tau pathology induces loss of GABAergic interneurons leading to altered synaptic plasticity and behavioral impairments

Tau solubility was analyzed using a modified protocol from [31‐33]. Briefly frozen cortex sections were homogenized without thawing in 5 X vol/wg of RIPA buffer (50 mM Tris–HCl, pH 7.4; 1% Nonidet P-40; 0.25% Na-deoxycholate; 150 mM NaCl; 1 mM EDTA; 1 mM PMSF; 1 mM Na₃VO₄; 1 mM NaF; Complete protease inhibitor cocktail, Roche, IN, USA), with a mechanical homogenizer (TH; Omni International, USA), and centrifuged at 20,000 X g for 20 min at 4°C. An aliquot of the supernatant representing the total tau fraction was kept for protein quantification and western blot analysis. The rest of the supernatant was adjusted to 1% sarkosyl (N-lauroylsarcosine), incubated for 30 min at room temperature with constant rotating, and centrifuged at 100,000g for one hr at 20°C. After high speed centrifugation, the pellet was washed with 1% sarkosyl and centrifuged again at 100,000g for one h at 20°C. The post wash pellet containing sarkosyl-insoluble, aggregated tau was resuspended and analyzed by SDS-PAGE. Tau in the sarkosyl pellet has been shown by immuno-electron microscopy to be filamentous [32], and it is synonymous with that identified by immunohistochemistry in NFTs. All fractions were diluted in O+ buffer (62.5 mM Tris–HCl, pH 6.8; 10% glycerol; 5% 2-mercaptoethanol; 2.3% SDS; 1 mM EGTA; 1 mM EDTA; 1 mM PMSF; 1 mM Na₃VO₄; 1 mM NaF; Complete protease inhibitor cocktail, Roche), a modified O buffer, boiled for 5 min, and kept at −20°C. Depending on the antibody used, 10 to 20 μg of protein were analyzed by western blotting. Equal amount of protein (BCA assay, Promega) was loaded and the samples were electrophoresed on 10-12% SDS-PAGE gels and transferred to nitrocellulose membranes. All blots were blocked (5% nonfat milk and 0.1% Tween 20 in TBS) and then incubated with various primary antibodies overnight. Subsequently, the blots were washed and incubated for 2 h at room temperature with peroxidase-conjugated, goat anti-rabbit (Thermo Scientific) or anti-mouse IgG (1:2000; Jackson ImmunoResearch). Immunoreactive bands were visualized and analyzed by enhanced chemiluminescent reagent (Pierce ECL, Thermo Scientific) using a Fujifilm LAS4000 imaging system and the Multi Gauge software (Fujifilm Life Science). To compare the relative amount of tau protein, the densities of the immuno-reactive bands corresponding to phospho-tau were normalized and reported relative to the amounts of total tau protein or α-tubulin.

Transverse hippocampal slices (400 μm) were prepared from one hemisphere of age-matched mice (12–18 months or 6–7 months of age) using a vibratome (from part of the cohort, the other hemisphere was used to perform immunostaining or in situ hybridization, see below). Slices were maintained in oxygenated ACSF containing the following (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 25 D-glucose, 2 CaCl₂, and 1 MgCl₂ at room temperature. For electrophysiology experiments, slices were transferred to interface recording chambers (preheated to 30°C) perfused with oxygenated ACSF. Extracellular field EPSPs (fEPSPs) were evoked by stimulation of Schaffer collateral pathway afferents and were measured by recording in stratum radiatum of area CA1. In order to determine the response range for each hippocampal slice, the stimulus range was divided into 10 arbitrary units. The slices were stimulated at each level, and the fEPSP was recorded. In each recording, the fiber volley amplitude was as a measure of the input stimulation to the fEPSP. The range of input values and their respective output values (measured as the fEPSP slope) were plotted as a mean to characterize basal synaptic transmission in transgenic and wild-type mice across all experiments. Baseline responses were calculated using the stimulation intensity that elicited 40-50% of the maximal fEPSP response as determined by the input–output relationship. Paired-pulse facilitation (PPF), an assay of normal presynaptic function, was induced with two stimuli of equal intensity (same as baseline intensity) presented in rapid succession at variable interpulse intervals, ranging from 10 ms to 300 ms. PPF was measured by examining the ratio of the fEPSP slope in response to stimulus 2 and that of stimulus 1. Before LTP-inducing high-frequency stimulation (HFS), stable baseline synaptic transmission was established for 20–30 min with a stimulus intensity of 40–50% of the maximum fEPSP. Stimulus intensity of the HFS was matched to the intensity used in the baseline recordings. LTP was induced by either one or four trains (2 min intertrain interval) of 100 Hz HFS for 1 s. Data were collected and presented as the average slope of the fEPSP from six individual traces collected over 2 min and then normalized to baseline recordings of fEPSPs. Hippocampal slices from Tg and WT mice were prepared simultaneously and placed in a chamber outfitted with dual-recording equipment, thereby minimizing day-to-day variability in slice preparations and recordings. For zolpidem treatment, slices were incubated 20 min prior to HFS with a subthreshold concentration of zolpidem (1 μM). Student’s t-test, Repeated Measures ANOVA or N-way ANOVA (where appropriate) were used for electrophysiological data analysis with p < 0.05 as significance criteria.

Brains (one hemisphere from mice also used for electrophysiology) were fixed overnight in buffered 2% paraformaldehyde (PFA) at 4°C. Next day, hemispheres were stored in 20% glycerol, 2% DMSO in phosphate buffer (PB) until the brains were sliced as consecutive serial coronal sections (5 series, 50 μm thickness). Brain sections were stored in cryoprotectant (300 ml ethylene glycol, 550 ml PB, 300 g Sucrose, volume to 1000 ml with H₂O; pH7.2) until used. Serial brain sections were used to perform in situ hybridizations as previously described in [34]. Briefly, RNA probes were prepared using dioxygenin (DIG) RNA labeling kits (Roche). Sections were postfixed in 4% PFA for 10 min followed by a wash in phosphate-buffered saline (PBS). Sections were treated with 1.5% H₂O₂ in methanol and rinsed in PBS. Then, sections were treated with 0.2 M HCl and washed in PBS. Proteinase K (Roche) digestion (20 μg/mL in PBS) was carried out followed by a wash in PBS, and the sections were refixed for 5 min in 4% PFA and washed with PBS. The sections were acetylated for 10 min (2.2 g triethanolamine hydrochloride (Sigma), 540 μL of 10 N NaOH (Fisher Scientific), 300 μL of acetic anhydride (Sigma) in 60 mL water) and washed in PBS. RNA probes, prepared at a dilution of 2 μL/mL in hybridization solution (50% formamide, 10% dextran sulfate, 1% 100× Denhart's, 250 μg/mL yeast tRNA, 0.3 M NaCl, 20 mM Tris–HCl, pH8, 5 mM EDTA, 10 mM NaPO4, 1% sarkosyl), were incubated at 80°C for 2 min. Thereafter, 500 μL of the probe mix was applied to the brain sections and incubated at 55°C overnight. The next day, sections were subjected to high stringency wash in pre-warmed 50% formamide, 2× SSC at 65°C. Next, the sections were rinsed in RNase buffer (0.5 M NaCl, 10 mM Tris–HCl, pH 7.5, 5 mM EDTA), followed by an RNaseA (Roche) treatment (20 μg/mL in RNase buffer) for 30 min and followed by a wash in RNase buffer, all at 37°C. The high stringency washes were repeated twice for 20 min each at 65°C, followed by a 15 min rinse in 2× SSC, then 0.1× SSC, both at 37°C. Sections were then washed in Wash Buffer (WB, 100 mM maleic acid, 150 mM NaCl, 0.5% Tween-20) and blocked with Blocking Buffer (1% Boehringer Manheim in WB) followed by incubation with anti-DIG-POD antibody (Roche) overnight at 4°C. Next day, slices were washed in WB and then incubated with Blocking Buffer 2 (0.5% Casein, 150 mM NaCl, 100 mM Tris, pH 7.5). Next biotinyl-tyramide was added to the sections, followed by washes in WB. Then sections were incubated with Streptavidin – AP for 1h at room temperature followed by 3 washes in WB and a wash in NTMT buffer (100 mM NaCl, 100 mM Tris–HCl, pH 9.5, 50 mM MgCl₂, 0.1% Tween-20). The sections were then placed in a light-protected environment with approximately 400 μL of BM-purple AP substrate (Roche) until satisfactory staining was achieved. Finally, the sections were rinsed twice in PBS, coverslipped using Crystal mount aqueous mounting media (Sigma) and images were acquired using a Leica DM 5000B light microscope. Using Image J software (National Institutes of Health), regions of interests (ROI) were defined in all consecutive slices of one mouse. The size of ROI varied to some extent from anterior to posterior based on each individual slice. General criteria for the CA1 ROI were determined by boundaries starting at the CA2-CA1 border till the subiculum-CA1 boundary, including the stratum oriens and stratum radiatum. Within these ROIs, the number of positive cells was counted. The number of positive cells was corrected to the volume of the ROIs. All cell counts were made by observers blind to genotype.

For fluorescent immunostaining to count number of SST or PV positive cells from mice used for electrophysiology, consecutive serial brain sections (50 μm thick) of freshly fixed brain stored in cryoprotectant (see above), were washed in PBS for 3× 10 min. For blocking and permeabilization we used “staining buffer” containing 0.05 M Tris, 0.9% NaCl, 0.25% gelatin, and 0.5% Triton X-100, pH 7.4. Primary antibodies, rabbit anti-somatostatin (1:1000; Peninsula Laboratories/Bachem) and mouse anti-parvalbumin (1:500; Sigma-Aldrich) were diluted in staining buffer and incubated overnight at 4°C. The next day, the brain sections were washed in PBS and incubated with donkey anti-mouse Cy3 antibody (1:200; Jackson Immunoresearch) and donkey anti-rabbit Alexa 488 antibody (1:200; Jackson Immunoresearch), or with donkey anti-mouse Alexa 488 (1:200; Jackson Immunoresearch) and donkey anti-rabbit Cy3 antibody (1:200; Jackson Immunoresearch) diluted in staining buffer with Hoechst for one h at room temperature. Finally, the brain slices were washed in PBS and mounted in Mowiol mounting solution (Mowiol 4–88). In order to examine if pathological tau was expressed in GABAergic interneurons, two other aged JNPL3 (BL6) mice (age 13 months) were perfused with 0.9% saline, followed by perfusion of 4% paraformaldehyde (PFA) in PBS. Brains were fixed overnight in buffered 2% paraformaldehyde at 4°C. Next day, hemispheres were stored in 20% glycerol, 2% DMSO in phosphate buffer (PB) until the brains were sliced as consecutive serial coronal sections (5 series, 50 μm thickness). Brains slices were immunostained as described above and incubated with rabbit anti-somatostatin (1:1000; Peninsula Laboratories/Bachem) or rabbit anti-parvalbumin (1:1000, Swant) and with mouse anti-PHF1 (1:500; gift from Dr. Peter Davies) or mouse anti-MC1 (1:100; gift from Dr. Peter Davies). Images of brains were acquired using a Zeiss LSM510 confocal microscope. To quantify the number of GABAergic interneurons per hippocampal region, the complete hippocampus was imaged using Zeiss Zen2009 software. Next, Z-stacked images were stacked to a maximum intensity projections and ROIs were defined in all consecutive slices of one mouse, i.e. hippocampal CA1 and dentate gyrus (DG) in Image J. ROI varied from posterior to anterior. General criteria for the DG ROI were determined by the boundaries of the granular layers. Criteria for the CA1 ROIs were determined as described above. Within these ROIs, the number of positive cells was counted. The average of all positive cells was corrected to the volume of the ROIs.

For DAB staining, consecutive serial brain sections (50 μm thickness) stored in cryoprotectant, were washed in TBS (10 mM Tris, 140 mM NaCl, pH 7.4), then incubated with 3% H₂O₂/0.25% Triton X-100 (Sigma) for 30 min and followed by incubation in 5% milk in TBS. Brain slices were then incubated overnight with mouse anti-PHF1 (1:1000) or mouse anti-MC1 (1:100) in 5% milk in TBS. Slices were then washed in TBS +0.05% Triton-X100 (TBS-T), incubated with goat anti-mouse-Biotin (1:1000; M.O.M. kit, Vector laboratories) secondary antibody diluted in 20% Superblock (Pierce)/TBS-T for 2 hours, washed with TBS-T, and incubated with ABC reaction (M.O.M. kit) in Superblock in TBS-T. Brain slices were washed in 0.2 M sodium acetate and developed in DAB + nickel ammonium sulfate (17mg of DAB, 1.25 g Nickel ammonium sulfate in 50 ml of 0.2 M sodium acetate and 0.3% H₂O₂). Slices were washed in sodium acetate and TBS and mounted. Images of brains were acquired using a Leica DM 5000B light microscope.

Biotin surface labeling was performed as described in [35]. Briefly, hippocampal slices were generated as described for electrophysiology. Slices were maintained in oxygenated ACSF containing the following (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 25 D-glucose, 2 CaCl₂, and 1 MgCl₂ at 32°C for at least 1 h to recover. Then slices were treated with 0.5 mg/ml Sulfo-NHS-SS-biotin (Pierce) for 45 min on ice to label surface proteins. Slices were then washed in Tris-ACSF (25mM Tris pH 7.2 + ACSF). Slices were snap frozen and lysed in homogenization buffer (in mM: 40 HEPES pH 7.5, 150 NaCl, 10 pyrophosphate, 10 glycerophosphate, 1 EDTA) containing protease inhibitor and phosphatase inhibitor cocktail II and III (Sigma). Homogenates were cleared by centrifugation at 13000 rpm for 10 min at 4°C. Then samples were run on a SDS-page gel and immunoblotted against GABA_AR alpha1 (1:1000, Neuromab), GABA_Abeta2/beta3 (1:1000, Upstate) and GAPDH (1:10000, Cell Signaling).

JNPL3 mice, originally generated in a mixed genetic background, develop severe pathological phenotypes, such as neurofibrillary tangles [36]. In our study, we used JNPL3 mice in a C57BL/6 background JNPL3 (BL6) (young: 5–6 months or aged: 12–18 months). To determine if markers of pathological tau were present in aged mice, we examined levels of total tau and phosphorylated tau in the cortex of JNPL3 (BL6) and age-matched C57BL/6 wild-type (WT) control mice. We found that the soluble protein fraction (low speed supernatant) derived from Tg mice showed increased total tau levels compared to WT mice (Figure 1a). In addition, early and late-stage pathological phosphorylation sites of tau, recognized by CP13, AT180, and PHF1 antibodies, were significantly increased in the Tg mice (Figure 1a). Tau protein levels in the sarkosyl pellet, which contains insoluble tau aggregates, were also increased in Tg mice (Figure 1b). Next we analyzed brain tissue from aged Tg mice with the pathological-tau antibodies MC1 and PHF1. MC1 immunostaining, an early-stage conformational tau pathology marker, was found throughout the brain of Tg mice, including in the molecular layer of the hippocampus (Figure 1c). PHF1 staining, a late-stage phospho-tau pathology marker, also showed staining in Tg mice but was less abundant than MC1 staining (Figure 1c). In WT mice, no positive cells were identified (Additional file 1: Figure S1). Combined these data clearly demonstrate that tau pathology is present in the aged JNPL3 (BL6) mice used for this study.

Having verified tau pathology in aged Tg mice, we next examined synaptic plasticity in the CA3-CA1 circuit in hippocampal slices from Tg and WT animals. Late phase-LTP (L-LTP) was significantly enhanced in aged Tg mice compared to age-matched WT mice (Figure 2a, 2b). An examination of several basal synaptic functions such as input/output (I/O), paired-pulse facilitation (PPF), and early phase-LTP revealed no difference between Tg and WT mice (Additional file 2: Figure S2). The L-LTP enhancement was age-dependent, as no effect was observed in younger (6 months old) Tg mice compared to WT controls (Additional file 3: Figure S3a). When we divided the mice in three age groups of 6 months, 12 months and 18 months, we found that enhanced L-LTP was detected from 12 months of age (Additional file 3: Figure S3a, 3b and 3c). Because primary neurons are not lost in aged mice of this model [37], we hypothesized that inhibitory GABAergic signaling was impaired in aged Tg mice leading to enhanced potentiation of the hippocampal CA3-CA1 network. To test the idea that GABAergic function was compromised in our pathological tau mouse model, we treated slices with a sub-threshold (no effect on basal synaptic transmission) dose of the GABA_A receptor (GABA_AR) agonist, zolpidem (1 μM) [38]. Zolpidem treatment 10 min prior to stimulation rescued the enhanced L-LTP in hippocampal slices from Tg mice, and importantly, had no effect on L-LTP in WT slices (Figure 2c,d). We also found a significant positive correlation between the potentiated fEPSP slope and the levels of phospho-tau markers PHF1 (53 kD) and CP13 in the cortex (Additional file 4: Figure S4), linking the severity of hippocampal dysfunction to pathological tau expression. These findings support the idea that the exaggerated synaptic plasticity in Tg mice results from reduced GABAergic function, perturbing normal inhibitory-excitatory balance in the CA3-CA1 circuit.

One explanation for our results is that JNPL3 (BL6) mice have fewer surface GABA_A receptors, reducing neuronal responses to GABA. However, using biotin-surface labeling, we found that total levels and surface levels of several GABA_AR subunits (α1, β2, and β3) were unchanged in Tg mice compared to WT mice (Additional file 5: Figure S5a, 5b). Another possibility is that hippocampal GABAergic interneurons are reduced or lost in the hippocampus of JNPL3 (BL6) mice. As shown in previous studies, the cell body layer and the stratum oriens in the hippocampus contain GABAergic interneurons that control excitation [39, 40]. To test this notion, we used in situ hybridization with a GAD67 riboprobe to quantify the number of GABAergic interneurons in the hippocampus. In agreement with this idea, we found significantly fewer (~20%) GAD67-positive cells in hippocampal area CA1 of Tg mice compared to controls (Figure 3a, 3c). Several classes of GABAergic interneurons have been identified using molecularmarkers, including parvalbumin (PV) or somatostatin (SST) positive cells [41]. Because PV- and SST-positive neurons represent the predominant populations of GABAergic interneurons in the hippocampus [42], we next determined whether the GABAergic interneurons reduced in Tg mice are represented by these two different interneuronal types. We found that the number of both PV and SST interneuron subtypes were significantly decreased (~20%) in area CA1 of Tg mice compared to WT mice (Figure 3b, 3c). Additionally, we found significantly fewer SST-positive neurons in the dentate gyrus (DG) (Additional file 6: Figure S6). There was also a trend toward fewer PV-positive cells in the DG, but it was not statistically significant (Additional file 6: Figure S6). The combined results of these experiments indicate that hippocampal GABAergic function in JNPL3 (BL6) mice is compromised as a result of interneuronal cell loss, derived from the most widely represented interneuronal classes in area CA1.

After finding that pathological tau staining is abundantly present and that GABAergic interneurons are reduced in the hippocampus of JNPL3 (BL6) mice we next examined if the pathological tau markers were localized in GABAergic interneurons in aged JNPL3 (BL6) mice. To do this, we perfused mice and stained fixed brain slices for MC1 (an early pathological tau marker) and PV or SST or for PHF1 (a late pathological tau marker) and PV or SST. Strikingly, both GABAergic interneuron subtypes co-localized with MC1 and PHF1 in the hippocampus (Figure 4). We found that a subset of PV-positive neurons in the CA1 region co-localized with MC1 and PHF1 (Figure 4a, arrows). In the DG however, where PV-positive neurons are less abundant, we did not find PV-positive neurons that co-localized with either MC1 or PHF1 (Figure 4a, arrowheads). In contrast, we found that SST-positive interneurons co-localized with MC1 and PHF1 staining in both the CA1 and dentate gyrus (Figure 4b, arrows). These results show that pathological tau markers are present in GABAergic interneurons, which might promote interneuronal loss in the hippocampus.

A characteristic feature of tauopathy is dementia, so next we examined memory in aged JNPL3 (BL6) mice. Because we found age-dependent deficits in hippocampal synaptic plasticity in these mice (Additional file 3: Figure S3), we assessed both young and aged JNPL3 (BL6) mice in hippocampus-dependent behavioral tasks. Several hippocampal-dependent behavioral paradigms have been described including the Morris Water Maze or Barnes Maze. However these tasks are highly dependent on locomotor competency. Aged JNPL3 mice in the mixed background develop severe motor deficits [36], and while JNPL3 (BL6) mice are less impaired (data not shown) they do exhibit some locomotor deficits. Therefore, to reduce possible locomotor confounds, we decided to test Tg mice and WT controls using the contextual fear conditioning paradigm, a test that depends on hippocampal function but is less reliant on locomotor performance. Mice were trained by placing them in a training context and exposing them to a brief aversive unconditioned stimulus (US, footshock). Then fear memory assessed by freezing behavior was measured at three time points following training (Figure 5a). Before the administration of the US, no difference was detected in the freezing levels or baseline exploratory locomotor activity between Tg and control mice (Figure 5b). Although aged Tg mice displayed levels of acquisition that was indistinguishable from WT mice, they had severely impaired short-term (1 h) and long-term memory (24 h and 7 days) (Figure 5b). The observed memory deficits were not due to differences in nociception, as aged Tg and WT mice displayed indistinguishable response curves in the hot plate assay (Additional file 7: Figure S7). To determine if the onset of this deficit was age-dependent, we also tested younger mice (5–6 months of age). In contrast to aged Tg mice, young Tg mice showed no contextual fear memory impairment at 24 h (Figure 5b). We also tested prepulse inhibition (PPI) of startle response, a behavioral assay dependent on GABAergic function [43]. Consistent with the loss of the GABAergic interneurons we observed, aged Tg mice had a significantly impaired PPI response compared to WT controls (Figure 5c), whereas young Tg and WT mice did not differ (Figure 5d). These results show that synaptic and behavioral deficits in JNPL3 (BL6) mice are a function of age-related pathological tau expression and may be related to GABAergic dysregulation.