Macroautophagy deficiency mediates age-dependent neurodegeneration through a phospho-tau pathway

Genetically altered mice that are deficient in an essential component of the macroautophagy machinery, Atg7 [18], specifically within mature forebrain neurons, were generated using a Cre-loxP strategy [19]. Briefly, we interbred mice that express bacterial Cre recombinase (CRE) under the control of the CamKIIα gene regulatory sequences (CamK-Cre) [20] with Atg7 ^flox/flox mice [19]. CRE expression was limited to CA1 field pyramidal neurons of the hippocampus and glutamatergic neurons within the cerebral cortex [20], leading to ATG7 loss and prominent macroautophagy defects including the accumulations of LC3, GABARAP, GABARAPL1, and p62 in forebrain specific Atg7 conditional knockout (CamK-Atg7 cKO) mice (Figure 1a,b). Quantification of CA1 pyramidal neuron number revealed a significant reduction of approximately 25% in CamK-Atg7 cKO mice at 1-year of age, while 3-month-old cKO mice maintained a normal complement of CA1 neurons (Figure 1c). Consistent with the neurodegenerative process, hippocampal CA1 neurons of 8-month-old CamK-Atg7 cKO mice stained positively for cleaved caspase-3 (Figure 1d). In contrast, neither neuronal loss nor caspase-3 positive signal was observed in the cerebral cortex of 1-year-old CamK-Atg7 cKO mice.

Furthermore, numerous ubiquitin-positive inclusions were apparent in essentially all Atg7-deficient CA1 cell bodies from 2-month of age, whereas these were never seen in the control CamK-Atg7 cWT mice (Figure 1e). These inclusions were stained positive for p62 [17, 21], which is a component of the macroautophagy machinery pathway (Additional file 1), and further confirmed the macroautophagy defect in forebrain neurons. In contrast, such inclusions were absent from the CA3 neurons (data not shown). Further analysis by electron microscopy revealed that these inclusions were composed of both filamentous and vesicular elements (Figure 1f).

We further compared CamK-Atg7 cKO neurodegeneration with the effect of Atg7 deficiency in a second population of mature CNS neurons, midbrain dopamine (DA) neurons. To this end, we generated animals that express CRE under the control of the dopamine transporter (Dat) gene regulatory elements, and are homozygous for the floxed Atg7 allele (Dat ^Cre/+ Atg7 ^flox/flox ; Dat-Atg7 cKO mice rather than CamK-Atg7 cKO mice). Dat-Atg7 cKO mice displayed a very similar pathological progression to CamK-Atg7 cKO mice with cytoplasmic ubiquitin- and p62-positive inclusions, albeit the process is selective for midbrain DA neurons as expected (Additional file 2c,d). Neurodegeneration progresses appeared more rapid in the Dat-Atg7 cKO mouse model than the CamK-Atg7 cKO mouse model (25% midbrain DA neuron lost at 2-months of age and 38% lost at 4-month; Additional file 2a,b).

We further examined the physiological and behavioral consequences of Atg7-deficiency within forebrain neurons. Extracellular recording of field potentials were performed at Schaffer collateral synapses in area CA1 of acutely prepared hippocampal slices from 3-month-old male CamK-Atg7 cKO mice and control CamK-Atg7 cWT littermates. CamK-Atg7 cKO mice showed normal input/output amplitudes in response to single stimuli (Figure 2a), as well as intact paired-pulse facilitation (PPF) at a variety of interpulse intervals (Figure 2b). These findings suggest that there are no gross differences in synaptic organization or baseline synaptic transmission in the cKO mice at this age. In contrast, early long-term potentiation (E-LTP) induced by a single high-frequency tetanic stimulation - a long-lasting protein synthesis-independent form of synaptic potentiation - was impaired in CamK-Atg7 cKO slices (Figure 2c). In contrast, we note that long-term depression was intact in the cKO mice (data not shown). The relatively selective physiological impairment is unlikely to be secondary to the limited cell loss.

Next, we assessed forebrain-dependent fear conditioning in CamK-Atg7 cKO mice and CamK-Atg7 cWT mice. CamK-Atg7 cKO mice did not show any increase in the ratio of freezing at their basal level. However, CamK-Atg7 cKO mice showed a significant impairment in contextual fear conditioning relative to control CamK-Atg7 cWT animals (Figure 2d). Furthermore, the cKO mice showed significant reduced freezing ratio in cued fear conditioning, whereas the basal freezing (‘Pre-Test’) was not changed (Figure 2e). Taken together, these data demonstrate forebrain physiological dysfunction, consistent with the selective forebrain pathology of CamK-Atg7 cKO mice.

We investigated whether neurodegeneration caused by Atg7-deficiency is associated with typical pathological hallmarks of human neurodegenerative syndromes. Macroautophagy has previously been implicated in the clearance of various proteins implicated in human neurodegenerative syndromes including Alzheimer precursor protein (APP), α-synuclein, TDP-43, tau, and huntingtin [22‐29]. However, direct in vivo evidence of an essential role for macroautophagy in the degradation of these proteins in forebrain is lacking. No accumulation of APP (or the APP-derived peptide fragmant β-amyloid), α-synuclein, or TDP-43 was detected in CamK-Atg7 cKO mouse brain (Additional file 3a, b). However, cytoplasmic inclusions in Atg7-deficient CA1 pyramidal neurons and cerebral cortex neurons were prominently stained with multiple well-characterized antibodies to phospho-tau including AT8 (epitope at Ser202/Thr205), AT100 (epitope at Ser212/Thr214), and TG3 (epitope at Thr231/Ser235) [30, 31] (Figure 3a-c). Similarly, electron microscopic analysis confirmed TG3-positive staining in the cytoplasmic inclusions of Atg7-deficient neurons (Figure 3d). We note that the inclusions were not stained with other antibodies for mature phospho-tau positive inclusions in human pathology, AT270 (epitope at Ser181) and PHF1 (epitope at Ser396/Ser404). Furthermore, the cytoplasmic inclusions did not stain with Thioflavin S, which marks mature NFTs in human tauopathies (Additional file 3c).

Quantitative Western blotting of forebrain extracts revealed that phospho-tau protein epitopes were broadly increased in forebrain tissues from CamK-Atg7 cKO mice, whereas total tau protein appeared unaltered (Figure 3e). Several epitopes, including AT8, AT100, and TG3, were increased in both 0.5% TritinX-100-soluble and insoluble brain extracts (relative to CamK-Atg7 cWT controls; Figure 3e), whereas AT180 accumulated only in insoluble extracts, and accumulation was not altered for AT270 and PHF1 (Figure 3e). The phospho-tau epitope staining pattern appeared very similar in midbrain DA neurons of Dat-Atg7 cKO mice (Additional file 2e, Figure 4e). A similar phospho-tau pattern has previously been suggested to represent an early ‘pre-tangle’ state [32]; this may reflect an early stage of non-fibrillar tau aggregation prior to its assembly into paired helical filaments (PHF). Taken together, these data implicate phospho-tau accumulation in Atg7-deficiency-mediated neurodegeneration. However, the phospho-tau aggregates in the context of Atg7-deficient neurons do not replicate aspects of mature human tauopathy pathology.

Given the accumulation of phosphorylated -- but not total -- tau in Atg7-deficient neurons (Figure 4e), we hypothesized that a kinase that is known to phosphorylate tau, such as GSK3β, may be altered. Immunostaining of cortical neurons revealed dramatic re-localization of GSK3β, including both active (epitope at Tyr216) and inactive (epitope at Ser9) phosphorylated forms, to phospho-tau-positive and ubiquitin/p62-positive inclusions in Atg7-deficient neurons (Figure 4a-c). Western blot analysis confirmed that total and phosphorylated forms of GSK3α/β were increased in forebrain tissue extracts from CamK-Atg7 cKO mice, compared to CamK-Atg7 cWT mice (Figure 4d). Another kinase implicated in phosphorylation of tau, CDK5, did not appear to be re-localized to the inclusions in Atg7-deficient neurons [33] (Additional file 4d). Inclusions in Atg7-deficient neurons stained positively for a second microtubule-associated GSK3β substrate, phospho-CRMP2 [34] (Additional file 4a,b). In contrast, β-Catenin, a well-described GSK3β substrate in the context of Wnt signaling pathway, did not appear altered in staining in Atg7-deficient neurons (Additional file 4c). Thus, accumulated GSK3β in the context of Atg7-deficiency appears to display substrate specificity, perhaps related to subcellular re-localization at inclusions.

To examine the causality between phospho-tau and neurodegeneration in the context of Atg7-deficiency, we sought to determine whether neurons deficient in Atg7 could be effectively protected in vivo through the modulation of phospho-tau production. We focused these ‘rescue’ studies on Dat-Atg7 cKO mice (rather than CamK-Atg7 cKO mice) because the neurodegeneration progresses more rapidly in Dat-Atg7 cKO mouse model than CamK-Atg7 cKO mouse model, as noted above, and the degenerative and pathological processes are restricted to a single cell type in the Dat-Atg7 cKO mice (midbrain DA neurons; Additional file 2a,b). Dat-Atg7 cKO mice also displayed a very similar pathological progression to CamK-Atg7 cKO mice with cytoplasmic ubiquitin- and p62-positive inclusions (Additional file 2c,d) that further stain for phospho-tau and GSK3β (Additional file 2e,f). Thus, analysis of pathology in Dat-Atg7 cKO mice affords a more facile and accurate quantification of the cell autonomous impact of macroautophagy on the loss of mature CNS neurons.

To investigate the role of phospho-tau accumulation in Atg7-deficiency-induced neurodegeneration, Dat-Atg7 cKO or Dat-Atg7 cWT mice were treated chronically with a potent GSK3β/CDK5 inhibitor, Alsterpaullone (5 mg/kg/d, i.p.) for a period of 3 weeks starting at 5-week of age [35]. Alsterpaullone can inhibit the activities of GSK3β, as well as several other tau kinases (CDK1/2/5, GSK3α, and, to lesser extent, ERK1/2 and PKA) to suppress tau phosphorylation (Additional file 5a) [36]. At the end of the treatment course (8-weeks of age), pathological examination of the mice revealed that Alsterpaullone treatment led to a significant increase in the survival of midbrain DA neurons in Dat-Atg7 cKO mice (24.3% increased survival, p < 0.01), whereas Alsterpaullone-treated control Dat-Atg7 cWT mice appeared unaltered (Figure 5a, b). In contrast, ubiquitin-positive inclusions were unchanged in size and number in Alsterpaullone-treated Dat-Atg7 cKO mice, whereas no inclusions were seen in Alsterpaullone-treated Dat-Atg7 cWT mice (Additional file 5b, c). This is consistent with the previous report that the inclusion formation and neurodegeneration are independent in the context of macroautophagy deficiency [17]. These in vivo results are suggesting a protective effect by phospho-tau inhibition in the context of macroautophagy deficiency-induced neurodegeneration. As Alsterpaullone does display some inhibitory activity at kinases in addition to GSK3β, such as CDK5 [36], we cannot exclude additional in vivo kinase targets. But we note that unlike GSK3β, CDK5 did not appear modified or re-localized in Dat-Atg7 cKO neurons (Additional file 4d).

Next, we examined the effect of tau-deficiency [37] in Dat-Atg7 cKO mice. We generated Dat-Atg7/tau double cKO (Dat ^Cre/+ Atg7 ^flox/flox tau ^-/- ) mice, and compared the loss of midbrain DA neuron in Dat-Atg7 single cKO (Dat ^Cre/+ Atg7 ^flox/flox tau ^+/+ or Dat ^Cre/+ Atg7 ^flox/flox tau ^+/- ) and Dat-Atg7/tau double cKO mice. The loss of midbrain DA neurons in Dat-Atg7 cKO mice was significantly rescued in Dat-Atg7/tau double cKO mice at the age of 3-month (Figure 5c,d). Again, the formation of ubiquitin-positive inclusion was not changed in Dat-Atg7/tau double cKO mice (Additional file 5d,e). Consistent with the previous report that tau-deficiency alone led to no abnormality in the brain [37, 38], neither neurodegeneration nor ubiquitin/p62-positive inclusions was seen in the midbrain DA neurons of tau KO mice (Figure 5c,d and Additional file 5d,e). Taken together, these approaches support a model whereby accumulation of phospho-tau contributes to neurodegeneration in the context of macroautophagy-deficiency, whereas the formation of ubiquitin/p62-positive inclusions is independent of phospho-tau signaling.

CamK-Cre transgenic mice, Dat ^Cre/+ mice, Atg7 ^flox/flox mice, hAPP-Tg and tau KO mice, used in this study were generated previously [19, 20, 37, 60‐62]. CamK-Cre Tg and tau KO mice were purchased from Jackson Laboratories. All animals were maintained in the animal facility of the Columbia University Medical Center. Experimental protocols were approved by the Institutional Animal Care and Use Committee. Genomic DNA extracted from mouse tails was amplified by PCR for genotyping using standard methods. The PCR primers are the followings: 5’-AGA TGT TCG CGA TTA TC-3’, 5’-AGC TAC ACC AGA GAC GG-3’ for Cre transgene; 5’-TGC TCT GTG AAC TGC CCT GTT T-3’, 5’-TGT TCC TGT GCA CTG CCT CAT T-3’ for Atg7 wild-type allele; 5’-CTT GGG TGG AGA GGC TAT TC-3’, 5’-AGG TGA GAT GAC AGG AGA TC-3’ for Atg7 floxed allele.

Mice were perfused and fixed in 4% paraformaldehyde and post-fixed at 4°C overnight, 50 μm coronal brain sections were generated using a vibratome. The sections were blocked with PBS containing 5% normal donkey serum [NDS], 0.2% Triton X-100 [Tx] for 1 h, and incubated with the solution (PBS, 1% NDS, 0.2% Tx) containing primary antibody at 4°C overnight. The following antibodies were used; anti-TH (P60101, Pel-Freez), anti-TuJ1 (MMS-435P, COVANCE), anti-MAP2 (AB5622, Millipore), anti-cleaved caspase-3 [Asp175] (#9661, Millipore), anti-active caspase-3 (AB3623, Cell Signaling Technology), anti-ubiquitin (Sigma-Aldrich), anti-p62 (03-GP62-C, American Research Products), anti-Aβ [4G8] (SIG39200, COVANCE), anti-Aβ [6E10] (SIG39300, COVANCE), anti-αSynuclein (610786, BD Bioscience) (AB5038, Millipore) (ab1903, ab24715, Abcam), anti-phosph-tau TG3 and PHF1 (gifts from Dr. Peter Davies, Alberts Einstein College of Medicine), anti-phospho-tau AT8, AT100, AT180, and AT270 (Pierce), anti-total GSK3β (#9315, Cell Signaling Technology), anti-phospho-GSK3α/β [Y279/Y216] (ab52188, Abcam), anti-phospho-GSK3β [S9] (ab30619, Abcam), anti-total CRMP2 (#9393, Cell Signaling Technology), anti-phospho-CRMP2 [T514] (#9397, Cell Signaling Technology), anti-Cdk5 (MAB5410, Millipore), anti-p35/25 (#2680, Cell Signaling Technology), anti-β-catenin (#9581, 9587, Cell Signaling Technology), and anti-β-catenin (#610154, BD Biosciecnes). For secondary detection, Cy3- or FITC-conjugated antibodies were incubated for 1 h (Jackson ImmunoResearch). Photographs were taken using a Zeiss LSM 510 Meta confocal microscope.

To obtain neuronal cell count, 50 μm coronal brain sections were made using a vibratome. In order to count CA1 neurons, the first 30 sections from the rostral hippocampus were stained with rabbit anti-MAP2 antibody (AB5622, Millipore) at a dilution of 1:500, as well as NeuroTrace^TM Fluorescent Nissl stain (N21480, Invitrogen). MAP2-positive neurons were visualized using a Cy3-conjugated secondary antibody (Jackson ImmunoResearch). MAP2 and Nissl double-positive neurons in the CA1 regions were counted manually. In order to count TH-positive neurons, sections covering the entire substantia nigra (25-30 sections / mouse) were stained with sheep anti-TH antibody (P60101, Pel-Freez) at a dilution of 1:250. TH-positive neurons were visualized using the ABC Kit (PK6106, Vector Laboratories) and DAB (SK4100, Vector Laboratories). TH-positive neurons in the substantia nigral regions were counted manually under the light microscope.

Electron microscopic analysis was as described [61]. Anesthetized mice were perfused and fixed in PBS containing 4% paraformaldehyde and 0.5% gultaralaldehyde. The brains were post-fixed at 4°C for 2 h, and the 80 μm vibratome sections were made. The sections were treated in 1% osmium tetroxide, then dehydrated in pure ethanol and infiltrated overnight with Epon 812. Epon was polymerized at 60°C for 24 h, cooled and embedded in a larger Epon capsule. Ultrathin sections were cut with an MT5000 ultramicrotome, stained with uranyl acetate and lead citrate. Images were taken with a JOEL 100S Electron Microscope (JOEL USA).

Preparation of soluble and insoluble fractions was performed as described with some modifications [14]. Cortical and hippocampal tissues from mouse brains were homogenized in 5× volume of ice-cold 0.25M sucrose buffer (50mM Tris-HCl [pH7.6]) containing protease inhibitors (P8340, Sigma) and phosphatase inhibitors (#78420, Thermo Scientific). The homogenized tissues were centrifuged at 500× g for 10 min at 4°C. The supernatants were lysed with an equal volume of cold sucrose buffer containing 1% Triton X-100. The lysates were centrifuged at 13,000× g for 15 min at 4°C. The supernatants contained the soluble fraction. The pellets were resuspended in 1% SDS in PBS (insoluble fraction). Both fractions were subjected to standard Western Blotting analysis. The antibodies used here are: anti-phospho-tau AT8, AT100, AT180, AT270, TG3 and PHF1, anti-Tau1 and anti-Actin (ab3280, Abcam). Horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch) and SuperSignal West Pico or Dura (#34077, 34075, Pierce) were used for detection.

Brains from CamK-Atg7 cWT and cKO mice littermates (~12 weeks of age) were quickly removed and transverse hippocampal slices (400 μm) were isolated with a Leica VT1200 Vibratome (Leica, Bannockburn, IL), and placed in ice-cold cutting solution (CS: 110 mM Sucrose, 60 mM NaCl, 3 mM KCl, 1.25 mM NaH₂PO₄, 28 mM NaHCO₃, 0.5 mM CaCl₂, 7 mM MgCl₂, 5 mM Glucose, 0.6 mM Ascorbate. Slices were placed in an interface chamber (Scientific Systems Design, Mississauga, Ontario, Canada) and maintained at 32°C in ACSF (2 ml/min) containing 125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃, 25 mM D-glucose, 2 mM CaCl₂, and 1 mM MgCl₂. All solutions were constantly caboxygenated with 95% O₂ + 5% CO₂. Slices were allowed to recover for 120 min on the electrophysiology rig prior to experimentation. Bipolar stimulating electrodes (92:8 Pt:Y) were placed at the border of area CA3 and area CA1 along the Schaffer-Collateral pathway. ACSF-filled glass recording electrodes (1–3 MΩ) were placed in stratum radiatum of area CA1. Basal synaptic transmission was assessed for each slice by applying gradually increasing stimuli (0.5–15V), using a stimulus isolator (A-M Systems, Carlsborg, WA) and determining the input:output relationship. All subsequent stimuli applied to slices was equivalent to the level necessary to evoke a fEPSP that was ~40% of the maximal initial slope that could be evoked. Synaptic efficacy was continuously monitored (0.05 Hz). Sweeps were averaged together every 2 min. fEPSPs were amplified (A-M Systems Model 1800) and digitized (Digidata 1440, Molecular Devices, Sunnyvale, CA) prior to analysis (pClamp, Molecular Devices, Sunnyvale, CA). Stable baseline synaptic transmission was established for 30 min. Slices were given high-frequency stimulation (HFS) to induce long-term potentiation (LTP) using one train of 100 Hz for one second. Stimulus intensity of the HFS was matched to the intensity used in the baseline recordings. fEPSP initial slopes from averaged traces were normalized to those recorded during baseline. Two-way RM-ANOVA were used for electrophysiological data analysis with p < 0.05 as significance criteria.

10-13-mon-old male CamK-Atg7 cWT or CamK-Atg7 cKO mice were used (n = 8 - 10). The mice were placed in a conditioning chamber (Med Associates) for 2 min before the onset of a tone (conditioned stimulus) (30 s, 85 dB sound at 2800 Hz) and conditioned by a single electrical foot shock (0.45 mA) in the last 2 s. The mice were left in the chamber for another 30 s and placed back into their home cage. Contextual fear learning was measured in the same chamber 24 h after the training by monitoring the freezing for 5 min without electrical shock. Cued fear learning was measured 24 h after the contextual testing. The mice were placed in a novel chamber for 2 min (pre-conditioning). After that, the mice were exposed to the conditioned stimulus for 3 min, and the freezing was monitored. Freezing behavior was scored using FreezeView software (Med Associates Inc.).

Five-week-old Dat-Atg7 cWT and Dat-Atg7 cKO mice were treated with Alsterpaullone (A1136, A.G. Scientific) [35]. The drug was dissolved in saline containing 20% DMSO/ 25% Tween80, sonicated, and injected intraperitoneally at a dose of 5 mg/kg every day for 3 weeks. After the final injection, the mice were perfused and processed for histological analyses. We used Dat-Atg7 cWT mice as controls for Dat-Atg7 cKO mice, to address potential phenotypes due to Cre transgene inserted at the DAT locus [62].