|Home | About | Journals | Submit | Contact Us | Français|
Tau inclusions are a prominent feature of many neurodegenerative diseases including Alzheimer’s disease. Their accumulation in neurons as ubiquitinated filaments suggests a failure in the degradation limb of the Tau pathway. The components of a Tau protein triage system consisting of CHIP/Hsp70 and other chaperones have begun to emerge. However, the site of triage and the master regulatory elements are unknown. Here we report an elegant mechanism of Tau degradation involving the co-chaperone BAG2. The BAG2/Hsp70 complex is tethered to the microtubule and this complex can capture and deliver Tau to the proteasome for ubiquitin-independent degradation. This complex preferentially degrades sarkosyl insoluble Tau and phosphorylated Tau. BAG2 levels in cells are under the physiological control of the microRNA miR-128a, which can tune PHF Tau levels in neurons. Thus we propose that ubiquitinated Tau inclusions arise due to shunting of Tau degradation toward a less efficient ubiquitin-dependent pathway.
Tau inclusions are classic hallmarks of many neurodegenerative diseases in which phosphorylated Tau proteins self-assemble into massive polymeric fibrils that are held together in an unknown linkage and resist the normal degradation mechanisms of the cell. Tau inclusions are ubiquitinated, both in poly-ubiquitinated (Cripps et al., 2006) and mono-ubiquitinated linkages (Ii et al., 1997; Iqbal and Grundke-Iqbal, 1991; Morishima-Kawashima et al., 1993). The presence of ubiquitinated inclusions points to a defect in Tau protein triage, a single molecule decision system that assesses each Tau protein and decides whether the protein is functional, or misfolded and salvageable, or beyond repair and subject to proteolysis. Currently, the site of Tau triage is unknown. In the Tau inclusion diseases, at a point beyond this early triage decision, the cell makes an ensemble decision regarding its continued survival in the face of the overall burden due to inclusions or toxic oligomers.
The Tau protein triage system utilizes a complex consisting of the E3 ligase CHIP (carboxyl terminus of Hsp70-interacting protein) in conjunction with several chaperones to refold Tau or direct it toward the degradation machinery (Shimura et al., 2004; Petrucelli et al., 2004). Thus the Tau/CHIP/heat shock complex lies at a pivotal decision node directing the fate of each Tau molecule recognized by the complex toward degradation or restitution. The accumulation of ubiquitinated Tau proteins in cells could arise due to the blockade of a ubiquitin-dependent Tau degradation pathway or inappropriate shuttling to the ubiquitin-dependent pathway from a preferred route. Indeed, other Tau degradation pathways have been described including ubiquitin-independent Tau degradation and caspase-mediated degradation (Berry et al., 2003; Ding et al., 2006).
The observation that the CHIP/Hsp70 complex specifically ubiquitinates AD-type hyperphosphorylated Tau (p-Tau) suggested that phosphorylation is the signal for at least one Tau ubiquitination pathway (Shimura et al., 2004). Increasing Hsp70 reduced total Tau levels and attenuated Tau aggregation suggesting that the CHIP/Hsp70 can promote Tau degradation (Petrucelli et al., 2004). In vivo evidence supports the attenuation of Tau aggregation by CHIP up-regulation (Sahara et al., 2005). Small molecule inhibition of Hsp90, which binds the Hsp70/substrate complex, can decrease p-Tau in a mouse tauopathy model (Dickey et al., 2007).
Our findings reveal a novel and highly efficient pathway of Tau degradation that operates in proximity to the microtubule, is ubiquitin-independent, and is regulated by miR-128, a microRNA that is increased in Alzheimer’s disease (Lukiw, 2007). This pathway is mediated by the co-chaperone BAG2. Members of the BAG family interact with the ATPase domain of Hsp70 through their BAG domains (Takayama et al., 1999) and stimulate the degradation of the chaperone clients in the proteasome. In the case of BAG1, degradation of the glucocorticoid hormone receptor (Demand et al., 2001) occurs in a ubiquitin-dependent manner via the BAG1 ubiquitin-like domain. However, BAG2 lacks the ubiquitin-like domain (Luders et al., 2000; Alberti et al., 2002), and therefore, may be suited to triage client proteins independently of ubiquitin.
The following antibodies were used for immunoblotting and/or immunofluorescence: Tau-5 antibody (1:1000, Biosource), which recognizes phosphorylated and non-phosphorylated forms of Tau. Phosphorylation-dependent Tau antibodies included PHF-1 monoclonal antibody, which recognizes Ser-396 and Ser-404 residues (1:500, provided by P. Davies, Albert Einstein College of Medicine); T181 monoclonal antibody (1:1000, Sigma); S199/202 rabbit monoclonal antibody (1:1000, Sigma). Also used were Flag antibody (1:1000, Sigma); rabbit polyclonal BAG2 antibody (1:500, Abcam, clone ab58682); mouse anti-alpha tubulin (1:50, Sigma), mouse mono- and poly ubiquitinylated proteins, (1:20, clone FK2, BIOMOL), mouse β-actin monoclonal (1:10000, Sigma), rabbit anti-CHIP (N-terminal) (Sigma-Aldrich, C9118), and mouse anti-Hsp70/Hsc70 mAb (Stressgen, BB70). Lactacystin, a proteasome inhibitor (Fenteany and Schreiber, 1998), was used at 10 μM (Calbiochem). Benzyloxycarbonyl-valinyl-alaninyl-aspartyl fluoromethyl ketone (Z-VAD.FMK), an interleukin-1β-converting enzyme (ICE)-like protease inhibitor was used at 20 μM.
Human 4R Tau and mouse BAG2-Flag cDNAs were cloned into pEYFP-C1, pDsRed2-C1 and pECFP-C1 vectors (Clontech). The RNAi sequences were obtained by running an algorithm for picking siRNA sites (Heale et al., 2005) and cloned into pSilencer™ 4.1-CMV puro vector (Ambion). The negative control construct was altered so that the sequence was no longer complementary to BAG2 mRNA. The BAG2 shRNAi sequences synthesized were: Sense strand GCCGGACCCUCACGGUUGAgg and antisense strand UCAACCGUGAGGGUCCGGCcc (overhang in lower case). The wild type human ubiquitin UBC expression plasmid (plamid # 11928) and the Ub-KO plasmid with all seven lysines of ubiquitin mutated to arginines (plasmid # 11934) were purchased from Addgene (Dantuma et al., 2006; Bergink et al., 2006). The K48R ubiquitin mutant was prepared by site-directed mutagenesis on a plasmid expressing mVenus-UBB (Quick-Change II site-directed mutagenesis Kit, Stratagene). Forward primer 5′-gctcatctttgcaggccggcagctggaagatggc, and reverse primer 5′-gccatcttccagctgccggcctgcaaagatgagc were used to introduce a lysine codon (aag) for an arginine codon (cgg) at position 48 of the ubiquitin protein. The mutagenesis was confirmed by sequencing using the primer 5′-cttaccggcaagaccatc.
Monkey kidney COS-7 cells were grown in Dulbeco’s Modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and penicillin/streptomycin (Invitrogen) in a 5% CO2 humidified incubator at 37 °C. Cells were transfected with Lipofectamine (Invitrogen) and lysed with Ripa Buffer (1% Triton X-100, 0.5% Sodium deoxycholate, 0.1% sodium dodecyl sulfate, 150 mM NaCl, 50 mM TrisHCl pH 7.4). Protein concentration was estimated by the BCA protein assay kit (Pierce) and was adjusted to 1 μg/μl.
Pregnant embryonic day 18 (E18) Sprague Dawley rats were sacrificed by CO2 incubation, and embryos were removed immediately by Cesarean section. Hippocampi were removed in dissection media without Calcium and Magnesium (HEPES Buffered Hanks’ Balanced salt solution (HBSS), HEPES, 10 mM, pH 7.3, and Pen/Strep) and digested in 0.25% trypsin with the same dissection media at 37°C for 15 min. The tissue was then washed 2X with HBSS and manually dissociated with a fire-bored Pasteur pipette. Cells were plated at 250,000 cells per six well plate for immunoblot analyses and 90,000 cells per six well plate for immunofluorescence. The plates were previously coated overnight with poly-L-Lysine and incubated with Glial medium (MEM, 20% glucose, pyruvate, Pen/Strep and 10% Horse serum) until plating. 3 h after plating, the medium was changed to Neurobasal medium containing B27 supplement and 0.5 mM glutamine. Very few glial cells were observed in these cultures.
COS7 cells were cotransfected with TAU and BAG2 or transfected with TAU in the absence of BAG2. 16 hours postransfection, cells were incubated for 30 min in DMEM methionine/cysteine Free, supplemented with dyalized FBS and L-Glutamine. Cell were incubated in the same media with Expre35S35S protein labelling mix (200uCi/mL; PerkinElmer, MA) for one hour. After this pulse period, media was removed and cells were washed twice with pre-warmed PBS and incubated up to 28 hs in the same media with unlabelled L-cysteine-HCl (500 ug/mL; Sigma-Aldrich, MO) and L-methionine (100 ug/mL; Sigma-Aldrich). Cells were harvested at 0, 4, 8, 24, and 28 hrs after the pulse period. For harvesting, cells were washed once in pre-warmed PBS, scraped from the well, centrifuged at 2300 rpm/5 min/4 C. The pellet of cells were resuspended in lysis buffer (150 mM Kcl, 25 mM trisHCl, 2 mM EDTA, 0.5 mM DTT, 0.5% NP-40 and 1X Protease Inhibitor Coctail-Sigma Aldrich), and incubated at 4C for 30 min, in a head-to-tail mixer. Lysates were centrifuged at 13200 rpm/1 min/4 C and the supernatant were transferred to a microcentrifugue tube. The lysates were immunoprecipitated with PHF-1 antibody, using protein-G sepharose. Spin Column (Sigma-Aldrich) were used to minimize sepharose beads lost during washing times. Immunoprecipitated p-Tau was eluted using 1X Laemmli sample buffer and separated in an SDS-PAGE 10% gel (1.0 mm). Acrylamide gels were Coomassie-blue stained and then incubated for 30 minutes with EN3HANCE enhancer solution (PerkinElmer), washed on cold water with 1% glycerol and dried. A Phosphorimager screen was exposed for 3 days at −80C and scanned images were quantified with quantity-one software (Bio-Rad Laboratories, CA)
To over-express Tau and BAG2 protein, COS-7 cells were transfected with 4 μg of BAG2 RNAi or a non-silencing RNAi vector in a 6 well plate with Lipofectamine (Invitrogen). 48 h later, the cells were co-transfected with 2 μg of 4R Tau (pDsRed2-C1) and 2 μg of BAG2-Flag pEYFP-C1. 24 h later cells were lysed for immunoblotting. All experiments in primary neurons were initiated five days after plating, and the amount of transfected plasmid was half of the amount used in COS-7 cells. To over-express mir-128a we transfected 250,000 DIV 5 neurons with 75 nM or 250 nM pre-mir-128a or pre-scrambled (Ambion) using Lipofectamine. For transfections, lysates were harvested 48 h later. For RNA isolation and qPCR, precipitates were added to cells and medium was renewed after 5 h. After the indicated time, RNA was extracted by using the miRVana Isolation Kit (Ambion) and DNase treated (DNA-free, Ambion). Reverse transcription was performed by using Superscript III (Invitrogen). Quantitative real-time PCR was performed in an Applied Biosystems PRISM 7900HT Fast Real-Time PCR System with SYBR green PCR master mix (Applied Biosystems). BAG2 Ct was normalized to that of GAPDH.
Downstream of the Firefly luciferase reporter vector pMIR-Report (Ambion) we inserted a 60 bp sequence of the BAG2 3′UTR that contained the miR-128a predicted target site. Our negative control construct was altered so that three bases at the seed region were no longer complementary to miR-128a. We seeded 50,000–60,000 HeLa cells 24 h prior to transfection in 24-well tissue culture plates. The next day 200 ng of the pMIR-Report vector, 20 ng of the transfection control Renilla vector phRLTK (Promega) and 30 pmol pre-mir-128a or pre-srambled were transfected using 3 μl of lipofectamine. Lysates were harvested 24 h after transfection and reporter activity was measured using the Dual Luciferase Assay (Promega). The Firefly luciferase units from every sample were normalized to the units of the transfection control Renilla.
Preparations of sarkosyl insoluble Tau were previously decribed (Cho and Johnson, 2004; DeTure et al., 2002). Briefly, cells were prepared in lysis buffer containing 50 mM TrisHCl pH 7.4, 0.15 M NaCl, and 1% Sarkosyl and scraped off the plate after 30 min incubation at 4 °C. The samples were vortexed, incubated for 30 min at RT, centrifuged 20 min at 3000 g. The supernatants were recovered and centrifuged at 170,000 g for 2 h at 4°C. Recovered supernatants were mixed with electrophoresis sample buffer. Pellets containing sarkosyl insoluble material were mixed with 2X sample buffer. All the samples were boiled at 100 °C for 5 min. Protein was separated on a 4–20% gradient sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), transferred to a Protran membrane, incubated with the relevant antibodies and detected with horseradish peroxidase conjugated antibodies. Bands were visualized with chemiluminescence (Pierce).
COS-7 cells were transfected with 4 μg of each plasmid (BAG2 and Tau). 24 h later, cells were washed in cold PBS and harvested in immunoprecipitation buffer (150 mM NaCl, 0.5% TritonX100, 50 mM TrisHCl pH 7.5, 5 mM EDTA, 1 mM DTT, 1 mM PMSF and protease inhibitor cocktail). The lysate was pre-cleared for 1 h at 4 °C with 25 μl of protein G (Sigma) and centrifuged at 14,000 rpm. The supernatant was incubated with 5 μg of antibody against TAU-5 and 60 μl of protein G and rocked at 4°C overnight. The protein G beads were pelleted and washed with immunoprecipitation buffer. The precipitates were resolved on SDS-PAGE gel and immunoblotted. Beta-actin was used to normalize Tau loading. Statistical analysis were carried out with PRISM Software (GraphPad software) using the Dunnetts’s multiple comparison test after one-way analysis of variance (ANOVA).
Primary rat hippocampal neuronal cultures were infected at DIV 1 with BAG2-FLAG lentivirus or a GFP control lentivirus and grown for one week. The cells were then washed twice with pre-warmed PBS and incubated for 30 min in lysis buffer (150 mM KCl, 25 mM TrisHCl, 2 mM EDTA, 0.5 mM DTT, 0.5% NP-40 and 1X Protease Inhibitor Cocktail (Sigma Aldrich)) at 4 C. They were scraped from the well and the lysates were centrifuged at 13,200 rpm/10 min/4 C, and the supernatant was recovered. Approximately, 600 uL of lysate were immunoprecipitated with PHF-1 antibody or with anti-FLAG M2 agarose affinity gel (Sigma-Aldrich, MO). Immunoprecipitates were eluted with 1X loading sample buffer and western blot was performed with anti-FLAG, PHF-1, anti-CHIP, anti-HSP70.
For immunofluorescent labeling, neurons or COS-7 cells were grown on glass coverslips and gently washed (3×1 min) in warmed MEM-H (1.423% of MEM with HEPES, 0.22% NaHCO3), pH 7.2. The cells were fixed in 4% (wt/vol) paraformaldehyde (15min, 37°C), washed (3×1 min) with PBS, permeabilized with 0.1% triton X-100 (15 min), washed 2X in PBS and blocked for 20min with blocking medium (MEM-H, 1% BSA, 10% Horse serum). Cells were incubated with primary antibodies in blocking medium for 2 h at room temperature, washed (3×5 min) with PBS and incubated for an additional 1 h with secondary antibodies. Finally, cells were washed in PBS and visualized by fluorescence microscopy (Nikon eclipse TE300).
To analyze the effect of BAG2 over-expression, BAG2 RNAi over-expression or BAG2 RNAi non-silencing on Tau levels, BAG2-pEYFP-C1 vector, BAG2RNAi + pEYFP-C1 vector and BAG2 RNAi non-silencing-pEYFP-C1 vector were separately transfected into 5 day old neurons plated on glass coverslips. After three days neurons were fixed and labeled with antibody. In each transfection experiment, 100 BAG2 positive and 100 BAG2 negative neurons were randomly chosen in the same plate. Optical Density (OD) values of PHF-1 stained neurons were compared among the transfected and non-transfected neurons in the same plate. Quantification of pixel intensity was performed with Metamorph software and normalized to background. The number of cells in each group with OD >200 and OD <200 were binned.
For co-localization studies, COS-7 cells were transfected with BAG2 and Tau for 24 h, fixed in 4% paraformaldehyde, and stained with mouse monoclonal anti-α-tubulin and counterstained with Alexa 594 antibody. Data was acquired on a spinning disc microscope (Olympus) and then post-processed by ImageJ software (http://rsb.info.nih.gov/ij/). Images were processed using successive median and Wiener filtering to reduce noise.
COS-7 cells were transfected with BAG2-pEYFP-C1 plasmid and/or with Tau-pECFP-C1 plasmid in 1:4 ratio. 6 hr after transfection, 100 μM 3,4-Methylenedioxy-γ-benzylidine-γ-butyrolactam (KNK437) was added. After 20 h, the cells were imaged on an Olympus IX50 inverted microscope with an Olympus UPLSAPO 60× NA=1.2 water immersion objective and a DVC-1310 CCD camera (DVC Company). With the 8-bit camera used for this study, gain was set to avoid saturated pixels in the image. Images were analyzed with custom-written Matlab (Mathworks) programs in order to extract puncta trajectories. Images were first segmented to extract the puncta, their center was determined by adjusting the background- and intensity-corrected intensity profile to a Gaussian curve. Puncta trajectories were then determined by matching puncta positions between frames using ascending pair wise distances and using puncta diameter and intensity as criteria to confirm the match. See Supplemental Data for additional details.
Insoluble Tau is operationally defined as the fraction of Tau that pellets in the presence of the detergent sarkosyl. When the levels of this insoluble pool of Tau are increased, neurons are predisposed to neurofibrillary tangle formation (Ballatore et al., 2007). The over-expression of BAG2 in primary neurons and the expression of BAG2 and Tau in COS-7 cells both resulted in a marked and selective decrease in the sarkosyl insoluble Tau fraction as detected by two Tau antibodies—Tau-5 and PHF-1 (Fig 1A). Tau-5 recognizes total Tau and PHF-1 recognizes a form of phosphorylated Tau thought to be vulnerable to misfolding.
COS-7 cells were used to investigate the effects of BAG2 on phosphorylated and non-phosphorylated Tau levels (Fig 1B and Supplemental Data). Phosphorylated Tau was detected with PHF-1 (detects Ser-396 and Ser-404, Jicha et al., 1999), Tau181 (Vanmechelen et al., 2000), and S199/202 (Jenkins et al., 2000), and total Tau, both phosphorylated and non-phosphorylated isoforms, was detected with Tau 5 (Berger et al., 2006). Transfection of BAG2 markedly decreased the levels of exogenous phosphorylated Tau in COS-7 cells as detected by these three different phospho-tau markers (Fig 1B). The change in PHF-1, T181 and S199/202 immunoreactivity compared to Tau-5 suggested that the predominant effect of BAG-2 was on phosphorylated Tau. While total Tau was reduced only 40%, we observed a 3–4 fold decrease for PHF-1, a 10–12 fold decrease was observed for Tau181 and an 8–10 fold decrease for S199/200. Thus the degradation rate of phosphorylated Tau induced by BAG-2 is at least four times larger than total Tau, a sufficiently large difference that all phospho-Tau can be degraded without significantly affecting the pool of non-phospho-Tau. The effects of BAG2 over-expression on Tau levels were reversed in the presence of BAG2 RNAi (Fig 1B). Knock down of BAG2-Flag protein was confirmed by immunoblot (Fig 1B).
Pulse-chase experiments were conducted to test whether the effect of BAG2 on Tau was due to enhanced degradation or decreased synthesis (Fig 1D and Supplemental Data). When co-transfected with BAG2, Tau had a half-life of 3 hrs, which corresponds to a degradation rate of 0.23±0.04 h−1; whereas in the absence of BAG2, Tau had a half life of 20 hrs, which corresponds to a degradation rate of 0.035±0.017 h−1. These control rates approximate values in the literature (David DC et al., 2002). Using Welch’s t test we calculated this change to be statistically significant (p < 0.001). The enhanced degradation of Tau was not accompanied by any change in Tau synthesis.
In neurons, over-expression of BAG2 also decreased the endogenous levels of phosphorylated Tau. These biochemical results were confirmed by immunocytochemistry (Fig 1C). Transfection of BAG2-pEYFP-C1 into neurons decreased PHF-1 signal compared to BAG2 negative neurons. Of 100 BAG2 positive neurons analyzed, 13±11% had an OD above 200. BAG2-pEYFP-C1 negative neurons in the same plate showed the opposite result: 92±6% of neurons had an OD above 200. Using the same quantification for the experimental control, empty vector pEYFP-C1 did not change PHF-1 immunoreactivity. Analysis of the neurons by immunoblot gave results that were consistent with the immunocytochemistry (Fig 1C). The decrement in Tau in neurons following BAG2 expression affected both the phosphorylated and total pools of Tau similarly (Fig 1C) probably because phosphorylated Tau represents a greater proportion of total Tau in neurons.
BAG2 RNAi was used to suppress endogenous BAG2 in neurons. This treatment increased the levels of endogenous PHF-1 immunoreactivity by western blot (Fig 1B) and by immunohistochemistry (Fig 1C). Out of 100 neurons, the number of neurons with PHF-1 staining above an OD of 200 was increased by 1.7±0.3 fold. Over-expression of pEYFP-C1 non-silencing did not change the OD values with PHF-1 when compared to non-transfected neurons on the same plate (data not shown).
The association of BAG2 and Tau was validated by their co-immunoprecipitation. Primary neurons from rat hippocampus or cortex were infected with BAG2-Flag Lentivirus (+) or an empty Lentivirus (−) at DIV1 and incubated for 1 week. An immuno-complex was brought down by PHF-1 that included BAG2-Flag, Hsp70 and small amounts of CHIP (Fig 2A). In the absence of the BAG2-Flag infection, the amount of co-immunoprecipitated Hsp70 was markedly reduced and the amount of co-immunoprecipitated CHIP remained negligible. In the reverse immunoprecipitation, the M2 Flag antibody brought down PHF-1 Tau, Hsp70, and interestingly considerably more CHIP (Fig 2A). Taken together, these data point to the existence of an immuno-complex that contains PHF-1 Tau, BAG2, Hsp70, and small amounts of CHIP, as well as BAG2-CHIP complexes that do not contain Tau, but may serve to inhibit CHIP-mediated ubiquitination of Tau.
BAG2, in its role as a CHIP inhibitor (Arndt et al., 2005), prevents the ubiquitination of Tau and would be expected to inhibit ubiquitin-dependent targeting of Tau to the proteasome. If CHIP-mediated Tau degradation were the only Tau degradation pathway, one would expect to see an increase of Tau in BAG2-expressing cells. Instead, the observed decrease in Tau levels suggested that BAG2 may be shuttling Tau to a ubiquitin-independent pathway. To determine whether ubiquitin was involved in BAG2-mediated Tau degradation, blots from the COS-7 cells were also probed for ubiquitinated Tau. High molecular weight Tau bands corresponding to ubiquitinated Tau were confirmed by immunoprecipitating the cell lysate with Tau-5 antibody and labeling the precipitant with ubiquitin antibody. A band of exactly the same size as the high molecular weight Tau-immunoreactive band was observed. Co-expression with BAG2 markedly reduced the ubiquitinated Tau (Fig 2B).
To prove that the Tau degradation enhanced by BAG2 was independent of ubiquitin, we blocked ubiquitin-mediated degradation with the K48R ubiquitin mutant and the Ub-KO mutant (see Experimental Procedures). Neither mutant blocked BAG2-mediated degradation (Fig 2C), thereby providing strong evidence that the BAG2-degradation pathway was independent of ubiquitin. Interestingly, both mutants not only failed to increase the amount of Tau, they actually enhanced Tau degradation in comparison to BAG2 alone. This increased Tau degradation is probably due to caspase activation (see Supplemental Data). To test this possibility, cells expressing Tau and BAG2 were treated with the caspase inhibitor Z-VAD, which inhibited the Tau clearance effects of BAG2 in the presence of the K48R mutant (data not shown). Caspase-3/7 cleavage of Tau occurs in AD (Berry et al., 2003; Ding et al., 2006), and as suggested here, this Tau degradation route may result from blocking Ub-dependent degradation.
Ubiquitin-independent pathways may or may not be independent of the proteasome (Li et al., 2007; Chen et al., 2004). To resolve these possibilities cells expressing BAG2 and Tau were treated with lactacystin to inhibit the proteasome (Fig 2D). Lactacystin blocked the Tau-clearance effects of BAG2. This treatment increased both phosphorylated and total Tau, but with a greater effect on phospho-Tau. This increase in Tau occurred despite the fact that blockade of proteasomal function by lactacystin can activate caspase (Lang-Rollin et al., 2004). On the other hand, Z-VAD did not block the effects of BAG2 on Tau clearance (Fig 2D), suggesting that in the presence of otherwise intact degradation pathways, Tau is not triaged to a caspase pathway by BAG2 in a major way. These data suggest that the mechanism by which BAG2 promotes Tau degradation is mediated by the proteasome in a ubiquitin-independent manner.
The microRNA, miR-128a, is predicted by TargetScan and PicTar to target the BAG2 3′UTR with a seven nucleotide complementarity in the seed region (Fig 3A) (Krek et al., 2005; Grimson et al., 2007). The site, located between 597–619 bp of the human BAG2 3′UTR, is conserved in mouse, rat and dog. DIV 5 neurons were treated with pre-miR-128a and RNA was harvested at day 8. BAG2 RNA levels fell 4-fold upon addition of mir-128a compared to a scrambled control (Fig 3B). The addition of pre-miR-128a to COS-7 cells transfected with FLAG-BAG2 also decreased the protein levels of the fusion protein (Fig 3B). To test whether the regulation of BAG2 by mir-128a is direct, we fused a 60 bp sequence of the BAG2 3′UTR containing the mir-128a site downstream of a luciferase gene. miR-128a addition decreased the luciferase levels of the reporter by 32%, whereas the 3′UTR construct with a mutated seed region rescued the repression (Fig 3C). Finally, treatment with miR-128a in DIV 5 neurons induced a two-fold increase in PHF Tau (p<0.05), detected by PHF-1 antibody (Fig 3D). Upregulation of miR-128a had no effect on the protein levels of Hsp70 or CHIP, suggesting that the increase in PHF Tau is not due to changes in CHIP or Hsp70 (Fig 3D). By deregulating the control pathway through miR-128a, we independently validated the relationship between BAG2 and Tau.
Transfection of BAG2 into COS-7 cells resulted in discrete BAG2 puncta at 15–24 h after transfection (Fig 4A). 28±11% of these puncta co-localized with microtubules. Co-expression of Tau and BAG2 showed a qualitatively similar pattern (Fig 4B); however the proportion of BAG2 puncta that co-localize with microtubules increased to 72±11%. Similarly, BAG2 co-distributes with microtubules in neurons and appeared as puncta in both axons and dendrites (Fig 4C).
To observe the dynamic relationship between BAG2, Tau, and the microtubules, COS-7 cells were transfected with a BAG2-YFP plasmid and in some cases with a Tau-CFP plasmid. BAG2 puncta positions were tracked both with and without Tau at 250 ms intervals. Most of the puncta moved within 1 μm span of their starting point, when observed over a 75 s acquisition period (Fig 5A–C), a region compatible with their diffusion coefficients, D, and the observation time (see Supplemental Movie). With Tau, we measured D=0.044±0.025 μm2.s−1 and without Tau D=0.096±0.060 μm2.s−1 (Fig 5D).
To characterize BAG2 mobility in more detail, we measured the angular distribution of puncta displacements between every three successive puncta positions taken at 250 ms intervals. Near 180°, a significant departure occurred from a flat distribution (Fig 5E). While a flat distribution is the signature of a Brownian motion, the enrichment near 180° indicated a propensity for puncta to return to their approximate starting point after the second 250 ms step as well as a linear bias in the movement. The most likely explanation for the 180° bias is that puncta are spring-like tethered and, taken together with the co-localization, the tethering is probably to a site on the microtubule. The spread of the angle distribution around 180° can be linked to the diffusion coefficient D of the puncta and the length of the tether (see Supplemental Data). For any given D, the longer the tether, the wider the spread. Tau widened the spread of angles compared to cells without Tau suggesting that the leash tethering BAG2 puncta has increased in length. In the absence of Tau, that length approximated the derived size of the puncta, suggesting that puncta lie very near the site where the leash is bound, probably the microtubule. However, in the presence of Tau, the leash was 5 times longer and measured 60±20 nm, a value comparable to the size of Tau (Ruben et al., 1991).
An analysis of reversal rates further supported BAG2 tethering. The probability of observing a reversal after a given number i of forward steps was much higher for the first step and then abruptly fell to random (Fig 5F). Thus for i=1 (one step), reversal events were over-represented, and for i>1 the motion was indistinguishable from Brownian. This finding suggests that BAG2 puncta bear two mobility behaviors: unbiased Brownian motion with a probability Pbrown and tethering with a probability Ptether, with the relation Pbrown+ Ptether=1. Taking all i>1 steps, we determined those steps which were attributable to Brownian motion and computed the duration that each punctum spent moving in Brownian motion. When cumulated over all the puncta, we computed the totality of the tethering events as a ratio of the cumulated observation duration. Tau increased the fraction of time over which BAG2 puncta were tethered, Ptether, from 35.8±0.5% to 68.0±0.6% (see Supplementa Data).
BAG2 is known to bind to Hsp70 which might be the tethering element. To test this hypothesis, we inhibited Hsp70 using KNK437 (Yokota et al., 2000). Interestingly, this treatment induced a dramatic loss of puncta during tracking due to their enhanced diffusion outside the focal plane. For example, upon Hsp70 inhibition, tracking failed in almost 98% of the cases upon 40 frames, this figure was only 64% with normal levels (Fig 6A). The dependence of BAG2 localization on Hsp70 was also suggested by the change of diffusion coefficients, D. Upon inhibition of Hsp70 with KNK437, these values were D=0.23±0.12 μm2.s−1 with Tau and D=0.31±0.14 μm2.s−1 without Tau (see Supplemental Data). After correcting for changes in puncta size, KNK437 treatment resulted in a two-fold increase in D in the presence of Tau and a seven-fold increase in the absence of Tau (Fig 6B). These data suggest that regardless of Tau, the Hsp70-BAG2 complex is restrained from diffusing by a putative tether. When Hsp70 was inhibited with KNK437, the probability of a reversal after a single forward step decreased (see Supplemental Data). In the absence of Tau, Ptether dropped modestly to 31.1±2.6%. In the presence of Tau this percentage dropped more dramatically to 46.5±1.5%. This effect of Hsp70 inhibition on tethering is consistent with the increased diffusion coefficient of BAG2 puncta after KNK437 treatment.
The co-chaperone, BAG2 markedly increased Tau degradation and selectively reduced the levels of sarkosyl insoluble Tau. The acquisition of sarkosyl insolubility is believed to represent a step in the misfolding of Tau that leads to inclusions. Among the regulatory elements that control BAG2 is the microRNA, miR-128a. Upon miR-128a treatment, phosphorylated forms of Tau were more abundant consistent with a functional role for this miRNA as a BAG2 regulator. miR-128a is upregulated in Alzheimer’s disease (Lukiw, 2007) and the resulting decreased strength of BAG2-mediated Tau degradation pathways could confer risk for neurodegeneration. BAG2 enhanced degradation of Tau was not impeded by dominant negative Ubiquitin mutants; however, its degradation was impeded by lactacystin, an inhibitor of the proteasome. Taken together these findings suggest that BAG2 mediates ubiquitin independent degradation of Tau through the proteasome. A pathway such as that outlined here may allow access to internal folding defects, and thereby effectively degrade natively disordered substrates at internal peptide bonds (Weinreb et al., 1996). Examples of proteins degraded in this pathway are the cyclin-dependent kinase inhibitor p21cip1 (Touitou et al., 2001; Sheaff et al., 2000), translation initiation factors eIF4G and eIF3a (Baugh and Pilipenko, 2004), p53 (Asher et al., 2002), and α-synuclein (Tofaris et al., 2001; Liu et al., 2003). Tau is also natively unfolded and evidence of Ubiquitin-independent Tau degradation has been published (David et al., 2002; Cardozo and Michaud, 2002).
The complexity of Tau degradation pathways might be viewed as a hierarchy. Degradation of misfolded Tau occurs preferentially in a ubiquitin-independent manner. Under various conditions Tau can get shunted to a ubiquitin-dependent pathway, but this pathway may be less efficient as suggested by the finding that over-expressed CHIP did not reduce Tau levels (Dickey et al., 2006). Thus CHIP mediated ubiquitin-dependent degradation may become more readily saturated than the BAG2 mediated ubiquitin-independent degradation. Under these circumstances, Tau is ubiquitinated, and based on the longstanding observation that Tau inclusions are ubiquitinated, Tau ubiquitination may make Tau prone to aggregation. While inclusions may represent a temporizing protective measure to create an inert body of non-degradable protein, a riskier strategy to rid the cell of Tau is caspase activation. When the proteasome becomes inhibited in the presence of inclusions, caspases are activated (Bence et al., 2001), and Tau is an efficient substrate for caspase as well as further downstream enzymes such as calpains (Johnson, 2006; Park et al., 2007). However, caspases turn off protective pathways and lead to cellular destruction. The findings here suggest that Tau may become a caspase substrate only under stressful conditions that make proteasomal degradation pathways inaccessible.
BAG2 undergoes two types of mobility: spring-like tethering and Brownian excursions between tethering events. Previously, Tau was reported to move along the axon at rates consistent with slow transport (0.1–8 mm/d, or 0.001–0.08 μm/s; Mercken et al., 1995). More recently, unexpectedly fast diffusion-like movements of Tau at approximately 1 μm/s have been observed along microtubules that were explained by the highly dynamic interaction between Tau and the microtubule (Konzack et al., 2007). However the increased number of BAG2 tethering events induced by Tau indicates that the BAG2/Hsp70/Tau complex is relatively static compared to the reported Tau velocities. Thus BAG2 captures and tethers Tau and possibly makes it available for delivery to the proteasome (Fig 7).
Given the nanomolar affinity of Tau protein for microtubules (Makrides et al., 2004) and its cellular abundance (μM), a large fraction of Tau species is found on the microtubules. Thus positioning the triage decision at the microtubule is a highly expedient strategy. By reducing the relative abundance of phospho-Tau species, BAG2 may also reduce the tendency toward neurofibrilllary tangle formation by specifically limiting the pool of Tau that has lower affinity for microtubules and are prone to aggregate. BAG2 can associate with microtubules in the absence of Tau (Fig 4A; Gache et al., 2005). The addition of Tau to a cell results in further recruitment of BAG2 to the microtubules. Thus BAG2 targets misfolded Tau on the microtubules, a pool of Tau that has been considered a source of Tau filaments (Ackmann et al., 2000; Lu and Kosik, 2001).
Inhibition of Hsp70 using KNK437 resulted in a significant decrease of BAG2 tethering suggesting that Hsp70 plays a key role in anchoring BAG2 on the microtubule. Hsp70 is also involved in Tau triage by binding through its ATPase domain to the C-terminal region of BAG family members (Takayama et al., 1997; Zeiner et al., 1997). Binding in this region gives rise to the chaperone inhibitory properties of BAG family members by preventing the ATP-dependent release of the Hsp70-associated substrate (Bimston et al., 1998). The effect of this complex formation may maintain the substrate in a soluble nonnative state at the point of triage. The recent report that BAG1 over-expression impedes ubiquitin-independent Tau degradation (Elliot et al., 2007) can be explained in the light of our results: BAG1 and BAG2 compete for the same binding site, and are pivotally positioned to determine alternative fates of misfolded Tau, i.e. either proteolysis in a ubiquitin independent or dependent manner. The E3 ligase CHIP can also mediate Tau degradation by interacting with Hsp70. By inhibiting CHIP (Qian et al., 2006), BAG2 directs the Hsp70-Tau complex away from ubiquitination. In contrast to BAG1, the lack of an N-terminal ubiquitin-like domain in BAG2 makes it well suited for ubiquitin-independent delivery of substrates to the proteasome.
In summary, our data suggest that BAG2 plays a critical role in shuttling a pool of microtubule-associated Tau, prone to misfolding, to a ubiquitin-independent pathway for degradation. Because phosphorylation controls both Tau binding to the microtubule and its tendency to misfold, the localization of this degradation pathway to the microtubule is well positioned to divert Tau from ubiquitination, and ubiquitin-dependent delivery to the proteasome, and possibly aggregate formation. Hsp70 plays a key role in Tau triage. We hypothesize that Tau triage decisions are related to the relative concentrations of the co-chaperones, BAG1, BAG2, and CHIP for competitive binding to Hsp70. Routing Tau to a less efficient degradation pathway, will lead to excess misfolded Tau and ultimately the formation of neurofibrillary tangles.
Support for this work came from the National Institutes of Health, the Hillblom Foundation, the Everett Fisher Foundation and the National Science Foundation under Grant No. PHY05-51164. Carrettiero D.C. was a visiting fellow and received his fellowship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES; BEX4494-05-9). Thanks to Stuart Feinstein and DeeAnn Hartung for their help with the pulse chase experiments.
The authors declare no competing financial interests.