Tau – an inhibitor of deacetylase HDAC6 function


  • Mar Perez,

    1. Centro de Biología Molecular “Severo Ochoa” (CSIC-UAM), Universidad Autonoma de Madrid, Madrid, Spain
    2. Facultad de Medicina, Universidad Autonoma de Madrid (UAM), Madrid, Spain
    Search for more papers by this author
    • These two authors contributed equally to this work.

  • Ismael Santa-Maria,

    1. Centro de Biología Molecular “Severo Ochoa” (CSIC-UAM), Universidad Autonoma de Madrid, Madrid, Spain
    2. CIBERNED, Madrid, Spain
    Search for more papers by this author
    • These two authors contributed equally to this work.

  • Elena Gomez De Barreda,

    1. Centro de Biología Molecular “Severo Ochoa” (CSIC-UAM), Universidad Autonoma de Madrid, Madrid, Spain
    Search for more papers by this author
  • Xiongwei Zhu,

    1. Department of pathology, Case Western Reserve University, Cleveland, Ohio, USA
    Search for more papers by this author
  • Raquel Cuadros,

    1. Centro de Biología Molecular “Severo Ochoa” (CSIC-UAM), Universidad Autonoma de Madrid, Madrid, Spain
    Search for more papers by this author
  • Jose Roman Cabrero,

    1. Departamento de Biología Vascular e Inflamación, Centro Nacional de Investigaciones Cardiovasculares, Madrid, Spain
    2. Servicio de Inmunologia, Hospital de la Princesa, Universidad Autonoma de Madrid, 28006 Madrid, Spain
    Search for more papers by this author
  • Francisco Sanchez-Madrid,

    1. Departamento de Biología Vascular e Inflamación, Centro Nacional de Investigaciones Cardiovasculares, Madrid, Spain
    2. Servicio de Inmunologia, Hospital de la Princesa, Universidad Autonoma de Madrid, 28006 Madrid, Spain
    Search for more papers by this author
  • Hana N. Dawson,

    1. Division of Neurology, Duke University, Durham, North Carolina, USA
    Search for more papers by this author
  • Michael P. Vitek,

    1. Division of Neurology, Duke University, Durham, North Carolina, USA
    Search for more papers by this author
  • George Perry,

    1. Department of pathology, Case Western Reserve University, Cleveland, Ohio, USA
    2. College of Science, University of Texas at San Antonio, Texas, USA
    Search for more papers by this author
  • Mark A. Smith,

    1. Department of pathology, Case Western Reserve University, Cleveland, Ohio, USA
    Search for more papers by this author
  • Jesus Avila

    1. Centro de Biología Molecular “Severo Ochoa” (CSIC-UAM), Universidad Autonoma de Madrid, Madrid, Spain
    2. CIBERNED, Madrid, Spain
    Search for more papers by this author

Address correspondence and reprint requests to Jesus Avila, Facultad de Ciencias, Centro de Biología Molecular “Severo Ochoa”, Universidad Autónoma de Madrid, Campus de Cantoblanco, Madrid 28049, Spain. E-mail: javila@cbm.uam.es


Analysis of brain microtubule protein from patients with Alzheimer’s disease showed decreased alpha tubulin levels along with increased acetylation of the alpha tubulin subunit, mainly in those microtubules from neurons containing neurofibrillary tau pathology. To determine the relationship of tau protein and increased tubulin acetylation, we studied the effect of tau on the acetylation-deacetylation of tubulin. Our results indicate that tau binds to the tubulin-deacetylase, histone deacetylase 6 (HDAC6), decreasing its activity with a consequent increase in tubulin acetylation. As expected, increased acetylation was also found in tubulin from wild-type mice compared with tubulin from mice lacking tau because of the tau-mediated inhibition of the deacetylase. In addition, we found that an excess of tau protein, as a HDAC6 inhibitor, prevents induction of autophagy by inhibiting proteasome function.

Abbreviations used

Alzheimer’s disease


acetylated lysine


bovine serum albumin


Green Fluorescent Protein


histone deacetylase 6


human embryonic kidney cells


microtubule-associated proteins


neurofibrillary tangles


paired helical filaments

Neurons are highly polarized cells with a clear asymmetric morphology that is needed for their function. A major determinant of neuronal morphology is the cytoskeleton composed of microfilaments, neurofilaments, and microtubules. Microtubules, polymers of the protein tubulin, are especially abundant in neurons and they have a direct participation in the maintenance of neuronal polarity, through the stabilization of the microtubule structure (Avila 1990). That stabilization takes place, in part, by the presence of a group of proteins that co-purify with tubulin, the microtubule-associated proteins (MAPs) (Caceres and Kosik 1990). One of these MAPs, the tau protein (Kanai et al. 1992), is a major microtubule-stabilizing factor (Drubin and Kirschner 1986; Caceres and Kosik 1990; Kanai et al. 1992; Takemura et al. 1992); although its function can be complemented by other MAPs as mice lacking tau are viable (Harada et al. 1994; Dawson et al. 2001). In addition to MAPs, microtubule stabilization is correlated with the presence of post-translational modifications of tubulin, one of them being tubulin acetylation at residue 21 of the alpha subunit of tubulin (LeDizet and Piperno 1987). Acetylated tubulin has been found in very stable microtubules (Maruta et al. 1986; Piperno et al. 1987); however, acetylation, in vitro nor in vivo, does not seem to be the cause of microtubule stabilization, but rather a consequence of this process (Schulze et al. 1987). Analysis of tubulin acetylation in vitro showed that the specific enzyme that acetylates tubulin has a greater affinity for the polymeric form of tubulin than for the monomeric form (Maruta et al. 1986). More recently, it has been indicated that this enzyme could be a component of the histone acetyltransferase elongator complex (Creppe et al. 2009). On the other hand, the level of tubulin acetylation is the consequence of the equilibrium between acetylases and deacetylases. At least two types of deacetylases act on acetylated tubulin (North et al. 2003; Zhang et al. 2003), the most studied being histone deacetylase 6 (HDAC6) (Zhang et al. 2003). HDAC6 is expressed in most neurons, but mainly in Purkinje cells (Southwood et al. 2007). Tubulin acetylation, regulated by HDAC6, enhances recruitment of kinesin 1 to microtubules and leads to an increased vesicle transport and subsequent release of brain-derived neurotrophic factor (Reed et al. 2006; Kazantsev and Thompson 2008). However, while HDAC6 knockout mice have hyperacetylated tubulin, they are viable and develop normally, without neurological abnormalities (Zhang et al. 2008). Another function of HDAC6 is as a component of the aggresome (Kawaguchi et al. 2003); cells deficient in HDAC6 fail to clear misfolded protein aggregates from the cytoplasm as they cannot form functional aggresomes.

In some neurodegenerative disorders, known as tauopathies [i.e., Alzheimer’s disease (AD)], there are changes in neuronal morphology. In these disorders, hyperphosphorylated tau no longer binds with the same affinity to tubulin, and instead aggregates into filaments (Avila et al. 2004). Recently, increased tubulin acetylation has been demonstrated in neurons containing neurofibrillary (tau) pathology in cases of AD (Zhu et al. 2007). Increased tubulin acetylation, which could stabilize microtubules, may reflect a compensatory response to the reduced tau binding to microtubules. Alternatively, another mechanism could be responsible for tau acetylation that is related to the impairment in tau function. In the latter case, it would be of interest to know what the consequence of tubulin acetylation in the absence of tau protein.

In the present study, we compared the level of tubulin acetylation in mice lacking tau protein with wild-type mice. Our results suggest that tau protein can act as a deacetylase inhibitor by directly interacting with HDAC6. At the time of preparation, a report indicating a direct interaction of tau with HDAC6 in human cells was published (Ding et al. 2008). HDAC6 has two main functions, both as a deacetylase and as a component of the aggresome (Yang and Seto 2008). In this study, we found that tau protein can interfere with both HDAC6 activities.

Materials and methods


The antibodies used in this study were the phosphorylation-independent 7.51 monoclonal antibody (1 : 100; kindly provided by Dr. C. M. Wischik, Aberdeen, UK); Tau-5, a monoclonal antibody against the non-phosphorylated form of tau (1 : 1000; Calbiochem, San Diego, CA, USA); Tau-1, a monoclonal antibody against unphosphorylated tau (1 : 1000; Chemicon, Temecula, CA, USA); the monoclonal antibodies directed against α-tubulin and acetylated α-tubulin from Sigma (St Louis, MO, USA); anti-HDAC6 (H-300; rabbit polyclonal, 1 : 10, Santa Cruz Biotechnology, Santa Cruz, CA, USA); LC3 antibody (Sigma) and rabbit polyclonal directed against acetylated lysine (AL, Upstate Biotechnology, Lake Placid, NY, USA) residues.

Human brains

Tissue section preparation

CNS tissue samples were obtained from postmortem of patients (= 8, age: 72–83 years) with histopathologically confirmed AD as well as from non-AD young and aged-matched controls (= 7, age: 69–85 years), both groups with comparable postmortem intervals. Tissue was fixed in methacarn (methanol : chloroform : acetic acid in a 6 : 3 : 1 ratio) by immersion for 24 h at 4°C. Tissue was subsequently dehydrated through graded ethanol and xylene solutions and embedded in paraffin. Six μm thick microtome sections were prepared and placed on silane-coated slides.


Following hydration, sections were immunostained by the peroxidase–antiperoxidase procedure (Zhu et al. 2000) using affinity-purified rabbit polyclonal antibody to AL residues (Upstate Biotechnology). Immunoreactivity was enhanced by formic acid (70%, 5 min) pre-treatment, presumably because of the unmasking of a previously occult epitope. Adjacent sections were also immunostained with a tau antibody to locate the neurofibrillary tangles (NFTs). To verify the specificity of immunolabeling, adsorption experiments were performed by pre-incubating the antibody with 1 mg/mL of acetylated bovine serum albumin (BSA) protein. Furthermore, in order to assess the specificity of the adsorption, AT8 antisera was also incubated with the acetylated BSA protein and then applied to the section. Double immunolabeling with AL and AT8 was performed to resolve the temporal and spatial nature of the acetylation modifications. A mouse monoclonal antibody to alpha tubulin (Sigma) was also used in adjacent sections along with anti-AL double staining to confirm the specificity of the AL antibody for tubulin.


Generation of the mouse line tau−/− was previously described (Dawson et al. 2001). Transgenic mice were genotyped by PCR. Transgenic animals as well as wild-type mice (C57BL/6) were bred at the Centro de Biología Molecular ‘Severo Ochoa’ (Madrid, Spain) and maintained following the institutional guidelines. Four to five mice were housed per cage with food and water available ad libitum. Mice were maintained in a temperature-controlled environment on a 12/12 h light–dark cycle with light onset at 7:00 am.

Cell cultures

Primary cultures of hippocampal neurons were prepared according to established procedures with slight modifications (Hernandez et al. 2004). Briefly, hippocampal tissue was obtained from E18 mouse embryos, dissected, and dissociated with the Papain Dissociation System (PDS; Worthington Biochemical Corporation, Lakewood, NJ, USA). Neurons were maintained in Neurobasal medium (Gibco, Grand Island, NY, USA) supplemented with 1 % B-27, 1% N2, 0.5 mM glutamine, 100 units/mL penicillin, and 100 mg/mL streptomycin, and grown on poly-l-lysine (100 mg/mL) plus laminin (10 μg/mL) coated plastic dishes. The cells were incubated in 95% air/5% CO2 in a humidified incubator at 37°C. Human embryonic kidney (HEK293) and tau-expressing HEK293 cells were cultured as previously described (Santa-Maria et al. 2007). To induce autophagy, cells were incubated with 25 μM MG262 (Biomol International, Plymouth Meeting, PA, USA). The induction was assessed by measuring LC3 levels (Iwata et al. 2005). For cultured cells and neurons Tubacin was added to a final concentration of 2 μM (Cabrero et al. 2006).

Staining of autophagosomes

Human embryonic kidney cells and HEK293 tau cells were transfected with a Green Fluorescent Protein (GFP)-LC3 expression plasmid (Garcia-Escudero et al. 2008). After 24 h, the cells were treated with MG262, a proteasome inhibitor, and the GFP-LC3 pattern was observed under an LSM510 confocal microscope (Carl Zeiss S.A., Madrid, Spain).

Lentiviral vector and virus preparation

The lentiviral vector was constructed on the backbone of pLVFrat (Fleming et al. 2005). The Frataxin cDNA was removed from this vector using XhoI and SalI restriction enzymes. The cDNA encoding the tau isoform 4R2N (Spillantini and Goedert 1998) was excised from plasmid pLVGFPTau (previously constructed) as an XhoI-SalI fragment. This fragment was then subcloned into the XhoI-SalI sites of the pLVFrat vector. The correct orientation and integrity of the tau open reading frame in the final construct was confirmed by restriction fragment analysis and sequencing. The resultant lentiviral vector was designated pLVTau. The viral particles were produced as described previously for both pLVEGFP and pLVTau (Deglon et al. 2000).

HDAC activity assay

Histone deacetylase activity was measured using a flurometric HDAC assay kit (ab1438; Abcam, Cambridge, United Kingdom) according to manufacturer’s instructions. Depletion of HDAC6 activity in the cell extract was carried out by incubation of the extract with an HDAC6 antibody (Santa Cruz Biotechnology) absorbed on nitrocellulose, overnight at 4°C. In parallel, the same treatment was carried out by incubating the extract with albumin absorbed into nitrocellulose. After incubation, HDAC activity in the extract was measured by fluorimetry. In addition, HDAC activity in vitro was measured in the absence or presence of 1 μM tubacin, a specific HDAC6 inhibitor (Estiu et al. 2008; Cabrero et al. 2006). Purified human HDAC6 (Gu and Roeder 1997; Ito et al. 2002) was obtained from Biomol International and assayed with the kit of the same company.

Protein extraction and western blotting

One half of the mouse hippocampus was disrupted in a Dounce homogenizer in 3 mL of buffer H (Tris–HCl 10 mM, EGTA 1 mM, NaCl 0.8 M, 10% sucrose, pH 7.4) containing a protease inhibitor cocktail, orthovanadate, NaF, and okadaic acid. This homogenate was stored frozen at −70°C and used for immunoblot analysis. Samples were run on 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and electrophoretically transferred to a nitrocellulose membrane (Schleicher & Schuell GmbH, Dassel, Germany). The membrane was blocked by incubation with 5% semi-fat dried milk in phosphate-buffered saline and 0.1% Tween 20, followed by incubation for 1 h at 20–22°C with the primary antibody in phosphate-buffered saline and 0.1% Tween 20. The dilutions of the primary antibodies are as follows: α-tubulin, 1 : 1000, and acetylated α-tubulin, 1 : 1000. After three washes, the membranes were incubated with a horseradish peroxidase anti-rabbit or anti-mouse Ig conjugate (DAKO, Barcelona, Spain), followed by several washes in phosphate-buffered saline and Tween 20. The membranes were then incubated for 1 min in Western Lightning reagents (Perkin Elmer Life Sciences, Boston, MA, USA).

Isolation of PHFs

Paired helical filaments (PHF) were isolated by preparation of sarkosyl-insoluble extracts and the isolated PHF were characterized by electron microscopy as previously indicated (Greenberg and Davies 1990; Hernandez et al. 2002).


For isolation of the tau/HDAC6 complexes from mouse brain tissue, 100 μL of brain extract plus 40 μL of Protein G Sepharose beads were incubated at 4°C. After 1 h, the mixture was centrifuged for 10 min at 4°C. For immunoprecipitation, the supernatant was incubated overnight with the following tau antibodies: 7.51, Tau-1, and Tau-5. Protein G-Sepharose-isolated immune complexes were washed three times with assay buffer, boiled in 2× sodium dodecyl sulfate–polyacrylamide gel electrophoresis sample buffer for 5 min, and further analyzed by western blot with either 7.51, tau-1, or HDAC6 antibodies.

Statistical analysis

Comparisons were performed using Student’s t-test, and values were considered significant at < 0.05.


Increased tubulin acetylation in the brain of Alzheimer’s disease patients

The mechanism underlying tau pathology in neurons from AD patients is not fully understood. It has been proposed that tau pathology involves tau phosphorylation and the subsequent formation of tau aggregates. This may result in microtubule destabilization, and presumably, a decrease in the number of acetylated microtubules. However, as indicated in Fig. 1, the opposite was found; tubulin acetylation was increased in AD. Acetyl lysine (AL) immunolabeling in AD was present in two distinct forms, first, within intracellular NFTs (Fig. 1c), and second, as homogenous cytoplasmic staining in hippocampal pyramidal neurons (Fig. 1a). The cytoplasmic phenotype was more prominant in the CA3/4 region while the neurofibrillary pathology predominated in the CA1 region. In the age-matched control cases, only two cases displayed AL immunoreactivity. Both cases contained mild cytoplasmic staining in neuronal cell bodies but lacked NFTs and NFT-associated staining with double-label immunocytochemistry (Fig. 1b).

Figure 1.

 Presence of acetylated lysine in neurons from patients with Alzheimer disease (AD). (a and b) Representative cytoplasmic anti-AL staining in AD (a) and control (b). In addition to the cytoplasmic labeling, all cases of AD showed distinct staining of neurofibrillary tangles (NFTs) (c), which is completely abolished following adsorption with antigen (d, *marks the same vessel in adjacent sections). Serial sections stained with AL (e) and AT8 (f) show that AL is present in greater numbers of NFT (*marks the same vessel in adjacent sections). In the inset of (f) double-immunostaining shows that AL (blue) is present alone in some NFT (arrow) while other NFT display both AT8 (brown) and AL (arrowheads).

Absorption using acetylated BSA protein completely abolished the AL immunoreactivity (Fig. 1d). Double immunolabeling with AT8 and AL showed the presence of AL immunoreactivity alone in some neurons, and the colocalization of AL and AT8 in some neurons containing neurofibrillary pathology (Fig. 1f, inset). The presence of AT8 immunoreactivity alone was seen in some neurons, but AL was found in greater numbers of neurons than AT8 as shown in adjacent sections stained with AL and AT8 (Fig. 1e and f).

One possible explanation for this result is that the tau protein participates in the regulation of tubulin acetylation, which is the consequence of equilibrium between acetylases and deacetylases. Tubulin deacetylases have been well characterized, so we determined whether tau could bind to any of the tubulin deacetylases, as recently reported (Ding et al. 2008).

Histone deacetylase 6 co-immunoprecipitates with tau protein in a mouse brain extract

There are several deceatylases, but HDAC6 is one of the few that deacetylates acetylated alpha-tubulin and is present in the cytoplasm of neurons (Hubbert et al. 2002). In a recent study using human cells, an interaction between tau and HDAC6 was reported (Ding et al. 2008). In that work, the interaction of tau and HDAC6 was shown to take place through the SE14 domain present in HDAC6, which binds to the tubulin binding domain present in the tau protein. The SE14 domain is located in the HDAC6 molecule between the catalytic moiety involved in deacetylation (Yang and Seto 2008) and the region involved in the binding to polyubiquitin that has been related to aggresome formation (Hook et al. 2002). Although the whole SE14-repeat domain was described in human HDAC6, a partially homologous sequence is present in mice (Yang and Seto 2008).

Consequently, we determined whether tau co-immunoprecipitates with the major tubulin deacetylase (HDAC6) (Hubbert et al. 2002; Zhang et al. 2003) in a mouse brain extract. HDAC6 co-immunoprecipitated with tau from a cellular wild-type mouse brain extract (Fig. 2a). In contrast, HDAC6 remained in the soluble fraction when the immunoprecipitation analysis was performed with tau antibody but using a brain extract from tau−/− mice (Fig. 2b). This result suggests that there is an interaction between tau and HDAC6, as previously suggested using human cells (Ding et al. 2008). In addition, our results suggest that the partial SE14 sequence domain present in mice may be sufficient for tau binding.

Figure 2.

 Co-immunoprecipitation (IP) of HDAC6 and tau protein. A brain cell extract from wild-type tau (a) or from tau−/− (b) mice was immunoprecipitated with tau antibodies (ab). After IP, the immunoprecipitate was isolated by centrifugation (pelleted fraction) and the presence of HDAC6 and tau protein in the non-pelleted (s) and pelleted (p) fractions was determined by western blotting (WB).

Next we tested whether the binding of tau to HDAC6 results in the modification of either of the two HDAC6 functions, its deacetylase activity or its role in aggresome formation. To that end, we used two model systems, mouse neuronal cells lacking or containing tau, and for biochemical studies that require a higher amount of cells, a non-neuronal human cell line lacking or expressing human tau protein. These studies were complemented with some in vitro analysis.

Decrease in tubulin acetylation in mice lacking tau protein neurons

Primary cultures of hippocampal neurons from mice lacking or containing tau protein were obtained and the level of tubulin acetylation was determined by immunofluorescence. Figure 3a shows a decrease in acetylated tubulin in the cells obtained from tau-deficient mice compared with those from wild type. If the tau protein inhibits tubulin deacetylation, tubulin will be more acetylated in the presence of tau protein. This result was confirmed by western blot analysis as shown in Fig. 3b. There was a higher level of tubulin acetylation in cell extracts from tau containing mice compared with those lacking tau protein; quantitation of total tubulin and tau protein for wild-type and tau−/− mice is shown (Fig. 3b and c) (Matsuyama et al. 2002; Tran et al. 2007).

Figure 3.

 Decreased level of tubulin acetylation in neurons from tau-deficient mice. (a) Cortical neurons from wt and tau−/− mice were immunostained with an antibody against tubulin that recognizes the whole α-tubulin subunit (left panel) or with an antibody that only reacts with acetylated α-tubulin (right panel). (b) Western blot (WB) and quantitation for acetylated and total tubulin from wild-type (wt) and tau-deficient mice (tau −/−). The average of at least three separate experiments are indicated. *< 0.05 compared with wt neurons. (c) The levels of acetylated tubulin in neurons from wt and tau−/− mice in the absence or presence of 2 μM tubacin are shown. Quantitation of (c) is indicated. The average of at least three separate determinations is indicated. *< 0.05 compared with control neurons without tubacin treatment.

In addition, cortical neurons from wild-type and tau−/− mice were infected with lentivirus expressing GFP protein or tau protein. Only when the tau protein was expressed through lentivirus transduction was an increase in tubulin acetylation observed (Fig. S1). This result agrees well with the previous data.

On the other hand, the effect of tau on tubulin acetylation could be mimicked by the addition of tubacin, a specific HDAC6 inhibitor. Figure 3c shows this effect of treatment with tubacin, which suggests that tau can act as a HDAC inhibitor (see below).

Increase in tubulin acetylation in human cells over-expressing tau protein

To determine whether an increase in tau expression can cause the opposite effect of tau depletion on tubulin acetylation, we have compared the levels of tubulin acetylation in HEK293 cells, which do not express tau, with those of tau-expressing HEK293 cells. We tested human cells as the binding of human tau to human HDAC6 may be stronger than that of the mouse homologue (Ding et al. 2008) (Yang and Seto 2008).

Figure 4 shows a clear increase in the level of tubulin acetylation in tau-expressing HEK293 cells compared with the level of tubulin acetylation in tau lacking cells. Again, addition of tubacin to HEK293 cells mimics the effect of tau expression (Fig. 4c).

Figure 4.

 Increased level of tubulin acetylation in cells over-expressing tau protein. (a) The levels of acetylated tubulin in human HEK293 or tau-expressing HEK293 cells were measured by WB together with those of tau protein, total tubulin or actin. (b) The quantitation of the results in Panel a are shown. (c) The levels of acetylated tubulin in human HEK293 cells in the absence or presence of 2 μM tubacin are also shown. (d) Quantitation of (c) is indicated. The average of at least three separate determinations is indicated. *< 0.01 compared with HEK293 cells.

Tau is a deacetylase inhibitor

A consequence of the tau–HDAC6 interaction could be the direct inhibition of the deacetylase activity of HDAC6. To test this possibility, we produced a human cell extract (HeLa) that contains HDAC6 (Fig. 5b, inset) and mixed it with human tau. Then we have determined whether tau has an inhibitory effect on deacetylase activity (Fig. 5). To determine whether tau action on HDAC activity was mainly through HDAC6, this enzyme was immunodepleted from the extract. Figure S2 shows that after incubation with HDAC6 antibody (see Materials and methods), a decrease in total HDAC activity to around 52% of the initial activity was found. When tau was added to the HDAC6 depleted fraction, an additional inhibition of HDAC activity was observed (Fig. S2).

Figure 5.

 Tau inhibits HDAC activity. (a) Scheme of human HDAC6 molecule showing its different domains and protein binding sites. (b) HeLa cell extract, showing a high HDAC activity, was incubated with increasing amounts of monomeric tau protein or tau assembled into PHFs. Inset shows the presence of HDAC6 in the HeLa cell extract. (c) Effect of increasing amounts of tau or tubacin addition on purified HDAC6 (0.1 mg/mL). The average of at least three separate experiments is indicated. *< 0.05, **< 0.01 compared with control.

Histone deacetylase 6 has two catalytically active deacetylase domains (Fig. 5a), DAC1 and DAC2, which both have HDAC activity. However, only DAC2 is responsible for alpha-tubulin deacetylase activity, the activity that is specifically inhibited by tubacin (Haggarty et al. 2003).

Tau protein decreased deacetylase activity in a concentration-dependent manner (Fig. 5). A maximum inhibition of 65% of total HDAC6 was found at 0.2 mg/mL of tau protein. At higher concentrations the effect was partially reversed and less HDAC inhibition was found (not shown). As mentioned in Fig. 1, intracellular NFTs composed of tau and an increase in the level of protein acetylation in AD neurons was found. Therefore, we determined whether aggregated tau (from PHF) had an effect on HDAC activity. Figure 5b indicates that PHF-tau inhibits HDAC activity with a slighter higher efficiency that monomeric tau. Although, the HDAC6 interaction with monomeric tau takes place through the microtubule binding region of tau protein (Ding et al. 2008) which is partially masked in tau aggregates (Perez et al. 1996), we suggest that an alternate interaction of PHF-tau with HDAC6, because of the presence of ubiquinated tau in PHF protein, may result from the interaction of ubiquinated residues with the HDAC6 C-terminal region.

All these data suggest that tau behaves as an HDAC6 inhibitor, although we have ruled out the possibility that tau inhibits other HDACs. Finally, we tested the action of purified tau on purified HDAC6. Figure 5c shows that HDAC6 deacetylase activity decreases in the presence of tau protein as well as in tubacin (positive control).

Tau inhibits HDAC6 function in aggresome formation

Histone deacetylase 6 has two functions: a deacetylase activity and a role as a component of the aggresome complex (Kawaguchi et al. 2003). It has been suggested that the two functions are coupled (Caron et al. 2005). It is known that polyubiquinated proteins, together with a dynein complex, bind to HDAC6 prior to aggresome formation (Kawaguchi et al. 2003), a process that may be required for autophagic degradation of aberrant aggregated proteins like huntingtin (Iwata et al. 2005). Indeed, when the proteasome, a complex that degradaes cytosolic proteins, was inhibited, undegraded proteins could interact with other proteins, leading to the formation of a complex of protein aggregates, aggregates that can be destroyed via the aggresome (autophagic) pathway (Pandey et al. 2007). After formation, the pre-aggresomal particles were transported close to the nuclear membrane, forming a cellular garbage bin-like structure known as the aggresome; one component of this pathway is HDAC6 (Kawaguchi et al. 2003). Thus, inhibition of the proteasome may induce autophagy if HDAC6 is not inhibited. Formation of protein aggregates that will be located close to the nuclear membrane (aggresomes) can be induced by proteosome inhibition resulting in autophagy.

In HEK cells we followed autophagy induction by proteosome inhibition (Rubinsztein 2007) using LC3II, an established indicator of autophagy activation (Kabeya et al. 2000). Thus, LC3II levels were assessed in HEK cells (in the presence or absence of tau expression), exposed to the proteasome inhibitor MG262. Figure 6 shows an increase in the level of LC3II in those cells treated with the proteasome inhibitor not expressing the tau protein. In contrast, LC3II was not increased in tau-expressing cells, suggesting that the effect of tau on HDAC6 prevents autophagy induction. Also, the formation of protein aggregates located close to the nuclear membrane (aggresome) was reduced in cells over-expressing the tau protein, as determined by the localization of LC3II in the presence of MG 262 (Fig. S3).

Figure 6.

 Differences in LC3II levels between cells lacking or over-expressing tau protein, upon proteasome inhibition. (a) Autophagy was induced by proteosome inhibition after addition of MG262 to HEK293 or tau-expressing HEK293 expressing cells. After incubation for 8 h in the presence of the inhibitor, the cells were disrupted and the levels of LC3II were measured by western blot. The levels of HDAC6, tau, and actin were measured in parallel. (b) Quantitation of the data for LC3II in HEK293 and HEK293 tau-expressing cells is shown. *< 0.05 compared with control cells without MG262 treatment.


In the present study, we described a novel function for the tau protein through its interaction with HDAC6. Tau acts as both a deacetylase inhibitor and as an inhibitor of the aggresome pathway. Using in vitro assays of deacetylase activity or by measurement, from the amount of acetylated tubulin by immunofluorescence and western blot in cells from mouse lacking tau protein or in human cells over-expressing tau protein, we observed that tau inhibits the deacetylase activity of HDAC6. These results are consistent with previous observations (Takemura et al. 1992), indicating increased tubulin acetylation in cells transfected with tau cDNA and with those demonstrating that over-expression of tau in transgenic mice dramatically increases the level of acetylated microtubules, especially in neuronal cell bodies (Nuydens et al. 2002). These results may explain the increase of acetylated tubulin in the brain of AD patients, where an increase of total tau was observed (Kopke et al. 1993). However, they disagree with the observation that in AD brains, tubulin acetylation is reduced in neurons containing NFTs (Hempen and Brion 1996). However, that study did not investigate acetylated tubulin in non-AD age-matched controls.

As our marker to follow acetylation and deacetylation has been tubulin, and one of the main deacetylases described for acetylated tubulin is HDAC 6 (Zhang et al. 2003), we determined whether deacetylase interacts with the tau protein. Our results indicate an interaction between tau and HDAC6 proteins in mouse cells. This observation supports a recent report indicating a similar interaction in human cells (Ding et al. 2008).

Nevertheless, the consequences of that interaction were more evident in human than in mouse cells. This may be because of a weaker binding of tau to mouse HDAC6 than to human HDAC6 as the tau binding region in human HDAC6 is only partially present in the mouse (Yang and Seto 2008). Four classes of HDAC have been identified (Gregoretti et al. 2004), HDAC6 is in class II, although the key catalytic residues have been conserved in class I, class II, and class IV of HDAC (Fantin and Richon 2007). Class III consists of homologues of yeast Sir 2, which is also a tubulin deacetylase (North et al. 2003; Southwood et al. 2007); however only HDAC6 is expressed in most neurons whereas the Sir 2 homologue is targeted to myelin sheaths (Southwood et al. 2007). Thus, HDAC6 is likely to be the tubulin deacetylase inhibited by the tau protein. This interaction, HDAC6 with tubulin, has been reported to be in a region located between two HDAC6 functional domains, those involved in tubulin deacetylase activity and in the binding to ubiquitinated proteins (Ding et al. 2008).

One of the consequences of the binding of tau to HDAC6 is a decrease in HDAC6 activity. As HDAC6 is the main tubulin deacetylase (Hubbert et al. 2002), in the presence of tau which inhibits the deacetylation of tubulin, the level of tubulin acetylation should increase as we observed. In a tauopathy like AD in which the level of total cytoplasmic tau increases (Kopke et al. 1993), an increase in tubulin acetylation should be expected as we found in this work. In addition, the binding of tau to HDAC6 will also impair autophagy of protein aggregates, measured by the level of LC3II (see model in Fig. 7). This, in turn, would result in the accumulation of tau aggregates as shown in Fig. 1.

Figure 7.

 Tau inhibition of HDAC6. Tau inhibition of HDAC6 will result in an increase in the level of acetylated tubulin and in a decrease in proteasome-induced autophagy, measured by LC3II levels, both features could take place in AD. It should be noted that among the compounds that could inhibit the proteasome, to induce autophagy, is beta amyloid peptide (Tseng et al. 2008). MT, Microtubule; MTBD, Microtubule Binding Domain.

Thus, we propose a novel function for the tau protein as an inhibitor of HDAC6. Interestingly, the consequences of increased tau on HDAC6 activity would not be expected to directly affect neuronal viability as HDAC6 knock-out mice are viable and develop without neurological abnormalities (Zhang et al. 2008). However, HDAC 6 inhibition may result in a toxic gain of function, for example, through the formation of aberrant protein aggregates that will accumulate when HDAC6 is inhibited (Iwata et al. 2005). Thus, we can hypothesize that the tau pathology found in AD results first in an increase in the amount of tau protein that is not bound to microtubules (Kopke et al. 1993). This increase in free cytoplasmic tau protein could facilitate its aggregation into PHF and, in addition, tau could inhibit HDAC6 resulting in an increase in tubulin acetylation and a decrease in autophagy that will result in the accumulation of tau aggregates. The tau aggregates will remain in those neurons, neurons that are characterized by the presence of acetylated tubulin and tau aggregates. Thus, we found that the presence of tau could regulate the different HDAC6 functions, resulting in an increase in tubulin acetylation and in a decrease in degradation of aggregated protein.


This work was supported by grants from Spanish Plan Nacional, Comunidad de Madrid, Fundación Botín, CIBERNED, and an institutional grant Fundación Areces.