Unraveling the biochemical and genetic alterations that control the aggregation of protein tau is crucial to understand the etiology of tau-related neurodegenerative disorders. We expressed wild type and six clinical frontotemporal dementia with parkinsonism (FTDP) mutants of human protein tau in wild-type yeast cells and cells lacking Mds1 or Pho85, the respective orthologues of the tau kinases GSK3β and cdk5. We compared tau phosphorylation with the levels of sarkosyl-insoluble tau (SinT), as a measure for tau aggregation. The deficiency of Pho85 enhanced significantly the phosphorylation of serine-409 (S409) in all tau mutants, which coincided with marked increases in SinT levels. FTDP mutants tau-P301L and tau-R406W were least phosphorylated at S409 and produced the lowest levels of SinT, indicating that S409 phosphorylation is a direct determinant for tau aggregation. This finding was substantiated by the synthetic tau-S409A mutant that failed to produce significant amounts of SinT, while its pseudophosphorylated counterpart tau-S409E yielded SinT levels higher than or comparable to wild-type tau. Furthermore, S409 phosphorylation reduced the binding of protein tau to preformed microtubules. The highest SinT levels were found in yeast cells subjected to oxidative stress and with mitochondrial dysfunction. Under these conditions, the aggregation of tau was enhanced although the protein is less phosphorylated, suggesting that additional mechanisms are involved. Our results validate yeast as a prime model to identify the genetic and biochemical factors that contribute to the pathophysiology of human tau.
Protein tau is a microtubule-associated protein that is widely present in neurons and is important for the assembly and stabilization of microtubules. Tau is expressed as six isoforms derived from a single gene by alternative mRNA splicing. These isoforms differ by 29 or 58 amino acid N-terminal insertions encoded by exons 2 and 3 (0N, 1N and 2N isoforms, respectively) and by an extra C-terminal domain encoded by exon 10 (3R and 4R, respectively). The N-terminal inserts have no defined function, while the C-terminal domains are denoted as microtubule-binding domains, whereby isoform tau-4R binds with a greater affinity to microtubules than tau-3R. Besides, by alternative mRNA splicing, the physiological role of protein tau is further regulated by dynamic phosphorylation through the interplay of a variety of protein kinases and phosphatases (Sergeant et al., 2008; Gendron & Petrucelli, 2009; Iqbal et al., 2009). The longest isoform, i.e. tau-2N/4R, has 85 putative phosphorylation sites, of which 71 have been reported to become phosphorylated, although many are only demonstrated in isolated systems. The majority of these sites are present in the proline-rich and C-terminal regions flanking the microtubule-binding domains (Sergeant et al., 2008; Tremblay et al., 2009). Phosphorylation of tau, especially by MARK kinases within the microtubule-binding repeats, triggers disengagement of tau to regulate microtubule assembly dynamics. This is thought not only to decrease microtubule stability, important for the establishment of cell polarity, but also to be required for proper intracellular trafficking along the axons, because tau attached to microtubules may form a physical obstacle for vesicles moving along the microtubule tracks (Mandelkow & Mandelkow, 1998; Drewes, 2004; Gendron & Petrucelli, 2009). In addition, tau phosphorylation is developmentally regulated, as it is substantially higher in fetal brain and decreases with age (Avila et al., 1994).
The importance of phosphorylation of tau gained momentum following the finding that hyperphosphorylated forms of tau are the major constituents of intraneuronal tau deposits that define neurodegenerative diseases such as Alzheimer's disease (AD), Pick's disease, progressive supranuclear palsy and frontotemporal dementia, collectively termed tauopathies (Sergeant et al., 2008; Gendron & Petrucelli, 2009; Iqbal et al., 2009). In these tauopathies, hyperphosphorylation of tau is believed to abrogate the physiological function of tau and to introduce conformational changes that direct tau to form paired helical filaments (PHF), which in turn further aggregate into neurofibrillary tangles (NFT) (Mandelkow et al., 2003; Drewes, 2004). Interestingly, different states of tau phosphorylation appear to be associated with different stages of disease progression, as demonstrated for AD (Kimura et al., 1996; Augustinack et al., 2002). Both increased phosphorylation and decreased dephosphorylation may thus influence the aggregation process (Sergeant et al., 2008; Gendron & Petrucelli, 2009; Iqbal et al., 2009), but the exact kinases and phosphatases responsible in vivo and the phosphorylation sites targeted to induce aggregation of tau remain largely elusive.
Direct genetic evidence that implicated protein tau in neuronal loss and dementia came with the discovery of a diverse range of dominant mutations associated with frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17). To date, >40 pathogenic mutations have been identified, some of which are intronic and affect mRNA splicing to alter the ratio of expression of tau-4R over tau-3R isoforms, while others are exonic and cause single amino acid substitutions or deletions that affect the ability of tau to regulate microtubule dynamics (Rademakers et al., 2004; Gendron & Petrucelli, 2009). Analysis of postmortem brain samples of FTDP-17 patients confirmed the accumulation of hyperphosphorylated tau fibrils and suggested that mutations promote tau phosphorylation site-specifically (Spillantini et al., 1998; van Swieten et al., 1999). This is further supported by the analysis of various transgenic models (Götz et al., 2009). However, studies performed to comparatively analyze the effect of tau mutations on phosphorylation and its causal relation to aggregation failed to produce consistent data. For example, when expressed in human neuroglioma H4 cells, both the tau-V337M and the tau-R406W mutants displayed a drastically reduced phosphorylation at several epitopes as compared with wild-type tau-2N/4R (DeTure et al., 2002), while in CHO cells, reduced phosphorylation was seen with tau-R406W and not with tau-V337M (Matsumura et al., 1999; Vogelsberg-Ragaglia et al., 2000). On the other hand, phosphorylation of recombinant tau in vitro with rat brain extract revealed that the R406W and V337M mutations enhanced phosphorylation at most epitopes (Alonso Adel et al., 2006). This study also demonstrated that distinct tau mutants display a similar enhanced propensity to aggregate, but a more recent study presented a different picture and, for instance, failed to confirm enhanced aggregation for tau-R406W (Chang et al., 2008).
The many advantages of a less complex cellular system to study in detail the molecular mechanisms leading to hyperphosphorylation and aggregation of protein tau led us to express and study human tau in yeast. Previously, we reported that the humanized yeast cells recapitulated robustly the most important aspects of a tauopathy, i.e. hyperphosphorylation, conformational change and self-aggregation of wild-type tau-2N/3R or tau-2N/4R isoforms. The ease and rapidity of genetic modification of yeast cells was then capitalized upon the finding that major pathogenic phosphoepitopes on human tau are produced by Mds1 and Pho85, the yeast orthologues of the two major mammalian tau kinases, i.e. GSK3β and cdk5, respectively (Vandebroek et al., 2005, 2006).
In the present study, we analyzed the expression of a series of clinical FTDP-17 mutants of tau in yeast, to extend the findings of the wild-type tau isoforms and to define the molecular and phenotypic similarities as well as differences among the mutants. This uncovered the importance of serine-409 (S409)-phosphorylation to induce tau self-assembly. In addition, we studied the influence of oxidative stress and mitochondrial dysfunction on tau aggregation and found both conditions to drastically induce the formation of tau filaments in yeast by mechanisms that appear to act additional to tau phosphorylation.
Materials and methods
Yeast strains and media
The Saccharomyces cerevisiae strains used in this study were W303-1A (Mat a leu2-3 112 ura3-1 trp1-1 his3-11 15 ade2-1 can1-100 GAL SUC), BY4742 (S288C Mat αhis3Δ1 leu2Δ0 lys2Δ0 ura3Δ0) and the isogenic single- or double-deletion mutants mds1Δ, pho85Δ, mds1Δpho85Δ, sod2Δ and rim1Δ. The single pho85Δ and double mds1Δpho85Δ mutants were obtained by disruptions of PHO85 using PCR-derived cassettes containing KanMX, HIS3 or TRP1 markers as described previously (Wach et al., 1994). The sod2Δ and rim1Δ strains were obtained from the genome-wide yeast deletion collection (Winzeler et al., 1999).
The plasmids expressing tau-2N/4R and tau-P301L were described previously (Vandebroek et al., 2005, 2006). Other FTDP-17 mutations, i.e. G272V, N279K, ΔK280, V337M and R406W, as well as the synthetic mutants S409E and S409A were introduced into the tau-2N/4R isoform using site-directed mutagenesis. All tau constructs were fully sequenced to ensure that no additional mutations were introduced during amplification. Human Gsk3β was inserted as an EcoRV–PvuII fragment and cdk5 as a BamHI fragment into the yeast expression vector pKT10 (Tanaka et al., 1990).
Standard yeast transformation techniques were applied (Gietz et al., 1992). Cells were grown at 30 °C in a selective minimal medium containing 2% glucose. To induce ROS production and oxidative stress, FeSO4 was added to a final concentration of 20 mM to early exponential cells (OD=0.5) and the cultures were allowed to grow till the mid-exponential phase (OD=2) before harvesting and determination of the sarkosyl-insoluble tau (SinT) fraction.
Yeast cells were grown until OD600 nm=2.0 and then harvested by centrifugation. Samples were prepared as described previously, separated with standard sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), both under reducing and nonreducing conditions, or native PAGE and further analyzed using standard Western blotting techniques (Towbin et al., 1979; Vandebroek et al., 2005). The tau antibodies used in this study are listed in Table 1. Relative immunoreactivity was determined by densitometric comparison (software: imagemaster1d, Amersham; tina 2.08, Raytest) and normalized for total tau amounts as measured by pan-tau antibody Tau-5.
Table 1. List of tau antibodies and their specificities
Sarkosyl insolubility levels were determined as described before (Vandebroek et al., 2005). Cells were homogenized with glass beads and extracts were centrifuged (30 min, 20 000 g). After N-lauroyl-sarkosine was added to the supernatant (final concentration 1%) and samples were incubated for 1 h, sarkosyl-soluble and -insoluble fractions were separated by centrifugation (150 000 g, 40 min). Relative quantities were determined by Western blotting using antibody Tau-5.
Microtubule interaction studies
The method to purify phosphorylated tau from yeast has been described previously (Vandebroek et al., 2005). Also, the methods to prepare taxol-stabilized microtubules from pig tubulin and to perform tau–microtubule-binding filter assays have been reported earlier (Vandebroek et al., 2006).
Statistical analysis was performed using one-way or two-way anova, followed by multiple comparison tests of Tukey and Fisher.
Phosphoepitope mapping reveals altered phosphorylation profiles in two FTDP-17 tau mutants
The human tau-2N/4R wild-type isoform and the clinical mutants tau-G272V, tau-N279K, tau-ΔK280, tau-P301L, tau-V337M and tau-R406W, all in the tau-2N/4R isoform, were expressed constitutively from high-copy-number plasmids in the W303-1A wild-type yeast strain. Their expression was evaluated using Western blot analysis with a polyclonal tau antiserum (data not shown) and the pan-tau monoclonal antibody (Mab) Tau5. This revealed that the proteins are expressed as a mixture of isoforms varying in molecular weights from 64 to 72 kDa (Fig. 1). These isoforms originate from differences in phosphorylation as demonstrated previously by dephosphorylation studies (Vandebroek et al., 2005). When quantified over all subforms and normalized vs. the levels of endogenous yeast alcohol dehydrogenase II, the expression levels of wild-type and mutant tau proteins appeared to be similar (data not shown). Interestingly, a detailed comparison of the electrophoretic mobility patterns detected with Tau5 for the wild-type tau protein and the clinical mutants revealed that the most phosphorylated isoform, with the lowest electrophoretic mobility, was only very weakly present in the case of tau-P301L, while being completely absent in the case of tau-R406W (Fig. 1a). This was even more prominent when immunodetection was performed with the antibody MC1, which detects a discontinuous epitope characteristic for tauopathy (Jicha et al., 1997) (Fig. 1a). Combined, our data suggested that the tau-P301L and tau-R406W mutants displayed phosphorylation patterns deviating from the other FTD mutants. Importantly, neither the wild-type tau-2N/4R nor the clinical FTDP-17 mutants markedly influenced the growth properties of the transformed yeast cells (data not shown).
To assess the phosphorylation status of protein tau in detail, phosphoepitope scanning was performed with Mabs Tau-1, AT8, AT100, AT180, AT270, AD2 and PG5 (Fig. 1a; Table 1). Normalized quantification against the total tau levels, measured with Tau5, demonstrated that the epitopes defined by Tau-1, AT8, AT180 and AT270 were similarly present on wild-type and mutant tau (data not shown). In contrast, the AD2 epitope (P-S396, P-S404) was markedly lower in tau-R406W, while the PG5 epitope (P-S409) was almost absent in tau-R406W (P<0.001) and drastically reduced in tau-P301L (P<0.05) (Fig. 1a, c, d).
We previously identified the protein kinase Mds1 as the valid orthologue of GSK3β by its capacity to generate the AD2 and PG5 epitopes on tau-2N/4R. We then also reported that phosphorylation of the AD2 and PG5 epitope on tau-2N/4R was negatively influenced by Pho85, the orthologue of mammalian cdk5 (Vandebroek et al., 2005, 2006). This then led us to the hypothesis that Pho85/cdk5 influences the phosphorylation of tau indirectly by acting as a negative regulator of Mds1/Gsk3β, in line with data obtained in transgenic mice (Hallows et al., 2003; Plattner et al., 2006). This hypothesis is also based on our current observation that the capacity of human GSK3β to restore the AD2 and PG5 epitopes on tau2N/4R in the yeast mds1Δpho85Δ double-deletion strain is strongly reduced upon coexpression with human cdk5 (Fig. 1b).
To document the contribution of the two yeast kinases on phosphorylation of mutant tau, we expressed the different FTDP-17 tau mutants in strains lacking either Mds1 or Pho85 and again quantified the immunoreactivity for AD2 and PG5. This confirmed that the level of the AD2 epitope was in general twofold lower for the wild-type and mutant tau in the mds1Δ strain as compared with the wild-type strain (P<0.05), while being slightly higher in the pho85Δ strain (Fig. 1c). The level of the PG5 epitope in the different tau proteins was moderately affected in the mds1Δ strain, but significantly enhanced in the pho85Δ strains (P<0.05 to <0.01), except for tau-R406W (Fig. 1d). Intriguingly, the PG5 immunoreactivity of the mutants in the pho85Δ strain tended to correlate with the distance of their mutation to the C-terminus, the exceptions being tau-P301L, which still displayed limited PG5 immunoreactivity, and tau-R406W, which still lacked this phosphoepitope. Consistent with their hampered C-terminal phosphorylation, both tau mutants also failed to display high levels of the pathologic and aggregation-prone MC1 conformation upon native-PAGE analysis (Fig. 1e).
Phosphorylation of S409 is crucial for aggregation of tau in yeast
We demonstrated previously that a small, but consistent fraction of tau-2N/4R is aggregated in yeast as determined by its sarkosyl insolubility (Vandebroek et al., 2005). Quantitative analysis of this sarkosyl-insoluble fraction (SinT) in the W303-1A wild-type strain suggested that all FTDP-17 mutants aggregated more than wild-type tau-2N/4R. However, the levels remained low, which makes assessment of the biological or pathological significance difficult (Fig. 2a). In line with our previous report, the deletion of MDS1 did not have a major impact on tau insolubility. In contrast, the deletion of PHO85 led to a significant increase in SinT (P<0.001) and this for the wild-type protein as well as all clinical tau mutants (Fig. 2a). Nonetheless, the SinT levels in the pho85Δ strain remained rather low for tau-P301L and tau-R406W, i.e. the two mutants with the lowest PG5 and MC1 immunoreactivity. This suggested that the phosphorylation of S409 could be a determining factor for tau aggregation.
To confirm this hypothesis, we mutagenized the epitope recognized by PG5 and expressed the synthetic tau-S409A mutant and its pseudophosphorylated counterpart tau-S409E in wild-type and pho85Δ strains. Western blotting and subsequent immunodetection with PG5 demonstrated that both synthetic mutants had lost all PG5 immunoreactivity (Fig. 2b). While this result was expected for the tau-S409A mutant, it proves that in the tau-S409E mutant, glutamic acid does not substitute perfectly for a phosphorylated serine residue. Interestingly, immunodetection with AD2 was also affected in the synthetic mutants, even though the amino acid composition was not altered at this epitope (Fig. 2b). Indeed, we observed increased AD2 immunoreactivity for tau-S409E, particularly in the wild-type strain, and a decreased immunoreactivity for tau-S409A. These results indicate that the phosphorylation of both AD2 and PG5 is closely linked and that priming, i.e. the facilitation of phosphorylation of a residue through previous phosphorylation at other residues, also occurs in yeast.
We then monitored SinT levels in both strains. As compared with tau-2N/4R, these levels were drastically reduced for tau-S409A (P<0.001), while it was comparable or even increased for tau-S409E expressed (Fig. 2c, d). These data underscore the importance of S409 phosphorylation for the aggregation of tau. Furthermore, Western blot analysis with Tau5 of the soluble and insoluble tau fractions clearly demonstrated that mutant tau-S409A failed to display the slow-migrating hyperphosphorylated isoform, which we identified previously as being crucial for tau to adopt the MC1 conformation and to aggregate (Vandebroek et al., 2005).
Phosphorylation of S409 is detrimental for tau–microtubule interaction
The commonly accepted hypothesis, although yet to be proven in vivo, posits that hyperphosphorylation of tau disrupts its binding to, and assembly of, microtubules. To elaborate on this assumption, we performed in vitro binding assays with taxol-stabilized porcine microtubules and recombinant tau-2N/4R isolated from the wild-type yeast strain. The qualitative analysis aimed to identify phosphoepitopes that differentiate tau bound to microtubules from soluble tau present in the unbound fraction. Our analysis clearly revealed the presence of epitopes recognized by AT180, AD2, AT8, AT270 and AT100 on protein tau bound to microtubules. In contrast, the epitope defined by PG5 was almost completely absent in the microtubule-bound tau fraction, while being abundantly present in the unbound tau fraction (Fig. 3a, b). This signifies that mainly the tau species that are not phosphorylated on S409 interact with microtubules. Combined with the observations described above, the data led us to conclude that the presence of the PG5 epitope is a decisive factor for tau aggregation and for the disengagement of tau to bind microtubules.
Oxidative stress and mitochondrial dysfunction enhance tau aggregation independent of phosphorylation
Recent studies highlighted the involvement of oxidative stress and mitochondrial dysfunction in the etiology of different tauopathies (Melov et al., 2007; Moreira et al., 2008; Martinez et al., 2009), while others revealed the interplay of tau phosphorylation and oxidative stress for the formation of NFT (Schweers et al., 1995; Takeda et al., 2000; Tremblay et al., 2009). To assess the importance of oxidative stress for SinT formation in our yeast system, cells expressing the wild-type or mutant tau were challenged with the addition of ferrous sulfate in the growth medium to increase free radical production (Stadler et al., 2001). This treatment increased SinT levels drastically in wild-type cells, particularly for the mutants tau-G272V, tau-N279K, tau-ΔK280 and tau-V337M (Fig. 4a). As these SinT increments were considerably higher than those observed with tau-2N/4R, the data demonstrate that FTDP-17 mutations render protein tau more vulnerable to oxidative stress-induced aggregation. Interestingly, the SinT levels of tau-P301L and tau-R406W were also significantly enhanced (P<0.001) and were comparable to, or even higher than, those of tau-2N/4R (Fig. 4a). In cells lacking Pho85, the addition of ferrous ions to the culture medium also induced SinT for the wild-type and mutant tau (Fig. 4b), but this effect was less pronounced than in wild-type cells and, surprisingly, most obvious for tau-P301L (P<0.01) and tau-R406W (P<0.001). Combined, the data suggested that oxidative stress increases the aggregation of tau by mechanisms acting mainly in parallel to tau phosphorylation. In fact, Western blot analysis of extracts prepared from wild-type and pho85Δ cells expressing tau-2N/4R indicated that tau becomes dephosphorylated upon ferrous ion treatment, especially at the AD2 and PG5 epitopes (Fig. 4c). Moreover, the treatment of wild-type cells with ferrous ions significantly induced the formation of SinT by tau-S409A (P<0.001) to a level that is comparable to that found for tau-2N/4R and tau-S409E in untreated cells (Fig. 4d, e). This confirms that the presence of the PG5 epitope is not strictly essential, although it facilitates tau aggregation under oxidative stress conditions.
It has been shown previously that site-specific nitration and oxidation of tau influences its oligomerization and polymerization in vitro (Schweers et al., 1995; Barghorn & Mandelkow, 2002; Landino et al., 2004; Reynolds et al., 2007). Especially, the oxidation of cysteine residues and the formation of intramolecular and intermolecular disulfide bridges were proposed as determining factors for PHF assembly because intramolecular bonds would lead to compacted monomers that do not assemble further, while intermolecular bonds would give rise to stabilized dimers and thereby accelerate further oligomerization (Schweers et al., 1995; Barghorn & Mandelkow, 2002). To assess whether tau dimerization would explain the observed enhanced formation of SinT in yeast cells upon oxidative stress, we performed SDS-PAGE under nonreducing conditions of extracts prepared from wild-type cells. Immunodetection with the pan-antibody Tau5 revealed that in untreated cells, a significant fraction of the wild-type and mutant tau was indeed present as dimers with an approximate molecular weight of 130 kDa (Fig. 4f). In line with our data reported previously for the purification of recombinant tau from yeast strains (Vandebroek et al., 2006), tau-2N/4R, tau-G272V, tau-N279K, tau-ΔK280 and tau-V337M produced markedly more dimers than tau-P301L and tau-R406. This was also the case for the pseudophosphorylated tau-S409E mutant when compared with its nonphosphorylated tau-S409A counterpart. Hence, it appears that dimer formation by protein tau is, at least in part, dependent on conformations determined by phosphorylation in the C-terminal domain. To our surprise, the treatment of the cells with ferrous ions reduced the level of tau dimers significantly and led to an oligomerization pattern characterized by a protein of tau species of approximately 120 kDa and the presence of higher-order tau oligomers. This pattern was also obvious with tau-P301L and tau-R406W and thus appeared to be independent of the capacity of tau to form dimers. Similar to that for SinT, the presence of higher-order tau oligomers was most pronounced with the synthetic tau-S409E mutant and least with the tau-S409A mutant, again indicating that phosphorylation at the PG5 epitope facilitates tau polymerization. Interestingly, we noticed a distinct band at around 35 kDa in each sample isolated from the iron-treated yeast cells, indicating that the stress treatment affected the breakdown or the clearance of tau in yeast.
Because oxidative stress eventually leads to mitochondrial dysfunction, we analyzed two mutants from the yeast genome-wide deletion collection, i.e. the sod2Δ strain that lacks mitochondrial manganese-dependent superoxide dismutase activity (van Loon et al., 1986) and the rim1Δ strain that lacks a single-stranded DNA-binding protein essential for mitochondrial genome maintenance (Van Dyck et al., 1992). Compared with the isogenic wild-type BY4742 strain (Fig. 5a), SinT levels for tau-2N/4R were on average 16-fold higher in the sod2Δ strain (P<0.001) and 10-fold higher in the rim1Δ strain (P<0.01), which is similar to the increments observed upon ferrous sulfate treatment. Also, the increase in SinT for tau-S409A and for tau-S409E was within this range (compare Fig. 5b, c with Fig. 4d, e), again suggesting that the effects of tau-S409 phosphorylation and mitochondrial dysfunction on tau aggregation are synergistic. Note that in none of the experiments described above did we observe strong tau-related growth phenotypes.
An increasing number of mutations in the gene encoding protein tau are linked to the autosomal dominant-inherited form of FTDP-17. Although these mutations have been described to affect tau–microtubuli interaction and tau aggregation, neither their mode of action in vivo nor how exactly the mutations lead to tau filament formation is known (Mandelkow et al., 2003; Sergeant et al., 2008; Gendron & Petrucelli, 2009; Iqbal et al., 2009). Because all pathological aggregates of tau in the inheritable as well as in sporadic cases of tauopathy contain hyperphosphorylated tau isoforms, aberrant or increased phosphorylation is thought to be the necessary step for aggregation. We systematically and comparatively analyzed the biochemistry of six different clinical FTDP-17 tau mutants when expressed in W301-A wild-type yeast or in the isogenic mutant strains lacking Mds1 or Pho85, i.e. the orthologues of the tau kinases Gsk3β or cdk5, respectively. We focused on these kinases because they produce major pathogenic phosphoepitopes on tau as reported previously (Vandebroek et al., 2005). However, note that other yeast kinases phosphorylate tau as well. One such example is the DYRK orthologue, Yak1, which phosphorylates tau at Y18 (data not shown).
In the strains examined, the mutant proteins tau-G272V, tau-N279K, tau-ΔK280 and tau-V337M displayed phosphorylation patterns similar to those of wild-type tau-2N/4R, while tau-P301L and tau-R406W deviated from this pattern. Particularly, tau-R406W was less phosphorylated at the AD2 epitope and both tau-P301L and tau-R406W displayed hampered phosphorylation at the PG5 epitope. The observed specific reduction in immunoreactivity of tau-R406W for AD2 is consistent with reports demonstrating decreased phosphorylation at S396/S404 of this mutant when phosphorylated in vitro with GSK3β and cdk5 (Connell et al., 2001; Sakaue et al., 2005), when expressed in different cell lines (Matsumura et al., 1999; Perez et al., 2000; DeTure et al., 2002) and transgenic mice (Zhang et al., 2004), or when injected in Xenopus oocytes (Delobel et al., 2002). To our knowledge, the phosphorylation of S409 was previously not analyzed systematically in FTDP-17 mutants.
Mechanistically, the reduced AD2 reactivity of tau-R406W can be explained by the close proximity of the R406W mutation to the S396/S404 residues in the AD2 epitopes, i.e. the mutation can interfere directly with the recognition of these residues by GSK3β (Perez et al., 2000; Li & Paudel, 2006; Tatebayashi et al., 2006). Whether the R406W mutation directly interferes with the recognition of S409 to establish the PG5 epitope has not been studied, but our current data make this very likely because the tau-R406W failed to generate this epitope under conditions where other mutants displayed enhanced S409 phosphorylation. Furthermore, the formation of the AD2 and PG5 epitopes is interdependent, as demonstrated with the synthetic S409A and S409E mutants. This suggests that the phosphorylation of S409 (PG5 epitope) primes tau for or at least facilitates the subsequent phosphorylation of S396/S404 (AD2 epitope). These observations are in line with data from the brain of AD patients, demonstrating that the formation of the PG5 epitope on tau is an early event in the pretangle stage and precedes the phosphorylation at S396, which is characteristic for NFT (Kimura et al., 1996).
Besides by direct interference, mutations in tau apparently also affect phosphorylation of more distant residues, most likely by altering the conformation of tau. Evidence for this comes from our observation that the P301L mutation reduces phosphorylation of the PG5 epitope, although the mutation and phosphorylation site are separated by 107 amino acids. Recent studies show that even soluble tau can adopt the so-called ‘paperclip’ conformation, whereby the C-terminus folds over the microtubule-binding domain, with the N-terminus approaching the C-terminal domain (Jeganathan et al., 2006). In that conformation, the C-terminus of tau is in close proximity of the P301 residue and the mutation can be envisaged to affect the paperclip conformation and affect the exposure and phosphorylation of S409. Similar long-distance effects may apply to various extents to other FTDP-17 mutations, explaining our observed inverse relation of PG5 immunoreactivity with the proximity of the mutation to the C-terminal end of protein tau.
Studies that combined FRET analysis with pseudophosphorylation of tau demonstrated that the paperclip conformation of tau became more compact when pseudophosphorylation sites were introduced both N- and C-terminally of the microtubule-binding region and that this generated the aggregation-prone MC1 conformation. Unfolding or opening of the paperclip conformation was shown to occur upon the imbalanced introduction of pseudophosphorylation sites, whereby pseudophosphorylation alone in the N-terminal moves this region away from the C-terminal region and, likewise, pseudophosphorylation alone in the C-terminal region moves this region away from the microtubule-binding repeats (Jeganathan et al., 2008). Based on these data, we predicted that FTDP-17 mutants that solely affect phosphorylation in the C-terminal region would be less aggregated under normal physiological conditions. This was particularly the case for the tau-P301L and tau-R406W mutants because they combined hampered S409 phosphorylation with reduced immunoreactivity for MC1 and reduced SinT production. The clinical counterpart of this finding is reflected by the fact that despite the heterogeneity of clinical symptoms manifested by FTDP-17 mutations, most have an age of onset between 40 and 60 years. The exceptions are patients carrying the P301L and R406W mutations, in which the disease develops later, i.e. after 60 years of age (Spillantini et al., 2000). In this respect, these genetic, primary tauopathies can be mechanistically compared, or even related, with the secondary tauopathy in sporadic AD patients.
Additional support for the importance of the PG5 epitope in determining the switch in the function of tau comes from our observation that tau isoforms lacking this epitope preferably interact with prestabilized porcine microtubules. This extends our previous data on differential microtubule binding of phosphorylated tau isoforms produced in yeast (Vandebroek et al., 2006) Consistent with the notion that FTDP-17 mutations compromise the ability of tau to regulate microtubule dynamics in vitro and in cellulo (Bunker et al., 2006; Han et al., 2009), we demonstrated previously that the binding of the tau-P301L mutant to stabilized porcine microtubules is less sensitive to phosphorylation than the wild-type protein. We then also showed that the mutant readily formed aggregates on the surface of these microtubules (Vandebroek et al., 2006), a property similar to the inherent aggregation propensity of different tau mutants as seen in the diseased brain of tauopathy patients (Gendron & Petrucelli, 2009; Iqbal et al., 2009) as well as in the brain of transgenic mice (Götz et al., 2009). At first glance, the enhanced propensity to aggregate the tau-P301L mutant on prestabilized microtubules in vitro appears to be contradictory to our data, indicating that this mutant formed less SinT than wild-type tau-2N/4R in yeast cells under normal growth conditions. However, the conditions used for in vitro tau–microtubule binding do not compare with those in cellulo, where the reducing environment has important consequences on aggregation as discussed below. In addition, the binding of tau to microtubules was reported to facilitate its oligomerization (Makrides et al., 2003), and this is known to be a driving force for subsequent PHF formation and aggregation (Sahara et al., 2008). Despite considerable efforts, we were unable to demonstrate binding of tau to yeast microtubules. The reason for this might be simply because yeast microtubules differ from their mammalian counterparts by the lack of a typical taxol-binding site (Gupta et al., 2003), which is known to partially overlap with the tau-binding site (Kar et al., 2003). Therefore, SinT produced in our yeast cells is concluded to reflect the aggregation capacity of unbound soluble tau proteins in mammalian systems.
Increasing evidence indicates that oxidative damage and mitochondrial dysfunction significantly impact the development of neurodegenerative disorders, including AD and primary tauopathies (Melov et al., 2007; Moreira et al., 2008; Martinez et al., 2009). We monitored the effect of oxidative stress and mitochondrial dysfunction on SinT formation of wild-type and mutant tau. Our data indicated that both conditions markedly increased tau insolubility through mechanisms that are not strictly dependent on hyperphosphorylation of tau, but rather act in parallel. In wild-type yeast cells, the increase in SinT upon oxidative stress was most pronounced for the FTDP-17 mutants, suggesting that the FTDP mutations render tau more vulnerable to oxidative stress. In the pho85Δ cells, the increment in SinT was less pronounced than that in wild-type cells; however, nonstressed pho85Δ cells already contained elevated levels of insoluble tau, leaving less substrate with the proper conformation for oxidative stress to further induce aggregation. We addressed the question of whether the increased insolubility of tau upon oxidative stress coincided with the enhanced formation of tau dimers as suggested previously by in vitro experiments (Schweers et al., 1995; Barghorn & Mandelkow, 2002; Landino et al., 2004). The answer was apparently negative, because we observed a reduction in the level of tau dimers under conditions of oxidative stress. This was not due to a shift of preexisting tau dimers to higher-order oligomers because those FTDP mutants with low dimer formation, i.e tau-P301L and tau-R406W, also readily formed higher oligomeric structures when challenged with ferrous ions. Although the underlying mechanisms remain to be identified, we noticed that oxidative stress led to the appearance of a distinct degradation product of protein tau of 35 kDa, indicative of altered processing or diminished clearance under this condition. Recent studies highlighted the importance of both ubiquitin-proteasome and autophagy-lysosome pathways for tau clearance and the generation of aggregation-prone tau fragments (Poppek et al., 2006; Dickey et al., 2007; Wang et al., 2009). Consequently, we are interested in examining whether these systems cause oxidative stress-induced aggregation of tau in our model.
We further noticed that oxidative stress led to decreased phosphorylation of specific epitopes in tau-2N/4R in wild-type yeast cells and to a lesser extent in pho85Δ mutant cells. This is consistent with studies showing iron or peroxide ions to induce dephosphorylation of tau in primary neuronal cultures from embryonic rat and mouse brain, as well as in neuroblastoma cells (LoPresti & Konat, 2001; Zambrano et al., 2004; Galas et al., 2006). Those studies implicated the prolyl peptidyl isomerase, Pin1, to facilitate dephosphorylation of P-T231 in tau via PP2A (Galas et al., 2006), and pointed to a role of the cdk5/p35 complex to inactivate inhibitor-2, which acts as a negative regulator of the phosphatase PP1 (Zambrano et al., 2004). Both modulatory pathways are conserved in yeast. The Pin1 function is represented by the yeast orthologue Ess1 (Lu et al., 1996), and our preliminary results, indeed, confirmed that disruption of the Ess1 activity leads to increased hyperphosphorylation of tau (data not shown). The cdk5/p35 complex is functionally equivalent to the yeast Pho85/Plc6,7 complex, which phosphorylates Glc8, the orthologue of mammalian inhibitor-2, thereby controlling the activity of the Glc7 phosphatase, the orthologue of mammalian PP1 (Tan et al., 2003). Hence, it appears that similar mechanisms may govern oxidative stress-induced tau dephosphorylation in yeast and mammalian cells. Notably, the requirement of Pho85 for the activation of PP1 may explain why the oxidative stress-induced dephosphorylation of tau is less extensive in pho85Δ cells as compared with wild-type cells. Whether dephosphorylation of tau is an essential step for tau aggregation upon oxidative stress remains to be clarified. Tau apparently regains its hyperphosphorylation status upon prolonged stress exposure (LoPresti & Konat, 2001) and this seems to involve specific stress-responsive kinases (Tremblay et al., 2009; Su et al., 2010). Interestingly, hyperphosphorylated tau was shown to be a poor substrate for proteasomal degradation (Poppek et al., 2006; Dickey et al., 2007). Therefore, it is feasible that the observed dephosphorylation of tau during oxidative stress reflects an attempt of the cells to improve the clearance of aggregation-prone tau proteins.
Oxidative stress and mitochondrial integrity are closely linked to the life span of yeast cells (Laun et al., 2001; Kaeberlein et al., 2007), and our data thereby relate the aggregation of tau to the problem of aging, which remains the major risk factor in AD and other sporadic tauopathies. However, we did not observe strong tau-related growth phenotypes in fermenting yeast cells treated with ferrous ions or suffering from mitochondrial dysfunctions despite higher SinT levels. Obviously, aggregation of tau is not per se correlated with toxicity, a conclusion that corroborates directly findings in mammalian systems and the human brain (Santacruz et al., 2005; Castellani et al., 2008; Terwel et al., 2008). The assembly of tau into PHF and NFT must then be considered as a mechanism that converts hyperphosphorylated tau into rather inert polymers to protect or even act as a buffer against oxidative damage (Castellani et al., 2008; Moreira et al., 2008; Jaworski et al., 2010).
We thank Peter Davies for advice and for the generous gifts of reagents and Marleen Michels for technical assistance. This investigation was supported by grants from FWO-Vlaanderen, IWT-Vlaanderen, the KU Leuven Research Fund (KU Leuven-BOF and KU Leuven-IOF), KU Leuven R&D, the International Alzheimer's Research Foundation (SAO) and the Marie Curie Graduate School NEURAD.
T.Vh. and T.Vdb. contributed equally to this work.