Iron (III) induces aggregation of hyperphosphorylated τ and its reduction to iron (II) reverses the aggregation: implications in the formation of neurofibrillary tangles of Alzheimer's disease

Authors


Address correspondence and reprint requests to Ryong-Woon Shin, Department of Neurological Science, Tohoku University School of Medicine, 2–1 Seiryo-machi, Sendai 980–8575, Japan. E-mail: shin@mail.cc.tohoku.ac.jp

Abstract

Iron as well as aluminum is reported to accumulate in neurons with neurofibrillary tangles (NFTs) of Alzheimer's disease (AD) brain. Previously we demonstrated that aluminum (III) shows phosphate-dependent binding with hyperphosphorylated τ (PHFτ), the major constituent of NFTs, thereby inducing aggregation of PHFτ. Herein we report that iron (III) can also induce aggregation of soluble PHFτ. Importantly, for the aggregation of PHFτ to occur, iron in the oxidized state (III) is essential since iron in the reduced state (II) lacks such ability. Furthermore, iron (III)-induced aggregation is reversed by reducing iron (III) to iron (II). Thus the iron-participating aggregation is mediated not only by τ phosphorylation but also by the transition of iron between reduced (II) and oxidized (III) states. Further incubation of insoluble PHFτ aggregates isolated from AD brain with reducing agents produced liberation of solubilized PHFτ and iron (II), indicating that PHFτ in association with iron (III) constitutes the insoluble pool of PHFτ. These results indicate that iron might play a role in the aggregation of PHFτ leading to the formation of NFTs in AD brain.

Abbreviations used
AD

Alzheimer's disease

BCIP

5-bromo-4-chloro-3-indolylphophate p-toluidine

βME

β-mercaptoethanol

DFO

desferrioxamine mesylate

DTT

dithiothreitol

NBT

nitro blue tetrazolium

NFT

neurofibrillary tangle

PBS

phosphate-buffered saline

PHFs

paired helical filaments

PHFτ

τ protein in paired helical filaments

PVDF

polyvinylidene difluoride

SDS–PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

TTBS

1% Tween 20 in Tris-buffered saline.

One of the major histopathologic abnormalities of brains with Alzheimer's disease (AD) is the neurofibrillary lesions consisting of neurofibrillary tangles (NFTs), dystrophic neurites associated with senile plaques, and neuropil threads (Goedert et al. 1997). These lesions contain filamentous structures called paired helical filaments (PHFs) and related straight filaments, which are formed from the microtubule-associated protein τ of the CNS. τ constituting the PHFs, often referred to as PHFτ, is characterized by several post-translational protein modifications. In the normal adult brain, τ is dynamically phosphorylated to a state resulting from balanced activities of protein kinases and phosphatases, while in AD brain τ is in a highly and stably sustained state of phosphorylation (Matsuo et al. 1994). Another prominent feature of PHFτ is its ability to form insoluble aggregates, in contrast to the unusual solubility of normal τ (Cleveland et al. 1977; Lee et al. 1988).

To understand the mechanism(s) involved in the formation of NFTs, it is important to know the causal relationship between phosphorylation and aggregation of τ. In the brain the association of τ with microtubules is normally regulated by altering its phosphorylation state; τ in a non-phosphorylated form associates with microtubules and, on phosphorylation, dissociates from microtubules (Drechsel et al. 1992). In AD brain hyperphosphorylated τ shows diminished association with microtubules and is aggregated to assemble into PHFs (Bramblett et al. 1993; Yoshida and Ihara 1993). Together with the insoluble pool, there is a soluble pool of hyperphosphorylated τ that is not aggregated into assembled PHFs (Kopke et al. 1993), indicating that aggregation of τ follows its phosphorylation. Taken together, it is commonly assumed that increased phosphorylation dissociates τ from microtubules, following which hyperphosphorylated τ is aggregated and assembled into PHFs. The effects of phosphorylation on the aggregation of τ into assembled filamentous structures remain to be elucidated. In vitro studies have produced conflicting results, describing inhibitory (Schneider et al. 1999) and stimulatory (Alonso et al. 2001) effects of phosphorylation of τ on its aggregation.

On the other hand, several factors that promote the aggregation of τ have been reported (Scott et al. 1993; Ledesma et al. 1994; Shin et al. 1994; Yan et al. 1994; Miller and Johnson 1995; Gamblin et al. 2000). Among them is aluminum, an element accumulated in NFTs (Markesbery and Ehmann 1993; McLachlan 1995). Specifically, aluminum is involved in the phosphorylation-dependent aggregation of τ (Shin et al. 1994; Shin et al. 1995; Murayama et al. 1999; Shin 2001). Iron may have a similar role because iron and aluminum share the same cationic trivalency and ligands including phosphoproteins (Webb et al. 1973), and have common mechanisms for both extracellular and intracellular distribution (Roskams and Connor 1990). Moreover iron is the major redox active transition metal distributing between a redox-active state (II) and redox-stable state (III) (Reif 1992; de Silva and Aust 1993). A disturbed oxidative balance, whereby reactive oxygen production exceeds cellular antioxidant defenses, leads to oxidative stress which is possibly involved in the pathogenesis of AD (Markesbery 1997; Behl 1999; Christen 2000; Smith et al. 2000).

This paper reports a biochemical study of the in vitro interaction between iron and τ proteins. Iron (III) but not iron (II) shows phosphorylation-dependent aggregation of PHFτ, and the iron (III)-induced aggregation is reversed by reducing iron (III) to iron (II). Iron might therefore be implicated in the aggregation of τ leading to the formation of NFTs under pathological conditions favoring hyperphosphorylation of τ and increased oxidative stress with concomitant decreased reducing potential.

Materials and methods

Isolation of PHFτ

Aqueous-soluble and sarkosyl-insoluble fractions of PHFτ were prepared from post-mortem AD brain and normal adult human τ from post-mortem control brain, as described previously (Shin et al. 1993; Shin et al. 1994; Murayama et al. 1999). The aqueous-soluble fraction of PHFτ is defined as that remaining in the supernatant after centrifugation at 100 000 g for 30 min. The protein sample of aqueous-soluble PHFτ contains co-existing normal adult τ, while that of sarkosyl-insoluble PHFτ is removed from co-existing normal τ. The protein concentration of PHFτ was determined using a semiquantitative immunoblot assay by comparison with a recombinant human τ 441 amino-acid isoform (Wako, Osaka, Japan) as a standard. The standard recombinant τ at a concentration of 2 mg/mL and the test sample containing PHFτ were serially diluted, subjected to 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and transferred to polyvinylidene difluoride (PVDF) membranes (Amersham, Poole, UK). The immunoblot membranes were probed with phosphorylation-independent monoclonal antibody T14 (Kosik et al. 1988; Trojanowski et al. 1989) and alkaline phosphatase-linked anti-mouse immunoglobulins, and were developed with nitro blue tetrazolium (NBT)/5-bromo-4-chloro-3-indolylphophate p-toluidine (BCIP) salt (Promega, Madison, WI, USA). The amount for the detection limit of the standard was obtained and this value was assigned to PHFτ loaded in an amount giving the detection limit.

Assessment of binding of metal ions to PHFτ on immunoblots

Protein samples containing PHFτ and normal τ were resolved by 8% SDS–PAGE and transferred to PVDF membranes. A series of membrane strips including each of the protein samples was preincubated in solutions of various metal salts for 2 h at room temperature (25°C). The metal salts were used at a concentration of 10 mmol/L and included FeCl3[iron (III)], AlCl3[Al (III)], FeCl2[iron (II)], CaCl2, CuSO4, MgCl2, and ZnSO4 (Wako). The metal solutions were freshly prepared for each experiment in 0.2 mol/L glycine-HCl buffer, pH 3.0. Glycine buffer was employed because high solubility of free iron (III) and aluminum (III) was achieved at and around pH 3.0 (Martin 1986). After washing in TTBS buffer (0.1% Tween 20, 50 mmol/L Tris-HCl pH 7.6, 0.5 mol/L NaCl), the membrane strips were probed using monoclonal antibodies including phosphorylation-dependent PHF1 (Greenberg et al. 1992; Lang et al. 1992) and AT8 (Innogenetics, Zwijndrecht, Belgium) (Goedert et al. 1993; Goedert et al. 1995), dephosphorylation-dependent Tau1 (Cedar Lane, Hornby, Canada) (Binder et al. 1985;Papasozomenos and Binder 1987) and phosphorylation-independent T14. To analyze dephosphorylated PHFτ, in some experiments the membrane strips including PHFτ were treated with 12 U/mL of Escherichia coli alkaline phosphatase type III (Sigma, St Louis, MO, USA) at 65°C for 3 h. Previously we developed the ‘chelating autoclave method’ (Murayama et al. 1999), which allows aluminum removal by pretreating tissue sections or immunoblot membranes with trivalent cationic chelators in the hydrated autoclaving procedure (Shin et al. 1991; Shin et al. 1992). To determine whether the chelating autoclave method can remove iron (III) that is added exogenously, presumably to bind to PHFτ on immunoblots, the membrane strips were treated by autoclave heating in a solution of 0.1 mol/L desferrioxamine mesylate (DFO) (Sigma), pH 4.6, or 0.1 mol/L EDTA, pH 4.6, as described (Murayama et al. 1999). The effects of the metal ions used on the access of the primary antibodies to their antigens as well as alterations caused by the chelating autoclave method were visualized by alkaline phosphatase-linked anti-mouse immunoglobulins with NBT/BCIP as a color development substrate.

Assessment of in vitro interaction of iron (III) versus aluminum (III) and iron (II) with PHFτ

A protein sample including each of the aqueous-soluble PHFτ (∼3–10 ng/µL) and normal adult τ was incubated at 37°C for 1 h with or without 1 mmol/L FeCl3, AlCl3 or FeCl2. The reaction mixture (9 : 1) of PHFτ dissolved in 0.1 mol/L morpholine ethansulfonic acid buffer pH 6.8 and 10 mmol/L FeCl3, AlCl3 or FeCl2 dissolved in 0.2 mol/L glycine-HCl, pH 3.0, yielded a final pH of 6.6. Following the incubation, each sample was centrifuged at 100 000 g for 30 min. To avoid contamination by trace amounts of the supernatant, the resulting pellets were carefully washed with the corresponding buffers and cleared by centrifugation at 100 000 g for 10 min. The supernatants and washed pellets were boiled in Laemmli sample buffer with or without β-mercaptoethanol (βME) and were subjected to 8% SDS–PAGE, followed by immunoblot analysis using PHF1 and Tau1. Additional examination for in vitro interaction between PHFτ and iron (III) was performed with FeCl3 at concentrations of 0, 0.01, 0.1, 1 and 10 mmol/L.

Iron (III)–maltol solutions were also prepared, according to a method commonly used for preparing neutral aluminum (III) solutions (Langui et al. 1990). An aqueous solution of 20 mmol/L FeCl3 or AlCl3 was mixed at a volume ratio of 1 : 1 with 20 mmol/L maltol (Wako) dissolved in 0.1 mol/L phosphate-buffered saline (PBS), pH 6.8, to give 10 mmol/L iron (III)–maltol or aluminum (III)–maltol, pH 6.8. Using these neutral complexes instead of FeCl3 and AlCl3, pH 3.0, we repeated some of the experiments described above.

Electron microscopy

To investigate the ultrastructural morphology of PHFτ aggregates induced by iron (III) or aluminum (III), the protein sample containing the aqueous-soluble PHFτ was further purified by extraction with 2.5% perchloric acid (Lindwall and Cole 1984). The acid buffer was replaced by PBS, pH 7.0, using Centricon (Amicon, Beverly, MA, USA). The sample containing purified PHFτ (∼10 ng/mL) was incubated at 37°C with or without 10 mmol/L FeCl3 or AlCl3 for 1 or 72 h. The reaction mixtures were spread on carbon-coated grids, negatively stained with 1% phosphotungstic acid, pH 7.0, and examined under a Hitachi H-7000 electron microscope (Hitachi, Hitachinaka, Japan) with an acceleration voltage of 75 kV.

Assessment of effects of reductants on the interaction of iron (III) versus Al (III) with PHFτ

Membrane strips containing PHFτ that had been incubated with a solution of FeCl3 or AlCl3 as described above were incubated with 20% βME in 0.5 mol/L Tris-HCl buffer, pH 6.8, or an equivalent amount of dithiothreitol (DTT) for 1 h at room temperature. After washing with TTBS, the membrane strips were probed with PHF1. The protein sample containing aqueous-soluble PHFτ was incubated with 1 mmol/L FeCl3 or AlCl3 at 37°C for 1 h and then centrifuged at 100 000 g for 30 min. Following washing as described above, the resulting pellets were incubated with 20% βME in 0.5 mol/L Tris-HCl buffer, pH 6.8, for 1 h and then centrifuged again at 100 000 g for 30 min. The supernatants and pellets obtained from the first centrifugation and those from the second centrifugation were subjected to non-reducing SDS–PAGE, followed by immunoblot analysis using PHF1.

Assessment of reduction of iron (III) to iron (II)

We tested the abilities of βME and DTT to reduce iron (III) to iron (II) by using an established qualitative analysis based on the colorimetric Turnbull blue reaction. This method allows detection of iron (II) by its reaction with potassium ferricyanide to form the complex iron (II) ferricyanide (Turnbull blue) (Culling 1976). To detect the residual amount of iron (III) after completion of its reduction, we used Perls' Prussian blue reaction by which iron (III) is detected by its reaction with potassium ferrocyanide to form iron (III) ferrocyanide (Prussian blue) (Culling 1976). Some 10 µL of either 10% potassium ferricyanide (for detection of iron (II)) or 10% potassium ferrocyanide [for detection of iron (III)] was added to premixed solutions of FeCl3 and βME to give a total volume of 500 µL with final concentrations of 1 mm FeCl3 and 1% βME. Control experiments were performed in parallel in which one of the three reagents was omitted.

Assessment of in vitro interaction of iron (III) and aluminum (III) with wild versus cysteine-free τ expressed in COS7 cells

The recombinant τ cDNA htau40 encoding the longest 441-amino acid isoform (Goedert et al. 1989; Goedert and Jakes 1990) was tagged by a newly established epitope sequence at its carboxyl terminus. The epitope tag referred to as BE7 consists of seven amino acid residues in the sequence SEPYHHW, which is derived from VP2 protein of human parvovirus B19 (Sato et al. 1991). Cysteine at 291 and 322 residues of htau40 was mutated to alanine according to the manufacturer's instructions (QuickChange, Stratagene, La Jolla, CA, USA). XbaI–SalI fragments encompassing full-length wild τ or doubly mutated (C291A and C322A) τ were ligated into XbaI–SalI-digested pCIneo mammalian expression vector (Promega). COS7 cells grown on 10-cm Petri dishes were transiently transfected with 10 µg of expression construct DNA using liposomal transfection reagent (DOTAP, Roche, Mannheim, Germany). Cells were harvested 48 h after transfection with 0.1 mol/L Tris-HCl buffer, pH 7.6, containing 0.1 mol/L NaCl and protease inhibitor cocktail (Boehringer-Mannheim, Mannheim, Germany) or 50 mmol/L Tris-HCl buffer, pH 8.0, containing 0.1 mol/L NaCl and protease inhibitor cocktail for subsequent treatment with alkaline phosphatase. Following sonication and boiling for 5 min, the cell homogenates were centrifuged at 100 000 g for 30 min. The resultant supernatants containing the τ-BE7 and its dephosphorylated preparation (Shin et al. 1993; Murayama et al. 1999) were examined for an in vitro interaction with 1 mmol/L FeCl3 and AlCl3 as described for PHFτ. The reaction mixture (9 : 1) of τ-BE7 dissolved in 0.1 mol/L Tris, pH 7.6, or 50 mmol/L Tris, pH 8.0, and 10 mmol/L FeCl3 or AlCl3 in 0.2 mol/L glycine-HCl, pH 3.0, yielded a final pH of 7.6. Similar experiments were repeated by incubating wild and mutant τ-BE7 with 1 mmol/L iron (III)–maltol or aluminum (III)–maltol, pH 6.8. To specifically detect τ-BE7 and to distinguish it clearly from endogenous protein, monoclonal antibody BE11 was used (Sato et al. 1991) which recognizes the C-terminal BE7 epitope tag.

Reduction of the sarkosyl-insoluble aggregates of AD brain

Aliquots of the sarkosyl-insoluble pellets were resuspended in 0.5 mol/L Tris buffer, pH 6.8, containing βME or DTT at concentrations of 0, 10, 20, 30, 40 and 50%, and were incubated with shaking at 37°C for 3 h. Following centrifugation at 100 000 g for 30 min, the resultant supernatants and pellets were subjected to 10% SDS–PAGE and immunoblot analysis with PHF1. The sarkosyl-insoluble pellet suspended in 0.5 mol/L Tris buffer, pH 6.8, was divided into two portions according to its differential degrees of aggregation by stepwise centrifugation, initially at 20 000 g for 5 min and subsequently at 100 000 g for 30 min. Aliquots of the first and second pellets were resuspended in 0.5 mol/L Tris buffer, pH 6.8, with or without 20% βME and were examined as described above. The 100 000 g supernatant resulting from the incubation without or with 20% βME was also analyzed for the presence of iron (II) and iron (III) by the colorimetric assays.

Results

Phosphate-dependent binding of iron (III) to PHFτ on immunoblots

We tested whether pre-incubation of the immunoblot membrane strips with iron (III), iron (II) and several other divalent metal ions induces immunoreactive alterations of PHFτ. Iron (III) was found to abolish the characteristic immunoreactive bands and smear corresponding to PHFτ when probed with phosphorylation-dependent antibodies PHF1 (Fig. 1a, lanes 1 and 2) and AT8 (Fig. 1c, lanes 1 and 2). The iron (III)-induced abolishment was reversed when the membrane strips were treated by autoclave heating in a solution of DFO (Fig. 1a, lane 3) or EDTA (Fig. 1a, lane 4), but remained unchanged when treated without DFO or EDTA (not shown). Dephosphorylation of PHFτ immobilized on immunoblots caused elimination of the epitopes recognized by PHF1 and AT8 and concomitant acquisition of that by dephosphorylation-dependent Tau1 (Fig. 1d). The dephosphorylated PHFτ did not undergo immunoreactive alterations following pre-incubation with FeCl3 or AlCl3 as revealed by Tau1 (Fig. 1d). Similarly, no immunoreactive alteration revealed by Tau1 was observed for normal adult human τ (Fig. 1e). When probed with phosphorylation-independent T14, only minimal immunoreactive alterations were observed for PHFτ (Fig. 1f). All these immunoreactive alterations elicited by iron (III) were identical to those produced by aluminum (III) (Figs 1b, d and e) as was shown previously (Shin et al. 1994; Murayama et al. 1999). Pre-incubation with iron (II), calcium (II), copper (II), magnesium (II) or zinc (II) induced no changes in immunoreactivity of PHFτ as recognized by PHF1 (not shown). Thus the abolishment of the phosphorylated epitopes in PHFτ was seen selectively for iron (III) and aluminum (III). Pre-incubation with iron (III) or aluminum (III) did not induce dephosphorylation of PHFτ as revealed by its negative acquisition of the Tau1 epitope (not shown), excluding the possibility that dephosphorylation of PHFτ is the cause of the abolishment of immunoreactivity. Furthermore, the retrieval of PHFτ immunoreactivity as revealed by PHF1 and AT8 excludes the occurrence of dephosphorylation of PHFτ. These findings indicate that the abolishment of immunoreactivity of the phosphorylated epitopes in PHFτ is due to masking of these epitopes through binding of iron (III) and aluminum (III). Thus iron (III) and aluminum (III) are metal ions that uniquely exhibit phosphorylation state-dependent binding with PHFτ on immunoblots.

Figure 1.

Phosphate-dependent binding of iron (III) and aluminum (III) to PHFτ on immunoblots and its abolition by trivalent cationic chelators. Aliquots of each of the aqueous-soluble PHFτ (a, b and c), the sarkosyl-insoluble PHFτ (d and f) and autopsy-derived normal adult human τ (e) were subjected to SDS–PAGE and transferred to PVDF membranes. In (d) some of the membrane strips (lanes 2, 4–6) were pretreated with E. coli alkaline phosphatase. Each of the membrane strips was incubated without FeCl3 and AlCl3 (lane 1 in a, b, c; lanes 1–4 in d; lane 1 in e and f) or with either FeCl3 (lanes 2–4 in a, lane 2 in c, lane 5 in d, and lane 2 in e and f) or AlCl3 (lanes 2–4 in b; lane 3 in c, lane 6 in d, lane 3 in e). Following washing with TTBS, some of the membrane strips were treated by the hydrated autoclave procedure using DFO (lane 3 in a and b) or EDTA (lane 4 in a and b). The membrane strips were probed with PHF1 (lanes 1–4 in a, b; lanes 1–2 in d), AT8 (lanes 1–3 in c), Tau1 (lanes 3–6 in d, lanes 1–3 in e), and T14 (lanes 1–2 in f). Note that incubation with FeCl3 or AlCl3 totally abolished the immunoreactive bands and smear corresponding to PHFτ when probed with PHF1 and AT8. The iron (III)- and aluminum (III)-induced abolishment of immunoreactivity was reversed when the membrane strips were treated by the chelating autoclave method using DFO or EDTA. In contrast, the immunoreactive bands of dephosphorylated PHFτ or normal τ were preserved with FeCl3 and AlCl3 when probed with Tau1. The immunoreactivity of dephosphorylated PHFτ (lane 5 in d) and normal τ (lane 2 in e) appeared weakly attenuated following the incubation with FeCl3 due to yellowish staining of the whole membrane strips with the yellow solution of FeCl3. The position of Mr markers is indicated in kilodaltons at the left of panels (a) and (d).

In vitro iron (III)-induced aggregation of PHFτ

Alterations in solubility of PHFτ (∼3–10 ng/µL) were examined following in vitro incubation with FeCl3 versus AlCl3 and FeCl2 at a fixed concentration of 1 mmol/L. In previous studies (Shin et al. 1994; Murayama et al. 1999) and here aluminum (III) was shown to induce aggregation of PHFτ(Fig. 2a, lanes 7–10) but not of dephosphorylated PHFτ or normal adult τ (Fig. 2b, lanes 7–10). Iron (III) also formed aqueous-insoluble aggregates of PHFτ (Fig. 2a, lanes 1–6) but not of normal adult τ (Fig. 2a, lanes 1–6). There were differences noted, however, in the electrophoretic pattern of the PHFτ aggregates between reducing and non-reducing conditions. With βME in the sample buffer, the iron (III)-induced aggregates were capable of entering SDS gel with the characteristic electrophoretic pattern corresponding to that of PHFτ, indicating SDS-solubility of the aggregates (Fig. 2a, lanes 3 and 4). When βME was excluded from the sample buffer, the PHFτ aggregates did not enter the gel and remained on top of the stacking gel (Fig. 2a, lanes 5 and 6), indicating SDS-insolubility. On the other hand, the aluminum (III)-induced aggregates remained on the top of the stacking gel with some smear regardless of the reducing and non-reducing conditions (Fig. 2a, lanes 7–10). In contrast to iron (III), iron (II) did not cause significant alterations in the solubility and electrophoretic pattern of PHFτ or normal adult τ (Fig. 2c). When various concentrations of FeCl3 ranging from 0.01 to 10 mmol/L were used, the amount of PHFτ formed into aggregates and recovered in the pellets increased with the concentrations of FeCl3 (not shown). Thus the aggregation of PHFτ by iron (III) occurs in a dose-dependent manner as shown for aluminum (III) (Shin et al. 1994; Murayama et al. 1999). Similar in vitro experiments repeated using 1 mmol/L iron(III)–maltol or aluminum (III)–maltol, pH 6.8, instead of FeCl3 or AlCl3, pH 3.0, gave the same results (not shown).

Figure 2.

Immunoblot analysis of the in vitro effects of iron (III) versus aluminum (III) and iron (II) on the solubilities of τ. (a, b) Aliquots of each of the soluble PHFτ and normal τ were incubated without the metal salts (lanes 1–2 in a, b; lanes 1–2, 5–6 in c) or with FeCl3 (lanes 3–6 in a and b), AlCl3 (lanes 7–10 in a, b) or FeCl2 (lanes 3–4, 7–8 in c), followed by centrifugation at 100 000 g for 30 min. For each of these samples, the resulting supernatants (S) (lanes 1, 3, 5, 7, 9 in a and b; lanes 1, 3, 5, 7 in c) and pellets (P) (lanes 2, 4, 6, 8, 10 in a and b; lanes 2, 4, 6, 8 in c) were subjected to SDS–PAGE under reducing (R) (lanes 1–4, 7–8 in a and b; lanes 1–8 in c) or non-reducing (NR) (lanes 5–6, 9–10 in a, b) conditions. After transfer to PVDF membrane, the gel replicas were probed with PHF1 (lanes 1–10 in a, lanes 1–4 in c) or Tau1 (lanes 1–10 in b, lanes 5–8 in c). Both FeCl3 and AlCl3 induced aggregation of PHFτ but not normal τ, while FeCl2 did not alter the solubilities of PHFτ and normal τ. Note that there are differences in the electrophoretic pattern between the iron (III)- and aluminum (III)-induced aggregates. Under non-reducing conditions both the iron (III)-and aluminum (III)-induced aggregates were incapable of entering SDS gel (lanes 6, 10 in a), while under reducings condition the iron (III)- but not aluminum (III)-induced aggregates were capable of entering SDS gel (lanes 4, 8 in a). The arrow in panel (a) indicates the position of the top of the stacking gel. The position of Mr markers is indicated in kilodaltons at the left of each panel.

The sample containing the aqueous-soluble PHFτ was examined by electron microscopy. Before addition of FeCl3 or AlCl3, the sample showed no evidence of protein aggregation or filamentous assembly into PHFs and straight filaments. On incubation with FeCl3 or AlCl3, PHFτ formed granular amorphous aggregates with no detectable filamentous structures (Fig. 3).

Figure 3.

Electron micrographs showing granular aggregates formed of PHFτ incubated with FeCl3 at 37°C for 72 h. Some of the granules appeared in a cluster (arrowhead). Scale bar, 100 nm.

Reductants modulate interaction of iron (III) with PHFτ

Further experiments were performed to test the effects of the reductants in modulating interactions between PHFτ and iron (III) as observed on immunoblots and in vitro. Incubation of the membrane strips including PHFτ with iron (III) or aluminum (III) was followed by additional incubation with or without βME. The abolishment PHFτ immunoreactivity induced by iron (III) was retrieved with βME (Fig. 4a), while the aluminum (III)-induced abolishment remained unchanged under the same conditions (Fig. 4b). Similar results were produced when βME was replaced by DTT (not shown). Thus iron (III) that is bound to PHFτ on immunoblots may be removed under reducing conditions.

Figure 4.

Immunoblot analysis of the effects of βME on the interaction of iron (III) versus aluminum (III) with PHFτ on immunoblots and in vitro. (a, b) Each of the membrane strips including PHFτ was incubated without (lane 1 in a and b) or with either FeCl3 (lane 2–3 in a) or AlCl3 (lane 2–3 in b), followed by further incubation with βME (ME) (lane 3 in a and b). Note that the iron (III)-induced abolishment of PHFτ immunoreactivity was reversed following incubation with βME (lane 3 in a), while the aluminum (III)-induced abolishment of immunoreactivity remained unchanged under the same conditions (lane 3 in b). (c) Aliquots of the aqueous-soluble PHFτ were incubated without (lanes 1–2) or with FeCl3 in duplicate (lanes 3–6), followed by centrifugation at 100 000 g for 30 min. One of the resultant pellets from the duplicate incubation with FeCl3 was further incubated with βME and centrifuged again at 100 000 g for 30 min. The supernatant (S; lane 1) and pellet (P; lane 2) obtained from control incubation without FeCl3 and βME, the supernatant (S; lane 3) and pellet (P; lane 4) from the incubation with FeCl3 alone, and the supernatant (PS; lane 5) and pellet (PP; lane 6) from the incubation of iron (III)-induced pellet with βME were subjected to non-reducing SDS–PAGE and immunoblot analysis using PHF1. Note that iron (III)-induced aggregates remained on the top of the stacking gel before incubation with βME (lane 4, indicated by an arrow), while these insoluble aggregates were resolubilized and recovered in the supernatant following the incubation with βME (lane 5). The position of Mr markers is indicated in kilodaltons at the left of panel (a) and (c).

To determine whether the reductants can recover the solubility of PHFτ once formed as aggregates by iron (III), pellets resulting from the incubation of the PHFτ samples with FeCl3 or AlCl3 were further incubated in 0.5 mol/L Tris-HCl buffer, pH 6.8, with or without βME. Following centrifugation at 100 000 g for 30 min, the resulting supernatants and pellets were subjected to non-reducing SDS–PAGE. Iron (III)-induced insoluble aggregates were recovered in the supernatants following subsequent incubation with βME (Fig. 4c), while the aluminum (III)-induced aggregates remained in the pellets under the same conditions (not shown). Use of DTT instead of βME in the experiments gave the same results (not shown). These findings indicate that iron (III)-induced aggregates of PHFτ were solubilized in the presence of reductants.

Conversion of iron (III) to iron (II) by reductants

The solubilization of iron (III)-induced aggregates must have occured through the action of the reductants directed either to iron (III) or to PHFτ. As the aluminum (III)-induced aggregates of PHFτ were resistant to the solubilization by the reductants, it is less likely that the reductants exert their action on PHFτ. The possibility was considered that the solubilization is attributable to the reduction of iron (III) to iron (II). This assumption was supported by the observation that the reductants were indeed capable of converting iron (III) to iron (II) as qualitatively demonstrated by the Turnbull blue reaction (Fig. 5). These results indicate that the iron (III)-induced aggregates undergo solubilization by reducing iron (III) to iron (II).

Figure 5.

Turnbull blue reaction used to demonstrate βME-assisted reduction of iron (III) to iron (II). Addition of potassium ferricyanide K3[Fe(CN)6] to a premixed solution of FeCl3 and βME (ME) formed royal blue product iron (II) ferricyanide, indicating generation of iron (II) (tube 4). No residual iron (III) was detected after FeCl3 solution was mixed with βME as revealed by a negative Perls' Prussian blue reaction (not shown). Control experiments in which one of the three reagents was omitted did not form positive blue products (tubes 1–3).

Iron (III)/aluminum (III)-induced aggregation occurs independently of the presence of cysteine residues in τ

The cytosolic reducing environment normally prevents the formation of disulfide bonds by maintaining the cysteine residues of cytosolic proteins in reduced form. However, in in vitro milieu, a recombinant τ construct containing three internal repeats was shown to form PHF-like structures through disulfide bridges at the cysteine residue (Schweers et al. 1995). The solubilization of the PHFτ aggregates described above might be attributed to the action of the reductants directed at PHFτ itself if the iron (III)-induced aggregation of PHFτ occured by cross-linking through disulfide bonds at cysteine residues of the τ molecule. To test this possibility we produced a τ construct in which two cysteine residues were mutated to alanine and examined the ability of iron (III) to induce aggregation for such cysteine-free τ molecules. Immunoblot analysis with monoclonal antibody BE11 revealed that wild and cysteine-free τ-BE7 expressed in COS7 cells migrated as two major and several minor bands of 64–72 kDa (Fig. 6a and c, lane 1), all of which merge, on dephosphorylation, to a single band with the highest electrophoretic mobility (Fig. 6b, lane 1). The τ-BE7 reacted with Tau1 but not with PHF1 or AT8 (not shown), indicating lack of phosphorylation at Ser202, Thr205, Ser396 and Ser404, all of which form part of the known phosphorylation sites of PHFτ (Hasegawa et al. 1992; Morishima-Kawashima et al. 1995). Thus the wild and mutant τ-BE7 are heterogeneously phosphorylated at certain but at fewer sites than PHFτ. Incubation with FeCl3 or AlCl3 induced wild τ-BE7 to form aqueous-insoluble aggregates, which were capable of entering SDS gel regardless of the reducing and non-reducing conditions (Fig. 6a, lanes 3–6, 9–12). The iron (III)-induced aggregates recovered their aqueous-solubility by subsequent incubation with βME (Fig. 6a, lanes 7–8), while aluminum (III)-induced aggregates remained unchanged under the same conditions (Fig. 6a, lanes 13–14). Such aggregating effects of iron (III) and aluminum (III) were lost on dephosphorylation of the wild τ-BE7 (Fig. 6b). The aggregates formed from τ-BE7 expressed in COS7 cells appeared to differ from those of PHFτ in terms of SDS-solubility; the former was SDS-soluble and the latter SDS-insoluble. The differential SDS-solubility of iron (III)/aluminum (III)-induced aggregates might reflect the different phosphorylation states between the COS7-expressed τ-BE7 and PHFτ. This is again a finding in support of the phosphate-dependent interaction of τ with iron (III) and aluminum (III). Thus the wild τ-BE7 displays an interaction with iron (III) and aluminum (III) that is not exactly identical to, but significantly comparable to, that observed for PHFτ. Therefore the experimental system using τ-BE7 expressed in COS7 cells serves as a tool to give insight into the molecular mechanisms for the interaction of τ with Fe (III) and Al (III). The cysteine-free τ was then investigated for its interaction with iron (III) and aluminum (III). Following incubation with FeCl3 or AlCl3, the mutant τ-BE7 was converted to form aqueous-insoluble and SDS-soluble aggregates regardless of the reducing and non-reducing conditions (Fig. 6c, lanes 1–6 and 9–12). Additional incubation with βME produced resolubilization of iron (III)-induced aggregates (Fig. 6c, lanes 7–8) but not of aluminum (III)-induced aggregates (Fig. 6c, lanes 13–14). Thus there were no differences noted in the interaction with iron (III) and aluminum (III) between the wild and cysteine-free τ-BE7. Similar in vitro incubation experiments repeated using 1 mmol/L iron (III)–maltol or aluminum (III)–maltol, pH 6.8, instead of 1 mmol/L FeCl3 or AlCl3, pH 3.0, gave the same results (not shown). These results indicate that the iron (III)/aluminum (III)-induced aggregation of τ occurs independently of the presence of cysteine residues in the τ molecule.

Figure 6.

In vitro interaction of wild τ versus cysteine-free τ expressed in COS7 cells with iron (III) and aluminum (III). The protein samples including each of the longest isoform of wild τ (Wild τ) (a) and its dephosphorylated preparation (Dp. Wild τ) (b) and mutant cysteine-free τ (Cys-free τ) (c) were incubated without (lanes 1–2 in a, b and c) or with either FeCl3 (Fe) (lanes 3–8 in a, c; lanes 3–4 in b) or AlCl3 (Al) (lanes 9–14 in a and c; lanes 5–6 in b) for 1 h at 37°C. One of the triplicate samples incubated with FeCl3 (lanes 7–8 in a, c) and of those with AlCl3 (lanes 13–14 in a and c) were further incubated with 5% βME (ME) for 15 min at 37°C. Following centrifugation at 100 000 g for 30 min, the resultant supernatants (S) and pellets (P) were subjected to 10% SDS–PAGE under reducing (R) (lanes 1–4, 9–10 in a and c; lanes 1–6 in b) or non-reducing (NR) (lanes 5–8, 11–14 in a and c) conditions. The immunoblots were analyzed with monoclonal antibody BE11, which recognizes the BE7 epitope tag connected to the C-terminus of the COS7-expressed τ. Incubation with FeCl3 or AlCl3 induced wild τ to form aqueous-insoluble and SDS-soluble aggregates regardless of the reducing and non-reducing conditions (a). The iron (III)-induced aggregates were resolved to become aqueous-soluble by subsequent incubation with βME, while the aluminum (III)-induced aggregates remained unchanged under the same conditions. Such effects of iron (III) and aluminum (III) were eliminated on dephosphorylation of τ (b). Following incubation with FeCl3 or AlCl3, cysteine-free τ also underwent aggregation and its reversal by βME in exactly the same way as the wild τ. The arrowhead in panel (b) indicates the bands corresponding to alkaline phosphatase, which was included to dephosphorylate the wild τ and was recognized by the NBT/BCIP color development substrates for the enzyme. A fraction of alkaline phosphatase consistently formed aggregates by AlCl3 for some unknown reason. The position of Mr markers is indicated in kilodaltons at the left of each panel.

Reductant-assisted extraction of PHFτ from sarkosyl-insoluble aggregates of AD brain

All the results described above indicate that iron in oxidized form (III) participates in the aggregation of soluble PHFτ, and that the aggregation is reversed by converting iron to its reduced form (II). To confirm that such effects of iron indeed underlie the aggregation of PHFτ occuring in AD brain, we performed experiments to solubilize and extract PHFτ from insoluble PHFτ aggregates by treating these aggregates with reductants. Incubation of the sarkosyl-insoluble aggregates with βME yielded solubilized PHFτ that was recovered in the supernatant. The solubilization of PHFτ was dose dependent, at concentrations of βME ranging from 0 to 30% (Fig. 7a, lanes 1–8), and was maximally obtained at 20–30% βME; there was no apparent increase in the solubilized PHFτ at higher concentrations up to 50% (Fig. 7a, lanes 9–12). A considerable amount of solubilized PHFτ was derived from the less aggregated fraction (Fig. 7b, lanes 5–8), and a much smaller amount from the highly aggregated fraction (Fig. 7b, lanes 1–4). Use of DTT in an equivalent amount to βME gave the same results (not shown). Thus the sarkosyl-insoluble aggregates of PHFτ can be resolved by incubation with reductants. The supernatant was also analyzed by a colorimetric assay to examine whether iron is dissociated from the sarkosyl-insoluble aggregates. The supernatant resulting from the incubation with βME contained iron (II) but not iron (III) as revealed by a positive Turnbull blue reaction and negative Prussian blue reaction, while that without βME did not contain both (not shown). Thus reduction of the sarkosyl-insoluble aggregates liberated iron (II) concomitantly with solubilized PHFτ.

Figure 7.

βME-assisted extraction of PHFτ from the sarkosyl-insoluble aggregates of AD brain. (a) Aliquots containing the same amount of sarkosyl-insoluble PHFτ were incubated with βME at concentrations of 0% (lanes 1–2), 10% (lanes 3–4), 20% (lanes 5–6), 30% (lanes 7–8), 40% (lanes 9–10) and 50% (lanes 11–12)% and then centrifuged at 100 000 g for 30 min. The resultant supernatants (S) and pellets (P) were subjected to SDS–PAGE and immunoblot analysis with PHF1. (b) Aliquots of the highly aggregated fraction of the sarkosyl-insoluble PHFτ obtained from centrifugation at 20 000 g (20k Plt) and the less aggregated fraction from the subsequent centrifugation at 100 000 g (100k Plt) were incubated without or with 20% βME and centrifuged again at 100 000 g for 30 min. The resultant supernatants (S) and pellets (P) were subjected to SDS–PAGE and immunoblot analysis with PHF1. Half of the sample derived from each of the resulting pellets was loaded on to the gel (a, b). Note that incubation of the sarkosyl-insoluble pellet with βME yielded solubilized PHFτ in a dose-dependent manner at concentrations of βME ranging from 0 to 30%. βME-assisted solubilization of PHFτ from the less aggregated 100 000 g pellet was considerably greater than that from the highly aggregated 20 000 g pellet (b).

Discussion

The present study has established a new pathobiological feature of iron that might extend its putative role in the pathogenesis of AD. Specifically, iron (III) has phosphate-dependent ability to bind with hyperphosphorylated τ(PHFτ), thereby inducing its aggregation, a property originally designated to aluminum (III) (Shin et al. 1994; Murayama et al. 1999). It should be noted, however, that the resultant PHFτ aggregates generated by iron (III) differed from those produced by aluminum (III) in that the aqueous-solubility of the former was recovered by βME and DTT. The reversed aggregation of PHFτ was attributed to the conversion of iron (III) to iron (II) by the reductants because (1) the absence of the reductants eliminated the reversed aggregation, (2) the reductants were shown to reduce iron (III) to iron (II), and (3) iron in reduced form (II) lacks the ability to induce aggregation of τ. Nevertheless there might be an alternative pathway for the solubilization of PHFτ aggregates; the reducing agent might exert its action on PHFτ itself when the iron (III)-induced aggregates were formed by cross-linking through disulfide bonds in the τ molecule, as has been reported to occur in vitro under specific conditions (Schweers et al. 1995). This possibility was unlikely, however, because the wild and mutant cysteine-less τ molecules that were expressed in COS7 cells as heterogeneously phosphorylated forms underwent aggregation by iron (III) in an identical manner. Thus the iron-participating aggregation is mediated not only by τ phosphorylation but also by the transition of iron between reduced (II) and oxidized (III) states.

Such in vitro features of iron were further shown to underlie the in vivo aggregation of τ occuring in AD brain. Reduction of the sarkosyl-insoluble aggregates prepared from AD brain with βME or DTT liberated solubilized PHFτ and iron (II). Although it remains to be confirmed that the dissociated iron (II) was derived from PHFτ, the concomitant liberation of PHFτ and iron (II) following the reducing procedure suggests that complexes of PHFτ and iron (III) constitute the insoluble pool of PHFτ aggregates. These findings, however, do not necessarily exclude other factors in the aggregation of τ, as a significant amount of the insoluble PHFτ remained unresolved after completion of its extraction with the reductants. Among such factors implicated in τ aggregation might be specific post-translational modifications such as enzymatic transglutamination (Miller and Johnson 1995), glycation (Ledesma et al. 1994; Yan et al. 1994), fatty acid oxidation (Gamblin et al. 2000) as well as aluminum (III) interaction (Scott et al. 1993; Shin et al. 1994).

By the Fenton reaction the redox-active form (II) of iron generates reactive oxygen species, accompanied by its oxidation to the redox-stable form (III) (Reif 1992; de Silva and Aust 1993). When the oxidative balance is disturbed such that the production of reactive oxygen species exceeds cellular antioxidant defenses, oxidative stress could occur. In recent years considerable data have indicated that the AD brain is under increased oxidative stress, which may have a role in the pathogenesis of neuronal degeneration in AD (Markesbery 1997; Behl 1999; Christen 2000; Smith et al. 2000). One piece of the direct evidence supporting increased oxidative stress in AD includes accumulation of iron in neurons with NFTs (Good et al. 1992; Smith et al. 1997). The lesion-associated iron was shown to consist of both iron (II) and iron (III), and to be capable of catalyzing a Fenton reaction in situ (Smith et al. 1997). Normally the intracellular pool of physiological iron is mostly stored within ferritin as a ferric (III) form and is released upon reduction from ferritin as a ferrous form (II). The present study identified PHFτ as a possible pathological ligand for iron (III) in affected neurons with NFTs. The pretreatment of AD brain sections with the trivalent cationic chelator DFO produced immunoreactive enhancement in PHFτ of NFTs and other neurofibrillary lesions (Murayama et al. 1999), indicating the association of trivalent cations with these lesions. Considering the increased level of iron in NFTs (Good et al. 1992; Smith et al. 1997), the present observations of in vitro and in vivo interactions between PHFτ and iron (III) indicates that iron is associated with PHFτ in these lesions. Electron microscopy revealed that iron-induced aggregation of hyperphosphorylated τ contributes to amorphous granules but not to filamentous species related to PHFs and straight filaments. Further investigation of how such amorphous aggregates play a role in the formation of NFTs remains to be addressed.

The present study has explored biochemical features of the in vitro interaction of isolated PHFτ with iron (III) in concentrations of 1 mmol/L and ranging from 0.01 to 10 mmol/L at physiological pH. The normal human brain contains iron metal at 42.1–131 µg per g fresh weight (Thompson et al. 1988), which approximately corresponds to 0.8–2.3 mmol/L. Therefore the concentrations of iron (III) used in this study could cover the range of physiological concentrations of iron (III) in the human brain. Thus the in vitro interaction presented here is likely to reflect a substantial aspect of the biologically relevant in vivo phenomenon. This study provides a hypothetical scenario for a new pathobiological role of iron. In the iron homoeostasis under regulation, generation of hyperphosphorylated τ might elicit its interaction with redox-stable iron (III), but the interaction could be reversed by endogenous associated reductants. In situations of abnormally deregulated homoeostasis of iron, where oxidative stress is increased possibly with decreased reducing potential, the interaction of iron (III) with hyperphosphorylated τ might lead to the formation of insoluble aggregates, which might persist irreversibly in the affected neurons. As generation of reactive oxygen species by iron (II) is accompanied by its oxidation to iron (III), oxidative stress and iron (III)-induced aggregation of τ might be linked to each other. The dual role of iron as a mediator for oxidative stress as well as for τ aggregation should provide further insight into the importance of this metal in the pathogenesis of AD.

Acknowledgements

The authors thank Dr Peter Davies (Department of Pathology and Neuroscience, Albert Einstein School of Medicine, Bronx, NY, USA) for providing monoclonal antibody PHF1, Drs John Q. Trojanowski and Virginia M-Y. Lee (Department of Pathology and Laboratory Medicine, University of Pennsylvania, Philadelphia, PA, USA) for monoclonal antibody T14, Dr Michel Goedert (Medical Research Council, Laboratory of Molecular Biology, Cambridge, UK) for cDNA clone htau40, and Dr Theo P. A. Kruck (University of Toronto, Toronto, Canada) for reading the manuscript. This research was supported by a Grant-in Aid for Scientific Research on Priority Areas (C)-Advanced Brain Science Project-from the Ministry of Education Culture, Sports, Science and Technology, Japan (R-WS), by a Grant-in Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan (R-WS), and by the Research Grant for Longevity Sciences (11C-01) from the Japanese Ministry of Health, Labour and Welfare (R-WS).

Ancillary