|Home | About | Journals | Submit | Contact Us | Français|
In Alzheimer disease (AD), rod-like cofilin aggregates (cofilin-actin rods) and thread-like inclusions containing phosphorylated microtubule-associated protein (pMAP) tau form in the brain (neuropil threads) and the extent of their presence correlates with cognitive decline and disease progression. The assembly mechanism of these respective pathological lesions and the relationship between them is poorly understood, yet vital to understanding the causes of sporadic AD. We demonstrate that during mitochondrial inhibition, activated actin-depolymerizing factor (ADF)/cofilin assemble into rods along processes of cultured primary neurons that recruit pMAP/tau and mimic neuropil threads. Fluorescence Resonance Energy Transfer (FRET) analysis revealed co-localization of cofilin-GFP and pMAP in rods, suggesting their close proximity within a cytoskeletal inclusion complex. The relationship between pMAP and cofilin-actin rods was further investigated using actin-modifying drugs and siRNA knockdown of ADF/cofilin in primary neurons. The results suggest that activation of ADF/cofilin and generation of cofilin-actin rods is required for the subsequent recruitment of pMAP into the inclusions. Additionally we were able to induce the formation of pMAP-positive ADF/cofilin rods by exposing cells to exogenous Aβ peptides. These results reveal a common pathway for pMAP and cofilin accumulation in neuronal processes. The requirement of activated ADF/cofilin for the sequestration of pMAP suggests that neuropil thread structures in the AD brain may be initiated by elevated cofilin activation and F-actin bundling that can be caused by oxidative stress, mitochondrial dysfunction or Aβ peptides, all suspected initiators of synaptic loss and neurodegeneration in AD.
Alzheimer disease (AD) is a progressive, degenerative dementia histopathologically characterized by neurofibrillary tangles of tau protein and amyloid-β (Aβ) plaques. In early stages of AD, hyperphosphorylated microtubule associated protein (pMAP) tau forms striated thread-like structures in neurites, so-called ‘neuropil threads’ (Augustinack et al., 2002; Velasco et al., 1998), that correlate with cognitive decline and comprise >85% of end-stage cortical tau pathology (Velasco et al., 1998; Braak et al., 2006; Giannakopoulos et al., 2007).
Tau, like other MAPs, stabilizes neuronal microtubules (MT) and facilitates MT dynamics through its phosphorylation and dephosphorylation (Timm et al, 2003; reviewed in Garcia and Cleveland, 2001). While normal adult neurons exhibit low levels of tau phosphorylation, neurons of AD brain and other tau-related neurodegenerative diseases show high levels of tau phosphorylation at both physiological and pathological disease-specific residues. This tau hyperphosphorylation prevents binding and stabilization of MT and causes abnormal translocation of tau from axonal MT tracks to neuropil thread inclusions, dendritic processes and cell bodies where it accumulates and aggregates (Terry, 1998; Garcia and Cleveland, 2001). The phosphorylation of tau at Ser262 in the microtubule-binding domain is one of the earliest markers of AD neuropathology, readily detected in ‘pre-tangle’ neuropil threads (Augustinack et al., 2002).
Another prominent feature widespread in the AD brain is abnormal aggregates of the actin associated protein cofilin that forms punctuate and rod-like linear arrays through the neuropil (Minamide et al., 2000). Neuronal cofilin plays important roles in learning and memory pathways by modulating actin-rich dendritic spine architecture (Hotulainen et al., 2009; reviewed in Bamburg and Bloom, 2009). The activity of cofilin and related protein actin depolymerizing factor (ADF) is negatively regulated by phosphorylation of the conserved Ser3 by LIM and other kinases and reactivated upon its dephosphorylation by slingshot or chronophin phosphatases (Huang et al., 2008; reviewed in Bamburg and Bloom, 2009) allowing it to actively bind and sever filamentous actin (F-actin), thus regulating actin turnover (Bamburg and Bloom, 2009, Carlier et al., 1997).
ADF/cofilin-actin rods comparable to those observed in the AD brain are inducible in neuronal cell culture through inhibition of mitochondrial ATP generation and other neurodegenerative stimuli such as oxidative stress or exposure to Aβ peptides (Minamide et al., 2000; Maloney et al., 2005; Davis et al., 2009). Since actin dynamics in neurons are purported to use ~50% of total cellular ATP (Bernstein and Bamburg, 2003), ADF/cofilin-actin rods have been proposed to represent an early neuroprotective mechanism during times of transient stress since virtually all ADF/cofilin is sequestered into non-dynamic polymers of ADF/cofilin-actin, inhibiting actin turnover and thereby preserving ATP (Bernstein et al., 2006). While mitochondrial dysfunction has been linked to AD (Smith et al., 2005; Wang et al., 2009), the relationship between mitochondrial dysfunction, the generation of tau inclusions and their relationship to cofilin aggregates remains elusive.
In this study, we aimed to determine the effects of mitochondrial dysfunction on cellular pMAP/tau distribution compared to ADF/cofilin-actin rod distribution (Minamide et al., 2000; Huang et al., 2008). Using primary neuronal cell culture models, we demonstrate that cytoskeletal rods containing ADF/cofilin sequester and bind pMAP. The resulting striated pMAP-positive rods bare striking resemblance to neuropil threads observed in postmortem AD brain labeled with the same pMAP antibody. This process may well represent an early pathogenic event in AD leading to synaptic loss and neurodegeneration.
Mouse monoclonal antibodies are actin (1A4; Dako), β-actin (Abcam), β(III)-tubulin (Abcam), tau phosphorylated at Ser202/Thr205 (AT8; Pierce USA) and Ser262/356 (12E8; Elan USA) (Seubert et al., 1995). The monoclonal antibody 12E8 raised against Ser262-phosphorylated tau, is known to cross react with other phosphorylated MAPs such MAP2c, MAP4 (Timm et al., 2003) and doublecortin (Schaar et al., 2004) since these MAPs also contain the 12E8-specific KXGS motifs in their microtubule-binding domains. However 12E8 has strong affinity for tau, where the microtubule-binding domain contains several KXGS motifs, the target sequence for 12E8 (Timm et al., 2003; Schaar et al., 2004; Seubert et al., 1995; Yoshida and Goedert, 2002). Rabbit polyclonal antibodies are ADF (1439), ADF (D8815; Sigma), Ser3-phosphorylated ADF/cofilin (Shaw et al., 2004), cofilin (C8736; Sigma), actin (A2066; Sigma) and total tau (A0024; Dako). Phalloidin 488 (Invitrogen) was used to visualize F-actin (Invitrogen). Anti-mouse and anti-rabbit secondary antibodies included Alexa Fluor-conjugated 488, 555 and 647 (Invitrogen) for immunofluorescence and horseradish peroxidase-conjugated (Amersham) for immunoblotting. Lyophilized amyloid peptides were from Bachem.
Primary chick neurons were prepared from freshly dissected chicken embryos (E7), as previously described and cultured for 7 days in vitro (d.i.v) on polylysine-coated 30 mm culture dishes or glass coverslips (Goldsbury et al., 2008). Primary human neurons were prepared and cultured for 14–20 days as previously described (Guillemin et al., 2007). Primary (E18) rat hippocampal neurons were cultured for 5 days as previously described (Minamide et al., 2000). For ATP depletion studies, cells were treated with 1–2 µM AM (a mitochondrial complex III inhibitor; Sigma), 3 µM Carbonyl cyanide 3-chlorophenylhydrazone (CCCP: a mitochondrial uncoupling agent; Sigma), or 100 µM hydrogen peroxide (Sigma) in PBS containing 0.5 mM CaCl2 and 1 mM MgSO4. For ADF/cofilin phosphorylation studies, cells were treated for 15 min with 1 µM AM. Actin drugs were used at 1 µg/ml jasplakinolide (Calbiochem 420127, CA) or latrunculin B (Calbiochem 428020, CA). Cells left without a medium change or treated with PBS or DMSO acted as controls. Stock solutions of lyophilized Aβ peptides were solubilized to 2 mM in DMSO, aliquoted and stored at −20°C. Immediately before application to cells, the stock solutions were diluted to 100 µM in PBS and agitated at 1200 rpm on a laboratory shaker for 30 minutes to generate mixtures of polymorphic oligomeric and fibrillar aggregates as previously described (Goldsbury et al., 2000). The peptide assemblages were then applied to cells in 24 well plates at final concentrations of 1 or 2 µM. Control wells were treated with the same volume of DMSO in PBS. After treatments, cells were immediately fixed (cells on coverslips) or prepared for Western blot analysis (plated cells).
Following treatments (in triplicate), primary chick neurons were harvested and assessed for total ATP (ATPLite, Perkin Elmer) using a Victor III Multi-label plate reader. Protein measurements were determined using the identical samples and the Bicinchoninic acid assay (Sigma). Total ATP was normalized against total protein to account for any difference in cell density. Statistical analysis used Prism software (v3.0, GraphPad Software Inc).
Following treatments, cells were lysed and prepared for SDS-PAGE. For equal gel loading, protein concentrations were determined by the Lowry assay (BioRad). Proteins transferred to nitrocellulose membranes were detected with ECL Western Blotting Detection System (Amersham) on a ChemiDoc XRS (BioRad). For analysis, band densities were measured using Image J, background intensity was subtracted and normalized to individual β-actin or tubulin loading controls. An average and s.e.m. for each treatment condition was determined and results presented as a percentage of the mean control band intensities.
Plasmid-mediated expression of wild-type human cofilin has been previously described (Davis et al., 2009). Human cofilin cDNA in a pET vector (a gift from Alan Weeds, MRC Laboratory of Molecular Biology, Cambridge, UK) was modified by PCR with a 5’ PCR primer containing an EcoRI site and a 3’ primer that removed the stop codon and introduced an XmaI site at the 3’ end. EcoRI and XmaI were then used to cut the PCR product and the cDNA was ligated into pEGFP-N1 (Clontech) in frame with the green fluorescent protein to give pCofilin-GFP. A plasmid vector for expressing small interfering RNAs for chick ADF was made by inserting DNA oligonucleotides (Macromolecular Resources, Fort Collins, CO) into a plasmid expression vector (pSuper; (Brummelkamp et al., 2002) containing the H1 polymerase III promoter. The oligonucleotide product from the pol III promoter is a double-stranded hairpin RNA (antisense-linker-sense) that is processed into a functional siRNA in the cell. The modified inserts including the H1 pol III promoter from the pSuper vector were excised and ligated into the pAdTrack vector (He et al., 1998). The siRNA sequence contained within the hairpin used for chick ADF is 5'-GTGGAAGAAGGCAAAGAGATT-3'. A plasmid made identically but which makes a hairpin RNA to silence human Pak2 (5’-GTCTCTGGGTATCATGGCTAT-3’) was used as a transfection control in the chick cells. The ability of the shRNA-expressing plasmid to effectively knockdown ADF in chick cells was tested in cultures of chick skin fibroblasts (Marsick et al., submitted). For plasmid-mediated expression, neuronal cultures at 3 days were transfected using Lipofectamine 2000 (Invitrogen). Four days following transfection, neurons were treated with AM (1 µM or 2 µM for 10 or 20 min).
Cells were fixed with 4% paraformaldehyde at 37°C for 30 minutes, permeabilised with 0.05% Triton X-100 or 100% ice cold methanol, blocked in 5% goat serum and stained for immunofluoresence. To optimize visualization of ADF/cofilin-actin rods, 0.1% EM-grade glutaraldehyde was added to the fixative. Epifluorescence images were obtained on a Zeiss Axioplan 2 microscope, captured with a CCD camera driven by AxioVision software. Single labeled cells/sections were used to check for bleed-through in all double-label immunofluorescence studies. All captured images were converted to Tagged Image Files for subsequent presentation. For FRET analysis, cells (plated on glass bottomed Mattek dishes) were imaged in PBS using a Zeiss LSM 510 and C-Apochromat 40× water immersion objective and the Argon (488 nm) (for GFP donor) and HeNe1 (543) (for Alexa fluor 555 acceptor) laser lines. For photobleaching of acceptor, regions of interest in the transfected cells were bleached using the 543 nm laser line at 100% power. FRET efficiency was calculated from the increase in the fluorescence intensity of the donor after the acceptor was selectively photobleached (Gervásio et al., 2008). The donor fluorescence was measured from at least five bleached and five unbleached rods per cell.
For quantification of cells containing rod structures, cells with rods were counted for each treatment condition from randomly selected fields on each coverslip and cells containing rods structures were then expressed as a percentage of the total cell population or mean number of rods per cell. Treatments and measurements were repeated in triplicate using independently prepared cell cultures. For determination of relative fluorescence intensities for phalloidin or ADF, immunofluorescent intensity was measured using Image J (v1.38x, National Institutes of Health freeware; http://rsb.info.nih.gov/ij). Cells with abnormal nuclei, as indicated by DAPI staining, were excluded from the data. Following adjustment to background, mean and standard error of the mean (s.e.m.) intensities were calculated for each condition. Significance was measured using the student t-test.
Free-floating sections (45 µm) of formalin fixed superior frontal cortex and basal forebrain from normal adult and confirmed Alzheimer’s disease patients obtained from the New South Wales Brain Bank were immunostained as previously described (Cullen et al., 2005). Sections were incubated overnight at 4°C in primary 12E8 (Fig. 1a, b) or AT8 antibody (Supplementary Fig. S1a, b). Bound antibody was visualized using Alexa Fluor-conjugated secondary antibody (Invitrogen) or using ABC-peroxidase (Vector) and DAB (Sigma).
Organotypic rat hippocampal slices were grown for 14 days on membrane as described (Davis et al., 2009). Slices were treated with 2 µM AM in PBS for 1 h, fixed in 4% formaldehyde for 1 h, permeabilized 90 sec in 0.05% Triton X-100 in PBS and immunostained overnight with 1439 (cofilin) (2 µg/ml) and mouse 12E8 (4 µg/ml) antibodies. Secondary antibody incubations were for 2 h.
We used the monoclonal phosphorylation-dependent 12E8 antibody raised against the pMAP tau, an established early marker for neuropil threads in AD (Augustinack et al., 2002), to determine whether pMAP accumulates in primary neuronal models derived from human, rat or chick following mitochondrial inhibition. Probing AD brain sections with the 12E8 antibody (Fig. 1a, b) or AT8 antibody (Supplementary Fig. S1a, b) showed numerous striated rod-like inclusions within neuropil threads and cytoplasmic accumulations of pMAP, indistinguishable from previous observations in human AD (Augustinack et al., 2002). Additionally, linear striations of rod accumulations labeled with cofilin antibodies were also observed in AD brains (Supplementary Fig. S1c).
Treatment of primary human CNS neurons cultured for 7 days in vitro, with the mitochondrial complex III inhibitor antimycin (AM) elicited a rapid accumulation of 12E8-labeled protein into rod inclusions (Fig. 1c, arrows) comparable to those seen in the AD neurons (Fig. 1b, arrows). The rods sequestered the fraction of MAP serine-phosphorylated in the microtubule-binding-domain KXGS motifs specific for the 12E8 antibody, rather than total MAP/tau because a polyclonal antibody against total tau yielded a more uniform labeling along neurites of both human and chick neurons (Fig. 1c, d). During AM-treatment, primary chick CNS neurons derived from embryonic tectum also generated pMAP-positive rod-like structures (Fig. 1d, e) morphologically identical to those from human neurons. Thus, the readily accessible chick neurons are a useful model system for studying the mechanism by which pMAP accumulates into rods. The conserved 12E8 epitopes in the pMAP sequence reside within the microtubule binding domain in human tau (Ser262/Ser356 residues) and chicken tau (Ser253/Ser378) (Supplementary Fig. S2a, b) (Yoshida and Goedert, 2002). Primary chick tectal neurons express five isoforms of tau that are highly homologous to human tau isoforms and exhibit conservation of phosphorylation-specific epitopes recognized by antibodies against human tau (Supplementary Fig. S2c) (Yoshida and Goedert, 2002).
Actin dynamics are highly dependent on ATP availability (Bernstein and Bamburg, 2003) and ATP depletion is associated with an increase in cellular F-actin and an increase in ADF/cofilin activity (Minamide et al., 2000). It follows that acute inhibition of mitochondrial function used here to generate pMAP-immunostained rods, may result in changes to the actin cytoskeleton. Consistent with this, AM-treated chick neurons exhibited a 1.9-fold increase in overall phalloidin staining in cell bodies indicative of increased F-actin content (Supplementary Fig. S3). The pMAP-positive rods however, did not overlap with phalloidin-labeling (Supplementary Fig. S3b).
In ATP-depleted rat hippocampal neurons, the rapid dephosphorylation of cofilin leads to the development of cofilin-actin rods in neurites (Minamide et al., 2000). These rods contain actin, as evidenced by both antibody immunostaining and ultrastructure, but do not stain with phalloidin, indicating they are probably saturated with ADF/cofilin which stabilizes the “twisted” form of the filament and eliminates the phalloidin binding site (McGough et al., 1997). We therefore asked whether rods induced by ATP-depletion contain both pMAP and ADF/cofilin. Double labeling revealed that ADF/cofilin-actin rods in part co-localize with pMAP in chick (Fig. 2a–c, asterisks), human (Fig. 2d, asterisk) and rat hippocampal neurons (Fig. 2e, asterisk). Co-localization of cofilin and pMAP in rod structures in ATP-depleted organotypic rat hippocampal slice cultures was also revealed (Fig. 3). Moreover, staining for β(III)-tubulin did not accumulate at rods, further suggesting the specificity of pMAP and ADF/cofilin in formation of neuritic rods (Supplementary Fig. S4 b, c). However, ADF/cofilin-stained rods that do not stain for pMAP (arrowheads in Fig. 2c, e) and pMAP-stained rods that do not stain for ADF (arrow in Fig. 2b) were both observed. Because optimal immunostaining for pMAP requires Triton X-100 permeabilization (avoiding methanol) and optimal immunostaining for ADF/cofilin requires methanol (avoiding Triton), structures that stain for one and not the other may be sub-optimally permeabilized or immunostained. However, we cannot exclude that the detected inclusions may represent a heterogeneous population of rod structures. Surprisingly, in hippocampal neurons derived from rat there was no change in the diffuse and uniform distribution of pMAP staining following ATP-depletion, fixation in 4% paraformaldehyde and 0.1% glutaraldehyde, and brief (90 s) 0.05% Triton X100 permeabilization, which looked identical to the staining of pMAP in untreated human, rat and chick neurons (Fig. 1f). However, rat neurons subjected to AM-induced ATP-depletion, followed by fixation in 4% formaldehyde and permeabilization as above did exhibit rod-like 12E8 immunostaining (Fig. 2e), suggesting that glutaraldehyde fixation of pMAP when it is in a rod-like structure, masks its epitope from 12E8 binding. Together, these results link ADF/cofilin-actin and pMAP inclusions within the same rod structures in neurites, which can be induced in cell culture by the common mechanism of acute ATP depletion through mitochondrial dysfunction.
To establish the activation state of ADF/cofilin during pMAP accumulation, we used an antibody specific for phosphorylated (inactive) ADF/cofilin (Minamide et al., 2000; Maloney et al., 2005). ADF/cofilin actin-binding activity is negatively regulated by phosphorylation of the N-terminal Ser3 (Huang et al., 2008; Kim et al., 2009). Upon ATP depletion with AM, ADF/cofilin was rapidly dephosphorylated (activated) in chick tectal neurons, as early as 2 min after treatment (Fig 4a, b), consistent with studies in primary rat hippocampal neurons (Minamide et al., 2000). Peroxide (H2O2)-treated tectal neurons also showed nearly complete ADF/cofilin activation after 30 minutes (Fig. 4a, b). The time course for the redistribution of pMAP immunostaining into the rod-like inclusions (Fig. 4c) correlated with the rapid dephosphorylation of ADF/cofilin and the formation of ADF/cofilin-actin rods.
Based on the above results, we expected that the extent of pMAP incorporation into rods would be proportional to the level of intracellular ATP. To explore this relationship, we measured ATP levels in tectal neurons before and after treatment with AM, CCCP (carbonyl cyanide 3-chlorophenylhydrazone, a mitochondrial uncoupling agent) or H2O2 (Fig. 4d). Consistent with the notion that activation of ADF/cofilin induced by intracellular ATP depletion correlates with pMAP accumulation, there was an inverse relationship between the % of cells in the culture that contained pMAP-stained rods (Fig 4c) and the level of ATP detected in the cell treatment groups (Fig. 4d). Cell viability tests revealed 95% cell survival in 10 min AM-treated chick neuronal cultures (compared to control cultures) and 77% survival in 30 min treated cultures (Supplementary Fig. S4d). Likewise, abundant pMAP-positive rod formation was recapitulated in primary human neurons, following 30 min AM treatment (24±2 rods per field for AM-treated, compared to 8±1 rods per field in controls; mean ± SD, p<0.001). Together, these results suggest ATP-depletion is an important correlate for redistribution of pMAP into rods.
Since ADF/cofilin-actin rods co-stained for pMAP and rod formation depends on activation of ADF/cofilin, we asked whether transfected cofilin-GFP could sequester pMAP into rods. Previous work has shown that expression of cofilin-GFP leads to increased pools of activated cofilin and cofilin-actin rod formation in neurites of rat hippocampal neurons (Minamide et al., 2000; Bernstein et al., 2006). We investigated the effects of ATP depletion on the reorganization of pMAP in relation to cofilin-GFP rods by inhibiting mitochondria. ATP depletion of neuronal cultures led to rapid cofilin-GFP rod formation in transfected cells and strong pMAP co-labeling (Fig. 5a, b; Supplementary Fig. S5a). Conversely, cofilin-GFP rods did not sequester β(III)-tubulin label when all other conditions were identical (Fig. 5c). When non-neuronal cells in culture formed cofilin-GFP rods, these were negative for both pMAP and β(III)-tubulin (Fig. 5d), neither of which is expressed in non-neuronal cells. Taken together, these results suggest that activated cofilin first accumulates into rods and subsequently sequesters pMAP into these cytoskeletal inclusions.
To further evaluate the interaction between cofilin and pMAP in rods, we exploited fluorescence resonance energy transfer (FRET) between the GFP (donor) and Alexa-555 (acceptor) fluorophors. FRET between co-localized cofilin-GFP and pMAP/Alexa-555 was measured using acceptor photobleaching (Fig. 5b). The cofilin-GFP fluorescence signal in rods increased significantly after bleaching the Alexa-555 fluorophor (Fig. 5b, right, arrows), with a measured FRET efficiency of 41 ± 3 % (mean ± s.e.m.; n=52; p<0.0001) as measured by the % increase in donor fluorescence after acceptor bleaching. By contrast, cofilin-GFP fluorescence did not significantly increase in adjacent rods in the same cells where the Alexa-555 fluorophor was not bleached (Fig. 5b, right, asterisks). Furthermore, a much reduced FRET signal was detected in neurons between cofilin-GFP and Alexa-555 immunostained pMAP in adjacent non-rod forming regions of neurites and cell bodies (Supplementary Fig. S5b). Control experiments demonstrated no FRET between cofilin-GFP and control antibody labeled epitopes – β(III)-tubulin/Alexa-555 or Src/Alexa-555 - even in regions where the fluorescence appeared co-localized (Supplementary Fig. S5c, d). Cells expressing free GFP and treated with AM to generated pMAP-positive rods labeled with 12E8/Alexa-555 also never produced FRET (Supplementary Fig. S5d). These results demonstrate that cofilin-GFP and the secondary antibody labeling pMAP are co-localized <10 nm apart within the cytoskeletal rod complex indicating a close proximity of pMAP and cofilin in these inclusions.
Since induction of ADF/cofilin rods leads to the reorganization and sequestering of pMAP, we asked whether silencing ADF or cofilin would prevent sequestering of pMAP under the same conditions. During chick brain development, ADF comprises about 75% of the total ADF/cofilin from embryonic day 14 onwards (Devineni et al., 1999). We therefore chose to knockdown the total cellular pool of ADF specifically, using a plasmid for expression of a small hairpin RNA (shRNA) that yields an siRNA when expressed and processed. The shRNA was transfected into chick neurons and co-expressed from the same plasmid as GFP for visualization of transfected cells. Four days post-transfection, cells were immunolabeled for ADF (Fig. 6a) and quantified, revealing 75% knockdown of ADF (Fig 6e). Treating transfected cultures with AM (1µM for 10 min) and staining for pMAP, revealed that significant silencing of ADF inhibited accumulation of pMAP into rod structures while surrounding untransfected cells contained an abundance of rods (Fig. 6d, asterisks). In control chick neuronal cultures transfected for 4 days with shRNA specific for human PAK2, immunolabeling revealed no decrease in total cellular ADF (Fig. 6b, e). pMAP accumulated in rod-like inclusions both in surrounding non-transfected neurons and in shRNA PAK2 transfected neurons (Fig. 6c, asterisks and arrowhead). Together, these results suggest that the presence of ADF/cofilin is necessary for sequestration of pMAP into rod-like structures.
Since activation of ADF/cofilin coincided with pMAP recruitment into ADF/cofilin-actin rods, we asked whether other manipulations of F-actin assembly could influence pMAP sequestration to rods. To address this, we used pharmacological manipulation (enhancing or suppressing) of F-actin pools and stained for pMAP. Latrunculins sequester monomeric G-actin to form a nonpolymerizable 1:1 complex, thereby inhibiting F-actin reassembly and thus promoting overall F-actin depolymerization (Coue et al., 1987). Latrunculins however, compete weakly with ADF/cofilin for actin binding (Bernstein et al., 2006) and can actually induce ADF/cofilin-actin rod formation (Pendleton et al., 2003). The decline in the phalloidin-stainable F-actin pool after latrunculin B (Lat B) treatment is indeed evident (Fig. 7a) although it should be noted that phalloidin cannot stain ADF/cofilin saturated F-actin. By contrast, jasplakinolide (Jasp) binds and stabilizes F-actin resulting in a net decrease in the G-actin pool (Bubb et al., 1994). Since Jasp competes for the phalloidin-binding site on F-actin, stabilized F-actin cannot be visualized with fluorescent phalloidin in the presence of Jasp (Fig. 7a). However, ADF/cofilin cannot bind to phalloidin-stabilized F-actin (Minamide et al., 2000) and thus also would not likely bind to the Jasp-stabilized actin filaments.
Antibody labeling of Lat B-treated neurons revealed formation of both pMAP- and ADF-positive rods which frequently co-localized in double-labeling experiments (Fig. 7b). By contrast, neurons treated with Jasp (or co-treated with Jasp and AM) exhibited a total absence of pMAP and ADF rods (Fig. 7a). We then compared the rate and abundance of pMAP-stained rods following Lat B compared to AM treatments and found the chronology of their formation indistinguishable between conditions (Fig 7c). That Lat B treatment only moderately reduced ATP in neurons (Fig. 7d) yet still induced ADF/cofilin-actin and pMAP rod formation comparable to mitochondrial inhibitors suggests that ATP depletion is not a direct cause of rod formation, but an upstream event. These data collectively suggest that subunit release from F-actin is required for the generation of pMAP-positive rods and that in neurons this can be induced by ADF/cofilin activation. It also appears that ADF/cofilin binding to F-actin is essential for the formation of cytoskeletal rods. Dissociated actin subunits saturated with activated ADF/cofilin thereby form cytoskeletal rod inclusions that appear to sequester and accumulate pMAP.
In AD, the relationships between pMAP/tau pathologies, cofilin-actin pathologies and amyloid pathologies arising from Aβ peptide oligomerization and/or aggregation are poorly understood. Synthetic Aβ1–42 has been shown to induce cofilin-actin rod pathology in up to 20% of neurons in dissociated rodent hippocampal cultures (Maloney et al., 2005), with the majority of these neurons located in the dentate gyrus (Davis et al., 2009). We therefore asked whether Aβ could influence the sequestering of pMAP to ADF/cofilin-actin rods. Primary chick neurons were exposed to 1 or 2 µM of Aβ1–40 or Aβ1–42. Before applying to cells, peptide solutions were agitated on a shaker for 30 minutes to generate mixtures of oligomeric and fibrillar structures (Goldsbury et al., 2000). The solutions were then diluted into cell culture medium and added to cells at a final concentration of 1 or 2 µM. After exposure to the peptides for 20 hours, the cells were fixed and immunostained for pMAP. Less rods were generated in cultures treated with Aβ compared with those treated with the mitochondrial inhibitors, consistent with previous studies of ADF/cofilin-actin rod generation in hippocampal neurons treated with Aβ peptides (Maloney et al., 2005). However, when mean numbers of pMAP-positive rods per cell were quantified on vehicle control and Aβ-treated coverslips, a significant increase in the number of rods was revealed in Aβ-treated cells (Fig. 8a). No difference was observed between cells treated with 1 and 2 µM peptide or between cells treated with Aβ1-40 and Aβ1-42. The presence of F-actin rich active growth cones in Aβ-treated neurons containing pMAP rods in the neurite shafts indicates the continuing viability of the neurons, despite the formation of pMAP-positive rods (Fig 8b, white arrows). We confirmed that rods in Aβ-treated neurons contained both pMAP and ADF (Fig. 8c, arrows). In conclusion, these results show that synthetic Aβ aggregates can induce the recruitment of pMAP to ADF/cofilin-actin rods in a subset of neurons in these primary cultures.
Neuropil threads of the AD brain are comprised of linear arrays of cytoskeletal inclusions containing the pMAP tau (Augustinack et al., 2002; Velasco et al., 1998). These structures are positive for Ser262/356 tau labeled with the 12E8 antibody prior to extensive staining by antibodies marking hyperphosphorylation of tau that is observed in late-stage disease (Augustinack et al., 2002). Corresponding to this, formation of neuropil threads is associated with the early clinical stages of dementia (Giannakopoulos et al., 2007). The structure of neuropil threads implicates a disrupted cytoskeletal network that spans the width of the neurite likely contributing to dementia by blocking cargo trafficking to synapses that underlie memory formation and cognition, causing subsequent retraction of distal neurites (Velasco et al., 1998; reviewed in Terry, 1998; Bamburg and Bloom, 2009). Our data open the possibility that abnormal activation of cofilin, an actin-binding protein, in the AD brain and generation of cofilin-actin cytoskeletal rods is a triggering factor for the sequestration and accumulation of Ser262/356 phosphorylated tau in some neuropil threads. This process may well represent an early pathogenic event in AD neurodegeneration.
Neurodegenerative stimuli including oxidative stress, mitochondrial dysfunction, excitotoxic glutamate, ischemia, and soluble forms of Aβ lead to activation of cofilin and the related protein ADF and the generation of ADF/cofilin-actin rods in primary neuronal cell culture (reviewed in Bamburg and Bloom, 2009). These rods resemble the cofilin aggregates that are widely distributed in the AD brain (Minamide et al., 2000). At a low ratio to actin and in an active (dephosphorylated) state, ADF/cofilin maximally enhance subunit turnover of F-actin, dynamically remodeling the actin cytoskeleton and thus playing an integral role in regulating actin-dependent synaptic stabilization (reviewed in Bamburg and Bloom, 2009). ADF/cofilin bind cooperatively to F-actin and, at higher ratios to actin, they can saturate regions of filaments and stabilize the pieces of filaments that remain after severing (Chan et al., 2009, Andrianantoandro and Pollard, 2006). Because ADF/cofilin bind to a minor, slightly twisted conformation of F-actin, they stabilize this “twisted” form, which prevents binding of the commonly used F-actin stain, phalloidin (McGough et al., 1997). In the brain, mitochondrial dysfunction and energy depletion associated with AD (Smith et al., 2005; Wang et al., 2009) could feasibly serve as a pathway for F-actin remodeling in neurons, since up to 50% of neuronal energy is dedicated to actin dynamics (Bernstein and Bamburg, 2003). We propose mitochondrial dysfunction is one potential pathway upstream of the assembly of cytoskeletal rods because direct electron transport chain inhibitors elicit cofilin activation, concomitant with rod assembly in primary neurons and organotypic rat hippocampal brain slices. This pathway results in a precipitous drop in ATP, and the release of the cofilin phosphatase chronophin from an inhibitory complex with Hsp90, resulting in cofilin-actin rod formation (Huang et al., 2008). However, cofilin-actin rods can also be induced while maintaining higher levels of cellular ATP, such as with peroxide (Fig. 4), which activates the cofilin phosphatase slingshot by oxidizing and removing inhibition by 14-3-3 and leads to almost complete cofilin dephosphorylation and cofilin-actin rod formation (Kim et al., 2009). The common thread is an increased pool of active ADF/cofilin with respect to the amount of F-actin such that ADF/cofilin-saturated pieces of F-actin are available and coalesce into rods.
What is the role of pMAP at the cofilin-actin rods? Rods isolated from neurons and non-neuronal cell lines contain ADF/cofilin:actin in a 1:1 complex and indeed can be formed in vitro from these purified proteins (LS Minamide, S Maiti, JA Boyle, RC Davis, JA Coppinger, Y Bao, T Huang, J Yates, G Bokoch and JR Bamburg, unpublished results). Thus, as is also shown here, pMAP is not essential for cofilin-actin rod formation. Nevertheless, pMAP association with most cofilin-actin rods occurs very early in the rod formation process. Rods isolated from cortical neurons and those formed from endogenous proteins from a non-neuronal cell line both have similar stabilities to alterations in pH, ionic strength, reducing agents, Ca2+/EGTA, ATP and detergents (LS Minamide and colleagues unpublished results). Therefore the presence of pMAP on the neuronal rods does not appear to confer in them any special property related to their stability. The question then arises – do the rods confer any special property on the associated pMAP? The tandem arrays of the rapidly formed cofilin-actin rods are similar in size and distribution to the neuritic striations of the neuropil threads in AD. The ultrastructure of most striated neuropil threads from AD brain clearly shows these to consist of the ~20 nm diameter paired helical filaments (PHFs; Velasco et al., 1998), not of the cofilin-actin bundles containing approximately 10 nm diameter filaments that we observe in organotypic hippocampal slices treated with Aβ1–42 peptide oligomers (Davis et al., 2009). Thus, we suggest that the cofilin-actin rods serve as a template for recruitment and binding of pMAP, and that this association facilitates further phosphorylation of pMAP and its self-assembly leading to the eventual replacement of the cofilin-actin rods with PHF bundles. Further long-term studies are needed to test this hypothesis.
Supporting the concept that pMAP/tau accumulation and toxicity is linked to interactions with the actin cytoskeleton is work in Drosophila models that showed that toxicity of over-expressed hyperphosphorylated tau could be modified by genetic ablation or co-expression of actin associated proteins (Fulga et al., 2007). Interestingly, tau phosphorylation at the 12E8 antibody epitope was shown to be essential for the initiation of tau toxicity and its downstream hyperphosphorylation in Drosophila (Nishimura et al., 2004). Mutated tau non-phosphorylatable at the 12E8 epitope did not confer toxicity in Drosophila and exhibited reduced downstream phosphorylation at other AD-relevant phospho-tau epitopes (Nishimura et al., 2004). These findings are consistent with the premise that the sequestration of 12E8-specific pMAP to cofilin-actin rods demonstrated here could represent an essential early event in a neurodegenerative pathway. The recent finding that tau reduction can ameliorate toxicity induced by amyloid precursor protein (APP) overexpression or exposure of cells to Aβ oligomers is also consistent with the key involvement of pMAP/tau in response to toxic events (King et al., 2006; Roberson et al., 2007). Potential deleterious effects of cytoskeletal rods on neurons are multifold. Firstly, rods that assemble in axons and dendrites of cultured neurons form transport blockades inhibiting the free movement of vesicles and organelles leading to synaptic dysfunction (Maloney et al., 2005, 2008; Jang et al., 2005). Secondly, the accumulation of actin and cofilin in rod structures could directly impinge on the availability of cofilin for its function in modulating actin dynamics at the synapse (Hotulainen et al., 2009). Likewise the sequestration of pMAP/tau to the rod matrix could adversely affect the role of tau in stabilizing and regulating the axonal MT network of tracks for fast axonal transport.
The generation of cytoskeletal rods in response to stressors may be an initial compensatory response to slow down energy-consuming actin dynamics and thereby protect neurons under conditions where ATP supply is compromised (Bernstein et al., 2006). However repeated or prolonged stress would likely be detrimental. Increasing evidence suggests declining mitochondrial function and reduced energy metabolism to be early events in the AD brain, preceding severe pathological changes (Smith et al., 2005; Wang et al., 2009). Ageing is the major risk factor for sporadic AD and on the one hand, electron transport chain activity declines with age while on the other, oxidative stress increases, potentially impacting on the availability of ATP in cells (Lin and Beal, 2006). Moreover, the reduction of glucose uptake in AD brain cells and concomitant decline in glycolysis would also negatively impact ATP levels. Precedents of local mitochondrial dysfunction could be multifold and varied in the ageing brain, for example low brain perfusion due to vascular insufficiency, stroke and Aβ peptide-mediated perturbation of mitochondria (Cullen et al., 2005; Du et al., 2008; Cho et al., 2009). Mechanisms linking this metabolic dysfunction with the formation of neuropathological lesions in the AD brain need to be established. Interestingly, it was demonstrated that environmental toxin inhibitors of mitochondrial respiratory chain complexes induce somatic inclusions of phosphorylated tau in neuronal cell culture and neurodegeneration in tau-related disease (Escobar-Khondiker et al., 2007).
Advanced neuropathology involving pMAP and cofilin accumulation and aggregation may be a result of increased vulnerability to normal environmental stress in brains expressing mutant tau (i.e in hereditary Frontotemporal Dementias), or from elevated oxidative/mitochondrial stress in brains expressing wild-type endogenous tau (in sporadic AD). In a healthy brain, a fine balance between these stressors and compensatory defense mechanisms must normally occur, given that development of tau pathology and AD are not an inevitable part of ageing. Recent evidence showing that neuronal cell bodies attached to dystrophic neurites in AD-related transgenic mice are still viable and otherwise healthy (Adalbert et al., 2009) holds promise for future work to develop methods to disrupt the formation of the cofilin-actin rods and/or the interaction between cofilin and pMAP for establishing new therapeutic strategies to combat this disease early in its development.
Supplementary Figure S1. Neuritic striations in human AD. (a) An AD forebrain neuron, still containing a nucleus, shows accumulations of tau phosphorylated at Ser202/Thr205 (as indicated by AT8 immunoreactivity) throughout the cell body and neurites. Inset (magnified in (b)) shows striation of the neurites with linear arrays of AT8-positive inclusions (arrows). This AD case was classified with early stage (Braak Stage II) accumulation of tau pathologies. (c) Paraffin section (6–7 µm) of frontal cortex from a confirmed AD brain immunolabelled with 1439 (cofilin). Linear striations of rod accumulations are evident (arrows). Scale bars = 10 µm (a, c); 3 µm (b).
Paraffin sections (6–7 µm) of frontal cortex from a confirmed AD brain were obtained from the Alzheimer Disease Research Center, University of California, San Diego. Post-mortem time to fixation in formalin was 3 h. Sections were deparaffinized at 60 °C for 40 min followed by 2 × 10 min in Hemo-De (Fisher Scientific). The sections were then rehydrated sequentially in 50% Hemo-De/50% ethanol (1 min), 100% ethanol (3 × 1 min), 95% ethanol (2 × 1 min), and 70% ethanol (1 min). Sections were rinsed in several changes of water, microwaved in water for 4 min, rinsed in PBS, and blocked with 5% goat serum in 1% BSA in TBS for 1 hr before immunostaining. The ADF 1439 antibody diluted to 2 µg/ml in 1% BSA in TBS was applied for 1 hr followed by TBS washes. Fluorescein tagged goat anti-rabbit IgG (Invitrogen) diluted 1:400 in 1% BSA/TBS was applied for 45 min followed by PBS washes. The sections were then drained, 75 µL Prolong Antifade (Invitrogen) applied, and then covered with a clean coverslip and allowed to dry overnight with the and mounting as described elsewhere (Minamide et al., 2000).
Supplementary Figure S2. Chicken tau is highly homologous to human and rodent tau. (a) Epitope map of phosphorylated sites recognized by tau antibodies in human, chicken (Yoshida and Goedert, 2002) and rat brain. Schematic representations of the longest tau isoforms are shown (441 amino acid residues in human tau, 463 in chicken and 432 in rat). The phosphorylation-dependent anti-tau antibodies (AT270, AT8, AT180, 12E8 and pS396) are shown with their corresponding target residues. Positions of the various alternatively spliced inserts are shown in blue. Green boxes denote tandem amino acid sequence repeats, which constitute the microtubule-binding domain (MTBD). (b) Amino acid sequence alignment of MTBD regions of human, chicken and rat tau demonstrate strong homology in this region. (c) Western blots of primary chicken neuronal cultures were probed with anti-tau antibodies. Six tau bands ranging from 50–64 kDa are identified with a phosphorylation-independent antibody (total tau). Probing with phosphorylation-dependent antibodies (12E8, AT8, AT180, AT270 and pS396) reveals preferential phosphorylation of specific epitopes on different isoforms. Marker positions (kDa) indicated by arrows.
Supplementary Figure S3. Cell body F-actin labeling increases following ATP depletion. (a, b) Chick neurons (7 d.i.v) were co-stained with phalloidin (green) and 12E8 (red) before treatment (a) and after treatment (b) with AM for 15 min. Merged images from double-labeling are shown on the right. Rods forming in the neurite shaft did not stain with phalloidin (b, arrows). (c) Intensity of phalloidin staining in AM-treated cells increased significantly compared to controls (*p<0.001, n = 25 cells for each treatment condition. Error bars represent s.e.m.). Scale bar = 20 µm.
Supplementary Figure S4. pMAP and ADF/cofilin rods form after mitochondrial inhibition but tubulin remains uniformly distributed. (a) Single stainings of chick neurons with primary antibodies against pMAP (12E8), cofilin (C8736; Sigma), and ADF (1439) (as indicated) were conducted on cells following 10 min treatment with AM. Formation of rod-like structures was apparent in each staining condition. (b) Whereas ADF (D8815; Sigma) (green) accumulates in rod-like structures in cells treated with AM for 15 min, β(III)-tubulin (red) staining remained evenly distributed throughout neurites (arrows). (c) In non-treated cells, both β(III)-tubulin and ADF staining was evenly distributed except for a characteristic ADF increase and lack of tubulin in growth cones (asterisk). (d) To determine cell viability, cells were stained with Trypan blue following 0, 10, 30 or 60 min AM treatment. 94.6 ± 0.1% cells were viable after 10 min AM treatment (normalized to controls), 77.2 ± 0.1% after 30 min AM and 71.5 ± 0.14% after 60 mins (mean ± standard deviation expressed as % control, n>10 fields per treatment condition). Scale bars = 10 µm (a); 20 µm (b, c).
Supplementary Figure S5. pMAP is selectively sequestered into cofilin-GFP rods following ATP-depletion. (a) Co-localization of cofilin-GFP and pMAP (indicated by yellow in merged images) occurs following AM treatment. Transfected cells appear to accumulate more pMAP rods compared to surrounding non-transfected cells (arrows). (b) In contrast to rods, significant FRET between cofilin-GFP and Alexa 555 was not observed in non-rod containing regions of cell bodies (left arrow) or neurites (right arrow). (c) FRET was not observed between cofilin-GFP and β(III)-tubulin/Alexa-555. (d) Summary of quantification of FRET efficiency (% change in donor fluorescence) in experiment and controls. Only acceptor-bleached pMAP-positive cofilin-GFP rods gave rise to a significant FRET signal. Scale bars = 20 µm (a); 5 µm (b, c).
We thank Prof. Peter Seubert, Elan Pharmaceuticals for providing the 12E8 antibody. We are grateful for support from The Sir Zelman Cowen Universities Fund, The Judith Jane Mason & Harold Stannett Williams Memorial Foundation, The Rebecca Cooper Foundation and Sydney Medical School (to CG) and the Alzheimer Drug Delivery Foundation (grant 281201 to JRB), the National Institutes of Health, National Institute of Neurological Diseases and Stroke (grants NS43115, NS40371 to JRB).