Rapid Tau Aggregation and Delayed Hippocampal Neuronal Death Induced by Persistent Thrombin Signaling*

Tau hyperphosphorylation, leading to self-aggregation, is widely held to underlie the neurofibrillary degeneration found in Alzheimer's disease (AD) and other tauopathies. However, it is unclear exactly what environmental factors may trigger this pathogenetic tau hyperphosphorylation. From several perspectives, the coagulation serine protease, thrombin, has been implicated in AD and activates several different protein kinase pathways but has not previously been shown how it may contribute to AD pathogenesis. Here we report that nanomolar thrombin induced rapid tau hyperphosphorylation and aggregation in murine hippocampal neurons via protease-activated receptors, which was followed by delayed synaptophysin reduction and apoptotic neuronal death. Mechanistic study revealed that a persistent thrombin signaling via protease-activated receptor 4 and prolonged downstream p44/42 mitogenactivated protein kinase activation are at least in part responsible. These results pathogenetically linked thrombin to subpopulations of AD and other tauopathies associated with cerebrovascular damage. Such knowledge may be instrumental in transforming therapeutic paradigms.

Dephosphorylated microtubule-associated protein tau binds to tubulin dimers and stabilizes microtubule structure. Phosphorylation of tau at its microtubule-binding domain disassociates tau from microtubules and leads to microtubule disassembly. Moreover, abnormally hyperphosphorylated tau tends to self-aggregate and is likely transported to neuronal soma for either degradation or aggregation into inclusion bodies (IBs) 1 if incapable of degradation (see Ref. 1 for review). The IBs composed of primarily undegraded hyperphosphorylated tau aggregates are pathologically revealed as neurofibrillary tangles (NFTs), which are pathological hallmarks in Alzheimer's disease (AD) and other tauopathies (2). Although discrepancies exist regarding the causality between NFTs and neurodegeneration (3), certain types of neurodegeneration, particularly those in AD and other tauopathies, are characterized by accompanying NFT formation, referred as neurofibrillary degeneration.
Most tauopathies like AD are multifactorial disorders. Although significant efforts have been devoted to elucidate disease-specific risk factors or possible pathogenetic causes, the specific factors/causes responsible for induction of NFTs and neurofibrillary degeneration in different tauopathies remain to be established. Genetic mutations in the tau gene may render tau protein more prone to aggregate and form NFTs in frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17) (4 -6). Although evidence indicates that multiple tau mutations are clearly pathogenic in FTDP-17 and some polymorphisms may increase risk in other tauopathies, neither AD nor the majority of sporadic tauopathy patients have tau mutations, whereas they still express typical neurofibrillary degeneration in various populations of neurons. This points out that certain environmental pathogenetic risk factors may be sufficient to induce NFTs and/or neurofibrillary degeneration in these conditions. With AD as an example, cerebrovascular degeneration is consistently observed and the laminar and regional distribution of the vascular alterations are closely correlated with the presence of NFTs (7)(8)(9)(10)(11)(12). Moreover, antecedent traumatic brain injury, a known risk factor for AD (13)(14)(15), produces NFTs (13, 14, 16 -19). The pathological observations indicated that the NFTs in these patients were consistently situated around blood vessels in the heavily affected regions (16). In dementia pugilistica, the recurrent and severe form of traumatic brain injury seen in career boxers, the molecular characteristics of tau in terms of its six isoforms and precise phosphorylation profiles have been recently shown to recapitulate those of AD (13). These facts raise the question as to whether certain blood-derived factors that might leak from damaged cerebral vessels are capable of inducing neurofibrillary degeneration.
Among various blood-derived factors, the serine protease thrombin is not only critical in peripheral coagulation cascades (20) but may also affect a broad spectrum of central nervous system glial and neuronal cell functions (21)(22)(23)(24)(25). These include astrocytic stellation and apoptosis (26 -28), microglial activation and proliferation (29 -31), as well as synaptic collapse, neuritic retraction, and apoptosis in motor and sensory neu-rons (32)(33)(34). In fact, thrombin is elevated (35,36), whereas the activity of the cognate thrombin inhibitor, protease nexin I, is sharply reduced (37,38) in AD brains, particularly around degenerating cerebral vessels (39). Taken together, these facts have therefore led us to propose and test whether thrombin is capable of inducing neurofibrillary degeneration in hippocampal neurons, and if so, what extra-and intracellular mechanisms might be involved.
Preparation of Soluble and Insoluble Cellular Fractions-Treated cells were harvested in pre-chilled (4°C) H buffer (10 mM Tris, 1 mM EGTA, 0.8 M NaCl, 10% sucrose, protease, and phosphatase inhibitor mixtures, pH 7.4), and homogenized with a Teflon/glass homogenizer by 10 strokes at 200 rpm for each sample. After centrifugation at 27,200 ϫ g for 20 min at 4°C, the supernatants were saved, and the pellets were resuspended in 10 volumes of H buffer, homogenized, and centrifuged again at the same conditions. The resulting pellet (P1) was saved, the supernatants were combined with the earlier supernatants, and then centrifuged at 100,000 ϫ g for 20 min at 4°C. The final pellet (P2), along with P1, was resupended in 2ϫ reducing sample buffer (125 mM Tris, pH 6.8, 4% SDS, 10% glycerol, 0.006% bromphenol blue, 2% ␤-mercaptoethanol), sonicated with high power for 30 s and boiled for 5 min before SDS-PAGE analysis. An equal volume of the final supernatants (S) from each sample was directly mixed with 2ϫ reducing sample buffer, and boiled for 5 min before SDS-PAGE analysis.
RT-PCR-Cultured cells were washed and then lysed in Trizol and total RNAs were extracted according to the manufacturer's recommendations. RNA concentrations were based on absorbance at 230, 260, and 280 nm. Reverse transcription-PCR was performed with 0.5-1.0 g of total RNA, downstream primer sequence 5Ј-GCCATCACCCAAATGAC-CAC, and upstream primer 5Ј-GTGCGGTCCCTTGCTGTCTT in a 50-l volume with 5 units of rTth polymerase (PerkinElmer Life Sciences), 0.3 mM dNTPs, 2.5 mM MnOAc, 0.45 M primer incubated at 57°C for 30 min, 94°C for 3 min followed by 40 cycles of 94°C for 1 min and 57°C for 1 min. PCR of 2 l the products was performed with the upstream primer 5Ј-CAGCCAGAATCAGAGAGG in 100-l reactions with 2.5 units of Taq polymerase (Bioline, London, UK), 0.2 mM dNTPs, 1.5 mM MgCl 2 , and 0.5 M each primer. cDNA products were visualized, after agarose gel electrophoresis and staining with ethidium bromide, under UV light and images were captured and analyzed with the aid of a CCD camera and NIH Image software.
Thioflavin-S Staining-Fixed cells on glass coverslips were washed with PBS and followed by incubation in 0.01% thioflavin-S for 5 min. After washing 3 times with 70% ethanol, glass coverslips were mounted with Fluoromount and observed under a Nikon Bio-Radiance confocal microscope.
Statistical Analysis-All qualitative analysis (Western blot and ICC) was repeated at least three times. Quantitative data (LDH assay) are expressed as mean Ϯ S.E. and analyzed by ANOVA using StatView 6.0 (Abacus Systems, Mountain View, CA). Post-hoc comparisons of means were made using Scheffe's or Tukey's method where appropriate.

Thrombin Induced Tau Hyperphosphorylation and Aggregation in HT22 Hippocampal
Neurons-Potential impact of thrombin on the cellular status of tau protein was first examined using ICC and Western blot in differentiated HT22 cells, an immortalized mouse hippocampal neuronal cell line (43). As seen in Fig. 1, a phosphorylation-independent, total tau pAb, A24, revealed a homogenous staining pattern of tau in cytosol and processes of control HT22 cells, whereas it decorated intensified perinuclear and cytoplasmic IB-like tau profiles after 24 h treatment with 100 nM ␣-thrombin. These data indicated that thrombin induced significant tau translocation and possibly tau aggregation.
Hyperphosphorylation of tau has been widely believed to promote tau aggregation both in cultured cells and in vivo (see Ref. 1 for review). To determine whether thrombin treatment increases tau phosphorylation levels, we used a panel of phosphorylation-specific tau Abs and further examined the influence of thrombin on the cellular state of tau in HT22 cells. In contrast to the distinct IB-like tau staining revealed by the A24 total tau Ab, tau-1, a mAb recognizing only dephosphorylated tau, homogenously decorated neuritic and cytosolic tau in both thrombin-treated and control HT22 cells (data not shown). This failure to reveal the IB-like tau staining patterns in thrombintreated cells implied that the thrombin-induced IB-like tau staining profiles might be primarily comprised of phosphoryl-ated tau proteins. In support of this assumption, we next used AT8, a mAb specifically recognizing hyperphosphorylated tau at pSpT 202/205 , a phospho-epitope that may appear in both control and AD brains (1). AT8 positively decorated both cytoplasmic and neuritic IB-like tau structures in thrombintreated, but not control, HT22 cells. Moreover, a phosphoepitope-specific tau pAb for pS 422 , a phosphorylation site only found in AD but not control brains (1), also revealed IB-like tau profiles in HT22 cells treated with thrombin, but not the control cells (Fig. 1). Furthermore, a pAb (AB1518, Chemicon) raised against purified NFT extracts, and which is capable of recognizing multiple AD-related hyperphosphorylated tau isoforms, highlighted specific cytoplasmic and neuritic IB-like structures in thrombin-treated, but not control HT22 cells (Fig.  2, A-F). Nevertheless, at least two other phospho-epitope specific Abs, a pAb for pS 396 and a mAb for pS 409 tau (PG-5), failed to stain either cytoplasmic or neuritic structures in control or thrombin-treated cells (data not shown). Taken together, these results indicate that thrombin treatment of HT22 cells increased phosphorylation levels of tau proteins at certain specific sites that are believed relevant to NFT formation (Table I). Going further, these hyperphosphorylated tau proteins in mouse hippocampal cells were translocated to perinuclear and cytoplasmic and accumulated in IB-like structures.
To further characterize the property of the thrombin-induced IB-like tau profiles in HT22 cells, we also performed a routine thioflavin-S staining. This revealed the similar IB-like profiles that were co-localized with those stained by the AB1518 NFT pAb (Fig. 2, A and B). Because thioflavin-S staining preferentially labels protein aggregates containing ␤-sheet structure, these results suggest that the thrombin-induced hyperphosphorylated tau proteins in the IB-like structures formed ␤-sheetcontaining protein aggregates, and these IB-like structures might represent "pre-tangle" tau aggregates (46 -48). Therefore, in summary, these morphological studies suggest that thrombin is capable of inducing tau translocation and aggregation in HT22 hippocampal neurons and that hyperphosphorylation of tau may participate in mediating the tau aggregation.
In addition to the morphological characterization of the thrombin-induced IB-like tau aggregates, we also performed Western blot analyses with these tau Abs. As seen in Fig. 3A, 100 nM ␣-thrombin induced time-dependent (as early as 2 h) tau oligomerization in HT22 hippocampal neurons, as evidenced by the appearance of high molecular weight (HMW) bands immunopositive for tau using the total tau Ab, A24, in reduced and denatured electrophoresis conditions. In contrast, although the tau-1 Ab revealed monomeric tau isoforms it failed to show the HMW tau bands as revealed by A24 (Fig. 3B), supporting the notion that the HMW tau bands were comprised mainly of phosphorylated tau. In addition, under native conditions, Western blot with the NFT-specific Ab, AB1518, failed to detect immunopositive bands in control but numerous bands were stained intensively in HT22 cells treated with thrombin (Fig. 3C). As early as 2 h after treatment, thrombin induced HMW tau bands, and this was true for Western blots employing both A24 and AB1518 Abs. This is even more clearly shown by Western blot using AB1518, where thrombin initially induced several AB1518-positive bands with molecular weights between 50,000 and 285,000, which may include both monomeric and oligomeric phosphorylated tau proteins. With increasing time, these bands were more intensified and eventually became a smear with molecular weight greater than 200,000, whereas the smaller bands essentially disappeared after 24 or 48 h treatment. These results suggest that thrombin induced a rapid and time-dependent increase of tau phosphorylation and oligomerization, and that the time course for developing this change could be as short as 2 h and reached its peak at the latest by 24 h. Aside from the results using tau-1 and AB1518 NFT Abs, Western blots with several other phospho-specific tau Abs under native conditions were also performed. As shown in Fig. 6B, phospho-specific Ab against pSpS 199/202 tau revealed that control HT22 cells showed signif- icantly stained bands, particularly at sizes corresponding to the monomers, while thrombin treatment resulted in a strong increase of the immunoreactivity especially those corresponding to HMW tau oligomers. These data suggest that the pSpS 199/202 tau may be physiologically presented in these cells, but that thrombin strongly increased the tau phosphorylation levels at this particular site. In addition, the AD-specific phosphoepitope of tau, pS 422 , was not found in control HT22 cells but presented in thrombin-treated cells at high levels primarily as HMW bands (Fig. 6C). Probing with PG-5 (pS 409 -specific) and anti-pS 396 -specific tau failed to reveal significant differences between control and thrombin-treated cells. Taken together, except for PG-5 and pS 396 -tau, all other phospho-specific tau Abs tested revealed a common feature that thrombin treatment in HT22 cells not only increased phosphorylation levels of tau but the increased hyperphosphorylated tau proteins primarily integrated into HMW tau oligomers under native non-reducing conditions. Therefore, these Western blot data confirmed the morphological results, and support the conclusion that thrombin induced hyperphosphorylation and oligomerization of tau in HT22 hippocampal neurons.
Up to this point, we have shown that thrombin increased the phosphorylation levels of tau proteins in HT22 cells, and these hyperphosphorylated tau proteins formed oligomers and translocated to perinuclear and cytoplasmic compartments and accumulate in IB-like ␤-sheet-containing structures. All these observations support an assumption that thrombin not only induced tau hyperphosphorylation but also led to formation of insoluble tau protein "aggregates." To demonstrate that thrombin treatment in HT22 cells indeed induced insoluble tau protein aggregates that are sufficient to sediment after centrifugation, we attempted to separate the cell lysates into three fractions that were relatively enriched in large protein aggregates (P1), small protein aggregates (P2), and soluble protein extracts (S) using multistep ultracentrifugation (see "Experimental Procedures" for details). As analyzed by SDS-PAGE and Western blot (Fig. 4), in the P1 fraction, we found no noticeable tau immunoreactivity in control HT22 cells, whereas thrombin-treated cells showed strong tau immunoactive signals for molecular sizes corresponding to monomeric tau isoforms in addition to two weak HMW (between 200 and 285 kDa) tau bands. In the P2 fraction, there appeared faint but noticeable monomeric (but not HMW) tau staining in control HT22 cells, whereas thrombin-treated cells not only showed relatively stronger monomeric tau staining but also revealed several weakly stained HMW (between 90 and 285 kDa) tau bands. In large contrast, analysis of S fraction showed very strong monomeric tau signals for control but sharply reduced tau signals for thrombin-treated cells. Semi-quantitative analysis of the total tau immunoreactivity and its distribution in different fractions are plotted in Fig. 4B and listed in Table II. As indicated, although thrombin treatment resulted in an ϳ60% increase of the total tau as compared with control cells, the majority (Ͼ70%) of the tau immunoreactivity in thrombintreated cells was distributed in the insoluble P1 and P2 fractions, whereas only slightly more than a quarter remained in the soluble fraction. This is in vast contrast to that for control in which over 90% tau remained soluble. Therefore, the increase of tau by thrombin is likely to be a result from posttranslational accumulation of insoluble tau aggregates (more resistant to degradation compared with soluble tau) rather than up-regulation of tau expression at the transcriptional level. Taken together, these results confirmed our assumption that thrombin not only induced tau hyperphosphorylation but also led to formation of insoluble tau protein aggregates.
Thrombin-induced Tau Aggregation Is Mediated by Proteaseactivated Receptors (PARs)-Thrombin induces its cellular effects primarily by activation of PARs, a small subfamily of novel G protein-coupled receptors (20,49,50). PARs are distinct from typical G protein-coupled receptor-ligand interactions, in that their activation requires thrombin (or other protease) proteolytic activity to cleave an N-terminal exodomain  Phosphoryation-specific tau pS 422 ϩϩϩ that unmasks a new receptor N terminus serving as a tethered peptide ligand. The latter binds intramolecularly to the receptor body to self-activate and transmit signals into cells (20). In addition to PARs, a non-PAR receptor, potentially activated by the loop B region of thrombin that does not require the proteolytic activity of thrombin, has also been proposed (51)(52)(53)(54). Therefore, to understand the molecular mechanisms underlying the thrombin-induced tau aggregation in HT22 hippocampal neurons, we first determined whether the proteolytic activity of thrombin is required. In comparison with active ␣-thrombin-treated cells (Fig. 3C, lane 5), treatment with denatured ␣-thrombin (100 nM boiled for 10 min) for 24 h failed to induce NFT-immunopositive staining (Fig. 3C, lane 7). Furthermore, pre-treatment for 5 min with protease nexin I, the cognate tissue-based thrombin-inhibiting serpin, at 100 nM, or with the thrombin-specific leech inhibitor hirudin (11 units/ ml), dramatically reduced NFT-positive staining induced by 100 nM (equivalent to 11 units/ml) ␣-thrombin (Fig. 3C, lanes 8  and 9, respectively). These results suggest that proteolytic activity is required for thrombin to induce tau aggregation, and also imply that one or more PARs, and not a non-PAR thrombin receptor, might mediate this thrombin effect. Among the four PARs (PAR1-4) identified so far, thrombin activates PAR1, PAR3, and PAR4, but not PAR2 (49,(55)(56)(57)(58)(59)(60). Because PAR3 has been suggested to mainly function as a chaperone for PAR4 rather than transducing signals on its own (20, 61), we have concentrated on determining the roles of PAR1 and PAR4 in the thrombin-induced tau aggregation in murine hippocampal neuronal cells. For this purpose, we first examined expression of PAR1 and PAR4 in HT22 cells using both Western blot and RT-PCR (Fig. 5), and found that HT22 cells expressed both PAR1 and PAR4 mRNA and proteins. Moreover, when HT22 cells were treated with PAR1 agonist peptide (PAR1AP, SFLLRN) or PAR4AP (GYPGKF) at 100 M, ICC with AB1518 and thioflavin-S staining (Fig. 2, G-L), the data revealed intensified IB-like staining similar to that in-duced by thrombin following both PAR1AP and PAR4AP treatments. Moreover, Western blots with A24 (Fig. 4), AB1518, pSpS 199/202 , and pS 422 (Fig. 6, A-C, respectively) tau Abs confirmed that similar to thrombin, PAR1AP and PAR4AP both were able to induce tau hyperphosphorylation and aggregation, although subtle differences between effects of PAR1AP and PAR4AP were observed (e.g. PAR1AP appeared to be less effective than PAR4AP). These data suggested that the HT22 cell not only express PAR1 and PAR4, but that PAR1AP and PAR4AP can mimic the effects of thrombin on tau phosphorylation and aggregation without PAR cleavage. Taken together, these results suggest that both PAR1 and PAR4 activation may mediate thrombin-induced tau hyperphosphorylation and aggregation in HT22 hippocampal neurons.
Prolonged p44/42 MAPK Activation Contributed to Thrombin-induced Tau Hyperphosphorylation and Aggregation via PAR4 Activation-As indicated above, thrombin-induced tau aggregation in murine hippocampal neurons is likely to be promoted by tau hyperphosphorylation. A number of tau protein kinases (TPKs), such as TPK I (glycogen synthase kinase-3␤, GSK3␤) and TPK II (cyclin-dependent kinase 5, CDK5) as well as MAPKs, have been suggested to be responsible for tau hyperphosphorylation under various conditions (62)(63)(64)(65)(66)(67)(68)(69)(70)(71)(72). The earlier ICC studies have revealed several specific phosphorylation sites that are likely to be involved, which were further confirmed by Western blots. To identify potential responsible TPKs in thrombin-induced tau aggregation, we treated HT22 cells with thrombin (100 nM), PAR1AP (100 M), and PAR4AP (100 M) compared with control media in the absence and presence of specific inhibitors for several kinases, Western blots with the panel of phospho-specific tau Abs under native conditions were performed to evaluate the treatment effects on tau phosphorylation and aggregation simultaneously. These results indicated that neither the TPK I inhibitor, LiCl (20 mM), nor TPK II inhibitors, BL-1 (10 M) or roscovitine (10 M), had dramatic effects on thrombin-induced tau aggregation (data not shown). In stark contrast, a specific p44/42 MAPK inhibitor, PD98059 (10 M), almost completely abolished PAR4AP-induced tau hyperphosphorylation and aggregation (Fig. 6, A-C). PD98059 also appeared to significantly reduce the HMW tau bands induced by thrombin, although the low molecular weight tau bands appeared to be slightly increased. In addition, although PAR1AP induced relatively less intensified HMW tau bands, as compared with thrombin and PAR4AP, a mild reduction of the HMW tau bands induced by PAR1AP was also noticed when p44/42 MAPKs were inhibited by PD98059. Supporting a role for p44/42 MAPKs in tau hyperphosphorylation and aggregation induced by thrombin signaling, we found that thrombin, PAR1AP, and PAR4AP activated p44/42 MAPKs in 30 min while the p44/42 MAPK activation was prolonged for at least 24 h in HT22 cells treated with thrombin and PAR4AP, but not PAR1AP (Fig. 7A). The persistence of PAR4, but not PAR1, signaling agrees with previous observations of differential desensitization of PAR1 and PAR4 in other cell types (73). 2 Therefore, these results suggest that persistent p44/42 MAPK activation may be responsible for many, if not most effects, of PAR4AP and thrombin mediated via PAR4 activation on tau hyperphosphorylation and aggregation in murine hippocampal neuronal cells.
Thrombin-induced Tau Aggregation Is followed by Delayed Neuronal Degeneration-Hyperphosphorylation of tau is known to cause microtubule disassembly which, in turn, can lead to neuronal death (74 -76). Whether or not thrombin- induced tau hyperphosphorylation and aggregation had any impact on neuronal viability was further evaluated using the LDH assay. We found that thrombin treatment by 24 h appeared to partially neuroprotect against spontaneous cell death in serum-free differentiation medium (p Ͻ 0.05), whereas PAR4AP marginally increased the neurotoxicity (p Ͻ 0.05), and PAR1AP had no significant effect on cell viability (Fig. 8A). However, by 48 h, the neurotoxicity of PAR4AP was significantly increased (p Ͻ 0.01), whereas both thrombin and PAR1AP showed a trend to increase toxicity but neither was statistically significant. By 72 h exposure, thrombin (p Ͻ 0.01), PAR1AP (p Ͻ 0.01), and PAR4AP (p Ͻ 0.001) caused a significant increase of the cell toxicity as compared with control. These results suggest that the rapid tau hyperphosphorylation and aggregation induced by thrombin and PAR agonist peptides were followed by a delayed increase in hippocampal neuronal death.
Consistent with the LDH release, we also found a reduction of synaptophysin levels using Western blots in PAR4AP-treated cells by 48 h and in thrombin and PAR1AP-treated cells by 72 h (Fig. 7B). These results imply that neuronal death may be initiated at synapses (77), although such contacts are not clearly evident in these cultures. In addition, when cells were co-treated with 10 M PD98059 (Fig. 8B), the toxicity induced both by thrombin and PAR4AP was largely prevented (p Ͻ 0.01, thrombin versus PD98059; p Ͻ 0.001, PAR4AP versus PD98059). A trend of decrease in toxicity caused by PAR1AP was also noted when PD98059 was used, but such a change was not statistically significant. Moreover, Western blot with a mAb (Ab127) made against a synthetic peptide KGDEVD (a caspase cleavage domain) (33,78) revealed increased caspase cleavage products in thrombin-, PAR1AP-, and PAR4AP-treated cells by 72 h. Cotreatment with PD98059 essentially abolished the increase of Ab127-immunopositive products in cells treated with thrombin  and PAR4AP, but not PAR1AP (Fig. 7C). These data indicate that at least part of the thrombin/PAR-mediated neurotoxicity was associated with apoptosis, and that both the rapid tau hyperphosphorylation and aggregation, as well as the delayed neurodegeneration induced by thrombin signaling, particularly via PAR4, can be attributed, at least in part, to the persistent p44/42 MAPK activation. DISCUSSION In this study, we demonstrate for the first time that exogenous thrombin is sufficient to induce rapid formation of tau protein aggregates that contain hyperphosphorylated tau and are positive for thioflavin-S staining in HT22 mouse hippocampal neurons. Moreover, this tau aggregation is followed by a delayed apoptotic hippocampal neuronal degeneration. Therefore, we conclude that thrombin is capable of inducing neurofibrillary degeneration in hippocampal neurons in vitro. Mechanistic studies revealed that the proteolytic activity of thrombin is required, and that activation of both PAR1 and PAR4 may mediate thrombin effects. The persistent p44/42 MAPK activation appears to be the critical intracellular signaling event responsible for the neurofibrillary degeneration when PAR4 is activated by either PAR4AP or thrombin, whereas critical signaling events via PAR1 remain to be identified. In this regard, we did use specific inhibitors for other TPKs, such as GSK3␤ (e.g. LiCl), in the present study and found little or no effect. This is relevant because a previous report indicated that in human SH-SY5Y cells, lysophosphatidic acid, operating through a G protein-coupled receptor similar to PARs but not activated by cleavage, caused neurite retraction mediated via Rho activation and tau hyperphosphorylation by activating GSK3␤ (79). On the other hand, LiCl only partially prevented lysophosphatidic acid-mediated neurite retraction in the same cells (80). Although lysophosphatidic acid and thrombin are similar in their effects on neurons, this tau protein kinase difference may be significant.
The critical role of thrombin as a coagulation enzyme in the blood clotting system is well established. However, its roles in cellular signaling, especially in the central nervous system, are still being established. Accumulating evidence suggests that thrombin may regulate the functions of most, if not all, types of nervous system cells, perhaps as a natural response to cerebrovascular damage. These include reversal of stellation of astrocytes (26, 81) and activation of microglia (29 -31, 82), perhaps to propagate proinflammatory responses and direct actions on neurons. Regarding the direct effects of thrombin on neurons, in vitro studies have shown that thrombin can induce rapid growth cone collapse, neurite retraction, and neuronal degeneration in neuroblastoma cells (83)(84)(85), motor (33,86), and sensory neurons (34,87). Curiously, hippocampal neurons were reported to be relatively more resistant to thrombin treatment (85,88). Accompanying the neurite retraction in motor neurons, we recently found that tau proteins appeared to be rapidly degraded, in contrast to their aggregation and promotion of neurofibrillary degeneration in hippocampal neurons, in response to thrombin. 3 Thus, these results agree with previous observations suggesting a differential neuronal vulnerability response to thrombin in different neuronal subpopulations, and provide further insights into the mechanisms underlying hippocampal neuronal degeneration induced by thrombin. Moreover, induction of neurofibrillary degeneration in hippocampal neurons by thrombin directly links thrombin, a blood-derived vascular damage-associated risk factor, for the first time, to neurofibrillary degeneration, a typical pathological change in AD, traumatic brain injury, and other tauopathies, implicating an important pathogenetic role of thrombin in these diseases.
In addition to this phenomenologic discovery that thrombin induced neurofibrillary degeneration in hippocampal neurons, our mechanistic studies also indicate that this phenomenon is largely dependent on the proteolytic activity of thrombin and is mediated by thrombin-activated PARs. A previous report suggested that a naturally occurring proteolytic fragment of apolipoprotein E (apoE), cleaved by thrombin (89,90), induced NFTlike inclusions in cultured neurons (91). This finding provided an indirect link between thrombin and neurofibrillary degeneration. Such a mechanism may be additional to results from our experiments in this study, in which we used serum-free cultured murine hippocampal neurons, thus excluding the possibility of involvement of serum-derived apoE in this particular experimental paradigm. Moreover, the fact that PAR1AP and PAR4AP mimicked induction of neurofibrillary degeneration by thrombin suggests that thrombin can directly activate one or more of its cellular receptors to induce neurofibrillary degeneration. Among the three PARs that can be activated by thrombin, PAR1 is sensitive to low dose of the serine protease, whereas PAR4 only responds to high dose thrombin (20). Although PAR3 can also be activated by low dose thrombin, it does not transduce intracellular signals (20,61). Moreover, intra-brain injection of thrombin demonstrated that low dose thrombin may be neuroprotective, whereas high dose thrombin is likely to be destructive (92)(93)(94). Supporting this notion, our recent studies with murine microglia in culture also suggested that although microglial cells express both PAR1 and PAR4, only PAR4 activation can induce potentially neurotoxic tumor necrosis factor-␣ release (31,95). Therefore, it is likely that PAR4 may have more significant roles in various pathogenetic conditions than PAR1. Along these lines, the current study demonstrated that the neurofibrillary degeneration mediated via PAR4 could be largely attributed to persistent p44/42 MAPK activation. This finding calls attention for pathogenetic roles of PAR4 signaling in cerebrovascular damageassociated neurodegenerative disorders, and highlights potential therapeutic values of reagents that can inhibit PAR4 signaling or the persistent p44/42 MAPK activation in these diseases.
In practical terms, lack of in vitro neurofibrillary degeneration models has significantly limited the related therapeutic development using high throughput screening technology. When lysosomal function in apoE-deficient mouse hippocampal slice cultures is inhibited by cathepsins B and L, the hippocampal neurons develop AD-like NFTs (96). This model requires somewhat more complicated hippocampal slice culture, perhaps being practically less convenient. Moreover, inhibition of lysosomal function may cause nonspecific accumulation of multiple neuronal proteins in addition to tau. Therefore, the specificity and relevance of such a model to neurofibrillary degeneration in AD and other tauopathies are doubtful. Other recently described in vitro NFT models, such as overexpression of truncated apoE (91), human tau (97), or GSK3␤ (98) and heat shock (99), each may mimic certain limited aspects for neurofibrillary degenerative processes, making them potentially useful but currently incomplete. The present model in this study clearly demonstrates neurofibrillary degeneration in an immortalized hippocampal neuronal cell line induced by a known disease-associated risk factor, thrombin. It does so reproducibly and in a relatively short time frame. Although studies in transgenic mice indicate that murine tau does not seem to form typical mature NFTs, hyperphosphorylation of tau is often present (100 -103), and could form pre-tangle under certain circumstance (104). This in vitro study recapitulates most of the critical events in neurofibrillary degeneration including tau hyperphosphorylation, formation of NFT-immunopositive, and ␤-sheets containing (thioflavin-S-positive and SDS-resistant) tau aggregates as well as the delayed neuronal apoptosis. Therefore, it complements existing in vitro NFT models and represents a more ideal and convenient model with the potential to become an high throughput screening model for neurofibrillary degeneration.
Taken together, this study demonstrates that the blood-derived, vascular damage-associated factor, thrombin, can directly act on hippocampal neurons to induce neurofibrillary degeneration. This process is dependent on proteolytic activity of thrombin, mediated by activation of both PAR1 and PAR4, and intracellularly, at least in part, related to the prolonged p44/42 MAPK activation. These findings suggest that thrombin may not only be a risk factor but also a causal factor for neurofibrillary degeneration, not only in AD but other tauopathy subpopulations associated with cerebrovascular damage. Such knowledge, and the potential targets implicated in this study, may be instrumental in transforming therapeutic paradigms for AD and related disorders.