Truncation of tau selectively facilitates its pathological activities

Neurofibrillary tangles of abnormally hyperphosphorylated tau is a hallmark of Alzheimer’s and related tauopathies. Tau is truncated at multiple sites by various proteases in AD brain. While many studies have reported the effect of truncation on the aggregation of tau, these studies mostly employed highly artificial conditions, using heparin sulfate or arachidonic acid to induce aggregation. Here, we report for the first time the pathological activities of various truncations of tau, including site-specific phosphorylation, self-aggregation, binding to hyperphosphorylated and oligomeric tau isolated from AD brain tissue (AD O-tau), and aggregation seeded by AD O-tau. We found that


INTRODUCTION
Tau is a major neuronal microtubuleassociated protein. Aggregation of tau into neurofibrillary tangles (NFTs) is a hallmark brain lesion of Alzheimer's disease (AD) and related tauopathies (1-4). The number of NFTs correlates with the severity of dementia (5)(6)(7), and the regional distribution of tau pathology is apparently associated with the progression of AD (8,9).
In AD brain, tau is abnormally hyperphosphorylated at many sites (10)(11)(12). Hyperphosphorylation of tau leads to its loss of normal function, gain of neurotoxicity, and aggregation into NFTs (13)(14)(15). Tau pathology can be induced in the mouse brain by injection of tau aggregates isolated from AD or tau transgenic mouse brains or produced in vitro (16)(17)(18)(19)(20). Abnormally hyperphosphorylated and cytosolic tau isolated from AD brain (AD P-tau) by sedimentation and ion-exchange chromatography (10) can sequester/capture normal tau and template it into filaments in vitro (13). We recently found that like AD P-tau, oligomeric tau isolated from AD brain (AD Otau) by sedimentation also effectively induces tau aggregation in cultured cells (21), and templates tau pathology in vivo (22-24) in a prion-like fashion, which may underlie the amplification and propagation of tau pathology throughout AD brain.
In addition to hyperphosphorylation, tau is abnormally truncated at multiple sites in AD brain (25,26). Many proteases, including calpains and caspases, proteolyze tau in vitro and in vivo (27)(28)(29). Tau can be cleaved by caspase 6 at Asp13 and Asp402, by caspase 2 at Asp314, by caspase 3 at Asp25 and Asp421, by chymotrypsin at Tyr197, by an unknown thrombin-like cytosolic protease at Lys257, by asparaginyl endopeptidase at Asn255 and Asn368, and by calpain at Lys44 and Arg230 (30)(31)(32). At Glu391, tau is cleaved by an unknown protease (30). Puromycinspecific aminopeptidase (PSA) proteolyzes residues stepwise from the N-terminus of tau (4,30). In AD brain, various tau fragments have been identified (29,33,34). In addition to these protease-mediated cleavages, 21 novel proteolytic fragments of tau have been identified (35), which, as of yet, have not had a protease identified for their generation (33,35,36). Among all truncations of tau, truncations at Glu391 and Asp421 were the most reported in AD brain. We recently found that SDS-and reducing agent-resistant high molecular weight (HMW) aggregates of tau from AD brain lack the N-terminal portion (21,37). Many studies showed that truncation of tau promotes its aggregation (30,38), implying that tau truncation may play a critical role in tau pathogenesis (25,33). However, these studies were carried out employing mostly heparin or arachidonic acid for induction of aggregation, which are highly artificial conditions. The impact of tau truncation on its pathological activities, including hyperphosphorylation, selfaggregation and binding to and aggregation seeded by AD O-tau is not well documented. Based on the terminal acidic regions of tau and the truncations reported in AD brain (25,26,39,40), in the present study, we generated tau truncations with deletions from both the N-and C-termini and determined their pathological activities. We found that tau truncation from either the N-or C-terminal toward MTBR modulated the site-specific phosphorylation and enhanced its selfaggregation, its binding to AD O-tau and aggregation seeded by AD O-tau. Among these truncations, Tau151-391 showed the highest pathological activities and aggregation induced by AD O-tau, both in vitro and in cultured cells.

Construction and expression of various truncation forms of tau
In AD brain, tau is truncated at multiple sites by various proteases (25,33). Normally, the acidic termini of tau interact with the microtubule-binding repeats to prevent its aggregation (41). Based on the truncation sites reported in AD brain and on the acidic terminal regions of tau (Fig. 1A), we constructed 12 truncation forms of tau from both the N-and the C-termini of the protein (Fig. 1A), expressed them in HEK-293FT cells, and analyzed the expression by Western blots developed with a battery of monoclonal tau antibodies (Fig. 1B). Western blots showed that the truncation forms with the first 50 amino acid (a.a.) deletion-Tau51-441, Tau51-421, and Tau51-391-were not recognized by antibody 43D (6-18 a.a.). Truncation with the first 150 a.a. deletion-Tau151-441, Tau151-421, and Tau151-antibody Tau46.1 (428-441 a.a.) (Fig.1C). Monoclonal antibody (mAb) 77G7 against MTBR was able to detect all truncation forms (Fig. 1C). All truncation forms were tagged with HA at the N-terminus and were detected by anti-HA (Fig. 1C). These results indicate that the 12 truncation forms of tau were constructed and expressed as expected.
To learn the toxicity of these taus, we overexpressed tau1-441 and 11 truncates of tau in HEK-293T cells and analyzed lactate dehydrogenase (LDH) activity in the medium leaked from cells 48 hr after transfection. We found similar levels of released LDH activity from cells overexpressed with above 12 forms of tau (Fig. 1D). We further analyzed cell viability in HEK-293T cells by Cell Counting kit 8 (CCK8) and found similar cell viability in cells with overexpression of the 12 taus (Fig. 1D). Thus, tau truncates showed similar effect in cytotoxicity and cell viability as the full-length protein.

N-and C-terminal truncations modulate tau phosphorylation
Abnormal hyperphosphorylation is central to tau pathogenesis (3). To determine the effect of tau truncation on its phosphorylation, we overexpressed the above truncated taus in HEK-293FT cells and analyzed phosphorylation by Western blots developed with site-specific and phosphorylationdependent tau antibodies. We found no detectable tau phosphorylation at Ser199, Thr205, Thr212, Ser214, or Thr217 in Tau231 Tau-421 and Tau-391 truncation forms (Fig. 2K), confirming the sitespecific phosphorylation of tau.
To simplify the comparison, we first analyzed the impact of N-terminal truncation on tau phosphorylation. We found that deletion of the first 50 a.a. did not affect tau phosphorylation at Ser199 ( Fig. 2A), Thr212  (Fig. 2K), regardless of C-terminal truncations, but slightly suppressed tau phosphorylation at Thr205 (Fig. 2B) and Thr231 (Fig. 2F).
Deletion of the first 150 a.a. suppressed tau phosphorylation at Ser199, but significantly increased it at most sites, except at Ser396 and Ser422, independent of the C-terminal truncation (Fig. 2), suggesting that deletion of the first 150 a.a. enhances tau phosphorylation at most sites. Deletion of the first 230 a.a. did not change (Tau-441 and Tau-421) or decreased (Tau-391) Ser235 phosphorylation (Fig. 2G), but significantly increased tau phosphorylation at Ser262 ( Thus, deletion of the first 50 a.a. did not affect, but deletion of the first 150 a.a. significantly enhanced, tau phosphorylation regardless of the C-terminal truncation. The increase in phosphorylation in tau with deletion of the first 230 a.a. was dependent on intact or partially intact C-terminus. Then, we determined the effect of the Cterminal truncations on tau phosphorylation. We found that the C-terminal truncation also modulated tau phosphorylation sitespecifically (Fig. 3). Deletion of the last 20 a.a. decreased Ser199 phosphorylation in Tau1-, but not in the Tau51-and Tau151-truncation forms (Fig. 3A). Deletion of the last 20 a.a. did not significantly affect tau phosphorylation at Thr205, Thr212, Ser214, Thr217 or Ser262 ( Fig. 3B-E), increased at Ser235 (Fig. 3G), and decreased at Ser396 and Ser404 ( Fig. 3I and J) in most truncation forms. Deletion of the last 50 a.a. increased or tended to increase tau phosphorylation at Thr205, Thr212, Ser214, Thr 217, and Ser235, but did not significantly affect phosphorylation at Ser262 in the most Nterminal truncates (Fig. 3B-E). Interestingly, deletion of the last 20 or 50 a.a. suppressed Ser262 phosphorylation in Tau231-truncation forms (Fig. 3H), and deletion of either the last 20 or 50 a.a. almost abolished or dramatically suppressed tau phosphorylation at Thr231, Ser396, and Ser404 regardless of N-terminal truncation (Fig. 3F,I,J), suggesting that the last 20 a.a. play a critical role in the regulation of phosphorylation at these sites. These findings 4 suggest that generally, C-terminal truncation can increase the phosphorylation of tau at the sites located in proline-rich domain (Thr205, Thr212, Ser214, Thr217), but decrease at sites in microtubule-binding repeat (Ser262) and Cterminal domains (Ser396 and Ser404).

Truncation of tau promotes its aggregation
Tau is aggregated into NFTs in AD brain. To learn the impact of truncation in tau aggregation, we overexpressed various HAtagged truncated taus in HEK-293FT cells for 48 hr, separated RIPA buffer-soluble andinsoluble fractions, and analyzed tau levels by Western blots developed with anti-HA. We found a significant amount of RIPA-insoluble tau in cells expressed with tau truncations, except Tau231-391 (Fig. 4A). Interestingly, we found ~70 kDa HMW-tau only in the RIPAinsoluble fraction of Tau151-391 ( Fig. 4A), suggesting that Tau151-391 formed the AD-like SDS-and reducing agent-resistant aggregates described previously (21,37).
We determined the impact of N-terminal truncations on tau aggregation by analyzing the ratio of RIPA-insoluble/-soluble tau. We found that in general, deletion from the Nterminus toward MTBR increased the ratio of RIPA-insoluble/-soluble tau (Fig. 4A,B), suggesting that N-terminal truncation increases tau's capacity to aggregate. Deletion of the first 50 a.a. did not affect significantly the ratios of RIPA-insoluble/-soluble tau, regardless of the C-terminal truncations (Fig.  4A,B). However, deletion of the first 150 a.a. significantly increased the ratio of insoluble/soluble tau in Tau-441 and Tau-421, but not in Tau-391 truncation forms (Fig. 4A,B), suggesting that the first 150 a.a. inhibit tau aggregation, but this inhibition is dependent on the intact or partially intact C-terminus. Compared with Tau151-, Tau231-forms did not further enhance the ratios of RIPA-insoluble/soluble tau (Fig. 4A,B), suggesting that the first 150 a.a. are essential to inhibit tau selfaggregation. The expression level of Tau231-391 was very low, which may lead to undetectable aggregation of Tau231-391. These findings suggest that N-terminal truncation enhances tau aggregation, and the first 150 a.a. are critical to prevent tau from self-aggregation.
Deletion of C-terminal 20 a.a. did not affect the ratio of RIPA-insoluble/-soluble tau in Tau1and Tau50-, but decreased in Tau151-and Tau231truncation forms (Fig. 4A,C), suggesting that the last 20 a.a. may not protect tau from the self-aggregation. However, deletion of the last 50 a.a. significantly increased the aggregation regardless of N-terminal truncations except Tau231-truncation forms (Fig. 4A,C), suggesting that the last 50 a.a. are also critical to prevent tau from self-aggregation. Taken together, these findings indicate that both the first 150 a.a. and the last 50 a.a. protect tau from self-aggregation, and deletion of either the first 150 a.a or the last 50 a.a. promotes tau aggregation. Tau151-391 showed the highest self-aggregation capacity (Fig. 4).
To investigate the cause of low expression of Tau231-391 we determined mRNA levels of taus by real-time quantitative PCR 48 hr after transfection. We found similar mRNA levels in these 12 tau forms expressing cells (Fig. 4E), suggesting lower level of Tau231-391 could have resulted from instability of this protein fragment.

Truncation of tau enhances its capture by AD O-tau
Isolated AD O-tau was previously found to sequester/capture normal tau (13,14), which is a basis of tau aggregation induced by pathological tau. To learn the effect of tau truncation on its capture by AD O-tau, we performed overlay assay. We overexpressed HA-tagged truncated taus in HEK-293FT cells. The cell extracts containing similar levels of truncated taus in PBS were incubated with nitrocellulose membrane pre-dotted with various amounts of AD O-tau overnight. AD Otau-captured truncated taus were detected by development with anti-HA followed by HRPanti-IgG and ECL. We found that tau with various truncations was captured differentially by AD O-tau (Fig. 5A). Levels of truncated taus captured by AD O-tau were increased with truncation from N-terminal toward MTBR (  Tau-441 and Tau-421, but not in Tau-391 truncation forms (Fig. 5A,B). In the C-terminal truncations, deletion of the last 20 a.a. did not affect the capture of tau by AD O-tau (Fig. 5A,C). Although deletion of the last 50 a.a. did not enhance tau capture by AD O-tau in Tau1-and Tau51-truncations, it enhanced it in Tau151-and decreased it in Tau231-forms (Fig. 5A,C). These results suggest that the first 150 a.a. and the last 50 a.a. protect tau from being captured by AD O-tau, and their deletions enhance tau's capture by AD O-tau; Tau151-391 is the tau most potently captured by AD O-tau.

Truncation of tau enhances its aggregation seeded by AD O-tau
Misfolded tau templates tau aggregation in vitro and in vivo, which is a basis of propagation of tau pathology (14,42). To determine the effect of tau truncations on AD O-tau-seeded aggregation, we overexpressed tau truncations in HEK-293FT cells and treated these cells with AD O-tau for 42 hr after 6 hr transfection. RIPA-soluble and -insoluble taus from these cells were analyzed by Western blots. We found that the AD O-tau treatment caused the reduction of RIPA-soluble tau (  6A). The expression level of Tau231-391 was extremely low, and the RIPA-insoluble tau was undetectable in the cells treated with AD O-tau at this condition (Fig. 6A). Thus, collectively these findings suggest that AD O-tau induces tau aggregation, which is modulated by truncations.
To compare the efficiency of aggregation of various truncated taus seeded by AD O-tau, we analyzed the ratio of RIPA-insoluble/soluble tau. We found that deletion of the Nterminus toward MTBR increased AD O-tau-seeded aggregation regardless of C-terminal truncation (Fig. 6A,C). Similar to in the selfaggregation, deletion of the first 50 a.a. did not significantly affect tau aggregation templated by AD O-tau, but deletion of the first 150 a.a. enhanced AD O-tau-seeded aggregation in all C-terminal truncations (Fig. 6A,C). Deletion of the N-terminal 230 a.a. enhanced AD O-tauseeded aggregation in Tau-441 and Tau-421, but not in Tau-391 truncation forms (Fig. 6A,C). Tau151-391 was the most potent truncation to the aggregation induced by AD O-tau ( Fig. 6A-D). Deletion of the last 20 a.a. did not affect AD Otau-induced aggregation, but suppressed it in Tau231-truncations (Fig. 6A,D). Deletion of the last 50 a.a. enhanced tau aggregation templated by AD O-tau, except in Tau231truncations ( Fig. 6A,D). These data suggest that consistently, the first 150 a.a. and the last 50 a.a. of tau prevent AD O-tau-seeded aggregation. Collectively, these findings reveal that deletions of either the first 150 a.a. or the last 50 a.a. can enhance AD O-tau-seeded tau aggregation, and that Tau151-391 is the most prone to be seeded to aggregate by AD O-tau.

AD O-tau induces Tau151-391 aggregation in vitro and in cultured cells
Our studies described above indicated that Tau151-391 appears to have the highest activity to self-aggregate and bind to AD O-tau. Tau aggregation can be induced in vitro by polyanionic molecules, such as heparin, heparan sulfate, and RNA (43)(44)(45). We further studied its aggregation induced in vitro and in cultured cells. We first determined the aggregation of Tau151-391 induced by heparin. We incubated recombinant Tau1-441 (rTau1-441) or rTau151-391 with heparin in the presence of Thioflavin T for various time periods at room temperature. Aggregated tau was detected by fluorescence density and plotted against the incubation time. We found that heparin induced aggregation much faster and to a much higher extent in Tau151-391 than in Tau1-441 (Fig. 7A). Then, we studied tau151-391 aggregation induced by AD O-tau. We incubated rTau151-391 or rTau1-441 with AD O-tau for various time periods at room temperature and detected aggregation by Thioflavin T fluorescence. We found that AD O-tau induced aggregation of Tau151-391, but not Tau1-441, in a time-dependent manner (Fig. 7B). These results reveal that Tau151-391 has a higher capacity than Tau1-441 to be induced to aggregate by either heparin or AD O-tau in vitro.
To further verify AD O-tau-induced aggregation, we overexpressed HA-tagged Tau1-441 or Tau151-391 in HeLa cells and treated the cells with AD O-tau for 42 hr after 6 hr transfection.
The cells were then immunofluorescence-stained with anti-HA. Whereas we found <1% of Tau1-441 cells with aggregates (Fig. 7C,G), ~ 6% cells with atypical aggregates were seen in the Tau151-391 cells (Fig. 7D,G). However, we found that the treatment with AD O-tau markedly increased the number (~22%) of Tau151-391 cells with aggregates in the cytoplasm and the nucleus (Fig. 7F,G). This increase was only modest (`6%) in Tau1-441 cells (Fig. 7E,G). The aggregates of Tau151-391, but not Tau1-441, seeded by AD O-tau were thioflavin T positive (Fig. 7H). Thus, Tau151-391 was found to display a high potency for templation and aggregation by AD O-tau.

DISCUSSION
Since the first report on tau truncation in AD paired helical filament in 1993 (36), more than 50 fragments of tau were identified and over 30 were found present in AD brain (33). However, the role of truncation in tau pathogenesis in AD and related tauopathies was not well understood. Based on several reported cleavage sites, in the present study we studied effects of different tau truncations on pathological activities. We found that truncation at N-and C-termini both enhanced pathological activities of tau, including an increase of site-specific hyperphosphorylation, self-aggregation, capture by AD O-tau, and AD O-tau-seeded aggregation (Fig. 8). Among the truncated fragments, Tau151-391 was found to be the most pathogenic.
Prion-like activity of pathological tau may underlie the progression of AD and related tauopathies. Our group previously found that recombinant tau can be sequestered by AD Ptau by overlay assay (13). In the present study, we developed a novel assay with AD O-tau as capture to quantitively validate the ability of tau "seeds" to capture pro-aggregation truncated tau fragments, which is a significant advance over artificial conditions using heparan or other methods to induce aggregation. Here we found that N-terminal truncation is particularly significant to pathogenesis, while prior work focused on C-terminal truncation. Our present study also challenges the dogma of the importance of caspase cleavage at Asp421 since we did not find any increase in pathogenic activity in tau truncated at this site. The capture assay identified pathological cores for assembly, which mimic the early steps of tau aggregation. Similarly, the method also can be used to detect prion-like activity of pathological tau by overlay with crude extract of HEK-293FT/HA-Tau151-391.
Tau RD P301S FRET Biosensor, in which a HEK-293 cell line that stably expresses the repeat domain (amino acids 244-372) of 2N4R tau with two disease-associated mutations (P301L and V337M), is widely used to assess prion-like seeding activity (Sanders et al., 2014). We here expressed various forms of HA-tau in HEK-293FT cells, treated the cells with AD O-tau for 42 hr after 6 hr transfection and harvested the cells with RIPA buffer. Aggregated tau induced by AD O-tau was yielded in RIPA buffer insoluble fraction by 130,000 x g for 45 min and quantified by immuno-blots. By using this method, we found different efficacies of various tau truncates in AD O-tau seeded aggregation in cultured cells and Tau151-391 was found to be the most prone to be induced to aggregate. Consistently, we found that N-terminal truncation is more significant to pathogenesis than C-terminal truncation at Asp421. This method also can be used to evaluate the seeding activity of pathological tau from AD and 3xTg-AD mouse brains (21).
Heparin and other polyanionic molecules had been shown to induce tau aggregation in vitro (44)(45)(46). Therefore, heparin-induced tau aggregation assay combined with Thioflavin T fluorescence detection was most commonly used to access tau aggregation (47,48). Nevertheless, heparin-induced filament core is structurally different from that seen in the AD brain (48). In this study, tau151-391 aggregation in vitro was induced by incubating with AD Otau. However, the prion-like activity of AD Otau induced-tau aggregates in vivo remains to be investigated.
Tau molecule shows a preference to form a paperclip-like conformation by folding the acidic N-and C-terminal portions back on the basic MTBR region (49) (Fig. 8), which might protect tau from self-aggregation. It was reported that two sequences, 275VQIINK280 and 306VQIVYK311, within repeat (R) 1 and 2, respectively, of the MTBR are required for selfaggregation of tau protein (50,51). In the present study, we found that deletion of either the first 50 a.a. or the last 20 a.a., which removes part of the acidic N-and C-termini, did not significantly affect tau's self-aggregation and its binding to, and aggregation seeded by, AD O-tau. However, deletion of either the first 150 a.a. or the last 50 a.a., which removes the acidic portions of N-or C-terminus completely, markedly increased the pathological activities of tau (Fig. 8). Thus, both the N-terminal and C-terminal acidic portions of tau appear to protect tau from aggregation. Partial deletion of the N-or C-terminal acidic regions may not be able to disrupt the paperclip structure, but complete removal of these acidic regions probably disrupts the paperclip structure and increases the aggregation tendency. Thus, Tau151-391 is the most prone to (a) selfaggregation, (b) binding to AD O-tau, and (c) AD O-tau-seeded aggregation. Interestingly, deletion of 230 a.a. increased self-aggregation, its capture by AD O-tau, and AD O-tau-seeded aggregation in Tau-441 and Tau-421 truncation forms as compared with deletion of the first 150 a.a. However, while Tau231-391 was captured by AD O-tau efficiently, no self-aggregation and AD O-tau-induced aggregation of tau were detected in Tau231-391 cells. We found a very low expression level of Tau231-391 in cells, which could be a reason why we could not find any detectable tau aggregation with or without AD O-tau induction. However, mRNA level of Tau231-391 was similar to other tau truncates in cells, suggesting a lower stability of the protein.
Aggregation of hyperphosphorylated tau into NFTs is apparently central to the pathogenesis of AD and related tauopathies (3). In AD brain, tau is hyerphosphorylated and truncated. Caspase cleavage of tau may precede the hyper-phosphorylation, where especially caspase cleavage at Asp421 has been shown to initiate the cascade leading to tau aggregation (29,52,53). In the present study, we found that deletion of the first 50 a.a. did not affect significantly tau phosphorylation. However, deletion of the first 150 a.a. increased tau phosphorylation at Thr205, Thr212, Ser214, Thr217, Thr231, and Ser235, but not at Ser262, Ser396, Ser404 and Ser422, suggesting that deletion of the 150 a.a. makes the region upstream of MTBR, but not MTBR and downstream of MTBR, a favorable substrate for tau kinases and/or an unfovorable substrate for tau phosphatases. It is also possible that facilitation of tau phosphorylation by the truncation may result from the molecule opening or/and from reduction of acidic nature of the molecule by deletion of the N-terminal portion. Similarly, deletion of N-terminal 230 a.a. enhanced tau phosphorylation at the sites in MTBR and downstream of it. However, deletion of Cterminal either 20 a.a. or 50 a.a. did not increase or weakly increased tau phosphorylation at the majority of phosphosites, but suppressed that at Thr231, Ser262, Ser396, and Ser404. Thus, N-terminal truncation generally enhanced, but C-terminal truncation suppressed, tau phosphorylation. Hyperphosphorylation leads to tau aggregation (54). Pseudo-phosphorylation at Ser396 and Ser404 stimulates the rate of tau polymerization (55). Thus, truncation at the Nterminus may also enhance tau aggregation via hyperphosphorylation.
Many proteases are expressed in the infant brain, including ADAM10, thrombin, caspase and calpain. Under physiological circumstances, they are involved in the control of brain development via regulating multiple substrates in the signal cascades, brain inflammation, apoptosis and so on. In AD, upregulation and/or mislocalization of these proteases could lead to increased levels and/or inappropriate positioning of truncated tau. Unlike in infant or normal adult brain, in AD brain probably hyperphosphorylation opens tau's paper clip structure which makes its cleavage at aa 150 by thrombin and/or ADAM 10 possible (56). Furthermore, elevated level of tau in AD brain provides abdundant substrate for the proteases and aggravates the pathological accumulation of truncated fragments.
Two proteases, thrombin and A disintegrin and metalloproteinase 10 (ADAM10), which cleave tau at Arg155 and Ala152 respectively, might be responsible for N-terminal truncation of tau around Lys150 in vivo (33,57,58). Both enzymes are expressed in adult human brain (57,59). ADAM10, first identified as the primary -secretase of amyloid- protein precursor (APP) (60), showed a two-fold increase of mRNA levels in AD hippocampus (61). It is mainly localized to cell membrane and functions as a transmembrane protease, but its activity may not be limited to the membrane. For example, ADAM10 might exert proteolytic activity in its trafficking route to the plasma membrane and ADAM10 still functions at the early stages after endocytosis when it is not fully degraded (62,63). Although tau is mostly cytoplasmic, it is released in the extracellular space through multiple pathways including direct leakage, exocytosis, exosomes, synaptic vesicles or translocation across the plasma membrane (64). In addition, we can not exclude the possibility that tau might be cleaved extracellularly and then internalized. It was reported that ADAM10cleaved tau fragment (Tau-A) was elevated 10fold high in the brains of the Tg4510 mouse model of tau deposition (65). Although high levels of serum Tau-A were associated with lower risk of AD and dementia in a prospective study (66), Serum Tau-A levels did show a significant increase in AD and mild cognitive impairment (MCI) (67). In AD patients, Tau-A level in serum correlated inversely with the cognitive assessment score (Mattis Dementia Rating Scale) (65) and was related to the change of cognitive function over time in early AD patients (68), supporting that Tau-A could indicate the rate of clinical progression of AD.
Thrombin stains extra-and intracellular NFTs implying the enzyme could function both outside and inside cells (57). Nevertheless, the current evidence is not sufficient to reveal whether ADAM10 and thrombin cleave tau intracellularly or extracellularly or both.
NFTs in AD contains full length tau (69). The truncations probably occur after NFTs are formed and after cell death become extracellular. However, we recently reported that HMW-tau in AD brains lacked N-terminal portion (37). The truncations could play a significant role as aggregation-promoting seeds and thus in this way become important in promoting and accelerating secondary spread of tau pathology. Thus, it remains to be investigated whether these truncations occur prior to aggregation of tau in one or more tauopathies other than AD.
Truncations of tau at Asp421 (TauD421) and Glu391 (TauE391) are the most studied truncation forms in AD brain. Tau is cleaved at Asp421 by caspase-3 and at Glu391 by an unknown protease. TauD421 and TauE391 aggregate more rapidly and to a greater degree than full-length tau in the presence of heparin or arachidonic acid (38,55). TauD421 and TauE391 detected by the mAb C3 and mAb423 are also present at early stages of tau aggregation (70,71). Immunohistochemical studies indicate that cleavage at Asp421 occurs after formation of the Alz50 epitope but prior to truncation at Glu391. Thus, TauD421 appears to occur relatively early in the disease state, contemporaneous with the initial Alz50 folding event that heralds the appearance of filamentous tau in NFTs, neuropil threads, and the dystrophic neurites surrounding amyloid plaques (72). Levels of tau fragment cleaved at Asp421 (Tau-C) in serum were not significantly related to the change of cognitive function over time in early AD patients (68). In AD brain, the level of TauD421 is increased significantly (29,73,74), but was not found in HMW-tau (37). In the present study, we found that truncation of tau at Asp421, a partial deletion of the acidic C-terminal, did not enhance the selfaggregation in cultured cells and AD O-tauseeded aggregation in cultured cells, which may be due to the reduction of tau phosphorylation at Thr231, Ser262, Ser396, and Ser404. In a previous study we showed that hyperphosphorylation of tau at Thr231 and Ser262 are required for its self-aggregation into paired helical filaments in vitro (75). In contrast to TauD421, TauE391, the deletion of the whole acidic C-terminal region dramatically enhanced tau self-aggregation, capture by AD O-tau, and AD O-tau-seeded aggregation. Further deletion of N-terminal 150 a.a., which results in Tau151-391, the tau fragment of the protease-resistant core of PHFs isolated from AD brain (76), showed the highest pathological activities. Rats transgenic for tau151-391 develop neurofibrillary pathology and also display hyperphosphorylation and production of HMWtau species (25). In addition, SDS-and -ME resistant HMW-tau in AD brain homogenates was immunoreactive with anti-pS422-tau strongly. Thus, these findings argue against the impact of tau truncation at Asp421 in tau pathogenesis and suggest that N-terminal truncation is particularly significant to pathogenesis of AD.
Our previous studies showed that hyperphosphorylation of tau inhibits its binding to tubulin and promotion of the assembly and stabilization of microtubules and increases its self-aggregation and its capture of normal tau, the characteristics that suggest the opening of the paperclip structure of tau (13)(14)(15). The present study suggests that truncation of tau generating Tau151-391 results in similar characteristics. Furthermore, truncation of tau at Lys150 enhances tau phosphorylation. Thus, in addition to the inhibition of the hyperphosphorylation, inhibition of the proteases responsible for the truncation of tau at Lys150 and/or Glu391 could be required to effectively inhibit tau pathology.
We analyzed here tau truncation and its pathological changes in vitro and in cultured HEK-293FT cells which do not contain high concentration of assembled microtubules or axonal processes. Also, unique tau kinases sensitive phosphorylated sites and chaperones like HSP90 influence normal folding and what is exposed to proteases. Thus, the role of tau truncation and its phosphorylation, interaction with and aggregation seeded by AD O-tau remain to be studied in neuronal cells and in vivo.
In summary, the acidic N-and C-termini of tau suppress its pathological activities. Truncation of tau from the N-or C-terminal toward MTBR enhances its site-specific phosphorylation, self-aggregation, capture by AD O-tau, and AD O-tau-seeded aggregation. Deletion of either the first 150 a.a. or the last 50 a.a., but not the first 50 a.a. or the last 20 a.a., markedly increases the pathological activities, which may initiate tau pathogenesis in AD brain and related tauopathies. Tau151-391 is the most potent tau fragment for pathological activities, which can be used for seeding activity in vitro and in vivo. Inhibition of tau truncation may be a therapeutic approach to prevent or inhibit tau pathogenesis.

Plasmids, antibodies, and other reagents
pCI/HA-Tau truncations were generated by PCR amplification with the primers listed in Table 1 by using pCI/HA-tau as a template and were constructed into pCI-neo. All constructs were verified by DNA sequence analysis. The primary antibodies used in the present study are listed in Table 2. Peroxidase-conjugated anti-mouse and anti-rabbit IgGs were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Alexa Fluor 555-conjugated goat anti-mouse IgG was from Life Technologies (Rockford, IL, USA). Heparin and Thioflavin T were from Sigma-Aldrich Corp. (St. Louis, MO, USA). CyQUANT ™ LDH Cytotoxicity Assay Kit and the ECL kit was from ThermoFisher Scientific (Rockford, IL, USA). Cell Counting Kit 8 was from Dojindo molecular technologies (Rockville, MD, USA).

Western blots and immuno-dot blots
Samples were diluted with Laemmli SDS sample buffer, followed by boiling for 5 min. Protein concentration was measured using the Pierce™ 660 nm Protein Assay kit (ThermoFisher Scientific). Samples were subjected to SDS-PAGE and transferred onto polyvinylidene fluoride membrane (MilliporeSigma, Burlington, MA, USA). The membrane was subsequently blocked with 5% fat-free milk-TBS (Tris-buffered saline) for 30 min, incubated with primary antibodies (Table  2) in TBS overnight, washed with TBST (TBS with 0.05% Tween 20), incubated with HRPconjugated secondary antibody for 2 hr at room temperature (RT), washed with TBST, incubated with the ECL Western Blotting Substrate (ThermoFisher Scientific), and exposed to HyBlot CL® autoradiography film (Denville Scientific Inc., Holliston, MA, USA). Specific immunostaining was quantified by using the Multi Gauge software V3.0 from Fuji Film (Minato, Tokyo, Japan).
Tau level in samples was assayed by immuno-dot blots as described previously. Briefly, various amounts of a sample were applied onto nitrocellulose membrane (Schleicher and Schuell, Keene, NH, USA) at 5 μl per grid of 7 × 7 mm in size. The blot was placed in a 37 °C oven for 1 hr to allow the protein to bind to the membrane. The membrane was processed as Western blots as described above by using the mixture of R134d and 92e pan tau antibodies as primary antibodies.

Cytotoxicity and viability assay
Cell cytotoxicity was assessed by measuring LDH release from damaged or dead cells using a CyQUANT LDH Cytotoxicity Assay Kit (ThermoFisher Scientific) following the manufacturer's instructions. Briefly, HEK-293T cells in 96-well plate were transfected with pCI/HA-taus with FuGENE HD for 48 hr. The culture mediums were collected and incubated the Reaction Mixture for 30 min at RT. Maximum was yielded by adding lysis buffer to release all LDH to medium. The OD490 and OD680 were recorded by universal microplate spectrophotometer (BioTek, VT, USA) after adding stop solution. Relative cytotoxicity to tau1-441 was calculated according to the formulas in the manufacturer's instructions.
For cell viability assay, HEK-293T cells cultured in 96-well were transfected with pCI/HA-taus for 48 hr. After adding 10 μl CCK-8 to each well and incubating at 37 °C for 1 hr, OD450 was recorded and relative cell viability to tau1-441 was calculated.

Tau captured by AD O-tau assay
Various tau truncations tagged with HA at the N-terminus were overexpressed in HEK-293FT cells for 48 hr. The cells were lysed in phosphate-buffered saline (PBS) containing 50 mM NaF, 1 mM Na3VO4, 1.0 mM AEBSF, 5 mM benzamidine, and 10 μg/ml each of aprotinin, leupeptin, and pepstatin by probe sonication at 20% power for 2 min. The cell lysates were centrifuged for 10 min at 10,000 × g. The supernatants containing HA-tau truncations were stored at −80 °C until used.
Various amounts of AD O-tau isolated from AD brains were dotted on nitrocellulose membrane after serial dilution in TBS and dried at 37 °C for 1 hr. The membrane was blocked with 5% fat-free milk in TBS for 2 hr and incubated with the above cell extracts containing HA-tagged truncated taus overnight. After washing three times with TBST, the membrane was incubated with anti-HA in 5% milk in TBST overnight and processed as immune-dot blot as described above. Fluor 555-conjugated-secondary antibody for 2 hr at RT. After washing with PBS, the cells were mounted with ProLong™ Gold antifade reagent (ThermoFisher Scientific) and observed with a Nikon confocal microscope. Tau1-441 and Tau151-391 BL21/pGEX-6P1/Tau1-441 and BL21/pGEX-6P1/Tau151-391 cultured in LB medium were treated with 0.5 mM IPTG for 3 hr at RT to induce the protein expression, after which cells were centrifuged at 5,000 × g for 10 min at 4 °C. The resulting pellets were resuspended in TBS with 1 mM dithiothreitol (DTT) and cocktail of protease inhibitors for bacteria (Sigma) and probe-sonicated on ice with 70% power for 20 min. The bacterial lysate was adjusted to 1%Triton X-100 and centrifuged at 10,000 x g for 15 min. The supernatant was incubated with glutathione-Sepharose beads for 1 hr at 4 °C. The beads were agitated with PreScission TM Protease (Sigma-Aldrich) in buffer (50 mM Tris-HCl, pH 7.0, 150 mM NaCl, 1 mM EDTA, and 1 mM DTT) to cleave Tau from GST after extensive washing with TBS. Purified Tau was dialyzed against 5 mM MES, pH 6.8, and 0.5 mM EGTA and lyophilized.

Tau aggregation induced by heparin or AD O-tau in vitro
Recombinant tau (25 M) and heparin (0.5 mg/ml) or AD O-tau (0.53 mg/ml) in 50 l of PBS containing 50 M Thioflavin T and 1mM DTT were incubated by shaking at RT. Fluorimetry was performed by using a Spectra Max M5 fluorescence spectrophotometer (set at 440 nm excitation/538 nm emission) every 10 min.

Statistical analysis
The GraphPad Prism 6 software was used for statistical analysis. Results were analyzed by one-way ANOVA followed with Tukey's multiple comparisons test or by two-way ANOVA followed by Sidak's multiple comparisons test for multiple-group analysis.

Data Availability
All data are contained within the manuscript.

ACKNOWLEDGMENTS
We are thankful to Ms. Maureen Marlow for copy-editing the manuscript. This work was supported in part by funds from New York State Office for People With Developmental Disabilities and Nantong University and by grant from the U.S. Alzheimer's Association (DSAD-15-363172).

AUTHOR CONTRIBUTIONS
J.G., W.X., N.J., L.L., Y.Z., and D.C. performed experiments and analyzed results. C.-X. G. and K. I. provided the reagents, discussed results, and edited the manuscript. F. L. designed, supervised, and performed the experiments, analyzed and interpreted results, and wrote the manuscript.