Caspase-6-cleaved Tau fails to induce Tau hyperphosphorylation and aggregation, neurodegeneration, glial inflammation, and cognitive deficits

Active Caspase-6 (Casp6) and Tau cleaved by Casp6 at amino acids 402 (Tau∆D402) and 421 (Tau∆D421) are present in early Alzheimer disease intraneuronal neurofibrillary tangles, which are made primarily of filamentous Tau aggregates. To assess whether Casp6 cleavage of Tau contributes to Tau pathology and Casp6-mediated age-dependent cognitive impairment, we generated transgenic knock-in mouse models that conditionally express full-length human Tau (hTau) 0N4R only (CTO) or together with human Casp6 (hCasp6) (CTC). Region-specific hippocampal and cortical hCasp6 and hTau expression were confirmed with western blot and immunohistochemistry in 2–25-month-old brains. Casp6 activity was confirmed with Tau∆D421 and Tubulin cleaved by Casp6 immunopositivity in 3–25-month-old CTC, but not in CTO, brains. Immunoprecipitated Tau∆D402 was detected in both CTC and CTO brains, but was more abundant in CTC brains. Intraneuronal hippocampal Tau hyperphosphorylation at S202/T205, S422, and T231, and Tau conformational change were absent in both CTC and CTO brains. A slight accumulation of Tau phosphorylated at S396/404 and S202 was observed in Cornu Ammonis 1 (CA1) hippocampal neuron soma of CTC compared to CTO brains. Eighteen-month-old CTC brains showed rare argentophilic deposits that increased by 25 months, whereas CTO brains only displayed them sparsely at 25 months. Tau microtubule binding was equivalent in CTC and CTO hippocampi. Episodic and spatial memory measured with novel object recognition and Barnes maze, respectively, remained normal in 3–25-month-old CTC and CTO mice, in contrast to previously observed impairments in ACL mice expressing equivalent levels of hCasp6 only. Consistently, the CTC and CTO hippocampal CA1 region displayed equivalent dendritic spine density and no glial inflammation. Together, these results reveal that active hCasp6 co-expression with hTau generates Tau cleavage and rare age-dependent argentophilic deposits but fails to induce cognitive deficits, neuroinflammation, and Tau pathology.


Introduction
Tau is expressed as six isoforms in the human adult brain 1 , is principally located in axons, and promotes microtubules (MT) assembly and stabilization 2 . In Alzheimer disease (AD), Tau aggregates as paired helical filamentsforming neurofibrillary tangles (NFT) that accumulate in neuropil threads and neuritic plaques 3 . NFT initially occur in the trans-entorhinal region, progressively invade the brain through the subiculum to the hippocampus, and then cortical areas 4 . NFT density correlates with cognitive decline 5 , supporting a central role of Tau in AD.
Evidence suggests that caspase-cleaved Tau influences Tau pathogenesis and is involved in cognitive deficits.
TauΔD421 is generated by Caspase-3 (Casp3), Casp6 (Casp6), and Caspase-7 (refs. 7,23,33 ), whereas TauΔD402 is generated by Casp6 (ref. 8 ). Active Casp3 and Caspase-7 are sparse in AD brains 9,34 . However, active Casp6 and TauΔD402 are present in NFT, neuritic plaques, and neuropil threads in sporadic and familial AD, mild cognitively impaired individuals, and some non-cognitively impaired aged brains 8,10,35 , but absent in brains without AD pathologies 10,11 . In non-cognitively impaired individuals, active Casp6 is observed only in the entorhinal cortex and the CA1 region, the first areas to present with NFT according to Braak staging 4 , and its levels inversely correlate with cognitive and episodic memory scores 10,11 . Transgenic expression or injection of active human Casp6 (hCasp6) in mouse CA1 neurons is sufficient to induce age-dependent cognitive impairment, synaptic transmission deficits, neuroinflammation, and neurodegeneration, but does not generate NFT 12,36,37 . Active Casp6 causes axonal degeneration in neuron cultures [38][39][40] and increases amyloid beta production 41,42 . Consistent with its role in axonal degeneration, Casp6 cleaves numerous cytoskeleton or cytoskeleton-associated proteins including Tau, α-tubulin, and post-synaptic density proteins 8,43 . Furthermore, Casp6 impairs proteasomal degradation of misfolded proteins by cleaving p97 (ref. 44 ). Together, these data suggest that the active Casp6 in AD brains could be a major contributor to cognitive decline.
To assess whether Casp6-cleaved Tau is toxic or produces Tau abnormalities, we generated new transgenic mice that conditionally express full-length hTau alone (CTO) or with hCasp6 (CTC) in hippocampal CA1 and cortical neurons. Caspase-cleaved Tau was observed by 3 months of age, but significant Tau hyperphosphorylation, altered Tau binding to MT, or NFT pathology were not detected even in 25-month-old brains. Rare argentophilic deposits were observed in old mice. Surprisingly, CTC and CTO mice displayed normal memory and dendritic spine density, and did not show signs of neuroinflammation, indicating the absence of neurodegeneration. Together, these data suggest that Casp6-cleaved Tau is not toxic to CA1 and cortical neurons and is insufficient to generate obvious Tau pathology.

Results
Co-expression of hTau with hCasp6 generates cleaved Tau Human Casp6 is specifically observed in the CTC KI/Cre hippocampus and cortex Transgenic animals with knocked-in (KI) hTau or hTau-hCasp6 cDNA were crossed with mice expressing Cre recombinase (Cre) under the calcium/calmodulin-dependent protein kinase II alpha (CaMKIIα) promoter (Supplementary Fig. 1a). Mice carrying the KI and Cre transgenes (KI/Cre) were compared to wild-type (WT/ WT), KI transgene in the absence of Cre (KI/WT), or Cre only (WT/Cre) littermates. Compared to WT/WT, WT/ Cre, or CTC KI/WT control animals, CTC KI/Cre mice showed increased Casp6 protein levels in the hippocampus and cortex but not cerebellum ( Fig. 1a and b) or peripheral tissues ( Supplementary Fig. 1b), indicating region-specific Casp6 overexpression in CTC KI/Cre mice. The overexpressed Casp6 was confirmed to be human with the LS-B477 antibody (Fig. 1c). Multiplex PCR analyses showed specific STOP-cassette excision in the hippocampus and cortex of CTC KI/Cre mice, but not in other tissues or genotypes ( Supplementary Fig. 1c), confirming specific Casp6 overexpression under the CaMKIIα-Cre action.
Casp6 protein (Fig. 1b) and mRNA ( Supplementary Fig.  1d) levels were the same in CTC KI/Cre hippocampus and cortex, indicating similar expression of hCasp6 cDNA in these two regions. Hippocampal Casp6 protein levels were the same in CTC and ACL KI/Cre mice ( Supplementary  Fig. 1e), indicating that ACL KI/Cre were appropriate controls for hCasp6 expression alone.
Casp6 protein overexpression was not observed in CTO brains ( Fig. 1a and b) and peripheral tissues (Supplementary Fig. 1b). Similarly, hCasp6 mRNA was not detected in the CTO KI/WT and KI/Cre hippocampus and cortex ( Supplementary Fig. 1d), confirming the absence of hCasp6 in CTO mice.
Hippocampal hCasp6 was detected in CTC KI/Cre throughout aging (Fig. 1c). Both hCasp6 and total Casp6 protein levels increased with aging ( Supplementary Fig.  1f). Hippocampal hCasp6 levels were~2.3 fold higher at 25 months compared to 3 months of age (p = 0.0351). Brain sections from 3 to 25-month-old mice showed hCasp6 immunopositivity in the cell soma and neurites of CTC KI/Cre CA1 neurons at all ages ( Fig. 1d and Supplementary Fig. 2a) and in the cortex at 9, 18, and 25 months (Fig. 1e and Supplementary Fig. 2b).
Anti-his-tag immunopositivity in CTC KI/Cre sections was more abundant than hCasp6 immunostaining (Supplementary Fig. 2a and c). This may be related to different avidity and affinity of the antibodies or could suggest that hCasp6 underwent auto-processing; the anti-hCasp6 epitope being located in the hCasp6 linker domain. Tubulin cleaved by Casp6 (TubΔCasp6) immunopositivity confirmed Casp6 activity in CTC KI/Cre hippocampus as early as 3 months of age (Fig. 1f), and throughout aging ( Supplementary Fig. 2d). TubΔCasp6 was detected in CA1 pyramidal neurons and in punctate structures in the stratum radiatum, stratum oriens and stratum lacunosum moleculare ( Fig. 1f and Supplementary Fig. 2d), similar to the pattern observed with anti-hCasp6 ( Supplementary Fig. 2a). TubΔCasp6 was not detected in CTO KI/Cre and control genotypes ( Fig. 1f and Supplementary Fig. 2d), confirming that active hCasp6 is specifically expressed in the CTC KI/Cre hippocampus and cortex.
Human Tau is expressed in the CTO and CTC KI/Cre cortex and hippocampus Human Tau was detected in hippocampal and cortical lysates from CTO and CTC KI/Cre, but not from control genotypes ( Fig. 1a and b) or peripheral tissues , cerebellum (CB), or cortex (Cx) with anti-Casp6, anti-hTau, anti-hCasp6, and anti-β-Actin antibodies. In a, proteins from Casp6 KO hippocampus or Tau KO brainstem (KO) were used as negative controls for Casp6 and Tau analyses, respectively. Cx proteins from 3xTgAD were used as positive control for hTau. In b, the histograms represent densitometric analyses of the western blot shown. Data are expressed as ratios of Casp6 or hTau over β-Actin, relative to CTC KI/Cre hippocampus (n = 3/genotype/structure). One-way ANOVA followed by Bonferroni's post-hoc test was performed. ***p < 0.001. d-f Representative micrographs from n = 5/genotype/structure/age/antibody for 3 M hippocampus (d) and 9 M cortex (e) with anti-hCasp6, and 3 M hippocampus with anti-TubΔCasp6 (f).
( Supplementary Fig. 3a). The STOP cassette was deleted in CTO KI/Cre hippocampus and cortex, but not in other tissues or genotypes ( Supplementary Fig. 3b), confirming Cre-mediated hTau expression in KI/Cre brains.
Human Tau protein levels were equivalent in the hippocampus of 3-25-month-old KI/Cre mice ( Fig. 1c and Supplementary Fig. 3c). However, hTau levels were higher in CTO KI/Cre compared to CTC KI/Cre at all ages tested (Fig. 1a-c and Supplementary Fig. 3c). Furthermore, hTau protein levels were similar in the hippocampus and cortex of either CTO or CTC KI/Cre mice (Fig. 1b). Human Tau mRNA levels were increased by 3.5 fold in CTO KI/Cre cortex and hippocampus compared to CTC KI/Cre tissues at 2 months of age ( Supplementary  Fig. 3d), suggesting that higher hTau protein levels in CTO KI/Cre were likely due to higher amount of mRNA.
Together, these data confirm conditional genotype-and tissue-specific hCasp6 and hTau expression in the CTO and CTC models.

Cleaved Tau is detected in hCasp6-hTau expressing brains
Casp6 cleaves Tau at D13, D402, and D421. TauΔD13 was assessed with the Tau-C6g antibody whose epitope is absent in murine Tau. Our results indicated that this antibody was not specific under native conditions. Tau-C6g immunopositivity was observed in brain sections from 9-month-old WT/WT, CTC and CTO KI/Cre ( Supplementary Fig. 4a). Adsorption on recombinant hTau (rTau) digested by recombinant active hCasp6 (rCasp6) did not completely eliminate Tau-C6g immunopositivity in human AD brain ( Supplementary Fig. 4b).
In contrast, under denaturant western blot conditions, Tau-C6g detected TauΔD13 in rTau incubated with rCasp6 ( Fig. 2a and Supplementary Fig. 4c), but not in cortical and hippocampal proteins from 2-and 9-monthold CTO and CTC KI/Cre (Fig. 2a). However, Casp6 cleaved the N-terminal 9-18 amino acids of hTau because this epitope disappeared after a 4 h incubation with Casp6 in human heat-resistant proteins (Fig. 2b). Since Tau-C6g recognized Casp6-cleaved rTau (Fig. 2a), posttranslational modifications might have masked the epitope, therefore not allowing us to conclude whether TauΔD13 was present in CTC brains.
Brain sections from 3 to 25-month-old KI/Cre were negative for anti-TauΔD402 despite abundant immunopositivity in human AD brains ( Supplementary Fig. 4d). However, TauΔD402 was immunoprecipitated in CTC KI/Cre brain lysates, but not in Tau Knock-Out (KO) or WT/WT brains (Fig. 2c). Unexpectedly, TauΔD402 was also immunoprecipitated from CTO mice brains, although at much lower levels than in CTC brains, suggesting that Tau overexpression alone promoted TauΔD402. Hippocampal sections were stained with anti-ΔCasp6 to determine whether Tau overexpression in CTO mouse brains activated Casp6 ( Supplementary Fig.  4e). ΔCasp6 was detected in human AD and CTC KI/Cre, but not CTO sections, suggesting that active Casp6 in CTO KI/Cre was absent or below the detection limit, or that TauΔD402 was generated by another protease.
To confirm hCasp6 cleavage of Tau at D421, human or Casp6 KO mouse heat-resistant proteins were incubated with or without rCasp6 (Fig. 2d). TauΔD421 was detected only in samples treated with rCasp6. Casp3, however, was undetected in heat-resistant proteins, despite being easily detected in unheated lysate, heat-treated pellet, and normal and tumor colon tissues, as expected ( Supplementary  Fig. 4f). Furthermore, TauΔD421 was generated when rTau was incubated with rCasp6 ( Supplementary Fig. 4g), confirming direct Tau cleavage at D421 by hCasp6. TauΔD421 was observed mainly in the corpus callosum, and to a lesser extent in the hippocampus and cortex of CTC KI/Cre brain sections, but not in other genotypes ( Fig. 2e and Supplementary Fig. 4h), thus indicating specific Tau cleavage at D421 in CTC KI/Cre brains. At all ages tested, TauΔD421 immunopositivity appeared as punctate structures in neuritic cytoplasm of the hippocampus, corpus callosum and cortex (Fig. 2e). In addition, at 9 months TauΔD421 was occasionally observed in CA1 pyramidal neuron soma.
Despite active Casp3 immunopositivity in positive controls, active Casp3 was not detected in the CTC KI/ Cre hippocampus, corpus callosum and cortex (Fig. 2f). Together, these results suggest that Casp6 generates TauΔD402/D421 in the CTC KI/Cre brains.
CTC KI/Cre mice are cognitively normal CTC KI/Cre mice do not display locomotor impairments To assess whether the presence of cleaved Tau was associated with hippocampal and cortical dysfunction, 3-25-month-old mice were submitted to cognitive tests. Animals first underwent an openfield session to ensure normal locomotor function ( Supplementary Fig. 5). At each age, the distance travelled ( Supplementary Fig. 5a), total number of entries in each openfield zone (Supplementary Fig. 5b), and percentage of time moving (Supplementary Fig. 5c) of transgenic animals were similar to those of WT/WT mice, indicating normal overall activity and motor function. CTC KI/Cre mice explored the same percentage of cells ( Supplementary Fig. 5d) and spent the same amount of time in the periphery ( Supplementary  Fig. 5e) than control mice of the same age. Therefore, CTC KI/Cre mice did not display locomotor defects or anxiety at any age.
CTC KI/Cre mice show normal episodic and spatial memory CTC and CTO KI/Cre mice performed normally in the novel object recognition test, significantly spending more time touching the novel object than the familiar object at 3, 9, 18, and 25 months of age ( Fig. 3a and Supplementary  Fig. 6a). The discrimination index was equivalent among the six genotypes groups of the same age ( Fig. 3a). At 3 months, the percentage of impaired CTC KI/Cre and KI/WT, and CTO KI/Cre mice seemed higher than WT/ WT (Fig. 3b); however, there was no significant difference between these genotypes (Supplementary Table 1). There was also no interaction between age and genotype on discrimination index and percentage of impaired mice. Together, these results indicate normal episodic memory in KI/Cre animals.
Spatial memory was evaluated with the Barnes maze test. The number of primary errors and the primary latency to find the escape hole during learning acquisition significantly decreased over time and to the same extent for the six genotypes at all ages ( Supplementary Fig. 6b), indicating effective learning of the task. During the probe test, all genotypes performed equally well. No significant differences were observed in the number of primary errors and the latency to reach the target between KI/Cre mice and control littermates (Fig. 3c), and all genotypes made more pokes to the target compared to the other holes at all ages tested (Fig. 3d), confirming memory recall. Further analysis of individual performance during the probe test showed that CTC KI/Cre performed better than WT/WT mice at 3 months of age (p = 0.0421) but not after ( Fig. 3e and Supplementary Table 2). Consistent with previous observation of poor spatial memory in old WT/Cre mice 12 , more WT/Cre animals were impaired than mice with other genotypes at 25 months of age.
Together, these data indicate that neither CTC nor CTO KI/Cre mice are impaired in episodic and spatial memory.
Dendritic spine density is maintained in the old CTC KI/Cre hippocampus Dendritic spines loss is a feature of Tau transgenic mice. Therefore, 25-month-old brains were evaluated by Golgi-Cox impregnation (Fig. 4a, b). CA1 dendritic spine density was the same in CTC and CTO KI/Cre, and their control genotypes, which is consistent with the absence of cognitive deficits in KI/Cre mice.

Neuroinflammation is absent in old CTC KI/Cre mice
Since previous hCasp6 or TauΔD421 mouse models showed age-related neuroinflammation, 25-month-old brains were immunostained for glial fibrillary acidic protein (GFAP) and ionized calcium binding adaptor molecule 1 (Iba1) (Fig. 5). CA1 astrocytic GFAP + staining area was similar in the six genotypes ( Fig. 5a, b), indicating the absence of astrogliosis. Iba1 + microglia density in KI/Cre was equivalent to that of controls ( Fig. 5c and d). Furthermore, the percentage of types I-IV Iba1 + microglia did not differ between genotypes (Fig. 5e, f), indicating the same microglial activation profile. Together, these results indicate that the CTC KI/Cre hippocampus is devoid of neuroinflammation.
Old CTC KI/Cre mice display slight intraneuronal accumulation of phosphorylated-Tau, and rare hippocampal argentophilic deposits Intracellular Tau accumulation was analyzed in hippocampal sections from 25-month-old mice with antibodies commonly used to detect Tau pathology ( Fig. 6 and Supplementary Fig. 7). AT8 anti-pTauS202/T205, RZ3 anti-pTauT231, anti-pTauS422, and MC1 anti-misfolded-Tau immunostainings did not show any immunopositivity in the mouse sections, despite being detected in AD brains (Fig. 6a). In contrast, CP13 anti-pTauS202 and PHF1 anti-pTauS396/S404 immunopositive CA1 pyramidal cell layer (PCL) soma were detected in the brain of all mice genotypes. However, these were not as abundant as CP13 + and PHF1 + CA1 neurons from 14-month-old 3xTgAD mice (Fig. 6b). The number of CP13 + CA1 neurons was increased 2 fold in CTC KI/Cre compared to WT/WT (Fig. 6c). CTC KI/Cre PHF1 + CA1 PCL neurons were also increased compared to those in CTO KI/Cre. These differences were not detected by western blot ( Supplementary Fig. 7b), suggesting the dilution of the epitope in the extracted proteins. Together, these results (see figure on previous page) Fig. 2 Cleaved Tau is observed in KI/Cre brains. a Western blot analyses of 20 μg total proteins from 2-and 9-month-old CTO KI/Cre, CTC KI/Cre, and WT/WT hippocampal (Hc) and cortical (Cx) proteins with anti-TauΔD13. b Western blot analyses with anti-TauΔD13, anti-hTau (N-term Tau), and anti-total Tau antibodies of heat-resistant proteins (HR) from human cerebellum lysate incubated with recombinant human active Casp6p20p10 (rCasp6) (+) or vehicle (−) for 1 h or 4 h. c TauΔD402 fragments were immunoprecipitated from 19-month-old Tau KO, WT/WT, CTO, and CTC KI/Cre brain lysates with the anti-TauΔD402 antibody, and analyzed by western blotting with anti-total Tau antibody. Recombinant hTau digested with rCasp6 (TauΔCasp6) was used as positive control. d Western blot analyses with anti-total Tau and anti-TauΔD421 antibodies of HR from human cerebellum or Casp6 KO mouse brainstem lysate incubated with rCasp6 (+) or vehicle (−) for 1 h or 4 h. e, f Representative micrographs of e CTC KI/ Cre aged from 3 to 25 months (3-month-old: n = 5, 9-month-old: n = 5, 18-month-old: n = 4, 25-month-old: n = 3), and 25-month-old CTO KI/Cre (n = 4) and WT/WT (n = 3), or f 18-month-old CTC KI/Cre (n = 4) hippocampus (Hc), corpus callosum (CC), and cortex (Cx) tissue sections immunostained with e anti-TauΔD421 or f anti-ΔCasp3 antibody. Sections from AOM-DSS-treated mouse colon and human fetal stomach were used as controls.
indicate that the extent of Tau phosphorylation load in the CTC KI/Cre brains was modest.
Bielschowsky staining was performed to assess the presence of neurodegeneration ( Fig. 6d and Supplementary Fig. 7c). Argentophilic fibrillar structures, resembling the stained NFT in AD brains ( Supplementary Fig. 7c), were observed in the CA1 PCL of all 25-month-old CTC KI/Cre (n = 3) and in one out of three CTO KI/Cre, but not in WT/WT. Only 3-5 argentophilic inclusions were observed per CA1 in CTC and CTO KI/Cre brains. At 18 months, only one CTC KI/Cre displayed silver-stained CA1 inclusions and no staining was observed in 18month-old or younger CTO KI/Cre and WT/WT (Fig.  6d). The levels of soluble ( Supplementary Fig. 8a) and aggregated Tau (Supplementary Fig. 8b) were similar in 25-month-old CTC and CTO KI/Cre, and control hippocampi, as expected due to the rarity of silver-stained structures. Overall, these data suggest that hTau expression induces modest age-dependent neurodegeneration, which is enhanced by Casp6 activity.
Tau binding to MT is maintained in old CTC KI/Cre hippocampus Hippocampal MT-free and -bound Tau levels were identical in 25-month-old CTC and CTO KI/Cre, and littermates (Fig. 7a). Furthermore, hippocampal Tubulin acetylation, a hallmark of stable MT, was similar in all genotypes (Fig. 7b). Together, these data indicate that the presence of TauΔD402/D421 in the hippocampus does not alter the ability of Tau to interact with MT.
CTC mice are particularly relevant for AD research. Tau 0N4R is expressed 3-4 weeks after birth when Tau expression naturally shifts from 3 to 4 R isoforms 50 . Tau 0N4R overexpression mimics the increase of this isoform in AD 51,52 . The CTC model co-expresses hCasp6 in tandem with hTau to assure proteins co-localization in the same cell, as observed in AD neurons 8 . Furthermore, hTau and active Casp6 expression occurs in a cell-typeand regional distribution-specific manner as observed in AD brains 11 . Lastly, the long-term survival of CTC mice allows studies in aging animals. The difference between CTC and hTauΔD421 expressing mice could additionally be due to the absence of Tau N-terminus in TAU62 mice 29 , massive TauΔD421 levels 28,29,31 , or precocious Tau 0N4R expression 28,50,53 . Murine Tau unlikely affects TauΔD402/D421 in CTC mice since TAU62, TauC3, and AAV-TauΔD421 infected mice also express endogenous Tau 28,29,31 . Furthermore, the mouse background can be   excluded since CTC, AAV-TauΔD421 mice, and TAU62 models are all on the C57BL/6 genetic background.
Nevertheless, our results do not exclude that additional stressors could implicate Casp6-cleaved Tau in AD pathogenesis. Tau fragments cleaved at D25, K44, R230, N255, D314, and N368 that have been identified in AD brains remain to be investigated 6,15,16 . In addition, other post-translational modifications, especially Tau hyperphosphorylation, might be required to convert Casp6cleaved Tau into a pathological entity.
The lack of phenotype in CTC mice was surprising since the ACL mouse, expressing the same amount of Casp6, develops age-dependent neurodegeneration, neuroinflammation, and cognitive deficits that are reversed by Casp6 inhibitors 12,36,54 . Consistent with our results, Tau overexpression prevents cell death and decreases Casp6 activity in serum-deprived neurons 55 . Moreover, TauΔD421 binds MT more efficiently than full-length Tau 47 , suggesting that TauΔD421 may prevent neuronal cytoskeleton destabilisation. Together, these results suggest that Tau overexpression in CTC may have directly buffered Casp6 overactivity by limiting Casp6 accessibility to other substrates or activating protective signalling pathways against Casp6 toxicity. Alternatively, TauΔD402/D421 may be beneficial under certain conditions. However, Casp6 was not completely inhibited by Tau since TubΔCasp6 was observed in CTC mice. Comparing CTC and ACL brain proteomes may help identify substrates critical for Casp6-mediated toxicity and design new Casp6 inhibitors.
The absence of Tau pathology, neurodegeneration, neuroinflammation, and cognitive deficits in CTO mice was also unexpected since these features are commonly observed in transgenic models overexpressing nonmutated full-length hTau [56][57][58][59][60][61][62] . Random genomic insertions, expression of some Tau isoforms at the wrong developmental stage and at extra-physiological levels, could explain Tau toxicity in some mice. Our models avoid these potential problems by conditionally expressing Tau 0N4R, the predominant Tau species in adult murine axons 63 , from a knocked-in transgene in the hypoxanthine-guanine phosphoribosyl transferase (Hprt) locus. Our results are consistent with data from transgenic mice in which the entire genomic sequence of hTau has been inserted to allow normal physiological expression and alternative splicing. These mice are exempt from Tau pathology, neurodegeneration and memory deficits 64,65 .
The CTC and CTO models could be useful to assess region-and time-specific hTau expression in brains simply by using another Cre mouse. These may also be useful to study Tau propagation. Furthermore, hCasp6 cDNA in the CTC could be replaced by other genes to assess their implication in Tau function, structure and pathology.
In conclusion, this study suggests that in vivo Tau cleavage by Casp6 in CA1 and cortical neurons is insufficient to induce Tau pathogenesis and might not be an appropriate AD therapeutic target. Since Casp6 has many neuronal protein substrates, it is reasonable to conclude that Casp6-mediated damage occurs in many pathways that contribute to neurodegeneration and it may be more important to target Casp6 rather than its substrates in AD.   Generation of a mouse model conditionally expressing hTau in tandem with hCasp6 in cortical and hippocampal CA1 pyramidal neurons: the CTC model CaMKIIα-Cre-dependent hTau and hCasp6 expressing mouse, or CTC, was created (genOway, Lyon, France) to express the hTau 0N4R isoform with a self-activated form of hCasp6 under Cre/loxP recombination in the cortical and hippocampal CA1 pyramidal neurons. The transgene was composed of the ubiquitous cytomegalovirus immediate early enhancer fused to the chicken β-Actin promoter, a floxed STOP cassette, and hCasp6 cDNA followed by hTau cDNA (Supplementary Fig. 1a). An internal ribosomal entry site sequence was inserted between hCasp6 and hTau cDNA to allow the translation of both proteins from a bicistronic mRNA. The hCasp6 cDNA lacked its pro-domain to promote self-activation upon expression in mammalian cells 66 , was his-tagged and was flanked with Flippase recognition target sites to allow flippase (Flp)-dependent conditional deletion. The transgene was inserted into the Quick knock-In vector from genOway, and then introduced into the Hprt locus of 129Ola (E14) embryonic stem (ES) cells. E14 ES cells display a deletion of 35 kb upstream of the Hprt gene intron 2, which renders these cells unable to grow in culture medium containing hypoxanthine, aminopterin and thymidine. Since the targeting vector contained the wild-type Hprt sequence, targeted insertion of the vector repaired the Hprt gene deletion. Consequently, correctly recombined E14 ES cells were selected using medium containing hypoxanthine, aminopterin, and thymidine, and correct homologous recombination was validated by PCR and southern blot. Targeted insertion into the Hprt locus was chosen to allow insertion of a single transgene copy, to avoid potential defects due to random genomic insertions and to protect the transgene from gene silencing. After transfection and selection, E14 ES cells were injected into C57BL/6 J blastocysts. The transgenic animals were then crossed with CaMKIIα-regulated Cre T29.1 heterozygous mice (Jackson Laboratories, Bar Harbour, ME, USA) to generate mice expressing hTau and hCasp6 exclusively in the CA1 pyramidal neurons of the hippocampus and scattered forebrain neurons 2-3 weeks after birth, according to Cre expression pattern of the T29.1 model 67 . Mice were produced by in vitro fertilization to avoid eventual germ line transmission of the excised-STOP-cassette transgene related to Cre expression in the testis of T29.1 mice 36,68 . Since the hCasp6 and hTau cDNA were inserted into the Hprt locus of the X chromosome, only males were used to avoid lionization effects. Human Casp6-hTau expressing progeny, carrying both the CaMKIIα-Cre and the hCasp6-hTau transgenes, were called CTC KI/Cre mice. Control littermates were identified as: CTC KI/WT: animals carrying the hCasp6-hTau cDNA without the CaMKIIα-Cre transgene, WT/Cre: heterozygous CaMKIIα-Cre mice devoid of the hCasp6 and hTau cDNA, and WT/WT: wild-type animals.

Materials and methods
Generation of a mouse model conditionally expressing hTau only in the cortex and CA1 region of the hippocampus: the CTO model CaMKIIα-Cre-dependent hTau only expressing mouse, or CTO, was created to express hTau 0N4R under Cre/ loxP recombination in the cortex and hippocampal CA1 pyramidal neurons. Mice carrying the hCasp6-hTau cDNA knocked-in in the Hprt locus (described above) were crossed with C57BL/6 J Flp-deleter mice to specifically remove the hCasp6 cDNA sequence. The offsprings were screened by PCR and southern blot for the presence of the Flp transgene and analysis of the Hprt locus. Animals devoid of Flp and carrying the Flp-excised recombined Hprt locus were then crossed with CaMKIIα-Cre T29.1 heterozygous mice. Male progeny was identified as: CTO KI/Cre: hTau expressing mice carrying both the CaMKIIα-Cre and the hTau transgenes, CTO KI/WT: animals carrying the hTau cDNA without the CaMKIIα-Cre transgene, WT/Cre: heterozygous CaMKIIα-Cre mice, and WT/WT: wild-type animals.
Animals were produced by in vitro fertilizations over a period of one month, and randomly assigned for behavioural testing at 3, 9, 18, and 25 months of age. Mice were then sacrificed and randomly assigned to either histological or biochemical analysis, independent of behavioural performance. Two-month-old animals were also sacrificed for molecular biology and biochemistry analyses. Sample sizes were based on previous studies showing cognitive deficits in the ACL mice expressing equivalent levels of hCasp6 only.

Other mouse models
The ACL mouse model expressing hCasp6 under CaMKIIα-Cre-dependent expression was previously generated to express a self-activated form of hCasp6 under Cre/loxP recombination in the cortex and hippocampal CA1 pyramidal neurons 12,36 . Briefly, a transgene composed of the ubiquitous cytomegalovirus immediate early enhancer fused to the chicken β-Actin promoter, hCasp6 cDNA and an upstream loxP-flanked STOP sequence was inserted into the Hprt locus with the Quick Knock-In technology from genOway. The resulting transgenic mice were then crossed with T29.1 CaMKIIα-Cre animals. ACL mice were on a C57BL/6 J background. ACL KI/Cre mice showed age-dependent episodic and spatial memory impairments, neurodegeneration, and neuroinflammation.
Casp6 KO #006236 and Tau KO #004779 mice on a C57BL/6 J background were obtained from Jackson Laboratories.
Tissues from the 3xTgAD mice were received from Dr Emmanuel Planel (U. Laval, QC, Canada) who obtained the mice from Dr Frank LaFerla (University of California, Irvine, USA) and are on a C57BL/6;129 × 1/SvJ;129S1/Sv background. These transgenic mice harbour knocked-in human presenilin-1 with the M146V mutation, and the human Amyloid Precursor Protein (APP695 isoform) cDNA with the Swedish double mutation (KM670/ 671NL) and the hTau 0N4R with the P301L mutation cDNA, both under the control of the Thy1.2 promoter.

Genotyping of experimental mice
Tail biopsies were incubated in 300 μL of 50 mM NaOH for 20 min at 99°C, then 25 μL of 1 mM Tris pH: 8.0 were added to the samples. Tail debris were pelleted by 10 min centrifugation at 15900 × g and the supernatant was used for PCR with GoTaq ® DNA polymerase according to the manufacturer's protocol (New England Biolabs, Whitby, ON, Canada). The primers 5'-ATCCGAAAAGAAAA CGTTGA-3' and 5'-ATCCAGGTTACGGATATAGT-3' were used to amplify a 650 bp sequence from the Cam-KIIα-Cre transgene. To screen samples for the presence of the hCasp6-hTau or hTau cDNA, the primers were: forward for wild-type Hprt: 5'-CCTCCAGAAGCTCC AATGTGGATGC-3', forward for recombined Hprt: 5'-CGGTGGGACATTTGAGTTGCTTGC-3', and reverse: 5'-ACGTCAGTAGTCATAGGAACTGCGGT CG-3'. This PCR generated a 1 219 bp product corresponding to the wild-type Hprt allele, and a 340 bp product corresponding to the recombined Hprt alleles. To discriminate between the CTC and CTO models, the primers were: forward: 5'-CATGGTAAGTAAGCTTGGGCTGCAGG-3', reverse for recombined Hprt from CTC: 5'-C ATCTGCGTGGCTAACAGTTGACACC-3', and reverse for recombined Hprt from CTO: 5'-GCCAAAAGAC GGCAATATGGTGG-3'. This PCR generated a 488 bp amplicon corresponding to the recombined Hprt from CTC and a 228 bp amplicon corresponding to the recombined Hprt from CTO.

Behavioural tests
Two days after transfer from the IRIC to the LDI, the animals were handled 5 min daily for 12 days to reduce fear and anxiety due to the presence of the experimenter. Groups . They were subjected to the openfield test on day 1, and to the novel object recognition test, 24 h later. Barnes maze testing was performed on day 5 to day 10. The animals were sacrificed on day 14. The experimenter was blinded to the genotype. All animals were re-genotyped after death.

Openfield
Mice locomotor activity was video recorded during a 5 min exploration session in an empty box (#60111, Stoelting Co, Wood Dale, IL, USA). The HVS 2100 automated video tracking system (HSV Image, Hampton, UK) divided the field in 16 virtual zones (4 rows × 4 columns) and analyzed the distance travelled, the percentage of time moving, the percentage time spent in the periphery, the percentage of virtual zones used and the number of entries into each zone. One 3-month-old CTO KI/Cre mouse was excluded because it moved less than 20% of the time.

Novel object recognition
Mice were placed into the openfield box and were preexposed during 5 min to two identical objects (62020, Stoelting Co) located in the middle of the northwest and southeast quadrants of the box, equidistantly from the box corners and from each other. Two hours later, the animals were presented with the familiar object and a novel object for 5 min. The position of the novel object was interchanged between the animals to avoid any bias related to a preference in the location of the new object and the use of potential spatial cues. The number of times touching each object was manually recorded and the mice performances were expressed as the discrimination index, which represents the number of time touching the novel object minus the touches to the familiar object divided by the total number of touches. A mouse was considered as impaired when the percentage of touches to the novel object was ≤55%, which represents a discrimination index ≤0.1. The comparison of the number of impaired mice between two genotypes of the same age was done with the two-tailed Fisher's exact test. One 3-month-old CTO KI/ Cre mouse that did not touch either object was excluded.

Barnes maze
Spatial memory of the ACL mouse model was initially characterized with the Morris water maze 12 . However, Barnes maze testing was used in this study to avoid artificial increase in Tau phosphorylation levels related to water immersion-induced hypothermia. This protocol has previously been successfully used to assess spatial memory in rodents, including in Casp6 overexpressing mice 36,37 . The platform consisted in a 91 cm diameter circular dry-land maze containing 20 holes of 5 cm in diameter and elevated to a height of 90 cm from the ground. Spatial cues consisting in geometrical black pictures were placed on the white walls surrounding the maze. Mice were trained to escape from an elevated circular platform to a small static dark chamber located under one of the 20 holes around the platform perimeter (target hole). For each trial, the mice were placed in the middle of the platform under a dark box, and 10 s after the box was removed and the mice were exposed to bright light provided by a 150-watt incandescent light bulb, and a buzzer was turned on. Once in the escape chamber, the buzzer was turned off and the target hole covered. Barnes maze was conducted in three phases: adaptation (day 0), spatial acquisition (days 1-4) and test (day 5). The locations of the spatial cues and escape chamber were kept constant throughout the study. On day zero, the mice were free to explore the maze for 60 s and to stay in the escape chamber for 120 s. On days 1-4, the mice were subjected to 4 daily trials of 180 s each with an inter-trial interval of 15 min. If the mouse did not successfully escape the maze within 180 s, it was gently led into the escape chamber and left inside for 60 s. The latency to reach the target hole (primary latency), the number of errors to the first encounter of the escape hole (primary errors), the number of total errors and the latency (time spent before the mouse enters the escape hole) were recorded. On day 5, the mice were submitted to a single probe test. The target hole was blocked and the mice were allowed to explore the maze for 90 s under the same aversive conditions. The primary latency, the number of primary errors and the number of nose pokes for each hole, reflecting the hole preference, were recorded. Mice were tracked with the HVS 2100 automated video software throughout the testing. A mouse was considered as impaired if the percentage of touches to the target hole during the probe test was less than the touches to any other hole. The comparison of the number of impaired mice between two genotypes of the same age was done with the two-tailed Fisher's exact test. Two 25month-old CTC KI/Cre, one 3-month-old WT/WT, and one 25-month-old WT/WT were excluded because they froze during the training sessions and/or the probe test.
Aggregated Tau was isolated as described 72 . Total RIPA proteins were adjusted to 1% N-lauroylsarcosine (sarkosyl), incubated for 30 min at room temperature with constant orbital shaking, and centrifuged at 100 000 g for 60 min at 20°C. The supernatants were discarded and the sarkosyl-insoluble pellets re-suspended in sample buffer (NuPAGE LDS (Invitrogen) containing 5% of 2β-mercaptoethanol, 1 mM sodium orthovanadate, 1 mM NaF, 1 mM PMSF, 10 μL/mL P8340 Sigma protease inhibitors cocktail), boiled for 5 min and kept at −20°C until western blot analyzes. Protein extractions were done blinded to genotype.

Endogenous microtubule-binding assay
The association of endogenous Tau to endogenous MT was analyzed according to the endogenous MT binding assay 73 . Mice were killed by cervical dislocation without anaesthesia and the brains dissected out. Fresh cortices and hippocampi were immediately homogenized together in 37°C extraction buffer (80 mM MES pH: 6.8, 1 mM MgCl 2 , 2 mM ethylenediamine-tetraacetic acid (EGTA), 30% glycerol, 0.1% Triton X-100, 1 mM PMSF, 1 mM sodium orthovanadate, 1 mM NaF, 10 μL/mL P8340 Sigma Protease Inhibitor Cocktail) with a pre-warmed (37°C) mechanical homogenizer. The samples were immediately centrifuged at 3000 × g for 2 min at 25°C, and an aliquot of the supernatant, identified as the total fraction, was centrifuged at 100000 × g for 20 min at 25°C. The supernatant corresponding to the MT-free protein fraction was removed, and the remaining pellet corresponding to the MT-bound protein fraction was resuspended in extraction buffer. Total, MT-free and MTbound protein fractions were stored at −80°C until processing.

In vitro digestion by recombinant active Casp6
Human cerebellum (obtained with written informed consent from the University of Toronto Institutional Review Board by Dr C. Bergeron) 8 , or murine Casp6 KO brainstem RIPA proteins were boiled for 5 min at 95°C and then centrifuged for 20 min at 18440 × g at 4°C. This method is used to isolate proteins resistant to heat, including Tau. Thus, other endogenous proteins, like caspases, were precipitated during the boiling process and eliminated from the supernatant. The pellets were resuspended in sample buffer, boiled for 5 min and stored at −20°C. The heat-resistant proteins in the supernatant were precipitated with acetone (1:4 ratio, −20°C overnight) 74 , re-suspended in Stennicke's buffer (SB) (20 mM piperazine-N, N-bis (2-ethanesulfonic acid) (PIPES) pH: 7.2, 30 mM NaCl, 1 mM EDTA, 0.1% 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS), 10% sucrose) and submitted to in vitro digestion by recombinant hCasp6 (generated by Dr Agne Tubeleviciute-Aydn in our lab) for 1 or 4 h at 37°C. The reaction mix consisted of 330 ng active site titrated recombinant Casp6p20p10 prepared from the pET23b cDNA construct 43 , 100 μg brain heat-resistant proteins extract or 189 ng rTau 0N4R (Biolegend, San Diego, CA, USA), SB and freshly added 10 μM dithiothreitol. The digestion was replicated four times for rTau and three times for heat-resistant extracts.

Western blot
Samples were prepared in sample buffer. Proteins were separated on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels, blotted onto nitrocellulose or polyvinyl fluoride membranes, and blocked 1 h at room temperature with 5% non-fat dry milk in PBS containing 0.1% Tween 20 (PBS-T). Membranes were then washed three times 10 min with PBS-T, and probed overnight at 4°C with primary antibodies diluted in SuperBlock ® blocking buffer (ThermoFisher Scientific) ( Table 1). After three washes of 10 min with PBS-T, the membranes were incubated for 1 h at room temperature with the corresponding horseradish peroxidase (HRP)linked secondary antibody diluted in PBS-T containing 5% non-fat dry milk. For monoclonal anti-Tau antibodies, an HRP-coupled antibody directed against the light chain of murine immunoglobulins was used as secondary antibody (#AP200P, Millipore). Membranes were revealed by enhanced chemiluminescence (GE Healthcare Life Sciences, Mississauga, ON, Canada) in a Fujifilm LAS4000 imaging system (Fujifilm USA, Valhalla, NY, USA). Densitometric analyses were performed with Image Gauge analysis software 3.0. (Fujifilm USA).
Astrocytic GFAP and microglial Iba1 immunopositivity were quantitated on three brain sections per slide and on three slides spaced 56 μm apart. The area of GFAP staining was quantified with Image J software (NIH, Bethesda, MD, USA) and expressed as μm 2 staining/μm 2 of tissue. The number of total Iba1 + cells/mm 3 tissue was quantified with Image J software, and Iba1 + cells were submitted to morphological characterization according to scoring schemes 76 . Quantifications were blinded to genotype. Bielschowsky silver staining Bielschowsky silver staining was performed with the ab245877 Bielschowsky Silver Stain Kit from Abcam (Cambridge, UK) according to the manufacturer's protocol. Briefly, formalin-post-fixed, paraffin-embedded 4-μm thick brain sections were deparaffinized, hydrated, incubated in a warm 20% silver nitrate solution for 15 min at 40°C, and rinsed three times 3 min in water. The slides were then incubated for 30 min in a warm (40°C) ammonia hydroxide-cleared silver nitrate solution before incubation for 1-2 min in freshly-made developer solution (8 drops of 20% formalin, 8 drops of citric acid, 4 drops of nitric acid, in 50 mL H 2 O). The slides were then immediately placed in 0.2% ammonium hydroxide for 2 min. The sections were rinsed three times 2 min in water and incubated for 2 min in 5% thiosulfate solution. The slides were then dehydrated, coverslipped with Permount mounting medium (Thermo Fisher Scientific) and digitally scanned with the MIRAX SCAN scanner (Zeiss). Brain sections from human AD were used as positive control (Dr D.A. Bennett). Staining and analyses were blinded to genotype.

Golgi-Cox staining
Golgi-cox staining was performed with the FD Rapid GolgiStain™ kit from FD NeuroTechnologies (Columbia, MD, USA) according to the manufacturer's instructions. Briefly, animals were sacrificed by cervical dislocation without anaesthesia, the brains rapidly dissected out and immerged in impregnation solution for 2 weeks at room temperature in the dark. Tissues were then transferred into solution C and incubated for 3 days in the dark. Onehundred micrometre thick sections were cut with a frozen cryostat, mounted on gelatine-coated slides (#PO101, FD NeuroTechnologies) and air-dried for 2 days in the dark. The slides were then washed two times 4 min in distilled water, incubated for 10 min in the staining solution, rinsed two times 4 min with water, dehydrated and coverslipped with Permount mounting medium (Thermo Fisher Scientific). Brightfield microscopy images of CA1 pyramidal dendrites were acquired using the MIRAX SCAN scanner (Zeiss) with an ×100 objective. For each mouse, a minimum of 15 neurons and 1-3 dendrites per neuron of 25-35 μm length was used for analysis. The analyzed dendritic areas belonged to neurons with visible soma, were located at least 25 μm from the soma and did not overlap with other cells. Results were expressed as total number of spines per 10 μm of uninterrupted section. Tissue processing and quantifications were blinded to genotype.

Statistical analyses
Data are expressed as means ± SEM. Data were tested for equivalence of variance with Bartlett's test.
Comparison between two groups was done using the unpaired Student's two-tailed t-test. Statistical analyzes of more than two groups were performed with one-way analysis of variance (one-way ANOVA) followed by the Bonferroni's post-hoc test. For repeated measures, the two-way ANOVA with genotype as one factor and time for the other, followed by the Bonferroni's post-hoc test was used. Statistical calculations were performed using GraphPad Prism 5.0 software (GraphPad, La Jolla, CA, USA). Results were considered significantly different at p < 0.05.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.