In the present study, we demonstrated that overexpressed human tau was secreted by Hela cells through an unconventional secretory pathway. The pool of secreted tau was cleaved at the C-terminal and was less phosphorylated than intracellular tau. Both hyperphosphorylation and cleavage at D421 enhanced tau secretion by Hela cells.
Our results demonstrating that tau was secreted by an unconventional secretory pathway is consistent with recent studies reporting that tau was found in exosomes in culture medium from MC1 cells overexpressing human tau and that secreted tau by COS-7 and HEK-293 cells was present in microvesicles 
. It remains to be determined whether tau utilizes other non-conventional pathways since a portion of tau in the culture medium of MC1 cells was not associated with exosomes. In the present study, tau secreted by Hela cells could be immunoprecipitated from the culture medium without using any detergent indicating that it was not included in microvesicles/exosomes. Consistent with this, no decrease of tau in the medium was observed after the culture medium was deprived of microvesicles by ultracentrifugation. All together the above results indicate that tau is most likely secreted by more than one pathways as shown for other proteins involved in neurodegenerative diseases such as SOD1 associated with Amyotrophic lateral sclerosis and the prion protein 
. In our previous study, we showed that hyperphosphorylated tau was preferentially associated with the rough endoplasmic reticulum (RER) membranes in AD brain and in the tau transgenic mice JNPL3 
. An increase of hyperphosphorylated tau was also noted at the Golgi apparatus in the JNPL3 mice 
. RER and Golgi have been showed to be involved in non-conventional secretory pathways 
. For example, COPII vesicles budding from the ER and containing tau at their surface could directly fuse with the plasma membrane for secretion 
. This pathway is also used by the signal-peptide-containing protein, cystic fibrosis transmembrane conductance regulator (CFTR) 
. Another possibility is that tau secretion could occur through non-COPII-coated vesicles forming at the ER or vesicles forming at the Golgi having tau attached at their surface 
. We reported that Tau was found at the surface of RER membranes but this does not exclude the possibility that it could end up on the extracellular surface of the plasma membrane during the fusion process occurring between tau-containing vesicles and the plasma membrane.
Our results demonstrated that cleavage of tau at D421 increased its secretion. The fact that wild-type tau and tauΔ413–441 were secreted in a similar way by Hela cells strongly suggests that the major pool of secreted wild-type tau could be cleaved close to S412 in Hela cells. To further demonstrate this, Hela cells overexpressing wild-type tau were treated with a caspase-3 inhibitor since tau is preferentially cleaved at D421 by this caspase 
. When Hela cells were treated with a caspase-3 inhibitor, a small but significant decrease of wild-type tau secretion was observed. This could indicate that as mentioned above, the major pool of wild-type tau secreted by Hela cells was not cleaved at D421. From our results, it was not possible to conclude whether tauΔ413–441 and tauΔ422–441 underwent further cleavage during the process of secretion. The fact that they migrated in a similar way in the cell lysate and culture medium could signify that if they were cleaved it was only by a few amino acids. The secretion of tau cleaved mutants indicates that wild-type tau was most likely cleaved before its trafficking in the secretory pathway. Based on a recent study reporting that an increase of caspase activity is an early event in AD and our results showing the enhanced secretion of tau cleaved at the caspase-3 site, one can speculate that the secretion of tau would be enhanced at the initial stage of the disease 
Mimicking of hyperphosphorylation significantly enhanced the secretion of tau by Hela cells. However, tau found in the culture medium was dephosphorylated compared to the pool of tau that remained intracellular. This is consistent with a recent study reporting that released tau in the culture medium upon cell lysis was dephosphorylated compared to intracellular tau 
. In this study, extracellular tau was not phosphorylated at the epitopes recognized by the phospho-tau antibodies, AT8 and PHF-1 whereas intracellular tau was. Tissue non-specific alkaline phosphatases (TNAP) present in the plasma membrane were shown to be responsible for the dephosphorylation of extracellular tau 
. Interestingly, TNAP were shown to be increased in AD brain 
. All together, the above data indicate that CSF-tau might be less phosphorylated than intracellular tau. Several studies have examined the phosphorylation of tau found in the CSF of patients affected by a tauopathy. T181 and T231 are the sites that have been extensively used as a diagnostic tool for AD 
. Although in most studies, the phosphorylation of T231 was shown to be increased in AD, some studies reported that its phosphorylation was reduced at later stages of the disease 
. However, the phosphorylation of other sites such as the epitopes of AT8 and PHF-1 remains controversial 
. Our data revealed that these epitopes are preferentially dephosphorylated in secreted tau. The AT8 epitope seems to play a central role in the hyperphosphorylation cascade of tau. Indeed, an increase in phosphorylation of the AT8 epitope is detected at an early stage of AD 
. In our previous study, we reported that the phosphorylation of the AT8 epitope had the most significant effects on the phosphorylation of other sites in primary hippocampal neurons 
. The results of the present study highlight the possibility that the dephosphorylation of this epitope could be regulated in a distinct manner.
In a previous study, it was shown that dephosphorylated tau in the culture medium could act as an agonist of muscarinic M1 and M3 receptors inducing a robust and sustained increase of intracellular calcium that triggered cell death in SH-SY5Y cells 
. Most importantly, the increase in intracellular calcium induced by dephosphorylated tau in the culture medium was associated with an increase of TNAP expression 
. Based on these observations and our present data, one could speculate that tau found in the extracellular space in AD brain would be dephosphorylated and thereby would contribute to the aberrant homeostasis of calcium noted in this tauopathy.
From our data and that of other groups, it appears that both extracellular and intracellular tau could contribute to the process of neurodegeneration linked to AD. Furthermore, our data indicate that in AD, hyperphosphorylation of tau would induce a vicious circle that would result in the amplification of its secretion (). Indeed, our data revealed that hyperphosphorylation of tau would enhance its secretion and this would in turn increase the amount of dephosphorylated tau in the extracellular space. Dephosphorylated extracellular tau would then induce an increase of intracellular calcium, an event linked to the increase of tau hyperphosphorylation 
. This increased hyperphosphorylation of tau would further enhance its secretion leading to the emergence of a vicious circle that would promote the propagation of tau pathology in the brain and its accumulation in the CSF. The accumulation of total and phospho-tau in the CSF is used as a diagnostic biomarker for tauopathies 
. Our data highlight the possibility that the distinct phosphorylation and cleavage pattern of tau could account for its differential accumulation in the CSF among the tauopathies. The characterization of this pattern for each tauopathy could become a powerful tool for their early detection and to distinguish them from one another.
A schematic representation of the vicious cycle leading to the amplification of tau secretion in AD.