key: cord-0335582-zg9prmu3
authors: Tuck, Benjamin J.; Miller, Lauren V. C.; Wilson, Emma L.; Katsinelos, Taxiarchis; Cheng, Shi; Vaysburd, Marina; Knox, Claire; Tredgett, Lucy; Metzakopian, Emmanouil; James, Leo C.; McEwan, William A.
title: Tau assemblies enter the cytosol in a cholesterol sensitive process essential to seeded aggregation
date: 2021-09-15
journal: bioRxiv
DOI: 10.1101/2021.06.21.449238
sha: 05f87d8b842db0947aed808915f2bc054669b18d
doc_id: 335582
cord_uid: zg9prmu3

Accumulating evidence supports a prion-like mechanism in the spread of assembled tau in neurodegenerative diseases. Prion-like spread is proposed to require the transit of tau assemblies to the interior of neurons in order to seed aggregation of native, cytosolic tau. This process is poorly understood and remains largely hypothetical. Here, we develop sensitive techniques to quantify the cytosolic entry of tau in real-time. We find that tau does not promote its own entry but, rather, is wholly dependent on cellular machinery. We find that entry to the widely used reporter cell line HEK293 requires clathrin whereas entry to neurons does not. Cholesterol depletion or knockdown of cholesterol transport protein Niemann-Pick type C1 in neurons renders cells highly vulnerable to cytosolic entry and seeded aggregation. Our findings establish entry as the rate-limiting step in seeded aggregation and demonstrate that dysregulated cholesterol, a feature of several neurodegenerative diseases, potentiates tau aggregation. Graphical Abstract

Tauopathies are a group of neurodegenerative diseases characterised by the conversion of microtubule-associated protein tau into highly ordered fibrils 1 . Most prominent among these diseases is Alzheimer's disease, which accounts for the majority of the ~50 million dementia cases worldwide. Other tauopathies include fronto-temporal dementia, progressive supranuclear palsy and chronic traumatic encephalopathy. A causal role for tau in neurodegeneration is indicated by more than 50 nonsynonymous and intronic point mutations which lead to dominantly inherited, early onset forms of neurodegenerative disease characterized by the accumulation of tau assemblies in the brain 2 . Two non-mutually exclusive mechanisms are proposed to explain the occurrence of tau fibrils in the diseased brain. First, that nucleation of aggregation occurs within individual cells in a cell autonomous manner.

Alternatively, that tau assemblies transit between cells, promoting the aggregation of tau in recipient cells in a 'prion-like' manner. This latter model of aggregation could help explain the observed spatio-temporal spread of tau misfolding in the human brain and is consistent with observations of seeded aggregation in cultured cells [3] [4] [5] [6] , and in in vivo disease models [7] [8] [9] [10] .

Nonetheless, the relative contributions of these two mechanisms to human disease progression remains unknown 11 . Spread between cells may occur through the uptake of naked tau assemblies, or via extracellular membrane-bound vesicles. Naked tau assemblies are taken up into membrane bound vesicles following interactions between tau and cell-surface heparan sulphate proteoglycans (HSPGs), as well as the recently identified low-density lipoprotein receptor, LRP1 12, 13 . Tau is then taken up to membrane-bound compartments via endocytosis and macropinocytosis 12, [14] [15] [16] . For a prion-like mechanism to occur, tau assemblies must subsequently gain access to the cytosol, somehow breeching these intracellular membranes. While the process of tau filament uptake has been comparatively well studied, the transfer of tau from intracellular membrane-bound vesicles to the cytosol is largely unexplored and remains a critical missing step for assessing the relevance of seeded aggregation to disease progression 11, 17 .

Cholesterol is a critical determinant of membrane bilayer structural integrity, and a known risk factor in neurodegenerative disease [18] [19] [20] [21] . Cholesterol is depleted from the brain in an agedependent manner, resulting in impaired intracellular signaling and synaptic plasticity [22] [23] [24] .

Variations in APOE, which encodes a cholesterol transporting protein, is the strongest genetic risk factor for AD with inheritance of the e4 variant being associated with a substantially increased risk of AD 25, 26 . Though a link between APOE and Abeta pathology is well established, experimental evidence also suggests that APOE exacerbates tau pathology independently of Abeta [27] [28] [29] [30] . Supporting a role of cholesterol biology in tau pathology, APOE was also identified as a risk factor in primary age-related tauopathies, where Abeta is not implicated 31 . In the primary tauopathies, the ε2 allele, which is protective against AD, is associated with higher risk. Niemann-Pick type C (NPC) is a rare autosomal recessive disorder characterised by the aberrant accumulation of intracellular cholesterol and glycolipids. Approximately 95% of NPC cases are caused by loss-of-function mutations in the Niemann-Pick C1 gene (NPC1), which encodes a ubiquitous trafficking protein important for transport of cholesterol to organelles and the plasma membrane 32, 33 . NPC is characterized by progressive childhood neurological disease often with abundant tau pathology. Thus, cholesterol abundance and localization are strongly associated with tau pathology and may play a direct role in its progression.

In this study we develop highly sensitive methods for the detection of tau entry to the cytosol, permitting analysis of entry at physiological concentrations of tau. We find that entry of tau to the cytosol represents the rate-limiting step in seeded aggregation. Accordingly, changing entry levels through pharmacological or genetic means results in a concomitant change in seeded aggregation. We find that the entry pathway of tau assemblies into human embryonic kidney cells (HEK293), a widely used reporter line, differs substantially to entry in mouse or human neurons. We observe that tau does not mediate its own entry to the cytosol of neurons but, rather, enters through a clathrin-and dynamin independent pathway. Tau entry to neurons relies on cell surface receptor LPR1, heparin sulphate proteoglycans and membrane cholesterol. Depletion of cholesterol from the plasma membrane, or its mislocalisation by NPC1 depletion, enhances tau entry in multiple cell models, including organotypic slice culture and human neurons. The results establish tau entry as a distinct event from uptake that is essential for seeded aggregation. They further demonstrate that tau entry is modulated by cellular determinants rather than by tau itself and is therefore potentially susceptible to pharmacological inhibition.

The study of tau entry to cells has been complicated by the difficulty in reliably distinguishing cytosolic populations from vesicular populations 17 . In order to specifically detect the cytosolic fraction of exogenously supplied tau assemblies, we established a live-cell assay relying on split luciferase, NanoLuc (Nluc) (Fig. 1a) . NanoLuc is a 19 kDa luminescent protein engineered from the luciferase of a deep-sea shrimp Oplophorus gracilirostris. The enzyme has been split into an 18 kDa subunit (LgBiT) of Nluc and an 11 amino acid peptide (HiBiT) which interact with sub nanomolar affinity. Reconstitution results in the complementation of activity and luminescence in the presence of substrate 34 . We expressed recombinant P301S tau (0N4R isoform) in fusion with a HiBiT tag at the C-terminus in E. coli (Fig. 1b,c) .

Assemblies of tau-HiBiT were produced by incubation with heparin and the aggregation kinetics were quantified via thioflavin T fluorescence, a dye which fluoresces when bound to β-sheet amyloid structures 35 . We observed similar aggregation kinetics compared to a tagless variant of tau, suggesting that the HiBiT tag does not interfere with aggregation (Fig. 1d) .

Negative stain electron microscopy revealed filamentous structures with smaller assemblies present after sonication (Fig. 1e) . To assess the ability of tau-HiBiT assemblies to reconstitute Nluc, we titrated tau-HiBiT assemblies into a fixed concentration of recombinant LgBiT, resulting in a concentration-dependent increase in luminescent signal upon the addition of substrate (Fig. 1f) . We found the HiBiT tag was as readily accessible in assemblies as it was in monomer, as signal was unchanged between these states (Fig. S1a) . These results confirm that tau-HiBiT can form filaments similar to tagless variants and these assemblies yield a dosedependent luminescent signal when complexed with LgBiT.

In order to template aggregation, tau assemblies are proposed to cross cell limiting membranes. To measure the entry of tau assemblies to the cell, we expressed LgBiT in the cytosol of human embryonic kidney HEK293 cells, a cell line widely used in the study of seeded tau aggregation 3,6,36,37 . We expressed LgBiT by lentiviral transduction from the vector pSMPP 38 which drives expression from a viral (spleen focus forming virus; SFFV) promoter.

We applied sonicated tau-HiBiT assemblies to the cell exterior. This resulted in high luminescence signal that was sensitive to the addition of trypsin, a protease which completely abrogates the luciferase signal of cell free tau-HiBiT:LgBiT complexes (Fig. S1b,c) . We interpret this as an indication that the bulk of the interaction was extracellular, likely due to leakage or secretion of LgBiT to the media.

We therefore examined if low cytosolic LgBiT concentrations were necessary to establish an assay that reports solely on intracellular tau entry. We reduced cytosolic concentrations of LgBiT first by replacing the SFFV promoter of pSMPP with a weak mammalian housekeeping promoter 39 (PGK) to generate the vector pPMPP. The cytosolic concentration was further reduced by expressing the protein in fusion with a nuclear localisation signal (nls) and eGFP.

The resulting construct (nls-eGFP-LgBiT, abbreviated as NGL) was expressed at ~85:15 nuclear:cytoplasmic ratio and yielded a luciferase signal that was trypsin-insensitive following treatment of cells with tau-HiBiT and addition of a cell-penetrant luciferase substrate (Fig. 1gi) . We next titrated tau-HiBiT onto HEK293 cells expressing the construct (HEK-NGL) and observed dose and time-dependent uptake that saturated at around 6 h (Fig. 1j,k) . These results demonstrate that the assay reports on the real-time entry of tau to live cells with a broad and linear dynamic range. 

Tau seeding can be dramatically increased by the use of transfection reagents 12 , likely owing to enhanced delivery of tau assemblies to the cell interior. To test whether transfection reagents result in increased tau entry to the cytosol, we titrated tau-HiBiT assemblies in the presence and absence of Lipofectamine 2000 (LF) onto HEK-NGL cells. We observed a 50to-100-fold increase in entry of tau to the cytosol when supplemented with LF (Fig. 2a) . In HEK293 cells expressing P301S tau-venus 6 , we observed an attendant increase in seeded aggregation ( Fig. 2b-d) . These data demonstrate that LF promotes entry to the cell and implicates breeching of intracellular membranes as rate-limiting to seeded aggregation. We next tested whether entry could be inhibited by reducing uptake to the cell. Previous studies have shown that tau uptake occurs via endocytosis and relies on the vesicle coat protein clathrin, as well as the GTPase dynamin 14, [40] [41] [42] . We pre-treated cells with the clathrin inhibitor PitStop 2 or the dynamin inhibitor Dyngo 4a. This reduced the uptake of fluorescently labelled transferrin to the cell (Fig. S1c ) and resulted in a stark reduction in tau entry to the cytosol ( Fig. 2e) . We also tested the routinely used Na + /H + import channel inhibitor, dimethyl-amiloride (DMA) 43, 44 , which inhibits macropinosome establishment, but observed no change in entry to the HEK-NGL cells (Fig. 2e) . Incubation of cells at 4ºC, a temperature non-permissive to endocytosis, also prevented tau entry to the cell (Fig. 2f) . Heparan sulphate proteoglycans (HSPGs) mediate the uptake of tau to intracellular compartments and promote seeded aggregation 12, 45 . Pre-treatment with surfen hydrate, an agent that binds heparan sulphate, strongly diminished entry in our assay (Fig. 2g ) as did addition of exogenous heparin to the cell media (Fig. S1d) . These data together confirm that HSPGs promote entry of tau assemblies to the cytosol of HEK293 cells, and that entry relies upon clathrin and dynamin. 

After observing the dependency of tau entry on clathrin coat protein and dynamin GTPase, we next investigated the intracellular compartment from which tau assemblies escape.

Clathrin-and dynamin-dependent endocytosis results in the enclosure of cargo in a primary endocytic vesicle, which undergoes homotypic fusion to form an early sorting endosome that subsequently traffics throughout the endosomal network with the intraluminal pH simultaneously decreasing 46 . Rab GTPases are crucial effectors of this endosome maturation.

To investigate the nature of the compartment from which tau escapes in HEK293 cells, we used siRNA against early endosomal RAB5A or late endosomal RAB7A and performed tau entry assays over the course of 4 h (Fig. 3a,b) . Interestingly, we observed an increase in entry when RAB7A, but not RAB5A, was depleted, suggesting a sensitivity of entry to impaired late endosomal compartments.

Recently, it was shown that compromised function of VPS13D, a ubiquitin binding protein 47 , promoted the seeded aggregation of tau 48 . It was proposed that the potential mechanism of augmented tau aggregation may be a result of increased endolysosomal escape of tau seeds.

To determine whether VPS13D is involved in maintaining tau assemblies from accessing the cytosol, we depleted VSP13D in HEK-NGL cells and observed a marked increase in tau entry ( Fig. 3c,d) . This corroborates the finding that VPS13D is essential to maintaining vesicular integrity and confirms that its depletion promotes entry. We next investigated the role of VPS35, a core constituent of the endosome-to-golgi retromer complex, an essential component of proper endosome sorting. Deficiencies in VPS35 are linked to increased tau burden and late-onset AD 49,50 as well as being a known genetic risk in Parkinson's disease 51 .

We treated cells with VPS35 siRNA and, similar to VPS13D depletion, we observed a significant increase in tau entry in knockdown cells but not in control treated cells over 4 h ( Fig. 3e,f) . Given that VPS35 interacts with RAB7 53 , and that genetic silencing of VPS35 increases tau accumulation 49 , our results suggest a mechanism by which mutations associated with the gene may alter the progression of tau pathology. To further support our observations in the entry assay, we depleted VPS35 in HEK293 cells expressing P301S tauvenus and performed seeding assays in the absence of transfection reagents. We observed a significant increase in fluorescent puncta corresponding to tau aggregation in VPS35depleted cells relative to control cells (Fig. 3g,h) . These data demonstrate an essential role for endosome sorting and repair machinery in preventing tau escape to the cytosol. 

To investigate the mechanism of tau entry in a more physiologically relevant system, we adapted our assay to primary neurons derived from wild-type (WT) C57BL/6 mice. We used chimeric particles of adeno-associated virus produced with capsids of types 1 and 2 (AAV1/2) to deliver a self-cleaving variant of the LgBiT reporter from a neuron-specific synapsin (hSyn) promoter relying on the T2A peptide 54 . This construct, hSyn::eGFP-P2A-LgBiT-NLS (GPLN), provides eGFP as a fluorescent marker and low levels of nuclear-localised LgBiT (Fig. 4a-d) .

An optimised protocol was established with AAV challenge at 2 days in vitro (DIV) for tau entry assays at 7 DIV, where we see close to 100% of neurons expressing GPLN (Fig. 4d, Fig. 

We titrated tau assemblies onto GPLN-neurons and observed a concentration-dependent luciferase signal that was resistant to trypsin, confirming that luminescence originates from an intracellular tau-HiBiT:LgBiT interaction (Fig. 4e, Fig. S2d ). Investigating the entry mechanisms of tau in neurons, we were surprised to find that neither clathrin nor dynamin inhibition with PitStop 2 or Dyngo 4a reduced the entry of tau to the cytosol (Fig. 4f,g) .

Similarly, treatment with DMA to block Na + /H + import channels was unable to prevent entry ( Fig. 4h) . We also investigated the requirement for low endosomal pH in entry. We treated neurons acutely with the V-ATPase inhibitor bafilomycin-A at commonly used concentrations to inhibit acidification but avoid downstream effects on autophagy 55 . We found no requirement for V-ATPase mediated endosomal acidification in tau entry (Fig. 4i) . In all cases, neuronal viability was monitored to ensure that experiments reflect entry to healthy cells (Fig. S2e) . We confirmed that seeded aggregation is similarly insensitive to these inhibitors by challenging DIV 7 neurons with 100 nM tau-HiBiT assemblies in the presence or absence of PitStop 2 or Dyngo 4a for 7 days. We observed formation of endogenous mouse tau (mTau) aggregates in the presence of inhibitors which showed no significant deviation from the solvent treated control neurons (Fig. 4j,k) . Given that tau can spread in a transsynaptic manner 56-58 , we considered whether the observed independence was a result of performing entry assays in cells with improperly formed synapses, which occurs around 7-10 DIV 59,60 . We therefore performed entry assays in DIV 14 neurons but consistently found no role for clathrin or dynamin in entry of tau to the cytosol of neurons (Fig. S2f) . To further interrogate the entry pathway of tau to neurons, we knocked down mouse Rab5a or Rab7a via siRNA and monitored tau-HiBiT entry over 4 h (Fig. 4l) . Despite a knockdown efficiency of >90% for both proteins and minimal toxicity, we found no role for either protein in tau entry to neurons (Fig.   4m, Fig. S3g,h) . Our results suggest that tau entry to the cytosol of neurons occurs via an alternative mechanism, independent of clathrin-mediated endocytosis. The results in neurons are in direct contrast to observations in HEK293 cells, suggesting profound differences in the underlying biology between these systems.

One possibility is that tau assemblies mediate their own escape from vesicles following uptake by destabilising endosomal membranes. To test this, we titrated tagless tau assemblies in the presence of a constant (50 nM) concentration of tau-HiBiT. If the tagless tau assemblies promoted membrane rupturing, we predicted this would result in an increase in tau-HiBiT entry. However, in both HEK293 cells and primary neurons, we found no such increase in signal, suggesting that tau does not mediate its own entry (Fig. 4n, Fig. S3a) . We then used a second membrane integrity assay wherein a plasmid encoding luciferase is supplied to the extracellular media of HEK293 cells. Addition of membrane rupturing agents results in plasmid transfer to the intracellular environment, promoting luciferase expression. The addition of LF or human adenovirus type 5, which enters cells via endosomal rupture, resulted in 100-to 1,000-fold increases in the levels of luminescence. In contrast, tau assemblies and tau monomer at high concentration (1 µM) resulted in no observable signal above background ( Fig. 4o) . We next investigated the receptor-dependency of tau entry. We first depleted the low density lipoprotein receptor, Lrp1, recently identified as a receptor for tau uptake 13 . Knockdown of this protein substantially decreased tau entry to cells at 4 h after addition (Fig. 5a,b) . However, entry at 1 h was less affected by Lrp1 depletion, potentially reflecting residual membrane Lrp1 remaining available for uptake followed by receptor internalisation upon tau binding. Addition of heparin to the cell media reduced tau entry in a dose-dependent manner, consistent with a role for cell-surface HSPGs in promoting tau attachment and uptake to neurons (Fig. 5c) . As cholesterol biology is heavily implicated in susceptibility to tauopathies, we next investigated the dependency of cholesterol in neuronal tau entry. We hypothesised that changes in membrane cholesterol may modulate the entry of tau to the cytosol. First, we treated neurons with the cholesterol extracting agent methyl-betacyclodextrin 61 (MbCD). Doses in excess of 3 mM of this compound were found to be cytotoxic to neurons (Fig. S4a) . However, treatment with 2 mM MbCD for 2 h before addition of 50 nM tau-HiBiT for 1 h was not toxic but effectively extracted cholesterol from neurons, as visualised by staining with the cholesterol-binding dye, filipin (Fig. 5f, Fig. S4b ). These conditions resulted in a significant increase in tau entry to neurons (Fig. 5d) . We verified that the effect was due to cholesterol extraction and not a result of off-target effects of cyclodextrins, as treatment with gamma-cyclodextrin (γCD), which does not extract cholesterol from membranes, resulted in no significant change in entry compared to the control. observed a significant reduction in entry (Fig. 5e,f) . Mutations in the Niemann-Pick C1 protein (NPC1) are causative of NPC 65 , a primary tauopathy with mis-sorted cholesterol. We questioned whether depletion of Npc1 in primary neurons could modulate tau pathology, potentially informing the mechanism of action of NPC1 mutation in human disease. Npc1 depletion by siRNA resulted in a significant increase in tau entry (Fig. 5g,h) . These results suggest that intracellular cholesterol transport and membrane cholesterol content are critical to the entry pathway of tau to the cytosol, and may therefore potentially modulate seeded aggregation of tau. 

Our findings have revealed the relationship between cellular co-factors and tau entry to the cytosol in immortalised cells and primary culture models. We next sought to investigate whether changes in tau entry leads to changes in seeded aggregation in a physiological setting. For these experiments, we used organotypic hippocampal slice cultures (OHSCs) which maintain authentic neural architecture and cell-type diversity 66 . We prepared OHSCs from P301S tau transgenic mice, which develop intracellular tau inclusions by 6 months of age 67 . OHSCs prepared from these mice do not develop detectable tau pathology unless challenged with tau assemblies 68 . Tau assemblies were provided to OHSCs following a 1-day pre-treatment period with endocytosis inhibitors or cholesterol modulators. Tau assemblies and drug were removed after 3 days and the OHSCs were incubated for a further 3 weeks before detection and quantification of seeded aggregation using the phospho-tau specific antibody AT8 (Fig 6a-f) . Consistent with our observations for tau entry to the cytosol, inhibition of clathrin or dynamin with PitStop 2 or Dyngo 4a resulted in no significant change in seeded aggregation (Fig. 6a) . Supporting our proposed role for cholesterol in the maintenance of the unseeded state, extraction of cholesterol with MbCD substantially increased seeded aggregation whereas the γCD control yielded no statistically significant change (Fig. 6b) .

Moreover, treatment with 25-HC resulted in a significant reduction in seeded aggregation, again consistent with its effect on tau entry (Fig. 6c) .

We have previously shown that seeded aggregation in mouse neural tissue conforms to nonlinear dynamics, with low levels of extracellular tau assemblies failing to produce detectable seeded aggregation 68 . We titrated tau assemblies in the presence or absence of 200 µM MbCD in OHSCs and observed a significant increase in tau entry when MbCD was present ( Fig. 6d, Fig. S5a) . Remarkably, seeded aggregation was observed at 10 and 30 nM supplied tau when MbCD was present. These concentrations completely failed to induce detectable seeded aggregation in slices that had not been treated with MbCD. We also found that aggregation-promoting effect of MbCD is titratable, as 1 mM MbCD further increased seeded aggregation relative to lower concentrations (Fig. 6e) . These data taken together demonstrate an essential role of cellular cholesterol levels in maintaining membrane integrity in mouse neurons and neural tissue. Moreover, they suggest that dysregulated cholesterol permits seeded aggregation of tau when neurons are exposed to levels of extracellular tau that would otherwise remain safely in the extracellular environment.

We next tested whether aggregation of endogenous mouse tau was sensitive to cholesterol levels, by performing seeding assays in primary non-transgenic C57BL/6 mouse neurons.

Cholesterol was extracted with a low, tolerated, concentration of MbCD, which potentiated seeding of endogenous mouse tau (Fig. 6g, h) . These findings demonstrate that seeding of both transgenic human tau and endogenous tau are sensitive to cholesterol levels in mouse primary neurons. 

We next asked whether the tau entry pathway to mouse neurons is conserved with humans.

We generated human cortical neurons derived from pluripotent stem cells (iNeurons) using a robust protocol 69 . Cytosolic LgBiT was expressed by transduction with AAV1/2-hSyn-GLPN.

We verified LgBiT expression by western blot and flow cytometry, where we observed >95% transduction efficiency (Fig. 7a, S6a) . We confirmed that day 14 iNeurons expressed essential neuronal and synaptic genes indicative of fully differentiated human cortical neurons (Fig. 7b,   S6b ). We first investigated the entry dynamics of tau assemblies in day 14 LgBiT expressing iNeurons (GPLN-iNeurons), where entry occurred in a concentration-and time-dependent manner, similar to the other cellular systems we have studied (Fig. 7c, d) . We next questioned whether the human neurons would recapitulate the clathrin and dynamin dependence observed in human immortalised cell lines, or conversely display the endocytosis independence observed in mouse neurons. We pre-treated GPLN-iNeurons with clathrin inhibitor PitStop 2 or dynamin inhibitor Dyngo 4a and performed a 1 h entry assay with 50 nM tau-HiBiT assemblies. We found no role for clathrin or dynamin in tau entry in human neurons, consistent with observations in mouse neurons and OHSCs (Fig. 7e) . Similar to both mouse neurons and human immortalised cell lines, we found a strong dependence on HSPGs, as pre-treatment with 20 µg/ml heparin blocked entry by an average of ~67% (Fig. 7f) . We then investigated the cholesterol sensitivity of tau entry in human neurons, and we found that cholesterol extraction by pre-treating for 2 h with MbCD results in a >2-fold increase in entry compared to untreated cells. Again, we observed no significant change in entry with the inactive cyclodextrin, γCD (Fig. 7g) . Finally, we pre-treated GPLN-iNeurons with 25hydroxycholesterol and we saw a significant reduction in entry of an average of ~47% (Fig.   7h) . These data further support our findings from mouse model systems and corroborate the role of cholesterol in tau entry and seeded tau aggregation. 

In this study we have developed sensitive assays capable of detecting the entry of tau to the cytosol following its application at nanomolar concentrations to the cell exterior. The high sensitivity of the assay means that entry biology can be studied without the use of extraphysiological concentrations of tau. Using these assays, we have described the entry of tau to the cytosol in two cell-based models: HEK293 cells, which are widely used as reporters for seeded aggregation, and neurons, the major site of tau pathology in the brain of AD patients.

Entry occurred in a dose-dependent manner over several orders of magnitude, suggesting that extracellular tau concentrations contribute linearly to entry. Levels of entry in HEK293 and neurons were a critical determinant for seeded aggregation, consistent with entry, rather than uptake, being the rate-limiting step to seeded aggregation. We thus quantitatively described a link between entry and seeded aggregation and establish a causal relationship between these processes.

Surprisingly, we observed that the mechanism of entry was highly divergent between HEK293 cells and neurons. We found that tau entry in HEK293 cells is dependent on the coat protein clathrin and the GTPase dynamin, consistent with the frequently described role of canonical endocytosis in human immortalised cell lines uptake 17 . When we induced endolysosomal dysfunction in HEK293 cells via genetic knockdown of critical endosomal genes RAB5A and RAB7A, we found only knockdown of RAB7A increased entry. Furthermore, knockdown of VPS35, a core constituent of the retromer complex and a RAB7 effector, similarly increased entry. We therefore consider it likely that tau escapes from late endosomal compartments in HEK293 cells. This dependency was lost, however, in primary mouse neurons. Here, entry of tau occurred through a clathrin-and dynamin-independent pathway with no observable effect of Rab5 or Rab7 depletion. This was not an artefact of using neurons with immature synaptic biology, as 14 DIV cells with highly connected morphology displayed a similar phenotype.

Clathrin-and dynamin-independent endocytosis has been described for particular substrates suggesting that tau entry occurs from a specific subset of endosomal compartments 70 , or potentially at the plasma membrane. Further supporting these findings, cytosolic entry of tau also occurred in a clathrin-and dynamin-independent manner in iPSC derived human neurons which form highly connected synaptic networks and express an array of mature neuronal markers. This therefore indicates that this mechanism is a result of neuron biology, rather than mechanism specific to mouse systems. Previous studies that have examined the uptake of tau to neurons have demonstrated involvement of clathrin-mediated endocytosis 14 , macropinocytosis 71 and a clathrin-independent, dynamin-dependent pathway 72 . It is therefore possible that several pathways are responsible for tau uptake but tau escape to the cytosol occurs from a subset of compartments. This raises the prospect that molecular dissection of the entry pathway may permit the identification of inhibitors that prevent seeded aggregation by blocking entry of tau to the cytosol.

Cholesterol homeostasis has been repeatedly implicated in the risk and pathogenesis of Alzheimer's disease 73 Of potential significance therefore, is the finding that cholesterol content of the plasma membrane is depleted in an age-dependent manner 23 . Given that age is the most important risk factor for dementia, the results potentially indicate a mechanism that permits tau pathology to accumulate in aged cohorts. Furthermore, identification of low cholesterol as a feature that renders neurons highly susceptible to seeded aggregation may permit future therapeutic intervention in tauopathies.

95% of NPC patients harbour a loss-of-function mutations in NPC1, and many patients present with accumulation of tau filaments at an early age. In the present study, we found that depletion of the mouse homologue, Npc1, in primary neurons resulted in enhanced tau entry.

Loss-of-function NPC1 mutations impair the lysosomal export of cholesterol and subsequent intracellular trafficking, and mutations in the sterol sensing domain of NPC1 result in impaired delivery of cholesterol to the plasma membrane 79 . Taken together with our findings that cholesterol extraction via MbCD results in enhanced entry and increased seeded aggregation, these data indicate that the cholesterol content of the neuronal plasma membrane is critical in maintaining the unseeded state. Further, the results provide a potential mechanistic underpinning to the observed accumulation of tau pathology in NPC patients.

It has been suggested that tau escape to the cytosol is dependent on membrane destabilising activities of tau itself. This interpretation is based on the use of cytosolic galectin reporters, which bind to carbohydrates on the luminal leaflet of ruptured vesicles 15, 56, 80, 81 . The addition of tau assemblies, or cell lysate containing tau assemblies, was found to increase the number of galectin-positive structures in those studies. Another study demonstrated that tau assemblies can bind to cell membranes via interactions between the repeat region and phospholipid bilayers, potentially consistent with direct lysis of membranes 82 . However, in our study, which directly measures entry to the cytosol, we found no evidence of tau-mediated entry in two separate assays and in both HEK293 and primary neurons. The reasons for these differences are not clear, though may reflect different cell types and fibril preparations, or may reflect inherent differences in directly measuring tau cytosolic entry versus galectin puncta. Our study used recombinant fibrils of a single isotype whereas fibrils prepared from the AD brain comprise all six CNS isoforms. We therefore cannot rule out that other isoforms or tau structures may bear membrane destabilising activity. However, the membrane-interacting domain of tau was mapped to the R2 repeat region 82 , which is present in the tau isoform used in the present study. Alternatively, membrane destabilisation may be a property of small oligomers, or monomers, rather than the filamentous assemblies used here 83, 84 . Our results nonetheless favour an alternative model in which cellular cofactors, rather than tau, are the critical determinants in maintaining the integrity of membrane-bound compartments. Loss of this homeostatic control results in increased entry of tau to the cytosol and a concomitant increase in seeded aggregation. This suggests that cellular/genetic determinants will likely play a pivotal role in this rate-limiting step to tau pathology. The potential dynamic range of this rate-limiting step seeding is demonstrated by lipofectamine, which increased entry and seeded aggregation by >50-fold. In support of this interpretation, Chen et al. also found no evidence of tau mediating its own escape to the cytosol and instead suggested genetic factors influence membrane integrity 48 . Their results implicated vacuolar sorting proteins as playing a role in maintaining tau out of the cytosol, which we confirm here. Genome-wide association studies and certain dominantly inherited proteopathies implicate several endosomeassociated proteins in neurodegeneration [85] [86] [87] . Early endolysosomal genes such as BIN1, PICALM and CD2AP are among the most common genetic polymorphisms associated with late-onset AD, yet there is no consensus on how they influence AD 88, 89 . Future work should therefore ascertain whether membrane integrity is governed by such genetic risk factors and whether this predisposes to the transit of tau assemblies between cells of the brain.

Our study provides the first quantitative description of tau entry to the cytosol. We establish that entry is essential for and rate-limiting to seeded aggregation of intracellular tau pools. Tau transits to the neuronal cytosol in a cholesterol sensitive manner following interactions with HSPGs and Lrp1. Entry is independent of Rab5/Rab7 GTPases and clathrin in neurons, suggesting that entry occurs from a specific subpopulation of vesicles that lack these cofactors. Dysregulated cholesterol, by extraction with MbCD or through Npc1 depletion, renders neurons vulnerable to tau entering the cytosol and permits seeded aggregation at concentrations of extracellular tau assemblies that fail to induce seeding in untreated cells.

Our study enables the investigation of tau entry as distinct from uptake and will permit investigation of the genetic and pharmacologic factors that modify this rate-limiting step in seeded tau aggregation.

Lister 

The following rabbit polyclonal antibodies were acquired from Proteintech; RAB5A (11947-1 and Alexa-fluor conjugated secondary antibodies were purchased from Thermo Fisher.

The following plasmids were purchased from addgene: pAAV-hSyn-eGFP (#50465), 

Protein purification was performed as previously described 90 ) and primers designed to detect a range of genes (Fig. S6c) . qPCR was performed on the QuantStudio5 (Thermo Scientific). All samples were analysed in technical triplication from 3 independent differentiations. qPCR data was normalised to housekeeping gene 18s

and results analysed using the ΔΔCt method.

The following Human onTARGETplus SMARTpool siRNA were purchased from Horizon 

The seeding assay was performed as previously described 6 . Briefly, 20,000 cells were seeded into poly-L-lysine (Sigma, P4707) coated black 96-well plates in 50 µl OptiMEM (Thermo Fisher, 31985062). Tau assemblies were diluted in OptiMEM and 50 µl added to cells to final concentration in the presence or absence of LF. Cells were incubated at 37ºC and 5% CO2 for 1 h before the addition of 100 µl complete medium to stop the transfection process. Cells were incubated at 37ºC in an IncuCyte S3 Live-Cell Analysis System for 48-72 h after the addition of tau assemblies.

Neuronal seeding assays were performed as previously described 94 . Briefly, neurons were prepared as described and supplemented with a final concentration of 100 nM tau-HiBiT assemblies at 7 DIV in maintenance medium and incubated at 37ºC and 5% CO2 until 14 DIV.

At 14 DIV, cells were fixed with methanol, nuclei stained with DAPI and epitopes immunofluorescently labelled with anti-MAP2 and anti-mTau (T49) antibodies. Stained neurons were subjected to high-content imaging of 15 fields per well and tau aggregation quantified. mTau puncta per field were normalised to cell count per field. Any drugs were left on throughout the duration of the seeding assay.

Slices were prepared from P301S tau transgenic pups aged 6-9 days as previously and 25% horse serum. Slices were maintained at 37ºC and 5% CO2 in a humid atmosphere.

Drugs were diluted in 1 ml culture medium and applied to the underside of the slices for 1 day before addition of 100 nM tau. 3 days later, media was removed and fresh medium added without drug or tau. 3 weeks post-challenge, OHSCs were fixed in 4% paraformaldehyde, nuclei stained with DAPI, and MAP2, phopsho-tau (AT8) and tau (DAKO) epitopes immunofluorescently labelled. Stained slice images were acquired using a Zeiss LSM780

confocal microscope with either a 20X or 63X objective lens and images stitched using ZEISS Zen software package.

HEK 293 cells were plated at 1x10 4 cells per well in 96 well plates. The next day, media was exchanged for serum-free complete medium with pNL1.1 (Promega) plasmid at 10 ng/µl.

Recombinant tau monomers or heparin-assembled tau were added to the media.

Lipofectamine 2000 and endosome-destabilising adenovirus type 5 (ViraQuest) were added as positive controls for entry. Plates were examined for NanoLuc luminescence after 24 h on the ClarioSTAR plate reader.

Transferrin uptake assay 5x10 5 HEK 293T cells were suspended in serum-free complete medium and pre-treated with PitStop 2 (20 µM), Dyngo 4a (20 µM) or DMSO (0.1%) for 30 minutes at 37ºC and 5% CO2.

50 µg/ml alexa-flour-647-conjugated human transferrin (Thermo Fisher, T23366) was added to the cell suspension and incubated on ice for 5 minutes, followed by 10 minutes at 37ºC and 5% CO2. Cells were pelleted (500 x g, 5 minutes), acid washed twice (100 mM glycine, 150 mM NaCl pH 2.5 in PBS), fixed in 4% PFA (10 mins at room temperature) and subsequently resuspended in FACS buffer (0.5% bovine serum albumin in PBS). Transferrin uptake was then quantified via flow cytometry (CytoFLEX flow cytometer, Beckman Coulter).

Cells were rinsed once with 1X PBS and incubated with accutase for 12-15 minutes and allowed to detach. Cells were then resuspended in complete medium (HEK293) or maintenance medium (neurons/iNeurons) and flow cytometry performed in 96-well format on the Beckman Coulter CytoFLEX. 1*10 4 events per condition were the minimum recorded events.

Media was aspirated and cells were rinsed once with PBS, then fixed with 4% PFA for 15 minutes at room temperature. Fixed cells were then rinsed 3 times with PBS followed by permeabilization with 0.1% triton X-100 (Sigma, X100) in PBS for 10 minutes at room temperature. Cells were then washed 3 times with PBS, and blocked with 1% BSA in PBS (IF block) for 1 h at room temperature. Primary antibody was diluted to required concentration in IF block and incubated overnight at 4ºC. Antibody was removed, cells were then rinsed 3 times with PBS and incubated with alexa-flour conjugated secondary antibody in IF block for 1 h at room temp. Secondary antibody was removed, cells rinsed with PBS 3 times and imaged via fluorescence microscopy. Filipin III staining was performed according to manufacturer instructions (Sigma, SAE0087). For methanol fixation, cells were fixed and permeabilised with ice cold methanol on ice for 3 minutes, followed by 3 half washes with PBS (1X volume added, 0.5X total volume removed) followed by complete aspiration and 3 full washes with PBS.

Permeabilised cells were then blocked and stained as described.

Formation of Tau-HiBiT fibrils was confirmed by uranyl acetate negative stain transmission electron microscopy at the Cambridge Advanced Imaging Centre as previously described 95 .

Puncta corresponding to seeded aggregation in HEK 293T and primary neurons were quantified in Fiji using the ComDet plugin 96 . Total puncta per field was normalised to cell count per field via DAPI staining or cell confluency detected by the IncuCyte S3 Live-Cell Analysis software. In neuronal seeding assays, only small puncta colocalised to neurons were quantified. AT8 staining in OHSCs was segmented into a binary threshold and fields of 150 µM x 150 µM were analysed for % AT8 positive area from slices from N=6 mice per condition.

All statistical analyses were performed via GraphPad Prism software. Differences between multiple means were tested by one-way ANOVA, followed by Tukey's post hoc test unless otherwise indicated. Differences between two means were tested by unpaired t-test with Welch's correction. All data represent mean values ± s.e.m with the following significances:

ns p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; and **** p ≤ 0.0001.

Propagation of Tau Aggregates and Neurodegeneration

Tau filaments in neurodegenerative diseases

Propagation of Tau Misfolding from the Outside to the Inside of a Cell

Distinct Tau Prion Strains Propagate in Cells and Mice and Define Different Tauopathies

Trans-cellular propagation of Tau aggregation by fibrillar species

Cytosolic Fc receptor TRIM21 inhibits seeded tau aggregation

Transmission and spreading of tauopathy in transgenic mouse brain

Brain homogenates from human tauopathies induce tau inclusions in mouse brain

Synthetic tau fibrils mediate transmission of neurofibrillary tangles in a transgenic mouse model of Alzheimer's-like tauopathy

Unique pathological tau conformers from Alzheimer's brains transmit tau pathology in nontransgenic mice

What is the evidence that tau pathology spreads through prion-like propagation?

Heparan sulfate proteoglycans mediate internalization and propagation of specific proteopathic seeds

LRP1 is a master regulator of tau uptake and spread

Extracellular Monomeric and Aggregated Tau Efficiently Enter Human Neurons through Overlapping but Distinct Pathways

Galectin-8-mediated selective autophagy protects against seeded tau aggregation

Small Misfolded Tau Species Are Internalized via Bulk Endocytosis and Anterogradely and Retrogradely Transported in Neurons *

Knockin' on heaven's door: Molecular mechanisms of neuronal tau uptake

Cholesterol Metabolism in Neurodegenerative Diseases: Molecular Mechanisms and Therapeutic Targets

Serum total cholesterol, apolipoprotein E epsilon 4 allele, and Alzheimer's disease

Cholesterol dysfunction in neurodegenerative diseases: Is Huntington's disease in the list?

Membrane lipid rafts and neurobiology: age-related changes in membrane lipids and loss of neuronal function

Age-associated cholesterol reduction triggers brain insulin resistance by facilitating ligand-independent receptor activation and pathway desensitization

Aging Triggers a Repressive Chromatin State at Bdnf Promoters in Hippocampal Neurons

Genome-wide association study identifies variants at CLU and PICALM associated with Alzheimer's disease

Genome-wide association study identifies variants at CLU and CR1 associated with Alzheimer's disease

Apolipoprotein E and Alzheimer disease: risk, mechanisms, and therapy

Apolipoprotein E in sporadic Alzheimer's disease: allelic variation and receptor interactions

ApoE4 markedly exacerbates tau-mediated neurodegeneration in a mouse model of tauopathy

Association of Apolipoprotein E ε4 With Medial Temporal Tau Independent of Amyloid-β

A Brief Overview of Tauopathy: Causes, Consequences, and Therapeutic Strategies

Niemann-Pick C1 protein regulates cholesterol transport to the trans-Golgi network and plasma membrane caveolae

Genetic heterogeneity in Niemann-Pick C disease: a study using somatic cell hybridization and linkage analysis

NanoLuc Complementation Reporter Optimized for Accurate Measurement of Protein Interactions in Cells

Rapid assembly of Alzheimer-like paired helical filaments from microtubule-associated protein tau monitored by fluorescence in solution

Conformation Determines the Seeding Potencies of Native and Recombinant Tau Aggregates

Tau prions from Alzheimer's disease and chronic traumatic encephalopathy patients propagate in cultured cells

A Method for the Acute and Rapid Degradation of Endogenous Proteins

Systematic Comparison of Constitutive Promoters and the Doxycycline-Inducible Promoter

Loss of Bin1 Promotes the Propagation of Tau Pathology

Whole genome CRISPR screens identify LRRK2-regulated endocytosis as a major mechanism for extracellular tau uptake by human neurons

Clustering of Tau fibrils impairs the synaptic composition of α3-Na+/K+-ATPase and AMPA receptors

Inhibition of macropinocytosis blocks antigen presentation of type II collagen in vitro and in vivo in HLA-DR1 transgenic mice

Na+/H+ exchange and growth factor-induced cytosolic pH changes. Role in cellular proliferation

Tau Internalization is Regulated by 6-O Sulfation on Heparan Sulfate Proteoglycans (HSPGs)

To degrade or not to degrade: mechanisms and significance of endocytic recycling

Vps13D encodes a ubiquitin binding protein that is required for the regulation of mitochondrial size and clearance

Compromised function of the ESCRT pathway promotes endolysosomal escape of tau seeds and propagation of tau aggregation

VPS35 regulates tau phosphorylation and neuropathology in tauopathy

VPS35 haploinsufficiency increases Alzheimer's disease neuropathology

VPS35, the Retromer Complex and Parkinson's Disease

VPS35, the Retromer Complex and Parkinson's Disease -PubMed

Molecular insights into Rab7-mediated endosomal recruitment of core retromer: deciphering the role of Vps26 and Vps35

Cleavage of foot-and-mouth disease virus polyprotein is mediated by residues located within a 19 amino acid sequence -PubMed

Bafilomycin A1, a specific inhibitor of vacuolar-type H(+)-ATPase, inhibits acidification and protein degradation in lysosomes of cultured cells

Synaptic Contacts Enhance Cell-to-Cell Tau Pathology Propagation

Trans-synaptic spread of tau pathology in vivo

Neuronal activity enhances tau propagation and tau pathology in vivo

Synaptogenesis of hippocampal neurons in primary cell culture

Image-Based Profiling of Synaptic Connectivity in Primary Neuronal Cell Culture

Effects of cholesterol depletion by cyclodextrin on the sphingolipid microdomains of the plasma membrane

25-Hydroxycholesterol activates the expression of cholesterol 25-hydroxylase in an LXR-dependent mechanism

Effect of 25-hydroxycholesterol in viral membrane fusion: Insights on HIV inhibition

Cholesterol 25-hydroxylase suppresses SARS-CoV-2 replication by blocking membrane fusion

Identification of 58 novel mutations in Niemann-Pick disease type C: correlation with biochemical phenotype and importance of PTC1-like domains in NPC1

Organotypic brain slice cultures to model neurodegenerative proteinopathies

Abundant tau filaments and nonapoptotic neurodegeneration in transgenic mice expressing human P301S tau protein

Tau assemblies do not behave like independently acting prion-like particles in mouse neural tissue

Inducible and Deterministic Forward Programming of Human Pluripotent Stem Cells into Neurons, Skeletal Myocytes, and Oligodendrocytes

Clathrin-independent pathways of endocytosis

Small misfolded Tau species are internalized via bulk endocytosis and anterogradely and retrogradely transported in neurons

PIKfyve activity is required for lysosomal trafficking of tau aggregates and tau seeding

Amyloid-β-independent regulators of tau pathology in Alzheimer disease

APOE ε2 is associated with increased tau pathology in primary tauopathy

Microglia drive APOE-dependent neurodegeneration in a tauopathy mouse model

A glance at the structural and functional diversity of membrane lipids

Cholesterol Depletion by MβCD Enhances Cell Membrane Tension and Its Variations-Reducing Integrity

25-Hydroxycholesterol Increases the Availability of Cholesterol in Phospholipid Membranes

The Sterol-sensing Domain of the Niemann-Pick C1 (NPC1) Protein Regulates Trafficking of Low Density Lipoprotein Cholesterol*

Endocytic vesicle rupture is a conserved mechanism of cellular invasion by amyloid proteins

Exosomal and vesicle-free tau seeds -propagation and convergence in endolysosomal permeabilization

Discovery and characterization of stable and toxic Tau/phospholipid oligomeric complexes

Tau oligomers impair artificial membrane integrity and cellular viability

Interaction of tau protein with model lipid membranes induces tau structural compaction and membrane disruption

Genome-wide association study identifies variants at CLU and PICALM associated with Alzheimer's disease

Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease

Endosome function and dysfunction in Alzheimer's disease and other neurodegenerative diseases

Endo-lysosomal dysregulations and late-onset Alzheimer's disease: impact of genetic risk factors

Alzheimer's disease risk genes and mechanisms of disease pathogenesis

Unconventional Secretion Mediates the Trans-cellular Spreading of Tau Article Unconventional Secretion Mediates the Trans-cellular Spreading of Tau

Therapeutic efficacy of regulable GDNF expression for Huntington's and Parkinson's disease by a high-induction, background-free 'GeneSwitch' vector

REAP: A two minute cell fractionation method

Culturing pyramidal neurons from the early postnatal mouse hippocampus and cortex

Unique pathological tau conformers from Alzheimer's brains transmit tau pathology in nontransgenic mice

Tau proteins of alzheimer paired helical filaments: Abnormal phosphorylation of all six brain isoforms

Fiji: an open-source platform for biological-image analysis

We gratefully thank Dr Michel Goedert for provision of P301S tau transgenic mice and Prof Steven Paul for insightful conversations. We thank Dr Filomena Gallo of the Cambridge Advanced Imaging Centre for support and assistance in this work. We thank Annabel Smith and Sophie Sanford of the Cambridge UK DRI for preparation of tau assemblies. WAM is a