Neurodegenerative diseases are typified by the accumulation of specific proteins into fibrillar assemblies. In around twenty distinct neurodegenerative diseases, including the most common, Alzheimer’s disease, the protein tau forms hyperphosphorylated, filamentous inclusions within the cytoplasm of neurons. Tau pathology can be a direct cause of neurodegeneration. For instance, human genetic studies reveal that around 50 distinct mutations in MAPT, the gene that encodes tau, cause inherited forms of dementia with evidence of tau filaments [1]. The origin of tau assemblies in the human brain remains uncertain. Cell-autonomous processes may lead to the spontaneous nucleation of oligomeric forms of tau within the cytoplasm of neurons. Some of these assemblies adopt filamentous conformations that are able to undergo extension by the addition of tau monomers to the filament ends. Over the past decade it has been postulated that, in addition to these cell-autonomous mechanisms, tau pathology may occur through a spreading or ‘prion-like’ mechanism [2]. Several lines of evidence demonstrate that assemblies of tau can be taken up into cells, whereupon they seed the conversion of native tau to the assembled state. Addition of tau assemblies to the exterior of cells, or the injection of tau assemblies into the brains of tau-transgenic and wildtype mice, can induce intracellular tau assembly in the recipient [3,4,5,6].
Population cross-sectional studies demonstrate that the appearance of tau pathology follows a predictable pattern over time and space in the human brain, potentially indicating spreading via a prion-like mechanism. Immunoreactivity to antibodies such as AT8, which detects tau that is phosphorylated at positions S202 and T205 [7], progresses in a manner that can be systematically categorised into stages according to anatomical distribution (Braak stages 0–VI) [8, 9]. In young adults, some AT8 immunoreactivity is observed in the vast majority of brains by the third decade of life. However, it is generally confined to neurons within the locus coeruleus (LC) in the brainstem (Braak pretangle stages 0 a–c and 1a,b). Subsequently, AT8 staining is observed in the entorhinal cortex (EC) and hippocampus (HC) (Braak stages I–II). Later stages are characterised by progressive dissemination and increasing density of staining in neocortical regions (Braak stages III–VI). These late stages are associated with severe disease and the overall burden of tau pathology negatively correlates with cognitive function [10].
Though intracranial challenge experiments demonstrate that seeded aggregation can in principle occur, they provide little insight as to whether physiological concentrations of extracellular tau species might support prion-like activity. The concentration of tau in wildtype mouse interstitial fluid (ISF) is around 50 ng/ml total tau (equivalent to ~ 1 nM tau monomer). Wildtype mouse ISF levels typically exceed cerebrospinal fluid (CSF) tau levels by around tenfold [11]. In humans between ages 21–50 years, CSF total tau is below 300 pg/ml increasing to 500 pg/ml over age 70 [12]—approximately 7–12 pM if considering the average mass of full length tau isoforms. Levels are increased 2–threefold in Alzheimer’s disease [13]. If a similar relationship between ISF and CSF tau concentration exists in humans as in mice, ISF tau levels are likely in the order of 100 pM, rising to 300 pM in Alzheimer’s disease. Intracranial injection experiments typically supply tau in the high micromolar range. Even if this were distributed broadly across the brain, micromolar concentrations would be exceeded and local concentration at the injection site may plausibly be 100-fold greater. Thus, intracranial injection experiments likely exceed physiological concentrations of extracellular tau by two to seven orders of magnitude.
For classical infectious agents, infectivity is related to dose by a “one-hit” relationship wherein the amount of infectivity decreases linearly upon dilution until end-point [14]. This property is also evident in PrPSc prions, though it is complicated by the presence of multiple aggregation states and the size distribution of particles [15]. The relationship between dose and prion-like activity for tau has not been established. It is therefore currently not possible to reconcile high-dose challenge experiments with the low concentrations of tau observed in the extracellular spaces of the brain. To address this, we developed a model of seeded tau aggregation in mouse organotypic hippocampal slice cultures, allowing direct control of the concentration of tau neurons were exposed to. Brain slice cultures have been used for over 40 years [16], though developments in recent years have rendered them increasingly relevant for the study of neurodegenerative diseases [17,18,19,20,21]. We prepared slices from transgenic mice expressing human tau bearing the P301S mutation [22], which is causative of familial fronto-temporal dementia and displays accelerated fibrilisation compared to wildtype tau [23].
Using our system, we show that neurons within CA1 are preferentially susceptible to seeded aggregation, displaying intracellular hyperphosphorylated tau tangles. We find that seeding activity cannot be titrated down and only occurs at high concentrations of tau assemblies. The concentrations of tau assemblies required to initiate seeding in this model exceed reported physiological ranges of ISF and CSF tau. Our results imply that a model of tau spread via seeded aggregation requires these concentrations to be locally exceeded or requires other mechanisms not captured here to facilitate seeded aggregation.
Results
Sagittal hippocampal slices of ~ 300 µm thickness were prepared from homozygous P301S tau-transgenic mice at age 7 d, and cultured at the air–liquid interface on culture support membranes (Fig. 1a). Hippocampal structures are well developed at this age, yet the tissue exhibits plasticity that aids recovery from the slicing procedure [24]. We stained OHSCs with antibodies against markers of the major cell types of the brain: neurons (Map2), microglia (Iba1) and astrocytes (Gfap) (Fig. 1b). Similar to previous studies [20], neurons were found to maintain extensive arborisation with evidence of intact neuronal tracts. Microglia were observed with normal morphology with extensive processes, similar to quiescent cells in whole brains [25]. Astrocytes were also well represented throughout the cultures. Immunostaining for P301S tau revealed widespread expression in all regions of the hippocampus (Fig. 1c). After 5 weeks in culture no overt signs of tau pathology were apparent, as visualised by staining with AT8 (Fig. 1d). These results demonstrate that hippocampal architecture and cell types from P301S tau transgenic mice are maintained through the slicing and culture process. Importantly, they demonstrate that OHSCs from P301S tau transgenic mice do not undergo detectable spontaneous aggregation over this time period.
To investigate the response of slice cultures to challenge with tau assemblies, we prepared tau from two independent sources. First, we expressed the 0N4R isoform of tau bearing the P301S mutation in E. coli. Recombinant protein was incubated with heparin and, following a lag period, was found to give a fluorescent signal in the presence of thioflavin T, a dye whose fluorescence increases upon binding to β-sheet rich amyloid structures (Fig. 2a). Negative stain transmission electron microscopy revealed the presence of abundant filamentous structures (Fig. 2b). Second, we prepared the sarkosyl-insoluble (SI) fraction from aged P301S tau transgenic mice, a procedure that enriches insoluble tau species. Brain-derived assemblies were subjected to western blot, confirming the presence of hyperphosphorylated, insoluble tau (Fig. 2c). The samples were quantified using a dot-blot method using recombinant fibrillar tau as a standard (Additional file 1: Supplementary Fig. 1). Tau assemblies were added to HEK293 cells stably expressing 0N4R P301S tau-venus, a reporter cell line for seeded aggregation [26]. In this assay, transfection reagents are used to deliver tau assemblies into cells, whereupon tau-venus is observed to form puncta over 1–2 d. This aggregation was previously found to result in the accumulation of tau-venus in the sarkosyl-insoluble pellet [26]. In the present study, abundant venus-positive puncta were detected following challenge with recombinant fibrils or mouse brain derived tau (Fig. 2d). To investigate whether these seeded assemblies bore markers of tau hyperphosphorylation, we stained with the monoclonal antibody AT8 and AT100, which recognises tau phosphorylated at pT212 and pS214. These epitopes occur on tau filaments extracted from post-mortem tauopathy brains, including those from Alzheimer’s disease patients. We observed that challenge with recombinant tau assemblies resulted in colocalization between tau-venus puncta and AT8 or AT100 (Fig. 2e, f). We therefore concluded that our tau preparations contained species able to induce bona fide seeded aggregation in recipient cells.
To test whether OHSCs support seeded aggregation, we treated slices at DIV7 with recombinant tau assemblies or mouse brain derived sarkosyl insoluble (SI) tau. After three days a complete media change was performed to remove the tau, and the OHSCs were incubated for a further three weeks (Fig. 3a). Following treatment with tau assemblies, we observed pronounced AT8 staining, suggesting the presence of intracellular hyperphosphorylated tau structures (Fig. 3b, c). The addition of monomeric tau did not induce these same structures, indicating that the misfolded state of tau was responsible for seeded aggregation (Fig. 3b, d). Furthermore, addition of tau assemblies to OHSCs prepared from wildtype mice did not induce seeded aggregation suggesting that the transgenic P301S tau construct is responsible for the phenotype (Fig. 3b, c). The hyperphosphorylation of tau was accompanied by the accumulation of sarkosyl-insoluble species following seeding in P301S transgenic OHSCs but not in wildtype OHSCs (Fig. 3e, f) (Additional file 1: Supplementary Fig. 2). Whilst the human transgenic tau is therefore required for seeding to occur, we cannot exclude the possibility that endogenous mouse tau also contributes to aggregates in the transgenic slices since the AT8 epitope is identical between human and mouse tau. To test whether exogenously applied tau could be found within the slices, recombinant tau assemblies were supplied to the culture media of wildtype slices, beneath the membrane, for a period of three days. After this time, we observed human tau present in the slices by western blot, indicating the transfer of tau assemblies from the media to the slice (Fig. 3g) (Additional file 1: Supplementary Fig. 2). Taken together, these data demonstrate that exogenously supplied tau assemblies are taken up into OHSCs and can induce the formation of hyperphosphorylated, insoluble tau species.
To further characterise the induced aggregates, we investigated the subcellular and regional location of tau lesions. We used recombinant tau assemblies to induce seeding owing to the high confidence that AT8-reactive aggregates result from seeded aggregation rather than the input tau. Within cell bodies, we observed large aggregates in peri-nuclear regions (Fig. 4a). Additionally, numerous smaller tau puncta were found along the length of neurites. Puncta were interrupted by regions apparently devoid of hyperphosphorylated tau. In contrast, Map2 staining revealed the presence of intact neurites, indicating that the punctate distribution of tau is not a consequence of neuronal fragmentation. They further demonstrate that neurons are able to tolerate tau aggregation to a certain degree without gross loss of morphology or overt toxicity. We compared levels of seeding between regions of the hippocampal slices. We observed the presence of AT8 positive structures in neurons within all subdivisions (Fig. 4b). However, AT8 reactivity was considerably greater within the CA1 region compared to CA2 and CA3. Approximately 80% of AT8-postitive structures were found in CA1, compared to ~ 10% in each of CA2 and CA3 (Fig. 4c). We examined levels of tau as a potential underlying cause of CA1 susceptibility but observed comparable expression levels across different regions (Fig. 4d). In summary, these results demonstrate that challenge of OHSCs with assemblies of tau induces the accumulation of pathology in neurites and cell bodies, predominantly in CA1 neurons, resulting in widespread accumulation of intracellular hyperphosphorylated tau structures.
To determine the time-dependence of this tau pathology, we next performed a time course following the addition of seed. Slices were fixed at various time points following challenge with 100 nM recombinant assemblies or buffer only (Fig. 5a). As above, slices that were not exposed to tau assemblies developed no robust evidence of hyperphosphorylated tau puncta. However, challenge with tau assemblies resulted in increasing levels of bright AT8-positive structures over time, consistent with seeded aggregation of intracellular pools of tau (Fig. 5b). At 1 week after challenge, isolated AT8 positive puncta were observed as well as diffuse AT8 staining. A week later, puncta became more numerous and a few large aggregates were observed. At 3 weeks after challenge with tau assemblies, AT8 staining was widespread with the presence of numerous aggregates that occupied entire cell bodies. This level of staining was maintained at approximately the same level between 3 and 4 weeks, suggesting that the induced aggregation is complete at 3 weeks post-challenge in this system. This timepoint was therefore selected for all further experiments. During the first three weeks, AT8 staining followed an exponential curve with a doubling time of just under a week (Fig. 5c). The size of AT8-positive structures was similarly found to increase over time. Stained areas greater than 50 µm2, generally only present within cell bodies, were found to be largely absent at 1 week post-challenge, but subsequently increased in prevalence (Fig. 5d). This suggests that the amplification of aggregates within individual neurons contributes to the overall increase in AT8 signal. The results are therefore consistent with a model of growth of hyperphosphorylated tau structures via a process of templated aggregation following exposure to seed-competent tau assemblies.
The above results demonstrate that our OHSC model exhibits behaviour consistent with the prion-like spread of tau. However, the dose we used (100 nM monomer equivalent) represents a concentration in excess of ISF and CSF tau concentrations, which occupy the low nanomolar to picomolar region. We therefore investigated the response of OHSCs to varying doses of exogenously supplied tau assemblies. Remarkably, we found that a reduction of seed concentration from 100 to 30 nM resulted in virtually no seeded aggregation being detectable within the slice at three weeks post-challenge (Fig. 6a). Whereas cell bodies reactive for AT8 could be observed when challenged with 100 nM tau assemblies, only very rare and small AT8-positive assemblies in neurites were observed following challenge with 30 nM tau. Conversely, increasing exogenous tau concentration from 100 to 300 nM increased the AT8-immunoreactive area by almost tenfold, consistent with a non-linear effect of tau concentration in this range (Fig. 6b). To exclude any effect of the culture membrane on the efficiency of tau uptake, we applied tau at the same concentrations directly to the apical surface of the slices in a volume of 20 µl. Under both experimental set-ups we observed robust induction of seeding at 100 nM, but not at 30 nM (Fig. 6c). Thus, the local concentration of tau governs seeded aggregation and is independent of application route. The smaller volumes required for apical application of tau permitted challenge with an extreme concentration of supplied tau, at 1000 nM. Here we observed a plateauing of percent AT8-positive area and the presence of AT8 reactivity in almost all neurons (Additional file 1: Supplementary Fig. 3). These results demonstrate that tau seeding in OHSCs only occurs efficiently at concentrations above 100 nM of supplied assemblies, and plateaus in the low micromolar range.
Independently acting infectious particles such as viruses retain infectivity upon dilution until they are diluted out at endpoint. They display one-hit dynamics where the proportion of infected cells, P(I), can be described by the equation P(I) = 1 − e−m where m is the average number of infectious agents added per cell. To determine whether tau assemblies display these properties, we titrated tau assemblies onto HEK293 cells expressing P301S tau-venus. Here, where conditions have been optimised for sensitive detection of seeding, and tau assemblies are delivered directly to the cytoplasm with transfection reagents, we observe that seeding activity is proportional to dose and can be titrated down. The observed level of seeding closely follows a one-hit titration curve (Fig. 7a). Thus, tau assemblies have the intrinsic ability to act as independent particles when tested in reporter cell lines. This is in direct contrast to the results observed in OHSCs where seeding reduces much more rapidly as tau assemblies are diluted than would be expected under a single-hit model (Fig. 7b). One potential explanation for these differences is that clearance mechanisms in intact tissue inherently prevent single-particle activity. To test this, we titrated AAV1/2.hSyn-GFP particles expressing GFP and measured the percent of Map2-positive neurons that were transduced. We observed that AAV1/2 behaved in a manner consistent with one-hit dynamics (Fig. 7c). Thus, tau assemblies differ from classical infectious agents and do not titrate in a manner expected of independently acting particles in mouse neural tissue. Rather, in this system, seeding is a behaviour that only emerges at high concentrations of extracellular tau assemblies.