Pro-aggregant P301L/S320F-tau-Dendra2 forms authentic neurofibrillary inclusions in BSCs
rAAV-mediated expression of human P301L/S320F-tau induces rapid and robust fibrillar tau pathology in BSCs . Dendra2 has been used in previous studies to identify localized distributions of tau , but we sought here to exploit the photoswitchable properties of Dendra2 [6, 32] to gain insight to the dynamics underlying soluble and aggregated tau protein. We generated rAAVs of Dendra2 alone, and Dendra2 fused to the C-terminus of human 4R0N WT tau and human 4R0N P301L/S320F tau. Dendra2 is a photoswitchable fluorescent protein that emits only green fluorescence until it is exposed to short-wavelength light, which induces an irreversible conformation change to switch their spectral properties to emit red fluorescence and then reduce total green fluorescence  Any proteins synthesized post-conversion will emit green fluorescence above the residual green fluorescence allowing assessment of protein dynamics using long-term live microscopy [6, 32].
We characterized the development of tau pathology in BSCs that were transduced with Dendra2, WT-tau-Dendra2 and P301L/S320F-tau-Dendra2 beginning at 10 DIV, as a large proportion of cells expressing P301L/S320F-tau-Dendra2 already bear tau inclusions. Biochemical assessments  identify overexpression of total tau (3026)  in BSCs expressing WT-tau-Dendra2 and P301L/S320F-tau-Dendra2, as well as, accumulation of soluble tau phosphorylated at Ser202 (CP13) and Ser396/404 (PHF-1) (Fig. 1a). BSCs expressing P301L/S320F-tau-Dendra2 developed sarkosyl-insoluble, hyperphosphorylated tau that was not detected in BSCs expressing Dendra2 or WT-tau-Dendra2. We also identified no increases in 50 kDa tau, which may occur if the Dendra2 tag was being cleaved from tau (Fig. 1a). We also examined the distribution of tau inclusion pathology in BSCs by immunohistochemistry (Fig. 1b). Both WT-tau-Dendra2 and P301L/S320F-tau-Dendra2 transduced BSCs show overexpression of tau as detected by 3A6 staining , with WT-tau-Dendra2 transduced BSCs developing Ser396/404 phosphorylated tau throughout the cytoplasm of neurons. In contrast, P301L/S320F-tau-Dendra2 transduced BSCs show Ser396/404 phosphorylated tau exclusively in the soma by 10 DIV.
We examined fibrillar tau pathology by staining tau-Dendra2 transduced BSCs with the dye Thiazin Red which recognizes NFTs in human AD [40, 41]. BSCs transduced with P301L/S320F-tau-Dendra2 develop aggregates that are Thiazin red positive, with staining absent in BSCs transduced with Dendra2 or WT-tau-Dendra2 (Fig. 1c). Quantification identified that 86.72 ± 4.45% (mean ± SEM) of P301L/S320F-tau-Dendra2 cells are Thiazin Red positive (Fig. 1d). We further studied P301L/S320F-tau-Dendra2 transduced BSCs by immuno-electron microscopy (EM) with the 7F2 antibody to tau phosphorylated at Thr205  (Fig. 1e). These studies show that fibrillar tau accumulates in these cultures comparable to those we identified in BSCs transduced with untagged pro-aggregant tau  and those found in human AD [15, 29]. We also evaluated cytotoxicity in BSCs expressing Dendra2, WT-tau-Dendra2, P301L/S320F-tau-Dendra2 by an EthD-1 uptake assay at 10 DIV, 1 month and 2 months in culture and identified no increased toxicity at any of these time points compared to the Dendra2 control (Fig. 1f) similar to our previous findings using untagged human tau rAAVs .
These data confirm that tau pathology with a Dendra2 tag recapitulates previous studies using tau or tau-EGFP transduced BSCs, with P301L/S320F-tau-Dendra2 transduced BSCs developing mature neurofibrillary pathology by DIV 10 . Indeed, after 10 DIV, we have established that the majority of inclusions within cells in the rAAV P301L/S320F-tau-Dendra2 transduced BSCs are fibrillar aggregates.
Optical pulse labeling methodology reveals P301L/S320F-tau-Dendra2 shows turnover in BSCs
The precise kinetics and stability of tau inclusions bearing aggregated tau and their relationship to neurodegeneration remains to be determined. Here we examined the synthesis and turnover of WT tau compared to pro-aggregant P301L/S320F tau by executing long-term optical pulse labeling experiments to estimate the turnover of tau in these conditions. A schematic diagram of the long-term imaging timeline is shown in Fig. 2a. We tracked the red fluorescent photoconverted Dendra2-tagged tau pool using long-term imaging from 10 to 31 DIV, as well as newly synthesized non-photoconverted green fluorescent Dendra-2 tagged tau (Fig. 2b). Upon population level quantification, amounts of red photoconverted P301L/S320F-tau-Dendra2 was significantly higher than Dendra2 or WT-tau-Dendra2 throughout the long-term imaging study from 1 to 21 days post photo-conversion (Fig. 2c). When estimating half-lives from this data; Dendra2 alone shows a half-life of 2.47 days; 95% CI [2.24, 2.69], r2 = 0.99, WT-tau a half-life of 2.67 days; 95% CI [2.47, 2.87], r2 = 0.99 and P301L/S320F-tau shows a significantly longer half-life of 7.48 days; 95% CI [5.44, 11.06], r2 = 0.89. These results indicate that WT tau turns over in the same time frame as Dendra2 alone; whereas P301L/S320F-Dendra2 pro-aggregant tau is more stable but does decline over time albeit with a longer half-life compared to Dendra2 or WT-tau-Dendra2, suggesting a slower but apparent turnover. We also confirmed that the half-life of Dendra2 does not significantly depend on initial fluorescent intensity or expression levels (Supp. Figure 1), as per previous reports .
It is plausible that aggregated tau may accumulate in BSCs and in disease due to a higher overall production rate of tau. Previous studies have suggested that some MAPT mutations may increase overall tau production in tau transgenic mice [54, 55]. In human AD patients, CSF tau is also produced at a higher rate than controls . We therefore tracked the production of new pools of WT and P301L/S320F tau by quantifying total green fluorescence which would be expected to increase over time due to protein synthesis. After photoconversion at 10 DIV, we imaged and quantified the newly produced green non-photoconverted tau pool using long-term live imaging until 31 DIV (Fig. 2b, d). Dendra2 alone, WT tau and P301L/S320F tau all increased at a similar rate, with newly synthesized P301L/S320F tau significantly accumulating more at 14 DIV after the initial photoconversion (as predicted by the slower decline of red photoconverted tau), and WT tau and Dendra2 showing a faster turnover of the newly synthesized protein at these long-term imaging periods.
To examine the dynamics of tau turnover, we further evaluated, at an individual cellular level, the rate of tau turnover and replacement in BSCs. For this analysis, a sample of individual cells alive at 31 DIV were selected as any cell death could bias half-life calculations. We also demonstrated earlier no significant levels of toxicity above Dendra2 control up to 2 months in culture. This analysis of living individual cells identified the mean half-life of P301L-S320F-tau-Dendra2 as 7.99 days; 95% CI [6.87, 9.12], compared to Dendra2 at 3.27 days; 95% CI [3.02, 3.51] and WT-tau-Dendra2 at 2.70 days; 95% CI [2.41, 2.99] (Fig. 3a). We also performed a probability density analysis, confirming the P301L/S320F tau shows a significantly longer half-life (Fig. 3b). Representative images of individual inclusions turning over in pro-aggregant P301L/S320F-Dendra2 BSCs are shown in Fig. 3c. In many cases, the red photoconverted tau in the soma, decreased over time and was simply replaced by green fluorescence. If the amount of red fluorescence was relatively small, then this was often completely replaced by green fluorescence within 21 days. In contrast, larger initial levels of red cytoplasmic fluorescence, would show decreasing red fluorescence and replacement by green fluorescence, increasingly over 7–21 days. However, the overall size of the fluorescent ‘inclusion’ appears to remain consistent or only slightly increased.
Overall, by characterizing and then using optical pulse labeling methods in BSCs, we have identified that once a tau inclusion is formed, it can turn over and is replaced by newly synthesized pools of tau whilst the overall size of the inclusion is restricted.
Extended time in culture increases the half-life of P301L/S320F-tau-Dendra2 inclusions
After identifying that tau inclusions that were formed after 10 DIV turned over with a half-life of ~ 1 week, we sought to identify how ageing or duration in culture may affect half-life and turnover of these inclusions. A schematic diagram of the long-term imaging timeline is shown in Fig. 4a. We tracked the red fluorescent photoconverted Dendra2-tagged tau pool using long-term live imaging after photoconverting at ~ 1 month in culture, as well as newly synthesized non-photoconverted green fluorescent Dendra-2 tagged tau.
As for earlier analysis, we examined at an individual cellular level, the rate of tau turnover and replacement in BSCs that had been aged for ~ 1 month prior to photoconversion. This analysis of living individual cells identified the mean half-life of P301L-S320F-tau-Dendra2 as 15.80 days; 95% CI [14.42, 17.18], compared to Dendra2 at 3.67 days; 95% CI [3.44, 3.91] and WT-tau-Dendra2 at 3.97 days; 95% CI [3.74, 4.19] (Fig. 4b) after 1 month of aging. We also performed a probability density analysis, confirming the P301L/S320F tau shows a significantly longer half-life (Fig. 4c).
We then extended the time in culture prior to photoconversion and long-term imaging to ~ 2 months—a schematic diagram of this timeline is shown in Fig. 4d. We then quantified the rate of tau turnover and replacement in living cells in BSCs that had been aged for ~ 2 months prior to photoconversion. We identified the mean half-life of P301L-S320F-tau-Dendra2 after ~ 2 months of aging, as 22.80 days; 95% CI [20.68, 24.93], compared to Dendra2 at 4.17 days; 95% CI [3.72, 4.62] and WT-tau-Dendra2 at 4.13 days; 95% CI [3.72, 4.54] (Fig. 4e). Probability density analysis confirmed the P301L/S320F tau shows a significantly longer half-life (Fig. 4f).
Taken together, these findings suggest that extended culture periods or exposure to a tau inclusion increases the half-life of the inclusion turnover to ~ 2 weeks after 1 month of aging, and to ~ 3 weeks after 2 months of aging. This highlights the utility of this system for studying differing tau kinetics with age or different culture conditions.
Development of a seeded BSC model of tau inclusion pathology
Tau pathology progresses spatiotemporally after disease onset in AD  and the seeding and propagation of tau inclusions has been identified as a possible mechanism for the spread of tau pathology in tauopathies . To examine aggregation triggered by the addition of exogenous K18 tau preformed fibrils, we expressed different human tau variants in BSCs with and without fibrils. Biochemical analysis revealed that P301L 4R0N human tau can be robustly seeded in BSCs by the addition of exogenous K18 tau fibrils and phosphorylated insoluble tau accumulates (Fig. 5a). BSCs transduced with an EGFP control, WT or S320F tau and seeded with K18 tau fibrils only accumulated soluble phosphorylated tau (Fig. 5a). P301L/S320F tau seeded with K18 tau fibrils did not accumulate any further insoluble or phosphorylated tau (Fig. 5a). This data highlights the versatility of this platform to evaluate tau propagation and seeding.
To facilitate optical pulse labeling experiments of this seeded model, we generated rAAVs of Dendra2 fused to the C-terminus of human 4R0N P301L tau. We then biochemically characterized BSCs expressing P301L-tau-Dendra2 with and without K18 tau fibrils at the time point we planned to perform live imaging. P301L-tau-Dendra2 transduced BSCs were robustly seeded by K18 tau fibrils, showing the accumulation of insoluble, phosphorylated tau (Fig. 5b). We also confirmed no increased 50 kDa tau which could occur if the Dendra2 tag was being cleaved off.
We also performed immunohistochemistry to identify the distribution of inclusion pathology (Fig. 5c). Both P301L-tau-Dendra2 seeded and non-seeded BSCs show overexpression of tau . P301L-tau-Dendra2 BSCs without the addition of exogenous K18 tau fibrils develop phosphorylated tau throughout the cytoplasm of neurons, whilst P301L-tau-Dendra2 BSCs that were seeded with K18 tau fibrils show accumulation of this pathology almost exclusively in the soma. To identify any neurofibrillary pathology, we stained with Thiazin Red and identified seeded P301L-tau-Dendra2 BSCs develop aggregates that are Thiazin red positive, with staining absent in BSCs transduced with P301L-tau-Dendra2 without the addition of fibrils (Fig. 5d). Quantification revealed at 10 days post-seeding with K18 fibrils 84.77 ± 3.41% of P301L-Dendra2 expressing cells were also Thiazin Red positive (Fig. 5e). Upon further examination by immuno-EM (Fig. 5f), seeded P301L-tau-Dendra2 BSCs accumulate tau filaments that are 7F2 positive. We also evaluated cytotoxicity in BSCs expressing P301L-tau-Dendra2 with and without exogenous K18 tau fibrils by an EthD-1 uptake assay at 10 DIV, 1 month and 2 months in culture post the addition of fibrils. We identified some increased toxicity in the seeded P301L-tau-Dendra2 BSCs 2 months post addition of seeds compared to the Dendra2 control (Fig. 5g).
These data confirm that P301L tau BSCs can be seeded with exogenous K18 tau fibrils to develop mature neurofibrillary pathology and provide a novel three-dimensional model of ‘seeded’ tau aggregation.
Tau inclusions form rapidly in both intrinsic and ‘seeded’ BSC models
One of the main advantages of using ex vivo BSC models is their accessibility and ease for long-term imaging studies. As we identified that P301L-tau-Dendra2 tau can be induced to aggregate with the addition of exogenous K18 tau fibrils, we established a long-term imaging paradigm to track the formation and fate of inclusions in this model following addition of seeds. A schematic diagram of the imaging timeline to track inclusion formation with the addition of seeds is shown in Fig. 6a. P301L-tau-Dendra2 BSCs that were not seeded show turnover of tau after the initial and second repeated photoconversion suggesting a constant dynamic pool of tau turnover in this model (Fig. 6b). In P301L-tau-Dendra2 BSCs that undergo seeding with exogenous K18 tau fibrils, tau inclusions can be seen forming from both newly synthesized green fluorescent tau and initially photoconverted red fluorescent tau rapidly in the 10 day period after addition of seeds. When photoconverted for a second time, these inclusions remain more stable than the unseeded condition for the additional 10 day imaging period, highlighting that these stable aggregates rapidly formed in the previous 10 day period and may turnover more slowly.
To alternatively visualize and track tau inclusion formation, we used a MAP-2 promoter to restrict expression of P301L/S320F-tau-Dendra2 within neurons in BSCs. We have previously confirmed that tau inclusion pathology is driven in neurons using neuronal-specific promoters . MAP-2 driven P301L/S320F-tau-Dendra2 expression resulted in Thiazin Red positive inclusions in BSCs like those observed when tau is expressed under the hCBA promoter (Supp. Figure 2a). Furthermore, as tau expression from the MAP-2 promoter is slower, this paradigm can be used to image the formation of neuronal tau inclusions over time (Supp. Figure 2b). Imaging P301L/S320F-tau-Dendra2 in BSC neurons showed that inclusions can form rapidly in less than 24 h or more slowly over a period of 96 h (Supp. Figure 2b) similar to those forming rapidly in seeded P301L-tau-Dendra2 BSCs and in vivo in tau transgenic mice . Once formed, these inclusions also exist in culture for at least a further 7 DIV similar to our earlier findings using hCBA promoter.
Taken together, these findings in both ‘seeded’ and ‘intrinsic’ tau inclusion BSC models show that this system can be used to visualize the rapid formation of inclusions in neurons over a period 12–96 h, and their fate can continue to be tracked long term.
Seeded P301L-tau-Dendra2 inclusions also show appreciable turnover of tau
Tau seeding and propagation have been a major focus of recent studies , but the precise dynamics of inclusions induced by seeding has not been established. After observing that exogenous K18 tau fibrils can induce P301L-tau-Dendra2 to form inclusions rapidly over a 10 day period, we aimed to compare the longevity of soluble non-seeded P301L-tau-Dendra2 tau to aggregated, seeded P301L-tau-Dendra2 using long-term optical pulse labeling experiments. A schematic diagram of the imaging timeline is shown in Fig. 7a. After 14 DIV, P301L-tau-Dendra2 BSCs were seeded with K18 tau fibrils, or untreated for a further 10 DIV, and at this point BSCs underwent photoconversion. The red fluorescent photoconverted Dendra2-tagged tau pool and newly synthesized green fluorescent Dendra2-tagged tau were tracked using long-term live imaging (Fig. 7b). Population level quantification to assess protein turnover (Fig. 7c) identified levels of photoconverted P301L-tau-Dendra2 in the presence of K18 tau fibrils were significantly higher than non-seeded P301L-tau-Dendra2 BSCs throughout the live imaging study from 1 to 21 days post photoconversion. Estimated half-lives of the photoconverted Dendra2-tagged P301L tau reveal P301L tau without seeds had a half-life of 2.72 days; 95% CI [2.34, 3.17], r2 = 0.97, whilst seeded P301L tau showed a half-life of 8.16 days; 95% CI [5.51, 14.01], r2 = 0.88. These findings highlight that seeded P301L-tau-Dendra2 has a significantly longer half-life than non-seeded P301L-tau-Dendra2. We also confirmed that the half-life of P301L-Dendra2 alone or in the presence of K18 tau fibrils does not significantly depend on initial fluorescent intensity or expression levels (Supp. Figure 3), as per previous reports . Of note, the longevity and half-life of tau inclusions formed through the seeding of P301L-tau-Dendra2 was similar to that of inclusions formed through intrinsic P301L/S320F-tau-Dendra2 aggregation (Fig. 2c), suggesting inclusions comprised of aggregates formed either intrinsically or extrinsically show a similar turnover.
To estimate any effects of seeding on tau production, newly synthesized P301L-tau-Dendra2 was quantified (Fig. 7d) and found to accumulate at similar rates in the presence and absence of seeds. P301L-tau-Dendra2 seeded BSCs accumulated significantly higher newly synthesized tau 14 DIV after the initial photoconversion with non-seeded P301L-tau-Dendra2 showing a more rapid turnover of newly synthesized tau over these long-term imaging periods [similar to WT-tau-Dendra2 and Dendra2 alone (Fig. 2)].
These data suggest that seeding of P301L tau induces tau inclusions that show turnover at a slower rate than unseeded P301L tau, but the addition of K18 tau fibrils does not affect overall tau production.
We then examined the fate of inclusions under P301L-tau-Dendra2 seeded conditions, where population analysis suggests some clearance of seeded tau inclusions, at an individual cellular level. For this single cell analysis, a sample of individual cells alive and emitting green fluorescence at 45 DIV at the end of the imaging period to omit any bias from cells that had died. This analysis of living individual cells identified the mean half-life of P301L-tau-Dendra2 at 3.161 days; 95% CI [2.90, 3.43], compared to P301L-tau-Dendra2 + K18 fibrils at 7.73 days; 95% CI [6.26, 9.20] (Fig. 8a). We also performed a probability density analysis, confirming the seeded P301L tau shows a significantly longer half-life (Fig. 8b). Representative images of individual tau inclusions in cells turning over in seeded P301L-tau-Dendra2 BSCs are shown in Fig. 8c. As observed in P301L/S320F-tau-Dendra2 BSCs, the red photoconverted tau, can be seen to decrease over time and be replaced by green fluorescence whilst the overall size of the fluorescent ‘inclusion’ appears to remain consistent or slightly increased in the seeded P301L-tau-Dendra2 BSCs.
Extended time in culture increases the half-life of seeded P301L-tau-Dendra2 inclusions
After identifying that tau inclusions in P301L-tau-Dendra2 BSCs formed 10 DIV after the addition of K18 tau fibrils turned over with a half-life of ~ 1 week, we sought to determine how ageing or longer durations in culture may affect the half-life and turnover of these inclusions. A schematic diagram of the long-term imaging timeline after longer exposure to inclusions and K18 tau fibrils is shown in Fig. 9a. We tracked the red fluorescent photoconverted Dendra2-tagged tau pool using long-term live imaging after photoconverting at ~ 1 month after the addition of K18 tau fibrils.
We determined in individual living cells, the rate of tau turnover and replacement in BSCs that had been aged for ~ 1 month after the addition of K18 tau fibrils prior to photoconversion. This analysis identified the mean half-life of P301L-tau-Dendra2 at 3.80 days; 95% CI [3.53, 4.07], compared to P301L-tau-Dendra2 + K18 fibrils at 17.92 days; 95% CI [16.32, 19.52] (Fig. 9b) ~ 1 month after the addition of fibrils. We also performed a probability density analysis between the two conditions (Fig. 9c).
We then extended the time in culture after fibril addition prior to photoconversion and long-term imaging to ~ 2 months—a schematic diagram of this timeline is shown in Fig. 9d. This analysis identified the mean half-life of P301L-tau-Dendra2 at 4.36 days; 95% CI [4.02, 4.71], compared to P301L-tau-Dendra2 + K18 fibrils at 23.73 days; 95% CI [21.57, 25.90] (Fig. 9e) ~ 2 months after the addition of fibrils. Probability density analysis confirmed the seeded P301L-tau-Dendra2 shows a significantly longer half-life (Fig. 9f).
Taken together, these findings indicate that extended culture periods or exposure to a tau inclusion after seeding increases the half-life of the inclusion turnover 1 month post seeding to approximately 2 weeks, and 2 months post seeding to approximately 3 weeks. This further underscores the similarities between intrinsic and seeded tau inclusions and the slowing down of tau turnover over time in these models.