Development of a robust, physiologically relevant neuronal tau aggregation assay for tau-related drug discovery

Abstract

Therapeutic interventions to block extracellular tau seeding to prevent endogenous tau aggregation and progression of Alzheimer’s disease pathology are currently being investigated in clinical trials. However, the translation of promising preclinical findings to benefit clinical outcomes remains problematic due to the lack of pathophysiological models that recapitulate key features of sporadic Alzheimer’s disease-related tauopathies. We developed a primary neuronal tau (hTau) seeding and propagation model. Neurons expressing wild-type human tau protein at a physiological level, seeded with a sub-nanomolar tau derived from Alzheimer’s disease brain tissues, rapidly and robustly form tau aggregates and develop impaired mitochondrial function. Resulting aggregates are quantitatively measured using automated high-content algorithms. The considerable pathophysiological relevance, coupled with a highly sensitive dynamic range, makes this assay a valuable model system for studying tau pathobiology and an efficient screening tool for modulators of tau aggregation. Using this model, we demonstrate that by targeting a phosphorylation-specific epitope of tau, an antibody effectively stops tau aggregation.

In Alzheimer’s disease, the axonal protein tau changes from its native intrinsically disordered structure into insoluble, cross-β sheet-rich, paired helical filaments that accumulate as tangles in neurons (1). Accumulation and propagation of tau aggregates along neural circuits defines the disease pathology stage progression (2). In its physiological form, tau binds to and stabilizes microtubules in neurons. Under disease conditions, tau aggregates into tangles and becomes hyperphosphorylated, supporting the hypothesis that the binding of tau to microtubules is negatively regulated by post-translational modifications, which in turn lead to tau aggregation (3). Once formed, pathological tau aggregates can act as “seeds” to convert neighboring soluble tau into aggregates (4) and move from neuron to neuron to propagate tau pathology, either transynaptically or via extracellular vesicles (5, 6, 7). Growing evidence indicates that once aggregated, tau loses its physiological function and gains toxic functions, including disrupting axonal transport and long-term potentiation (LTP), clumping of mitochondria, and eventually causing cell death (3). These pathological features likely account for the observation that in AD patients, the number and extent of tangles closely correlate with cognitive decline (8, 9). Preclinical studies have shown that reducing total tau, and therefore decreasing associated tau aggregation, prevents both amyloid-beta β peptide (Aβ) associated LTP and behavior deficits in tau knockout transgenic mouse models (10, 11). A recent Phase 1B clinical trial reported using antisense oligonucleotide to reduce tau pathology in brains of patients with AD which led to a reduction in tau CSF biomarkers, including the hyperphosphorylated tau (pT181), a biomarker that is usually associated with accelerated cognitive decline (12), further strengthening the rationale for reducing tau aggregation as a disease-modifying treatment for AD patients.

Drug development for tau aggregation inhibitors and studies to understand the associated pathobiology require a physiologically relevant AD cell model. To date, multiple tau aggregation cell models have been developed and were used in preclinical development of drugs that have failed to meet their primary endpoints in clinical trials. These cell models mostly examine seeding of overexpressed, truncated human tau protein carrying aggregation-prone mutations in non-AD relevant kidney cells (HEK293) and rely on tagging tau protein with large fluorescent molecules as an aggregation readout (13, 14, 15, 16, 17). Additionally, tau seeds used to induce aggregation are in vitro prepared filaments formed with polyanion cofactors (18) that require lipofectamine delivery for internalization, representing potentially non-relevant mechanisms for AD. One study overcame these constraints by using AD brain-derived seeds in a primary murine neuronal model, but this study was specific to aggregation of murine tau protein (19) and may not fully represent the pathophysiology of human tau protein.

Learning from the limitations present in the existing cell models, we sought to develop a murine primary neuronal model system that closely recapitulates physiological human tau aggregation in AD. Our hTau neuronal seeding model uses a sub-nanomolar amount of human AD brain-derived tau aggregates to induce human tau protein aggregation in hippocampal and cortical neurons. Here, we systematically define components of the hTau seeding model, including the seed preparation, tau aggregate characteristics, aggregation kinetics, assay dynamic range, aggregation-induced mitochondrial deficit, and application of the model for assessing the ability of tau antibodies to block tau aggregation and propagation.

Results

AD brain tissue tau pathology characterization and extraction

The level of pathological tau aggregates varies between patients and dictates the quality and quantity of tau aggregates and paired helical filaments (PHFs) that can be extracted. Accordingly, AD patient brain samples with the highest level of tau pathology were selected by quantifying the amounts of tau aggregates in crude brain lysate. For this study, postmortem frontal cortical tissue at Braak stage VI of both male and female patients was used (Table S1). To minimize variability in tau pathology level between brain regions, we divided the frozen brain tissues (usually >10 g/piece) into smaller pieces (0.5–1 g/piece) and then prepared aliquots of stock tissue lysates consisting of multiple divided small pieces from the same donor (Fig. 1A). From our experience, brain lysates stored at −80 °C can be used for at least 6 months. As an experimental control, normal donor brains were characterized in the same way. Total tau protein (MSD) and tau aggregate-specific (HTRF) assays revealed that while total tau amount is similar between normal and AD brain lysates (MSD data not shown), AD lysates have significantly higher levels of tau aggregates as compared to brains without pathology (Fig. 1B). Based on these data, we selected AD brains with the highest levels of tau aggregates to be used for subsequent seeding studies and excluded the brains with the lowest level of pathology (Fig. 1B, red arrow). Two types of seeds, lysate (S1) representing total soluble and insoluble tau aggregates, and sarkosyl insoluble (SP) tau (S5 fraction) consisting of insoluble high molecular weight filaments, were prepared using a modified protocol based on previous publications (19, 20) (Fig. 1C). Specifically, SP tau prepared using this method shows the typical electrophoresis migration profile observed in AD with 3 major bands running at 60, 64, and 68 kDa and a minor band at 72 kDa that are made of hyperphosphorylated tau isoforms 1, 2 + 4, 3 + 5, and 6, respectively (Fig. 1D, AT8) (21).

Figure 1.

Tau pathology characterization of human AD and normal brain tissues.A, AD brain tissue processing schematic diagram. B, representative HTRF assay shows AD brains have a high level of tau aggregates compared to normal brains. AD brain with low tau pathology is excluded from the study (case no. 15, red arrow). Each bar represents mean ± s.d. C, schematic diagram detailing lysate preparation and sarkosyl insoluble tau (SP, S5) extraction. D, immunoblots (AT8) show SP tau purified from AD brains comprise of a typical profile: 3 major high molecular weight tau aggregates at 60, 64, 68 and a minor band at 72 kDa that are made of tau isoforms 0N3R, 1N3R + 0N4R, 2N3R + 1N4R, and 2N4R, respectively.

AD brain-derived tau induces tau aggregation in hTau neurons

Next, we examined whether SP tau induces tau aggregation in the hTau murine neurons. The hTau transgenic mice express six isoforms of human tau at close to physiological levels (22) while a mixed culture of neurons isolated from the hippocampus and from the cortex of hTau embryos (E16) primarily expresses the 0N3R isoform (data not shown). Importantly, in this model, GFP disrupts the expression of the endogenous murine tau gene, resulting in the deletion of murine tau protein (Fig. S1). Cultured neurons display deep cortical layer markers (Ctip2, green; Satb2, red) and have both excitatory (VGLUT1, red) and inhibitory types (VGAT, orange) (Fig. S2). To induce tau aggregation, neurons were incubated with AD SP tau (0.5 nM) for 48 h and cultured for an additional 9 days before fixing with cold methanol to remove soluble tau protein (19) and immunostaining with an anti-tau antibody that recognizes all isoforms of tau (CP27) for insoluble tau aggregates and MAP2 for dendrites (Fig. 2A). We found that seeding with pathological AD SP tau induces robust tau aggregation in neurons 2 weeks post-seeding, whereas seeding with normal (N) brain-derived SP tau does not induce an aggregation phenotype (Fig. 2B). Importantly, no visible tau aggregates are detected in tau KO neurons treated with AD SP tau seeds (Fig. 2B), confirming that exogenous AD-seeds taken up by neurons are below detection in the assay and that the observed aggregates in hTau neurons are formed from endogenous tau protein. These newly formed aggregates accumulate in the soma and along neuronal processes. High-resolution images show newly formed tau aggregates are filamentous, misfolded (7–9/312–322, MC1) and hyperphosphorylated at previously reported pathological epitopes (1), including pS396/S404 (PHF1), pS202/T205 (AT8), pT231/S235 (AT180), and pT212/S214 (AT100) (Fig. 2, C and D).

Figure 2.

AD-derived SP tau induces tau aggregation in hTau neurons.A, schematic diagram of the hTau neuronal seeding workflow. B, representative IF images of tau aggregates (CP27, human tau, green) in hTau neurons (MAP2, magenta) seeded with AD SP. No aggregates formed in normal SP tau seeds treated hTau neurons or in Tau-KO neurons seeded with AD SP. Scale bar, 50 μm. C, map of pathological tau antibody epitopes on 2N4R tau protein. D, representative IF images depicting seed-induced filamentous hTau aggregates are misfolded and hyperphosphorylated at pathological epitopes recognized by antibodies CP27, MC1, PHF1, AT8, AT180, and AT100 (gray). Scale bar, 25 μm.

Tau aggregation is a tightly regulated, time-dependent process comprised of an initial formation of seed nuclei from monomeric tau protein (nucleation), and then an exponential replication of aggregates (seeding) (23). We investigated the kinetics of tau aggregation in seeded hTau neurons by quantifying the levels of tau aggregates at 2 days (early), 7 days (middle), and 14 days (late) post seeding. At early stages of seeding (2 days), no aggregates are observed (Fig. 3A). By 7 days post-seeding, a modest number of small, distinct, and round aggregates formed in the processes and soma of selective neurons. At 14 days post-seeding, we observed more neurons with formed aggregates and, similar to observations of AD pathology progression in human brains, subsets of aggregates with a densely packed filamentous morphology are backed into the soma (Fig. 3A). Automated segmentation and quantification of tau aggregates (Fig. 3B) measured a 3-fold increase in aggregation from 7 days to 14 days post seeding (Fig. 3D). Aggregate size (area), brightness (concentration), and number all positively correlated (Fig. 3E). No tau aggregates are detected in hTau neurons seeded with control normal brain SP-derived tau up to 14 days (Fig. 3D). Seeding is not toxic as neurite analysis measured no reduction in total neuronal processes between seeded and non-seeded control neurons (Fig. 3C). Together, these data indicate that in seeding, small round transient intermediates are initially formed and then mature over time into morphologically distinct larger and brighter filaments that closely resemble mature fibrils of tangle pathology in AD.

Figure 3.

Seed-induced hTau aggregation is time-dependent.A, representative images of aggregates formed at different time points post seeding. CP27, grey; MAP2, red; merged, yellow. Hoechst 33342, blue. Scale bar, 50 μm. B, examples of neurites segmented and quantified using Find Neurites algorithm and tau aggregates measured by Find Spot or Find Image Region algorithms in Harmony software. C, AD SP tau seeds treatment does not cause neuronal toxicity as assessed by dendrite area (MAP2). Each bar represents mean ± SEM (N = 2–4 replicates, 18 images/replicate, 8 planes/image). D, seed-induced tau aggregation increases over time (mean ± SEM, N = 2–4 replicates, 18 images/replicate, 8 planes/image). E, Correlation graph shows good agreement between tau aggregation (2 and 14 days post AD SP seeding) represented in total intensity, number of aggregates, and total area. F, corresponding Pearson R values relating to the correlation graph in (E).

Comparison of AD SP and lysate on tau seeding potency and assay dynamic range

Next, we assessed the seeding potency between SP and lysate (S1) fraction tau seeds. Neurons were seeded with different tau concentrations of SP tau or lysates, and the resulting induced aggregation was measured. For direct comparison of the two types of seeds, both SP and lysate seeds from the same patient (case 1) were used in addition to seeds prepared from pooled patients (mix) to rule out variabilities between patients that could influence seeding. We found that SP tau consistently had a greater seeding potency in both single and mixed patient material as compared to patient-matched tau isolated from lysate only, with a maximal seeding efficacy difference (EC50 0.03 versus EC50 > 2 nM) (Fig. 4, A and B). The observed increase in seeding potency is likely due to the isolation of aggregated tau in SP, whereas lysate includes a high percentage of soluble, non-aggregated tau in addition to aggregated tau. No neuronal toxicity is observed in either SP or lysate-seeded hTau neurons, as total neurite area (MAP2) was the same (Fig. 4C).

Figure 4.

AD seeds potency, hTau neuronal seeding assay dynamic range, and in vivo validation.A, representative IF images comparing the seeding potency between AD lysate (top two panels) and SP (bottom two panels). Tau seeds were prepared from the same patient (case. no 1). CP27, green; MAP2, magenta. Scale bar, 50 μm. B, AD lysate and SP tau seed dose-response curves in hTau neurons. Each point represents mean ± SEM, n = 3, 18 images/replicate, 8 planes/image. AD SP EC50 = 0.03, lysate EC50 = 1.22 nM. C, seeding does not induce neuronal toxicity as assessed by neurite area (MAP2) except for a minor reduction in the mixed AD lysate-treated neurons (mean ± SEM, n = 3 replicates). D, schematic illustration of in vivo P301S seeding study design. Seeds were injected into the CA1 hippocampus and the induced pathology is examined 6 weeks post-inoculation. E, representative photomicrographs of AT100 IR showing both AD lysate and SP-induced tau tangles in the ipsilateral hippocampus (AT100, pT212/S214). (F) Quantification of AD lysate and SP-induced IR in the whole hippocampus shows approximately 4.4× difference in lysate-induced pathology. Each point represents average level of AT100 IR within the hippocampus of a single brain (n = 4 animals/group).

AD SP tau seeds were also consistently more potent than AD lysate when evaluated in tau transgenic mouse models. For this study, we used a transgenic mouse model that expresses human tau gene MAPT with the P301S mutation, which is associated with autosomal-dominant disease in humans, and develops widespread tau pathology in the brain at 5 to 6 months of age (P301S) (24) Here, we inoculated mice at a young age (2 M) before onset of tau pathology with either AD lysate or SP tau in the CA1 hippocampus and examined the seed-induced pathology 6 weeks post injection using immunohistochemistry (AT100, pS212/S214) (Fig. 4D). While both AD lysate and AD SP induced abundant AT100 immunoreactivity (IR) in the CA1 brain region (Fig. 4E), similar to the in vitro hTau seeding model, SP seeds inoculation resulted in approximately 6× more pathology as compared to AD lysate seeds (Fig. 4F). No tau pathology was observed in mice injected with control brain S1 lysate. These in vivo data strongly agree with the in vitro hTau neuronal seeding model results and provide confidence in in vitro to in vivo alignment for supporting application in drug discovery workflows.

Seed-induced tau aggregation in hTau neurons propagates from neuron to neuron

In AD, the progressive accumulation of tau pathology along synaptically connected brain regions suggests that misfolded tau protein appropriates elements of neuronal synaptic transmission to propagate pathology (1). To examine if this pathophysiological phenotype is recapitulated in the seeding model, we performed tau seeding in synaptically connected co-cultures of neurons grown in microfluidic chambers (Fig. S2). In this setup, cultures of three populations of neurons are compartmentalized into three chambers with their axons projecting through interconnecting microchannels to form synapses (Fig. S2A). Dendrites do not project through the microchannels due to the distance (900 μm) (25). To seed tau aggregation in this model system, AD SP tau is added to neurons grown in the first chamber (P1) only (Fig. S2B). Importantly, due to hydrostatic pressure created by higher volumes of culture media in connecting cell populations (P2, P3), SP tau is restricted to the P1 population of neurons. Additionally, 2days after seeding, SP tau-containing media is completely washed out, further decreasing the risk of seed diffusion into the neighboring P2 and P3 population of neurons. As such, tau aggregation in neurons within P2 and P3 are primarily due to propagation of endogenous, newly formed tau aggregates. We examined multiple timepoints post-seeding (data not shown) and found that only after 9 weeks of extensive culture does abundant measurable propagation occur. At this time point, widespread tau aggregation is observed in the seeded P1 neurons, abundant aggregation has built up in the first synaptically connected P2 neurons, and visible aggregation is detected in the neurites and soma of selective neurons in the second synaptically connected P3 neurons (Fig. S2, B and C). As retrograde uptake of SP seeds by axons of P2 neurons could occur and contribute to aggregation in P2 neurons but not by axons of P3 neurons due to distance, the formation of aggregates in the P3 neurons convincingly shows that SP tau seed-induced aggregates in hTau neurons propagated from synaptically connected neurons, further demonstrating the pathophysiological relevance of the hTau neuron model.

Seed-induced tau aggregation impairs mitochondrial respiration

Mitochondrial dysfunction and oxidative stress are associated with early pathology in dementia and Alzheimer’s disease (26, 27). Furthermore, hyperphosphorylated tau has been shown to directly compromise mitochondrial distribution, maintenance, and bioenergetic functions in cell models (28) as well as in tau transgenic mouse models such as P301L (29, 30) and hTau (31). To investigate the pathological relevance of the AD SP-induced tau aggregation in the hTau neuronal model, we assessed mitochondrial function. Hippocampal and cortical neurons that express either human tau protein (hTau) or have endogenous murine tau deleted (KO) were seeded with AD SP at 0.1 nM. Mitochondrial respiration was measured by Seahorse at 1, 2 and 3 weeks post-seeding (Fig. 5A). As expected, a time-dependent increase in tau aggregation recognized by the MC1 antibody stain was observed in seeded hTau neurons (Fig. 5, B and C). Tau aggregation did not form in seeded KO neurons, nor in unseeded hTau and KO neurons. Impairment of mitochondrial respiration was observed to correlate with tau aggregation level (Fig. 5, D–G). At 1 and 2 weeks post-seeding, where tau aggregation levels were either low or moderate, no impairments in mitochondrial respiration were observed (Fig. 5D). In contrast, a significant reduction in maximal respiration and ATP production occurred in seeded hTau neurons at 3 weeks post-seeding (Fig. 5, E–G). The impairment in mitochondrial activity could be due to either the high level of tau aggregation or prolonged accumulative exposure to tau aggregation over 3 weeks, or both. Mitochondrial impairments are specific to tau aggregation, as neurons without tau aggregation, such as unseeded hTau neurons, seeded or unseeded KO neurons, did not have altered mitochondrial activity. These data therefore support that the tau aggregates formed in the hTau seeding model recapitulate one of the key functional deficits of AD pathology.

Figure 5.

Tau aggregation in hTau primary neurons impairs mitochondrial respiration.A, schematic diagram illustrating neurons were seeded at DIV 3, washed at DIV 5, and analyzed for tau aggregation by IF or mitochondria activity by seahorse at 1, 2, and 3 weeks post seeding. B, representative IF images showing tau aggregation (MC1, green) in hTau neurons at 1, 2, and 3 weeks post seeding. No aggregation is observed in seeded KO neurons, unseeded hTau, unseeded KO neurons. C, Quantification shows aggregation level increases over time in seeded hTau neurons. D, seeded hTau neurons at 3 weeks post seeding exhibits impaired mitochondria activity as measured by Seahorse MitoStress and not at 1 and 2 weeks post seeding. Neurons without aggregation do not have impaired mitochondria activity. E–G, Quantifications reveal while tau aggregation cause a marginal mitochondrial basal respiration impairment (E), maximal respiration (F) and ATP production (G) are significantly reduced. Graphs represent mean ± s.d. (n = 15 replicates/group). (∗p ≤ 0.05, ∗∗p ≤ 0.01, one-way ANOVA with Sidak’s multiple comparisons test against unseeded hTau or KO neurons condition).

Anti-tau antibody PHF1 depletes seed-competent tau aggregates

We next evaluated the assay’s potential in characterizing tau antibodies for drug discovery. This is in contrast to previous reported cell aggregation models used by the tau antibody discovery program that use overexpressed tau protein with mutations in non-AD-related HEK293 kidney cells (13, 14, 16, 32). We measured the ability of two widely used anti-tau antibodies, MC1 and PHF1, to immunodeplete seeding competent tau species (Fig. 6A). For this study, we selected AD lysate as the seeding material, as it contains both soluble and insoluble tau seeds. AD lysate was incubated with antibodies, and bound antibody-tau seed complexes were removed with protein G beads. A dose-response immunodepletion was performed where varying concentrations of anti-tau antibody (0.3, 3, 30, or 100 nM) were incubated with 100 nM tau seeding fraction material. The immunodepleted lysate was then added to hTau neurons to test residual seeding efficiency. The PHF1 antibody, but not the MC1 antibody, significantly dose-dependently depleted seeding efficacy (Fig. 6A). PHF1 reduced tau aggregation by 90% at 100 nM (1:1 M ratio) and by 41% at 30 nM (0.3:1 M ratio). In contrast, MC1 immunodepletion shows a non-statistically significant small seeding reduction when compared to the control IgG-treated lysate. As PHF1 shows most potency in reducing tau seeding, we further characterized the antibody using a full dose-response depletion (0.03 nM–300 nM, or 0.0003–3 M ratio) and determined PHF1’s seed-competent immunodepletion IC50 is 9.75 nM (Fig. 6B). To determine the translatability of the in vitro cell model finding to the in vivo animal model, we performed a similar seeding inoculation of MC1 or PHF1 antibody immunodepleted AD lysates into a tau transgenic mouse model, rTg4510 (33) (Fig. 6C). We chose this model for its accelerated 3-week seeding pathology in contrast to the aforementioned P301S model. Immunodepletion of AD lysate by PHF1 significantly reduced AT100 IR pathology by 77% as compared to IgG immunodepletion, whereas MC1 immunodepletion non-significantly reduced pathology by 46% (Fig. 6D). These data collectively demonstrate that for therapeutic antibody development, candidate anti-tau antibodies can be pre-screened and ranked in their ability to immunodeplete seeding-competent tau species using this sensitive hTau seeding assay to increase the success of in vivo validation and reduce animal usage.

Figure 6.

Anti-tau antibody PHF1 depletes seed-competent tau aggregates.A, representative IF images showing depletion of tau aggregation (CP27, green) in hTau neurons at following seeding with PHF1 immunodepleted AD lysates. In contrast, no change in tau aggregation was observed with control IgG or MC1 depleted AD lysates. B, PHF1, but not MC1 immunodepleted AD lysates show concentration-dependent significant reduction in seeding hTau neurons (n = 3 replicates). Mock-depleting lysate with Protein G magnetics beads alone, or isotype control IgG does not reduce seeding. (∗p ≤ 0.05, ∗∗∗∗p ≤ 0.001, two-way ANOVA with Sidak’s multiple comparisons test against IgG condition). Bars are presented as mean values ± s.d. for n = 3 replicates. C, PHF-1 IP seeding dose-response curve (n = 9 replicates). PHF1 IC50 = 9.75 nM. best-fit bottom = 12.7% (R² = 0.72). Seeding inhibition curve was fit using the 4-parameter log-logistic Hill equation with top constrained to 100. Data are presented as mean values ± s.d. for n = 6 to 9 replicates. D, schematic showing in vivo study design. AD lysates immunodepleted with antibodies were injected into the CA1 hippocampus of 2 M old rTg4510 mice. Pathology was analyzed 3 weeks post-injection of lysate by AT100 IR area in the CA1 hippocampus. E, compared to AD lysate immunodepleted with control IgG, PHF1 immunodepleted seeds (1.2 M ratio) induced significantly less tau pathology. In contrast, MC1 immunodepletion resulted in marginal pathology reduction as determined by AT100 IR as compared to IgG-depleted seeds (p = 0.055). Graph represents mean ± SEM (n = 6–7 animals/group). ∗∗p ≤ 0.01, one-way ANOVA with Dunnett’s multiple comparisons analysis against IgG condition.

Discussion

In this study, we built a quantitative, physiologically relevant primary neuron seeding assay for tau pathobiology research and systematically defined individual components of the model. We chose neurons prepared from hTau mice as our cell model. Notably, these neurons 1) express wild-type human tau protein without contaminating murine tau, and 2) are derived from the hippocampus and cortex, two primary brain regions affected in AD. Together, these attributes make hTau neurons a suitable biological system for investigating tau protein expression, function, and aggregation pathology.

An important component of the hTau neuronal assay is the usage of AD brain-derived tau seeds. Multiple studies have reported that in AD, and when prepared in vitro with polyanions such as heparin, tau protein misfolds into conformationally distinct strains of aggregates (34, 35, 36). Further, recombinant fibrils have shown to be ineffective in seeding wild type tau expressed at physiological level in mouse model (36). Therefore, we created a neuronal assay where sporadic AD disease-relevant tau filaments are formed via a seeding mechanism. We demonstrated that two types of AD brain derived tau seeds can robustly induce tau aggregation in hTau neurons. The first type of seeds, lysate tau, is a total collection of tau protein encompassing both aggregated soluble oligomers and insoluble high molecular weight filaments. The EC50 of this seeding material is in the high nM range because tau is just one of many proteins present in the lysate. In contrast the second type of seeds, SP tau, was purified to contain mostly high molecular weight insoluble tau filaments and therefore, has a much more lower seeding EC50 in the 0.03 nM range. We propose both types of seeds are suitable for studying tau pathogenesis in AD. The choice of template material used within the hTau seeding assay should be based on experimental demands. AD lysate material is most suitable for assessment of unknown tau aggregate species, whereas SP tau provides specificity against more mature tau filaments.

Endogenous tau aggregates in neurons induced by either AD lysate or SP tau seeds contain features reminiscent of those found in AD tau pathology. For example, in both AD lysate and SP tau seeding, the newly formed tau aggregates in the soma (Fig. 2) bear similar filamentous morphology as tau pathology drawings from Alois Alzheimer’s diagnosis (37). The aggregates in the hTau neurons are also hyperphosphorylated at key epitopes found in tau tangles (Fig. 2D). Importantly, these seed-induced aggregates can propagate aggregation across synapses, a key mechanism hypothesized to drive tau pathology spread in AD (Fig. S3). Like tau pathology progression in AD, both aggregation and propagation of tau aggregates in the hTau neuronal assay are time-dependent. We observed tau aggregates rapidly accumulate between 7 to 14 days, whereas propagation took over 9 weeks to manifest. In addition to recapitulating known tau aggregation phenotypes, we found that prolonged tau aggregation in hTau neurons induced mitochondrial functional deficits, consistent with observed mitochondrial impairment in AD. Mitochondrial maximal respiration and ATP production, as measured by Seahorse, were impaired in hTau neurons containing high levels of tau aggregates, but not in control-seeded KO neurons cultured for the same extended time (Fig. 3). Taken together, these data provide evidence demonstrating the disease physiological relevance of the hTau neuronal model.

When comparing the present model to previously described neuronal tau aggregation models, the hTau model’s sporadic AD relevance and broad applicability are the most distinctive attributes. For example, a seeding-based neuronal model specifically developed for identifying tau-targeted therapies required both the introduction of the human 2N4R isoform with a P301S mutation and the use of tau seeds extracted from a P301S transgenic mouse brain, which limits its relevance when sporadic AD is considered (38). The versatility of the hTau model is also an important aspect since it can be applied either to small-molecule tau aggregation inhibitors or anti-tau antibodies. In contrast, this versatility is not found in other rodent neuronal primary cultures, where aggregates of rodent tau might not be recognized by anti-human tau antibodies (39, 40). The hTau model also contrasts with our human induced pluripotent stem cell (hiPSC)-derived neuron models of tau aggregation (41) by enabling a shorter culture length (2 weeks for hTau neurons versus 4 weeks for hiPSC-neurons) and a lower seed quantity (0.1 nM for hTau neurons versus 0.5 nM for hiPSC-neurons). Finally, one limitation of the present model could be the exclusive expression of the 0N3R tau isoform that does not recapitulate the equal 3R:4R tau isoform ratio inherent to the human adult brain and would not be suitable for studies investigating the 4R tau isoform specifically. However, we have shown with our iPSC-neuron model that AD-brain-derived seeds induce 4-fold more aggregation in 0N3R expressing MAPT-WT neurons than in neurons expressing both 3R and 4R isoforms (41, 42, 43). Therefore, trying to gain in physiological relevance might be disadvantageous when a robust and quicker tau aggregation readout is required.

As an application example, we demonstrated that the hTau neuronal assay is an effective assay to differentiate tau antibodies for tau immunotherapy development. Tau immunotherapy is based on the hypothesis that anti-tau antibodies bind to specific epitopes unique to tau seeds, and in doing so, can prevent neuronal uptake, tau seeding and propagation in AD. To date, several immunotherapies targeting extracellular tau species were unsuccessful in reducing tau pathology progression (44, 45). A possible reason is that the therapeutic antibodies did not specifically target the seeding competent species responsible for pathological spread. We used the hTau seeding model to evaluate two tau antibodies, MC1 and PHF1 based on their disease relevance and clinical implications. Both are antibodies raised against AD brain derived PHFs and not against recombinant tau protein and therefore could be more disease relevant. The humanized MC1 zagotenemab was evaluated and failed in a phase 2 clinical trial. Concerning PHF1, several immunotherapies targeting the same epitope as PHF1 have also reached the clinical stage, such as LuAF87908, a humanized IgG1 antibody that has completed phase 1, and the anti-tau active immunotherapy ACI-35.030 (also known as JNJ-2056) that is currently in a phase 2 clinical trial for preclinical AD. No efficacy data in AD patients have yet been reported for these last two programs (46). In the hTau seeding model, of the two antibodies, we found PHF1 more efficiently immunodepletes tau seeding in cell and animal seeding studies, suggesting that the pS396/pS404 epitope recognized by PHF1 is more present or accessible in AD seeding-competent tau species than the folded N-terminal and central domain complex epitope recognized by MC1. It is interesting that even at 300 nM (3:1 M ratio), PHF1 does not completely remove all seeds from competent tau. This implies that while pS396/pS404 is an epitope highly presented in pathological tau species, there are species of tau seeds that do not contain this epitope. Additional follow up evaluation of different tau antibodies targeting other phosphorylation sites or conformations of tau aggregates would be required to further dissect out seeding-competent pathological tau species. The current example study effectively illustrates the hTau seeding assay’s application in characterizing and differentiating tau antibody’s function to block tau aggregation.

In addition to drug development, the hTau seeding assay also provides a platform to study tau pathobiology, such as the mechanism of selective vulnerability. Many studies have shown that in AD, tau pathology only accumulates and results in neurodegeneration of a subset of vulnerable neurons (47). These neurons are unique either by their spatial location (in vulnerable regions such as the entorhinal cortex) or are associated with specific molecular markers (47). In the hTau seeding assay, we notice that even at the late stage of aggregation, where high levels of tau aggregates have formed, only a subset of neurons within the culture develop aggregates, as evidenced by the buildup of tau filaments in the cell body (Figs. 3 and S2). This phenotype closely resembles the selective vulnerability observation in AD pathology. Novel technology such as CITE-seq (cellular indexing of transcriptomes and epitopes) could be used to label and sequence transcriptomics of tau aggregates-bearing neurons to elucidate unique molecular markers fingerprinting these vulnerable neurons to gain a deeper understanding of tau disease pathogenesis (48).

The presented hTau neuronal seeding assay has two technical limitations. First, this model is currently limited to assessing tau aggregation in a seeding assay, as tau propagation takes 9 weeks to develop in the microfluidic co-culture model (Fig. S3). While this phenotype highlights that tau pathology spreading occurs in later decades in AD and resonates with the hypothesis that aging is the highest risk factor for the disease, retaining good viability for the extensive culture duration for non-replicating post-mitotic primary neurons is challenging. The sensitive environment of microfluidics, created by multiple constraints such as small growth surfaces, shear force due to fluidic movement, and hydrophobic properties of PDMS material fabrication, adds further technical challenges for maintaining neuronal viability over an extensive period. Indeed, although we were to capture the propagation phenotype in our study, we were not able to quantitatively evaluate the tau antibody’s property to block propagation. Future studies establishing novel microfluidic technology would be necessary to enable tau propagation-related interrogations. Second, the hTau neuronal seeding assay is limited by the presence of non-neuronal CNS cells in the culture introduced during primary neuron isolation from regions of the murine brain. Since the establishment of primary neuron culture (49), efforts have been made to reduce non-neuronal cells, such as peeling away meninges during dissection to exclude endothelial and fibroblast cells and the use of serum-free culture media to reduce the growth of proliferating cells. These conditions reduce but do not eliminate the presence of non-neuronal cells in the culture. Lastly, although not a technical limitation, the hTau seeding assay does take over 3 weeks in experiment and analysis to complete, and therefore, may not be suitable for time-sensitive studies such as diagnostic tests.

In summary, we built a robust primary neuronal tau seeding model that recapitulates sporadic AD-relevant tau aggregation and associated functional deficit phenotypes. The physiological relevance of the model, coupled with its automated, sensitive quantitative capability, makes it a powerful assay for tau drug development research and pathobiology studies.

Experimental procedures

Human brain tissues

Frozen human donor brain tissues diagnosed with AD neuropathology or normal without pathology were purchased from Folio Biosciences and Banner Sun Health Research Institute. One brain was donated from Mass General Hospital. All tissues were approved for the study. Detailed patient demographic information is listed in Supporting Table S1.

Declaration of Helsinki principles: All human studies reported in this work abide by the Declaration of Helsinki principles.

Animals

Protocols and procedures used in these studies were approved by the AbbVie Institutional Animal Care and Use Committee (IACUC). Three transgenic mouse models were used for these studies. For hTau neuronal culture, transgenic mice expressing human tau transgene (MAPT) via a PAC cloning vector, driven by the tau promoter (8c) (strain: C57blk6.Cg-Mapt (tm1 eGFP Klt)tg(MAPT)8cdav/J) were crossed with tau knock-out mice that have cDNA of green enhanced fluorescent protein (EGFP) inserted into exon one of MAPT (strain: B6. Cg-Mapt (tm1 eGFP Klt)). The resulting mice will either express human MAPT but do not express mouse MAPT (hTau line), or conversely, do not express mouse MAPT (KO line) (22). Mice were obtained through license with Albert Einstein College of Medicine, Bronx New York USA. For direct comparison of AD lysate and SP seeds in vivo, P301S transgenic mouse expressing human MAPT carrying P301S mutation driven by the mouse prion promoter encoding 4R1N (strain: B6;C3-Tg(Prnp-MAPT∗P301S)PS19Vle/J) (24) were licensed through agreement with MRC Laboratory of Molecular Biology. To characterize tau antibody’s efficacy in immunodepleting seeding competent tau, rTg4510 expressing human tau with four microtubule-binding domain repeats (0N4R) and the P301L mutation under the control of mouse calcium-calmodulin kinase II–driven tetracycline-controlled transcriptional activator (Camk2a-tTA) were used (strain: 129S6.Cg-Tg(Camk2a-tTA)1Mmay/JlwsJ; Fgf14^{Tg(tetO-MAPT∗P301L)4510Kha}/J) (33). This mouse line was obtained under licensing agreement with Mayo Clinic.

Brain tissue lysate preparation

Randomized frozen human brain tissues were homogenized in PBS buffer (supplemented with protease and phosphatase inhibitors) at 5 μl/mg on a Precellys tissue homogenizer (7 ml CK mix bottles, 6600 rpm, 2 × 20 s/cycle, 6 cycles). Brain lysates were collected from supernatant fraction of spun homogenates (3000g, 5 min, 4 °C) in low-binding microcentrifuge tubes (Costar). Total protein was measured using the micro-BCA kit. Brain lysates were stored at −80 °C until usage.

Tau pathology measurement by HTRF

The Homogeneous time-resolved fluorescence (HTRF) assay was used to quantify the brain lysate’s tau pathology levels. This assay specifically quantifies aggregated tau species using donor and receptor FRET fluorophores conjugated to the same tau antibody (50). Assay was conducted according to the manufacturer’s protocol and within the assay’s dynamic range. Briefly, brain lysate samples were incubated with fluorophore-tagged antibodies (in 384-microplate, 24 h, 4 °C) and the fluorescent signal was captured on Cytation 5 reader (excitation 340 nm, emission at 620 nm and 665 nm). Tau aggregation level represented by the relative FRET signal was calculated as the ratio between the two fluorescence signals (665/620). Data were plotted in GraphPad Prism.

Tota tau measurement by MSD

Total tau levels of brain lysates and SP seeds were measured using the V-PLEX Human total tau sandwich immunoassay (Meso Scale Discovery). The assay was carried out according to the manufacturer’s protocol. Samples were prepared by first diluting (1:5, latter being 1 × 90% Laemmli 10% β-mercaptoethanol buffer), denaturing (95 °C, 5 min), and diluting again in diluent 35 buffer (1:3000). Prepared samples were added to diluent-35 pre-blocked and precoated plates (50 μl, 1 h, 25 °C), washed (0.05% PBST, 3×), incubated with 1× SULFO-tag antibody (25 μl, 1 h, 25 °C), washed (0.05% PBST, 3×), and after incubating with reading buffer (7 min, 25 °C), read on Meso Sector S600 plate reader. Tau calibrator was used as a standard for tau concentration calculation (3× serial dilution starting from 28,800 pg/ml). Blank wells were included to set the baseline background signal. Samples were measured in duplicates.

Sarkosyl insoluble tau seed extraction from AD brains

Sarksoyl insoluble tau seeds were isolated from high tau-pathology confirmed brain tissues as previously described (19)) with modifications. The workflow is detailed in the schematic diagram of Figure 1C. Briefly, randomized brain tissues were homogenized at TBS buffer (50 mM Tris, 150 mM NaCl, 20 mM NaF, 1 mM Ma3VO4, 0.5 mM MgSO4 supplemented with protease and phosphatase inhibitors) at 5 μl/mg on a Precellys tissue homogenizer (7 ml CK mix bottles, 6600 rpm, 2 x 20 s/cycle, 6 cycles). Lysates were centrifuged (27,000g, 20 min, 4 °C) to obtain supernatant (S1) while pellet (P1) was sonicated in salt/sucrose buffer (0.8 M NaCl, 10% Sucrose, 10 mM Tris/HCl, 1 mM EGTA, pH 7.4 supplemented protease and phosphatase inhibitors) and supernatant (S2) was collected after centrifugation (27,000g, 20 min, 4 °C). The combined S1 and S2 fractions were incubated in final 1% sarkosyl solution (made in same salt/sucrose buffer as above, 1.5 h, 25 °C). Incubated samples were then centrifuged (250,000g, 1.5 h, 4 °C). Resulting sarkosyl insoluble tau pellet was resuspended (TBS buffer), sonicated (30 s, qSonica Q125 model), and centrifuged (100,000g, 1 h, 4 °C). The pellet was resuspended in sterile 1× PBS and centrifuged (10,000g, 30 min, 4 °C) to obtain a supernatant that contains insoluble tau filaments. Enriched AD SP was further characterized by MSD and Western blot to confirm quality of tau seeds.

Immunoblot

Immunoblot to examine tau isoforms was prepared as previously described (5) Briefly, normalized samples (20–30 μg, 2 mg/ml protein) were dephosphorylated using lambda protein phosphatase (NEB, 3 h, 37 °C), mixed with 4× LDS sample buffer, and boiled for 5 min before separation on NuPAGE Bis-Tris gels (4–12%) using MOPS buffer (Invitrogen NP0001) with antioxidant (Invitrogen NP0005) on ice. To achieve maximum separation of tau isoforms, electrophoresis was conducted for 2.5 h at 100V. Precision Plus Protein Dual Color ladder (Thermo Scientific 1610374) and recombinant tau isoform ladder (Sigma) were used as references for tau protein molecular weight and isoforms. Proteins were transferred onto a PDVF membrane using the iBlot dry blotting system. Membranes were blocked in Odyssey blocking buffer (P/N927-40000, 2 h, 25 °C) before incubation with primary antibodies (CP27, 1:500, RD3, 1:500, RD4, 1:500, TauC, 1:2000, overnight, 4 °C). The following day, membranes were washed (3× TBST, 5 min), incubated with respective IRDye 680- and 800-conjugated donkey anti-rabbit, anti-mouse secondary antibodies (LiCor Biosciences) for 1 h and visualized using the Odyssey CLX imaging system (LiCor Biosciences).

Neuronal culture and tau seeding

Primary hippocampal and cortical neurons were prepared accordingly to a previously published protocol with some key modifications (5, 49). Importantly, hippocampus and cortex were dissected out from all embryos and maintained in special Hibernate-E media (Invitrogen) on ice during genotyping. Genotyping using tail tissues from embryos by qPCR was performed to separate hTau embryos from KO embryos. Two sets of probes were used for genotyping and are included in Supporting Table S3. This procedure takes approximately 2 h. Once the genotype was confirmed, hippocampus and cortices from either KO or hTau embryos were dissociated using papain enzyme (Neuronal Isolation Enzyme dissolved in 1× HBSS buffer free of Ca/Mg, 30 min, 30 °C). Tubes containing the digesting tubes were inverted every 5 min to ensure complete dissociation. Dissociated tissues were then washed and triturated to obtain single neurons. Total neuron number and viability were determined using Countess II Cell Counter (source of equipment). Dissociated neurons were plated in plating media (DMEM supplemented with 10% B27, 10% L-glutamine, 10% FBS, Invitrogen) at 35,000 cells/well on poly-d-lysine-coated 96-well plates (Corning). Three hours post-plating, the media was switched to serum-free neuronal culture media (Neurobasal Plus Media supplemented with 10% B27, 5% GlutMax, Invitrogen) for culturing. For seeding and propagation studies, tau seeds were added to and incubated with neurons (DIV 3) for 48 h and then washed off by complete media change.

Microfluidic culture and seeding

For propagation studies in microfluidic cultures, polydimethylsiloxane (PDMS) microfluidic triple chambers (Xona) were treated with plasma (75% power, 1 min, Diener plasma generator, Thierry Plasma) and bonded immediately to glass-bottom 6-well plates (Cellvis). Water was added to bonded chips to retain hydrophilicity at 6 min post-bonding. Bonded microfluidic plates were coated with PLO (0.01% poly-l-ornithine, overnight, 4 °C) and laminin (2 μg/ml, 25 °C, 2 h). Neurons were plated at high density (9 million/ml, 5 μl). Three hours post-plating, neuron culture media were added (20 μl/reservoir). In the case with seeding, P2 and P3 population neurons are maintained at a higher volume (45 μl) to create a hydrostatic pressure that restricts tau seeds in the P1 neuronal population for 2 days. Seeding was performed using AD SP tau (5 nM).

Immunofluorescences

Treated neurons were fixed with cold methanol at −20 °C for 15 min to remove soluble tau protein for immunofluorescent staining of tau aggregates as previously published (19). After fixation, cells were re-hydrated in 1× PBS, blocked in 5% BSA (1 h, 25 °C), and incubated in primary antibody cocktails (overnight, 4 °C). Antibodies used in this study are listed in Supporting Table S1. Fluorescent-conjugated secondary antisera mixtures containing Alexa 488 IgG, Alexa 568, and Alexa 647 IgG (anti-mouse, anti-rabbit, anti-chicken) were used (1 h, 25 °C, performed in dark protected from light). Nucleus were stained with Hoechst 33342 dye (1:1000 in PBS, 5 min, 4 °C). Controls were done where exclusion of one, or the other primary antisera to confirm specificity of the secondary antibodies.

Automated high-content confocal microscopy and quantitative analysis

Multiphoton images were acquired using the high-content PerkinElmer Opera Phenix confocal microscope with a 40× water objective. Images were captured at 8 planes with 1 μm/field, 18 fields/well for quantitative analysis. Images shown are pseudo-colored and presented as maximal projection encompassing all planes. All IF images were analyzed using algorithms available in Harmony Software (PerkinElmer). Tau aggregates were identified using either “Find Spots” or “Find Images” methods. The threshold was set to exclude background fluorescent noise (no seeds treated negative control condition) to select “real tau aggregates”. Quantitative features collected for tau aggregates include area (μm²), total intensity, and total number are used to represent tau aggregation level. Neurites (MAP2) are identified by “Neurite Analysis” in Harmony and represented as total area (μm²). Quantification algorithms were held constant for the analysis of all conditions within the same study.

Seahorse respirometry in primary neurons

Primary neurons (hTau or KO) were plated on Seahorse plates at 35,000 cells/well. AD SP at 0.1 nM was added to cells on DIV5, and cells were analyzed by Seahorse on DIV13, DIV19 and DIV26. To ensure no plate-related impact on assay readout, hTau and KO neurons were plated on alternating columns of the same plate with AD SP seeds addition in alternating rows. On the day of analysis, the number of cells was counted using brightfield imaging (ImageXpress Micro Confocal High-Content Imaging System) to normalize oxygen consumption rates (OCR) measured using Seahorse methodology. Mitochondrial OCRs were measured using the XF Cell Mito Stress Test kit (Agilent, 103015-100) following the manufacturer’s instructions. Briefly, on the day of analysis, cells were washed 2 × with and incubated in Mito Stress Test Assay Medium (1 mM pyruvate, 2 mM glutamine, and 10 mM glucose in XF base medium) in a 37 °C non-CO2 incubator for 1 h. OCRs were measured with an Agilent Seahorse XFe96 analyzer at baseline, after addition of 1.5 μM oligomycin to evaluate respiration associated with cellular ATP production, after addition of 1 μM FCCP to evaluate maximal respiration rate, and after addition of 0.5 μM antimycin/rotenone to measure non-mitochondrial oxygen consumption rate. OCR were then normalized to the number of cells counted using brightfield imaging employing ImageXpress Micro Confocal High-Content Imaging System.

AD lysate seed immunodepletion with tau antibodies

Immunodepletion of AD lysate seeds with tau antibodies was performed using protein G magnetic beads (ThermoFisher) according to manufacturer’s protocol with modifications. Beads were coated with either tau antibody or isotype control IgG (100 nM or desired test concentrations for dose-response, 30 min rotation on Hula mixer, 25 °C) in 0.05% TBST buffer. Samples were then centrifuged (pulse-spin), placed on the magnetic bead stand (1 min) to separate out the supernatant from the now antibody-coated bead pellet. AD lysates (100 nM) were added to the coated beads, mixed by vortex, incubated (1.5 h, 25 °C), spun (pulse-spin) and placed on the magnetic bead stand. Supernatant containing antibody-immunodepleted AD lysates was collected and used for in vitro and in vivo seeding studies.

In vivo tau seeding stereotaxic surgery

Male and female rTg4510 mice (8–9 weeks of age) or P301S mice (7–8 weeks of age) were treated with the analgesic compound, meloxicam, at least 1 h before surgery. Mice were then anesthetized with 3% isoflurane in an induction chamber. Once anesthetized, the fur on the top of the head was shaved, and the mice were placed in the stereotaxic frame fitted with a nosecone to allow for continuous exposure to 3% isoflurane during surgery. Once placed in the frame, the top of the head was cleaned, alternating three times with alcohol and betadine, followed by a topical analgesic (e.g., lidocaine) along the incision site. Using a scalpel, a small incision was made through the skin and muscle on the top of the head. A small hole was drilled at the appropriate location (A/P = −1.9 and M/L = −1.4). A Hamilton syringe containing the AD lysate was lowered into the brain to the appropriate depth (D/V = −1.6). After the syringe was left to settle, 2.5 μl AD lysate was infused at a rate of 0.2 μl/min. When the infusion was complete, the syringe remained in place for an additional 2 min and was then removed slowly. The incision was closed with suture, and the mice were allowed to fully recover from anesthesia in an animal incubator before being returned to their home cage.

Immunohistochemistry for tau pathology analysis

At the appropriate time point, mice were euthanized with sodium pentobarbital, perfused with PBS, and whole brains were drop-fixed in 10% formalin for 48 h before being switched to 70% ethanol. Brains were stored in 70% ethanol until all brains from the study were collected to allow all samples to be processed for paraffin embedding at the same time. Before processing, brains were trimmed into three coronal slices to remove the frontal cortex and the brainstem/cerebellum using a stainless-steel coronal mouse brain matrix (Harvard Apparatus). The 5 mm slice in the middle that contains the hippocampus was processed (ASP300, Leica) for paraffin embedding. For all studies, 3 to 4 brains from different treatment groups were randomized and embedded (Sakura Finetek) into paraffin blocks. Coronal plane and every fifth section (5 μm) were collected through the entire hippocampus and mounted onto glass slides.

Select slides through the entire hippocampus were stained for AT100 immunoreactivity using the BOND RX stainer (Leica) with the Refine Polymer Detection kit (Leica). In brief, the slides were deparaffinized, rehydrated, and underwent epitope retrieval 1 (Leica), and a series of blocking reagents, including peroxide block and 5% donkey serum blocking (Jackson Immunoresearch Labs, Leica), and mouse on mouse blocking (VectorLabs). Slides were then incubated overnight in the mouse monoclonal antibody to AT100 tau (MN1060, Thermo Fisher) at a concentration of 0.036 μg/ml. After incubation in primary antibody, slides underwent a series of washes before being incubated in Biotin-SP-conjugated F(ab')2 donkey anti-mouse IgG(H + L) secondary antibody (715-006-151, Jackson ImmunoResearch Labs, West Grove, PA, USA) at a concentration of 2 μg/ml for 20 min. After additional washing steps, the slides were incubated for 15 min in Streptavidin/Horseradish Peroxidase (RE7104, Leica) and the immunoreactivity was visualized using diaminobenzidine (DAB; Leica) and counterstained with hematoxylin (Leica).

Immunohistochemistry quantification

Immunostained slides were scanned with a Panoramic 250 Flash III scanner (3D Histotech), and matched sections were analyzed for area covered by immunoreactivity in the region of interest (ROI) using the Area Quantification module in HALO image analysis software (Indica Labs). For both in vivo studies, 3 matched sections (approximately 125 μm apart) through the CA1 hippocampus were analyzed per brain. Using the software, the "threshold" within the ROI was determined by an observer who was blind to the treatment of the animals. Once an appropriate threshold was set, the software measured the percentage of the area of interest (CA1 hippocampus) containing positive immunoreactivity (AT100). The percentages for the 3 sections analyzed were then averaged to obtain a single value for each animal.

Statistics

All statistics were performed in GraphPad Prism 10.3.1 software. Statistical analysis comparing multiple groups was performed using a one-way ANOVA with Dunnett’s multiple comparisons or a two-way ANOVA with Sidak’s multiple comparisons against the isotype IgG control condition as indicated in figure legends. Data shown in graphs are mean ± SEM unless otherwise indicated. Significance is defined as: ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗∗p ≤ 0.001. n represents an independent replicate of either a well or an individual animal.

Data availability

Data supporting findings in this study are included in main figures or in supporting materials. Additional data will be shared upon request to corresponding author. No data is deposited to databases.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflict of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Elizabeth J. Coulson

Supporting information

Figure S1.

Neurons isolated from hTau mice express human tau protein and not murine tau. Neurons isolated from hTau mice express GFP (green). In contrast, neurons isolated from wild-type CD-1 mice do not express GFP, Likewise, expression of human tau protein (HT7, white) is observed exclusively in hTau neurons whie mouse tau protein (T49, white) is observed only in CD1 neurons. Both CD1 and hTau neurons stain positive for a pan-tau protein marker (Tau5, white).

Figure S2.

Cultured hTau neurons (DIV 14) exhibit a comnplex neurite network expression fo mature neuronal markers. Neurons have dense neurites (MAP2, green) express human tau protein (CP27, red), and markers consistent with a deep layer cortical identity (Satb2, green; Ctip2, red). Neurons represent a mixture of both excitatory (VGLUT1, green) and inhibitory (VGAT, yellow) neurons, Scale bar: 100 μm.

Figure S3.

Seed-induced tau aggregation in hTau neurons propagates from neuron to neuron. (A) Schematic illustration of neuronal co-culture in compartmentalized microfluidic setup. Six individual chips are mounted in one cultured plate. Three populations of neurons (yellow, blue, yellow) are co-cultured into 3 compartments interconnected by microgrooves. In propagation assay, only P1 neurons are seeded with AD tau seeds. Higher culture volume in P2 and P3 creates hydrostatic pressure preventing diffusion of tau seeds. (B) Representative IF images showing overview of tau propagatin into P1 and P2 neurons. CP27, green; MAP2, magenta; DAPI, blue. (C) Inset images of neurons developed tau aggregation due to seeding (P1), propagation across first (2, 3), and second (4, 5) synapses.