Abstract
Neurofibrillary tangles, formed of misfolded, hyperphosphorylated tau protein, are a pathological hallmark of several neurodegenerations, including Alzheimer's disease. Tau pathology spreads between neurons and propagates misfolding in a prion-like manner throughout connected neuronal circuits. Tauopathy is accompanied by significant neuronal death, but the relationships between initial tau misfolding, propagation across connected neurons and cytotoxicity remain unclear. In particular the immediate functional consequence of tau misfolding for the individual neuron is not well understood. Here, using microfluidic devices to recreate discretely organized neuronal connections, we show that the spread and propagation of misfolded tau between individual murine neurons is rapid and efficient; it occurs within days. The neurons containing and propagating tau pathology display selective axonal transport deficits but remain viable and electrically competent. Therefore, we demonstrate that seed-competent misfolded tau species do not acutely cause cell death, but instead initiate discrete cellular dysfunctions.
SIGNIFICANCE STATEMENT Public awareness of progressive neurodegenerations such as dementias associated with aging or repetitive head trauma is rising. Protein misfolding underlies many neurodegenerative diseases including tauopathies, where the misfolded tau protein propagates pathology through connected brain circuits in a prion-like manner. Clinically, these diseases progress over the course of years. Here we show that the underlying protein misfolding propagates rapidly between individual neurons. Presence of misfolded tau is not directly cytotoxic to the neuron; the cells remain viable with limited deficits. This suggests that neurons with tau pathology could be rescued with a therapeutic disease modifier and highlights an under-appreciated time window for such therapeutic intervention.
Introduction
Neurofibrillary tangles (NFTs), a pathological hallmark of several neurodegenerations including Alzheimer's disease (AD), consist of structured insoluble aggregates of hyperphosphorylated tau (Grundke-Iqbal et al., 1986). The neuroanatomical localization of NFTs in AD brains suggests that tau pathology propagates through the brain along anterograde connected circuits (de Calignon et al., 2012; Ahmed et al., 2014). Indeed, it is well established that pathogenic tau spreads between cells in vitro (Frost et al., 2009; Sanders et al., 2014) and in vivo (Lasagna-Reeves et al., 2012; Liu et al., 2012; de Calignon et al., 2012) in a prion-like manner (Clavaguera et al., 2009; Frost et al., 2009; Kfoury et al., 2012; Kaufman et al., 2016). Tau oligomers of hyperphosphorylated tau have been isolated from patient brains (Köpke et al., 1993) and cause hyperphosphorylation and misfolding of native tau (Alonso et al., 1996; Li et al., 2007). The tau seeds that template the misfolded conformation to native tau are monomers or lower molecular weight oligomers (Michel et al., 2014; Falcon et al., 2015; Kim et al., 2015; Mirbaha et al., 2015; Jackson et al., 2016; Sharma et al., 2018), and as tau polymerizes into filaments it loses seeding activity (Alonso et al., 2006).
Pathogenic tau causes cell death both in vivo (de Calignon et al., 2012) and in vitro (Gómez-Ramos et al., 2006; Tian et al., 2013), and advancement of AD is associated with extensive neuronal loss (Braak and Braak, 1991). Exogenous addition of tau is toxic to neurons in vitro (Gómez-Ramos et al., 2006; Kopeikina et al., 2012; Tian et al., 2013), suggesting that tau misfolding and aggregation is associated with neuronal death. Therefore, it has been suggested that tau seeds are released following the disintegration of the tangle-bearing neurons (Guo and Lee, 2011; Hu et al., 2016). This is consistent with in vivo observations where loss of neurons is progressive within brain areas affected by degeneration. Interestingly, recent studies have detected tau species capable of seeding misfolding in human brain areas free of tangle pathology (DeVos et al., 2018), implying that tau seed release occurs from intact neurons and precedes neuronal death (Pickett et al., 2017). By their nature these in vivo studies are short of the resolution required to identify the individual neurons bearing and releasing misfolded tau species, and as a consequence the physiological state of these neurons remains unclear.
To propagate pathology in a prion-like manner tau seeds spread to connected neurons and interact with native tau in the cytosol to template its misfolding. Studies using exogenous tau preparations to investigate the mechanisms underlying tau pathology transmission have shown that aggregates are internalized into primary neurons, trafficked both anterogradely and retrogradely along axons, spread to connected cells (J. W. Wu et al., 2013; Wu et al., 2016; Takeda et al., 2015; Wang et al., 2017), and propagate tau pathology (Calafate et al., 2015; Wu et al., 2016; Nobuhara et al., 2017). However, the efficiency of propagation of tau misfolding between individual neurons and the consequence for the individual neuron's physiology have not been resolved.
In this study, we created a minimalistic neuronal circuit within a compartmentalized microfluidic device to investigate tau misfolding and propagation with single cell resolution. We show that a phosphomimetic tau, tauE14, in which 14 disease-relevant serine/threonine residues have been mutated to glutamate to mimic phosphorylation (Hoover et al., 2010), misfolds within primary neurons in the absence of exogenous seeds. Misfolded tauE14 seeds template a rapid and efficient prion-like misfolding of native tau and transmit the conformational change of tau between intact, connected neurons with high efficiency. This suggests that propagation of misfolded tau occurs between live, functioning neurons in very early stages before neuronal degeneration. Our findings imply that propagation of misfolded tau through the brain likely precedes detectable symptoms, strengthening the idea that targeting the spread of misfolded tau in as yet unaffected areas may present a disease modifying approach for mild cognitive impairment.
Materials and Methods
Plasmids.
The following plasmids were used: pRFP-N1, pEGFP-C3 (Clontech); pRK5-EGFP-tauWT and pRK5-EGFP-tauE14 were a gift from Karen Ashe (Hoover et al., 2010; Addgene, plasmids #46904 and #46907); GCaMP6 was a gift from Douglas Kim (Chen et al., 2013; Addgene, plasmid #40753); and R-GECO was a gift from Robert Campbell (J. Wu et al., 2013; Addgene, plasmid #45494). RFP-tauWT and RFP-tauE14 were created by excising the GFP fragment of pRK5-EGFP-TauWT at Cla1 and BamH1 sites, and replacing it with RFP, which was amplified by PCR with forward primer 5′-CATGATCGATATGGCCTCCTCC-3′ and reverse primer 5′-CATGGGATCCGGCGCCGGT-3′.
Cell culture and transfection.
All experiments were performed in accordance with the Animals (Scientific Procedures) Act 1986 set out by the UK Home Office. Primary cultures were prepared as described previously (Deinhardt et al., 2011) from embryonic day (E)15–E18 C57BL/6 mouse hippocampus. Dissociated neurons were plated in neurobasal medium supplemented with 2% B27 and 0.5 mm GlutaMAX (Invitrogen) at a density of 7000 cells/μl in microfluidic devices and 150,000 cells/ml in glass-bottom dishes. Partial medium changes were performed on the devices every 2–3 d. At 1 DIV, neurons were transfected using Lipofectamine 2000 as described previously (Deinhardt et al., 2011). Transfection mix was added to device channels sequentially, and channels were fluidically isolated from each other through a volume difference to ensure no diffusion of the solution across channels (Dinh et al., 2013).
Microfluidic devices.
Custom microfluidic devices were manufactured based on existing designs (Taylor et al., 2003; Peyrin et al., 2011). Devices were replicated as described previously (Holloway et al., 2019), washed in 70% EtOH for 1 h and dried before use. Devices were mounted onto 22 × 55 mm coverslips (Smith Scientific) pretreated with 0.1 mg/ml poly-d-lysine (Sigma-Aldrich). Device channels were filled with supplemented neurobasal medium and incubated overnight before addition of cells.
Immunocytochemistry.
Neurons were fixed in 4% paraformaldehyde in PBS for 10–15 min, washed with 50 mm ammonium chloride in TBS for 5 min, and permeabilized in 0.1% Triton X-100 in TBS for 5 min at room temperature (RT). The cells were then blocked for 30 min in 10% goat serum in TBS at RT. Cells were incubated for 1 h at RT or overnight at 4°C in the following primary antibodies: MC1 [1:300; gift from Peter Davies (Jicha et al., 1997)], synapsin-1 (D12G5, 1:1000; Cell Signaling Technology). Primary antibody was washed off in TBS, and Hoechst (33342) was added to the second wash to stain nuclei (1:3000; ThermoFisher Scientific). This was followed by incubation for 30 min at RT with fluorescently-conjugated secondary antibodies (Invitrogen).
Microscopy.
Fixed cell images for axonal length analysis were taken on a Zeiss Axioplan Fluorescence Microscope equipped with a HBO103 Mercury lamp for illumination, a QImaging Retiga 3000 monochrome CCD camera (Photometrics), 20×/0.4 NA and 40×/0.75 NA Plan-Neofluar objectives, using Micro-manager software (Vale Lab). Fluorescent and differential interference contrast (DIC) images of cells in devices were obtained using a 60×/1.42 NA Oil Plan APO objective on a DeltaVision Elite system (GE Life Sciences) with SSI 7-band LED for illumination and a monochrome sCMOS camera, using SoftWoRks software v6. Confocal images were taken on a Leica SP8 laser scanning confocal microscope using a 63×/1.30 NA HC Pl Apo CS2 glycerol immersion objective, with a PCO Edge 5.5 sCMOS camera. Lasers used for illumination were continuous wave solid-state lasers at 405 and 561 nm, and a continuous wave argon gas laser at 488 nm.
Live cell imaging was performed on the DeltaVision Elite system (GE Life Sciences). For imaging of lysosomes, 14 DIV neurons were incubated with 25 nm LysoTracker Deep Red (ThermoFisher Scientific) for 20 min at 37°C. LysoTracker solution was then removed and replaced with supplemented neurobasal medium containing 50 mm HEPES-NaOH, pH 7.4. Images were taken at 0.2 Hz for 5 min. For calcium imaging, neurons were cotransfected with GCaMP6 and RFP-TauE14 or RFP-TauWT, or R-GECO and GFP-TauE14 or GFP-TauWT and imaged at 14 d in vitro (DIV) at 2 Hz for 6 min. After 3 min, 1 μm tetrodotoxin (Sigma-Aldrich) was added, followed after 2 min by 100 mm KCl.
Image analysis.
Overview images were reconstructed from multiple single images using Autostitch software (University of British Columbia). Images were analyzed using ImageJ software (NIH), and its plugins NeuronJ (Meijering et al., 2004) and Iterative Deconvolution (Bob Dougherty, OptiNav Inc). Axonal length was defined as the longest axonal branch from the longest neurite. Distal axon was defined as a 75 μm stretch of axon at, or near to, the terminal of an axon branch, and proximal axon was defined as a 75 μm long stretch of axon measured beyond the first 50 μm of axon protruding from the cell body, therefore, beyond the axonal initial segment. Intensity profiles along a line were generated using plot profile, and kymographs using the MultipleKymograph plugin on ImageJ. Lysosomes which displaced >50 μm over the course of the time lapse were considered moving. Aggregate analysis was performed using MATLAB.
Aggregate analysis.
The fluorescence intensity values of control and experimental axons were measured. Plot profiles of 75 μm long axonal stretches were generated from 16-bit images, and the pixel intensity values were analyzed. To assess normal fluorescence fluctuations, a selection of intensity values derived from 50 control tauWT axons across the time course were randomly chosen, and each was zeroed to its 10th percentile. The mean + 5 SD of these values was calculated as 500 arbitrary units (a.u.). Therefore, tauWT control axons with values >500 a.u. were excluded from generating experimental control means, but included in the final analysis.
The fluorescence values for at least 12 control tauWT and experimental tauE14 75 μm long axonal stretches were analyzed per time point, with three separate experiments per time point. The mean of the SDs of the control axons not excluded by the 500 a.u. cutoff was calculated. Any individual fluorescence value of tauWT or tauE14 axons lying five times outside this mean was identified to be an aggregate-containing point. The sum of the aggregate-containing points was calculated for control tauWT and experimental tauE14 axonal stretches, and from this the percentage of aggregate-containing values along an axonal stretch was calculated. Any axonal stretch that contains >10% of its fluorescence values, i.e., over cumulative 7.5 μm of the analyzed length, as aggregate-containing values was identified as axon positive for tau aggregation. This cutoff allows for intensity variations because of e.g., crossing axons to be discounted. Finally, the percentage of cells positive for tau aggregation is calculated for tauWT and tauE14-expressing neurons.
Electrophysiology.
Cells were cultured on coverslips and transfected with RFP-TauWT or RFP-TauE14. For patch-clamp recording, cells were perfused with oxygenated (95% O2, 5% CO2) artificial CSF, which contained the following (in mm): 126 NaCl, 3 KCl, 1.25 NaH2PO4, 2 MgSO4, 2 CaCl2, 26 NaHCO3, and 10 glucose, pH 7.3–7.4, at a rate of 1–2 ml/min. Recordings were performed under visual control. Patch pipettes (4–6 MΩ) were pulled from thick-walled borosilicate glass tubing and filled with a solution containing the following (in mm): 110 K-gluconate, 10 KCl, 10 Na-phosphocreatine, 10 HEPES, 4 ATP-Mg, 0.3 GTP, pH 7.25 adjusted with KOH; osmolarity 280 mOsm/L. Recordings were performed at room temperature using an amplifier AxoPatch 200B. After measurement of intrinsic membrane potential, if necessary, current was injected to maintain the membrane potential −75 ± 5 mV. All membrane potentials recorded were corrected off-line for liquid junction potential of −10 mV measured directly. Current pulses of increasing amplitude were used to test excitability in current-clamp. Input resistance was measured in voltage-clamp with 2 mV pulses. Signals were low-pass filtered at 5 kHz and sampled at 20 kHz with 16-bit resolution, using a National Instruments analog card, and custom software written in MATLAB and C (MatDAQ, Hugh Robinson, Department of Physiology, Development and Neuroscience, University of Cambridge, 1995–2013). All analyses were performed in MATLAB.
Experimental design and statistical analysis.
All experiments contain data from a minimum of three independent dissections, with an individual experiment defined as the cells derived from embryos of one mouse. Statistical analysis was performed using GraphPad Prism 6 (GraphPad Software). All data in text are expressed as mean ± SD; graphs show mean ± SEM. Statistical analyses were performed using a two-tailed t test for comparison of two groups unless indicated otherwise, or an ANOVA for comparison of three or more groups, with details provided in the figure legends.
Results
Phosphomimetic tau spontaneously misfolds in cultured neurons
Hyperphosphorylation of tau is associated with aggregation and pathology, and phosphomimetic tau (tauE14) mislocalizes to dendritic spines and causes synaptic dysfunction (Hoover et al., 2010). To examine whether tauE14 spontaneously misfolds and forms visible aggregates, we cultured murine hippocampal neurons and transfected them at 1 DIV with fluorescently tagged human 0N4R tau: GFP-tauE14 or RFP-tauE14 and GFP-tauWT or RFP-tauWT, respectively. Expression of either tauWT or tauE14 did not adversely affect axonal outgrowth compared with control neurons (Fig. 1a), and both wild-type (data not shown) and mutant tau localized to intracellular structures (Fig. 1b). At 14 DIV, tauWT displayed a smooth and even distribution throughout the cell, including along the axon. In contrast, tauE14 expression resulted in a clustered tau distribution, particularly along distal axons (Fig. 1c,d). Using an antibody that selectively recognizes misfolded tau, MC1 (Jicha et al., 1997), we confirmed that the clustering of fluorescence reported tau misfolding within the cell. No misfolded tau was detected in axons exogenously expressing tauWT (Fig. 1d), confirming that it is not a consequence of the introduction of human tau per se. This suggests that the fluorescence accumulations in the mutant tau-expressing neurons represent either amorphous or structured buildups of misfolded tau that we refer to as aggregates. To analyze the appearance of tau aggregation we generated fluorescence distribution profiles along individual axons. This confirmed highly variable fluorescence intensities along axons of tauE14-expressing neurons, compared with a smooth fluorescence distribution in tauWT-expressing neurons (Fig. 1c). At 14 DIV, 56.0 ± 7.7% of the tauE14-expressing axons have developed aggregates, whereas the distal axons of tauWT-expressing neurons remained aggregate-free. This demonstrates that tau phosphorylation is sufficient to induce its misfolding and aggregation.
Tau aggregates spontaneously develop in GFP-tauE14-expressing hippocampal neurons. a, Hippocampal neurons were transfected with GFP, GFP-tauWT or GFP-tauE14 at 1 DIV and fixed and imaged at 7 DIV. No difference in axonal outgrowth was observed. Each data point is one axon, n ≥ 18 axons per condition from three experiments. One-way ANOVA, p = 0.98, F(2,55) = 0.017. Error bars indicate SEM. b, A higher-magnification view of a hippocampal neuron transfected with GFP-tauE14 at 1 DIV and fixed and imaged at 14 DIV. Scale bars, 10 μm. c, A line was drawn along distal axons expressing tauWT (top, magenta) or tauE14 (middle, green) to generate corresponding intensity profiles (bottom). Scale bar, 5 μm. d, TauE14 and surrounding untransfected axons (arrows), but not tauWT-expressing axons are positive for misfolded tau. Scale bar, 5 μm. ns = not significant.
Tau misfolding efficiently propagates to connected neurons
Untransfected neurites in the vicinity of a tauE14-expressing axon were positive for misfolded tau (Fig. 1d, arrows). This suggests that tauE14 expression is sufficient not only to induce misfolding and aggregation of tau within the axon, but also to generate seeds that spread to neighboring cells. To more precisely investigate the spread and propagation of tau misfolding and aggregation between neurons, we cocultured tauE14-expressing “donor” cells and tauWT “acceptor” cells within a microfluidic device that allows coculture of two spatially distinct neuronal populations that can be manipulated independently but are in contact via projecting axons (Taylor et al., 2005; J. W. Wu et al., 2013). This enabled us to separate tauE14 and tauWT-expressing neurons and identify individual connecting cells (Fig. 2a–c). Aggregates were first detected at 8 DIV in the axons of tauE14-expressing donor neurons (Fig. 2e), and the number of axons positive for tau aggregation steadily increased at a rate of 4.9 ± 0.6% per day, until a plateau was reached at 18 DIV with 75.5 ± 6.4% of axons containing visible tau aggregates (Fig. 2e). Next, we analyzed the percentage of aggregate positive acceptor neurons to assess whether propagation occurs in our system. We first detected aggregates in tauWT acceptor axons connected to tauE14-expressing donor neurons at 10 DIV, and their percentage increased over time by 4.9 ± 0.4% additional distal acceptor axons positive for tau aggregation per day (Fig. 2d,e). They thus followed the donor neurons with a delay of ∼3.7 d, until 77.5 ± 7.8% of distal axons contained tau aggregates, reaching the same plateau as donor distal axons (p > 0.999 between donor and acceptor axons at 24 and 26 DIV; Fig. 2e). No aggregation was detected when tauWT acceptor neurons were connected to GFP or GFP-tauWT-expressing control cells, demonstrating a clear difference in conformation of tauWT dependent on the type of donor cell to which it connects (p < 0.0001; Fig. 2d,e). This demonstrates that within a minimal neuronal circuit, expression of phosphomimetic tau in primary hippocampal neurons prompts tau misfolding and the generation of tau seeds, which rapidly and efficiently transfer to connected cells where they propagate aggregation of tau in a prion-like manner.
Tau misfolding efficiently propagates to connected neurons. a, Example image of donor and acceptor cells connected within microfluidic devices. Dashed lines indicate channel boundaries. Scale bar, 30 μm. b, Schematic of microfluidic device to investigate propagation from donor (tauE14; green) to acceptor (tauWT; magenta) neurons. c, Higher-magnification of white box in a showing intersections of donor axons and acceptor dendrites. Scale bar, 15 μm. d, TauWT expressing acceptor neurons form aggregates when connected to tauE14-expressing donor neurons. e, Neurons were transfected at 1 DIV and analyzed for aggregate formation. Time course of aggregate formation in axons of tauE14-expressing donor neurons (green circles), connected tauWT-expressing acceptor neurons (magenta squares), and tauWT-expressing neurons connected to control cells (orange triangles). N ≥ 12 axons per experiment, three independent experiments per time point. Two-way ANOVA with Tukey's test, F(2,60) = 851. TauE14 (green) compared with tauWT connected to tauE14 (magenta) at 24 and 26 DIV: ns, p > 0.999. TauWT connected to tauE14 (magenta) compared with tauWT connected to control (orange) at 10 DIV: **p = 0.0028; 12–26 DIV: ****p < 0.0001. Error bars indicate SEM. ns = not significant.
Distinct neuronal subcompartments display a differential vulnerability to tau misfolding and aggregation
In both the donor (data not shown) and acceptor neurons, tau aggregates first appeared in the distal axon, and were later detected in the somatodendritic compartment (Fig. 3a). For distal axonal misfolding and aggregation to occur in the acceptor neuron within the oriented setup of the microfluidic chamber (Fig. 2a,b), seeds that have been internalized at the somatodendritic compartment must have been transported to the distal axon, or alternatively, must have propagated the conformational change throughout the cell. Indeed, we detected GFP-tauE14-positive accumulations in the somata of RFP-tauWT-expressing neurons (Fig. 3b), indicating transfer of tauE14 to the connected neuron. No aggregates were visible within the somata of tauWT-expressing neurons connected to GFP-expressing cells (Fig. 3c). The aggregates within the tauWT-expressing somata were dual-positive for GFP-tauE14 and RFP-tauWT, demonstrating that the phosphomimetic tau had further recruited native tau into aggregates (Fig. 3b), thus suggesting a prion-like propagation of aggregation. We confirmed that the misfolding of tau was propagated to the distal axons of acceptor cells using MC1 staining (Fig. 3d). No GFP-tauE14 was detected in the tau accumulations of RFP-tauWT-expressing axons (data not shown); substantiating that tauE14 seeded the misfolding of tauWT. This demonstrates that as well as being sufficient for inducing tau aggregate formation, mimicking phosphorylation of tau is sufficient for the prion-like propagation of the conformational change to surrounding neurons. Interestingly, we never detected tau aggregation at the proximal axonal segment of either donor or acceptor neurons, and this axonal subcompartment also remained negative for MC1 staining indicating absence of misfolded species (Fig. 3d). Together with the observation that visible tau aggregates were first observed in the distal axon of acceptor cells, despite receiving the seeds at the somatodendritic compartment, this suggests that (1) smaller seeds distribute throughout the neuron before the formation of visible aggregates; (2) tau in the distal axon is most susceptible to aggregation; and (3) misfolded tau does not accumulate in, or is rapidly cleared from, the proximal axonal region.
Differential subcompartment vulnerability to tau misfolding. a, Acceptor somata (light gray circles) and distal axons (dark gray squares) containing tau aggregates at different time points. Each data point is the mean of one experiment, n ≥ 8 fixed cells per time point per experiment. Two-way repeated (region) -measures ANOVA, F(1,4) = 68.02, ns p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, with post hoc Sidak's multiple-comparison test. Error bars indicate SEM. b, c, Confocal microscopy shows that (b) GFP-tauE14 aggregates are detected in the somata of RFP-tauWT-expressing acceptor neurons and overlap with RFP-tauWT aggregates, whereas (c) RFP-tauWT-expressing neurons connected to GFP-expressing cells do not develop aggregates. Shown are Z-projections. Scale bars: 10 μm, imaged at 14 DIV. d, Proximal (1,3) and distal (2,4) axonal segments of tauE14 and connected tauWT-expressing neurons were counterstained for MC1 and imaged at 14 DIV. Scale bar, 5 μm. ns = not significant.
Misfolded tau is not acutely toxic
We monitored the formation and propagation of tau pathology at the single cell level, and thus were able to analyze the state of tauE14-expressing cells at the time of misfolding and active tau seed transmission and propagation of misfolding. Loss of synaptic connections has been identified as one of the earliest cellular changes in AD (Scheff et al., 1990; Masliah et al., 2001). Because tau misfolding and aggregation first occur in the distal axon, we visualized a presynaptic marker, synapsin, to assess the density of presynaptic sites in tauE14-expressing donor neurons at a time point of active tau seed propagation (14 DIV). There was no significant difference in the number of presynaptic sites between tauWT- and tauE14-expressing neurons and their untransfected surrounding axons (p = 0.82; Fig. 4a–c). In fact, the neurons appear intact and undistinguishable from surrounding untransfected cells, as judged by DIC imaging, exhibiting a smooth and intact plasma membrane and healthy nuclear morphology (Fig. 4a). This suggests that the presence and release of misfolded tau seeds are not acutely toxic to hippocampal neurons. Indeed, we did not observe disintegration or death of these neurons across the analyzed time course to 26 DIV and beyond.
Neurons transmitting tau pathology remain intact. a, b, A 14 DIV tauE14-expressing neuron. a, Overview of neuron stained with synapsin, nucleus stained with Hoechst 33342, and the DIC image showing an intact membrane at the cell body and along the axon. Scale bar, 30 μm. b, The distal axons of tauE14-expressing neurons counterstained with synapsin. Scale bar, 10 μm. c, Quantification of synapsin puncta per 20 μm stretch of axon. One-way ANOVA, F(2,6) = 0.06, p = 0.82. Each data point represents the mean of one experiment, with n = 45 axons total (untransfected, tauE14) and 35 axons total (tauWT).
We next investigated the physiological state of aggregate-containing neurons in more detail to assess the functional consequences. Because tau acts in microtubule stabilization (Drubin and Kirschner, 1986), and its dysfunction leads to axonal transport deficits (Alonso et al., 1994; Mandelkow et al., 2003), we first assessed lysosome dynamics. Lysosomes are transported bidirectionally along axons (Che et al., 2016). At 14 DIV there was a significant decrease in the number of lysosomes present at the distal axon of tauE14-expressing neurons compared with tauWT controls (p = 0.0017; Fig. 5a,b). Of those lysosomes present, a significantly lower percentage were moving within tauE14 axons (21.1 ± 3.9%) compared with tauWT controls (45.8 ± 5.8%, p = 0.0036; Fig. 5c–e), confirming an early axonal transport dysfunction in the presence of tau misfolding.
TauE14-expressing neurons show selective axonal dysfunctions at 14 DIV. Imaging of lysosomes at the distal axon reveals (a, b) a reduction in the number of lysosomes (white arrows) in tauE14-expressing distal axons compared with tauWT-expressing cells. N = 8 cells per experiment, each point is the mean of one experiment. t test, df = 4, t = 7.44, p = 0.0017. c–e, Analysis of lysosome dynamics in tauE14 and tauWT-expressing axons. c, d, Representative kymographs of tauWT and tauE14-expressing axons. e, Transport analysis of lysosomes. Each data point is the mean of one experiment; n ≥ 8 cells per experiment. Error bars indicate SEM. One-tailed t test, df = 4, t = 6.14, p = 0.0018. **p < 0.01.
To confirm our observation that neurons-expressing tauE14 remained viable in the presence of misfolded tau we then visualized spontaneous activity using calcium imaging (Fig. 6a,b). Both tauWT- and tauE14-expressing neurons displayed calcium fluxes that were sensitive to tetrodotoxin, showing that they were driven by voltage-gated sodium channels. This demonstrates that energy-dependent processes were functional and suggests that neurons remain electrically competent in the presence of misfolded and aggregated tau. To investigate the electrical competence in more detail we performed electrophysiological analysis on the cells with whole-cell patch-clamp. Both tauE14- and tauWT-expressing neurons were capable of responding to positive current injections with action potentials of a similar amplitude (Fig. 6c). There were no significant differences in the minimum current required to evoke a single action potential (rheobase); with tauE14- and tauWT-expressing neurons requiring 77 ± 40 pA and 83 ± 25 pA respectively (p = 0.82; Fig. 6d). We further found no significant differences in the input resistance between tauE14 (694 ± 367 MΩ) and tauWT (578 ± 101 MΩ) -expressing cells (p = 0.62; Fig. 6e), and the resting membrane potentials of tauE14 (−73.8 ± 2.4 mV) and tauWT (−73.3 ± 3.5 mV) -expressing neurons did not differ (p = 0.85; Fig. 6f). This shows that despite possessing and transmitting tau pathology, donor neurons maintain a normal resting membrane potential and fire action potentials to the same extent as tauWT control neurons.
TauE14 neurons are electrically competent. Basal calcium activity of a 14 DIV (a) tauWT (b) and tauE14-expressing neuron, which is silenced on application of 1 μm TTX. Addition of 100 mm KCl confirms viability at the end of acquisition. c–f, Electrophysiology on 14 DIV neurons. c, Sample traces of current injection into patch-clamped neurons. d, Rheobase of tauWT and tauE14-expressing neurons show a similar minimal current is required for action potential initiation. t test, df = 4, t = 0.24, p = 0.82. e, Input resistance of tauWT and tauE14-expressing neurons shows no significant difference. t test, df = 4, t = 0.53, p = 0.62. f, Resting membrane potentials recorded for tauWT and tauE14-expressing neurons. t test, df = 4, t = 0.20, p = 0.85. d–f, N ≥ 3 cells per experiment, each point = median of one experiment. ns = not significant.
Together, these data demonstrate that phosphomimetic tau misfolds and aggregates in the absence of exogenous seeds, and that transmission of tau misfolding to healthy neurons is an active and efficient process that is accompanied by selective neuronal dysfunction. The presence of misfolded and aggregated tau does not compromise neuronal excitability and is compatible with longer-term neuronal viability.
Discussion
What triggers the initial tau seed formation in vivo is unclear, and there is currently a debate as to the nature of this propagative species (Michel et al., 2014; Mirbaha et al., 2015; Sharma et al., 2018). The presence of mutations increases the propensity of seed formation (Gao et al., 2018), and tau aggregation is associated with hyperphosphorylation (Alonso et al., 1996; Wang et al., 1996, 2007). Indeed, AD is associated with an imbalance of kinase and phosphatase activities (Stoothoff and Johnson, 2005) that leads to hyperphosphorylation of tau, detectable within NFTs (Hanger et al., 2009). We here show that the negative charges conferred by a phosphomimetic tau, which simulates hyperphosphorylation at 14 disease-related sites, do not interfere with localization to subcellular structures, but are sufficient to initiate tau misfolding and aggregation within a living neuron in the absence of exogenous seeds. Previous work showed that when using recombinant seeds, the efficiency of seeding is decreased in the presence of hyperphosphorylation (Falcon et al., 2015), and tau seeds isolated from tgP301S tau mouse brains more potently seed misfolding than tau seeds formed in vitro. This increased potency of tgP301S seeds is retained upon amplification with recombinantly generated tau protein that is not post-translationally modified (Falcon et al., 2015). Together with our data, this suggests that hyperphosphorylation itself does not interfere with subcellular tau localization nor dictate misfolding. However, within a cellular environment the negative charges associated with hyperphosphorylation increase the propensity for folding into an alternative structure that can act as an efficient seed for native tau to misfold, aggregate, and propagate.
The compartmented setup of the device propagation assay allowed us to monitor individual neurons that formed or received misfolded tau species. This allowed for the first time, an analysis of propagation efficiency, and showed unexpectedly fast and robust tau propagation from neuron to neuron in vitro with a near complete transmission efficiency. Our study used pure murine hippocampal neurons, cultured in the absence of other cell types. The high efficiency of transmission may be because of a lack of a glial population, which display tau pathology in patients with tauopathy (Spillantini et al., 1997, 1998; Arai et al., 2001), as well as in in vivo (Clavaguera et al., 2013) and in vitro (Bolós et al., 2016) models of tauopathy. Therefore, glial populations may play a role in clearance of secreted pathogenic tau, and their absence in our setup revealed an under-appreciated intrinsic high efficiency of tau release and re-uptake in neurons. Synaptic contacts have been shown to enhance tau propagation in vitro (Calafate et al., 2015; Wang et al., 2017), however, our system shows tau propagation at earlier time points than mature synapse formation (Ichikawa et al., 1993), and no increased rate of propagation after this time point. This high efficiency of neuron-to-neuron tau propagation suggests that a physiological process of protein transmission may be at play, which is occurring in healthy neurons but only revealed under pathological conditions through transmission of a conformationally altered species. This idea is reinforced by the observation that healthy tau is secreted from intact neurons in an activity-dependent manner (Pooler et al., 2013). The rapid and efficient propagation of tau misfolding in our system is at odds with findings that show Braak staging progresses in patients over a matter of years to decades (Braak et al., 2011). This staging measures the presence of NFTs, which are highly structured end-stage tau assemblies within dead or dying neuronal cells. The discrepancy may thus result from the fact that we measured early events of tau misfolding in response to exogenously expressed mutant tau. The timing with which initial misfolding of tau converts to cellular degeneration, tangle formation, and cytotoxicity remains unclear. The connected compartmentalized setup used in this study allows direct investigation of individual neurons that initiate and propagate misfolding, or that receive transmitted tau seeds, at single-cell and subcellular levels. This revealed compartmentalized tau aggregation within individual neurons that begins within the distal axon, regardless of whether misfolding was originally initiated within the cell or transmitted to it. This spatial organization holds true despite the oriented setup of connected neurons, where tau seeds first enter the receiving neurons at the somatodendritic region: the site furthest from the distal axon where aggregation is initiated. Only later were aggregates visible within the somata. This confirms that the in vitro setup faithfully replicates in vivo observations describing pathological tau present in axonal tracts before its appearance in the somatodendritic compartment (Christensen et al., 2019), and before the neuropil threads seen in advanced AD (Braak et al., 1986). Furthermore, we did not observe tau aggregates in the proximal axonal segment, even at stages where clear aggregates were visible in both the soma and distal axon. This suggests either a selective vulnerability of the distal axonal segment, or protection against tau misfolding in proximal axonal regions. The axon initial segment has been shown to act as a barrier for select isoforms of tau proteins, potentially because of increased microtubule dynamics in this area (Sohn et al., 2016; Zempel et al., 2017), which may play a role in the lack of retention of misfolded tau within this region. However, it is clear from our data that misfolded tau is able to cross this barrier and cause distal axon pathology.
In vivo paradigms do not yet facilitate the direct visualization of neurons that are actively releasing and propagating tau pathology. Monitoring tau misfolding and aggregation within single cells allowed us to directly assess the physiological effects that tau misfolding has on neurons. We found that despite containing and propagating misfolded tau for prolonged periods of time, neurons remained alive, functional, and retained energy dependent processes, yet displayed a selective dysfunction in axonal transport. Previous studies showed that acutely added tau seeds can induce toxicity and cell death via calcium dysregulation (Gómez-Ramos et al., 2006; Kopeikina et al., 2012; Tian et al., 2013). However, we found that neurons containing self-generated tau aggregates had intact calcium fluxes, physiological resting membrane potentials, were able to elicit action potentials and remained viable for extended periods of time after initial tau misfolding, aggregate formation, and propagation. Our data therefore shows that misfolded tau and tau aggregates are not themselves lethal to neurons. Instead, they initiate a selective axonal transport dysfunction that precedes synaptic loss, and in isolation does not result in axonal degeneration or cell death. Furthermore, this compartmentalized assay provides the first direct evidence showing that tau pathology propagation can occur between live and functioning cells and precedes synaptic or neuronal degeneration. Interestingly, prion seeds are detected throughout all brain regions in mice infected with prion disease, whereas degeneration is region specific (Alibhai et al., 2016). Together, these data and our findings suggest that the idea that the presence of misfolded protein in isolation does not determine neurodegeneration (Alibhai et al., 2016) may hold true across different protein misfolding diseases. Combined with a potential physiological transmission of tau between neurons, this raises the possibility that at the stage of diagnosis, tau seeds may have spread throughout wider brain regions, having passed the time point where a sequestration of free tau seeds in isolation is effective in halting disease progression. However, our data imply that because the affected neurons are intact and viable at the early stages of tau misfolding, neurons containing early tau pathology could potentially be rescued.
Footnotes
This work was supported by Alzheimer's Research UK (Grants ARUK-PhD2014-10 and ARUK-PPG2017B-001) and the Biotechnology and Biological Sciences Research Council (BB/L007576/1). We thank Vincent O'Connor, Dr. Amrit Mudher, and Aleksandra Pitera for critically reading the paper.
The authors declare no competing financial interests.
- Correspondence should be addressed to Katrin Deinhardt at K.Deinhardt{at}soton.ac.uk
