Seizures are a druggable mechanistic link between TBI and subsequent tauopathy

Traumatic brain injury (TBI) is a prominent risk factor for dementias including tauopathies like chronic traumatic encephalopathy (CTE). The mechanisms that promote prion-like spreading of Tau aggregates after TBI are not fully understood, in part due to lack of tractable animal models. Here, we test the putative role of seizures in promoting the spread of tauopathy. We introduce ‘tauopathy reporter’ zebrafish expressing a genetically encoded fluorescent Tau biosensor that reliably reports accumulation of human Tau species when seeded via intraventricular brain injections. Subjecting zebrafish larvae to a novel TBI paradigm produced various TBI features including cell death, post–traumatic seizures, and Tau inclusions. Bath application of dynamin inhibitors or anticonvulsant drugs rescued TBI-induced tauopathy and cell death. These data suggest a role for seizure activity in the prion-like seeding and spreading of tauopathy following TBI. Further work is warranted regarding anti-convulsants that dampen post-traumatic seizures as a route to moderating subsequent tauopathy.


Summary:
Traumatic brain injury (TBI) is a prominent risk factor for neurodegenerative diseases 1 and dementias including chronic traumatic encephalopathy (CTE). TBI and CTE, like all 2 tauopathies, are characterized by accumulation of Tau into aggregates that 3 progressively spread to other brain regions in a prion-like manner. The mechanisms that 4 promote spreading and cellular uptake of tau seeds after TBI are not fully understood, in 5 part due to lack of tractable animal models. Here, we test the putative roles for excess 6 neuronal activity and dynamin-dependent endocytosis in promoting the in vivo spread of 7 tauopathy. We introduce 'tauopathy reporter' zebrafish expressing a genetically-8 encoded fluorescent Tau biosensor that reliably reports accumulation of human tau 9 species when seeded via intra-ventricular brain injections. Subjecting zebrafish larvae to 10 a novel TBI paradigm produced various TBI symptoms including cell death, 11 hemorrhage, blood flow abnormalities, post-traumatic seizures, and Tau inclusions. 12 Bath application of anticonvulsant drugs rescued TBI-induced tauopathy and cell death; 13 these benefits were attributable to inhibition of post-traumatic seizures because co-14 application of convulsants reversed these beneficial effects. However, one convulsant 15 drug, 4-Aminopyridine, unexpectedly abrogated TBI-induced tauopathy -this was due to 16 its inhibitory action on endocytosis as confirmed via additional dynamin inhibitors. These 17 data suggest a role for seizure activity and dynamin-dependent endocytosis in the prion-18 like seeding and spreading of tauopathy following TBI. Further work is warranted 19 regarding anti-convulsants that dampen post-traumatic seizures as a route to 20 moderating subsequent tauopathy. Moreover, the data highlight the utility of deploying 21 in vivo Tau biosensor and TBI methods in larval zebrafish, especially regarding drug 22 screening and intervention. 23 24 Introduction: 25 Traumatic brain injury (TBI) is a leading cause of mortality and disability worldwide 26 (Hay et al., 2016;Nguyen et al., 2016;Rimel et al., 1981). It also is a prominent risk 27 factor for neurodegeneration and dementia, such as chronic traumatic encephalopathy 28 (CTE) (Chauhan, 2014;Gardner and Yaffe, 2015;Uryu et al., 2007). TBI can result from 29 direct physical insults, from rapid acceleration and deceleration of the brain, or from targets. Progress on this front is hampered by lack of access to suitable models: 37 applying physical injury to a cell culture is difficult and poorly represents the complex 38 biopathology that intertwines many multifaceted aspects of brain physiology. 39 The progressive deposition of hyperphosphorylated tau protein in filamentous forms 40 is a defining hallmark of tauopathies, which includes Alzheimer's disease (AD), CTE,41 and several other dementias. Each of the tauopathies affects distinct brain regions and 42 has a unique clinical presentation (Kovacs, 2017;Orr et al., 2017). Early in CTE, 43 hyperphosphorylated tau is accumulated in a cluster of perivascular neurons and glia in 44 the depths of cortical sulci. Later in CTE, tau pathology is widespread and incorporates 45 cortical and subcortical grey-matter areas (Hay et al., 2016;Johnson et al., 2012;46 McKee et al., 2015). This broad spreading of tau pathology in CTE can also be 47 observed following TBI ascribed to single trauma events (Johnson et al., 2012). This 48 spreading of tauopathy is consistent with a prion-like mechanism; indeed brain 49 homogenates from mice subjected to TBI can initiate p-Tau pathology when injected 50 into healthy wildtype mice (Zanier et al., 2018). The recipient mice develop a p-tau 51 pathology similar to single severe TBI patients, which then spreads from injection sites 52 to distant regions, behaving similarly to bona fide prions (Zanier et al., 2018). 53 Beyond TBI, the self-propagation and prion-like spread of tau aggregates is thought 54 to play a key role in the progression of other tauopathies such as AD (Iba et   In this light, an intriguing aspect of TBI is the prominence of post-traumatic seizures 69 that might be predicted to initiate the aggregation and/or exacerbate the spread of tau 70 pathology. Seizures are one of the key consequences of all types of TBI, and they have 71 been more commonly reported in patients who suffered from blast injuries (Asikainen et 72 al., 1999;Salinsky et al., 2015). Though the exact prevalence remains undetermined 73 (Lucke-Wold et al., 2015), it is anticipated that over 50% of TBI patients with severe 74 injuries develop seizures or post-traumatic epilepsy (Kovacs et al., 2014). A link 75 between seizures and tau pathology is suggested by increased prevalence of seizures 76 in AD patients and animal models of AD (Sanchez et al., 2018;Yan et al., 2012). 77 Whether reducing post-traumatic seizures can delay or minimize the progression of 78 tauopathy has yet to be fully explored. 79 This knowledge gap is due in part to a lack of accessible in vivo models that can 80 report the progression and spread of tauopathy, or that allow neural activity associated 81 with TBI to be measured and manipulated. To address these issues, we engineered a 82 tauopathy biosensor transgenic zebrafish that develops GFP+ puncta when tau 83 aggregates within the brain or spinal cord. Additionally, we introduce a simple medium-84 throughput method to induce TBI in zebrafish larvae. Combining these novel 85 approaches, we found that post-traumatic seizures correlate strongly with spreading tau 86 pathology following TBI. Manipulating this seizure activity mitigated tau aggregation and 87 revealed a critical role for endocytosis in the prion-like spread of tau seeds in vivo 88 following TBI. The results from our novel in vivo TBI model implicate seizures and 89 dynamin-dependent endocytosis in the spread of tau seeds, thereby offering potential 90 therapeutic targets.

Immunoblotting of Cell Lysate and Zebrafish Brain Lysate 153
For cell lysate preparation, cells were washed with cold PBS, then collected and 154 incubated with cold lysis buffer (150mM NaCl, 50mM Tris-HCl (pH 8), 1 mM EDTA and 155 1% Nonidet P-40) supplemented with protease inhibitor (Cocktail Set III; Millipore) for 10 156 mins on ice. Cells were lysed using a bio-vortexer homogenizer for 20 sec for two 157 rounds. The lysate was centrifuged at 13000rpm for 10 mins at 4°C. The supernatant 158 was collected, and the protein concentration was determined using the Qubit™ Protein 159 Assay Kit (Invitrogen). 160 For zebrafish brain lysate preparation, the brains of adult zebrafish were dissected. 161 Brains were homogenized in cell lysis buffer (20mM HEPES, 0.2mM EDTA, 10mM 162 NaCl, 1.5mM MgCl2, 20% glycerol, 0.1% Triton-X) with protease inhibitor and 163 phospSTOP (Sigma-Aldrich) in the case of pt406 Tg. Brains were lysed using a bio-164 vortexer homogenizer and sonicated for 3 sec for one round. Samples were centrifuged 165 as above and concentration of the samples was assessed in a Qubit® fluorometer 166 (Invitrogen). 167 For immunoblotting, 30-40µg of the total protein was combined with 2X sample 168 buffer (Sigma-Aldrich) and boiled for 10 mins before loading in 11% SDS-PAGE. 169 Electrophoresis was performed using the Bio-Rad Power PAC system in running buffer 170 (25 mM Tris base, 192 mM glycine and 0.1% SDS). The gel was transferred to a PVDF 171 membrane using a wet transfer system. All membranes were blocked for one hour in 172 protein-free blocking buffer PBS (ThermoFisher) or TBST with 5% milk and then 173 incubated with primary antibody overnight at 4 o C with gentle agitation. The primary 174 antibodies used in this study include rabbit monoclonal GFP (abcam, EPR14104) at 175 1:3000 dilution, rabbit anti-β-actin (Sigma-Aldrich, A2066) at 1:10000. All membranes 176 were washed three times with 1X TBST before incubation with secondary antibody 177 (goat-anti-mouse HRP or HRP-conjugated anti-rabbit at 1:5000 dilution (Jackson 178 ImmunoResearch) for one hour at room temperature. The membranes were washed for 179 the final time before visualization using Pierce® ECL Western Blotting Substrate 180 (ThermoFisher) on a ChemiDoc (Biorad). For stripping and re-probing, the membranes 181 were stripped using mild stripping buffer (199.8 mM Glycine, 0.1% SDS and 1% Tween 182 20 with a pH of 2.2) before blocking them and repeating the methods described before. 183

Immunohistochemistry 184
Larvae were fixed overnight in 4% paraformaldehyde, either one day after being 185 subjected to TBI or following the subsequent application of drugs as indicated. 186 Immunostaining of Activated-Caspase3 on whole-mount larvae was carried out as 187  The phosphotungstate anion (PTA)-precipitated brain homogenate was prepared as 206 described (Woerman et al., 2016). Briefly, 10% (wt/vol) brain homogenate was prepared 207 as reported above and mixed with a final concentration of 2% sarkosyl (Sigma Aldrich) 208 and 0.5% benzonase (Sigma Aldrich, E1014), and then incubated at 37°C for two hours 209 with constant agitation in an orbital shaker. Sodium PTA (Sigma Aldrich) was dissolved 210 in ddH 2 O, and the pH was adjusted to 7.0 before it was added to the samples at a final 211 concentration of 2% (vol/vol). The samples were then incubated overnight under the 212 previous conditions. The next day, the samples were centrifuged at 16,000g for 30 mins 213 at room temperature. The supernatant was discarded, while the resulting pellet was 214 resuspended in 2% (vol/vol) PTA in ddH 2 O (pH 7.0) and 2% sarkosyl in DPBS. The 215 samples were next incubated for one hour before the second centrifugation. The 216 supernatant was removed and the pellet was re-suspended in DPBS. An aliquot of 5µl 217 of PTA purified brain homogenate was employed for electron microscopy (EM) analysis 218 to confirm the presence of fibrils in each sample. 219 220

Tau Fibrillization and EM Analysis: 221
Synthetic human tau protein (wildtype full-length monomers) was purchased as a 222 lyophilized powder (rPeptide, T-1001-2) and resuspended in ddH 2 O at a concentration 223 of 2mg/ml. The recombinant protein was fibrillized as described previously (Guo and 224 Lee, 2011). Recombinant tau was incubated with 40µM low-molecular-weight heparin 225 and 2mM DTT in 100 mM sodium acetate buffer (pH 7.0) at 37 o C, thereafter being 226 agitated for seven days. The fibrillization mixture was centrifuged at 50,000g for 30 227 mins, and the resulted pellet was resuspended in 100 mM sodium acetate buffer (pH 228 7.0) without heparin or DTT. Successful fibrillization was verified by EM. 229 Negative staining for EM analysis of fibrils was conducted as described elsewhere 230 (Eskandari-Sedighi et al., 2017). Briefly, 400 mesh carbon-coated copper grids 231 (Electron Microscopy Sciences) were glow-discharged for 40 sec before adding the 232 sample aliquots. PTA-purified brain homogenates or synthetic tau fibrils (5µL) were 233 applied on the top of the grid for 1 min. These grids were washed using 50µL each of 234 0.1M and 0.01M ammonium acetate and negatively stained with 2 × 50µL of filtered 2% 235 uranyl acetate. After removing excess stain and drying, the grids were examined with a 236 Tecnai G20 transmission electron microscope (FEI Company) with an acceleration 237 voltage of 200 kV. Electron micrographs were recorded with an Eagle 4k × 4k CCD 238 camera (FEI Company). 239 240

Cells 242
Polyclonal Tau4R-GFP cells were plated at 2 x 10 5 per well in 24-well plates. Cells were 243 transduced the next day, using 40µl of 10% clarified brain homogenate combined with 244 Opti-MEM to a final volume of 50µl. A further 48µl of Opti-MEM and 2µl of  To induce TBI, 10-12 unanesthetized larvae (3 dpf) were loaded into a 10ml syringe 301 with 1ml of E3 media. The syringe was blocked using a stopper valve to ensure no 302 larvae or media left the syringe upon compression of the plunger. The syringe was held 303 vertical using a metal tube holder at the bottom end of a 48" tube apparatus. A defined 304 weight (between 30 and 300g) was dropped manually from the top of the tube. The tube 305 diameter was matched to (slightly greater than) the weight's diameter to enhance 306 repeatability. This was either done once or repeated three times, with either 65 or 300g 307 weights. Once larvae were subjected to the traumatic brain injury, they were moved 308 back to a petri dish with fresh media and maintained for further analysis. To characterize the dynamic changes in pressure that occurred within the syringe 312 during the TBI events, the stopper valve attached to the syringe (described immediately 313 above) was replaced with a piezoresistive pressure transducer (#MLT844 AD 314 Instruments, Colorado Springs, CO). Events were monitored via a PowerLab 2/26 data 315 acquisition device and LabChart 7 software (AD Instruments). The pressure transducer 316 was zeroed to report gauge pressure (pressure changes relative to atmospheric 317 pressure) and was calibrated against a manometer (Fisherbrand Traceable from 318 Thermoscientific, Ottawa ON). After each weight drop, the syringe apparatus was reset 319 to remove any air bubbles and the pressure transducer was zeroed. Time courses of 320 induced pressure were reported over a 350 msec time frame with 50 msec of base line 321 recording, while mean and maximum pressure values were calculated from the initial 322 300 msec following the impact of the weight. 323 324

Recording blood flow following TBI 325
Abnormalities of blood flow and circulation resulted from TBI was detected 5 to 10 mins 326 after larvae was subjected to TBI. The blood flow in the tail area of zebrafish larvae, 327 either those subjected to TBI or uninjured controls, was recorded using Leica DM2500 328 LED optical microscope.

Measuring neuronal activity during TBI using CaMPARI 356
Bright green CaMPARI larvae were loaded into 20ml syringe containing 1ml E3 media 357 (prepared as per Westerfield 2007, but without ethylene blue) and were exposed to a 358 405 nm LED array (Loctite), which illuminated the syringe entirely (Fig. 4A). Larvae 359 were exposed for 10 sec, with the LED array at a distance of 7.5 cm from the syringe, 360 while subjected to traumatic brain injury using the 300g weight as described above. 361 Following this photoconversion of CaMPARI during TBI, larvae were anesthetized in 362 0.24 mg/mL tricaine (MS-222, Sigma Aldrich) and embedded in 2 % low-gelling agarose 363 (A4018, Sigma Aldrich) for analysis under confocal microscopy. 364 Maximum intensity projections were acquired from Z-stacks (8 µm steps) using a 365 20x/0.8 Objective and a laser-point scanning confocal microscope (Zeiss 700). The 366 hindbrain area was analyzed, as it was the brain region most responsive to traumatic 367 brain injury. To specifically isolate the brain regions and obtain data points, a 3D area 368 was isolated by creating a surface with Imaris® 7.6 (Bitman, Zuerich) and the mean 369 fluorescence intensities of the green and red channel intensities were calculated. Data 370 points were presented as a red/green ratio for each individual larva and interpreted as 371 relative neural activity, which is defined as red photoconverted CaMPARI in ratio to 372 indicated. For Retigabine (RTG) treatment, 10µM was used to treat TBI larvae 383 beginning six hours after TBI. Unless otherwise stated, KA, 4-AP and/or RTG were 384 applied to larvae for 38 hours, then a fresh drug-free E3 media was added. The 385 formation of GFP-positive puncta was analyzed at four to five days post injury. 386 Pyrimidyn-7™(P7), the dynamin inhibitor, was purchased at a 50mM concentration 387 supplied in DMSO (Abcam). Larvae that were subjected to TBI were treated within six 388 phenotypes when expressed in these cells and moreover, that prion-like mechanisms of 424 tauopathy spread are best modeled in an intact brain (e.g. vectored by blood and 425 glymphatic circulation, ventricles, axonal projections and immune systems). Therefore, 426 we engineered a tauopathy biosensor transgenic zebrafish that expresses a fluorescent 427 tau reporter protein. Our genetically encoded fluorescent reporter protein was 428 composed of the sequence of the human tau core-repeat domain fused to GFP with a 429 linker sequence and is referred to here as Tau4R-GFP ( Fig. 1A and S1A). Contrasting 430 previous in vitro models, our biosensor did not feature any pro-aggregation mutations in 431 the human tau repeats; this design was intended to minimize spontaneous aggregation 432 Wildtype GFP is also abundant in the heart, which serves as a marker of the transgene being present but is otherwise irrelevant to our analyses. (C) Western Blot on zebrafish brain confirmed production of Tau4R-GFP at the expected size, similar to a SOD1-GFP biosensor and coordinately larger than GFP alone. (D) Human Tau fibril precipitated from transgenic (Tg Tau P301L ) mouse brain homogenates using PTA and assessed by EM. (E) Application of PTA-purified brain homogenate induced the formation of tau inclusions similar to clarified brain homogenate (scale bar 50µm; compare to Fig S1D), but application of equivalent preparations from non-Tg mice produced no GFP+ inclusions. (F-I) Tau biosensor zebrafish detects diseaseassociated human tau fibrils following intraventricular injection of brain homogenate. Crude brain homogenates were microinjected into the hindbrain ventricle of Tau4R-GFP zebrafish larvae at two days post-fertilization, and tau inclusions were analyzed at several time points. (F) Tau biosensor zebrafish larvae developed readily apparent GFP+ inclusions in the brain and spinal cord (Fig. S2) when injected with brain homogenate burdened with tau pathology pathology (from Tg mice) but not from healthy brain homogenate (F', from non-Tg mice). F" inset shows many adjacent cells exhibiting GFP+ Tau aggregates. (G) Tau biosensor zebrafish injected with human tau fibrils (within Tg mouse brain homogenate) developed significantly more aggregates on the spinal cord compared to uninjected control and other control groups, including compared to wildtype mouse brain homogenate (**p ≤ 0.01, ***p ≤ 0.001) (H) Same data as in G, expressed as the percentage of larval fish showing Tau aggregates in the spinal cord, and (I) those same fish also showed Tau aggregates in the brain, over time. n = number of individual larvae.
events. The expression of the biosensor protein in zebrafish was under the control of 434 the pan-neuronal promoter neuronal enolase 2 (eno2, see Bai et al., 2007), which drives 435 expression throughout the CNS (Fig. 1B and S1C). We deployed the transgene in a 436 transparent zebrafish line (the 'Casper' background (White et al., 2008)) to facilitate 437 analysis beyond the early larval development stages (when pigmentation would 438 otherwise begin to obscure microscopy). We isolated a stable transgenic (Tg) line that 439 expresses the Tau4R-GFP biosensor reporter robustly and clearly in the CNS (Fig. 1B), 440 Tg(eno2:Hsa.MAPT_Q244-E372−EGFP) ua3171 , and assigned it allele number ua3171. 441 Simultaneously, we expressed the same biosensor in vitro to validate the construct 442 we deployed in vivo ( Fig. S1A and B). Both in HEK293T cells and Tg zebrafish, 443 immunoblotting using anti-GFP antibody detected our Tau-4R-GFP reporter protein at 444 the expected size of ~45 Kd, similar to a SOD1:GFP biosensor protein of similar 445 predicted size, and an appropriately larger size relative to GFP protein alone (Fig. 1C, 446 S1B'). 447 We assessed the capacity of our Tau4R-GFP biosensor to report the presence of 448 tau pathology via transducing brain homogenates into cells. Brain homogenates 449 burdened with tauopathy, from transgenic mice expressing mutant human tau (Tg 450 Tau P301L ), were compared to normal non-Tg mouse homogenates as a negative control. homogenate containing pathogenic human tau fibrils (from Tg Tau P310L mice) (Fig.  454   S1D). The in vitro assay detection rate was approximately 5% of cells having GFP+ 455 inclusions in total, with 2% of cells forming multiple nuclear puncta and ~3% forming 456 one cytoplasmic inclusion, whereas various negative controls consistently displayed 0% 457 of cells with inclusions (Fig. S1E). To verify that tau aggregates in the clarified brain 458 homogenate caused the GFP+ puncta, we purified tau aggregates from the tissue 459 samples using PTA precipitations (Woerman et al., 2016). Tau fibrils purified from these 460 preparations were characterized via EM analysis (Fig. 1D). Transducing these 461 preparations (in contrast to control preparations derived from non-Tg mice) produced 462 fluorescent puncta in the Tau4R-GFP reporter cells (Fig. 1E), confirming the ability of 463 our Tau4R-GFP chimeric protein to report tau aggregation. Tg mouse brain) developed GFP+ puncta, reflective of tau aggregation in the brain (Fig.  475 1F, F"). These tau inclusions were prominent near the ventricle wall as well as in 476 sensory neurons along the spinal cord, when injected with brain homogenate from 477 human-tau transgenic mouse (Fig. S2B). These puncta appeared to have either a lone 478 dot-like shape or were similar to the multiple nuclear puncta detected in vitro, in which 479 three to four small puncta are clustered together. Repeated assessment of the location 480 of tau aggregates on the spinal cord of the same individuals over multiple days, using 481 somite numbers as landmarks, suggested a movement of some of these puncta over 482 time ( Fig. S3A and B). 483 The abundance of GFP+ spinal cord inclusions was progressive and significantly 484 higher in larvae injected with pathogenic TAU brain homogenate compared with various 485 controls (p<0.0001 at 3dpi and 4dpi, Fig. 1G). Few larvae in the control groups 486 developed spontaneous inclusions but the number of the larvae and the abundance of 487 those inclusions were minimal (Fig. 1G). 80% and 35% of the larvae injected with 488 human tau fibrils developed puncta in the brain and spinal cord, respectively ( GFP+ inclusions (Fig. 2B). Following injection of mouse brain homogenate containing 507 human tau fibrils, applying the proteasome inhibitor MG-132 substantially enhanced the 508 percentage of larvae bearing Tau4R-GFP+ inclusions in the brain (to ~70%, Fig. 2B), 509 relative to equivalent larvae without MG-132 (~36%, Fig. 2B). 510 It was striking that the zebrafish tau biosensor was robustly able to discriminate 511 brain homogenates with human tau aggregates versus those that were not. However, 512 we considered an alternative explanation for the data: the difference may not depend 513 directly on human tau in the brain homogenate but could instead reflect other bioactive 514 components of the degenerating Tg mouse brain. To verify that the formation of GFP+ 515 puncta in zebrafish can be seeded by a protein-only injection, we delivered synthetic 516 human tau protein (2N4R). After confirming the recombinant tau proteins were 517 appropriately fibrillized via EM (Fig. 2C), we delivered them by intraventricular injections 518 as described above. Similar to previous data with brain homogenate, the larvae that 519 were injected with synthetic tau fibrils developed inclusions proximal to the brain 520 ventricles as well as along the spinal cord at 3-6 dpi. The abundance of tau aggregates 521 along the spinal cord was significantly higher in larvae injected with the synthetic tau 522 fibrils compared to larvae injected with tau monomers or to the non-injected group 523 (p<0.05) (Fig. 2D). The distribution of larvae based on the number of tau aggregates 524 they accumulated also supported these findings (Fig. 2E). In sum, the Tau4R biosensor 525 deployed in the CNS of larval zebrafish was able to report tau species, and further 526 revealed the prion-like induction of tauopathy via protein-only seeding in vivo. 527 528 529

Introduction of the first traumatic brain injury model for larval zebrafish 530
We next sought to deploy our tau biosensor in a tauopathy model that enables 531 higher throughput than can be achieved with intraventricular injection methods. We 532 considered traumatic brain injury (TBI) as an inducer of the tauopathy in Chronic 533 Traumatic Encephalopathy (CTE); further, we were encouraged that innovations in this 534 realm could fill an unmet need for a high-throughput, genetically tractable in vivo model 535 of these devastating concussive injuries. Although a few methods have been reported to 536 induce traumatic brain injury in adult zebrafish that are comparable to mammalian TBI available for zebrafish larvae. Here we introduce and validate a simple and inexpensive 539 method to induce traumatic brain injury in zebrafish larvae. Investigating traumatic brain 540 injury in larvae offers substantial benefits regarding experimental throughput, economy, 541 accessibility of drug and genetic interventions, and bioethics. 542 We devised a traumatic injury paradigm by loading zebrafish larvae (~12 individuals 543 in their typical E3 liquid growth media) into a syringe with a closed valve stopper, and 544 applying a hit on the plunger to produce a pressure wave through the fish body akin to 545 pressure or shock waves experienced during human blast injury (Nakagawa et al., 546 2011) (Fig. 3A). To challenge the method's reproducibility, and to permit manipulation of 547 injury intensity, a series of defined masses were dropped on the syringe plunger. Increased cell death in the brain of 4 dpf larvae subjected to TBI as indicated by immunostaining of activated Caspase-3 (magenta). Positive and negative controls for immunostaining are in Figure S4. Nuclei were stained with DAPI in gray gray for reference. These are dorsal views of larval zebrafish brains with anterior at the left. (G) Seizurelike clonic shaking is observed in a subset of larvae after TBI. Movie frames are displayed from Supplemental Video S2. These frames (left and right panels) are separated by ~400 msec in time, and are lateral views of the larval zebrafish trunk (akin to red box in G'). Control fish without TBI show little movement except obvious blood flow. Following TBI, larvae show bouts of calm (bottom left) interspersed (~400 msec later) with bouts of intense seizure-like convulsions (Stage III seizures; bottom right). (H) Larvae subjected to TBI also displayed Stage II seizures, i.e. weaker seizures that manifest as hypermotility and are detected using a previously optimized behavioural tracking software system -seizures are significantly more intense following TBI compared to the control group (***p<0.001; dots are raw data for each larva, mean is plotted ±SE). from pressure waves, we examined multiple markers known to be associated with blast-555 induced traumatic brain injury, including cell death, hemorrhage, blood flow 556 abnormalities, and tauopathy (Bir et  We established the TBI method via empirical testing of various parameters, 560 restricting ourselves to materials and methods that can be adopted inexpensively, with a 561 goal of consistently inducing a robust injury (see phenotypes below) vs. a tradeoff with 562 maximizing survival of the larvae. Subsequent to this optimization, we were able to 563 characterize the pressure induced within the syringe during each injury (Fig. 3B-D). The 564 maximum pressure induced was near 170 kPa (Fig. 3B). The dynamics of the pressure 565 change events during TBI (Fig. 3B) imply that dropping the heavier weights led to the 566 weight bouncing and producing a secondary increase in pressure (e.g. at ~175 or ~275 567 msec in Fig. 3B). The maximal pressure induced varied from ~130 to ~175 kPa in an 568 approximately linear fashion depending on the mass of the weight dropped (Fig. 3D). 569 The mean pressure change over the first 300 msec of the TBI also increased in a nearly 570 linear fashion, and increased by nearly an order of magnitude when dropping weights of 571 30g compared to 300g (Fig. 3C). 572 We evaluated TBI-induced hemorrhage via the use of Tg[gata1a:DsRed] larvae that 573 have red fluorescence in their blood cells (Traver et al., 2003). Hemorrhage was 574 observed variably in larvae when a heavy weight (300g) was used to induce the 575 traumatic injury (Fig. 3E). Further, approximately half of the TBI larvae showed 576 abnormalities in blood flow including a temporary reduction or complete absence of 577 blood circulation (video 1), consistent with abnormalities detected in rodent TBI models 578 (Bir et al., 2012). Subsequently, we assessed apoptosis in the TBI larvae, observing 579 that our TBI method induced cell death in larvae as detected by staining for active 580 Caspase-3 ( Fig. 3F and S4). The number of active-Caspase-3-positive cells was 581 negligible in the control groups compared to a mean of 62 apoptotic cells in TBI larvae 582 (SEM ± 9.17, n=3) and 75 (SEM ± 4, n=2) in positive-control-larvae (cell death induced 583 with camptothecin, CPT; Fig. S4). These data all align well with existing animal models 584 of TBI with respect to mimicking characteristic features of human TBI, and support the 585 effectiveness of our method in inducing traumatic brain injury in larval zebrafish. (approximately 40%) of zebrafish larvae after they were subjected to traumatic brain 593 injury. In some instances the activity was highly reminiscent of Stage III seizures 594 (defined previously in larval zebrafish as the most intense seizures; Liu and Baraban, 595

2019) with bouts of intense clonic convulsions and arrhythmic shaking (Supplemental 596
Video S2; exemplar frames from the movie are in Figure 3G). Other individuals 597 exhibited hypermotility that is exactly consistent with past definitions of less intense 598 Stage I or Stage II seizures. We quantified the latter seizure activity via behavioral 599 tracking software (which we had previously optimized and validated for quantifying 600 seizures in larval zebrafish Leighton et al., 2018)) and determined 601 that larvae subjected to TBI exhibited seizure-like activity that was significantly higher 602 than the control group (p<0.0007) (Fig. 3H). to TBI, coincident with brief application of photoconverting light (405nm light provided by 612 an LED array directed at the syringe, as described in Fig. 4A). A sharp increase in 613 neuronal activity during TBI was evident, especially in the hindbrain region as indicated 614 by enhanced red emission (Fig. 4B). CaMPARI allows robust quantification of neural 615 activity expressed as a ratio of red:green fluorescent emission, which confirmed that 616 neuronal excitability increases significantly in response to brain trauma ( Fig. 4C and D). 617 Notably, this combination of newly introduced methods of traumatic brain injury being 618 integrated with CaMPARI optogenetic methods (where the stable/irreversible changes 619 from green to red fluorescent reportage allows a ratiometric quantification in a 620 subsequent microscopy session) offers the rare ability to assess neural activity on un-621 restrained (free-swimming) subjects during TBI. In sum, our data reveal a substantial 622 burst of neural activity occurs during TBI, and that zebrafish larvae exposed to TBI 623 33 624 ). The ratio of red:green emission is stable such that it is quantifiable via subsequent microscopy. (B) Increased neural activity during TBI is represented by increased red:green emission (red pseudocoloured to magenta) in the hindbrain of larvae (B"), compared to larvae not receiving TBI (B') or fish not exposed to photoconverting light ("no PC" in panel B). These representative maximum intensity projection images show dorsal view of zebrafish brain (anterior at top), including merged, or red or green channels alone.
(C) Heatmaps encode the CaMPARI signal (higher neural activity = higher red:green = hotter colours), highlighting location of increased neural activity during TBI relative to control larvae not receiving TBI. (D) Quantification of CaMPARI output in the hindbrain area reveals a significant increase in the neuronal excitability during TBI compared to control group not receiving TBI (*p<0.05. Each data point is an individual larva).
subsequently exhibit a significantly higher propensity for spontaneous seizures. 625 626

Traumatic brain injury on Tau biosensor zebrafish larvae induced GFP+ puncta 627
After validating that our method was able to induce traumatic brain injury upon 628 zebrafish larvae, we next asked whether TBI induces tau aggregates in our tau 629 biosensor model. Initially, we evaluated if our TBI method would induce aggregation of 630 fluorescent proteins in models expressing GFP alone or other biosensor proteins such 631 as SOD1-GFP. Following TBI, and regardless of injury intensity, no GFP+ aggregates 632 were detected in these controls (Fig. S5 A-C). Similar results were obtained with other 633 transgenic zebrafish that express GFP in motor neurons (data not shown). Further, our 634 Tau4R-GFP fish additionally express an unmodified GFP variant in the active heart 635 muscle, and this robust GFP showed no sign of aggregation following TBI. Remarkably, 636 in these same individual Tau4R-GFP larvae we detected Tau4R-GFP biosensor GFP+ 637 puncta in both brains and spinal cords following TBI (Fig. 5A-B). The abundance of 638 GFP+ puncta increased with time following the injury (Fig 5C-D). To determine if the 639 severity of tauopathy varies coordinately with severity of the traumatic injury, we 640 assessed the impact of different masses. Although some variability is evident, a dose-641 response relationship is apparent such that the 65g, 100g and 300g weights induced 642 more tau aggregates compared to the control and 30g weight (Fig. S8E). The heaviest 643 weight (300g) induced significantly more tau aggregates versus the control group or the 644 group with the 30g weight (p<0.01 and p<0.001, respectively). Therefore, we decided to days, perhaps reminiscent of repetitive sports injury. We observed an increase in the 650 abundance of tau aggregates when the weight was dropped multiple times during one 651 day, or over three consecutive days, but this increase was not statistically significant 652 (Fig. S8 B-D). 653 The GFP+ tau aggregates formed in the brain region following TBI tend to form 654 fused shapes (Fig. S8A) reminiscent of the spontaneous aggregates described above. 655 The aggregates on the spinal cord, however, had similar shapes to aggregates detected 656 post brain-injections, but with qualitatively less brightness in some instances. Multi-day 657 monitoring of individual larvae (Fig. S6 & S7) revealed variation in formation of tau 658 aggregates amongst individual TBI larvae. We monitored the abundance of tau 659 aggregates within individual fish over time following TBI and found that the average 660 tauopathy significantly increased compared to the control group (p<0.05 at 3 dpti and 661 p<0.01 at 4 dpti (days post traumatic injury) (Fig. 5C,D). Analysis of distribution of larvae 662 binned into the number of Tau4R-GFP+ puncta at 3 dpti showed that more larvae 663 developed Tau4R-GFP+ puncta compared to the control group (inset in Fig 5D). 664 Considering that many of the larvae subjected to TBI formed Tau4R-GFP+ puncta in the 665 brain that had a fused pattern (Fig. S8A), we focused on tau aggregates that formed on 666 the spinal cord as their abundance could be most efficiently quantified compared to 667 aggregates that formed in the brain. tauopathy. We first asked if a correlation exists between seizure intensity and extent of 675 tauopathy. Following TBI, some larvae exhibited seizure-like movements, while some 676 did not seem to move abnormally relative to untreated fish (Fig. 3H). We sorted the 677 larvae subjected to TBI into groups exhibiting the seizure-like behavior and those that 678 displayed no overtly abnormal movement. Larvae exhibiting seizure-like behavior after 679 TBI went on to develop abundant spinal cord aggregates (5-fold increase, p<0.001) in 680 comparison to larvae that showed no seizure-like response to TBI (Fig. 5E). 681 To assess the hypothesis that seizure activity has a causal role in increasing the 682 abundance of tau aggregates in our TBI model, we employed convulsant and anti-683 convulsant drugs to modulate the seizure intensity and in vivo neural activity. We 684 selected drugs that are well-established to behave similarly in zebrafish as in mammals, 685 though it is perhaps notable that the multi-day drug application used here is longer than 686 the acute applications typically considered in zebrafish (Ellis et al., 2012). Our 687 hypothesis predicted that decreasing seizure-like activity following TBI would reduce 688 tauopathy. Indeed, applying the anti-convulsant drug Retigabine (RTG), that opens 689 voltage-gated potassium channels (KCNQ, Kv7), resulted in a significant decrease in 690 the abundance of GFP+ puncta (p<0.05) with many TBI larvae not developing any 691 Tau4R-GFP aggregates (Fig 5F). 692 Similarly, intensifying post-traumatic seizures via application of the convulsant 693 kainate increased the abundance on tauopathy 4-fold (p<0.001. Fig 5G) in a dose-694 dependent manner (Fig S9). Kainate did not increase Tau4R-GFP+ puncta in the 695 absence of TBI. Surprisingly, the convulsant 4-aminopyridine did not increase tauopathy 696 (explored below). 697 To assess if the impacts of kainate and retigabine on tauopathy were directly due to 698 their modulation of post-traumatic seizures, we applied effective doses of each in 699 concert. Co-application of kainate and retigabine following TBI produced an abundance 700 of Tau4R-GFP+ puncta that was indistinguishable from larvae receiving TBI without 701 pharmacology (Fig 5G). 702

TBI-induced cell death was likewise correlated with the intensity of post-traumatic 703
seizures. Co-application of kainate and retigabine following TBI increased or decreased, 704 respectively, the abundance of cell death in a manner coordinate with the tauopathy 705 ( Fig 5H). 706 Overall, convulsant and anti-convulsant drugs acted to increase and decrease TBI-707 induced tauopathy, respectively. The drugs appear to be specific -their individual 708 impacts on tauopathy and cell death are largely attributable to their epileptic and anti-709 epileptic modulation of post-traumatic seizures, because when kainate and retigabine 710 were applied concurrently they negated each others' effects. 711 712

Appearance of tauopathy following TBI requires endocytosis 713
To further examine increased seizure activity after TBI, we applied 4-aminopyridine 714 (4-AP), a K v channel blocker and convulsant drug. We predicted that raising the level of 715 seizure activity would elevate tauopathy abundance in our TBI model, aligning with our 716 observations following application of kainate (above). Surprisingly, higher doses of 4-AP 717 consistently abrogated the appearance of tau aggregates. Treating TBI larvae with 200 718 or 800µM of 4-AP for a prolonged period (38 hours, beginning 24 hours post traumatic 719 injury) significantly inhibited the abundance of Tau4R-GFP+ puncta in the TBI group 720 ( Fig. 6A-B, Fig. S10A-B). Analysis of the distribution of larvae linked to the number of 721 tau aggregates supported this finding with no zebrafish larvae developing aggregates in 722 groups treated with 4-AP (Fig. S10C). It is worth noting that 4-AP is commonly used in 723 zebrafish models of epilepsy, but rarely used for prolonged treatment. To evaluate if the 724 time at which treatments are administered plays a role in this unexpected result, we 725 treated larvae with 200 µM 4-AP at earlier time points, specifically during traumatic brain 726 injury and 1.5 hours later. We kept the duration of 4-AP treatment the same as previous 727 experiments (38 hours). We found that administering 4-AP during different time 728 windows relative to the traumatic brain injury did not measurably alter the inhibitory 729 action of 4-AP on the abundance of tau aggregates (Fig. S10D). A similar observation 730 was made when the duration of the 4-AP treatment was reduced to 24 hours (Fig.  731   S10E). 732 Next, we considered if this unexpected inhibition of tauopathy by high-dose 4-AP 733 convulsant is a direct consequence of increased neural activity (e.g. perhaps via neural 734 exhaustion). We found that larvae receiving TBI and 4-AP continued to exhibit a lack of 735 tau aggregates when co-treated with anti-convulsant retigabine (p<0.0001) (Fig. 6B). 736 This suggested that high doses of 4-AP block the formation of tau aggregates via a 737 mechanism independent of its convulsant activity. 738 To resolve a mechanism whereby high doses of 4-AP reduced tau pathology, 739 contrary to our predictions above regarding neural hyperactivity, we considered   treatments significantly reduced tauopathy in our TBI model (Fig. 6D). To determine if 754 these results are applicable to human tau, we induced traumatic brain injury on double-755 transgenic larvae expressing both human tau (the 0N4R human Tau isoform, see Bai et 756 al., 2007) and our tau biosensor reporter, followed by treatment with either 4-AP or P7. 757 Apart from the untreated control, both groups treated with 4-AP or P7 exhibited a 758 noticeable reduction in abundance of tau aggregates. While the decrease in the case of 759 P7 was not statistically significant, statistical analysis showed significance after 4-AP 760 treatments (p<0.001) (Fig 6E and F). These findings confirmed the ability of 4-AP and 761 dynamin inhibitors of reducing human tau aggregates in our TBI larvae. human TBI and tauopathy, but by addressing many of these challenges in an accessible 780 and vibrantly active vertebrate brain, we offer an innovative approach for the study of 781 prion-like events, tauopathy and/or TBI. 782 Here, we introduce a simple method for delivering TBI to larval zebrafish that can be 783 scaled to high-throughput and adopted at low expense. The tractability and (that are accessible for TBI only at postnatal stages) when considering highly invasive 857 procedures like TBI. The latter conclusion relies on the assumption that the knowledge 858 gained is of value, i.e. relevant to appreciating disease etiology. 859 Our data argue that our TBI methods are germane to clinical aetiology, because 860 (akin to existing animal models of TBI) we were able to confirm the presence of various 861 markers associated with brain injury, such as cell death, abnormalities in blood flow, 862 hemorrhage and the occurrence of post-traumatic seizures. 863 The post-traumatic seizures apparent in our TBI model led us to consider the neural 864 events occurring during the TBI, and their potential bearing on the correlation between 865 neural activity and tauopathy. Few studies examine how TBI impacts neuronal circuits, 866 especially in vivo, and these typically consider events several hours or days after brain 867 trauma (Bugay et al., 2019). This may be of importance when considering evaluating 868 the reasons behind the developments of post-traumatic seizures and epilepsy. In a 869 controlled cortical impact model of TBI, an initial decrease or loss in neuronal activity is 870 recorded after injury before a rise in neuronal activity is noted (Ping and Jin, 2016). 871 Whether this occurs in different types of TBI, like blast TBI, was unexamined. To 872 address this, we performed TBI on larval zebrafish expressing CaMPARI, a genetically 873 encoded optogenetic reporter of neural activity. CaMPARI is particularly ideal for this 874 question, as its reportage of neural activity (a stable and quantifiable shift from green to 875 red fluorescence) occurs only during user-defined times and that reportage is relatively 876 permanent. This allowed us to quantify the CNS activity that had occurred during TBI, 877 by characterizing the ratio of red:green fluorescent emission using confocal microscopy 878 after the TBI injury was completed. This approach therefor allows relatively easy access 879 to quantifying neural activity during injury in an unencumbered freely-swimming animal. abundance and cell death in our TBI model (Fig. 5G,H). This finding is in line with 906 observations in a patient with epilepsy and a history of head injury, in which progressive 907 tau pathology was noted (Geddes et al., 1999;Thom et al., 2011). Intriguingly, reducing 908 seizure activity after TBI via anti-convulsant drugs was able to significantly reduce 909 tauopathy and cell death, providing further evidence of the relationship between 910 seizures and tauopathy in TBI (Fig 5F,G and H). The mechanism of drug action 911 appears to be dominated by its anticonvulsant properties, because its effects were 912 reversed by co-application of convulsants. Thus, anti-convulsants are intriguing as a 913 route to slowing progression of tauopathy following TBI, and it is encouraging that they 914 are already commonly deployed to prevent post-traumatic seizures. 915 916 Endocytosis mediates prion-like spread of tauopathy 917 One particular convulsant drug, 4-AP, inhibited tauopathy in our TBI model (Fig 6 A  918 and B), contrary to our hypothesis that seizure intensity is positively correlated with 919 tauopathy following TBI. 4-AP is a voltage-gated potassium channel blocker that 920 enhances neuronal firing activity and has been used often in zebrafish seizure studies 921 treatments such as those we deployed here, e.g. past studies rarely exceed one hour of 924 4-AP (Winter et al., 2017). Thus, we considered that our high dose and prolonged 925 stimulation with 4-AP may have led to off-target effects; we confirmed this insomuch 926 that the 4-AP's inhibition of tauopathy was not related to its convulsant properties (as 927 determined by 4-AP's effects being unaltered by potent anti-convulsants (Fig. 6B)). 928 Indeed previous in vitro work revealed high concentrations or prolonged stimulation with 929 4-AP has off-target effects via inhibiting dynamin, which is important for the endocytosis 930 of synaptic vesicles at the nerve terminals (Cousin and Robinson, 2000). The inhibition 931 of endocytosis observed in that study was independent of 4-AP-dependent seizure 932

activity. 933
We further queried the potential role of dynamin-dependent endocytosis in the 934 prion-like progression of tau pathology after TBI by applying endocytosis inhibitors that 935 target dynamin. Dynamin is a GTPase involved in two mechanisms of endocytosis that 936 are important for synaptic vesicle transport (Singh et al., 2017). Empirical work on 937 human stem cell-derived neurons has indicated that tau aggregates are internalized via 938 dynamin-dependent endocytosis and that blocking other endocytosis pathways 939 independent of dynamin, such as bulk endocytosis and macropinocytosis, did not 940 disrupt tau uptake (Evans et al., 2018). On the contrary, inhibiting dynamin significantly 941 decreased the internalization of tau aggregates. Our results are in line with the 942 previously mentioned findings that show tau progression in TBI models depends on 943 dynamin-dependent endocytic pathways -blocking them with two different inhibitors 944 (and with 4-AP) dramatically lessened the abundance of tau seeds (Fig 6C-F). Hence, 945 our findings not only provide in vivo validation of past in vitro works, but also suggest 946 mechanisms underlying prion-like spreading of tau seeds in TBI and CTE that could aid 947 in developing therapeutic strategies. 948 949

Limitations of our approach 950
Our tauopathy biosensor, human Tau4R-GFP, was deployed in vivo and uniquely 951 able to detect significant increases (and decreases) in the abundance of tau aggregates 952 following various insults and treatments, typically in a dose-dependent manner and in 953 harmony with expected trends. Considering this success, it remains intriguing that a 954 subset of larvae exhibit GFP+ tau puncta despite receiving no known tauopathy-955 inducing insults. This suggests the larvae express the transgene at a level near to a 956 threshold for producing spontaneous aggregates. We performed selective breeding to 957 minimize these occurrences and tentatively believe, after too few generations, that 958 genetics of the fish is a factor -substantial genetic variation exists in zebrafish in-bred 959 lines (Balik-Meisner et al., 2018; Guryev et al., 2006). However, we acknowledge the 960 variation could be a minor technical artefact rather than biological. More optimistically, 961 this inter-individual variation and stochastic appearance of tauopathy, in a high-962 throughput model, could be leveraged to newly appreciate aspects of spontaneous AD 963 or other non-familial tauopathies. Regardless, future work will also need to characterize 964 the biochemistry and biophysics of the human Tau inclusions in zebrafish compared to 965 patients or rodent models. 966 Regarding our TBI methods, further refinements may yet be able to improve 967 consistency of the injury and reduce the apparent variability between individuals. This 968 variability is real, but somewhat offset by the large sample sizes attainable: our TBI 969 methods offer the potent advantages of zebrafish larvae with respect to genetic and 970 drug accessibility in high-throughput formats, while also retaining the critical in vivo 971 complexity required to investigate disease etiology and treatments. Further, it remains 972 to be established if the mechanisms we reveal are ubiquitous across the various forms 973 of TBI: our model fills a gap by supplying a rare 'closed head' TBI model (as opposed to 974 the majority of animal models that access the brain by removing skull elements prior to 975 brain injury, see exceptions by (Meconi et al., 2018;Mychasiuk et al., 2014). Our model 976 might be most relevant to brain trauma experienced by the human foetus (e.g. during 977 car collisions or domestic abuse), considering the developmental stage and aqueous 978 media. Further work is also needed to appreciate how the physics of our blast injury is 979 altered by occurring at a small scale (e.g. larval brain is <500 µm). At this point we are 980 left to assume that the cellular and physiological aspects of TBI we consider here are 981 sufficiently similar across all classes of TBI, and thus the knowledge gleaned may be 982 variably applicable. 983 Finally, we have chosen to restrict our analysis to study of larval fish. While this 984 offers many logistical and ethical advantages detailed above, it limits our study to acute 985 effects occurring over the course of several days. Conclusions from such work, once 986 refined and validated using the power of the in vivo zebrafish larva model, should be 987 tested in rodent models where it is equally time-consuming to assess the long-term 988 efficacy of treatments on these progressive late-onset dementias. 989 990

Conclusion 991
Currently, no available treatments are applicable to all tauopathies, which remain as 992 devastating and inevitably fatal dementias. Zebrafish larvae, fostered by appropriate 993 innovations, now offer a potent complement both to rodent models of TBI and to cellular 994 models of tauopathy. Our engineered fish allowed us to reveal post-traumatic seizures 995 as a druggable mechanistic link between TBI and the prion-like progression of 996 tauopathy. Intriguingly, our conclusions have potential for translation to TBI clinics 997 where anti-convulsants are already in use as prophylactics for post-traumatic epilepsy, 998 though further work remains to address if they mitigate (the risk or severity of) later 999 progression of CTE, AD or other tauopathies. 1000