Modeling Alzheimer’s disease in primary neurons reveals DNA damage response coupled with MAPK-DLK signaling in wild-type tau-induced neurodegeneration

Background: Alzheimer’s disease (AD) is the most prevalent form of neurodegeneration. Despite the well-established link between tau aggregation and clinical progression, the major pathways driven by this protein to intrinsically damage neurons are incompletely understood. Methods: To model AD-relevant neurodegeneration driven by tau, we overexpressed wild-type human tau in primary mouse neurons and characterized the subsequent cellular and molecular changes. RNAseq profiling and functional investigation were performed as well. A direct comparison with a mutant human tau was conducted in detail. Results: We observed substantial axonal degeneration and cell death associated with wild-type tau, a process accompanied by activated caspase 3. Mechanistically, we detected deformation of the nuclear envelope and increased DNA damage response in tau-expressing neurons. Gene profiling analysis further revealed significant alterations in the mitogen-activated protein kinase (MAPK) pathway; moreover, inhibitors of dual leucine zipper kinase (DLK) and c-Jun N-terminal kinase (JNK) were effective in alleviating wild-type human tau-induced neurodegeneration. In contrast, mutant P301L human tau was less toxic to neurons, despite causing comparable DNA damage. Axonal DLK activation induced by wild-type tau potentiated the impact of DNA damage response, resulting in overt neurotoxicity. Conclusions: We have established a cellular tauopathy model highly relevant to AD and identified a functional synergy between DNA damage response and the MAPK-DLK axis in the neuronal degenerative process.

developing abundant amyloid plaques in conjunction with neuro brillary tangles (NFT) that contain insoluble tau aggregates in their brains [8,9]. Through studying animal and cell culture models expressing mutated MAPT, many groups over the years have elucidated the pathogenic potential of FTLD-associated mutant tau [10][11][12]. However, a critical question that remains to date is whether the pathogenic role of tau in AD is in any way dissimilar from that in primary tauopathy. If so, the ndings from the mutant tau studies might be inadequate to guide the therapeutic development for AD.
Elevated tau protein level is causally linked to disease pathogenesis; in particular, intraneuronal accumulation of aggregated tau closely correlates with the clinical progression of AD as well as other tauopathies [13,14]. In rodent brains, neuronal overexpression of mutant MAPT is su cient for the development of NFT pathology and the onset of neurodegeneration, leading to the successful construction of numerous tauopathy disease models [12,15]. In vitro, however, tau overexpression has been mostly carried out in immortalized non-neuronal cells, such as HEK293 and HELA cells, to facilitate studies on the tau interactome, transcriptomic in uence by tau, or self-aggregation and seeding properties of tau [16][17][18]. Surprisingly little is known about the functional impact of increased intraneuronal tau on differentiated primary neurons. Consequently, the major cellular pathways that govern tau-mediated neuropathology relevant to AD remain poorly understood.
To investigate AD-relevant tau-dependent neurodegeneration, we developed a cellular tauopathy model by overexpressing human tau in mouse primary neurons. This system afforded us to comprehensively assess neuronal intrinsic responses to full-length wild-type (WT) tau, and enabled a direct comparison with an FTLD-associated mutant form. Remarkably, we have detected heightened neurotoxicity triggered by WT tau and obtained insights on two major signaling pathways that promote the degenerative process that is pertinent to AD.

Generation of AAV vectors
Recombinant AAV1/2 vectors containing the longest tau isoform (2N4R) of human MAPT cDNA of wildtype and mutant P301L under human SYN1 promoter were used [19]. To construct a control vector, a stop codon arisen from single nucleotide mutation was generated at the N-terminus of the MAPT sequence. The detailed sequences were provided in supplemental data 4. Recombinant AAV stocks were produced by Gene Vector Core at Baylor College of Medicine. Vector genomes were titrated by quantitative PCR and purity validated by SDS-PAGE. The titer of 4 x 10 9 GC/ml was selected for neuronal cell infection.

Primary cultures of mouse cortical and hippocampal neurons
Primary neurons were harvested and cultured as previously described [20]. Brie y, P0 pups of mixed sexes were decapitated into ice-cold dissection buffer (1x HBSS supplemented with 10 mM HEPES, pH 7.5, 0.6% glucose, 20 U/ml penicillin, 20 µg/ml streptomycin). Forebrain hemispheres were dissected and stripped of meninges. The tissues of cortex with hippocampus were isolated using a dissecting microscope and digested in trypsin at 37°C for 15 minutes with gentle swirling. Following addition of 500 µl trypsin (2.5%), 400 µl soybean trypsin inhibitor (1 mg/ml) and 100 µl DNase I (1%) were then added. After tissue pieces decanted off, the supernatant was removed and replaced with 2 ml of DMEM and 20 µl DNase I (1%). The digested tissues were triturated with a P1000 pipette tip 8-10 times. After allowing the remaining pieces to settle, the supernatant was collected into a fresh tube for centrifuging at 1200 rpm for 5 min. The cell pellet was resuspended with 5ml of DMEM and then centrifuged again. After carefully removed the supernatant, the dissociated cells were resuspended in 2 ml completed neuronal culture media (Neurobasal medium supplemented with 2% B27, 0.5 mM L-glutamine, 40 U/ml penicillin, 40 µg/ml streptomycin), and passed through a 70 µm lter. The single suspended cells were plated into poly D-lysine-coated culture plates or glass coverslips (1-2 x 10 5 cells/cm 2 ). Cells were maintained in incubators at 37°C, 5% CO 2 , and half the culture medium was replaced every 5-7 days. LDH cytotoxicity assay 50 µL of culture media was analyzed by the standard procedure provided by the manufacturer (Cat# C20301, CyQUANT™ LDH Cytotoxicity Assay, Thermo sher, USA). The absorbance by excitation of 560 nm and emission of 590 nm was measured on the SpectraMax® ABS Microplate Reader (Molecular Devices, USA).
Neuronal culture treatments

Immuno uorescent staining
The cultured neurons on the coverslips were xed with 4% paraformaldehyde (Santa Cruz, cat# sc-281692) for 20 minutes at 4°C. After washing with 1x PBS for 3 times, the cells were incubated in 0.2% Triton X-100 for 20 minutes. After rinsing the cells with PBS twice and pre-incubating them with a blocking buffer of 10% normal donkey serum (Cat# S30-100ML, Millipore) and 0.5% Triton X-100 in PBS for 1 h, the cells were incubated with primary antibodies anti-HT7 (Cat# MN1000, Invitrogen), Total tau were then incubated with uorescent secondary antibodies diluted in blocking buffer for 1 hour at room temperature. Following 3 additional PBS washes for 10 minutes each, the cells were counterstained with DAPI, mounted with mounting medium, and photographed using the confocal laser scanning microscopy (Leica, Germany) or EVOS uorescence microscopy (Life Technologies, USA).

Immunoblotting
Total protein from cultured neurons were extracted with a cold RIPA lysis buffer composed of protease and phosphatase inhibitor mixtures. The protein concentration was quanti ed using the BCA assay. The tissue or cell lysate with equal amounts of proteins were subjected to electrophoresis on 8 to 15% SDS-PAGE and then electrophoretically transferred to a PVDF membrane. After blocking in 5% milk for 1 h, the membranes were incubated with primary antibodies Total tau, HT7, AT8, CP13, AT180, PHF1, p-c-Jun (Ser73), Synaptophysin, PSD-95, p-JNK, pH2AX, Active Caspase 3, γ-tubulin (Cat# T6557, Sigma), β-Actin (Cat# sc-47778, Santa Cruz) overnight at 4°C. After three washes with Tris-buffered saline containing 0.05% Tween 20 for 10 minutes each, the membranes were incubated with uor-conjugated Donkey anti-Mouse IgG (Cat# 926-68072, LI-COR), Donkey anti-Rabbit IgG (Cat# 926-68073, LI-COR), or Donkey anti-Goat IgG (Cat# 926-68074, LI-COR) for 1 h. After three times of washing, the signals were visualized by on a LI-COR Odyssey blot imager or Bio-Rad ChemiDoc™ Imagers. The band intensities were normalized to the corresponding value of γ-Tubulin or β-actin expression as a loading control. For tau oligomers detection, the method from Dr. Kayed's lab was followed [21] and a pre-cast NuPAGE 4-12% Bis-Tris Gels for SDS-PAGE (NP0335BOX, Invitrogen) was used for oligomeric tau detection.  Table S1. The results of relative quantitative PCR were analyzed using the comparative threshold cycle (Ct) method and normalized to Hprt1 expression as an endogenous reference.

RNA-seq
Total RNA was extracted with Direct-zol RNA Microprep Kits. Novogene Co. (CA, USA) performed mRNA sequencing and data analysis. Basically, the RNA quality was evaluated as follows: RNA integrity number > 7.0 and 28S:18S ratio > 1.8. Messenger RNA was puri ed from total RNA using poly-T oligo-attached magnetic beads for library construction. The library cDNA was subjected to paired-end sequencing with a pair end 125-base pair reading length on an Illumina HiSeq 2500 sequencer (Illumina, San Diego, CA, USA). For quanti cation of gene expression level, featureCounts v1.5.0-p3 was used to calculate the reads numbers mapped to each gene. And then FPKM of each gene was determined based on the length of the gene and reads count mapped to this gene. Differential expression analysis was performed using the DESeq2 R package (1.20.0). The P-values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate. Genes with an adjusted P-value < = 0.05 found by DESeq2 were considered as differentially expressed. ClusterPro ler R package was used to test the statistical enrichment of differential expression genes in KEGG pathways. The local version of analysis tool (http://www.broadinstitute.org/gsea/index.jsp), GO, KEGG, Reactome, DO and DisGeNET data sets were used for Gene Set Enrichment Analysis (GSEA).

Quanti cation of axon degeneration and neuro lament fragments in vitro
The cultured neurons were immuno-stained with β-tubulin III antibody to visualize the microtubule structure, or with Phospho-Neuro lament H (pNF-H) to visualize the neuro lament. For per image, the total number of spheroids on the microtubule and neuro lament fragments was counted by ImageJ software. Then the number was normalized by dividing to the area (pixels) of β-tubulin III or pNF-H positive expression. More than 6 independent areas from 3 individual slides were analyzed for each experimental condition.
Thio avin S staining Fresh Thio avin S (ThioS) solution was prepared by dissolving 1 g of ThioS (Cat# T1892, Sigma) in 100ml 80% ethanol, and stirring overnight at 4°C, and ltering for nal use. The xed cells were rinsed in PBS, then transferred to solution containing 0.0002% thio avin S in PBS for 8 mins, rinsed in 40% ethanol in PBS twice for 2 mins, followed by two rinses in PBS, then mounted on slides.

Quanti cation and statistical analysis
Statistical analysis was performed using GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA). All data in bar plots are presented as mean ± SEM. Data are representative of two or three independent experiments. Unless otherwise noted, differences between two groups were analyzed by two-tailed Student's t-tests, and differences between three or more groups were analyzed by one-way ANOVA with Tukey's multiple comparisons test, as indicated in gure legends. The rate of nuclear envelop invagination between groups was compared by Fisher's exact test. p < 0.05 was considered statistically signi cant (noted as * P < 0.05, * * P < 0.01, * * * P < 0.001 in plots), and those over 0.05 were considered non-signi cant ("ns", or numerical P values listed in certain plots). All n values are listed in gure legends for each respective plot. All micrographs shown are images representative of multiple replicates as indicated.

Results
Neuronal wild-type human tau overexpression prompts axonal and neuronal degeneration To enable human tau expression, primary mouse neurons were infected at DIV5 with an AAV vector containing the human WT MAPT gene (2N4R) under the control of the human SYN1 promoter (Fig 1A). Full-length human tau (hTau) protein was produced in the neurons in a time-dependent manner and reached an estimated ratio of 2:1 relative to endogenous mouse tau protein (Fig S1A-B). Neuronal hTau expression was further con rmed by immunostaining with a hTau-speci c antibody (clone HT7) and a polyclonal antibody that recognizes total tau protein (both human and mouse) ( Fig 1B). In addition, we detected hTau phosphorylation at multiple amino acid positions (Ser396/Ser404 (PHF-1 + ), Thr231 (AT180 + ), Ser202/Thr205 (AT8 + ), and Ser202 (CP13 + )) ( Fig 1C), which were localized in the soma as well as neurites of the MAPT-transduced neurons ( Fig 1D). Hyperphosphorylation can promote tau aggregation and NFT formation [1,2]. We detected high molecular weight multimeric species of hTau protein by western blot (Fig 1E) and upregulation of a conformational epitope of aggregated hTau ( Fig  1F); however, no signal was obtained by Thio avin S staining ( Fig S1C). These ndings suggest that WT full-length MAPT overexpression in primary neurons leads to human tau hyperphosphorylation and aggregation in the absence of NFT.
Several days after AAV infection, WT hTau+ neurons started to display morphological changes indicative of degeneration. By day 7 post-infection, we observed gross axonal degeneration manifested by the development of numerous axon swelling or spheroids (Fig 2A) and the accumulation of fragmented neuro laments in culture (Fig 2B). In conjunction, synaptic markers were signi cantly reduced (Figs 2C and 2D). By day 9 post-infection, neuronal density was decreased by > 80% in WT hTau+ culture ( Fig 2E).
Moreover, by assaying the culture medium longitudinally, we detected a time-dependent escalation of lactate dehydrogenase (LDH) release, corroborating a remarkable neurotoxic phenotype of intracellular WT human tau in primary neurons (Fig 2F).

Caspase 3 activation is involved in WT hTau-dependent neurodegeneration
To investigate the neurotoxicity elicited by WT hTau, we rst examined apoptosis, a programmed cell death pathway that has been implicated in the loss of neurons in AD [22]. Apoptotic death involves the function of caspases, among which caspase 3 activation serves as a central effector [23]. In WT hTau + neurons, we detected increased abundance of cleaved caspase 3, the signal of which was present both inside the nuclei and in the neuritic areas rich with axonal spheroids (Figs 3A and 3B). Consistently, WT hTau + neurons expressed more Casp3 mRNA (Fig 3C). PUMA, encoded by the Bbc3 gene, is critically involved in the intrinsic apoptotic signaling pathway [24]. We detected elevated expression of both mRNA and protein products of Bbc3 in neurons with WT hTau (Figs 3D and 3E). To test the importance of caspase activity in tau-induced degeneration, we treated WT hTau + neurons with pan-caspase inhibitor Z-VAD-FMK and observed a signi cant lessening of neurotoxicity (Figs 3G and 3H). These results suggest that caspase 3 activity contributes to WT human tau-driven neurodegeneration.
Wild-type hTau induces DNA damage response in neurons To probe the mechanistic connection between tau and apoptosis, we assessed the nuclei of neurons expressing WT hTau. Lamins are architectural proteins that confer mechanical stability to the nuclear envelope [25]. Visualization of nuclear lamins revealed that, in the presence of WT hTau, many neurons displayed abnormal invaginations in their nuclear membrane (Figs 4A and 4B). Further analysis revealed the colocalization of hTau with ru ed lamin A/C signals in the nuclear envelope (Fig 4C), suggesting a physical presence of WT hTau at sites of disruption.
Nuclear envelope disturbance can have serious consequences for a cell, particularly detrimental for chromatin and genomic stability [25]. We found that WT hTau + neurons displayed fewer nuclear foci of heterochromatin protein 1α (HP1α), an indication of the loss of chromatin compactness (Fig 4D). Phosphorylated Ser-139 of the histone variant H2AX (pH2AX) is a sensitive molecular marker of doublestrand DNA (dsDNA) damage and repair [26]. Inside the nuclei of WT hTau + neurons, we detected signi cantly increased pH2AX signal intensity, which correlated with the higher abundance of pH2AX protein in the cells (Figs 4E, 4F and 4H). Furthermore, we found heightened nuclear pH2AX signal in cells exhibiting nuclear membrane invagination (Fig 4G), suggesting a functional connection between nuclear envelope disruption and DNA damage response (DDR).
Activation of the well-known tumor suppressor p53 occurs in response to DNA damage and other cellular stresses and plays a critical role in apoptosis [27]. In conjunction with the pH2AX signal, phospho-p53 protein was detected in the nuclei of WT hTau + neurons (Fig 4I, J). During DDR, p53 induces cell cycle regulatory proteins to elicit intrinsic checkpoint control [28]. Accordingly, we found that WT hTau + neurons increased the transcription of Cdkn1a and Cdkn1b, which encode inhibitors for cyclin-dependent kinases ( Fig 4K). These results collectively pinpoint a prevalent DNA damage response that is triggered by WT human tau in differentiated neurons.

MAPK-DLK signaling partakes in WT hTau-dependent neurodegeneration
To gain deeper molecular insights into the cellular pathways that are affected by WT hTau, we performed RNAseq analysis on the AAV-transduced neurons. A signi cant number of genes were differentially regulated between the AAV control and WT hTau + neurons. Among these, multiple apoptosis and cell death-related genes displayed signi cantly higher levels of expression in WT hTau + neurons (Fig S2A).
KEGG pathway analysis further revealed multiple signi cantly affected cellular processes (Fig 5A; supplemental data S1). Consistent with the observed axonal and synaptic degeneration, "axon guidance" and "synaptic vesicle cycle" represented the topmost signi cantly altered processes. Interestingly, the "MAPK signaling pathway" contained the greatest number of genes signi cantly affected by WT hTau (Fig 5A). Among the genes differentially upregulated by WT hTau, Jun proto-oncogene ranked highest by the adjusted p value (Fig 5B; supplemental data S2).
c-Jun, the protein product of Jun, can be phosphorylated by c-Jun N-terminal kinases (JNKs), a subset of MAP kinases, and subsequently translocate to nucleus to take a part in the transcriptional activity of AP-1 [29]. Consistent with the upregulation of the Jun transcript, the c-Jun protein was expressed more abundantly and exhibited a nuclear enrichment in WT hTau + neurons (Fig 5C). Moreover, phosphorylated c-Jun at both serine 63 and serine 73 increased in these cells, which was exclusively present inside the nuclei (Figs 5D, 5F, 5G, and S2B-C). In addition, we detected highly elevated nuclear phospho-JNK expression in many WT hTau + neurons (Figs 5D, 5E, 5G). Together with the RNA pro ling results, these ndings uncover a substantially activated MAPK-JNK pathway in degenerating WT hTau + neurons. JNK has both pro-and anti-apoptotic functions, depending on many complex factors [30]. To gauge its involvement in our model, we deployed a selective JNK inhibitor at a dose effective to diminish the activities of JNK and c-Jun in the culture of WT tau + neurons (Figs 5D-5F) and observed substantial reduction of neurotoxicity with the treatment (Figs 5H-5J).
DLK is a MAP3K functionally involved in axonal degeneration as well as regeneration, generally in sync with the activities of JNKs and c-Jun [31][32][33][34]. We hypothesized that DLK activity may serve as a key node of the MAPK pathway in WT hTau + neurons thus decided to examine the potency of an established DLK inhibitor to curb JNK/c-Jun signaling [35]. Similar to the JNK inhibitor, DLK inhibition effectively repressed the activities of JNK and c-Jun in hTau+ neurons (Figs 5D-5F). More potent than the JNK inhibitor, DLK inhibition signi cantly rescued the neurons from WT hTau-induced neurotoxicity (Fig 5H-J). Altogether, we have identi ed MAPK-DLK signaling as a signi cant contributor to WT hTau-induced neurodegeneration.

P301L hTau overexpression differentially affects primary neurons
The P301Lmissense mutation in MAPT is causally associated with human FTLD and has been extensively studied in various tauopathy models [12,36]. To compare its functional impact relative to WT hTau, we similarly infected primary neurons with P301L MAPT packaged into an identical AAV vector to achieve equivalent overexpression. By morphological examination and measurement of LDH release, we unexpectedly observed a reduced extent of neurodegeneration by P301L hTau in side-by-side examination with WT hTau (Figs 6A-C). Consistently and signi cantly, P301L hTau induced less cytotoxicity than WT hTau on primary neurons in all time points examined (Fig S3A). In line with these observations, we detected lower amounts of activated caspase 3 protein in mutant hTau + neurons (Figs  6D and 6E). To comprehend the underlying mechanism, we carefully compared the hTau protein expression and found that P301L hTau was somewhat less phosphorylated, even though the total abundance of transgenic tau protein was slightly higher than in cells expressing WT hTau (Figs 6D-6F, S1A). Despite a ubiquitous distribution of the mutant hTau protein, phosphorylated P301L hTau was largely restricted to somas of neurons, an intriguing contrast to phosphorylated WT hTau which abundantly associated with axons and dendrites (Fig 6F). No thio avin S + signal was detected in these cells (Fig S3B), suggesting the lack of NFT deposition, similar to WT hTau.
For a deeper understanding of changes initiated by mutated tau, we pro led the transcriptome of P301L hTau + neurons and obtained a list of genes that were differentially regulated by the mutant tau. Pathway analysis revealed that P301L hTau signi cantly affected processes such as "electron transport chain", "mitochondrial protein complex", and "ribosome/ribosomal subunit" (Fig 6H; supplemental data S3). By contrast, WT hTau altered the expression of over 1600 genes in neurons, more than 3 times over mutant hTau (Fig 6G; supplemental data S2). A direct genome-wide comparison between WT and P301L hTau+ neurons uncovered selective alterations of "axon guidance" and "apoptosis" by WT hTau, and "ribosome" and "oxidative phosphorylation" by P301L hTau (Fig S3C). In terms of DNA damage response, the two hTau forms elicited similar gene expression pro les ( Fig 6I); however, P301L hTau did not disrupt MAPK pathway genes as did the WT hTau ( Fig 6I).
To con rm these ndings, we examined P301L hTau + neurons and detected many cells with invaginated nuclear envelope, similar to WT hTau + neurons (Figs 6J and 6K). Likewise, increased pH2AX signal, comparable with WT hTau, was detected in P301L hTau + neurons (Figs 6L-6N). By contrast, P301L hTau + neurons contained much lower detectable levels of phospho-c-Jun in their nuclei in comparison with WT hTau (Figs 6M and 6O), consistent with RNAseq pro ling showing a major difference in MAPK pathway between the two forms of hTau. Moreover, P301L hTau + neurons showed little sign of axonal degeneration (Fig S3D). Altogether, our investigation reveals that P301L hTau overexpression differentially affected primary neurons other than WT hTau in many important aspects.

DNA damage response and MAPK-DLK signaling synergistically promote neurodegeneration
To comprehend the relationship between DDR and MAPK-DLK activation in WT tauopathy, we took a closer look at the WT hTau + neurons by dual immunostaining of pH2AX and phospho-c-Jun and detected more cells expressing the marker of DDR than those with MAPK signaling (Fig 7A). We did not nd a signi cant correlation between the levels of the two markers on a per cell basis, but noticed a general lack of cells solely expressing nuclear phospho-c-Jun (Figs 7A and 7B). Interestingly, WT hTau + neurons treated with DLK inhibitor did not diminish the pH2AX signal (Figs 7C and 7D), suggesting that DNA damage response is either upstream or independent of MAPK-DLK signaling.
Pathogenic tau is known to modify cytoskeletal functions by dissociating from microtubules and intercepting F-actin [37,38]. Having observed distinct distribution patterns of WT and P301L phospho-hTau ( Fig 6F) and selective accumulation of activated caspase 3 in axonal spheroids in WT hTau + neurons (Fig 2A), we hypothesized that 1) axonal WT hTau may stimulate MAPK-DLK signaling by inducing axonal degeneration independently from somatic hTau-induced DDR; and 2) MAPK-DLK signaling and DDR function synergistically in WT hTau + neurons, whereas lack of axonal DLK activation renders P301L tau less neurotoxic. To test this hypothesis, we adopted a selective inhibitor of microtubule formation [39], as disruption of microtubule dynamics is known to cause DLK-mediated axonal swelling and degeneration [40]. Low-dose colchicine treatment robustly induced nuclear phospho-c-Jun accumulation without any sign of DDR in primary neurons (Figs 7E and 7F). We also used etoposide, a chemotherapy agent capable of inducing dsDNA breaks: primary neurons exposed to low-dose etoposide readily upregulated nuclear pH2AX expression (Fig S4). Individually, these compounds caused limited toxicity in neurons after 48 hours; however, combined treatment resulted in severe cell loss and LDH release, indicating a powerful synergistic effect of axonal stress and DDR in neurodegeneration (Figs 7G-7I). To further demonstrate the interaction between these pathways in the tau-relevant situation, we treated P301L hTau + neurons with low-dose colchicine and observed signi cantly expedited degeneration in culture (Figs 7J-7L). In summary, these ndings uncover a striking synergistic interplay between DNA damage response and the MAPK-DLK axis in AD-relevant neurodegeneration.

Discussion
In this study, we investigated the consequence of human tau overexpression in primary neurons and discovered a pathogenic coupling of DNA damage response and MAPK-DLK signaling in wild-type hTaumediated neurotoxicity. By contrast, P301L hTau, which is associated with FTLD, elicited primarily the DNA damage response, with less aggressive neurodegeneration. The overt WT hTau-induced pathology was highly signi cant and reproducible, yet somewhat unexpected. Historically, non-mutant MAPTexpressing mouse strains either develop no CNS NFT or limited tangle pathology only when the animals are aged [12,36]. Instead, overexpression of mutant MAPT genes results in robust NFT deposition in rodent brains, which led to the wide adaptation of the mutant models to study tauopathy. It becomes clear now that transgene expression in vivo is grossly affected by the promoter of choice as well as transgene copy number and insertion sites, making it di cult to directly compare different transgenic lines. Interestingly, Gamache et al constructed genetically matched transgenic mice overexpressing WT or P301L 0N4R hTau and observed greater pathogenicity in WT hTau + animals, including exaggerated tau hyperphosphorylation and early cognitive impairment [41]. Although we studied the 2N4R isoform in vitro, the phenotypic disparity between WT and mutant hTau is highly analogous.
AD and other tauopathies may affect different brain regions, thus their pathogenesis can be mediated by different types of neurons, circuits, and other cell types under the in uence of distinct local signals. Nevertheless, our cortical neuron culture revealed a remarkable contrast between WT and mutant hTau in altering key cellular processes, highlighting a keen in uence by the identity of the tau protein itself in disease progression. Extensive biochemical and biophysical characterization has revealed structural distinctions between NFTs formed under different tauopathy conditions, of which sensitive tracers can be applied to detect their deposition by positron emission tomography [42,43]. In addition to altered binding to microtubules, WT hTau interacts with a long list of cellular proteins not observed with FTLD-associated mutants [16,38,44,45]. For example, Tracy et al performed an in-depth analysis of the tau interactome in human iPSCs and reported preferential interaction of WT tau with mitochondria and impaired bioenergetics by FTLD-tau [16]. Not so surprisingly, WT and mutant hTau also differentially affect the global transcriptome. Using tau-inducible HEK cells, Montalbano et al reported that WT hTau affected a higher number of genes than P301L mutant, including those involved in cytoskeleton-dependent processes, while P301L hTau perturbed pathways associated with reactive oxygen species [17]. In fully differentiated neurons, we similarly observed more profound in uence of WT hTau on transcriptome (Fig. 6H) and detected selective effects on "axon guidance" and "apoptosis" by WT hTau, and "ribosome" and "oxidative phosphorylation" by P301L hTau (Fig S3).
AD and tauopathies are associated with numerous nuclear irregularities, which include excessive DNA damage [46,47], altered DNA repair [21, 48], cell cycle re-entry [19,49], chromosomal defects [50] and senescence [51,52]. Long linked to brain aging and neurodegeneration, neuronal DNA damage, manifested by pH2AX + foci, accumulates early in AD brain [51,52]. Here, we detected comparable levels of pH2AX signal associated with WT and mutant hTau in primary neurons. Very low levels of tau localizes inside the nuclei of healthy neurons, where it plays a role in regulating genome stability and nucleolar function [53]. Although we did not observe increased nuclear tau, our characterization revealed close proximity of hTau with the nuclear membrane, which is accompanied by signi cant nuclear envelope deformation (Figs. 4 and 6). This nding is consistent with earlier reports on human AD as well as the aberrant interaction between pathogenic tau and nuclear pore components [54][55][56][57][58]. Nuclear envelope disruption may also impair the structures that anchor heterochromatin and cause genomic injury [59]. Our examination revealed abnormal heterochromatin relaxation and dsDNA damage response in hTau + neurons (Fig. 4). While tau-associated global chromatin relaxation has been reported in human AD and identi ed as a toxic effector of neurodegeneration in a tauopathy model, tau accumulation was also shown to trigger DDR by various studies [21, 58, 60, 61]. DNA damage is a well-known stress inducer of apoptosis [62]. In WT hTau + neurons, we observed an enrichment of the "apoptosis" pathway, the upregulation of several prototypical executors of apoptosis, such as caspase-3 and PUMA, and detected the functional involvement of caspase activation (Figs. 3, S2 and S3).
Neurons are highly sensitive to stress, which can activate multiple signaling pathways and eventually lead to cell death governed by a variety of mechanisms [63,64]. Axonal degeneration is of particular importance, as axonal cytoskeleton integrity is essential for many neuronal functions, such as long-range cargo transport and transmission of action potentials [65]. We have detected several hallmarks of axonal degeneration in WT hTau + neurons: prominent formation of axonal spheroids colocalized with activated caspase 3 and accumulation of fragmented neuro laments, accompanied by the loss of synapses and eventual release of LDH and cell death (Figs. 2 and 3). In contrast, little axonal degeneration was detectable in P301L hTau + neurons (Fig S3D). Coincidently, we found that WT hTau, hyperphosphorylated and aggregated, is abundantly distributed throughout the axons and dendrites. This is in sharp contrast with P301L hTau, which was largely restricted to the soma of neurons when phosphorylated (Figs. 2 and 6). Consistent with our in vitro ndings, axonopathy represents the most robust pathological feature exhibited by various WT hTau animal models in vivo, which was remarkably absent in mutant MAPTexpressing brains [66-69].
The activity of DLK, a neuron-speci c MAP3K, is centrally involved in both axonal degeneration and regeneration, and its pathogenic role has been implicated in neurodegeneration [35,70]. Intriguingly, DLK's downstream target JNK and its substrate c-Jun were increased in WT hTau + neurons, mirrored by the enrichment of the MAPK pathway in these cells (Fig. 5). Furthermore, we showed that pharmacological inhibition of DLK was highly effective in rescuing WT hTau + neurons, implying a prominent role played by the MAPK-DLK axis in AD-relevant neurodegeneration (Fig. 5). Axonopathy and transport de cits occur early in the pathogenesis of AD when axonal swellings are abundant near the amyloid plaques [71]. Not merely appearing in Braak I-III, phospho-tau + axonal spheroids continuously and robustly build up in later stages (Braak IV-VI) of the disease [71][72][73]. Remarkably, expression of phospho-JNK and phospho-c-Jun correlates positively with phosphorylated pathogenic tau in AD brains [35]. Notably, the MAPK pathway has been independently identi ed by recent proteomic studies as one of the most signi cantly affected processes in AD [74,75]. Hence, our ndings here offer mechanistic insight into how the DLK-JNK-c-Jun axis, one discrete branch of the MAPK cascade, directly partakes in neuron-intrinsic degeneration in AD.
We have focused on investigating the neuronal intrinsic effects of human tau in this study. As a result, the major limitation is the absence of glial cells in the model. We aim to develop in vivo tauopathy models using the same set of AAV vectors to dissect the interaction between hTau + neurons and different glial populations that play critical roles in neuronal support as well as pathogenesis [81,82]. We have only characterized the 2N4R form so far, but both 3R and 4R tau isoforms are implicated in AD. Since the hTau protein is expressed in the presence of endogenous mouse tau, we cannot rule out a potential interference by mouse tau. Therefore, we plan to elucidate the impacts of WT hTau isoforms and mutant forms in differentiated human neurons in the future.

Conclusions
Tau-mediated pathogenesis of CNS diseases is exceedingly complex in terms of its underlying mechanism, which may involve losses and/or gains of function. Pathogenic tau species formed de novo in disease are believed to play a major role in tauopathies. Despite being the disease's pathological hallmark, insoluble NFT per se may not necessarily be neurotoxic [1,2]. Multiple studies have highlighted soluble oligomeric assemblies or fragments of tau protein as the more harmful forms of abnormal tau [76][77][78][79]. In our cellular tauopathy model, hyperphosphorylated and aggregated full-length hTau was su cient to induce neurotoxicity in the absence of NFT formation. Remarkably in WT hTau + neurons, soluble pathogenic tau simultaneously initiates two major pathways from distinct compartments: DNA damage response in the perinuclear area, and MAPK-DLK activation in distal axons. We have demonstrated that these two pathways synergistically promote cytotoxicity in neurons, an exciting discovery with both scienti c and therapeutic implications. Stress signaling has been shown to couple with DLK to drive degenerative process [80]. Here our ndings identify a functional coupling of DNA damage response, a major cellular stressor, with DLK activity in AD-relevant tauopathy. The vital Page 14/35 pathogenic involvement of MAPK-DLK axis solidi es DLK as a potential therapeutic target, especially in the later stages of the disease, and provides a rationale to explore DLK inhibitors in treating AD. Availability of supporting data • In addition to results included in supplementary data 1-4, RNAseq data will be deposited at GEO and publicly available as of the date of publication.
• Any additional information related to the data reported in this paper is available from the corresponding author upon reasonable request.

Competing interests
Nothing to declare.
(A) Representative images of pH2AX and p-c-Jun (Ser73) co-staining in WT tau group. The scale bar represents 10 μm.