Introduction

Alzheimer’s disease (AD) is the most prevalent form of neurodegeneration with ominous societal impacts. The AD brain accumulates neurofibrillary tangles (NFTs) comprised of aggregated tau protein (encoded by MAPT gene), a core pathology centrally associated with the clinical manifestation1,2. Among “tauopathies”, neurodegenerative disorders that share excessive tau aggregation, primary tauopathies are caused by the genetic predisposition of various MAPT mutations3. A secondary tauopathy, AD brains harbor abundant amyloid plaques and build up 3–8 times more tau protein than non-diseased controls, leading to the development of NFTs containing non-mutated tau4,5,6,7,8.

Abundantly expressed in neurons, tau primarily associates with and stabilizes microtubules and regulates various key neuronal processes1,2. To investigate AD-relevant tau-dependent neurodegeneration, we recently developed a cellular tauopathy model by overexpressing human tau (hTau) in mouse primary neurons9. This system allowed us to comprehensively assess neuronal-intrinsic responses to wild-type hTau and enabled a direct comparison with a frontotemporal lobar degeneration (FTLD)-associated mutant form. Remarkably, we observed heightened neurotoxicity triggered by wild-type hTau and revealed joined MAPK-DLK signaling and DNA damage response in the degenerative process pertinent to AD9. Given the cellular complexity of the central nervous system (CNS), it is imperative to recapitulate the neurotoxic phenotype associated with non-mutated hTau in vivo. In the brain, glial cells, especially microglia, have been critically implicated in the pathogenesis of tauopathy as a result of mutant MAPT transgene overexpression10,11,12. At this time, the contribution of non-neuronal cells or factors to AD-relevant neurodegeneration instigated by wild-type hTau is unknown.

In this study, we established a murine tauopathy with the AAV vectors used in our in vitro study to express human wild-type or P301L tau selectively in neurons in vivo. We systemically characterized the histological, structural, cellular, and molecular changes that occurred in the brains expressing these hTau forms. Furthermore, we assessed the functional requirement of microglia in wild-type hTau-induced neurodegeneration by applying this tauopathy model to Csf1rFIRE mice, a transgenic mouse strain devoid of microglia13. This investigation confirmed a greater neurotoxic effect of wild-type over P301L hTau to the brain and demonstrated an essential role of microglia in the progression of AD-relevant tau-mediated neurodegeneration.

Results

AAV-mediated human Tau expression in the mouse brain

To establish a model of tau pathology in vivo, we infected neonatal C57BL/6J mice intracerebroventricularly with AAV vectors containing different forms of human MAPT: one with an early STOP codon preventing translation of tau protein (AAV-STOP), full-length wild-type 2N4R form (AAV-Tau), and FTLD-associated P301L 2N4R mutant (AAV-P301L tau) (Fig. 1A). We chose to express four-repeat (4R) tau to recapitulate neurodegeneration caused by tauopathy in AD, wherein 4R tau outperforms 3R tau1,2. At multiple timepoints, positive expression of transgenic hTau was detected throughout the hippocampal formation as well as in some surrounding regions, such as the cortex and subcortical regions like the thalamus (Fig. 1B). Driven by the human SYN1 promoter9, hMAPT expression was confined to neurons (Fig. 1C) and expressed at roughly physiological levels within hippocampal excitatory neurons, as compared to the level of endogenous mouse Mapt (Fig. S1A). Given the moderate dose of AAV administered, between 10 and 20% of hippocampal neurons expressed hTau, whereas cortical neurons did so at a lower percentage (Fig. 1C). Within neurons, hTau was localized mostly to the soma, with some expression in dendrites and axons as well (Fig. 1C).

Fig. 1
figure 1

(A) Schematic showing AAV-mediated overexpression of transgenic human MAPT forms in neurons in vivo. (B) Fluorescence microscopy images of brain sections from animals (6 weeks of age) overexpressing transgenic human tau (HT7), persisting into adulthood. Scale bar, 1.5 mm. CTX, cortex; HPC, hippocampus; TH, thalamus. Images representative of 3–6 animals per group. (C) Representative confocal image of HT7+ total transgenic human tau expression in the CA1 of B6 mice (scale bar, 50 μm), with quantification of frequency of hTau+ neurons in multiple brain regions. N = 6 animals. (D) Panel of antibodies targeting tau forms, including total transgenic tau (HT7), phospho-tau (pT205, AT8), and a conformational epitope (MC1), expressed by either wild-type or P301L mutant tau constructs. Representative images of 3–6 animals per group were taken from the dentate gyrus of animals from different time points. Scale bars, 50 μm.

Full size image

Human tau protein undergoes post-translational modifications at various sites to aggregate and form toxic species while eventually assembling into NFTs3,14,15,16,17. To gain understanding of pathological tau formation, we assessed multiple histological markers of hyperphosphorylated tau (p-tau), including pT205, AT8 (which detects both p-S202 and p-T205), and MC1 (which detects a conformational epitope within amino acids 312–322). Robust expression of these markers was detected in multiple subregions of the hippocampal formation, especially in the dentate gyrus (Fig. 1D), the CA1, and the CA3, with both wild-type and P301L mutant tau, and at all the time points assessed. Taken together, these results demonstrate a robust, persistent, and pathological tau expression from both wild-type and mutated MAPT sequences in multiple brain regions.

Brain pathologies induced by human Tau

To assess how wild-type and mutated tau affect the mouse brain on the macroscopic scale, we first gathered brain weight data on a large cohort of animals harvested at different time points. While no significant difference was noticed at 3 weeks, there was a significant decrease in brain weight (~ 6.7%) brought by wild-type AAV-Tau at 6 weeks, which continued at 12 weeks (Fig. 2A). By contrast, animals expressing P301L tau did not show brain weight changes at 3 weeks, and only a very small (~ 3.1%) decrease at 6 weeks (Fig. S1B), suggesting a less acute pathogenic impact by mutant tau protein despite comparable expression levels, consistent with our findings in vitro9. Of note, wild-type hTau expression in CNS had no impact on the body weight of the animals (Fig. S1C).

Fig. 2
figure 2

(A) Quantification of brain weight at various time points in animals overexpressing wild-type tau (“AAV-Tau”) compared to non-coding control vector (“AAV-STOP”). 3wk AAV-STOP, N = 9; 3wk AAV-Tau, N = 10; 6wk AAV-STOP, N = 10; 6wk AAV-Tau, N = 17; 12wk AAV-STOP, N = 24; 12wk AAV-Tau, N = 21. Unpaired t-tests were used for the 3- and 6-week timepoints, and Mann-Whitney test was used for 12-weeks, depending on results for normality testing. (B) 3D volumetric renderings of lateral ventricles (top) and hippocampi (bottom) from AAV-Tau and AAV-STOP animals at 3 months of age based on MRI images. Quantification of these volumes is shown. 12wk AAV-STOP, N = 5; 12wk AAV-Tau, N = 6. (C) Representative confocal images of NeuN+ neurons in three brain subregions, at three time points after neonatal infection. Scale bar, 50 μm. Quantification of relative neuronal density in AAV-Tau compared to AAV-STOP animals by region and age. In all regions: 3wk AAV-STOP, N = 5; 3wk AAV-Tau, N = 5; 6wk AAV-STOP, N = 6; 6wk AAV-Tau, N = 6; 12wk AAV-STOP, N = 12; 12wk AAV-Tau, N = 12. (D) Representative confocal images of DNA damage in hippocampal neurons (CA3) marked by nuclear γH2AX+ foci. Quantification of γH2AX+ neurons in brain subregions at 6 weeks of age. 6wk AAV-STOP, N = 6; 6wk AAV-Tau, N = 6.

E) Representative confocal images of DLK stress pathway in hippocampal neurons (CA3) marked by nuclear p-cJun signal. Quantification of p-cJun+ neurons in brain subregions at 6 weeks of age. 6wk AAV-STOP, N = 6; 6wk AAV-Tau, N = 6.

Full size image

Because non-mutated tau has direct biological relevance to AD, we decided to focus primarily on wild-type tau (referred to throughout as “Tau”) for the remainder of this study. To examine the overall changes in brain structure brought by tau, we performed a non-biased analysis via resting-state volumetric MRI on a cohort of mice 12 weeks after AAV infection. 3D reconstruction of structural volumes revealed a significant increase (~ 13.4%) in total lateral ventricle volume and a concomitant decrease (~ 12.2%) in total hippocampal volume (Fig. 2B). Ventricular enlargement is a prominent biomarker of AD progression and most neurodegenerative disorders18. The hippocampus represents the utmost damaged brain region in AD, and its dysfunction underlies the core feature of cognitive impairment19. Therefore, these findings signify a gross degenerative effect of wild-type human tau on the murine brain that is highly relevant to AD.

We next examined cellular and molecular changes in tau-expressing animals at the histological level. Using NeuN as a pan-neuronal marker, we measured the density of neurons in the hippocampus, specifically in CA1 and CA3 areas, and other brain regions, including layer V of the cortex. While the cortex did not show a tau-dependent decrease in neuronal density at any age, the CA1 and CA3 displayed progressive tau-dependent neuron loss beginning at 6 weeks (Fig. 2C). This observation is consistent with the abundant hTau and p-tau expression in the hippocampus region and in line with the structural changes observed with brain weight and volumetric MRI.

We previously identified a functional synergy between DNA damage response and the DLK-MAPK pathway in tau-induced neurotoxicity on cultured primary neurons9. Phosphorylated histone variant H2AX (γH2AX) is a core molecular marker of double-strand DNA damage and repair20. Staining for γH2AX revealed neuronal nuclear foci, indicating genomic DNA damage, at a higher frequency in AAV-Tau animals compared to control in both the CA3 and cortex (Fig. 2D). Phosphorylated by c-Jun N-terminal kinases (JNKs), c-Jun protein acts as a potent transcriptional factor downstream of MAP kinase signaling21. In a similar manner, CA1, CA3, and cortical neurons in AAV-Tau animals expressed a higher frequency of nuclear p-cJun, indicating activation of the DLK-MAPK injury signaling (Fig. 2E). Interestingly, these pathways were elevated at 6 weeks, but not at 3 weeks (Fig. S1D), suggesting a delayed molecular response to early tau expression in the brain. Taken together, these results demonstrate the toxic effect of wild-type human tau in mouse brains, which potently infringes on brain structure, reduces neuronal density in the hippocampus, and activates intrinsic pathways involved in genomic stress and injury response.

Molecular alterations in hippocampal neurons brought on by non-mutated hTau

To investigate specific molecular changes in neuronal populations during pathology induced by tau, we performed single-nucleus RNA sequencing (snRNA-seq) on pooled hippocampi from animals receiving AAV-STOP or AAV-Tau at 6 weeks of age (n = 3 per group). A total of 39,861 cells were sequenced. After preprocessing steps including quality control and normalization (see Methods), 39,568 cells with 22,374 genes were kept for the downstream analysis. Clustering with Seurat, we detected 15 distinct cell clusters, visualized by Uniform Manifold Approximation and Projection (UMAP) (Fig. S2A). Using established marker genes22,23, six major cell types were identified in the tissue, the most prominent among them by cell counts being neurons. Other cell types in the dataset included oligodendrocytes, oligodendrocyte precursor cells (OPCs), astrocytes, microglia, and vascular cells (Fig. 3A).

Fig. 3
figure 3

(A) UMAP plot of single nuclei isolated from hippocampi of AAV-infected animals. Cells clustered into six groups based on genome wide expression patterns. (B) Cell counts, expressed as a proportion, in hippocampi of 6-week-old animals injected with AAV-STOP and AAV-Tau. (C) UMAP plot of single neuronal nuclei isolated from hippocampi of AAV-infected animals, showing 10 clusters of neurons. (D) Cell counts of neuronal subtypes, expressed as a proportion, among the 10 neuron clusters identified in hippocampi of 6-week-old animals injected with AAV-STOP and AAV-Tau. (E) Top 10 pathways (KEGG48,49,50 most significantly regulated in cluster 0 neurons from AAV-Tau animals compared to AAV-STOP. (F) Top 10 pathways (KEGG48,49,50 most significantly regulated in cluster 1 neurons from AAV-Tau animals compared to AAV-STOP.

Full size image

Remarkably, AAV-Tau hippocampus tissue displayed a significant reduction in the neuron cell population, by percent of total cells, when compared with AAV-STOP (Fig. 3B), a finding consistent with our structural and histological analyses (Fig. 2). We further interrogated the neuronal cells and came up with 10 subclusters (Fig. 3C). Based on marker gene profiles, they fell into four subtypes: excitatory neurons from CA1 (Ex_CA1), excitatory neurons from CA2/3, excitatory neurons from the dentate gyrus (Ex_DG), and inhibitory neurons (Fig. S2B). Populational analysis revealed selective loss of Cluster 0 and Cluster 1 neurons, corresponding to Ex_DG and Ex_CA1, respectively (Fig. 3D). Pathway analysis on the differentially expressed genes between these neurons showed remarkable upregulation of multiple disease-related pathways such as “Parkinson’s Disease,” “Alzheimer’s Disease”, and “Pathways of neurodegeneration,” as well as cell death process, such as “Necroptosis” in the affected neuronal subsets (Fig. 3E, F). In addition, several calmodulin genes, such as Calm1, Calm2, and CamK1d, were differentially affected by wild-type tau in these hippocampal neurons, implicating a profound effect of tau on neuronal calcium signaling (Fig. S2C, D). Thus, snRNA-seq analysis confirmed a severe loss of hippocampal neurons by wild-type tau and identified molecular processes relevant to neurodegeneration in the neuronal populations.

Glial responses to non-mutated hTau

It is well known that tau pathology is associated with a neuroinflammatory response, manifesting in the brain as reactive gliosis10,11. GFAP+ astrocytes showed subtle increases, most significantly in the CA1 field of the hippocampus, as early as 3 weeks post-infection with AAV-Tau (Fig. 4A). While the complement pathway is activated in various neurological diseases, the central component C3, a product of reactive astrocytes, has been functionally implicated in the pathogenesis of tauopathy models overexpressing mutant hTau24,25. We found that hippocampal astrocytes displayed higher C3 expression in tau-expressing brain at 6 weeks (Fig. 4B).

Fig. 4
figure 4

(A) Representative confocal images and quantification of GFAP+ astrocytes in brain subregions at 3 and 6 weeks of age in AAV-Tau compared to AAV-STOP animals. Scale bar, 100 μm. 3wk AAV-STOP, N = 5; 3wk AAV-Tau, N = 5; 6wk AAV-STOP, N = 6; 6wk AAV-Tau, N = 6. (B) Confocal images showing co-localized C3+ signal within GFAP+ hippocampal astrocytes at 6 weeks of tau expression. Insets show isolated C3 channel. Scale bar, 10 μm. Quantification of relative percent occupancy of C3+ signal within GFAP+ astrocytes. AAV-STOP, N = 4; AAV-Tau, N = 5. (C) Representative confocal images and quantification of Iba1 + microglia in brain subregions at 3 and 6 weeks of age in AAV-Tau compared to AAV-STOP animals. Scale bar, 100 μm. 3wk AAV-STOP, N = 5; 3wk AAV-Tau, N = 5; 6wk AAV-STOP, N = 6; 6wk AAV-Tau, N = 6. (D,E) Confocal images showing co-localized Clec7a+ (D) or CD68+ (E) signal within Iba1+ hippocampal microglia at 6 weeks of tau expression. Scale bars, 10 μm. Quantification of relative percent occupancy of given signals within Iba1+ microglia. AAV-STOP, N = 4; AAV-Tau, N = 5. (F) A representative confocal image of several CA1 microglia expressing CD68+ lysosomal vesicles. AT8+ tau aggregates can be seen within these vesicles (arrowhead inside inset). The highlighted vesicle is further enlarged and presented as an orthogonal projection at the right. All scale bars, 10 μm.

Full size image

Microglia are brain-resident immune phagocytes that fulfill a multitude of critical roles in the CNS26,27. Fairly subtle increases in Iba1+ microglia were detected at 3 weeks of tau expression but more dramatic increases were evident by 6 weeks (Fig. 4C). Activated microglia are a feature of the brains of tauopathy models overexpressing mutant hTau10,11. Likewise, we observed elevated expression of microglial activation markers, namely Clec7a (Fig. 4D) and CD68 (Fig. 4E), in the hippocampus by 6 weeks. We also found examples of AT8+ p-tau localized within CD68+ microglial vesicles (Fig. 4F), indicating a continuous engulfment of tau by microglia. Our examination thus suggests an early astrocyte reaction, followed by microglial activation, in response to wild-type human tau expression in the rodent brain.

Resistance to hTau-induced hippocampal neurodegeneration in mice lacking microglia

Depending on specific conditions, microglia activation may be reactive, protective, restorative, or pathogenic in the brain26. To examine the functional role of microglia in AD-relevant tauopathy, we adopted the Csf1rFIRE line (referred to here as “FIRE”), which completely lacks microglia throughout the lifespan due to the ablation of a super-enhancer element within the Csf1r gene (Fig. 5A)13. After infecting with AAV-Tau, we observed no significant brain weight loss in FIRE mice at either 6 or 12 weeks (Fig. 5B), implying that microglia may be involved in neurodegeneration caused by non-mutated human tau. Indeed, the neuronal density decreases seen in tau-expressing wild-type animals, especially in the CA1, were abrogated in the brains of FIRE mice (Fig. 5C). These findings point to a prerequisite involvement of microglia in tau-associated neurodegeneration.

Fig. 5
figure 5

(A) Representative confocal images of Iba1 staining in the dentate gyrus of B6 wild-type mice and FIRE mice. Scale bar, 50 μm. (B) Quantification of brain weight at various time points in B6 and FIRE animals overexpressing wild-type tau compared to non-coding control vector. 6wk B6 AAV-STOP, N = 10; 6wk B6 AAV-Tau, N = 17; 6wk FIRE AAV-STOP, N = 6; 6wk FIRE AAV-Tau, N = 7; 12wk B6 AAV-STOP, N = 24; 12wk B6 AAV-Tau, N = 21; 12wk FIRE AAV-STOP, N = 10; 12wk FIRE AAV-Tau, N = 14. One-way ANOVA and Kruskal-Wallis test were used for the 6- and 12-week time points, respectively, depending on results for normality. (C) Representative confocal images of NeuN+ CA1 neurons in B6 and FIRE mice 6 weeks post-infection with AAV-STOP or AAV-Tau. Scale bar, 50 μm. Quantification of relative neuronal density in B6 and FIRE animals overexpressing wild-type tau compared to non-coding control vector animals by region at 6 weeks of age. B6 AAV-STOP, N = 6; B6 AAV-Tau, N = 6; FIRE AAV-STOP, N = 6; FIRE AAV-Tau, N = 7. (D) Quantification of p-cJun+ neurons in brain subregions at 6 weeks of age in both B6 and FIRE mice expressing tau or control vector. B6 AAV-STOP, N = 6; B6 AAV-Tau, N = 6; FIRE AAV-STOP, N = 6; FIRE AAV-Tau, N = 7.

Full size image

To assess the impact of microglial processes on tau-induced alterations, we examined a another marker of neuronal processes, namely p-cJun, which is a key molecule in the DLK-MAPK stress pathway. While increased p-cJun was observed elevated in B6 animals expressing tau (Fig. 2E), despite having no microglia, FIRE mice expressing tau displayed similar levels of neuronal p-cJun (Fig. 5D). This result suggests that microglia-mediated neuronal loss and DLK-MAPK activation due to tau expression are likely independent processes in the hippocampus. Taken altogether, these data suggest that microglia may play a crucial role in the hippocampal neurodegenerative process prompted by wild-type hTau.

Tau- and microglia-modulated neurodegeneration in the thalamus

During examination, we noticed a particularly concentrated expression of transgenic tau in the thalamus (Fig. 1B). In fact, whereas the percentage of neurons in other areas that expressed the transgene were normally under 20% (Fig. 1C), thalamic neurons expressed tau at a level twice as frequently (Fig. 6A). Within the thalamus, tau appeared to be expressed almost exclusively in excitatory neurons (marked by CAMKIIα), whereas inhibitory neurons (marked by PV) were devoid of human tau (Fig. 6B). Consistent with greater tau expression, the frequency of neurons displaying DNA damage in the form of γH2AX expression, and those with elevated p-cJun, were both significantly elevated (Fig. 6C), surpassing other regions (Figs. 2D, E). In addition, GFAP+ astrocytes were far more drastically induced in the thalamus, while the level of Iba1+ microglia was largely comparable to other regions (Fig. 6D). However, no significant change of neuronal density was detected in the thalamus (Fig. 6E), a contrast to the hippocampal regions (Fig. 2C). Thus, thalamic neurons abundantly express tau and display heightened stress response amidst significant gliosis, yet without undergoing neurodegeneration in wild-type mice.

Fig. 6
figure 6

(A) Representative confocal images of thalamus in AAV-Tau and AAV-STOP animals at 6 weeks of age. Scale bar, 50 μm. Quantification of % HT7+ neurons in the thalamus of 6-week-old AAV-Tau animals. (B) Distribution of transgenic human tau (HT7) among neuronal subtypes, including excitatory (CamKIIα+) and inhibitory (PV+) neurons in the cortex and thalamus. Scale bar, 30 μm. (C) Quantification of γH2AX+ and p-cJun+ neurons in the B6 thalamus at 6 weeks of age in AAV-Tau compared to AAV-STOP animals. AAV-STOP, N = 6; AAV-Tau, N = 6. (D) Quantification of GFAP+ astrocytes and Iba1+ microglia in the B6 thalamus at 3 and 6 weeks of age in AAV-Tau compared to AAV-STOP animals. 3wk AAV-STOP, N = 5; 3wk AAV-Tau, N = 5; 6wk AAV-STOP, N = 6; 6wk AAV-Tau, N = 6. (E) Representative confocal images of NeuN+ neurons in thalamus, at two time points after neonatal infection with AAV-STOP or AAV-Tau. Scale bar, 50 μm. Quantification of relative thalamic neuronal density. 3wk AAV-STOP, N = 5; 3wk AAV-Tau, N = 5; 6wk AAV-STOP, N = 6; 6wk AAV-Tau, N = 6. (F,G) Quantification of relative thalamic neuronal density (F) and p-cJun+ neurons (G) at 6 weeks in FIRE animals expressing AAV-Tau compared to AAV-STOP. AAV-STOP, N = 6; AAV-Tau, N = 7. (H) Quantification of HT7+ neuron percentage in thalamus of B6 and FIRE mice expressing AAV-Tau. AAV-STOP, N = 6; AAV-Tau, N = 7.

Full size image

Given the profound ongoing gliosis, we sought to determine the effects of microglia on tau-induced cellular changes in the thalamus of FIRE mice. Surprisingly, thalamic neuronal cell density was significantly and dramatically reduced in the absence of microglia at 6 weeks (Fig. 6F), differing from the observation in wild-type animals (Fig. 6E). Accompanying this, the elevation of neuronal stress and injury marker p-cJun was exacerbated in the thalamic neurons of FIRE mice (~ 25%, Fig. 6G) compared to wild-type animals (~ 15%, Fig. 6C). Additionally, the overall population of tau-expressing neurons was significantly increased in thalamus in the absence of microglia (Fig. 6H). Altogether, we have detected a prominent wild-type hTau expression in the thalamic neurons and uncovered an unexpected protective role of microglia in restraining tau-mediated neurodegeneration in the thalamus.

Discussion

By comparing neuronal expression of non-mutated versus FTLD-related human tau in the mouse brain, we have demonstrated the capacity of wild-type tau to induce neurodegeneration in vivo. The findings align with our previous observations from primary neurons and highlight an AD-relevant tau-mediated pathogenesis that has not been sufficiently recognized and investigated. Beyond neuronal intrinsic changes, our results further demonstrated that glial responses accompanied the course of tauopathy with wild-type hTau and revealed contrasting roles of microglia in mediating or restraining neurodegenerative processes from different brain regions.

Unlike primary tauopathies, AD patients possess higher levels of tau proteins in their brains without genetic mutations in the MAPT gene. However, rodents overexpressing FTLD-related tau mutations are commonly used to model clinical AD tau pathology28. Our finding that wild-type hTau induces severe degenerative sequelae over mutant tau in the brain, especially the hippocampal region, is consistent with recent reports. Middle-aged mice expressing AAV-mediated wild-type hTau exhibited more profound brain atrophy than those expressing P301L mutant tau despite the latter accumulating higher tau protein aggregation29. Further, a direct comparison of transgenic mice genetically matched in overexpressing wild-type versus P301L hTau has revealed a greater pathogenicity of wild-type hTau, which was accompanied by exaggerated tau hyperphosphorylation and early cognitive impairment30. Altogether, these studies highlight the importance to establish an AD-relevant tauopathy model and delineate the underlying mechanism that is likely disease-specific.

AD and other tauopathies affect distinct brain regions, which results in diverging clinical presentations. In AD, tau pathology is commonly believed to initiate in the entorhinal cortex, spread to limbic regions, which includes the hippocampus, and then migrate to the neocortex31,32. Our wild-type hTau model displays not only neuronal tau expression in the hippocampus but also, together with signs of brain atrophy, selective neuronal loss in the hippocampal areas. Beyond the brain regions mentioned above, thalamic neuropathology has been implicated in early AD33,34. Most recently, Sárkány et al. reported that tau filaments accumulated early and progressively inside the glutamatergic neurons at the anterodorsal thalamus of AD patients35. Here, we have uncovered abundant hTau expression, enhanced stress response, and elevated gliosis in the thalamus, a regional pathology not present in transgenic models. As increasingly recognized, the thalamus critically participates in memory and cognitive functions via various network connections, such as the Papez circuit33,36. Therefore, our hTau model offers a unique opportunity to dissect various aspects of AD-relevant tau pathology.

In primary neuronal cultures, non-mutated hTau triggers neurotoxicity via synergistic MAPK-DLK signaling and DNA damage response9. Yet, characterization of in vivo neuronal wild-type hTau expression has revealed intriguing complex molecular and cellular events, both neuron-intrinsic and neuron-extrinsic. Wild-type hTau expression was sufficient to cause the accumulation of phosphorylated tau, together with MAPK-DLK signaling and DNA damage response in hippocampal neurons. Our snRNAseq analysis validated the selective hippocampal neuronal loss and revealed the enrichment of pathways related to neurodegenerative diseases in the mostly affected excitatory neuron populations. Tau upregulated the necroptosis pathway in both Ex_DG and Ex_CA1 neurons that were significantly diminished in numbers due to wild-type hTau expression. Of note, necroptosis was recently shown to play a central role in mediating the loss of human tangle+ neurons in an AD xenograft model37. Moreover, dysregulated neuronal calcium signaling by hTau is consistent with earlier findings that wild-type hTau induces memory deficits by calcineurin-mediated inactivation of nuclear CaMKIV/CREB signaling38.

Glia cells are increasingly recognized to play crucial roles in a wide range of neurological diseases, especially neurodegeneration. In the hippocampus expressing wild-type hTau, we have detected an early astrocyte response followed by a subsequent microglial activation. Moreover, a robust glial reaction to wild-type hTau manifested in the thalamus. These findings echo the early and persistent glial responses in transgenic mouse models expressing mutant tau39. By examining FIRE mice, we discovered that microglia seemingly exert contrasting regional functional influences: while wild-type hTau-mediated hippocampal neuronal loss decreased in the FIRE brain, neuronal loss only incurred in the thalamus in the absence of microglia. This process, in part, is comparable to mutant hTau-mediated hippocampal degeneration fostered by aberrantly activated microglia10,12,40.

However, microglia’s role in brain pathophysiology is highly complex. In the lifelong absence of microglia, FIRE mice offer a novel and valuable tool to investigate the function of these brain immune cells. Intriguingly, spontaneous age-related thalamic degeneration, such as atrophy and neuronal loss, has been a prominent feature of aged FIRE mice41. Crossed with a xenotolerant strain, adult FIRE mice also display a selective pathology in the thalamus42. Comparatively, the FIRE mice we examined were fairly young – up to 6 weeks old. Given the high tau load and stress response in our model, these observations indicate a key protective function of thalamic microglia in maintaining the viability of resident neurons, which otherwise would be particularly vulnerable to neuropathological changes during brain aging or AD.

Taken together, we have established an AD-relevant tauopathy model in young mice and uncovered neuron-intrinsic and -extrinsic responses associated with wild-type hTau-mediated neurodegeneration. These findings may inspire further research on the regional functions of microglia and the delineation of protective versus pathogenic roles of brain glia cells.

Limitations

Our work introduces a new murine model of tauopathy, yet several caveats merit consideration. First, all experiments were performed in very young mice. Although the neonatal AAV-mediated transduction strategy ensures brainwide transgene expression, it also raises the possibility that early expression of tau alters neurodevelopment. A preliminary survey of doublecortin, an established marker of precursor and immature hippocampal neurons, showed no obvious difference between AAV-STOP and AAV-Tau cohorts (data not shown). Nevertheless, subtler developmental effects cannot be excluded and should be considered. Second, our microglial ablation approach relies on FIRE mice, which lack microglia throughout life, which is essential because tau expression begins shortly after birth in this model. The FIRE line used here was maintained on a mixed CBA/J × C57BL/6J background 13. Attempts to backcross FIRE onto a pure C57BL/6J background have resulted in poor health and reduced lifespan 43. Thus, we opted for the original FIRE mice and compared them with C57BL/6J controls, which renders a somewhat genetic background mismatch. Finally, we observed more apparent neurodegeneration associated with non-mutated human tau than with P301L tau at an young age. Extended longitudinal studies and inclusion of additional disease-associated tau mutations will be important to define how different tau variants drive pathology over time. Because AD and related dementias are highly age-dependent, a particularly relevant next step is to express human tau in adult or aged hosts and evaluate its impacts thoroughly.

Methods

Mice

C57BL/6J mice were obtained from the National Institute on Aging (NIA). FIRE mice in a mixed C57B/6:CBA background were generated by Clare Pridans13. Mice with ad libitum access to food and water and were housed in mixed-genotype groups of 3–5 per cage under specific pathogen-free conditions and standard light/dark cycle. Both male and female mice were used in experiments. Mice were analyzed at specific timepoints as noted in the study, and precise ages of all animals used in experiments are listed in respective figure legends. All animal procedures were performed in accordance with NIH guidelines and the study received approval from the Institutional Animal Care and Use Committee at University of Texas Health Science Center- Houston. This study is reported in compliance with ARRIVE guidelines.

AAV delivery

The generation and preparation of all AAV vectors used in this study were described previously9. For in vivo infection, frozen stock solutions of AAV particles were thawed and diluted with vehicle composed of sterile PBS with 10% trypan blue (for visual confirmation of correct injection) to the appropriate concentration and kept on ice. Neonatal mouse pups were chilled on an ice-cold aluminum block until fully cryo-anesthetized. AAVs were delivered into the intracerebroventricular spaces of neonatal mouse pups (on p0, the day of birth) using a Hamilton Gas-tight syringe (Hamilton Company, Reno, NV) fitted with a steel 32-gauge needle (point style 4 with a 12° bevel). We used a dose of 5 × 109 gc in 2 µL vehicle into each hemisphere of the brain, for a total dose of 1 × 1010 gc in 4 µL vehicle per animal. Injected neonates were allowed to regain body temperature on a heating pad before being placed back into parent cages. This method of cryo-anesthesia, which is approved for use in neonatal mice under 10 days old in AVMA Guidelines for the Euthanasia of Animals 2020 Edition, resulted in 100% survival and recovery rate for all animal cohorts in this study.

Volumetric MRI procedure

In vivo MRI was performed on a 7T Bruker BioSpec MRI system (Bruker BioSpin, Billerica, MA) at the MRI core facility of UTHealth Houston. Specifically, T2-weighted images were acquired using the RARE sequence with the following imaging parameters: TR/TE: 6000/60 ms, in-plane resolution: 100 × 100 µm2, slice thickness: 0.5 mm, and a total of 30 slices to cover the whole brain. Brain structures of interest were manually segmented using ITK-SNAP software (version 3.8.0, https://www.itksnap.org)44. Briefly, appropriate structures were delineated on each 2D image slice in which the structure appeared, and 3D volumes were calculated based on this segmentation, which yielded a total 3D volume in mm3 for each structure.

Single-nuclei RNA sequencing

The mouse hippocampus was dissected and snap-frozen at -80 °C. Tissue (100 mg) was powdered on dry ice using a scalpel and transferred to a 1.5 mL microtube. Homogenization was performed by adding 0.5 mL of freshly prepared Homogenization Buffer (320 mM Sucrose, 5 mM CaCl2, 3 mM Mg(Ac)2, 10 mM Tris pH 7.8, 0.1 mM EDTA, 0.1% NP40, 0.1 mM Protease inhibitor, 1 mM Beta-mercaptoethanol, 1 U/µL RNase inhibitor), followed by gentle pipetting and incubation on ice for 10 min. The homogenate was passed through a 30 μm strainer and centrifuged at 300 g for 3 min at 4 °C. The pellet was resuspended in 0.5 mL Homogenization Buffer and mixed with 0.5 mL Upper Gradient Centrifugation Buffer (50% OptiPrep Gradient, 5 mM CaCl2, 3 mM Mg(Ac)2, 10 mM Tris pH 7.8, 0.1 mM Protease inhibitor, 1 mM Beta-mercaptoethanol, 1 U/µL RNase inhibitor). The mixture was gently layered on top of 1 mL of Lower Gradient Centrifugation Buffer (29% OptiPrep Gradient, 1 U/µL RNase inhibitor) in a 2 mL microtube.

The gradient was centrifuged at 7,200 g for 15 min at 4 °C. The supernatant was carefully discarded. The nuclei pellet was resuspended in the PBS containing 0.04% BSA and 1 U/µL RNase inhibitor to achieve a concentration of 700–1200 nuclei/µL.

Nuclei (∼20,000 per sample) were loaded onto a 10x Genomics Next GEM chip G. Libraries were prepared using the Chromium Next GEM Single Cell 3ʹ Library and Gel Bead kit v3.1 (10x Genomics), following the manufacturer’s protocol, and sequenced on an Illumina NovaSeq 6000. The sequencing data was generated by the UTHealth Houston Cancer Genomics Core and deposited into the GEO repository (GSE285945). Token ID: stwfqgoavjcxzuf Contact: Qiang.wang@uth.tmc.edu.

Single-cell RNA preprocessing and cell type annotation

The raw scRNA-seq data were processed using CellRanger v4.0 (https://www.10xgenomics.com), which performed initial data alignment, filtering, and UMI counting. The mouse reference genome (mm10) was utilized for accurate mapping of sequencing reads to genes. After preprocessing with CellRanger, the resulting gene-barcode matrices were imported into R for downstream analyses. The quality control and preprocessing steps were performed using Seurat (v4.4.0), and batch effect removal was performed using Harmony45,46. To preprocess the scRNA-seq data, we applied stringent quality control thresholds to ensure the reliability and biological relevance of the downstream analysis. Cells were filtered based on the following criteria: a minimum of 200 detected genes (nFeature_RNA > 200) to exclude low-quality or empty droplets, and a total RNA count between 1,000 and 20,000 (1000 < nCount_RNA < 20000) to eliminate cells with extremely low or high RNA content, which could indicate technical artifacts or doublets. Additionally, cells with a mitochondrial gene percentage exceeding 5% (percent.mt < 5) were excluded to remove cells with potential stress or apoptosis signals.

Cell clustering for the scRNA-seq data was performed using the Seurat package in R, utilizing normalized gene expression profiles to identify distinct cell populations. The resolution was set as 0.1 for the clustering. Cell type annotation was carried out using the UCell package, which computes gene set enrichment scores based on marker genes curated from established literature and databases47. Each cluster was assigned to a corresponding cell type based on the highest enrichment score.

Differential gene expression analysis and enrichment analysis for each cell type

We performed differential expression analysis to identify aging-related genes across different mouse types (i.e., AAV-Tau vs. AAV-STOP) using the ‘FindMarkers’ function in the Seurat package for each cell type. Differentially expressed genes (DEGs) were defined as those with an adjusted p-value < 0.05 and |logFC| greater than the threshold calculated as mean(abs(logFC)) + 2 × sd(abs(logFC)).

To investigate the potential biological roles of these DEGs, we performed KEGG48,49,50 pathway enrichment analysis using the clusterProfiler v4.10.1 package51. Separate analyses were conducted for upregulated and downregulated DEGs within each cell type. Only pathways with an adjusted p-value < 0.05 were considered significant.

Identification of subpopulation for major cell types

After the cell type annotation for the major cell types, we extract the certain cell type for the subpopulation analysis. Similarly with the major cell type annotation, Seurat and UCell packages were used in this step45,47.

Single-cell trajectory reconstruction and analysis

Single-cell pseudotime trajectories were constructed with Monocle (version 2.30.1, https://bioconductor.org/packages/release/bioc/html/monocle.html)52. Briefly, we first selected the cell type and the subpopulations with their gene expression matrix. Monocle applies reversed graph embedding, a machine learning approach, to construct a parsimonious principal graph that reduces high-dimensional expression profiles into a low-dimensional space. Each cell is projected onto this space and organized into a trajectory with distinct branch points with a pseudotime value.

Immunofluorescence

Immunofluorescence was performed as previously described53. Mice were perfused with ice-cold saline (0.9% NaCl) after deep anesthesia with ketamine/xylazine, and brains were extracted, fixed overnight at 4 °C in 4% paraformaldehyde (Santa Cruz, cat# sc-281692), and dehydrated in 30% sucrose until sectioning. Brains were sectioned into 30 μm tissue sections using a freezing microtome, and tissue was stored in cryoprotectant at -20 °C. For staining, floating sections were washed in phosphate-buffered saline (PBS) and blocked for 1 h at RT in a blocking buffer of 10% normal donkey serum (Millipore, cat# S30-100ML) and 1% Triton X-100 in TBS. Primary antibodies were diluted in blocking buffer and applied to sections overnight at 4 °C. Tissues were washed with PBS-T (PBS with 0.1% Tween-20) three times, then incubated with fluorescent secondary antibodies diluted in blocking buffer for 1 h at RT. After final washing in PBS-T, sections were mounted on glass slides, allowed to dry, and coverslipped with ProLong Glass Antifade mountant (Life Technologies, cat# P36982).

Image quantification

All imaging in this study was performed with a Leica SP8 system as previously described53. For all quantitative analyses in this study, multiple images were used as technical replicates from each animal. For counting of neurons, NeuN antibody was applied to tissues to visualize individual neuronal nuclei. In some experiments, other markers were co-labelled with NeuN, such as p-cJun. Single-plane confocal images from regions of interest were subjected to manual counting in the Leica confocal software (version 3.7.5, https://www.leica-microsystems.com), and data were recorded for total NeuN+ cells per field, and those with and without co-expression of markers pertinent to the experiment.

For area fraction calculation, Z-stacks were converted to maximum intensity projections and loaded into ImageJ (NIH) for analysis. Images were thresholded manually to overlap fluorescence signals, and percent area was measured. For assessment of hippocampal layer thickness, confocal images of NeuN+ cell layers (CA1 and CA3) were acquired using a 20X lens. For each field of view, the thickness of the layer at three to five points was collected and averaged, and multiple sections were imaged per animal.

Expression of various markers by specific cell types (i.e. microglial CD68, microglial Clec7a, astrocytic C3) were quantified on Imaris 10.0 software (Bitplane, https://imaris.oxinst.com/) using the “Co-loc” feature. Briefly, Z-stacks were captured by confocal microscopy, and fluorescence channels for cellular markers (i.e. Iba1, GFAP) were used to create a mask within which the occupancy of the target marker was calculated as a percentage of the 3D space. Within all co-localization experiments, consistent thresholds were used for each signal to prevent biased measurements.

Quantification and statistical analysis

All data in bar plots are presented as means ± s.e.m. When used, individual datapoints in plots represent mean values of technical replicates per animal analyzed. N is provided in each figure legend. GraphPad Prism (v10.0.0, https://www.graphpad.com) was used for all statistical analyses and graph plotting. All data was tested for normality using the Shapiro-Wilk test. For normally distributed data, differences between two groups were analyzed by two-tailed Student’s t-tests (with Holm-Šídák’s post-hoc multiple comparisons tests, where applicable), and differences between three or more groups were analyzed by one-way ANOVA with Šídák’s post-hoc multiple-comparisons tests. For comparisons including non-normally distributed data, differences between two groups were analyzed by Mann-Whitney U Tests (with Holm-Šídák’s post-hoc multiple comparisons tests, where applicable), and differences between three or more groups were analyzed by Kruskal-Wallis tests with Dunn’s post-hoc multiple-comparisons tests. P or Padj values less than 0.05 were considered significant, and those over 0.05 were considered non-significant. All exact P values are listed within plots. Female and male mice were used for all experiments. All micrographs shown are images representative of multiple replicates as noted in legends. No statistical methods were used to predetermine sample sizes, but similar publications were used as guidelines. Biorender.com was used to generate schematics within the figures.