Glucocorticoid stress hormones stimulate vesicle-free Tau secretion and spreading in the brain

Chronic stress and elevated levels of glucocorticoids (GCs), the main stress hormones, accelerate Alzheimer’s disease (AD) onset and progression. A major driver of AD progression is the spreading of pathogenic Tau protein between brain regions, precipitated by neuronal Tau secretion. While stress and high GC levels are known to induce intraneuronal Tau pathology (i.e. hyperphosphorylation, oligomerization) in animal models, their role in trans-neuronal Tau spreading is unexplored. Here, we find that GCs promote secretion of full-length, vesicle-free, phosphorylated Tau from murine hippocampal neurons and ex vivo brain slices. This process occurs via type 1 unconventional protein secretion (UPS) and requires neuronal activity and the kinase GSK3b. GCs also dramatically enhance trans-neuronal Tau spreading in vivo, and this effect is blocked by an inhibitor of Tau oligomerization and type 1 UPS. These findings uncover a potential mechanism by which stress/GCs stimulate Tau propagation in AD.


Introduction
Stressful life events and high circulating levels of glucocorticoids (GCs), the primary stress hormones, are known risk factors for Alzheimer's disease (AD) [1][2][3] . Indeed, epidemiological and clinical studies suggest that prolonged psychosocial stress signi cantly elevates AD risk, and high GC levels are associated with faster cognitive decline in AD patients [1][2][3][4][5] . Moreover, stress interacts with genetic risk factors to hasten the onset of AD symptoms and pathology in both animal models and humans 6-12 . Stress and AD appear to share a common pathological driver: the microtubule-associated protein Tau. Not only do stress and high GC levels trigger Tau pathology similar to that seen in AD brain tissue (i.e., Tau hyperphosphorylation and aggregation) [13][14][15] , but Tau depletion is protective against both amyloid beta-and stress-induced neurotoxicity and cognitive impairment in animal models [15][16][17] , indicating Tau's essential role as a mediator of neurodegeneration in the context of AD and chronic stress.
A key feature of Tau pathology in AD is its stereotypical spreading pattern between anatomically connected brain regions (entorhinal cortex to hippocampus to prefrontal cortex) 18 . This spreading is highly correlated with the severity of cognitive impairment in AD patients and appears to be a major driver of AD progression [18][19][20] . Given the relationship between Tau propagation and clinical AD symptoms, there has been tremendous interest in elucidating the mechanisms of Tau secretion and spreading in the brain.
Numerous studies have shown that Tau is secreted from neurons in extracellular vesicles, including ectosomes that derive from the plasma membrane and exosomes that derive from multivesicular endosomes of the endolysosomal pathway, and also as vesicle-free protein [21][22][23] . While vesicle-mediated mechanisms of Tau spreading have been a focus of study for over a decade 21,24,25 , the vast majority of Tau secreted by neurons (~ 90%) is vesicle-free 23,25−32 , and considerably less is known about this mode of secretion and its contribution to pathogenic Tau propagation. Similarly, although chronic stress and high GC levels are known to induce Tau pathology in the hippocampus and cortex, precipitating synaptic loss and behavioral impairment in animal models (i.e. anxiety, anhedonia, learning/memory de cits) 14,15,33 , it is unclear whether or how stress/GCs stimulate the spreading of Tau pathology between these brain regions.
In the current study, we investigate the effects of GCs on neuronal Tau secretion and spreading in murine hippocampal neurons, brain slices, and in vivo hippocampus. We nd that GCs induce secretion of vesicle-free Tau through type 1 unconventional protein secretion (UPS), in an activity-and glycogen synthase kinase 3b (GSK3b)-dependent manner. Moreover, GC administration stimulates Tau spreading through the hippocampus, and this process is prevented by inhibiting Tau aggregation and type 1 UPS with the catechin EGCG. Together, these ndings demonstrate that elevated GC levels promote Tau propagation, and suggest a mechanism by which stress/GCs speeds cognitive decline in AD.

Results
To determine whether GCs stimulate neuronal Tau secretion, we measured extracellular Tau levels by immunoblot and/or ELISA in three preparations: 1) media from 14 day in vitro (DIV) murine hippocampal neurons treated for 48 hours with vehicle control, the synthetic GC dexamethasone (DEX), or DEX + GC receptor (GR) antagonist mifepristone (MIF) (Fig. 1A-D, I), 2) arti cial cerebrospinal uid (ACSF) from ex vivo murine brain slices of 4 months old mice, perfused for 4 hours with vehicle, DEX, or DEX + MIF ( Fig. 1E-H, J), and 3) CSF from 4-5 months old mice administered vehicle, DEX, or DEX + MIF for 15 days (Fig. 1K). The e cacy of DEX treatment was con rmed by immunoblotting hippocampal lysates for phospho-GR and by immunostaining for phospho-and oligomeric Tau (Fig. S1A-E), as in our recent study 34 . In all three preparations, DEX signi cantly increased Tau concentration compared to vehicle and DEX + MIF (Fig. 1A-K). This increase in extracellular Tau did not result from cell death or disruption of plasma membrane integrity, as lactate dehydrogenase (LDH) levels were unaltered by DEX +/-MIF in both the in vitro and ex vivo preparations (Fig. 1L), and the abundant cytoskeletal proteins actin and tubulin were not detected in these uids (Fig. 1A, E). We also found that extracellular Tau was predominantly fulllength, phosphorylated at multiple sites, as indicated by immunoreactivity for AT8 (Ser202/Thr205) and PHF1 (Ser396/Ser404) antibodies ( Fig. 1A-C, E-G), and vesicle-free rather than associated with extracellular vesicles (EVs), which were depleted from media and ACSF by a well-established centrifugation procedure (Fig. S1F-H) 35,36 . Extracellular Tau levels were similarly increased in media containing cortical and hippocampal brain slices from mice subjected chronic unpredictable stress (CUS) compared to control conditions (Fig. S1I), demonstrating that CUS and GC exposure have comparable stimulatory effects on Tau secretion.
Since stress/GCs promote Tau accumulation, these ndings could re ect similar levels of Tau secretion from a larger intraneuronal pool. To determine whether GCs alter the fractional amount of Tau secreted from neurons, we measured Tau concentration in media versus neuronal lysate for control and DEX conditions, using ELISA kits to detect total or pS199 phospho-Tau. Interestingly, DEX treatment did not change the secreted versus intracellular ratio for total Tau (Fig. 1M), but signi cantly increased this ratio for pS199 Tau (by two-fold; Fig. 1N). These results indicate that DEX preferentially stimulates secretion of this phospho-Tau species.
Secretion of vesicle-free Tau has been shown to occur through type 1 UPS, wherein Tau is directly translocated across the plasma membrane through interactions with heparin sulfate proteoglycans (HSPGs), cholesterol, and sphingolipids 37,38 . To determine whether GCs stimulate Tau secretion via this pathway, we rst treated hippocampal neurons with vehicle or DEX +/-NaClO 3 , an inhibitor of HSPG synthesis previously shown to decrease Tau secretion via type 1 UPS 37,38 . Following media collection, EV depletion, and measurement of phospho-and total Tau levels by immunoblot and ELISA, respectively, we found that NaClO 3 almost completely blocked the DEX-induced increase in extracellular Tau levels ( Fig. 2A-D). Treatment with methyl-b-cyclodextrin to extract membrane cholesterol similarly inhibited DEXinduced Tau secretion ( Fig. 2E-H), demonstrating its HSPG-and cholesterol-dependence. Comparable results were seen in brain slices treated with NaClO 3 and methyl-b-cyclodextrin ( Fig. S2A-H). Since type 1 UPS is ATP-independent, we tried to con rm this aspect of DEX-induced Tau secretion. Unfortunately, the different time courses of DEX treatment vs. ATP depletion with 2-deoxyglucose, and the toxicity of this latter treatment, prevented us from testing both conditions simultaneously. However, we veri ed that baseline Tau secretion in our neuronal cultures was ATP-independent, by brie y incubating hippocampal neurons from PS19 mice (overexpressing human P301S mutant Tau; hTau) with 2-deoxyglucose (30 mM, 1 hr). While this treatment reduced cellular ATP production by ~ 70%, it did not change the concentration of extracellular Tau in EV-depleted medium (Fig. S2I), con rming the overall ATP-independence of Tau secretion measured in our assays.
GCs are known to induce Tau hyperphosphorylation via activation of Tau kinases (e.g. GSK3b, CDK5) 13,39−41 and also to stimulate neuronal ring [42][43][44] , both of which are reported to enhance Tau secretion 26,37,38 . We therefore treated hippocampal neurons with DEX +/-the GSK3b inhibitor TDZD-8 or the Na + channel blocker tetrodotoxin (TTX) to inhibit neuronal ring. Both treatments not only decreased Trans-cellular spreading of pathogenic Tau is regarded as a key driver of AD progression 22 . We therefore examined whether Tau secreted in response to high GC levels is internalized by neighboring neurons. For this experiment, media from PS19 'donor' hippocampal neurons treated with vehicle or DEX (1 µM, 48h) was incubated for 48 hours with 'recipient' neurons from wild-type mice (Fig. 3A), and Tau uptake quanti ed by immunostaining with human-speci c anti-Tau13 antibodies. As anticipated based on previous studies 37,38 , hTau secreted by both control and DEX-treated donor neurons was readily taken up by recipient neurons (Fig. 3B, C), indicative of its ability to spread trans-cellularly. However, hTau levels were three-fold higher in recipient neurons incubated with medium from DEX-treated donor cells versus vehicle-treated cells (Fig. 3C). This nding likely re ects increased hTau levels in media following DEX treatment in donor cells, but could also indicate a stimulatory effect of DEX on Tau uptake by recipient cells, or DEX-related toxicity leading to increased membrane permeability to Tau. To investigate these latter possibilities, we treated recipient neurons with DEX for 48 hours during their incubation with (control) donor cell medium and subsequently quanti ed hTau levels. Interestingly, DEX-treated recipient neurons exhibited similar levels of hTau as their vehicle-treated counterparts ( Fig. 3B-C), and no difference in LDH release (Fig. 3D). These ndings indicate that GCs do not stimulate Tau internalization or alter plasma membrane permeability, but rather facilitate Tau spreading by stimulating its secretion.
Finally, we evaluated whether high circulating GC levels promote Tau spreading in vivo. Here, 4-5 monthold wild-type male mice (3/group) were pre-treated with vehicle, DEX, or DEX + MIF for 7 days, then injected in hippocampal area CA1 ( Fig. 4A) with an adeno-associated virus (AAV) that enables visualization of trans-cellular Tau spreading (AAV.CBA.eGFP.2A.P301L-Tau) 45 . Animals were then treated for an additional 14 days with vehicle (CON), DEX, or DEX + MIF prior to tissue harvest. DEX administration caused a ~ 10% loss of body weight during this time period, demonstrating its ability to promote an endocrine response mimicking stress (Fig. S2L). After brains from each treatment group were harvested and sectioned, human P301LTau spreading was evaluated by immunostaining with anti-Tau13 antibodies (Fig. 4B, C). Tau propagation was quanti ed as in previous studies 45,46 , by counting hTau + /GFP − neurons per mm 2 near the injection site and calculating the ratio of hTau + cells expressing GFP (GFP/hTau colocalization) (Fig. 4D, E). Remarkably, the number of hTau + /GFP − neurons was dramatically increased in DEX-treated animals compared to CON or DEX + MIF conditions (Fig. 4B, D), while GFP/hTau colocalization was signi cantly decreased (Fig. 4B, E). AAV transduction e ciency (number of GFP + cells per mm 2 ) was similar across treatment conditions (Fig. 4F). Moreover, in DEXtreated animals, hTau was detected in brain areas more than 1000 µm away from GFP + neurons, a phenomenon not observed in the other two groups (Fig. 4C, G). These data demonstrate that GCs strongly promote Tau secretion and spreading in vivo. To assess whether this spreading occurs via type 1 UPS, we initiated a second hTau spreading experiment with epigallocatechin gallate (EGCG), a potent inhibitor of Tau aggregation that attenuates its secretion via type 1 UPS 37,47 and can be used in vivo, unlike other inhibitors of type 1 UPS 48, 49 . We rst veri ed the ability of EGCG to reduce GC-induced Tau secretion in brain slices ( Fig. S2M-P). Animals were then subjected to the same experimental paradigm as above, but with EGCG instead of MIF. We again found that DEX provoked a ~ 10% loss of body weight, and this phenotype was not rescued by EGCG (similar to MIF treatment; Fig. S2L). As predicted by its ex vivo e cacy, EGCG administration almost completely prevented DEX-induced Tau spreading in the hippocampus (Fig. 4H-M), showing that this process occurs via Tau oligomerization and secretion via type 1 UPS.

Discussion
This work provides the rst demonstration that GCs stimulate Tau spreading in the brain, implicating these stress hormones in both the initial stages of Tau pathogenesis, by inducing Tau hyperphosphorylation and aggregation within neurons, and subsequently in the transmission of pathogenic Tau between neurons. We show that GCs stimulate Tau secretion via type 1 UPS, an ATPindependent process requiring interactions between phosphorylated/oligomeric Tau and plasma membrane-associated HSPGs and lipids. Notably, Tau secretion and spreading have also been shown to occur via extracellular vesicles (i.e. exosomes and ectosomes) and to be mediated by other brain cell types including microglia 24,50 . Additional work will be required to determine whether GCs also stimulate Tau propagation via these mechanisms.
An intriguing nding of this study is that EGCG, a catechin found at high levels in green tea leaves, blocks GC-induced Tau spreading in vivo. EGCG is an inhibitor of Tau oligomerization and aggregation as well as its secretion via type 1 UPS, suggesting that this is an important mode of GC-driven Tau propagation.
However, EGCG also alters lipid membrane properties 51,52 and could alter Tau secretion/uptake via this mechanism. Other drugs that block type 1 UPS, such as NaClO 3 and methyl-b-cyclodextrin, have similarly pleiotropic effects (and further cannot be used in vivo due to their blood brain barrier impermeability and toxicity, respectively 48,49 ), making it di cult to de nitively demonstrate that secretion via type 1 UPS is the primary driver of GC-induced Tau propagation in vivo. However, given the relative amount of Tau reported to undergo secretion in vesicle-free form (~ 90%) 22 , we think it is reasonable to assume that type 1 UPS contributes substantially to GC-induced Tau spreading in vivo.
Our experiments further reveal that GCs promote Tau secretion by stimulating GSK3b-mediated Tau phosphorylation. GSK3b, a brain-enriched serine/threonine kinase implicated in Tau pathogenesis in AD, phosphorylates multiple Tau residues, including those detected by our ELISA (S199) and immunoblotting (Ser202/Thr205, S396/S404) assays 53,54 . Indeed, we observe that GCs selectively increase secretion of pS199 Tau compared to total Tau, and the GSK3b inhibitor TDZD-8 effectively blocks GC-mediated Tau secretion in hippocampal neurons. Interestingly, we also nd that neuronal activity is critical for this process, as treatment with TTX to inhibit action potential ring prevents GC-induced Tau phosphorylation as well as its secretion. These data are in line with other studies reporting that neuronal activity, in the form of depolarization or NMDA receptor activation, stimulates Tau phosphorylation 55,56 . On the other hand, phosphorylated Tau can also exert effects on neuronal activity. In particular, the mislocalization of phospho-Tau species to dendritic spines in response to stress/GCs has been suggested to induce aberrant neuronal ring/excitotoxicity via Fyn kinase-mediated opening of NMDA receptors, leading to Ca 2+ in ux 14,15,56 . These ndings suggest the existence of a positive feedback loop, wherein GC-induced neuronal activity promotes Tau phosphorylation, which in turn induces the synaptic mistargeting of phospho-Tau species that stimulate additional neuronal activity to continue this cycle. However, since GCs are known to activate multiple Tau kinases, including CDK5 and GSK3b 13,39−41 , and also to stimulate the ring of cortical and hippocampal glutamatergic neurons on a rapid timescale (1-4h after application) [42][43][44] , it may be challenging to fully disentangle the causality of these events.
Cumulatively, our data show that GC-mediated phosphorylation and oligomerization of Tau stimulates its vesicle-free secretion and trans-cellular spreading via type 1 UPS. While questions remain about how Tau phosphorylation is precipitated by GCs, and how stress/GCs impact other forms of Tau propagation in the brain, this work provides some of the rst mechanistic insight into how high GC levels accelerate pathogenic Tau spreading in AD and other tauopathies. Chronic unpredictable stress, brain tissue collection, and media harvest Three-to four-month-old wild-type animals (C57BL/6J) were housed in groups of 5-6 per cage under standard environmental conditions with ad libitum access to food and water. For the chronic unpredictable stress (CUS) protocol, animals were subjected to different stressors (i.e. 3 hours overcrowding, 3 hours rocking platform, 3 hours restraint, 30 min hairdryer; one stressor per day) that were chosen randomly to prevent habituation, over a period of six weeks. Following the CUS protocol, animals were euthanized, brain tissue was immediately macrodissected and incubated in EV-release medium (Neurobasal medium, 1% Glutamax, 1% Anti-anti; ThermoFisher) for 16h at 37ºC, 5% CO 2 . Five hemi-cortices were pooled to obtain each cortical sample while hippocampi from 5 mouse brains were pooled into each hippocampal sample. After the incubation period, media was collected and subject to extracellular vesicle depletion as described above (Media/ACSF preparation).

ExoView Imager Analysis
The characterization and quanti cation of exosomes in hippocampal culture media were performed according to the manufacturer's instructions 59 . Brie y, chips containing capture probes coated with antibodies against two exosome-enriched tetraspanins, CD81 and CD9, were pre-scanned to acquire baseline particle adhesion prior to sample incubation. Media samples were diluted to fall within the dynamic range of the Exoview R100 instrument (Unchained Labs), and incubated overnight at room temperature on the pre-scanned chips in a sealed 24-well plate. The chips were then washed to remove any non-captured material, incubated for 1 hour at room temperature with uorescently-conjugated antibodies against CD9, CD63, and CD81, washed again, dried, and then scanned with the ExoView R100 system to obtain data on particle counts, size, and exosome surface membrane protein pro les. For each capture probe (CD9 and CD81), background particle readout is subtracted from the nal particle count to produce a nal exosome count readout.

Immunoblotting
The concentrated media/ACSF with extracellular vesicle (EV) depletion were prepared in 4x Laemmli buffer and then boiled for Tau uptake assay Media was collected from donor WT or PS19 neurons treated with vehicle control or dex (1µM) for 48 hours. The media from these cultures was then depleted of EVs as described above and transferred to naïve recipient wild-type neurons for a 48-hour incubation. For one condition, recipient neurons were also treated with dex (1µM) during this time. Following incubation, recipient cells were washed three times with cold 1x PBS and xed with 4% paraformaldehyde as previously described 57 . The uptake of hTau was then detected by immunostaining with MAP2 and Tau13 antibodies as described below.
Immuno uorescence staining of brain slices, cultured neurons. Floating brain sections or xed primary neurons were immunostained as previously described 57  injection) +/-EGCG (20 mg/kg, i.p. injection) for 7 days. Stereotactic AAV injections were performed under standard aseptic surgery conditions as previously described 46 . Brie y, mice were anaesthetized with iso urane (2%), placed in a stereotactic frame (digital stereotaxic device, Stoelting Co.), and injected bilaterally with 2 ml of AAV in hippocampal region CA1 (at the following coordinates relative to Bregma: A/P −2.7 mm, M/L ±2 mm, D/V −1.5 mm) with a 10 μl Hamilton syringe at a rate of 0.25 μl/min by a Nano-injector system (Stoelting microsyringe pump, Stoelting Co.). The needle was kept in place for an additional 5 min. Afterwards, the skin over the injection site was sutured and mice were placed on a warming pad during their recovery from anesthesia. Mice were then administered dex with or without mifepristone or EGCG for an additional 14 days prior to euthanasia and brain harvest. Control animals received daily i.p. injections of 50% PEG400 in PBS (dex/mifepristone vehicle) or PBS (dex/EGCG vehicle).

Quanti cation and statistical analysis
All values were expressed as the mean ± SEM. All graphing and statistical analyses were performed using GraphPad Prism (GraphPad Prism9.Ink). Statistical details of experiments are provided in the gure legends. Statistical analyses were performed with unpaired, two-tailed t-test or one-way ANOVA, with appropriate corrections for unequal variances and multiple comparisons. A minimum of 3 independent replicates were used for all experiments. Values of p < 0.05 were considered statistically signi cant. *p<0.05, **p<0.01, ***p<0.001,****p<0.0001.