Constitutive XBP-1s-mediated activation of the endoplasmic reticulum unfolded protein response protects against pathological tau

To endure over the organismal lifespan, neurons utilize multiple strategies to achieve protein homeostasis (proteostasis). Some homeostatic mechanisms act in a subcellular compartment-specific manner, but others exhibit trans-compartmental mechanisms of proteostasis. To identify pathways protecting neurons from pathological tau protein, we employed a transgenic Caenorhabditis elegans model of human tauopathy exhibiting proteostatic disruption. We show normal functioning of the endoplasmic reticulum unfolded protein response (UPRER) promotes clearance of pathological tau, and loss of the three UPRER branches differentially affects tauopathy phenotypes. Loss of function of xbp-1 and atf-6 genes, the two main UPRER transcription factors, exacerbates tau toxicity. Furthermore, constitutive activation of master transcription factor XBP-1 ameliorates tauopathy phenotypes. However, both ATF6 and PERK branches of the UPRER participate in amelioration of tauopathy by constitutively active XBP-1, possibly through endoplasmic reticulum-associated protein degradation (ERAD). Understanding how the UPRER modulates pathological tau accumulation will inform neurodegenerative disease mechanisms.

P ersisting for decades, human neurons face unique metabolic challenges, including the need to maintain long-term protein quality control. Cellular systems such as the ubiquitinproteasome system (UPS), autophagy, the heat shock response, and the unfolded protein responses in the endoplasmic reticulum (UPR ER ) and mitochondria (UPR mt ) enable neurons to maintain protein homeostasis (proteostasis) within different cellular compartments. Proper folding of all secretory proteins occurs within the endoplasmic reticulum (ER), comprising~25% of all proteins in the mammalian proteome 1,2 . Proteostasis imbalance in the ER triggers the UPR ER , which causes the molecular chaperone binding immunoglobulin protein (BiP) to dissociate from the three ER transmembrane stress sensors [protein kinase RNA-like ER kinase (PERK), inositol-requiring enzyme 1 α (IRE1α), and activating transcription factor 6 (ATF6)] (Fig. 1a). This results in immediate inhibition of general translation to decrease the protein folding load in the ER and activation of genes for restoring protein folding homeostasis. PERK attenuates global translation to reduce entrance of newly synthesized proteins into the ER lumen and activates translation of activating transcription factor 4 (ATF4), which controls the expression of genes involved in autophagy, amino acid metabolism, antioxidant responses, and apoptosis 3 . IRE1α catalyzes non-canonical splicing of X-box binding protein 1 (XBP1) mRNA into the constitutively active form XBP1s, which becomes the master UPR ER transcription factor controlling a wide range of gene targets required for ER proteostasis. These include genes related to protein folding, ERassociated protein degradation (ERAD), protein translocation into the ER, and lipid synthesis 4 . ATF6 translocates to the Golgi apparatus where it is cleaved into C-and N-terminal fragments. The N-terminal fragment (ATF6n) translocates to the nucleus to activate transcription of ERAD genes and UPR ER master transcription factor XBP1s 3 . If the stress is prolonged and unresolved, the UPR ER switches from a pro-survival role to a pro-apoptotic role, activating downstream signaling events of all three stress sensors to induce cell death.
The UPR ER is an evolutionarily conserved adaptive signaling pathway. The most conserved UPR ER signaling branch, and the only one present in yeast, is initiated by IRE1α 5,6 . After BiP dissociates from the ER luminal domain of IRE1α, IRE1α undergoes dimerization and trans-autophosphorylation to activate its cytosolic endoribonuclease activity. Activated IRE1α catalyzes the excision of a short non-canonical intron from XBP1 mRNA. This causes a frameshift in the XBP1 coding sequence, translation into XBP1s protein, and entrance into the nucleus as an active transcription factor. UPR ER malfunctions are associated with a wide range of disease states, including tumor progression and diabetes, as well as immune, inflammatory, and neurodegenerative diseases 7 . Although the UPR ER plays many important physiological roles, we still do not fully understand the role of the UPR ER in post-mitotic neuronal survival.
Aging leads to a decline in proteostasis capacity, which can particularly affect post-mitotic cells such as neurons. Agingrelated declines in proteostasis contribute to abnormal protein aggregation and downstream cellular degeneration. In neurodegenerative disease, age-dependent pathological protein accumulation drives synaptic impairment and neuronal loss. Several studies have highlighted the importance of the UPR ER in the aging brain. Under basal conditions, UPR ER chaperones such as BiP are decreased in several brain regions of aged mice [8][9][10] . When these animals experience cellular stress, the UPR ER fails to properly activate, leading to the upregulation of apoptotic pathway mediators [8][9][10] . The IRE1α/XBP1 branch may be particularly critical during aging. Dietary restriction extends lifespan in multiple organisms and requires IRE-1/XBP-1 activity in Caenorhabditis elegans (C. elegans) 11 . Interestingly, constitutive XBP-1s activation in neurons can extend lifespan in C. elegans via XBP-1s-mediated non-cell autonomous signaling events 12 . Age strongly drives risk for many protein aggregate-deposition diseases of the nervous system; UPR ER dysfunction has been implicated in several of these diseases, including Alzheimer's, Parkinson's, and Huntington's disease, and amyotrophic lateral sclerosis 7,13 .
Pathological tau protein serves as both a marker of brain aging and a hallmark of several neurodegenerative diseases termed tauopathies. Tauopathies exhibit cytoplasmic accumulation of neuronal and/or glial lesions containing hyperphosphorylated tau. As the most prevalent tauopathy, Alzheimer's disease (AD) neuropathology consists of both extracellular amyloid plaques composed of fibrillar amyloid beta (Aβ) peptides and intracellular neurofibrillary tangles composed of pathological tau protein.
Mutations in the microtubule-associated protein tau (MAPT) gene encoding the protein tau cause FTDP-17, providing a direct link between tau dysfunction and disease [15][16][17] . Interventions for tauopathies are limited to treatment of symptoms, as no approved therapies impact tau pathology or concomitant neurodegeneration.
To study tauopathies in a genetically tractable system, we have generated transgenic C. elegans that express human tau panneuronally 18 . These models exhibit behavioral abnormalities, accumulation of insoluble phosphorylated tau protein, progressive neurodegeneration, and shortened lifespan. We previously conducted a genome-wide RNA interference (RNAi) screen to identify modifiers of tau toxicity 19 . From this screen, we found RNAi targeting the UPR ER master transcription factor xbp-1 enhanced FTDP-17 mutant tau-induced behavioral deficits, and strong xbp-1 loss of function causes synthetic lethality with tau expression, nominating the UPR ER as a regulator of tau proteostasis 19 . The relationship between UPR ER function and pathological tau accumulation remains poorly characterized. To study the consequences of UPR ER loss of function on pathological tau, we have utilized loss of function mutations and gain of function transgenes to genetically dissect the role of the UPR ER in tau proteostasis.

UPR ER branches differentially affect tauopathy phenotypes.
The UPR ER consists of three branches that regulate signaling to the nucleus and control the responses to unfolded protein within the ER lumen (Fig. 1a). Using the genetically tractable model organism C. elegans, we examined the effects of loss of function mutations ablating signaling through each of the three branches of the UPR ER [pek-1 (−/−), atf-6 (−/−), and xbp-1 (−/−)] on tau-mediated behavioral dysfunction in a mild model of human wildtype tau toxicity, which will be referred to as Tau (low) transgenic C. elegans 20 . Compared to non-transgenic (non-Tg) animals, Tau (low) animals exhibit mild behavioral deficits and no significant accumulation of pathological tau species.
To investigate whether atf-6 loss of function in a Tau (low) background altered tau levels, we analyzed total tau protein expression by immunoblot. We found atf-6 loss of function did not significantly affect total tau protein levels (Fig. 1d, e).

ns
Relative density Fig. 1 The UPR ER PEK-1/PERK and ATF-6/ATF6 branches are differentially involved in tauopathy in transgenic C. elegans. a Model figure of the mammalian UPR ER , which acts to restore ER proteostasis. Cellular stress can disrupt ER proteostasis, triggering the UPR ER stress response. Three transmembrane stress sensors (PERK, IRE1α, and ATF6) are bound by the ER molecular chaperone BiP, holding them in inactive conformations. Misfolded proteins in the ER causes dissociation of BiP from the stress sensors, allowing adoption of active conformations. PERK dimerization and autophosphorylation allows phosphorylation of eIF2α and other substrates in the cytoplasm. Active PERK attenuates overall protein translation to reduce nascent protein folding load in the ER and activates transcription factor ATF4 to control target genes involved in restoring proteostasis. IRE1α dimerization promotes non-canonical splicing of XBP1 mRNA in the cytoplasm to produce XBP1s mRNA encoding the UPR ER master transcription factor XBP1s. ATF6 translocates to the Golgi apparatus for cleavage into N-and C-terminal fragments, with the N-terminal fragment becoming an active transcription factor (ATF6n). Additionally, XBP1s and ATF6n can form heterodimers to regulate transcriptional target genes essential for ER proteostasis restoration. b pek-1 loss of function in Tau (low) animals does not alter mild behavioral defects observed in a liquid environment [n = 100 animals; N = 5 biologically independent experiments; statistical analysis is by unpaired t-test, two-tailed (ns: p = 0.2261)]. c atf-6 loss of function in Tau (low) animals enhances mild behavioral defects observed in a liquid environment [n = 100 animals; N = 5 biologically independent experiments; statistical analysis is by unpaired t-test, two-tailed (***p < 0.0001)]. d, e atf-6 loss of function in Tau (low) animals does not affect soluble tau protein levels. Representative immunoblots for total tau and tubulin are shown, and densitometry analysis of chemiluminescence signals for tau normalized to tubulin are plotted [N = 3 biologically independent experiments; statistical analysis is by paired t-test, two-tailed (ns: p = 0.4041)]. All bar graphs represent mean + SEM. Arrowhead on y-axis of all liquid thrashing bar graphs denotes non-Tg animals average~70 thrashes/min under standard laboratory conditions. Source data are available as a Source Data file Therefore, atf-6 loss of function enhances tau locomotion dysfunction independent from a role in tau protein turnover. C. elegans with an xbp-1 loss of function mutation [xbp-1 (−/−)] had mild locomotion defects when compared to non-Tg animals ( Supplementary Fig. 2); however, Tau (low); xbp-1 (−/−) C. elegans exhibited dramatically exacerbated tau-dependent locomotion dysfunction ( Supplementary Fig. 3). The mild behavioral deficits seen in Tau (low) animals were significantly enhanced with xbp-1 loss of function. This phenotype was confirmed using an independent tau transgenic C. elegans (Fig.  2a). To determine the effects of xbp-1 loss of function on tau protein turnover, we assayed tau protein levels by immunoblot. Tau (low); xbp-1 (−/−) animals exhibited increased tau protein compared to Tau (low) animals (Fig. 2b, c).
Neurodegeneration is the ultimate consequence of pathological tau in tauopathy disorders. To assess whether xbp-1 loss of function promotes tau-dependent neurodegeneration, we used a transgenic C. elegans reporter strain expressing green fluorescent protein (GFP) in D-type gamma-aminobutyric acid (GABA) positive neurons (unc-47p::gfp) 21 . This strain allows in vivo assessment of the integrity of GABAergic motor neurons along the ventral nerve cord. At day 1 of adulthood, Tau (low) animals did not exhibit significant neurodegeneration (Fig. 2d, e). Fewer than one in four Tau (low) animals were missing one neuron (Fig.  2e). However, at day 1 of adulthood, xbp-1 (−/−) exacerbated neurodegeneration in Tau (low) animals, including loss of neuronal cell bodies and breaks in the ventral nerve cord (Fig.  2d). Nearly every Tau (low); xbp-1 (−/−) animal exhibited neurodegeneration (Fig. 2e). Loss of GABAergic motor neurons was more pronounced at day 4 of adulthood ( Fig. 2f). Importantly, xbp-1 (−/−) animals did not exhibit neurodegeneration on their own in the absence of tau ( Supplementary Fig. 4). Taken together, xbp-1 loss of function exacerbates tau-mediated behavioral deficits, increases accumulation of tau protein, and drives tau-dependent neurodegeneration in C. elegans models of tauopathy.
Neuronal xbp-1s activation ameliorates tauopathy phenotypes. Given xbp-1 loss of function enhances tauopathy phenotypes, we hypothesized xbp-1 gain of function might suppress tauopathy. To address this question, we utilized a severe model of human statistical analysis is by unpaired t-test, two-tailed (**p = 0.0020)] and day four of adulthood [n = 24 animals; N = 2 biologically independent experiments; statistical analysis is by unpaired t-test, two-tailed (***p < 0.0001)] are plotted. All bar graphs represent mean + SEM. Arrowhead on y-axis of all liquid thrashing bar graphs denotes non-Tg animals average~70 thrashes/min under standard laboratory conditions. Source data are available as a Source Data file wildtype tau toxicity, referred to as Tau (high) transgenic C. elegans 20 . Compared to non-Tg animals, Tau (high) animals exhibit strong behavioral and neurodegenerative phenotypes driven by pathological tau protein 20 , enabling us to readily detect potential improvement in tauopathy phenotypes. We crossed Tau (high) transgenic animals with transgenic animals expressing a pan-neuronal constitutively active xbp-1s gain of function transgene (xbp-1s Tg), which promotes activation of the UPR ER in the absence of increased ER protein folding load 12 . C. elegans expressing xbp-1s Tg have mild locomotion defects when compared to non-Tg animals ( Supplementary Fig. 5), while Tau (high) C. elegans have severe locomotion dysfunction (Fig.  3a). However, Tau (high); xbp-1s Tg animals had robust rescue of the tau-mediated behavioral deficits compared to Tau (high) x b p -1 s  (Fig. 3a). To test whether xbp-1s Tg promotes tau protein turnover, we examined tau levels by immunoblot. Tau (high); xbp-1s Tg animals accumulated less total tau protein when compared to Tau (high) animals (Fig. 3b, c). To address the possibility xbp-1s hyperactivity decreases tau protein levels by an effect on tau transgene expression rather than protein turnover, we assessed human tau mRNA levels by quantitative reversetranscription PCR (qRT-PCR). We found no significant difference in tau-encoding mRNA ( Supplementary Fig. 6), indicating reduced levels of tau protein are a result of increased tau protein clearance rather than decreased transgenic tau mRNA expression. In tauopathy disorders, tau becomes hyperphosphorylated, driving pathological protein aggregation and neurodegeneration 22 . We measured tau phosphorylation by immunoblot at four sites commonly associated with tauopathy (Threonine 181, Serine 202, Serine 396/404, and Serine 422) using phosphorylated tauspecific antibodies. Tau (high); xbp-1s Tg C. elegans had less phosphorylated tau at all four epitopes compared to Tau (high) animals (Fig. 3d, Supplementary Table 3). We then tested whether xbp-1s gain of function would protect against tau-driven neurodegeneration. At day 1 of adulthood,~50% of Tau (high) C. elegans lost one GABAergic neuron (Fig. 3e, f). However, Tau (high); xbp-1s Tg worms were protected against this early tauinduced neuronal loss (Fig. 3e, f). By day 4 of adulthood, Tau (high) worms showed progressive neuronal loss with an average of almost three neurons lost per animal, while Tau (high); xbp-1s Tg worms that exhibited neurodegeneration lost approximately one neuron (Fig. 3g). Altogether, we have demonstrated xbp-1s gain of function significantly suppresses behavioral, biochemical, and neurodegenerative tauopathy phenotypes observed in the C. elegans Tau (high) background.
ATF-6 mediates xbp-1s suppression of tauopathy. The XBP-1s transcription factor mediates one of three branches of the UPR ER (Fig. 1a). To understand whether the other two UPR ER branches participate in xbp-1s-mediated tauopathy suppression in C. elegans, we again utilized pek-1 and atf-6 loss of function mutant strains. Tau (high); xbp-1s Tg animals were crossed with either pek-1 or atf-6 loss of function mutant strains.
Because xbp-1s-mediated tauopathy behavioral suppression requires atf-6, we also measured atf-6 loss of function effects on total tau protein by immunoblot in Tau (high) animals using two independent atf-6 loss of function alleles. Loss of function of atf-6 alone did not change tau protein levels in Tau (high) animals (Fig. 4c, d; Supplementary Fig. 7b). We have shown xbp-1s gain of 3b, c; 4c, d; Supplementary Fig. 7b). In contrast, Tau (high); xbp-1s Tg; atf-6 (−/−) animals accumulated total tau protein similar to the level seen in Tau (high) animals (Fig. 4c, d; Supplementary  Fig. 7b), indicating functional ATF-6 is necessary for xbp-1smediated tauopathy detoxification. Taken together, these data demonstrate distinct, but overlapping roles for the three branches of the UPR ER in regulating tau proteostasis in transgenic C. elegans.
ERAD is required for xbp-1s-mediated tauopathy suppression. Dysregulated proteostasis within the ER triggers the UPR ER , resulting in activation of ER-associated protein degradation (ERAD). This provides a means for the direct removal of misfolded proteins inside the ER lumen and facilitates their degradation through either autophagy or the proteasome 23 . To determine whether tau protein resides within the ER, we crossed Tau (high) animals with transgenic animals expressing a fluorescently-tagged ER integral membrane protein, CP450:: mCherry 24 . We co-immunostained transgenic animals with antibodies targeting tau and mCherry to visualize subcellular localization of these proteins (Fig. 5a-c). Using colocalization immunofluorescence analysis, we showed no significant overlap of tau protein and the ER protein CP450 (Fig. 5d). Thus, the mechanism of tau detoxification by UPR ER activation in transgenic C. elegans does not rely upon removal of tau protein directly from the ER. Previous work in cultured human cells demonstrated tau impairs ERAD function 25 . We hypothesized abnormal tau also inhibits ERAD in C. elegans, and neuronal tau toxicity in C. elegans may depend upon ERAD impairment. In support of this hypothesis, previous work has shown UPR ER activation stimulates ERAD through XBP-1s target genes 26 . Therefore, XBP-1s suppression of tauopathy phenotypes in C. elegans may be through an ERAD-dependent mechanism. To test the involvement of ERAD in tau clearance, we examined whether ERAD loss of function could block XBP-1s suppression of tauopathy in C. elegans. SEL-11, the C. elegans homolog of the ERAD-associated E3 ubiquitin ligase HRD1, is required for ERAD 27 . We crossed sel-11 loss of function mutants [sel-11 (−/−)] with Tau (high); xbp-1s Tg animals and measured behavioral function. We found sel-11 (−/−) fully blocked XBP-1s-mediated suppression of tauopathy behavioral deficits (Fig. 5e). Taken together, these data suggest ERAD activation via XBP-1s promotes clearance of pathological tau in C. elegans.

Discussion
Although aging both increases AD risk and decreases UPR ER function, the specific contribution of UPR ER activity to disease initiation and/or progression remains poorly understood. Likewise, whether UPR ER induction causes or results from AD pathology remains unclear. Most studies to date have made correlative associations or relied upon immortalized cell culture experiments, which fail to model key aspects of tauopathy 28 . Here, we have taken a more direct approach by genetically dissecting the role of the UPR ER in tauopathy in vivo using human wildtype tau transgenic C. elegans. We found xbp-1s expression to be a central control point in tauopathy, although this role requires involvement of all three branches of the UPR ER .
To study the contributions of each UPR ER branch, we systematically eliminated the function of the three UPR ER branches in tau transgenic C. elegans. The PERK branch of the UPR ER mainly functions to attenuate protein synthesis in the ER. Interestingly, we found C. elegans pek-1 does not regulate tau proteostasis on its own (Fig. 1b), indicating a non-essential role for this branch in our tauopathy model. However, the PERK branch of the UPR ER plays a modest role in modulating tauopathy risk, as human PERK variants increase risk for developing PSP by~25% 29 . In addition, human post-mortem brain analyses found increased PERK activation in a subset of PSP 30 and AD disease-affected brain regions [30][31][32] . Our results indicate either the consequences of PEK-1 modulation are too modest to be detected in tau transgenic C. elegans, or there might be a critical speciesspecific difference between humans and C. elegans regarding PERK/PEK-1 function in the UPR ER .
The XBP1 and ATF6 branches of the UPR ER mediate transcriptional regulation of genes for restoring proteostasis in the ER. Our results show both atf-6 and xbp-1 loss of function enhance tauopathy-induced behavior defects in C. elegans (Figs. 1c, 2a; Supplementary Fig. 3), with xbp-1 loss of function additionally increasing tau protein levels and neurodegeneration (Fig.  2b-f). In humans, a polymorphism in the promoter of the XBP1 gene was found to be a risk factor for AD in a Chinese Han population 33 . Post-mortem tissue analyses have revealed a more complex scenario for XBP1 branch changes in AD, possibly related to disease stage, patient cohort characteristics, or experimental methodology. One group found the constitutively active form XBP1s is downregulated in AD brains 34 . Another group found increased XBP1s, but no increase in its key regulatory target BiP, indicating a disturbance in UPR ER activation 35 . Work from our laboratory and others has demonstrated loss of UPR ER function via XBP1 exacerbates pathological tau phenotypes in simple model organisms such as C. elegans ( 19 , this study) and Drosophila melanogaster (D. melanogaster) 36 .
Several recent studies highlight the potential for the UPR ER , specifically XBP1, as a therapeutic target for dementia. First, upregulation of XBP1s rescues Aβ-mediated neurotoxicity in D. melanogaster, mammalian cultured neurons, and C. elegans 37,38 . Second, Xbp1 gain of function in the hippocampus of mice can improve spatial memory 39 . To dissect the role of UPR ER induction in protection from tauopathy, we expressed constitutively active xbp-1s in C. elegans neurons. We found XBP-1s protects neurons from pathological tau by promoting proteostasis and clearance of abnormal tau species, resulting in amelioration of both behavioral dysfunction and neurodegenerative changes in C. elegans (Fig. 3). Taken together, these data clearly support a critical role for UPR ER transcriptional induction in mediating the clearance of abnormal protein species such as Aβ and tau, while shedding light on a novel active tau clearance mechanism in AD and other tauopathies.
Given the highly orchestrated nature of the UPR ER (Fig. 1a), we next investigated whether the other two branches of the UPR ER were genetically required for XBP-1s-mediated protection against pathological tau in C. elegans. Interestingly, this experiment identified interplay between the three branches of the UPR ER . We observed C. elegans PEK-1 loss of function limits, but does not eliminate xbp-1s-mediated suppression of tauopathyinduced behavioral deficits (Fig. 4a). In C. elegans, the ire-1/xbp-1 and pek-1 branches play partially complimentary roles in eliminating ER stress, where ire-1/xbp-1 signals to activate UPR ER transcription and pek-1 signals to attenuate protein synthesis 40 . In addition, pek-1 appears to be dispensable for the induction of genes directly involved in ER protein folding and ERAD 41 . The limited role of pek-1 in xbp-1s-mediated tauopathy suppression indicates a fully functioning UPR ER is required to completely overcome cytoplasmic pathological tau protein accumulation in C. elegans. Nonetheless, transgenic overexpression of XBP-1s can partially suppress tauopathy even in the absence of pek-1 activity.
In contrast, we have shown atf-6 loss of function completely abolished xbp-1s-mediated tauopathy suppression in C. elegans (Fig. 4b-d; Supplementary Fig. 7). This demonstrates an absolute requirement for ATF-6 in mediating tauopathy suppression by transgenic constitutively active XBP-1s expression in C. elegans and a previously unappreciated critical crosstalk between ATF-6 and XBP-1 branches of the UPR ER in tau proteostasis. Given the known importance of the XBP-1 and ATF-6 branches in UPR ER transcriptional regulation, a plausible mechanism of tauopathy suppression involves regulation of specific target gene(s) coordinately upregulated by ATF-6n and XBP-1s transcription factors. In agreement with our results, previous studies suggest ATF-6n might have evolved as a backup mechanism to the XBP-1s transcriptional pathway in C. elegans, since many genes regulated by ATF-6n overlap with those regulated by XBP-1s 41 . In mammals, ATF6 activates the transcription of XBP1 and can form heterodimers with XBP1s to control the induction of specific patterns of gene expression 42,43 . Additionally, mouse XBP1s/ ATF6n heterodimers exhibit a higher binding affinity to target genes than XBP1s homodimers, indicating the strong influence of ATF-6n on XBP-1s transcriptional regulation 42 . Currently, it is unknown whether C. elegans XBP-1s and ATF-6n transcription factors can form heterodimers, but they do exhibit a high degree of sequence and functional conservation with mammalian XBP1 and ATF6 homologs. Taken together with the requirement for ATF-6 function in XBP-1s-mediated tauopathy suppression, we hypothesize these two transcription factors also form The ER does not contain tau protein, but ERAD is required for xbp-1s-mediated tauopathy suppression in transgenic C. elegans. a-c Tau protein does not colocalize with the ER in the Tau (high) C. elegans nervous system. Tau protein subcellular localization in relation to the ER integral membrane protein CP450 was measured by co-immunofluorescent staining and analyzed using structured illumination microscopy (SIM). a Representative SIM image of the ER protein CP450 fused to mCherry shows concentrated ER membrane localization primarily surrounding nuclei in the nerve cord [yellow = mCherry (ER); scale bar = 10 µm]. b Representative SIM image of tau protein shows diffuse tau localization throughout the nerve cord (red = tau; scale bar = 10 µm). c Representative merged SIM image of tau and ER membrane proteins, showing distinct subcellular localizations [yellow = mCherry (ER); red = tau; blue = DAPI (nuclei); scale bar = 10 µm]. d Colocalization analysis of ER reporter staining intensity (y-axis) with tau staining intensity (x-axis) shows no significant colocalization of the two proteins [statistical analysis is by Pearson correlation coefficient (r = 0.23)]. e sel-11 loss of function abolishes the ability of neuronal overexpression of xbp-1s in Tau (high) animals to suppress severe behavioral defects observed in a liquid environment [n = 75 animals; N = 5 biologically independent experiments; bar graphs = mean + SEM; statistical analysis is by one-way ANOVA, followed by Tukey's post-test (***p < 0.0001)]. Arrowhead on y-axis denotes non-Tg animals average~70 thrashes/min under standard laboratory conditions. Source data are available as a Source Data file heterodimers in C. elegans to promote tauopathy suppression. Alternatively, other scenarios involving an ordered pathway, where atf-6 acts downstream of xbp-1, could explain the dependence of xbp-1s suppression on atf-6. Similarly, atf-6 loss of function could cause synergistic deficits with pathological tau that xbp-1s overexpression cannot ameliorate. Regardless, the parallel findings for atf-6 and pek-1 loss of function suggests full UPR ER functionality promotes xbp-1s-mediated tauopathy suppression in C. elegans. We found upregulating the UPR ER master transcription factor XBP-1s can increase tau protein turnover in the cytoplasm. Cytoplasmic tau can be degraded by at least two mechanisms: the UPS and autophagy. Soluble monomeric tau species are recognized by molecular chaperones and directed to the 20S proteasome without ubiquitination or 26S proteasome with ubiquitination for degradation, while aggregation-prone tau species, such as hyperphosphorylated tau, are degraded via autophagy 44,45 . Both of these protein degradation mechanisms are utilized by the UPR ER via ERAD, which shuttles proteins from the ER to the cytoplasm for degradation. Although tau does not reside in the ER, ERAD loss of function via sel-11 (−/−) is essential for xbp-1s-mediated behavioral tauopathy suppression in C. elegans (Fig. 5), highlighting the importance of transcompartmental trafficking mechanisms for global proteostasis.
We propose the following mechanism explaining the relationship between cytoplasmic tau and the UPR ER in neurons (Fig.  6). Although tau is not an ER resident protein, abnormal tau in the cytoplasm can nevertheless trigger the UPR ER by altering overall neuronal proteostasis through impaired ribosome function 46 , ER/cytoplasmic protein translocation, and disruption of ERAD 25 , leading to misfolded proteins in the ER. The main goal of the UPR ER is to restore proteostasis within the ER lumen, but to achieve this requires integrated stress relief within both cytoplasm and ER compartments. This is highlighted by a recent study demonstrating soluble tau impairs ERAD, which results in increased ER protein folding load and activation of the UPR ER 25 (Fig. 6a). Nascent proteins in the ER failing to properly fold are degraded through ERAD, a branch of the UPS in which proteins are ubiquitinated at the ER membrane for subsequent degradation. Although this is an ER-associated degradation pathway, the ERAD machinery can degrade abnormal proteins located outside the ER, such as in the cytoplasm or nucleus 47 . XBP-1s UPR ER branch loss of function inhibits ERAD, leading to pathological accumulation of abnormal tau as it is converted from soluble normal tau in the cytoplasm (Fig. 6b). In the presence of pathological cytoplasmic tau, XBP-1s UPR ER branch gain of function stimulates high levels of ERAD, leading to turnover of pathological tau species, such as hyperphosphorylated tau (Fig.  6b). Taken together with our findings, these studies indicate the importance of understanding the role of abnormal tau protein and the UPR ER in age-associated neurodegenerative disease.
Future translational studies will need to address whether UPR ER activation can protect against pathological tau in the mammalian brain. More broadly, exploring how different cellular compartments communicate to achieve neuronal proteostasis remains an important mechanistic question with therapeutic implications. Specifically, the mechanism of how cytoplasmic pathological tau feeds back on ER folding load needs to be identified. Given the importance of the two transcriptional branches of the UPR ER in tauopathy, our next step will be to identify xbp-1s target genes driving tau proteostasis in C. elegans, which could help inform more specific neuroprotective strategies in mammals. Likewise, involvement of the UPR ER in tauopathy needs to be explored in the mammalian brain to clarify whether targeting the UPR ER in tauopathy could have neuroprotective benefit.

Methods
C. elegans strains and transgenics. C. elegans strains used are listed in Supplementary Table 1. All strains were maintained at 20°C on standard nematode growth media plates containing OP50 Escherichia coli (E. coli) 48 . C. elegans were grown on nematode growth media plates containing five times more peptone (5X PEP) prior to collection for protein and RNA extraction. The C. elegans husbandry and experimentation was conducted in accordance with all relevant ethical and safety regulations for animal testing and research.
C. elegans behavioral analysis. C. elegans swimming behavior was performed as described previously 18 . C. elegans were developmentally synchronized using a timed egg-lay and grown until all reached day one of adulthood at 20°C. Individual developmentally synchronized animals were placed in a 10 µL M9 buffer droplet on a 10-well Teflon-printed glass slide and allowed to acclimate to a liquid environment for 10 s. One thrash was defined as a bend across the midline or two consecutive bends from the midline toward the same side. The number of thrashes were counted during a 1-min period and averaged for each strain. At least three biological replicate assay sessions with at least 10 animals per assay session were analyzed for statistical significance. Behavioral analysis was conducted by S.M.W., who was blinded to genotype to avoid biases. However, when the strains to be assayed were visibly distinguishable, blinding was not feasible. For comparison of two groups, an unpaired t-test, two-tailed, was used. For comparisons of three or more groups, a one-way analysis of variance (ANOVA), followed by Tukey's multiple comparison post-test was used.
C. elegans neurodegeneration analysis. Analysis of C. elegans neurodegeneration was performed using a transgenic reporter strain with fluorescently marked VD and DD type inhibitory motor neurons 18 . C. elegans were crossed with the EG1285 strain (unc-47p::gfp) 21 to visualize the ventral nerve cord green fluorescent protein (GFP) marked gamma-aminobutyric acid (GABA)ergic neurons via fluorescence microscopy. C. elegans were developmentally synchronized using a timed egg-lay and grown until all reached day one or four of adulthood. C. elegans were mounted on glass slides with a 2% agarose gel pad and 10 µL 0.1% sodium azide. The number of intact ventral nerve cord GABAergic D-type motor neurons were counted under fluorescence microscopy on a DeltaVision Elite (GE, Issaquah, WA, USA) imaging system using an Olympus × 60 oil objective. Representative images of neuronal loss were also taken using an Olympus × 60 oil objective. The three most rostral GABAergic motor neurons were excluded from counts due to occlusion from pharyngeal GFP co-injection markers. Therefore, the number of visibly accessible ventral nerve cord D-type motor neurons were counted out of a total of 16. Neurodegeneration was plotted as the average number of neurons lost from the 16 possible neurons analyzed. Two biological replicate assay sessions with at least 10 animals per assay session were analyzed for statistical significance. For comparison of two groups, an unpaired t-test, two-tailed, was used. Representative anatomical images were created with ImageJ Java 49 to overlay images from the POL and GFP channels, with the GFP channel image re-colored green and the POL channel image re-colored grey for emphasis. Images from the GFP channel were also shown below merged images for visual clarity of neuronal loss.
C. elegans protein extraction. Protein was extracted from C. elegans using a rigorous lysis approach 18 . C. elegans were synchronized using hypochlorite treatment and grown until adulthood at 20°C on 5X PEP plates. Synchronized adult C. elegans were washed off 5X PEP plates in M9 buffer, washed an additional four times in M9 buffer to remove excess OP50 E. coli, pelleted by a 1-min centrifugation at 800 × g, snap frozen in liquid nitrogen, and stored at −70°C until protein extraction. A total of 2 µL of high-salt reassembly (RAB) buffer [0.1 M 2-(N-morpholino)ethanesulfonic acid, 1 mM ethylene glycol bis-2-aminoethyl ether-N,N',N",n'-tetraacetic acid, 0.5 mM MgSO 4 , 0.75 M NaCl, 0.02 M NaF, pH 7.0] containing phenylmethylsufonyl fluoride and protease inhibitors was added per milligram of packed worm pellet wet weight and homogenized via sonication four times at 70% power for 8 s. This total soluble worm lysate was reserved for subsequent immunoblotting.
Protein immunoblotting. C. elegans protein immunoblotting was performed using Criterion apparatus (Bio-Rad) as recommended by the manufacturer (Bio-Rad Laboratories, Hercules, CA, USA) 18 . C. elegans protein preparations were diluted 5:1 with sample buffer (0.046 M Tris, 0.005 M ethylenediaine tetraacetate, 0.2 M dithiothreitol, 50% sucrose, 5% sodium dodecyl sulfate, 0.05% bromophenol blue), boiled for 10 min, and centrifuged at 16,100 × g for 2 min. A total of 10 µL of diluted protein preparations were loaded and resolved on precast 4-15% gradient sodium dodecyl sulfate polyacrylamide (SDS-PAGE) gel and transferred to polyvinylidene difluoride (PVDF) membrane as recommended by the manufacturer (Bio-Rad Laboratories, Hercules, CA, USA). PVDF membranes were blocked in 5% milk in phosphate-buffered saline (PBS) for 1 h before overnight incubation with primary antibody at 4°C. The next day, PVDF membranes were washed in PBS with 0.05% Tween, incubated at room temperature with horseradish peroxidase (HRP)-coupled secondary antibody for 2 h, and washed in PBS with 0.05% Tween before visualization. The dilutions and pertinent details for all primary and secondary antibodies used are listed in Supplementary Table 2. Enhanced chemiluminescence substrate (Bio-Rad Laboratories, Hercules, CA, USA) was added to the PVDF membrane, and chemiluminescence signals were detected with ChemiDoc-It [®2 510 Imager (UVP LLC, Upland, CA, USA). Relative intensity of chemiluminescence signals was measured with ImageJ Java 49 . At least three biological assay replicates were analyzed for statistical significance. For comparison of two groups with biological replicates run on independent or single immunoblots, a paired t-test, two-tailed, was used. For comparisons of three or more groups run on independent or single immunoblots, a repeated measures analysis of variance (ANOVA), followed by Tukey's multiple comparison post-test was used. Fulllength immunoblot images are available for all experiments as a Source Data file called: Immunblots.
C. elegans freeze cracking for immunofluorescent staining. C. elegans immunofluorescent staining was performed according to standard protocols as described by Janet Duerr 18,50 . Tau (high) animals were crossed with the CX10344 kyEx2454 strain (unc-25::calf-1::gfp, unc-25::CP450::mCherry, odr-1::dsRED) 24 to visualize the subcellular localization of tau protein relative to the endoplasmic reticulum via fluorescence microscopy. C. elegans were developmentally synchronized using a timed egg-lay and grown until all reached day one of adulthood at 20°C . Individual developmentally synchronized animals were placed in a 20 µL M9 buffer droplet on a 10-well Teflon-printed glass slide and washed to remove excess OP50 E. coli by moving to successive 20 µL M9 buffer droplets three times. Washed animals were transferred to a 10 µL 1% paraformaldehyde solution droplet on a poly-L-lysine coated glass slide and freeze cracked 50 . Adherent animals were exposed to a light fixation using methanol-acetone for 5 min each at −20°C. Freeze cracked, fixed animals were stained and washed as described. Animals were counterstained with 300 nM DAPI and mounted with ProLong™ Gold antifade (Invitrogen, Carlsbad, CA, USA) before super-resolution microscopy analysis. The dilutions and pertinent details for all primary and secondary antibodies used are listed in Supplementary Table 2. Three-dimensional structured illumination microscopy (3D-SIM) images of animals were acquired on a Nikon N-SIM-E Ti2 microscope (Tokyo, Japan) using a CFI SR HP Apochromat TIRF 100XC Oil (NA 1.49) objective, 561/647 nm diode lasers, and an ORCA-Flash 4.0 sCMOS camera (Hamamatsu Photonics K.K., Naka-ku, Hamamatsu City, Shizuoka, Japan). Image reconstruction and colocalization analysis was conducted using NIS-Elements (Nikon, Minato, Tokyo, Japan).
C. elegans RNA extraction. RNA was purified from snap-frozen packed C. elegans pellets using TRIzol Reagent as directed by the manufacturer's instructions 51 . RNA was resuspended in 50 uL sterile water. RNA concentration and purity were assessed using a NanoPhotometer® NP80 spectrophotometer (Implen GmbH, Munich, Germany). RNA integrity was assessed by 1% Tris/Borate/EDTA (TBE) agarose gel electrophoresis.
Quantitative RT-PCR. cDNA was prepared using iScript™ Reverse Transcription Supermix for RT-qPCR (Bio-Rad Laboratories, Inc., Hercules, CA, USA) as recommended by the manufacturer 51 . The primer set used to measure human Activation of the UPR ER XBP-1 transcriptional branch mediates detoxification of pathological tau via ERAD. a The UPR ER master transcription factor XBP-1 s controls transcription of ERAD genes. ERAD degrades proteins via the proteasome and autophagy, and ERAD loss of function can inhibit degradation of pathological tau. The presence of pathological tau in the cytoplasm is a stressor that impairs ERAD, leading to the accumulation of ER resident misfolded proteins and induction of the UPR ER25 . b Loss of the UPR ER XBP-1 branch promotes conversion of non-pathological tau to pathological tau, potentially by inhibition of ERAD, resulting in toxicity and subsequent neuronal cell death in C. elegans. Constitutive activation of the UPR ER master transcription factor XBP-1 s stimulates ERAD and promotes clearance of pathological tau, leading to reduction of toxicity and subsequent neuronal cell survival in C. elegans MAPT gene was (forward primer: 5′-GTGTGGCTCATTAGGCAACATCC-3′, reverse primer: 5′-CGTTCTCGCGGAAGGTCAG-3′). The primer set used for normalization was designed to detect C. elegans rpl-32 gene (forward primer: 5′-GGTCGTCAAGAAGAAGCTCACCAA-3′, reverse primer: 5′-TCTGCGGA-CACGGTTATCAATTCC-3′). Each genotype was tested with three biological replicates (including three technical replicates within each experiment) for each primer set. qPCR was performed using the iTaq™ Universal SYBR ® Green Supermix kit (Bio-Rad Laboratories, Inc., Hercules, CA, USA) in a 384-well plate on a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). Data were normalized within samples using an internal reference control gene (rpl-32).
Schematic illustrations. Original schematic illustrations were created using an academic license with the online application BioRender (Toronto, ON, CA).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data analyzed for this study are published in this manuscript and associated supplementary information. The source data underlying Figs