Alleviating chronic ER stress by p38-Ire1-Xbp1 pathway and insulin-associated autophagy in C. elegans neurons

ER stress occurs in many physiological and pathological conditions. However, how chronic ER stress is alleviated in specific cells in an intact organism is an outstanding question. Here, overexpressing the gap junction protein UNC-9 (Uncoordinated) in C. elegans neurons triggers the Ire1-Xbp1-mediated stress response in an age-dependent and cell-autonomous manner. The p38 MAPK PMK-3 regulates the chronic stress through IRE-1 phosphorylation. Overexpressing gap junction protein also activates autophagy. The insulin pathway functions through autophagy, but not the transcription of genes encoding ER chaperones, to counteract the p38-Ire1-Xbp1-mediated stress response. Together, these results reveal an intricate cellular regulatory network in response to chronic stress in a subset of cells in multicellular organism.


Author summary
The accumulation of unfolded proteins triggers the ER stress response (UPR), which allows cells to fight against fluctuations in protein expression under both physiological and pathological conditions. Severe acute ER stress responses can be induced by drug treatment. However, such intense ER stress rarely occurs ubiquitously in every cell type in vivo. Here, we designed a genetic system in the nematode C. elegans, which allows us to induce ER stress in specific cells, without drug treatment or any other external stimuli, and then to monitor the stress response. The p38 MAPK directly acts on the phosphorylation of IRE-1 to promote the stress response. Meanwhile, the insulin receptor function through autophagy activation to counteract the p38-IRE-1-XBP-1 pathway. Together, these results reveal an intricate cellular regulatory network in response to chronic stress in multicellular organism. Introduction Proteins destined for secretion or insertion into membranes enter the endoplasmic reticulum (ER) in an unfolded form and generally leave only after they have reached their native states. During normal function, a secretory cell may experience dramatic variations in the flux of new proteins through its ER in response to changes in demand. Disruption of the balance between secretory protein synthesis and the folding capacity of the ER activates a signaling network called UPR (the Unfolded Protein Response) in an attempt to maintain homeostasis [1]. The UPR reduces protein translation, increases expression of ER chaperones and enzymes to facilitate protein folding, and clears misfolded proteins for degradation [1]. Extensive or prolonged UPR activity signals that the accumulation of misfolded proteins has overwhelmed the compensatory mechanisms of the UPR, and an apoptotic response may be elicited [2]. In contrast, certain types of cancer use the protective role of the UPR to sustain their rapid growth [3,4]. Therefore, the UPR can serve either as an apoptotic executor or as a cytoprotector, depending on the cellular context. In multicellular organisms, the UPR is comprised of three branches, mediated by the transmembrane ER luminal sensors IRE1 (inositol requiring enzyme 1), PERK (double-stranded RNA-activated protein kinase [PKR]-like ER kinase), and ATF6 (activating transcription factor 6) [1]. Of the three, IRE1 is the only ER-stress sensor protein that is conserved from yeast to human [5]. In response to ER stress, IRE1 oligomerizes to activate an endoribonuclease domain that splices the mRNA encoding the transcription factor HAC1 (homologous to ATF/ CREB) in yeast [6,7] or XBP1 (X-box binding protein 1) in metazoans [8,9]. The spliced xbp1 mRNA produces a short, activated form of XBP1, called XBP1s, which activates expression of genes encoding chaperones and other ER-associated degradation proteins that expand the folding capacity of the ER and increase the breakdown of misfolded proteins [10]. In various experimental systems, the induction of ER stress is usually achieved by treatment with drugs, including the glycosylation inhibitor tunicamycin and the ER calcium ATPase inhibitor thapsigargin. It is worth noting, however, that the rapid induction of ER stress by toxins results in a massive and synchronous perturbation of ER function. In contrast, in a variety of human diseases, including Pelizaeus-Merzbacher disease, Huntington's disease, and osteogenesis imperfecta, the mutant or wild-type forms of secretory proteins gradually accumulate in the ER and cause pathological conditions in a subset of tissues or cell types [11,12]. In vivo models in which misfolded or/and unfolded proteins in the overloaded ER can be directly monitored in limited cell types will certainly help us to explore the molecular mechanisms underlying chronic ER stress responses [13].
Macroautophagy (hereafter referred to as autophagy) is a cellular degradation process initiated in response to stress. It attempts to restore metabolic homeostasis through lysosomal degradation of cytoplasmic organelles or cytosolic components [14]. Autophagy requires a core set of conserved proteins known as autophagy-related (Atg) proteins and is mediated by an organelle called the autophagosome [15]. Autophagosomal membranes originate from the ER, the Golgi complex, mitochondria, endosomes, and the plasma membrane [16,17]. During formation and expansion of the autophagosome precursor structure, called the phagophore, cytosolic material and damaged subcellular organelles are captured and enclosed. The complete closed autophagosome then fuses with lysosomes to become an autolysosome, inside which degradation and recycling of its content occurs. Autophagy is activated under ER stress conditions and many of the components that mediate autophagy have been identified as UPR target genes and are important for cells to survive severe ER stress [18,19].
The nematode C. elegans is transparent throughout its life cycle. We reasoned that if we could induce the ER stress response by overloading the cell with a non-secreted plasma membrane protein, we would then be able to detect the occurrence of UPR by directly observing the sub-cellular localization of this protein in vivo. Gap junction channels direct cell-to-cell communication and defects in gap junction formation and/or function are the cause of at least 10 human pathological conditions [20]. Moreover, the disease-related gap junction proteins accumulate in the ER and trigger the ER stress response [21][22][23][24]. Here, we found that overexpressing the gap junction protein UNC-9 in C. elegans neurons activates the IRE-1-XBP-1-mediated UPR in a cell-autonomous manner. Using this model, we further identified that the p38 MAP kinase PMK-3 functions through IRE-1 phosphorylation to regulate UPR. Loss of function of daf-2 (encoding the insulin-like receptor) suppresses the mis-localization defect of overexpressed UNC-9 in pmk-3, ire-1 or xbp-1 mutants. Intriguingly, although the reduced transcription of ER chaperone Bip/hsp-4 (immunoglobulin heavy chain-binding protein/heat shock protein 4) is not rescued by daf-2 mutation, autophagy induction is largely restored. This study reveals the intricate molecular interactions underlying the ER stress response in a limited subset of cells in an intact organism, and provides insights into the pathophysiology of many human degenerative diseases.

Excess UNC-9::GFP in DD/VD neurons triggers the ER stress response
We selected the DD/VD motor neurons in C. elegans to establish the in vivo ER stress model. The D-type motor neurons, including 6 DDs and 13 VDs, are a set of GABAergic neurons with cell bodies distributed along the ventral nerve cord and neuronal processes projecting both along the ventral cord and to the dorsal cord [25]. unc-25 encodes the GABA biosynthetic enzyme glutamic acid decarboxylase [26]. Using the Punc-25 promoter, the GFP (green fluorescent protein)-fused gap junction protein UNC-9 could be specifically overexpressed in DD/ VD neurons. We found that the ectopically expressed UNC-9::GFP is distributed in a punctate pattern along the neuronal processes and in the cell body region of DD/VD neurons (Fig 1A).
To understand in which cell compartments these puncta are distributed, we performed double staining with markers labeling various subcellular compartments. We found that there was very little overlap between the ectopic UNC-9::GFP puncta and the ER marker CYTB-5.1:: mCherry ( Fig 1B) or the Golgi marker mCherry::MANS ( Fig 1C). Instead, a large proportion of the UNC-9::GFP signal was distributed on the plasma membrane (co-localized with Myr:: mCherry) (Fig 1D and S1A  In an UNC-9::GFP knock-in (KI) line created using the CRISPR-Cas9 technique, we identified that the UNC-9::GFP KI displays a punctate pattern on neuronal processes and the cell surface, but not within the cell body region of DD/VDs (Fig 1G and S1B Fig). The fact that UNC-9 puncta on the cell membrane and processes appear in UNC-9 over-expressing worms (Fig 1A), but also in worms expressing UNC-9 from the endogenous promoter (the UNC-9::GFP KI strain) ( Fig 1G) suggests that UNC-9 puncta on the cell membrane and processes represent the normal localization pattern of the protein.
To test whether ER stress response and its regulators are required to achieve this proper localization in UNC-9 over-expressing worms, we firstly examined the hsp-4 induction. hsp-4 encodes the worm homolog of the ER chaperone protein Bip and its transcription is strongly induced by the stress response [9]. In the absence of ectopic UNC-9::GFP, the Phsp-4:: mCherry is widely but weakly distributed in multiple tissues (Fig 1H). When UNC-9::GFP was overexpressed in DD/VD neurons, we found that the mCherry signal was not increased in any other neurons or other tissues (S1C, S1D and S1E Fig). Instead, Phsp-4::mCherry was specifically induced in DD/VD neurons (Fig 1I and Fig 1J). There are more than 60 neurons with their cell bodies distributed along the ventral cord [25]. As shown in Fig 1I, only the cell bodies (asterisks) (18±0.7, N = 22) and the commissures (white arrows) of the 19 DD/VD neurons ( Fig 1I) could be distinctly visualized by the Phsp-4::mCherry signal. Double-labeling analysis further confirmed that the mCherry signal was co-distributed specifically with the Punc-25 promoter driven UNC-9::GFP expressing cells (Fig 1J and S1G Fig).

PLOS GENETICS
Alleviating chronic ER stress in C. elegans The IRE1-XBP1 UPR branch regulates the subcellular localization of overexpressed UNC-9::GFP in neurons The stress response triggered by UNC-9::GFP overexpression is likely employed to facilitate the proper localization of excess UNC-9::GFP protein. We suspected that the disruption of components involved in the underlying stress response process would abolish the punctate distribution pattern of UNC-9::GFP in DD/VD neurons. Hence, we conducted a genetic screen for mutants with altered UNC-9::GFP distribution. From this screen, we isolated an xbp-1 allele, xd131 (S2A Fig). In xbp-1(xd131) animals, the UNC-9::GFP signal no longer takes the characteristic punctate form. Instead, it is diffusely distributed in the cell body region of DD/ VD neurons. Except for a few small residual dots, there is no GFP signal on the neurites (Fig  2A). Other xbp-1 alleles, including zc12 and tm2482, display a similar phenotype to xd131 (Fig  2A). Introducing a wild-type copy of the active but not the inactive xbp-1 into DD/VD cells significantly rescues the UNC-9::GFP localization defect (Fig 2B), which suggests that loss of function of xbp-1s is responsible for the abnormal UNC-9::GFP distribution. ire-1 mutants show an almost identical phenotype to xbp-1 mutants (Fig 2A). In contrast, mutation of PERK (pek-1 mutant) or ATF6 (atf-6 mutant) did not affect the punctate pattern of UNC-9::GFP in DD/VDs (Fig 2A and S2B Fig). No obvious reduction on UNC-9::GFP mRNA level could be detected in ire-1, xbp-1, pek-1 or atf-6 mutant animals (S2C and S2D Fig).
The unc-9 gene is also expressed in body wall muscles [27]. myo-3 encodes the body wall muscle-specific myosin [28,29]. Using the Pmyo-3 promoter, we drove UNC-9::GFP protein expression in body wall muscle cells. In accordance with the observation that the body wall muscle cells are extensively coupled by gap junction connections [30,31], most of the UNC-9:: GFP signal is distributed in a punctate pattern along the muscle edge, particularly at the region where two neighboring muscle cells contact each other (S3A Fig) [27]. When the Pmyo-3:: UNC-9::GFP marker was introduced into xbp-1 or ire-1 mutant animals, however, we found that the UNC-9::GFP distribution was not affected by xbp-1 or ire-1 (S3A Fig). When we examined the UNC-9::GFP distribution in UNC-9::GFP KI strain created by CRISPR-Cas9, we found that the UNC-9::GFP signal retains its punctate pattern in both ire-1 and xbp-1 mutant animals, and is distributed on the neuronal processes and the cell surface of DD/VDs in a pattern that is indistinguishable from wild type (S3B Fig). The different requirement of ire-1 or xbp-1 in UNC-9::GFP distribution in DD/VD neurons versus muscle cells or knock in line implies that ire-1 and xbp-1 does not affect the protein localization of UNC-9 in general.
The Punc-53 promoter drives gene expression in a set of sensory neurons and interneurons [32]. Among them, PVP and PVQ, for instance, display extensive gap junction connections [31]. Indeed, characteristic UNC-9::GFP puncta were observed on the neurites and cell bodies with this Punc-53::UNC-9::GFP marker (S4A Fig). Interestingly, when either ire-1 or xbp-1 is mutated, the UNC-9::GFP signal becomes diffusely distributed to the perinuclear region of those neurons (S4A Fig). Meanwhile, the GFP signal along neurites is significantly decreased (S4A Fig). Together, the Ire1-Xbp1 pathway is required for the sub-cellular localization of ectopically expressed UNC-9 proteins in both DD/VD and Punc-53 expressing neurons.

The chronic stress triggers the IRE-1-mediated xbp-1 splicing
The excess UNC-9::GFP triggers an ER stress response in DD/VD neurons. We wondered whether ire-1 and xbp-1 regulate the subcellular localization of UNC-9::GFP in DD/VDs through their roles in the ER stress response. To address this question, we firstly examined in which subcellular compartments the UNC-9::GFP is localized in ire-1 and xbp-1 mutants. In wild-type animals, a few UNC-9::GFP puncta co-localize with the ER marker CYTB-5.1:: mCherry ( Fig 1B and Fig 2C). In ire-1 or xbp-1 mutant animals, however, the diffuse UNC-9::

PLOS GENETICS
Alleviating chronic ER stress in C. elegans GFP signal completely overlaps with the CYTB-5.1::mCherry signal ( Fig 2C). The ER localization of UNC-9::GFP signal suggests that the ectopically expressed UNC-9::GFP proteins probably are misfolded or unfolded in ire-1 and xbp-1 mutants. Under ER stress conditions, the transcription of genes encoding ER chaperones is upregulated to facilitate the correct folding of ER-synthesized proteins. Therefore, we further examined the expression of Phsp-4:: mCherry in ire-1 or xbp-1 mutant animals, and found that the specific induction of Phsp-4:: mCherry in DD/VD neurons was dramatically reduced (Fig 2D). We noticed that the basal expression level of the hsp-4 reporter is also significantly reduced by ire-1 or xbp-1 mutation (S1C, S1D and S1E Fig). Therefore, ire-1 and xbp-1 are required for both induced and basal expression of hsp-4 gene.
The induction of ER chaperone genes relies on activation of the transcription factor XBP-1s. Next, we tested whether the excess UNC-9::GFP activates the alternative splicing of xbp-1 mRNA. We fused a cDNA encoding mCherry in-frame with the inactive form of xbp-1 driven by the Punc-25 promoter ( Fig 2E) [33]. In the absence of ER stress, the regulatory intron cannot be spliced out, which results in "mCherry on". In contrast, the ER stress response will trigger the alternative spicing of xbp-1 mRNA and the regulatory intron can be removed, resulting in "mCherry off". When UNC-9::GFP was overexpressed in DD/VDs, we found that the "mCherry off" phenotype was induced in DD/VDs ( Fig 2F). In the absence of ire-1, the "mCherry off" phenotype was strongly suppressed and the mCherry signal reappeared in DD/ VDs ( Fig 2F). Together, these results provide evidence that UNC-9::GFP overexpression in DD/VDs triggers IRE-1-mediated xbp-1 splicing.

Beneficial effect of the IRE-1-XBP-1 branch during prolonged chronic stress
Utilizing the distinct UNC-9::GFP localization pattern, we explored at which stage the ire-1-xbp-1 branch is required for the stress response triggered by excess UNC-9::GFP. From the embryonic stage to larval stage 3 (L3), we found that the majority of UNC-9::GFP signal in ire-1 mutants displays a punctate pattern similar to wild type (Fig 3A), which suggests that the ire-1-mediated stress response is not required for the proper localization of excessive UNC-9 protein during this period. However, when ire-1 animals progress to the L4 stage, the UNC-9:: GFP signal on the neuronal processes is significantly decreased ( Fig 3A). Meanwhile, more UNC-9::GFP signal became diffusely distributed in the cell body region ( Fig 3A). Thus, the absence of the stress response has no obvious effect on the subcellular localization of ectopically expressed UNC-9::GFP until the late developmental stage.
Next, we examined whether prolonged failure of the UPR in DD/VD neurons leads to cell death. With the Punc-25::mCherry marker, around 17 DD/VD neurons could be unambiguously identified in wild-type animals ( Fig 3B). A similar number of DD/VDs was observed when excess UNC-9::GFP was present ( Fig 3B). In contrast, when ire-1 or xbp-1 was removed, the total number of DD/VDs dropped significantly ( Fig 3B). Notably, the reduced cell number occurs only under stress conditions. Neither ire-1 nor xbp-1 mutation caused neuronal loss in the absence of excess UNC-9::GFP ( Fig 3B). Introducing a wild-type copy of the ire-1 or xbp-1 gene into DD/VDs significantly rescued the neuronal loss in ire-1 or xbp-1 mutant animals respectively ( Fig 3B). Hence, the IRE1-XBP1-mediated stress response has a beneficial effect during prolonged chronic stress in vivo.

pmk-3 is required for the ER stress response
In addition to xbp-1(xd131), the unbiased genetic screen also identified a pmk-3 allele, xd74. In xd74 animals, the UNC-9::GFP signal is diffusely distributed in the cell body region of DD/ VDs and is greatly reduced in DD/VD neuronal processes (Fig 4A), which mimics the ire-1 or xbp-1 mutant phenotype. Whole-genome sequencing identified a G-to-A mutation which affects the splicing donor site of the fourth intron in the pmk-3 gene (S5A Fig). RT-PCR combined with sequencing analysis further showed that the xd74 mutation leads to two distinct splicing alterations (S5B Fig), which both result in a premature stop codon in the pmk-3 gene product. Other pmk-3 alleles, including pmk-3(ok169) and pmk-3(tm745), display similar UNC-9::GFP localization defects in DD/VD neurons to pmk-3(xd74) animals (Fig 4A and 4B). Introducing a wild-type copy of pmk-3 into DD/VDs using the Punc-25 promoter greatly rescued the UNC-9::GFP distribution defect in pmk-3 mutant animals (Fig 4A and 4B), which indicates that loss of function of pmk-3 is responsible for the UNC-9::GFP mis-localization. pmk-3 encodes one of the three p38 MAPKs in C. elegans and the kinase domain is essential for PMK-3 function in UNC-9::GFP localization in DD/VDs (Fig 4A and 4B). Because animals carrying a single mutation of any of the other MAPK components do not show any obvious UNC-9::GFP distribution abnormality (S1 Table), we concluded that p38/PMK-3 alone plays an important role in the protein localization of ectopic UNC-9::GFP in DD/VD neurons.
pmk-3 mutation does not affect the punctate distribution of UNC-9::GFP at the cell-cell junction region in muscle cells (S3A Fig). Therefore, pmk-3 is probably not involved in the subcellular localization of UNC-9 protein in general. The similarity between pmk-3 and ire-1 or xbp-1 mutant animals strongly hinted that pmk-3 is also involved in mediating the stress response triggered by UNC-9::GFP-overexpression. The hsp-4 expression is specifically induced in DD/VD neurons by UNC-9::GFP overexpression. Hence, we examined the induction of the Phsp-4::mCherry reporter in DD/VDs in pmk-3 mutants. In wild type, the commissures of DD/VD neurons could be easily detected with Phsp-4::mCherry. In contrast, the 19 DD/VD commissures were hardly visible in pmk-3 mutants (Fig 4C). We noticed that the individual DD/VD cell bodies were still identified with Phsp-4::mCherry in pmk-3 mutants ( Fig  4C), which suggests that the stress response is partially affected by pmk-3. We further examined the UNC-9::GFP distribution and found that the diffuse UNC-9::GFP signal in pmk-3 mutants completely overlaps with the ER marker CYTB-5.1::mCherry in DD/VD cell body regions (Fig 4D), which is also similar to the ire-1 or xbp-1 mutants.

PLOS GENETICS
Alleviating chronic ER stress in C. elegans
We showed above that the kinase-defective (KD) mutant form of PMK-3 (pmk-3 KD) lost its capability to rescue the UNC-9::GFP localization defect in pmk-3 mutants (Fig 4A and 4B). This suggests that the kinase activity of PMK-3 is crucial for its function. The cytosolic region of IRE-1 contains three putative p38 MAPK phosphorylation sites: Ser519 (S519), Ser802 (S802) and Thr949 (T949). T949 is located on the C-terminal RNase domain of IRE-1 and phosphorylation of the corresponding site in human (T973) contributes to the full RNase activity of Ire1 in vivo [34]. To test whether PMK-3 regulates the stress response through phosphorylation of IRE-1, we performed the following experiments. Firstly, we tested whether the presence of PMK-3 altered the phosphorylation level of IRE-1. By incubating IRE-1 protein with either the wildtype or kinase-inactive form of PMK-3, we found that the wild-type but not the KD PMK-3 enhanced the phosphorylation of IRE-1 (Fig 5B). Second, we tested whether PMK-3 can associate with IRE-1. After affinity purification, we found that the Flag-tagged PMK-3 protein was coimmunoprecipitated with IRE-1-Myc (Fig 5C), which is suggestive of direct binding between PMK-3 and IRE-1. Catalytically inactive MAP kinases are often used to detect interactions between MAP kinases and their substrates or interacting partners [35]. Indeed, the constitutively inactive form of PMK-3 associated with IRE-1 with even higher affinity ( Fig 5B). Third, we purified PMK-3 and the GST-tagged intracellular domain of IRE-1 protein and performed in vitro kinase assay. Incubating IRE-1 with PMK-3, a significant increase of kinase activity was detected compared to control (without PMK-3) (S6C Fig). We further examined the phosphorylation level of IRE-1 with antibody against p(phosphorylated)-Thr-Pro site. We found that the phosphorylation of IRE-1 was significantly increased by PMK-3 (Fig 5D and Fig 5E). Fourth, we mutated the S519, S802 or T949 residues individually to alanine or glycine and found that the T949A substitution significantly decreased the rescuing activity of IRE-1 (Fig 5F). Further mutating the S519 and S802A sites (S519G/S802G/T949A triple mutation) completely blocked the rescuing activity of PMK-3 ( Fig 5F). Together, these results provide evidence that PMK-3 phosphorylates IRE-1 to regulate the ER stress response.

daf-2 suppresses the ER stress response defect in pmk-3, ire-1 and xbp-1 mutants
Considering that ER stress underlies many pathological conditions, we utilized the xbp-1 (xd131) strain and performed a further genetic suppressor screen to explore whether the malfunctioning ER stress response can be repressed. From the screen, we identified the xd211 mutation. In the presence of xd211, the punctate pattern of UNC-9::GFP was restored along the neurites and in the cell body region of DD/VD neurons (Fig 6A and 6B). Genetic mapping and whole-genome sequencing identified a threonine-to-lysine change at codon 926 in the insulin/IGF-1 receptor DAF-2. The UNC-9::GFP distribution defect of xbp-1(xd131) or xbp-1 (zc12) was also suppressed by the daf-2(e1370) allele (Fig 6A and 6C) (Materials and methods).  6. daf-2 suppresses pmk-3, ire-1 and xbp-1
We then tested whether daf-2 could suppress the pmk-3, ire-1 or xbp-1 mutant phenotypes by acting through autophagy induction. In the absence of excess UNC-9::GFP, the basal level of autophagy in DD/VD cells is slightly increased by daf-2 mutation (Fig 7F). In the presence of UNC-9::GFP, the autophagy induction in DD/VD cells in daf-2 mutants remains at a similar level to that in wild type (Fig 7F). Intriguingly, when the daf-2 mutation was introduced into either ire-1, xbp-1 or pmk-3 mutants, the number of mCherry::LGG-1 puncta in DD/VD neurons is significantly increased compared to the ire-1, xbp-1 or pmk-3 single mutants (Fig 7F). Because the reduction of autophagy activity in ire-1 or xbp-1 or pmk-3 mutant animals could be counteracted by loss of daf-2 function, we wondered whether increasing the autophagy activity alone could suppress the UPR failure caused by the defective pmk-3-ire-1-xbp-1 Fig 7. Autophagy induction is restored by daf-2 in pmk-3, ire-1 and xbp-1 mutants. (A) The reduction of Phsp-4::mCherry expression (red) in ire-1 or xbp-1 mutants is not rescued by daf-2. (B) The partial reduction of Phsp-4::mCherry expression pathway. Starvation treatment activates autophagy [36,37]. In well-fed ire-1 mutants, most of the DD/VD cell bodies (more than 95%) contain diffusely distributed UNC-9::GFP signal. After starvation, the UNC-9::GFP adopted a punctate distribution pattern in a significant proportion (about 58%) of DD/VD cells (Fig 7G). This implies that starvation-activated autophagy suppresses the UPR failure in ire-1 mutant animals. Furthermore, when the autophagy process was disrupted by lgg-1 mutation, we found that the number and intensity of UNC-9::GFP puncta was significantly increased (Fig 7H). Our attempts to introduce autophagy mutations into either daf-2 xbp-1 or daf-2;ire-1 animals failed, probably because disturbed autophagy has a severe effect on the viability of these double mutants. Therefore, we turned to the autophagy inhibitor chloroquine. After chloroquine treatment [38], we found that the daf-2 suppression on pmk-3 was significantly reduced (Fig 7I). The daf-2 suppression on xbp-1 was also reduced by chloroquine treatment (Fig 7I). In contrast, the reduction of the daf-2 suppression on ire-1 was not obvious (Fig 7I). We suspected that the relatively weak potency of drug treatment on C. elegans neurons may be responsible for the various response from different genetic backgrounds [38]. Alternatively, autophagy maybe only part of the mechanisms by which daf-2defficiency improves UNC-9 processing. Taken together, the autophagy activation may contribute to the suppression effect of daf-2 on the defective pmk-3-ire-1-xbp-1-mediated ER stress response.

Discussion
Accumulation of exogenous or abnormal misfolded proteins in neurons contributes to the pathology associated with neurodegeneration. Our understanding of the molecular events linking ER stress to neurodegeneration requires deeper knowledge of the mechanisms that underlie ER homeostasis, including how gene transcription is activated and autophagy is regulated in specific neuronal cells in vivo.
In response to ER stress, transcription of the ER chaperone gene is activated by the IRE-1/ XBP-1 pathway [39]. Indeed, when worms are exposed to tunicamycin or dithiothreitol (DTT), or to global protein misfolding caused by heat shock, a GFP reporter driven by the hsp-4 promoter is strongly induced in a wide range of tissues and cells [9,40]. Pan-neuronal ER stress also triggers hsp-4 expression cell-non-autonomously in gut and hypodermis [41]. In contrast to previous observations though, UNC-9::GFP overexpression in DD/VD motor neurons induces ER stress only within DD/VD cells. C. elegans contains 302 neurons and the Punc-25 promoter limits UNC-9::GFP overexpression to less than 30 neurons (mostly DD and VD neurons). Although the detailed mechanisms underlying this restricted effect remain elusive, it is possible that the broadness or/and the severity of ER stress in the nervous system determines whether the stress response can spread to neighboring tissues.
The cell-autonomous stress response triggered by overloading of UNC-9::GFP protein ensures that UNC-9::GFP adopts its distinctive punctate pattern in DD/VD neurons. In

PLOS GENETICS
Alleviating chronic ER stress in C. elegans contrast, when UPR fails, such as in ire-1 or xbp-1 mutants, a diffuse UNC-9::GFP signal accumulates in the ER. Evidently, with no need for drug treatment or any other external stimuli, the distinct UNC-9::GFP subcellular localization offers a dual marker to monitor functional versus defective UPR at single-cell resolution in an intact animal. Notably, the subcellular localization of endogenous UNC-9 and of ectopic muscle-expressed UNC-9::GFP is unaffected by ire-1 or xbp-1. Thus, the IRE-1/XBP-1 UPR is not involved in the UNC-9 protein localization under normal growth condition. Previous studies demonstrated that that the UPR is also function in absence of external ER stress and the UPR genes are critical for maintaining secretory protein metabolism [33,42]. Here, the UNC-9 puncta on the neuronal cell membrane and processes appear in UNC-9 over-expressing worms, but also in worms expressing UNC-9 from the endogenous promoter (the UNC-9::GFP Knock-in strain). To achieve this proper localization in UNC-9 over-expressing cells, the ER stress response and its regulators are specifically required. Thus, by comparing UNC-9::GFP localization patterns in different systems, we revealed a scenario that when there is no ER stress (normal UNC-9 expression), there is no need for IRE-1/XBP-1. When ER stress is produced (UNC-9 overexpression), the IRE-1/XBP-1-mediated UPR is needed for protein localization.
Aided by unbiased genetic screens, we further identified that pmk-3/p38 MAPK is involved in the ER stress response triggered by excess UNC-9::GFP. Among the three p38 MAPKs in C. elegans [43], PMK-1 is extensively involved in innate immunity and subsequently in DAF-2 or/ and DAF-16-related longevity [44][45][46][47][48][49]. While the ire-1-xbp-1 pathway protects animals from Pseudomonas aeruginosa infection, inactivation of the PMK-1-mediated immune response suppresses the lethal effect of P. aeruginosa infection in xbp-1 mutants [50]. Here, in the context of UNC-9::GFP overexpression, pmk-3 functions similarly to ire-1 and xbp-1. In contrast, neither PMK-1 nor the upstream MAPKK (MAPK kinase) SEK-1 nor MAPKKK (MAPK kinase kinase) NSY-1 [49] is required for UNC-9::GFP-induced UPR (S1 Table). What determines the differential roles of p38 MAPKs in various biological processes? Compared to the wide range of reactions from multiple tissues during pathogen infection, we suspect that protein overloading in a handful of neurons triggers a rather mild stress response. Although the core components (for instance, ire-1 and xbp-1) are shared, other regulatory components are employed in a tissueor/and intensity-dependent manner. How does PMK-3/p38 participate in this cell-autonomous chronic ER stress response? A previous study showed that p38 MAPK directly phosphorylates the activated form of XBP1 and enhances its nuclear migration [51]. Here, however, we found that the xbp-1 mutants in which the putative phosphorylation sites of p38 MAPK (T191 and S283) were disrupted could fully rescue the UNC-9::GFP mis-localization defect in xbp-1 (the percentage of DD/VD cell bodies showing defective UNC-9::GFP distribution was dropped to 8.7% and 11.3% respectively). In addition, overexpressing ire-1 suppresses the pmk-3 mutant phenotype, while upregulation of pmk-3 does not bypass the functional requirement for ire-1. This suggests that pmk-3 functions upstream of ire-1. Consistent with this idea, PMK-3 binds to IRE-1 and the phosphorylation status of IRE-1 is affected by the presence of functional PMK-3. Among three putative p38 MAPK phosphorylation sites in IRE-1, T949 is located in the C-terminal RNase domain. Phosphorylation of the corresponding site does not have a direct impact on the splicing activity in vitro, but has a noticeable effect in vivo [34]. Here, we showed that the alternative splicing of xbp-1 mRNA in DD/VD neurons is partially affected by pmk-3 and mutation of this T949 site significantly decreased the rescue activity of IRE-1. Given that phosphorylation on the RNase domain may be involved in the recruitment of unknown cofactors to maximize the RNase activity of IRE-1 [34], it is fascinating to speculate that the regulatory role of PMK-3/p38 MAPK is particularly important in fighting against chronic ER stress caused by gradual accumulation of mutated or overexpressed protein, which is common under pathogenic conditions in many human diseases.
The relationship between insulin signaling and UPR regulation is rather complicated [52][53][54]. Worms with loss-of-function mutations in the insulin/IGF-1 receptor daf-2 are longlived, and both ire-1 and xbp-1 make a large contribution to the long lifespans of daf-2 mutants and also increase their ER stress resistance [36,55]. In addition, the daf-2 could restore protein secretion in ire-1 and xbp-1 mutants in a daf-16 dependent manner [56]. The daf-2 mutation also improved the ER stress resistance of xbp-1 and ire-1 mutant animals when they were treated with low but not high tunicamycin concentrations, and DAF-16 acts through ER-associated degradation systems independently of ire-1 [56]. A more recent study further showed that the complexity of PVD neuron dendrite arborization is greatly reduced by a loss-of-function mutation in ire-1, and further reducing insulin signaling restored the normal PVD neuronal morphology [57]. However, xbp-1 is not involved in PVD morphogenesis at all, which suggests that the antagonistic role of insulin signaling may not act through the non-canonical splicing of xbp-1 mRNA as in the UPR process [57]. Here, we found that daf-2 mutation suppresses the defective ER stress response in both ire-1 and xbp-1 animals. Because the cell autonomous reduction of hsp-4 expression in ire-1 or xbp-1 mutants is not restored by daf-2, we suspected that daf-2 may act on alternative path to regulate the stress response. Intriguingly, autophagy induction was restored in ire-1 or xbp-1 mutant cells by loss of daf-2 function. It has been shown that the daf-2 mutants exhibit increased levels of autophagy, and autophagy is required for their extended lifespan [37]. In addition, the basal level of autophagy is increased by daf-2 mutation in both wild-type and ire-1 mutant animals [56,58,59]. Here, we found that although the excess UNC-9-triggered autophagy up-regulation in DD/VD neurons is not further enhanced by daf-2 mutation, the reduction of autophagy in ire-1 or xbp-1 mutants was indeed suppressed by daf-2. The FOXO forkhead transcription factors are key transcriptional regulators of autophagy [60]. Indeed, the autophagy induction by excess UNC-9::GFP is completely suppressed by daf-16 mutation (the number of LGG-1::mCherry puncta in DD/VDs was reduced to 2.58 ± 0.12 in daf-16 animals, N = 148).Therefore, we suspected that in ire-1 or xbp-1 single mutants, the DAF-16/FOXO remains inactive due to the presence of DAF-2. Since neither transcription of ER chaperone genes (for instance hsp-4/Bip) nor autophagy induction could occur in ire-1 or xbp-1 single mutants, the proper localization of overexpressed protein could not be achieved. In contrast, when both ire-1/xbp-1 and daf-2 are removed, the DAF-16 is activated due to the absence of DAF-2. Hence, although the transcriptional upregulation of ER chaperone genes (such as hsp-4/Bip) is missing-because of the lack of ire-1/xbp-1 pathway, the active DAF-16 could act through autophagy or other protein degradation process or ER homeostasis to facilitate the proper localization of overexpressed protein. Intriguingly, the accumulation of misfolded proteins in the ER appears not to directly activate DAF-16 [56]. Therefore, the DAF-16 function is crucial only under conditions in which both DAF-2/insulin receptor and the ire-1-xbp-1 branch of the UPR are removed. Nevertheless, the availability of two parallel systems (S7A Fig) suggests that induction of autophagy using small molecules may be an alternative approach to treat ER stress response failure in many degenerative diseases.

DNA constructs and transgenic animals
The unc-25 promoter was inserted between the BamHI and SphI sites of the pSM-GFP vector (a generous gift from Dr. Kang Shen) and the UNC-9 cDNA was inserted into the BamHI site. xbp-1, ire-1 and pmk-3 cDNAs were amplified by reverse transcription. The ER stress reporter plasmid Punc-25::XBP-1us::mCherry was modified from XB3 plasmid (a generous gift from Dr. Christopher Rongo). Punc-25::cytb-5.1::mCherry was kindly provided by Dr. Yingchuan Qi. Punc-25:: mCherry::MANS, Punc-25:: mCherry::RAB-7, Punc-25:: mCherry::RAB-5, and Punc-25::mCherry::LGG-1 were constructed by the recombination method. Transgenic animals were produced as previously described [62]. The wild type pmk-3 and the kinase dead form of pmk-3(pmk-3KD) were amplified by PCR and then inserted between the EcoRI and XbaI sites of pFLAG-CMV-2 plasmid. The cytosolic C-terminus of IRE-1 was amplified by PCR and inserted between the NheI and KpnI sites of pcDNA 3.1/myc-His vector. Integrated strains were obtained by UV irradiation. The strain BCN1071 which expresses Phsp-4:: mCherry was kindly provided by Ben Lehner. All integrated transgenic animals were outcrossed at least 3 times. The corresponding transgenes are listed in S2 Table. CRISPR/Cas9-mediated gene editing The UNC-9::GFP knock-in strain was created using sg1:CGTATGGTTGCAACTCACGCC GG and sg2: CCGGAGAACTACCCTGTTACGAG. The GFP sequence was inserted before the stop codon TGA of unc-9 gene. This UNC-9::GFP knock in strain (PHX727) was generated by SunyBiotech and was verified with DNA sequencing. This strain was outcrossed with N2 for 4 times before use.

Image collection and quantification
Animals were mounted on 2% agar pads in M9 buffer containing 1% 1-phenoxy-2-propanol and examined by fluorescence microscopy unless indicated otherwise. Fluorescence photographs were taken using a Zeiss Axioimager A1 with an AxioCam digital camera and Axiovision rel. 4.6 software (Carl Zeiss) or an IX81 Olympus inverted confocal microscope. All images were taken at the young adult stage unless specifically indicated. The UNC-9::GFP distribution defect in the cell body region was quantified by counting the percentage of cell bodies with diffusely localized UNC-9::GFP. To quantify the UNC-9::GFP puncta distribution in DD/VD cell body region, the size of UNC-9::GFP co-localized with either myr::mCherry, mCherry::RAB-5, mCherry::RAB-7 or mCherry::LGG-1 was measured using ImageJ. The ratio of the co-labeled UNC-9::GFP to the total UNC-9::GFP was calculated. Signals from 5 DD/VD cells were scored for each genotype. To quantify the number of DD/VD neurons, the Punc-25:: mCherry marker was used. Phsp-4::mCherry was used to detect induction of the chaperone gene hsp-4 in the ER stress response. The average Phsp-4::mCherry fluorescence intensity in the anterior and posterior intestine regions was measured and normalized with the wild type. To quantify the xbp-1 splicing activity, the Punc-25::XBP-1us::mCherry plasmid was used and DD/VD cell bodies with mCherry signal were counted. To characterize the distribution of UNC-9::GFP under starvation conditions, well-fed L4 ire-1(v33);xdIs15 worms were collected and washed 8 times with M9 buffer. After sedimentation, worms were plated on NGM plates without food. After 24 hours, the starved adult worms were examined using Leica TCS SP8 confocal microscopy. All graphical data are presented as mean ± s.e.m. Two-tailed unpaired and paired Student's t-tests were performed for comparison between two groups of samples. To compare multiple groups, one-way ANOVA was used with an appropriate multiple comparisons post hoc test (the test used is stated in each Fig legend). � P < 0.05; �� P < 0.01; ��� P < 0.001; NS, not significant.

Quantification of autophagy
To analysis autophagy activity in DD/VD neurons, Punc-25::mCherry::LGG-1 was constructed. Images at the young adult stage were collected and the numbers of mCherry::LGG-1 foci in DD/VD cell body regions were counted. To characterize mCherry::LGG-1 puncta in strains carrying daf-2((e1370 ts), embryos were collected and raised at 25˚C until L2/L3 stage before phenotypic examination.

Immunoprecipitation and western blotting
HEK293T (human embryonic kidney) cells were maintained in Dulbecco's modified Eagle's medium containing 12% FBS. For cell transfection, polyethylenimine (Polysciences) was used according to the manufacturer's instructions. The clones Myc-tagged cytoplasmic region of IRE-1 and the Flag-tagged full-length wild-type or kinase inactive form of PMK-3 were cotransfected into HEK293T cells. 24 hours after transfection, cells were harvested and lysed for 30 min at 4˚C. After centrifugation, the supernatants were incubated with anti-FLAG M2 affinity gel beads at 4˚C overnight and then washed three times with washing buffer and incubated with SDS-sample buffer. Samples were resolved by standard immunoblotting techniques. All the co-immunoprecipitation experiments were repeated at least three times.

IRE-1 phosphorylation
The Myc-tagged cytoplasmic region of IRE-1 and the Flag-tagged full-length wild-type or kinase inactive form of PMK-3 were co-expressed in HEK293T cells and then the IRE-1 protein was immunoprecipitated with Myc antibodies. The phosphorylation level of IRE-1 was examined with P-Thr-Pro-101antibody (Cell Signal Technology #9391), which detects the conserved phosphorylation site (S/TP) for p38 MAPK. The signal intensity of individual protein bands was analyzed with ImageJ. The relative protein level was determined by subtracting the respective FLAG control. All the co-immunoprecipitation experiments were repeated at least three times.

Protein purification and in vitro kinase assay
The pGEX-4T-1-GST::IRE-1 plasmid was transformed into BL21(DE3) competent cells. The fusion protein was induced by 1mM IPTG at 18˚C for 24hr and then purified by Glutathione Sepharose 4B. The Flag::PMK-3 and myc::SEK-1 plasmids were transfected into HEK293T cells. 24 hours after transfection, cells were lysed and purified by anti-FLAG M2 affinity beads. Subsequently, the protein samples were concentrated and replaced with the kinase buffer (50 mM Tris-HCl (pH 7.6), 10 mM MgCl, 1 mM DTT). The phosphorylation of IRE-1 by active PMK-3 was monitored using a Universal Kinase Activity Kit (Catalog Number EA004, R&D Systems). Purified IRE-1 and 200μM ATP were mixed with purified PMK-3 and coupling phosphatase CD39L2 in phosphatase buffer (250 mM HEPES, 1.5 M NaCl, 100 mM MgCl2, and 100 mM CaCl2). As a control group, PMK-3 was replaced by 1x kinase buffer. The reaction is carried out in 50 μL for 10 minutes at room temperature. The coupling phosphatase CD39L2 can generate inorganic phosphate from ADP resulting from the kinase reaction. Malachite Green Reagent was added to terminate the reaction. After incubation for 20 minutes at room temperature to stabilize the color development. The amount of phosphorylated protein was determined using a microplate reader set to 620 nm. The experiments were repeated three times independently. To further determine the phosphorylation of IRE-1, anti-P-Thr-Pro-101 antibody was used to detect the phosphorylation site (S/TP) for p38 MAPK. The GST::IRE-1 was incubated with Flag::PMK-3 in kinase buffer (50 mM Tris-HCl pH 7.6, 10 mM MgCl, 1 mM DTT, 400μΜ ATP) at 25˚C for 30 minutes. As a control group, the Flag::PMK-3 was replaced by 1x kinase buffer. The samples were boiled with loading buffer to terminate the reaction. The experiments were repeated three times independently.