The ATF6β-calreticulin axis promotes neuronal survival under endoplasmic reticulum stress and excitotoxicity

While ATF6α plays a central role in the endoplasmic reticulum (ER) stress response, the function of its paralogue ATF6β remains elusive, especially in the central nervous system (CNS). Here, we demonstrate that ATF6β is highly expressed in the hippocampus of the brain, and specifically regulates the expression of calreticulin (CRT), a molecular chaperone in the ER with a high Ca2+-binding capacity. CRT expression was reduced to ~ 50% in the CNS of Atf6b−/− mice under both normal and ER stress conditions. Analysis using cultured hippocampal neurons revealed that ATF6β deficiency reduced Ca2+ stores in the ER and enhanced ER stress-induced death. The higher levels of death in Atf6b−/− neurons were recovered by ATF6β and CRT overexpressions, or by treatment with Ca2+-modulating reagents such as BAPTA-AM and 2-APB, and with an ER stress inhibitor salubrinal. In vivo, kainate-induced neuronal death was enhanced in the hippocampi of Atf6b−/− and Calr+/− mice, and restored by administration of 2-APB and salubrinal. These results suggest that the ATF6β-CRT axis promotes neuronal survival under ER stress and excitotoxity by improving intracellular Ca2+ homeostasis.


Results
Expression of ATF6β in the CNS and other tissues. We first verified the tissue distribution of ATF6β in mice. Quantitative real-time PCR (qRT-PCR) revealed that Atf6b mRNA was broadly expressed, but was most highly expressed in the hippocampus of the brain among the tissues analyzed (Fig. 1A). Further analysis of cultured cells revealed that expression of Atf6b mRNA was higher in hippocampal neurons than in cortical neurons and astrocytes under normal conditions (Fig. 1B). Consistently, in situ hybridization revealed that Atf6b mRNA was highly expressed in hippocampal neurons (Fig. 1C). These patterns were in contrast with those of Atf6a mRNA, which was more ubiquitously expressed (Fig. S1A,B). There was no significant difference in Atf6b mRNA levels between male and female mice (Fig. S1C).
We next analyzed expression of Atf6b mRNA under ER stress. Treatment of cultured hippocampal neurons with the ER stressors tunicamycin (Tm) and thapsigargin (Tg) significantly increased expression of Atf6b mRNA (1.5 to twofold increase) (Fig. 1D), although these increases were smaller than those in expression of Atf6a mRNA (5-6.5-fold increase) (Fig. S1D). At the protein level, both the full-length 110 kDa protein (FL) and a cleaved N-terminal 60 kDa fragment (NTF) of ATF6β were detected in primary hippocampal neurons. The level of this fragment was low under normal conditions, but increased as early as 2 h after Tg treatment or 1 h after dithiothreitol (DTT) treatment, the latter another ER stressor (Fig. 1E). These results suggest that ATF6β functions in neurons especially under ER stress.
Calr is a unique target gene of ATF6β in the CNS. To identify downstream molecules of ATF6β in the CNS, RNA-sequencing was performed using hippocampal brain samples from wild-type (WT) and Atf6b −/− mice. A total of 55,531 genes were examined. We filtered genes in two ways. When filtering genes stringently with FPKM values in WT mice higher than 10 and q values smaller than 0.05, only 2 downregulated genes and 4 upregulated genes were identified in Atf6b −/− mice (Table 1). Although expression of Atf6b mRNA was observed to some extent in Atf6b −/− mice, this may be due to the presence of the 5' Atf6b transcript with exon 1-9 in these mice, as exon 10 and 11 were deleted by homologous recombination 7 (Fig. S2). Besides Atf6b, only Calr, which encodes CRT, a molecular chaperone in the ER with a high Ca 2+ -binding capacity, was downregulated in Atf6b −/− mice (Table 1). By contrast, in case filtering genes less stringently with FPKM values in WT mice higher than 10 and p values smaller than 0.05, 22 downregulated genes and 27 upregulated genes were identified in Atf6b −/− mice (Table S1). Calr was again identified as a gene downregulated in Atf6b −/− mice, and interestingly, six ER stress-responsive genes, namely, Hpsa5 (GRP78), Pdia4 (ERP72), Dnajb11, Atf4, Wfs1, and P4ha1 were upregulated in Atf6b −/− mice (Table S1 right columns). These results suggest that Atf6b deficiency may cause mild ER stress in the brain under normal conditions. RNA-sequencing also indicated that expression level of Calr was highest among the major molecular chaperones in the ER in WT brains (Table S2). Taken together with previous reports demonstrating possible roles of ATF6β in the expression of molecular chaperones in the ER 7,8,16 , we decided to focus on ATF6β-CRT axis in further experiments.
The effect of Atf6b deletion on CRT expression was next analyzed in different tissues under normal conditions ( Fig. 2A,B). Both qRT-PCR ( Fig. 2A) and western blotting (Fig. 2B) revealed that CRT expression was significantly lower in the CNS, but not in other tissues tested, in Atf6b −/− mice than in WT mice.
Effect of ATF6β deletion on CRT promoter activity. To analyze the role of ATF6β in CRT expression at the promoter level, reporter assays were performed with chloramphenicol acetyltransferase (CAT) plasmids containing 1763 bp (pCC1), 415 bp (pCC3), and 115 bp (pCC5) of the mouse CRT promoters 17
Effect of ATF6β deletion on expression of molecular chaperones in the ER in primary hippocampal neurons. The effect of ATF6β deletion on the expression of molecular chaperones in the ER was next examined under normal and ER stress conditions using WT and Atf6b −/− hippocampal neurons. Consistent with the results obtained using mouse tissues under normal conditions, RT-qPCR revealed that Calr expression was significantly lower in Atf6b −/− neurons than in WT neurons under both control and ER stress conditions, with the latter induced by Tg (Fig. 3A upper row) and Tm ( Fig. 3B lower row). By contrast, expression of other molecular chaperones in the ER such as Canx (calnexin), Hsp90b1 (GRP94), and Hspa5 (GRP78) was temporally lower in Atf6b −/− neurons than in WT neurons after stimulation with Tg ( Fig. 3A upper row) or Tm (Fig. 3A lower row). Similarly, western blot analysis revealed that expression of CRT protein was constitutively lower in Atf6b −/− neurons than in WT neurons, while protein expression of other molecular chaperones in the ER was similar in Atf6b −/− and WT neurons under both normal and ER stress conditions (Fig. 3B).
Finally, gene-silencing experiments were performed to exclude the possibility that the effect of Atf6b deletion on CRT expression was indirect due to the long-term absence of Atf6b. Transfection of Neuro 2a cells with two sets of ATF6β-targeting siRNAs (ATF6β-siRNA1 and ATF6β-siRNA2) reduced Atf6b expression to 30% and 43% and reduced Calr expression to 62% and 66%, but did not affect Hspa5 (GRP78) expression, compared with that in control-siRNA-transfected cells (Fig. S6A).
Effect of ATF6β deletion on Ca 2+ homeostasis in primary hippocampal neurons. As CRT is involved in the regulation of the ER Ca 2+ capacity 19 , Ca 2+ levels in the ER were measured by the green fluorescence-Ca 2+ -measuring organelle-entrapped protein indicator 1 in the ER (G-CEPIA1er) 20 . Consistent with the reduced level of CRT expression in Atf6b −/− neurons, Ca 2+ levels in the ER were lower in Atf6b −/− neurons under both normal and ER stress conditions (Fig. 3C left). By contrast, basal Ca 2+ levels in the cytosol, which were measured by GFP-based Ca 2+ calmodulin probe 6f (GCaMP6f) 21 , were higher in Atf6b −/− neurons (Fig. 3C middle). Ca 2+ levels in the mitochondria, which were measured by CEPIA2 in the mitochondria (CEPIA2mt) 20 , were at similar levels between two genotypes under both normal and ER stress conditions (Fig. 3C right). Because these Ca 2+ measurements may be methodologically affected by expression levels of Ca 2+ -indicator proteins, Table 1. Differentially expressed genes in Atf6b −/brain (q < 0.05). Pink and blue colors indicate Upregulatedgenes and downregulated-genes in Atf6b −/− brains.
Upregulated genes in Atf6b -/brain  Data are shown as mean ± SEM. **p < 0.01 by the Mann-Whitney U test. (C) Schematic Representation of the promoters used. Triangles indicate the locations and orientations of ERSE motifs that completely or considerably match the consensus CCAATN9CCACG 18 . Numbers indicate nucleotide positions from transcription start site. ERSE2 and ERSE3 of the human CRT promoter were disrupted by mutating their sequences (marked by crosses). (D) Reporter assays using cultured hippocampal neurons. The CAT ELISA and luciferase assay were performed using cells transfected with the mouse CRT promoters, pCC1, pCC3, and pCC5 (upper graph), or with the human CRT promoters, huCRT(wt) and huCRT(mt) (lower graph). n = 4. Data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, by a two-way ANOVA followed by the Bonferroni test.  , and qRT-PCR was performed with the indicated primers. Data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001 between two genotypes by a two-way ANOVA followed by the Bonferroni test. Note that expression of Calr mRNA was significantly lower in Atf6b −/− neurons than in WT neurons under both normal and ER stress conditions. (B) Protein samples extracted from WT and Atf6b −/− hippocampal neurons exposed to control or ER stress conditions for 16 h (n = 5-6) were analyzed by western blotting using antibodies against the indicated proteins. Data are shown as mean ± SEM. ***p < 0.001 between two genotypes and # p < 0.05, ## p < 0.01, ### p < 0.001 compared to normal conditions by a two-way ANOVA followed by the Bonferroni test. Note that expression of CRT protein was significantly lower in Atf6b −/− neurons than in WT neurons under both normal and ER stress conditions. (C) Ca 2+ levels in the ER, cytosol and mitochondria of hippocampal neurons were measured using G-CEPIA1er, GCaMP6f. and CEPIA2mt, respectively under normal and ER stress (Tm for 3 h) conditions. n = 60-250 cells in each condition from two independent experiments. Data are shown as mean ± SEM. ***p < 0.001 between two genotypes and # p < 0.05, ## p < 0.01, ### p < 0.001 compared to normal conditions by a two-way ANOVA followed by the Bonferroni test. www.nature.com/scientificreports/ immunocytochemical staining was performed with anti-GFP antibody (Fig. S4) and with anti-Myc antibody (data not shown), the latter antibody recognizes the Myc epitope inserted in G-CEPIA1er and CEPIA2mt. The expression levels of Ca 2+ -indicator proteins were at similar levels in both genotypes. Furthermore, to see the floor levels of Ca 2+ in the ER, neurons were treated with Tg for 3 min. G-CEPIA1er-derived, but not CEPIA2mtderived, fluorescence disappeared in both genotypes (data not shown).

Neuroprotective role of the ATF6β-CRT axis against ER stress-induced neuronal death.
To evaluate whether ATF6β has a neuroprotective role against ER stress, WT and Atf6b −/− hippocampal neurons were treated with Tg or Tm and the cell death/survival was evaluated in two ways. Staining of living and dead cells with the fluorescent dyes, calcein-AM (green) and ethidium homodimer-1 (EthD-1, red), respectively, revealed that almost all cells are alive in normal conditions. Treatment with ER stressors induced death and reduced viability of both WT and Atf6b −/− neurons, but this effect was more pronounced in Atf6b −/− neurons (Fig. 4A). Nuclear localization of EthD-1 was confirmed by the images with single fluorescence for Hoechst 33342 and EthD-1 (Fig. S5). Consistently, immunocytochemical staining using an antibody against the apoptosis marker cleaved caspase-3 (red) and neuronal marker βIII tubulin (green) indicated that activation of caspase-3 and loss of βIII tubulin expression were more prominent in Atf6b −/− neurons than in WT neurons under ER stress (Fig. 4B). Similarly, transient silencing of Atf6b gene using siRNA enhanced ER stress-induced death of Neuro 2a cells (Fig. S6B,C).
To confirm the neuroprotective role of the ATF6β-CRT axis against ER stress, rescue experiments were performed by transfecting ATF6β and ATF6α cDNAs (Fig. 4C), or by using a lentivirus-mediated CRT overexpression system (LV-CRT) (Fig. 5A,B). When cells were co-transfected with ATF6β and GFP cDNAs, the number of cleaved caspase-3-positive cells in GFP-positive cells was reduced in Atf6b −/− neurons. By contrast, the number was not changed in Atf6b −/− neurons when cells were co-transfected with ATF6α and GFP cDNAs (Fig. 4C). Consistent with the neuroprotective effect of ATF6β, overexpression of CRT in Atf6b −/− neurons restored its expression to a similar level as that in control WT neurons (Fig. 5A), and improved the survival of neurons upon Tm treatment (Fig. 5B).

Effects of Ca 2+ -modulating reagents and an ER stress inhibitor on ER stress-induced neuronal death.
It has been reported that the CRT-mediated Ca 2+ regulation is critical for modulating neuronal death in a neurodegeneration model 22,23 ; therefore, the effects of the Ca 2+ -modulating reagents O,O′-bis(2aminophenyl)ethyleneglycol-N,N,N′,N′-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), a cell-permeable Ca 2+ chelator, and 2-aminophosphate borate (2-APB), an inhibitor of IP3Rs and store-operated channels, and an ER stress inhibitor salubrinal were analyzed in our model. Immunocytochemical analysis revealed that all reagents significantly improved survival of Atf6b −/− neurons under ER stress (images are shown in Fig. S7A and quantified data are shown in Fig. 5D). Although it was reported that BAPTA-AM alone caused ER stress and neuronal death at the concentrations higher than 13 µM 24 , we confirmed no toxicity of BAPTA-AM and 2-APB at the concentrations used in this study (5 µM and 2 µM, respectively) (Fig. S7B). Furthermore, we analyzed the effect of salubrinal with different concentrations. Significant neuroprotection was observed at 5 and 10 µM, and higher concentration (50 µM or higher) enhanced ER stress-induced neuronal death in our model (Fig. S7C).
Effect of ATF6β deletion on expression of molecular chaperones in the ER after kainate (KA) injection in mice. Kainate (KA), an agonist of glutamate receptors, causes Ca 2+ -dependent hyperactivation of neurons, followed by the induction of ER stress and neuronal death in the hippocampus 10 . We and other groups have demonstrated the protective role of UPR signaling in KA-injected mice 2,9,10 . In this study, we investigated the role of ATF6β-CRT axis in these mice. RT-qPCR revealed that expression of Atf6b, Calr, Canx, and Hspa5 mRNAs mildly, but significantly, increased after injection of KA into the mouse hippocampus (Fig. 6A). Consistent with the results in cultured neurons, the level of Calr mRNA, but not of other mRNAs, was reduced to ~ 50% in Atf6b −/− mice under both sham and KA-injected conditions (Fig. 6A). Western blot analysis confirmed that the level of CRT protein was decreased in the Atf6b −/− hippocampus under both sham and KAinjected conditions (Fig. 6B).
The neuroprotective role of ATF6β-CRT axis against KA-induced neuronal death. The neuroprotective effect of ATF6β in vivo was evaluated using KA-injected mice. Consistent with our previous reports 9,10 , Nissl staining (Fig. S8A) and immunohistochemical staining for cleaved caspase-3 (Fig. 7A) revealed that KA caused neuronal death in the CA3 region of the hippocampus, which is one of the most KA-sensitive areas. The level of neuronal death was significantly higher in Atf6b −/− mice than in WT mice at 1 and 3 days after KA injection ( Supplementary Fig. S8A, Fig. 7A). To analyze the involvement of CRT in KA-induced neuronal death, Calr +/− mice, which developed normally and showed no gross phenotypes with a reduced level of Calr expression (Fig. S8B,C), were injected with KA and neuronal death was evaluated. Consistent with the results obtained with Atf6b −/− mice, the level of neuronal death in the hippocampus was significantly higher in Calr +/− mice than in WT mice (Fig. 7B).
ATF6β-mediated regulation of neuronal activity and Ca 2+ homeostasis after KA administration. To elucidate the mechanism underlying the enhanced level of neuronal death in Atf6b −/− hippocampus after KA injection, earlier events following KA injection were investigated. qRT-PCR (Fig. S9A) and immunohistochemistry (Fig. S9B)   mice. Data are shown as mean ± SEM. ***p < 0.001 between two genotypes and # p < 0.05, ## p < 0.01, ### p < 0.001 compared to normal conditions by a two-way ANOVA followed by the Bonferroni test. (B) Protein samples were extracted from the CA3 region of hippocampi from KA-injected WT and Atf6b −/− mice and subjected to western blotting using antibodies against CRT, calnexin, GRP94 and GRP78. n = 4-5 mice. Data are shown as mean ± SEM. **p < 0.001, ***p < 0.001 between two genotypes and # p < 0.05, ## p < 0.01, ### p < 0.001 compared to normal conditions by a two-way ANOVA followed by the Bonferroni test.  www.nature.com/scientificreports/ and Bdnf was induced in both genotypes after KA-injection, but was higher in Atf6b −/− mice, suggesting that hyperactivity is involved in the enhanced level of neuronal death in the Atf6b −/− hippocampus. The effects of the Ca 2+ -modulating reagent 2-APB and ER stress inhibitor salubrinal were next analyzed. They did not cause neuronal death at the doses used in this study (Fig. S9C). Immunohistochemical analysis revealed that both reagents significantly improved neuronal survival in the Atf6b −/− hippocampus after KA injection (Fig. 7C, Supplementary Fig. S9D), suggesting that temporal dysregulation of Ca 2+ homeostasis in Atf6b −/− neurons enhances ER stress, leading to increased ER stress-induced neuronal death. Finally, the Morris water maze test was performed to analyze the effect of ATF6β deletion on the spatial memory, which is typically associated with the hippocampal function. No significant differences were observed between WT and Atf6b −/− mice in either escape latency (Fig. S10B) or the time spent in the approach (Fig. S10C) and evacuation zone (Fig. S10D), suggesting the maintenance of hippocampal function in Atf6b −/− mice in normal condition.

Discussion
The major findings of the current study are that ATF6β specifically regulates CRT expression in the CNS and that the ATF6β-CRT axis plays an important role for the survival of hippocampal neurons upon exposure to ER stress and excitotoxicity.
CRT is a Ca 2+ -binding molecular chaperone in the ER that functions in diverse cellular processes such as Ca 2+ homeostasis, protein folding, gene expression, adhesion, and cell death 19,25 . It is also important for organogenesis especially in the heart, brain, and ventral body wall 26 . Deficiency of CRT leads to defects in myofibrillogenesis and thinner ventricular walls in the heart 27 . Interestingly, overexpression of CRT in the heart also causes severe phenotypes such as arrhythmias and sudden heart block following birth 28 . Therefore, the transcription of CRT in the heart is strictly controlled by several transcriptional factors such as Nkx2.5, COUP-TF1, GATA6 and Evi-1 29 . CRT is also highly expressed in the developing brain and retina, and its deficiency leads to a defect in closure of neural tubes 26 . Although it is unclear whether overexpression of CRT is toxic in the CNS, it is possible that the similar strict regulation of CRT expression is required and that ATF6β is utilized in addition to ATF6α for this purpose.
Our results suggest a novel role for ATF6β in the regulation of molecular chaperones in the ER. CRT expression was constitutively reduced to ~ 50% in the CNS of Atf6b −/− mice at both mRNA and protein levels. This was in contrast with other molecular chaperones in the ER. Expressions of calnexin, GRP94, and GRP78 were only temporally reduced in the Atf6b −/− neurons under ER stress condition at the mRNA level (Fig. 3A), but not at the protein level (Fig. 3B). These observations may raise a scenario that, in the CNS, the expression of molecular chaperones in the ER is generally governed by ATF6α as previously described 7 and that ATF6β functions as a booster if their levels are not high enough. However, neurons may require a high level of CRT expression even under normal conditions as described in Table S2, meaning that ATF6β are required to enhance CRT expression. This may increase the dependency of the CRT promoter on ATF6β, which would explain why CRT expression was not reduced in the brain of Atf6a −/− mice (Fig. S3A). Alternatively, it may be possible that some regulatory molecules suppress the transcriptional activity of ATF6β to maintain ATF6α-dependency for molecular chaperones in the ER except CRT. Further studies, especially those to find binding partners of ATF6β, are required to clarify the molecular basis how this unique system is regulated.
The current study also suggest a neuroprotective role of ATF6β which is associated with CRT-mediated Ca 2+ homeostasis. Atf6b deficiency reduced CRT expression (Fig. 3A,B), decreased/increased basal Ca 2+ levels in the ER/cytosol (Fig. 3C), and enhanced ER stress-induced death of cultured hippocampal neurons (Fig. 4A,B) and Neuro 2a cells (Fig. S6). Overexpression of ATF6β and CRT, but not ATF6α, rescued Atf6b −/− hippocampal neurons against ER stress-induced death (Figs. 4C, 5A,B). The lack of rescuing effect by ATF6α may be due to the fact that this molecule enhances the expression of different genes including cell death-related molecule CHOP in addition to molecular chaperons in the ER 30 . Consistent with the close relationship between ATF6β and Ca 2+ homeostasis, treatment with Ca 2+ -modulating reagents such as BAPTA-AM and 2-APB and with an ER stress inhibitor salubrinal restored the survival of Atf6b −/− neuronal cells under ER stress (Fig. 5D, Supplementary Fig. S7A,C). Our results in vivo also demonstrated that reduced level of CRT in Atf6b −/− neurons enhanced Ca 2+ -mediated hyperactivity and ER stress after KA injection (Fig. 7, Supplementary Fig. S9). To our knowledge, this is the first report to demonstrate the impaired Ca 2+ homeostasis and susceptibility of Atf6b −/− and Calr +/− hippocampal neurons to ER stress and excitotoxicity (Fig. 7B). Because Atf6a deficiency also enhances Kainate-induced neuronal death in vivo 10 , it is intriguing to study the expression of CRT and other Ca 2+ regulating molecules in Atf6a −/− brains after KA injection.
Accumulating evidence suggests that a reduced level of CRT is associated with the pathologies of neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) 23,31 and Alzheimer's disease (AD) 22,32 . In a mutant superoxide dismutase (mSOD1) model of ALS, activation of the Fas/nitric oxide (NO) pathway reduced CRT expression in motoneurons, which further activated Fas/NO signaling on one hand and enhanced ER stress and neuronal death on the other hand 23 . Consequently, the level of CRT was drastically decreased in 50% of fast fatigable motoneurons 31 . Although it is unclear whether ATF6β is involved in the Fas/NO-mediated reduction of the CRT expression, it will be intriguing to analyze the expression profile of ATF6β in these models.
Consistent with the observations in ALS, the level of CRT was reduced in the brains 22 and sera 32 of AD patients, raising the possibility that CRT is a good biomarker of AD 32 . However, it was also reported that a portion of CRT was located on the cell surface membrane, and acted as a receptor for C1q, the recognition subunit of the first component of complement. The C1q-CRT complex then induced oxidative neurotoxicity 33 ; therefore, CRT may also have a pathological role in AD. Further analysis is required to elucidate the precise role of CRT and involvement of ATF6β in AD. www.nature.com/scientificreports/ Although the function of ATF6β has been considered to be very limited or redundant compared with that of ATF6α, our results emphasize the critical and beneficial roles of ATF6β in the CNS. A recent study also demonstrated that ATF6β was functional in the heart, especially during the pressure overload-induced cardiac hypertrophic response 16 . The role of ATF6β may be determined by the need for specific molecular chaperones in the ER such as CRT, which may differ between tissues. Further studies dissecting the cell-and tissue-specific roles of ATF6β will help to elucidate the function of the UPR in pathophysiological conditions.

Materials and methods
Animals. All animal experiments including behavioral study were approved by the Animal Care and Use Committee of Kanazawa University (Approval No. AP-184013) and by the institutional review committee of Kanazawa Medical University (Approval No. 2018-21). They were conducted in accordance with the Fundamental Guidelines for Proper Conduct of Animal Experiment and Related Activities in Academic Research Institutions under the jurisdiction of the Ministry of Education, Culture, Sports, Science and Technology, as well as in compliance with the ARRIVE guidelines. Atf6a +/− and Atf6b +/− mice were generated as previously described 7 , and backcrossed with the C57BL/6 strain for more than eight times at the Institute of Laboratory Animals, Graduate School of Medicine, Kyoto University. Atf6a +/− and Atf6b +/− mice were intercrossed to obtain wild-type (WT), Atf6a −/− , and Atf6b −/− mice. These lines were maintained by mating mice of the same genotype at the Institute for Experimental Animals, Advanced Science Research Center, Kanazawa University. Mice used in behavioral study were propagated by mating Atf6b +/− mice, thereby producing Atf6b +/+ , Atf6b +/− , and Atf6b −/− offspring. Atf6b +/+ mice were used as WT controls. Calr +/− mice were generated as previously described 34 , and provided by the RIKEN BioResource Research Center (Tsukuba, Ibaraki, Japan). Calr +/− mice were maintained by mating mice with WT mice in the C57BL/6 background. WT, Atf6a −/− , Atf6b −/− , and Calr +/− mice (age, 10-12 weeks; weight, 25-30 g) were used for experiments expect behavioral study. In behavioral study, age-matched WT and Atf6b −/− mice (18-21 weeks) were used.
Morris water maze test. The Morris water maze test was conducted as described previously 37 . A plastic cylindrical tank (120 cm φ) surrounded by a wall of 45 cm high and filled with opaque water (25 °C) was used. A transparent plastic platform (10 cm φ) was hidden below the water surface. Four differently-shaped and -colored objects were placed above the edge of the tank as geographical external cues. On each of 5 consecutive days, mice were given 4 sessions of swim. For each session, the mice were released from a starting point pseudo-randomly chosen from the 4 positions, and the time spent to reach the platform (escape latency) was measured. If mice were not able to reach the platform within 60 s, they were placed on the platform by the experimenter and allowed to stay there for 20 s. The average of the time spent over the 4 sessions yielded the latency score for a particular day for an individual mouse. The day-by-day averages were then calculated for each group. After the last goal-seeking session on the last test day, the platform was removed to perform the probe test. The mice were placed into the water from the edge of the pool located opposite to the former platform position, and allowed to swim for 60 s. The swimming trajectory was digital video-recorded and analyzed offline (SMART, Panlab Harvard Apparatus, Barcelona, Spain). The time spent in the approach and evacuation zone were calculated and expressed as percentage over the total time of 60 s. Cell cultures. Primary hippocampal and cortical neurons were isolated from embryonic day 17.5 (E17.5) WT, Atf6b −/− , Calr +/− and WT mice, as previously described, with minor modifications 38 . Briefly, hippocampi and cerebral cortices were harvested from prenatal mice, and digested using neuron dissociation solution (FUJFILM Wako Pure Chemical Co.). After isolation, neurons were plated into 24-well culture plates precoated with poly-l-lysine (10 µg/ml; Sigma) at a density of 8 × 10 5 cell/well, and cultured in Neurobasal Medium (Life Technologies, Carlsbad, CA, USA) supplemented with 2% B-27 serum free supplement (Life Technologies), 0.4 mM l-glutamine (Sigma), 5% fetal bovine serum (FBS)(Sigma), 100 U/ml penicillin and 100 μg/ml streptomycin (Nacalai Tesque, Kyoto, Kyoto, Japan). After 3 days, neurons were used for experiments. Hippocampal neurons were treated with the ER stressors Tg (300 nM; Sigma), DTT (1 mM; Nacalai Tesque) and Tm (1 µg/ ml; FUJFILM Wako Pure Chemical Co.). In some cases, they were treated with BAPTA-AM (5 µM; Dojindo Molecular and Technologies Inc., Mashiki-machi, Kumamoto, Japan), 2-APB (2 µM) or salubrinal (5 µM) in addition to Tm for the indicated durations.
Astrocytes were isolated from the cerebral cortex of postnatal day 1-3 WT mice, as previously described 39 , and cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% FBS and penicillin/ streptomycin. Cells were used for experiments after achieving full confluency.
MEFs were isolated from the skin of E15.5 WT and Atf6b −/− mice, as previously described 7 , and were cultured in DMEM supplemented with 20% FBS and penicillin/streptomycin. Cells were used for experiments after achieving full confluency. www.nature.com/scientificreports/ Neuro 2a cells were plated at a density of 5 × 10 4 cells/well in 24-or 12-well culture plates, and cultured in DMEM supplemented with 10% FBS and penicillin/streptomycin. Cells were used for experiments after achieving 70% of confluency.
Marek Michalak (University of Alberta) 17 . Both of pCC1 and pCC3, but not pCC5, contain ERSE, a consensus of CCAATN 9 CCACG 18 (Fig. 3A). Luciferase plasmids containing huCRT(wt) and huCRT(mut), with the two ERSEs mutated in the latter, and a plasmid containing the WT human GRP78 promoter (huGRP78) were constructed as previously described 18 . The pRL-SV40 plasmid was obtained from Promega (Madison, WI, USA). Plasmids for Ca 2+ imaging such as pCMV G-CEPIA1er, pGP-CMV-GCaMP6f and pCMV CEPIA2mt were obtained from Addgene (Watertown, MA, USA). Cells were transfected with each plasmid for 5 h using Lipofectamine 2000 (Life Technologies) and further incubated for 24-48 h. In our model, transfection efficiency was approximately 5% in primary neurons.

Preparation and infection of lentivirus vectors. The lentivirus vector expressing full-length mouse
CRT under the control of the human eukaryotic translation elongation factor 1 α1 promoter and the lentivirus vector alone was purchased from VectorBuilder (Chicago, IL, USA). Viral stocks had titers of ~ 10 9 plaque-forming units/ml. Hippocampal neurons were infected with the CRT-expressing (LV-CRT) or control (LV-control) lentivirus vector at a multiplicity of infection 10 for 16 h and further incubated for 48-72 h.

qRT-PCR. Total RNA was extracted from the indicated mouse tissues and cultured cells using RNeasy Lipid
Tissue Mini Kit (Qiagen, Valencia, CA, USA). Reverse transcription reactions containing 1 μg of total RNA were performed using PrimeScript (Takara, Otsu, Shiga, Japan). Individual cDNAs were amplified with THUN-DERBIRD SYBR qPCR Mix (TOYOBO CO, LTD, Osaka, Osaka, Japan) using specific primers for Atf6b, Atf6a, Calr, Canx, Hspa5, Hsp90b1, Fos, Fosb, Bdnf and Gapdh. The primers are listed in Table S3. The comparative Ct method was used for data analyses with MxPro 4.10 (Agilent Technologies, Santa Clara, CA, USA). Values for each gene were normalized against the Gapdh expression level.