Cell-based HTS identifies a chemical chaperone for preventing ER protein aggregation and proteotoxicity

The endoplasmic reticulum (ER) is responsible for folding secretory and membrane proteins, but disturbed ER proteostasis may lead to protein aggregation and subsequent cellular and clinical pathologies. Chemical chaperones have recently emerged as a potential therapeutic approach for ER stress-related diseases. Here, we identified 2-phenylimidazo[2,1-b]benzothiazole derivatives (IBTs) as chemical chaperones in a cell-based high-throughput screen. Biochemical and chemical biology approaches revealed that IBT21 directly binds to unfolded or misfolded proteins and inhibits protein aggregation. Finally, IBT21 prevented cell death caused by chemically induced ER stress and by a proteotoxin, an aggression-prone prion protein. Taken together, our data show the promise of IBTs as potent chemical chaperones that can ameliorate diseases resulting from protein aggregation under ER stress.


Introduction
Proteins are responsible for many tasks in cellular life, and having a correct three-dimensional protein structure is essential for their function. Molecular chaperones are proteins that assist the covalent folding or unfolding of proteins and the assembly or disassembly of protein complexes. One major function of chaperones is to prevent both newly synthesized polypeptide chains and assembled subunits from aggregating into non-functional structures. The accumulation of unfolded or misfolded proteins in the endoplasmic reticulum (ER) is called ER stress. The cell has an adaptive system against ER stress called the unfolded protein response (UPR), which is the coordinated transcriptional upregulation of ER chaperones and folding enzymes that prevents the aggregation of unfolded and incompletely folded proteins (Mori, 2000;Walter and Ron, 2011;Han and Kaufman, 2017). The UPR is initiated by the activation of the following three signal transducers located in the ER: IRE1a, PERK and ATF6. Disturbance of the UPR has important pathological consequences, including diabetes, neurodegenerative disease and cancer (Hotamisligil, 2010;Martínez et al., 2018;Clarke et al., 2014). containing ten tandem repeats of ERSE2 (10X ERSE2), ten tandem repeats of UPRE (10X UPRE) and five tandem repeats of AARE (5X AARE) in the region upstream of a minimal promoter ( Figure 1A). The EUA-EGFP reporter cells can simultaneously monitor the activation of the three branches of the UPR with high sensitivity and specificity using the specific ER stressor tunicamycin (Tm), an inhibitor of N-linked glycosylation ( Figure 1B and Figure 1-figure supplement 1).
ER chemical chaperones are small molecules that prevent protein aggregation, thereby theoretically suppressing UPR activation under ER stress conditions. We hypothesized that chemical chaperones could be identified by monitoring UPR activation for reductions under ER stress conditions. The EUA-EGFP reporter cells allowed us to perform 384-well HTS via a sensitive (signal-to-background ratio = 5.09 ± 1.28) and robust (Z' factor = 0.78 ± 0.06) homogeneous assay using a microplate reader ( Figure 1C). We screened the 217,765-compound chemical library from the Drug Discovery Initiative at the University of Tokyo. The small molecule inhibition of EUA-EGFP cells was normalized to the untreated control (assigned to be 100% inhibition), as was the cell viability, allowing comparisons between the screening plates. A first screening was carried out with all compounds at a concentration of 5 mM. We identified 1920 compounds for which the % EUA-EGFP inhibition was greater than the average + 5 SD of the Tm-treated control (hit cut-off = 37%) and the % viability was greater than the average + 3 SD of the Tm-treated control (hit cut-off = 78%) ( Figure 1D and E). Confirmation screening was carried out with all compounds at a concentration of 0.5 mM in the second screening round. To decrease the number of compounds for follow-up, we chose the top 10 inhibition hits. A maximum common substructure search identified a common scaffold in 4 of 10 compounds: the 2-phenylimidazo[2,1-b]benzothiazole skeleton (shown in red) ( Figure 1D and F). Although the other six compounds had the same high activity, they were all singletons with no common chemical structure. Therefore, we prioritized the analysis of the compounds with the 2-phenylimidazo[2,1-b]benzothiazole scaffold. The other six compounds remain to be analyzed in future studies.

Structure-activity relationship study of imidazo[2,1-b]benzothiazole derivatives (IBTs)
To define the common scaffold of the hits, a structure-activity relationship (SAR) study was conducted for the 28 compounds considered imidazo[2,1-b]benzothiazole derivatives (IBTs). The 50% inhibitory concentration (IC 50 ) for UPR activation was calculated using a five-point dose-response curve for the 28 compounds. Hierarchical structure clustering analysis revealed four major clusters in the dendrogram (Figure 2 and Figure 2-figure supplement 1). The active compounds belonged to two clusters, suggesting that both R1 and R2 in the IBTs are important for the observed activity. Removal of the phenyl group at the R1 position completely abolished UPR inhibition, indicating that the phenyl group is necessary for potency. Preferred substituents at the p-position of the R1 phenyl group include hydrogen groups (IBT6, IBT9, IBT15, IBT19 and IBT21), halogen groups (IBT2, IBT17 and IBT22), an ethyl group (IBT14) and a methoxy group (IBT3, IBT5, IBT8 and IBT12). In contrast, methyl groups (IBT4 and IBT10) and amino groups (IBT18) were not preferred. To investigate whether the activity decrease at R2 is caused by an amide bond or by bulkiness, we designed and synthesized a methyl-substituted amide-containing IBT (IBT29). IBT29 exhibited higher activity than IBT21 (Figure 2-figure supplement 2).Thus, for R2, a short substituent, particularly methylamide, endowed the scaffold with the desired activity. We observed that the preferred substituents at R2 are hydrogen groups (IBT1, IBT11, IBT12, IBT13, IBT17 and IBT20), methylsulphonyl groups (IBT21 and IBT22), a methoxycarbonyl group (IBT14 and IBT15) and a methylamide group (IBT29). Non-preferred substituent is carboxy group (IBT16).
For the next step, we needed to select one compound from the top five inhibitory IBTs (IBT14, IBT15, IBT17, IBT21 and IBT22) as a lead compound for further characterization of the IBT mechanism of action. One of the limitations of the current screening is that the target proteins of the hit compounds were unidentified. Since small molecule-target interaction profiles are useful for structure-based drug design to improve the lead compound, this compound should be capable of linker attachment to pull down a target protein. To obtain a bioactive IBT probe for pulldown, R1 could be an appropriate site for linker attachment because the SAR results predicted that linker attachment at the R2 position might lead to loss of activity. IBT21 is the only compound that has no substituent at the R1 phenyl group 2-position among the top five compounds. In terms of pharmacokinetics and bioavailability, a compound with a lower logP value is preferred, and IBT21 was the compound with   the lowest logP value. Given these considerations, we decided to further characterize IBT21, and synthesized IBT21 ourselves and confirmed its structure and activity.

Comparison of IBT21 and the validated UPR modulator and chemical chaperone
Several compounds have been reported to mitigate ER stress, such as UPR modulators or chemical chaperones. To the best of our knowledge, IBT21 has a unique structure compared with previously reported UPR modulators or chemical chaperones ( Figure 3A). According to a database search, IBT21 was registered in PubChem (CID: 735314, https://pubchem.ncbi.nlm.nih.gov/) and SciFinder (CAS Registry Number 692744-81-3, https://scifinder.cas.org/). Moreover, there were 3042 types of 2-phenyl IBT scaffold compounds in PubChem and 9878 types in SciFinder as of February 24, 2019, but no chemical chaperone-like activity has been reported. To assess the effects of IBT21, we next aimed to compare the activity of IBT21 to that of a validated UPR modulator and a chemical chaperone, azoramide (AZO) and 4PBA, respectively. To evaluate the activity with more information including the cell morphology, we first established a UPR inhibition assay using the EUA-EGFP reporter and high content image analysis. Cell morphology and EUA-EGFP reporter activation can be evaluated simultaneously with high sensitivity (signal-to-background ratio = 15.3) and robustness (Z' factor = 0.75) ( Figure 3B,C and D). At 10 mM IBT21, Tm-induced UPR activation was inhibited 91.8% by IBT21 in 293A cells ( Figure  . We utilized our high content image analysis to investigate the potential UPR modulation by small molecules. IBT21 demonstrated an IC 50 value of 0.95 mM, consistent with the IC 50 (0.61 mM) obtained from the HTS assay ( Figure 3E). However, compared with IBT21, AZO and 4PBA showed much higher IC 50 values of 86 mM and greater than 5,000 mM, respectively ( Figure 3E). We concluded that our phenotypic screening approach identified a new and more potent small molecule, IBT21.

Inhibitory effects of IBT21 on activation of the three UPR branches under ER stress conditions
To rule out the possibility that IBT21 exerts its UPR inhibitory effect by suppressing one major branch of the UPR, we explored its effects on the three UPR branches. To directly monitor each branch of the UPR, we established EGFP reporter cell lines containing twenty-five tandem repeats of ERSE2, UPRE or AARE in the region upstream of a minimal promoter to detect the activation of the ATF6, IRE1 or PERK branches of the UPR, respectively. As expected, IBT21 suppressed the Tminduced activation of all three reporters in a dose-dependent manner ( Figure 4A-C). Indeed, the IC 50 values for ERSE2 (0.24 mM), UPRE (0.33 mM) and AARE (0.46 mM) were similar to that of EUA-EGFP (0.61 mM), producing a similar dose-response curve. To further rule out off-target effects through which IBT21 might induce a pseudo-response to the reporter, we assessed the effects of IBT21 by examining the expression levels of UPR downstream target genes by immunoblot analyses. As expected, the levels of ATF4, GADD34 and CHOP, markers of PERK branch signalling, were increased by Tm, and IBT21 treatment abolished these increases in a dose-dependent manner ( Figure 4D,E and Figure 4-figure supplement 1). Additionally, IBT21 decreased the levels of DNAJB9, a marker of IRE1 branch signalling, and SEL1L and HERPUD1, markers of ATF6 branch signalling, which were induced by Tm. By contrast, the UPR modulator AZO did not decrease any    Figure 2 continued on next page markers in the UPR by Tm, consistent with the previous report (Fu et al., 2015). These results indicated that IBT21 works as a chemical chaperone rather than a UPR modulator. However, one might say that IBT21 may influence many of the observed readouts by reducing transcription/translation, rather than by functioning as a chemical chaperone. To demonstrate the chemical chaperone activity of IBT21, we conducted several experiments. First, we overexpressed active UPR transcription factor and examined the effects of IBT21 on transcriptional induction of UPR target genes (Figure 4-figure supplement 2). IBT21 did not change the expression of CHOP or 4E-BP1, downstream targets of ATF4, in the ATF4-overexpressing cells. Additionally, IBT21 did not change the expression of BiP in the cells overexpressing an active form of ATF6 [ATF6 (N)], These data show that IBT21 does not affect the ability to induce UPR target genes. Next, we assessed the effects of IBT21 on protein translation by measuring protein synthesis based on the incorporation of puromycin into newly synthesized proteins through its detection with anti-puromycin antibodies (Figure 4-figure supplement 3). We evaluated that IBT21 does not affect protein translation. Furthermore, we tested whether IBT21 can also reduce ER stress caused by dithiothreitol (DTT), which does not require protein translation to induce ER stress (Figure 4-figure supplement 4). We showed that IBT21 modestly reduced DTT-induced ER stress by monitoring the activation of EUA-EGFP reporter cells. In contrast, IBT21 could not suppress the induction of heat-shock response by heat shock (Figure 4-figure supplement 5). These data support a chemical chaperone function of IBT21 during ER stress but not all proteostasis stress.

In vivo effects of IBT21 on protein aggregation during ER stress
To directly address whether IBT21 exerted its effects though chemical chaperone activity, we investigated the effects of IBT21 on reducing protein aggregation during ER stress. To monitor protein aggregation in vivo, we first took advantage of the ProteoStat dye, a molecular rotor that binds selectively to aggregated proteins (Shen et al., 2011). Proteasomal inhibition is known to lead to the accumulation of ubiquitin-conjugated proteins organized in perinuclear structures termed 'aggresomes'. As expected, aggregated protein cargo was visualized with a punctate red fluorescence pattern in cells treated with the proteasome inhibitor MG132. When we treated cells with Tm, diffuse cytoplasmic staining was observed that differed between the control cells and MG132treated cells with discrete punctuate staining, indicating the accumulation of aggregated proteins in the ER ( Figure 5A). Quantification of the ProteoStat dye staining within cells revealed increased fluorescence intensity in the Tm-treated cells (2.29-fold of untreated (UT)) and MG132-treated cells (2.33-fold of UT), and lower fluorescence intensity in the cells co-treated with Tm and IBT21 (0.87fold of UT), showing that IBT21 prevented protein aggregation induced by Tm ( Figure 5B). To validate the direct effect of IBT21 on protein aggregation inhibition, an in vitro protein aggregation assay was performed using human insulin and the ProteoStat dye. Human insulin was challenged to be denatured by overnight incubation of DTT with or with IBT, 4PBA or AZO. As expected, IBT21 and 4PBA inhibited protein aggregation, indicating a direct function as chemical chaperones, though the UPR modulator AZO did not ( Figure 5C). Notably, IBT21 inhibited protein aggregation at a concentration of 10 mM, which is 100-fold lower than that of 4PBA.
For further assessment of IBT21, we performed a biochemical assay to detect misfolded or aggregated proteins. Misfolded or aggregated proteins are prone to forming stable, detergent-insoluble, high molecular weight (HMW) oligomers together with molecular chaperones (Johnston et al., 1998). We previously developed and modified a method to estimate the amount of misfolded or aggregated protein in the ER based on the levels of the molecular chaperone BiP (also known as HSPA5) in SDS-resistant HMW complexes (AGGREGATED) (Marciniak et al., 2004;Hisanaga et al., 2018) ( Figure 5D). Compared with the levels of BiP in the INPUT fraction, the levels of BiP in the  AGGREGATED fraction corresponding to misfolded or aggregated proteins in the ER were increased by Tm treatment (3.1-fold of UT), and this increase was significantly suppressed by IBT21 treatment (1.7-fold of UT) ( Figure 5E and F). IBT21 treatment did not change the levels of BiP under ER stress conditions ( Figure 5E and F) or under basal conditions ( Figure 4-figure supplement 2C, D), indicating that IBT21 did not lead to reduced expression or enhanced degradation of BiP. Having established that IBT21 reduced the levels of the SDS-resistant HMW BiP-containing complex induced by Tm, we next determined the profile of proteins in the SDS-resistant HMW BiP-containing complex to directly monitor proteostasis changes. The AGGREGATED fraction samples containing the SDSresistant HMW BiP-containing complex were subjected to mass spectrometry analysis, and 1233 proteins were detected by this LC-MS/MS analysis. To avoid false positive protein identification, we filtered 598 proteins identified with !2 unique peptides per protein and !5 peptide-spectrum matches. Then, to determine the proteins modulated by IBT21, we analyzed the differentially expressed proteins among the 598 proteins with the following criteria: 1) 1.5-fold enrichment in the Tm-treated samples compared to the UT samples and 2) 1.5-fold enrichment in the Tm-treated samples compared to the Tm-and IBT21-treated samples. We found that 211 of 598 proteins (35.3%) were affected by Tm, and 125 of those 211 proteins (59.2%) were affected by IBT21 (IBT-SENSITIVE proteins), suggesting that IBT21 prevents Tm-induced protein misfolding ( Figure  . These enrichments support that IBT21 acts on the folding of proteins synthesized in the ER because secreted and membrane-bound proteins are synthesized in the ER. For example, the proteins we identified included Annexin A1 (ANXA1) and Transgelin-2 (TAGLN2), and these proteins represent the protein group affected by Tm and IBT21, as ANXA1 (Goulet et al., 1992) and TAGLN2 (Liu et al., 2018) were reported or predicted to have N-linked glycosylation sites for secretion ( Figure 5G  . In addition to ER client proteins, proteins required for proteostasis could also be affected by Tm and IBT21. As expected, protein folding was highly enriched in the category of biological function, as the representative proteins from the AGGREGATED fraction included the molecular chaperone HSPA5 and the ER folding enzyme PDIA4 ( Figure 5G,I, Figure 5figure supplement 1A,C). Thus far, our observations support that IBT21 has chemical chaperone activity against ER stress.

In vivo target protein identification of IBT21 during ER stress
To examine the direct binding of IBT21 to unfolded or misfolded proteins in vivo, we took advantage of diazido photoaffinity labelling probes, which allow a chemical biology approach for profiling IBT21-binding proteins (Hosoya et al., 2004;Hosoya et al., 2005). We designed a diazido-functionalized IBT21 analogue, diazido-IBT21, as a candidate photoaffinity labelling probe ( Figure 6A). The activity of diazido-IBT21 was comparable to that of IBT21, and we used the more potent compound IBT22 as a competitor ( Figure 6B). With a promising diazido-IBT21 probe in hand, we performed in vivo target protein identification experiments ( Figure 6C). In principle, the target proteins were specifically captured by covalent bonding via photoreaction of the probe aromatic azido group. The remaining probe aliphatic azido group was subsequently subjected to biotin labelling via Cu-catalysed azide alkyne cycloaddition (CuAAC) for pulldown of the target proteins. In total, 1441 proteins were detected as IBT-BOUNDED proteins and were classified with the following criteria: 1) 1.5-fold change in the probe-treated samples compared to the probe-and competitor-treated samples (competed) and 2) 1.5-fold change in the probe-treated samples compared to the UT samples (selective) ( Figure 6D). The IBT-BOUNDED proteins were categorized into the following four groups: Competed and Selective (1205 proteins), Competed (11 proteins), Selective (185 proteins)   and Non-Specific (40 proteins) ( Figure 6E). The Competed and Selective proteins were the majority (84%) of the IBT-BOUNDED proteins. To characterize the IBT21-targeted proteins to corresponding chemical chaperone activity, we analyzed the Overlapped proteins observed in both the IBT-SENSI-TIVE proteins and the Competed and Selective proteins. As expected, the Overlapped proteins included ANXA1 and TAGLN2, which were also identified as IBT21-rescued aggregation-prone proteins (IBT-SENSITIVE proteins), as shown in Figure 5 ( Figure 6G). The GO enrichment analysis revealed that groups of extracellular components were highly enriched in the Overlapped proteins ( Figure 6-figure supplement 1A). GO enrichment results of the IBT-SENSITIVE proteins were different from that of the IBT-bonded Competed and Selective proteins ( Figure 6-figure supplement 1B-C) because IBT21 is stabilizing many more proteins that cannot be detected in the experiment that determined IBT-sensitive compounds. Nevertheless, the Overlapped proteins contained approximately 60% of the IBT-SENSITIVE proteins in each enriched GO group, indicating that the Competed and Selective proteins largely cover the IBT21-targeted proteins corresponding to chemical chaperone activity ( Figure 6H). Taken together, these results support that IBT21 directly binds unfolded or misfolded proteins to prevent protein aggregation in vivo.

Cytoprotective effect of IBT21 during ER stress
We finally tested whether IBT21 could ameliorate ER stress. We first evaluated the effects of IBT21 on cell proliferation under ER stress conditions, as measured by cell confluency using the high content imaging system. Tm treatment almost completely inhibited cell proliferation, as was observed in cells treated with the cytotoxic alkaloid camptothecin (cell death control). Treatment with 1 mM IBT21 restored the cell growth inhibited by Tm up to 69.8% of the growth of the UT cells ( Figure 7A and B). Surprisingly, cell growth was significantly stimulated to 112.7 or 115.8% of that of UT cells by the addition of 5 or 10 mM IBT21, respectively, even in the presence of Tm. In contrast, 40 mM AZO treatment led to a significant but modest growth recovery up to 58.1% of that of the UT cells. We next evaluated the effects of IBT21 on cell viability or cytotoxicity as measured by intracellular ATP levels and lactate dehydrogenase release. The cytoprotective effects of IBT21 were also evident in biochemical measurements. Consistent with the results of the proliferation assays, IBT21 treatment attenuated viability loss and cell death in a dose-dependent manner ( Figure 7C and D).
To corroborate the cytoprotective effects of IBT21, we next evaluated the effects of IBT21 against an ER proteotoxin. Prion disease is a neurodegenerative disorder characterized by the accumulation of an abnormally folded and protease-resistant form of prion protein (PrP). As a secretory protein, PrP is synthesized in the ER, and approximately 10% of PrP is naturally misfolded (Yedidia et al., 2001). Accumulating evidence suggests that perturbations in ER homeostasis may contribute to neurodegeneration in prion disease (Shi and Dong, 2011;Torres et al., 2015). Therefore, we chose mutant PrP, which was previously reported to cause ER stress, as an ER proteotoxin (Satpute-Krishnan et al., 2014). To monitor the expression levels of PrP, EGFP was fused to wild-type PrP (WT PrP) or mutant PrP (mut PrP). Overexpression of mut PrP inhibited cell proliferation, and 5 mM           Figure 7E and F). Consistent with the results from the proliferation assays using the high content imaging system, the recovery of viability loss by IBT21 treatment in the mutant PrP-overexpressing cells was also confirmed by a biochemical assay with CellTiter-Glo. Collectively, these data provide in vivo evidence that IBT21 protects cells against ER stress conditions.

Discussion
In the current study, we aimed to identify a potent chemical chaperone through the HTS of a small molecule chemical library. We discovered that IBTs function as chemical chaperones and directly bind unfolded or misfolded proteins to inhibit protein aggregation using the compound IBT21. Consistent with the protective nature of chemical chaperones, IBT21 protected cells from death caused by ER stress. The observed effects cannot be explained by a simple mechanism because these compounds have a planar conjugated ring system similar to many DNA intercalators; however, IBTs are promising scaffolds for chemical chaperones. Furthermore, based on our SAR studies, we might predict more potent IBT compounds. The combination of the methoxy group at the m-position of IBT3 or the chloro group at the p-position of IBT17 for R1 and the methoxycarbonyl group of IBT15 or methylamide group of IBT29 for R2 seems to be ideal. In addition to assessing the absorption, distribution, metabolism, and excretion properties of IBTs, we will pursue lead compound optimization in future studies. Chemical chaperones can be divided into the following two groups: osmoprotectants and pharmacological chaperones. Osmoprotectants are generally considered to stabilize protein structure by altering solvent properties. The major osmoprotectants can be categorized in the following three chemical classes: polyols, such as glycerol and sucrose; free amino acid derivatives, such as taurine and b-alanine; and methyl-amines, such as DMSO and trimethylamine N-oxide (TMAO) (Cortez and Sim, 2014). Among these osmoprotectants, TMAO is the most efficient and has been the most studied over the past two decades. For example, TMAO has been shown to prevent the conversion of the normal cellular form of prion protein (PrP C ) into its infectious form (PrP SC ) in scrapies-infected mouse neuroblastoma ScN2a cells (Tatzelt et al., 1996). However, osmoprotectants are effective at high concentrations which may lead to toxicity. Thus, despite the high efficiency of the anti-aggregation properties of osmoprotectants, very few osmoprotectants have been applied in clinical trials. Another class of chemical chaperones is pharmacological chaperones, which are able to bind to and stabilize unfolded or misfolded proteins. Structurally, IBT21 was not particularly similar to Figure 5 continued stained with ProteoStat dye with or without the indicated concentration of IBT21, 4PBA or AZO. Error bars show the mean ± SD (n = 3). One-way ANOVA with Tukey's post hoc test, **p<0.01. (D) Protocol for high molecular weight (HMW) detergent-resistant complex separation. HEK293A cells treated with 0.2 mg/mL Tm in the presence or absence of 10 mM IBT21 were lysed in lysis buffer, adjusted to 0.8% SDS, loaded on separation buffer and then ultracentrifuged at 100,000 Â g for 55 min. The pellet fraction (SDS-resistant HMW complex) was considered aggregated proteins (AGGREGATED). (E) Representative immunoblots of BiP in the AGGREGATED SDS-resistant HMW complexes and in the INPUT. (F) Densitometry quantification of the immunoblot analysis in E. The protein levels of BiP in the AGGREGATED and INPUT were normalized to b-actin levels in the INPUT. Error bars show the mean ± SD (n = 3). One-way ANOVA with Tukey's post hoc test, *p<0.05, **p<0.01. n.s: not significant. (G) Bubble plot representation of HMW detergent-resistant aggregated proteins. The X-axis denotes the Tm to UT abundance ratio in the absence of IBT21. The Y-axis denotes the Tm to Tm+IBT21 abundance ratio. Each threshold is 1.5-fold-change. Representative hit proteins, such as HSPA5, PDIA4, ANXA1 and TAGLN2, are indicated. Peptide-spectrum matches (PSMs) are correlated with the size of the circle. (H-I) Functional classification of proteins protected from aggregation by IBT21. Aggregated proteins that were suppressed by IBT21 (Tm/UT > 1.5 and Tm/Tm+IBT21 >1.5) were categorized on the basis of cellular components (H) and biological processes (I) through Gene Ontology (GO) annotation. The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Dataset for Figure 5B, C and F. Source data 2. Dataset for Figure 5G. Source data 3. Dataset for Figure 5H. Source data 4. Dataset for Figure 5I.      Figure 6 continued on next page osmoprotectants, and IBT21 was observed to preferably bind to unfolded or misfolded proteins in combined analyses such as the ER stress-induced aggregated protein profile and the IBT21-targeted protein profile. Therefore, we considered that IBT21 acted as a pharmacological chaperone. In unfolded or misfolded proteins, the hydrophobic region tends to be exposed, and protein aggregation occurs when the exposed hydrophobic portions of a protein interact with the exposed hydrophobic segment of other unfolded proteins. Pharmacological chaperones may prohibit protein aggregation by binding and masking hydrophobic parts of unfolded or misfolded proteins. The IBT21 core possesses hydrophobicity comparable to that of 4PBA and TUDCA, which are known to be hydrophobic compounds. This similarity supports the hypothesis that the hydrophobic core of IBT21 may interact with the exposed hydrophobic segments of unfolded or misfolded proteins. Currently, 4PBA and TUDCA are the only two pharmacological chaperones for proteotoxins approved by the FDA for human use. Notably, IBT21 shows much higher chaperone-like activity than 4PBA. In future studies, we will further examine the therapeutic effects of IBTs in model mice with a disease involving ER stress. In theory, pharmacological chaperones for proteotoxin have one of the following three mechanisms of action: preventing protein aggregation by masking the aggregation-prone region (group I), increasing the folding rate by stabilizing the folding transition state (group II) or decreasing the misfolding rate by stabilizing the native state (group III). We demonstrated that IBT21 directly prevents protein aggregation in an in vitro aggregation inhibition assay, but the second and third mechanisms remain elusive. A recent report suggests that the action of chemical chaperones is determined by the proteins to be folded and the folding environment, as a chemical chaperone that behaves as a group II chaperone for one protein might behave as a group III chaperone for another (Dandage et al., 2015). Notably, heat shock response is not influenced by IBT21. It would be difficult to generalize the mechanism of action of pharmacological chaperones based on their effect in one test stress condition or on one test protein. Hence, further characterization of IBT21 appears to require testing on disease-related misfolding-prone proteins and on disease-related stress conditions.
The chemical properties of IBTs that enable them to preferably bind to unfolded or misfolded proteins might be useful for developing diagnostic tools for ER stress-related diseases. Several positron emission tomography (PET) and single photon emission computed tomography (SPECT) tracers for imaging b-amyloid deposits in Alzheimer's disease have been used clinically or in a clinical trial. Alagille et al. designed and synthesized a series of 2-aryl-imidazo[2,1-b]benzothiazoles by a combination of the benzothiazole motif and the 2-arylimidazo motif from two potent b-amyloid imaging agents, PIB and IMPY, respectively (Alagille et al., 2011). Yousefi et al. demonstrated that 11 Clabelled IBTs have high in vitro and in vivo binding affinities for b-amyloid aggregates (Alagille et al., 2011) and that 18 F-labelled IBTs allow high-contrast PET imaging of b-amyloids in Figure 6 continued IBT competitor to no competitor. The Y-axis denotes the abundance ratio of diazido-IBT21 probe to no probe. Each threshold represents a 1.5-foldchange. Peptide-spectrum matches (PSMs) are correlated with the size of the circle. (E) Pie chart representation for the categorization of the IBT-BONDED proteins identified in experiments comparing cells treated with probe/no-probe and probe/probe + competitor. Each threshold represents a 1.5-fold-change. (F) Flow diagram for the comparison between the IBT-SENSITIVE and Competed and Selective proteins. (G) Bubble plot representation of the Overlapped proteins. The X-axis denotes the abundance ratio of diazido-IBT21 probe to no probe. The Y-axis denotes the abundance ratio of IBT competitor to no competitor. Each threshold represents a 1.5-fold-change. Representative hit proteins, such as ANXA1 and TAGLN2, are indicated in blue. Peptide-spectrum matches (PSMs) are correlated with the size of the circle. (D) Functional classification of the Overlapped proteins. The online version of this article includes the following source data and figure supplement(s) for figure 6: Source data 1. Dataset for Figure 6. Source data 2. Dataset for Figure 6D and E. Source data 3. Dataset for Figure 6G. Source data 4. Dataset for Figure 6H.     Alzheimer's disease model mice (Yousefi et al., 2011). Together with our results, these findings indicate that IBT21 could be effective for PET and SPECT imaging of ER stress-related diseases. Targeting improper protein folding and aggregate accumulation, which are the most upstream events in ER stress-related diseases, is an obvious therapeutic strategy with chemical chaperones. It is generally accepted that neurodegenerative disorders such as Alzheimer's disease, Huntington disease and prion disease are caused by the accumulation of neurotoxic protein aggregates, and chemical chaperones have the potential to separate these neurotoxic aggregates. In the case of prion disease, the osmoprotectant TMAO was first applied for preventing PrP SC formation in vitro as a chemical chaperone in the late 1990 s (Tatzelt et al., 1996). PrP SC toxicity was reported to cause ER stress by increasing the intracellular calcium released from the ER (Hetz et al., 2003). Indeed, a reduction in the levels of the ER molecular chaperone BiP accelerates prion pathogenesis in vivo, indicating that the accumulation of the misfolded form of PrP in the ER is a key event in prion propagation (Park et al., 2017). Prion replication has been shown to cause sustained UPR induction with persistent eIF2a phosphorylation (Moreno et al., 2012). Repression of protein synthesis diminished the load of unfolded proteins in the ER, but overactivation of eIF2a phosphorylation leads to synaptic failure, spongiosis and neuronal loss. Pharmacological inhibition of PERK by the compound GSK2606414 mitigated neurodegeneration and clinical prion disease in mice (Moreno et al., 2013).
In this study, we demonstrated that IBT21 presents the most effective treatment effect and that the best treatment regimen may involve a combination of compounds, some of which fine-tune protein synthesis, such as GSK2606414, and others that reduce protein aggregates, such as the chemical chaperone IBT21. More studies with these promising compounds need to be performed to design therapies to treat prion diseases and possibly other neurodegenerative diseases associated with ER stress.

Chemical compounds
A total of 217,765 library compounds were provided by the Drug Discovery Initiative at the University of Tokyo (Tokyo, Japan). 7-Methylsulphonyl-2-phenylimidazo[2,1-b]benzothiazole (IBT21) was purchased from Butt Park (Devon, UK). 2-(4-Chlorophenyl)À7-methylsulphonylimidazo[2,1-b]benzothiazole (IBT22) was purchased from Specs (Zoetermeer, Netherlands). Azoramide was purchased from Axon Medchem (Groningen, Netherlands). 4-Phenylbutyric acid was purchased from Tokyo Chemical Industry (Tokyo, Japan). Camptothecin was purchased from FUJIFILM Wako Pure Chemical (Osaka, Japan). PERK Inhibitor I (GSK2606414) was purchased from Merck Millipore (MA, USA). IRE1 Inhibitor I (STF083010) was purchased from Toronto Research Chemicals (Ontario, Canada). Site one protease inhibitor (PF429242) was purchased from R and D Systems (MN, USA). Twenty-eight compounds for the SAR study were provided by the Drug Discovery Initiative at the University of Tokyo and quality controlled using an ACQUITY UPLC (Waters, MA, USA) equipped with an ACQUITY UPLC BEH C18 column to determine the purity. The column temperature was 40˚C. Mobile phase A consisted of 5.0 mM ammonium acetate in water. Mobile phase B consisted of acetonitrile (ACN). The gradient conditions were as follows: 5% mobile phase B at the time of injection; linear increase to 95% B at 1.00 min; and maintain 95% B until 1.50 min. The flow rate was 0.7 mL/min. The purity measurements were determined by measuring the peak area at 230 nm, and the purity was Cell viability was determined by an intracellular ATP assay (C) or LDH activity (D). Error bars show the mean ± SD (n = 4). One-way ANOVA with Tukey's post hoc test, *p<0.05 and **p<0.01. (E) Representative bright-field and YFP images of HEK293T cells overexpressing YFP-tagged WT or YFP-tagged mutant prion protein (PrP) in the presence or absence of 5 mM IBT21 for the indicated times. The scale bar represents 100 mm. (F) Cell growth curve of HEK293T cells overexpressing YFP-tagged WT or YFP-tagged mutant prion protein (PrP) in the presence or absence of 5 mM IBT21 for the indicated times (n = 6). Two-way ANOVA with Holm's post hoc test, *p<0.05. (G) Intracellular ATP assay of HEK293T cells overexpressing YFP-tagged WT or YFPtagged mutant prion protein (PrP) in the presence or absence of 5 mM IBT21 for the indicated times. Error bars show the mean ± SD (n = 6). One-way ANOVA with Tukey's post hoc test, *p<0.05 and **p<0.01. The online version of this article includes the following source data for figure 7: Source data 1. Dataset for Figure 7. calculated based on the observed target peak area as a percentage of the total area. All compounds assayed were !95% pure. 7-(methylaminocarbonyl)À2-phenylimidazo[2,1-b]benzothiazole (IBT29) was synthesized from 4-(methylaminocarbonyl)aniline and phenacyl bromide. More detailed information is shown in the Supplemental information.

Establishment of reporter cell lines
The HEK293A cell line was provided by Thermo Fisher Scientific (MA, USA) and maintained in DMEM (Nissui Pharmaceutical, Tokyo, Japan) supplemented with 10% (v/v) foetal bovine serum (Thermo Fisher Scientific). The Hap1 cell line was provided by Horizon Genomics (Vienna, Austria) and maintained in IMDM (GE Healthcare, IL, USA) supplemented with 10% (v/v) foetal bovine serum. Cells were cultured at 37˚C in a humidified incubator continuously flushed with a mixture of 5% CO 2 and 95% air. Cells were authenticated with a morphology check by microscopy and using growth curve analysis with an Operetta CLS instrument (PerkinElmer, MA, USA). All cell lines were confirmed to be mycoplasma-free.

Overexpression of UPR transcription factors
The pcDNA3.1-human ATF4 plasmid was introduced into the HEK293T cell line (ATCC, VA, USA) by transfection with polyethylenimine (Polysciences, PA, USA), followed by incubation at 37˚C for 24 hr, after which the cells were incubated in the presence or absence of 10 mM IBT21 for 24 hr. The pcDNA3.1-human ATF6 (N) plasmid was introduced into the HEK293A cell line by transfection with polyethylenimine, followed by incubation at 37˚C for 24 hr, after which the cells were incubated in the presence or absence of 10 mM IBT21 for 12 or 24 hr.

Nonradioactive measurements of protein synthesis
To measure protein synthesis, we used the nonradioactive surface sensing of translation (SUNsET) technique. Briefly, 293A-EUA-EGFP cells were pre-treated with 10 mg/mL cycloheximide (Nacalai Tesque) or 10 mM IBT21 for 1-4 hr, followed by 2 mg/mL puromycin (GoldBio, MO, USA) treatment for 30 min. Total protein was extracted and immunoblot analysis was performed with an anti-puromycin antibody (EQ0001, Kerafast, MA, USA). Ponceau S (Sigma-Aldrich, MO, USA) was used as a loading control.

Aggregated protein detection
An aggregated protein detection assay was performed with the ProteoStat Aggresome Detection Kit (Enzo Life Sciences) according to the manufacturer's instructions. For cell analyses, the specimens were viewed with a BZ-X700 device (Ex 545 ± 12.5 nm, Em 605 ± 35 nm, Keyence, Osaka, Japan) and Operetta CLS (Ex 475 ± 15 nm, Em 610 ± 40 nm). For in vitro analysis, 5 mg/mL human recombinant insulin (FUJIFILM Wako Pure Chemical) was incubated at 37˚C overnight with 100 mM DTT and 10% (v/v) DMSO in the presence or absence of IBT21, azoramide or 4PBA. Insulin aggregation was quantified with an EnVision2102 (Ex 480 nm, Em 595 nm, PerkinElmer).

High molecular weight detergent-resistant complex detection
High molecular weight detergent-resistant complex detection was performed as described previously (Hisanaga et al., 2018). Briefly, cultured cells were incubated for 5 min in ice-cold PBS containing 20 mM N-ethylmaleimide (Sigma-Aldrich) and then collected by scraping in lysis buffer (0.5% Triton X-100, 20 mM HEPES [pH 7.4], 250 mM sucrose, 100 mM NaCl and 2.5 mM CaCl 2 ) with protease inhibitors. Equal amounts of protein were adjusted to 0.8% SDS, layered upon separation buffer (0.5% Triton X-100, 0.8% SDS, 20 mM HEPES [pH 7.4], and 20% glycerol) and centrifuged at 100,000 Â g for 55 min with an Optima Ultracentrifuge (Beckman Coulter, CA, USA). For immunoblot analysis, the resulting pellet was dissolved in urea loading buffer (9.6 M urea, 1.36% SDS, 40 mM Tris [pH 6.8], 12% glycerol, 100 mM DTT and 0.002% [w/v] bromophenol blue), boiled for 5 min and then subjected to SDS-PAGE. For mass spectrometry analysis, the pellet was rinsed with icecold 50 mM ammonium bicarbonate (Sigma-Aldrich) and centrifuged at 100,000 Â g for 80 min. Proteins in the pellet were dissolved in 8 M urea and 50 mM Tris (pH 8.0), reduced, alkylated, diluted to 1 M urea and digested with trypsin/Lys-C mix (Promega, WI, USA) at 37˚C for 16 hr. The digests were desalted using GL-Tip SDB (GL Sciences, Tokyo, Japan). LC-MS/MS analysis of the resultant peptides was carried out on an EASY-nLC 1200 UHPLC connected to a Q Exactive Plus mass spectrometer through a nanoelectrospray ion source (Thermo Fisher Scientific). The peptides were separated on a 75 mm inner diameter Â150 mm C18 reversed-phase column (Nikkyo Technos, Tokyo, Japan) with a linear gradient ranging from 4-28% ACN for min 0-100, followed by an increase to 80% ACN for min 100-110. The mass spectrometer was operated in data-dependent acquisition mode with a top 10 MS/MS method. MS1 spectra were measured with a resolution of 70,000, an automatic gain control (AGC) target of 1 Â 10 6 and a mass range from 350 to 1,500 m/z. MS/MS spectra were triggered at a resolution of 17,500, an AGC target of 5 Â 10 4 , an isolation window of 2.0 m/z, a maximum injection time of 60 ms and a normalized collision energy of 27. Dynamic exclusion was set to 10 s. Raw data were directly analyzed against the SwissProt database restricted to H. sapiens using Proteome Discoverer version 2.2 (Thermo Fisher Scientific) for identification and labelfree precursor ion quantification. GO enrichment was performed with Genomatix Genome Analyzer (Genomatix, Mü nchen, Germany).
Identification of proteins interacting with the diazido-IBT21 probe HEK293A cells were treated with 0.2 mg/mL Tm for 18 hr, followed by the addition of 1 mM diazido-IBT21 probe in the presence or absence of 10 mM IBT22 (competitor) for 3 hr at 37˚C. Photo-irradiation was performed at room temperature for 180 s using UVG-54 (254 nm, 6 W, from a distance of 1 cm, UVP, CA USA). Cell lysates were prepared in 20 mM HEPES (pH 7.4), 150 mM NaCl, 1.5 mM MgCl 2 , 1% NP-40, 2 M urea and protease inhibitors on ice. For each sample, 800 mg of lysate was reacted with the Click-iT Protein Reaction Buffer Kit (Thermo Fisher Scientific) and biotin-alkyne (PEG4 carboxamide-Propargyl Biotin, Thermo Fisher Scientific) according to the manufacturer's instructions. Labelled proteins were precipitated with Streptavidin Sepharose High Performance (GE Healthcare), resuspended in 50 mM ammonium bicarbonate and digested with trypsin/Lys-C mix for 16 hr at 37˚C. The digests were desalted using GL-Tip SDB. LC-MS/MS analysis of the resultant peptides was carried out on a Dionex UltiMate 3000 UHPLC system connected to an Orbitrap Fusion mass spectrometer through a nanoelectrospray ion source (Thermo Fisher Scientific). The peptides were separated on a 75 mm inner diameter Â120 mm C18 reversed-phase column with a linear gradient ranging from 4-40% ACN for min 10-110, followed by an increase to 80% ACN for min 110-120. The mass spectrometer was operated in data-dependent acquisition mode with a maximum duty cycle of 3 s. MS1 spectra were measured with a resolution of 120,000, an AGC target of 4 Â 10 5 and a mass range from 375 to 1,500 m/z. HCD MS/MS spectra were acquired at a resolution of 15,000, an AGC target of 5 Â 10 4 , an isolation window of 4.0 m/z, a maximum injection time of 54 ms and a normalized collision energy of 30. Dynamic exclusion was set to 10 s. Raw data were directly analyzed against the SwissProt database restricted to H. sapiens using Proteome Discoverer version 2.2 for identification and label-free precursor ion quantification.

Viability assay
HEK293A cells were seeded on a 384-well plate. Tunicamycin was added in the presence or absence of compounds and then incubated for 48 hr. Cell viability was measured using CellTiter-Glo (Promega) and Cytotoxicity LDH Assay Kit-WST (Abs 490 nm, FUJIFILM Wako Pure Chemical) with Cyta-tion3. Cell images were viewed with Operetta CLS and analyzed with Harmony 4.6.

Overexpression of mutant prion protein
The mutant PrP (YFP-PrP C179A/S232W) expression vector was kindly gifted by Dr. Ramanujan S. Hegde (MRC Laboratory of Molecular Biology, Cambridge, UK). Each plasmid was introduced into the HEK293T cell line by transfection with polyethylenimine in the presence or absence of 5 mM IBT21. Cell images were viewed with Operetta CLS and analysed with Harmony 4.6.

Data availability
All data generated or analysed during this study are included in the manuscript and supporting files.