An autophagy enhancer ameliorates diabetes of human IAPP-transgenic mice through clearance of amyloidogenic oligomer

We have reported that autophagy is crucial for clearance of amyloidogenic human IAPP (hIAPP) oligomer, suggesting that an autophagy enhancer could be a therapeutic modality against human diabetes with amyloid accumulation. Here, we show that a recently identified autophagy enhancer (MSL-7) reduces hIAPP oligomer accumulation in human induced pluripotent stem cell-derived β-cells (hiPSC-β-cells) and diminishes oligomer-mediated apoptosis of β-cells. Protective effects of MSL-7 against hIAPP oligomer accumulation and hIAPP oligomer-mediated β-cell death are significantly reduced in cells with knockout of MiTF/TFE family members such as Tfeb or Tfe3. MSL-7 improves glucose tolerance and β-cell function of hIAPP+ mice on high-fat diet, accompanied by reduced hIAPP oligomer/amyloid accumulation and β-cell apoptosis. Protective effects of MSL-7 against hIAPP oligomer-mediated β-cell death and the development of diabetes are also significantly reduced by β-cell-specific knockout of Tfeb. These results suggest that an autophagy enhancer could have therapeutic potential against human diabetes characterized by islet amyloid accumulation. Islet amyloid polypeptide (IAPP) deposition is associated with islet cell loss in diabetes. Here the authors show that a small molecule autophagy enhancer reduces IAPP accumulation in vitro, and also improves glucose tolerance in hIAPP+ mice fed high-fat diet, accompanied by reduced hIAPP accumulation, in vivo.

W hile most studies on the basic pathogenic mechanism of diabetes employ mouse models in which genetic modulation is possible, human diabetes and mouse diabetes are different in several key features. For instance, the structure and function of pancreatic islets are different between mouse models and humans 1 . Such differences may be responsible for the failure of antidiabetic drugs developed using mouse models in human clinical trials. One of the most intriguing differences between mouse models of diabetes and human diabetes is the deposition of amyloid in islets of human patients with diabetes but not in those of mouse models. Islet amyloid deposition is observed in over 90% of human diabetes patients but not in murine diabetes, which is due to the differences in the amino acid sequences of the islet amyloid polypeptide (IAPP) 2 . Others and we previously reported that human-type IAPP (hIAPP) oligomer or amyloid accumulation in transgenic mice expressing hIAPP (hIAPP + mice) or hIAPP knock-in mice in which endogenous murine IAPP gene was replaced by hIAPP markedly increases when β-cell autophagy is deficient, suggesting that autophagy is critical in the clearance of hIAPP oligomer or amyloid [3][4][5] . These data are consistent with a proposition that aggregate or amyloid-prone proteins are preferentially cleared by autophagy rather than proteasomal degradation 6 . These results also suggest the possibility that autophagy enhancers may be employed as a therapeutic modality against human diabetes characterized by islet amyloid accumulation. Indeed, we have observed that trehalose that has been reported to have autophagyenhancing activity 7 could ameliorate metabolic profile of hIAPP + mice fed high-fat diet (HFD). However, trehalose has been developed as a chaperoning agent or hydrating agent, and its autophagy enhancing activity has been questioned in a recent paper 8 .
We recently identified and developed bona fide autophagy enhancer small molecules in a high-throughput screening of a chemical library that can have beneficial metabolic effects in obese mouse models by enhancing clearance of lipid and ameliorating metabolic inflammation. These effects were attributed to calcineurin-dependent dephosphorylation and activation of TFEB that belongs to the MiTF/TFE family and is a master regulator of lysosome biogenesis and autophagy gene expression 9 . Here, we studied whether our autophagy enhancer small molecule, MSL-7, could expedite clearance of hIAPP oligomer in pancreatic islets and protect β-cells in transgenic mice showing hIAPP accumulation. We found that MSL-7 improved the glucose tolerance and β-cell function of hIAPP + mice rendered diabetic by HFD feeding, which was accompanied by reduced accumulation of hIAPP oligomer and islet amyloid. These results suggest that an autophagy enhancer acting in a TFEB-dependent manner could have a therapeutic potential against human diabetes characterized by islet amyloid accumulation.

Results
Dephosphorylation and nuclear translocation of TFEB and TFE3 by autophagy enhancer, MSL-7. We first studied whether MSL-7 can induce autophagy in INS-1 insulinoma cells. Confocal microscopy after transfection of INS-1 cells with mRFP-GFP-LC3 tandem construct showed that the numbers of both autophagosomes (yellow puncta) and autophagolysosomes (red puncta) were significantly increased after treatment with MSL-7 (Fig. 1a), indicating that MSL-7 induces autophagic activity in INS-1 cells. Conversion of LC3-I to LC3-II in the presence of bafilomycin A1was also increased by MSL-7 ( Fig. 1b and Supplementary  Fig. 1a), supporting increased autophagic flux by MSL-7. Since TFEB, a member of MiTF/TFE family, is a master regulator of autophagy gene expression and lysosomal biogenesis 10 and TFEB is expressed in primary murine islet cells ( Supplementary Fig. 2a,  b), we studied TFEB activation in INS-1 cells. As expected, a significant increase of the number of INS-1 cells with nuclear TFEB was observed after treatment with MSL-7 (Fig. 1c), indicating that MSL-7 activates autophagy of INS-1 cells through TFEB nuclear translocation. Probably because of TFEB activation, expression of SQSTM1 (also known as p62), a target of TFEB 11 , was not reduced but increased by MSL-7 despite activation of autophagy ( Fig. 1b and Supplementary Fig. 1a). We next studied TFE3, another member of the MiTF/TFE family regulating autophagy gene expression and lysosomal biogenesis 12 , which is also expressed in primary murine islets ( Supplementary Fig. 2a,  b). Translocation of TFE3 was also well observed after MSL-7 treatment of INS-1 cells (Fig. 1d). Since the phosphorylation status of MiTF/TFE family members is critical in nuclear translocation and induction of their target genes, we studied phosphorylation of MiTF/TFE family members. When we studied phosphorylation of S142 of TFEB, one of the most important phosphorylation sites of TFEB 13 , S142 phosphorylation was markedly reduced by MSL-7 ( Fig. 1e and Supplementary Fig. 1b), which was consistent with previous data using other types of cells 9 . We also studied phosphorylation of S211 of TFEB, another important phosphorylation site of TFEB, using immunoprecipitation assay based on the binding of the phospho-S211 motif of TFEB to 14-3-3 protein 14 . Band intensity of 14-3-3 protein identified by immunoblotting with anti-14-3-3 antibody in TFEB immunoprecipitate, thus TFEB-bound 14-3-3 protein, was markedly reduced by MSL-7 ( Fig. 1f and Supplementary Fig. 1c), indicating decreased phosphorylation of S211 of TFEB by MSL-7. Band intensity of TFEB with 14-3-3 binding motif identified by immunoblotting with anti-phospho-(Ser) 14-3-3 binding motif antibody in TFEB immunoprecipitate, thus phospho-S211-TFEB 14 was also markedly reduced by MSL-7 ( Fig. 1f and Supplementary Fig. 1c), again supporting decreased phosphorylation of S211 of TFEB by MSL-7. We next studied the phosphorylation of TFE3 using a similar immunoprecipitation assay based on the binding of phospho-S321 motif of TFE3 to 14-3-3 protein 12 . Band intensity of 14-3-3 protein identified by immunoblotting with anti-14-3-3 antibody in TFE3 immunoprecipitate, thus TFE3-bound 14-3-3 protein, and that of TFE3 with 14-3-3 binding motif identified by immunoblotting with anti-phospho-(Ser) 14-3-3 binding motif antibody in TFE3 immunoprecipitate, thus phospho-S321-TFE3 12 , were notably reduced by MSL-7 ( Fig. 1f and Supplementary Fig. 1c). Markedly increased nuclear translocation of TFEB and TFE3 likely due to reduced phosphorylation upon treatment of INS-1 cells with MSL-7 was confirmed by immunoblot analysis after nuclear fractionation ( Fig. 1g and Supplementary Fig. 1d). Reduced phosphorylation of S211-TFEB or S321-TFE3 by MSL-7 was also observed when Tfeb-GFP or Tfe3-GFP was overexpressed in INS-1 cells using the same immunoprecipitation assay employing anti-GFP antibody ( Fig. 1h and Supplementary Fig. 1e). Consequently, nuclear translocation of TFEB-GFP or TFE3-GFP was markedly increased after MSL-7 treatment (Fig. 1i), similar to the nuclear translocation of endogenous TFEB or TFE3. Treatment with MSL-7 inducing nuclear translocation of TFEB and TFE3 upregulated expression of several autophagy genes and lysosomal genes downstream of MiTF/TFE family members such as Uvrag, Becn 1, Sqstm1, Map1lc3, Lamp1, Mcoln1, Ctsa, or Atp6v1h ( Fig. 1j and Supplementary Table 1). Expression of Tfeb or Tfe3 themselves was also induced by MSL-7 (Fig. 1j).
Expedited clearance of hIAPP oligomer by MSL-7. We previously reported that pro-human IAPP (hIAPP) dimer or trimer accumulates when prepro-hIAPP encoding amyloidogenic hIAPP is expressed in INS-1 insulinoma cells 3 which could be the initial seed for hIAPP oligomer, fibril or amyloid 15,16 . Thus, we studied whether MSL-7 could reduce the pro-hIAPP dimer or trimer accumulation. When INS-1 cells were transfected with prepro-hIAPP-HA, pro-hIAPP dimer was observed in addition to hIAPP monomer ( Fig. 2a and Supplementary Fig. 1f), consistent with a previous report 3 . When INS-1 cells were transfected with prepromurine IAPP (mIAPP)-HA, only mIAPP monomer was observed, as expected. When INS-1 cells were incubated with MSL-7, accumulation of the pro-hIAPP dimer following transfection with prepro-hIAPP-HA was notably decreased compared to control cells ( Fig. 2a and Supplementary Fig. 1f). Decreased accumulation of pro-hIAPP dimer by MSL-7 was reversed by bafilomycin A1 inhibiting lysosomal V-ATPase ( Fig. 2a and Supplementary  Fig. 1f), suggesting that enhanced autophagic activity or lysosomal proteolysis is responsible for reduced pro-hIAPP dimer accumulation. In addition to pro-hIAPP dimer, pro-hIAPP trimer was also observed in the presence of bafilomycin A1( Fig. 2a and Supplementary Fig. 1f), consistent with an important role of constitutive autophagy in the clearance of pro-hIAPP dimer and trimer 3 . We also studied whether MSL-7 can ameliorate cell death by hIAPP oligomer accumulation which can be augmented by 3-MA increasing hIAPP oligomer accumulation 3 . When INS-1 cells were incubated with MSL-7, cell death determined by oligonucleosome release following transfection with prepro-hIAPP in the presence of 3-MA was significantly attenuated (Fig. 2b)  could be due to decreased accumulation of toxic hIAPP oligomer 17 by MSL-7. Without 3-MA, cell death was not observed after transfection with prepro-hIAPP which is probably due to the effective clearance of hIAPP oligomer by constitutive autophagy (Fig. 2b), consistent with a previous report 3 .
To study effect of MSL-7 on IAPP oligomer clearance in a more physiological setting, we next studied whether MSL-7 can expedite clearance of endogenous IAPP oligomer. To this end, we employed monkey islet cells expressing amyloidogenic simian IAPP (sIAPP) that is almost identical to hIAPP 3,18 . When we studied whether MSL-7 induces the nuclear translocation of MiTF/TFE family members in monkey islet cells using confocal microscopy, markedly increased nuclear TFEB and TFE3 was observed after treatment with MSL-7 (Fig. 2c, d), similar to the results employing INS-1 cells. When we examined whether MSL-7 can induce autophagic activity in monkey islet cells using confocal microscopy upon transfection with mRFP-GFP-LC3 tandem construct, the numbers of both autophagosomes and autophagolysosomes were significantly increased after treatment with MSL-7 (Fig. 2e), indicating that MSL-7 increases autophagic activity of monkey islet cells, similar to the results employing INS-1 cells. When primary monkey islet cells were cultured in the presence of 3-MA, sIAPP oligomer stained with A11 antibody recognizing hIAPP oligomer 19 significantly accumulated which was not seen without 3-MA ( Fig. 2f), suggesting that amyloidogenic sIAPP oligomer in monkey islet cells is constitutively cleared by autophagy. When monkey islet cells cultured in the presence of 3-MA were treated with MSL-7, accumulation of sIAPP oligomer was significantly reduced (Fig. 2f), suggesting that MSL-7 could expedite clearance of endogenous sIAPP oligomer. Furthermore, confocal microscopy after immunofluorescence revealed markedly increased colocalization of LC3 with A11 + sIAPP oligomer after treatment with MSL-7 in the presence of bafilomycin A1 inhibiting lysosomal degradation of sIAPP oligomer (Fig. 2g), supporting that sIAPP oligomer is cleared by autophagy. Apoptosis of monkey islet cells incubated with 3-MA probably due to accumulation of endogenous sIAPP oligomer 3 was also significantly attenuated by MSL-7 ( Fig. 2h), supporting that sIAPP oligomer is cleared by MSL-7-induced autophagy and that MSL-7-mediated reduction of sIAPP oligomer accumulation leads to the amelioration of cell death.
While behavior of sIAPP oligomer would be similar to that of hIAPP oligomer, we further investigated effect of MSL-7 on hIAPP oligomer clearance employing human β-cells or β-cell lines. We chose 1.1B4 cells that have been generated by electrofusion of human primary pancreatic islet cells to human pancreatic adenocarcinoma cells and express hIAPP 20 , instead of EndoC-βH1 cells that were derived from human fetal pancreas by SV40LT and human TERT transduction but virtually express no hIAPP 21 . MSL-7 treatment induced nuclear translocation of TFEB or TFE3, and enhanced formation of autophagosomes or autophagolysosomes in 1.1B4 cells ( Supplementary Fig. 3a-c). MSL-7 reduced hIAPP oligomer accumulation and apoptosis in 1.1B4 cells incubated with 3-MA enhancing hIAPP oligomer accumulation ( Supplementary Fig. 3d, e). MSL-7 induced expression of autophagy genes, lysosomal genes and TFEB or TFE3 in 1.1B4 cells (Supplementary Fig. 3f and Supplementary Table 2), showing that MSL-7 effectively induced target genes of TFEB or TFE3, and enhanced clearance of hIAPP oligomer in human β-cells as well.
body weight was not significantly changed ( Fig. 5b-d). Glucose tolerance test (GTT) showed that glucose intolerance in HFD-fed hIAPP + mice was significantly ameliorated by MSL-7, accompanied by reduced area under the curve (AUC) (Fig. 5e). Insulinogenic index (the difference between serum insulin 0 and 15 min after glucose infusion divided by the difference between blood glucose 0 and 15 min after glucose infusion) representing β-cell function which was reduced in hIAPP + mice on HFD (Fig. 5f), was also significantly improved by MSL-7 (Fig. 5f), suggesting that improved pancreatic β-cell function could explain amelioration of glucose intolerance of hIAPP + mice on HFD by MSL-7 treatment. We also studied possible effect of MSL-7 on mitochondrial dysfunction caused by hIAPP oligomer 26,27 which can affect glucose-induced insulin release 28 . Seahorse respirometry showed that basal, glucose-stimulated, ATP-coupled or maximal O 2 consumption was suppressed in islets of hIAPP + mice fed HFD (Fig. 5g), suggesting mitochondrial dysfunction of β-cells in these mice. Here, MSL-7 treatment in vivo ameliorated suppression of basal, glucose-stimulated, ATP-coupled or maximal O 2 consumption of pancreatic islet cells of hIAPP + mice on HFD, which could be due to increased turnover of dysfunctional mitochondria through autophagy activation and lead to improved β-cell function in these mice (Fig. 5g). We did not observe improvement of metabolic indexes by MSL-7 administration to hIAPP − mice on HFD for 8 weeks (Fig. 5a-f), which is different from significant metabolic improvement in C57BL/6 mice on HFD for 16 weeks by MSL-7 administration 9 . This discrepancy could be due to strain difference between C57BL/6 and FVB/N mice 29 or duration of HFD feeding (8 vs. 16 weeks). Indeed, glucose tolerance determined by AUC of GTT curves was significantly impaired by HFD feeding of C57BL/6 mice for 8 weeks but not by that of FVB/N mice ( Supplementary Fig. 7a). Furthermore, nonfasting blood glucose in FVB/N mice fed HFD for 8 weeks was not significantly elevated compared to that in the same mice fed normal chow diet (NCD) for 8 weeks (Supplementary Fig. 7b), suggesting that significant insulin resistance was not induced by HFD feeding of mice of FVB/N background and that effect of MSL-7 in hIAPP + mice on HFD is due to its effect on β-cell function associated with hIAPP clearance but not due to improved insulin sensitivity. Consistently, we observed no effect of MSL-7 administration on insulin sensitivity of hIAPP − or hIAPP + mice on HFD as evidenced by no change of AUC of insulin tolerance test (ITT) curves by MSL-7 administration to hIAPP − or hIAPP + mice on HFD ( Supplementary Fig. 7c), again indicating that metabolic improvement by MSL-7 administration to hIAPP + mice is due to improved β-cell function but not due to improved insulin sensitivity. AUC of ITT curves was reduced in HFD-fed hIAPP + mice compared to HFD-fed hIAPP − mice both before (compare gray and blue curve or bar) and after (compare green and red curve or bar) MSL-7 administration showing reduced insulin sensitivity of HFD-fed hIAPP + mice (Supplementary Fig. 7c), probably because elevated blood glucose level in HFD-fed hIAPP + mice itself can reduce insulin sensitivity 30 .
To investigate whether improved glucose profile and β-cell function in vivo by MSL-7 administration were due to the increased clearance of toxic hIAPP oligomer, we conducted immunofluorescence of pancreatic islets using A11 antibody. A11 + hIAPP oligomer accumulation in pancreatic islets of hIAPP + mice on HFD was significantly reduced by in vivo administration of MSL-7 for 8 weeks (Fig. 6a). Some islet cells with A11 + hIAPP oligomer had pyknotic nuclei, suggesting occurrence of apoptosis by hIAPP oligomer accumulation (Fig. 6a). Amyloid staining using (E,E)-1-fluoro-2,5-bis(3-hydroxycarbonyl-4-hydroxy) styrylbenzene (FSB) showed that pancreatic islet amyloid accumulation observed in hIAPP + mice on HFD was also significantly reduced by in vivo administration of MSL-7 for 8 weeks (Fig. 6b). The number of TUNEL + apoptotic β-cells, which was increased in HFD-fed hIAPP + mice, was significantly reduced by MSL-7 treatment for 8 weeks as well (Fig. 6c). Accordingly, β-cell mass determined after insulin immunohistochemistry which was reduced in hIAPP + mice on HFD was significantly recovered by MSL-7 treatment for 8 weeks (Fig. 6d). Insulin content of the pancreas which was reduced in hIAPP + mice on HFD was also restored by MSL-7 treatment for 8 weeks (Fig. 6e). When we studied whether MSL-7 administration could change the distribution of MiTF/TFE family members in pancreatic islets in vivo, nuclear translocation of TFEB and TFE3 in islet cells was clearly observed after 8 weeks of MSL-7 administration as evidenced by cyan nuclei due to colocalization of TFEB and TFE3 fluorescence with nuclear DAPI (Fig. 6f), suggesting that increased activity of MiTF/TFE family members leading to increased autophagic activity is responsible for the increased clearance of islet hIAPP oligomer in vivo and improved β-cell function.
We also studied whether MSL-7 could enhance in vivo clearance of hIAPP oligomer that accumulates through mechanisms other than HFD feeding of hIAPP + mice. When we administered MSL-7 to hIAPP + mice treated with a low dose of streptozotocin (STZ) that has been reported to induce accumulation of hIAPP amyloid without causing extensive β-cell destruction 5 , contents of hIAPP oligomer and islet amyloid in pancreatic islets of STZ-treated hIAPP + mice were significantly reduced ( Supplementary Fig. 8a, b), suggesting that MSL-7 is able to expedite clearance of hIAPP oligomer and amyloid accumulating through diverse pathways. Consistently, elevated nonfasting blood glucose and glucose intolerance after STZ administration to hIAPP + mice were significantly reduced by MSL-7 treatment in vivo ( Supplementary Fig. 8c, d).
To finally confirm that effect of MSL-7 improving metabolic profile of hIAPP + mice on HFD is due to enhanced autophagy or TFEB activation in pancreatic β-cells, we generated hIAPP + mice with β-cell-specific Tfeb-KO (hIAPP + Tfeb Δβ-cell ) by breeding hIAPP + mice with RIP-Cre × Tfeb F/F mice. Nonfasting or fasting blood glucose level after HFD feeding was further increased in hIAPP + Tfeb Δβ-cell mice compared to HFD-fed control hIAPP + Tfeb F/F mice probably due to reduced autophagy or lysosomal function in β-cells, while body weighs were not different between groups ( Supplementary Fig. 9a-c). HFD-induced impairment of glucose tolerance measured by AUC of GTT curves and HFDinduced decrease of insulinogenic index in control hIAPP + Tfeb F/F mice were also aggravated in hIAPP + Tfeb Δβ-cell mice (Supplementary Fig. 9d, e), again probably due to reduced constitutive autophagy or lysosomal function by β-cell-specific Tfeb-KO.
In these mice, we studied whether in vivo effects of MSL-7 is affected by Tfeb of β-cells. We started MSL-7 treatment of hIAPP + Tfeb Δβ-cell mice at the age of 8 weeks (Fig. 7a) because nonfasting blood glucose began to rise in 8-week-old hIAPP + Tfeb Δβ-cell mice on NCD (see Supplementary Fig. 9a), and it might be difficult to compare MSL-7 effect between hIAPP + Tfeb Δβ-cell and hIAPP + Tfeb F/F mice on HFD if basal nonfasting glucose levels before MSL-7 treatment were significantly different between these mice. In hIAPP + Tfeb Δβ-cell mice on HFD, reduction of nonfasting or fasting blood glucose by MSL-7 treatment (differences between orange and violet lines or bars in Fig. 7b, c) was statistically insignificant and less compared to that in hIAPP + Tfeb F/F mice on HFD (differences between blue and red lines or bars in Fig. 7b, c) without differences of body weight between groups (Fig. 7d), supporting that enhanced TFEB activation or autophagy contributes to the improved glucose profile of MSL-7-treated hIAPP + mice on HFD. Improvement of glucose tolerance measured by AUC of GTT curves or that of insulinogenic index after MSL-7 administration to hIAPP + Tfeb Δβ-cell mice on HFD (differences between orange and violet lines or bars in Fig. 7e, f) was also statistically insignificant and less compared to that in hIAPP + Tfeb F/F mice on HFD (differences between blue and red lines or bars in Fig. 7e, f), suggesting the role of enhanced TFEB activation or autophagy in the improvement of glucose tolerance and β-cell function by MSL-7 administration to hIAPP + mice on HFD. Consistently, reduction of hIAPP oligomer or FSB-stained islet amyloid accumulation after MSL-7 administration to hIAPP + Tfeb Δβ-cell mice on HFD (differences between orange and violet bars in Fig. 7g, h) was statistically insignificant and less compared to that in hIAPP + Tfeb F/F mice on HFD (differences between blue and red bars in Fig. 7g, h).

Discussion
We here have demonstrated that the autophagy enhancer small molecule, MSL-7 that could improve metabolic profile of obese mice by activating TFEB and enhancing lysosomal biogenesis or autophagy gene expression, can also improve glucose profile and pancreatic β-cell function of transgenic mice expressing amyloidogenic hIAPP fed HFD. Improvement of pancreatic β-cell function in vivo can be explained by expedited clearance of hIAPP oligomer and amelioration of pancreatic β-cell death caused by hIAPP oligomer accumulation. MSL-7-induced reduction of amyloidogenic IAPP oligomer accumulation and resultant cell death in the presence of 3-MA was observed not only when hIAPP was overexpressed by genetic methods but also when primary monkey islet cells expressing endogenous . e Insulin content of pancreatic islets from HFD-fed hIAPP + and hIAPP − mice treated with MSL-7 was measured by ELISA, which was normalized to the total pancreas weight (F = 9.5, df treatment = 3, df residual = 18). (n = 5 for Veh-treated hIAPP − mice, n = 5 for MSL-7-treated hIAPP − mice, n = 6 for Veh-treated hIAPP + mice and n = 6 for MSL-7-treated hIAPP + mice). f Immunfluorescence of pancreatic sections from HFD-fed hIAPP + and hIAPP − mice treated with MSL-7 using anti-TFEB (left) and anti-TFE3 antibodies (right). Insets were magnified to show colocalization of TFEB/TFE3 with nuclear DAPI. (n = 6 each) All data in this figure are the means ± SEM from more than 3 independent experiments. (scale bar, 100 μm) **P < 0.01; ***P < 0.001 by one-way ANOVA with Tukey's test.
amyloidogenic sIAPP or hiPSC-derived β-cells expressing hIAPP were employed, suggesting physiological relevance of MSL-7 effect improving pancreatic β-cell viability. In addition to the reduction of hIAPP oligomer-mediated β-cell death, ameliorated mitochondrial dysfunction by MSL-7 may also contribute to the improved β-cell function of HFD-fed hIAPP + mice because hIAPP oligomer can cause mitochondrial damage and mitochondrial function is crucial in the insulin release in response to glucose 27,28 . While we observed similar cell death following inhibition of autophagy in both genetic overexpression of hIAPP and endogenous amyloidogenic hIAPP models, accumulated hIAPP species were different between them because pro-hIAPP dimer or trimer was observed in INS-1 cells transfected with prepro-hIAPP-HA, while hIAPP oligomer stained with A11 antibody was observed in hiPSC-derived β-cells expressing endogenous amyloidogenic hIAPP. We have no direct biochemical evidence showing the progression from pro-hIAPP dimer to high-n hIAPP oligomer or amyloid. However, we have observed the proportional relationship between pro-hIAPP dimer and hIAPP oligomer in the same cellsmonkey islet cells transfected with prepro-hIAPP-HA 3 , and it is likely that (pro-)hIAPP dimer or trimer acts as a seed for high-n hIAPP oligomer 16 . Our study also suggests the possibility that while pro-hIAPP dimer may exist in unstressed β-cells and is not cytotoxic, further progression into high-n oligomer occurs in the presence of autophagy impairment or metabolic stress 3 . It should also be noted that A11 antibody that was raised against human Aβ oligomer and has been used to detect hIAPP oligomer, can recognize other proteins such as heatshock proteins 31 . Metabolic effect of MSL-7 appears to be due to the activation of the MiTF/TFE family members enhancing autophagic activity and lysosomal function. In our CRISPR/Cas9 KO experiments, both Tfeb and Tfe3 appear to contribute to the expedited clearance of hIAPP oligomer and reduced cell death by MSL-7 because the reduction of hIAPP dimer accumulation and hIAPPassociated cell death by MSL-7 became statistically insignificant by KO of either Tfeb or Tfe3, suggesting additive effects of Tfeb and Tfe3 in the protection of insulinoma cell death by MSL-7. Consistently, dephosphorylation and nuclear translocation by MSL-7 occurred in both TFEB and TFE3 proteins, which could be dependent on calcineurin 9 . Nuclear translocation of both TFEB and TFE3 by MSL-7 was observed not only in vitro but also . All data in this figure are the means ± SEM from more than 3 independent experiments. (scale bar, 100 μm) # P or *P < 0.05; ## P or **P < 0.01; ### P or ***P < 0.001 by two-way ANOVA with Bonferroni's test (b, e) and one-way ANOVA with Tukey's test (c, e, f, g, h). ( # , comparison between MSL-7-treated hIAPP + Tfeb F/F and Veh-treated hIAPP + Tfeb F/F mice; *, comparison between Veh-treated hIAPP + Tfeb Δβ-cell and Veh-treated hIAPP + Tfeb F/F mice in b, e). (n.s. not significant).
in vivo, which appears to be responsible for reduced hIAPP oligomer/amyloid accumulation in pancreatic islets, ameliorated pancreatic β-cell death, enhanced the pancreatic β-cell function and improved the glucose tolerance of HFD-fed hIAPP + mice. Direct evidence supporting the role of pancreatic β-cell TFEB in the clearance of hIAPP oligomer was obtained using β-cellspecific Tfeb-KO mice crossed to hIAPP + mice. In those hIAPP + Tfeb Δβ-cell mice on HFD, blood glucose profile, glucose tolerance and insulinogenic index were further deteriorated compared to control hIAPP + Tfeb F/F mice on HFD. Moreover, improvement of these metabolic parameters by MSL-7 administration was also diminished in hIAPP + Tfeb Δβ-cell mice on HFD, indicating in vivo role of β-cell TFEB in MSL-7-mediated clearance of hIAPP oligomer.
We have previously reported that autophagy enhancers such as MSL-7 could be a therapeutic agent against diabetes or metabolic syndrome associated with obesity based on our data using mouse models. However, human diabetes is different from murine diabetes in that islet amyloid is found in most patients with type 2 diabetes due to the amyloidogenic propensity of hIAPP 2,32 . Since the aggregate-prone or amyloid proteins are cleared preferentially by autophagy rather than by the proteasomal degradation pathway 6 , the role of autophagy or the effect of autophagy modulators may be more important in human type 2 diabetes than in murine type 2 diabetes, which was strengthened by our observation that MSL-7 can expedite clearance of hIAPP oligomer accumulation in hiPSC-β-cells and protect those cells from hIAPP oligomerinduced cell death. MSL-7 or other autophagy enhancers could be therapeutic agents against human diabetes associated with islet amyloid accumulation.
Cynomolgus monkeys purchased from Guangxi Grandforest Scientific Primate Company Ltd. were maintained in the Orientbio Animal Facility 3 .
All animals were maintained in a specific pathogen free (SPF) facility accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International (AAALAC). All animal experiments were approved by the Institutional Animal Care and Use Committee of Yonsei University Health System (IACUC of YUHS) and were conducted in accordance with the Public Health Service Policy on Humane Care and Use of Laboratory Animals. Monkey experiments were approved by the IACUC of Orientbio, another AAALACaccredited unit. MSL-7 was dissolved in DMSO to make 125 mg/ml stock solution, and diluted with PBS to 5 mg/ml for injection (50 mg/kg, intraperitoneally 3 times a week for 8 weeks). To study MSL-7 effect on STZ-induced hIAPP oligomer accumulation, STZ (Sigma) dissolved in 0.1 M citrate buffer, pH 4.5, was injected into mice intraperitoneally to 2-month-old male hIAPP + mice at a dose of 80 mg/kg. Twenty-four h later, MSL-7 administration was started and continued every other day for a total of 3 times.
Isolation of pancreatic islets. Primary monkey islet cells were isolated using the modified Ricordi's method 34 as previously described 3 . In brief, the pancreas dissected from a 60-month-old male monkey was placed in 4°C HTK solution for distention with 3-4 ml/g pancreas cold Liberase MTF C/T solution (Roche Custom Biotech). The distended pancreas was placed in an isolation chamber equipped with a peristaltic pump, and the temperature was raised to 37°C. The isolation chamber was gently shaken during digestion, and serial samples were examined under a light microscope after dithizone (Sigma Aldrich) staining. When liberated pancreatic islets were observed, Dilution Solution (Corning) was added to inactivate Liberase MTF C/T. Digested cells collected into 200 ml conical tubes were washed with Washing Solution (Corning). Islets were then isolated by continuous density gradient centrifugation using a COBE 2991 Cell Processor (Terumo BCT Inc.) in iodixanol density gradient media (Optiprep TM Axis-Shield, Alere Technologies AS) for culture in CMRL1066-10% FCS at 37°C.
Primary murine islets were isolated from fasted mice using the collagenase digestion technique as previously described 35 . In brief, after injection of 2.5 ml of collagenase P (0.8 mg/ml) into the common bile duct, the pancreas was procured and incubated in a collagenase solution at 37°C for 13 min 20 s. After cessation of enzymatic digestion with cold HBSS-5% FCS, tissue was passed through a 400 μm sieve and then centrifuged on 1.10, 1.085, 1.069 and 1.037 g/ml Biocoll gradients (Biochrom). Islets were collected from the interface using micropipettes.
hiPSC generation and differentiation into insulin-producing cells. hiPSCs were derived by umbilical cord blood mononuclear cells, and reprogramming was conducted using CytoTune-iPS Sendai Reprogramming Kit (Life Technologies). Healthy human cord blood had been obtained from newborn after obtaining informed parental consent at Catholic Hematopoietic Stem Cell Bank of Korea. hiPSCs were adapted to feeder-free conditions on vitronectin (VTN-N) (Gibco)coated plates in TeSR TM -E8 TM medium (STEMCELL Technologies). These cells were incubated at 37°C in a 5% CO 2 atmosphere, and the medium was changed daily. To initiate differentiation, confluent hiPSCs were dissociated into single cell suspension using TrypLE and seeded at 1.0 × 10 6 cells per well in Aggrewell TM 400 plates (STEMCELL Technologies) for incubation in AggreWell TM medium with 10 μM Rho kinase inhibitor (Y-27632). After 24 h of incubation, aggregated cells were differentiated into insulin-producing cells using STEMdiff TM Pancreatic Progenitor Kit (STEMCELL Technologies). At the stage for pancreatic progenitor generation, 0.25 μM SANT1 (Sigma-Aldrich) was added for the last 2 days. Cells were then cultured in CMRL medium (Welgene) containing 25 ng/ml human HGF (PeproTech), 10 mM nicotinamide (Sigma-Aldrich), 20 ng/ml Exendin-4 (Pro-Spec), 1 μM T3 (Sigma-Aldrich) and 10 μM Alk5 inhibitor (Biogems) for 6 days to produce insulin-producing cells. For immunofluorescence or TUNEL staining of iPSC-β-cells, islet-like clusters were pre-embedded in agar before paraffin embedding. All experiments using hiPSC were conducted in accordance with the protocol approved by the IRB of the Catholic University of Korea.
TUNEL staining and β-cell mass. Deparaffinized pancreatic sections were incubated with TUNEL reagent (Roche Applied Science) and DAB (Life Technologies) as the color substrate. Insulin immunochemistry was then conducted by serial incubation with anti-insulin antibody (Cell Signaling Technology, 1:150), biotinylated anti-rabbit antibody (Vector Laboratories, 1:100), streptavidin-alkaline phosphatase, and then with Vector® Blue alkaline phosphatase substrate 3 (Vector Laboratories). The percentage of TUNEL + cells among total β-cells was determined in more than 30 islets per mouse (more than 180 islets per group) by manual counting under BX43 microscope (Olympus). To detect TUNEL + cells among hiPSC-β-cells, TUNEL staining was followed by incubation with anti-insulin antibody and then with Alexa 594-anti-mouse IgG (Life Technologies, 1:200) which was subjected to fluorescent confocal microscopy using LSM700 microscope (Carl Zeiss). The percentage of TUNEL + cells among insulin + cells was determined in more than 20 islet-like clusters by manual counting. Relative β-cell mass was determined by analyzing more than 30 islets per mouse (more than 180 islets per group). After insulin immunohistochemistry using anti-insulin antibody and DAB, point counting morphometry was conducted as previously described 3 .
IAPP expression in vitro. The transient transfection of prepro-hIAPP-HA and prepro-mIAPP-HA 3 was conducted using jetPEI® DNA transfection reagent (Polyplus Transfection) 3 . To block autophagy of transfected cells, after 24 h of transfection, the culture medium was changed to a new medium containing 5 mM 3-MA (Sigma Aldrich) or 10 nM bafilomycin A1 (Sigma Aldrich). After another 16 h of incubation with or without MSL-7, cells were subjected to electrophoretic separation after lysis, immunofluorescence or oligonucleosome detection by ELISA 3 .
Confocal microscopy. Immunofluorescence of MiTF/TFE family members in INS-1 cells, 1.1B4 cells or primary pancreatic islet cells was conducted using anti-TFEB (Bethyl Laboratories, 1:150), anti-TFE3 (Sigma Aldrich, 1:150), or anti-MiTF (Thermo Fisher Scientific, 1:150) as the primary antibodies. For TFEB immunostaining of INS-1, 1.1B4 or primary islet cells, cells with nuclear TFEB were considered to be activated even though residual cytosolic TFEB is detected because TFEB staining was observed only in the cytoplasm and nuclear TFEB was almost completely absent in most of untreated cells, while nuclear TFEB staining with residual cytosolic TFEB staining was seen after treatment with MSL-7. For TFE3 immunostaining, cells with bright nuclear TFE3 without cytosolic TFE3 were considered to be activated because faint nuclear TFE3 was observed in untreated cells, while bright nuclear TFE3 staining without cytosolic TFE3 staining was observed in most cells after MSL-7 treatment. The percentage of cells with nuclear TFEB or TFE3 among total DAPI + cells was determined by counting more than 100 cells from more than 5 fields manually.
hIAPP oligomer in frozen tissue sections or in cells cultured in a Lab-Tak® II Chamber Slide (Thermo Fisher Scientific) was identified by immunostaining using A11 antibody (Merck Millipore, 1:150) and Alexa 488-anti-rabbit IgG (Life Technologies, 1:200). To analyze IAPP oligomer accumulation in cultured cells, confocal microscopy was conducted using LSM700 microscope, and the numbers of A11 + puncta in more than 20 cells were counted manually. To study hIAPP oligomer accumulation in the pancreas, the percentage of A11 + cells among total DAPI + islet cells in pancreatic sections was determined by manually counting more than 20 islets per mouse (more than 132 islets per group). To investigate colocalization of A11 puncta with autophagosome, immunostaining using A11 antibody was followed by immunostaining using anti-LC3 antibody (MBL, 1:150) and then with Alexa 594-anti-mouse IgG (Life Technologies, 1:200). Confocal microscopy was conducted using LSM700 microscope, and the number of yellow puncta (autophagosome colocalized with A11) was counted manually in more than 20 cells.
Amyloid staining. FSB staining for tissue amyloid was conducted using a modification of previously reported methods 36,37 . Briefly, pancreatic sections were deparaffinized in xylene and hydrated down to 30% ethanol. Deparaffinized sections were incubated in 70% formic acid for 20 min. After incubation with 10 μM FSB (Merck Millipore) for 1 h at room temperature with light protection, sections were subjected to fluorescent microscopy using BX53F fluorescent microscope (Olympus). Mean pixel intensity per islet area was determined in more than 25 islets per mouse (more than 151 islets per group) using ImageJ software (NIH).
Transfection and plasmids. Cells were transiently transfected with plasmids such as Tfeb-GFP or Tfe3-GFP using jetPEI® DNA transfection reagent. Nuclear translocation of Tfeb-GFP or Tfe3-GFP was determined after treatment of transfected INS-1, 1.1B4 or primary monkey islet cells with MSL-7 for 4 h by confocal microscopy. To examine autophagosomes and autophagolysosomes, tandem mRFP-GFP-LC3 was transfected using lipofectamine 2000 (Life Technologies). Confocal microscopy was conducted after treatment of transfected cells with MSL-7 for 4 h using LSM700 microscope. GFP or RFP puncta were defined as nonnuclear punctate structures with discrete fluorescent signal on a dark background which was clearly visible with fluorescent confocal microscopy. The numbers of yellow puncta (autophagosomes) and red puncta (autophagolysosomes) in more than 20 cells were counted manually.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data generated or analyzed during this study are available from the corresponding author on reasonable request. Source data are provided with this paper.