The Oral Hypoxia-inducible Factor Prolyl Hydroxylase Inhibitor Enarodustat Counteracts Alterations in Renal Energy Metabolism in The Early Stages Of Diabetic Kidney Disease

Contact: emily.li@wecistanche.com

Introduction

Division of Nephrology and Endocrinology, The University of Tokyo Graduate School of Medicine, Tokyo, Japan; Research Fellowship for Young Scientists, Japan Society for the Promotion of Science, Tokyo, Japan; Biological and Pharmacological Research Laboratories,

Central Pharmaceutical Research Institute, Japan Tobacco Inc., Takatsuki, Japan; Department of Systems Pharmacology, The University of Tokyo Graduate School of Medicine, Tokyo, Japan; Laboratory for Synthetic Biology, RIKEN Center for Biosystems Dynamics Research, Suita, Japan; and WPI International Research Center for Neurointelligence, The University of Tokyo Institutes for Advanced Study (UTIAS), The University of Tokyo, Tokyo, Japan.

Hypoxia-inducible factor (HIF) prolyl hydroxylase inhibitors, also known as HIF stabilizers, increase endogenous erythropoietin production and serve as novel therapeutic agents against anemia in chronic kidney disease. HIFinduces the expression of various genes related to energy metabolism as an adaptive response to hypoxia. However, it remains obscure how the metabolic reprogramming internal tissue by HIF stabilization affects the pathophysiology of kidney diseases. Previous studies suggest that systemic metabolic disorders such as hyperglycemia and dyslipidemia cause alterations of renal metabolism, leading to renal dysfunction including diabetic kidney disease. Here, we analyze the effects of enarodustat (JTZ-951), an oral HIF stabilizer, on renal energy metabolism in the early stages of diabetic kidney disease, using streptozotocin-induced diabetic rats and alloxan-induced diabetic mice. Transcriptome analysis revealed that enarodustatcounteracts the alterations in diabetic renal metabolism. Transcriptome analysis showed that fatty acid and amino acid metabolisms were upregulated in diabetic renal tissue and downregulated by enarodustat, whereas glucose metabolism was upregulated. These symmetric changes were confirmed by metabolome analysis. Whereasglycolysis and tricarboxylic acid cycle metabolites were accumulated and amino acids reduced in renal tissue of diabetic animals, these metabolic disturbances were mitigated by enarodustat. Furthermore, enarodustatincreased the glutathione to glutathione disulfide ratio and relieving oxidative stress in renal tissue of diabetic animals. Thus, HIF stabilization counteracts alterations in renal energy metabolism occurring in incipient diabetic kidney disease.

Cistanche is very good for Kidney function

Translational Statement

Hypoxia-inducible factor (HIF) prolyl hydroxylase inhibitors (also known as HIF stabilizers) increase endogenous erythropoietin production and serve as novel therapeutic agents against anemia in chronic kidney disease. Our transcriptome and metabolome analyses of renal tissue in rat and mouse diabetic models have revealed that enarodustat (JTZ-951), an oral HIF stabilizer, counteracts renal energy metabolism alterations in the early stages of diabetic kidney disease. The results provide important data for extrapolating the effects of HIF stabilizers on renal energy metabolism in clinical settings, although further studies are needed to clarify how this renal metabolism reprogramming by HIF stabilizers affects the progression of diabetic kidney disease.

The introduction

Hypoxia-inducible factor (HIF) prolyl hydroxylase inhibitors increase endogenous erythropoietin production and serve as novel therapeutic agents against anemia in chronic kidney disease. Cells are endowed with a defensive mechanism against hypoxia, and HIF is a master regulator of this defense. Kidneys are physiologically exposed to hypoxia, and chronic hypoxia is recognized as a final common pathway leading to end-stage kidney disease. Considering that HIF induces the expression of various genes related to hypoxia responses, HIF stabilizers might have pleiotropic effects on the progression of kidney diseases as well as improvement in anemia in chronic kidney disease. Interestingly, HIF induces the expression of glycolytic genes and pyruvate dehydrogenase kinase, which inhibits pyruvate dehydrogenase from using pyruvate to fuel the mitochondrial tricarboxylic acid (TCA) cycle.6,7 This metabolic reprogramming from the TCA cycle to glycolysis represses oxygen consumption and is critical for the adaptation of cells exposed to hypoxic environments. However, it remains obscure how the metabolic reprogramming of renal tissue by HIF stabilization affects the pathophysiology of kidney diseases. Diabetic kidney disease (DKD) is the major cause of end-stage kidney disease. Systemic metabolic disorders such as hyperglycemia and dyslipidemia cause renal metabolism alterations, leading to renal dysfunction including DKD. Previous studies have shown increased metabolic flux and accumulation of glucose and TCA cycle metabolites in diabetic renal cortical tissue, which might be related to mitochondrial dysfunction and DKD progression. We hypothesized that HIF stabilizers might reverse the metabolism alterations in diabetic renal cortical tissue, considering that HIF, as an adaptive response to hypoxia, reduces metabolic flux in cells to repress oxygen consumption. Thus, utilizing transcriptome and metabolome analyses, we conducted a proof-of-concept study to comprehensively understand how enarodustat (JTZ-951), an oral HIF stabilizer, affects renal metabolism alterations occurring in the early stages of DKD.

RESULTS: Enarodustat induces metabolic reprogramming from the TCA cycle to glycolysis in renal proximal tubule cells

As the renal cortex is mainly composed of proximal tubules, we first examined the effects of enarodustat on the metabolic flux of renal proximal tubule cells in vitro. First, Mito Stress Test (Agilent Technologies, Inc., Santa Clara, CA) and Glycolytic Rate Assay (Agilent) were conducted in cultured HK-2 cells, a human proximal tubule epithelial cell line (Figure 1a). Enarodustat significantly reduced mitochondrial respiration (TCA cycle) and increased basal glycolysis, indicating metabolic reprogramming from the TCA cycle to glycolysis (Figure 1; Supplementary Figure S1). We also conducted an experiment using small interfering RNA (siRNA) for HIF-1 (Figure 2; Supplementary Figure S2). HIF-1 knockdown by siRNA reversed metabolic alterations (basal respiration, maximal respiration, spare respiratory capacity, and adenosine triphosphate [ATP] production) induced by enarodustat, which showed that the metabolic reprogramming was mainly through HIF-1 stabilization (Figure 2; Supplementary Figure S2). Pyruvate dehydrogenase activity, an important factor for cells to use pyruvate to fuel TCA cycle, was also reduced by enarodustat (Figure 2f), which was compatible with the previously published observations.

Background data of streptozotocin-induced diabetic rats

From the results of in vitro experiments, we hypothesized that HIF stabilizers might alleviate metabolism alterations in diabetic renal cortical tissue through metabolic reprogramming from TCA cycle to glycolysis. We first chose streptozotocin (STZ)-induced diabetic rats as the model for the proof-of-concept experiment to test the above-mentioned hypothesis. In this model, diabetes is rapidly induced, allowing us to observe the net effects of diabetes and HIF stabilizers on the energy metabolism in renal tissue within a short period of time. The study protocol and basic data from experiments in STZ-induced diabetic rats are shown in Figure 3a. We divided the rats into 3 groups: group A (sham), group B (DKD), and group C (DKD þ enarodustat). Blood plasma glucose, glycosylated hemoglobin HbA1c, triglyceride, and total cholesterol levels on day 14 were significantly increased in diabetic groups as compared with group A, whereas there were no significant differences between groups B and C (Figure 3d–g). Although plasma creatinine levels were not different between groups, urinary albumin excretion was significantly increased and glomerulomegaly was noticeable in group B as compared with group A, and enarodustat tended to reverse these changes (Figure 4). Blood urea nitrogen levels were higher in diabetic groups, reflecting dehydration due to diabetes. In summary, the kidneys of STZ-treated rats in our study represent the early stages of DKD. Transcriptome and metabolome analyses were conducted using the renal cortical tissue of these rats.

Transcriptome analysis of renal cortical tissue

The results of transcriptome analysis of renal cortical tissue are shown in Figures 5 and 6. Principal component analysis and hierarchical clustering analysis indicated that groups A, B, and C were separated into different clusters, respectively (Figure 5a and b). Differentially expressed genes were selected by │log2 fold-change (FC) │ ≥0.5 and Q value < 0.05. Gene ontology and canonical pathway analyses revealed that genes related to fatty acid metabolism were upregulated in group B compared with group A. In contrast, genes related to glucose metabolism and hypoxia response including the HIF-1 network were upregulated in group C compared with group B (Figure 5c and d). We also conducted gene set enrichment analysis (GSEA) using the transcriptomic data (Figure 6, Tables 1 and 2). Gene sets of fatty-acid and amino-acid metabolism were upregulated in group B compared with group A (Figure 6). In contrast, these gene sets were downregulated, and gene sets of glucose metabolism were upregulated in group C as compared with group B, showing that enarodustat reversed metabolism alterations induced by diabetes (Figure 6). Moreover, gene sets of TCA cycle were downregulated in group C compared with group B (Table 2), which is compatible with the notion of enarodustat-induced metabolic reprogramming in proximal tubules observed in our in vitro study. In summary, enarodustat counteracted diabetic renal metabolism alterations from transcriptomic perspectives.

Metabolome analysis of renal cortical tissue

We measured the absolute concentrations of 116 energy-related metabolites in renal cortical tissue of rats (n =4, for each group) randomly selected from each group (Supplementary Table S1; Supplementary Figure S3). Partial least squares discriminant analysis (PLS-DA) was performed to assess the significance of class discrimination (Figure 7). We conducted metabolite set enrichment analysis (MSEA) using the metabolites with PLS-DA variable importance in projection (VIP) score $ 1. Differences in metabolism of amino acids, such as glycine, serine, methionine, aspartate, glutamate, arginine, proline, and b-alanine between groups A and B were noted. Differences in amino-acid metabolism were also observed between groups B and C. Moreover, glucose metabolism processes, such as glycolysis and gluconeogenesis, showed different trends between groups B and C (Figure 7).

Visualization of transcriptome and metabolome data on the energy metabolic pathway map

We visualized transcriptome and metabolome data for a comprehensive understanding of energy metabolism alterations (Figure 8). Glycolysis and TCA cycle metabolites were found to be accumulated in group B compared with group A, which might be due to the excessive glucose inflow and upregulation of fatty-acid metabolism in DKD. The amino acid concentration was reduced in group B compared with group A, reflecting the upregulation of amino-acid metabolism (Figure 8a). In contrast, the accumulation of glycolysis metabolites was relieved by enarodustat, due to the facilitated flow of glycolysis. Enarodustat also reversed diabetes-induced changes in TCA cycle metabolites and amino acids (Figure 8b). Moreover, enarodustat alleviated the accumulation of glutathione disulfide (GSSG) in diabetic renal tissue and thus showed a higher glutathione/GSSG ratio, which suggested that enarodustat relieved oxidative stress in DKD (Figure 8a and b). The reduction in oxidative stress was confirmed by the levels of the lipid peroxidation marker malondialdehyde in renal cortical tissue: enarodustat reversed the accumulation of malondialdehyde in diabetic renal cortical tissue (Supplementary Figure S4). In summary, integration of transcriptome and metabolome data has demonstrated that enarodustat counteracts renal energy metabolism alterations occurring in the early stages of DKD.

Symmetric metabolism alterations (diabetes vs. enarodustat) are confirmed in an alternative model

We conducted transcriptome analysis in alloxan-induced diabetic mice, another animal model of diabetes, to confirm our findings in the rat model. Study protocols and background data for the mouse model are shown in Figure 9. Urinary albumin excretion was significantly increased in group B compared with group A, which was reversed by enarodustat (Figure 9d). In addition, we applied comprehensive 3-dimensional analysis (Clear, Unobstructed Brain/ Body Imaging Cocktails, and Computational analysis [CUBIC]–kidney)16 to visualize glomeruli in the kidney. Glomerulomegaly was noticeable in group B compared with group A, which was reversed by enarodustat (Figure 9e). Transcriptome analysis of renal tissue in alloxan-induced diabetic mice showed symmetric metabolism alterations (diabetes vs. enarodustat) in the same way as in the STZ-induced diabetic rat model (Figure 10): fatty-acid metabolism was upregulated by diabetes, whereas glucose metabolism was upregulated by enarodustat. Furthermore, amino-acid metabolism was upregulated by diabetes and downregulated by enarodustat. Thus, enarodustat counteracted renal energy metabolism alterations occurring in the early stages of DKD in the alloxan-induced diabetic mouse model as well as in the STZ-induced diabetic rat model.

DISCUSSION

In this study, we have demonstrated that enarodustat (JTZ-951), an oral HIF stabilizer, counteracts renal energy metabolism alterations occurring in the early stages of DKD in rat and mouse models of diabetes. Transcriptome and metabolome analyses have shown symmetric metabolism alterations in renal tissue (diabetes vs. enarodustat): fatty-acid and amino-acid metabolisms were upregulated in DKD, whereas enarodustat downregulated these pathways and additionally upregulated glucose metabolism (Figures 5–10).

It has not been fully elucidated whether HIF stabilization has protective effects on the pathophysiology of DKD or not. Previous studies have shown that diabetic renal tissue is exposed to hypoxia17 and HIF expression in diabetic renal tissue is insufficient in responding to their hypoxic conditions, which is partially caused by oxidative stress.18 HIF stabilization by cobalt chloride improved oxidative stress status and reduced proteinuria and tubulointerstitial damage in STZ-induced DKD.19 Animal study results also indicated that HIF stabilization protected against the development of obesity,20 improved insulin sensitivity,20 and lowered serum cholesterol levels.21 These results suggest the protective effects of HIF stabilization in the development of DKD. In contrast, supraphysiologic HIF stabilization by von Hippel-Lindau deletion was reported to induce renal fibrosis.22 The role of HIF stabilization in DKD progression may be context-dependent given the pleiotropic effects of HIF and the existence of varied DKD phenotypes. The aim of this study was to clarify the net effects of HIF stabilization on energy metabolism in diabetic kidney. Previous reports have indicated that energy metabolism alterations occur in DKD. Sas et al.10 showed increased energy metabolic flux and accumulation of glucose metabolites in diabetic renal cortical tissue, which might be related to mitochondrial dysfunction. You et al.23 reported that fumarate, a TCA cycle metabolite, was accumulated in diabetic renal tissue and it directly contributed to the progression of DKD. We have previously shown that TCA cycle metabolites accumulated in diabetic renal tissue, which was reversed by sodium-glucose cotransporter 2 inhibition or calorie restriction.11 Moreover, Qi et al.24 showed that enzymes in glycolytic pathways including pyruvate kinase M2 (PKM2), were upregulated in type 1 diabetic patients without DKD compared with patients with DKD. PKM2 activation reversed the accumulation of glycolysis metabolites and restored mitochondrial function, partially by increasing glycolytic flux.24,25 These previous reports suggest that accumulation of glycolysis and TCA cycle metabolites might affect the progression of DKD, which can be mitigated by facilitated glycolysis including PKM2 activation. In our study, HIF stabilization by enarodustat reversed energy metabolism alterations and mitigated the accumulation of glycolysis and TCA cycle metabolites in diabetic renal cortical tissue (Figure 8). The expression of glycolytic enzymes including PKM2 was also upregulated by enarodustat (Supplementary Figure S4B). Moreover, enarodustat alleviated GSSG accumulation and increased the glutathione/GSSG ratio, which suggests that enarodustat relieved the oxidative stress in DKD (Figure 8). This enarodustat-induced reduction in oxidative stress was also confirmed by changes in malondialdehyde levels in renal cortical tissue (Supplementary Figure S4A). These results suggest that HIF stabilization should have protective roles on the pathophysiology of DKD at least from metabolic perspectives. It is certain that we cannot examine whether the metabolic reprogramming by HIF stabilization has direct effects on DKD progression, because it may be impossible to discern the net effects of metabolism alterations on the renal outcome due to the pleiotropic role of HIF. However, in our study, mild urinary albumin excretion and renal pathological abnormalities (glomerulomegaly and glomerular basement membrane thickening) induced by diabetes were mitigated by enarodustat, in association with the normalization of renal energy metabolism. We extracted the genes whose expression changes correlated with urinary albumin levels (257 probes; 232 genes) from the microarray data in the alloxan-induced diabetic mouse model and conducted pathway enrichment analysis (Supplementary Figure S5). As a result, pathways related to the mitochondrial membrane and respiratory electron transport are upregulated associated with urinary albumin levels, indicating that mitochondrial burden is closely related to urinary albumin excretion. Thus, there is a possibility that the metabolic normalization of diabetic renal tissue by HIF stabilizers might reduce the mitochondrial burden and mitigate the progression of DKD. Our data suggest the pathophysiology of DKD from metabolic perspectives as follows: Hyperglycemia creates energy demand for glucose reabsorption in renal proximal tubules; thus, TCA cycle in mitochondria is forcibly activated to meet the energy demand in the early stages of DKD (upregulation of fatty-acid and amino-acid metabolism). And HIF stabilizers might mitigate this mitochondrial burden by the metabolic reprogramming from TCA cycle to glycolysis (downregulation of fatty-acid and amino-acid metabolism and upregulation of glycolysis), which might have protective roles against DKD progression. Further studies including single-cell omics are needed to clarify how the mitochondrial burden in renal tissue directly affects the pathophysiology of DKD. Another interesting question is whether HIF stabilizers directly affect glucose absorption in proximal tubules. Our in vitro data showed that ATP production is significantly reduced by enarodustat through metabolic reprogramming from the TCA cycle to glycolysis (Figures 1 and 2). Considering that ATP is required for glucose reabsorption by sodium-glucose cotransporter 2, we first hypothesized that glucose reabsorption might be inactivated by enarodustat. However, urinary glucose levels were not significantly different between group B and group C in both animal models (Supplementary Figure S6). Probably, the reduction in ATP production by enarodustat is much milder in a physiological situation compared with its effect in vitro; thus, HIF stabilization does not affect glucose absorption through sodium-glucose cotransporter 2, at least in our animal experiments. Our study had 2 limitations. First, we utilized STZinduced diabetic rats and alloxan-induced diabetic mice as early-stage DKD models and observed the short-term effects of diabetic conditions. However, in the clinical setting, HIF stabilizers would be administered to patients with anemia in late-stage DKD. More studies are needed to understand the effect of HIF stabilization on renal metabolism alterations in late-stage DKD. Second, the study lacked sufficient statistical power with respect to the dispersion of metabolites’ concentrations in the metabolome data, whereas the transcriptome data had statistical power large enough to clarify the whole picture of energy metabolism. Because not many metabolites showed a significant difference between groups, we conducted PLS-DA and selected the metabolites with a PLS-DA VIP score ≥1 for MSEA. Although it is difficult to cross-verify a metabolic change using metabolomics data alone, the results of metabolome analysis were compatible with the transcriptomic data (Figures 7 and 8), providing additional support for the observed effects of transcriptome alterations on renal energy metabolism. In conclusion, enarodustat (JTZ-951), an oral HIF stabilizer, counteracts renal energy metabolism alterations in the early stages of DKD. Our study suggests that HIF stabilization may serve as a potential intervention that targets the dysregulated renal energy metabolism in DKD.

METHODS

Mito Stress Test and Glycolytic Rate Assay

HK-2 cells were seeded onto a 96-well microplate at a density of 1 * 10⁴ cells/well. On the next day, metabolic flux was measured in real-time by Seahorse XFe96 Analyzer (Agilent) using Mito Stress Test Kit (103015-100; Agilent) or Glycolytic Rate Assay Kit (103344- 100; Agilent). Concentrations of glucose, pyruvate, and glutamine in culture media were 10 mmol/l, 1 mmol/l, and 2 mmol/l, respectively.

siRNA transfection

HIF-1 knockdown was conducted by Stealth RNAi for human HIF1A (HSS104774 [#1] and HSS104775 [#2]; Thermo Fisher Scientific, Waltham, MA). Stealth RNAi siRNA Negative Control Med GC Duplex #2 (12935112; Thermo Fisher Scientific) was used as the negative control.

Western blotting

Primary antibodies used for staining were anti-human HIF-1α antibody (rabbit polyclonal, 1:500; Novus Biologicals, Littleton, CO) and anti-human actin antibody (rabbit polyclonal, 1:1000; Sigma-Aldrich, St. Louis, MO). A secondary antibody was the horseradish peroxidase-conjugated goat anti-rabbit IgG antibody (170-6515, 1:5000; Bio-Rad Laboratories, Hercules, CA).

Animal experiments

Crl: CD (Sprague Dawley) rats and Crl: CD1 (Institute of Cancer Research) mice were obtained from Charles River Laboratories Japan Inc. (Yokohama, Japan). All experiments were approved by the University of Tokyo Institutional Review Board (approval number P17-110). All animal procedures were performed according to the National Institutes of Health guidelines (Guide for the Care and Use of the Laboratory Animals). Study protocols are shown in Figures 3 and 9.

Transcriptome analysis

Total RNA from renal cortical tissue was isolated using GenElute Mammalian Total RNA Miniprep Kit (RTN70; Sigma-Aldrich). Total RNA (100 ng) was labeled using Low Input Quick Amp Labeling Kit (Agilent) and then hybridized to SurePrint G3 Rat GE v2 8x60K Microarray (for the rat; Agilent) and SurePrint G3 Mouse GE v2 8x60K Microarray (for mouse), respectively. All microarray experiments were performed by DNA Chip Research Inc. (Tokyo, Japan). Raw data were processed with the R package limma in Bioconductor (http://www.bioconductor.org/) to perform background correction and data normalization using the quantile normalization method. Batch effects were removed by ComBat (Bioconductor) in the rat experiment. Differentially expressed genes were determined by│log2 FC│ ≥ 0.5 and Q value < 0.05. Gene ontology and canonical pathway analyses were conducted using BaseSpace Correlation Engine (Illumina, San Diego, CA). Microarray data have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus as series GSE131221 (rat) and GSE139317 (mouse).

Gene Set Enrichment Analysis

GSEA is a computational method that determines whether an a priori–defined set of genes shows statistically significant, concordant differences between 2 biological states. GSEA was performed using GSEA v.3.0 (Broad Institute, Cambridge, MA). The pathways with a false discovery rate < 0.25 were considered significant.

Metabolome analysis

Metabolome measurements were performed by Human Metabolome Technologies Inc. (Tsuruoka, Japan). Metabolite concentrations were calculated by normalizing the peak area of each metabolite with respect to the area of the internal standard and by using standard curves, which were obtained from 3-point calibrations. See also the Supplementary Methods.

Metabolite set enrichment analysis

The analysis of metabolomics data was conducted using the integrated web-based platform MetaboAnalyst 4.0. PLS-DA was performed and associated VIP scores were calculated. The metabolites with PLS-DA VIP score≥ 1 were used for the MSEA.

Quantification of glomerular areas in renal pathologies

We randomly took 3 pathological pictures from each sample’s periodic acid–Schiff staining image and measured every glomerular area respectively using Fiji (a distribution of ImageJ; National Institutes of Health, Bethesda, MD).

CUBIC-kidney 

Comprehensive 3-dimensional imaging of glomeruli in renal tissue was performed using CUBIC according to our previous papers. In brief, fixed mouse kidneys were immersed in CUBIC-L for delipidation and then subjected to immunofluorescent staining. Finally, the refractive index was matched by the placement of the samples in CUBIC-R+. Comprehensive 3D–dimensional images from transparent kidneys were acquired with custom-built light-sheet fluorescence microscopy (MVX10-LS; Olympus, Tokyo, Japan). The primary antibody used for staining was the anti-podocin antibody (rabbit polyclonal, 1:100, P0372; Sigma-Aldrich). The secondary antibody was Alexa Flour 555- conjugated donkey anti-rabbit IgG (1:100, A-31572; Invitrogen, Thermo Fisher Scientific).

Statistical analysis

For multiple comparisons, 1-way analysis of variance followed by the post hoc Tukey multiple comparisons tests, if appropriate, was applied. All statistical analyses were performed with GraphPad Prism 7 software (GraphPad Software, San Diego, CA). P < 0.05 was considered statistically significant. Data are presented as the mean±SD.

DISCLOSURE

MN has received honoraria, advisory fees, or research funding from KyowaHakko-Kirin, Akebia, Astellas, Chugai, GlaxoSmithKline, Japan Tobacco Inc. (JT), Torii, Tanabe-Mitsubishi, Daiichi-Sankyo, Takeda, Ono, Bayer, Boehringer Ingelheim, and Alexion. TT has received a research grant from JT. Enarodustat (JTZ-951) was provided by JT, Tokyo, Japan. EAS and HRU are coinventors of patents owned by RIKEN. Part of this study was done in collaboration with Olympus Corporation and kind software supported by Bitplane. The other author declared no competing interests.

ACKNOWLEDGMENTS

We are grateful to Dr. Shuhei Yao and Dr. Shinji Tanaka for valuable discussions. We also thank Dr. Tsuyoshi Inoue, Dr. Satoru Fukuda, and Ms. Kahoru Amitani for their technical support. This work was partially carried out under the support of the Isotope Science Center, the University of Tokyo. This work was supported by Grant-in-Aid for Japan Society for the Promotion of Science (JSPS) Research Fellow (JSPS KAKENHI grant 19J11928 to SH), Grant-in-Aid for Scientific Research (B) (JSPS KAKENHI grant 18H02824 to MN), Grant-in-Aid for Scientific Research (C) (JSPS KAKENHI grant 17K09688 to TT), and Grant-in-Aid for Scientific Research on Innovative Areas (KAKENHI grant 26111003 to MN).