Insulin-derived amyloidoma, also called an insulin ball, is a skin-related complication of insulin therapy caused by repeated insulin injections at the same site, where native folded insulin changes into amyloid fibrils and forms a mass with a granulomatous reaction. Insulin-derived amyloidoma is a clinically important condition because of its association with subcutaneous insulin resistance, but the precise effect and mechanism of the insulin absorption impairment have not been clarified. We generated insulin-derived amyloidomas in mouse skin, with the amyloidomas large enough to perform insulin tolerance tests in the mass by repeated injections of highly concentrated insulin amyloid fibrils. We demonstrated that the insulin-derived amyloidomas inhibit insulin absorption. By simultaneous administration of insulin and insulin amyloid fibrils, we showed that this effect is due to the amyloid fibril itself in the absence of a granulomatous reaction. In vitro studies revealed that insulin amyloid fibrils have extremely strong adhesion to native human insulin and various insulin analogs. Furthermore, we showed that native insulin that had adhered to insulin amyloid forms amyloid fibrils at physiological pH. These results suggest that the extreme adhesion of insulin amyloid to native insulin is the main mechanism of impaired insulin absorption and amyloidoma growth.
Introduction
Amyloidosis comprises a heterogeneous group of diseases characterized by extracellular deposition of amyloid fibrils in various tissues and organs. To date, more than 36 different proteins have been determined to cause amyloidoses, ranging from localized amyloidosis such as Alzheimer disease and islet amyloidosis to systemic amyloidoses such as Ig light chain amyloidosis and transthyretin amyloidosis (1). Amyloid fibrils result from pathological misfolding and polymerization of precursor proteins and have common structural and biochemical features: nonbranching fibrils with diameters of ∼10 nm and apple-green birefringence when stained with Congo red and visualized with polarized light (2–4).
Localized insulin-derived amyloidosis, also called an insulin-derived amyloidoma or insulin ball, is a unique iatrogenic localized amyloidosis at sites of repeated insulin injections in patients with diabetes (5–7). Human insulin is a 51-residue peptide hormone that consists of primarily an α-helical A chain (21 residues) and a B chain (30 residues) linked by two disulfide bridges, and dissociation from a hexamer to a dimer or monomer is required for absorption from a subcutaneous site into the bloodstream (8). In vitro studies revealed that the kinetics of insulin amyloid fibril formation showed a nucleation-dependent polymerization. Elevated concentrations, high temperatures, high ionic strength of the solution, low pH (9), and preformed amyloid “seeds” reportedly promote formation of amyloid fibrils (10). The precise pathogenesis by which insulin is polymerized into amyloid fibrils in certain patients is still not clear.
Störkel et al. (11) first reported an insulin-derived amyloidoma in 1983; the amyloidoma was believed to be a rare skin complication of insulin therapy. However, in recent years, the number of reported amyloidoma cases has continued to increase, in keeping with the worldwide increase in the number of insulin-dependent patients with diabetes (5,7). Moreover, insulin-derived amyloidoma is possibly underestimated due to misdiagnosis as lipohypertrophy, which is a common skin complication of insulin therapy (6,12). Both conditions appear as subcutaneous lumps, and histopathological examination is necessary to distinguish between the two.
Insulin-derived amyloidoma was initially thought to be an exceptional situation caused by using animal-derived insulin for humans (11,13). However, even human insulin and various types of insulin analogs cause amyloidomas to develop in patients with type 1 and type 2 diabetes (5–7,14–22). Islet amyloid polypeptide (IAPP) (also referred to as amylin) amyloidosis is frequently associated with type 2 diabetes (23). However, there is no evidence of a direct relationship between IAPP amyloidosis and localized subcutaneous insulin-derived amyloidoma.
Insulin-derived amyloidoma is clinically important because it destabilizes glycemic control by inhibiting absorption of freshly injected insulin into the bloodstream (5–7,14–16). Insulin injections at amyloidoma sites cause a significant increase in the insulin requirement (14–16). However, severe hypoglycemia can be caused by moving the injection site to outside the amyloidoma (5,7,16,24). The reported degree of the inhibitory effect of insulin absorption by amyloidomas varies, probably because the degree and distribution of amyloid deposits differ greatly in particular cases (25–28).
With regard to the mechanism by which amyloidomas prevent absorption of freshly injected insulin, the following four hypotheses have been proposed (5,7,16) (Fig. 1). First, amyloid deposits may act as a physical barrier against injected insulin. Second, native insulin may adhere to amyloid fibrils. Third, inflammatory cells and insulin-degrading enzyme (IDE), which are induced by the amyloidomas, may degrade injected insulin. Fourth, injected native insulin may rapidly form amyloid fibrils in the presence of preformed amyloid seeds. However, the mechanism by which amyloidomas prevent absorption of freshly injected insulin has not yet been clarified.
Hypothesis about the mechanism by which amyloidomas inhibit insulin absorption into the bloodstream.
Hypothesis about the mechanism by which amyloidomas inhibit insulin absorption into the bloodstream.
In this study, we therefore generated mouse models to evaluate the effect of subcutaneous insulin resistance specifically caused by an insulin-derived amyloidoma. We also clarified the pathophysiological mechanism of the inhibition of insulin absorption caused by the amyloidoma.
Research Design and Methods
Human Recombinant Insulin and Insulin Analogs
We used a recombinant human insulin (Humulin R; Eli Lilly and Company, Kobe, Japan) and three insulin analogs: insulin lispro (Humalog; Eli Lilly and Company), insulin aspart (NovoRapid; Novo Nordisk Pharma, Tokyo, Japan), and insulin glulisine (Apidra; Sanofi, Frankfurt, Germany).
Insulin Amyloid Fibril Formation In Vitro
Recombinant human insulin solution (3.6 mg/mL) was mixed with an equal volume of 1 mol/L glycine-HCl buffer at pH 2.5 and was incubated at 55°C with agitation for 48 h. After the incubation, the reaction solution was centrifuged (20,000g, 20 min, 4°C), and the pellet of insulin amyloid was washed with saline. This procedure was repeated five times to remove acidic buffer.
Thioflavin T Fluorescence Assay
Amyloid fibril formation was monitored by measuring thioflavin T (ThT) fluorescence. Samples (1 μL) of reaction solutions were mixed with 3 mL of ThT solution (5 μmol/L in glycine-NaOH buffer at pH 9.5). Fluorescence intensity was measured with a spectrofluorometer (F-2700; Hitachi, Tokyo, Japan) with excitation and emission wavelengths of 444 and 482 nm, respectively. Each measurement was done in triplicate.
Congo Red Staining and Electron Microscopy
Congo red staining and electron microscopy confirmed the presence of amyloid fibrils. A 10-μL incubated sample was placed on a glass microscope slide and stained with phenol Congo red. For electron microscopy, a 2-μL incubated sample was placed on a Formvar-coated grid and allowed to adhere for 1 min, after which it was drained by using a strip of filter paper. The sample was then stained with a drop of 0.2% uranyl acetate for 1 min. After excess stain was drained, the grid was air dried and viewed with an electron microscope (H-7700; Hitachi High Technologies) at an 80-kV accelerating voltage.
Subcutaneous Insulin–Derived Amyloidoma Formation in Wild-Type Mice
Insulin-derived amyloidomas were generated by modifying the method reported by Chinisaz et al. (29). Male C57BL/6 mice 3–6 months old received gluteal subcutaneous injections of insulin amyloid fibrils (1.2 mg/200 μL) prepared as described above for seven consecutive days. After the 7 days, all mice manifested an elastic, hard, palpable mass.
Histopathological Evaluation of Amyloidomas
An amyloidoma was excised with surrounding tissues and subsequently fixed in formalin and embedded in paraffin. Sections were stained with hematoxylin and eosin, Congo red, or, for immunohistochemistry, anti-insulin antibody (Abcam, Cambridge, U.K.).
Insulin Tolerance Tests at Amyloidoma Sites
Mice were fasted for 2 h before insulin tolerance tests. Several doses of native human insulin were injected subcutaneously at the site of the amyloidoma in amyloidoma model mice and at the gluteal site in control mice. Blood glucose levels were measured at 0, 30, 60, 90, and 120 min by using a blood glucose meter (Accu-Chek; Roche, Mannheim, Germany).
Evaluation for Insulin Autoantibody
Insulin autoantibodies in the serum from the amyloidoma mouse model and control mice were evaluated using an ELISA kit (MyBioSource, San Diego, CA).
Coadministration of Insulin Amyloid and Native Insulin
Concomitant subcutaneous administration of regular native insulin (1 unit/kg body weight) and insulin amyloid fibrils (1.2 mg/kg body weight) into the gluteal area of mice was performed after mice were fasted for 2 h. Just before injection (<1 min), native and amyloid insulin were mixed. Blood glucose levels were measured at 0, 30, 60, 90, and 120 min.
Evaluation of Adhesion of Insulin Amyloid to Native Insulin In Vitro
Insulin amyloid fibrils (1.9 mg/mL) were mixed with equivalent amounts of native human insulin or insulin analogs in phosphate buffer at pH 7.4. Reaction solutions were centrifuged (20,000g, 20 min, 4°C), after which concentrations of native insulin in the supernatant that was not adsorbed by insulin amyloid were measured by using an ELISA. Briefly, diluted samples in coating buffer (15 mmol/L Na2CO3, 35 mmol/L NaHCO3) were applied to a 96-well ELISA plate and incubated overnight at 4°C. After the plate was washed with washing buffer (0.05% Tween 20 in 50 mmol/L phosphate buffer), it was blocked in Blocking One (Nacalai Tesque, Kyoto, Japan). After additional washes, rabbit anti-insulin antibody (1:5,000) was added and the plate was incubated at 37°C for 1 h. After additional washes, horseradish peroxidase–conjugated goat anti-rabbit IgG (1:2,000; Dako, Glostrup, Denmark) was added and incubation continued at 37°C for 1 h. After extensive washing, the plate was incubated with a substrate (SureBlue; KPL, Gaithersburg, MD) at room temperature for ∼3 min, after which a stop solution (1 N HCl) was added and absorbance was measured at 450 nm via a plate reader (Bio-Rad, Hercules, CA).
Seeding Effect of Amyloid Fibrils at Neutral pH In Vitro
To test the seeding effect of preformed amyloid fibrils, preformed insulin amyloid (0.19 mg/mL) was mixed with an equivalent amount of native human insulin or insulin analogs, and these solutions were incubated at 37°C. The seeding effect was evaluated by measuring ThT fluorescence.
Statistical Analysis
Data are presented as means ± SE. Statistical differences were determined by using the Student t test or one-way ANOVA (with Tukey honestly significant difference) for multiple groups, with the value of P < 0.05 considered to be statistically significant.
Results
Rapid Amyloidoma Formation Caused by Frequent Injections of Highly Concentrated Insulin Amyloid Into Wild-Type Mice
High concentrations of human recombinant insulin incubated under acidic conditions (pH 2.5) and high temperature (55°C) with agitation rapidly formed amyloid fibrils, which were evaluated by using ThT fluorescence, Congo red staining, and electron microscopy (Fig. 2). Repeated subcutaneous injections of highly concentrated (1.2 mg/200 μL) insulin amyloid for seven consecutive days produced an elastic, hard, palpable mass (Fig. 3A). The mass had a spheroid shape (longest diameter 8 mm, thickness 3 mm) and was large enough to allow the performance of insulin tolerance tests. Macroscopically, the mass had a yellow-white, waxy appearance (Fig. 3B). Microscopic analyses showed an acellular material, which was confirmed as amyloid by Congo red staining (Fig. 3C and D), and this material demonstrated positive results for anti-insulin antibody (Fig. 3E). The amyloid was surrounded by granuloma composed of inflammatory cells and fibrous tissue. No significant insulin autoantibody was detected in the serum samples from the amyloidoma mouse model. Quantification of the IDE protein concentration and mRNA expression level showed no significant increase in the mouse skin at the site of the amyloidoma compared with the skin of control mice (Supplementary Fig. 1).
Insulin amyloid fibril formation in vitro monitored by ThT fluorescence, Congo red staining, and electron microscopy. A: ThT fluorescence intensity of human recombinant insulin incubated under acidic (pH 2.5) and high-temperature (55°C) conditions. B: Congo red staining of insulin amyloid formed in vitro. Scale bar = 50 µm. C: Electron microscopic image of insulin amyloid fibrils. Scale bar = 200 nm. A.U., arbitrary units; BF, bright field; PL, polarized light.
Insulin amyloid fibril formation in vitro monitored by ThT fluorescence, Congo red staining, and electron microscopy. A: ThT fluorescence intensity of human recombinant insulin incubated under acidic (pH 2.5) and high-temperature (55°C) conditions. B: Congo red staining of insulin amyloid formed in vitro. Scale bar = 50 µm. C: Electron microscopic image of insulin amyloid fibrils. Scale bar = 200 nm. A.U., arbitrary units; BF, bright field; PL, polarized light.
Insulin amyloidoma formation in mouse skin and insulin tolerance test results at the site of insulin amyloidoma. A: Appearance of a subcutaneous amyloidoma (arrow) after subcutaneous injections of insulin amyloid for seven consecutive days (total 8 mg/mouse). Scale bar = 1 cm. B: Cross-sectional surface of an amyloidoma. Arrows indicate the superior and inferior edges of the mass. Scale bar = 5 mm. C: Congo red staining of an amyloidoma. Scale bar = 500 µm. D: High-power image of C. Scale bar = 50 µm. E: Anti-insulin antibody-stained image of an amyloidoma. Scale bar = 500 µm. F: Electron microscopic image of an amyloidoma. Scale bar = 200 nm. G: No significant hypoglycemic effect was shown after insulin injections (0.5 units/kg body weight) at amyloidoma sites in model mice (n = 3) (closed circle). Control mice (n = 3) (open circle) received the same dose of insulin at gluteal sites. *P < 0.05. H: Blood glucose levels after injections of saline or several doses of insulin (1, 5, and 25 units/kg body weight) (n = 3 per group). *P < 0.05, compared with the saline group. bw, body weight; BF, bright field; PL, polarized light.
Insulin amyloidoma formation in mouse skin and insulin tolerance test results at the site of insulin amyloidoma. A: Appearance of a subcutaneous amyloidoma (arrow) after subcutaneous injections of insulin amyloid for seven consecutive days (total 8 mg/mouse). Scale bar = 1 cm. B: Cross-sectional surface of an amyloidoma. Arrows indicate the superior and inferior edges of the mass. Scale bar = 5 mm. C: Congo red staining of an amyloidoma. Scale bar = 500 µm. D: High-power image of C. Scale bar = 50 µm. E: Anti-insulin antibody-stained image of an amyloidoma. Scale bar = 500 µm. F: Electron microscopic image of an amyloidoma. Scale bar = 200 nm. G: No significant hypoglycemic effect was shown after insulin injections (0.5 units/kg body weight) at amyloidoma sites in model mice (n = 3) (closed circle). Control mice (n = 3) (open circle) received the same dose of insulin at gluteal sites. *P < 0.05. H: Blood glucose levels after injections of saline or several doses of insulin (1, 5, and 25 units/kg body weight) (n = 3 per group). *P < 0.05, compared with the saline group. bw, body weight; BF, bright field; PL, polarized light.
Subcutaneous Insulin Resistance at Sites of Insulin-Derived Amyloidomas
Subcutaneous insulin administration into preformed amyloidomas at 0.5 units/kg body weight resulted in no significant hypoglycemia (Fig. 3G), but the same dose of insulin administered to control mice produced a rapid and significant decrease in blood glucose levels. The blood glucose level did not decrease significantly when a twofold amount of insulin was administered, and a fivefold dose of insulin produced a weak hypoglycemic effect (Fig. 3H).
Subcutaneous Insulin Resistance Without Inflammatory Cell Infiltration
To elucidate the mechanism of subcutaneous insulin resistance at the site of an amyloidoma, which was composed of amyloid and surrounding granuloma, we performed experiments without inflammatory cell infiltration. Consecutive histopathological analyses after a single subcutaneous administration of amyloid fibrils revealed that inflammatory cell infiltration occurred between 1 and 8 h after the amyloid injection (Fig. 4A) and demonstrated the minimal effect of inflammatory cells immediately after amyloid injection. Coadministration of native insulin at 1 unit/kg body weight with insulin amyloid (1.2 mg/kg body weight) produced no significant decrease in blood glucose levels (Fig. 4B), a result that was similar to the kinetics of administration of only amyloid fibrils.
Subcutaneous insulin resistance without inflammatory cell infiltration. A: Congo red staining of subcutaneous tissues at 0–24 h after a single injection of insulin amyloid. Scale bar = 50 µm. Inflammatory cell infiltration occurred between 1 and 8 h after the injection. B: Blood glucose levels after administration of native insulin (1.0 unit/kg body weight) alone or insulin amyloid (1.2 mg/kg body weight) alone and coadministration of native insulin (1.0 unit/kg body weight) and insulin amyloid (1.2 mg/kg body weight). n = 3 per group. *P < 0.05, compared with the amyloid group.
Subcutaneous insulin resistance without inflammatory cell infiltration. A: Congo red staining of subcutaneous tissues at 0–24 h after a single injection of insulin amyloid. Scale bar = 50 µm. Inflammatory cell infiltration occurred between 1 and 8 h after the injection. B: Blood glucose levels after administration of native insulin (1.0 unit/kg body weight) alone or insulin amyloid (1.2 mg/kg body weight) alone and coadministration of native insulin (1.0 unit/kg body weight) and insulin amyloid (1.2 mg/kg body weight). n = 3 per group. *P < 0.05, compared with the amyloid group.
Strong Adhesion of Insulin Amyloid Fibrils to Native Insulin and Insulin Analogs
The adhesion activity of insulin amyloid fibrils induced by human insulin was analyzed in vitro. When equivalent amounts of insulin amyloid derived from native human insulin and native insulin analogs were mixed, the amyloid adhered to 98% of native human insulin and ∼60% of native insulin analogs (Fig. 5).
Adsorption effect of insulin amyloid on human insulin and insulin analogs. After equal parts of native insulin or insulin analogs and insulin amyloid derived from human insulin were mixed, ELISA was used to calculate adsorption ratios of human insulin and insulin analogs (insulin lispro, insulin aspart, and insulin glulisine) to insulin amyloid. All experiments were performed in triplicate.
Adsorption effect of insulin amyloid on human insulin and insulin analogs. After equal parts of native insulin or insulin analogs and insulin amyloid derived from human insulin were mixed, ELISA was used to calculate adsorption ratios of human insulin and insulin analogs (insulin lispro, insulin aspart, and insulin glulisine) to insulin amyloid. All experiments were performed in triplicate.
Effects of Preformed Amyloid Seeds on Insulin Amyloid Fibril Generation at Neutral pH
To clarify whether native insulin that had adhered to insulin amyloid forms amyloid fibrils under physiological conditions, we incubated native insulin with insulin amyloid at neutral pH and 37°C. Native human insulin formed amyloid fibrils at pH 7.4 in the presence of preformed insulin amyloid but did not form amyloid fibrils in the absence of amyloid (Fig. 6). The seeding effect of human insulin–derived amyloid was found not only for human insulin but also for all insulin analogs.
Insulin amyloid fibril generation at neutral pH with preformed amyloid seeds. Kinetics of amyloid fibril formation of human insulin (A) and insulin analogs (insulin lispro [B], insulin aspart [C], and insulin glulisine [D]) in phosphate buffer at pH 7.4 with (black circles) and without (white circles) 10% preformed amyloid seeds derived from human insulin. White squares indicate only preformed amyloid without native insulin. All experiments were performed in triplicate. A.U., arbitrary units.
Insulin amyloid fibril generation at neutral pH with preformed amyloid seeds. Kinetics of amyloid fibril formation of human insulin (A) and insulin analogs (insulin lispro [B], insulin aspart [C], and insulin glulisine [D]) in phosphate buffer at pH 7.4 with (black circles) and without (white circles) 10% preformed amyloid seeds derived from human insulin. White squares indicate only preformed amyloid without native insulin. All experiments were performed in triplicate. A.U., arbitrary units.
Discussion
In this study, we elucidated the degree and the mechanism by which amyloidoma inhibits insulin absorption into the bloodstream by using mouse models. We generated subcutaneous insulin-derived amyloidomas by daily administration of highly concentrated amyloid fibrils for 7 days. Chinisaz et al. (29) previously generated amyloidomas in mice by continuously administering 0.057 mg/day of insulin amyloid for 21 days. We shortened the preparation period and increased the size of the mass so that we could perform insulin tolerance tests by administering much higher doses (1.2 mg/day) of insulin amyloid. Our mice with quickly formed insulin-derived amyloidomas exhibited pathological features similar to those of patients (13–15,18,24,30): reactive inflammation and foreign body reactions with various types of cells, including plasma cells, macrophages, lymphocytes, and multinucleated giant cells in and around the amyloid deposits (Fig. 3).
Insulin therapy for diabetes can cause various skin complications, including lipoatrophy, lipohypertrophy, edema, allergy, acanthosis nigricans, and localized amyloidosis (31). Lipohypertrophy, which is a subcutaneous lump caused by an accumulation of excessive fat tissue, is the most common cutaneous complication of repeated insulin injections. Lipohypertrophy reportedly impairs the absorption of insulin into the bloodstream and yields a 25% lower maximum concentration of plasma insulin (32). Insulin-derived amyloidoma reportedly caused additional impairments in insulin absorption and led to poor glycemic control. Nagase et al. (16). investigated insulin absorption by comparing serum insulin levels after insulin injections into insulin-derived amyloidosis sites versus injections into normal sites in four patients with diabetes and calculated that the mean insulin absorption at insulin-derived amyloidosis sites was 34% of that at normal sites, although differences in the degree of decrease occurred in each case. Variations in the degree of insulin absorption may be attributed to histopathological diversity, because the degree and distribution of amyloid deposits and fibrous tissue differ greatly case by case (25–28). In addition, insulin-derived amyloidomas and lipohypertrophy can reportedly coexist in the same lesion (26). Our mouse model with a pure amyloidoma pathology, as shown in Fig. 3, demonstrated severe subcutaneous insulin resistance, and a minimal hypoglycemic effect was observed even when we administered a 10-fold amount of insulin (5 units/kg) (Fig. 3G and H), which suggests that typical amyloidomas inhibit insulin absorption in critical ways.
Our simultaneous administration of insulin amyloid and native insulin showed no significant decrease in blood glucose level (Fig. 4), which suggests that insulin amyloid itself has a strong inhibitory effect on the absorption of native insulin when inflammatory cells are not present. Moreover, no significant induction of insulin autoantibodies or IDE was observed in the amyloidoma mouse model (Supplementary Fig. 1). Results with dye-labeled insulin suggested that amyloidoma does not act solely as a physical barrier against the injected insulin (Supplementary Fig. 2). In vitro experiments, in which equivalent amounts of amyloid fibrils induced by human insulin and native human insulin or insulin analogs were mixed, showed that ∼98% of native human insulin and >60% of insulin analogs adhered to amyloid fibrils (Fig. 5). These results suggest that this strong adhesion of amyloid fibrils is a major pathogenic pathway of subcutaneous resistance at the site of insulin-derived amyloidomas.
We also investigated whether native insulin adsorbed to insulin amyloid forms amyloid fibrils at physiological pH. Additional amyloid fibril formation derived from native insulin was indeed induced in the presence of preformed amyloid fibrils after a lag phase. However, no significant amyloid fibril formation occurred at pH 7.4 without preformed amyloid. These results suggest that amyloid fibrils have a cross-seeding effect on various insulin analogs at neutral pH, and similar situations may occur in patients receiving several types of human insulin and insulin analogs. Also, the presence of a lag phase in the fibrillation kinetics suggests that freshly injected native insulin first adhered to local amyloid and then changed to an amyloid structure after a certain time period, thus causing an increase in amyloidoma size.
We studied the mechanism of the inhibitory effect of amyloidomas on insulin absorption into the bloodstream by using mouse models. However, several unanswered questions remain, including the mechanism that initiates the first amyloid fibril formation in particular patients. Fibrosis, inflammation, and insulin fragmentation at an injection site may be involved in amyloid fibril formation (5,30), but additional research is required to elucidate the mechanism of initiation of insulin-derived amyloidosis. Regression of an insulin-derived amyloidoma reportedly requires long time periods (5,7), whereas lipohypertrophy usually regresses soon after the insulin injection site is changed, but the precise rate and mechanism of insulin amyloid degradation have not been determined.
In conclusion, we demonstrated that amyloid fibrils adhered strongly to native insulin, which resulted in critical impairment in insulin absorption into the bloodstream and led to amyloid fibril formation and amyloidoma enlargement.
Article Information
Funding. This work was supported by Grant-in-Aid for Scientific Research (B) 16K19516 from the Ministry of Education, Science, Sports, and Culture of Japan.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. M.N. researched data and wrote the manuscript. Y.M. researched data and wrote, reviewed, and edited the manuscript. T.N., W.O., A.I., and K.K. researched data. T.M., T.Y., and Y.I. contributed to the discussion and reviewed the edited manuscript. Y.A. and M.U. wrote, reviewed, and edited the manuscript. Y.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.