Role of Plasminogen Activator Inhibitor-1 in Glucocorticoid-Induced Diabetes and Osteopenia in Mice

Glucocorticoids (GCs) have strong anti-inflammatory properties and are highly effective in the treatment of allergies and inflammatory and autoimmune conditions, such as rheumatoid arthritis, asthma, inflammatory bowel disease, and collagen diseases (1,2). Despite the high efficacy of GC treatment, its clinical use is limited due to adverse effects, such as diabetes, hyperlipidemia, osteoporosis, muscle wasting, and immunosuppression, which depend on the administered dose and duration of GC treatment (3–5).

GC-induced diabetes and osteoporosis are manifestations of adverse metabolic effects because of the high incidence. A clinical study showed that >11% of GC users develop diabetes within 3 years of GC therapy (6), and a larger number of patients transition into a prediabetic state, such as insulin resistance and impaired glucose tolerance. In muscles, GCs have been shown to suppress a number of steps in the insulin signaling network (3,7). GCs directly promote hepatic gluconeogenesis, leading to hyperglycemia (7). Moreover, GCs promote proteolysis, lipolysis, free fatty acid production, and fat accumulation in the liver that contributes to insulin resistance.

GC treatment is the most common cause of secondary osteoporosis (8). GCs affect osteoblasts, osteoclasts, and osteocytes. Osteoblasts are generally considered the main skeletal target (4,9). Several studies suggest that GCs suppress bone formation by inhibiting differentiation, proliferation, and apoptosis of osteoblasts, leading to osteoporosis (4,9); however, the mechanism is not well understood. In addition, the patients treated with GC frequently suffer from muscle wasting (10). GCs are also important mediators of muscle wasting in many pathological conditions, such as sepsis, cachexia, starvation, and metabolic acidosis (11).

Experimental evidence suggests that the above-mentioned effects of GCs are induced in the cytoplasm by the direct action of GC binding to GC receptors (4). Nonetheless, the precise mechanisms are not fully understood, and the evidence of systemic mediators, in the adverse effects of GCs, is lacking.

Plasminogen activator inhibitor-1 (PAI-1) is a serine protease inhibitor that primarily inhibits tissue-type and urokinase-type plasminogen activators; hence, it is an inhibitor of fibrinolysis. PAI-1 is a well-known adipocytokine, being upregulated along with fat accumulation (12). It has been suggested that elevated levels of circulating PAI-1 is a risk factor in cardiovascular diseases (atherosclerosis), obesity, and diabetes (13–15). Moreover, we recently demonstrated in female mice with streptozotocin-induced type 1 diabetes that PAI-1 is involved in bone loss (16). PAI-1 is upregulated in atrophic skeletal muscle (17), and this change is associated with impaired muscle regeneration (18,19). Several clinical studies showed that circulating PAI-1 concentration is elevated in patients with Cushing syndrome or during corticosteroid treatment (20,21). Nonetheless, the role of PAI-1 in GC-induced glucose/lipid abnormalities, osteoporosis, and muscle wasting is unknown. Therefore, we examined the effects of PAI-1 deficiency on GC-induced glucose/lipid and bone metabolism abnormalities as well as muscle wasting in mice.

Forty-four female and 24 male mice with a mixed C57BL/6J (81.25%) and 129/SvJ (18.75%) genetic background were analyzed as described in the figure legends. We included 22 female and 12 male mice with PAI-1 gene deficiency (PAI-1 knockout [KO]) and the corresponding wild-type (WT) control mice (22). These mice were provided by D. Collen (University of Leuven, Leuven, Belgium). Nine-week-old female WT and PAI-1 KO mice received a subcutaneous implant with slow-release pellets containing either 1.5 mg of corticosterone or placebo (Innovative Research of America, Sarasota, FL). These pellets were implanted on days 0, 7, 14, and 21 (23,24). The numbers of female and male mice in each group were 11 and 6, respectively. The animals were maintained in metabolic cages in the 12-h light/12-h dark cycle; they received food and water ad libitum. Four weeks after the first implant of corticosterone or placebo pellets, insulin and glucose tolerance tests were performed. Quantitative computed tomography (qCT) was used to measure bone mineral density (BMD) in the tibia. Mice (placebo- and corticosterone-treated groups) were starved for 6 h before euthanization; blood and tissue samples were collected from the dead mice. All experiments were performed according to the guidelines of the National Institutes of Health and as per the institutional rules put forth for the use and care of laboratory animals at Kinki University.

The plasma concentrations of insulin, total PAI-1, triglyceride (TG), total cholesterol (T-Chol), osteocalcin (OCN), cross-linked C-telopeptide of type I collagen (CTX), and TG content in liver and muscle were measured (25,26). The plasma concentration of uncarboxylated OCN (ucOCN) was measured using Mouse Undercarboxylated Osteocalcin ELISA Kit (MyBiosource, San Diego, CA). Glucose and insulin tolerance tests were performed as previously described (25).

For the qCT analysis of body fat composition, BMD, and muscle volume, the mice were scanned using a LaTheta LCT-200 experimental animal CT system (Hitachi Aloka Medical, Tokyo, Japan).

The tibia, muscle, and liver tissues were fixed in 4% paraformaldehyde for 16 h at 4°C, and the tibia was further fixed for 7 days in 80% ethanol. Muscle and liver tissues were embedded in paraffin right away, and the tibia was embedded in paraffin after dehydration with formic acid. A paraffin block was sliced into 4-μm sections. For visualization of osteoclasts, the slices were stained using a TRAP/ALP Staining Kit (Wako Pure Industries, Osaka, Japan). Immunostaining with alkaline phosphatase (ALP) was performed (27). TUNEL staining was performed to identify apoptotic cells, using In Situ Cell Death Detection Kit, Fluorescein (Roche Diagnostics, Tokyo, Japan). The numbers of osteoclasts and TUNEL-positive and ALP-positive cells were counted in selected visual fields under a microscope in a blinded evaluation (16,27).

Total RNA was extracted from the homogenized tissue samples and the cultured cells using the RNeasy Mini Kit (Qiagen, Tokyo, Japan). Real-time PCR was performed using StepOnePlus and the Fast SYBR Green PCR Master Mix (Life Technologies, Tokyo, Japan) as previously described (25). The primer sets are shown in Supplementary Table 1. The mRNA levels in tissues of the mice and in the cultured cells were normalized to β-actin and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA, respectively.

3T3-L1 cells and C2C12 myoblasts were induced to differentiate into adipocytes and myotubes, respectively (28,29). The differentiated C2C12, 3T3-L1, and HepG2 cells were preincubated with human active PAI-1 (Molecular Innovations, Novi, MI) and dexamethasone (Sigma-Aldrich Japan, Tokyo, Japan) in the serum-free DMEM for 24 h. The cells were incubated with 100 nmol/L insulin for 15 min. Protein levels of phosphorylated Akt (Ser⁴⁷³) and total Akt were measured by Western blotting (30). The expressions of gluconeogenesis-related genes in hepatocytes were measured; HepG2 cells were incubated with 20 nmol/L human active PAI-1 and 100 nmol/L dexamethasone for 24 h in the presence of as well as without 1 nmol/L insulin. Insulin-stimulated 2-deoxyglucose (2DG) uptake in HepG2, C2C12, and 3T3-L1 cells was assessed using 2DG Uptake Measurement Kit (Cosmo Bio, Tokyo, Japan).

Primary osteoblastic cells were obtained from female WT and PAI-1 KO mouse calvaria (27). Cell proliferation, apoptosis, and cell death rates of primary osteoblasts and MC3T3E1 cells were analyzed using the BrdU Cell Proliferation Assay Kit (Exalpha Biologicals, Shirley, MA); In Situ Cell Death Detection Kit, Fluorescein; and trypan blue staining, respectively.

All data were expressed as mean ± SEM. Two-way ANOVA and two-way repeated-measures ANOVA were used to compare the effects of the placebo and corticosterone pellet on both mouse genotypes (PAI-1 WT and PAI-1 KO) for non-repeated and repeated measures, respectively. When significant differences were observed, individual means were compared using Tukey post hoc test. For the other simple comparisons between two groups, the unpaired Student t test was used. Statistical values at P < 0.05 were considered significant. All calculations were performed in the StatView software, version 5.0 (SAS Institute).

Corticosterone treatment increased body weight in female and male WT mice, compared with placebo-treated WT mice (Fig. 1A and B). The qCT analysis showed that the percentage of fat mass, visceral fat, and subcutaneous fat mass were markedly increased by corticosterone in both sexes of WT mice, compared with the corresponding control WT mice (Fig. 1C–F). The PAI-1 deficiency did not affect the body weight gain, percentage of fat mass, visceral fat, and subcutaneous fat mass either in placebo- or corticosterone-treated female and male mice (Fig. 1A–F). The levels of plasma PAI-1 were markedly higher in both sexes of corticosterone-treated WT mice, compared with the corresponding control WT mice (Fig. 1G and H). The levels of PAI-1 mRNA were markedly and significantly elevated in adipose and muscle tissues but significantly decreased in the liver and spleen of corticosterone-treated WT mice compared with placebo-treated WT mice. There were no differences in the levels of PAI-1 mRNA in tibia, heart, lung, and kidney between the placebo- and corticosterone-treated WT mice (Fig. 1I).

The PAI-1 deficiency significantly suppressed the fasting levels of blood glucose and the fasting plasma levels of insulin that were elevated by corticosterone in both sexes of mice (Fig. 2A–D).

The decrease in blood glucose levels after intraperitoneal insulin injection was attenuated in corticosterone-treated WT mice; the PAI-1 deficiency significantly reversed the insulin intolerance of corticosterone-treated WT mice (Fig. 2E and F), although the reversion of blood glucose levels by the PAI-1 deficiency was not significant 90 and 120 min after the insulin injection in male mice. The levels of blood glucose after intraperitoneal glucose injection were markedly elevated in corticosterone-treated WT mice compared with placebo-treated WT mice (Fig. 2G and H). The PAI-1 deficiency significantly suppressed the elevation of blood glucose levels in corticosterone-treated mice 60, 90, and 120 min after the glucose injection (Fig. 2G and H). Moreover, the PAI-1 deficiency significantly suppressed the levels of plasma insulin in the mice 30 min after the glucose injection (Fig. 2I and J).

Active PAI-1 significantly suppressed Akt phosphorylation and glucose uptake induced by insulin in HepG2 cells (hepatocytes), although active PAI-1 did not affect them in differentiated C2C12 cells (myotubes) and differentiated 3T3-L1 cells (adipocytes) (Fig. 3A–F). On the other hand, dexamethasone suppressed Akt phosphorylation and glucose uptake induced by insulin in HepG2, differentiated C2C12, and differentiated 3T3-L1 cells (Fig. 3A–F), whereas the effects of both dexamethasone and PAI-1 treatment were not additive in HepG2 cells (Fig. 3A and D). Moreover, active PAI-1 and dexamethasone did not affect the mRNA levels of key enzymes of gluconeogenesis such as glucose-6-phosphatase (G6Pase) and PEPCK in the presence of insulin in HepG2 cells, although insulin suppressed them in the absence of PAI-1 and dexamethasone (Fig. 3G and H).

Corticosterone markedly increased the levels of plasma TG and T-Chol in female and male WT mice compared with placebo-treated WT mice, whereas the PAI-1 deficiency did not affect the levels of these plasma lipids in corticosterone-treated mice (Fig. 4A–D). The PAI-1 deficiency did not seem to affect corticosterone-induced lipid accumulation in the adipose tissue and liver of the mice in histological analysis (Supplementary Fig. 1A and B). The PAI-1 deficiency did not affect hepatic and muscular TG contents that were elevated by corticosterone (Supplementary Fig. 1D and E).

The qCT analysis revealed that corticosterone treatment significantly reduced BMD, bone volume fraction (trabecular bone volume/total bone volume [BV/TV]), and trabecular area of the tibia in female and male WT mice, compared with placebo-treated WT mice (Fig. 5A–H, O, and P), although it did not affect cortical thickness, cortical area, and total cross-sectional area in the tibia of WT mice (Fig. 5I–N). In contrast, the PAI-1 deficiency blunted the corticosterone-induced decrease in BMD, BV/TV, and trabecular area in both sexes of mice (Fig. 5A–H, O, and P).

Corticosterone treatment markedly suppressed the levels of osteogenic genes, such as Runx2, Osterix, Alp, Ocn, and type 1 collagen (Col1) mRNA in tibia as well as the levels of plasma OCN and ucOCN in WT mice, compared with the placebo-treated WT mice (Fig. 6A–C). The PAI-1 deficiency did not affect these osteogenic markers when they were downregulated by corticosterone in mice (Fig. 6A–C). In contrast, immunohistochemical analysis revealed that the PAI-1 deficiency attenuated the corticosterone-induced decrease in the number of ALP-positive osteoblastic cells on the bone surface of tibia in mice (Fig. 6D). When compared with the tibia of placebo-treated WT mice, corticosterone suppressed the levels of receptor activator of nuclear factor-кB ligand (RANKL) mRNA, osteoprotegerin (OPG) mRNA, the RANKL/OPG ratio, and the number of tartrate-resistant acid phosphatase (TRAP)–positive osteoclasts in the tibia of WT mice (Fig. 6E). In contrast, corticosterone did not affect the levels of plasma CTX, a marker for bone resorption, in WT mice (Fig. 6E). The PAI-1 deficiency did not affect these bone resorption markers in either placebo- or corticosterone-treated mice (Fig. 6E).

Corticosterone treatment increased the number of TUNEL-positive cells on the bone surface of tibia of WT mice, compared with placebo-treated mice (Fig. 7A). The PAI-1 deficiency significantly blunted the corticosterone-induced increase in the number of TUNEL-positive cells on the bone surface of the tibia of the mice (Fig. 7A). Moreover, an in vitro experiment showed that active PAI-1 significantly increased the number of trypan blue–stained and TUNEL-positive MC3T3E1 cells and primary osteoblasts obtained from mouse calvaria (Fig. 7B–E). In contrast, active PAI-1 did not affect BrdU incorporation into MC3T3E1 cells and into primary osteoblasts obtained from mouse calvaria (Fig. 7F and G), indicating that exogenous PAI-1 did not affect the proliferation of osteoblasts in vitro. Moreover, the PAI-1 deficiency did not affect the proliferation impaired by dexamethasone in primary osteoblasts obtained from mouse calvaria (Fig. 7H).

The qCT analysis revealed that corticosterone treatment significantly reduced muscle mass in the whole body of WT mice compared with placebo-treated mice (Fig. 8A and B). Corticosterone significantly decreased the tissue weight of the gastrocnemius muscle, but not of the soleus muscle, in WT mice (Fig. 8C and D). This result was suggestive of the predominant involvement of type II muscle fibers in the GC-induced muscle wasting. The PAI-1 deficiency attenuated the corticosterone-induced decrease in tissue weight of the gastrocnemius muscle in both sexes of mice (Fig. 8C and D). Moreover, the PAI-1 deficiency attenuated the corticosterone-induced decrease in the cross-sectional area of myofibers in the gastrocnemius muscles in female mice (Fig. 8E and Supplementary Fig. 1C). Corticosterone significantly decreased the levels of muscle differentiation–related genes, such as MyoD mRNA, in the gastrocnemius muscle tissues of WT female mice. In contrast, corticosterone did not affect the mRNA levels of myogenin and myosin heavy chain (MHC) in the gastrocnemius muscle tissue of female mice (Fig. 8F). The PAI-1 deficiency significantly attenuated the corticosterone-induced decrease in the levels of MyoD mRNA in mice (Fig. 8F), suggesting that PAI-1 was involved in muscle differentiation suppressed by GCs at an early differentiation stage.

GC treatment simultaneously induces multiple metabolic disorders in an individual (4). Therefore, it is important to identify a common therapeutic target for GC-induced adverse effects. We found that PAI-1 levels are associated with major GC-induced adverse metabolic effects in mice, for example, insulin resistance, osteoporosis, and muscle wasting.

Studies suggest that PAI-1 is linked to insulin resistance as well as metabolic abnormalities (15,31). We recently reported that PAI-1 deficiency ameliorates insulin resistance and hyperlipidemia in obese female mice (25). For the first time, we demonstrated that PAI-1 deficiency reduces insulin resistance and glucose intolerance but not hyperlipidemia in GC-treated mice.

GCs impair glucose metabolism by affecting insulin-sensitive organs (the liver, muscle, and adipose tissue) (3). Several studies revealed that GCs inhibit insulin signaling in 3T3-L1 adipocytes, L6 myotubes, and HepG2 hepatocytes (30,32,33). In this study, dexamethasone inhibited the phosphorylation of Akt and the glucose uptake induced by insulin in 3T3-L1 adipocytes, HepG2 hepatocytes, and C2C12 myotubes in vitro. Nevertheless, incubation with exogenous PAI-1 suppressed insulin signaling in HepG2 hepatocytes but not in 3T3-L1 adipocytes and C2C12 myotubes. Moreover, active PAI-1 seemed to blunt the levels of key enzymes of gluconeogenesis suppressed by insulin in HepG2 cells. Thus, PAI-1 participates in GC-induced insulin resistance by influencing hepatocytes.

Seki et al. (34) reported that dexamethasone increases PAI-1 production in 3T3-L1 adipocytes. Our study showed that corticosterone treatment elevates the levels of PAI-1 mRNA in adipose and muscle tissues of mice. The extent of PAI-1 upregulation was much higher in adipose tissues than in the muscle tissues. This suggests that circulating GC-induced PAI-1 produced in adipose and muscle tissues might cause GC-induced whole-body insulin resistance by impairing insulin signaling in the liver via the bloodstream but not in adipose and muscle tissues. The target organs for inhibition of insulin signaling by GC and PAI-1 seem to be different. PAI-1 circulates in plasma as a complex with vitronectin, an extracellular matrix glycoprotein (35), stabilizing active conformation of PAI-1. López-Alemany et al. (36) showed in vitro that PAI-1 inhibits insulin signaling by competing with integrin αvβ3 for vitronectin binding. Possibly PAI-1 deficiency ameliorated GC-induced insulin resistance via enhancement of the interaction with vitronectin and αvβ3 integrin in the liver. PAI-1 suppression of insulin signaling in hepatocytes, but not in myotubes and adipocytes, is still unknown. Further studies are needed to clarify the molecular mechanism underlying PAI-1 effects on hepatocytes and different PAI-1 tissue sensitivities.

Recent studies suggest that osteoblast-derived OCN, a bone matrix protein, is a potent regulator of glucose metabolism, acting by modulating the insulin release and peripheral insulin sensitivity (37,38). Brennan-Speranza et al. (23) showed that bone-derived OCN release, which is impaired by GC, is involved in GC-induced abnormalities in glucose/lipid metabolism. Our study demonstrated that the PAI-1 deficiency does not affect levels of carboxylated and uncarboxylated plasma OCN when suppressed by GC in mice; this deficiency significantly reduces GC-induced insulin resistance. This suggests that the involvement of PAI-1 in GC-induced insulin resistance is independent of OCN in mice.

This study is the first to report that PAI-1 deficiency ameliorates GC-induced bone loss and decrease in osteoblast numbers in mice. Long-term GC treatment causes bone loss mainly due to the impaired osteoblastic bone formation (4). GCs decrease the number of osteoblasts because of the impaired osteoblast differentiation and proliferation as well as enhanced apoptosis of osteoblasts, resulting in impairment of bone formation (4). According to our data, corticosterone markedly suppressed levels of osteogenic genes in the tibia of mice, indicating that GCs impair osteoblast differentiation. The PAI-1 deficiency did not affect the osteoblast differentiation suppressed by GC. This suggests that PAI-1 is not linked to the impaired GC-induced osteoblast differentiation. Also, either endogenous or exogenous PAI-1 does not affect the proliferation of osteoblasts in vitro. In contrast, the PAI-1 deficiency suppressed corticosterone-induced apoptosis in the tibia of mice. Moreover, active PAI-1 induced apoptosis in primary osteoblasts and MC3T3E1 cells. Overall, this indicates that GCs induce osteopenia through PAI-1 presumably via the enhancement of apoptosis in osteoblasts.

The enhancement of osteoclastic bone resorption that is induced by excess GC is also associated with GC-induced bone loss (4). During the initial stage of high-dose GC therapy, a rapid but transient increase in bone resorption caused by increase in osteoclast number and activity can be observed in humans and in animal models (39,40). Nonetheless, corticosterone decreased the number of osteoclasts, the levels of RANKL mRNA, and the RANKL/OPG ratio in the tibia of mice. It is suggested that excess GC for a prolonged period can downregulate osteoclast numbers and function (41). Furthermore, GC suppresses the formation of osteoclast precursors (41,42). Therefore, we can speculate that the suppression of osteoclast formation by GC in this study might partly be due to the dose and/or duration of GC administration. The PAI-1 deficiency did not affect either osteoclast numbers or the levels of tibia RANKL and OPG mRNA (suppressed by the GC), suggesting that PAI-1 is not involved in the changes of GC-induced bone resorption.

Clinical evidence suggests that GCs negatively affect both muscle and the bone; this phenomenon may lead to muscle wasting as well as osteoporosis (43–45). Our data suggest that PAI-1 deficiency blunts the decrease in muscle mass and the GC-induced changes in muscle phenotypes in mice, indicating that PAI-1 is also involved in GC-induced muscle wasting in mice. Because GCs significantly enhanced endogenous PAI-1 levels in muscle tissues, GCs might induce muscle wasting in mice through both muscle’s endogenous PAI-1 and circulating PAI-1.

GCs directly enhance transactivation of PAI-1 through binding of a GC receptor to a GC response element in PAI-1’s promoter (46). Several studies showed that GCs enhance PAI-1 expression in various cell types (34,47). However, the source tissue of the PAI-1 production enhanced by GC in vivo has not been identified. In the current study, GC significantly decreased the levels of PAI-1 mRNA in the liver and spleen, although the GC elevated PAI-1 mRNA levels in adipose and muscle tissues in mice. Oishi et al. (48) reported that serum GC concentrations are correlated with PAI-1 expression in adipose tissues but not in the liver of mice. In the same study, it was shown that GC-induced adipose tissue–derived PAI-1 possibly had negative effects on the GC-induced liver-derived PAI-1 production. Additionally, our previous study showed that the diabetic state increases PAI-1 expression in the liver of female mice (16). Thus, we can speculate that there are tissue-specific differences in the PAI-1 induction by GC. The response of different tissue-specific factors related to GC may negatively affect the PAI-1 induction by GC, depending on an organ. Alternatively, the change in the metabolic state may modulate the PAI-1 expression in response to GC in different ways, depending on a tissue.

The reversion of blood glucose levels by the PAI-1 deficiency was not significant at 90 and 120 min after insulin injection in male mice (Fig. 2F). The blood glucose levels after insulin injection seemed to be higher (not significant) in PAI-1–deficient mice than in the WT mice, suggesting that slight sex differences might exist in PAI-1 involvement in the pathogenesis of GC-induced insulin resistance in mice. We previously reported that PAI-1 is involved in the pathogenesis of type 1 diabetic osteoporosis in female mice but not in male mice (16). Studies have suggested that a protein linked to the sex chromosomes is associated with the sex differences in the prevalence of metabolic abnormality and osteoporosis (49,50). This suggests that a protein linked to a sex chromosome might be responsible for the sex differences observed in the current study. Further studies are needed to clarify the above issues.

PAI-1 is considered as a humoral factor that affects systemic metabolism. Our previous study suggested that a type 1 diabetic state increases PAI-1 expression in the liver, resulting in increased circulating PAI-1 levels in female mice (16). Upregulation of circulating PAI-1 impairs osteoblast differentiation and mineralization and promotes adipogenesis in bone tissues; the latter effect leads to diabetic osteoporosis. Moreover, PAI-1 is involved in obesity-induced insulin resistance and hyperlipidemia but not bone loss in female mice, although the role of PAI-1 in bone metabolism in the type 2 diabetic state remains completely unknown at present (25). Our study suggested that GC increases PAI-1 production mainly in adipose tissues, thereby leading to increased blood PAI-1 levels. The upregulation of circulating PAI-1 may lead to osteoporosis as well as insulin resistance and muscle wasting induced by GC (Fig. 8G). Thus, PAI-1 possibly plays an important role as a humoral factor to regulate various metabolic states in the presence of excess GC as well as in diabetes.

In conclusion, for the first time, we demonstrated that PAI-1 is involved in the metabolic adverse effects of GC treatment, for example, in insulin resistance, bone loss, and muscle wasting in mice. PAI-1 may be a novel therapeutic target that can help to decrease the risk of GC-induced adverse outcomes. PAI-1 may serve as a diagnostic marker of GC-induced diabetes, osteoporosis, and muscle wasting.

Funding. This study was supported by grants-in-aid 26860152 and 24590289 from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to Y.T. and H.K., respectively) and grants from the Japan Osteoporosis Foundation, the Takeda Science Foundation, and Kinki University.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.T. researched data, contributed to the discussion, and wrote, reviewed, and edited the manuscript. N.K., M.Y., K.Oka., K.Oku., Y.C., and O.M. contributed to the discussion and reviewed and edited the manuscript. H.K. contributed to the discussion and wrote, reviewed, and edited the manuscript. H.K. 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.