|Home | About | Journals | Submit | Contact Us | Français|
Osteoblasts/osteocytes are the principle sources of fibroblast growth factor 23 (FGF23), a phosphaturic hormone, but the regulation of FGF23 expression during osteoblast development remains uncertain. Because 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3) and inorganic phosphate (Pi) may act as potent activators of FGF23 expression, we estimated how these molecules regulate FGF23 expression during rat osteoblast development in vitro. 1,25(OH)2D3-dependent FGF23 production was restricted largely to mature cells in correlation with increased vitamin D receptor (VDR) mRNA levels, in particular, when Pi was present. Pi alone and more so in combination with 1,25(OH)2D3 increased FGF23 production and VDR mRNA expression. Parathyroid hormone, stanniocalcin 1, prostaglandin E2, FGF2, and foscarnet did not increase FGF23 mRNA expression. Thus, these results suggest that 1,25(OH)2D3 may exert its largest effect on FGF23 expression/production when exposed to high levels of extracellular Pi in osteoblasts/osteocytes.
Inorganic phosphate (Pi) contributes to multiple cell pathways and processes by acting as a component of mineralized matrices, nucleic acids and phospholipid bilayers; as a source of energy in the hydrolysis of ATP; as a substrate for various kinases/phosphatases; and as a regulator of intracellular signaling. The 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3)–parathyroid hormone (PTH) axis plays a major role in phosphate homeostasis, but clinical features of, for example, vitamin D-resistant rickets suggest the existence of additional phosphaturic factor(s). Recently, intensive studies of several such putative factors (e.g. matrix extracellular phosphoglycoprotein, dentin matrix protein 1 (DMP1), and fibroblast growth factor 23 (FGF23) have provided new insights into phosphate homeostasis (Quarles 2003, Qin et al. 2007). Among these molecules, FGF23 has been studied most extensively in both basic and clinical studies, since it was identified as the factor responsible for autosomal dominant hypophosphatemic rickets (The ADHR Consortium 2000). More recently, this newest FGF family member has been shown to be involved in multiple inherited/acquired hypophosphatemic and chronic kidney disease–mineral bone disorders (Yu & White 2005).
FGF23 is expressed primarily in bones, most notably in osteoblasts and osteocytes (Riminucci et al. 2003, Kolek et al. 2005, Yoshiko et al. 2007b). Indeed, analyses of Fgf23-null (Fgf23−/−) mice on a Hyp (a mouse model of X-linked hypophosphatemia with loss-of-function mutations in phosphate-regulating gene with endopeptidase activity on the X chromosome (Phex)) background (Sitara et al. 2006) and Dmp1−/− mice (Feng et al. 2006) indicated that osteocytes are the major sources of FGF23 at least under these pathological conditions. Klotho, a 130kDa single transmembrane protein having β-glucuronidase activity, appears to form a complex with FGF23 and FGF receptors to support FGF23-dependent signaling in target cells (Urakawa et al. 2006, Kurosu et al. 2007). Indeed, FGF23 is released into the circulation, and it acts on renal proximal tubules to prevent phosphate reabsorption by suppressing the expression of the type IIa and type IIc sodium-dependent phosphate cotransporters (NPT2a,c; Larsson et al. 2004). The polypeptide also suppresses the expression of vitamin D 1α-hydroxylase (1α(OH)ase) and PTH, resulting in a reduction in serum 1,25(OH)2D3 (Shimada et al. 2004) and PTH (Ben-Dov et al. 2007) levels respectively. Given these effects, an excess of active FGF23 in the circulation causes hypophosphatemia with resultant onset of rickets/osteomalacia (Liu & Quarles 2007). Thus, elucidation of the mechanisms of the regulation of FGF23 expression may facilitate the development of new therapies for abnormal phosphate metabolism involving FGF23.
1,25(OH)2D3 appears to be a stimulator of FGF23 expression/production in humans (Burnett-Bowie et al. 2009) and rodents (Shimada et al. 2004, Ito et al. 2005, Kolek et al. 2005, Saito et al. 2005). Klotho−/− mice show extremely high serum FGF23 levels with increased serum 1,25(OH)2D3 and Pi, and decreased serum PTH levels (Yoshida et al. 2002), which are traits that are significantly reversed by ablating 1α(OH)ase as well as Klotho (Ohnishi et al. 2009), suggesting that 1,25(OH)2D3 plays a key role in FGF23 production. However, in Dmp1−/− mice, serum FGF23 levels are high, despite low serum Pi and normal 1,25(OH)2D3 levels (Liu et al. 2008). A contribution of Pi to serum FGF23 levels has been described in rodent models controlled by dietary phosphorus under 5/6 nephrectomized conditions (Saito et al. 2005). Normalization of serum Pi levels by diet increases serum FGF23 levels in vitamin D receptor (Vdr)−/− mice exhibiting hypocalcemia and hypophosphatemia secondary to hyperparathyroidism (Yu et al. 2005).
With respect to in vitro studies, Pi at an optimum concentration (3mM) acts synergistically with 1,25(OH)2D3 to increase FGF23 promoter activity as well as endogenous FGF23 mRNA expression in the K562 human chronic myelogenous leukemia cell line, but not in the MC3T3-E1 mouse osteoblastic cell line (Ito et al. 2005). The 1,25(OH)2D3 effect on FGF23 mRNA expression is observed in UMR-106 osteosarcoma cells (Kolek et al. 2005, Barthel et al. 2007) and fetal rat calvarial cells (Yoshiko et al. 2007b). However, Pi (1–4mM) alone does not change FGF23 promoter activity in ROS17/2.8 rat osteosarcoma cells (Liu et al. 2006a). PTH has also been indicated as a regulator of FGF23; serum FGF23 levels and FGF23 mRNA expression in bones increase in transgenic mice with parathyroid-targeted overexpression of the human cyclin D1 oncogene, a model of primary hyperparathyroidism, and in parathyroidectomized mice (Kawata et al. 2007). In contrast, PTH decreases FGF23 promoter activity in ROS17/2.8 cells (Liu et al. 2006a, Barthel et al. 2007). Taken together, the data indicate a need for additional studies to clarify whether and how FGF23 expression is regulated during osteoblast development. Herein, we have used a well-established rat calvaria (RC) osteoblast developmental model in vitro, and shown that 1,25(OH)2D3 acts predominately in mature cells to increase FGF23 expression in response to high levels of extracellular Pi.
His-tagged human stanniocalcin 1 (STC1) was prepared as described (Yoshiko et al. 2003). Human FGF2 were obtained from R&D Systems (Minneapolis, MN, USA). Selective prostaglandin E receptor subtype EP2 agonist (ONO-AEI-259; Suzawa et al. 2000) was a gift from Ono Pharmaceutical Co. (Osaka, Japan). 1,25(OH)2D3 and synthetic PTH1–34 peptide were purchased from BIOMOL International (Plymouth Meeting, PA, USA) and BACHEM AG (Bubendorf, Switzerland) respectively. All other chemicals, unless otherwise specified, were purchased from Sigma–Aldrich. Stock solutions were prepared in an appropriate vehicle and diluted with a medium (1000 times or more) before use. We used 1,25(OH)2D3 and foscarnet, a competitive inhibitor of NPT at 10nM and 0·5mM respectively according to our previous observations (Yoshiko et al. 2007a,b).
Animal use and procedures were approved by the Committee of Research Facilities for Laboratory Animal Science, Hiroshima University.
Calvariae were obtained from 21-day-old fetal rats as described (Bellows et al. 1986). Briefly, calvariae were minced and digested using collagenase (type I) for 10, 20, 30, 50, and 70min at 37°C. Cells retrieved from the last four of five digestion fractions were separately grown in αMEM containing 10% fetal bovine serum (HyClone, Logan, UT, USA) and antibiotics. After 24h, the cells were recovered, pooled, and grown in multi-well plates or 35mm dishes (0·3×104/cm2) in the same medium supplemented additionally with 50μg/ml ascorbic acid (osteogenic medium). To obtain osteoblast/osteocyte-rich fractions, cells at day 12 were treated with collagenase until cells in osteoid-like nodules were selectively dispersed (Yoshiko et al. 2007a). Recovered cell suspension was then replated at a high cell density (~5×104/cm2), and grown in osteogenic medium for a week. Cells were treated with or without agents including β-glycerophosphate (βGP) either alone or in different combinations for 2 days; βGP was also used as an inducer of matrix mineralization in this model. Medium was changed every 2–3 days, and cultures were maintained at 37°C in a humidified atmosphere with 5% CO2.
Conditioned media were collected, and Pi concentrations were determined colorimetrically (Phospha-C test, Wako Pure Chemical Industries Ltd, Osaka, Japan) according to the manufacturer's directions.
Conditioned media containing cells grown under appropriate conditions were stored at −80°C until use. Levels of FGF23 were measured using an FGF23 ELISA kit (Kainos Lab, Tokyo, Japan) according to the manufacturer's directions.
Total RNA was isolated from cells using TRIzol reagent (Invitrogen) according to the manufacturer's directions. cDNA was synthesized from ≤3μg of total RNA using ReverTra Ace (TOYOBO, Osaka, Japan) at 50°C for 40min. Primer sets for rat osteoblast markers and ribosomal L32 as the internal control are described elsewhere (Yoshiko et al. 2003): alkaline phosphatase (ALP), 5′-GAT AGG CGA TGT CCT TGC AG-3′ and 5′-TTA AGG GCC AGC TAC ACC AC-3′; bone sialoprotein, 5′-CGC CTA CTT TTA TCC TCC TCT G-3′ and 5′-CTG ACC CTC GTA GCC TTC ATA G-3′; osteocalcin (OCN), 5′-AAC GGT GGT GCC ATA GAT GC-3′ and 5′-AGG ACC CTC TCT CTG CTC AC-3′; osteopontin, 5′-AGA GGA GAA GGC GCA TTA CA-3′ and 5′-GCA ACT GGG ATG ACC TTG AT-3′; and ribosomal protein L32, 5′-CAT GGC TGC CCT TCG GCC TC-3′ and 5′-CAT TCT CTT CGC TGC GTA GCC-3′. Primer sets for rat FGF23, VDR, PHEX and DMP1 were designed using Primer Picking (Primer 3) as follows: FGF23, 5′-TAA TAG GGG CCA TGA CCA GA-3′ and 5′-CCT TCC TCT GCA CTC GGT AG-3′; VDR, 5′-ACA GTC TGA GGC CCA AGC TA-3′ and 5′-CTG GTC ATC GGA GGT GAG AT-3′; PHEX, 5′-CCG AAC CAG TGA GGC TAT GT-3′ and 5′-TCA GAG TCC ACA GAC CAC CA-3′; and DMP1, 5′-AGT TCG ATG ATG AGG GGA TG-3′ and 5′-GTC CCT CTG GGC TAT CTT CC-3′. Real-time RT-PCR was carried out using the Light Cycler system (Light CyclerTM DNA Master SYBR Green I; Roche Diagnostics) as described earlier (Wang et al. 2008). To confirm the authenticity of DNA products, each amplicon was sequenced, and qRT-PCR melting curve analysis was performed. Controls containing no reverse transcriptase or no cDNA were also used.
Cells were fixed in neutral buffered formalin for 15min, washed, and incubated with AS MX phosphate/blue LB in 0·1M Tris–HCl (pH 8·3), followed by incubation with 2·5% silver nitrate solution.
Data obtained from at least three samples are expressed as the mean±s.d., and minimum two independent experiments were performed. Statistical differences were evaluated by one-way ANOVA and post hoc Tukey's test.
FGF23 expression appears to be restricted to osteoblasts and osteocytes in normal human (Mirams et al. 2004), rat (Yoshiko et al. 2007b) and Hyp mouse (Liu et al. 2006b) skeletal tissues, but the expression is low throughout osteoblast development in non-1,25(OH)2D3-treated RC cells in vitro (Yoshiko et al. 2007b). To determine whether 1,25(OH)2D3 differentially regulates FGF23 at different osteoblast developmental times, as it does other genes (Gurlek et al. 2002), we compared the effect of 1,25(OH)2D3 on FGF23 levels in conditioned media containing rat calvarial cells at three typical developmental times/stages: d3 (proliferation stage in primary culture), d9 (differentiation stage in primary culture), s-d7 (cells at d12/mature stage in primary culture subcultured and grown for an additional 7 days). Cells in each developmental time window were either treated or not treated with 1,25(OH)2D3 for 2 days in the presence or absence of βGP, the latter was used to stimulate matrix mineralization in the RC model. Based on the relative levels of DMP1 versus ALP and OCN mRNAs, we confirmed that cells at d3 were immature, those at d9 were differentiated, and those at s-d7 were most mature/fully differentiated (i.e. exhibited osteocyte-like features; Toyosawa et al. 2001, Kalajzic et al. 2004; Fig. 1A). FGF23 levels of the cells grown in media without exogenous 1,25(OH)2D3 were very low at all stages, and treatment with 1,25(OH)2D3 increased FGF23 levels at d9 and s-d7 but not at d3; 1,25(OH)2D3-induced FGF23 levels were highest in s-d7 cultures (Fig. 1B).
To determine whether Pi increases FGF23 production, we treated cells at s-d7 with βGP, a substrate known to be hydrolyzed by ALP in the RC cultures, as described earlier (βGP was hydrolyzed within 8h (Bellows et al. 1992)). In fact, Pi levels in the medium containing s-d7 (mature) cells treated with 10mM βGP for 2 days were comparable to those when 10mM Pi was added exogenously (Fig. 2A). βGP alone increased FGF23 levels in the medium, but to a lesser extent than 1,25(OH)2D3 treatment (Fig. 2B). Cotreatment with both reagents led to FGF23 levels in the medium that were 3 and 500 times higher than those obtained with 1,25(OH)2D3 alone or βGP alone respectively (Fig. 2B), suggesting that βGP plus 1,25(OH)2D3 exerted a synergistic effect on FGF23 production. Treatment of s-d7 cells with either βGP or Pi dose-dependently increased FGF23 levels in the medium; the maximum effect (6–9 times higher than those obtained with vehicle alone) was observed at 10mM in each case (Fig. 2C). Correspondingly, βGP and Pi nearly equally increased FGF23 mRNA expression, but less effectively than 1,25(OH)2D3 (Fig. 2D). In contrast to the effects on FGF23 levels in the medium (see Fig. 2B and C), combined 1,25(OH)2D3 and βGP or Pi treatment increased FGF23 mRNA expression only slightly; however, the effects were abolished by cotreatment with foscarnet, a competitive inhibitor of NTP (Fig. 2D). Thus, we concluded that both 1,25(OH)2D3 and Pi increase FGF23 production in RC osteoblast/osteocyte cultures, but the effect of 1,25(OH)2D3 is much larger than that of Pi alone. However, we also established that Pi potently enhances the effect of 1,25(OH)2D3 on FGF23 production.
Consistent with the results showing that 1,25(OH)2D3 increases FGF23 expression via classical VDR/nuclear receptor-mediated pathways (Ito et al. 2005, Liu et al. 2006a), we found that VDR mRNA levels in RC cells were highest in s-d7 cells (Fig. 3A), a temporal pattern that paralleled the FGF23 expression/production profile in response to 1,25(OH)2D3 (cf. Fig. 1B). Moreover, treatment of s-d7 cells with βGP increased VDR mRNA expression, which was further enhanced by cotreatment with 1,25(OH)2D3 (Fig. 3B). Therefore, we concluded that RC osteoblasts/osteocytes but not less mature cells produce a large amount of FGF23 when exposed to 1,25(OH)2D3 concomitant with high levels of extracellular Pi. Because serum FGF23 levels were high in Hyp and Dmp1−/− mice, even with low or normal levels of serum Pi and 1,25(OH)2D3, we also assessed whether βGP and/or 1,25(OH)2D3 downregulate the expression levels of PHEX and DMP1 mRNAs in s-d7 cells. βGP increased both PHEX and DMP1 mRNA levels, but it lost the stimulatory effect on PHEX but not on DMP1 when combined with1,25(OH)2D3 (Fig. 3C and D).
Because PTH is a stimulator of FGF23 expression in a mouse model of primary hyperparathyroidism as well as in parathyroidectomized mice (Kawata et al. 2007), we examined whether PTH increases FGF23 mRNA expression in the RC model. In addition, we tested a number of potential mediators of PTH actions/pathways and factors involved in matrix mineralization. STC1 is suggested to be involved in FGF23 in cartilage organ cultures (Wu et al. 2006) and osteoblast cultures (Yoshiko et al. 2003). Prostaglandin E2 (PGE2) acts diversely on osteoblasts, for example, by increasing receptor activator of NF-κB ligand (RANKL) secretion (Tat et al. 2008). FGF2 inhibits matrix mineralization through the regulation of Pi handling in the MC3T3-E1 mouse calvaria osteoblastic cell line (Hatch et al. 2005). Like foscarnet that decreases NPT activity (Yoshiko et al. 2007a), some of these factors increase or decrease NPT activity in osteoblasts (see, for example, Selz et al. 1989, Veldman et al. 1998, Yoshiko et al. 2003). We also examined forskolin (not shown) and the prostaglandin E receptor subtype EP2, which, like PTH, activates the adenylate cyclase/protein kinase A (PKA) pathway (Narumiya & FitzGerald 2001). However, in contrast to 1,25(OH)2D3, none of these factors increased FGF23 mRNA expression in s-d7 cells in the presence of βGP (Fig. 4A). Notably, there was also no correlation between FGF23 levels and the changes in mineralization elicited by any of these reagents (Fig. 4B and C).
By using the RC cell culture model, we have shown that 1,25(OH)2D3 acts mostly on mature cells to increase FGF23 secretion/mRNA expression during osteoblast development. βGP, apparently via its ability to increase extracellular Pi, enhances the 1,25(OH)2D3 effect on FGF23 secretion, but has only a small effect on its own. The upregulation of VDR mRNA expression in mature cells in response to cotreatment with 1,25(OH)2D3 and βGP provides one plausible explanation for the specificity of their effects on mature osteoblast/osteocyte stages. On the other hand, PTH and other factors involved in NPT activity and/or the PKA pathway did not alter FGF23 mRNA expression. These results suggest that 1,25(OH)2D3 may act specifically on osteoblasts/osteocytes exposed to high levels of extracellular Pi to increase FGF23 production.
Although recent expression profiling of RNA from cortical bones of Hyp mice points towards a potential relationship between FGF23 mRNA expression and molecules involved in Wnt signaling and the FGF family members FGF1 and FGF7 (Liu et al. 2009), to date, 1,25(OH)2D3 and Pi are the most unequivocal stimulators of FGF23 expression/production (Ito et al. 2005) (see Introduction). However, FGF23 responses to either 1,25(OH)2D3 or Pi are not identical across different osteoblastic/non-osteoblastic cell models (Ito et al. 2005, Kolek et al. 2005, Liu et al. 2006a, Barthel et al. 2007, Yoshiko et al. 2007b). For example, in some genetically engineered mouse strains, serum FGF23 levels do not respond to increases or decreases in 1,25(OH)2D3 and Pi levels (Yu et al. 2005, Liu et al. 2008); similar discrepancies exist among culture models (Ito et al. 2005, Liu et al. 2006a) (cf. Introduction). Our data show that there is a synergistic effect of 1,25(OH)2D3 and Pi on FGF23 production, but that the effect is much less on FGF23 mRNA expression. Thus, we speculate that in contrast to 1,25(OH)2D3, which enhances transcriptional activation of FGF23 (Barthel et al. 2007), Pi may be involved in post-translational control of FGF23. Further analysis of the several signaling pathways activated in osteoblasts by Pi uptake (Beck 2003) is needed to elucidate how Pi contributes to 1,25(OH)2D3-induced FGF23 production. Differences in the levels of other systemic and local factors or serum components that participate in Pi homeostasis may also contribute to the variations observed in Pi, 1,25(OH)2D3 and PTH effects on FGF23 expression. These include, for example, factors associated with DMP1 (Feng et al. 2006), PHEX (Liu et al. 2006b) or Klotho as observed in osteoblasts from Klotho mutant mice (Kawaguchi et al. 1999). It is also worth noting that we did not detect downregulation of PHEX and DMP1 concomitant with increased FGF23 production, as observed in Hyp and Dmp1-null mice. However, the increased PHEX mRNA levels, but not DMP1 mRNA levels, induced by βGP were completely abolished by cotreatment with 1,25(OH)2D3 in our model, which are effects that will require further assessment. In this regard, it is interesting that the Hyp bone phenotype is fully rescued by crossing Hyp mice with PHEX transgenic mice, despite FGF23 expression remaining high in bones and uncorrected hypophosphatemia (Erben et al. 2005). DMP1 decreases 1,25(OH)2D3-induced FGF23 mRNA expression in UMR cells (Samadfam et al. 2009). Klotho is not expressed in osteoblasts (Takeshita et al. 2004). Thus, our data, taken together with the previous studies (Takeshita et al. 2004, Erben et al. 2005, Samadfam et al. 2009), support the view that neither PHEX, DMP1 nor Klotho is directly involved in the effect of βGP and/or 1,25(OH)2D3 on FGF23 production/expression.
Our results showing that the magnitude of FGF23 levels in response to 1,25(OH)2D3 varies markedly during osteoblast development may also explain, at least in part, the diverse responses to 1,25(OH)2D3, Pi and/or other factors in functionally and phenotypically different osteoblastic models. Our results are also consistent with in vivo data showing developmental stage-associated differences in the intensity of FGF23 expression by immunohistochemistry and in situ hybridization (Riminucci et al. 2003, Yoshiko et al. 2007b) and with the results of Fgf23-deficient eGFP reporter activity in Hyp mice (Liu et al. 2006b).
We reported that not only 1,25(OH)2D3 (Yoshiko et al. 2007b) but also adenoviral overexpression of FGF23 (Wang et al. 2008) inhibits mineralization in osteoid-like nodules when βGP is present in the RC cell model. Osteoblasts may be exposed to high levels of extracellular Pi during bone resorption and formation, and Pi uptake via NPT3 in osteoblasts may be crucial for matrix mineralization (Suzuki et al. 2006, Yoshiko et al. 2007a). However, old rats overexpressing POU class 1 homeobox 1 via the β-actin promoter exhibit a decrease in bone mineral density with disruption of mineral metabolism (Suzuki et al. 2010). Thus, imbalances in serum levels of either 1,25(OH)2D3 and Pi or both may lead to an overproduction of FGF23. PTH and other factors are capable of regulating serum 1,25(OH)2D3 or Pi levels (Wortsman et al. 1986, Hoppe et al. 1991, Murer et al. 1996, Nakajima et al. 2009) whether FGF23 is involved or not. Further studies are needed to dissect these pathways. In any case, our previous data showing the significance of Pi handling by osteoblasts (Yoshiko et al. 2007a) suggest that Pi levels not only in the serum but also in the bone microenvironment may be crucial for FGF23 expression/production. Our data on the expression pattern of VDR also support the importance of the microenvironment. In any case, the mechanism(s) underlying the ability of Pi and 1,25(OH)2D3 to act cooperatively to increase FGF23 expression/production is unclear. However, taken together with the combined effect of 1,25(OH)2D3 and Pi on FGF23 promoter activity in K562 cells (Ito et al. 2005), it seems likely that the mechanisms are not restricted to osteoblasts/osteocytes. In this regard, VDR mRNA levels are increased in K562 cells treated with 1,25(OH)2D3, but not in those treated with Pi (Ito et al. 2005). Thus, intracellular Pi levels in particular cells such as osteoblasts/osteocytes may play a critical role in 1,25(OH)2D3-dependent events. Collectively, the data support that FGF23 may act as a phosphaturic factor and/or an inhibitor of bone mineralization under the influence of extracellular Pi and 1,25(OH)2D3 locally and systemically.
In summary, the stimulatory effects of 1,25(OH)2D3 on FGF23 mRNA expression/production were observed primarily in mature osteoblasts exposed to high levels of extracellular Pi in the RC cell culture model. VDR mRNA expression was also upregulated in an osteoblast developmental stage-specific manner, and expression was further increased by 1,25(OH)2D3 and extracellular Pi. However, similar to what has been reported in previous in vitro experiments (Liu et al. 2006a), we found no stimulatory effect of either PTH or other molecules known or thought to be downstream of PTH and/or Pi uptake on FGF23 mRNA expression, suggesting that PTH is not a direct stimulator of FGF23 expression at least in cultured RC cells. Thus, we conclude that 1,25(OH)2D3 acts predominately in osteoblasts/osteocytes under conditions of high levels of extracellular Pi to increase FGF23 production in the RC cell culture model, an observation worth further evaluation.
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.
This work was supported in part by grants from the Ministry of Education, Science, Sports and Culture of Japan (18592001 to YY) and the Canadian Institutes of Health Research (FRN 83704 to JEA).
We thank Sayaka Suzuki for her technical assistance.
*(R Yamamoto and T Minamizaki contributed equally to this work)