|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: KU TN KM HY. Performed the experiments: KU MA MN CY. Analyzed the data: KU MA. Contributed reagents/materials/analysis tools: SM TN KF TT KM HY. Wrote the papefr: KU KM HY. Organized and supervised the project: KM TT HY.
Inositol 1,4,5-trisphosphate receptors (IP3R1, 2, and 3) are intracellular Ca2+ release channels that regulate various vital processes. Although the ryanodine receptor type 2, another type of intracellular Ca2+ release channel, has been shown to play a role in embryonic cardiomyocytes, the functions of the IP3Rs in cardiogenesis remain unclear.
We found that IP3R1−/−-IP3R2−/− double-mutant mice died in utero with developmental defects of the ventricular myocardium and atrioventricular (AV) canal of the heart by embryonic day (E) 11.5, even though no cardiac defect was detectable in IP3R1−/− or IP3R2−/− single-mutant mice at this developmental stage. The double-mutant phenotype resembled that of mice deficient for calcineurin/NFATc signaling, and NFATc was inactive in embryonic hearts from the double knockout-mutant mice. The double mutation of IP3R1/R2 and pharmacologic inhibition of IP3Rs mimicked the phenotype of the AV valve defect that result from the inhibition of calcineurin, and it could be rescued by constitutively active calcineurin.
Our results suggest an essential role for IP3Rs in cardiogenesis in part through the regulation of calcineurin-NFAT signaling.
Intracellular Ca2+ signaling is crucial for cardiac functions . Two types of Ca2+ release channels on the sarcoplasmic/endoplasmic reticulum (SR/ER) serve to regulate Ca2+ release from intracellular Ca2+ stores: the ryanodine receptor (RyR) and inositol 1,4,5-trisphosphate receptor (IP3R). RyR is mainly required for physiologic excitation–contraction coupling in the heart, whereas IP3R mediates Ca2+ mobilization, in response to IP3 produced by phospholipase C activation, not only in most non-excitable cells but also in excitable cells including cardiomyocytes . There have been identified three subtype of IP3Rs (IP3R1, IP3R2 and IP3R3), derived from three distinct genes in mammals , . We previously generated mice that lacked IP3R1, IP3R2 and IP3R3 by disrupting the corresponding gene within the first exon , , and reported the cerebellar phenotype of IP3R1−/− mice  and the pancreatic phenotype of IP3R2−/−-IP3R3−/− mice , thereby demonstrating the specific and redundant roles of IP3Rs in organ development and function. Regarding the heart, each IP3R1, IP3R2, and IP3R3 single-mutant mouse showed normal cardiogenesis, in contrast to the ryanodine receptor type 2 single-mutant mouse, which showed embryonic lethality owing to dysfunction of the SR in the embryonic cardiomyocyte .
Extracellular ligands binding to many receptors, including G-protein coupled receptors and tyrosine-kinase coupled receptors, lead to a transient release of Ca2+ from ER/SR, through IP3Rs. IP3-induced Ca2+ release concurrently results in depletion of intracellular Ca2+ store, which triggers Ca2+ release activated Ca2+ (CRAC) channels . Subsequent increase of cytosolic [Ca2+] through CRAC channels activates several Ca2+- binding proteins, including calcineurin, which in turn dephosphorylates and induces the nuclear localization of the nuclear factor of activated T cells (NFAT) transcription complexes . During heart development, NFATc1 is expressed in the endocardium of the AV canal that will make up the endocardial cushion . NFATc1 knockout embryos show abnormal valvulogenesis , , while NFATc2/3/4 triple knockout embryos and calcineurin-deficient embryos demonstrate impaired endocardial cushion formation, thinning of ventricular myocardium and dysregulation of vascular development , , . To determine the function of the intracellular Ca2+ signaling cascade via IP3Rs in the embryonic hearts, here we generated and analyzed IP3R1 and IP3R2-deficient mice. Our findings support an essential redundant role of IP3R1 and IP3R2 during cardiogenesis, possibly implicating the calcineurin/NFAT signaling pathway.
Firstly, we examined the normal pattern of expression of the IP3Rs by RNA in situ hybridization. Consistent with a previous report , expression of IP3R1 mRNA was detected at embryonic day (E)8.5 in the heart, where it was enhanced in the posterior part of the primitive heart, including the atrium (Fig. 1A). IP3R1 mRNA expression extended to the ventricles through the AV canal at E9.5–10.5 (Fig. 1A). In contrast, IP3R2 mRNA expression was not detected at E8.5 but was detected at E9.5 throughout the heart, including the atrium, AV canal, and ventricles (Fig. 1B). These expression signals were not detected by sense riboprobes as a control (data not shown). Quantitative RT-PCR using total RNA extracted from hearts at E8.5 to E16.0 and western blot analysis of proteins at E9.5 and E12.5 indicated that the IP3R1 and IP3R2 transcripts and proteins were expressed at significant levels in the developing hearts (Fig. 1C and D), consistent with the results of the RNA in situ hybridization experiments. We performed an immunohistochemical analysis on sections of the heart at E9.25, E9.75 and E10.5 to determine the cell types in the embryonic heart that express the IP3R proteins. At E9.25, IP3R1 was expressed in both endocardial cells and myocardial cells, whereas IP3R2 was expressed dominantly in endocardial cells (Fig. 1E). Co-expression of IP3R1 and IP3R2 was observed in the endocardial cells of the AV canal (Fig. 1E). At E9.75 to E10.5, the expression of IP3R2 expanded to the myocardium and IP3R1 and IP3R2 were co-expressed not only in the endocardial cells but also in the myocardial cells (Fig. 1E). These expression signals with anti- IP3R1 and anti- IP3R2 antibodies were not detected in the sections of the IP3R1−/−-IP3R2−/− double-mutant embryos as a control (data not shown).
To explore further the roles of the IP3Rs in cardiac development, we delineated the cardiac phenotype of the IP3R mutant mouse. IP3R1−/−-IP3R2−/− double-mutant mice showed embryonic lethality by E11.5 (see supporting information (SI) Table S1) with heart defects, while either IP3R1+/−-IP3R2−/− or IP3R1−/−-IP3R2+/+ mouse developed normally through E11.5. The IP3R1−/−-IP3R2−/− double-mutant mice began to show growth retardation at E9.5, although they appeared normal by E9.25 and had regular cardiac contractions comparable to those of wild-type. Detailed analysis of mouse hearts at E9.75 showed that the ventricles of IP3R1−/−-IP3R2−/− mice were more transparent than those of IP3R1+/−-IP3R2−/− mice, even though comparable 29 pairs of somites were developed, cardiac looping occurred normally, and morphologic atria and ventricles were apparent in both IP3R1−/−-IP3R2−/− mice and IP3R1+/−-IP3R2−/− mice (Fig. 2A). Histological analysis revealed thin myocardial walls and poor trabeculation in the ventricles of IP3R1−/−-IP3R2−/− mice at E9.75 (Fig. 2B). Together with the ventricular abnormalities, hypocellularity in the cushion of AV canal was documented in the IP3R1−/−-IP3R2−/− hearts (Fig. 2B). Taken together with the evidence of overlapping expression of IP3R1 and IP3R2 in the developing ventricles and AV canal at E9.5 or later, these results suggest that these two genes play redundant roles in cardiogenesis. Although histologic abnormalities were noted, analysis using electron microscopy found no structural defects in the subcellular organelles, in the IP3R1−/−-IP3R2−/− hearts at E9.5 (Fig. S1).
We performed a molecular analysis to investigate the association between the IP3R1−/−-IP3R2−/− double-mutant phenotype and the differentiation of atrial and ventricular cardiomyocytes. Nkx2.5, Mlc2a, and Mlc2v are the earliest markers of cardiomyocyte differentiation in the atrium and ventricle during cardiogenesis , , . These markers were expressed at normal levels in IP3R1−/−-IP3R2−/− hearts at E9.5 (Fig. S2). Markers for the right ventricle (Hand2 ), left ventricle (Hand1 ) and both ventricles (Hrt2 ) were also expressed normally in IP3R1−/−-IP3R2−/− hearts at E9.5 (Fig. S2). These results suggest that cardiomyocyte differentiation and the specification of both ventricles occur normally in IP3R1−/−-IP3R2−/− hearts despite poor formation of the ventricles.
In spite of the well-developed organs other than hearts in IP3R1−/−-IP3R2−/− embryos the early embryonic lethality of IP3R1−/−-IP3R2−/− mice and a previous report on the requirement for certain isotypes of phospholipase Cδ that produce IP3 for placentation ,  led us to investigate the possibility of placental defects in IP3R1−/−-IP3R2−/− mice. Histologic analysis revealed that the embryonic vessels elongated and contact with the maternal blood vessels in the IP3R1−/−-IP3R2+/− or IP3R1+/−-IP3R2−/− placentas at E9.5. The area of the labyrinth layer, which comprises a network of embryonic and maternal blood vessels, was significantly smaller in the IP3R1−/−-IP3R2−/− placenta than that in the IP3R1+/−-IP3R2−/− placenta at E9.5 (Fig. S3).
Next, we looked in the ventricles of IP3R1−/−-IP3R2−/− mice at E9.5 for cell proliferation markers using the anti-phospho-histone H3 (PH3) antibody and 5-bromo-2′-deoxy-uridine (BrdU), and performed the TUNEL assay for apoptosis (Fig. 3 A and B). The numbers of PH3-positive cells and BrdU-incorporated cells were lower in the IP3R1−/−-IP3R2−/− ventricular endocardium and myocardium than in those of IP3R1+/−-IP3R2−/− hearts, whereas the numbers of proliferating cells in the pharyngeal arch and the numbers of apoptotic signals detected by the TUNEL assay were similar between IP3R1−/−-IP3R2−/− and IP3R1+/−-IP3R2−/− embryos (Fig. 3A and B). These results indicate that the redundant functions of IP3R1 and IP3R2 are essential for ventricular cell proliferation. The IP3R1−/−-IP3R2−/− mouse phenotype is similar to that of mice lacking both Ca2+-sensitive transcription factors, NFATc3 and NFATc4, where the proliferation of the ventricular cells is considerably reduced .
Histological analyses of the developing hearts were performed using alcian blue staining, which detects acidic mucosubstances, as well as staining with biotinylated hyaluronan binding protein for visualizing hyaluronan production (Fig. 4A). These analyses revealed the absence of mesenchymal cells in the cushion of the AV canal of IP3R1−/−-IP3R2−/− mice at E9.5, whereas the volume and content of the cardiac jelly in the IP3R1−/−-IP3R2−/− hearts were comparable with those in the IP3R1+/−-IP3R2−/− hearts. Expression of the transcription factor Tbx2 was also normal in the AV canals of IP3R1−/−-IP3R2−/− mice, which suggests that specification and differentiation of the AV canal is unaffected in the mutant mice (Fig. 4A). These findings led us to hypothesize that IP3R1 and IP3R2 were required for the epithelial-mesenchymal transformation (EMT) in the developing AV cushion. To determine whether IP3R-mediated Ca2+ signaling could affect the process of EMT in the AV cushion, we performed an in vitro EMT assay . The E9.5 wild-type AV canal tissues were explanted and treated with 2-aminoethoxydiphenyl borate (2APB) onto a collagen gel, which is a non-selective membrane-permeable inhibitor of IP3Rs and CRAC channels . After 24 h of treatment with 2APB, EMT was inhibited, resulting in a significant decrease in the number of cells that transformed and migrated into the collagen gel compared with control explants treated with DMSO (Fig. 4B). Similarly, EMT was significantly inhibited in the AV canal explants from IP3R1−/−-IP3R2−/− embryo at E9.5 (Fig. 4B).
Of the several signaling pathways implicated in AV cushion and myocardial development, we focused on calcineurin/NFAT signaling because of the similarities noted between the phenotypes of the IP3R1−/−-IP3R2−/− mice and calcineurin- or NFATc-deficient mice , , . To determine whether IP3R1 and IP3R2 regulate intracellular Ca2+ signaling upstream of calcineurin/NFATc, we examined the cellular localization of NFATc1 in the AV canals of IP3R1−/−-IP3R2−/− mice. Immunohistochemical analysis revealed that NFATc1 failed to translocate into the nuclei of the endocardial cells in the AV canals of IP3R1−/−-IP3R2−/− mice at E9.5, and that the expression of NFATc1 was downregulated by E10.5 (Fig. 4C). Nuclear translocation of NFATc1 appeared normal at E9.25 but was inhibited from E9.5 (Fig. S4). Next, we examined the activation of NFATc4 by western blot analysis using the lysates from IP3R1−/−-IP3R2−/− hearts at 9.75. The quantity of the dephosphorylated form of NFATc4 was decreased in the IP3R1−/−-IP3R2−/− heart compared to the IP3R1+/−-IP3R2−/− heart (Fig. 4D). These observations are reminiscent of those made for calcineurin B*/* mice, in which the activation of calcineurin B is disrupted . To provide further evidence for the possible interaction between IP3Rs and calcineurin/NFATc signaling, we tested whether constitutively active calcineurin could rescue the phenotype of IP3R1−/−-IP3R2−/−. Infection of cytomegalovirus-associated constitutively active calcineurin, which can be activated independently of Ca2+ increase by deletion of its calmodulin binding domain and autoinhibitory domain , significantly restored the EMT activity in IP3R1−/−-IP3R2−/− AV explants (Fig. 4E). Taken together, our results indicate that IP3R1 and IP3R2, at least in part, redundantly activate and maintain calcineurin/NFATc signaling.
To investigate further whether IP3R-mediated Ca2+ signaling is required for AV cushion development, we utilized a relatively simple model of heart development in zebrafish treated with an inhibitor of IP3R (Fig. 5A). As described previously , treatment with the calcineurin inhibitors cyclosporine A (CsA) or FK506 between 24 h and 33 h postfertilization (hpf) affects AV cushion development and heart valve formation in zebrafish. Intriguingly, treatment with 2APB between 24 hpf and 33 hpf disrupted AV cushion development in zebrafish (Fig. 5A) in a dose-dependent fashion (Fig. 5B). This results in a massive regurgitation of the AV valve and heart failure with pericardial effusion reminiscent of treatment with CsA (Fig. 5A and Video S1). Although myocardial development appeared normal by staining of the sarcomeric protein MF20, disruption of AV canal cushion development, and possibly decreased myocardial function, was common in CsA-treated and 2APB-treated zebrafish (Fig. 5A and B). These data further support that IP3R-mediated signaling is essential for AV cushion development, possibly implicating the calcineurin/NFAT signaling across species.
In the present study, we used a series of molecular, pharmacologic and genetic manipulations in mice and zebrafish to demonstrate that IP3Rs are involved in the local control of Ca2+ that is necessary to activate the calcineurin-NFATc signaling pathway required for cardiogenesis. To date, it has been shown that calcineurin/NFATc signaling has an important role for AV cushion formation, valvulogenesis, and cardiomyocyte development , , , , however, the process by which this signaling is activated by Ca2+ has not been elucidated. In Xenopus embryos, NFAT is a downstream effecter of the IP3-Ca2+ signal during dorsoventral axis formation . In mammals, IP3Rs act upstream of calcineurin/NFAT signaling in T lymphocytes in vitro . In this study, we have indentified that IP3R1 and IP3R2 may exist, at least in part, upstream of calcineurin/NFATc signaling during heart development. Our data provide new insights into the way how Ca2+ signals regulate Ca2+-binding molecules essential for cardiogenesis, such as the Ca2+ dependent phosphatase, calcineurin.
Here, we have shown several lines of findings that support that the IP3R-mediated Ca2+ signaling may be implicated in the calcineurin/NFATc signaling pathway although it is difficult to clarify whether IP3Rs may directly activate calcineurin during cardiogenesis. We showed that NFATc1 and NFATc4 were not activated in IP3R1−/−-IP3R2−/− mice at E9.5 (Fig. 4C and D), and that the expression of NFATc1 was downregulated by E10.5 (Fig. 4C), using immunohistological and biochemical experiments. Moreover, we showed that the disturbance of endocardial cushion development in IP3R1−/−-IP3R2−/− embryos could be rescued by addition of constitutively active calcineurin (Fig. 4E). These data indicate that IP3R1 and IP3R2 may, directly or indirectly, function redundantly in the activation of calcineurin/NFAT signaling. Vascular endothelial growth factor (VEGF), inducing NFATc1 nuclear translocation, markedly increases the proliferation of valve endothelial cells in vitro . The VEGF receptor tyrosine kinase activating phospholipase Cγ- IP3 signal triggers intracellular Ca2+ increase , suggesting that VEGF might be a candidate of extracellular signals upstream of IP3Rs-NFATc1 activation in the endocardium. Failure of nuclear translocation and downregulation of NFATc1 is probably due to disruption of some of the following positive feedback mechanisms: 1) NFATc1-binding to the promoter/enhancer of the NFATc gene ; 2) activation of calcineurin transcription by NFATc ; and 3) induction of the IP3R1 gene as previously documented in neurons .
We also showed that the phenotype of IP3R1−/−-IP3R2−/− embryos involved that of the NFATc2/3/4 knockout mice , ,  and was much more severe than that of the NFATc1 null mice , . Therefore, it is likely that IP3R-mediated Ca2+ signaling during cardiogenesis regulates NFATc family members in addition to NFATc1 as well as unknown downstream targets other than the calcineurin/NFATc signaling pathway. It is also unclear whether endocardial or myocardial cells are primarily affected in the IP3R1−/−-IP3R2−/− embryos. In characterizing the role of this pathway in heart development, our study revealed the co-expression and involvement of IP3R1 and IP3R2 in the development of both endocardial and myocardial cells. Unequivocal answer about the functions of IP3R1 and IP3R2 for cardiac development awaits further study of cell-specific conditional knockout mice.
Although our experiments in AV explants and in zebrafish using 2APB support our hypothesis led by our experiments using IP3R1−/−-IP3R2−/− mice, there is a limitation for using 2APB. Because 2APB is non-selective inhibitor, it is a potent CRAC channel inhibitor in addition to having an inhibitory effects on not only IP3R1 and IP3R2, but also other subtype of IP3R . Unfortunately, there is no selective IP3R inhibitor presently available. The observed defect in AV explants and in zebrafish with 2APB should include the possible effects of 2APB on CRAC channel activity, although AV explants from IP3R1−/−-IP3R2−/− mice showed similar defect. There also is a limitation to determine the cause of death in the IP3R1−/−-IP3R2−/− mice in this study. The placental defect, in conjunction with the cardiac defect, are likely to account for embryonic lethality in these mice, however, we could neither correctly estimate impact of each defect nor rule out defects of other organs. Further strategy for inactivation of the specific IP3R in specific tissues and organs would be required to overcome these limitations appeared in this study.
The DSCR1/MCIP1 gene encoding a calcineurin inhibitor is located on human chromosome 21 and a reduction of NFATc activity is associated with a 1.5-fold increase in gene dosage of DSCR1 and many of the features of Down (Trisomy 21) syndrome, including AV canal defects . The balance between Ca2+-sensitive positive (e.g., IP3Rs) and negative (e.g., DSCR1) regulation of NFATc may be crucial for organogenesis. Understanding the novel roles of IP3R-mediated Ca2+ signaling opens new avenues of research regarding the molecular embryology of the heart, which may lead to regenerative interventions for patients with congenital heart defects.
IP3R1/2 double-knockout mice were generated by intercrossing IP3R1+/−-IP3R 2−/− paired mice , . Genomic DNA samples prepared from tail biopsies or yolk sacs were subjected to PCR for genotyping, as described previously , . Zebrafish (Danio rerio) embryos were obtained from the natural spawning of a wild-type ABK line. CsA (50 µg/mL) or 2APB (5 µM) was added between 24 hpf and 33 hpf, and the zebrafish were harvested at 72 hpf for analysis . All experimental procedures and protocols were approved by the animal care and use committees of Keio University (the approval number 09122-(3)) and conformed to the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals.
Total heart extracts (10 µg) from E9.5 and E12.5 mice were subjected to 5% SDS-PAGE and analyzed by western blotting with anti-IP3R1 (18A10)  and anti-IP3R2 (KM1083) antibodies . For quantitative RT-PCR analysis of IP3R1 and IP3R2 expression, cDNAs were synthesized using the High Capacity cDNA Archive kit (Applied Biosystems) from total RNA samples extracted from embryonic hearts or placentas using the RNeasy kit (Qiagen). These cDNAs were used as templates in a TaqMan Real-Time PCR with the ABI 7500 Real-Time PCR system (Applied Biosystems). The data are normalized to the levels obtained for GAPDH. The TaqMan probes used for IP3R1 and IP3R2 were Mm00439917_m1 and Mm00444937_m1 (Applied Biosystems), respectively. The whole-mount RNA in situ hybridization and immunohistochemistry were performed as previously described , . The antibodies used for immunohistochemistry on the sections were: mouse anti-NFATc1 (7A6) monoclonal antibody (1100; Santa Cruz Biotechnology), anti-phospho-histone H3 (Ser10) antibody (1100; Upstate), anti-BrdU antibody (175; Becton Dickinson), anti-NFATc4 (H-74) antibody (1250; Santa Cruz Biotechnology) and anti-α smooth muscle actin (1A4) antibody (1400; Sigma).
The EMT assay was performed as reported previously . Endocardial cushions from the atrioventricular canal were explanted onto rat-tail collagen gel (BD Biosciences). After 24 h of incubation with 2APB, which is a non-selective inhibitor of IP3Rs , or with DMSO as a control, the total number of mesenchymal cells in each dish was counted. AV canal explants were infected using inactivated adenovirus harboring the gene encoding cytomegalovirus-driven constitutively active calcineurin  (Adeno-X ViraTrak Expression System 2, Clontech). After overnight incubation to allow attachment, each explant was incubated with 6.4×104 viral particles in culture medium. Whereupon, the AV explants were cultured for an additional 5 days and then assessed as indicated.
All data are expressed as mean ± s.d. (n≥3). Statistical analysis was performed using the Student's t-test. The results shown are representative of more than three independent experiments.
(0.03 MB DOC)
Genotype Distributions of Embryos from IP3R1+/−-IP3R2−/− Intercrosses.
(0.03 MB DOC)
The ultrastructures of the subcellular organelles of the IP3R1−/−-IP3R2−/− mice are comparable with those of the IP3R1+/−-IP3R2−/− mice. Scale bars, 1 µm. g, golgi; m, mitochondrion; mf, myofilament; n, nucleus; rer, rough endoplasmic reticulum; sr, sarcoplasmic reticulum.
(1.74 MB TIF)
Expression of site-specific markers in embryonic hearts. Whole-mount in situ hybridization images of the left- or right-side of the hearts of E9.5 IP3R1+/−-IP3R2−/- (upper panels) and IP3R1−/−-IP3R2−/− (lower panels) embryos. The expression patterns of Nkx2.5, MLC2a, and MLC2v (earliest markers of the embryonic heart), and of Hand2, Hand1 and Hrt2 (markers of the right, left, and both ventricles, respectively) are shown. Scale bars, 0.2 mm. a, atrium; lv, left ventricle; oft, outflow tract; pa, pharyngeal arch; rv, right ventricle.
(5.26 MB TIF)
Cross-sections of E9.5 placentas from IP3R1−/−-IP3R2+/− (upper panels), IP3R1+/−-IP3R2−/− (middle panels) and IP3R1−/−-IP3R2−/− (lower panels) mutant mice. The widths of the labyrinth area indicated in parentheses. Higher-magnification images of the boxed areas are shown in the right panels. Scale bars, 0.5 mm. The graph shows quantification of the labyrinth areas of the IP3R1+/−-IP3R2−/− and IP3R1−/−-IP3R2−/− placentas at E9.5. The area of the IP3R1−/−-IP3R2−/− labyrinth is significantly lower (*P<0.05, n=3). Error bars indicate standard deviations. ev, embryonic vessel; mv, maternal vessel.
(5.68 MB TIF)
The percentages of cells with nuclear translocation of NFATc1 in the IP3R1−/−-IP3R2−/− hearts (white bars) are significantly lower than those in the IP3R1+/−-IP3R2−/− hearts (black bars) at E9.5 to E10.0 (*P<0.01, n=3). Error bars indicate standard deviations.
(2.97 MB TIF)
Atrioventricular regurgitation in the zebrafish hearts treated with DMSO (top), calcineurin inhibitor, cyclosporine A, (middle) and IP3R inhibitor, 2APB, (bottom).
(0.32 MB MOV)
We are grateful to D. Srivastava for providing probes and for scientific discussion; J. Chen for comments; H. Kokubo for Tbx2 probe; T. Nagai for TEM; K. Matsuo for construction of adenoviral expression vector of constitutively active calcineurin; M. Hirose for technical assistance with Real-Time PCR, western blotting and immunohistochemistry; K. Kimura for technical assistance with pathology; K. Shimura, E. Tsutsui and M. Mitani for their assistance.
Competing Interests: The authors have declared that no competing interests exist.
Funding: This work was supported by a Grant-in-Aid for Scientific Research (H.Y.) and a Grant-in-Aid for Young Scientists (K.U.) from the Ministry of Education, Culture, Sports, Science and Technology, Japan, the Pfizer Fund for Development Research (H.Y.), and a Research Fund from the Japan Automobile Association (H.Y.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.