In order to determine the molecular weight cut-off (MWCO) for the atrial wall filtration into kidneys of the Mytilus galloprovincialis, we employed five magnetic resonance (MR) tracers: manganese chloride (Mn2+), gadolinium chloride (Gd3+), manganese-ethylenediaminetetraacetic acid (MnEDTA), gadolinium-diethylenetriamine pentaacetic acid (GdDTPA) and oligomer-based contrast agent (CH3-DTPA-Gd). After injection of the MR tracers (1 or 2 mmol l−1×0.1 ml) into the visceral mass, T1-weighted MR imaging (T1w-MRI) and the longitudinal relaxation rates (1/T1=R1) were measured at 20°C. The MR tracers were distributed uniformly in the visceral mass within 1 h after injection. The T1w-MRI intensity and R1 of the kidney (R1K) were increased by Mn2+ and MnEDTA, with urine concentrations estimated at 210 and 65 µmol l−1, respectively. The rest of the tracers showed only minimal or no increase. When the mussels were additionally incubated in seawater with 10 µmol l−1 MnCl2, R1K was increased in the GdDTPA group, but not in the GdCl3 group. Therefore, Gd3+ might have inhibited renal accumulation of Mn2+ and Gd3+. Incubation in seawater with 10 µmol l−1 MnEDTA showed no increase in the R1K, but additional incubation with 10 µmol l−1 MnCl2 caused an increase in R1K. It is suggested that injected MnEDTA was filtrated as MnEDTA per se, and not likely separated into free Mn2+. Thus, we concluded that the MWCO of the atrial wall of the M. galloprovincialis is around 0.5 kDa, which is almost 1/100 of that for vertebrate animals, and suggests a reduction in efforts to reabsorb metabolites and osmolytes from the urine.
The excretion system of mussels consists of kidneys and the pericardium (Martin and Harrison, 1966; Bayne, 1976). As shown in Fig. 1, hemolymph is filtered into the pericardial cavity by the atrial wall or the pericardial gland in the heart. Then, the filtrate flows to the nephridia sac (kidney), and secretion, absorption and storage might be attributed to the kidneys (Martin and Harrison, 1966; Bayne, 1976). Morphological evidence of filtration was obtained by previous electron microscopic studies on the atrial wall of the Mytilus heart (Pirie and George, 1979; Andrews and Jennings, 1993). Physiological experiments of bivalves were initially conducted by Picken (1937). Body fluid directly collected from the pericardial cavity of Anodonta cygna by needle was approximately isotonic (Picken, 1937). According to the constant-volume hypothesis (Ramsay, 1952; Krijgsman and Divaris, 1955; Seo et al., 2014a), it was expected that there would be filtration pressure, and this was confirmed by the atrial and pericardial pressure in Anodonta anatine (Brand, 1972) and also by the flow direction in the renopericardial canal of Mytilus galloprovincialis (Seo et al., 2014a). Therefore, it can be considered that filtration occurs via the atrial wall.
oligomer-based contrast agent
gadolinium-diethylenetriamine pentaacetic acid
relaxivity value of MR tracers
stability constant of metal complex
magnetic resonance imaging
molecular weight cut-off
p-amino hippuric acid
longitudinal relaxation rate (1/T1)
longitudinal relaxation time
T1-weighted gradient-echo magnetic resonance imaging
The characteristics of the filtration wall of bivalves have not been well analyzed to date. As far as we found in a search of investigative reports, the excretion system of the abalone Haliotis rufescens could filtrate p-amino hippuric acid (PAH, 0.19 kDa), phenolsuphonphthalein (PSP, 0.35 kDa) and inulin (5.5 kDa) (Harrison, 1962). Filtration of inulin was also reported in the octopus and the African snail (Martin and Harrison, 1966). Inulin is the common substance for measurement of the glomerular filtration rate in vertebrates (Guyton and Hall, 2006). However, the evidence available is not convincing in regard to whether the atrial wall of Mytilus filtrates inulin, because of the following reasons: amino acids, such as taurine and betaine, are important osmolytes for the hemolymph in the Mytilus edulis (Bayne, 1976), and the cell release/uptake of osmolytes varies when seawater shows lower or higher salinity (Burton, 1983; Hoyaux et al., 1976). If the mussel filters out large substances such as inulin (5 kDa), (1) the mussel may lose a lot of metabolites and osmolytes from the hemolymph, and (2) the kidney has to reabsorb a lot of metabolites and osmolytes from the urine. However, the structure of the canal between the pericardial cavity and the nephridia is different from the tubular formation seen in the vertebral kidney, being comparably short and not convoluted. It is not clear whether the canal actively reabsorbs osmolyte to maintain osmolarity of body fluid. Considering that the nephridia or kidney of the mussel has a luminal structure that is drained by the auricle filtrates via the canal, the physiological function of the atrial barrier for the substrate should be different from that seen in vertebrates. Therefore, we were motivated to determine the molecular size barrier of the atrial wall. In human renal tubules, small molecules such as glucose and oligopeptides (up to 4 aa) are transported by the specific transporters. Larger molecules such as polypeptides and immunoglobulins are reabsorbed by endocytosis, which is slower than the oligopeptide transport, and much more energy is required in the process for the kidney epithelia (Boron and Boulpaep, 2004). Therefore, we assumed that the auricle wall could not filter molecules larger than 0.5 kDa.
MATERIALS AND METHODS
The Mytilus galloprovincialis Lamarck 1819 used in this study were collected from a subtidal zone of the shore of Hisanohama, Fukushima, Japan, on 27 February 2019. At the laboratory, in three separate 5 liter baths, 15–20 mussels were kept in each bath in aerated synthetic seawater (salinity 36‰) at room temperature (20–24°C) (Seo et al., 2014b). Salinity of the seawater was measured by a refractometer (Master-S28α, Atago, Tokyo, Japan), and half of the seawater was replaced every week. A total of 47 mussels were used in this MRI study. The length, height and width of the mussels was 35.7±0.4, 19.4±0.2 and 14.1±0.2 mm (means±s.e.m.), respectively. All of the animal experiments in this study were carried out under the rules and regulations of the ‘Guiding Principles for the Care and Use of Animals’ set by the Physiological Society of Japan, and approved by the Animal Research Councils at Dokkyo University School of Medicine.
MR tracers and measurement of relaxivity
Five kinds of MR tracers were used for the tracer study as follows: Mn2+ as manganese chloride (nacali, Kyoto, Japan); Gd3+ as gadolinium chloride (Wako, Osaka, Japan); MnEDTA, made from MnCl2 and Na2EDTA (DOJINDO, Kumamoto, Japan); GdDTPA (Magnevist, Schering, Berlin, Germany); and CH3-DTPA-Gd (NMS-60, Nihon Medi-Physics, Chiba, Japan). These tracers were dissolved in 36‰ NaCl solution, and the R1 values of the solutions (0–1 mmol l−1) were measured by inversion recovery pulse sequences with a 20-mm 1H coil at 20°C. The relaxivity of the tracers was calculated from the slope of a linear regression line using Excel 2016 (Microsoft, Redmond, WA, USA).
Magnetic resonance imaging
The MRI examination of the M. galloprovincialis in this study used procedures noted in our previous reports (Seo et al., 2014a, 2016; Wakashin et al., 2018). In brief, the mussels were placed in a plastic tube (inner diameter of 22.5 mm), and were immersed in 12 ml of synthetic seawater without aeration, and the temperature was kept at 20°C. Seawater was exchangeable through another tube set in the bottom of the tube holding the mussel. The 1H MR images were obtained by a 7 T vertical magnetic resonance imaging system (AVANCE III, Bruker Biospin, Ettlingen, Baden-Württemberg, Germany) with ParaVision operating software (version 5.1) and equipped with an active shielded gradient (micro2.5) and a 25-mm 1H birdcage radiofrequency coil.
The R1 relaxation rate was measured by a two-dimensional saturation–recovery imaging method with five relaxation delays from 0.1 to 4 s. The pixel size was 180×180 µm and the slice thickness was 1 mm. The total image acquisition time was 9 min 21 s. The R1 image was calculated using Image Sequence Analysis Tool in ParaVision. Images with a large motion artifacts were omitted from the following data analysis. In order to include the whole-body structure of the mussel, three-dimensional T1-weighted gradient-echo imaging (3D T1w-MRI) was used. The MR signal in the kidney was analysed by 3D T1w-MRI with a voxel size of 240×240×240 µm, and a combination of TR/TE/θ=50 ms/2.5 ms/22.5 deg, where TR, TE and θ are relaxation delay, echo-time and flip angle, respectively. The total image acquisition time was 7 min 40 s. Images were Fourier-transformed with a data matrix of 256×128×128 after zero-filling of data, and the final voxel size was 180×180×180 µm. The volume of the kidney was obtained using the 3D-MRI analysis package in ParaVison. The image intensity of the T1w-MR image was quantified compared with the image intensity of reference capillary containing 0.5 mmol l−1 MnCl2 solution. Statistical analysis of R1 was performed using unpaired or paired t-tests in Excel 2016. P-values less than 0.05 were regarded as significant.
Injection of MR tracers
Mussels were anesthetized by 4% MgCl2, 0.1 ml 36‰ NaCl solution containing 1 or 2 mmol l−1 MR tracers were injected into the visceral mass near the root of the foot by a 30 G needle (Nipro, Osaka, Japan). Injection of the MR tracers was confirmed by noting a higher signal intensity of the visceral mass in scout T1w-MR images.
Protocols of MR experiments
The protocols of the three types of MR experiments are summarized in Fig. 2. Experiments on the accumulation of the MR tracers in the kidney (Fig. 2A) started from the injection of the MR tracers. Then, the mussels were incubated in the control seawater. The R1 relaxation rate was measured at 1, 2 and 3 h after the injection, and 3D T1w-MR images were measured between the R1 measurements. When the atrial wall could filter the MR tracer, the T1w-MRI intensity and the R1 of the kidneys increased, and vice versa (Fig. 1).
Experiments on the effects of GdDTPA (Fig. 2B) or Gd3+ (Fig. 2C) on the renal function also started from the injection of GdDTPA or Gd3+, followed by three or one set of measurements of the 3D T1w-MR image and the R1 relaxation rate, respectively. Then, the seawater was replaced by seawater containing 10 µmol l−1 MnCl2, and 3D T1w-MR images and R1 relaxation rates were observed at 1 h after incubation. If GdDTPA or Gd3+ inhibited the renal function, we could not detect any increase in the T1w-MRI intensity or R1 even with the additional incubation in seawater containing Mn2+ (Fig. 1).
The experiments on the stability of the MnEDTA and the toxicity of Gd3+ (Fig. 2D) were started from the incubation of the mussels in seawater containing 10 µmol l−1 of MnEDTA or GdCl3 for 1 h. Then, seawater was replaced by seawater containing 10 µmol l−1 of MnCl2. The R1 relaxation rate was measured before and 1 h after the Mn2+ incubation. If the Gd3+ inhibited Mn2+ uptake from the seawater (supposed by digestive and/or gill epithelia) (Fig. 1), we could not detect any increase in the T1w-MRI intensity or R1 with the additional incubation in the seawater containing Mn2+. When the MnEDTA released free Mn2+ in the seawater, we could detect an increase in the T1w-MRI intensity and R1.
In order to minimize the length of the MR experiments, the R1 for the control condition was taken as the mean R1 value obtained in separate experiments.
Relaxivity of MR tracers
First, we examined the concentration dependency of the MR tracers on the R1 relaxation rate. The results are shown in Fig. 3. As mentioned in the Introduction, the R1 value depends on the concentration of the MR tracer; the relaxivity values calculated from the slope of the linear regression line are shown in Table 1.
Distribution and accumulation of the MR tracers in the kidney
The injection of the MR tracers was confirmed using scout T1w-MR images. When MR tracers were injected into the visceral mass near the root of the foot, the T1w-MR image intensity of the visceral increased owing to the increase in R1 caused by the MR tracers. Therefore, the image intensity of the foot was depicted as a higher signal intensity at 10 min after the injection of GdDTPA (Fig. 4A,B). Then, the image intensity of the foot decreased, dependent on the time course (Fig. 4C). Usually, the image intensity reached an almost uniform intensity until 1 h after the injection (Fig. 4D). As shown in Fig. 4D, the T1w-MR image intensity of the visceral mass and also the mantle were almost the same degree, representing a uniform distribution of GdDTPA in the hemolymph of the mussel. As shown in Fig. 4H, in a few cases, the MR tracers remained at a high concentration in the tip of foot, suggesting that the dynamics of the tracers could be affected by a local factor.
In order to indicate the position of the kidneys, pairs of longitudinal slices at the kidney and a transverse slice at 2 mm posterior to the atrioventricular valve of a mussel fixed in paraformaldehyde are shown in Fig. 4E (Wakashin et al., 2018). When Mn2+ was injected into the mussel, only the lumen inside of the kidneys showed a higher intensity, which represents the accumulation of Mn2+ in the urine (Fig. 4F). Similar results were also obtained after the injection of MnEDTA (Fig. 4G). The mean±s.e.m. volume of kidney was 9.2±0.49 µl (n=23), as estimated from a 3D reconstructed image of the kidney.
GdDTPA and CH3-DTPA-Gd were also distributed equally in the interstitial space (Fig. 4H,I). However, the T1w-MR image intensity of the kidney was not enhanced, and remained at almost the same intensity as that of the seawater. These low-intensity regions could be enhanced by additional incubation with 10 µmol l−1 MnCl2 (Fig. 4J). Therefore, we confirmed that these regions were kidneys, and that GdDTPA and CH3-DTPA-Gd are not accumulated in the urine. The injected Gd3+ showed a different distribution compared with that of GdDTPA (Fig. 4K). The T1w-MR image intensity of the borders of the kidney, intestine, stomach and gills were enhanced. Urine in the kidney was not enhanced, even after 1 h of additional incubation with 10 µmol l−1 MnCl2.
Changes in the R1 relaxation rate of the kidney and foot
We also examined time-dependent changes of R1 in the mussel, and the results were summarized at 1, 2 and 3 h after the injection of the MR tracers (Fig. 5). The proximal side of the foot (Fig. 4E) was used for the typical point of the tracing in the visceral mass. In the kidney, the injection of MnCl2 and MnEDTA caused an increase in R1 compared with the control (P<0.01), which remained at the same level for 3 h (P>0.05). The injection of CH3-DTPA-Gd and GdDTPA showed no or a minimal increase in R1 for 3 h compared with the control (P>0.05). In the foot, R1 increased significantly after the injection of MnCl2 and CH3-DTPA-Gd (P<0.01 and 0.05, respectively).
The means±s.e.m. of R1 at 1 h after the injection of the MR tracers are summarized in Fig. 6A. The concentrations of the MR tracers were estimated by C=(R1−Rc)/K, where Rc is the R1 of the control and K is the relaxivity value of the MR tracer. As shown in Fig. 6B, the kidney concentrations of Mn2+ and MnEDTA were higher than those of the foot (P<0.01), and were estimated as 213±22 and 65±8 µmol l−1 (mean±s.e.m., n=14), respectively. There were no differences in the concentration of GdDTPA between the kidney and the foot (P>0.05). The concentrations of CH3-DTPA-Gd and Gd3+ in the kidney were lower than those in the foot (P<0.01). Therefore, GdDTPA, CH3-DTPA-Gd and Gd3+ were not concentrated in the kidney, even though the interstitial concentration of the MR tracers was approximately 50 µmol l−1.
The dose dependency of the CH3-DTPA-Gd injection is shown in Fig. 7. There was a linear relationship with a slope of 0.236±0.032 (l (s mmol)−1) (coefficient±s.e.m.). The slope was one-twentieth of the relaxivity of the CH3-DTPA-Gd [4.8 (l (s mmol)−1)]. Therefore, the injected tracers (0.1 ml) were diluted 20-fold in the mussel. Because CH3-DTPA-Gd could not enter cells and was not filtered into the urine, the total volume of the hemolymph could be estimated as 2.0 ml (1.6–2.8 ml 95% confidence limits).
Effects of the MR tracers on renal function
In order to check the effect of the interstitial MR tracers (50 µmol l−1) on the function of the kidney, seawater was replaced by fresh seawater containing 10 µmol l−1 of MnCl2 3 h after the injection of the GdDTPA. As shown in Fig. 8A, the R1 of the kidney did not increase with the injection of 1 mmol l−1 GdDTPA as expected, but it was increased by the additional incubation with Mn2+ (P<0.05). Therefore, GdDTPA does not affect the accumulation of Mn2+. On the contrary, when 1 mmol l−1 GdCl3 was injected, no elevation of the renal R1 was observed, and importantly, the expected enhancement of renal R1 owing to the additional incubation with Mn2+ (P>0.05) was not reproduced. In addition, as shown in Fig. 8B, the R1 of the kidney was not increased by incubation in seawater containing 10 µmol l−1 of GdCl3, nor was it increased by the additional incubation in 10 µmol l−1 of Mn2+ (P>0.05). These data indicated that Gd3+ functionally inhibited Mn2+ uptake from seawater and the concentration in the kidney (Fig. 1).
The stability constant of the metal complex is represented by pKM=log ([ML]/[M] [L]), where [ML], [M] and [L] are the concentrations of chelate, free metal and free ligand, respectively. The pKM of GdDTPA (22.5) is much higher than the pKM of CaDTPA (10.7) and MgDTPA (9.3) (https://www.dojindo.eu.com/Images/Product Photo/Chelate_Table_of_Stability_Constants.pdf). Thus, the binding competition of Ca2+ and Mg2+ in the hemolymph or seawater should be negligible. Indeed, in a separate experiment, the T1w-MR image intensity of the kidney epithelia was not increased by incubation in seawater containing 1 mmol l−1 of GdDTPA for 24 h. However, the KM of MnEDTA (14.0) is higher, but closer to the KM of CaEDTA (11.0) and MgEDTA (8.7). It may be the case that MnEDTA releases free Mn2+ ion owing to binding competition (Seo et al., 2013). In order to test the stability of MnEDTA, mussels were incubated in seawater containing 10 µmol l−1 of MnEDTA for 1 h. There was no increase in R1 in the kidney owing to the MnEDTA incubation, but R1 was increased by the additional incubation with Mn2+ (P<0.05) (Fig. 8B). Therefore, MnEDTA is stable even at the concentration of 10 µmol l−1. It is also true that MnEDTA does not inhibit Mn2+ uptake and concentration in the kidney.
Distribution of injected MR tracers
Tracer injection studies on the kidney started at the end of the 19th century (Martin and Harrison, 1966), and the accumulation of indigo sulfonate (0.34 Da) was reported in the kidney of bivalves (Martin, 1983). Then, PAH (0.19 Da), PSP (0.35 Da) and inulin (5.5 kDa) appeared in the urine of Haliotis rufescens (Harrison, 1962). However, little knowledge was obtained owing to the difficulty in obtaining consecutive samples of urine and blood from healthy, unrestrained bivalves (Martin, 1983). In the present study, the concentration of MR tracers was estimated using the R1 value, which precluded the need to insert a catheter in the mussel and to extract hemolymph or urine from the mussel. Except for the injection of MR tracers (5% of the volume of the hemolymph), the circulation of the hemolymph was not impeded. Therefore, this MRI technique is advantageous compared with conventional invasive techniques. In regard to the results, we conclude that smaller MR tracers (Mn2+, MnEDTA) appeared in the urine in the kidney, while the larger MR tracers (CH3-DTPA-Gd and GdDTPA) stayed in the hemolymph.
In this study, we measured the whole body of the mussel using 3D T1w-MRI. In the beginning, the injected part of the visceral mass was depicted at a higher image intensity (Fig. 4A). Then, from 1 to 3 h after the injection of the MR tracers, the image intensity had reached almost the same value in the visceral tissues, except for the kidney, digestive organs and the tip of the foot in some cases (Fig. 4F–I). Indeed, the R1 values presented the same level for 3 h (P>0.05) (Fig. 5). Therefore, MR tracers were distributed fairly uniformly in the whole body, and were also stable for a time range of 1–3 h after the injection. We estimated the volume of the hemolymph (2.0 ml) using CH3-DTPA-Gd (Fig. 7). From the length of the shell, the wet mass of the mussel without the shell was estimated as 3.93 g (Hosomi, 1985), and the percentage of the hemolymph volume was estimated as 51% of the wet mass. This value is in good agreement with the previously reported values, 50.8±7.6% (mean±s.d.), of M. californianus measured by inulin (Martin et al., 1958). Thus, CH3-DTPA-Gd might be distributed in the same space as the inulin.
Molecular size dependency for filtration at the atrial wall
CH3-DTPA-Gd (2.1 kDa) and GdDTPA (0.55 kDa) were not accumulated in the urine of the kidney, but Mn2+ (0.055 kDa) and MnEDTA (0.35 kDa) were accumulated in that urine. Because there can be uptake of Mn2+ from the seawater, epithelial cells (probably in the digestive organ or gills) can transport Mn2+ (Fig. 1). Therefore, Mn2+ was filtrated at the atrial wall and it also might be secreted from the renal epithelial cells. As mentioned in the Introduction, Ca2+, Mn2+ and Gd3+ share a close chemical similarity. Therefore, Mn2+ could be transported by the Ca2+ channel and Gd3+ could inhibit Ca2+-dependent processes. If kidney epithelial cells uptake and secrete Mn2+, there should be Gd3+ uptake in the epithelial cells, which may inhibit the epithelial functions. Indeed, Gd3+ was accumulated in the kidney epithelia (Fig. 4K), but was not concentrated in the urine (Fig. 6B). These results support that Gd3+ inhibits the epithelial functions. In contrast, there was no uptake of MnEDTA and GdDTPA from the seawater in the kidney. Therefore, epithelial cells, not only in the digestive organ but also in the kidney, could not transport MnEDTA and GdDTPA. This concept is supported by the lower MnEDTA concentration in the urine, compared with the concentration of Mn2+ (Fig. 6B). Thus, MnEDTA must be filtered at the atrial wall, but GdDTPA was not. These results are in good agreement with the results of previous reports: indigo sulfonate (0.34 kDa) appeared in the urine of the kidney of bivalves such as Anodonta, Cardium and Tellina (Martin, 1983), and injected PAH (0.19 kDa) and PSP (0.35 kDa) appeared in the urine of H. rufescens (Harrison, 1962). Therefore, molecules smaller than 0.35 kDa can be filtered by the atrial wall. Harrison (1962) reported that the concentration of urinary inulin was the same as the plasma in a wide range of the plasma inulin concentration (20-500 mg l−1). This is not the case for GdDTPA or CH3-DTPA-Gd. The R1 of the urine in the kidney remained the same as the R1 in the control seawater (Fig. 6A). It is also true that MnEDTA in the urine is higher than that in the hemolymph (Fig. 6B). Therefore, it is not likely that inulin (5.5 kDa) is filtered by the atrial wall in Mytilus. Alternatively, H. rufescens could filtrate larger molecules, and it is possible that Harrison's (1962) methods might have lacked sufficient resolution or his experiments might have been prone to fluid contamination. We concluded that the MWCO for filtration by the atrial wall of Mytilus is around 0.5 kDa, which is almost 1/100 (Guyton and Hall, 2006) of the MWCO for filtration by the glomerulus of the kidney in vertebrates, such as humans. The low MWCO may promote the maintenance of metabolites and osmolytes in the hemolymph, and save energy to reabsorb useful substances, such as glucose, as reported for Achatina fulica (Martin and Harrison, 1966). In vertebrates, it is considered that a protein named nephrin controls the MWCO for filtration by the glomerulus (Kestilä et al., 1998), which forms the slit diaphragm. We are not sure whether nephrin can cover the wide range of the MWCO. Indeed, the MWCO at around 0.5 kDa is somewhat similar to that controlled by the tight junction (Nitta et al., 2003). Therefore, in a future study, we plan to investigate the molecular-barrier mechanism in the atrial wall of Mytilus in which nephrin and claudin might be involved, and determine the protein responsible for the low MWCO.
In summary, we (1) injected five MR tracers in M. galloprovincialis, and detected the accumulation of the MR tracers in the urine of the kidney by MRI; (2) MR tracers smaller than 0.35 kDa appeared in the kidney, but MR tracers larger than 0.5 kDa did not. Thus, the MWCO is around 0.5 kDa, which is almost 1/100 of that of vertebrate animals, suggesting a reduction in efforts to reabsorb metabolites and osmolytes from the urine.
We offer our sincere thanks to Dr A. Kinjo-Kakumura and Prof. K. Inoue (AORI, UT) for providing the mussels. We also express our thanks to Drs D. Gross, V. Lehman and T. Oerther (Bruker Biospin), as well as Ms. Y. Imaizumi-Ohashi, Ms. M. Yokoi-Hayakawa and Ms. Kazuyo Mashiyama (DSUM) for their technical assistance. We must also thank Prof. S. Kojima (AORI, UT) for his helpful suggestions and encouragement to E.S.
Conceptualization: H.W., E.S., Y.S.; Methodology: H.W., E.S., Y.S.; Investigation: H.W., E.S., Y.S.; Resources: Y.S.; Writing - original draft: H.W., E.S., Y.S.; Writing - review & editing: H.W., E.S., Y.S.
Parts of this study were supported by grants from the Japan Society for the Promotion of Science KAKENHI (JP24659102 and JP15K08185 to Y.S.).
The authors declare no competing or financial interests.