CCL8 is a potential molecular candidate for the diagnosis of graft-versus-host disease

Hematopoietic stem-cell transplantation (HSCT) can be curative for many hematologic, oncologic, metabolic, and immune disorders. However, despite 2 decades of great advances in posttransplantation immunosuppressive therapy, graft-versus-host disease (GVHD) remains a major life-threatening posttransplantation complication. Thirty to eighty percent of hematopoietic stem cell (HSC) transplant recipients develop GVHD, the risk depending on the degree of histoincompatibility between donor and recipient, the type of transplantation, the number of T cells in the graft, the underlying disease, and the immunosuppressive treatment.¹  Currently the diagnosis of GVHD is mainly clinical, based on skin rash, hyperbilirubinemia, and diarrhea. A simple noninvasive means of early and more precise diagnosis of GVHD would facilitate its early treatment and improve the outcome of HSCT

Recent advances in proteomics have created new techniques to examine the global expression of proteins in biologic fluids and to identify novel biomarkers in disease or pathologic states. One such technique is surface enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS), a high-throughput and sensitive proteomic approach to segregate proteins from complex bodily fluids such as plasma and to generate comparative protein profiles. In SELDI, proteins from complex biologic samples bind selectively to chemically modified affinity surfaces on a ProteinChip (Ciphergen Biosystems, Fremont, CA), with nonspecifically bound impurities washed away. Captured proteins are then analyzed by TOF-MS, producing spectra of the molecular mass (m/z) and relative concentration (intensity) of each protein. Recently this technology has been successfully applied to the diagnosis of cancer and other diseases.^2-5 

Previous reports have described proteomic analysis of human biofluids in GVHD.^6-8  However, in human clinical studies unavoidable experimental artifacts related to genetic and environmental factors may confound the discovery of novel biomarkers. This is especially true in patients undergoing HSCT, who have a wide variety of preexisting diseases, undergo diverse conditioning regimens and GVHD prophylaxis, and have other variables. Mouse models provide an invaluable experimental system for studying human GVHD pathogenesis. To reduce these confounding variables, we looked first for diagnostic biomarkers for GVHD in a mouse model. These experiments suggested several candidate proteins linked to GVHD. One peak with 8972 Da m/z clearly discriminated between GVHD mice and normal controls with increased expression in the disease. Furthermore, administration of cyclosporin A to overt GVHD mice decreased the plasma level of this peak. In syngeneic transplantation, the 8972-Da peak was barely detectable. Purification and mass spectrometric analysis identified this molecule as CCL8, a member of a large chemokine family derived from a lineage of macrophages and other cell types. In human studies, we examined the serum concentration of CCL8 in relation to GVHD severity. CCL8 proved to be highly up-regulated in GVHD serum and 2 patients with severe, eventually fatal, GVHD had extremely elevated serum levels of CCL8. Our data suggest that serum CCL8 might be a highly specific test for the early and accurate clinical diagnosis of GVHD.

Male C57BL/6(H-2^b) mice and female BALB/c(H-2^d) mice were obtained from Sankyo Labo Service (Tokyo, Japan) and were bred in the Institute of Animal Experiment at Sapporo Medical University. Mouse ages ranged between 7 and 12 weeks at the start of experiments. The Sapporo Medical University Animal Experimental Ethics Committee approved the studies.

All reagents were purchased from SIGMA/ALDRICH (Tokyo, Japan) unless otherwise indicated.

On the day of bone marrow transplantation (BMT), donor mice (C57BL/6 for allogeneic BMT; BALB/c for syngeneic BMT) were killed by cervical dislocation. Donor bone marrow cells were collected into Dulbecco modified Eagle medium with 2% fetal calf serum/1% penicillin-streptomycin by flushing the femurs and tibias, and a single-cell suspension was prepared. Cells were washed and resuspended with RPMI 1640 medium for intravenous injection through the caudal vein. BM inoculum consisted of 2 × 10⁷ BM cells for allogeneic and syngeneic BMT. Recipient BALB/c mice were fed with acidified water for at least 7 days before BMT to prevent sepsis after lethal irradiation. Recipient mice were given 8.5 Gy total body irradiation at a rate of 0.34 Gy/min and injected with donor BM cells within 3 hours.

Recipient mice were monitored daily for clinical manifestations of GVHD including weight loss, hunched posture, skin erythema, alopecia, and diarrhea.

Recipient mice were killed on day 7, day 14, day 21, or day 28 after transplantation. Skin, liver, and small intestine were removed, fixed in 10% buffered formalin, and stained with hematoxylin and eosin for microscopy. Histologic changes in representative organs considered compatible with GVHD were as follows: skin (mononuclear infiltration into the dermo-epidermal junction and injury to hair follicles or sebaceous glands); liver (periportal mononuclear infiltration and hepatocellular necrosis); and small intestine (crypt cell apoptosis and dilatation or flattening of the villi). Findings were scored and given an overall interpretation of positive (+), indefinite (±), or negative (−) for GVHD.^9,10  Representative histologic findings are shown in Figure S1 (available on the Blood website; see the Supplemental Materials link at the top of the online article).

Cyclosporin A (CsA; Novartis Pharma AG, Wilmington, DE) was diluted to 1.67 mg/mL with 0.9% NaCl. CsA 20 mg/kg daily was administered intraperitoneally from day 8 through day 13 after transplantation.

BALB/c mice received a transplant of 2 × 10⁷ syngeneic marrow cells. On day 7 after transplantation, mice were injected intraperitoneally with either 5 μg lipopolysaccharide (LPS, Escherichia coli) or 5 μg poly(I:C) (GE Healthcare BioSciences, Tokyo, Japan) plus 20 mg D-GalN.¹¹  Four hours after injection, blood was collected. Doses of LPS or poly(I:C) were determined by preliminary experiments.

Plasma was sampled before BMT on day 0, and after BMT on days 7, 14, 21, and 28. Blood was obtained from the tail vein of living mice using a capillary tube coated with heparin and then centrifuged at 2400g for 7 minutes within 30 minutes after bleeding. Aliquots of plasma were stored at −80°C until assayed.

We added 20 μL of a solution containing 9 M/L urea and 10 g/L CHAPS in 10 mM Tris-HCl, pH7.4, to 10 μL of each sample. The mixture was vortex-mixed at 4°C for 15 minutes and diluted 1 to 40 in Tris-HCl. Eight-spot immobilized metal-affinity capture arrays (IMAC-30) were activated with 50 mM/L CuSO_4. Diluted samples (50 μL) were applied to each spot on the protein chip array and incubated for 1 hour on a shaker. After washing with the same Tris-HCl followed by a quick water rinse, 0.5 μL saturated sinapinic acid (SPA) was applied twice to each spot and allowed to air-dry. Mass/charge (m/z) spectra of proteins bound to the chelated metal (through tryptophan, cysteine, histidine, or phosphorylated amino acids) were generated in a Ciphergen Protein Biology System II Time-of-Flight mass spectrometer (PBS II; Ciphergen Biosystems). Data were collected by averaging 65 laser shots with a laser intensity of 200 and a detector sensitivity of 8.

All spectra were compiled and we performed preliminary data analysis using Ciphergen ProteinChip Software 3.2.0 (Ciphergen Biosystems). The m/z values less than 2.0 kDa, corresponding to the signal from the SPA matrix, were omitted. Raw data were normalized and aligned by Biomarker Wizard (Ciphergen Biosystems). A classification tree was developed with Biomarker Pattern's Software (BPS; Ciphergen Biosystems) as previously described.^12-14  Briefly, classification trees split the data into 2 nodes, using one rule at a time in the form of a question. The splitting decisions in this case were based on the normalized intensity levels of peaks or clusters identified from the SELDI protein-expression profile. Each identified peak or cluster becomes a variable in the classification process. Splitting is continued until terminal nodes are reached, and further splitting gives no gain in data classification. Multiple classification trees were generated using this process, and the best performing tree was chosen.

The 3 most abundant plasma proteins (albumin, IgG, transferrin) in the pooled plasma were removed by immunodepletion chromatography (Multiple Affinity Removal Column MS-3, 4.6 mmID × 50 mm; Agilent, Wilmington, DE). Plasma (50 μL) was diluted 5-fold in Buffer-A (Agilent), then injected onto the immunodepletion column. The flow-through fractions were collected and further resolved by high-performance liquid chromatography (HPLC, PU-2089plus; JASCO Engineering, Tokyo, Japan). The separation column used in HPLC was an Inertsil Ph column (5 μm, 4.6 mmID × 150 mm; GL Sciences, Tokyo, Japan). The elution gradient profile was as follows: (1) elution solvent A: 2% ACN/0.1% TFA, solvent B: 80% ACN/ 0.1% TFA; (2) liner gradient: 0% to 100% for solvent B for 50 minutes; flow rate: 1.0 mL/min. Fractions were collected every 30 seconds and their composition was monitored by SELDI-TOF MS; 2 μL of each fraction was applied on an Au chip (Ciphergen Biosystems) and processed with SPA matrix as described above. HPLC fractions containing candidate markers were selected according to SELDI-TOF MS monitoring. The fractionated samples were lyophilized and dissolved in 200 μL solubilization buffer (7 M urea, 2 M thiourea, 50 mM DTT, 2% ampholine, 3% CHAPS, 1% Triton X100). After sonication, sample solutions were loaded onto an IPG gel strip (pH 3-11 NL, 11 cm long; GE Healthcare BioSciences, Piscataway, NJ), and the strip was rehydrated for 10 hours at 30 V. The first-dimensional separation by isoelectric focusing (IEF) was carried out using the IPGphor system (GE Healthcare BioSciences, Tokyo, Japan) at 20°C for a total of 12 kV/h. After IEF, the IPG strips were equilibrated for 15 minutes in 50 mM Tris-HCl (pH 8.8) containing 6 M urea, 2% sodium dodecyl sulfate (SDS), 30% glycerol, 0.002% bromophenol blue, and 1% dithiothreitol and then re-equilibrated for 15 minutes in the same buffer except 2.5% iodoacetamide replaced dithiothreitol. For the second-dimensional separation, SDS–polyacrylamide gel electrophoresis (PAGE) was performed on homemade 8% to 20% gradient polyacrylamide gels and electrophoresed with a constant current of 40 mA/gel. After 2-dimensional electrophoresis (2DE), proteins were visualized with a silver staining kit (EzStain; ATTO, Tokyo, Japan).

Gel spots were washed with 100% ACN and 100 mM NH₄HCO₃, vacuum-dried, and then incubated in 5 μL trypsin solution (12.5 ng/μL in 50 mM NH₄HCO₃, 5 mM CaCl₂) for 16 hours at 37°C. The resultant peptides were extracted once with 20 μL of 20 mM NH₄HCO₃ and 3 times with 20 μL of 5% formic acid/50% ACN. The collected extracts were vacuum-dried to approximately 40 μL and then analyzed by nanoflow HPLC-ESI-MS/MS. HPLC was performed using a DiNa system (KYA Technology, Tokyo, Japan), and tryptic digest samples were separated on a HiQsilC18W-3 column (75 μmID × 50 mm; KYA Technology). Elution solvent A was 0.1% formic acid, while solvent B was 70% ACN/0.1% formic acid. The gradient separation was 0% to 50% solvent B over 40 minutes at a flow rate of 200 nL/min. The separated peptides were characterized using a QSTAR XL Q-TOF mass spectrometer (Applied Biosystems, Foster City, CA). Data were acquired in information-dependent acquisition mode using Analyst Applied Biosystems, and only multiple charged ions were chosen for MS/MS. Each cycle was composed of 1-s MS and 2-s MS/MS. The acquired mass spectral data were automatically processed and searched against the Swiss Prot database using MASCOT software (Matrix Science, London, United Kingdom) and ProteinPilot software (Applied Biosystems). The search parameters were as follows: MS accuracy: 0.2 Da; MS/MS accuracy for the MASCOT search: 0.2 Da.

PS20 (preactivated surface) ProteinChip surface (Ciphergen Biosystems) is composed of epoxides that form stable covalent linkages with amino groups of biomolecules (such as Abs). Ab (0.1 μg) was added to each spot of a PS20 chip and incubated for 2 hours at room temperature. After blocking residual active sites with 5 μL 1 M ethanolamine (pH 8.0) for 30 minutes, the spots were washed 3 times with 0.5% Triton X100 in PBS and twice with PBS. Mouse plasma samples were diluted 1:75 and human sera 1:25 with PBS and incubated for 2 hours under gentle mixing at room temperature. Each spot was washed twice with 0.5% Triton X100 in PBS and twice with PBS. After a rapid wash with 5 mM HEPES, SPA matrix was added and MS analysis performed using a PBS II ProteinChip reader.

Fourteen patients underwent HSCT between April 1993 and December 2005 in the Department of Pediatrics, Sapporo Medical University Hospital. The median age was 8.9 years (range: 1-20 years, Table 3). The grading of GVHD was based on that used by Przepiorka et al¹⁵  Detailed clinical features of the patients in Figure 6 were as follows.

A 5-year-old boy with Fanconi anemia received a cord blood stem-cell transplant from an unrelated donor. He received pretransplantation conditioning composed of 3 Gy total body irradiation (TBI) with 200 mg/m² fludarabine (Flu) and 40 mg/kg cyclophosphamide (CY), and GVHD prophylaxis with CsA and mycofenolate mofetil. GVHD first developed on day 13 after HSCT with skin rash (stage 2, grade II GVHD), and on day 26 diarrhea increased (grade III GVHD). Methyl prednisolone (mPSL) was started on day 13 (Figure 6A).

A 10-year-old boy with chronic myelogenous leukemia received a bone marrow transplant from an unrelated donor. He received pretransplantation conditioning composed of 12 Gy TBI with 120 mg/kg CY, and GVHD prophylaxis with tacrolimus and methotrexate. GVHD developed on posttransplantation day 19 with skin rash (stage 2, grade I GVHD), and diarrhea increased on day 27 (stage 3, grade III GVHD). mPSL was started on day 19 (Figure 6B).

A 3-year-old girl with acute lymphoblastic leukemia received a bone marrow transplant from an HLA-matched sister. Her pretransplantation conditioning was composed of 12 Gy TBI and 120 mg/kg CY, and her GVHD prophylaxis was methotrexate. She had no evidence of GVHD throughout her course (Figure 6C).

The Sapporo Medical University Ethics Committee approved human sera studies. All volunteers and patients provided written permission and for young children we obtained written permission from their parents in accordance with the Declaration of Helsinki. Sera were aliquoted and stored at −80°C until assayed.

Seven independent samples from 7 patients without GVHD and 7 from 7 patients with GVHD (more than grade II) were assayed by ELISA in duplicate. After HSCT, sera were obtained at the time of clinical diagnosis of GVHD. The concentration of CCL8 was quantified using an ELISA kit for human CCL8.

Polyclonal antibodies against mouse or human CCL8 were purchased from Peprotech (London, United Kingdom). The human CCL8 ELISA kit was purchased from RayBiotech (Norcross, GA) and the mouse IL-6 kit from R&D Systems (Minneapolis, MN). These were used according to the manufacturer's protocol. Plates were read with a plate reader (Multiskan JX; Thermo Labsystems, Helsinki, Finland) at 450 nm.

The results are expressed as the mean plus or minus SE. Statistical significance analyses were performed by either paired or unpaired t test as indicated in the figure legends. A significant difference was set at P less than .05. The Bonferroni correction for multiple comparisons was applied. The results were representative data from a set of experiments.

Acute GVHD was induced in BALB/c mice by injection of BM grafts from C57BL/6. In these mice, clinical symptoms of acute GVHD, such as diarrhea and ruffled fur, became apparent within 7 days, and skin erythema and alopecia developed within 21 days after transplantation. For controls, an identical number of syngeneic grafts were transplanted. The recipient mice who received allogeneic marrow cells but not those given syngeneic cells had pathologic evidence of GVHD at all time points after transplantation (Table 1; Figure S1).

The peak intensity values of 169 differentially expressed peaks identified in all samples from 2.0 to 200 kDa mass ranges were used for further analysis. Classification trees were created using all 169 peaks, and on analysis a single peak with average mass of 8972 Da that distinguished GVHD plasma from normal plasma was identified (Table 2; Figure S2).

Figure 1A shows changes of intensity of the 8972-Da peak from an individual mouse at each time point. Compared with day 0 (before transplantation), we first detected an apparent increase of the 8972-Da peak on day 7 after BMT. The intensity of the 8972-Da peak was 4.6 at day 0, which increased to 47.3, 51.4, 25.2, and 55.9 on days 7, 14, 21, and 28, respectively. All increases in peak intensity of GVHD samples were significantly higher than those in syngeneic controls at posttransplantation sampling days 7, 14, 21, and 28 (Figure 1B). The increase on days 7, 14, 21, and 28 of allogeneic samples was also significantly higher than the day-0 samples.

Figure 2 shows the effects of CsA treatment on the 8972-Da peak. Its intensity was 5.8 on day 0 and it increased to 69.4 on day 7. CsA was administered from day 8 to day 13, and the 8972-Da peak decreased to 13.8, 15.7, and 14.3 on days 14, 21, and 28, respectively (Figure 2A). This decrease was statistically significant (Figure 2B).

Figure 3A shows the effect of LPS or poly(I:C) on the 8972-Da peak in syngeneic BMT. Neither LPS nor poly(I:C) increased the 8972-Da peak, although allogeneic marrow transplantation apparently increased it. LPS or poly(I:C) increased serum IL-6 concentration dramatically, although allogeneic marrow transplantation did not increase serum IL-6 at day 7 after transplantation (Figure 3B).

A fraction with an elution time of 31.0 to 31.5 minutes (approximately 38% acetonitrile) contained abundant 8972-Da protein in GVHD samples, but little in controls (Figure 4A,B). The fractions of GVHD and control samples were collected, concentrated, and subsequently loaded onto 2DE. By comparing the 2 gel images of GVHD and control samples, we identified a highly expressed spot with apparent molecular mass of between 6500 Da and 12 300 Da in the gel of GVHD samples (Figure 4C and arrow in 4E), suggesting a relationship between the 8972-Da marker and the 2DE candidate spot. After in-gel digestion, this spot was identified by nanoLC-MS/MS as a CCL8 precursor with MASCOT score 236 (data not shown). One of the MS/MS spectra leading to the identification of CCL8 is presented in Figure 5A,B. Eleven peptides were matched to the theoretic masses with 52% coverage of the amino acid sequence (data not shown). The predicted mass of CCL8 precursor is 11 017 Da and its predicted pI is 8.64. The CCL8 precursor contains a signal peptide of 19 amino acids followed by the mature CCL8 sequence of 78 amino acid residues. The predicted mass of mature CCL8 is 8972.65 with a predicted pI of 8.45, which is consistent with the data acquired by SELDI-TOF MS and 2DE. The identity of the 8972-Da marker as CCL8 was further verified by SELDI immunoassay using a specific rabbit antimouse CCL8 Ab on a PS20 ProteinChip. The assay showed that while CCL8 was captured and detected from GVHD plasma samples, very little was obtained from control samples (Figure 5C).

To examine the clinical significance of CCL8 in GVHD sera, we first studied the relative intensity of CCL8 in patients treated with HSCT with or without GVHD. Changes in serum CCL8 peak intensities during the course of HSCT are depicted in Figure 6. Patient A developed clinical GVHD on day 13 after HSCT and started mPSL the same day (Figure 6A). The peak profile of CCL8, which was very weak on day 0 and obviously increased on day 10, had declined by day 21 after mPSL treatment. The peak was again up-regulated by day 33. In patient B, the CCL8 peak was undetectable on day 0, increased on days 9 and 19, and decreased on days 21 and 30 (Figure 6B). Clinical GVHD was recognized on day 19 and mPSL was started. In marked contrast, patient C remained GVHD free throughout her course (Figure 6C). The CCL8 peak was very low in all samples from this patient. In Figure 6D, an example of a healthy control is shown. The CCL8 peak is barely detectable. These data suggest a strong relationship between CCL8 peak and GVHD status in humans.

Since the SELDI technique is semiquantitative, we next quantitated CCL8 concentration in patient's sera using ELISA. Patient profiles are summarized in Table 3. As shown in Figure 7, ELISA studies suggested that CCL8 concentration correlates with GVHD in HSCT patients. All non-GVHD plasma contained less than 48 pg/mL (mean ± SE: 22.5 ± 5.5 pg/mL, range: 12.6-48.0 pg/mL, n = 7). In contrast, CCL8 is highly up-regulated in GVHD plasma ranging from 52.0 to 333.6 pg/mL (mean ± SE: 165.0 ± 39.8 pg/mL, n = 7). Of note were 2 patients with severe, eventually fatal, GVHD who had dramatically high levels of CCL8 (333.6 and 290.4 pg/mL). In healthy controls, the mean concentration of CCL8 was 18.9 pg/mL (range: 0.00 to 32.6 pg/mL, n = 8).

HSCT can be curative for various hematologic, oncologic, metabolic, and immune disorders. However, GVHD remains a major life-threatening complication after transplantation. Early and sensitive detection of GVHD and the ability to monitor treatment efficacy should improve the outcome for patients after HSCT.

Recently, proteomics has emerged as a key technique to identify biomarkers and disease targets. To discover novel biomarkers for GVHD, we used SELDI-TOF MS, which is one of several high-throughput technologies able to generate peptide and protein profiles of complex biologic samples with high sensitivity, reproducibility, and simplicity. Using the SELDI technique, we first compared plasma protein profiles between control and GVHD mice. Since human samples have intrinsic person-to-person variability that may obscure the discovery of new biomarkers, we considered that the relative simplicity of an animal model of GVHD would give better analytic power including estimates of specificity and sensitivity. We selected an 8972-Da protein as a potential plasma biomarker for GVHD in mice and also determined its expression in syngeneic transplantation, and after CsA, LPS, or poly(I:C) treatment. The 8972-Da expression was not increased in syngeneic combination mice, suggesting that it is not induced by pretreatment regimens given before bone marrow transplantation including radiation. Decrease of the 8972-Da peak expression following CsA administration in GVHD mice was strongly correlated with GVHD activity. Furthermore, neither LPS nor poly(I:C) induced an 8972-Da peak. These results suggest the probable existence of a protein or proteins linked to the onset and severity of GVHD at least in mouse plasma. We identified this peak as CCL8, one of the CC chemokines.

As observed in the mice, we found elevated serum levels of CCL8 in humans with GVHD but not in non-GVHD patients who had undergone HSCT. These results suggest that CCL8 is a potential biomarker for human GVHD. Human CCL8 has 2 forms: one has 76 amino acids (1-76) with biologic activity, and the other has 71 amino acids (6-76) and no activity.¹⁶  We detected a single peak with 8920 Da (average mass). Since the biologic role of human CCL8 is not well characterized, it should be determined whether this peak is derived from 1-76 or 6-76 CCL8. Before concluding that CCL8 is a valid biomarker for GVHD, we must determine whether it differentiates GVHD from other complications with overlapping clinical features, such as veno-occulusive disease (VOD), or viral reactivation. In preliminary studies, we found that CCL8 was not elevated in 2 cases of VOD (data not shown). A previous study demonstrated that CCL8 levels were increased in patients with Gram-positive sepsis.¹⁷  Additional studies with larger populations of human patients after HSCT are currently under way and are essential to verify the role of CCL8 as a biomarker for human GVHD.

Chemokines are predominantly small molecules (8 to 12 kDa) that bind to G protein–coupled receptors (chemokine receptors) and function primarily in leukocyte migration.¹⁸  The chemokine superfamily currently comprises more than 50 members subclassified into 4 families based on the arrangement of their cysteine residues: CXC, C, CX₃C, and CC.¹⁹  CCL8 is also known as monocyte chemoattractant protein-2 (MCP-2). CCL2/MCP-1, CCL7/MCP-3, CCL8/MCP-2, and CCL13/MCP4 constitute a small subfamily within the CC chemokine group, which acts as a major chemoattractant for monocytes, activated T cells, and dendritic cells.^20,21  Chemokine and chemokine receptor interactions play a crucial role in donor T-cell migration in GVHD.^22-27  However none of the above studies addressed the role of CCL8. The only report connecting CCL8 to GVHD was by Sugerman et al who showed up-regulated gene expression of CCL8 in mouse cutaneous GVHD.²⁸ 

GVHD is initiated by allogeneic activation of donor T cells by host-derived antigen presenting cells (APCs).⁹  Dendritic cells (DCs) are professional APCs that activate naive T cells most efficiently, and CCL8 is one of the chemokines secreted by mature DCs. These findings are consistent with our demonstration that CCL8 expression is up-regulated during GVHD. Although the complex interactions of chemokines involved in the induction of GVHD remain poorly understood, CCL8 is thought to be pivotal. CCL8 as a biomarker may facilitate early diagnosis of GVHD and assessment of treatment response, since it seems to be engaged in the development and progression of GVHD.

We thank Dr Peter M. Olley for helpful discussion and English revision of our paper.

This study was supported by grants from the Ministry of Health, Labor and Welfare of Japan and from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Contribution: T.H., Y.N., H.S., and Y.M. performed research; T.S. performed statistical analysis; N.S., N.H., M.Y., K.I., and H.T.contributed vital new reagents or analytic tools; Y.K. designed research; T.H. and Y.K. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Yasuo Kokai, Department of Biomedical Engineering, Sapporo Medical University School of Medicine, S1W17, Chuo-ku, Sapporo, 060-8556, Japan; e-mail: kokai@sapmed.ac.jp.