Apoptotic cells suppress mast cell inflammatory responses via the CD300a immunoreceptor

Several receptors for PS are expressed on phagocytes and are involved in clearing apoptotic cells in physiological and pathological settings (Scott et al., 2001; Hanayama et al., 2002; Kobayashi et al., 2007; Miyanishi et al., 2007; Park et al., 2007, 2008). Because PS mediates an “eat-me” signal to phagocytes such as macrophages (Fadok et al., 1992; Savill et al., 2002), we generated CD300a-deficient (Cd300a^−/−) mice (Fig. 1). Cd300a^−/− mice lacked the expression of Cd300a mRNA and CD300a protein but showed normal expression of neighboring genes to Cd300a in the Cd300 family gene cluster on chromosome 11 (Fig. 1), indicating that Cd300a^−/− mice were correctly generated. We examined whether CD300a is involved in phagocytosis of apoptotic cells by using macrophages from Cd300a^−/− mice and apoptotic thymocytes from caspase-activated DNase (CAD)^−/− mice. CAD^−/− cells do not undergo apoptotic DNA fragmentation, but their DNA is degraded in phagocytes after they are engulfed (Nagata, 2005). WT or Cd300a^−/− peritoneal resident macrophages were cocultured with dexamethasone-induced apoptotic CAD^−/− thymocytes for 2 h, stained with TUNEL, and analyzed by laser-scanning confocal microscopy. The frequencies of WT and Cd300a^−/− peritoneal macrophages carrying a given number of apoptotic cells were comparable (Fig. 2 A). Peritoneal resident macrophages also express several receptors, including TIM-1, TIM-4, and α_Vβ₃ (Kobayashi et al., 2007; Miyanishi et al., 2007; Fig. 3 A), that directly or indirectly bind PS and are involved in engulfment of apoptotic cells (Kuchroo et al., 2008). To examine whether these PS receptors had functional redundancy that masked the CD300a phagocytic function, we established NIH 3T3 transfectants stably expressing either CD300a or TIM-4 (Fig. 2 B). We labeled apoptotic thymocytes with pHrodo succinimidyl ester, a pH-sensitive fluorescent dye, which can be detected when they are engulfed as a result of the increased post-phagocytic light emission (Miksa et al., 2009). The NIH 3T3 transfectants or parental NIH 3T3 cells were cocultured with pHrodo dye–labeled apoptotic thymocytes and analyzed by laser-scanning confocal microscopy. In contrast to the parental NIH 3T3 cells, both CD300a and TIM-4 transfectants bound to apoptotic cells (Fig. 2 C). However, although TIM-4 transfectant expressed a pHrodo signal, we could not detect any signals from CD300a transfectant (Fig. 2 C). Flow cytometry analyses also demonstrated that a pHrodo signal was detected only in TIM-4, but not CD300a, transfectant (Fig. 2 D). Together, these results indicated that CD300a does not mediate phagocytosis of apoptotic cells by macrophages.

CD300a is also expressed on mast cells. To address whether apoptotic cells affect mast cell activation via CD300a, we generated BM-derived mast cells (BMMCs) from Cd300a^−/− or WT mice. Unlike macrophages, mast cells do not express the PS receptors TIM-1, TIM-4, and stabilin-2, although they do express CD300a and α_V integrin (Fig. 3 A). We examined whether PS binding to CD300a has any effect on LPS-induced cytokine or chemokine production from BMMCs by proteome analyses. WT or Cd300a^−/− BMMCs were cocultured with apoptotic cells at a ratio of 1:0.1 in the presence of LPS for 4 h. Cd300a^−/− BMMCs produced greater amounts of several chemokines and cytokines, including MCP-1 and IL-13, compared with WT BMMCs (Fig. 3 B). We therefore performed quantitative analysis of MCP-1 and IL-13 in addition to a proinflammatory cytokine TNF by ELISA. Although we did not detect any cytokines or chemokines in the culture supernatants in the absence of stimuli, stimulation with 1 µg/ml LPS induced both WT and Cd300a^−/− BMMCs to produce TNF, IL-13, and MCP-1; however, Cd300a^−/− BMMCs produced them at significantly higher concentrations (Fig. 3 C). Flow cytometry analyses also detected significantly more of these cytokines and chemokines in the cytoplasm of Cd300a^−/− BMMCs than in the cytoplasm of WT BMMCs 4 h after stimulation with LPS (Fig. 3 D). However, in the presence of a variant of MFG-E8 containing a point mutation (D89E) in the RGD motif (D89E MFG-E8), which is able to bind PS but is not able to bind α_Vβ₃ integrin (Hanayama et al., 2002), the levels of these cytokines and chemokines became comparable in WT and Cd300a^−/− BMMCs (Fig. 3 C), suggesting that CD300a–PS interaction inhibits these cytokines' and chemokines' production from BMMCs.

CD300a contains the five tyrosines at residues 233, 258, 270, 299, and 313 in the cytoplasmic domain, and Y²⁵⁸ and Y²⁷⁰ consist of the consensus ITIM sequence ((I/V/L/S)xYxx(L/V)). We previously reported that Y²⁵⁸ and Y²⁷⁰ are responsible for phosphorylation of CD300a and recruitment of SHP-1 in a rat basophilic leukemia cell line (RBL-2H3) upon stimulation with pervanadate (Okoshi et al., 2005). We found that Y²⁵⁸ and/or Y²⁷⁰ were phosphorylated and involved in SHP-1 recruitment in RBL-2H3 transfectants expressing WT or mutated CD300a when these cells were cocultured with apoptotic cells in the presence of LPS (Fig. 3, E–G). We have recently reported that CD300a also recruited SHP-1 in BMMCs when BMMCs were cocultured with apoptotic cells in the presence of LPS (Nakahashi-Oda et al., 2012). To examine whether SHP-1 is involved in CD300a-mediated signaling, we knocked down Ptpn6 with a small interfering RNA (siRNA) to create SHP-1–deficient (Ptpn6-KD) WT and Cd300a^−/− BMMCs (Fig. 3 H). Whereas the control Cd300a^−/− BMMCs produced significantly more TNF than did WT BMMCs after stimulation with LPS for 4 h, there was no significant difference in TNF production between Ptpn6-KD WT and Ptpn6-KD Cd300a^−/− BMMCs (Fig. 3 I). Collectively, these results indicate that upon binding to PS, CD300a recruits SHP-1 and mediates an inhibitory signal in BMMCs, resulting in the suppression of cytokines and chemokines secretion.

We next examined whether CD300a also recruits SHP-1 and delivers an inhibitory signal in macrophages upon binding to PS. WT or Cd300a^−/− peritoneal resident macrophages were cocultured with apoptotic thymocytes at 1:0.1 ratios in the presence of LPS, demonstrating that CD300a also recruited SHP-1 in peritoneal resident macrophages (Fig. 3 J). Moreover, the production of TNF and IL-10 by Cd300a^−/− peritoneal macrophages was significantly greater than that by WT peritoneal macrophages (Fig. 3 K), indicating that PS binding of CD300a inhibits cytokine production by macrophages as well as mast cells.

TNF, IL-13, and MCP-1 produced by mast cells are chemoattractants for neutrophils and play an important role in bacterial clearance in a CLP peritonitis model in mice (Echtenacher et al., 1996; Malaviya et al., 1996; Baumhofer et al., 1998; Matsukawa et al., 1999; Marshall, 2004). Therefore, we hypothesized that CD300a affects immune regulation by mast cells at the site of peritonitis, where large numbers of cells undergo apoptosis in the peritoneal cavity (Hotchkiss and Nicholson, 2006). In fact, we observed that >10% of cells in the peritoneal cavity 4 h after CLP were apoptotic cells, as determined by staining with Annexin V (Fig. 4 A). To test the hypothesis, we subjected WT and Cd300a^−/− mice to CLP and observed that Cd300a^−/− mice survived significantly longer than did WT mice after CLP (Fig. 4 B).

To elucidate the mechanism responsible for the longer survival advantage of the Cd300a ^−/− mice in this model, we measured bacterial content in the peritoneal cavity 4 h after CLP. Cd300a^−/− mice had significantly fewer bacterial colony-forming units (CFUs) than did WT mice (Fig. 4 C). Because we did not observe any differences in the number and type of commensal bacteria between the two mouse genotypes (unpublished data), these results suggested that Cd300a^−/− mice had more efficient bacterial clearance. We therefore examined the number of phagocytes in the peritoneal cavity after CLP. Cd300a^−/− mice had significantly greater numbers of neutrophils than did WT mice 4 h after CLP (Fig. 4 D). However, macrophage numbers were comparable between the two genotypes (Fig. 4 D). Because Cd300a^−/− peritoneal resident macrophages did not differ in the phagocytic function of Escherichia coli from those from WT mice (Fig. 4 E), these results suggested that the increased numbers of neutrophils in Cd300a^−/− mice contributed to the enhanced bacterial clearance, which in turn led to the decreased mortality rates after CLP.

To verify the important role of neutrophils, but not macrophages, in the lower mortality rates in Cd300a^−/− mice, we generated WT and Cd300a^−/− mice lacking neutrophils as a result of injection with anti–Gr-1 monoclonal antibody or fewer macrophages as a result of injection with clodronate liposomes. Flow cytometric analyses demonstrated that the neutrophils or macrophages were efficiently depleted in these mice (Fig. 4 F). Depletion of neutrophils rendered the survival rates of both mouse genotypes comparable (Fig. 4 G), indicating that neutrophils played an important role in improving the mortality rate of Cd300a^−/− mice after CLP. In contrast, there was still a significant difference in survival duration between WT and Cd300a^−/− mice in which macrophages were depleted (Fig. 4 H). These results indicated that neutrophils play an important role in the survival of mice after CLP.

In the early phase of the CLP peritonitis model, mast cells play an important role in recruitment of neutrophils into the peritoneal cavity by secretion of chemoattractants for neutrophils, such as TNF, and prolong the survival (Echtenacher et al., 1996; Malaviya et al., 1996). We have demonstrated that Cd300a^−/− BMMCs produced more chemoattractants for neutrophils—including TNF, IL-13, and MCP-1—in response to LPS than WT BMMCs in vitro (Fig. 3, C and D). To examine whether CD300a on mast cells is involved in neutrophil recruitment and longer survival, WT and Cd300a^−/− BMMCs were or were not transferred into the peritoneal cavities of Kit^W-sh/W-sh mice—which are deficient in mast cells (Grimbaldeston et al., 2005)—before CLP. After CLP, Kit^W-sh/W-sh mice that received WT BMMCs survived significantly longer than those that did not receive any BMMCs (Fig. 5 A), which is consistent with previous studies (Echtenacher et al., 1996; Malaviya et al., 1996). However, Kit^W-sh/W-sh mice that received Cd300a^−/− BMMCs showed significantly longer survival after CLP and significantly greater bacterial clearance than those that received WT BMMCs (Fig. 5, A and B), suggesting that Cd300a^−/− BMMCs produced a greater amount of chemoattractants for neutrophils than did WT BMMCs in the peritoneal cavity after CLP. To directly confirm this idea, Kit^W-sh/W-sh mice received a 1:1 mixture of CFSE-labeled WT and Cd300a^−/− BMMCs via the peritoneal cavity and were subjected to CLP. Although both WT and Cd300a^−/− BMMCs transferred into the peritoneal cavity showed a low level of TNF expression, Cd300a^−/− BMMCs, defined as CD300a⁻CFSE⁺ cells, showed significantly higher frequency in the BMMCs producing a high level of TNF than did WT (CD300a⁺CFSE⁺) BMMCs (Fig. 5 C). Together, these results indicate that CD300a inhibits production of chemoattractants for neutrophils from BMMCs in the peritoneal cavity after CLP. Although BMMC transfer into Kit^W-sh/W-sh mice does not completely reconstitute the immune system in Kit^W-sh/W-sh mice (Dudeck et al., 2011), our results suggested that CD300a on mast cells was primarily responsible for the phenotype of prolonged survival of Cd300a^−/− mice after CLP.

Because Cd300a^−/− mice survived longer after CLP, we examined whether an antibody against mouse CD300a (TX41) could have a prophylactic effect on CLP-induced sepsis. TX41 does not deplete myeloid cells, including mast cells. Intraperitoneal injection of mice with TX41 1 h before and 18 h after CLP significantly increased neutrophil numbers in the peritoneal cavity, improved bacterial clearance 4 h after CLP, and prolonged survival compared with treatment with a control antibody (Fig. 6, A–C). Similarly, intraperitoneal injection with D89E MFG-E8 also prolonged survival of WT mice compared with treatment with PBS. In contrast, D89E MFG-E8 did not change the survival of Cd300a^−/− mice, indicating that D89E MFG-E8 blocked the interaction between PS and CD300a, but not other PS receptors (Fig. 6 D). Moreover, D89E MFG-E8 improved bacterial clearance of WT mice (Fig. 6 E). These results provided the formal evidence that PS on apoptotic cells affected the bacterial clearance. Thus, blocking the interaction between PS and its receptor CD300a is potentially a useful therapy for prophylaxis against peritonitis-induced sepsis.

A large number of cells undergo apoptosis every day in physiological and pathological settings. PS receptors play a central role in engulfment of apoptotic cells and are crucial to avoiding the development of autoimmune diseases (Hanayama et al., 2004). Pathological conditions such as microbial infections significantly augment cell death by apoptosis (Hotchkiss and Nicholson, 2006) and trigger the first line of inflammatory responses by mast cells through receptors for pathogen-associated molecular patterns such as Toll-like receptors (Marshall, 2004; Bischoff, 2007; Abraham and St John, 2010). Mast cells therefore play an important role in the early phase of immunity against microbial pathogens. Our results indicate that PS not only delivers “eat-me” signals to phagocytes via PS receptors but also suppresses inflammatory responses by mast cells and macrophages via CD300a.

Although CD300a is expressed on professional phagocytes, such as macrophages and dendritic cells (Yotsumoto et al., 2003), we did not observe CD300a-mediated phagocytosis of apoptotic cells by macrophages. Because CD300a associates with SHP-1 upon ligand binding and inhibits activation signals, CD300a may not have a functional property of phagocytosis. However, it cannot be excluded that the CD300a-mediated signal suppresses, rather than augments, phagocytosis by other PS receptors such as TIM-4 (Kobayashi et al., 2007), if PS binding could be shared by CD300a and TIM-4 on a macrophage. This may depend on the binding affinity of these receptors to PS. At present, the affinity of CD300a binding to PS has remained undetermined. Very recently, Choi et al. (2011) have reported that CD300f also recognizes PS and exogenous expression of CD300f enhanced phagocytosis of apoptotic cells by fibroblast cell line L929 cells. Like CD300a, CD300f contains ITIM in the cytoplasmic portion. Upon cross-linking with antibody, the ITIM of CD300f is phosphorylated, recruits SHP-1 and SHP-2, and mediates an inhibitory signal in myeloid cells (Izawa et al., 2009). It is unclear, however, how CD300f-mediated signaling involves phagocytosis of apoptotic cells.

One of the key biological features of apoptotic cell engulfment by phagocytes is the noninflammatory nature (Fadok et al., 1998; Ravichandran and Lorenz, 2007). The molecular mechanism under this feature has still remained unclear. The suppressive effect of CD300a on the proinflammatory cytokine and chemokine production by mast cells does not necessarily account for this feature because mast cells are not phagocytes and their activation is induced by pathological conditions such as microbial infections. However, it may contribute to avoiding excessive inflammations by suppressing mast cell–mediated triggering of inflammation. Our data showed that CD300a deficiency also affected the production of both the proinflammatory and antiinflammatory cytokines TNF and IL-10, respectively, without selectivity by peritoneal resident macrophages, when these cells were cocultured with apoptotic cells in the presence of LPS. However, because SHP-1 recruitment by CD300a that is essential for CD300a-mediated inhibitory signal requires both PS binding and LPS stimulation in macrophages as well as mast cells, it is unlikely that CD300a on macrophages is involved in the noninflammatory nature of phagocytosis of apoptotic cells mediated by PS receptors other than CD300a in the physiological conditions. Further studies should be required to clarify the regulatory role of CD300a in activation of myeloid cells, including dendritic cells and neutrophils as well as mast cells and macrophages. Nonetheless, CD300a is the first PS receptor identified on mast cells and suppresses inflammatory responses by mast cells.

Sepsis, the systemic inflammatory syndrome induced by severe infections, remains a leading cause of death worldwide (Angus and Wax, 2001; Cohen, 2002). Despite extensive basic and clinical research, the pathophysiology of sepsis is still poorly understood, and trials of several therapies have failed to improve the prognosis of septic patients in large multicenter clinical trials (Riedemann et al., 2003; Russell, 2006). The best established mouse model of sepsis, CLP, has shown that mast cells play essential roles in host defense against CLP (Malaviya et al., 1996). Here, we provide compelling experimental evidence supporting the efficacy of anti-CD300a monoclonal antibody and D89E MFG-E8 adjunctive therapy for polymicrobial peritonitis-induced sepsis. Both anti-CD300a and D89E MFG-E8 blocked the binding of PS to CD300a, resulting in the increased production of chemoattractants for neutrophils from mast cells induced by LPS in the peritoneal cavity. Long-term treatment with D89E MFG-E8 leads to the development of autoimmune diseases as a result of the defect of apoptotic cell clearance (Hanayama et al., 2004). However, we injected D89E MFG-E8 only twice before CLP and 18 h after CLP, which was enough for the prophylaxis of peritonitis-induced sepsis. Mast cell activation is important in the very early phase after infection, only when the blocking CD300a-mediated signaling by D89E MFG-E8 or anti-CD300a antibody should be required.

The classical function of mast cells is involved in allergic responses, in which FcεRI-mediated signals lead to degranulation from mast cells (Gilfillan and Tkaczyk, 2006). In vitro analyses demonstrated that cross-linking CD300a with monoclonal antibody suppressed FcεRI-mediated signals, resulting in the decreased degranulation in human and mice (Yotsumoto et al., 2003; Okoshi et al., 2005; Munitz et al., 2006; Karra et al., 2009). Because many cells undergo apoptosis at the site of allergic inflammation, CD300a may play an important role in allergic responses by mast cells and may be a potential prophylaxis and/or therapeutic target for treatment of allergic diseases.

To construct the CD300a targeting vector, we replaced exon 1 to exon 6 of the Cd300a gene with a neomycin-resistance gene cassette (PGK-GB2-neo) using the bacterial artificial chromosome (BAC) system (Yang and Seed, 2003). BALB/c embryonic stem cells were electroporated with linearized Cd300a targeting vector, and cells were selected for G418 antibiotic resistance. Correctly targeted embryonic stem cell clones with normal karyotypes were injected into 3.5-d-old BALB/c blastocysts to create chimeric mice (Cd300a^+/−). Cd300a^+/− mice were intercrossed to generate Cd300a^−/− mice. Homozygous mice were obtained at the expected Mendelian frequencies and developed normally. Cd300a^−/− mice and control WT mice were bred under specific pathogen-free conditions in the same room of our animal facility. CD300a-deficient (Cd300a^−/−) mice in the BALB/c background were backcrossed onto the C57BL/6J genetic background for 12 generations. C57BL/6J-Kit^W-sh/W-sh mice were provided by Riken BioResource Center (Tsukuba, Japan). CAD-deficient mice have been previously described (Nagata, 2005). C57BL/6J mice were purchased from Clea Japan. All experiments were performed in accordance with the guidelines of the animal ethics committee of the University of Tsukuba Animal Research Center.

The fusion protein of CD300a with the Fc portion of human IgG (CD300a-Fc) was generated from chimeric cDNA containing the entire extracellular domain of CD300a and the human IgG1Fc, as previously described (Tahara-Hanaoka et al., 2004). MFG-E8 and D89E MFG-E8 were provided by M. Tanaka (Research Center for Allergy and Immunology, Yokohama, Japan) and also generated in our laboratory, as previously described (Hanayama et al., 2002).

Control rat antibodies and antibodies to CD11b (M1/70), F4/80, c-Kit (2B8), Gr-1 (RB6), FcεRI (MAR-1), and TNF were purchased from BD. The CD300a-specific monoclonal antibody TX41 (rat IgG2a) was generated in our laboratory, as described previously (Yotsumoto et al., 2003). The antibody to MCP-1 was purchased from BioLegend, and the antibody to IL-13 was purchased from eBioscience. Antibodies to SHP-1 (C-19) and SHP-2 (C-18) were purchased from EMD Millipore. To generate BMMCs, 2 × 10⁶ BM cells were cultured in a 10-cm dish in complete RPMI 1640 medium containing 10% FBS in the presence of 10 ng/ml stem cell factor and 4 ng/ml IL-3 for >5 wk. Cells were passaged with fresh medium every week. Flow cytometry analyses showed that >99% of the cultured cells were c-kit⁺FcεRI⁺ cells. NIH 3T3 cells were transfected with a pMX-neo retrovirus vector plasmid containing Flag-tagged CD300a cDNA. An NIH 3T3 transfectant stably expressing TIM-4 was provided by T. Kitamura (University of Tokyo; Yamanishi et al., 2010). RBL-2H3 transfectants expressing WT CD300a and mutated CD300a tagged with Flag at the N terminus, in which the tyrosines were replaced with phenylalanine or the cytoplasmic region was deleted, were previously described (Okoshi et al., 2005).

Thymocytes from C57BL/6 or CAD-deficient mice were incubated with 10 µM dexamethasone (Sigma-Aldrich) in RPMI medium for 6 h. Induction of apoptosis was determined by labeling PS expressed on the plasma membrane with APC-conjugated annexin V (BD) and propidium iodide (Sigma-Aldrich).

2 × 10⁵ thioglycollate-elicited peritoneal macrophages were cocultured with apoptotic thymocytes at a ratio of 1:5 for 1 h at 37°C in 8-well Laboratory-Tek II chamber slides (Thermo Fisher Scientific). The cells were then washed with cold PBS and fixed with 1% paraformaldehyde, subjected to TUNEL staining with FITC-labeled dUTP (Roche), and then analyzed by laser-scanning confocal microscopy, as described previously (Miyanishi et al., 2007; Yamanishi et al., 2010).

Apoptotic thymocytes were labeled with pHrodo succinimidyl ester (Invitrogen), as described previously (Miksa et al., 2009). In brief, 10⁶ thymocytes were incubated at room temperature in pHrodo succinimidyl ester solution at a concentration of 20 ng/ml for 30 min, washed twice with PBS, and resuspended in OPTI-MEM 1. 6 × 10⁴ NIH 3T3 transfectants expressing CD300a or TIM-4 were cocultured with pHrodo succinimidyl ester–labeled apoptotic thymocytes at a ratio of 1:5 for 1 h at 37°C in 8-well Laboratory-Tek II chamber slides, washed with PBS and fixed as described in the previous paragraph, and then analyzed by laser-scanning confocal microscopy or flow cytometry.

CLP was performed as previously described (Echtenacher et al., 1996). In brief, the cecum was exposed by a 1- to 2-cm midline incision on the ventral abdomen, ligated at its distal portion, and punctured twice with a 27-gauge needle in the ligated segment. We then returned the cecum to the abdomen, injected 1 ml of sterile saline subcutaneously for rehydration, and closed the incision by suture.

The concentrations of cytokines and chemokines were measured in triplicate using ELISA kits from BD (TNF) and R&D Systems (MCP-1 and IL-13). Intracellular cytokines and chemokines were analyzed by flow cytometry after fixation and permeabilization using a Fix & Perm cell fixation and permeabilization kit (Invitrogen). The supernatant of the co-culture of BMMC with apoptotic cells was subjected to cytokine and chemokine proteome analyses with a Proteome Profiler Array (R&D Systems) in accordance with the manufacturer's instructions.

Aerobic bacterial CFUs were quantified by plating serial dilutions of peritoneal lavage fluid and culturing them at 37°C for 48 h on brain-heart-infusion agar plates. CFUs were then enumerated by calculating the number of colonies in 1 ml peritoneal lavage fluid, as previously described (Plitas et al., 2008).

For macrophage depletion, clodronate liposomes and control PBS liposomes (Encapsula NanoSciences) were prepared as previously described (Van Rooijen, 1989). Mice were injected i.p. with 0.5 ml liposomes 24 h before CLP. For neutrophil depletion, mice were injected i.p. with 5 µg anti–Gr-1, 24 h before CLP. WT or Cd300a^−/− neutrophils (1 × 10⁶/mice) purified from the BM were injected into some neutrophil-depleted mice just after CLP.

To target SHP-1 in BMMCs, 0.5 mM SHP-1 siRNA (siGENOME SMARTpool; Thermo Fisher Scientific) was mixed with 1 µl X-tremeGENE siRNA transfection reagent (Roche) and transfected into 5 × 10⁵ BMMCs, as previously described (An et al., 2006). After 48 h, the supernatant was removed, and fresh medium and 0.3 mM SHP-1 siRNA were added. The cells were cultured for another 16 h before further experiments.

The Kaplan-Meier log-rank test was used for survival analyses. All other statistical analyses were performed using an unpaired Student's t test. P < 0.05 was considered statistically significant.

We thank S. Mitsuishi and F. Abe for secretarial and technical assistances, respectively.

This research was supported in part by grants provided by the Ministry of Education, Culture, Sports, Science and Technology of Japan and by the Program for Promotion of Fundamental Studies in Health Science of the National Institute of Biomedical Innovation (NIBIO), Japan.

The authors have no competing financial interests.