Background Signalling cross talk provides a molecular basis for modulating a given signalling pathway by another, and it is often critical for regulating cellular responses elicited by cytokines. Previously, we reported on the critical role of the IFN-α/β signalling complex, generated by spontaneously produced IFN-α/β, in efficient IFN-γ signalling.
Results In the present study, we have demonstrated that the IFN-α/β signalling complex also contributes to efficient IL-6 signalling. In fact, IL-6-induced activation of the Stat1 and Stat3 transcription factors is markedly diminished in the absence of the IFN-α/β signalling complex. The induction of several target genes for these factors is also diminished, both in vitro and in vivo. We provide evidence that the cytoplasmic tyrosine residues of IFNAR-1, which remains phosphorylated by a weak IFN-α/β stimulation, provide docking sites for Stat1 and Stat3 to form homo- or heterodimers following IL-6 stimulation. Furthermore, a chemical cross-linking experiment revealed that IFNAR-1 and gp130, a common signal transducer for the IL-6 family of cytokines, exist in close proximity.
Conclusions The constitutive weak IFN-α/β signal provides a foundation for strong cellular responses to IL-6, IFN-γ, and possibly other cytokines. Our results also suggest the assembly of cytokine receptor subunits, which may represent a ‘receptosome’-like structure, allowing the unique signalling cross talks to occur.
Common features among many cytokines lie in their functional redundancy and pleiotropy (Darnell et al. 1994; Durum & Oppenheim 1993; Howard et al. 1993; Ihle & Kerr 1995; Miyajima et al. 1992; Paul & Seder 1994; Taniguchi 1995). In this context, many cytokine receptors consist of more than two subunits, and that, in some cases, one receptor is shared among different cytokine receptor complexes. One example is gp130, which was originally discovered as a subunit of the interleukin-6 (IL-6) receptor; this molecule is utilized for signal transmission by many cytokines, such as the leukaemia inhibitory factor, and oncostatin M (Taga & Kishimoto 1997). Loss-of-function mutations in gp130 and other shared receptors lead to a complete absence of cellular responses to the respective cytokines, indicating that they are indispensable for the signalling to occur (Hibi & Hirano 1998). To further elaborate on the complexity of the cytokine receptors, one may also consider the presence of receptor subunits belonging to another category, which may help enhance (or suppress) the magnitude of signal transmission via signalling cross talk. In fact, such examples have been reported in erythropoietin and IL-6 signalling (Wu et al. 1995), whereby the Kit receptor and ErbB2 receptor (Qiu et al. 1998) are, respectively, involved.
Recently, we reported on a unique signalling cross talk between interferon-α/β (IFN-α/β) and IFN-γ (Takaoka et al. 2000). In fact, low levels of IFN-α/β mRNA expression were detected in mouse embryonic fibroblasts (MEFs) in the absence of virus infection, and a constitutive subthreshold signalling by spontaneously produced IFN-α/β is critical to the efficient IFN-γ signalling. This cross talk is dependent on the association between two receptor subunits, i.e. IFN-α/β receptor subunit-1 (IFNAR-1) and IFN-γ receptor subunit-2 (IFNGR-2), and on the cytoplasmic tyrosine residues of IFNAR-1. Thus, tyrosine residues of IFNAR-1 provide ‘docking sites’ for signal transducer and activator of transcription 1(Stat1) and Stat2 after stimulation of IFNGR, brought into close proximity through the association of the above two subunits (Takaoka et al. 2000).
These findings prompted us to examine whether or not similar IFNAR-1-mediated cross talk is operational for other cytokines, such as IL-6, which also activates similar transcription factors of the Stat family. The IL-6 receptor is a heterodimeric complex, consisting of an IL-6 specific ligand-binding subunit, an α chain, and a signal-transducing subunit gp130 (Hirano & Kishimoto 1990; Taga & Kishimoto 1997). IL-6 bound to the α chain receptor causes the association of the complex with gp130, resulting in activation of Janus kinase1 (Jak1), Jak2 and Tyk2. Subsequently, phosphorylation of Stat1 and Stat3, mediated by these tyrosine kinases, is induced for their activation (Hirano & Kishimoto 1990; Taga & Kishimoto 1997). Here we report that IL-6 also requires a constitutive subthreshold IFN-α/β signalling for efficient activation of these transcription factors by a mechanism which is similar to that of IFN-γ signalling (Takaoka et al. 2000). The significance of the current findings will be discussed with respect to the similar but distinct functions of these classes of cytokines.
IL-6-induced activation of Stat1 and Stat3 in cells lacking the IFNAR-1 subunit
To investigate the contribution of the IFN-α/β signalling complex in IL-6 signalling, we examined the signalling events induced by IL-6 in wild-type (WT) and IFNAR-1-deficient (AR1−/−) MEFs. We performed an electrophoretic mobility shift assay (EMSA) for the IL-6-induced DNA-binding activities of Stat1 and Stat3, using an interferon regulatory factor-1 (IRF-1)-IFN-γ-activated site (GAS) probe (Takaoka et al. 2000). As shown in Fig. 1A, IL-6 stimulation of WT MEFs resulted in the formation of three bands, representing the DNA-binding activities of the Stat3-Stat3 homodimer (upper band), Stat3-Stat1 heterodimer (middle band) and Stat1-Stat1 homodimer (lower band) (Sadowski et al. 1993). The identities of these complexes were confirmed based on their reactivity to the anti-Stat1 or anti-Stat3 antibody (data not shown). Interestingly, these DNA-binding activities induced by IL-6 were found to be much lower (approximately by sevenfold) in AR1−/− MEFs than in WT MEFs, although the kinetics of the induction remained unaltered (Fig. 1A). The IL-6-induced DNA-binding activities of both Stat1 and Stat3 increased in a dose-dependent manner in AR1−/− MEFs (25∼400 ng/mL), but they remained lower than those in WT MEFs at any concentrations (data not shown). Similar results were obtained in splenocytes from AR1−/− mice (Fig. 1B), indicating that the observation is not restricted to MEFs.
Immunoblot analysis revealed that the expression levels of endogenous gp130, Stat1 and Stat3 in AR1−/− MEFs were essentially the same as those in WT MEFs (Fig. 1C). Furthermore, the levels of the IL-6-induced tyrosine phosphorylation of Stat1 and Stat3 in AR1−/− MEFs also remained the same as those in WT MEFs (Fig. 1C). On the other hand, when NF-κB activation, induced by unrelated cytokines (i.e. TNF-α or IL-1β) was examined, no difference was found between WT and AR1−/− MEFs (Fig. 1D), indicating that IFNAR-1 is not involved in these events.
Induction of the target genes by IL-6, in vivo and in vitro
We next examined the induction of two target genes for IL-6, i.e. IRF-1 and junB genes in MEFs (Abdollahi et al. 1991; Kojima et al. 1996). The IL-6-response elements have been identified as IRF-GAS and JRE-IL6, respectively, for these two genes. It is known that both Stat1 and Stat3 bind to the former element, whereas only Stat3 binds to the latter element (Coffer et al. 1995; Zhang et al. 1995). As shown in Fig. 2, the induction of IRF-1 mRNA in IL-6-stimulated AR1−/− MEFs was approximately 10-fold lower than that in WT MEFs (a reduction from an approximately 629-fold induction level to a 60-fold induction level at 0.5 h). Furthermore, the IL-6-induced junB mRNA expression level was also reduced in AR1−/− MEFs, compared with that in WT MEFs (a reduction from an approximately 195-fold induction level to a 70-fold induction level at 0.5 h). These results are consistent with the above results in the case of the reduced Stat1/Stat3 activation by IL-6 in AR1−/− MEFs.
We further examined the contribution of IFNAR-1 in IL-6-mediated gene induction for acute-phase proteins in the liver (Stadnyk & Gauldie 1991), an event which is critical for eliciting innate immune response (Kopf et al. 1995). It has been reported that, among the acute-phase response genes, α2-macroglobulin (α2M) and serum amyloid P-component (SAP) are regulated by the Jak-Stat pathway involving mainly Stat3 (Wegenka et al. 1994). WT and AR1−/− mice were injected intraperitoneally with IL-6, and the hepatic mRNA induction of these genes was examined. As shown in Fig. 3, the mRNA levels of both α2M and SAP increased after IL-6 treatment in the WT mice. In the AR1−/− mice, α2M mRNA induction was much lower than that in the WT mice (a reduction from an approximately 8.4-fold induction level to a 2.8-fold induction level at 4 h). Furthermore the IL-6-induced SAP mRNA expression level was also reduced in AR1−/− mice (a reduction from an approximately 86-fold induction level to a 55-fold induction level at 4 h).
Requirement of IFNAR-1 phosphorylation by IFN-α/β signalling for efficient IL-6 signalling
It has been shown that low levels of spontaneously produced IFN-α/β are detected in MEFs and splenocytes, and that the IFNAR-1 phosphorylation by the subthreshold IFN-α/β signalling is critical for providing docking sites for Stats (Takaoka et al. 2000). We then examined the IL-6-induced DNA-binding activities of Stat1 and Stat3 in IFN-β−/− MEFs, in which such IFN-α/β signalling can be barely detected (Takaoka et al. 2000). As shown in Fig. 4A, the IL-6-induced Stat1/Stat3 activation was also reduced in IFN-β−/− MEFs, the same as in AR1−/− MEFs (Fig. 1A). This deficiency was rescued by exogenously adding IFN-β at a low concentration (0.1 U/mL), which, by itself, did not activate the Stats (Fig. 4A). These results indicate that the constitutive subthreshold IFN-α/β signalling contributes to the activation of Stat1 and Stat3 by IL-6.
It is likely that the subthreshold IFN-α/β signalling results in a phosphorylation of IFNAR-1, which provides docking sites for Stats in IL-6 signalling. The IL-6-induced activation of Stat1 and Stat3 in AR1−/− MEFs, which were transfected with wild-type murine IFNAR-1 or two different types of mutant IFNAR-1 that lack some tyrosine residues, was examined (Takaoka et al. 2000) (Fig. 4B). In these cells, the expression levels of each transfected IFNAR-1 and endogenous gp130 were shown to be nearly similar (data not shown). It was shown that Stat3 binds to the YXXQ motif present in gp130, and that similar motifs are present within the cytoplasmic domain of human and murine IFNAR-1 (Pfeffer et al. 1997; Stahl et al. 1995). It was also shown that, in addition to Stat1, Stat3 also binds to human IFNAR-1 (Su & David 2000; Yang et al. 1996). As shown in Fig. 4B, the expression of mutant IFNAR-1 (AR1Y455F) by the retrovirus mediated gene transfer resulted in only about 20% restoration of the IL-6-induced activation of Stat1, as compared with the complete restoration by the expression of wild-type IFNAR-1 (AR1). Furthermore, mutant IFNAR-1, which had mutations in all four cytoplasmic tyrosine residues (AR1YF), was completely inactive (Fig. 4B). Thus, these IFNAR-1 residues contribute to the activation of Stats by IL-6.
These results support the view that IFN-α/β signalling is essential for maintaining IFNAR-1 in its phosphorylated form, thereby providing docking sites for Stat1 and Stat3, to efficiently dimerize when tyrosine phosphorylation of these Stats is induced by IL-6. Consistent with this view, the immunoprecipitation assay revealed that Stat1 and Stat3 were co-immunoprecipitated with IFNAR-1, even in the absence of IL-6 stimulation in WT MEFs but not in IFN-β−/− MEFs (Fig. 4C). These results suggest that the phosphorylated tyrosine residues of IFNAR-1 are required for this association. It is of note that the association of the tyrosine-phosphorylated Stat1 with IFNAR-1 was observed after IL-6 stimulation in WT MEFs, but not in IFN-β−/− MEFs (Fig. 4C). Thus, these observations provide a molecular basis for the requirement of IFNAR-1 for the enhanced IL-6 signalling. Interestingly, activation of the IFN-stimulated gene factor 3 (ISGF3), the critical transcription factor for the antiviral response of the cell, is observed in IFN-γ-stimulated cells but not in IL-6-stimulated cells (Fig. 4D). In agreement with this observation, Stat2, which participates in the formation of ISGF3 upon tyrosine phosphorylation, is not phosphorylated by IL-6 stimulation (Fig. 4E).
Proximal localization of IFNAR-1 and gp130
In view of the above signalling cross talk, in which seemingly futile IFN-α/β signalling contributes to IL-6 signalling, one may speculate that IFNAR-1 may be located close to gp130. We previously reported that the IFN receptor components and Janus protein tyrosine kinases (Jak PTKs) are found exclusively in membrane fractions rich in caveolin (termed as caveolar membrane fractions) (Anderson 1998; Brown & London 1998; Simons & Ikonen 1997) in MEFs (Takaoka et al. 2000). To investigate whether gp130, which is known to be associated with Jak PTKs, is localized in caveolar membrane fractions, these fractions from WT MEFs were prepared by the detergent-free method (Mineo et al. 1996; Smart et al. 1995). Similar to IFNAR-1, gp130 was found to be localized in the caveolar membrane fractions (Fig. 5A), which is consistent with a previous report (Koshelnick et al. 1997). It is not known, however, if a similar localization is found in other cells such as lymphocytes which lack caveolin expression. On the other hand, Stat1 was found in both caveolar fractions and noncaveolar fractions; the latter fraction was reported to be rich in paxillin proteins (Smart et al. 1995; Xavier et al. 1998). These observations prompted us to examine the association of these two components by immunoprecipitation analysis in the presence or absence of the chemical cross-linker 3,3′dithiobis(sulfosuccinimidylpropionate) (DTSSP). As shown in Fig. 5B, the association between IFNAR-1 and gp130 could be detected by immunoprecipitation analysis after cross-linking, suggesting that IFNAR-1 and gp130 exist in close proximity, thereby making the above signalling cross talk possible.
Signalling cross talk is an aspect which is critical to the enhancement or suppression of cellular responses, thereby modulating the magnitude and diversity of a response to a given stimulus. In the present study, we provide evidence for the cross talk between the IFN-α/β and IL-6 signalling pathways, in that weak signalling by spontaneously produced IFN-α/β contributes to an efficient IL-6 signalling. Since we have previously shown that the weak IFN-α/β signalling is also critical for efficient IFN-γ signalling (Takaoka et al. 2000), our current results suggest a broad function of the weak IFN-α/β signalling in the regulation of cellular responses to cytokines.
In fact, full activation of Stat1 and Stat3 by IL-6 is not observed in either AR1−/− MEFs or IFN-β−/− MEFs, and the target genes of IL-6 are not fully activated in the absence of IFNAR-1, either in vitro and in vivo. It is worth noting that the mRNA induction levels of the IL-6-induced genes are differentially affected in the absence of IFNAR-1 (Figs 2 and 3). Thus, we infer that each of these genes may require different Stat activity threshold levels for full induction by these cytokines. In this context, the activation of promoters usually requires the assembly of a highly specific nucleoprotein complex (Tjian & Maniatis 1994). Thus, it is possible that these Stats need to collaborate with other factors, which may be distinct from one promoter to another, to efficiently activate each of the promoters, and that the Stat activity threshold level for a given promoter is subject to the activities of the collaborating factors. This is an interesting issue, to be addressed in future studies.
In view of the previous observation that maximal IFN-γ-induced activation of Stat1 requires IFNAR-1 for its dimerization, in which IFNAR-1 provides ‘docking sites’ for Stat1 (Takaoka et al. 2000), it is likely that the defect is due to the inefficient dimerization of Stat1 and Stat3 in AR1−/− MEFs. In fact, the levels of tyrosine phosphorylation in Stat1 (and Stat3) are similar between the wild-type and AR1−/− MEFs (Fig. 1C), but the IL-6-induced association of the phosphorylated Stat1 to IFNAR-1 observed in WT cells cannot be observed in IFN-β−/− cells (Fig. 4C). In addition, the IFNAR-1 mutants lacking the cytoplasmic tyrosine residues failed to procure the full blown Stat activation by IL-6 in AR1−/− MEFs (Fig. 4B). In this context, there is evidence that, in addition to Stat1 and Stat2, Stat3 also associates with IFNAR-1 (Su & David 2000; Yang et al. 1996). Thus, weak IFN-α/β signalling is needed to maintain the intracellular tyrosine residues of IFNAR-1 in a phosphorylated form so as to provide niches where these Stats can dimerize efficiently. In this regard, there has also been a report that Stat1 and Stat2 are also associated with IFNAR-2 before ligand stimulation (Li et al. 1997). It is not currently known if a similar association occurs in MEFs, but the contribution of IFNAR-2 to recruit these Stats may be marginal, if any, for the enhancement of IL-6 or IFN-γ signalling.
Although both IFN-γ and IL-6 utilize the IFN-α/β signalling complex, the latter shows no antiviral activity (Hirano et al. 1988; Reis et al. 1988). In fact, unlike IFN-γ stimulation, IL-6 stimulation failed to activate ISGF3 (Fig. 4D); Stat2 phosphorylation, which is required for ISGF3 formation, is not induced by IL-6 (Fig. 4E). These results may offer an explanation as to why IL-6, unlike the IFNs, cannot induce antiviral activity in cells.
It was reported that gp130-deficient mice are embryonically lethal (Akira et al. 1995), possibly due to the loss of response to IL-6 and cardiotrophin-1 (CT-1), which also utilize gp130 for signalling. Since AR1−/− mice develop normally (Muller et al. 1994), the IFN-α/β signal is not apparently essential in this gp130-mediated process. On the other hand, our results showing that the extent of gene induction for acute-phase proteins by IL-6 in the liver decreases in AR1−/− mice (Fig. 3) suggest the involvement of this signal in the innate immune response. Apparently, it is important to further examine which of the other IL-6-mediated biological responses are affected by absence of the IFN-α/β signal.
IFNAR-1, which was found to be in close proximity to gp130, was previously shown to be associated with IFNGR2 (Takaoka et al. 2000). In addition, it was reported that ErbB2 formed a complex with gp130 in an IL-6-dependent manner (Qiu et al. 1998). Taking our results together with those of previous reports, cytokine receptors or growth factor receptors may exist in close proximity and form a multimeric complex, which may be termed a ‘receptosome.’ Such a complex may be critical to the enhancement or sometimes suppression of a given cytokine signalling pathway by cross talk with other signalling pathways. The proposed receptosome may be reminiscent of the signalling complex for lymphocyte activation, in which numerous signalling molecules, collaborating with antigen receptors, are brought into proximity through lipid rafts, so as to allow the molecules to function efficiently (Brown & London 1998; Dustin & Chan 2000; Langlet et al. 2000).
Taken together with previous results of the cross talk with IFN-γ signalling, the weak IFN-α/β signal appears to provide a foundation for more efficient signalling in different cytokine signalling pathways such as those for IL-6 and IFN-γ. It is interesting to examine whether IFN-α/β signalling affects other signalling pathways. Many other cytokines, including those belonging to the gp130 family, IL-10 and growth hormone, also activate Stat1 and Stat3, which are usually recruited to IFNAR-1 (see Fig. 4C); it is therefore possible that cells also use the IFNAR-1-associated Stats in these signalling systems. Moreover, Stat4 is recruited to IFNAR-1 via Stat2 in human cells (Farrar & Schreiber 1993; O'Shea & Visconti 2000), suggesting the possibility that the activation of Stat4 by IL-12 is affected by IFN-α/β signalling. These are interesting issues to be addressed in future studies.
Mice, cells and cell cultures
IFNAR-1-deficient mice were purchased from B&K Universal Group (North Humberside, UK). The generation of IFN-β-deficient mice was as previously described (Takaoka et al. 2000). MEFs and splenocytes were prepared using a standard procedure (Takaoka et al. 1999). MEFs were grown in Dulbecco's modified Eagle's medium (Nissui Pharmaceutical, Tokyo, Japan) supplemented with 10% foetal calf serum.
Electrophoretic mobility shift assay (EMSA)
Cells were stimulated with cytokines for the indicated times and then lysed. Whole-cell extracts were subjected to EMSA, as previously described (Takaoka et al. 1999). Cytokines used with MEFs in this assay are as follows: recombinant human IL-6 (specific activity: 1.0 × 105 U/µg) (Genzyme/Techne, Cambridge, MA); soluble human IL-6 receptor (R&D systems, Minneapolis, MN); recombinant mouse TNF-α (specific activity: 2.0 × 105 U/µg) (Genzyme/Techne, Cambridge, MA); recombinant mouse IL-1β (specific activity: 1.0 × 106 U/µg) (Genzyme/Techne, Cambridge, MA); recombinant mouse IFN-β (specific activity: 3.0 × 104 U/µg) (kindly provided by Toray Inc., Shiga, Japan). Recombinant mouse IL-6 (specific activity: 1.0 × 105 U/µg) (Genzyme/Techne, Cambridge, MA) was used with splenocytes.
DNA transfection, immunoprecipitation and immunoblotting
Retrovirus-mediated gene transfer was performed as previously described (Takaoka et al. 2000). Cell lysis, immunoprecipitation and immunoblotting were carried out as previously described (Takaoka et al. 2000). The following antibodies were purchased: anti-IFNAR-1 antibody (Research Diagnostics, Flanders, NJ); anti-caveolin and anti-paxillin antibodies (Transduction Laboratories, Lexington, KY); anti-Stat1 p84/p91 and anti-Stat3 antibodies (Santa Cruz Biotechnology, Santa Cruz, CA); anti-phospho-Stat1 (Y701) and anti-phospho-Stat3 (Y705) antibodies (New England Biolabs, Beverly, MA); anti-phosphotyrosine and anti-Stat2 (Upstate Biotechnology). The anti-gp130 antibody was a gift from Dr T. Taga (Tokyo Medical and Dental University).
Preparation of caveolar membrane fractions
Caveolar membrane fractions were prepared from MEFs by the detergent-free method previously described (Mineo et al. 1996; Smart et al. 1995). Gradient fractions were collected from the bottom of the gradient and aliquots of these fractions were analysed by SDS–polyacrylamide gel electrophoresis.
RNA blotting (Northern hybridization)
MEFs were plated at a density of 2 × 106 cells/10 cm plate. After a 12 h incubation, the cells were stimulated with IL-6 for the indicated times. RNA preparation and RNA blotting were performed as previously described (Takaoka et al. 1999). In the case of in vivo induction of acute-phase proteins, mice (8–12 weeks old) were intraperitoneally injected with 10 µg of recombinant human IL-6 (specific activity: 5 × 103 U/µg; specific activity was determined using SKW6-C14 cells (Hirano & Kishimoto 1990)) (kindly provided by Ajinomoto Inc., Tokyo, Japan). Total RNA was extracted from the liver at the indicated times and subsequently subjected to RNA blotting as previously described (Takaoka et al. 1999). The cDNA probes for interferon regulatory factor-1 (IRF-1) (Takaoka et al. 1999), junB (Shibuya et al. 1992), and serum amyloid P-component (SAP) (kindly provided by Dr Takeda, Osaka University) were used in RNA blotting; and that for α2-macroglobulin (α2M) was synthesized by transcriptase-polymerase chain reaction from total RNA obtained from liver stimulated with lipopolysaccharide.
MEFs (5 × 107 cells) were suspended in a Cell Dissociation Buffer (Life Technologies, Gaithersburg, MD). The cells were then incubated with IL-6 (400 ng/mL) together with sIL-6Rα (400 ng/mL) for 1 h at 4 °C with rotation. The cells were then treated with a 2-mm solution of cross-linker 3,3′-dithiobis (sulfosuccinimidylpropionate) (DTSSP) (Pierce, Rockford, IL) for 30 min, and the reaction was quenched with 20 mm Tris-HCl (pH 7.5) for 15 min on ice. The cells were lysed in a lysis buffer (50 mm Tris-HCl pH 7.5, 150 mm NaCl, 1 mm EDTA, 0.5% (v/v) NP-40, 15% (v/v) glycerol, 30 mm NaF, 1 mm sodium orthovanadate, 1 mm PMSF, 5 µg/mL aprotinin, 5 µg/mL leupeptin, 2 mmβ-glycerophosphate and 1.5 µg/mL pepstatin). The lysates were subjected to immunoprecipitation with an antibody for gp130 and then to electrophoresis under reducing conditions (with 280 mm 2-mercaptoethanol). Immunoprecipitates were immunoblotted using an anti-IFNAR-1 antibody.
We thank Dr T. Hirano for his invaluable suggestions; Drs K. Takeda and A. Okano for helpful advice, support and discussion; Dr G. Uze for the murine IFNAR-1 cDNA; Dr T. Taga for the anti-gp130 antibody; Dr K. Takeda for the murine SAP cDNA; Drs S. Noguchi and H. Suemori for the generation of IFN-β-deficient mice; Toray Inc. for the mouse IFN-β; Ajinomoto Inc. for the human IL-6; and Ms M. Isobe for technical assistance. This work was supported by grants from the Research for the Future Program (96L0037), the Japan Society for the Promotion of Science, Advanced Research on Cancer, the Human Frontier Science Program Organization, the Ichiro Kanehara Foundation, and the Mochida Memorial Foundation for Medical and Pharmaceutical Research.