Delta-like protein (Dlk) is a transmembrane protein with 6 epidermal growth factor (EGF)-like repeats in its extracellular domain (Figure1). Our laboratory identified Dlk as a cell surface marker molecule of hepatoblasts, progenitor cells in fetal liver. Previous in vitro studies have shown that Dlk is involved in differentiation and development of various tissues and organs such as adipogenesis, osteogenesis, hematopoiesis, B-cell development, and thymocyte development. The extracellular domain of Dlk is also known as fetal antigenl (FAI), which is cleaved off from the membrane and secreted into the serum in embryonic stages. Furthermore, it is well known that Dlk is one of the imprinted genes, and expressed only from the paternal allele. In general, the imprinted genes play indispensable roles in embryonic development. Consistent with these facts, targeted disruption of Dlk in mice clearly showed its importance in normal development in vivo. Dlk-deficient mice show growth retardation, increased adipose tissues, abnormal lipid metabolism, skeletal abnormality, and abnormal B-cell development. Moreover, about half of Dlk-deficient mice die within 2 days after birth. Therefore, Dlk seems to play essential and fundamental roles in normal development and differentiation of various cell types and tissues. However, functions of Dlk are still largely unknown despite of its importance. The extracellular domain of Dlk shows sequence homology with that of Delta, a transmembrane ligand of Notch, but lacks the Delta/Serrate/LAG-2 (DSL)-motif required for activation of Notch. Therefore, Dlk may not be a ligand of Notch. In addition, its intracellular domain does not show any sequence homology with known domains. These molecular features may be a reason preventing elucidation of molecular functions of Dlk. Therefore, identification of a Dlk-binding partner molecule should help understanding functions of Dlk.

For this purpose, I first screened for Dlk-binding molecules by expression cloning. As I found that hepatic progenitor cells proliferating on laminin (HPPL), a cell line of hepatoblasts, bound to the extracellular domain of Dlk fused to the Fc domain of human IgG (Dlk-Fc), I constructed a cDNA library of HPPL. Then, I searched for Dlk-binding molecules in the cDNA library, resulting in identification of cysteine-rich fibroblast growth factor receptor (Cfr) as a Dlk-binding molecule. Cfr is also a transmembrane protein with a long extracellular domain. The extracellular domain of Cfr contains 16 repeats of a unique motif only found in Cfr, which is called Cfr-repeats (Figurel). Specific interaction of Dlk and Cfr was confirmed by immunoprecipitation assay utilizing protein expression system with COS7 cells. The extracellular domain of Cfr (Cfr-EC) was coimmunoprecipitated with Dlk-Fc from culture supernatant, but not with the extracellular domain of CD4 fused to the Fc domain of human IgG (CD4-Fc) as a control.

Originally, Cfr was biochemically identified as a fibroblast growth factors (FGFs)-binding molecule, and it is known that Cfr binds to FGFs by its extracellular domain. However, Cfr has only a short peptide without a kinase domain in its intracellular region (Figure1). Cfr does not show any sequence homology to fibroblast growth factor receptors (FGFRs) with a tyrosine-kinase or other known FGF receptors such as Klothos. Thus, Cfr is a unique FGF receptor. Interestingly, Cfr is also known as E-selectin ligand-1 (Esl-1) for its affinity for E-selectin. Furthermore, Cfr is identical to Golgi apparatus proteinl (G1g1) or MG-160, one of the proteins that reside in the Golgi apparatus. Consistent with this, I also confirmed localization of endogenous Cfr in the Golgi apparatus in mouse embryonic fibroblasts (MEFs) by immunofluorescent staining. Surprisingly, I found that endogenous Dlk was also predominantly localized in the Golgi apparatus, while some in the cell surface in MEFs. Intracellular localization of endogenous Cfr and Dlk in 3T3-L1 cells, a preadipocyte cell line that expresses Dlk, as well as those exogenously introduced in COS7 cells was basically the same. The colocalization of Cfr and Dlk strongly suggests a functional link between them.

The FGF family consists of 22 members in mice, and their signal transducing receptors are the FGFRs with a tyrosine-kinase. There are 4 FGFRs in mice and each binds to its specific set of FGF ligands. Because the FGF-FGFR signaling pathways play indispensable roles in almost all developmental processes, the binding of Cfr to FGFs may play a significant role for the FGF signaling to accomplish normal development. However, there has been no evidence for involvement of Cfr in the FGF signaling or developmental processes so far. Therefore, I investigated functions of Cfr in the FGF signaling pathway.

For this purpose, I generated Cfr-deficient mice and examined their phenotypes to reveal physiological functions of Cfr and relation between Dlk and Cfr. I utilized a gene-trapped ES cell line in which a β-geo cassette with a splice acceptor was inserted in the 1st intron of Cfr gene. This insertion causes splicing of the 1st exon of Cfr to the β-geo cassette and disrupts the normal splicing of Cfr. Using this ES cell line, I obtained a chimeric mouse with the mutated Cfr gene in the germ line. I mated this chimeric mouse with wild-type mice and obtained heterozygotes. Inverse PCR using the heterozygote mouse genome confirmed the gene-trapping of the Cfr gene. Western blot analysis also showed the absence of the wild-type Cfr protein in whole embryo lysate from mutant Ell.5 embryos. These results indicate that this gene-trapping resulted in a null mutation. To observe phenotypes of Cfr-deficient mice, I mated the heterozygote mice. As a result, genotypic ratio for Cfr-deficient mice was Mendelian till El 8.5, but about 90% of Cfr-deficient mice died within 2 days after birth. Moreover, Cfr-deficient mice showed growth retardation, tail distortion, bloated abdomen, and cleft palate, indicating that Cfr is also an important molecule for normal development as Dlk. First, I considered the possibility that the observed phenotypes were due to alteration of Dlk expression by loss of Cfr, a binding partner of Dlk. Quantitative PCR of Dlk mRNA showed no significant difference in the amount of Dlk mRNA between wild-type and Cfr-deficient embryos at E9.5. Whole-mount immunohistochemistry also showed no difference in the level and pattern of Dlk protein expression. I also examined the intracellular localization of Dlk in Cfr-deficient MEFs. In Cfr-deficient MEFs, Dlk was still localized in the Golgi apparatus in the absence of Cfr. From these results, I concluded that the intracellular localization of Dlk is independent of Cfr, and the observed phenotypes of Cfr-deficient mice are caused by a mechanism distinct from the alteration of Dlk expression.

To address the functional link between Cfr and Dlk, I compared the phenotypes of Cfr-deficient mice with those of Dlk-deficient and Dlk-transgenic mice, and noticed that the phenotypes of Cfr-deficient mice are similar to those of Dlk-transgenic mice, i.e. both show growth retardation throughout their life span and tail distortion. The fact that loss of Cfr and gain of Dlk result in similar phenotypes indicates that Cfr and Dlk may function in the opposite directions. Furthermore, I addressed functional relation between Cfr and the FGF signaling by comparing phenotypes of the known Fgf-deficient mice with those of Cfr-deficient mice, and noticed that Fgfl8-deficient mice show phenotypes very similar to those of Cfr-deficient mice, such as perinatal death, growth retardation, skeletal abnormality including tail distortion, and cleft palate. These similarities suggest cooperation between Cfr and FGF18.

To reveal genetic interaction between Cfr and Fgf18, Cfr+/ -mice were crossed with Fgf18+/- mice, and I observed phenotypes of the offspring. As a result, Cfr+/-;Fgf18+1- double heterozygotes were viable and fertile, and showed no obvious phenotype at birth compared to other genotypes. However, the double heterozygotes developed tail distortion by postnatal day 7, while neither Cfr+/- nor Fgf18+/- littermates showed such a phenotype. These results clearly demonstrate that Cfr and FGF18 cooperatively function. It is well known that the major receptor for FGF18 is FGFR3, and targeted disruption of Fgfr3 causes skeletal abnormality including tail distortion. The tail distortion of Ffgr3-deficient mice is observed by postnatal day 3, though they have normal tails at birth. The close similarity between Fgfr3-deficient and Cfr-deficient mice further indicates that Cfr is positively involved in the FGF18-FGFR3 signaling pathway. Next, I addressed binding of Cfr to FGF18 by immunoprecipitation assay, because previous reports showed binding of Cfr to only FGF1, FGF2, FGF3, and FGF4 so far. I found that recombinant FGF18 fused to GST tag (GST-FGF18), but not GST tag alone was also coimmunoprecipitated with Cfr-EC, demonstrating that Cfr and FGF18 interact not only genetically but also physically. The physical interaction of Cfr with FGF18 may be necessary for their cooperative function.

Above results indicated that Cfr is a Dlk-binding molecule that positively regulates the FGF18 signaling pathway and Cfr binds to both Dlk and FGF18 in its extracellular domain. In addition, the similar phenotypes of Cfr-deficient and Dlk-transgenic mice suggest that Dlk and Cfr function in the opposite directions. Therefore, I suspected that Dlk could affect the physical interaction between Cfr and FGF18. To address this possibility, I examined the binding of Cfr and FGF 18 in the presence or absence of Dlk. I expressed Cfr-EC along with Dlk-Fc or CD4-Fc as a control in COS7 cells, and assessed binding of Cfr-EC to GST-FGF 18 in the culture supernatants by immunoprecipitation assay. GST-FGF18 was not coimmunoprecipitated with Cfr-EC when Dlk-Fc was coexpressed with Cfr-EC, while CD4-Fc had no effect. From these results, I concluded that one of molecular functions of Dlk is to inhibit the functions of Cfr and FGF 18 by interrupting their physical interaction.

In conclusion, by identification of Cfr as a binding partner molecule of Dlk and demonstration of a positive regulatory role of Cfr for the FGF18 signaling which is inhibited by Dlk, this study has revealed a novel regulatory mechanism for the FGF signaling pathway by Cfr and Dlk.

Figurel. Structure of Dlk and Cfr.

本論文は4章からなる。第1章は、イントロダクションであり、本論文で解析を行ったCysteine-rich fibroblast growth factor receptor (Cfr)、Delta-like protein (Dlk)、Fibroblast growth factor (FGF)、及びFibroblast growth factor receptor (FGFR)の、これまでに明らかとなっている分子的特徴と機能が述べられている。また、本研究でDlkの機能解明を目指してDlk結合分子を探索しCfrを同定したこと、またCfrが生理的機能の不明なFGF受容体の一種であったためFGFシグナルに着目してCfrとDlkの機能解明を目指したことについて言及し、本研究の目的を示している。

第2章は本研究で用いた実験材料及び方法について述べられている。

第3章は本研究から得られた結果が述べられている。第1節では発現クローニング法によるDlk結合分子の探索の過程と、その結果としてのCfrの同定について述べられている。DlkとCfrの特異的な結合を免疫沈降実験により確認し、さらに両者が共にゴルジ体に強く局在することを示してDlkとCfrの機能的関連を示唆している。これまでに知られていなかったDlkの新たな結合分子を同定できたことに大きな意義がある。

第2節ではCfr遺伝子の機能欠損マウスの作製の過程とその表現型について述べられている。作製したマウスが生育遅延、尾の形成異常および口蓋裂などの発生異常を生じ、約90%が生後2日以内に死亡することが示されている。Cfrの遺伝子機能欠損マウスはこれまでに報告がなく、Cfrの生理的機能を初めて明らかにした点が重要な成果である。また、このマウスの表現型がこれまでに報告のあるFgf18遺伝子の機能欠損マウスおよびDlkの過剰発現マウスの表現型と類似していることにも言及し、これらの分子の間の機能的関連を推測している。

第3節では、Cfr遺伝子の機能欠損マウスの表現型から推測されたCfr、DlkおよびFGF18の分子間の機能的関連を実際に検討した結果について述べられている。まずCfrとFgf18の遺伝学的相互作用をCfrへテロマウスとFgf18へテロマウスをかけ合わせ、その仔を観察することで検討している。その結果CfrとFgf18のダブルへテロマウスでのみCfr遺伝子機能欠損マウスやFgf18遺伝子機能欠損マウスなどで認められるような、尾の形成異常を示すことを明らかにしている。したがって、CfrとFgf18が遺伝学的相互作用を持ち、協調的に機能するということを示している。そこでCfrとFGF18の物理的な結合を検討し、両者が結合することも明らかにしている。CfrとFGF18の相互作用はこれまでに報告がなく、新しい知見である。さらに、DlkがこのCfrとFGF18の結合を阻害することも示している。これらの結果から、DlkがCfrとFGF18の結合を阻害し、両者の協調的機能を抑制していると推測される。この知見により、これまでほとんど明らかになっていなかったDlkの分子機能の少なくとも一部を明らかにしている。またCfrとFgf18の両遺伝子の機能を欠損させたマウスを作製することにより、CfrとFgf18の遺伝学的相互作用をさらに検討している。その結果、両方の遺伝子機能を欠損させたマウスでは、CfrとFgf18のそれぞれの遺伝子機能を欠損させたマウスよりも強い表現型が認められた。したがってCfrとFGF18の対応は1対1ではなく、他にもCfrの機能に関与するシグナル経路が存在することが示されている。

第4章では得られた結果に基づいた考察を行っている。CfrとFgf18との遺伝学的相互作用が認められたことと両者が物理的に結合することから、CfrとFGF18が協調的な機能を持つと考えられる。しかしDlk存在下ではCfrとFGF18の結合がDlkによって阻害され、両者の協調的な機能をDlkが抑制するというモデルを提唱している。このモデルによるとCfrとDlkが適切な時期と部位に発現することにより、FGF18シグナルの強弱が調節され、正常な発生が進行していくと考えられる。これは今までにない全く新しいFGFシグナルの制御機構であり、大変意義深い。また、これまで不明な点が多かったCfrとDlkの機能として、FGFシグナルの制御という新たな役割を発見したという点も大きな成果である。

なお、本論文第3章は、田中 稔・高田 慎治・宮島 篤との共同研究であるが、論文提出者が主体となって実験、解析及び検証を行ったもので、論文提出者の寄与が十分であると判断する。

したがって、博士(理学)の学位を授与できると認める。