Molecular cloning of POEM: a novel adhesion molecule that interacts with alpha8beta1 integrin.

Cell adhesion molecules are involved in a number of biological functions, such as cell survival, cell differentiation, tissue repair, and development. A novel molecule, POEM (preosteoblast epidermal growth factor-like repeat protein with meprin, A5 protein, and receptor protein-tyrosine phosphatase μ domain), was isolated by reverse transcription-polymerase chain reaction using a set of degenerate primers designed after other known epidermal growth factor (EGF)-like motifs. From its structure, POEM was suggested to be a novel adhesion molecule with five EGF-like domains, an Arg-Gly-Asp (RGD) cell binding motif, and a meprin, A5 protein, and receptor protein-tyrosine phosphatase μ (MAM) domain. By in situhybridization using embryonic day 16.5 (E16.5) mouse embryos, strong expression of POEM mRNA was observed in developing kidney renal tubules, parathyroid and thyroid glands, developing bone, tooth germ, and endocrine organs of the brain. The inner ear, skeletal muscle, smooth muscle (except for the vascular system), and skin were also positive for POEM expression. Bacterial recombinant POEM protein containing the RGD sequence and MAM domain showed strong cell adhesion, spreading, and survival-promoting activities. By mutating the RGD sequence to RGE, the cell spreading and survival activities were significantly decreased, but the MAM domain was shown to contribute only to cell adhesion and not to cell spreading and survival-promoting activities. The distribution of POEM in several tissues was close to that of α8β1integrin. Therefore, we conducted cell adhesion assays using KA8 cells, a K562 leukemia clone stably expressing α8integrin. Parental K562 cells, which expressed α5β1 integrin, bound to fibronectin but not to POEM. On the other hand, KA8 cells showed strong binding and spreading on both fibronectin and POEM. These results suggest that POEM is a novel ligand for α8β1integrin and that POEM may be involved in the development and function of various tissues, such as kidney, bone, muscles, and endocrine organs.

The epidermal growth factor (EGF) 1 -like repeat structure is present in a number of extracellular matrix (ECM) proteins and cell surface receptors (1). Fibrillin (2) and laminin (3) are typical ECM proteins with multiple EGF-like repeats. Notch and Delta are cell surface molecules that act as a receptor and ligand, respectively, and interact via their EGF-like domains to determine cell fates (4). Transforming growth factor ␣ (5) and heparin-binding EGF-like growth factor (6) are generated by cleavage of transmembrane precursors and act through the EGF receptor. Recently, small matrix proteins with several EGF-like repeat structures have been reported. For example, Pref-1 is produced as a membrane protein and controls adipocytic cell differentiation (7). Del-1 is an adhesion molecule with an Arg-Gly-Asp (RGD) sequence and induces angiogenesis (8). DANCE is another secreted molecule with an RGD cell binding motif; it is expressed predominantly in developing arteries and supposedly acts as a cell adhesion molecule in tissue development and repair (9).
Interactions between cells and ECM play roles in morphogenesis, tissue homeostasis, and remodeling. The ECM presents much information to the cells by working as a multifunctional ligand for cell adhesion receptors (10,11). Integrins are a large family of heterodimeric cell surface proteins that serve as receptors for various ECM proteins (12)(13)(14). Integrins are involved in tissue repair, development, and immune responses. Integrins also serve important functions in bone development and remodeling. Recently, it was reported that ␤ 1 integrin was significantly involved in osteoblastic function (15). Expression of dominant negative ␤ 1 subunit in osteoblasts significantly reduced the bone-forming activity of osteoblasts. Matrix proteins produced by osteoblastic cells are the major target molecules not only for osteoblasts themselves but also for osteoclastic cells. Therefore, identification of a novel adhesion molecule produced by osteoblastic cells provides a new insight into the biology and development of bone tissue.
In this study, we screened a cDNA library derived from preosteoblastic MC3T3-E1 cells and isolated a cDNA clone coding for a novel adhesion protein, POEM.

MATERIALS AND METHODS
Cell Lines and Culture-A cell population enriched in osteoblasts was prepared from the calvaria of a 4-week-old C57BL/6 mouse as described previously (16). MC3T3-E1, a mouse calvaria-derived osteoblast-like cell line, was maintained in ␣-modified essential medium * This work was supported by Research Fellowships of Japan Society for the Promotion of Science for Young Scientists, Sumitomo Marine Welfare Foundation, Kanae Foundation, the Ministry of Health and Welfare of Japan, and the Ministry of Education and Science of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ/GenBank TM /EBI Data Bank with accession number(s) AB059656.
(Life Technologies, Inc.). COS-7 cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.). KA8 cells, which express the chicken integrin ␣ 8 subunit (17), and control K562 cells were kindly provided by Dr. Louis F. Reichardt (University of California, San Francisco, CA). KA8 cells were cultured in RPMI 1640 medium containing 400 g/ml G418 (Calbiochem-Novabiochem Corp., San Diego, CA), and K562 cells were cultured in the same medium but without the G418. Each medium was supplemented with 10% fetal calf serum (Roche Molecular Biochemicals) and antibiotic antimycotic solution (Life Technologies, Inc.). Cells were incubated at 37°C in a humidified atmosphere of 5% CO 2 in air.
cDNA Cloning and Sequencing-A pair of degenerate primers used for polymerase chain reaction (PCR) amplification was designed based on the conserved amino acid sequences found in EGF-repeat structures (18). Complementary DNA was synthesized from the mRNA of MC3T3-E1 cells and amplified by use of a reverse transcription-PCR kit (Stratagene, La Jolla, CA). The PCR program consisted of 30 cycles of 94°C for 1.5 min, 48°C for 2 min, and 72°C for 2 min. The amplified PCR products were subcloned into pBluescript plasmid (Stratagene) and subjected to sequencing by the dideoxy-chain termination method with an automatic DNA sequencer (Model 373; Applied Biosystems, Foster City, CA). A cDNA library was constructed from the mRNA of MC3T3-E1 cells cultured for 6 days by using a UniZAP cDNA library construction kit (Stratagene).
A 270-base pair cDNA fragment of a gene that was not represented in the GenBank TM data base was obtained and used to screen a MC3T3-E1 (cultured for 6 days) cDNA phage library constructed in UniZAP (Stratagene). The single positive plaque was subjected to in vivo excision of the pBluescript SK(Ϫ) phagemid as described by the manufacturer. A homology search was performed with BLAST and FASTA against public sequence data bases, and a motif search was performed on-line with PROSITE. The nucleotide sequence data reported in this study will appear in the DDBJ/EMBL/GenBank TM nucleotide sequence data base with the accession number AB059656.
RNA Preparation and Northern Blotting Analysis-Total RNA was prepared from brain, kidney, liver, spleen, testis, and thymus of 4-weekold mice. In addition, 14.5, 16.5, and 18.5 days post coitus whole embryos and newborn mice were also used for RNA preparation. The total RNA was prepared by the acid guanidinium thiocyanate-CsCl method (19) or with ISOGEN (Nippon Gene, Tokyo, Japan). Polyadenylated RNA samples were extracted from cells by using a Quick Prep Micro mRNA Purification Kit (Amersham Pharmacia Biotech). The RNA samples were electrophoresed in 1.2% agarose gel and then transferred to nylon membranes (Hybond-N ϩ ; Amersham Pharmacia Biotech). The blots were hybridized with radiolabeled probes and washed twice in 2ϫ SSC, 0.1% SDS at room temperature followed by 0.2ϫ SSC, 0.1% SDS at 55°C. The equivalent loading of the RNA samples was confirmed by probing with a human ␤-actin probe.
In Situ Hybridization-Both sense and antisense probes were labeled by transcription from full-length mouse POEM cDNA with a digoxigenin RNA labeling kit (Roche Molecular Biochemicals). The labeled probes were alkaline-hydrolyzed and then purified.
Kidney, liver, and spleen were collected from 22-day-old mice, and spleen was collected from adult mice. These specimens and 16.5-day-old embryos were fixed in 4% paraformaldehyde, embedded in Tissue-Tec OCT (Miles, Inc., Elkhart, IN), frozen in a liquid nitrogen, and cryosectioned at 8-m thickness. Three serial sections were made (two for hybridization with sense or antisense probe and one for staining with hematoxylin and eosin).
In situ hybridization with digoxigenin-labeled sense and antisense RNA probes was performed as described previously with a slight modification. In brief, sections were permeabilized with proteinase K (3 g/ml) in phosphate-buffered saline (PBS) for 30 min at 37°C and hybridized with probes (0.3 g/ml) overnight at 45°C. The sections were then blocked with 1% blocking buffer, incubated with a 1:750 dilution of anti-digoxigenin antibody for 1.5 h at room temperature, and developed for 6 h at room temperature in coloring solution (nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate) containing levamisol (10 g/ml).
Construction of Expression Vectors and Immunoblotting-A cDNA fragment (nucleotide positions 144-1931 as shown in Fig. 1) encoding amino acid residues from 1-561 of POEM was amplified by PCR amplification using primers 5Ј-CCCAAGCTTTGTGCGTCCCAGCAGCCGCGG-3Ј (forward direction) and 5Ј-GAAGATCTGCAGCGACCTCTTTT-3Ј (reverse direction). Another cDNA fragment corresponding to nucleotide positions 144-1511 (amino acids 1-421; POEM⌬MAM) of POEM was also amplified using the same forward primer and reverse primer 5Ј-GAAGATCTGCTGT-GTATGAGAATACCTG-3Ј. PCR products were digested with HindIII and BglII and used to replace the osteopontin signal sequence and mouse CTLA4 cDNA of expression vector pcDNA3-OSM-mCTLA4-Ig (kindly provided by Dr. R. Abe; Science University of Tokyo, Tokyo, Japan) to generate expression vectors for fusion proteins of POEM and human immunoglobulin G (IgG) constant region. A single amino acid substitution (RGD3RGE) was introduced into POEM cDNA by PCR using a primer containing a single nucleotide mutation. The sequence of the primer was 5Ј-CGCGGATCCT-CAGAAACCCAGAGGAGAAGTG-3Ј. As a negative control, pcDNA3 mock vector was used. COS-7 and MC3T3-E1 cells were transfected with the expression vectors by using TransIT-LT1 (Mirus, Madison, WI) according to the manufacturer's protocol. Two days later, conditioned medium of these cells was concentrated with Protein G Fast Flow (Amersham Pharmacia Biotech). In the COS-7 cells, the cells were harvested with 1 mM EDTA in PBS, and cell samples were suspended at a concentration of 2 ϫ 10 5 cells/20 l Laemmli reducing SDS loading buffer. ECM samples were stripped from the cell culture dish with a small volume of Laemmli buffer after the cells were removed. The MC3T3-E1 samples containing the cells and ECM proteins were harvested with Laemmli reducing SDS loading buffer. These samples were separated by 7.5% SDS-polyacrylamide gel electrophoresis. Proteins were blotted onto Immobilon-P (Millipore, Bedford, MA) and blocked with 3% skim milk in Tris-buffered saline for 2 h. Proteins were directly detected with a horseradish peroxidase-conjugated goat anti-human IgG antibody (Southern Biotechnology Associates, Birmingham, AL). The membrane was exposed to x-ray film with chemiluminescence enhancement (Amersham Pharmacia Biotech). , and MBP proteins were produced by Escherichia coli strain XL1-Blue transformed with pMAL-c2-POEMc, pMAL-c2-POEMc(RGE), or pMAL-c2, respectively. The recombinant proteins were purified by amylose resin column chromatography and analyzed by Coomassie Brilliant Blue staining after SDS-polyacrylamide gel electrophoresis. The protein concentration was determined with a Coomassie Plus Protein Assay Reagent (Pierce) using bovine serum albumin as a standard.
Non-tissue culture 96-well plates were coated with 1-5 g of recombinant MBP-POEMc, MBP-POEMc(RGE), or MBP proteins diluted with PBS and incubated for 24 h at 4°C. Bovine fibronectin (Sigma) was used as a positive control, and MBP and bovine serum albumin were used as negative controls. Coated wells were washed with PBS and blocked for 1 h at 37°C with a solution of 10 mg/ml bovine serum albumin in PBS. Subconfluent MC3T3-E1 cells were harvested by trypsin-EDTA, suspended in serum-free medium (3 ϫ 10 4 cells/100 l), and placed in the wells. After incubation for 90 min at 37°C, the wells were washed several times with serum-free medium until almost all of the cells had been washed out of the negative control wells. Attached cells were fixed and stained with 0.1 ng/ml 4Ј,6-diamidino-2-phenylindole dihydrochloride (Roche Molecular Biochemicals) in methanol, and the cell number was counted as the number of fluorescent nuclei seen under a fluorescence microscope. Experiments were performed in duplicate.
For the cell adhesion assay using K562 and KA8 cells, glass-bottomed dishes (Mastunami, Osaka, Japan) were coated with recombinant proteins, bovine fibronectin, or MBP at a concentration of 5 g/ml at 4°C overnight and then blocked as described above. These cells were then plated at a concentration of 3 ϫ 10 4 cells/well, incubated for 20 min at 37°C, and observed under a phase-contrast microscope.
The reverse transcription-PCR amplification was conducted as described above using cDNA of MC3T3-E1 cells as a template. The PCR program consisted of 30 cycles of 94°C for 1 min, 55°C for 2 min, and 72°C for 3 min. The amplified PCR products were electrophoresed on a 2% agarose gel and visualized with ethidium bromide staining.

Molecular Cloning of a Gene Encoding a Novel EGF-like
Repeat Protein from MC3T3-E1-MC3T3-E1 is a cell line with preosteoblastic phenotypes. We generated a pair of degenerate PCR primers based on the EGF-like repeat structures found in various ECM and receptor proteins (18). Using a single-strand cDNA of MC3T3-E1 (cultured for 3 days) as a template, we obtained a clone, OB-9, whose sequence was not found in the GenBank TM data base. By screening a MC3T3-E1 cDNA library with OB-9 partial cDNA as a probe, we isolated four independent and overlapping clones. By restriction enzyme mapping and DNA sequencing, we isolated a full-length 3.6kilobase cDNA with an open reading frame of 561 amino acid residues (Fig. 1A). The translation initiation site, Met, was assigned at nucleotide positions 249 -251 according to the Kozak consensus sequence (21). The deduced amino acid se-quence of this protein had an N-terminal hydrophobic domain, which was presumed to be a signal sequence (22), but no other hydrophobic regions that could serve as transmembrane domains were found, suggesting that OB-9 is a secreted protein.
Additionally, five EGF-like repeat domains, a proline-rich domain, one putative Asn glycosylation site, an RGD cell adhesion sequence motif, and a MAM domain were found. Therefore, we named this potential protein POEM (preosteoblast EGF-like repeat protein with MAM domain).
The alignment of the five EGF-like repeats and MAM domain of POEM showed significant similarity with other known proteins, as shown in Fig. 1, B and C. The EGF-like repeats were located at the N terminus in tandem following the signal sequence. The MAM domain was about 160 amino acids long and had four conserved cysteines and hydrophobic and aromatic amino acids matching the consensus sequence of MAM domains found in other proteins.
The overall POEM sequence showed significant similarity to mouse MAEG (GenBank TM accession number AJ245672) or its human counterpart, EGFL6 (GenBank TM accession number AF186084), which contains a predicted signal sequence, an array of five EGF-like repeats, an RGD motif, and a MAM domain (Fig. 2). However, homology was about 39% at the amino acid sequence level, and MAEG/EGFL6 did not contain the proline-rich domain found in POEM. Recently, we found a putative human counterpart of POEM (hCG20771, hCP38387) in the Celera data base (23) that shows ϳ80% homology at the amino acid level with the corresponding region of mouse POEM. From these facts, we concluded that POEM and MAEG/ EGFL6 are distinct molecules belonging to same family.
POEM mRNA Expression in MC3T3-E1 Cells and in Mouse Embryos-In long-term cultures, MC3T3-E1 cells spontaneously differentiate into cells with mature osteoblastic phenotype. To gain further insight into POEM function, we used Northern blotting to examine the expression of POEM mRNA in MC3T3-E1 cells, mouse tissues, and mouse whole embryo. Approximately 4-kilobase POEM mRNA was highly expressed at the proliferation stage of MC3T3-E1 cells (Fig. 3A). Subsequently, the expression decreased as the differentiation proceeded. Expression of POEM was also detected in primary calvarial osteoblastic cells and kidneys of 4-week-old mice by Northern blotting (data not shown). In mouse development, expression of POEM mRNA could be detected as early as day 11.5 of gestation (data not shown) and subsequently increased (Fig. 3B), suggesting that POEM is involved in embryonic development as well as in the postnatal function of bone and kidney.
Tissue Distribution of POEM mRNA in Mouse Embryos-To investigate the expression of POEM in developing tissues, we performed in situ hybridization using frozen sections of 16.5day-old embryos.
In the kidneys, abundant expression of POEM mRNA was observed in most of the epithelium of the ureter branches, the epithelium of collecting tubules, and the podocytes surrounding glomeruli (Fig. 4, A and B). Weak or less intense expression was also observed in both the comma-shaped bodies and the S-shaped bodies toward the peripheral region of the kidney.
Hybridization of tissue sections with the sense probe of POEM showed no positive reaction (Fig. 4A). In 22-day-old mice, POEM expression was also observed in many tubules of the kidney (data not shown).
In the parathyroid gland, remarkable expression of POEM mRNA was observed in the developing parenchymal cells of cell cords (Fig. 4, CϪE). However, in the capsule, no expression was detected (Fig. 4, C and D).
Thyroid tissue of 16.5-day-old embryos showed disordered cords of cells without obvious organization into follicles or the presence of colloid. However, some follicular structures were becoming clearly defined at the same time (Fig. 4E). The relatively high level of POEM mRNA was restricted to the developing follicular epithelial cells and potential parafollicular cells of the thyroid gland (Fig. 4, C and D).
A significant level of POEM mRNA was also detected in the tooth germs (Fig. 4, FϪH). POEM mRNA expression was detected in the outer enamel epithelium. Higher expression was frequently observed in the outer part on the frontal side of the outer enamel epithelium and its stalk region. The entire region of the dental papilla was negative. POEM expression was also observed in some of the inner enamel epithelium and in the stellate reticulum.
In the intramembranous ossification sites, such as the ramus of the mandible (Fig. 4, IϪK), the membranous primordium of the frontal bone, and calvaria, osteoblasts were identified adjacent to the mineralized tissues as cuboidal cells. Weak expression of POEM mRNA was detected in these mature osteoblasts lining the bone matrix, however, more intensive expression was observed in the condensed mesenchymal cells, surrounding developing bones in both the ramus of the mandible (Fig. 4, I and J) and the calvaria (data not shown). In the case of the intracartilaginous ossification site in vertebral bones, weak expression of POEM mRNA was detected both in cuboidal osteoblasts on the trabecular bone surface and in cells surrounding the bony collar showing the early stage of osteogenesis (Fig. 5, AϪC).
POEM mRNA was also detected in the cells covering the outer side of the Eustachian tube and tubo-tympanic recess (Fig. 5, DϪF). In the cochlea, two types of cells were recognized: support cells and sensory epithelium (hair cells; Fig. 5I). POEM mRNA was localized apically in some of the developing hair cells next to the cluster of mature hair cells as seen in a sagittal in the plane (Fig. 5, G and H).
POEM was widely expressed in the skeletal muscles (Fig. 5,  JϪO). Generally, the expression was relatively high around the nucleus in the skeletal muscles and near the tendon. POEM expression was not detected in the tendon. In the lingua (Fig. 5,  MϪO), POEM was detected not only in the lingual muscle but also in the mucosal epithelium (Fig. 5, M and N). In the lingual muscle, POEM expression was also high around the nucleus, especially in the region just under the lingual aponeurosis and near the lingual septum, where the muscle made contact. In the mucosal epithelium, a relatively low level of POEM was expressed and restricted to the upper surface of the lingua. POEM expression could not be observed in the lingual aponeurosis or lingual aorta.
POEM expression was detected in the smooth muscle as well as in the skeletal muscle. In the digestive tract, POEM was expressed in the muscle layer of both the esophagus (data not shown) and stomach (Fig. 5, PϪR). In the esophagus, expres-sion was observed in both the inner and outer muscle layers. On the other hand, abundant expression was localized restrictedly in the inner muscle layer of the stomach (Fig. 5, P and Q). Through the remaining part, i.e. from the duodenum to the rectum, expression was barely detected, even in the muscle layer. The smooth muscle layer under the skin (Fig. 6, JϪL) and the trachea (Fig. 6, PϪR) were also positive for POEM mRNA. Vascular smooth muscle of arteries and veins showed no expression of POEM mRNA. The expression was not detected in the endothelial cells of blood vessels either.
In the brain, expression of POEM was detected in the pineal body (Fig. 6, AϪC), choroid plexus, hypophysis (Fig. 6, DϪF),  (Fig. 6, GϪI), midbrain, pons, and medulla oblongata. Obvious POEM expression was localized in the cells of the pineal body (Fig. 6, A and B) and dorsal region of the choroid plexus within the pineal recess of the third ventricle. In addition to this, relatively weak expression was observed in the other regions of the choroid plexus. In the hypophysis (Fig. 6, DϪF), POEM-positive nerve fibers were localized in the entire region of the neurohypophysis. POEM expression was also observed in the cells in the adenohypophysis and in a part of the pars intermedia, and the expression level was relatively high in the cells near the lumen. In some of the regions of the parenchyma, POEM expression was detected in large cell bodies, small cell bodies, and their fibers. These large cell bodies were identified as neurons, but it was hard to determine whether small cell bodies were neurons or glial cells. In the diencephalon, POEM expression was distributed in the pretectal area, in the posterior thalamus, in the medial and lateral regions of the periaquaductal gray (Fig. 6, GϪI), in the medial hypothalamus, and in the caudal portion of the lateral hypothalamus. In the midbrain, POEM expression was restricted to the cell bodies and their fibers of the superior colliculus. In the dorsal portion of pons, POEM was expressed in cell bodies and their fibers and distributed from the rostral region to the caudal region. Scattered expression was also detected in the medulla oblongata.
In the skin, expression of POEM mRNA was located in the basal layer (Fig. 6, JϪL). The most abundant expression was observed in the basal layer of the glans clitoris (Fig. 6, MϪO), and the expression level decreased in the prepuce. In the clitoris, POEM expression was also observed in the corpus cavernosum clitoridis, and the expression was higher in the glans clitoridis than in the prepuce (Fig. 6, M and N). In the respiratory system, POEM expression was localized in the epithelial cells surrounding the lumen of the nasopharynx and trachea (Fig. 6, PϪR). In the lung, expression was observed in the cells surrounding the developing alveoli (data not shown).
In the 16.5-day-old embryo, POEM mRNA was hard to detect in the spleen, and few cells, if any, were faintly stained. However, in 22-day-old mice, POEM was weakly expressed in unidentified cells present in the white pulp (data not shown). In adult mice, expression could not be detected in the spleen.
No obvious POEM expression was detected in the liver, pancreas, adrenal gland, blood-vascular system, atrium, ventricle, ovary, cartilage, thymus, spinal cord, or ganglion in the 16.5day-old embryo. In 22-day-old mice, POEM expression was not observed in the liver.
Subcellular Localization of POEM Protein-Most of the EGF-like repeat-containing proteins reported thus far are extracellular or membrane-bound proteins. To characterize POEM, we generated an expression vector for a fusion protein of POEM and human immunoglobulin constant region (POEM-Fc) and transfected COS-7 and MC3T3-E1 cells with it for transient expression (Fig. 7).
For examination of the localization of POEM-Fc in these cells, Western blotting was performed on the conditioned medium, cell extract, and ECM fraction of COS-7 cells transfected with the POEM-Fc expression vector (Fig. 7A). An 80-kDa POEM-Fc protein was detected in both cell and ECM fractions. Very little POEM-Fc protein was found in culture medium, suggesting that POEM-Fc protein was trapped on the cell surface and/or in ECM. Interestingly, the absence of the RGD cell binding motif did not seem to affect the subcellular localization. Additionally, a significant portion of the POEM protein seemed to be dimerized or trimerized, as suggested by nonreducing gel electrophoresis and Western blotting of each fraction (data not shown).
When expressed in MC3T3-E1 cells, POEM-Fc could hardly be detected in the medium fraction, as observed for the COS-7 cells. In the cell/ECM fraction, a significant amount of the POEM-Fc protein showed a smaller size than that in COS-7 cells (Fig. 7B), suggesting that POEM-Fc was processed by some unidentified protease in MC3T3-E1 cells.
To determine the cell binding domain of POEM, we performed FACS analysis using COS-7 cells transfected with expression vectors for POEM-Fc fusion proteins (Fig. 8). Wildtype POEM-Fc protein was localized on the cell surface, and the cell binding activity was not affected by the mutation introduced into the RGD sequence. However, the binding activity was significantly reduced by the deletion of the MAM domain, suggesting that this domain plays an important role in the localization of POEM at the cell surface.
RGD Sequence of POEM Promotes Cell Attachment and Spreading via ␣ 8 Integrin-To assess the role of the RGD motif and MAM domain in the cell adhesion and spreading activity, we produced recombinant MBP fusion protein of POEM C terminus portion (POEMc; amino acids 377-561) in E. coli and purified it by amylose resin column chromatography until it appeared as a single band by Coomassie Brilliant Blue staining on SDS-polyacrylamide gel electrophoresis gel (Fig. 9A). MBP-POEMc was detected at the expected size of 60 kDa. MC3T3-E1 cells adhered to and spread on MBP-POEMc protein and fibronectin in a dose-dependent manner (Fig. 9, B and C). To investigate the role of the RGD motif in this cell adhesion activity, we generated recombinant protein MBP-POEMc(RGE) encoding a mutant form of MBP-POEMc with a substitution of the RGD motif by RGE (Fig. 9A). Significantly less cell adhesion activity was observed with MBP-POEMc(RGE) (Fig.  9B). However, MBP-POEMc(RGE) stimulated cell attachment at a higher concentration (Ͼ5 g/ml) without cell spreading activity (Fig. 9C), consistent with the result of FACS analysis (Fig. 8). When the culture period was prolonged to 2 days in the absence of serum, MC3T3-E1 cells could survive on either MBP-POEMc or fibronectin, whereas the cells plated on MBP-POEMc(RGE) showed an apoptotic phenotype with condensed nuclei (Fig. 9C). These results indicate that not only the MAM domain but also the RGD motif in MBP-POEMc protein plays an important role in adhesion, spreading, and survival of MC3T3-E1 cell.
We suspected POEM to be a ligand for the ␣ 8 ␤ 1 integrin receptor based on the following information. (a) The RGD motif is known as an ␣ 8 ␤ 1 integrin-binding motif. (b) We confirmed that ␣ 8 integrin subunit was expressed in MC3T3-E1 cells by reverse transcription-PCR (data not shown). Furthermore, the expression is diminished with differentiation during primary osteoblasts cultured (24). This expression profile is similar to that of POEM observed in MC3T3-E1 cells (Fig. 2). (c) The mRNA expression profile of POEM in mouse embryos was similar to that of ␣ 8 ␤ 1 integrin in chick embryos (25). (d) Localization of POEM in kidney and bone tissues was almost identical to that of postulated ligands for ␣ 8 ␤ 1 identified as binding sites of ␣ 8 ␤ 1 -alkaline phosphatase fusion protein (26). To confirm this possibility, we carried out a cell adhesion assay using KA8 cells, a K562 clone stably transfected with an ␣ 8 integrin expression vector and expressing both ␣ 5 ␤ 1 and ␣ 8 ␤ 1 integrins (17). Parental K562 cells, which were used as a control, express only ␣ 5 ␤ 1 . KA8 cells adhered remarkably to both MBP-POEMc and fibronectin (Fig. 10). In contrast, K562 cells adhered only to fibronectin. Neither KA8 cells nor K562 cells attached to MBP-POEMc(RGE). These observations suggest that POEM is a specific ligand for ␣ 8 ␤ 1 integrin, whereas fibronectin is a multiligand for both ␣ 8 ␤ 1 and ␣ 5 ␤ 1 integrins, as reported previously (17,27).

DISCUSSION
After the molecular cloning of the chick integrin ␣ 8 subunit, a number of biological functions of this molecule were reported. Mice deficient in ␣ 8 showed severe defects in kidney morphogenesis (28) and disorganization of inner ear cells (29). We found POEM to be abundantly expressed in kidney cells, especially in the glomerulus and epithelial cells of the ureter. This expression continued after birth, and these results indicate that POEM is involved in kidney morphogenesis and function. Expression in the inner ear is another unique feature of POEM. Integrin molecules are usually expressed as heterodimers of ␣ and ␤ subunits. The ␣ 8 integrin subunit associates with ␤ 1 subunit (25). Thus far, fibronectin, vitronectin, tenascin-C, and osteopontin have been shown to be ligands for ␣ 8 ␤ 1 integrin (17,26,30,31). Among these ligands, binding of ␣ 8 ␤ 1 integrin to vitronectin showed great variability, depending on the preparation of the protein (17). Denda et al. (27) generated a soluble ␣ 8 ␤ 1 integrin heterodimer-alkaline phosphatase fusion protein to search for ligands of ␣ 8 ␤ 1 integrin. Far Western blotting of mouse embryo extracts with the ␣ 8 ␤ 1 -alkaline phosphatase fusion protein suggested the presence of several ligand molecules with molecular masses ranging from 65-85 kDa. These investigators also stained developing kidney and bone tissues with ␣ 8 ␤ 1 -alkaline phosphatase and found that localization of ligand molecules was similar to that of osteopontin. However, in a recent study (32), mice deficient in osteopontin did not show any significant defects in kidney morphogenesis, suggesting the presence of other ligand molecules for ␣ 8 ␤ 1 integrin. Very interestingly, the predicted size of POEM and the distribution of POEM mRNA are quite similar to those of the ligand molecule suggested by using ␣ 8 ␤ 1 -alkaline phosphatase, and these facts indicate that POEM is a novel ligand molecule for ␣ 8 ␤ 1 integrin. We confirmed this possibility by using KA8 cells expressing ␣ 8 ␤ 1 integrin, and we also found that K562 cells, which express ␣ 5 ␤ 1 (a fibronectin receptor), did not bind to POEM. Therefore, although the binding of POEM to other integrins has not yet been well investigated, POEM seems to have a unique binding activity distinct from that of other ligand molecules for ␣ 8 ␤ 1 integrin. From these results, we conclude that POEM is one of the important ligand molecules for ␣ 8 integrins.
By amino acid alignment and data base search, POEM was found to encode several potential functional domains, such as five EGF-like repeats and a MAM domain, as well as an RGD integrin binding motif. These structures are commonly found in secreted proteins; however, by Western blotting and immunocytostaining, we found that the majority of the POEM proteins were localized on the cell surface or in the ECM, but not in the culture medium. Interestingly, this localization was not affected by the mutation of the RGD cell adhesion motif. This finding raised the possibility that POEM may interact not only with RGD-dependent ␣ 8 ␤ 1 integrin but also with other cell surface molecules, including other integrins and/or other family receptor proteins. The FACS analysis showed that a mutant POEM-Fc molecule without the MAM domain detached from the cell surface. Moreover, this mutant POEM-Fc molecule without the MAM domain was detected mainly in the culture medium rather than in cell extracts when expressed in COS-7 cells (data not shown). Although there is a possibility that the deletion of the MAM domain affected the entire structure of POEM protein, these results suggest that the MAM domain plays a significant role for cell surface localization of POEM as well as the RGD motif in COS-7 cells. It has been reported that the MAM domain family, including meprin, A5 protein, neuropilin-1, neuropilin-2, and receptor protein-tyrosine phosphatases, mediates cell adhesion activities via homo-or heterophilic MAM domain interactions (33)(34)(35)(36)(37). This fact supports the hypothesis that MAM domain is involved in the cell surface binding via protein-protein interaction, and these molecules could serve as candidate receptor molecules for POEM. Additionally, MC3T3-E1 cells bound to MBP-POEMc(RGE), but K562 and KA8 cells did not. These results also suggest that the binding molecule for the MAM domain of POEM is expressed in a cell type-specific manner. The identification of this molecule will help us to prove the precise function of POEM.
In the developing mouse embryo, POEM was expressed in the endocrine organs (parathyroid gland, thyroid gland, hypophysis, and pineal organ). These endocrine organs are closely related to growth, bone metabolism, and calcium and phosphorus homeostasis. In the mouse embryo, POEM mRNA showed a unique distribution in and around the developing bone, in the tooth germ, and in muscle. These data also suggest the relationship between POEM and calcium metabolism. Osteoblastic cells produce a number of ECM proteins. For example, type I collagen and osteopontin are up-regulated after osteoblast maturation (38). On the other hand, we noted that the expression of POEM was down-regulated as osteoblastic cell differentiation proceeded. The distribution of POEM expression in the mouse embryo also suggested the role of POEM in the early stage of osteoblastic cell differentiation. POEM has an RGD cell adhesion motif, which is known to interact with integrins (13,14). In this study, we found POEM to be a novel candidate ligand molecule for ␣ 8 ␤ 1 integrin. In osteoclasts, integrins have been shown to play important functional roles by regulating cell attachment and bone resorbing activity (39). POEM was preferably expressed in preosteoblastic cells, which do not seem to interact with directly osteoclastic cells; however, integrins have been shown to play significant roles not only in osteoclasts but also in osteoblasts (15). Mice deficient in the ␤ 1 integrin subunit showed significantly less bone-forming activity. Moursi et al. (24) also reported that integrins, such as ␣ 5 ␤ 1 , ␣ 8 ␤ 1 , ␣ 3 ␤ 1 , and ␣ 4 ␤ 1 are critical for mineralized nodule formation and osteoblast differentiation. Therefore, POEM may play important roles in osteoblastic function by sending survival signals via ␣ 8 ␤ 1 integrin and mediating cell-cell interaction. POEM is a novel class of adhesion molecule with multiple motifs. Recently, a number of novel small EGF-like repeat proteins with RGD cell binding motifs have been reported (8,9,40,41). Each of these molecules shows a unique tissue distribution and integrin binding specificity. Further characterization of these molecules will lead to a better understanding of matrix proteins involved in tissue development and function.
After this manuscript was submitted, Brandenberger et al. (42) also reported detailed characterization of a novel ␣ 8 ␤ 1 integrin ligand named nephronectin. The amino acid sequence of nephronectin was completely identical with that of POEM. We believe that these findings will provide new insights into kidney morphogenesis and the development of other tissues expressing POEM/nephronectin.