Molecular Characterization of the MouseTem1/endosialin Gene Regulated by Cell Density in Vitro and Expressed in Normal Tissues in Vivo *

Human tumor endothelial marker 1/endosialin (TEM1/endosialin) was recently identified as a novel tumor endothelial cell surface marker potentially involved in angiogenesis, although no specific function for this novel gene has been assigned so far. It was reported to be expressed in tumor endothelium but not in normal endothelium with the exception of perhaps the corpus luteum. Here we describe the cDNA and genomic sequences for the mouseTem1/endosialin homolog, the identification and characterization of its promoter region, and an extensive characterization of its expression pattern in murine and human tissues and murine cell lines in vitro. The single copy gene that was mapped to chromosome 19 is intronless and encodes a 92-kDa protein that has 77.5% overall homology to the human protein. The remarkable findings are 1) this gene is ubiquitously expressed in normal human and mouse somatic tissues and during development, and 2) its expression at the mRNA level is density-dependent and up-regulated in serum-starved cells. In vitro, its expression is limited to cells of embryonic, endothelial, and preadipocyte origin, suggesting that the wide distribution of its expression in vivo is due to the presence of vascular endothelial cells in all the tissues. The ubiquitous expression in vivo is in contrast to previously reported expression limited to corpus luteum and highly angiogenic tissues such as tumors and wound tissue.

Human tumor endothelial marker 1/endosialin (TEM1/ endosialin) was recently identified as a novel tumor endothelial cell surface marker potentially involved in angiogenesis, although no specific function for this novel gene has been assigned so far. It was reported to be expressed in tumor endothelium but not in normal endothelium with the exception of perhaps the corpus luteum. Here we describe the cDNA and genomic sequences for the mouse Tem1/endosialin homolog, the identification and characterization of its promoter region, and an extensive characterization of its expression pattern in murine and human tissues and murine cell lines in vitro. The single copy gene that was mapped to chromosome 19 is intronless and encodes a 92-kDa protein that has 77.5% overall homology to the human protein. The remarkable findings are 1) this gene is ubiquitously expressed in normal human and mouse somatic tissues and during development, and 2) its expression at the mRNA level is density-dependent and up-regulated in serum-starved cells. In vitro, its expression is limited to cells of embryonic, endothelial, and preadipocyte origin, suggesting that the wide distribution of its expression in vivo is due to the presence of vascular endothelial cells in all the tissues. The ubiquitous expression in vivo is in contrast to previously reported expression limited to corpus luteum and highly angiogenic tissues such as tumors and wound tissue.
Angiogenesis, a tightly regulated formation of new blood vessels, plays an essential role in embryonic development, normal growth, wound healing, and other physiological processes (1)(2)(3). In addition, this process is critically involved in patho-logical conditions including tumor progression, where there is formation of new vessels that differ from normal vasculature in terms of both structure and function (4,5). It has become clear that there are quantitative differences between normal and tumor vasculature, evidenced by changes in the balance of angiogenesis inducers and inhibitors. However, it is less clear whether there are qualitative differences. In any case, the critical role of vascular endothelium in tumor growth has led to an intense search for potential targets for angiogenesis-based therapies (6,7).
TEM1, a novel gene that may play a role in angiogenesis, was recently identified by St. Croix and colleagues (8). They found it to be one of several genes elevated in human tumor endothelium when they carried out a comparison of gene expression profiles of endothelial cells derived from normal or malignant colorectal tissues. Furthermore, by in situ hybridization, they confirmed its expression in neoplastic colon tissue.
The same year, Christian et al. (9) reported the cloning of TEM1 and showed that it is identical to endosialin, a 165-kDa glycoprotein previously described as an antigen selectively expressed in vascular endothelial cells (10). Based on molecular analysis, TEM1/endosialin has been classified as a C-type lectin-like membrane receptor. Its extracellular portion contains five globular domains (C-type lectin domain, sushi-like domain, and three EGF 1 -like repeats) and a mucin-like region. The core protein is heavily sialylated and displays similarity to thrombomodulin, a receptor involved in control of blood coagulation (11), and to complement receptor C1qRp (12). The role of human TEM1/endosialin remains unclear.
In this work, we provide a molecular characterization of the mouse Tem1/endosialin cDNA and its genomic structure. In conjunction with this, we report the mapping of its location on the mouse chromosome and identification of its functional promoter. In addition, we present results from an extensive analysis of Tem1/endosialin mRNA and protein expression in vivo and in vitro. Our results provide new and intriguing data regarding the broad expression of Tem1/endosialin in both normal somatic cells and embryonic tissue. These findings alter the previous view that this gene is preferentially associated with tumor endothelium; however, they do not rule out the possibility that the expression of this gene is elevated in tumor tissue. Of further significance, and perhaps even more important, is our finding that Tem1/endosialin mRNA is regulated by cell culture density.
Human umbilical vein endothelial cells and media for their maintenance were obtained from Clonetics Corp. (Walkersville, MD). Endothelial cell growth medium contained endothelial cell basal medium supplemented with human recombinant epidermal growth factor, hydrocortisone, gentamycin, amphotericin B, bovine brain extract, and 2% FCS. Endothelial cell growth medium-2 contained endothelial cell basal medium supplemented with human recombinant epidermal growth factor, basic human fibroblast growth factor, human vascular endothelial growth factor, ascorbic acid, R3-IGF-1 human recombinant insulin-like growth factor, heparin, gentamycin, amphotericin B, and 2% FCS. Cells were used at passage 3.
NIH3T3 fibroblasts transformed with v-ras, v-raf, and v-src were provided by Douglas Lowy and William Vass. NIH3T3 cells transformed with v-mos were prepared using plasmid pM1 (15). pM1 plasmid DNA was transfected into the ⌿-2 packaging cell line, and the supernatant was used for infection of NIH3T3 cells.
To generate sparse, semiconfluent, and confluent cell cultures, logarithmically growing cells were seeded on 100-mm diameter tissue culture dishes (4 ϫ 10 5 cells/dish) and incubated at 37°C for 24, 48, or 72 h. Serum-starved cell cultures were incubated for 24 h in 10% FCS/ DMEM, washed and resuspended in 0.2% FCS/DMEM, and incubated for 24 h. For some experiments, the cells were refed with 10% FCS/ DMEM and grown for an additional 24 h. At the end of the incubation period, the cells were harvested and immediately subjected to RNA isolation as described below.
Screening of cDNA and Genomic DNA Libraries-A mouse thymus cDNA library cloned in pCMV6-XL4 (Rapid-Screen Arrayed cDNA Library Panel; OriGene Technologies, Rockville, MD) was screened by PCR using primers F1 (5Ј-TTCCCAGTCCCAAATCAGTG-3Ј) and R1 (5Ј-GAGTCCTGGTGTGTCCTTTG-3Ј) specific to the 3Ј-end of the mouse Tem1/endosialin gene and HotStarTaq Master Mix (Qiagen, Valencia, CA). Conditions of the PCR amplification for both Master Plate and Sub-plates were as follows: 95°C for 15 min, followed by 35 cycles of denaturation at 94°C for 30 s, annealing at 56°C for 30 s, and extension at 72°C for 90 s. For the third round of screening, PCR was performed on single bacterial colonies. DNA from positive clones was sequenced using both vector-and gene-specific primers. Sequencing was performed using the BigDye Terminator Cycle Sequencing kit (PerkinElmer Life Sciences), and gel separation was carried out at the NCI, National Institutes of Health, Sequencing Facility (Bethesda, MD).
Plasmids and Primers-An expressed sequence tag clone with IM-AGE ID 335868 was ordered from Research Genetics (Huntsville, AL) and was designed as clone 10/5. Tem1/endosialin cDNA from library clone A2 was cloned as a 2.6-kb NotI fragment into pBluescript II SK (Stratagene, La Jolla, CA) to generate pSK-Tem. A 13-kb XbaI fragment from a positive BAC clone RPCI-22 364N1 contained the whole genomic sequence of Tem1/endosialin and was subcloned into pBluescript SK to create pSK-Xb13. Preparation of several constructs was based on PCR using Pwo DNA polymerase (Roche Molecular Biochemi-cals), and the products were verified by sequencing. To create the pMH-Tem construct for expression of myc-tagged Tem1/endosialin, the ORF of the Tem1/endosialin cDNA was subcloned from pSK-Tem into the eukaryotic expression pcDNA 3.1/Myc-His A (Invitrogen, Carlsbad, CA) by PCR using primers L2 (5Ј-cttgaattcATGCTGCTGCGCC-3Ј) and R2 (5Ј-taattctagaCACACTGGTTCTACAG-3Ј) harboring EcoRI (5Ј) and XbaI (3Ј) sites, respectively. For preparation of Tem-green fluorescent protein (GFP), the coding region of the Tem1/endosialin gene was amplified from pSK-Tem using vector-specific T7 primer (5Ј-TAATAC-GACTCACTATAGGG-3Ј) and gene-specific R3 primer (5Ј-tttagatctCC-ACACTGGTTCTACAGGTC-3Ј) harboring a BglII (3Ј) site. The PCR product was digested with EcoRI and BglII and fused in-frame to GFP using pEGFP-N2 vector (CLONTECH, Palo Alto, CA).
PCR products were generated using 50 ng of pSK-Xb13 as a template DNA. PCR conditions were as follows: one cycle for 2 min at 94°C; five cycles at 94°C for 15 s, 60°C for 30 s, and 72°C for 45 s; five cycles at 94°C for 15 s and 72°C for 45 s; 20 cycles at 94°C for 15 s and 72°C for 45 plus 5 s/cycle and 1 cycle at 72°C for 5 min. The PCR products were digested with MluI and XhoI restriction enzymes and ligated into pGL3-Basic vector. Deletion mutants generated by PCR were sequenced to assure the absence of nucleotide incorporation errors.
Bioinformatics-Searches for homologous sequences and motifs were performed with the programs BLAST (17), FASTA (18), PROSITE (19), SMART (20), and Pfam (21). Comparison of the mouse and human TEM1/endosialin protein was executed using ClustalW (1.74). The DNA sequence upstream of the transcription start site was analyzed for eukaryotic transcription factor-binding sites using the program MatInspector in the TRANSFAC data base (22). Analysis for the presence of CpG islands was performed using the CpGplot program, available on the World Wide Web at www.ebi.ac.uk/emboss/cpgplot/ (23). Promoter prediction was done using NNPP/Eukaryotic at the BCM Search Launcher (available on the World Wide Web at dot.imgen.bcm. tmc.edu:9331/seq-search/gene-search.html).
In Vitro Transcription and Translation-The TNT T7-coupled reticulocyte lysate system (Promega, Madison, WI) was used for in vitro translation as recommended by the manufacturer. Briefly, 1 g of pSK-Tem plasmid DNA was incubated with a rabbit reticulocyte lysate for 1 h at 30°C in the presence of T7 RNA polymerase and 40 Ci of [ 35 S]methionine. The reaction was terminated by the addition of SDS sample buffer, and aliquots were analyzed by SDS-PAGE and visualized by fluorography.
Southern Blot Analysis-Mouse liver genomic DNA was isolated using the Wizard genomic purification kit (Promega). Ten g of DNA was digested with different restriction enzymes, and the digested DNA fragments were separated on a 0.8% agarose gel and processed for Southern hybridization. A 2.6-kb insert of A2 mouse cDNA was used as a probe.
Interspecific Mouse Backcross Mapping-Interspecific backcross progeny was generated by mating (C57BL/6J ϫ Mus spretus)F 1 females and C57BL/6J males as described (24). A total of 205 N 2 mice were used to map the Tem1/endosialin locus. DNA isolation, restriction enzyme digestion, agarose electrophoresis, Southern blot transfer, and hybridization were performed essentially as described (25). All blots were prepared with Hybond-N ϩ nylon membrane (Amersham Pharmacia Biotech). The probe (2.6-kb cDNA) was labeled with [␣-32 P]dCTP using a random primed labeling kit (Stratagene); washing was done to a final stringency of 1.0ϫ SSCP, 0.1% SDS, 65°C. Fragments of 1.9 and ϳ0.5 kb were detected in TaqI-digested C57BL/6J DNA, and a fragment of 2.7 kb was detected in TaqI-digested M. spretus DNA. The presence or absence of the 2.7-kb TaqI M. spretus-specific fragment was followed in the backcrossed mice. A description of the probes and RFLPs for the loci linked to Tem1/endosialin including Gal, Adrbk1, and Cd5 has been reported previously (26). Recombination distances were calculated using Map Manager, version 2.6.5. Gene order was determined by minimizing the number of recombination events required to explain the allele distribution patterns.
Northern Blot Analysis-Poly(A) ϩ and total RNA were prepared using an Oligotex Direct mRNA Midi/Maxi kit and RNeasy Mini kit (Qiagen, Valencia, CA), respectively, according to the instructions of the manufacturer. Samples containing 2 g of poly(A) ϩ or 5 g of total RNA were electrophoresed on 1.2% agarose gels containing 0.7% formaldehyde. RNA was blotted onto Nytran SuPerChargeNylon membrane (Schleicher & Schuell), UV-cross-linked using Stratalinker 1800 (Stratagene), and hybridized by standard techniques with 32 P-labeled probes.
A human 12-lane MTN blot was purchased from CLONTECH and used as recommended by the manufacturer. A 1.1-kb human TEM1/ endosialin probe was prepared by PCR screening of a human kidney cDNA library (OriGene Technologies) using human gene-specific primers hF1 (5Ј-GAGCAAGAGCCCAACAGAAC-3Ј) and hR1 (5Ј-TGGTGT-ATCCTTGGTCACGA-3Ј).
RT-PCR Analysis-Expression of the mouse Tem1/endosialin was analyzed by RT-PCR using the Mouse Rapid-Scan Gene expression panel (OriGene Technologies). For this analysis, primers F1 and R1 were utilized under conditions similar to those used above for screening of the cDNA library. Four panels containing 1000ϫ, 100ϫ, 10ϫ, and 1ϫ concentrations of cDNA from 24 different mouse tissues were used as templates. Five l of PCR products was separated on a 1% agarose gel followed by Southern blot analysis using the 10/5 insert as a probe.
Expression of the human TEM1/endosialin was assessed by RT-PCR analysis of human I, II, and cardiovascular multiple tissue cDNA panels (CLONTECH) using the HotStarTaq Master Mix kit (Qiagen). One ng of cDNA was used in combination with primers hF1 and hR1 under the following conditions of PCR amplification: 95°C for 15 min and 30 cycles consisting of 94°C for 30 s, 63°C for 30 s, and 72°C for 90 s. Aliquots from reactions were taken after 30, 34, 38, and 42 cycles, and the presence of the 272-bp DNA fragment was determined followed by electrophoresis on agarose gels. Reactions with GAPDH-specific primers were used as a control according to the manufacturer's instructions.
Luciferase Reporter Assays-Plasmid DNA for transient transfections was isolated using the Plasmid Maxi Kit (Qiagen). NIH3T3, PA6, and IP2-E4 cells were plated at a density of 3 ϫ 10 5 cells/60-mm plate and grown in appropriate medium overnight prior to transfections. Transfections were carried out using the Calcium Phosphate Transfection System (Life Technologies) according to the manufacturer's instructions. A total of 10 g of TEM1/endosialin promoter construct and 0.1 or 0.5 g of pRL-TK Renilla luciferase vector (Promega) were used for each transfection. The pRL-TK luciferase activity was used to control transfection efficiency. Each transfection experiment was performed in triplicate and repeated at least two or three times with two different DNA isolates. For experiments performed to determine the sensitivity of the promoter to cell density, cells were split 18 -20 h after transfection and plated in 12-well plates to achieve rapid confluence (dense cultures) or 60-mm plates to obtain sparse cultures. For transfection of SR-4987 cells, 10 g of promoter construct along with 0.1 g of pRL-TK was introduced into cells by electroporation with the addition of DEAE-dextran at a final concentration of 10 g/ml. Cells (2 ϫ 10 6 ) in a volume of 0.3 ml were mixed with DNA, transferred into 0.4-cm electroporation cuvettes, and transfected by electroporation using Gene Pulser II (Bio-Rad) (250 V, 950 microfarads). After electroporation, cells were plated on 60-mm tissue culture dishes with fresh medium and analyzed 24 h later. Firefly luciferase and Renilla luciferase assays were performed using the Dual-Luciferase Reporter Assay System (Promega) 24 -48 h after transfections. Both luciferase activities were measured in the same tube by a luminometer TD-20e (Turner Designs, Sunnyvale, CA).
Western Blot Analysis-Cells (5 ϫ 10 5 ) were transfected with 20 g of the pMH-Tem construct using the Calcium Phosphate Transfection System (Life Technologies) according to the manufacturer's instructions. 48 h after transfection, the cells were washed and lysed in radioimmune precipitation buffer (10 mM Tris, pH 7.4, 0.15 M NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS) with 10 mM iodoacetamide, 1 mM phenylmethylsulfonyl fluoride, and Complete protease inhibitor mixture (Roche Molecular Biochemicals). The amount of proteins was determined by the Bio-Rad DC (detergent-compatible) Protein Assay kit. Twenty-five g of whole-cell lysate protein was resolved on 7.5% Tris-HCl gel (Bio-Rad). After gel electrophoresis, the proteins were transferred to a nitrocellulose membrane (Bio-Rad). Membrane was blocked with 5% nonfat milk for 1 h and probed with either anti-Myc or anti-His 6 antibody conjugated with horseradish peroxidase (Invitrogen) at a 1:5000 dilution for 1 h. The signal was detected using the Super Signal West Pico Chemiluminiscent Substrate kit (Pierce) and Kodak Biomax MR film.
The rabbit polyclonal Tem1/endosialin antiserum specific to C-terminal part of the protein was used at a 1:800 dilution. Secondary goat anti-rabbit horseradish peroxidase-conjugated IgG (Pierce) was used at a 1:100,000 dilution. Blots were developed as described above.
Intracellular Localization and Confocal Microscopy-To determine intracellular localization of Tem-1, NIH3T3 cells were grown on coverslips in 35-mm dishes and transfected with 1.5 g of Tem-GFP, pMH-Tem plasmid, and control GFP vector, respectively, using the calcium phosphate method. For observation by confocal microscopy of the signal produced by GFP-Tem protein, the cells were washed twice with PBS, fixed for 30 min in 4% paraformaldehyde, and washed twice with PBS before mounting in Antifade (DAKO Corp., Carpinteria, CA).
For immunocytochemistry, cells were washed three times with PBS and fixed with ice-cold methanol for 15 min. After thorough washings, primary anti-Myc (1:200 dilution) or anti-His 6 (1:200 dilution) (Invitrogen) and secondary fluorescein isothiocyanate-conjugated anti-mouse IgGI and IgG 2B antibodies (1:200 -1:5000 dilution; Roche Molecular Biochemicals) in PBS were used for detection of Myc-or His 6 -tagged proteins by 30-min incubations at 37°C followed by three washings with PBS. Coverslips were then incubated with PI solution (0.3 g/ml propidium iodide and 10 g/ml RNase A in PBS) for 30 min at 37°C, washed three times with PBS, and mounted in Antifade.
GFP, fluorescein isothiocyanate, and propidium iodide fluorescence were analyzed using a Bio-Rad MRC-1024 confocal scan head mounted on a Nikon Optiphot microscope with a 60ϫ planapochromat lens. A krypton-argon gas laser provided excitation at 488 and 568 nm. Emission filters of 598/40 and 522/32 were used sequentially for acquiring red and green fluorescence, respectively. Images were collected using LaserSharp software (Bio-Rad) and evaluated using Adobe Photoshop 5.0 and Corel Draw 9.0.

RESULTS
Mouse Tem1/endosialin cDNA-The GenBank TM data base contains several mouse expressed sequence tag clones that possess high homology to human TEM1/endosialin (accession number AF279142) and originate from different tissues. We used an expressed sequence tag clone, IMAGE number 335868, to design primers F1 and R1 for screening of a mouse thymus cDNA library (Rapid-Screen Arrayed cDNA library panel; Ori-Gene Technologies). This approach allowed us to isolate two clones designated A1 and A2 that contained the mouse Tem1/ endosialin cDNA inserts with approximate sizes of 2.5 and 2.6 kb, respectively. Sequence analysis revealed that the 5Ј-end of the A2 cDNA insert is extended by 139 nucleotides compared with A1. As depicted in Fig. 1, the assembled mouse Tem1/ endosialin cDNA contains 2553 bp and is composed of a short 5Ј-untranslated region of alanines (9.41%), glycines (8.24%), and leucines (8.1%). It appears to be a type I membrane protein with a putative signal peptide of 17 amino acids (aa) and a long extracellular part of 676 aa followed by a single membrane spanning domain of 21 aa and a short cytoplasmic tail of 51 aa. The extracellular region contains several well known protein motifs: C-type lectin domain (aa 22-157), sushi domain (aa 164 -230), three EGF-like domains (aa 234 -272, 274 -311, 315-351) and sialomucin-like region. This modular architecture, as well as the homology to both thrombomodulin and C1qRp complement receptor, have already been described for the human TEM1/ endosialin (9).
Using the PROSCAN program, we found that the mouse Tem1/endosialin extracellular domain contains a potential N-glycosylation site (at position 636) and many putative Olinked glycosylation sites, while the cytoplasmic tail possesses a number of potential phosphorylation sites and several N-myristoylation sites. The utilization and importance of these modification sites remain to be assessed. Since the human TEM1/endosialin protein has been shown to be heavily O-glycosylated, we would expect that many of the predicted sites of O-glycosylation in the mouse homolog would also be true sites of post-translational modification (9).
Alignment of the mouse and human amino acid sequences showed an overall 77.5% identity (Fig. 1). The least conserved part of the molecule is a region from aa 478 to 610 on the mouse sequence. Interestingly, the putative transmembrane region is 100% identical between human and mouse Tem1/endosialin, and there are only four amino acid substitutions out of a total of 51 amino acids in the cytoplasmic tail, suggesting that these two regions may possess conserved function.
To prove that the Tem1/endosialin cDNA described here encodes a protein with the expected molecular weight, we subcloned the full-length cDNA into the pBluescript SK vector and used this plasmid as a template for T7-primed TNT-coupled in vitro transcription and translation. The reaction yielded a unique translation product migrating on SDS-polyacrylamide gel electrophoresis as a band of ϳ82 kDa, which is in excellent agreement with the predicted molecular weight of the mouse Tem1/endosialin core protein ( Fig. 2A). To determine the molecular weight of protein in eukaryotic cells, we transiently transfected NIH3T3 cells with an expression vector carrying the mouse cDNA fused to a C-terminal Myc-His 6 tag. Western blot analysis using anti-Myc antibody revealed a band of ϳ95 kDa (Fig. 2B). Since the fusion protein contains a 3.1-kDa tag, the estimated molecular mass for the Tem1/endosialin protein is 92 kDa. Western blotting of untransfected NIH3T3 cells using rabbit polyclonal antiserum raised against the intracellular domain of the mouse Tem1/endosialin demonstrated that the cells contain an endogenous protein of 92 kDa (Fig. 2C). However, the analysis also revealed several additional bands of higher molecular weight presumably representing different glycosylated forms of the protein as has been described previously for the human TEM1/endosialin (9).
Distribution of Tem1/endosialin in Vivo and in Vitro-Availability of the cDNA allowed us to perform a survey of a broad range of mouse and human tissues. This was important with respect to the previously reported absence of the human TEM1/ endosialin expression in normal tissues (9,10). First, we analyzed a collection of tissues in a mouse Rapid-Scan panel (Ori-Gene). F1 and R1 gene-specific primers were employed in an RT-PCR amplification, and the products were detected by Southern blot analysis with a labeled cDNA fragment (Fig. 3A). Surprisingly, we found that Tem1/endosialin mRNA can be detected in all mouse tissues involved in the panel, with the highest levels present in the heart, kidney, stomach, skin, transmembrane region are in italic, boldface, underlined letters. Extracellular structural domains are indicated by shaded backgrounds in dark (C-type lectin), light (sushi), or middle gray (EGF). The putative N-glycosylation site and phosphorylation sites are shown in white letters in black boxes. Amino acid residues that are identical in mouse and human proteins are marked by inverted triangles. pancreas, uterus, embryo at day 19, and virgin breast. Our data indicate a role for the gene in embryonic development, since expression was detected in all samples spanning embryonic days 8.5-19. The finding of high expression in numerous tissues is contradictory to data published before for human TEM1; therefore, we also looked at the expression of TEM1/ endosialin in a collection of normal human tissues. Hybridization of a multiple tissue Northern blot, with a 1.1-kb human cDNA fragment as a probe, proved that TEM1/endosialin is transcribed in most normal human tissues (Fig. 4A). Similar data were obtained by RT-PCR analysis (Fig. 4B). The highest levels of TEM1/endosialin mRNA were detected in the placenta, ovary, heart, skeletal muscle, small intestine, and various parts of cardiovascular system.
Although Tem1/endosialin is widely expressed in preparations of whole tissue, examination of various murine cell lines by a Northern blot analysis revealed that its in vitro expression is limited to cells of very specific origins (Fig. 3B). These include the cell lines derived from embryonic fibroblasts (NIH3T3, BALB/3T12-3, and SC1), preadipocytes (3T3L1 and PA6), and endothelial cells (SVEC4 -10 and IP2-E4). The cell lines that represent other cell types, such as hepatocytes (AML12); carcinoma cells of mammary gland (C127:LT), breast (JC), and lung (KLN205); stromal cells of the bone marrow (SR-4987); keratinocytes; and various cell lines of hematopoietic origin were negative (data not shown). These data suggest that the ubiquitous expression of Tem1/endosialin in tissues is conferred by their endothelial cell components.
Based on this assumption, we tested two available primary endothelial cell lines for the presence of Tem1/endosialin transcripts and found that the rat primary brain capillary endothelial cells express Tem1/endosialin specific RNA. On the other hand, the human umbilical vein endothelial cells do not express the gene, even when grown in the presence of angiogenic growth factors (data not shown). This lack of expression is in agreement with previously published data (10). These results suggest that the expression of TEM1/endosialin might be restricted to microvascular endothelial cells.
In summary, these analyses of Tem1/endosialin steady state expression indicate that mRNA for this gene is produced in a diversity of tissues in the adult and throughout development. However, the only cell lines identified here as positive for Tem1/endosialin expression were of either embryonic, endothelial, or preadipocyte origin. The distribution of Tem1/endosialin and its human relative as determined by us is much broader than originally reported.
Intracellular Localization of C-terminally Tagged Tem1/endosialin-The domain composition of Tem1/endosialin predicts that it is a type 1 receptor subject to intracellular transport along the secretory pathway and exposure at the plasma membrane. In order to investigate whether the mouse Tem1/endosialin is indeed associated with the plasma membrane, we generated a fusion antigen with the GFP at its carboxyl terminus. Confocal microscopy revealed that the Tem1/endosialin-GFP transiently expressed in NIH3T3 cells was predominantly located in a perinuclear area of the cytosol, while a minor fraction of the fusion protein was observed in association with the plasma membrane (Fig. 5A). The cytosolic fluorescence was not uniformly spread but showed a granular pattern reminiscent of that seen with the proteins associated with endoplasmic reticulum or Golgi vesicles. On the other hand, the wild-type GFP displayed a rather different distribution, being evenly spread over the whole cytosolic and nuclear areas (data not shown).
Because the addition of a relatively large GFP moiety to the C terminus of Tem1/endosialin protein could affect its conformation and thereby the capacity for proper transport and cell surface localization, we also examined a Tem1/endosialin fused to a shorter Myc-His 6 tag. However, the immunofluorescence using anti-Myc antibody (Fig. 5B) or anti-His 6 antibody (Fig.  5C) displayed a similar pattern as described above for the GFP fusion protein, supporting the view that only a small portion of the C-terminally tagged Tem1/endosialin was associated with the plasma membrane.
Cell Density-dependent Expression Tem1/endosialin in Culture-The domain architecture of Tem1/endosialin suggests that the protein might be involved in cell-cell interaction; in particular, it has a C-lectin domain and EGF domains, both of which are found in selectins (27,28). Since many intercellular communication events rely upon direct contact, their regula- tory pathways frequently exert density-dependent modulation. On this basis, we decided to investigate a possible effect of the cell density on expression of endogenous Tem1/endosialin in NIH3T3 cells. Cells were plated at a density of 5 ϫ 10 4 per 100-mm dish and harvested for analysis after 24 h (while they were still sparse), after 48 h (at medium density), and after 72 h (at full confluence). Northern blot analysis revealed that the sparse NIH3T3 cells express a relatively low level of Tem1/ endosialin RNA, but transcription is strongly induced when the cells reached full confluence (Fig. 6A).
Similar cell density-dependent induction of Tem1/endosialin transcription was observed in the IP2-E4 mouse endothelial cell line (Fig. 6A) as well as in other cell lines including BALB3T12-3 and the bone marrow-derived preadipocyte cell line PA6 (not shown). The effect of cell density on expression of the Tem1/endosialin protein was confirmed by Western blot analysis of whole cell extracts prepared from the sparse, semiconfluent, and confluent NIH3T3 cells using the rabbit polyclonal Tem1/endosialin antiserum (Fig. 6B).
Molecular signals involved in the induction of Tem1/endosialin expression in dense cells could be related either to establishment of direct contacts between the plasma membranes of the neighboring cells or to depletion of serum growth factors during the extended cultivation period. Interestingly, serum starvation for 24 h resulted in elevated transcription of Tem1/ endosialin in NIH3T3 cells, although the induction was not so dramatic as that caused by increased density. On the other hand, the refed cells displayed a similar level of the transcript as the nonstarved control culture (Fig. 6A).
Potent viral oncogenes such as v-src, v-ras, v-mos, and v-raf, when expressed in NIH3T3 cells, cause both a diminished requirement for growth factors and abnormal intercellular contacts. In particular, they are refractile to normal contact-inhibitory signals. We wished to see if any of these oncogenes, which are involved in signal transduction, could prevent the contactinhibited up-regulation of Tem1/endosialin. NIH3T3 cells expressing these oncogenes were examined for density-dependent induction of Tem1/endosialin mRNA. Cells expressing v-src, v-ras, and v-raf, still showed density dependent expression of Tem1/endosialin (Fig. 6C). In contrast and to our surprise, v-mos-transformed NIH3T3 cells expressed high levels of Tem1/endosialin transcript regardless of the cell density (Fig.  5C). This suggests that v-mos, instead of blocking the upregulation of Tem1, activates a pathway leading to the Tem1/ endosialin expression.
Density-stimulated expression of Tem1/endosialin could be related to the proliferative status of the cells, since the sparse cells proliferate more rapidly than the cells approaching confluence. This raised the possibility that Tem1/endosialin could be cell growth-inhibitory. To investigate this, we performed a series of experiments in which we analyzed growth properties of NIH3T3 cells that overexpressed Tem1/endosialin. Cells transfected with pcDNA3.1 containing Tem1/endosialin cDNA, in sense and antisense orientation, and mock-transfected cells were compared for their growth capacity in colony formation assays. In addition, Tem-1/endosialin was expressed conditionally using a lac operon system (Stratagene). Although we observed high induction of Tem1 mRNA, we were unable to show any growth inhibitory effects of the gene (data not shown). Moreover, clonal cell lines that were derived from transfected SR4987 cells and constitutively overexpressed Tem1/endosialin did not display any changes in their proliferative rate. Thus, at present we have no evidence that Tem1/ endosialin is directly involved in the regulation of cell proliferation; however, these experiments do not completely rule out this possibility.
Isolation and Structure of the Mouse Tem1/endosialin Gene-To further our knowledge of the molecular properties FIG. 4. Expression of Tem1/endosialin mRNA in human tissues. A, human multiple tissue Northern blot was hybridized with the 1.1-kb fragment of human TEM1/endosialin cDNA. The blot was reprobed with ␤-actin as a loading control. B, expression pattern of human TEM1/ endosialin as determined by ethidium bromide-stained PCR products amplified from human Multiple Tissue cDNA panels I and II and a cardiovascular panel. Primers were specific for the 3Ј-end of the human gene or specific for GAPDH. The ratio of signal intensity normalized to GAPDH is graphically illustrated below. He, heart; Br, brain; Pl, placenta; Lu, lung; Li, liver, SM, skeletal muscle; Ki, kidney; Pa, pancreas; Sp, spleen; Th, thymus; Pr, prostate; Te, testis; Ov, ovary; Sl, small intestine; Co, colon; PBL, peripheral blood leukocytes; AH, adult heart; FH, fetal heart; Ao, aorta; Ap, apex; AL, atrium left, AR, atrium right; AD, auricle dextra; AS; auricle sinistra; VL, ventricle left; VR, ventricle right; IS, interventricular septum; AN, atrioventricular node. and regulation of mouse Tem1/endosialin expression, we began a characterization of its genomic sequence including the related transcriptional control region. For the isolation of the Tem1/endosialin gene, we used a 650-bp fragment from the original expressed sequence tag clone (accession number AF279142) to screen a mouse BAC library. Positive clones were analyzed by Southern blot hybridization using the fulllength Tem1/endosialin cDNA as a probe and all displayed a similar digestion pattern (data not shown). This indicated that the isolated BAC clones contained the Tem1/endosialin gene and that no major rearrangements had occurred during cloning and propagation of the BAC library.
Interestingly, the entire coding sequence appeared to be located on a 13-kb XbaI fragment that was subcloned into pBluescript II SK for in depth characterization. Genomic sequence of 8245 bp was obtained from both strands using the primers initially selected from the full-length Tem1/endosialin cDNA and then completed by primer walking. The sequence was submitted to the GenBank TM data base under accession number AF388573. A BLAST search against the current Gen-Bank TM /EMBL data base revealed substantial sequence homology only to human TEM1/endosialin. Alignment of the genomic sequence with the full-length Tem1/endosialin cDNA revealed that the gene is intronless throughout its coding region (Fig. 7A). Two differences were detected between the genomic and cDNA sequences. One was a substitution of T for C in position 459 of the cDNA sequence, which did not result in an amino acid change and was present only in clone A2 of the cDNA, suggesting that this substitution was introduced during a preparation of the cDNA library. A second difference was a substitution of T for C at position 1637 of the cDNA sequence and resulted in a change of proline to serine at position 525. This substitution, which introduced a SalI site, was found to be present in both cDNA clones and presumably represents a polymorphism between different mouse strains used for construction of cDNA and genomic libraries. Digestion by SalI can be potentially used as a marker to identify different mouse strains.
Southern blot analysis revealed that the Tem1/endosialin gene is a single copy gene (Fig. 7B). The single 4.3-kb band in NheI-digested DNA corresponded to the size predicted from the genomic sequence (Fig. 6A). Furthermore, the presence of a single Tem1/endosialin positive band in EcoRI, EcoRV, or SpeI-digested DNA confirmed that the gene is intronless and present in only one copy.
Chromosomal Localization of Tem1/endosialin Gene-The mouse chromosomal location of Tem1/endosialin locus was FIG. 5. Subcellular localization of C-terminally tagged Tem1/ endosialin. A, fluorescence microscopy of NIH3T3 cells expressing Tem1-GFP fusion protein is depicted in two panels. Cells transfected in parallel with GFP showed a uniform distribution in the nucleus and cytoplasm (not shown). B, immunofluorescence analysis of NIH3T3 cells transfected with pMH-Tem. The images obtained by confocal laserscanning microscopy show the staining pattern obtained with anti-Myc primary antibody followed by fluorescein isothiocyanate-conjugated secondary antibody (left). Nuclei of the same cells were stained by propidium idodide (right). All images are single optical sections. C, same as in B using anti-His 6 primary antibody.
FIG. 6. Density-dependent expression of Tem1/endosialin in vitro. A, poly(A) ϩ RNA was prepared from sparse (S), confluent (C), serum-starved (SE), and refed (RE) NIH3T3 cells or IP2-E4 cells that were sparse or confluent. Two g of RNA was examined by Northern blot hybridization with Tem1/endosialin cDNA and reprobed with GAPDH. B, whole cell extracts were prepared from sparse (P), semiconfluent (SC), and confluent (C) NIH3T3 cells. The samples were analyzed by immunoblotting with anti-Tem1/endosialin rabbit polyclonal antiserum. C, Northern analysis of Tem1/endosialin expression in sparse (S) and confluent (C) NIH3T3 cells transformed by viral oncogenes v-mos, v-src, and v-raf was performed as described above. determined by interspecific backcross analysis using progeny from matings of (C57BL/6J ϫ Mus spretus)F 1 ϫ C57BL/6J mice (24). C57BL/6J and M. spretus DNAs were digested with several enzymes and analyzed by Southern blot hybridization for informative restriction fragment length polymorphisms using a mouse cDNA probe. The 2.7-kb TaqI M. spretus restriction fragment length polymorphism was used to follow the segregation of the Tem1/endosialin locus in the backcrossed mice. The mapping results indicated that Tem1/endosialin is located in the proximal region of mouse chromosome 19 linked to the Gal, Adrbk1, and Cd5 genes (Fig. 8). Although 118 mice were analyzed for every marker and are shown in the segregation analysis (Fig. 8), up to 191 mice were typed for some pairs of markers. Each locus was analyzed in pairwise combinations for recombination frequencies using the additional data. The ratios of the total number of mice exhibiting recombinant chromosomes to the total number of mice analyzed for each pair of loci and the most likely gene order are: centromere-Gal-2/191-Adrbk1-0/190-Tem1-3/121-Cd5. Recombination frequencies (expressed as genetic distances in centimorgans Ϯ S.E.) are: No recombinants were detected between Adrbk1 and Tem1 in 190 animals typed in common suggesting that the two loci are within 1.6 centimorgans of each other (upper 95% confidence limit).
We have compared our interspecific map of chromosome 19 with a composite mouse linkage map that reports the map location of many uncloned mouse mutations (provided from the Mouse Genome Data base, a computerized data base maintained at the Jackson Laboratory, Bar Harbor, ME). Tem1/ endosialin mapped in a region of the composite map that lacks mouse mutations with a phenotype that might be expected for an alteration in this locus (data not shown).
The proximal region of mouse chromosome 19 shares a region of homology with human chromosome 11q, and accordingly the human TEM1/endosialin gene was recently mapped to chromosome 11q13 (9,10).
Sequence Analysis of 5Ј-flanking Region and Promoter Activity-In an effort to determine the position, structure, and activity of the transcriptional control region of the mouse Tem1/ endosialin gene, we analyzed the sequence of its 5Ј-flanking region and examined the capacity of deletion fragments to drive transcription of a reporter gene. Nucleotide sequence analysis of the 1.8-kb 5Ј-flanking fragment revealed the presence of a CG-rich region overlapping the proposed ATG initiation codon (Fig. 9). Further analysis using the CpG plot program has shown that this region displays features of a CpG island, with an average observed-to-expected CpG ratio of 0.79 and C ϩ G content of 75%. Because the 288-bp CpG island covered the region immediately upstream of the 5Ј-end of the Tem1/endosialin gene, we proposed that it contained the promoter and/or was involved in regulation of the gene expression.
The transcription initiation site for the human TEM1/endosialin was reported to reside 17 nucleotides upstream of the first ATG (9). Since our A2 cDNA clone had only three nucleotides in front of ATG, it was reasonable to expect that a portion of the 5Ј-end of Tem1/endosialin is missing. Promoter prediction using NNPP/Eukaryotic suggested the presence of a promoter in the region Ϫ124/Ϫ75 relative to the ATG with a potential transcription start site at position Ϫ83. However, in accordance with the GC-rich character of this region, we were unable to identify an exact transcription initiation site(s) either by 5Ј-rapid amplification of cDNA ends or by primer extension analysis (not shown). Therefore, we decided to study the promoter activity of the whole 1.8-kb flanking region starting immediately upstream of ATG.
Toward this aim, we performed PCR amplification of the Ϫ1829/Ϫ5 fragment and generated a series of nested deletion mutants fused to the luciferase reporter gene in the promoterless pGL3-Basic vector. Altogether, we prepared nine constructs containing the fragments in the sense orientation (pGL-TemA to -F and pGL-TemJ to -L) and two constructs with the fragments in the antisense orientation (pGL-TemG and -H). The constructs were transiently transfected into cells that express the endogenous Tem1/endosialin gene (NIH3T3, PA-6, and IP2-E4) as well as into nonexpressing SR-4987 cells, and luciferase activity was measured after 24 h.
An evaluation of the complex profile of the luciferase activity produced by different constructs in NIH3T3 cells (Fig. 10A) shows that the promoter of the mouse Tem1/endosialin gene is located in the region Ϫ263/Ϫ65 and that this region contains important positive regulatory elements. In support of this, elimination of the promoter region from the longest fragment almost fully abrogated its activity. The antisense fragments were inactive.
Interestingly, the cell lines that express a lower level of Tem1/endosialin RNA displayed lower promoter activity, albeit with an analogous profile (not shown), suggesting that the initiation of transcription critically contributes to regulation of Tem1/endosialin expression. The transfected promoter constructs were also active in SR-4987 cells that do not express Tem1/endosialin mRNA (Fig. 10B), suggesting that negative regulatory mechanism(s) operative at the endogenous chromosomally localized gene were not functional on the nonintegrated constructs.
Insight into the promoter sequence with regard to the presence of putative regulatory elements revealed binding sites for several transcription factors, including SP1, AP2, RREB1, WT1, and Ets1 sites and one site for the serum response factor (Fig. 10A). Some of these elements have been shown to play an important role in transcriptional control of angiogenesis-related genes (29,30), but their actual contribution to the regulation of the Tem1/endosialin gene remains to be determined.
Since the Tem1 gene is up-regulated in high density cells, it was important to determine if this expression is due to regulation of transcription at the level of the promoter. Therefore, NIH3T3 cells were transfected individually with pGL-TemA and pGL-TemF as well as the pGL3-Basic vector and separated into dense and sparse cultures. The luciferase assay demonstrated a significant increase in promoter activity in dense cultures compared with sparse cultures as shown in Fig. 10C. Similar density-dependent luciferase activity was also demonstrated for construct pGL-TemC (data not shown). This strongly indicates that transcription initiation contributes at least partially to the augmented expression of Tem1 observed in dense cell cultures. DISCUSSION The data presented here indicate for the first time that Tem1/endosialin plays not only a role in tumor-specific endothelium, but a more general role in vascular endothelium both in the adult and during embryonic development. Although previous data suggested that its expression is restricted to tumors (9), we propose that it may simply be expressed at higher levels during the formation of new vessels in tumors. In support of this idea, others have found that it is expressed during angiogenic states, such as in the corpus luteum and granular tissue of healing wounds (8). Although this may not represent a qualitative difference between normal endothelium and tumor endothelium as suggested previously, it may certainly reflect a quantitative difference. Quantitative differences are likely in tumor tissue known to have imbalances in the levels of regulators of angiogenesis. Since Tem1/endosialin's expression is most likely elevated in tumor tissues undergoing formation and reorganization of vessels, it could provide a target for antiangiogenic therapy for neoplastic disease.
Endothelial cells demonstrate heterogeneity among and within tissues (1). Interestingly, the distribution of Tem1/ endosialin varied among tissues with higher levels in the heart and pancreas and lower expression in the spleen and liver in both mice and humans. In addition, the placenta, ovary, and uterus were demonstrated to have high levels. This observation, along with the fact that the gene was expressed in rat capillary cells, but not in large vessel cells (human umbilical vein endothelial cells), suggests that this gene may be important in some subtype(s) of microvascular endothelial cells.
As we have demonstrated in this study, Tem1/endosialin mRNA is up-regulated as cells increase in density. This was shown for NIH 3T3 cells as well as a mouse endothelial cell line, IP2-E4, and was confirmed at the protein level in NIH3T3 cells. Because its up-regulation may be associated with contact inhibition, we performed experiments to test whether, under conditions of forced expression, the gene could induce growth arrest. However, we were not able to demonstrate such a function for Tem1/endosialin. We hypothesize, therefore, that upregulation of this gene occurs either in conjunction with increases in cell density or growth arrest but does not induce growth arrest itself. The fact that Tem1/endosialin is also up-regulated in serum-starved cells suggests that this induction is not simply triggered by the contact between cells per se. Alternatively, there may be more that one mechanism for its induction. Our data using luciferase reporter constructs show that the density-dependent expression is at least in part due to differences in activity at the promoter level. Many potential transcription factor-binding sites have been identified in the sequence of the promoter region (Fig. 9), and it will be of interest to determine which of these sites functions to activate Tem1 transcription.
The Tem1/endosialin protein might serve one of several roles in the vasculature including signaling from cell to cell or serving as a receptor for soluble ligands. Its potential role in cell-cell interactions is suggested by the structure within its extracellular domain. Like many selectins that are involved in cell-cell interactions, it has a C-lectin domain and EGF domains (28). Its potential cell-cell interactions could include homotypic interactions or heterotypic interactions with periendothelial cells (pericytes in small vessels) that lie adjacent to endothelial cells in vivo (3). One cannot rule out the additional possibility that Tem1/endosialin is involved in interactions with components within the extracellular matrix. In any case, its specific function in the vasculature is beyond the scope of this discussion and will be subject of future investigations.
Tem1/endosialin shares, in its extracellular domain, homology with thrombomodulin. This is a cell surface molecule that is widely expressed on normal endothelial cells and under certain conditions in smooth muscle cells. It is best known for its role in the interaction with thrombin in the coagulation cascade (9, 31). Tem1/endosialin, however, lacks homology to FIG. 9. DNA sequence of the 5flanking region of the Tem1/endosialin gene. The sequence was analyzed for transcription binding sites using the Mat-Inspector Professional program. Position ϩ1 refers to A in the ATG translation initiation codon. Predicted regulatory elements are shown in boldface underlined letters with the relevant putative transcription factors indicated above. The basic promoter region as determined by the promoter activity analysis below is indicated by a shaded background. thrombomodulin in the thrombin interaction domains (EGF repeats 5 and 6). An additional role for thrombomodulin is evident from its ability to inhibit arterial smooth muscle cell proliferation (32,33) and modulate mitogenic responses of endothelial cells to thrombin (34). Indeed, embryonic lethality in homozygous thrombomodulin knockout mice has been proposed to be due to thrombin-independent functions (35). Parallel investigations on this related protein may help us to begin to understand the function of Tem1/endosialin.
Our findings concerning the localization of Tem1/endosialin in the cell support its predicted cell surface association. Following transfection and immunofluoresence of permeabilized cells, we found a proportion of the protein in the vicinity of the cell membrane, although a large proportion was also found in the perinuclear region. It is interesting to point out that this localization is similar to that seen for thrombomodulin following transient transfection and permeabilization of cells (36). We realize that the GFP or Myc-His 6 tag at the carboxyl terminus used in this study could potentially alter the localization of the protein. Although we wanted to confirm our result by looking at endogenous Tem1/endosialin, our antibody raised to the protein did not function well in this assay.
The intracellular tail of Tem1/endosialin contains several putative phosphorylation sites that could serve in transduction of extracellular signals. This domain could also function in the subcellular localization of the protein, as was shown to be the case for thrombomodulin (37).
An interesting outcome of our analysis of the genome structure of murine Tem1/endosialin was the finding that the coding region for the gene is intronless. Only a small proportion (at FIG. 10. Transcriptional activity of the 5-flanking region of the Tem1/endosialin gene. A, promoter activity was examined using reporter constructs in the pGL3-Basic vector. Deletion fragments of the 1.8-kb 5Ј-flanking region were placed upstream of the luciferase gene (left). NIH3T3 cells were transiently transfected with the reporter constructs. pRLTK vector was co-transfected as a control for transfection efficiency. Luciferase activity was normalized using Renilla luciferase. The data are presented in relative values compared with the promoterless pGL3-Basic vector. The results represent mean values of the three separate experiments with two different isolates of plasmid DNA. B, normalized luciferase activities obtained with the reporter constructs in transiently transfected SR-4987 cells that do not express Tem1/endosialin mRNA. C, luciferase activities in NIH3T3 cells separated after transfection into dense and sparse cultures. The data for sparse or dense cultures are presented in relative values compared with those in pGL3-Basic in sparse or dense cultures, respectively. most 5%) of genes lack introns. A family whose diverse members are usually intronless are the G-protein-coupled receptor family genes. Among other human cell surface receptor genes, about 35% are found to be intronless (38). It has been proposed that an advantage of being intronless may be that transcription can occur more efficiently and with greater abundance, but in some cases experimental evidence indicates that introns actually increase gene expression levels (39).
In conclusion, our data indicate that Tem1/endosialin may be functionally involved in angiogenesis or vascular function, not only in pathological situations, but under physiological conditions as well. Although information about pathways leading to its activation and about its function is scarce, our work provides important clues, comprehensive molecular data, and useful tools that will facilitate further studies.