Endosialin is a C-type lectin-like cell surface receptor of unknown function, with a distinctive pattern of endothelial expression in newly formed blood vessels in human cancers. The murine orthologue of endosialin has been identified, opening up the analysis of developmental regulation in the embryo and in aberrant tissue remodeling, notably cancer angiogenesis. To advance these studies we have generated an antibody to the extracellular domain of mouse endosialin and mapped protein expression from embryonic day E10.0 to the adult stage, complemented by mRNA quantification and co-typing for standard endothelial markers. Four main findings emerged. First, endosialin protein is restricted to vascular endothelium and fibroblast-like cells in developing organs, and largely disappears in the adult. Second, endothelial expression varies markedly between organs regarding spatial and temporal patterns. For instance, in the E10.0 embryo, endosialin is prominent in the endothelium of the dorsal aorta and, from E11.0 to E14.5, in vessels sprouting from the dorsal aorta, in perineural vascular plexuses, and in brain capillaries. Third, circumscribed mesenchymal expression in fibroblast-like cells was evident throughout development, most pronounced adjacent to certain budding epithelia, as exemplified by the lung and kidney glomeruli, but unrelated to the endothelial expression. The endosialin protein persists in the stromal fibroblasts of the adult uterus. Finally, in subcutaneous cancer xenograft models endosialin re-appears in the host-derived tumor stroma, both in neo-angiogenic vascular endothelium and in activated stromal fibroblasts. In future studies, the search for intrinsic or extrinsic signals contributing to endosialin induction in cancer stroma will be of interest.
This article was published in Cancer Immunity, a Cancer Research Institute journal that ceased publication in 2013 and is now provided online in association with Cancer Immunology Research.
Human endosialin is a highly sialylated, C-type lectin-like cell surface receptor structurally related to thrombomodulin and complement receptor C1qRp (1, 2). First identified with a monoclonal antibody, mAb FB5 (1), endosialin was discovered independently through large-scale expression profiling of human cancer endothelial cells with the SAGE (serial analysis of gene expression) method (3), leading to the alternative designation of tumor endothelial marker 1 (TEM1). Both lines of investigation suggested that human TEM1/endosialin RNA and protein expression distinguish tumor endothelium from the endothelium of most normal adult tissues. The initial immunohistochemical studies with mAb FB5 pointed to a marked heterogeneity of antigen expression within a given tumor nodule and among similar cancer types derived from different patients (1), a finding extended by subsequent studies (4), while others using different anti-human endosialin antibodies described antigen expression in vascular pericytes in a small number of human breast cancer samples (5), and in tumor endothelium and pericytes in invasive brain tumors (6). Why some vascular endothelial cells show induction of endosialin in vivo has remained an open question, and commonly used in vitro systems, such as human umbilical vein endothelial cell cultures, fail to express endosialin under standard growth conditions or following stimulation with known endothelial activators (1, 5, 7).
The molecular cloning of the mouse orthologue of endosialin (mEs), with 77 percent overall sequence homology and comparable domain structure, has been described (7, 8). Mouse endosialin comprises a 92-kDa cell surface glycoprotein with highly sialylated, O-linked oligosaccharide side chains similar to the human protein. While non-quantitative RT-PCR studies suggested the presence of endosialin RNA in most tissues of the adult and embryonic mouse (7), a competing approach with in situ hybridization found endosialin mRNA specifically localized to tumor endothelium in experimentally-induced mouse cancers, in subsets of blood vessels in E15.5 mouse embryos (8), but not in most adult mouse tissues.
The present study was designed to generate an antibody against mouse endosialin and to clarify the pattern of protein expression in fetal and adult mouse tissues, as well as blood vessels of experimental cancer models in the mouse. The emphasis was placed on a systematic tissue analysis, complemented by mRNA quantification and co-typing with standard endothelial markers for selected tissues.
Generation of a polyclonal antibody to mouse endosialin
A series of polyclonal rabbit antibodies (Ab) was raised against unique peptide epitopes of the extracellular domain of mouse endosialin, and the best candidate, Ab 171, was selected for affinity-purification and detailed profiling. Ab 171 recognizes the predicted, differentially glycosylated forms of mouse endosialin (Figure 1, panel A) in immunoblot experiments, notably the 90-kDa, 110-kDa and 160-kDa protein species, using extracts of mEs-transfected 293T cells, but not in non-transfected 293T control cells. Pre-absorption of Ab 171 with the cognate peptide abolished this reactivity. Specific binding to cell surface-expressed mouse endosialin was detected by immunocytochemistry (Figure 1, panel B), using cytospin preparations of intact mEs-transfected 293T cells, as compared to non-transfected 293T cells. Importantly, we found that Ab 171 is suitable for immunohistochemical detection of endosialin in paraffin-embedded mouse tissue sections, again showing specific blocking with cognate peptides (Figure 1, panel C), as this allows higher resolution mapping of the endosialin protein localization.
Immunohistochemical analysis of embryonic mouse tissues
The analysis of endosialin protein expression in the mouse embryo covered stages E10.0 to E17.5. At embryonic stage E10.0, endosialin expression was exclusively seen in the endothelial cells of the dorsal aorta (Figure 1, panel D). The endothelial cells covering the cavity of the heart did not show any immunoreactivity with Ab 171 (Figure 1, panel E), in contrast to the general endothelial marker PECAM-1/CD31, which was present in the endothelium of the dorsal aorta (Figure 1, panel F) as well as in endocardium (Figure 1, panel G). Co-typing with antibodies to platelet-derived growth factor receptor-β (PDGFR-β) and α-smooth muscle actin (αSMA) identified the presence of pericytes expressing both markers, surrounding the endothelium of the dorsal aorta at this stage (not shown).
During stages E10.5 and E12.0, a prominent perineural vascular plexus develops in the head region (Figure 2, panels A and B) which is endosialin positive (Figure 2, panel B). Angiogenic sprouts from this perineural plexus invade the proliferating neuroectoderm by embryonic day E11.0-E12.0. These newly formed vessels are elongated, give rise to many branches that anastomose with adjacent sprouts to form a plexus of immature capillaries in the developing brain. The extent of the perineural vascular plexus and its angiogenic sprouts were visualized by staining with an Ab directed to the basement membrane component collagen type IV, as illustrated in Figure 2, panel A. At E10.5, the majority of the vessels of the perineural plexus and those in close proximity to the outer layer of the brain showed strong endosialin expression (Figure 2, panel B, arrows), whereas only a small subset of vessels was endosialin-positive deeper in the brain parenchyma (Figure 2, panel B, arrowheads).
In the E12.0 embryo, the vascular system has developed considerably in the entire embryo and in particular in the brain. Numerous vessels sprouting from the perineural plexus into the brain can be observed at this stage (Figure 2, panels C-E). A comparison of the endosialin expression pattern (Figure 2, panel C) with the endothelial marker VEGFR-2 (Figure 2, panel D) revealed that all vessels in the perineural space, as well as those inside the brain, are positive for both markers. In contrast, αSMA was only observed in the larger vessels of the perineural space, and not in the small vessels inside the brain (not shown). The lack of αSMA in these vessels does not indicate that they are immature, since an Ab to PDGFR-β showed the expected pericyte labeling (Figure 2, panel E) in subsets of capillaries at this developmental stage. These findings are in agreement with previous reports indicating the lack of αSMA staining on most capillary-sized vessels in the mouse brain (9), and the presence of solitary PDFGR-β-positive pericytes in proximity to capillaries in mouse embryos by in situ hybridization (10).
Double-labeling studies on brain sections from E12.5 embryos (Figure 3) revealed that the endothelial cells lining the vessels co-expressed endosialin and VEGFR-2, whereas PDGFR-β was expressed by perivascular pericytes that did not show endosialin expression. For endosialin, we noted a predominantly abluminal distribution on the endothelial cells, compared with the preferentially apical localization of VEGFR-2 on the endothelial cells (Figure 3, panel A). Furthermore, we found as a consistent feature that PDGFR-β expressing, endosialin-negative pericytes (marked by arrowheads in Figure 3, panel B) come into close contact with the endosialin-positive, PDGFR-β negative endothelial cells, yet are spatially separated as revealed by the merged immunofluorescence images (Figure 3, panel B; right panel). Finally, in more mature vessels of the perineural plexus, endosialin-positive endothelial cells were found to be embedded in the basal membrane, as identified by labeling of collagen type IV (Figure 3, panel C).
By mid-gestation (E13.5 - E14.5), endosialin was detected in clusters of mesenchymal cells in the head region and surrounding the developing structures of the genitourinary system as shown in Figure 4, panel A. Furthermore, expression was observed in organs such as lung and salivary gland, where endosialin-positive, fibroblast-like cells were seen at the epithelial-mesenchymal interface, as shown for lung development in Figure 4, panel B. In some of the buds, a gradient of staining intensity was seen with strong expression by fibroblasts at the tip of the bud, and weaker expression at the stem. The epithelial cells lining the lung buds, as well as the blood vessels in the interstitial tissues, were endosialin-negative (Figure 4, panel B). Prominent endosialin expression at epithelial-mesenchymal interfaces was also observed in areas of epithelial folding, for instance during the formation of the optic cup (Figure 4, panel C), in the oropharynx, and surrounding the hair follicles (not shown).
By late-gestation (E17-17.5), clusters of endosialin-positive mesenchymal cells are seen in the folding mucosa of the gastric cavity (Figure 4, panel D). These cells are located in the lamina propria, underneath the gastric epithelium. Endosialin-expressing mesenchymal cells were also observed in the dermis, notably around hair follicles, and as dense bands of mesenchymal cells separating skeletal muscle fibers (Figure 4, panel D). In addition, endosialin staining was detected in subsets of cells in immature glomeruli (Figure 4, panel E), putatively identified as mesangial cells, a specialized type of pericyte in the kidney. No endosialin was detected at this stage in the lung (Figure 4, panel F), in contrast to αSMA, which was present in the smooth muscle layer of blood vessels and in myoepithelial cells surrounding the bronchi (Figure 4, panel G).
Immunohistochemical analysis of endosialin in newborn mouse tissues
Expression of endosialin in brain capillaries (Figure 5, panel D) of newborn mice coincides with the peak of angiogenesis in the early postnatal period (11). In addition, small clusters of endosialin-positive cells are seen in the renal glomeruli at this stage (Figure 5, panel E). By comparison, co-typing for CD31 showed a broader expression of this general endothelial marker in peritubular vessels and glomerular capillaries (Figure 5, panel F). Most other newborn organ systems tested lacked any endosialin expression, as illustrated for the liver, small intestine, and heart in Figure 5, panels A-C.
Immunohistochemical analysis of adult mouse tissues
In the adult mouse, endosialin was undetectable in all of the blood vessels of the organs and tissues examined, including cerebellum, cerebral cortex, liver and lung (Figure 5, panels G-L). In the kidney, the putative mesangial cells of the glomeruli remained endosialin-positive (Figure 5, panel K), but the proportion of labeled cells appeared to decrease compared to earlier developmental stages, and some labeled cells mapped to the juxtaglomerular apparatus. In female mice, strong endosialin staining was observed in the mesenchyme of the uterus (Figure 5, panel L).
Real-time RT-PCR for quantification of global endosialin mRNA patterns
Endosialin mRNA levels determined by quantitative, real-time RT-PCR corroborated the results of our immunohistochemical analysis. As shown in Figure 6, endosialin mRNA expression increases during development in whole embryos covering stages E11.5 to E14.5, with a drop in levels for most newborn and adult tissues, except the newborn kidney and adult uterus.
Immunohistochemical analysis of neoangiogenic endothelium in cancer models
Subcutaneously growing xenografts of human epithelial cancers in mice were analyzed for the expression of endosialin in the host-derived tumor stroma. In total, 15 different xenograft models were analyzed, including colorectal (Colo205, HCT116, HT29, LoVo and DLD1), breast (MCF7, MDA-MB231, MDA-MB468), lung (Calu6, NCI-H520, NCI-H157), prostate (DU145, PC3), pancreatic (AsPc1), and ovarian (SKOV3) carcinomas. In all cancer models, endosialin expression was observed in subsets of tumor-associated blood vessels, as shown in Figure 7, panels A and D. Co-typing with the standard endothelial marker, CD31 (Figure 7, panel B), suggested that endosialin can be assigned to neo-angiogenic capillary endothelial cells in these models, rather than to pericytes, as illustrated in Figure 7, panels A and B, for a colorectal cancer. In addition, scattered activated tumor stromal fibroblasts, located in close proximity to tumor cell clusters, also express endosialin (Figure 7, panel C). While similar patterns were observed in the various cancer models studied, the extent of endosialin induction in tumor angiogenesis appeared to vary widely, with no correlation to cancer type or histological growth patterns apparent from this survey.
The study of endosialin or TEM1/Tem1 (1, 3), a putative C-type lectin-like cell surface receptor (2), as a marker of neo-angiogenesis and tumor stroma formation in human cancers and experimental mouse tumors (1, 3, 4, 6) is hampered by the lack of understanding concerning its molecular function, but has gained momentum with the recent description of a targeted disruption of Tem1 in mice (12). Thus, Nanda et al. provide initial evidence that this protein is dispensable for normal development in Tem1-/- mice, for wound healing, and for subcutaneous tumor growth, but markedly modulates invasiveness and metastatic progression in an orthotopic, abdominal xenograft model of HCT116 colorectal cancer cells in immunodeficient mice.
There is currently no simple mechanistic model to accommodate these observations of Nanda et al. (12) regarding growth and progression of abdominal tumors in the Tem1-/- mice, and several levels of complexity have been identified and are of relevance for our present report on endosialin patterns in developing mouse organs and mouse cancer models.
First of all, as pointed out in the initial description of endosialin expression in human cancers, this molecule is expressed selectively in tumor endothelial cells but also, with a variable pattern, in activated tumor stromal fibroblasts (1), a cell type distinct from resting fibroblasts and previously shown to express the cell surface serine protease FAPα (4). We now show that this principle also holds for mouse endosialin, which can be found in tumor vascular endothelial cells and tumor stromal fibroblasts. Accordingly, it remains to be determined whether one or both of these cell types mediate the observed effects in the abdominal tumor model. Double-labeling for distinct stromal cell markers has allowed us to confirm our previous conclusion, based on human cancer studies, that endosialin is expressed by the endothelial cells proper and not by the pericytes that are found immediately juxtaposed to the endothelial cells. The fact that endosialin shows a predominantly abluminal localization in the endothelial cells, which differs from an apical endothelial marker such as VEGFR-2, has meant that the high resolution and tissue preservation possible in the formalin-fixed paraffin-embedded samples, the availability of our polyclonal antibody suited to study paraffin-embedded tissues, and the analysis of merged immunofluorescence patterns were all necessary to make this distinction.
Secondly, a distinctive feature of endosialin expression in human cancer tissues has been its heterogeneous presence on tumor endothelial cells, with marked differences between samples of a given histological type and also between different areas of a given cancer lesion (1, 13); this heterogeneity in expression in human cancers did not show any obvious correlation to clinico-pathological features. We now show that endosialin expression in developing mouse tissues and, importantly, in a range of experimental cancer models in the mouse, recapitulates this level of complexity. For embryonic mouse tissues (no data available for the corresponding human tissues), this diversity includes differences among vascular structures in distinct regions, with endosialin expression in the endothelium of the dorsal aorta, intersomitic vessels, perineural plexus, and brain capillaries on the one hand, and its absence in many other developing organs and most of the adult organs tested on the other hand. At present, we cannot exclude completely that technical factors contribute to these results; for example, immunohistochemical methods may detect only high levels of endosialin, with lower levels being present more widely on vascular endothelium yet escaping detection. However, the presence of strongly labeled and unlabeled segments of the vasculature in close proximity in the same tissue section are more likely due to clear differences in expression than to subtle quantitative variations; if this is true, it will be of interest in future studies to determine the locally acting factors that control endosialin expression in a spatial and temporal manner. Since endosialin is not a constitutive endothelial cell surface marker, it comes as no surprise that one of the well-explored in vitro cell culture models of endothelial biology, namely human umbilical vein endothelial cells, fails to express endosialin, and shows no induction when stimulated with standard peptide growth and differentiation factors (1). The search for more appropriate cellular models, perhaps involving microvascular endothelial cells, should be rewarding.
Finally, by extending the tissue analysis of endosialin to the developing mouse, we have been able to substantiate the restricted expression pattern seen in normal human tissues (1, 4, 13) and to document the general disappearance of mouse endosialin during early postnatal development, with re-appearance in cancer tissues. In humans as well as in mice, the stromal fibroblasts of the uterus and specialized cells of the kidney glomeruli continue to express endosialin at the adult organ stage. The observation of normal development in Tem1-/- mice (12) does not necessarily mean that endosialin has no developmental function; instead, its role may be taken over by redundant physiological mechanisms in the gene knockout animals. By contrast, the critical steps in abdominal cancer progression identified by Nanda et al. (12) may simply have fewer redundant molecular controls, thus leading to a phenotype that becomes apparent more readily under experimental conditions. Targeted disruptions of other endothelial markers, such as PECAM and CD34 in mice, have also failed to disrupt vasculogenesis on their own (14, 15, 16). Thus, a full understanding of endosialin function in normal development and cancer may not be possible without a clear understanding of its molecular function, overlapping functions carried out by other endothelial and tumor stromal markers, and the identification of the local factors that modulate the distinctive endosialin tissue patterns.
Materials and methods
Animals and cell lines
Animal experiments were conducted in accordance with all applicable institutional guidelines. Mice of C57/Bl6, 129/Bl6 and OF1/SPF background were obtained from the Institute of Biomedical Research (Medical University of Vienna). Pregnant animals were sacrificed on day 10, 10.5, 11, 11.5, 12.5, 13.5, 14, 14.5 and 17.5 after the vaginal plug was determined, and embryos were fixed in 4% buffered formalin at 4˚C overnight, either in toto (E10.0 to E12.5) or after sagittal sectioning, and embedded in paraffin. The following newborn and adult organs were analyzed: cerebrum, cerebellum, lung, heart, liver, stomach, small and large intestine, pancreas, spleen, kidney, adrenal gland, uterus, and placenta.
Human colorectal carcinoma (Colo205, HCT116, HT29, LoVo, DLD1), breast carcinoma (MCF7, MDA-MB231, MDA-MB468), lung carcinoma (Calu-6, NCI-H520, NCI-H157), prostate carcinoma (DU145, PC3), pancreatic carcinoma (AsPc1) and ovarian carcinoma (SKOV3) cell lines were obtained from the ATCC.
For xenograft studies, 1 to 10 x 106 cancer cells were inoculated subcutaneously into the flanks of NMRI nu/nu mice and allowed to form tumors of about 1 cm3. Animals were sacrificed under narcosis, and the tumors were removed, fixed in 4% buffered formalin, and embedded in paraffin.
Antibody generation and purification
For Ab generation ten different peptides from the extracellular domain of mouse endosialin were selected using the DNA Star Protean Sequence Analysis Software (version 4.03). Rabbits were immunized with these peptides using standard immunization protocols. Ab specificity was determined by Western blot and immunocytochemistry, followed by peptide blocking experiments on mouse endosialin-transfected HEK 293T cells and on mouse embryos. The selected Ab 171 was affinity-purified using the SulfoLink™ Kit (Pierce, Rockland, IL) with the cognate peptide, HLDPGDTTSKAHQHP.
The avidin-biotin complex (ABC) immunoperoxidase procedure was used as previously described (1). Briefly, 5 µm-thick sections were cut and mounted on poly-(L-lysine)-coated slides. After deparaffinization and epitope retrieval in 10 mM citrate buffer, the sections were blocked with 10% serum from the host of the secondary antibody and incubated with the primary antibody (Ab 171) at a 1:1000 dilution in PBS/2% BSA for 1 hour at room temperature or the other primary antibodies indicated below. Biotinylated secondary antibodies were added at a 1:100 dilution, followed by Vectastain ABC solution 1:100 (Vector Labs, Burlingame, CA). For the mouse monoclonal antibody to αSMA, the M.O.M. Kit (Vector Labs) was used. Staining with anti-CD31 (BD Pharmingen, Palo Alto, CA) at a 1:500 dilution required amplification with the Dako GenPoint Kit (Dako, Carpinteria, CA) after proteinase K treatment. Finally, slides were incubated in DAB solution (0.06% 3,3’-diaminobenzidine in PBS, 0.003% H2O2) for 2-5 min, dehydrated, and counterstained with hematoxylin. The other primary antibodies used were as follows: anti-VEGFR-2 (Upstate, Lake Placid, NY) at a 1:1000 dilution; anti-PDGFR-β (R&D Systems, Minneapolis, MN) at a 1:100 dilution; anti-collagen type IV (Chemicon, Temecula, CA) at a 1:500 dilution and anti-α smooth muscle actin (Dako, Carpinteria, CA) at a 1:100 dilution. Epitope heat retrieval in 10 mM citrate buffer was carried out for all these antibodies.
Co-localization studies for endothelial and pericyte markers were performed by double immunofluorescence methods. The following antibodies were used: Ab 171 at a 1:200 dilution, anti-VEGFR-2 (Abcam, ab10972, Cambridge, UK) at a 1:15 dilution, anti-PDGFR-β (R&D Systems, Minneapolis, MN) at a 1:50 dilution and anti-collagen type IV (Chemicon, Temecula, CA) at a 1:500 dilution. Epitope heat retrieval was carried out for all of these antibodies as described above. Primary antibodies were incubated for 1 hour at room temperature. Detection was performed with the following secondary antibodies: Alexa 594 donkey anti-goat; Alexa 594 goat anti-rabbit and Alexa 488 goat anti-rabbit (Molecular Probes, Eugene, OR).
Tissue samples were homogenized and total RNA was subsequently prepared with TRI reagent (MRC, Cincinnati, OH). Traces of genomic DNA were digested with Turbo-DNA free (Ambion, Austin, TX). First strand cDNA was synthesized with an input of 1 µg total RNA using oligo d(T)16 primers and Transcriptor Reverse Transcriptase (Roche, Basel, Switzerland). TaqMan probes and primers for mEs and mGAPDH were obtained from Applied Biosystems, Foster City, CA as inventoried assays. TaqMan PCR was done with an ABI PRISM 7000 Sequence Detection System (Applied Biosystems) according to the manufacturer’s instructions. The relative expression of endosialin mRNA was normalized to the amount of mGAPDH in the same cDNA by using the comparative CT method described by the manufacturer.
We thank Anke Baum, Christine Hartmann and Herbert Lamche for valuable contributions.
- Received April 25, 2006.
- Accepted June 2, 2006.
- Copyright © 2006 by Pilar Garin-Chesa