Culture and characterization of sinusoidal endothelial cells isolated from human liver

Summary Although most vascular models use large vessel endothelial cells from human umbilical veins, there is marked heterogeneity among endothelial cells from different vascular beds and organs. More accurate modeling of endothelial involvement in liver diseases, including metastasis, may result from the use of human hepatic sinusoidal endothelial cells. Liver resection specimens were sectioned, then treated with a 1.2 U/ml dispase solution. The tissue slurry was mechanically disaggregated and separated by centrifugation on a Percoll density gradient. Cells were then cultured in an endothelial-specific media with growth factors. These techniques resulted in a homogeneous monolayer consistent with endothelial cells by light microscopy. An endothelial origin was further confirmed by the expression of Factor VIII, binding of Ulex lectin, and uptake of acetylated low density lipoprotein. Electron microscopy showed transcellular fenestrations consistent with a sinusoidal origin. These human hepatic sinusoidal endothelial cells were then studied for expression of the adhesion molecules CD31/PECAM, CD34, E-selectin, ICAM-1, L-selectin, LFA-3, P-selectin, and VCAM-1 plus the binding of wheat germ agglutinin lectin. The patterns of adhesion molecule expression and lectin binding by these cells are characteristic of hepatic sinusoidal endothelia. In this paper, we have described a method for isolation and culture of human cells with the morphologic and phenotypic characteristics of hepatic sinusoidal endothelia.     

Optimization of use of UEA-1 magnetic beads for endothelial cell isolation

Most endothelial cells (EC) in the body belong to the microvasculature. Isolation and subsequent culture of these microvessel EC contributes greatly to our understanding of the heterogeneity and vascular specificity that exist between one organ site and another. However, a major obstacle is the overgrowth of contaminating cells (fibroblasts, pericytes, smooth-muscle cells) in cultures. Since 1990 the use of magnetic beads in combination with either a lectin, Ulex europaeus agglutinin-1 (UEA-1), or a monoclonal antibody has represented a powerful tool for the isolation/purification of microvessel EC. In the former case, operative conditions remain to be optimized to obtain pure cultures of EC.We have performed studies to optimize conditions of use for magnetic beads coated with UEA-1. Incubating beads with cells, the influences are studied of time, temperature, cell concentration, and number of beads per target cell for two cell types, human umbilical vein EC (HUVEC) and skin fibroblasts (HSF), either isolated or mixed. The effect of the last parameter was also checked on the behavior of cells undergoing proliferation after isolation. Results, expressed as isolation efficiency (from 40% to 90%) allowed us to select a 15-min incubation time at 4°C with rotary agitation, an optimal concentration of 4 x 10⁵ cells/ml, and an optimal cell:bead ration of 1:3. From a mixed cell population and in these conditions, even very low HUVEC:HSF proportions of 2.5:97.5 allowed us to obtain a pure HUVEC population in subsequent culture.Abbreviations UEA-1 Ulex europaeus agglutinin-1 - EC endothelial cells - HUVEC human umbilical vein endothelial cells - HSF human skin fibroblasts - MPC magnetic particle concentrator - IE isolation efficiency     

Targeted gene delivery to pulmonary endothelium by anti-PECAM antibody

To achieve efficient systemic gene delivery to the lung with minimal toxicity, a vector was developed by chemically conjugating a cationic polymer, polyethylenimine (PEI), with anti-platelet endothelial cell adhesion molecule (PECAM) antibody (Ab). Transfection of mouse lung endothelial cells with a plasmid expression vector with cDNA to luciferase (pCMVL) complexed with anti-PECAM Ab-PEI conjugate was more efficient than that with PEI-pCMVL complexes. Furthermore, the anti-PECAM Ab-PEI conjugate mediated efficient transfection at lower charge plus-to-minus ratios. Conjugation of PEI with a control IgG (hamster IgG) did not enhance transfection of mouse lung endothelial cells, suggesting that the cellular uptake of anti-PECAM Ab-PEI-DNA complexes and subsequent gene expression were governed by a receptor-mediated process rather than by a nonspecific charge interaction. Conjugation of PEI with anti-PECAM Ab also led to significant improvement in lung gene transfer to intact mice after intravenous administration. The increase in lung transfection was associated with a decrease compared with PEI-pCMVL with respect to circulating proinflammatory cytokine (tumor necrosis factor-alpha) levels. These results indicate that targeted gene delivery to the lung endothelium is an effective strategy to enhance gene delivery to the pulmonary circulation while simultaneously reducing toxicity.     

Absence of OX-43 antigen expression in invasive capillary sprouts: identification of a capillary sprout-specific endothelial phenotype

Endothelial cells exhibit a number of unique phenotypes, some of which are angiogenesis dependent. To identify a capillary sprout-specific endothelial phenotype, we labeled angiogenic rat mesentery tissue using a microvessel and capillary sprout marker (laminin), selected endothelial cell markers (CD31, tie-2, and BS-I lectin), and the OX-43 monoclonal antibody, which recognizes a 90-kDa membrane glycoprotein of unknown function. In tissues that were stimulated through wound healing and compound 48/80 application, double-immunolabeling experiments with an anti-laminin antibody revealed that the OX-43 antigen was expressed strongly in all microvessels. However, the OX-43 antigen was completely absent from a large percentage (>85%) of the capillary sprouts that were invading the avascular tissue space. In contrast, sprouts that were introverting back into the previously vascularized tissue retained high levels of OX-43 antigen expression. Double-labeling experiments with endothelial markers indicated that the OX-43 antigen was expressed by microvessel endothelium but was absent from virtually all invasive capillary sprout endothelial cells. We conclude that the absence of OX-43 antigen expression marks a novel, capillary sprout-specific, endothelial cell phenotype. Endothelial cells of this phenotype are particularly abundant in capillary sprouts that invade avascular tissue during angiogenesis.     

Identification of UEA I-binding surface glycoproteins of cultured human endothelial cells

Ulex europaeus agglutinin (UEAI) binds mainly to endothelial cells in human tissues. In cultured human umbilical vein endothelial cells TRITC-UEAI gave an even surface staining but no binding to pericellular material. After permeabilization of the cells UEAI decorated the Golgi apparatus as a juxtanuclear structure. Electrophoresis of Triton X-100 lysates of 35S-methionine labeled cells bound to lectin agarose beads showed that a similar set of polypeptides was recognized by UEA-I and WGA while distinctly different polypeptides were bound to LcA-agarose. Surface labelling revealed major glycoproteins with Mr 220 kD, 160 kD, 140 kD, 120 kD, 80 kD and 50 kD, most of which could be extracted with Triton X-100. However, only the 140 kD gp, 120 kD gp and 80 kD gp showed binding to UEA 140 kD gp, 120 kD gp and 80 kD gp showed binding to UEA I-lectin. The results show that among a distinct set of surface glycoproteins in cultured human endothelial cells only a few have alpha-l-fucosyl moieties capable of binding to UEAI lectin.     

Identification of a subpopulation of human renal microvascular endothelial cells with capacity to form capillary-like cord and tube structures

Summary Endothelial specialization is a prominent feature within distinct capillary beds of organs such as mammalian kidney, yet immunological markers for functionally distinct subpopulations of cultured endothelial cells from tissue sources such as kidney have not been available. We developed a simple and reproducible isolation and culture procedure to recover human renal microvascular endothelial cells (HRMEC) from the cortex of unused donor kidneys. This procedure yields highly purified preparations of cells that display endothelial markers that include Factor VIII antigen, acetyl-LDL receptors, and determinants that bind Ulex europaeus lectin. HRMEC assemble into capillary-like cord and tube structures when plated on the surface of basement membrane-like matrix (BMM) in media containing phorbol myristate acetate. To further define subpopulations of HRMEC, we generated a panel of monoclonal antibodies and screened for those recognizing cell surface determinants. One monoclonal antibody recovered from this screen recognized a cell surface protein expressed on a subpopulation of HRMEC that we have designated PEC-1 (pioneer endothelial cell antigen-1). Cells expressing PEC-1 extended long, interconnecting filopodial processes in response to phorbol myristate acetate and assembled into capillary-like structures when plated on BMM. Anti-PEC-1 immunoprecipitated proteins of 25 and 27 kDa. Magnetic bead separation of PEC-1 (+) cells selected cells that assemble into capillary-like cord and tube structures. The remaining PEC-1 (−) HRMEC population formed matrix adherent patches. In the kidney, the PEC-1 determinant is expressed on a small subpopulation of microvascular glomerular cells and is prominently expressed on the apical membrane of proximal tubule cells. The PEC-1 determinant discriminates among subpopulations of HRMEC, identifying a subpopulation that contributes to assembly of capillary-like structures.     

An efficient,nonenzymatic method for isolation and culture of murine aortic endothelial cells and their response to inflammatory stimuli

Given the utility of murine models and the physiological and pathological significance of the aortic endothelium, we developed a simplified, nonenzymatic method for isolation and culture of murine aortic endothelial cells (MAECs). Aortic explants were initially cultured on fibronectin-coated plastic. Murine aortic endothelial cells migrated from the explants and proliferated. This expansion allowed for cultures to be established from the aortas of one or three mice. Murine aortic endothelial cells were then purified from expanded cultures by fluorescence-activated cell sorting for the uptake of 1,1'-dioctadecyl-3,3,3',3'-tetramethyl-indocarbocyanine perchlorate-labeled acetylated low-density lipoprotein. The majority of the cells in expanded cultures were as positive as human umbilical vein endothelial cells labeled in the same way. The most positive half of the labeled MAEC population was placed back in culture, and the cells formed "cobblestone" monolayers at confluence. Smooth muscle alpha-actin, which was present in aortic tissue and to a lesser extent in explant cultures before sorting, was not detected in selected MAECs. Western blotting and immunostaining also demonstrated the presence of the endothelial markers, platelet endothelial cell adhesion molecule-1, factor VIII-related antigen, and Bandeiraea simplicifolia lectin 1 binding. Murine aortic endothelial cells retained expected inflammatory functions: vascular cell adhesion molecule-1 protein was induced by bacterial endotoxin, and NO production was synergistically induced by the combination of endotoxin and interferon-gamma. Our simple, efficient method will facilitate investigations of aortic endothelial cell function in vitro using murine models.     

Long-term serial cultivation of arterial and capillary endothelium from adult bovine brain

Summary Cerebrovascular endothelial cells from adult bovine brain were carried successfully in long-term, serial culture. Endothelial cells were obtained from the middle and anterior cerebral arteries and from capillaries isolated from grey matter of the cerebral cortex or caudate nucleus. Capillary cells were found to grow best in RPMI 1640 with 20% fetal bovine serum. They did not require tumor-conditioned medium or matrix-coated surfaces, although fibronectin was used to enhance the initial plating efficiency of the primary cultures. The same conditions were used to support satisfactory growth of arterial endothelial cells; however they did not grow as rapidly as the capillary cells. Retention of endothelial-specific characteristics were shown for capillary-derived cells carried up to Passage 28, arterial-derived cells up to Passage 11, and after frozen storage of both types of cultured cells. Cultures of both arterial and capillary cells stained positively for Factor VIII antigen, exhibited a nonthrombogenic surface, and produced prostacyclin in response to arachidonic acid. Arterial endothelial cells produced more prostacyclin than capillary endothelium. The capillary cells had a unique tendency to assume a ringlike morphology after subculture and sometimes formed capillarylike networks of cell cords in dense cultures. When cultured in a three-dimensional plasma clot, capillary and arterial endothelial cells, but none of the other cell types studied, organized into tubelike structures reminiscent of capillary formation in vivo. The availability of long-term cultures of cerebrovascular endothelial cells provides an opportunity to compare properties of arterial and capillary endothelium from the same tissue and to investigate such processes as angiogenesis and blood-brain barrier induction. This work was supported in part by Public Health Service grant NS-18586 from the National Institute of Neurological and Communicative Disorders and Stroke and by a grant from The Council for Tobacco Research—U.S.A., Inc. C. E. was supported by the Universidad Autonoma of Madrid and the Ministerio de Universidades e Investigacion of Spain.     

Anionic sites,fucose residues and class I human leukocyte antigen fate during interaction of Toxoplasma gondii with endothelial cells

Toxoplasma gondii invades and proliferates in human umbilical vein endothelial cells where it resides in a parasitophorous vacuole. In order to analyze which components of the endothelial cell plasma membrane are internalized and become part of the parasitophorous vacuole membrane, the culture of endothelial cells was labeled with cationized ferritin or UEA I lectin or anti Class I human leukocyte antigen (HLA) before or after infection with T. gondii. The results showed no cationized ferritin and UEA I lectin in any parasitophorous vacuole membrane, however, the Class I HLA molecule labeling was observed in some endocytic vacuoles containing parasite until 1 h of interaction with T. gondii. After 24 h parasite-host cell interaction, the labeling was absent on the vacuolar membrane, but presents only in small vesicles near parasitophorous vacuole. These results suggest the anionic site and fucose residues are excluded at the time of parasitophorous vacuole formation while Class I HLA molecules are present only on a minority of Toxoplasma-containing vacuoles.     

mRNA expression levels of tight junction protein genes in mouse brain capillary endothelial cells highly purified by magnetic cell sorting

Tight junctions (TJs) are an important component of the blood-brain barrier, and claudin-1, -3, -5 and -12 have been reported to be localized at the TJs of brain capillary endothelial cells (BCECs). To understand the contribution of each claudin subtype to TJ formation, we have measured the mRNA expression levels of claudin subtypes (claudin-1 to -23) and other relevant proteins in highly purified mouse BCECs. Mouse BCECs were labeled with anti-platelet endothelial cellular adhesion molecule-1 antibody and 2.3 × 10⁶ cells were isolated from 15 mice by magnetic cell sorting. Expression of Tie-2, Mdr1a and GLUT1 mRNAs was concentrated in the isolated fraction, and contamination with neurons and astrocytes was substantially less than in the brain capillary fraction prepared by the standard glass-beads column method. Expression of occludin, junctional adhesion molecule and endothelial-specific adhesion molecule mRNAs was concentrated in the isolated fraction, suggesting that the corresponding proteins are selectively expressed in mouse BCECs. Among claudin subtypes, claudin-5 was most highly expressed, at a level which was at least 593-fold greater that that of claudin-1, -3 or -12. Expression of mRNAs of claudin-8, -10, -15, -17, -19, -20, -22 or -23 was also concentrated in the isolated fraction, suggesting these subtypes are expressed in mouse BCECs. The levels of claudin-10 and -22 mRNAs were comparable with that of occludin mRNA. These results indicate that claudin-5 is the most abundant claudin subtype in mouse BCECs, and are consistent with the idea that claudin-10 and -22 are involved in TJ formation at the blood-brain barrier in cooperation with claudin-5.     

Primitive endothelial cell lines from the porcine embryonic yolk sac

Endothelial cell lines have been established from cells that were isolated from porcine yolk sacs from day 18 and day 22 embryos and propagated in vitro under various growth conditions. After expansion in vitro, the general properties of the cells proved similar for the different media used. The endothelial cells expressed cell surface receptors for acetylated low-density lipoprotein and also expressed cell surface-associated angiotensin-converting enzyme. The cells showed a characteristically high level of binding for Bandeiraea simplicifolia lectin I and Dolichos biflorus agglutinin but did not bind significant amounts of Ulex europaeus lectin I. The cells expressed low but serologically detectable levels of Class I major histocompatibility complex (MHC) antigens but failed to bind antibodies directed against Class II MHC antigens. Alpha5beta1 integrins were weakly expressed, whereas vascular cell adhesion molecule-1 (CD106) and alphavbeta3 integrins were not detected. Three-dimensional tube formation was readily observed in cultures grown on Matrigel and occurred even in uncoated plastic dishes in the absence of Matrigel. In contrast to most of the adult porcine endothelial cells, yolk sac-derived endothelial cells did not possess serologically detectable receptors for porcine growth hormone (GH), an observation consistent with the finding that GH did not increase the proliferative rate of these cells. Electron microscopic examination demonstrated the presence of Weibel-Palade bodies, tight endothelial cell junctions, and typical rough endoplasmic reticulum. Exposure of the cells to either concanavalin-A-stimulated porcine splenocyte culture supernatants or to human tumor necrosis factor alpha did not cause upregulation of VCAM-1 or Class II MHC antigens. Addition of porcine interferon-gamma led to an increase in the level of expression of Class I MHC. Yolk sac endothelial cells from day 22 embryos showed a low but detectable level of expression of Class II MHC antigens, whereas the endothelial cells from day 18 embryos showed no expression of Class II antigens after interferon-gamma stimulation. The cells maintained competence to develop vascular structures in vitro and could do so after coinjection with murine tumor cells into adult, immunocompromised mice.     

Eosinophil tissue recruitment to sites of allergic inflammation in the lung is platelet endothelial cell adhesion molecule independent

Platelet endothelial cell adhesion molecule (PECAM or CD31) is a cell adhesion molecule expressed on circulating leukocytes and endothelial cells that plays an important role in mediating neutrophil and monocyte transendothelial migration in vivo. In this study, we investigated whether eosinophils, like neutrophils and monocytes, utilize PECAM for tissue recruitment to sites of allergic inflammation in vivo. Eosinophils express similar levels of PECAM as neutrophils as assessed by FACS analysis. RT-PCR studies demonstrate that eosinophils like neutrophils express the six extracellular domains of PECAM. Eosinophils exhibit homophilic binding to recombinant PECAM as assessed in a single-cell micropipette adhesion assay able to measure the biophysical strength of adhesion of eosinophils to recombinant PECAM. The strength of eosinophil adhesion to recombinant PECAM is the same as that of neutrophil binding to recombinant PECAM and can be inhibited with an anti-PECAM Ab. Although eosinophils express functional PECAM, anti-PECAM Abs did not inhibit bronchoalveolar lavage eosinophilia, lung eosinophilia, and airway hyperreactivity to methacholine in a mouse model of OVA-induced asthma in vivo. Thus, in contrast to studies that have demonstrated that neutrophil and monocyte tissue recruitment is PECAM dependent, these studies demonstrate that eosinophil tissue recruitment in vivo in this model is PECAM independent.     

Ulex europeus type I agglutinin detects carcinoembryonic antigen in extracts of human colorectal carcinoma

Recent interest has focused on fucosylated epitopes expressed on human neoplasms. The plant lectin Ulex europus agglutinin, Type I (UEA) binds fucosylated oligosaccharides, while UEA-reactive substances have a tissue distribution similar to carcinoembryonic antigen (CEA). We sought to determine if UEA reacted with CEA in extracts of fresh primary and metastatic colorectal carcinomas and paired normal tissues. The extracts were electrophoretically transferred to nitrocellulose membranes after the proteins were separated by SDS-PAGE in 10% polyacrylamide gels. The transfer membranes were then stained with peroxidase-conjugated UEA (UEA-P) or antibody to CEA (CEA-P). UEA-P reacted with a 170-190-kDa band in extracts of 22 of 30 primary tumors, 10 of 12 metastases, but only 1 of 5 villous adenomas. UEA-P generally did not react with normal colon or liver extracts. UEA-P also did not bind to 170-190-kDa molecules in Western transfers of a breast carcinoma metastatic to bowel and a focal nodular hyperplasia of liver. CEA-P displayed similar reactivity and detected CEA in a tumor extract negative for UEA. Fucose blocked binding of UEA-P to Western transfers of tumor extracts. CEA-P reacted with a 170-190-kDa substance in tumor extracts eluted with fucose from a column of immobilized UEA. Thus, UEA reacts with fucosylated oligosaccharides on most, but not all, species of CEA and may be a useful adjunct to anti-CEA immunohistochemistry.     

Bovine Brain Endothelial Cells Express Tight Junctions and Monoamine Oxidase Activity in Long-Term Culture

The passage of substances across the blood-brain barrier is regulated by cerebral capillaries which possess certain distinctly different morphological and enzymatic properties compared to capillaries of other organs. Investigations of the functional characteristics of brain capillaries have been facilitated by the use of cultured brain endothelial cells, but in most studies a number of characteristics of the in vivo system are lost. To provide an in vitro system for studies of brain capillary functions, we developed a method of isolating and producing a large number of bovine brain capillary endothelial cells. These cells, absolutely free of pericyte contamination, are subcultured, at the split ratio of 1:20 (20-fold increase of the cultured surface), with no apparent changes in cell morphology up to the fiftieth generation (10 passages). Retention of endothelial-specific characteristics (factor VIII-related antigen, angiotensin-converting enzyme, and nonthrombogenic surface) is shown for brain capillary-derived endothelial cells up to passage 10, even after frozen storage at passage 3. Furthermore, we showed that bovine brain capillary endothelial cells retain, up to the fiftieth generation, some of the characteristics of the blood-brain barrier: occurrence of tight junctions, paucity of pinocytotic vesicles, and monoamine oxidase activity.     

A VCAM-like adhesion molecule on murine bone marrow stromal cells mediates binding of lymphocyte precursors in culture

Two new mAbs (M/K-1 and M/K-2) define an adhesion molecule expressed on stromal cell clones derived from murine bone marrow. The protein is similar in size to a human endothelial cell adhesion molecule known as VCAM-1 or INCAM110. VCAM-1 is expressed on endothelial cells in inflammatory sites and recognized by the integrin VLA-4 expressed on lymphocytes and monocytes. The new stromal cell molecule is a candidate ligand for the VLA-4 expressed on immature B lineage lymphocytes and a possible homologue of human VCAM-1. We now report additional similarities in the distribution, structure, and function of these proteins. The M/K antibodies detected large cells in normal bone marrow, as well as rare cells in other tissues. The antigen was constitutively expressed and functioned as a cell adhesion molecule on cultured murine endothelial cells. It correlated with the presence of mRNA which hybridized to a human VCAM-1 cDNA probe. Partial NH2 terminal amino acid sequencing of the murine protein revealed similarities to VCAM-1 and attachment of human lymphoma cells to murine endothelial cell lines was inhibited by the M/K antibodies. All of these observations suggest that the murine and human cell adhesion proteins may be related. The antibodies selectively interfered with B lymphocyte formation when included in long term bone marrow cultures. Moreover, they caused rapid detachment of lymphocytes from the adherent layer when added to preestablished cultures. The VCAM-like cell adhesion molecule on stromal cells and VLA-4 on lymphocyte precursors may both be important for B lymphocyte formation.     

Characterization of CD34+ cells isolated from human fetal lung

The large capillary mass of the newborn lung demands the presence of endothelial cell precursors in lung tissue before development of the pulmonary capillary bed. The objective of this investigation was to isolate and characterize putative endothelial cell precursors from developing human lung. CD34, a cell surface marker for hematopoietic progenitor cells, endothelial precursor cells, and small vessel endothelial cells, was employed as an immunological "handle" for the selection of the desired cells. When CD34+ cells were isolated from midtrimester human fetal lung tissue, then maintained in culture, the isolated cells expressed immunoreactivity for the endothelial cell marker von Willebrand factor and the vascular endothelial growth factor receptors KDR and Flt-1. However, only 5% or fewer of the cells expressed PECAM, an important factor in cell-cell interactions and a marker for endothelial cells associated with vessels. The CD34+ cells endocytosed acetylated low-density lipoprotein and formed capillary-like structures when incubated in a cushion of Matrigel. RT-PCR analysis of mRNA for endothelial cell-related proteins Flt-1, Tie-2, and endothelial nitric oxide synthase demonstrated expression of these mRNAs by the isolated cells for at least 16 cell passages. These observations demonstrate that capillary endothelial cell precursors can be isolated from developing human lung and maintained in cell culture. These cells represent a potentially important tool for investigating the regulation of mechanisms governing development of the air-blood barrier in the human lung.     

Effects of endothelial basement membrane on neutrophil adhesion and migration

The influence of the sub-endothelial basement membrane (BM) on the adhesion and migration of leukocytes is not well-defined. We therefore investigated the behaviour of human neutrophils on purified BM proteins and on BM deposited by short- or long-term cultures of endothelial cells (EC). The adhesion, but not migration velocities, of neutrophils activated with interleukin-8 was dependent on the coating concentrations of purified collagen, laminin or fibronectin. In contrast, adhesion was similar on matrices deposited by 3-day or 20-day cultures of EC, but neutrophils migrated more slowly on the distinct BM that formed over 20 days. In addition, while adhesion on all surfaces was greatly reduced when neutrophils were treated with antibody against β₂-integrins, antibody against β₁-integrins only inhibited adhesion to the 20-day BM. Thus, the native BM has distinct effects on integrin usage and migration by neutrophils, which are not reproduced by purified proteins or matrix deposited early during endothelial culture.     

Histogenesis of Ewing's sarcoma. An evaluation of intermediate filaments and endothelial cell markers

Four cases of Ewing's sarcoma, three in bone and one from an extraskeletal site, were studied immunohistologically using monospecific antibodies against intermediate filament proteins of keratin, vimentin, desmin and neurofilament types. All cases were also evaluated for the presence of Factor VIII-related antigen (FVIIIR:Ag) and for the binding of Ulex europaeus I lectin (UEA I), both of which are endothelial markers. In all cases the tumor cells contained vimentin but not keratin, desmin or neurofilaments. The tumor cells could not be decorated with either anti-FVIIIR:Ag or UEA I, whereas the vascular endothelium was positive for both markers. The vimentin-positivity indicates a mesenchymal derivation of Ewing's sarcoma, while the lack of endothelial markers argues against the proposed endothelial origin of this tumor.