Differentiation and expansion of endothelial cells from human bone marrow CD133+ cells


Davide Soligo, Centro Trapianti di midollo, IRCCS Ospedale Maggiore, Via F. Sforza 35, 20122 Milano, Italy. E-mail: dsoligo@galactica.it


We report a method of purifying, characterizing and expanding endothelial cells (ECs) derived from CD133+ bone marrow cells, a subset of CD34+ haematopoietic progenitors. Isolated using immunomagnetic sorting (mean purity 90 ± 5%), the CD133+ bone marrow cells were grown on fibronectin-coated flasks in M199 medium supplemented with fetal bovine serum (FBS), vascular endothelial growth factor (VEGF), basic fibroblast growth factor (bFGF) and insulin growth factor (IGF-1). The CD133+ fraction contained 95 ± 4% CD34+ cells, 3 ± 2% cells expressing VEGF receptor (VEGFR-2/KDR), but did not express von Willebrand factor (VWF), VE-cadherin, P1H12 or TE-7. After 3 weeks of culture, the cells formed a monolayer with a typical EC morphology and expanded 11 ± 5 times. The cells were further purified using Ulex europaeus agglutinin-1 (UEA-1)-fluorescein isothiocyanate (FITC) and anti-FITC microbeads, and expanded with VEGF for a further 3 weeks. All of the cells were CD45 and CD14, and expressed several endothelial markers (UEA-1, VWF, P1H12, CD105, E-selectin, VCAM-1 and VE-cadherin) and typical Weibel–Palade bodies. They had a high proliferative potential (up to a 2400-fold increase in cell number after 3 weeks of culture) and the capacity to modulate cell surface antigens upon stimulation with inflammatory cytokines. Purified ECs were also co-cultivated with CD34+ cells, in parallel with a purified fibroblastic cell monolayer. CD34+ cells (10 × 105) gave rise to 17 951 ± 2422 CFU-GM colonies when grown on endothelial cells, and to 12 928 ± 4415 CFU-GM colonies on fibroblast monolayers. The ECs also supported erythroid blast-forming unit (BFU-E) colonies better. These results suggest that bone marrow CD133+ progenitor cells can give rise to highly purified ECs, which have a high proliferative capacity, can be activated by inflammatory cytokines and are superior to fibroblasts in supporting haematopoiesis. Our data support the hypothesis that endothelial cell progenitors are present in adult bone marrow and may contribute to neo-angiogenesis.

The close association in the development of haemopoietic and endothelial cells (ECs) during embryonic life (Smith & Glomski, 1982; Muller et al, 1994; Garcia Porrero et al, 1995) has led to the hypothesis that the two lineages may derive from a common precursor called haemangioblast. However, adult neo-vascularization is thought to be exclusively caused by angiogenesis, a process derived from the proliferation and formation of new vessels as a result of sprouting from pre-existing blood cells (Schaper et al, 1971; Folkman & Shing, 1992; Risau et al, 1995). Vasculogenesis (the differentiation of ECs from endothelial-haematopoietic progenitor cells) still awaits definite proof (Rafii, 2000).

The circulation of mature endothelial cells in peripheral blood has been demonstrated under normal conditions and those associated with vascular injury an more recently in sickle cell anaemia (Sowemimo-Coker et al, 1989; Gerorge et al, 1992; Hebbel & Vercellotti, 1997; Solovey et al, 1997). Although circulating cells have the capacity to colonize vascular grafts, the contribution of mature endothelial cells to this process is still unknown (Scott et al, 1994; Shi et al, 1994; Rafii et al, 1995; Wu et al, 1995). However, the existence of a bone marrow-derived endothelial cell precursor was suggested strongly by the results obtained in a canine bone marrow transplantation model by Shi et al (1998). Dacron grafts implanted 6–8 months after the transplantation were colonized by endothelial cells that showed only donor alleles at microsatellite analysis.

The isolation of putative endothelial cell progenitors (ECP) from peripheral blood was initially suggested by Asahara et al (1997). Peripheral blood CD34+ cells expressing vascular endothelial growth factor receptor (VEGFR-2) (flk-1/KDR) were cultured on fibronectin-coated plates for 4 weeks, and the attached cells showed a typical spindle-shaped morphology, the uptake of acetylated low-density lipoproteins (Ac-LDL) and the expression of several EC markers (CD34, CD31, Flk-1, Tie-2 and E-selectins). More recently, it has been confirmed that 0·1–0·5% of circulating CD34+ cells express VEGFR-2/KDR receptors and that pluripotent haematopoietic stem cells (HSC) are restricted to this fraction (Ziegler et al, 1999).

CD34+ cells co-expressing VEGFR-2 and the novel haematopoietic stem cell marker AC133 (Miraglia et al, 1997; Buhring et al, 1999; Gallacher et al, 2000) recently designated CD133 (National Center for Biotechnology Information, 2000) have been isolated from peripheral blood, cord blood and fetal liver. When grown in the presence of vascular endothelial growth factor (VEGF) and fibroblast growth factor (FGF-2), these cells give rise to endothelial cell colonies, thus suggesting that this subset of CD34+ cells (VEGFR-2+/CD133+) may play a role in neo-angiogenesis (Peichev et al, 2000). Similar results have been presented by Gehling et al (2000), who generated endothelial cells in liquid cultures supplemented with VEGF, and the novel cytokine stem cell growth factor (SCGF) from CD133+ precursor cells isolated in granulocyte colony-stimulating factor (G-CSF) mobilized peripheral blood. Further evidence of the presence of circulating angioblasts was provided by the work of Lin et al (2000) who, in a group of gender-mismatched bone marrow transplant patients, observed the growth of two different EC populations, one rapidly expanding at the onset of culture and the second with greater proliferative potential but a delayed (1 month) outgrowth. The first population had a recipient genotype (possibly mature endothelial cells), whereas the second had a donor genotype (putative angioblasts).

Endothelial cells are present in adult bone marrow and can be stained with various endothelial-specific antibodies in trephine biopsies, including CD34 (Fina et al, 1990), von Willebrand factor (VWF; Wagner et al, 1982) and UEA-1 lectins (Holthofer et al, 1982). Endothelial cells are also present in the adherent layer of long-term bone marrow cultures (LTBMC), where they have been found in percentages of up to 3% (Schweitzer et al, 1995). Bone marrow ECs have been isolated using a variety of techniques (Irie & Tavassoli, 1986; Fei et al, 1990; Masek & Sweetenham, 1994; Schweitzer et al, 1995). They can grow to confluence, have a typical spindle-shaped morphology and have several markers of mature endothelial cells, such as VWF, endothelial leucocyte adhesion molecule (ELAM-1), vascular cell adhesion molecule (VCAM-1), E-selectins and CD31. During culture, they progressively lose some markers such as CD34 (Delia et al, 1993) and, after 4–6 weeks, their viability decreases slowly (Masek & Sweetenham, 1994; Schweitzer et al, 1995). These phenotypic and functional features are consistent with the isolation of mature microvascular endothelial cells.

The aim of this study was to test whether anti-CD133 antibodies, which react with early haematopoietic progenitor cells but do not bind to mature ECs (Yin et al, 1997), can be used to isolate a subset of bone marrow cells that can give rise to a pure endothelial cell population when grown with an appropriate combination of growth factors. We demonstrate that these cells have an high proliferative potential and share the phenotypic and functional properties of cells derived from the putative EC progenitors (angioblasts).

Materials and methods

Isolation and culture of CD133+ bone marrow cells After informed consent, bone marrow samples were collected from healthy donors for allogeneic bone marrow transplantation, placed in sterile tubes containing preservative-free lithium heparin (20 U/ml) and layered on a Ficoll–Paque gradient (specific gravity 1·077 g/ml; Nycomed Pharma, Oslo, Norway). The low-density mononuclear cells (LD-MNCs) were washed twice in Hanks' balanced salt solution (HBSS) and resuspended in Iscove's modified Dulbecco's medium (IMDM) (BioWhittaker, Caravaggio, Italy) supplemented with 10% fetal bovine serum (FBS) (Sigma-Aldrich, Milan, Italy) and 5 mmol/l EDTA.

CD133+ cell separation The haemopoietic stem and progenitor cells were isolated using positive selection of CD133-expressing cells. Briefly, the bone marrow LD-MNCs were incubated for 30 min with the AC133/1 monoclonal antibody (mAb) directly labelled to microbeads (MACS, Miltenyi Biotec GmbH, Bergisch Gladbach, Germany), washed, filtered through a 50-µm nylon mesh to remove clumps and placed on a column in the midiMACS cell separator (Miltenyi Biotec). The labelled cells were separated using a high-gradient magnetic field, and eluted from the column after their removal from the magnet. The positive fraction was then placed on a new column and the magnetic separation step repeated. At the end of the separation, the cells were counted and assessed for viability using Trypan Blue dye exclusion; their purity was determined using a FACSCalibur flow cytometer (Becton Dickinson, San José, CA, USA).

Cultures of CD133+ bone marrow cells CD133+-enriched cells (3 × 105 ) were plated on fibronectin-coated (Boehringer Mannheim, Milan, Italy) 25-T tissue culture flasks, grown in M199 medium (Sigma-Aldrich) supplemented with 10% FBS, 50 ng/ml vascular endothelial growth factor (VEGF), 1 ng/ml basic-fibroblast growth factor (b-FGF) and 2 ng/ml insulin-like growth factor-1 (IGF-1) (all from Peprotech, London, UK), and cultured to confluence for 3–4 weeks at 37°C in a fully humidified atmosphere and 5% CO2.

Purification and expansion of UEA-1 endothelial cells To purify the endothelial cells further, the cultures were treated with trypsin-EDTA 2·5% (Sigma-Aldrich) for 6 min and washed twice in IMDM with the addition of 10% FBS. The recovered cells (an average of 2–4 × 106) were incubated with fluorescein isothiocyanate (FITC)-labelled lectins from Ulex europaeus agglutinin-1 (UEA-1) (Sigma-Aldrich) for 15 min at 4°C, washed, incubated with anti-FITC microbeads (MACS, Miltenyi) for 15 min at 4°C, and separated using immunomagnetic sorting. A mean of 8·1 ± 3·1 × 103 purified cells were plated on fibronectin-coated 25-T tissue culture flasks and cultured in M199 supplemented with 10% FBS, 10 ng/ml VEGF, 1 ng/ml bFGF and 2 ng/ml IGF-1. The cells were then cultured to confluence (generally 3–4 weeks), resuspended by trypsin digestion, and had their viability, expansion rate and phenotype assessed.

Immuno-cytochemistry and flow cytometry

Antibodies Biotin and FITC-labelled lectins (UEA-1) (Sigma-Aldrich) and mAbs to factor VIII/VWF (Immunotech, Marseille, France), KDR (Accurate Chemical, Westbury, NY, USA), Tie-2, E-selectin (CD62E), CD105, VCAM-1 (CD106), platelet endothelial cell adhesion molecule-1 (PECAM-1) (CD31) (HQ PharMingen Europe), CD34, CD34 HPCA-2/FITC and phycoerythrin (PE), CD14, CD14 FITC (all from Becton Dickinson, Milan, Italy), VE-cadherin (CD144) (Santa Cruz Biotechnology, Santa Cruz, CA, USA), CD45 (Dako, Milan, Italy), AC133/2 (CD133) (Miltenyi) and P1H12 (Chemicon International, Temecula, CA, USA) were used for immunocytochemical and flow cytometry analysis.

Immunocytochemistry After trypsin digestion, the cells were cyto-centrifuged on to glass slides (Cytospin II, Shandon) for phenotype analysis. The presence of surface markers was detected using the alkaline phosphatase-antialkaline phosphatase (APAAP) technique (Cordell et al, 1984). Briefly, the cytospin preparations were fixed with acetone for 10 min, incubated for 20 min with phosphate-buffered solution (PBS) containing 20% human decomplemented AB serum and 1% bovine albumin, and reacted for 30 min with the selected mAbs. After rinsing with Tris-buffered saline (TBS) pH = 7·6, the cells were incubated for 30 min with rabbit anti-mouse immunoglobulins (Dako) and for 15 min with the APAAP complexes (Dako); these steps were repeated at least three times for signal amplification. The cells were then incubated for 25 min with the alkaline substrate containing naphtol AS-BI phosphate (Sigma Chemical, St. Louis, MO, USA), neo-fuchsin (Fluka Chemie, Buchs, Switzerland), sodium nitrite (Fluka) and 1 mmol/l levamisole (Sigma-Aldrich). The slides were then counterstained with haematoxylin, air-dried, mounted with DPX (Fluka), and observed using a Nikon Labophot-2 light microscope.

Flow cytometry The immunophenotype analysis was performed by staining 2·5 × 105 cells in 50 µl of Roswell Park memorial Institute (RPMI) medium containing 20% FBS with selected primary mAbs for 30 min at 4°C. The samples were then washed twice with PBS. In the case of primary unconjugated mAbs, the cells were incubated with FITC-conjugated goat-anti-mouse antibody (Dako) for 30 min at 4°C and washed twice with PBS. Cell fluorescence was measured by flow cytometry using a FACSCalibur instrument (Becton Dickinson) and the data were analysed using the cellquest software (Becton Dickinson).

Transmission electron microscopy After trypsin digestion, the UEA-1 selected cells after 3–4 weeks of culture were washed in Cacodylate buffer and fixed in 2·5% glutaraldehyde in Cacodylate buffer at 4°C overnight. The cells were then post-fixed for 1 h in 1% OsO4 in Cacodylate buffer at 4°C, washed and rinsed in distilled water. The samples were dehydrated using a graded ethanol series (from 70° up to 100°) and propylene oxide. After embedding in Araldite (Fluka Chemie), ultra-thin sections were obtained using a Reichert Jung ultra-microtome, counter-stained with uranyl acetate and Reynhold's lead citrate, and then observed using a Philips CM10 transmission electron microscope (Philips, Heindoven, The Netherlands) at 80 kV.

Endothelial cell activation The endothelial cells were activated with interleukin-1α (IL-1α) and tumour necrosis factor-α (TNF-α) (both by Peprotech). After trypsin culture treatment, 1 × 105 cells were plated on fibronectin-coated 25 T tissue culture flasks and cultured in M199 supplemented with 10% FBS, 10 ng/ml VEGF, 1 ng/ml bFGF and 2 ng/ml IGF-1. At confluence, either IL-1α (10 ng/ml) or TNF-α (100 UI/ml) was added to the cultures for 4 and 24 h. The cells were then washed, trypsinized, counted and cyto-centrifuged on to glass slides. The expression of CD62E, CD105, CD106, VWF, Tie-2 and KDR was evaluated using immuno-cytochemistry.

Co-cultivation of CD34+ haematopoietic progenitors with purified endothelial and fibroblastic layers: BM fibroblastic layers Purified BM fibroblastic layers were established in T-25 tissue culture flasks by plating normal BM mononuclear cells in IMDM supplemented with 30% FBS, as described by Pittenger et al (1999). After 2–3 weeks of culture, confluent layers were treated with trypsin and 2 × 105 cells were seeded under the same culture conditions. A second trypsinization step was needed to purify fibroblastic cells from ‘contaminating’ haematopoietic cells, and then 2 × 105 selected cells were seeded in IMDM supplemented with 30% FBS and 10 ng/ml b-FGF for 7–10 d up to confluence.

Co-cultures Purified endothelial (UEA-1 selected) and fibroblastic cultures were treated with trypsin and inactivated with mitomycin A (25 µg/ml mitomycin A/1 × 106 cells for 30 min at 37°C); 3·6 × 105 cells/well were seeded in 6-well microplates (Costar, Cambridge, UK). One hundred thousand immunomagnetic-sorted (Miltenyi) normal peripheral blood CD34+ haematopoietic progenitor cells from G-CSF mobilized donors were plated in direct contact with the ECs and stromal layers or in a transwell insert with a 0·4-µm microporous filter membrane (Falcon, Becton Dickinson) placed above the feeder-layer. The cultures were maintained for 3 weeks at 37°C in a humidified atmosphere and 5% CO2, with weekly media changes. The cells harvested from the supernatant and the stroma of the contact and non-contact cultures were counted and replated in methylcellulose in order to determine the number of colony-forming cells (CFC).

Progenitor assays The colony forming unit-granulocyte/macrophage (CFU-GM) and burst forming unit-erythroblast (BFU-E) assays were carried out by plating 5 × 104 cells in a methylcellulose culture medium (MethoCult GF H4434, StemCell Technologies, Vancouver, Canada) containing 0·9% methylcellulose in IMDM, 30% pre-tested FBS, 1% pre-tested bovine serum albumin (BSA), 10−4M 2-mercaptoethanol, 2 mmol/l l-glutamine, 10 ng/ml granulocyte-macrophage CSF (GM-CSF), 50 ng/ml SCF, 10 ng/ml IL-3 and 3 U/ml erythropoietin. Triplicate dishes were incubated at 37°C in a fully humidified atmosphere and 5% CO2. After 14 d of culture, aggregates of geqslant R: gt-or-equal, slanted 40 cells were scored as colonies and counted.

The absolute number of CFU-GM and BFU-E colonies was obtained by normalizing to the total number of the cells harvested under the different culture conditions.

Statistical analysis The results are presented as the mean values ± SD of the data obtained from three or more experiments performed in triplicate. Statistical significance was determined using the Student's two-tail t-test for paired data.


Characterization of the CD133 selected bone marrow cells

After immunomagnetic sorting with anti CD133 mAbs, the low-density bone marrow mononuclear cells had a mean purity of 90 ± 5% at FACS analysis. The AC133+ fraction contained 95 ± 4% of CD34+ cells, and 3 ± 2% of cells expressing KDR (Fig 1), but no cells expressed mature EC (VWF, VE-cadherin, P1H12) or fibroblastic cell markers (TE7).

Figure 1.

Phenotypic flow cytometry characterization of immunomagnetically selected CD133+ cells. The scattergram represents log PE (AC133) versus log FITC (CD34) fluorescence activity. The open histograms are negative controls, whereas the solid histograms represent indirect FITC-labelling with P1H12 or KDR mAbs.

The cells were then grown on plastic fibronectin-coated flasks in the presence of VEGF, b-FGF and IGF-1. After 14 d of culture, the cells had formed isolated colonies, each containing an average of 60 cells whose morphology was characterized by a round central body with some cytoplasmic projections (Fig 2A). After 3–4 weeks, the cultures became confluent and consisted of a monolayer of spindle-shaped cells (Fig 2B), sometimes forming a focal area of cell aggregates with a typical ‘cobblestone’ morphology. At this time, the cells were harvested and counted: there was a mean 11 ± 5-fold expansion of CD133+ cells in three different experiments (Fig 3).

Figure 2.

Phase-contrast micrographs of EC cultures. (A) CD133+ selected cells after 2 weeks of culture. The cells grown on fibronectin form initially isolated colonies and have a central body and short cytoplasmic dendrites (x100). (B) The same cells after 4 weeks of culture reach confluence and have an elongated spindle-shaped morphology (magnification ×100). (C) UEA-1 selected cells at confluence after 3 weeks of culture. The cells have a typical spindle-shaped morphology with focal areas of cobblestone-like cells (magnification ×100). (D) Higher magnification of the cobblestone-like cells (magnification ×400).

Figure 3.

Expansion of the CD133+ selected bone marrow cells after 4 weeks of culture with VEGF, bFGF and IGF-1 (white columns), and of the cells further purified with UEA-1 and grown under the same culture conditions (dark columns). The numbers at the top of the columns indicate the population doublings (three different bone marrow samples).

To confirm the EC phenotype, the cells were analysed extensively on cytospin preparations immunostained using the APAAP method. In three independent experiments, all the cells were CD133 and the majority stained with EC markers (VE-cadherin, VWF, P1H12 and UEA-1) (Table I). In 2/3 experiments, weak CD34 staining was visible (5–37%). A variable percentage of the cells were still CD45+ and CD14+, thus suggesting the persistence of monocytes/macrophages in the different cultures (Table I).

Table I.  Immunophenotype of CD133+ cells after 4 weeks in culture.
  1. Reactivity of mAbs against CD133 sorted cells after culture was evaluated by APAAP staining (results from three different bone marrow samples). The percentage of positive cells was calculated counting a minimum of 400 cells at magnification ×400.

BM 10296560374034
BM 20408046023
BM 3070781351012

Characterization of UEA-1 enriched endothelial cells

To purify the endothelial cell cultures further, 2–4 × 106 cells were incubated with UEA-1/FITC conjugates and separated using immunomagnetic sorting. A mean of 8 ± 3 × 103 purified cells were plated on fibronectin 25-T tissue culture flasks and cultured in the presence of VEGF, bFGF and IGF-1. The cells rapidly adhered to the substrate and grew to confluence within 2–3 weeks (Fig 2C and D). The cells were very similar to those observed in the primary cultures, with a spindle-shaped morphology and a higher number of cobblestone-like areas. However, the expansion capacity of UEA-1 purified cells was much greater: the mean expansion in total cell number was 2228 ± 375-fold in three different experiments (Fig 3). In all cases, the cultures were continued for up to 3 months without any decay in cell viability.

After 3 weeks in culture, the cells were further analysed for the expression of EC-markers: 100% were CD45 and expressed VWF, UEA-1 lectins, CD105, KDR, CD31 and CD144; no cells with monocytic or stromal cell markers were detectable (Table II).

Table II.  Immunophenotype of UEA-1 + selected cells after an additional 3 weeks of culture.







  1. Reactivity of UEA-1 selected cells was evaluated by APAAP staining (results of three different bone marrow samples).

  2. The percentage of positive cells was calculated counting a minimum of 400 cells at magnification ×400.

BM 199·3 ± 0·710010099·6 ± 0·510000099·3 ± 0·700
BM 299·6 ± 0·310010099·3 ± 0·210000099·2 ± 0·600
BM 3100%10010099·2 ± 0·4100000100%00

Transmission electron microscopy showed that they had a typical EC morphology, with multiple pynocytic vesicles at the cell periphery and Weibel–Palade bodies (Fig 4). Finally, in order to detect the receptors and adhesion molecules that are typically modulated by endothelial cells upon stimulation with inflammatory cytokines, the cells were incubated for 4 and 24 h with IL-1α or TNF-α and analysed using immunocytochemistry: E-selectin (CD62E) and Tie-2 were expressed only after 4 and 24 h of incubation with both cytokines, whereas VWF and V-CAM (CD106) were upregulated upon stimulation. The expression of KDR was downregulated at 4 h and upregulated at 24 h (Table III).

Figure 4.

Transmission electron microscopy of (A) two cells with numerous blebs and small vesicles (magnification ×5000); at higher magnification (B), the cell on the left also has a typical Weibel–Palade body (magnification ×114000). (C) This single cell has an overall morphology similar to the previous cells (magnification ×7000) and, at higher magnification (×95300), shows a horseshoe-shaped Weibel–Palade body (D).

Table III.  Expression of ECs receptors upon stimulation with TNF-α.
 04 h24 h
  1. Cultured ECs, either unstimulated and after incubation with 100UI/ml TNF-α for 4 and 24 h, were cytocentrifuged, fixed with acetone, labelled with mAbs and analysed using the APAAP method (results from three different bone marrow samples). –, negative; +, weakly positive; ++, intermediate positivity; +++, strongly positive.


Co-cultivation of haematopoietic progenitor cells on purified endothelial and fibroblastic cells

To determine whether CD133+/UEA-1+ selected bone marrow cells support the growth of haematopoietic progenitor cells, peripheral blood CD34+ cells were seeded on purified endothelial and fibroblast cell monolayers and cultured under contact conditions or in microporous transwell plates for 3 weeks. Both in the contact and in the transwell experiments, the CFU-GM output was always higher on the EC layers than on the pure fibroblast monolayers, even if the differences were not statistically significant (Fig 5A).

Figure 5.

Co-cultures of CD34+ cells with purified endothelial cell (EC) or stromal cell layers (SC). Immuno-magnetically separated CD34+ cells (from G-CSF mobilized peripheral blood) were placed in direct contact with EC or SC (contact cultures) or separated by a 0·4-µm microporous transwell membrane (non-contact cultures). After 3 weeks of culture in Myelocult medium, the cells from the supernatant and the adherent layer (after enzymatic digestion) were seeded in methylcellulose to assess the CFU-GM (A) and BFU-E (B) output (mean value ±SD of three different bone marrow samples, each performed in triplicate). Dotted columns, mean value ±SD of CFU-C obtained from 1 × 105 CD34+ cells seeded in a 3·5-mm well under contact conditions; black columns, mean value ±SD of CFU-C obtained from 1 × 105 CD34+ cells seeded in a 3·5-mm well in transwell cultures.

The BFU-E output was always higher in transwell culture conditions than in contact cultures: in particular, the mean values ±SD of BFU-E colonies were 715 ± 282 versus 65 ± 53 on ECs (P < 0,05) and 388 ± 211 versus 139 ± 29 on stromal cells (Fig 5B). Furthermore, a significant difference between BFU-E output on ECs and SCs was also seen in transwell culture conditions (P < 0,05) (Fig 5B).


We report a method of differentiating, purifying and expanding endothelial cells from human bone marrow.

Microvessel endothelial cells are difficult to separate in adult bone marrow because they represent a very small percentage compared with haematopoietic cells (1–2% in rat bone marrow) (Irie & Tavassoli, 1986), are generally arranged in more complex structures such as sinusoids and can be obtained only after the collagen digestion and filtration of bone marrow spicules (Rafii et al, 1994). Furthermore, it is only recently that EC specific markers have become available; the majority of the previous markers were cross-reactive either with haematopoietic cells (monocytes/macrophages) or fibroblasts. These cells are mature endothelial cells and, once isolated, can only be grown for a limited time (4–5 weeks) and used only for a few passages (Masek & Sweetenham, 1994; Schweitzer et al, 1995).

Recent reports have shown that circulating CD34+ cells co-expressing the novel stem cell marker CD133 and/or KDR can be isolated and differentiated in ECs in vitro (Gehling et al, 2000; Peichev et al, 2000). Using a similar approach, we immunomagnetically separated bone marrow CD133+ cells, plated them on fibronectin-treated plastic flasks and grew them in the presence of VEGF, bFGF and IGF-1. These cells still contained a mixture of haematopoietic cells, fibroblasts and endothelial cells, and were therefore further selected to obtain a pure cell population. The cells grew with a typical morphology, expressed the majority of endothelial cell markers, up-regulated endothelial-specific and adhesion molecules upon stimulation with inflammatory cytokines and showed the presence of typical Weibel–Palade bodies at the ultrastructural level (Weibel & Palade, 1964; Wagner et al, 1982). The proliferation rate of this pure cell population was very high, leading to expansion rates of up to 2500-fold, much higher than those reported for ECs derived from peripheral blood CD34+ or CD133+ stem cells (Shi et al, 1998; Gehling et al, 2000; Peichev et al, 2000). These results may be caused by the intrinsic properties of the selected bone marrow cells giving rise to ECs or a more primitive phenotype of cells not exposed to G-CSF in vivo; however, the different cytokine combinations used (including SCGF)(Gehling et al, 2000) or the fact that the complete absence of accessory cells such as macrophages and fibroblasts leads to selective exposure to EC-specific growth factors may also be considered.

Long-term bone marrow culture studies have shown that BM stroma supports haemopoiesis and predominantly myelopoiesis (Dexter, 1982). However, BM stroma consists of a number of different cell types including fibroblasts, adipocytes, macrophages and a variable number of endothelial cells (Gartner & Kaplan, 1980). The contribution of mature ECs in supporting haemopoiesis and megakaryocytopoiesis is well established (Rafii et al, 1994), and they have been reported to produce a number of growth factors that are known to regulate haematopoiesis, such as FGF and transforming growth factor-β (TGFβ) (Sieff et al, 1987; Segal et al, 1988; Lewinsohn et al, 1990; Arkin et al, 1991) and to express receptors for IL-3, SCF, erythropoietin (Epo), thrombopoietin (Tpo) and bFGF (Voyta et al, 1984; Tavassoli & Aizawa, 1987; Bagby & Heinrich, 1991); they also have adhesion molecules known to interact with haematopoietic cells (Verfaillie, 1998; Roy & Verfaille, 1999). Recently, it has been shown that the co-cultivation of primitive CD34+/CD38 cord blood cells and porcine microvascular ECs leads to the extensive expansion of cells with multilineage engraftment capacity (Davis et al, 1995; Rosler et al, 2000). Our study shows that highly purified ECs support myelo-erythropoiesis better than purified stromal cells derived from mesenchymal cells. In addition, the transwell experiments with purified CD133 cells show that proliferation of committed progenitor cells (both from the myeloid and the erythroid lineage) is higher in non-contact cultures, suggesting that adhesive interactions between ECs and HCs may play a role in the preservation of quiescent primitive progenitors. The final proof obviously needs long-term culture assays (CAFC or LTC-IC) with EC monolayers.

Cells selected from CD133+ bone marrow cells and further purified with UEA-1 lectins may originate from mature endothelial cells, putative haemangioblasts or mesenchymal stem cells (MSCs). The first hypothesis seems unlikely for various reasons: the immunomagnetically separated CD133+ cells were highly purified and this antigen is not expressed on mature ECs (Gehling et al, 2000; Peichev et al, 2000); furthermore, the phenotype was different from that reported in mature ECs, which also have a lower proliferation potential (Masek & Sweetenham, 1994; Rafii et al, 1994; Schweitzer et al, 1995). The cells may therefore arise from primitive bone marrow ‘haemangioblasts’, although their origin in mesenchymal stem cells cannot be excluded a priori. They lack the markers of mature mesenchymal cells, but it is possible that primitive MSCs may be forced to ‘trans-differentiate’ into ECs by specific growth factors. Preliminary evidence that MSCs may be capable of giving rise to ECs has been provided by Reyes et al (1998).

Whatever their origin these cells can be obtained easily and expanded from bone marrow, and can be used for a variety of purposes, such as the ‘ex vivo’ expansion of ECs for cell and gene therapy or for the study of interactions with normal and leukaemic haematopoietic stem cells. ECs isolated from leukaemia bone marrows could further elucidate their contribution to leukaemia progression. Since the first reports by the Folkman group (Hanahan & Folkman, 1996) of an increase in BM microvessel density (MVD) in ALL patients (Perez-Atayde et al, 1997), enhanced angiogenesis has been found in a myeloma (Vacca et al, 1999a), non-Hodgkin's lymphoma (Vacca et al, 1999b), acute myeloid leukaemia (AML) (Hussong et al, 2000) and myelodysplastic syndromes (MDS) (Pruneri et al, 1999). However, although a correlation between angiogenesis and disease progression has been suggested (Folkman et al, 2000), it has never been formally proved. The isolation and purification of bone marrow ECs from haematological malignancies should make it possible to establish whether ECs are derived from neoplastic clonal cells and if they contribute to disease progression. Finally, the co-cultivation of ECs with neoplastic cells ‘in vitro’ will be useful for testing several antiangiogenic compounds, and will also provide an ‘in vitro’ model for testing ‘metronomic’ therapies (Hanahan et al, 2000).


The authors would like to thank to Mr Mario Azzini (Future Image, Milan, Italy) for the photographic artwork. This study was partially supported by the Milan section of the ‘Associazione Italiana contro le Leucemie’ (AIL), Milan, Italy.