Circulating mouse Flk1+/c-Kit+/CD45- cells function as endothelial progenitors cells (EPCs) and stimulate the growth of human tumor xenografts

Mouse tissue were isolated and homogenized to generate a single cell suspension as described. Cells were analyzed by flow cytometry by first gating on single cells and then on viable cells. Samples were then processed in a sequential manner to isolate cells that were positive for the endothelial marker Flk-1 (VEGFR2), then for the stem cell marker c-Kit positive fraction. The dual population of Flk-1+/c-Kit+ cells was then analyzed for CD45 expression and only those cells that were CD45 negative were selected for further assays. Figure 1A and B demonstrate isolation of unstained mouse aorta/vena cava suspensions and isotype control samples which do not show any viable populations of Flk-1+/c-Kit+/CD45- cells, as expected. Figure 1C demonstrates mouse aorta/vena cava samples with a small “tail” population positive for Flk-1. The Flk-1 positive fraction of cells demonstrated higher levels of c-Kit + staining, and when this subset of cells was analyzed for CD45, slightly less than half the cells were CD45 negative. Similar results were obtained for normal mouse lung tissue, Figure 1D. Overall, the Flk-1+/c-Kit+/CD45- cells made up a very small population of the total number of cells analyzed from the tissue; as seen in Figure 1E, quantification of this population showed it to be approximately 0.5% of the total cellular population. Given the small quantities of cells isolated from the mouse aorta/vena cava and lung samples, we screened a wide variety of mouse tissues in order to determine which organs would provide the highest yield of cells. Figure 1F shows that the adipose, aorta/vena cava and lung samples were the highest yielding tissues. Tissue such as brain, liver, muscle, and blood had the lowest levels of Flk-1+/c-Kit+/CD45- cells. Interestingly, the collected mouse adipose tissue had the highest level of Flk-1+/c-Kit+/CD45- cells found; however, subcutaneous fat was difficult to obtain from the young nude mice used in this study and fat collections from 3 different nude mice were used to generate one “sample” of fat tissue and three samples were then used for analysis. Furthermore, only 50,000 cells could be isolated from this conglomerate adipose tissue sample for FACS analysis in contrast to the 100,000 cells readily available from the other tissues.

After the identification of a small resident population of Flk-1+/c-Kit+/CD45- cells within most mouse organs, we determine whether human tumor xenografts also had baseline levels of this cellular subpopulation. A similar FACS analysis was performed on several human tumor xenografts. As shown in Figure 2A, in vitro samples of U251 human glioma cells were not reactive to either anti-mouse or anti-human antibodies. However, using in vivo U251 tumors, mouse antibodies directed against Flk-1+/c-Kit+/CD45- populations found baseline levels of approximately 0.35%. Anti-human antibodies directed against the same markers did not react with any cells, indicating that the population of interest was derived from the host animal and not derived from the tumor cells themselves. Further experiments (Figure 2B) demonstrated increased levels of the selected EPC populations of cells in A549 lung carcinoma and UMSCC-1 head and neck carcinoma cell line at approximately 1.45% and 1.40%, respectively, compared to approximately 0.50% found in the U251 human glioma cell lines. This result suggested that the resident population of Flk-1+/c-Kit+/CD45- cells is tumor type independent.

We performed additional studies to determine if the selected subpopulation of Flk-1+/c-Kit+/CD45- cells demonstrated characteristics consistent with endothelial progenitor cells. Figure 3A (panel 1) demonstrates the membrane labeling of CD31, a known extracellular membrane marker of mature endothelial cells, as well as the peri-nuclear uptake of acetylated low density lipoprotein (acLDL), which is readily taken up by the HUVECs endothelial cell control line. Figure 3A (panel 2) demonstrates that the tumor cell line A549 does not show the CD31 membrane marker or uptake of acLDL. From our flow cytometry studies we were able to isolate the mouse lung tissue Flk-1+/c-Kit+/CD45- cell line and grow them in culture. This initial population was referred to as the EPC cell line. As seen in Figure 3B (panel 1), only about 10-20% of these cells have the CD31 marker and acLDL dual-labeling. As the acLDL fluorescent marker is non-toxic to cells, we took the acLDL-labeled EPC cell line, trypsinized the cells, and re-analyzed them by flow cytometry to select for an enriched population of acLDL + cells. As seen in Figure 3B (panel 2), the majority of cells have dual staining for CD31 and acLDL. This enriched population is referred to as the EPC_acLDL cell line. Finally, when the EPC_acLDL cell line approached passages 16–20, the morphology of the cells changed to a more flat appearance, but the cell membranes still labeled with CD31 and continued to demonstrate peri-nuclear uptake of acLDL (Figure 3B, panel 3). We refer to this high passage cell line as the EPC_Late cell line. Thus all three isolated EPC cell lines demonstrated CD31+ staining and uptake of acLDL, both known characteristics of mature endothelial cells.

Additionally, we also sought to determine if our isolated cells lines demonstrated stem-cell like characteristics suggestive of self-replication. Figure 3C (panel 1) demonstrates that isolated mouse bone marrow grown in methylcellulose media had colony forming ability. In contrast, Figure 3C (panel 2) demonstrates the A549 tumor cell line generates a typical flat growth morphology, but does not generate the cluster-like cell formations seen in the control mouse bone marrow samples. Our generated cell lines (EPC and EPC_acLDL) demonstrate cell-cluster formations consistent with the ability to form colonies (Figure 3D, panel 1 and panel 2, respectively). Interestingly, the EPC_Late cell line demonstrated only the flat growth morphology with no cluster-like formations (Figure 3D, panel 3). Additionally, stem cell marker studies using alkaline phosphatase analysis and Oct-1 immunofluorescence were negative in all three cell lines (data not shown). Finally, a classic marker of endothelial function is the ability to form capillary-like structures when grown on matrigel coated plates. Figure 3E demonstrates the formation of capillary-like structures from HUVECs (panel 1) and lack of these structures in the A549 cell line when grown on matrigel. All three of our isolated cell lines (EPC, EPC_acLDL, and EPC_Late) were all able to generate capillary-like structures on matrigel plates.

After generating the three cell lines as above, we then checked the cell lines for the original markers they were selected for to determine if these markers stayed intact in vitro or changed over time. Figure 4A demonstrates that the EPC cell line had low levels of Flk-1 expression, low levels of c-Kit expression, no significant CD45 expression, and high levels of CD31 expression. The enriched population, EPC_acLDL, demonstrated increased Flk-1 expression, enhanced c-Kit expression, no significant CD45 expression, but maintained high CD31 expression. Finally, our high passage population, EPC_Late, demonstrated Flk-1 expression, lower levels of c-Kit, no significant CD45 expression, and also maintained CD31 expression. These results suggest that the EPC_acLDL cell line is enriched for EPC functionality and that the EPC_Late population has a more mature endothelial phenotype, but may have less stem-cell like properties.

EPC are considered to give rise to mature endothelial cells but also maintain the ability to self-replicate. However, EPCs are assumed to be downstream of a hemangioblast cell line and should be independent of hematopoietic cell lines. Thus, we undertook a negative study of our EPC cell lines to determine if they had the ability to repopulate the hematopoietic cell lines after lethal whole body irradiation. In Figure 5A, 9 Gy of whole body radiation (2 × 4.5 Gy, given 4 h apart) resulted in over 90% animal death on day 20. However, if normal healthy marrow (3×10⁶ cells) was transplanted by retro-orbital injection 2 hrs after the final dose of radiation, then 100% survival was achieved on day 20. Our hypothesis was that the isolated EPC cell lines would not be able to restore hematopoietic function; however, Figure 5B demonstrates that transplantation of 3×10⁶ cells of either the EPC cell line or the EPC_acLDL cell line were able to protect the animals from death and restore hematopoiesis. Interestingly, the EPC_Late cell line was only able to keep approximately 60% of the animals alive after whole body radiation, consistent with the above data that the EPC_Late cell line lacks some of the stem-cell characteristics demonstrated by the other isolated EPC cell lines.

In order to compare our isolated cell lines genetically, we performed a single-bead gene array analysis (Illumina, San Diego, CA) on mouse reference slides and found only 126 genes differentially regulated between the three cell lines (Figure 6A). Clustering studies showed that the EPC and EPC_acLDL cell lines appeared more closely related than the high passage EPC_Late cell line. A full list of genes up and down regulated between the cell lines is presented in Table 1. One gene of potential interest, Sox2 - often associated with stemness, was down regulated in the EPC_acLDL (3.6 fold) and EPC_Late (4.7 fold) samples as compared to the EPC parental cell line. A similar set of genes were up and down regulated between the EPC and EPC_acLDL cell lines and a greater variety of genes were changed in the EPC_Late cell line versus the EPC parental cell line. These data indicate a close relationship between all three cell lines and that the EPC_Late population is a more mature cell line with less stem cell-like characteristics.

Finally, we wanted to determine if circulating EPCs could enhance tumor growth. First, we co-injected U251 tumor cells (3×10⁶) with the EPC_acLDL cell line (3×10⁵) at a 10:1 ratio. Figure 7A demonstrates that local co-injection of the two cell lines enhanced tumor growth and increased the efficiency of tumor take (3/5 tumors [60%] in mice with U251 cells only and 5/5 tumors [100%] in the combined U251/EPC_acLDL mixture). Thus EPC appear to enhance tumor growth in a human xenograft model. However, local co-injection does not reflect a physiological situation. Therefore, we injected the EPC_acLDL cell line (3×10⁵ cells) systemically at 7 days after the original tumor was implanted subcutaneously (3×10⁶ cells). As seen in Figure 7B, systemic injection of EPC_acLDL cells enhanced tumor growth as compared to PBS vehicle controls. To evaluate the impact of EPC_acLDL injection on vascular growth, we stained tumor samples from the growth curve samples at day 28 using CD31 as a marker of endothelial cells and blood vessels. Figure 7C demonstrates increased numbers of blood vessels in the animals with the systemic injection of EPC_acLDL cells versus control tumors. Figure 7D is the quantification of mean vessel density between the two groups. Tumor blood vessel density was significantly increased in the EPC_acLDL injected animals compared to the control animals. Vascularity was enhanced in both central tumor regions as well as at the periphery (data not shown). These data suggest that systemic injection of EPC_acLDL cells enhances tumor cell growth by increasing the tumor vascular network.

Selected antibodies were acquired as indicated: Rat IgG Isotype control (553993, BD Pharmingen), Rat IgG Fluorescent Isotype controls (sc-2831, sc-3788, sc-2895, sc 2872; Santa Cruz Biotechnology), Rat Anti-Mouse Flk-1-PE (555308, BD Pharmingen), Rat Anti-Mouse c-Kit-PE-Cy7 (558163, BD Pharmingen), Rat Anti-Mouse CD45-APC-Cy7 (557659, BD Pharmingen), Rat Anti-Mouse CD31-FITC (553372, BD Pharmingen), Mouse Anti-Human VEGFR2-PE (560872, BD Pharmingen), Mouse Anti-Human CD117-APC (561118, BD Pharmingen), Mouse Anti-Human CD45-APC-Cy7 (557833, BD Pharmingen), Mouse Anti-Human CD31-FITC (560984, BD Pharmingen), Rat Anti-Mouse CD31 (550274, BD Pharmingen), Goat Anti-Rat FITC (sc-2011, Santa Cruz Biotechnology), Goat Anti-Rat AlexaFluor594 (A-11007, Life Technologies), Anti-Rabbit Oct-4 (AB3209, EMD Millipore), and Goat Anti-Rabbit FITC (sc-2012, Santa Cruz Biotechnology). Additional reagents used: Hoechst’s 33258 Stain (94403, Sigma-Aldrich), acLDL-Dil (L3484, Life Technologies), acLDL-BODIPY (L3485, Life Technologies), acLDL (L35354, Life Technologies), Fibronectin (F1141, Sigma-Aldrich), and Alkaline Phosphatase Stain (A14353, Life Technologies).

HUVECs were grown in human endothelial growth media with supplementation (PM211500, Genlantis, San Diego, CA). The tumor cell lines: A549 lung carcinoma, U251 glioblastoma, and UMSCC-1 head and neck squamous cell carcinoma were grown in DMEM supplemented with 10% inactivated fetal bovine serum (Life Technology) and 1% penicillin-streptomycin. All cell lines were acquired from ATCC (Manassas, VA).

Female athymic nu/nu nude mice (aged 8–12 weeks) were acquired from NCI-Fredrick (Fredrick, MD). Mice were maintained in a germ-free environment and had access to food and water ad libitum. All animal procedures were approved by Stanford University’s Administrative Panel on Laboratory Animal Care (APLAC).

Tumors were implanted subcutaneously on the backs of nude mice approximately 2 cm above the base of the tail. A549, U251, and UMSCC-1 cell lines were implanted at a concentration of 3×10⁶ cells in 100 μL PBS. Tumors reached a size of approximately 200 mm³ in 3–4 weeks with an efficiency of approximately 90%, 70%, and 90%, respectively.

EPC isolation was performed as reported previously with documented modifications [1, 55, 56]. Selected mouse tissues or tumor samples were surgically removed from euthanized nude mice. Tissues were kept in mouse endothelial media (M1168, CellBiologics, Chicago, IL) supplemented with the growth factor kit (VEGF, EGF, heparin, hydrocortisone, and L-Glumatine) and 10% fetal bovis serum (FBS) and then minced into small chunks with a #11 scalpel blade. Tissue fragments were then placed into a glass homogenizer and mechanically disrupted into cellular slurry. The slurry was transferred to 15 mL tubes and mixed with 1 mL of supplemented mouse endothelial media and combined with 2 mL of 0.1% Collagenase I (CLS-1, Worthington Biochem, Lakewood, NJ), 0.1% Collagenase IV (CLS-4, Worthington Biochem, Lakewood, NJ), and 0.001% Deoxyribonuclease I (DPRF, Worthington Biochem, Lakewood, NJ). Samples were incubated for 30 minutes at 37°C. Mouse endothelial medium was added to a total volume of 15 mL and solutions were poured through a 70 μM cell strainer into a new 15 mL tube. Tube were then spun at 4°C at 3000 rpm for 5 minutes. The supernatant was decanted and the resulting cell pellet was resuspended in 2 mL of RBC lysis buffer (1.5 M NH₄CL, 100 mM NaHCO3, 10 mM Na₂-EDTA; pH 7.4) and placed on ice. After 15 minutes, samples were spun down at 4°C at 3000 rpm for 5 minutes, supernatant removed, and resuspended in PBS and kept on ice. Live/Dead stain (L23105, Invitrogen) was added to the cells for 30 minutes with cells kept on ice and protected from light. Samples were then moved to eppendorf tubes and spun for 5 minutes, 4°C at 5000 rpm. Supernatant was removed and the cell pellet was washed twice in PBS, and resuspended in PBS and used for flow cytometry.

Sorted cell from nude mouse lung samples were placed in 5 μg/cm² fibronectin treated T25 flasks in mouse endothelial medium (M1168, Cell Biologics, Chicago, IL) supplemented with the growth factor kit (VEGF, EGF, heparin, hydrocortisone, L-Glumatine, and 10% FBS) and incubated at physiologic concentrations of FiO2 (5%) at 37°C. Flasks were examined every 2–3 days and once attached cells began to generate colonies of 10–20 cells, flasks were moved to normoxic incubators (FiO2 21% at 37°C). As flasks became confluent, cells were passaged as usual. The EPC cell line and EPC_acLDL cell lines were used from passages 3–6 and the EPC_Late cell lines was used between passages 16–24.

Tumor and EPC cell lines were grown and treated in chamber slides (C1782, Sigma-Aldrich). The reagent acLDL-Dil or acLDL-BODIPY was mixed with cells at 2.5 μg/mL concentration and allowed to incubate for 4 hours at 37°C. Medium was then aspirated and cells were fixed in 4% paraformaldehyde for 10 min at room temperature. Paraformaldehyde was aspirated and the cells treated with a 0.2% NP40/PBS solution for 15 min. HUVEC cells and the human tumor cell lines used acLDL-BODIPY and the Anti-CD31-APC antibody while the murine derived EPC-related cell lines used acLDL-Dil and the Anti-CD31-FITC antibody for staining. Cells were then washed in PBS twice, and the selected anti-CD31 antibody at a dilution of 1:50 in 1% BSA was added and incubated overnight at 4°C. Cells were again washed twice in PBS before incubating in the dark with the appropriately labeled secondary antibody at a dilution of 1:100 in 1% BSA for 1 h. The secondary antibody solution was then aspirated and the cells washed twice in PBS. Cells were then incubated in the dark with Hoechst’s 33258 (1 μg/ml) in PBS for 30 min, washed twice, and coverslips mounted with an anti-fade solution (Dako Corp., Carpinteria, CA) and sealed with clear nail polish. Slides were examined on a Lecica DM6000B fluorescent microscope (Lecica, Buffalo Grove, IL). Images were captured by a CCD camera using the Image Pro Premier v9.0 (MediaCybernetics, Rockville, MD) software package.

Tissues were harvest from mice and placed in cassettes and covered in OCT media. Cassettes were held at -80°C until they were process on a cryotome and cut to 2 μm tissue sections. Slides were then kept at -80°C until processing. Slides were allowed to air dry for 10 min at room temperature and then were kept in -20°C methanol for 10 minutes. Methanol was removed and samples were allowed to air dry for 30 minutes and washed with PBS twice for 5 minutes each. The rat anti-CD31 antibody was used at a dilution of 1:20 in 2% BSA and incubated overnight at 4°C. Slides were again washed twice in PBS before incubating in the dark with a goat anti-rat AlexaFluor594-labeled secondary antibody at a dilution of 1:100 in 1% BSA for 1 h. The secondary antibody solution was then aspirated and the cells washed twice in PBS. Cells were then incubated in the dark with Hoechst’s (1 μg/ml) in PBS for 30 min, washed twice, and coverslips mounted with an anti-fade solution (Dako Corp., Carpinteria, CA) and slides were sealed with clear nail polish. Slides were examined and images captured as above.

Selected cell lines were cultured in methylcellulose-containing medium (M3434, StemCell Technologies, Vancouver, BC, Canada) with 50 ng/mL vascular endothelial (VE) growth factor (R&D Systems, Minneapolis, MN, USA), 50 ng/mL basic fibroblast growth factor (Wako, Osaka, Japan) and 10% fetal bovine serum (FBS) on 35-mm dishes. Cell densities were 1×10³ for each cell lines. Dishes were plated in triplicate and kept in humidified incubator for 8–10 days until colonies started to develop. EPC colony forming units (CFUs) were defined as cluster-like collections of cells associated with attached spindle-shaped cells. The EPC-CFUs were identified by visual inspection with an inverted microscope under 20-40× magnification. Images were acquired on a Zeiss Observer. A1 (Zeiss Microscopy, Thornwood, NY) microscope using AxioVision 4.6.3 software (Zeiss Microscopy, Thornwood, NY).

As a modified protocol of Wu et al., matrigel (356237, BD Pharmingen) was applied in thin layers to wells of 24-well plates [57]. Cells of selected cell lines were placed at indicated cellular concentrations: 1×10⁴, 1×10⁵, and 1×10⁶ cells. Cells were incubated overnight and the next morning (18 hrs) evaluated under light microscopy at 20-40X magnification. Images were acquired on a Zeiss Observer. A1 (Zeiss Microscopy, Thornwood, NY) microscope using AxioVision 4.6.3 software (Zeiss Microscopy, Thornwood, NY).

Nude mice were exposed to two doses of 4.5 Gy given 4 hours apart and, as indicated, supplemental cells were given 2 hours after the final radiation dose [58]. Normal bone marrow was collected from unirradiated 12–16 week old mice by isolating the femurs and tibia, cutting the end from the bone, and then flushing out the central cavity using 28 gauge needles with 2% FBS in Hank’s balanced salt solution. The cell solution was collected, spun down, and cell concentration calculated. Harvested bone marrow or the selected ex vivo cultured EPC cell lines were injected into the retro-orbital venous plexus of an anesthetized mouse at 3×10⁶ cells in 100 uL PBS. Survival was assessed every two days and mice were sacrificed if weight had dropped more than 10% per animal protocol guidelines.

Sorted cell lines: EPC, EPC_acLDL, and EPC_late were grown out, collected, and RNA isolated (RNeasy mini kit, Qiagen). RNA concentration and quality was measured by NanoDrop analysis. RNA was stored in RNAase free water at -80°C until analysis. RNA was hybridized to MouseRef-8 v2.0 Beadchips (25 K, Illumina, San Diego, CA). Two separate experiments were performed independently and run in duplicate at the Stanford Gene Array Core Facility. Bead level intensity values were summarized without normalization and local background correction was applied by default using Beadstudio v3.1. Microarray gene expression data were processed and analyzed using Genespring VX (Agilent, Santa Clara, CA). Clustering was performed using non-centered Pearson correlation. Data was deposited at: http://​www.​ncbi.​nlm.​gov/​projects/​geo (GSE53681).

U251 cells at a concentration of 3 × 10⁶ were mixed with 3×10⁵ EPC_acLDL cells (10% ratio). For co-injection studies, the cell mixture was implanted on the back of nude mice and tumor measurements were taken approximately three times a week. Tumor volume (mm³) was calculated as (height²*length)/2. For the systemic injection studies, Tumors (U251; 3×10⁶) were implanted on the back of mice on day 0 and EPC_acLDL cells (3 × 10⁵) were given by 100 μL retro-orbital injection on day 7 with controls receiving an injection of normal saline in an identical volume. Tumor measurements were then taken approximately three times weekly and tumor volumes calculated as above. Numeric data are presented as the mean +/- standard error.

Images from the fluorescent immunohistochemical analysis were imported into ImageJ analysis software (http://​www.​rsbweb.​nih.​gov/​ij/​, NIH, Bethesda, MD). A threshold of CD31 pixel intensity was applied to all images and the total number of CD31 positive pixels was counted within the field of view. Total CD31 pixel counts from five random images from each tumor sample were analyzed and averages were calculated over a 0.1 mm² area [59, 60]. Two tumors from two different experiments were used for analysis. Numeric data are presented as the mean +/- standard deviation.