SEARCH

SEARCH BY CITATION

Keywords:

  • Cancer stem cells;
  • Chemotherapy;
  • Bone marrow-derived cells;
  • Angiogenesis

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

Tumor-initiating cells (TICs) are a subtype of tumor cells believed to be critical for initiating tumorigenesis. We sought to determine the angiogenic properties of TICs in different tumor types including U-87MG (glioblastoma), HT29 (colon), MCF7 (breast), A549 (non-small-cell lung), and PANC1 (pancreatic) cancers. Long-term cultures grown either as monolayers (“TIC-low”) or as nonadherent tumor spheres (“TIC-high”) were generated. The TIC-high fractions exhibited increased expression of stem cell surface markers, high aldehyde dehydrogenase activity, high expression of p21, and resistance to standard chemotherapy in comparison to TIC-low fractions. Furthermore, TICs from U-87MG and HT29 but not from MCF7, A549, and PANC1 tumor types possess increased angiogenic activity. Consequently, the efficacy of vascular endothelial growth factor-A (VEGF-A) neutralizing antibody is limited only to those tumors that are dependent on VEGF-A activity. In addition, such therapy had little or reversed antiangiogenic effects on tumors that do not necessarily rely on VEGF-dependent angiogenesis. Differential angiogenic activity and antiangiogenic therapy sensitivity were also observed in TICs of the same tumor type, suggesting redundant angiogenic pathways. Collectively, our results suggest that the efficacy of antiangiogenic drugs is dependent on the angiogenic properties of TICs and, therefore, can serve as a possible biomarker to predict antiangiogenic treatment efficacy. Stem Cells2012;30:1831–1841


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

For the last several years, antiangiogenic treatment strategy has been evaluated clinically for a number of malignancies, including colorectal [1], breast [2], non-small-cell lung [3], hepatocellular [4], and renal cell [5] cancers. Although the mechanism of action of antiangiogenic drugs is well investigated, limited success has been demonstrated, raising doubts regarding their therapeutic potential. For example, combined treatment with bevacizumab, the humanized antibody that neutralizes VEGF-A [6], and FOLFOX chemotherapy did not extend disease-free survival of stages II and III colorectal cancer patients, as compared to FOLFOX monotherapy [7]. Such studies accelerated research to identify new biomarkers that can predict treatment outcome in patients treated with antiangiogenic drugs.

Tumor angiogenesis relies on growth factor signaling that activates endothelial cell migration, proliferation, and survival, resulting in vessel sprouting. In addition, stromal cells and accessory bone marrow-derived cells (BMDCs) including endothelial precursor cells (EPCs) are recruited to the tumor site, substantially contributing to the proangiogenic microenvironment of tumors. Hemangiocytes (expressing VEGFR1 and CXCR4 surface markers) can promote revascularization of tumors through the secretion of stromal-cell-derived factor 1α (SDF-1α) [8, 9]. In addition, myeloid-derived suppressor cells (MDSCs) expressing CD11b and Gr-1 surface markers were found to play a proangiogenic role in tumors, thereby increasing the resistance of tumors to antiangiogenic therapy [10]. Therefore, antiangiogenic therapies, in addition to their impact on local angiogenesis, can also target various types of BMDCs known to contribute to systemic angiogenesis.

Recent studies have identified and characterized a subtype of cancer cells that exhibit self-renewal and multilineage differentiation capabilities as well as limitless proliferation potential. Such cells are termed cancer stem cells or tumor-initiating cells (TICs) due to their ability to initiate tumor growth [11, 12]. TICs were identified in several malignancies such as acute myeloid leukemia, melanoma, glioma, breast, colon, pancreas, prostate, lung, and head and neck tumors [13]. These cells possess several cellular and molecular properties that isolate them from the rest of the tumor cells, such as: growth in spheres in serum-free medium [14]; surface immunophenotypes characteristic of primitive stem or progenitor cells [15]; resistance to conventional anticancer therapies [16, 17]; and enrichment of P-glycoproteins and aldehyde dehydrogenase (ALDH) activity [18]. It has been demonstrated that in addition to their proliferative capabilities, TICs support tumor progression by promoting tumor angiogenesis. High expression of neoangiogenic factors were observed in breast carcinoma-initiating cells characterized as CD44+/CD24−/low [19]. Furthermore, Bao et al. found that a subpopulation of human glioma cells that share characteristics of neural stem cells (CD133+ fraction) potently generate highly angiogenic tumors characterized by increased microvessel density, hemorrhage, and necrosis when implanted into the brains of immunocompromised mice [20]. Putative TICs in brain tumors were also found to reside in proximity to blood vessels in a “vascular niche.” Treatment of orthotopic transplanted gliomas in mice with antibodies to VEGF-A disrupts the vascular niche and targets the TIC population [21]. Therefore, TICs play a major role in tumor angiogenesis, or at the very least, require angiogenic competence for initiation and maintenance of tumor growth. In such circumstances, antiangiogenic drugs may serve as an adequate treatment approach for these malignancies.

In this study, we characterize the angiogenic properties of TICs derived from various cancer cell lines. We demonstrate that TICs of different tumor types possess differential angiogenic properties, as evaluated by endothelial cell migration and invasion, BMDC recruitment, microvessel sprouting, and the relative expression of angiogenesis factors. Consequently, antiangiogenic therapy using VEGF-A neutralizing antibody has diverse effects on the different tumors. These findings may have clinical implications and suggest that TICs can serve as a biomarker to predict antiangiogenic treatment efficacy.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

Cell Culture

Human umbilical vein endothelial cells (HUVEC) were kindly provided by Dr. Neta Ilan (Faculty of Medicine, Technion) and were cultured using a previously described method [22]. Human U-87MG, U-373MG, A172 glioblastoma, HT29 colon carcinoma, A549 non-small-cell lung carcinoma, MCF7 breast carcinoma, and PANC1 pancreatic adenocarcinoma cell lines were purchased from the American Type Culture Collection and were used within 6 months of resuscitation. Cell lines were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum, 1% L-glutamine, 1% sodium-pyruvate, and 1% streptomycin. All cells were passed in culture for no more than 4 months after being thawed from authentic stocks.

The Generation of TIC-High and TIC-Low Cultures

U-87MG, U-373MG, A172, HT29, A549, MCF7, and PANC1 cells were maintained either as monolayers in DMEM with 10% fetal calf serum (TIC-low fraction) or as nonadherent tumor spheres in serum-free DMEM (TIC-high fraction) supplemented with stem cell factors as detailed in Supporting Information Materials online. Of note, using this technique, we only enriched the TIC population but did not grow a pure TIC population.

Measurement of ALDH Activity In Vitro

The ALDEFLUOR kit (StemCell Technologies, Vancouver, Canada) was used to determine ALDH enzymatic activity by flow cytometry, as previously described [18]. Briefly, cells were incubated in ALDEFLUOR assay buffer containing ALDH substrate (BAAA, 1 μmol/l per 1 × 106 cells) for 40 minutes at 37°C. In each experiment, a sample of cells was supplemented under identical conditions with 50 mmol/l of diethylaminobenzaldehyde, a specific ALDH inhibitor, as a negative control. The sorting gates were established using 7AAD-negative cells to identify viability.

Evaluation of Cell Viability Following Conventional Therapy

Cell viability was evaluated quantitatively with the metabolic indicator dye AlamarBlue (Serotec Ltd., Oxford, UK). A detailed protocol is available in Supporting Information Materials online.

Generation of Conditioned-Medium from Cultured TIC-High and TIC-Low Cells

TIC-high or TIC-low cultured cells (as described above) were dissociated using trypsin, washed twice with phosphate-buffered saline (PBS), and incubated in serum-free medium at a concentration of 106 cells per milliliter. After 48 hours, cultured medium was collected and residual cells were removed by centrifugation. Supernatants were stored in −20°C, until used. Protein was precipitated from the conditioned-medium by adding five volumes of 100% methanol followed by incubation in −80°C. Samples were then centrifuged at 4,000 rpm for 20 minutes. Pellets were dried and resuspended in PBS.

Modified Boyden Chamber Assay

The invasion and migration properties of HUVECs or the invasion properties of BMDCs were evaluated by Boyden chambers, using a previously described protocol [23]. A detailed protocol is available in Supporting Information Materials online.

Matrigel Plug Assay

Matrigel (0.5 ml) that contained precipitated conditioned-medium of either TIC-high or TIC-low cultures was injected subcutaneously into each flank of a C57Bl/6 mouse (n = 4 mice per group). Plugs were removed 10 days later and subsequently prepared for either histological evaluation or flow cytometric analysis following single-cell suspension as previously described [24, 25].

Sprouting Microvessels Using Aortic Ring Assay

Aortic ring assay was performed as previously described [8]. Detailed protocols are available in Supporting Information Materials online.

Evaluation of Surface Markers and BMDCs by Flow Cytometry

In order to identify the TIC population of the different tumor types and the BMDC types that colonized the Matrigel plugs, cells from culture or from the plugs were immunostained for flow cytometry analysis as described in Supporting Information Materials online.

Quantification of the Expression Levels of Angiogenesis-Related Factors and the Analysis by DAVID Software

The angiogenic protein array kit (R&D system, Minneapolis, MN) was used on conditioned-medium from tumor cells in accordance with the manufacturer's instructions. The angiogenesis protein array results were analyzed using the online database for annotation, visualization, and integrated discovery (DAVID, NIH) bioinformatics resources version 6.7. For details please see Supporting Information Materials online.

Transfection of Short-Hairpin RNA for VEGF-A

U-87MG and A172 cells were stably transfected with pGIPZ vector, containing short-hairpin RNA (shRNA) targeting VEGF-A or shRNA control (scrambled) vectors (Thermo, Open Biosystem, Lafayette, CO), using the FuGeNE 6 reagent (Roche Applied Science, Indianapolis, IN) in accordance with the manufacturer's instructions. Transfected cells were selected with puromycin (Sigma-Aldrich, Rehovot, Israel) for 2 weeks and subsequently expanded and pooled. VEGF-A concentration was measured in conditioned-medium of cultured cells using a VEGF-A ELISA kit (R&D Systems).

Immunoblotting

The expression levels of p21, p53, and actin were evaluated on cell extracts of TIC-high or TIC-low cultured cells, as described in Supporting Information Materials online.

Animal Tumor Models and Drug Concentrations

Five million U-87MG or PANC1 cells were injected subcutaneously into the flanks of 8–10 weeks old athymic-nude mice (Harlan, Israel). Tumor size was assessed regularly with Vernier calipers using the formula, width2 × length × 0.5. When tumors reached 500 mm3, mice were treated with 50 mg/kg paclitaxel, 500 mg/kg gemcitabine, the doses of which were previously determined as being the maximum tolerated dose (MTD) [26], and/or 5 mg/kg anti-human/mouse VEGF-A neutralizing antibody (B.20 antibody kindly provided by Genentech, San Francisco, CA). Control mice were treated with the appropriate vehicles. All animal studies and experimental protocols were approved by the Animal Care and Use Committee of the Technion.

Immunostaining

Tissue processing and immunohistochemistry were performed as described previously [27]. Detailed protocols are available in Supporting Information Materials online.

Statistical Analysis

Data are presented as means ± SEM. Statistical significant differences in mean values were assessed by one-way analysis of variance (ANOVA), followed by Newman-Keuls ad hoc statistical test using GraphPad Prism four software (La Jolla, CA). When applicable, statistical significance comparing only two groups was determined by two-tailed Student's t test. Significance was set at values of *, p < .05; **, p < .01; and ***, p < .001.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

The Generation and Validation of TIC-Enriched Cultured Cells

Previous studies have demonstrated that in addition to primary tumors, established cancer cell lines may contain a subpopulation of cells possessing TIC properties [28, 29]. We sought to enrich these cells from a variety of cancer cell lines. U-87MG, HT29, A549, MCF7, and PANC1 cancer cell lines were routinely cultured as monolayers under normal, “TIC-low” growth conditions. To enrich for a TIC population, the cells were maintained in long-term culture under “TIC-high” conditions. We chose to further investigate TIC-high cultured cells of the tumor types which could form spheres (Fig. 1A) and which demonstrated at least a threefold increase in the expression of stem cell surface markers in the TIC-high fractions as compared to the TIC-low fractions (Fig. 1B). We should note that other cell lines that could not grow in spheres and did not exhibit enrichment of TIC surface markers were eliminated from the study (data not shown). Next, to further evaluate TIC properties of the specific cell lines, we focused on ALDH enzymatic activity and on the expression of p21 [30, 31]. We found that ALDH activity was at least twofold higher in the TIC-high fractions of U-87MG, MCF7, and PANC1 cells as compared to the TIC-low counterparts. However, A549 and HT29 cells exhibited substantially high levels of ALDH activity in the TIC-low fractions as compared to the TIC-high fractions (Fig. 1C). Expression of p21 was higher in TIC-high fractions of all the cell lines and was associated with p53 expression in U-87MG and HT29 cells but not in PANC1 cells. No detectable expression of p53 was observed in A549 and MCF7 in either cell fraction (Fig. 1D). Since TICs are known to resist conventional therapy [32, 33], we also asked whether cells of TIC-high fractions are more resistant to paclitaxel and gemcitabine chemotherapies than cells of TIC-low fractions. We assessed cell viability in the presence of paclitaxel and gemcitabine chemotherapy by AlamarBlue. The results in Figure 1E demonstrate that cells from the TIC-high fractions of U-87MG, HT29, and PANC1 were more resistant to chemotherapy than cells from the TIC-low fractions, although the results were not always statistically significant. No changes in resistance to chemotherapy were observed in A549 and MCF7 TIC-high cells. Overall, our results demonstrate that TIC-high cells share a spectrum of TIC properties, albeit with some differences between the various tumor types.

thumbnail image

Figure 1. Generation and validation of TIC-high and TIC-low cultured cells. (A): Representative images of formed tumor spheres in suspension from a series of human established cell lines: U-87MG glioblastoma, HT29 colon carcinoma, A549 non-small-cell lung adenocarcinoma, MCF7 breast carcinoma, and PANC1 pancreatic adenocarcinoma, captured by ×40 objective-field. (B): Representative flow cytometry plots of phenotypic evaluation of TIC-low and TIC-high fractions from the various tumor cell lines. (C): Representative flow cytometry plots for the analysis of ALDH enzymatic activity in TIC-low and TIC-high fractions in the presence of ALDH substrate (BAAA) (left and middle plots, respectively) and in the TIC-high fraction in the presence of ALDH substrate and the ALDH enzyme inhibitor DEAB (right plots). The percentages shown in the polygons represent the ALDH-positive population after gating on viable cells using 7AAD staining. (D): Western blot analysis of p21, p53, and actin of TIC-low and TIC-high fractions of the indicated cell lines. (E): TIC-high and TIC-low cells of the indicated cell lines were exposed to 100 nM PTX or 10 nM GEM for 4 days in culture, and cell viability was evaluated using the AlamrBlue assay. Results were plotted as fold increase from untreated cells.*, 0.05>p > .01; **, 0.01>p > .001; ***, p < .001. Abbreviations: ALDH, aldehyde dehydrogenase; DEAB, diethylaminobenzaldehyde; GEM, gemcitabine; PTX, paclitaxel; TIC, tumor-initiating cell.

Download figure to PowerPoint

TICs of Various Tumor Types Exhibit Differential Angiogenic Properties

A number of studies have demonstrated the ability of TICs to induce angiogenesis and vasculogenesis [20, 34]. We therefore assessed the angiogenic ability of TIC-high and TIC-low conditioned-medium to induce HUVEC migration and invasion using the modified Boyden chamber assay. The results in Figure 2A and Supporting Information Figure S1 show that TIC-high conditioned-medium of U-87MG, HT29, and PANC1 induced HUVEC migration as compared to TIC-low conditioned-medium. In contrast, HUVEC migration was induced in the presence of A549 TIC-low conditioned- medium, whereas no difference was observed between TIC-high and TIC-low conditioned-medium of MCF7 cells. Furthermore, we found that TIC-high conditioned-medium of U-87MG and HT29 cells induced HUVEC invasion as compared to TIC-low conditioned-medium. In contrast, HUVEC invasion was induced in the presence of MCF7 TIC-low conditioned-medium as compared to TIC-high medium. No significant differences in HUVEC invasion were observed between TIC-high and TIC-low conditioned-medium of A549 and PANC1 cells (Fig. 2B and Supporting Information Fig. S1). Overall, these results demonstrate that only U-87MG and HT29 TIC-high conditioned-medium induced both endothelial cell migration and invasion.

thumbnail image

Figure 2. Endothelial and bone marrow-derived cell colonization in the presence of C.M from TIC-high and TIC-low cell fractions. The effect of C.M obtained from TIC-low and TIC-high fractions on the migration (A) and invasion (B) of endothelial cells as assessed by the modified Boyden chamber assay and were captured per ×200 objective-field. Quantification of cells invading the filter is presented in Supporting Information Figure S1. (C, D): Matrigel plugs that contained TIC-low or TIC-high C.M of the indicated cell lines were removed from C57Bl/6 mice after 10 days (n = 4 mice per group), and (C) sections were prepared for hematoxylin and eosin staining (upper micrographs) or for CD31 immunostaining (in red, lower micrographs) (scale bar = 200 μm). (D): Vessel structures were counted in each field and presented as the number of microvessels per field. Matrigel plugs were digested, and cells were prepared as single-cell suspensions. Cells were analyzed for hemangiocytes and MDSCs by flow cytometry. *, 0.05>p > .01; **, 0.01>p > .001; ***, p < .001. Abbreviations: C.M, conditioned-medium; MDSC, myeloid-derived suppressor cell; TIC, tumor-initiating cell.

Download figure to PowerPoint

TICs from Different Tumor Types Exhibit Differential Host Cell Colonization of Matrigel Plugs

Folkins et al. have recently demonstrated that TIC-high fraction of C6 glioma cells implanted in mice resulted in more EPCs and other BMDCs in the tumors, as compared to tumors from mice implanted with cells from a TIC-low fraction [34], suggesting that TIC-high fractions can induce vasculogenesis. To test this, we used a Matrigel plug assay using conditioned-medium from either TIC-high or TIC-low fractions. Ten days after Matrigel implantation, plugs were removed and evaluated for microvessel density and hemangiocyte and MDSC colonization [9, 10]. The results in Figure 2C show that microvessel density was substantially higher in plugs containing TIC-high conditioned-medium of both U-87MG and HT29 cells. However, it was significantly lower in the TIC-high MCF7 and PANC1 plugs and unchanged in the A549 plugs. In a parallel experiment, single-cell suspensions were prepared from the plugs, and hemangiocytes and MDSCs were evaluated by flow cytometry. The results in Figure 2D show that hemangiocytes and MDSCs are differentially colonized in the plugs containing conditioned-medium of either TIC-high or TIC-low fractions of the different cell lines. Notably, hemangiocytes, but not MDSCs, were significantly higher in Matrigel plugs containing conditioned-medium from the TIC-high fraction of U-87MG cells, an effect which was not observed in plugs containing conditioned-medium of neither TIC fraction of PANC1 cells. Overall, these results demonstrate that TICs of various tumor types have differential angiogenesis and vasculogenesis properties.

Anti-VEGF-A Antibody Inhibits BMDC Invasion and Sprouting Angiogenesis in U-87MG and HT29 but Not PANC1 TICs

The substantial differences in angiogenic and vasculogenic properties of TICs enriched from different tumor types prompted us to evaluate the impact of neutralizing VEGF-A on BMDC invasion and sprouting angiogenesis. We chose to work with the U-87MG and HT29 cell lines, since they exhibited potent angiogenic effects in the TIC-high fraction and with PANC1 cell line as a less potent representative. To this end, conditioned-medium of TIC-high and TIC-low fractions of U-87MG, HT29, and PANC1 cells was collected and subsequently tested for BMDC invasion capability in the presence or absence of bevacizumab, the humanized antibody that neutralizes VEGF-A [6]. Bevacizumab is known to block human VEGF-A, but not murine VEGF-A [35], therefore only paracrine VEGF-A, which is present in the conditioned-medium, is blocked. The results in Figure 3A show that in the presence of U-87MG and HT29 TIC-low conditioned-media, bevacizumab weakly but significantly inhibited BMDC invasion. This inhibition was significantly pronounced in the presence of TIC-high conditioned-medium, as evidenced by the dramatic decrease (approximately fivefold) in the number of invading BMDCs. In contrast, bevacizumab had no effect on BMDC invasion in the presence of TIC-low or TIC-high conditioned-medium of PANC1 cells.

thumbnail image

Figure 3. Anti-vascular endothelial growth factor-A (VEGF-A) antibody inhibits bone marrow-derived cell (BMDC) invasion and microvessel sprouting in the presence of U-87MG and HT29 but not PANC1 TIC-high C.M. C.M of cells from TIC-high and TIC-low fractions of U-87MG, HT29, or PANC1 cells in the presence or absence of 5 μg/ml BEV was evaluated for their impact on: (A) BMDC invasion using the modified Boyden chamber assay, and (B) microvessel sprouting using aortic ring assay (scale bar = 50 μm). *, 0.05>p > .01. Abbreviations: BEV, bevacizumab; BM, bone marrow; C.M, conditioned-medium; TIC, tumor-initiating cell.

Download figure to PowerPoint

We next evaluated the impact of bevacizumab on the inhibition of sprouting angiogenesis using the aortic ring assay. Again, murine endothelial cells from aortic rings are not directly affected by bevacizumab (Supporting Information Fig. S2), hence we tested only the neutralization of human VEGF-A from conditioned-medium on sprouting microvessels from aortic rings. Aortic rings were placed in Matrigel that contained conditioned-medium of TIC-high or TIC-low fractions of U-87MG, HT29, or PANC1 cells, in the presence or absence of 5 μg/ml bevacizumab. In the U-87MG and HT29 experiments, microvessel sprouting induced by TIC-high conditioned-medium was inhibited by the addition of bevacizumab. In contrast, bevacizumab induced microvessel sprouting in the presence of TIC-high and TIC-low PANC1 conditioned-medium (Fig. 3B). Overall, these results suggest that the angiogenic properties of TICs in U-87MG and HT29 are substantially dependent on VEGF-A pathways, consistent with the neutralizing effect of VEGF-A antibodies on angiogenesis and BMDC invasion. However, the angiogenic properties of PANC1, if existent, are presumably not VEGF-A-dependent, explaining why neutralizing VEGF-A failed to inhibit sprouting angiogenesis and BMDC invasion.

U-87MG but Not PANC1 TICs Reside in Proximity to Blood Vessels Following Chemotherapy

We next sought to determine the impact of antiangiogenic drugs on the proangiogenic activity of TICs of U-87MG and PANC1 chemotherapy-treated tumors in mice. We analyzed the distance between blood vessels and TICs in the tumor microenvironment. To do so, athymic-nude mice were implanted with either U-87MG or PANC1 cells (5 × 106). When tumors reached a size of 500 mm3, the mice were treated with paclitaxel or gemcitabine in the presence or absence of B20 antibody, which neutralizes both human and murine VEGF-A. Four days later, tumors were measured and then harvested. Tumor sections were evaluated for microvessel density, TICs, and the colonization of hematopoietic cells by staining for CD31+, CD133+, and CD45+, respectively. The results in Supporting Information Figure S3A show that the B20 antibody enhanced treatment efficacies of paclitaxel and to a lesser extent of gemcitabine in U-87MG but not PANC1 tumors. Moreover, we observed that the microvessel density in U-87MG tumors was significantly reduced in response to the combination therapies (Fig. 4A, Supporting Information Fig. S3B); however, the microvessel density in PANC1 tumors was not significantly different between the treatment groups. We then assessed the distance between TICs (CD133+) and blood vessels in sections of U-87MG and PANC1 tumors from all treatment groups. Figure 4B and 4C shows that TICs of U-87MG tumors were found in proximity to blood vessels with both chemotherapy treatments. This effect was significantly reversed upon combined treatment with chemotherapy and B20 antibody. In contrast, no significant changes in the distance between TICs and blood vessels were observed in PANC1 tumors in any of the treatment groups. Notably, more TICs in both U-87MG and PANC1 tumors were observed following treatment with either paclitaxel or gemcitabine. Taken together, TICs from U-87MG, but not PANC1, contribute to tumor angiogenesis and are therefore sensitive to VEGF-A targeted drugs.

thumbnail image

Figure 4. Antiangiogenic treatment decreases the proximity of U-87MG, but not PANC1, tumor-initiating cells (TICs) to blood vessels in chemotherapy-treated mice. U-87MG (5 × 106) or PANC1 cells were implanted subcutaneously into the flanks of 8–10 weeks old Athymic-nude mice (n = 3–5 mice per group). When tumors reached 500 mm3, mice were treated with PTX, GEM, anti-human/mouse vascular endothelial growth factor-A (VEGF-A) neutralizing antibody (B20), or the combination of either PTX or GEM chemotherapy and the antiangiogenic drug. Four days later, tumors were removed and evaluated for tumor growth volume (Supporting Information Fig. S3A). (A): U-87MG and PANC1 tumor sections were stained for CD31 as a marker for endothelial cells (scale bar = 100 μm); summary of microvessel density quantification is shown in Supporting Information Figure S3B. (B): Tumor sections were stained for CD31+ cells (red), TICs CD133+ cells (violet), hematopoietic CD45+ cells (green), and 4′-6-diamino-2-phenylindole (blue). White arrows represent TICs, and orange arrows represent vessels. Images were captured using Leica TCS SP5 confocal imaging system (scale bar = 50 μm). High-magnification images are shown in small squares for each micrograph (Zoom Factor X3). (C): A summary of the quantification of the distance between TICs (CD133+) and blood vessels (CD31+) in U-87MG and PANC1 tumors in all treatment groups. 0.05>p > .01; *, 0.05>p > .01; ***, p < .001, from control group, unless indicated otherwise. Abbreviations: GEM, gemcitabine; PTX, paclitaxel.

Download figure to PowerPoint

TICs of Different Cell Lines Secrete Differential Angiogenic-Related Factors

To test whether additional angiogenic pathways play a role in the insensitive of PANC1 to VEGF-A targeted drugs, we compared several angiogenic-related factors in conditioned-medium of TIC-high and TIC-low fractions of U-87MG and PANC1 cells, using angiogenic protein arrays. The results in Figure 5 show a Venn diagram of the changes in angiogenic protein expression between U-87MG and PANC1 cells following DAVID software analysis for angiogenesis category. In U-87MG cells, five angiogenic-related factors were upregulated, and nine angiogenic-related factors were downregulated in the TIC-high fraction as compared to the TIC-low fraction. In contrast, this pattern was reversed in PANC1 cells, in which 11 angiogenic factors were upregulated and only two angiogenic factors were downregulated. These results suggest the existence of compensatory bypass angiogenic pathways in PANC1 cells, which may counteract the antiangiogenic activity of VEGF-A neutralizing antibody. We further evaluated levels of VEGF-A in conditioned-medium of TIC-high and TIC-low fractions of all cell lines tested. Significantly, higher levels of VEGF-A was found in conditioned-medium of the TIC-high fractions of U-87MG and HT29 cells, as compared to the TIC-low fractions. In contrast, conditioned-medium of A549, MCF7, and PANC1 TIC-high fractions contained significantly and remarkably lower VEGF-A levels than that of their TIC-low fractions (Supporting Information Fig. S4).

thumbnail image

Figure 5. Relative expression levels of angiogenesis-related factors of tumor-initiating cell (TIC)-high versus TIC-low conditioned-media (C.M) from U-87MG and PANC1 cells. The results of the angiogenesis protein array following DAVID software analysis were illustrated in a stylized Venn diagram depicting the pattern of more than 10% change in the expression levels of angiogenesis-related factors between TIC-high and TIC-low C.M comparing U-87MG and PANC1 cells. Factors detected by specific ELISA are indicated by * (for absolute concentration of VEGF-A in TIC-high and TIC-low of all cell lines tested, Supporting Information Fig. S4). Abbreviations: PD-ECGF, platelet derived endothelial cell growth factor; FGF, fibroblast growth factor; PDGF, platelet derived growth factor; VEGF, vascular endothelial growth factor.

Download figure to PowerPoint

Differential Antiangiogenic Activity in TICs from Various Glioblastoma Tumors

We next assessed whether differences in angiogenic properties of TICs exist in the same tumor type. To this end, U-373MG and A172 additional glioblastoma cell lines were enriched for TICs in a similar way to that performed for U-87MG. In TIC-high culture conditions, both U-373MG and A172 cells formed spheres, enriched for CD133 expression, and increased ALDH enzymatic activity when compared with TIC-low culture conditions (Supporting Information Fig. S5A–S5C). In addition, U-373MG and A172 TIC-high cells were resistant to chemotherapy. Of note, the U-373MG TIC-low fraction also exhibited chemoresistant properties (Supporting Information Fig. S5D). Furthermore, the angiogenic properties of TICs of both A172 and U-373MG were found to be similar to the angiogenic properties of TICs of U-87MG by means of endothelial cell migration (Supporting Information Fig. S6). Taken together, TIC-high fraction from the various glioblastoma cell lines possess TIC properties and are found to have similar angiogenic activity.

Next, to evaluate the antiangiogenic activity of VEGF neutralizing antibodies in the various glioblastoma cell lines, conditioned-medium of TIC-high and TIC-low fractions from U-373MG and A172 cells was collected and subsequently tested for BMDC invasion in the presence or absence of bevacizumab. The results in Figure 6A show that bevacizumab inhibited BMDC invasion in the presence of conditioned-medium obtained from both TIC-low and TIC-high fractions of U-373MG cells but not A172 cells. Furthermore, the results in Figure 6B show that microvessel sprouting assessed by the aortic ring assay was induced in aortic rings in the presence of TIC-high conditioned-medium of U-373MG cells and was drastically inhibited by the addition of bevacizumab. In contrast, bevacizumab did not inhibit microvessel sprouting in the presence of either TIC-high or TIC-low conditioned-medium of A172 cells. Collectively, our results indicate that TICs derived from glioblastoma tumors exhibit differential angiogenic properties and, therefore, react differently to antiangiogenic therapy.

thumbnail image

Figure 6. Inhibition of VEGF-A decreases bone marrow-derived cell (BMDC) invasion, microvessel sprouting, and TIC properties in TIC-high of U-373MG and U-87MG but not in TIC-high of A172. C.M of cells from TIC-high and TIC-low fractions of either U-373MG or A172 cells (as indicated in Supporting Information Figs. S5, S6) in the presence or absence of 5 μg/ml BEV were evaluated for their impact on: (A) BMDC invasion using the modified Boyden chamber assay and (B) microvessel sprouting using aortic ring assay (scale bar = 50 μm). (C): Representative images of formed tumor spheres in suspension from U-87MG and A172 glioblastoma following shVEGF-A RNA manipulation (as indicated in Supporting Information Fig. S7), captured by ×40 objective-field. (D): Representative flow cytometry plots of CD133 expression in cells from TIC-low and TIC-high fractions of U-87MG and A172. *, 0.05>p > .01. Abbreviations: BEV, bevacizumab; BM, bone marrow; C.M, conditioned-medium; TIC, tumor-initiating cell; VEGF, vascular endothelial growth factor.

Download figure to PowerPoint

Inhibition of Endogenous VEGF-A Attenuates TIC Properties of U-87MG but Not A172

A recent study suggested that VEGF-A signaling is critical for controlling stemness of TICs. The deletion of neuropilin-1, a VEGF coreceptor, in skin tumor cells blocked the ability of VEGF-A to maintain TIC stemness and tumor cell renewal. Thus, in addition to its angiogenic effects in tumors, VEGF-A may also play a significant role in maintaining TICs [36]. We therefore sought to determine whether VEGF-A depletion affects the properties of TICs obtained from U-87MG and A172 tumors. To this end, U-87MG and A172 tumor cells were stably transfected with shRNA for VEGF-A or a scrambled shRNA. VEGF-A in conditioned-medium of TIC-high and TIC-low fractions from both cell lines was depleted more than 50% in the shVEGF-A-transfected cells (Supporting Information Fig. S7). As expected, VEGF-A depletion abrogated the ability of TIC-high U-87MG cells to grow in spheres. In contrast, VEGF-A-depleted TIC-high A172 cells were still able to form spheres (Fig. 6C). Consistently, whereas the CD133 surface marker was not enriched in VEGF-A-depleted TIC-high U-87MG cells, it was enriched in VEGF-A-depleted TIC-high A172 cells, as compared to the TIC-low counterparts (Fig. 6D). Overall, our results suggest that compensatory mechanisms for VEGF-A account for the TIC properties in A172 but not U-87MG cells, explaining the differential effect of antiangiogenic therapy in the two different tumors.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

In this study, we identified, characterized, and enriched for TICs from a variety of tumor cell lines, based on their morphological, phenotypic, and functional properties. We showed that glioblastoma, colon carcinoma, non-small-cell lung adenocarcinoma, breast carcinoma, and pancreatic adenocarcinoma grown in long-term, serum-free culture can form spheres. We defined the cultured spheres as TIC-high cells, whereas the same cells cultured in serum-containing medium were defined as TIC-low cells. It has already been demonstrated that U-87MG, A549, MCF7, and PANC1 cell lines cultured in TIC-high conditions are enriched in TICs based on the limiting dilution transplantation assay, which is the most definitive method to functionally identify TICs [37–40]. In this study, we used additional stem cell assays to demonstrate the successful enrichment of TICs from the various cell lines. We showed that cells from TIC-high fractions express higher levels of primitive, stem cell surface markers. We also found that TIC-high cells of U-87MG, MCF7, and PANC1 but not HT29 and A549 cell lines exhibit increased levels of ALDH1 activity, in line with several studies [31, 41]. However, the use of TIC-high cells does not mean a pure population of TICs but only an enriched population. This could explain why in some TIC assays the different cells did not always adhere to TIC properties. Furthermore, the expression levels of p21 were evaluated, since it has recently been demonstrated to maintain self-renewal properties of leukemic stem cells, and as such it was found to be critical for preventing excess DNA-damage accumulation [30]. The results of our study show that p21 was highly expressed in TIC-high cells in all the cell lines tested as compared to TIC-low cells. We also found that the expression of p21 was associated with high expression of p53 in both U-87MG and HT29, as opposed to PANC1 cells in which TIC-low cells highly expressed p53. The slow proliferation of the TIC population may contribute to DNA damage repair and therefore may explain the resistance of TICs to chemotherapy and radiation [30]. Indeed, most of the TIC-high cells exhibited some degree of resistance to chemotherapy, further reinforcing that such cells share stem cell-like properties, albeit with some degree of variability.

The efficacy of combination therapy in which antiangiogenic drugs are administered together with chemotherapy has been demonstrated in many studies. Several models have been proposed to explain why antiangiogenic drugs may sometimes act as chemosensitizing agents [42]. In this study, we offer an additional model to explain how antiangiogenic drugs can synergize with chemotherapy. The fact that TICs are resistant to chemotherapy may explain their presence in large numbers in chemotherapy-treated tumors. We showed that TICs of glioblastoma and colon adenocarcinoma can promote sprouting angiogenesis and BMDC colonization known to contribute to tumor angiogenesis and growth. In such circumstances, the addition of an antiangiogenic drug to chemotherapy may improve clinical outcome, by blocking TIC-induced angiogenesis. We demonstrated that this type of combination therapy was effective in U-87MG tumors but it failed in PANC1 tumors. One possible reason for this failure is that the angiogenic activity in PANC1 is VEGF-A-independent. Indeed, VEGF-A was present at low levels in the conditioned-medium of TIC-high PANC1 cells. Wang et al. explained the failure of antiangiogenic therapy in specific tumors, by demonstrating that glioblastoma stem-like cells can give rise to tumor endothelium due to CD133+CD144+ progenitor cells that were found in the tumor mass. Such cells are capable of forming endothelial-cell-like structures that are not VEGF-A dependent. Therefore, using antiangiogenic drugs or silencing angiogenic factors does not inhibit tumor growth [43]. Collectively, our results suggest that the activity of an antiangiogenic drug in combination with chemotherapy is dependent on the type of the tumor being treated and the angiogenic properties of its TICs.

An interesting question that arose from this study pertains to the fact that bevacizumab failed to inhibit sprouting angiogenesis in TICs of pancreatic cancer. We demonstrated that an abundant number of proangiogenic factors, except for VEGF-A, were upregulated in the TIC-high conditioned-medium of PANC1 cells when compared with TIC-high conditioned-medium of U-87MG cells, suggesting that angiogenesis in PANC1 cells is probably not mediated by VEGF-A but by different compensatory bypass pathways. Furthermore, the blockade of VEGF-A or VEGF pathways has been shown to increase a number of other proangiogenic factors, for example, granulocyte colony-stimulating factor (G-CSF), basic fibroblast growth factor, and SDF-1 [44, 45]. Therefore, in order to effectively block angiogenesis in tumors, which are not VEGF-A dependent, a combination of drugs blocking more than one angiogenic pathway is most likely necessary.

Our results also shed light on the possible use of TICs as biomarkers for antiangiogenic treatment outcome. We demonstrated that although TICs from different glioblastoma cell lines possess similar angiogenic properties, they respond differently to neutralize VEGF-A by bevacizumab. Specifically, bevacizumab treatment reduced angiogenic properties of the TIC-high fractions of U-87MG and U-373MG but not A172. Beck et al. have recently demonstrated that VEGF-A and neuropilin (the coreceptor of VEGF) may regulate stemness properties of skin tumors [36]. Indeed, the depleting VEGF-A in TICs of U-87MG but not of A172 resulted in reduced stemness properties. Taken together, our results suggest that TICs of glioblastoma tumors react differently to antiangiogenic drug therapies that not only target angiogenesis but also maintain TIC properties within the treated tumor.

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

In summary, our study suggests a new possible approach to predict the outcome of antiangiogenic therapy in certain types of cancers. The evaluation of the proangiogenic activity of TICs and the ability of neutralizing VEGF-A using a specific antibody to inhibit sprouting angiogenesis in a specific tumor may serve as a new biomarker for the efficacy of antiangiogenic drugs not only with respect to angiogenic inhibition mechanisms of such drug but also according to their ability to attenuate TIC properties within a treated tumor. The evaluation of the proangiogenic properties of TICs obtained from cancer patients is therefore worthy.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

B20 antibody was kindly provided by Genentech Inc. L.B. was supported by Israel Student Education Foundation, Fine, and Jacobs studentships. This work was supported by research grants from the Israeli Ministry of Health, Israel Science Foundation, European Commission under FP7 program (Marie Curie), and a sponsored research agreement with Hoffmann La Roche.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. Disclosure Of Potential Conflicts Of Interest
  10. REFERENCES
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
SC_12-0327_sm_SupplFig1.tif203KFigure S1: The quantification of endothelial cell migration and invasion in the presence of conditioned-medium of TIC-high and TIC-low fractions. Endothelial cells which invaded the bottom filter of the modified Boyden chamber assay presented in Figure 2A-B, were counted, and the number of cells per field is presented for migration (upper graph) and invasion (lower graph). (n>15 fields/group).*, 0.05>p>0.01; **, 0.01>p>0.001; ***, p<0.001.
SC_12-0327_sm_SupplFig2.tif343KFigure S2: Bevacizumab does not inhibit microvessel sprouting of murine aortic ring assay. M199 BT-203 medium known to induce microvessel sprouting of aortic ring was used in the presence or absence of 5 μg/ml bevacizumab. Images of microvessel sprouting from aortic rings were captured using light microscopy (Scale bar=50μm).
SC_12-0327_sm_SupplFig3.tif274KFigure S3: The effects of chemotherapy in combination with antiangiogenic drug on volume and microvessel density of either U-87MG or PANC1 tumors. Five million of either U-87MG or PANC1 cells were subcutaneously injected into 8-10 week old nude mice (n=3-5 mice/group) when tumors reached a size of 500mm3 treatment with paclitaxel, gemcitabine, B20 antiangiogenic drug, or the combination of paclitaxel+B20 and gemcitabine+B20 was initiated. Four days later, (A) tumors were measured for the evaluation of the percentage of changes in tumor volume (after normalization to control untreated group) for either U-87MG or PANC1 tumors. Subsequently, tumors were removed and sections were immunostained for CD31 (in red) to evaluate microvessel density (micrographs are presented in Figure 4A), and (B) quantification of the number of vessel structure per field are presented for both U-87MG and PANC1 tumors *, 0.05>p>0.01; **, 0.01>p>0.001; ***, p<0.001.
SC_12-0327_sm_SupplFig4.tif199KFigure S4: VEGF-A expression in TIC-high and TIC-low conditioned medium. The level of VEGF-A as part of the analysis of different angiogenic-related factors (Figure 5) in conditioned-medium of cells from TIC-high and TIC-low fractions of either U-87MG, HT29, A549, MCF7 or PANC1 cells was evaluated using specific human VEGF-A ELISA. **, 0.01>p>0.001; ***, p<0.001.
SC_12-0327_sm_SupplFig5.tif394KFigure S5: Generation and validation of TICs from U-373MG and A172 glioblastoma cultured cells. (A) Representative images of formed tumorspheres in suspension from a series of human glioblastoma cell lines: U-373MG and A172, captured by x200 objective-field. (B) Representative flow cytometry plots of phenotypic evaluation of TIC-low and TIC-high fractions from the glioblastoma cell lines. (C) Representative flow cytometry plots for the analysis of ALDH enzymatic activity in TIClow and TIC-high fractions in the presence of ALDH substrate (BAAA) (left and middle plots, respectively) and in the TIC-high fraction in the presence of ALDH substrate and the ALDH enzyme inhibitor diethylaminobenzaldehyde (DEAB) (right plots). The percentages shown in the polygons represent the ALDH-positive population after gating on viable cells using 7AAD staining. (D) Cells from TIC-high and TIC-low fractions of the indicated cell lines were exposed to 100nM paclitaxel (PTX), 10nM gemcitabine (GEM), and cell viability was evaluated using the AlamarBlue assay. Results were plotted as fold increase from untreated cells. The generation and validation of TICs of glioblastoma cultured cells were obtained for experiments conducted in Figure 6. *, 0.05>p>0.01; **, 0.01>p>0.001.
SC_12-0327_sm_SupplFig6.tif522KFigure S6: Endothelial cell migration in the presence of conditioned medium from glioblastoma TIC-high and TIC-low fractions. (A) The migratory effect of conditionedmedium obtained from TIC-low and TIC-high fractions of glioblastoma cells was assessed by the modified Boyden chamber assay. Images were captured per x200 objective-field. (B) Endothelial cell migration was quantified by counting the number of invading endothelial cells to the bottom filter of the Boyden chamber. The number of cells per field is presented (n>15 fields/group). The evaluation of the angiogenic properties of glioblastoma cultured cells were obtained for experiments conducted in Figure 6. **, 0.05>p>0.01; ***, p<0.001.
SC_12-0327_sm_SupplFig7.tif40KFigure S7: Inhibition of VEGF-A in TIC-high fraction of glioblastoma cells. Conditioned-medium of cells from TIC-high fractions of either U-87MG or A172 cells transfected with shRNA for VEGF-A was evaluated for VEGF-A expression by specific ELISA, for experiments performed in Figure 6C-D.
SC_12-0327_sm_SupplMaterials.pdf103KSupplementary data

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.