Chemoresistance acquisition induces a global shift of expression of aniogenesis-associated genes and increased pro-angogenic activity in neuroblastoma cells

Background Chemoresistance acquisition may influence cancer cell biology. Here, bioinformatics analysis of gene expression data was used to identify chemoresistance-associated changes in neuroblastoma biology. Results Bioinformatics analysis of gene expression data revealed that expression of angiogenesis-associated genes significantly differs between chemosensitive and chemoresistant neuroblastoma cells. A subsequent systematic analysis of a panel of 14 chemosensitive and chemoresistant neuroblastoma cell lines in vitro and in animal experiments indicated a consistent shift to a more pro-angiogenic phenotype in chemoresistant neuroblastoma cells. The molecular mechanims underlying increased pro-angiogenic activity of neuroblastoma cells are individual and differ between the investigated chemoresistant cell lines. Treatment of animals carrying doxorubicin-resistant neuroblastoma xenografts with doxorubicin, a cytotoxic drug known to exert anti-angiogenic activity, resulted in decreased tumour vessel formation and growth indicating chemoresistance-associated enhanced pro-angiogenic activity to be relevant for tumour progression and to represent a potential therapeutic target. Conclusion A bioinformatics approach allowed to identify a relevant chemoresistance-associated shift in neuroblastoma cell biology. The chemoresistance-associated enhanced pro-angiogenic activity observed in neuroblastoma cells is relevant for tumour progression and represents a potential therapeutic target.


Background
Neuroblastoma is the most frequent extracranial solid tumour of childhood. About half of all neuroblastoma patients are diagnosed with high-risk disease with overall survival rates below 40% despite intensive multimodal treatment [1]. Therapy failure is basically caused by acquired chemoresistance. Primary tumours usually respond to initial chemotherapy. However, a significant fraction of tumours reappear as chemoresistant recidives [2].
Here, differences in angiogenesis signalling were identified by bioinformatics pathway analysis of gene expression data from chemosensitive and chemoresistant neuroblastoma cells. Subsequently, cell culture and animal experiments using 14 human neuroblastoma cell lines indicated a consistently higher pro-angiogenic activity of chemoresistant neuroblastoma cells than of chemosensitive cells. The molecular mechanisms underlying the chemoresistance-associated increased pro-angiogenic potential were individual and differed between individual cell lines. Doxorubicin treatment of doxorubicin-resistant neuroblastoma xenografts resulted in impairment of tumour angiogenesis and growth suggesting the chemoresistance-associated pro-angiogenic phenotype to contribute to tumour progression.
For every microarray experiment, the expression pattern of 50 randomly chosen genes was verified by quantitative real-time PCR resulting in confirmation of expression of >80% of investigated genes (data not shown).

Signal transduction pathway bioinformatics
Statistical analysis to identify significant expression changes was focusing on a pathway analysis using the PANTHER database [13]http://www.pantherdb.org, which identifies global patterns in expression. For each expert-curated pathway in the database, potential differential expression was determined by a binomial test [14], using the PANTHER human gene reference list matching our microarrays (Human AB1700 genes) and lists of differentially expressed genes that passed a false discovery rate threshold [15] of 0.05 based on a t-test.
A total of 25,909 genes were annotated in the dataset, 3,125 of them included pathway information, and 223 of these (corresponding to 280 AB1700 ProbeIDs or 537 HGU133 Plus 2.0 ProbeIDs) were annotated as related to angiogenesis. For this list of angiogenesis-associated ProbeIDs and angiogenesis-associated genes signal intensities of UKF-NB-3, UKF-NB-3 r VCR 10 , UKF-NB-3 r CDDP 1000 , or UKF-NB-3 r DOX 20 cells were visualised as heatmaps using R http://www.r-project.org.

Viability assay
HUVEC viability was investigated using the CellTiter-Glo ® Luminescent Cell Viability Assay (Promega, Mannheim, Germany) following the manufacturer's instructions.

Tube formation assay
Endothelial cellular tube formation was investigated using HUVECs seeded on extracellular matrix (Matrigel, BD Biosciences, Heidelberg, Germany) as described before [21].

Western blot
Cells were lysed in Triton X-sample buffer and separated by SDS-PAGE. Proteins were detected using specific anti-bodies against β-actin (Sigma, Taufkirchen, Germany), ERK 1/2, the phosphorylated forms of ERK 1/2 (each from New England Biolabs, Frankfurt am Main, Germany), Akt, or the phosphorylated forms of Akt (all Millipore (Upstate), Schwalbach, Germany) and were visualised by enhanced chemiluminescence using a commercially available kit (Amersham, Freiburg, Germany).

Animal experiments
Experiments using the chick chorioallantoic membrane (CAM) were performed using described methods [23]. 10 6 cells were placed onto the CAM at day 8. Vessel formation was examined at day 12.
Mouse experiments were performed using female NMRI:nu/nu mice as described before [9]. 10 7 cells were injected subcutaneously together with Matrigel in a total volume of 100 μl. For doxorubicin treatment, the day when xenograft tumours became palpable was defined to be day 1. Tumour sections were stained for apoptotic cells by TUNEL staining and for cell proliferation by ki67 staining using established methods [24,25]. All animal experiments were performed in accordance with all relevant declarations on the use of laboratory animals and with the German Animal Protection Law.

Expression of angiogenesis-associated genes
A pathway analysis was performed in order to detect the most strongly influenced signalling pathways between UKF-NB-3 and its chemoresistant sub-lines UKF-NB-3 r VCR 10 and UKF-NB-3 r CDDP 1000 . Of the 153 pathways mapped at PANTHER, angiogenesis was found to be the fourth most significantly affected signalling pathway (p = Subsequently to these analyses, we compared angiogenesis signalling between UKF-NB-3 and UKF-NB-3 r DOX 20 cells. Since Applied Biosystems had stopped manufacturing of AB1700 arrays, HGU133 Plus 2.0 arrays (Affymetrix) were used. Results were similar to those obtained from the comparison of UKF-NB-3 with UKF-NB-3 r VCR 10 and UKF-NB-3 r CDDP 1000 cells. PANTHER pathway analysis indicated angiogenesis to be the fourth most significantly differentially regulated signalling pathway [see Additional file 4]. Hierarchical cluster analysis of angiogenesis-associated genes separated UKF-NB-3 from UKF-Hierarchical cluster analysis and heatmap showing expression of angiogenesis-associatd genes (taken from PANTHER pathway) in UKF-NB-3, UKF-NB-3 r VCR 10 , or UKF-NB-3 r CDDP 1000 cells Figure 1 Hierarchical cluster analysis and heatmap showing expression of angiogenesis-associatd genes (taken from PANTHER pathway) in UKF-NB-3, UKF-NB-3 r VCR 10 , or UKF-NB-3 r CDDP 1000 cells. Data were merged from two independent exeriments each comparing UKF-NB-3 with one chemoresistant cell line. UKF-NB-3a was analysed together with UKF-NB-3 r VCR 10 , UKF-B-3b was analysed together with UKF-NB-3 r CDDP 1000 .
To further investigate the influence of chemoresistance acquisition on the pro-angiogenic potential of cancer cells, a panel of chemsensitive and chemoresistant neuroblastoma cell lines was systematically investigated for their angiogenic phenotypes.

Influence of neuroblastoma cell line supernatants on endothelial cell growth and survival
Neuroblastoma cell lines were grown for seven days. Then medium was removed, cells were washed and protein-free medium was added. After 48 h incubation, supernatants were collected, adjusted to the same protein content, mixed in a 1:1 ratio with fresh IMDM, and FCS (resulting in 10% FCS) was added (unless otherwise noted, all cell culture supernatants were handled following this procedure). HUVECs were trypsinised and suspended in the mixtures of supernatants, fresh IMDM and FCS (without further addition of growth factors). 10 3 cells suspended in 100 μl of respective medium were seeded per well in 96well plates. After five days, HUVEC growth was examined by viability assay (Table 1). HUVECs suspended in IMDM plus 10% FCS did not grow (negative control). HUVECs suspended in IMDM plus 15% FCS, 5% pooled human serum, and basic fibroblast growth factor (bFGF) 2.5 ng/ ml formed vital, closely grown monolayers (positive control). Cell viabilities were calculated relative to positive control. Supernatants from cell lines adapted to cytotoxic drugs induced stronger HUVEC growth than supernatants from parental chemosensitive cells (Table 1). Moreover, the neuroblastoma cell lines UKF-NB-4 and Be(2)-C that were isolated as chemoresistant cell lines from patients materials induced stronger HUVEC growth than the chemosensitive parental cell lines UKF-NB-3, UKF-NB-2, or IMR-32. Subsequently, growth kinetics of HUVECs (determined by cell count) incubated with supernatants of UKF-NB-3, UKF-NB-3 r VCR 10 , UKF-NB-3 r CDDP 1000 , or UKF-NB-3 r DOX 20 cells were compared confirming increased growth of HUVECs incubated with supernatants of chemoresistant cells (Figure 2A).
Next, the influence of neuroblastoma cell culture supernatants was examined on HUVEC survival. Confluent HUVEC monolayers were washed and incubated for 48 h with supernatants of UKF-NB-3, UKF-NB-3 r VCR 10 , UKF-NB-3 r CDDP 1000 , or UKF-NB-3 r DOX 20 cells and HUVEC viability was determined. Results revealed increased Influence of neuroblastoma cell culture supernatants on endothelial cell growth and viability HUVEC viability in cultures incubated with supernatants of chemoresistant cells ( Table 1). Lack of growth factors or nutrients induces apoptosis in endothelial cells [26][27][28]. Therefore, we investigated caspase 3/7 activation as indicator of apoptosis in confluent HUVEC monolayers incubated for 48 h with supernatants of UKF-NB-3, UKF-NB-3 r VCR 10 , UKF-NB-3 r CDDP 1000 cells or UKF-NB-3 r DOX 20 cells. Results indicated decreased caspase activation in HUVECs incubated with supernatants from chemoresistant cells ( Figure 2B).

Influence of neuroblastoma cell line supernatants on endothelial cell tube formation
HUVECs were suspended with supernatants of neuroblastoma cell lines and seeded on extracellular matrix (Matrigel). After 16 h, tube formation was determined.

Influence of neuroblastoma cell line supernatants on activation of pro-angiogenic signalling events in endothelial cells
The phosphoinositide-3-kinase (PI3K) -Akt (also known as protein kinase B, PKB) signalling pathway, "classical" mitogen-activated protein kinase (MAPK) signalling via Ras-Raf-MEK-ERK, and activation of nuclear factor κB (NFκB) are involved in angiogenesis signalling in endothelial cells [29][30][31]. The influence of supernatants of chemoresistant cells on Akt phosphorylation or ERK 1/2 phosphorylation in HUVECs is shown in Figure 3C. Densitometric analysis of Western blot data is given in Additional file 8. Akt may be activated through phosphorylation at Ser473 and/or at Thr308. The supernatants of UKF-NB-3 r VCR 10 or UKF-NB-3 r CDDP 1000 cells induced enhanced Akt phosphorylation at Thr308 and ERK 1/2 phosphorylation in comparison to UKF-NB-3 supernatants. All supernatants of chemoresistant cells caused enhanced NFκB activation relative to supernatants of chemosensitive UKF-NB-3 cells ( Figure 3D).

Chemoresistant cancer cells induce increased vessel formation in animal models
Vessel formation was first investigated in vivo in the CAM of fertilised eggs. 10 6 tumour cells were seeded onto the CAM per egg (eight eggs per cell line) at day 10. Vessel for-mation was scored by two independent observers at day 14. Results indicated higher vessel formation in chemoresistant (UKF-NB-3 r VCR 10 , UKF-NB-3 r DOX 20 ) cells than in chemosensitive (UKF-NB-3) cells ( Figure 4A).

Increased pro-angiogenic activity of chemoresistant neuroblastoma cells is mediated by individual molecular mechanisms
VEGF is a pro-angiogenic factor that has frequently been associated with neuroblastoma angiogenesis [32,33]. However, increased VEGF levels were not consistently found in supernatants of chemoresistant cells [see Additional file 9]. Acute cisplatin treament has been described to induce tumour progression through VEGF expression in paediatric tumour cells including the neuroblastoma cell line SK-N-BE2 [34]. In cisplatin-resistant neuroblastoma cells, VEGF expression has not been investigated, yet. Expression of a number of further pro-and anti-angiogenic factors has been suggested to be relevant for neuroblastoma angiogenesis including platelet-derived growth factor α (PDGFα), matrix metalloproteinase 2 (MMP-2), MMP-9, erythropoietin (EPO), EPO receptor, activin A, interleukin-6 (IL-6), leukemia inhibitory factor (LIF), tissue inhibitor of metalloproteinase 2 (TIMP2), pigment epithelial-derived growth factor (PEDGF), secreted protein acidic and rich in cysteine (SPARC), thrombospondin-1, and thrombospondin-2 [33]. However, analysis of gene microarray data from neuroblast-oma cell lines did not reveal specific expression of these or other angiogenesis-related genes that would suggest a single common molecular event underlying increased neuroblastoma tumour angiogenesis in all chemoresistant cells (data not shown).
N-myc amplification has also been reported to result in increased neuroblastoma tumour angiogenesis through different mechanisms [33,[35][36][37]. However, UKF-NB-3 r DOX 20 cells showed enhanced pro-angiogenic potential compared to UKF-NB-3 cells although both cell lines do neither differ in N-myc amplification nor in N-myc expression [8,18]. This indicates that the N-myc status may not generally be critical for increased pro-angiogenic potential of chemoresistant cells. Furthermore, the loss of functional p53 during tumourigenesis has been correlated to a more pro-angiogenic tumour phenotype [38]. However, in our experiments pro-angiogenic activity was enhanced in both p53-mutated and p53-wild-type chemoresistant neuroblastoma cells (Table 1). Taken together, Influence of neuroblastoma cell culture supernatants on tube formation and activation of pro-angiogenic signalling pathways in endothelial cells the more pro-angiogenic phenotype observed in chemoresistant neuroblastoma cells appears to result from different individual shifts in the expression of angiogenesisassociated genes.

Doxorubicin inhibits tumour angiogenesis and growth of doxorubicin-resistant neuroblastoma xenografts
Data had indicated individual changes in the expression of angiogenesis-related genes to be responsible for the proangiogenic phenotype of chemoresistant neuroblastoma cells (see above). To investigate if the increased proangiogenic activity of chemoresistant neuroblastoma cells may be relevant for enhanced growth of chemoresistant neuroblastoma xenografts, doxorubicin-resistant UKF-NB-3 r DOX 20 neuroblastoma cells were treated with doxorubicin that is known to interfere with angiogenesis by direct influence on endothelial cells [39,40].
Administration of a single dose of doxorubicin 10 mg/kg i.v. into mice results in maximal doxorubicin plasma levels in the range of 500 -600 ng/ml that decline to doxorubicin plasma levels of 20 -30 ng/ml 24 h after injection [40][41][42].

Discussion
Here, we used a bioinformatics-based approach based on transcriptomics data to identify signalling pathways asso-ciated with increased malignant behaviour of chemoresistant neuroblastoma cells. Angiogenesis signalling belonged to the top 5 pathways most strongly differentially regulated between chemosensitive and chemoresistant neuroblastoma cells. Systematic evaluation of a panel of neuroblastoma cell lines in cell culture and animal models showed consitently increased pro-angiogenic acivity exerted by chemoresistant cells. These findings are in accordance with previous reports showing that human melanoma and breast cancer cells selected for resistance to chemotherapeutic agents produced higher levels of multiple angiogenic factors [44,45]. Moreover, an increased microvessel density (MVD) was detected in chemotherapy resistant xenograft tumours [44,45].
Selection of cancer stem cells has been suggested to play a role in the enhanced pro-angiogenic activity seen in chemoresistant cancer cells. In lung cancer cells, treatment with cisplatin, doxorubicin, or etoposide resulted in the selection of cancer stem cells as indicated by cell biology and analysis of expression of stemness genes [46]. These chemotherapy-selected cancer stem cells were responsible for the observed increased pro-angiogenic properties of lung cancer cells. In the absence of cytotoxic drugs, lung cancer cell lines returned to their initial phenotype and reacquired drug sensitivity [46]. In contrast, UKF-NB-3 r VCR 10 and UKF-NB-3 r CDDP 1000 cells remained chemoresistant and did not loose their pro-angiogenic phenotype even when they were cultivated for up to six months in the absence of drugs (data not shown). This suggests that chemoresistance and pro-angiogenic activity in these cell lines are not consequence of a simple chemotherapyinduced selection of cancer stem cells that are already present in the parental UKF-NB-3 cell line. Moreover, acute cisplatin treatment increased VEGF expression together with expression of the stemness genes Nanog, Bmi-1, and Oct 4 in osteosarcoma (HOS), rhabdomyosarcoma (RH-4) and neuroblastoma (SK-N-BE2) cell lines [46]. However, none of these stemness genes was found up-regulated in UKF-NB-3 r VCR 10  The view that individual chemoresistant neuroblastoma cell lines exert pro-angiogenic effects by individual mechanisms is supported by the results derived from the examination of pro-angiogenic signalling in endothelial cells incubated with supernatants from different neuroblast-oma cell lines. Supernatants of chemoresistant UKF-NB-3 r DOX 20 , UKF-NB-3 r VCR 10 , and UKF-NB-3 r CDDP 1000 cells enhanced NFκB activation compared to supernatants of chemosensitive UKF-NB-3 cells However, only supernatants of UKF-NB-3 r VCR 10 and UKF-NB-3 r CDDP 1000 cells but not UKF-NB-3 r DOX 20 cells elevated Akt and ERK 1/2 phosphorylation in endothelial cells. Based on these differences in the activation of pro-angiogenic signalling events in endothelial cells, it appears plausible that endothelial cell activation might be caused by different chemoresistant neuroblastoma cell lines by different molecular mechanisms resulting in up-or down-regulation of varying pro-or anti-angiogenic factors.
Possibly, there is an overlap between gene products involved in angiogenesis and gene products relevant in chemoresistance. Indeed, among between the aniogenesis-associated genes that were differentially expressed there are those that are also considered to contribute to chemoresistance. Three arbitrarily chosen examples are BIRC5, MAPK3, and AKT1. BIRC5 (increased expression in UKF-NB-3 r VCR 10 vs. UKF-NB-3, [see Additional file 2], and in UKF-NB-3 r DOX 20 vers. UKF-NB-3, [see Additional file 6]) encodes for a protein that is also named survivin and plays a prominent role in apoptosis inhibition and cancer cell chemoresistance [47]. Moreover, BIRC5 expression in cancer cells has been linked to tumour angiogenesis [48] and inhibition of BIRC5 expression in tumour cells decreased tumour angiogenesis [49,50]. MAPK3 (increased expression in UKF-NB-3 r VCR 10 vs. UKF-NB-3, [see Additional file 2]) encodes for a protein that is also called extracellular signal-regulated kinase 1 (ERK1) and is a constituent of the "classical" MAP kinase pathway Ras/Raf/MEK/ERK. ERK1 phosphorylation protects cancer cells from different entities against chemotherapy-induced apoptosis [51][52][53]. Moreover, MAPK3 activation/phosphorylation induces production of proangiogenic factors in renal carcinoma cells [54]. AKT1 (increased expression in UKF-NB-3 r DOX 20 vers. UKF-NB-3, [see additional file 6]) encodes for a protein also called protein kinase B (PKB) that is a central mediator of survival signals transduced by the phosphatidylinositol 3kinase and is involved in chemoresistance [55][56][57] as well as in cancer cell expression of pro-angiogenic factors [58][59][60]. Remarkably, an angiogenesis-associated gene expression signature had been described before to predict the sensitivity of cancer cells to artemisinins, an anti-cancer active group of anti-malaria drugs [61].
The complexicity of pro-angiogenic mechanisms observed in chemoresistant neuroblastoma cells is in accordance with other reports demonstrating that pro-angiogenic activity of cancer cells is commonly caused by complex changes in angiogenesis signalling and that inhibition of one pro-angiogenic event may not be enough to interfere with tumour vessel formation [62]. N-myc-amplified neuroblastoma cells that exert pro-angiogenic activity mainly through VEGF have very recently been shown to rapidly develop alternative pro-angiogenic mechanisms when VEGF signalling is inhibited [63]. In addition, up-regulation of multiple pro-angiogenic factors enabled carcinoma cells to escape from angiogenesis inhibition by the three endogenous anti-angiogenic molecules thrombospondin-1, endostatin, and tumstatin [64]. Notably, combination therapy of metastatic breast cancer with paclitaxel and the anti-VEGF-A antibody bevacizumab resulted in prolonged progression-free survival but did not influence overall survival relative to paclitaxel in a phase III trial [65]. In the light of the findings presented here, one may speculate that anti-angiogenic therapy may prolong progression-free survival but that resistance development (to chemotherapy and/or anti-angiogenic therapy) may result in a more aggressive cancer cell phenotype, which might be the reason for the decreased time period observed between tumour re-onset and patients' deaths.
High tumour angiogenesis and high-level expression of pro-angiogenic factors at diagnosis have previously been suggested to be correlated with advanced disease stages in neuroblastoma [32,66,67]. However, the prognostic value of angiogenesis in neuroblastoma at diagnosis is still a matter of debate [68,69]. Notably, analysis of two different data sets reporting on gene expression profiles in tumours from poor outcome or bad outcome N-myc amplified [70] or non-N-myc amplified [71] neuroblastoma patients indicated statistically significant differences in angiogenesis signalling between these groups [see Additional files 13,14]. To investigate if the increased proangiogenic phenotype observed in chemoresistant cells may contribute to tumour progression, xenografts grown from doxorubicin-resistant (UKF-NB-3 r DOX 20 ) cells were treated with doxorubicin, an anti-cancer drug that exerts anti-angiogenic activity by direct effect on endothelial cells [39,40]. Tumour vessel formation and growth were strongly reduced by doxorubicin in doxorubicin-resistant xenografts. Although it cannot be concluded without a doubt that the entire effect on xenograft growth can be attributed to inhibition of angiogenesis, microvessel density was statistically reduced supporting the view that inhibition of angiogenesis has definitely contributed. Therefore, these data suggest that increased pro-angiogenic activity of doxorubicin-resistant cells contributes to their more malignant phenotype and that anti-angiogenic strategies that target endothelial cells might represent a therapeutic option for neuroblastoma treatment.

Conclusion
Bioinformatics pathway analysis indicated differences in the expression of angiogenesis-associated genes between chemosensitive and chemoresistant neuroblastoma cell lines. Cell culture and animal data showed that acquired resistance to different anti-cancer drugs resulted in increased pro-angiogenic activity of neurobastoma cells.
The changes in angiogenesis signalling observed in chemoresistant neuroblastoma cells are very complex and differ between individual cell lines. Therefore, individual molecular mechanisms appear to be responsible for the enhanced pro-angiogenic activity that was consistently observed in all investigated chemoresistant neuroblastoma cell lines relative to chemosensitive cells. Doxorubicin treatment of doxorubicin-resistant neuroblastoma xenografts resulted in decreased vessel formation and tumour growth suggesting that the more pro-angiogenic phenotype of chemoresistant cells may contribute to increased malignancy of chemoresistant neuroblastoma cells and that endothelial cell targeting may represent a possibility for therapeutic intervention. The complex nature of the chemoresistance-associated changes responsible for the more pro-angiogenic phenotype strongly stresses the need for an improved understanding of biological processes like angiogenesis on a systems biology level.