1 Introduction
Glioblastoma (GBM) is a World Health Organization (WHO) defined grade IV astrocytoma, which is the most aggressive form of glioma [
1]. Despite notable achievements that have been made in the past, the clinical management of GBM patients is still a major challenge [
2], which may primarily be attributed to its heterogeneous nature [
3]. Temozolomide is an oral alkylating agent that is able to traverse the blood-brain barrier but, although promising results have been obtained in the standard care for GBM [
4‐
6]
, its efficacy ultimately diminishes due to the aquisation of resistance [
7]. Cisplatin has been used as neoadjuvant therapy with temozolomide [
8] and as adjuvant therapy with carmustine in patients with malignant gliomas [
9]. In heavily pre-treated patients with relapsed high-grade gliomas refractory to temozolomide alone, combinations of temozolomide and cisplatin have been moderately effective [
10]. When used as a first-line chemotherapy with fractionated temozolomide in chemotherapy-naïve patients with recurrent GBM, it was found to result in a progression-free survival (PFS) of 6 months [
11]. Resistance to cisplatin does, however, arise, potentially due to high levels of glutathione-S-transferase (GST) activity [
12]. Etoposide has been found to improve the survival of patients with high-grade glioma [
13], but finally most patients succumb to GBM as a result of relapse.
The tumour microenvironment consists of a myriad of factors, including extracellular cellular matrix (ECM) components [
14] and variations in oxygen tension [
15], that may affect tumour biology. It has also been shown that the ECM composition may affect cancer cell proliferation and motility [
16]. GBMs are characterized by regions of hypoxia [
17] ranging from 0.1% to 10% [
18] which are thought to play important roles in the aggressiveness of the tumours [
18]. As such, hypoxia has been considered to serve as a biomarker for a poor prognosis, and to be associated with radio-resistance and chemo-resistance, as well as tumour cell migration and invasion [
19]. Hypoxic regions have also been shown to harbour GBM stem cell populations [
20]. Hypoxia inducible factors (HIFs) have been postulated to be the primary mediators of hypoxic responses [
21]. The HIF factors HIF-1α and HIF-2α are known to be able to bind to the same DNA sequence, referred to as hypoxia responsive element (HRE), and to heterodimerise with HIF-1β to activate target genes. However, HIF-1α and HIF-2α may also exhibit functional differences, including different protein-protein interactions [
22] and tissue specificities [
23]. HIF-1α has been found to promote the survival of glioma cancer stem cells under hypoxic conditions [
24], and HIF-2α is thought to be linked to the GBM stem cell phenotype within the hypoxic niche [
20]. Additionally, it has been found that in neuroblastoma cells exposed to hypoxia, the expression of HIF-1α rapidly peaks and declines, whereas that of HIF-2α remains steadily upregulated, thereby mediating a long lasting response [
23].
Since cancer stem cells (CSCs) have been implicated in tumour recurrence and drug failure, their identification and characterization is of therapeutic relevance [
25,
26]. Several markers have been employed to differentiate CSCs from the vast majority of cells within a tumour sample [
27]. In GBM, CD133 is commonly used as a CSC marker [
28]. As yet, however, only a few studies have been performed on the biological significance of CD133 in GBM. In addition, the most appropriate environment in which to study GBM cells has been contentious. Most assays used so far do not take into account the influence of the tumour microenvironment (TME) and its role in regulating CSC numbers and phenotypes. Tumour cells, including GBM cells, are traditionally cultured in vitro with an oxygen tension of 20%. Since this may not accurately reflect the behaviour of tumours in the in vivo situation, it may explain why 95% of the anti-cancer drugs tested in vitro have failed their translation to the clinic [
29].
In the present study, we sought to understand the role of hypoxia on CD133 expression, and whether or not HIFs play any regulatory role in this expression. We also assessed the role of hypoxia in etoposide and cisplatin chemo-resistance and how HIF and CD133 downregulation may sensitize GBM cells to chemotherapy. Our results strongly indicate that CD133 upregulation under hypoxic conditions is mediated by HIFs, with a longer-lasting effect of HIF-2α. In addition, we found that CD133 knockdown sensitized GBM cells to cisplatin in a HIF-independent manner.
2 Materials and methods
2.1 Cell lines and culture conditions
U87 cells were derived from the European Collection of Authenticated Cell Cultures (ECACC), while U251 and SNB19 cells were derived from the National Cancer Institute USA (NCI-60). The GBM cells were grown under normoxic (20% oxygen) or hypoxic (1% oxygen) conditions as standard 2D cultures or as Cultrex-based 3D cultures (Trevigen, Gaithersburg, MD, USA).
2.2 Quantitative real-time PCR
RNA extraction, cDNA synthesis and quantitative real-time RT-PCR (qRT-PCR) were performed as previously reported [
30] using SYBR green (Eurogentec) for detection. The data were expressed relative to the housekeeping gene
HPRT and calculated using the 2
-∆∆Ct method. The primer sequences used were: HPRT (F) 5’-ATTATGCTGAGGATTTGGAAAGGG-3′ and (R) 5’-GCCTCCCATCTCCTTCATCAC-3′; CD133 (F) 5’-CAATCTCCCTGTTGGTGATTTG-3′ and (R) 5’-ATCACCAGGTAAGAACCCGGA-3′; VEGF (F) 5’-CCAAGTGGTCCCAGGCTGCA-3′ and (R) 5’-TGGATGGCAGTAGCTGCGCT-3′; HIF1A (F) 5’-CCTCTGTGATGAGGCTTACCATC-3′ and (R) 5’-CATCTGTGCTTTCATGTCATCTTC-3′, HIF2A (F) 5’-CCACCAGCTTCACTCTCTCC-3′ and (R) 5’-TCAGAAAAGGCCACTGCTT-3′.
2.3 Small interfering RNA transfections
GBM cells were transfected with CD133, HIF-1α and HIF-2α siRNAs (Eurogentec) using a Lipofectamine® RNAiMAX Transfection Reagent (Life Technologies) according to manufacturer’s instructions. The sequences used were: CD133siRNA- GAUCAAAAGGAGUCGGAAA, HIFIAsiRNA- GCCACUUCGAAGUAGUGCU and HIF2AsiRNA- GCGACAGCUGGAGUAUGAA.
2.4 3D cultures
Cultrex basement membrane extract (BME; Trevigen) was diluted to a concentration of 3 mg/ml on ice using phenol red-free modified RPMI-1640 medium (Life Technologies). Next, the cells were resuspended at appropriate densities and seeded into black-walled, low-adherent, clear-bottom 96-well culture plates (BrandTech) prewarmed to 37 °C.
2.5 Drug sensitivity assays
A cisplatin stock solution of 1 mg/ml was diluted to appropriate concentrations. GBM cells were subsequently incubated with drugs for 48 h after which an Alamar Blue cell viability assay (Invitrogen) was carried out (10% v/v, 37 °C, 1 h). The resulting fluorescence was measured using a fluorescence plate reader (Flex-Station II, Molecular Devices, CA, USA) and IC50 values were calculated relative to untreated cells using the Graphpad prism software tool. Drug sensitivities were calculated as percentages of matched untreated controls. IC50 curves were plotted and values determined using GraphPad Prism 6 (GraphPad Software Inc., USA; nonlinear curve fit of Y = 100/ (1 + 10(LogIC50-X)*HillSlope).
2.6 Flow cytometry
After harvesting, GBM cells were spun down in Eppendorf tubes and re-suspended in 80 μl buffer and 20 μl FcR blocker (Miltenyi Biotec), after which 10 μl anti-CD133/1 (AC133)-PE antibody (Miltenyi Biotec) or a mouse monoclonal IgG-PE isotype control was added according to the manufacturer’s instructions. Next, the cells were analysed using a Beckman Coulter flow cytometer. The resulting data were analysed using Weasel software (
http://www.frankbattye.com.au/Weasel/).
2.7 Statistic analyses
Graphpad Prism version 6 was used to analyse all data. Data comparisons were carried out using either Student’s t-test or one-way ANOVA (Turkey’s multiple comparison test), when appropriate.
4 Discussion
The emerging concept that tumour development largely relies on signals derived from the tumour microenvironment (TME) has led to a paradigm shift in cancer research [
31]. Locally advanced tumours are frequently characterized by hypoxia resulting from abnormal micro-vessel structure and function, and increased distances between blood vessels and tumour cells [
32]. Hypoxia is an important feature of the glioblastoma (GBM) TME [
20] and cells within their hypoxic niches have been shown to be chemo-resistant [
33]. Here, we observed CD133 expression upregulation in GBM cells cultured in 2D and 3D models under hypoxic conditions and, although largely consistent with other reports, different CD133 levels were observed in the different GBM cells tested [
25,
34]. CD133 expression upregulation under hypoxic conditions has been reported before [
35]. Our current data indicate that hypoxia acts on GBM cells irrespective of whether they are cultured in 2D or 3D. Although no significant CD133 upregulation was reached in the 3D model, a consistent higher level of CD133 expression was observed when the cells were grown in 3D, similar to what has recently been reported for GBM cells grown in a 3D chitosan-alginate scaffold [
36]. CD133 protein expression was also verified in the 2D models using flow cytometry. We found, however, that this expression rapidly decreased after harvesting of the cells, making it technically challenging to assess CD133 expression in the 3D cultures, which require additional manipulations to free and disaggregate the GBM cell clusters. This observation may also explain previous discrepancies noted in the analysis/sorting of CD133 positive cell populations via flow cytometry and why apparently CD133 negative cells may give rise to CD133 positive cells.
In both the 2D and 3D GBM cell models we found that HIF-1α and HIF-2α knockdown led to a reduced CD133 expression. While the effect of HIF-1α silencing on CD133 expression diminished rapidly, we found that the effect of HIF-2α silencing on CD133 expression became stronger over time, similar to its effect on VEGF expression, which is a known HIF downstream target. This finding is in keeping with a previous report on neuroblastoma cells [
23]. The underlying mechanism may involve the hypoxia-associated factor HAF, which becomes upregulated during acute hypoxia and selectively binds to HIF-1α, resulting in its degradation, while at the same time HIF-2α transactivation and stability are promoted [
37]. SIRT1, a histone deacetylase, also appears to be involved in this switch, i.e., it may deacetylate HIF-1α and reduce its activity while enhancing HIF-2α-mediated transcription through a HIF-dependent mechanism [
38].
Chemo-resistance plays a crucial role in GBM tumour progression [
39,
40] and it has been found that cells located in its hypoxic regions are more resistant to chemotherapy and may give rise to recurrences [
17]. Our current results also indicate that GBM cells become more resistant to temozolomide, cisplatin and etoposide under hypoxic conditions. This, however, is not a universal finding. While in one study breast cancer cells exposed to low oxygen levels (< 0.1%) were also found to be resistant to a range of chemotherapeutic agents, others have found differential responses of cells to therapeutic drugs under hypoxic conditions, with cisplatin being more effective in hypoxic breast cancer, small cell lung cancer and lymphoma cells compared to the same cells maintained under normoxic conditions [
41]. The mechanism underlying this drug resistance may be cell type dependent. In the breast cancer cells, it has been found that enhanced resistance to cisplatin under hypoxic conditions was mediated through HIF-1α [
42], whereas in osteosarcoma cells hypoxia-induced resistance to cisplatin, doxorubicin and etoposide has been found to be HIF-independent and to occur via hypoxia-driven attenuation of p53 activation [
43]. In the present study, we found that siRNA-mediated HIF-1α knockdown, but not HIF-2α knockdown, significantly sensitized GBM cells to temozolomide. Others have similarly found that HIF-1α downregulation by BMP2 may increase the responsiveness of GBM cells to temozolomide [
44]. Our findings are also in agreement with those from Li et al. who found that HIF-1α silencing increases the sensitivity of tumour cells to temozolomide in vivo [
45].
We also found that HIF-2α downregulation sensitized GBM cells to cisplatin more effectively than HIF-1α downregulation, substantiating the differential roles of HIF-1α and HIF-2α in tumour development [
46]. This effect of HIF downregulation on drug sensitivity was, however, relatively small compared to the effect of CD133 downregulation, which resulted in an increase in cisplatin sensitivity of at least 2-fold in all GBM cells tested and as much as 7-fold in U251 cells. In contrast, we found that CD133 downregulation did not significantly affect the temozolomide sensitivity of GBM cells. This latter observation is in accordance with that of Perazzoli et al. who concluded from their results that CD133 was unrelated to temozolomide resistance [
47], but disagrees with others that have shown that CD133
+ GBM cells are resistant to temozolomide [
48]. Our results suggest that the hypoxia-induced cisplatin sensitivity of GBM cells may be HIF independent and may be directly or indirectly induced via CD133 activation. One potential mechanism for CD133-dependent resistance may be CD133-mediated activation of the Erk pathway, which has indeed been observed after CD133 overexpression in GBM cells [
49]. Interestingly, phosphorylation of the tyrosine-828 residue in the C-terminal cytoplasmic domain of CD133 has been found to mediate direct interactions between CD133 and the phosphoinositide 3-kinase (PI3K) 85 kDa regulatory subunit (p85), resulting in preferential activation of the PI3K/protein kinase B (Akt) pathway in GBM stem cells compared to matched non-stem cells [
50]. Therefore, upregulation of CD133 in GBM cells under hypoxic conditions may lead to activation of the anti-apoptotic Akt pathway, resulting in drug resistance [
51].
Although cisplatin as a neoadjuvant agent, alone or in combination with other agents including temozolomide, has been explored and some benefits have been noted [
8‐
11], eventually resistance may occur resulting in a poor clinical outcome. While HIF-dependent and HIF-independent mechanisms of cisplatin resistance have previously been described, we identified a novel mechanism of hypoxia-induced cisplatin resistance that is mediated via CD133. This mechanism may also, at least partly, explain the observed association between CD133 expression and a poor prognosis [
52], and suggests a role for CD133 as a functional molecule in its own right rather than as a CSC marker. A further in-depth investigation of the mechanism underlying this hypoxia-induced CD133-dependent cisplatin resistance may facilitate the identification of cisplatin combination therapies to improve its efficacy for clinical use in neo-adjuvant and/or adjuvant settings.