Profiling anti-apoptotic protein BCL-xL expression in U-87 MG glioblastoma cells and patient-derived tumorspheres

Glioblastoma is one of the cancers with the worst prognosis, despite huge efforts to understand its unusual heterogeneity and aggressiveness. These are mainly attributable to glioblastoma stem cells, which are also responsible for the frequent tumour recurrence following surgery or chemo/radiotherapy. We report here that tumorspheres derived from the U-87 MG glioblastoma cells have an increased expression of the anti-apoptotic protein BCL-xL. Modulation of this expression in tumorspheres enlightened us on the potential role of apoptosis and BCL-2 proteins might play in the survival of glioblastoma stem cells. Moreover, increased BCL-xL expression appears to sensitise glioblastoma cells to the newly developed BH3 mimetics, opening new therapeutic perspectives for treating glioblastoma patients.


Introduction
Glioblastoma (GBM) is the most frequent primary malignant and invasive adult brain tumour, with an extremely devastating prognosis since the five-year survival rate does not exceed five per cent (1). The current therapy (STUPP treatment) combines surgical resection of the primary tumour, radiotherapy and chemotherapy with the alkylating agent temozolomide (TMZ) (2). However, despite these aggressive treatments, GBM invariably relapse, iterative surgical removal is rarely feasible and no consensual medical treatment is available leading to tumour progression and patient death (3).
GBM recurrence can partially be explained by the important cellular complexity of GBM tumours and by a high inter-and intra-tumoral heterogeneity (4). Due to their pronounced cellular and molecular heterogeneity, GBM cells can adapt to selective pressures and changes in their microenvironment, leading to increased tumour aggressiveness and treatment resistance (5).
GBM tumours contain a small subset of cells with both stemness (selfrenewing, proliferation, differentiation) and tumour-initiating properties, called glioblastoma stem cells (GSCs), driving tumour recurrence in GBM patients even after STUPP treatment (6) (7). Similarly to neural stem cells which have the capacity to form neurospheres in vitro, GSCs have the ability to form clonogenic structures, called tumorspheres, in serum-free non-adherent conditions in vitro (6) (8). These 3D culture models therefore represent models of choice to unravel and test new therapeutic strategies to specifically target GSCs and prevent GBM development (9).
Cancer cells must overcome several cellular stresses such as DNA damage, inflammation, oncogene activation, aberrant cell cycle progression and a harsh microenvironment that would normally trigger apoptosis in nontransformed cells (10). Apoptosis, the most extensively studied form of programmed cell death, is essential in maintaining the homeostatic balance between cell proliferation and cell death. Apoptosis can be triggered by two major signalling pathways, both leading to effector caspases activation and subsequent DNA fragmentation, plasma membrane blebbing and cell shrinkage (11). The extrinsic pathway is induced upon activation of death receptors (superfamily of tumour necrosis factor receptors) on the cell surface (12). The intrinsic pathway can be activated through various signals such as DNA damage, reactive oxygen species (ROS), or growth factor deprivation (11). Following mitochondrial outer-membrane permeabilization (MOMP), cytochrome c is thus released into the cytosol and engages the formation of the apoptosome that will first activate initiator caspase-9 and then effector caspases-3 and -7. These two pathways are controlled by BCL-2 family members: the pro-apoptotic (BAX, BAK, BID or BAD) and the anti-apoptotic proteins (BCL-2, BCL-xL, BCL-w, MCL-1) (11) (13) (14).
Cancer cells commonly share the ability to escape to programmed cell death. This evasion allows cancer cells to grow and develop into a tumour, while also contributing to treatment resistance (15). The blockade of cell death is a frequent cause of treatment resistance in GBM (16) (17) (18) (19).
Apoptosis escape commonly occurs through the deregulation of proand anti-apoptotic proteins in human cancers cells. The overexpression of anti-apoptotic proteins at transcriptional and protein levels has been observed in various cancers. BCL-2 was first described to be constitutively expressed in follicular lymphoma and the amplification of MCL1 and BCL2L1 (encoding BCL-xL) are the most frequent in solid cancers (13). In glioblastoma, MCL-1 is also overexpressed while high BCL-xL expression is often associated with poor prognosis and advanced disease (20) (21). Additionally, BCL-xL expression was also shown to increase with chemotherapy and ionizing radiation in lung cancer. Its role in stemness and aggressiveness is documented in melanoma and glioblastoma (22). Recently, BH3 mimetics developed to functionally imitate the pro-apoptotic BCL-2 proteins were shown to neutralize the anti-apoptotic proteins, enabling efficient apoptosis in cancer cells (23).
Considering its important function in regulating the apoptotic response in several cancers, we therefore focused on characterizing the expression and possible role of BCL-xL in GSCs growth and possible resistance to BH3 mimetics. The main finding of this short investigative work is that BCL-xL is highly expressed in tumorspheres originating from U-87 MG GBM cells, rendering them specifically sensitive to the BCL-xL inhibitor ABT-263.
Although these findings are in our hands limited to one GBM cell line, this study provides interesting preliminary data for future research into repurposing BH3 mimetics for GBM treatment.

High diversity of BCL-xL expression among GBM cell lines
As several research articles highlighted a link between resistance to apoptosis and cancer development in GBM, we speculated that anti-apoptotic proteins could be implicated in resistance to the therapy of CSC-like cells. Several protocols to isolate and culture GBM CSCs are described and currently used, however the most common one is culturing GBM cells in serum-free media, complemented by EGF and FGF which favours GBM tumorspheres formation ( Fig. 1 A).
We sought to investigate BCL-xL expression in several GBM cell lines, either commercially available or patient-derived GSC tumorspheres cultures. These cells were either grown as adherent, differentiated cells, or allowed to form spheres when deprived of serum ( Fig. 1 A). Given the widely accepted heterogeneity between GBM subtypes and even within the same tumour, it came as no surprise that all types of cells we tested show a different pattern of BCL-xL expression. This result highlights that the complexity of GBM is also mirrored in the different patterns of BCL-xL expression among various GBM cells. Of interest, the U-87 MG cells show a clear increase of BCL-xL expression when grown as tumorspheres (Fig. 1 B). We thus focused on this qRT-PCR ( Fig. 2 B). To exclude an indirect impact of EGF and bFGF added for tumorspheres culture, we tested the effect of these cytokines on the expression on BCL-xL, however no effect was noticed on its protein expression ( Fig. 2 C). These initial results therefore strongly suggest that U-87 MG tumorspheres support the growth of cells with GSC-like molecular signatures. Taken together, these results show that GSC-enriched U-87 MG-derived tumorspheres highly express the anti-apoptotic BCL-xL protein.

BCL-xL regulates the size of U-87 MG neurospheres
We next investigated the possible roles that BCL-xL might play in the biology of GBM tumorspheres. For this, we first took advantage of U-87 MG cells stably expressing a degradation-sensitive DD-FLAG-BCL-xL that is therefore constantly degraded in the absence of Shield-1 (24). To finely tune the overexpression of BCL-xL in spheres, we first tested various concentrations of Shield-1 and choose 100 nM as working concentration to induce a significant BCL-xL protein accumulation for future experiments (Figure 3 A). Next, we  (Fig. 3 F). Since GBM spheres are composed of a mixture of GSCs and differentiated cells, we speculated that BCL-xL might impact on the proliferation of the latter cellular population. We next tested the effect of BCL-xL downregulation by stably expressing two lentiviral vectors expressing BCL-xL-targeting shRNAs. First, the efficacy of BCL-xL knockdown was validated both in differentiated cells and U-87 MG tumorspheres (Fig. 4 A). Of note, BCL-xL knockdown does not affect cell proliferation in the differentiated growth conditions (Supp. Fig. 1 A), while it clearly sensitize U-87 MG cells to actinomycin D induced apoptosis (Supp. Fig. 1 B). Actinomycin D is a chemotherapy drug blocking DNA transcription and therefore inducing apoptosis. Interestingly, knock-down of BCL-xL has the opposite effect on U-87 MG tumorsphere growth compared to its overexpression, namely the size of tumorspheres is significantly impaired ( Fig. 4 B, C), again without a noticeable effect on the expression on stemness markers (Fig. 4 D). In summary, these results suggest that in a model of U-87 MG tumorsphere formation, BCL-xL plays a stemness-independent role in the modulation of GBM tumorspheres size.

BCL-xL upregulation in GBM tumorspheres enhances the therapeutic targeting of GBM using BH3 mimetics
BH3 mimetics are a novel class of drugs tailored to mimic the function of the BH3-only proteins and therefore targeting the pro-survival members of the BCL-2 family (23). Several BH3 mimetics are available, each displaying a different level of affinity for a specific anti-apoptotic protein (Fig. 5 A). To test whether the increased expression of BCL-xL in U-87 MG tumorspheres underlies a newly acquired sensitization to a specific BH3 mimetic, we treated these spheres with the ABT-263, ABT-737 or S63845 BH3 mimetics and assessed the induction of cell death using the IncuCyte live cell imaging coupled with SYTOX Green staining of apoptotic cells. As shown in Figure 5 B, C, U-87 MG tumorspheres undergo apoptosis mush faster (as early as one hour) and at a much higher rate when treated with ABT-263 and ABT-737, two of the BH3 mimetics targeting BCL-xL and BCL2/BCL-xL/BCL-w, respectively. Of interest, the MCL-1 inhibitor, S63845, has a lower killing efficacy (Fig. 5 B-C). In addition, a short treatment of only 3 hours with ABT-263, but not S63845, can efficiently activate the effector caspase-3, as shown by the caspase-3 processing and PARP-1 cleavage (Figure 5 D). Of note, the pan-caspase inhibitor Q-VD-OPh blocks this processing, therefore confirming an induction of apoptosis in tumorspheres treated with the BCL-xL inhibitor ABT-263. We have also performed a long-term tumorsphere survival assay, where U-87 MG spheres were briefly treated for 3 hours with either ABT-263 or S63854 and then allowed to grow in fresh media for 7 days. As depicted in the representative spheres pictures in and corresponding quantifications, the spheres obtained following ABT-263 treatment were both smaller in size of OLIG2 increased sharply (Fig. 5 G). Collectively, these results provide important insights into targeting GBM spheres using tailored BH3 mimetics, owing to their elevated anti-apoptotic expression.

Discussion
In this report, we provide an investigative approach into the pattern of shRNA, and these latter cells were extremely sensitive to actinomycin D, which is highly reminiscent of results obtained with MCL-1 knockdown glioblastoma cells exposed to BH3 mimetics (20).
As confirmed by the upregulation of stemness genes, the population of GSCs is clearly more abundant in U-87 MG tumorspheres compared to adherent cells. However, we cannot yet prove at this stage that BCL-xL is specifically upregulated only in the GSCs compartment, given that tumorspheres are certainly composed of a mix of GSCs, progenitors and differentiated cells (25). To address this issue in future experiments, we would need to isolate from tumorspheres the two populations (differentiated and GSCs) and compare their respective expression of BCL-xL. This could be achieved for instance via flow cytometry-based cell sorting according to the expression of a cell surface marker specifically expressed by GSCs. However, although commonly described as a marker of GSCs, CD133 is probably not the most appropriate stemness marker in our hands and accordingly to other studies since only the glycosylated cell-surface CD133 epitope is stem cellsspecific, while mRNA levels appear to be unrelated to stemness (26) (27) (28). However, OLIG2 and ITGA6 appear to be promising alternatives to faithfully characterize U87-MG GSCs, with ITGA6 being a possible plasma membrane marker that could allow cell sorting of GBM stem cells (29). In line with our findings in U87-MG cells, Trisciuglio and colleagues observed an increase in the amounts of tumorspheres number when BCL-xL is overexpressed in glioblastoma cells while the opposite was noticed when using BCL-xL depleted cells. They also noted that BCL-xL overexpression positively regulated the cancer stem cell phenotype in tumorspheres compared to control cells (22). These authors took their efforts further by demonstrating that exogenous BCL-xL controls several hallmarks of cancer aggressiveness such as migration, invasion and tumour cell plasticity, both in GBM and melanoma (22).
To improve our understanding of the transcriptional and/or translational regulation of BCL-xL in U87-MG neurospheres, an extensive study of various signalling pathway should be envisaged. Since BCL-xL can be regulated by a plethora of transcription factors ranging from STAT, Rel/NFkB or Ets, further insight into the mode of action of these regulators may unravel the specific regulation of BCL-xL in GBM tumorspheres (30).
A growing body of evidence suggests apoptosis-independent functions for several BCL2 family members, ranging from cell cycle, DNA damage response, metabolism or autophagy (31). In this article we focused on the anti-apoptotic function of BCL-xL and more specifically on revealing a therapeutic sensitivity to BH3 mimetics targeting BCL-xL. Indeed, the upregulation of BCL-xL in U-87 MG tumorspheres is linked to a higher sensitivity to BCL-xL specific BH3 mimetics. This finding corroborates a study by Liwak and colleagues that reported higher sensitivity of GBM cells to a combination of ABT-737 and doxorubicin, owing to increased levels of BCL-xL in a specific subset of glioblastoma cell lines (32). This is a highly promising finding, as it highlights means to kill GSCs, responsible for GBM tumours recurrence. However, these results should be interpreted with great caution as our findings are neither applicable to all patient-derived nor commercially available GBM cell line. BCL-xL can be targeted only if fresh tumour samples grown in tumorspheres permissive medium exhibit higher BCl-xL expression.
Nevertheless, the use of BH3 mimetics for the treatment of glioblastoma could alleviate other pro-oncogenic effects of BCL-xL. For instance, BCL-xL was shown to function as an epigenetic modifier and promote both epithelial to mesenchymal transition (EMT) and stemness in pancreatic, breast and lung cancer (33) (34). This is also consistent with a growing number of pre-clinical studies confirming the benefits of using BH3-mimetics when treating gliomas (35).          Apoptosis induction based on SYTOX Green incorporation following actinomycin D (1 µM) treatment.