Decrease of Nibrin expression in chronic hypoxia is associated with hypoxia-induced chemoresistance in medulloblastoma cells

Solid tumours are less oxygenated than normal tissues. This is called tumour hypoxia and leads to resistance to radiotherapy and chemotherapy. The molecular mechanisms underlying such resistance have been investigated in a range of tumour types, including the adult brain tumours glioblastoma, yet little is known for paediatric brain tumours. Medulloblastoma (MB) is the most common malignant brain tumour in children. Here we used a common MB cell line (D283-MED), to investigate the mechanisms of chemo and radio-resistance in MB, comparing to another MB cell line (MEB-Med8A) and to a widely used glioblastoma cell line (U87MG). In D283-MED and U87MG, chronic hypoxia (5 days), but not acute hypoxia (24 h) induced resistance to etoposide and X-ray irradiation. This acquired resistance upon chronic hypoxia was much less pronounced in MEB-Med8A cells. Using a transcriptomic approach in D283-MED cells, we found a large transcriptional remodelling upon long term hypoxia, in particular the expression of a number of genes involved in detection and repair of double strand breaks (DSB) was altered. The levels of Nibrin (NBN) and MRE11, members of the MRN complex (MRE11/Rad50/NBN) responsible for DSB recognition, were significantly down-regulated. This was associated with a reduction of Ataxia Telangiectasia Mutated (ATM) activation by etoposide, indicating a profound dampening of the DNA damage signalling in hypoxic conditions. As a consequence, p53 activation by etoposide was reduced, and cell survival enhanced. Whilst U87MG shared the same dampened p53 activity, upon chemotherapeutic drug treatment in chronic hypoxic conditions, these cells used a different mechanism, independent of the DNA damage pathway. Together our results demonstrate a new mechanism explaining hypoxia-induced resistance involving the alteration of the response to DSB, but also highlight the cell type to cell type diversity and the necessity to take into account the differing tumour genetic make-up when considering re-sensitisation therapeutic protocols.


Introduction
Medulloblastoma (MB) is a malignant embryonal brain tumour originating from neural stem cells or granule-cell progenitors of the cerebellum, due to a deregulation of signalling pathways involved in neuronal development such as Wnt or Sonic Hedgehog (SHH) 36 . It is one of the most common brain tumours in children accounting for 15-20% of all paediatric CNS tumours. MB is a heterogeneous cancer with distinct genetic variants, classified into 4 principal subgroups; Wnt, SSH, Group 3 and Group 4, based on their transcriptome 33,63 reviewed in 57 . More recent classification using genome wide DNA methylation and gene expression data resulted in the division of these 4 groups into 12 sub-groups 9 . Treatment of MB generally involves surgery followed by a combination of radiotherapy and cytotoxic chemotherapy. In very young children, surgery is followed by chemotherapy alone due to the severe effects of radiation therapy on the developing brain 48 . Recent studies have explored the use of proton beam therapy for MB, which triggers fewer detrimental side effects, as a result of reduction in irradiation of healthy brain tissue 68 .
Despite a marked improvement in the 5-year survival rate for MB patients, 40% succumb to the disease, demonstrating the limitations of the current therapies. Group 3 MB have the worst overall survival (~ 50%) even after extensive treatment 15,43,63 . One reason for treatment failure is the resistance to chemo-or radiotherapeutic interventions. In both MB patients and MB cell lines such resistance has been observed and attributed to mutations in specific signalling pathways in the case of targeted therapies 40 , or to mutations in the p53 signalling pathway 19, 23, 29, 39,62 . For example, in the SHH group p53 mutations correlate with poor treatment outcome 70 . In addition to the intrinsic mutations in cancer cells conferring resistance, extrinsic factors such as tumour hypoxia have been implicated in chemo-and radioresistance in a range of tumour types 8,31,47,61 .
Tumour hypoxia is generated by irregular and tortuous vasculature formed within solid tumours, hence causing a poor delivery of oxygen and nutrients to cells. Hypoxia is associated with malignancy and tumour aggressiveness, by increasing tumour cell proliferation, metastasis and treatment resistance 27,28,55 . Hypoxic markers are, therefore, commonly associated with poor prognosis in many tumour types including breast cancer, glioblastoma and neuroblastoma 14,24,32,65 . Interestingly, carbonic anhydrase IX (CaIX), a robust marker of hypoxic tumours 14 , is expressed in 23% of all MBs and is associated with poor prognosis 42 , suggesting that tumour hypoxia impacts on MB progression and/or management. The critical impact of tumour hypoxia on drug resistance and malignancy has been reported for a large number of cancers. However, very little is known about the effect of hypoxia on MB beyond, for example, a recent study showing that the hypoxia inducible factor (HIF-1) can influence the ability of MB cells to proliferate 18 .
Here we demonstrate that chronic hypoxia (>4 days) triggers resistance in a common MB cell line, D283-MED to cell death induced by etoposide or X-ray irradiation. The effect of chronic hypoxia on chemo-and radio-sensitivity was comparable to that observed in a classic glioblastoma cell line (U87MG). To

Long term hypoxia induces etoposide and X-ray irradiation resistance in MB and glioblastoma cells.
Hypoxia has previously been associated with increased drug resistance in a number of tumours including glioblastoma, hepatocellular carcinoma and breast tumours 1, 17, 61 .
Medulloblastoma exhibit poor response to chemo-and radiotherapeutic interventions 25, 58 , yet the possibility that this could be explained by tumour hypoxia is unexplored. We used two MB  Figure 1A). In contrast, 1 day of hypoxic preconditioning did not induce any significant resistance, suggesting that the duration of hypoxic exposure is critical to alter the cell response to drugs. The hypoxia-induced resistance to etoposide was also observed in MEB-Med8A cells, although to a lesser extent and with high variability thereby without reaching statistical significance ( Figure 1B; p=0.57). The U87MG glioblastoma cells displayed similar results to the D283-MED cell line ( Figure 1C), with 5 day preconditioning resulting in significantly higher cell viability compared to the normoxic controls. Results for the D283-MED cell line were reproduced with the cell viability assay ViaCount TM , to ensure that the MTS results were not biased by altered mitochondrial activity in hypoxia, and similar data were obtained ( Figure S1A). Acquired resistance was also investigated upon X-ray irradiation.
In this case, the cells were pre-incubated in 1% or 21% O 2 for 5 days and the cells were irradiated in normoxic conditions to prevent loss of the oxygen enhancement effect during irradiation treatment 12 . With this protocol we ensured that only the effects of hypoxia preconditioning were assessed. Similarly to etoposide treatment, yet to a lesser extent, resistance to X-ray irradiation upon hypoxic preconditioning was observed for the D283-MED and the U87MG cell lines ( Figure 1D). The MEB-Med8A cells showed an inverse trend ( Figure 1D). This counterintuitive result may be due to the fact that this cell line appeared stressed by the cooling procedure prior to irradiation, hence generating variable results.
We further assessed the ability of etoposide to induce cell cycle arrest in D283-MED cells preconditioned or not in 1% O 2 . Hypoxia preconditioning on its own did not induce any cell cycle alteration ( Figure S1B and C). This important control indicates that the reduced effect of etoposide in hypoxic preconditioned cells is not due to changes in cell proliferation agreeing with previous observations in glioblastoma 50  the sensitivity to X-ray irradiation. This resistance is not due to alteration of the cell cycle by hypoxia ( Figure S1) or to the absence of oxygen during the irradiation process, as irradiation was performed in atmospheric conditions. It, therefore, points to global changes in cell signalling, resulting in lack of sensitivity to the DNA damaging protocols used.

Chronic hypoxic exposure induces large transcriptional remodelling
To investigate the molecular mechanisms driving the observed hypoxia-induced cell death resistance in D283-MED, we used micro-array gene expression analysis to assess the global transcriptomic modifications triggered by long-term hypoxia. Hypoxia-inducible factors (HIF) are the main transcription factor family directly regulating gene expression in hypoxia. HIF-1 alpha (HIF-1α) is the primary isoform governing the expression of genes whose levels are directly dependent on oxygen availability. Hypoxia-induced gene expression was measured at a time point that correlates with the first peak in HIF-1α stabilisation, and at two later time points subsequent to additional peaks of HIF accumulation ( Figure 2A). Considering all three time points together, we found that 6,124 transcripts were significantly up-or down-regulated in hypoxia ( Figure 2B). This corresponds to 4,303 significantly regulated genes (and 23,611 non-regulated genes). Significantly regulated transcripts were further analysed globally by clustergram ( Figure 2C), which showed a large number of genes regulated at late time points We initially performed a biased analysis and specifically examined the expression of several multi-drug resistance genes. These genes were the obvious candidates potentially responsible for hypoxia-induced treatment resistance and had previously been described to have a role in hypoxia-induced drug resistance in other cellular models including glioblastoma 7 13 . However, none of the well described multi-drug resistance (MDR) genes abcb1 (mdr1), abcc1 (mrp1) and abcg1 were up-regulated in hypoxia ( Figure 3A).
For a non-biased analysis using the k-means clustering method, the large number of regulated transcripts were further grouped into 16 smaller clusters based on their expression profile pattern, to find potential correlation between groups of genes with similar expression dynamics ( Figure S2A). From this we identified that genes involved in the response to double strand breaks were strongly regulated at later hypoxic time points. The effectiveness of DNA damage response mechanisms can play a key role in response to chemo-and radiotherapy.
Therefore, we generated heat map clustergrams of the HRR and NHEJ pathways, which are primarily responsible for the repair of DSB, the most lethal form of DNA damage ( Figure 3B and C). Strong regulation was observed for a number of genes involved in both NHEJ ( Figure   3B) and HRR ( Figure 3C). We further focused on HRR pathways as key genes involved in signal transduction for HRR were down-regulated strongly by hypoxia. Down-regulated transcripts include brca2 and two members of the MRN complex, mre11A (MRE11) and Nibrin (NBN). NBN plays a key role in ATM activation and signal amplification of the DNA damage response 10 . Therefore, the hypoxia mediated down-regulation of key components of the HRR pathway could be responsible of the altered cellular response to etoposide.

DNA damage recognition and signal transduction is dampened in chronic hypoxia
Sensing of DNA damage is a crucial step in the initiation of double strand break (DSB) repair by the HRR machinery, and its alterations is likely to impact further downstream signalling of the repair pathway. The strong reduction in NBN transcription (70% decrease) ( Figure 4A) observed in the micro-array was confirmed using qPCR, which equally showed a decrease of NBN mRNA by 70% at 96 hours ( Figure 4B). Similarly, NBN protein levels were reduced by ~50% after 96 hours of hypoxic exposure ( Figure 4C). To further understand the effect of chronic hypoxia on each member of the MRN complex, mRNA levels of NBN, MRE11 and RAD50 were determined by qPCR upon 5 days of hypoxic incubation. Hypoxia had little effect on RAD50 mRNA levels, yet there was significant down-regulation of MRE11 (~53%) comparable to NBN ( Figure 4D), which was further confirmed at the protein level ( Figure 4E). Surprisingly, such decreases in NBN and MRE11 expression was not observed in the U87MG cells ( Figure S3A). stabilisation with three times lower levels of phosphorylation observed when normalised to the total amount of p53 ( Figure 7D). Similar results were obtained for the U87MG glioblastoma 9 cell line ( Figure S4A, B), suggesting that, whilst recruiting different mechanisms of resistance, there is a convergence on the activation of p53 protein.
In conclusion, we have shown that the chemo-and radio-resistance triggered by a chronic hypoxic exposure can be explained by a broad gene expression remodelling of components of the DNA DSB sensing machinery, subsequently affecting the activation of DNA repair and pro-apoptotic mechanisms. This suggests that reactivation of the signalling pathway could resensitise cells to treatment. However, the way cells adapt to the low oxygen levels is diverse with individual cell line or cell-type specific responses. In this study for example, whilst the U87MG cells exposed to chronic hypoxia, showed a similar resistance to treatment and a similar dampened p53 activation to the D283-MED cells, the mechanism of alteration of the DNA DSB sensing and repair pathways were different. Such diversity in the cellular adaptation, makes it difficult to define global clinical strategies for hypoxic cell sensitisation.

Discussion
Tumour hypoxia has been extensively correlated with acquired resistance to cytotoxic drugs and irradiation, yet the underlying mechanisms are variable between tumour types. The impact of oxygen deprivation during treatment as opposed to the cellular experience prior to treatment has not been untangled so far. Whilst the oxygen enhancement effect, whereby oxygen directly reacts with broken end of DNA preventing their repair, has been implicated in reduced effectiveness of radiotherapy in hypoxia 12, 20 , we here demonstrate that the cellular oxygenation history has a broad impact on cell adaptation and sensitivity to DNA damaging agents. To observe resistance to drug and irradiation a chronic hypoxic preconditioning of several days is necessary, in both medulloblastoma and glioblastoma cells. Although, the prolonged hypoxia exposure triggers different adaptive mechanisms between cell lines, they ultimately converge on the inability to activate the pro-apoptotic p53 protein, thereby mediating cell death and cell-cycle arrest resistance.

Heterogeneity of the cellular response to hypoxia between cell types.
We observed a strong hypoxia-induced chemo-and radio-resistance in the D283-MED cell line (Figure 1), however the MEB-Med8A cells did not show the same extent of resistance.
This differing response might be due to the differences in the molecular signature of these two The complex and varied response to hypoxia is also apparent when comparing in vitro cell line models to in vivo organisms and pre-clinical models. In this case, our results obtained in vitro mimicked the observations made in hypoxic tumours in vivo, in terms of resistance to cancer treatments. Moreover, hypoxia related changes to DNA repair proteins have been translated to an in vivo setting. For example, active phospho-ATM is co-localised with the hypoxic marker CAIX in tumour xenografts 45 and RAD51 down-regulation was observed in cervical and prostate cancer xenografts 5 as well as in a glioma model 37 .
In a clinical perspective, the development of HIF inhibitors targeting either HIF mRNA transcription, translation, or HIF stabilisation 38   Coverslips were mounted onto glass slides using Dako mounting medium. Images were obtained using Leica DM2500 microscope (Leica) and quantified using AQM Advance 6 software (Kinetic Imaging Ltd, UK).
Western Blotting: Conducted as described in 50  Comet assay: D283-MED cells were cultures on a 12 well plate directly before drug treatments. Cells were collected by centrifugation and re-suspended in 1 ml ice-cold PBS.
The samples were diluted with pre-warmed agarose and loaded on the slide provided by the OxiSelectTM comet assay kit. Subsequently, the Comet assay procedures were performed as per manufacturer's protocol, followed by alkaline electrophoresis. Vista dye was added immediately after electrophoresis for DNA staining. Slides were then imaged using Leica