Inhibition of ATR prevents macropinocytosis driven retraction of neurites and opposes invasion in GBM


 Glioblastoma (GBM) is the most common and aggressive type of primary brain tumour and remains incurable despite decades of research. GBM are characterised by highly infiltrative growth patterns that contribute to the profound cognitive and neurological symptoms experienced by patients, and to inevitable recurrence following treatment. Novel treatments that reduce infiltration of the healthy brain have potential to ameliorate clinical symptoms and improve survival. Here, we report a novel role of the Ataxia telangiectasia and Rad 3 related kinase (ATR) in supporting the invasive properties of GBM cells through the regulation of macropinocytosis-driven internalisation of integrin adhesion receptors. We demonstrate that inhibition of ATR opposes GBM migration in vitro, and correspondingly reduces infiltrative behaviour in orthotopic mouse models. These results indicate that ATR inhibition, in addition to its use as a radiosensitiser, may be effective in reducing GBM infiltration and its associated symptoms.


Abstract 28
Glioblastoma (GBM) is the most common and aggressive type of primary brain tumour and remains 29 incurable despite decades of research. GBM are characterised by highly infiltrative growth patterns 30 that contribute to the profound cognitive and neurological symptoms experienced by patients, and to 31 inevitable recurrence following treatment. Novel treatments that reduce infiltration of the healthy 32 brain have potential to ameliorate clinical symptoms and improve survival. Here, we report a novel 33 role of the Ataxia telangiectasia and Rad 3 related kinase (ATR) in supporting the invasive properties 34 of GBM cells through the regulation of macropinocytosis-driven internalisation of integrin adhesion 35 receptors. We demonstrate that inhibition of ATR opposes GBM migration in vitro, and 36 independent of its role in cell survival/DNA damage response. Interestingly, this role does not appear 125 to be linked to the RT-driven motility response that has previously been described [5][6][7]; although 126 irradiation of G7 cells further sensitised them to the anti-migratory effects of berzosertib (IC50 0.26µM 127 vs 96nM), this sensitisation was not observed in R15 or E2 cells, suggesting that ATR inhibition 128 effectively inhibits migration in both the presence and absence of radiation. 129 Berzosertib is a well characterised and highly specific inhibitor of ATR, and its efficacy as an anti-130 invasive compound at low concentrations suggests that our observations reflect on-target effects of 131 the inhibitor. However, to confirm that the anti-invasive effects of ATR inhibition were not due to an 132 off-target property of berzosertib, we measured migration speed of G7 cells by sub-confluent 133 migration assay and single cell tracking following exposure to siRNAs targeting ATR (Fig 2C). The results 134 demonstrated a significant decrease in migration speed following ATR knockdown, providing 135 confirmation of an on-target effect of ATR inhibition on motility. 136

Inhibition of ATR opposes GBM cell infiltration in vivo 137
To confirm the biological relevance of our in vitro migration studies, and to further test the clinical 138 potential of using ATR inhibitors as an anti-invasive strategy, we conducted a series of in vivo studies. 139 Initial pharmacokinetic (PK) studies demonstrated tumour penetration of berzosertib in two different 140 intracranial GBM models (U87MG and the more infiltrative G7) at levels sufficient to inhibit the kinase 141 and induce phenotypic effects (Fig 3A (i) and (ii); Fig S3). Importantly, delivery was enhanced in the 142 tumour compared to the contralateral brain, further enhancing tumour specificity. 143 Next, we utilised a GFP-labelled S24 GBM primary line in combination with an intracranial window 144 mouse model of GBM that allows intravital imaging of tumours in situ over time (Fig 3B (i)). Intracranial 145 windows were introduced into the skulls of immuno-compromised mice, followed by intracranial 146 injection of GFP-S24 cells. After 3-4 weeks, mice were treated twice with vehicle, followed by 147 multiphoton imaging of tumour cells within the brain. Mice were then treated twice with berzosertib 148 followed by repeat multiphoton imaging (Fig 3B (ii)). Measurement of TM length under control and 149 treatment conditions revealed a significant decrease in length, consistent with the destabilisation and 150 breakage of neurites observed in vitro, and indicative of reduced invasive potential (Fig 3B (iii)). 151 To test this hypothesis further, mice bearing U87MG intracranial tumours were treated with RT in 152 combination with vehicle or berzosertib. Mice were culled at clinical endpoint (symptomatic) and the 153 brains fixed, sectioned and stained for Ki67 to label proliferating GBM cells within the non-proliferative 154 brain tissue. Modest infiltration of tumour cells was observed beyond the tumour margins, as 155 expected following irradiation [5]. Although there was no survival benefit to berzosertib treatment 156 (data not shown), a significant decrease in number of infiltrating cells was observed in the berzosertib 157 cohort (Fig 3C (i) and (ii)). Mice were culled with a short time frame (5 days) and no correlation 158 between tumour size and infiltration was found (Fig 3B (iii)). This confirms that the effect of ATR 159 inhibition on tumour cell infiltration is due to loss of invasive potential, and not a function of tumour 160 burden or growth time. 161

Inhibition of ATR increases cytoplasmic vacuoles 162
In addition to the described defect in cell retraction/de-adhesion/cell migration, treatment with 163 berzosertib or siRNA mediated knockdown of ATR increased the appearance of enlarged vacuolar 164 structures within the cytoplasm of GBM cells (Fig 4A(i) and (ii)). We conducted a series of experiments 165 to elucidate the origin of these structures. Autophagy has been demonstrated to modulate cell 166 migration and integrin membrane recycling [27]. To investigate whether a dysregulation of autophagy 167 was underpinning our observations, we undertook autophagic flux assays as previously described [28]. 168 Western blot analysis for processed LC3B from cells treated with/without VE822 and chloroquine 169 indicated no change in the levels of LC3B II, indicating that the observed vacuoles were unlikely to be 170 derived from a block in autophagy ( Fig S4A). Indeed, electron microscopy revealed the vacuoles to be 171 universally single membrane bound, inconsistent with double membraned autophagosomes ( Fig S4B). 172 This data clearly demonstrates that autophagy is not blocked upon ATR inhibition and does not 173 contribute to the observed ATR dependent phenotype. 174 Next, we investigated whether these structures could be macropinosomes, which are 175 characteristically large (0.2 -5µm) and bound by a single membrane. G7 cells were incubated with a 176 Texas red labelled 70kDa dextran (red) in the presence of berzosertib. Dextrans of this size are 177 predominantly internalised via macropinocytosis, and thus mark any cellular structure that has 178 resulted from this endocytic process. The cells were fixed, stained and subject to confocal imaging (Fig 179 4B (i)). Indeed, a large proportion of the vacuoles contained dextran, identifying them as 180 macropinosomes. Furthermore, high-throughput confocal imaging and automated analysis of cells 181 treated concurrently with fluorescein labelled 70kDa dextran (green) and DMSO or berzosertib 182 revealed a significant increase in dextran positive macropinosomes in cells treated with berzosertib 183 (Fig 4B (ii)). This accumulation of dextran within the cell body suggests either an increase in 184 macropinocytosis or reduced processing of internalised macropinosomes. Strikingly, 185 immunofluorescence staining revealed a close association between ATR (green) and dextran positive 186 macropinosomes (red), suggesting a direct mechanistic link (Fig. 4C). 187

processing of internalised macropinosomes
To test whether the increase in cytoplasmic macropinosomes was due to an increase in active 190 macropinocytosis or a block in processing we conducted a pair of timed experiments. In the first 191 instance, cells were treated with either DMSO, berzosertib or an inhibitor of macropinocytosis (5-(N-192 Ethyl-N-isopropyl) amiloride; EIPA) for 15mins or 2 hours prior to the addition of 70kDa dextran 193 (green). Cells were allowed to internalise dextran for 30 mins before fixing, staining and high-194 throughput imaging and automated analysis (Fig 5A (i)). Rather than the expected increase in 195 macropinocytosis, we observed a significant dose-dependent decrease in dextran positive 196 macropinosomes in berzosertib pre-treated cells (Fig 5A (ii)), indicating a block in active 197 macropinocytosis at the plasma membrane and mirroring the EIPA effect of blocking macropinocytosis. 198 With inhibition of active macropinocytosis being observed upon pre-treatment with berzosertib ( Fig  199   5A (ii)), but a clear increase in dextran positive macropinosomes apparent when VE822 and dextran 200 are given concomitantly, a block in the processing of cytoplasmic macropinosomes must also be 201 occurring. To confirm this, the experiment was repeated, allowing cells to internalise dextran for 30 202 mins in the absence of compound, prior to its removal and addition of DMSO, berzosertib or EIPA for 203 15 mins ( Fig 5B). The results show that turnover of pre-internalised dextran is reduced in a dose 204 dependent manner in cells treated with berzosertib, strongly suggesting that a block in processing is 205 the cause of intracellular dextran accumulation. These effects differed from those of EIPA treatment,206 suggesting that the processing block is specific to ATR inhibition. 207 While macropinocytosis has been well characterised in vitro, in vivo studies have proven challenging. 208 The intracranial window model presents an opportunity to bridge this gap. In order to confirm that 209 our observations were not an in vitro artefact, we utilised the window model to look for dextran 210 accumulation in GFP-S24 cells in vivo. Using advanced imaging and processing with Imaris software 211 we were able to confidently detect internalised fluorescent 10kDa dextran within GFP-S24 cells within 212 the mouse brain ( Fig 5C (i)). 10kDa dextran was used in this instance to allow multiple subcutaneous 213 injections, and for efficient delivery across the blood-brain barrier. Unlike 70kDa molecules, 10kDa 214 dextran can be internalised by other endocytic processes, however a significant amount is internalised 215 via macropinocytosis [29]. We confirmed via in vitro uptake assays that its endocytosis was similarly 216 impacted by ATR inhibition and thus suitable for our in vivo study (data not shown). We then used this 217 technique to test whether berzosertib caused an accumulation of dextran in vivo by combining 218 vehicle/berzosertib with dextran labelled with two different fluorophores. Mice were initially treated 219 on two consecutive days with vehicle alongside cascade blue-dextran (CB-dex), before multiphoton 220 imaging. They were then treated twice with berzosertib alongside Texas red-dextran (TR-dex), before 221 repeat imaging (Fig 5C (ii)). By changing the colour of the dextran label, we ensured that only dextran 222 internalisation/processing that may have been affected by berzosertib treatment was detected in the second imaging session. To control for any variance in fluorophore detection affecting results, the 224 order of administration of dextran colours was switched in one mouse (TR-dex/veh then CB-dex/ 225 berzosertib),and an additional mouse was treated with berzosertib and CB-dex alone. Subsequent 226 colocalization analysis revealed that berzosertib caused an accumulation of dextran in GFP-S24 cells, 227 confirming the biological relevance of our in vitro findings (Fig 5C (iii). 228 Together, these data indicate that blocks in both active macropinocytosis and macropinosome 229 processing occur in tandem after ATR inhibition. These results are consistent with time-lapse data 230 from cells treated with berzosertib that show the number of macropinosomes remaining constant 231 over time (Supp Video 2). These observations are reminiscent of previous studies showing that ATR 232 inhibition causes a block in both endocytosis and processing of excitatory vesicles at neuronal 233 synapses [15], and suggest a similar role in endocytosis and vesicle trafficking in GBM cells. 234

de-adhesion, retraction and cell migration. 236
The parallel observations of a de-adhesion/retraction defect caused by an inability of TMs to release 237 from the extra cellular matrix (ECM), plus a block in macropinocytosis are intriguing, raise the 238 possibility of a mechanistic connection between the two phenomena. To test this theory, we treated 239 sub-confluent G7 and R15 GBM cells with 10µM EIPA and performed time-lapse microscopy followed 240 by single cell tracking to measure motility speed. Cells treated with EIPA showed a striking 241 morphological resemblance to cells treated with berzosertib ( Fig 1B; Fig 6A), with a similar increase in 242 vacuoles, although these turned over more rapidly than in cells treated with berzosertib, reflecting 243 the observations made in Fig  To investigate this hypothesis, we performed integrin internalisation assays for α5, α3 and α6 integrins. 257 Transferrin R (TrnR) is passively, as opposed to actively, internalised and was used as a control. The 258 data in Fig 7B (i)-(iv) clearly indicate that internalisation of all three integrins is reduced upon 259 treatment with Berzosertib, providing further evidence of a role of ATR in regulating integrin 260 internalisation. The reduction is modest, suggesting that internalization may be occurring in a discrete 261 location in the cell. 262 To investigate this in greater detail, we over expressed GFP-labelled α5 integrin in G7 cells and used 263 super-resolution, time-lapse microscopy to look at 70kDa dextran uptake at growth cone like 264 structures at the ends of neurites (Fig 8). In control cells we obtained clear evidence that 265 macropinocytosis was occurring at areas of membrane retraction, and that many of the resulting 266 macropinosomes were positive for both GFP-α5 integrin and Texas red labelled dextran (Fig 8A (i)). In 267 addition to a subset of larger macropinosomes that were largely static, we observed a population of 268 highly mobile smaller macropinosomes consistent with observations from studies in neuronal crest 269 cells [26]. Some large endosomal structures were dextran negative, but their size was indicative of 270 macropinosomes which had not internalised dextran. 271 Importantly, (Fig 8B) we observed that berzosertib downregulated de novo macropinocytosis, with all 272 pre-existing, internalised macropinosomes becoming static, regardless of size. This suggests that the 273 processing defect that causes accumulation of intracellular macropinosomes may be due to defective 274 trafficking of the structures through the cytoplasm, either to be recycled to the membrane or shuttled 275 to the lysosomes for degradation. 276 These data provide strong evidence to support our hypothesis that ATR is required for effective 277 internalisation of integrins and subsequent de-adhesion and retraction of TMs to enable effective 278 migration in vitro and in vivo. In this paper we present data that describe a novel, non-canonical role for ATR kinase in facilitating 282 tumour cell motility and invasion via macropinocytic internalisation of integrins and subsequent 283 growth cone de-adhesion. Importantly, we also present evidence for the potential clinical use of ATR 284 inhibitors as anti-invasive therapy. The widespread malignant infiltration observed in GBM is accepted 285 to be a major determinant of the poor clinical outcomes associated with this disease, however a 286 detailed understanding of the key mechanisms involved remains elusive. Abrogation of tumour 287 infiltration is a highly desirable clinical target which could improve symptom control and survival in 288 this tumour of unmet need. However, no clinically useful anti-invasive treatments currently exist. 289 Inhibiting invasion has been shown to correlate with increased mouse survival in previous studies 290 using different GBM models [5,6]. The absence of a survival benefit in our U87MG experiment may 291 be attributed to the rapidity of growth of the model, as well its comparatively restricted pattern of 292 invasion. A rapid increase in tumour burden and associated 'mass effect', is the major cause of clinical 293 symptoms in this model, with less dependence on infiltration of brain stem and other vital structures 294 that is seen in other GBM models. Nevertheless, our data provide strong evidence for a role of ATR in 295 facilitating GBM invasion within the brain and thus identify a potential therapeutic role of ATR 296 inhibition beyond its established function as a radiosensitiser. 297 Macropinocytosis is a key process in the bulk trafficking of integrins and plasma membrane to allow 298 active migration and is a prominent feature in GBM [21,22,31,32]. The importance of 299 macropinocytosis in growth cone de-adhesion and neurite retraction/redirection has previously been 300 described in neuronal cells [23,24]. Therefore, our observation that this process occurs in GBM cells 301 is perhaps unsurprising, considering their shared lineage, morphology and function. It has been clearly 302 demonstrated in the literature that the TMs of GBM cells share multiple properties with the neurite 303 projection of neuronal cells, including the presence of growth cones, the ability to probe and respond 304 to the local environment through dynamic movement, and even the ability to form functional 305 synapses [16][17][18]. GBM cells can protrude extraordinarily long TMs into the normal brain, pulling the 306 rest of the cell body behind them as they invade. One can postulate that these leading TMs are sensing 307 the new environment, before committing to invading fully. In this situation, the ability to de-adhere 308 rapidly to allow retraction or redirection in response to positive or negative environmental cues would 309 be advantageous. 310 The canonical functions of ATR in replication stress and DNA damage signalling are well characterised. 311 However, evidence for biological functions beyond the nucleus are slowly emerging [12,13,15]. We 312 demonstrate that ATR is required to mediate internalisation of integrins to allow active migration in 313 GBM cells using clinical samples and state of the art in vitro and in vivo techniques. On first 314 consideration, this observation may be surprising. However, the ability of ATR to trigger cell cycle 315 arrest and modulate kinesins and cytoskeletal dynamics provide a compelling mechanistic link [33] [34] 316 [12]. In addition to its role in bulk internalisation of integrins, macropinocytosis has also been 317 implicated in chemotaxic response, and nutrient sensing [19,20]. Placing ATR at the forefront of 318 invading TMs puts it in an ideal position to link nutrient and chemokine sensing via macropinocytosis 319 to the cytoskeleton network and cell cycle. In addition, the freezing effect of ATR inhibition on 320 macropinosomes at growth cones that we observed via super resolution microscopy indicates that 321 ATR can modulate kinesins and cytoskeletal dynamics in GBM cells beyond the previously described 322 context of the cell cycle. 323 ATR inhibitors are currently in development in early phase clinical trials and have been the subject of 324 intense interest due to their potentiation of DNA damage in combination with conventional cytotoxic 325 agents and radiation in a range of tumour sites. In the context of neuro-oncology, the ATR/replication 326 stress response axis is a particularly appealing target. Treatment resistance in GBM appears to be 327 driven by a subpopulation of glioma stem cells with high levels of DNA replication stress and ATR 328 activation and we and others have shown potent radiosensitisation by ATR inhibitors in vitro. Our 329 novel finding of an anti-invasive effect of ATR inhibition suggests that the clinical benefits of this 330 approach could be far reaching and extend beyond radiosensitisation. It could be envisaged that ATRi 331 could be employed as a neoadjuvant strategy to promote regression of invasive tumour edge prior to 332 surgery and irradiation as well as in an adjuvant or 'maintenance' manner to reduce or prevent 333 malignant infiltration into vital neural structures. A prolonged period of neo adjuvant anti-invasive 334 therapy with ATRi may be a particularly attractive strategy in managing patients with lower grade 335 gliomas, who often undergo pre-planned 'elective' tumour resection. 336 In summary we present a novel role for ATR in regulating GBM invasion which has significant relevance 337 for clinical translation. We predict that this previously unappreciated function of ATR will lead to 338 unexpected clinical benefits from therapeutic strategies combining ATR inhibition with radiotherapy. 339 340 341

Transmission Electron microscopy 343
Samples were fixed in 2.5% glutaraldehyde, 4% paraformaldehyde, in 0.1M cacodylate buffer, pH 7.2; 344 washed in 0.1M cacodylate buffer, pH 7.2 and post-fixed in 1% OsO4 for 1 hour. After several washes 345 in the same buffer, the samples were en bloc stained with 0.5% uranyl acetate in water for 30 minutes. 346 Afterwards, samples were washed with water, dehydrated in ascending acetone series and resin 347 embedded. Ultrathin sections (50nm thick) were collected and imaged on a JEOL 1200 Transmission 348 electron microscope (JEOL, Japan) operating at 80kV. 349 In Silico Analysis of ATR expression from public datasets 350 Expression of ATR and correlation to patient survival (Kaplan -Meier survival analysis) was performed 351 using the publically available RNA sequencing dataset from the Chinese Glioma Genome Atlas (CGGA) 352 project via the gliovis platform: http://gliovis.bioinfo.cnio.es/. 353

Derivation and maintenance of primary GBM cell lines 354
Primary GBM cell lines E2, G7, and R15 were derived from resected tumors and maintained as 355 In vitro irradiation was performed using an Xstrahl RX225 radiation cabinet (195 kV X-rays, dose rate 365

In vitro immunofluorescence 367
For confocal imaging sub-confluent GBM cells were plated on coverslips coated with Matrigel and 368 incubated at 37 o C for 24 hours. For high-throughput imaging, 1 x 10 4 cells were plated on 96 well black 369 sided plates (Perkin Elmer) precoated with Matrigel. Cells were incubated with anti-ATR (23HCLC 370 ThermoFisher) or anti-integrin a6 (CD49F 55734 BD Bioscience) antibodies overnight at 4°C followed 371 by incubation with secondary conjugated antibodies, DAPI and HCS Cell Mask Deep Red Stain 372 (ThermoFisher H32721) or Texas red-X Phalloidin (ThermoFisher T7471). For confocal imaging nuclei 373 were counterstained with Vectashield mount containing DAPI. Images were acquired using either a 374 Zeiss LSM 780 or 880 confocal microscope and analyzed using Zen 2012 (Zeiss). For High-throughput 375 analysis, images were captured using the Opera Phenix High-content Screening system and analysed 376 using Columbus Image Analysis software (Perkin Elmer). 377 Super resolution microscopy 378 0.5 x 10 5 G7 cells were plated on glass bottomed 35 mm dishes pre coated with 5µg/ml fibronectin 379 (ThermoFisher) and allowed to establish for 24hours. 1µg of plasmid containing α5-GFP was then 380 transiently transfected into the cells using SuperFect transcription reagent (Qiagen). After 24 hours, 381 0.1 mg/ml lysine fixable, 70kDa Texas-red dextran was added, and cells allowed to internalise for 30 mins before addition of either DMSO or Berzosertib. Timelapse, super resolution imaging was 383 performed using Zeiss 880 Airyscan (SR mode) with a Plan-Apochromat 40x/1.3 Oil objective and 384 images analysed using Imaris software. Cell were maintained at 37°C, 5% CO2 in an incubation 385 chamber for the duration of imaging. 386

Cell viability 387
Cell viability was carried out using CellTiter-Glo according to the manufacturer's protocol (Promega). experiments were housed separately following surgery to prevent damage to the implanted ring. 413

Intracranial window experiments 414
Male NMRI-Foxn1 nu mice (n=5) were used for these experiment using an adaptation of a surgical 415 procedure described by Winkler et al, 2004 [37]. Custom made titanium rings and frames for surgery 416 and imaging were supplied by VetTech. 2-3 weeks after window implantation, mice were injected 417 with 30,000 S24-GFP cells and tumours allowed to develop for a further 2-3 weeks. 418 Mice were given either vehicle (10% vitamin E) or Berzotersib (50mg/kg) by oral gavage according to 419 the schedules described in the results section. 200µl of 10mg/ml 10kDa dextran was administered 420 subcutaneously. Imaging was undertaken using a Zeiss880 multiphoton microscope using W Plan-421 Apochromat 20x/1.0 objective and subsequently analysed using Imaris software. Z-stacks were taken 422 to a depth of 60 -100 µm, at 1µm intervals. GFP and TR-dex were imaged with 890nm laser, CB dex 423 with 800nm. Multiple fields were taken each imaging session, and included individually, or as means 424 as shown by the overlayed plots in Fig 3B and 5B. Statistical analysis = paired student T Test. 425

Intracranial tumour experiments 426
Female CD1 nude mice were orthotopically injected with 1 × 10 5 U87MG or G7 cells as previously 427 described [5]. For PK studies, tumours were allowed to establish for 3 weeks (U87MG) or 10 weeks 428 (G7) before dosing with either vehicle (10% vitamin E) or 60mg/kg Berzosertib and mice culled at the 429 indicated time points. Tumors were sub-dissected and fresh-frozen specimens of tumour and 430 contralateral hemispheres sent for PK analysis (Vertex). For the phenotypic study, tumours were 431 allowed to establish for 12 days before treatment. Brain irradiation was performed on an XStrahl Small 432 Animal Radiation Research Platform (SARRP) as previously described [5]. Mice received 5 x 3Gy on 433 alternate days and were dosed via oral gavage with either vitamin E or 60mg/kg Berzosertib daily for 434 10 days. Formalin-fixed, paraffin-embedded sections were stained for Ki67, and scanned using a 435 Hamamatsu Nanozoomer Slide scanning machine with Leica SlidePath Slide imaging software.

Figure 1: ATR is found in the cytoplasm of GBM cells and correlates with invasive potential A) (i)
ATR is found in both the nucleus and cytoplasm of GBM cells, including distribution along neuritelike projections (red arrows). (ii) Sub-lethal radiation dose causes an increase in cytoplasmic ATR. G7 cells were irradiated with 2Gy and fixed after 6 hours and stained for actin (yellow) and ATR (green) followed by high-throughput imaging. Data from 3 biological repeats, 2000-4000 cells imaged per condition each repeat. Bars are mean ±-SEM. High-throughput imaging at 40x (A) and 20x (D) Blue-DAPI; Green-ATR; Yellow-Actin. B) ATR expression correlates with both (i) glioma grade and poorer patient survival (ii). C) Increased pATR expression can be found in the invasive tumour margin versus the tumour core in patient samples. (i) Core and margin samples were taken from 3 different patients, stained for pATR and scored by a neuropathologist for positive cells and the fold change between core and margin plotted. (ii) ATR expression is increased in primary Ox5 GBM cells derived from the tumour margin compared to matched cells from the core and correlates with increased neurite number and length. Western blot quantified from 4 independent biological repeats. Statistical analysis using student's T Test. * p<0.01, *** p<0.001