Selective toxicity of functionalised graphene oxide to patients-derived glioblastoma stem cells and minimal toxicity to non-cancerous brain tissue cells

Glioblastoma (GBMs) is an aggressive type of brain tumour, driven by immature neural stem cell-like cells that promote tumour growth and underlie resistance to conventional therapy. The GBM stem cells (GSCs) can exist in quiescent or dormant states and infiltrate widely into surrounding brain tissues, currently incurable with only around one-year median survival. Innovative therapeutic strategies for GBMs are urgently needed. Here we explore functionalized graphene oxide (GO) to assess their value as delivery vehicles for GBM therapeutics. Interactions and cellular responses were assessed in vitro using both classic cell lines and patient derived GSCs. Association between the functionalised GO and established GBM cell lines (serum grown ‘non-stem’ cells) was strong and resulted in decreased cell viability, increased cell oxidative stress, and changes in lipids composition in a concentration-dependent manner. Responses were more moderate in GSCs and were only observed at highest functionalised GO concentrations. However, no significant toxicity was detected in brain astrocytes and endothelial cells. These results indicate selective toxicity to highly proliferative GBM cell lines and patient GSCs, with minimal toxicity to normal neural cells and brain tissue. We conclude that a novel class of GBM-targeting graphene-based nanocarriers could be useful delivery vehicles for GBM therapeutics.


Introduction
Glioblastoma (GBM) is the most common and lethal type of primary brain tumour. GBM tumours invariably regrow after surgical resection, chemoand-radiotherapy. The median survival after standard treatments is around one year [1]. Accumulating evidence shows that recurrence of this type of tumour is driven by a subset of quiescent GBM stem cells, which share phenotypic and molecular programs with normal neural stem cells and resistant to therapy. These cells can self-renew, proliferate and differentiate (albeit corrupted and inefficient) [2,3]. GSCs were also shown to form tumours in animals upon xenotransplantation [3]. However, understanding the critical molecular and cellular alterations that underpin therapy resistance has yet to translate into improved clinical outcomes. To seek for more efficient treatments, GSCs can be expanded in vitro using NSC culture conditions while retaining tumour-initiating potential and disease-relevant genetic and epigenetic profiles, which serve as in vitro tractable cellular models for the testing of new therapeutics [4][5][6].
Over the last decade, graphene-based composites formed of a single layer of carbon atoms, have attracted significant interest in biomedicine, due to their exceptional physicochemical properties [7]. Since graphene has very low dispersibility in water, various surface modification strategies, a.k.a. functionalization, have been explored to improve biocompatibility for biomedical applications, e.g. nanoscale drug delivery systems for small molecular and siRNA [8][9][10], radio-frequency therapy [11] and simultaneous therapy, sensing and bio-imaging purposes [12]. One of the popular ways to functionalize nanomaterials for drug delivery is via biorthogonal chemistry, e.g. copper(i) catalysed azide-alkyne cycloaddition (a.k.a. CuAAC or click chemistry) [13]. Alkyne-functionalized graphene oxide (GO) has been used in different biological applications, from photoactivation to drug delivery. Double functionalised GO containing both azide and protected alkyne (Click 2 GO) has also been pioneered by our group and was exploited for drug delivery and biomedical applications [14][15][16][17][18]. Here we explore the potential use of Click 2 GO as GSCs targeting nanotherapeutics by studying the bio-nano interactions and cellular responses of cultured non-proliferating brain-derived cells (e.g. brain endothelial cells and astrocytes), patient-derived GSCs, and a classic non-stem GBM cancer cell line (which display astrocytic features, but not neural stem cell identity).

Synthesis of graphene oxide (GO) and Click 2 GO
The detailed procedure for the synthesis of GO and Click 2 GO is provided in supporting material. To reduce the sheet size, Click 2 GO was sonicated for 6 h on an ice water bath. Ice water was replaced every 30 min to keep the temperature low. Click 2 GO water dispersions (3 mg ml −1 × 30 ml per 50 ml tube) were bath sonicated for 10 min then frozen in liquid nitrogen (with gentle shaking during freezing). Samples were placed ∼5 cm above the condense chamber in a Lyotrap freeze dryer (LTE Scientific Ltd., UK. SN: J5467/5) at −55 • C under vacuum at <0.05 mbar (RV3 oil sealed rotary vane pump, Edwards, UK) to freeze dry. The final product was a highly porous, black spongy Click 2 GO cake.

Attenuated total reflectance Fourier transform infrared (ATR-FTIR) of the nanomaterials
ATR-FTIR analysis was performed in freeze-dried Click 2 GO using the PerkinElmer ® Frontier™ FT-IR equipped with ATR accessory (diamond ATR polarization accessory with 1 reflection top-plate and pressure arm). Samples were loaded onto the reflection top-plate (enough to cover the entire diamond surface), the pressure arm was set for force gauge between 100 ∼ 120 units, and 15 scans were performed for each sample.

Raman spectroscopy
For Raman spectroscopy analysis, aqueous dispersions of Click 2 GO were placed on a calcium fluoride (CaF 2 ) slide (Crystran Ltd, UK) and air-dried in a fume hood. Measurements were performed using a Renishaw ® inVia-Reflex spectrometer (UK) with an excitation wavelength of 785 nm, 0.1%-5 % laser power and spectral window of 500 cm −1 to 3200 cm −1 for each sample (n = 3). Data were acquired and analysed using Renishaw's WiRE 4.0 software.

Isolation of primary cells and culturing conditions
The isolation and culturing of porcine brain endothelial cells (PBECs) and rat astrocytes [23,24] is provided in detail in supporting material.

2.7.
Cytotoxicity assessment using the modified lactate dehydrogenase (LDH) assay GL261-luc (10 000 cells well −1 ), G7 (10 000 cells well −1 ), G26 (10 000 cells well −1 ), PBEC (10 000 cells well −1 ) and rat astrocytes (10 000 cells well −1 ) were seeded onto 96-well plates and allowed to set for 24 h. Click2 GO (1 mg ml −1 in DI H2O) was briefly sonicated before addition to the wells at a final concentration of 10, 50 or 100 µg ml −1 . DMSO (10% in cell culture media) was used as positive control for the assay, as it causes cytotoxicity by interfering with membrane permeability. Light microscopy images were captured during cell incubation (10X optical followed by 4X electronic amplification). After 24 or 72 h, cells were washed two to three times with PBS and lysed upon incubation for 1 h at 37 • C with fresh phenol-free DMEM containing 0.9% Triton X-100 (v/v). The cell lysates were subsequently centrifuged for 2 h at 4000 rpm (Eppendorf 5810 R), the supernatant was carefully recovered (to avoid the precipitated cells and Click2 GO) and the activity of LDH was measured CytoTox 96® cytotoxicity assay, following the manufacturer's instructions. Absorbance at 490 nm was measured in a FLUOstar Omega microplate reader (BMG Labtech, Germany). Cell viability was calculated as the percentage of control untreated cells using the following equation: {(A490 of treated cells-A490 of negative control)/(A490 of untreated cells-A490 of negative control)} × 100. Negative control: phenol-free DMEM containing 0.9% Triton X-100.

Evaluation of cell proliferation by flow cytometry
GL261-luc (60 000 cells well −1 ), G7 (55 000 cells well −1 ) and G26 (55 000 cells well −1 ) cells were seeded onto 12-well plates and allowed to set for 24 h. After dilution the CellTrace™ stock solution in HBSS (to a final concentration of 0.5 µM), the probe was incubated with the cells for 20 min at 37 • C. The cells then washed twice with fresh culture medium, and Click2 GO (1 mg ml −1 in DI H2O, briefly sonicated) was added to the cells in fresh culture medium at a final concentration of 10, 50 or 100 µg ml −1 . At 24 or 72 h of incubation, the cells were detached, washed, resuspended in 400 µl of cold PBS and immediately analyzed in a FACSCalibur flow cytometer (BD Biosciences). CellTrace Far Red fluorescence was evaluated in the FL-4 channel, and a total of 20 000 events were collected. The data were analyzed by Cell Quest software (BD Biosciences).

Quantification of graphene in cultured cells using Raman spectroscopy
For Raman spectroscopy studies of Click 2 GO in cultured cells, GL261-luc (70 000 cells well −1 ), G7 (70 000 cells well −1 ) and G26 (70 000 cells well −1 ) cells were seeded on circular CaF 2 coverslips (22 × 0.5 mm) in 6-well plates and allowed to settle for 24 h. The next day, Click 2 GO (1 mg ml −1 in DI H 2 O, briefly sonicated) was added to the cells at a final concentration of 10 or 50 µg ml −1 . After 48 h, cells (excluding floating cells) were washed twice with ice-cold PBS and fixed with paraformaldehyde (PFA, 4% in DI H 2 O) overnight at 4 • C. The cells were subsequently washed twice with DI H 2 O, to remove traces of PFA or PBS that could interfere with measurement and allowed to dry at RT. Raman spectroscopy measurements were performed using an inVia™ Raman microscope (Renishaw, UK) as described in supporting material.

Oxidative stress analysis using the Superoxide Dismutase (SOD) assay
For oxidative stress measurements, GL261-luc (10 000 cells well −1 ), G7 (10 000 cells well −1 ), and G26 (10 000 cells well −1 ) cells were seeded onto 12well plates and allowed to settle for 24 h. Cells were then incubated for 24 or 72 h with Click 2 GO (10, 50 or 100 µg ml −1 ; 1 ml total volume) or H 2 O 2 (2.43 mM). Subsequently, the cells were rinsed 1X with ice-cold PBS, detached using 0.05% Trypsin-EDTA (for 3 min) and, after media inactivation, transferred into 2 ml vials for centrifugation at 4000 rpm (Beckman Coulter Allegra X-22 R) for 10 min at 4 • C. The resulting pellets were washed 2X with ice-cold PBS, resuspended in 200 µl cold PBS followed by two freeze-thawing cycles (freezing in liquid nitrogen, thawing at 37 • C in a water bath) to obtain cell lysates. The lysates were centrifuged at 15 000 × g for 15 min at 4 • C and supernatants were collected to detect the SOD activity (using the SOD determination kit). Briefly, 20 µl of supernatant (or DI H 2 O for control samples) was mixed with 200 µl of WST working solution (WS) in a plate, followed by addition of 20 µl of dilution buffer and 20 µl of enzyme working solution (ES). The plate was then incubated at 37 • C for 20 min, and the absorbance was measured at 450 nm (UV/Vis Absorbance Spectra) in a FLUOstar Omega microplate reader (BMG Labtech, Germany) using the Omega 2.10 software. The SOD activity was calculated using the following equation: SOD  Representative scheme of the synthesis of Click 2 GO. Graphene oxide (GO), obtained by oxidation of graphite using a modified Kovtyukhova-Hammer's method, was epoxidized by reaction with meta-chloroperoxybenzoic acid (mCPBA). Azide-functionalised GO (GO-N3) was obtained by promoting epoxy ring opening via reaction with NaN3. Azide, alkyne double-functionalised GO (Click 2 GO) was obtained via Steglich esterification of carboxylic groups with trimethylsilyl (TMS)-protected propargyl alcohol.

Synchrotron radiation-based Fourier transform infrared microspectroscopy (SR-FTIR) studies
For SR-FTIR measurements in cultured cells and subsequent PCA analysis, GL261-luc (70 000 cells well −1 ), G7 (70 000 cells well −1 ) and G26 (70 000 cells well −1 ) cells were seeded in CaF 2 coverslips (22 × 0.5 mm) and allowed to settle for 24 h. The next day, Click 2 GO (1 mg ml −1 in DI H 2 O, briefly sonicated) was added to the cells at a final concentration of 10 or 50 µg ml −1 . After 48 h, cells (excluding floating cells) were washed 2X with ice-cold PBS and fixed with PFA (4% in DI H 2 O) overnight at 4 • C. The cells were subsequently washed 2X with DI H 2 O, to remove any traces of PFA and PBS that could interfere with the measurement, and allowed to dry at RT. SR-FTIR was performed at the MIRAS infrared beamline of the ALBA synchrotron (Spain), as described in Supplementary Material.

Statistical analysis and others
Data are presented as mean ± standard deviation (SD) of at least three different experiments unless otherwise stated. One-way ANOVA combined with the Tukey post-hoc test, was used for multiple comparisons (unless stated otherwise) and considered significant when p < 0.05. Statistical differences are presented at probability levels of p < 0.05, 0.01 and 0.001. Calculations were performed with Prism 5 (GraphPad, San Diego, CA, USA). Detailed information on GO size measurement by atomic force microscopy, primary porcine brain endothelial cells (PBEC) and rat astrocytes isolation can be found in supporting material.

Production and characterization of graphene oxide derivatives
The synthesis of GO via Mei's modified Kovtyukhova-Hummer's method and surface modification methods to produce Click 2 GO (Scheme 1) has been previously described in detail [14][15][16]. Major steps in the GO synthesis include pre-oxidation of graphite, low-, medium-and high-temperature treatment stages followed by washing and purification, yielding a water-dispersible GO suspension at a concentration of 6.84 mg ml −1 , determined by thermogravimetric analysis. GO epoxidation via reaction with meta-chloroperoxybenzoic acid (mCPBA) was performed before reaction with sodium azide to increase epoxy content. Alkyne groups were introduced via Steglich esterification using trimethylsilyl (TMS)-protected propargyl alcohol, to obtain azide and alkyne double functionalised GO (Click 2 GO) (Scheme 1).
Detailed characterization of the synthesized GO and Click 2 GO were reported in a recent publication from our group [16]. Briefly, the average surface area of the GO and Click 2 GO, measured by AFM and TEM, were 186 nm 2 and 62 nm 2 before and after sonication, respectively. [16] A 6 h sonication further reduced the medium surface area from 62 nm 2 to 50 nm 2 (figures 1(a), (b)). Raman spectroscopy confirmed the presence of the main graphene peaks with the G peak (1597-1600 cm −1 ) created by stretching of sp 2 atoms in rings and chains, and the D peak (1316-1325 cm −1 ) created by stretching of sp 2 atoms in surface defects. The I D /I G ratio, which indicates the degree of surface defects on the graphene flakes, increased after sonication (1.28 ± 0.04 pre-sonication to 1.36 ± 0.12 post-sonication, p = 0.132) A shift in the D peak was also detected for sonicated Click 2 GO, compared to the non-sonicated material (p = 0.059) ( figure 1(c)).
Infrared (IR) spectroscopy confirmed the introduction of azide groups (peak at 2118 cm −1 ) and the presence of enhanced C-OH peaks in Click 2 GO The alkyne peak could not be detected at 2100-2260 cm −1 as the C ≡ C stretching vibration has weak IR transmittance which overlaps with the azide peak as previously reported ( figure 1(d)) [25]. In addition, as shown in our recent publication, the dual-click modification significantly reduced the serum protein binding, i.e. hard protein corona (HC), from 1.44 mg to 0.80 mg HC per mg of GO, a 44% reduction compared to GO. The reduction of HC formation has linearly correlated with the increase in cellular uptake in phagocytic cells (r 2 = 0.99634), i.e. Click 2 GO was taken up ∼40%-50% more efficiently than GO at the same time frame [14].

Double functionalised graphene oxide compromises the viability of glioblastoma cells
The biocompatibility and cellular/molecular interactions and mechanisms of Click 2 GO on one GBM cell lines GL261 (serum grew), two GSCs G7 and G26 (neural stem cell media), non-dividing primary rat astrocytes, and porcine brain endothelial cells (PBEC) were investigated. Light microscopy images showed an association of Click 2 GO with all cell types (figure S1(https://stacks.iop.org/2DM/7/045002/mmedia)) at various level after 72 h incubation in the following order: rat astrocytes >GSCs >GL261 >PBEC.

GBM stem-like cells (GSCs)
Phenotypically-distinct tumor-derived human GSCs (G26, G7), previously characterized and tested in stem cell behavior and drug screening studies, were used in these experiments [19,21]. These cells are maintained as an adherent monolayer culture to provide a more uniform environment that suppresses spontaneous differentiation and to enable direct visualization of cells at single cell resolution [4,5,19,21]. Assessment of cell viability, using a modified LDH assay [26], revealed time-, dose-and GSC typedependent toxicity of Click 2 GO (figured 2 and S2) or the parent GO (figure S3): G26 cells were sensitive to Click 2 GO or GO when incubated at 50 or 100 µg ml −1 for 72 h, while toxicity towards G7 cells was only found after incubation with 100 µg ml −1 Click 2 GO for 72 h (∼67% cell viability, p < 0.01).

Non-proliferating brain-derived cells
Having shown that the toxicity of Click 2 GO in GL261 cells > G26 > G7, experiments were performed to evaluate its toxicity towards non-proliferating brainderived cells using an in vitro blood-brain barrier coculture model: endothelial cells (which form the brain vascular structure) and astrocytes (provide structural and metabolic support). PBEC cells tolerated well Click 2 GO (figures 2 and S2) or the parent GO (figure S3) when incubated for 24 h or 72 h at 10, 50 or 100 µg ml −1 (>80% cell viability). Average values of >90% cell viability were also obtained following incubation of astrocytes for 24 or 72 h with Click 2 GO (figures 2 and S2) or the parent GO (figure S3).

Double functionalised GO induces oxidative stress reversibly and irreversibly in human GSCs and GL261, respectively
Since an effect on cell viability was primarily found for Click2 GO, experiments tested whether Click 2 GO could affect the oxidative balance in cells by measuring the activity of the superoxide dismutase (SOD), an anti-oxidant enzyme that catalyses the dismutation of the superoxide anion (O 2 − ) into H 2 O 2 and O 2 . Similar to our previous observation with the cell viability assay, a concentration-dependent reduction in SOD activity (i.e. higher oxidative stress) was detected in G26, G7 and GL261 cells ( figure 3). Unlike cell viability, however, the most pronounced changes were found in G7 (>4-fold reduction in SOD activity vs control) and G26 cells (>2-fold reduction in SOD activity vs control) exposed to 100 µg ml −1 Click 2 GO for 24 h. Interestingly, prolonged incubation with Click 2 GO reversed this effect in G7 and G26 cells but not in GL261.

Double functionalised GO induces antiproliferative effects on GL261cells
To clarify the mechanisms by which Click 2 GO compromises the viability of proliferating cells (G26, G7 and GL261), initial experiments assessed cell proliferation using the CellTrace™ kit. In this assay, a cellpermeant non-fluorescent molecule is converted to a fluorescent derivative by cellular esterase and retained by the cells. Cell division results in daughter cells receiving half of the fluorescent label of parent cells, i.e. fluorescence intensity is diluted with cell proliferation. In this regard, a small but significant increase in mean fluorescence intensity (indicative of reduced cell division) was only detected in non-stem cancer cells GL261 treated with 100 µg ml −1 Click 2 GO for 72 h (∼115% MFI, p < 0.05) (figure 4).

Double functionalised GO alters lipid and protein signatures of GL261 and GSCs, respectively
To further explore the Click 2 GO toxicities, we assessed biochemical changes in cell components after exposure to Click 2 GO by synchrotron radiationbased Fourier-transform infrared (SR-FTIR) spectromicroscopy. Differences in the resulting infrared spectra were analysed using principal component analysis (PCA). Analysis of the PCA score plots for the lipid region of the IR spectra (3100-2800 cm −1 ) revealed significant separation in principal component 1 (PC-1) for GL261 exposed to 10 or 50 µg ml −1 (figure 5) (p < 0.05). The corresponding loading plots (figure S4) confirmed that biochemical changes in this cell line correspond well with changes in the symmetric and asymmetric CH 2 vibrational modes of the lipid molecules. The most pronounced separation in the protein region (1710-1470 cm −1 ) was found in G26 cells, followed by G7 and GL261 cells when exposed to Click 2 GO at 10 and 50 µg ml −1 (p < 0.05). The PCA loading plots show IR spectral changes in the Amide I and II regions due to biochemical changes caused by exposure to Click 2 GO ( figure S4). Spectral changes in the protein region of GL261 could, nevertheless, be the result of the infrared absorption of Click 2 GO at 1616 cm −1 . These data suggest that toxicities observed in GL26 and G26 are linked to changes in lipid and proteins structures, respectively.

GL261 exhibits the highest association with Click 2 GO compared to other cells
Due to expected variability in the degree of uptake among the cell types studied and taking advantage of the strong D and G bands of Click 2 GO, Raman spectroscopy was used to assess association/interaction between Click 2 GO and the tested cells. Association/interaction between cells and Click 2 GO appeared to vary with cell line and concentration ( figure 6(a), red signals). D-band intensities, normalized to the cell area, were quantified to compare the degree of cell association indirectly. A significant increase in intensity was found Figure 2. Viability of stem-like GBM cells (GSCs), non-stem brain cancer cells and primary brain-derived cells after exposure to Click 2 GO for 72 h. Human GSCs (G26, G7), non-stem glioma cells (GL261), porcine brain endothelial cells (PBEC) and rat astrocytes were incubated with Click 2 GO at concentrations of 10, 50 or 100 µg ml −1 for 72 h. DMSO (10% in cell culture media) was used as positive control for the assay, as it causes cytotoxicity by interfering with membrane permeability. Subsequently, cell viability was assessed via modified LDH assay. Cell viability (mean ± SD), expressed as a percentage to control untreated cells (n = 6, representative of 2 different experiments). * p < 0.05, * * p < 0.01, * * * p < 0.001 compared to control untreated cells.

Figure 3.
Oxidative stress levels in cells incubated with Click 2 GO. GSCs (G26, G7) and non-stem cancer cells (GL261) were incubated with Click 2 GO (10, 50 or 100 µg ml −1 ) or H2O2 (2.43 mM, positive control) for 24 or 72 h. Subsequently, the cells were recovered and prepared for the SOD assay. SOD inhibits the production of a purple-colored formazan (WST-1), the substrate measured in this assay. Higher levels of oxidative stress decrease the pool of SOD available to inhibit this reaction, thus reflecting on lower SOD activity levels. Results are presented as mean ± SD (n = 3, representative of 2 different experiments). * p < 0.05, * * p < 0.01, * * * p < 0.001 compared to control untreated cells.
in GL261 cells, compared to the other tested cells ( figure 6(b)). Decreased cell confluency (i.e. the number of cells per substrate area) was also detected in GL261 cells and, to a smaller extent, G26 cells incubated with 50 µg ml −1 Click 2 GO ( figure 6(c)). The results suggested that increased sensitivity of GL261 to Click 2 GO compared to other cells tested could be due to a combination of an enhanced association/interaction and type of biological responses exhibited by those cells.

Discussion
Driven by therapy-resistant quiescent stem-like cells (GSCs), GBM remains among the most lethal types of cancer. Despite the increased understanding of the molecular alterations involved in therapy resistance [27] and promising experimental therapeutic options clinical outcomes have yet to improve, thus emphasizing the critical need for new therapeutic strategies that can deliver therapeutics widely in the adult brain.
Graphene-based nanomaterials have drawn significant interest for brain application as biomedical devices, due to their exceptional electrical and conductive properties, with studies showing potentiation of synaptic function or downregulation of neuronal signalling without affecting cell viability. Similarly, graphene nanocomposites have been tested as carriers for drug delivery and multimodal GBM therapy [28][29][30][31][32]. Chowdhury and coworkers developed DSPE-PEG-coated GO for delivery of lucanthone to GBM, with results showing significant toxicity towards cultured U251 GBM cells and minimal toxicity to glial progenitor cells [28]. Song and colleagues constructed multifunctional GO-Fe 3 O 4 nanocomposites, targeted to brain cancer via conjugation to lactoferrin, which showed good biocompatibility, large doxorubicin loading capacity and enhanced intracellular delivery efficiency and cytotoxicity [32]. Dong and colleagues developed a multifunctional drug delivery system, combining transferrin-conjugated PEGylated GO and doxorubicin, for simultaneous chemo-and photothermal therapy, with results from intravenously treated gliomabearing rats showing enhanced tumour cell death and prolonged survival compared to single therapy [29]. There is a growing interest in utilizing GO-based nanocarriers for glioblastoma delivery. A systematic study on how these carriers interact with classic GBM cell lines and GSC and impact on their viability is, however lacking.
Cell uptake of graphene is known to be cellspecific and is dependent on particle size and surface chemistry [33]. Protein-coated small GO nanosheets (∼0.42 µm diameter) were shown to enter mesenchymal progenitor cells primarily by clathrinmediated endocytosis, whereas larger nanosheets (∼0.86 µm diameter) were predominantly phagocytosed [34]. Chatterjee and colleagues have also shown that GO could be internalized by HepG2 cells, while the more hydrophobic reduced GO (rGO) is prone to adsorption onto the cell surface without internalization [35]. The size of Click 2 GO used in this study was modulated by sonication for 6 h to achieve flake size reduction. While AFM analysis confirmed that Click 2 GO flakes were smaller than the precursor GO, a significant reduction in flake could not be found after sonication (compared to pre-sonication stocks), which suggests that the smaller size of Click 2 GO results from the functionalization reaction. The effect of sonication on Click 2 GO could also be determined by Raman spectroscopy. The observation of postsonication increases in I D /I G ratio (a measure of the defects presents in the structure) for Click 2 GO could suggest that this process introduces surface defects in the nanomaterial.
With the development of graphene-based products for medical and biological application, a wide range of in vitro and in vivo studies have tested the effect of these nanomaterials in different biological systems [30,[36][37][38][39][40][41][42]. In this regard, graphenebased nanomaterials were found to induce toxicity in target cells via different cellular mechanisms, including loss of membrane integrity [39], apoptosis and cell cycle alterations [43], oxidative stress and mitochondria activation [38,44], and immunotoxicity [45]. Graphene-mediated cytotoxicity was not only related to material concentration but especially to their physicochemical properties, such as surface area [46], layer number, [46] lateral dimensions [34], hydrophobicity and surface chemistry. Surface functionalization is essential for biomedical application as pristine graphene has poor water dispersibility. Chemical modification (via oxidation/amidation) or conjugation to biomedical-grade polymers (such as polyethylene glycol, poly(ε-caprolactone) or dextran) has shown to improve biocompatibility and decrease toxicity [41,47]. Interestingly, while studying the biocompatibility of GO and rGO towards cultured glioma cells, Jaworski and co-workers revealed uptake-and dose-dependent reduction in cell viability and proliferation, this effect was being more pronounced for rGO [30].
In our study, a stronger effect on cell viability was found for Click 2 GO on the non-stem glioma cell line GL261 followed by moderate effects on GSCs (G26) then no effect on the non-proliferating cells. GSCs culture media, for example, is serumfree (FCS is replaced by 0.011% BSA); we, therefore, anticipated higher toxicity for GSCs than serum grew cell lines (media containing 10% FCS). However, this was not the case. This suggests that differences in association/interaction and toxicity profiles are related to intrinsic variability how the different cells respond to GO derivativesboth association/interaction and intra-cellular responses.
GL261 (most positive for Click 2 GO association/interaction) revealed small changes in cell proliferation for the highest tested concentration (100 µg ml −1 ). More importantly, oxidative stress failed to return to normal levels at 72 h postincubation. Moreover, those changes were accompanied by alterations in the biochemical signature of lipids. Interestingly, while a considerable increase in cell oxidative stress, reflected on the decrease in SOD activity, could be detected in GSCs (namely G26), those returned to normal levels within 72 h. Since no alterations in the biochemical signature of lipids were found in GSCs, it could be hypothesised that the failure of GL261 cells to return oxidative stress to normal levels could be due, among others, to compromised membrane structures (reflected in alterations of the lipid profile). GSCs cells, however, showed significant changes in the biochemical signature of proteins (G26 > G7) suggesting that the effects detected at the cellular level could be caused by the interaction of Click 2 GO with these building blocks.
Findings in our study on interactions between GOs and GSCs differ from those reported by Fiorillo et al, which demonstrated GO treatment inhibited several stem cell-associated signal transduction pathways, e.g. WNT, Notch and STAT1/3, which affected tumour-sphere formation and induction of differentiation [48]; however, GO-related alterations in oxidative stress status and mitochondrial function could not be detected. An explanation could be that cancer stem cells were grown in suspension (as neurospheres) [48] in comparison to an adherent culture used in our study. The latter method has been previously validated and shown to provide a more uniform environment that suppresses spontaneous differentiation without the loss of stemness potential [4,5,19,21]. Our results are, however, in line with several other reports which have shown that GO is not toxic to normal stem cells and could be used as a growth substrate and to induce their differentiation into different lineages [49,50].

Conclusions
Our results confirmed selective toxicity of Click 2 GO to highly proliferative GBM cell lines and patient GSCs, and minimal toxicity to primary brain astrocytes and endothelial cells. We conclude that a novel class of GBM-targeting graphene-based nanocarriers could be useful delivery vehicles for GBM therapeutics.