The cytotoxic potential of GO and CXYG nanoplatelets was assessed in the human hepatoma cell line Hep G2 by means of various cell viability assays based on different toxicity endpoints.
In the CFDA-AM assay both graphene derivatives caused a dose-dependent decrease in fluorescence intensity. According to the assay principle a decrease in fluorescence intensity in the CFDA-AM assay may be indicative of various cytotoxic effects including plasma membrane damage, cell proliferation inhibition and cell death. However, observation in the light microscope and measurement of total protein contents did not disclose any significant differences in the amount of cells in treatments and controls. This suggests that the decrease in fluorescence reflected primarily plasma membrane damage.
Yet, attention must be paid to possible interference of GO and CXYG with the assay. Both graphene derivatives were able to act as fluorescence quenchers. However, considering that the nanomaterial-containing medium was entirely removed prior to adding the probe, the amount of GO and CXYG present at the time of analysis was probably much lower than the concentration at which a relevant degree of quenching was observed. Furthermore, it was observed that despite the high colloidal stability of the suspensions GO and CXYG platelets deposited on the plasma membrane forming a layer, which completely covered the cells’ surface at exposure concentrations ≥ 16 μg/ml (≈ 4.2 μg/cm2). The presence of such a layer may locally quench the fluorescence and/or prevent the uptake of CFDA-AM into the cells due to steric, electrostatic or chemical interaction with the probe so that cells would show lower fluorescence intensity, independent of whether or not their plasma membrane is damaged. However, in preliminary experiments in which three different hepatoma cell lines (Hep G2, H4IIE and RTH149) were treated with the same graphene suspension only Hep G2 cells demonstrated a decrease in fluorescence intensity (Additional file 4: Figure S2). This suggests that the observed effect was not due to interference as if this was the case all cell lines would show a similar trend. Moreover, at the lower concentrations at which membrane damage was observed (4 and 8 μg/ml) only a small area of the cell surface was covered with nanomaterial, so that it is unlikely that CFDA-AM diffusion over the cellular membrane was significantly impeded. Yet, the implications for assays that are based on measuring the leakage of cellular macromolecules (e.g. lactate dehydrogenase (LDH) or mRNA) into the medium may be more important than for assays using fluorescent probes of low molecular weight and thus have to be further investigated.
Plasma membrane damage can be the consequence of various cytotoxic effects. The results obtained in this study suggest that the observed loss in plasma membrane structural integrity was associated with a strong physical interaction of GO and CXYG nanoplatelets with the phospholipid bilayer. TEM micrographs of ultrathin sections demonstrated that GO and CXYG nanoplatelets were able to penetrate through the plasma membrane resulting in disruption of the phospholipid bilayer. If the capability of the nanoplatelets to penetrate through the plasma membrane depended on their relative orientation to the latter has to be further examined. Cells responded with the formation of thick intermediate filament bundles, most likely to countervail the tensile forces occurring at the site of interaction/disruption and thus mechanically enforce the plasma membrane and prevent further loss of structural integrity [57–59].
Graphene nanomaterials-caused plasma membrane damage has been reported previously, in both prokaryotic [41, 60] and eukaryotic cells [13, 38, 39, 42, 43, 45]. Liao et al. (2012) demonstrated that both pristine graphene and GO sheets were able to disrupt the plasma membrane of erythrocytes (hemolysis assay).The EC50 calculated for the hemolytic activity of GO platelets with similar dimensions to those used in our study was 30.5 μg/ml (after 3 h of incubation with agitation) . Chang et al. (2010), on the contrary, were not able to observe any adverse effect of GO nanoplatelets with lateral dimensions of about 200, 400 and 800 nm on plasma membrane integrity in the human lung cell line A459. In fact, at exposure concentrations ≥ 50 μg/ml the LDH activity was observed to be even lower than in the control . In a study by Sasidharan et al. (2011), in which carboxyl-functionalized graphene was compared with pristine graphene, no LDH leakage could be observed neither - even at concentrations as high as 300 μg/ml . Zhang et al. (2010) observed that graphene aggregates/agglomerates that had sedimented onto the surface of rat PC12 cells caused an increase in LDH leakage only at the highest exposure concentration (100 μg/ml) . These findings are partly conflicting with those obtained in our study. However, the LDH assay may not be the most appropriate one to assess the membrane disrupting potential of graphene nanomaterials. First, as discussed above, a graphene layer covering the cell surface may impede leakage of LDH into the medium. And second, any enzyme “successfully” released into the medium may adsorb to the suspended nanomaterial and thus be inactivated.
The observed effect of GO and CXYG on plasma membrane integrity was congruent with the concentration-dependent increase in alamarBlue reduction that was observed upon exposure. Although the exact mechanism through which alamarBlue is reduced still has to be elucidated, it is generally assumed that reduction occurs in the mitochondria. There, due to its relatively less negative redox potential, it can receive electrons from various components of the electron transport chain including NADPH, NADH, FADH2, FMNH2, and cytochromes . As a consequence, an increase in resazurine reduction could be indicative of an increase in the metabolic activity of the cells and/or an increase in cell number . Since no significant differences in the protein content of treated and not-treated wells were observed, the increased resazurine reduction could well be related with an augmented metabolic activity of the individual cells. The inverse correlation of metabolic activity and degree of membrane damage may suggest that cells have initiated energy-dependent processes involved in plasma membrane repair (e.g. rearrangement of cytoskeletal elements, biosynthesis of proteins and lipids, trafficking of exocytotic vesicles to injured sites at the plasma membrane) .
The increase in fluorescence observed in the alamarBlue assay could also be due to interference or autofluorescence properties of the graphene derivatives used in this study. However, neither acellular alamarBlue reduction by GO and CXYG nor autofluorescence at the excitation and emission wavelength used could be observed. Besides, incubation of resorufin (=the fluorescent reduction product of alamarBlue) with increasing concentrations of GO and CXYG demonstrated that both graphene derivatives were able to quench its fluorescence. Thus, interference of the tested nanomaterials with the assay would rather lead to an underestimation of the signal.
It must also be kept in mind that resazurine can be reduced by mechanisms different to those stated above. Gonzalez and Tarloff (2001), for example, demonstrated that resazurine can be reduced by cytosolic and microsomal enzymes (S9-fraction) . Thus, an increased expression/activity of the latter, as it is for example observed during detoxification, could also explain the increase in fluorescence intensity. Besides, Lancaster et al. (1996) suggested that resazurine reduction may occur through scavenging of electrons from lipid peroxidation cascades in dying cells [61, 64] and Prutz et al. (1996) demonstrated that resazurine reduction may occur through reaction with free radicals . Since in our study resazurine reduction was correlated with intracellular ROS levels, these mechanisms could also explain the elevated fluorescence intensity at high exposure concentrations.
Induction of oxidative stress is considered one of the principal mechanisms underlying nanomaterial toxicity [66, 67]. In our study, GO and CXYG nanoplatelets were observed to induce the generation of intracellular ROS in a concentration and time-dependent manner. In addition, GO and CXYG-induced ROS formation seemed to follow different kinetics. For GO, maximum ROS levels were reached after exposure to 16 μg/ml for 24 h. In cells treated with lower GO concentrations (1 – 8 μg/ml) intracellular ROS levels kept increasing in the lapse between 24 and 72 h and eventually reached levels comparable to those measured at 16 μg/ml. On the contrary, exposure to low concentrations of CXYG (< 8 μg/ml) did not result in significantly increased ROS levels (not even upon exposure for 72 h). Yet, ROS levels in cells treated with high CXYG concentrations (≥ 8 μg/ml) were observed to increase considerably in the lapse between 24 and 72 h.
Regarding the oxidant-generating potential of GO, the obtained results are consistent with those reported by other authors. As in the present study, Yuan et al. (2012) could not detect any significant increase in intracellular ROS levels in Hep G2 cells exposed to 1 μg/ml of single-layered GO for periods of less than 24 h . Yet, the results presented here demonstrate that exposure to such low concentrations can indeed lead to intracellular ROS formation in this cell line if the exposure duration exceeds 24 h. The ability of GO to induce the generation of intracellular ROS was also assessed in other cell lines. A549 cells exposed to 10 μg/ml GO for 24 h demonstrated comparable ROS levels to those determined in this study . In human skin fibroblasts, however, no significant increase with respect to the control could be detected after 24 h of exposure to concentrations as high as 25 μg/ml . The discrepancy between the results obtained in this study and those stated above (including ours) might be due to differences in the lateral size of the platelets tested (> 1 μm), the suspension protocol (serum-free medium), the assay protocol (loading of the cells with the dye DCFH-DA was carried out prior to treatment) or the sensitivity of the cell line. To our knowledge, no data on the oxidant-generating ability of CXYG have been reported in the scientific literature to this day.
The fact that GO and CXYG-induced ROS generation displayed different kinetics suggests that the underlying ROS-generating mechanisms are distinct. The exact mechanism(s) through which a nanomaterial exerts oxidative stress is relatively difficult to identify and still remains to be elucidated for most nanomaterials including graphene and graphene derivative nanoplatelets. An integrative consideration of results obtained by different assays, however, can help to get a first indication about the possible mechanisms involved. In general, it is distinguished between direct and indirect mechanisms of ROS generation. Direct ROS generation typically involves processes that are independent of the presence of biological systems (acellular ROS generation), i.e. are solely a function of the nanomaterial’s physico-chemical properties. Indirect ROS generation, on the contrary, typically involves cellular (i.e. biochemical) processes that were triggered by the nanomaterial beforehand . In non-inflammatory cells, one of the probably most important nanomaterial-triggered mechanisms leading to increased intracellular ROS formation is impairment of mitochondrial function. To assess whether or not the increased ROS levels may have originated from GO- and CXYG-induced alterations in mitochondrial processes, the nanomaterials´ effect on the mitochondrial integrity was investigated. It was observed that exposure to GO and CXYG nanoplatelets resulted in a decrease in fluorescence intensity in the MMP assay indicating mitochondrial membrane depolarization and/or a decrease in the amount of (functional) mitochondria. These findings are consistent with those of Li et al. (2012), who reported that the MMP decreased in a dose and time-dependent manner in the macrophage cell line RAW 264.7 exposed to increasing concentrations of pristine graphene . Depolarization of the mitochondrial membrane can be due to the loss of both structural and functional integrity of the mitochondrion . Mitochondrial dysfunction is known to be associated with oxidative damage of mitochondrial macromolecules including mtDNA, lipids and proteins caused by reaction with intracellular ROS . Structural damage of mitochondria can be provoked directly, i.e. by physical interaction of the nanomaterial with the mitochondrial membrane  or indirectly, e.g. by physically disrupting the membrane of other cell organelles, such as lysosomes, resulting in release of hydrolytic enzymes into the cytosol .
Both mechanisms would require prior internalization of the nanomaterial, which in the present study was observed. In a few cases interaction of the nanoplatelets with the plasma membrane was observed to be attended by invagination of the latter. Dutta et al. (2007) demonstrated that serum albumin adsorbed to the surface of carbon nanotubes facilitates their uptake via scavenger receptor-mediated endocytosis . Interaction of scavenger receptors in the plasma membrane of Hep G2 with serum proteins adsorbed to the surface of GO and CXYG nanoplatelets may explain the observed membrane invagination at the site of platelet/membrane interaction. However, in this study no evidence for successful uptake of GO or CXYG into endocytotic vesicles was found. TEM micrographs demonstrated that GO and CXYG nanoplatelets were able to penetrate through the plasma membrane and were freely localized in the cytosol. Besides, TEM images showed aggregates of different size and compactness, whereas most (but not all) were enveloped by intracellular membranes. This suggests that GO and CXYG nanoplatelets that entered the cytosol were recognized by the cell as foreign particle, concentrated in one or more defined areas in the cytosol and then packed into intracellular vesicles to isolate the nanomaterial and protect itself from further damage.
Yet, in an initial phase of the internalization process GO and CXYG nanoplatelets and aggregates were freely localized in the cytosol and thus potentially able to directly interact with cellular organelles including mitochondria and lysosomes. In the present work no direct interaction of individual nanoplatelets with lysosomes could be observed. In addition, no adverse effect of GO and CXYG nanoplatelets on lysosomal function was detected in the NRU assay. All these results together suggest that GO and CXYG nanoplatelet-induced ROS generation and mitochondrial damage were not related with release of lysosomal iron or hydrolytic enzymes into the cytosol.
Direct interaction of GO and CXYG nanoplatelets with mitochondria could be observed in one micrograph. Moreover, GO and CXYG-treated cells demonstrated an augmented number of autophagosomes, in some of which degraded mitochondria could be identified. Degraded mitochondria could also be observed in the cytosol, i.e. not yet enclosed in autophagocytotic vesicles. Thus, it may be possible that enhanced intracellular ROS levels originated from mitochondrial damage. If mitochondrial damage was caused by physical interaction of GO and CXYG nanoplatelets with the mitochondrial membrane or is a secondary effect of a possible oxidative damage of mitochondrial macromolecules due to elevated intracellular ROS levels remains to be elucidated [70, 73].
So far, uptake of graphene nanomaterials has been almost exclusively reported for phagocytotic cells [13, 40, 44]. To our knowledge, there is only one published study reporting accumulation of a graphene nanomaterial in the cytosol of a non-phagocytotic cell line . Yue et al. (2012), who studied the uptake of GO in four hepatoma cell lines including Hep G2 could not observe any internalization. They suggested that the negative surface charge of GO may have led to electrostatic repulsion of the platelets from the plasma membrane . The SEM micrographs in the present study however demonstrate that GO and CXYG have a rather high affinity to biological membranes. These findings suggest that other physico-chemical properties may determine whether graphene nanomaterials are internalized and that further research has to be carried out into this direction.