Graphene constitutes a new group of carbon-based nanomaterials defined as a high-aspect ratio material due to the single- or few layered carbon structured 2-dimensional layers arranged in a hexagonal lattice [Bianco et al., 2013]. The lateral size of individual graphene layers spans from nanoscale to microscale [Wick et al., 2014].
Production of 2-dimensional graphene carbon materials is currently increasing and more manufacturers produce graphene at industrial-scale [Ren and Cheng, 2014]. One atom thickness, high conductivity, and transparency are some of the properties that make the 2-dimensional carbon material graphene attractive in future applications, as component in electronics or in medical devices.
Graphene is a carbon-based nanomaterial and chemically similar to carbon nanotubes and carbon black nanoparticles, but with very different morphology. Graphene-based materials are nanosized in one dimension, whereas carbon nanotubes are nanosized in two and carbon black nanoparticles are nanosized in three dimensions. Graphene is usually manufactured by chemical vapor deposition [Li et al., 2009; Reina et al., 2009] or oxidation and exfoliation from graphite [Park and Ruoff, 2009]. In particular, production by graphite exfoliation has increased over the last few years [Ren and Cheng, 2014]. The graphene derivatives graphene oxide (GO) and reduced graphene oxide (rGO) are prepared from oxidation of graphite (GO) followed by, for example, N chemical reduction (rGO). GO is categorized as an insulator due to altered graphitic structure and up to 50% oxygen content [Bianco et al., 2013]. Electric conductivity can partially be restored during reduction to rGO resulting in significantly lowered oxygen content although a complete reduction has not yet been achieved [Park and Ruoff, 2009; Wick et al., 2014].
Graphene [Ambrosi and Pumera, 2014] and carbon black nanoparticles [Jacobsen et al., 2008b; Hogsberg et al., 2013] are usually chemically pure and primarily consist primarily of C, O, and H with low levels of metal impurities whereas carbon nanotubes often have varying levels of inorganic impurities [Jackson et al., 2015]. Due to attractive physicochemical properties of GO for biomedical applications, GO toxicity has been subject to multiple in vitro toxicity studies. Currently, most of the studies have addressed cytotoxicity of relatively small GO layers with lateral sizes below 0.5 µm [Chang et al., 2011; Ali-Boucetta et al., 2013; Lammel et al., 2013; Wang et al., 2013; Wang et al., 2014; Sydlik et al., 2015]. The potential genotoxicity of GO has only been assessed in a few studies [Liu et al., 2013; Wang et al., 2013; Chatterjee et al., 2014].
We have previously studied cytotoxicity and genotoxicity in vitro of both carbon black nanoparticles and 15 different carbon nanotubes in murine lung epithelial cells (FE1) [Jacobsen et al., 2007; Jacobsen et al., 2008b; Jacobsen et al., 2011; Poulsen et al., 2013; Jackson et al., 2015]. The advantage of using this cell line is that it contains 80 copies of the λgt10lacZ transgene, thus allowing for determination of the mutation frequency based on a positive selection assay for a defective functional cII repressor [Jacobsen et al., 2007]. We have previously reported that carbon black was not cytotoxic, but generated reactive oxygen species (ROS) in cellular and acellular assays [Jacobsen et al., 2008b], induced DNA strand breaks and FPG-sensitive sites in FE1 cells [Jacobsen et al., 2007]. Further, carbon black increased the mutant frequency to a similar level as NIST1650 diesel exhaust particles [Jacobsen et al., 2008a]. The mutation spectrum was consistent with being caused by ROS [Jacobsen et al., 2011]. We recently assessed cytotoxicity and genotoxicity of 15 different commercial multiwalled carbon nanotubes (MWCNT) with varying physicochemical properties in FE1 cells [Jackson et al., 2015]. None of the studied MWCNT induced cytotoxicity and only one MWCNT induced DNA strand breaks.
In this study, we compare the cellular response of GO and rGO to the cellular responses of MWCNTs and carbon black using FE1 cells. We conducted an in-depth physicochemical characterization of commercially available GO and rGO and assessed cytotoxicity and genotoxicity in the murine lung epithelial cell line FE1.
In this study, we present an in-depth physicochemical characterization of commercially available GO and rGO materials alongside assessment of cytotoxicity and genotoxicity in the murine lung epithelial cell line FE1.
The studied GO and rGO materials mainly existed as few-layered graphene with lateral sizes in the respirable size range (<5 µm). All materials were highly dispersible although lateral size and wrinkling differed between GO and rGO materials. Wrinkling may be due to lower level of surface oxidation resulting in increased electrostatic repulsion. Lower level of oxidation and subsequent higher degree of wrinkling has also previously been reported (Das et al. 2013). In addition, smaller hydrodynamic sizes than determined in TEM may be interpreted as wrinkling of layers in cell culture medium. Furthermore, TEM imaging indicated similar lateral sizes for rGO-s and rGO-l. In accordance with this, BET surface area of rGO-s and rGO-l were comparable. Although GO existed as few-layered graphene with larger size, less wrinkling of layers supports an estimated surface area comparable to those of rGO-s and rGO-l.
GO is typically produced with a C/O ratio of 2–4, while reduction will increase C/O ratio to approximately 12 [Wick et al., 2014]. In general, graphite exfoliation by Hummers method introduces a higher level of O than other methods [Chua et al., 2012]. As expected, GO had high contents of O and H, rGO-s, and rGO-l contained less. The C/O ratios for GO (∼1.4) and rGO (∼8.0) materials used in this study are similar to C/O ratios of graphene materials used in previous toxicological studies of 0.7–2.9 for GO [Mattevi et al., 2009; Das et al., 2013; Lammel et al., 2013; Sydlik et al., 2015] and 5-12.5 for rGO [Mattevi et al., 2009; Akhavan et al., 2012; Chng and Pumera, 2013; Das et al., 2013; Zhang et al., 2015], respectively. The estimated surface density of hydroxyl groups on the rGO-s and rGO-l (0.019 and 0.025 mmol/m2, respectively) are comparable to levels estimated for hydroxylated MWCNTs (< 0.029 mmol/m2) [Jackson et al., 2015].
The main inorganic impurities found in the present study (manganese and silicon) are most likely contaminants from the oxidation step with sulphuric acid and potassium permanganate, while the reduction step may have led to increased levels of, for example, iron, copper, and zinc due to the use of synthesis reagents that could have been already contaminated [Wong et al., 2014]. However, the level of impurities present in GO and rGO are comparable to or lower than levels found in MWCNTs [Jackson et al., 2015].
The dispersions of the graphenes differed greatly in stability due to the hydrophobic properties of rGO materials. Achieving a stable dispersion in cell culture medium and administration to cells were challenges that need to be highlighted. Immediately following sonication, rGO-s in particular, formed large agglomerates which precipitated (Supporting Information Fig. S3). This may have affected the measurements of the hydrodynamic size. Zeta-potential measurement supports the indication of colloid instability in cell culture medium. Higher instability rGO compared to GO has also been reported previously [Yue et al., 2012; Wang et al., 2013; Chatterjee et al., 2014]. Instability of rGO in cell culture medium correlates with previous findings showing that graphene with C/O ratio exceeding 3 is difficult to disperse in water and quickly sediments [Zhang et al., 2015]. Varying degree of sedimentation may result in differences in cellular uptake of GO and rGO. Cellular uptake was not quantified in the current study but would be highly relevant to assess. Twenty four hours exposures were included to allow uptake to take place, whereas the 3 hr time point was included to detect transient genotoxicity.
Cytotoxicity and genotoxicity in cultured FE1 cells were assessed following exposure to 5–200 µg/ml GO and rGO. Overall, we observed no effect on cytotoxicity (cell viability > 92%) or cell proliferation at any dose after 24 hr exposure to GO or rGO. The experimental setup used in this study is based on our previous toxicity studies with related carbon nanomaterials in FE1 cells at comparable dose-range (5–200 µg/ml); The spherical-shaped carbon black [Jacobsen et al., 2007; Jacobsen et al., 2008b] and a wide range of MWCNTs with varying lengths and functionalization levels [Poulsen et al., 2013; Jackson et al., 2015]. Jacobsen et al. [Jacobsen et al., 2008b] reported no effect on cell viability in spite of ROS generation by carbon black. However, dose-dependent reduction of cell proliferation was observed for some of the MWCNTs [Jackson et al., 2015].
GO is the most frequently investigated graphene-derivative in toxicological research due to the potential in biomedical applications. Cytotoxicity of graphite exfoliated GO with lateral size below ∼0.5 µm has commonly been studied in vitro in the human adenocarcinoma alveolar basal epithelial cells line A549. Chang et al [Chang et al., 2011] reported size and dose-dependent cytotoxicity of 160–780 nm GO (50-200 µg/ml) and GO with smaller lateral size was shown to be most cytotoxic. Dose-dependent cytotoxicity of GO (20–100 µg/ml) was also observed by Hu and colleagues [Hu et al., 2011] in a study where the lateral size of GO layers was not clearly determined. Cytotoxicity of 100 nm GO (125 µg/ml) was also observed in a later study [Ali-Boucetta et al., 2013]. By contrast, a single study reported no cytotoxicity for 300 nm GO (100 and 300 µg/ml) [Jin et al., 2014].
Cytotoxicity has also been reported in human lung fibroblasts and b-lymphocytes cells after exposure to 100–200 nm GO [Wang et al., 2013; Wang et al., 2014]. Cytotoxicity of > 1 µm GO and rGO with lateral size above 1 µm has only been reported at a dose of 1,000 µg/ml in monocytes and macrophages [Sydlik et al., 2015].
Our findings of no cytotoxicity of GO and rGO with lateral size above 1 µm are thus consistent with the current literature.
Introduction of hydroxyl groups has been suggested to increase cytotoxicity of MWCNTs [Magrez et al., 2006; Ursini et al., 2012]. Likewise, it has been suggested that the level of oxidation of GO and rGO may affect cytotoxicity in vitro. Das and colleagues [Das et al., 2013] investigated cytotoxicity of < 800 nm GO and rGO in human endothelial cells and human keratinocytes (10 µg/ml). Although the lateral size was reported to influence cytotoxicity, higher level of oxidation was the major contributor to the reported cytotoxicity. In contrast, Zhang et al.  recently reported increased cytotoxicity of GO in mouse embryo fibroblasts when decreasing level of oxygen. However, the study design was weak, since cell exposures were only performed once. In this current study, the oxidation level had no effect on cytotoxicity.
Generation of ROS by nanoparticles has been linked to cytotoxicity and genotoxicity [Halliwell and Whiteman, 2004]. A covariation of cytotoxicity and ROS generation was recently found following MWCNT exposure of FE1 cells [Jackson et al., 2015]. Current studies on toxicity of GO have reported a correlation between high ROS generation and cytotoxicity [Chang et al., 2011; Schinwald et al., 2012; Lammel et al., 2013; Wang et al., 2013]. In the present study, both GO and rGOs were shown to generate ROS without being cytotoxic. Our results are consistent with the observation that carbon black nanoparticles are efficient ROS generators without inducing cytotoxicity in FE1 cells [Jacobsen et al., 2007; Jacobsen et al., 2008b] and taken together, the results indicate that ROS is not a predictor of cytotoxicity in FE1 cells. Metal impurities such as iron in MWCNTs have been reported to induce toxicity in vivo [Koyama et al., 2009]. In addition, manganese, derived from potassium permanganate used during graphite exfoliation, induced cytotoxicity in, for example, murine macrophages [Yue et al., 2012]. However, the trace amount of metal impurities in the presently studied rGOs and rGO-l were comparable to levels in MWCNTs that were also non-cytotoxic [Jackson et al., 2015].
Overall, we show that ROS generating GO and rGO with larger lateral size (>1 µm) and relatively free of impurities were not cytotoxic even at relatively high doses.
Only few other studies have assessed the in vitro genotoxicity of graphite exfoliated graphene [Das et al., 2013; Wang et al., 2013]. Wang and colleagues [Wang et al., 2013] reported a dose-dependent increase in the level of DNA strand breaks of 200–500 nm GO at a dose-range of 1–100 µg/ml. DNA strand breaks was also reported by Das et al [Das et al., 2013] with reduced genotoxicity following further reduction of GO. In both studies ROS generation was proposed as the main mechanism of genotoxicity.
In the present study, GO and rGO did not induce DNA strand breaks at any of the studied doses and time points. 3 hr and 24 hr were chosen to reflect both transient and prolonged genotoxicity. In the previously mentioned MWCNT study [Jackson et al., 2015], and Printex90 was included as reference particle and induced a 20% increase in DNA strand breaks at 200 µg/ml, determined as TL. In the present study, we also found 20% increase in TL for Printex90 at 100 µg/ml, although the increase was not statistically significant. However, hydrogen peroxide, used as positive control, induced a dose-dependent increase in DNA strand breaks (Fig. 7). Carbon black Printex90 is both an efficient ROS generator and induces DNA strand breaks in the comet assay in FE1 cells. MWCNT induce less ROS and only one of 15 tested MWCNT induced DNA strand breaks in the comet assay in FE1 cells. Our results that rGO and GO induce less ROS compared to Printex90 and no genotoxicity, are consistent with the notion that ROS generation may be an important determinant for genotoxicity. Furthermore, the lack of genotoxicity in the current study suggests that relatively high levels of ROS generation are required for genotoxicity in FE1 cells.
Carbon black has been shown to be both genotoxic and mutagenic in vivo and in vitro and is classified as possibly carcinogenic to humans [International Agency for Research on Cancer, 2010]. The mutation spectrum of carbon black-induced mutations indicates that the mutations may be caused by ROS. For MWCNTs, results are less clear. Several types of MWCNT have been shown to induce ROS and two different MWCNT were shown to induce DNA strand breaks in lung tissue following pulmonary exposure [Poulsen et al., 2015], whereas limited genotoxicity was found in vitro [Jackson et al., 2015]. The biological mechanism underlying the MWCNT-induced genotoxicity in lung tissue is unclear. The genotoxicity may be caused directly by ROS or indirectly by inflammation as suggested for man-made mineral fibers [Topinka et al., 2006]. In our recent study of pulmonary exposure to ten different MWCNT in mice, large diameter of MWCNT was found to predict genotoxicity (Sarah Sos Poulsen et al., submitted for publication), supporting the notion that the observed in vivo MWCNT-induced genotoxicity was not ROS-dependent. The present results suggests that ROS generation by GO and rGO does not induce genotoxicity in FE1 cells. In a future study, it will be informative to compare the present data to in vivo genotoxicity following pulmonary exposure.
In conclusion, we report that few layered GO and rGO with lateral size above 1> µm were not cytotoxic or genotoxic to FE1 murine lung epithelial cells at concentrations up to 200 µg/ml.