Administration route

The common administration routes in animal models include airway exposure (intranasal insufflation, intratracheal instillation, and inhalation), oral administration, intravenous injection, intraperitoneal injection and subcutaneous injection. The major exposure route for GFNs in the working environment is airway exposure, thus inhalation and intratracheal instillation are used mostly in mice to simulate human exposure to GFNs. Though the inhalation method provides the most realistic simulation to real life exposure, instillation is more effective and time-saving method, and GFNs was found that causing longer inflammation period using instillation (intratracheal instillation, intrapleural installation and pharyngeal aspiration) than inhalation [24, 30, 47, 48]. GFNs were investigated to deposit in the lungs and accumulate to a high level, which retained for more than 3 months in the lungs with slow clearing after intratracheal instillation [49]. Intravenous injection is also widely used to assess the toxicity of graphene nanomaterials, and graphene circulates through the body of mice in 30 min, accumulating at a working concentration in the liver and bladder [32, 50–52]. However, GO derivatives had rather finite intestinal adsorption and were rapidly excreted in adult mice via oral administration [31, 53]. Nano-sized GO (350 nm) caused less mononuclear cells to infiltrate subcutaneous adipose tissue after subcutaneous injection in the neck region compared to micron-sized GO (2 μm) [34]. GO agglomerated near the injection site after intraperitoneal injection, and numerous smaller aggregates settled in the proximity of the liver and spleen serosa [31, 33]. Experiments on skin contact with or skin permeation of GFNs were not found in the papers reviewed here, and there is insufficient evidence available to conclude that graphene can penetrate intact skin or skin lesions. The route of nasal drops, which has been widely used to test the neurotoxicity or brain injury potential of other nanomaterials, was not mentioned in the papers reviewed here.

GFNs entry paths

GFNs reach various locations through blood circulation or biological barriers after entering the body, which results in varying degrees of retention in different organs. Due to their nanosize, GFNs can reach deeper organs by passing through the normal physiological barriers, such as the blood-air barrier, blood-testis barrier, blood-brain barrier and blood-placental barrier.

Distribution and excretion of GFNs in tissue

The absorption, distribution, and excretion of graphene nanoparticles may be affected by various factors including the administration routes, physicochemical properties, particle agglomeration and surface coating of GFNs.

The different administration routes influence the distribution of GFNs, for example, intratracheally instilled FLG passing through the air-blood barrier mainly accumulated and was retained in the lungs, with 47 % remaining after 4 weeks [61]. Intravenously administered GO entered the body through blood circulation and was highly retained in the lung, liver, spleen and bone marrow, and inflammatory cell infiltration, granuloma formation and pulmonary edema were observed in the lungs of mice after intravenous injection of 10 mg kg/body weight GO [49]. Similarly, high accumulation of PEGylated GO derivatives was observed in the reticuloendothelial (RES) system including liver and spleen after intraperitoneal injection. In contrast, GO-PEG and FLG did not show detectable gastrointestinal tract absorption or tissue uptake via oral administration [31].

The different properties of GFNs, such as their size, dose and functional groups, always lead to inconsistent results in the distribution profiles of graphene. For instance, Zhang et al. found that GO was mainly entrapped in mouse lungs [49]; however, Li et al. observed that GO accumulated in mouse liver [76]. Notably, small GO sheets, with diameters of 10–30 nm, were mainly distributed in the liver and spleen, whereas larger GO sheets (10–800 nm) mainly accumulated in the lungs [49, 52, 77]. If the size of GO is larger than the size of the vessels, GO usually becomes stuck in the arteries and capillaries in the proximity of the injection site. The accumulation of GO in the lungs was shown to increase with an increase in the injected dose and size, but that in the liver significantly decreased [78]. Coating biocompatible polymers onto GO also affects the biodistribution, for instance, the intravenous injection of GO-PEG and GO-dextran (GO-DEX) accumulate in the reticuloendothelial system (RES), including the liver and spleen, without short-term toxicity [31, 79]. Moreover, the charge of plasma proteins and adsorption of GO by plasma proteins also affects the biodistribution [34].

The excretion and clearance of GFNs vary in different organs. In the lungs, observations indicated that NGO is drawn into and cleared by AMs, which might be eliminated from the sputum through mucociliary clearance or other ways [57], and 46.2 % of the intratracheally instilled FLG was excreted through the faeces 28 d after exposure [61]. In the liver, nanoparticles can be eliminated thorough the hepato-biliary pathway following the biliary duct into the duodenum [80]. In addition, PEGylated GNS that mainly accumulates in the liver and spleen can be gradually cleared, likely by both renal and faecal excretion. As recently reviewed, GO sheets larger than 200 nm are trapped by splenic physical filtration, but small sizes (approximately 8 nm) can penetrate the renal tubules into the urine and be rapidly removed without obvious toxicity [81]. The excretion paths of GFNs have not yet been clearly explained, but renal and faecal routes appear to be the main elimination routes for graphene.

Recently, the distribution and excretion/toxicity strategy has become an important part of nano-toxicological studies. To date, several controversial results regarding the distribution and excretion of graphene in vivo have been reported in several papers, and a systematic evaluation of the toxicokinetics of GFNs is still needed. The metabolism and excretion of nanomaterials are long-period processes, however, the recent studies of GFNs had been limited to short-term toxicological assessments, and the long-term accumulation and toxicity of GFNs on different tissues remain unknown. Therefore, long-term studies on the deposition and excretion of GFNs need to be performed using different cells and animals to ensure the materials’ biosafety before utilization in human biomedical applications.

Uptake and location of GFNs in cells

The uptake and location of GFNs have also been observed to exert different effects in different cell lines. Graphene is taken up into cells via various routes [82, 83]. Basically, the physicochemical parameters such as the size, shape, coating, charge, hydrodynamic diameter, isoelectric point, and pH gradient are important to allow GO to pass through the cell membrane [84]. As stated previously, nanoparticles with diameters <100 nm can enter cells, and those with diameters <40 nm can enter the nucleus [85]. For example, GQDs possibly penetrate cell membranes directly, rather than through energy-dependent pathways [86, 87]. Larger protein-coated graphene oxide nanoparticles (PCGO) (~1 μm) enter cells mainly through phagocytosis, and smaller PCGO nanoparticles (~500 nm) enter cells primarily through clathrin-mediated endocytosis [88]. GO sheets could adhere and wrap around the cell membrane, insert in the lipid bilayer or be internalized into the cell as a consequence of interactions with cells [89]. Similarly, PEGylated reduced graphene oxide (PrGO) and rGO were shown to adhere onto the lipid bilayer cell membrane prominently due to the interaction of hydrophobic, unmodified graphitic domains with the cell membrane [90, 91]. Consequently, it was suggested that prolonged exposure to or a high concentration of graphene induces physical or biological damage to the cell membrane, along with destabilization of actin filaments and the cytoskeleton [92].

Current data demonstrates that GO sheets interact with the plasma membrane and are phagocytosed by macrophages. Three major receptors on macrophages take part in the phagocytosis of GNS: the Fcg receptor (FcgR), mannose receptor (MR), and complement receptor (CR). Furthermore, FcgR is a key receptor in the mediated phagocytic pathway [90, 93, 94]. The protein corona of GO promotes the recognition by macrophage receptors, especially the IgG contained within the protein corona. Macrophages were observed to undergo prodigious morphological changes upon contact with GO [34]. After internalization, graphene accumulated in the cell cytoplasm, perinuclear space, and nucleus, which induced cytotoxicity in murine macrophages by increasing intracellular ROS through depletion of the mitochondrial membrane potential and by triggering apoptosis through activation of the mitochondrial pathway [83]. The possible interactions and accumulation sites of GFNs are summarized in Fig. 1.

Toxicity of GFNs in organs

Toxicity of GFNs in cell models

The cytotoxicity of GFNs in vitro has been verified in various cells to change the cell viability and morphology, destroy the membrane integrity, and induce DNA damage [110–112]. GO or rGO decrease cell adhesion; induce cell apoptosis; and enter lysosomes, mitochondria, cell nuclei, and endoplasm [113]. GQDs entered cells and induced DNA damage by the increased expression of p53, Rad 51, and OGG1 proteins in NIH-3 T3 cells [87]. However, GQDs did not pose significant toxicity to human breast cancer cell lines (at a dose of 50 μg/mL) or human neural stem cells (at a dose of 250 μg/mL) [114, 115]. GO derivatives dramatically decreased the expression of differential genes that are responsible for the structure and function of the cell membrane, such as regulation of the actin cytoskeleton, focal adhesion and endocytosis [89]. In rat pheochromocytoma cells (PC12 cells), graphene and rGO caused cytotoxic effects and mitochondrial injury, such as the release of lactate dehydrogenase (LDH), an increase in the activation of caspase-3, and the generation of ROS [82, 116].

Graphene can increase cell viability [117] or cause cell death [118] depending on the cell line, type of graphene material and the doseage. GO cytotoxicity was observed in human fibroblasts and lung epithelial cells at concentrations above 20 μg/mL after 24 h, but minimal toxicity was found in A549 cells at concentrations higher than 50 μg/mL [119]. The biological responses induced by GO such as ROS, malondialdehyde (MDA), and LDH increased, whereas superoxide dismutase (SOD) decreased dose-dependently in HeLa cells [120]. However, GO-molecular beacon (GO-MB) showed low cytotoxicity even at 20 μg/mL in HeLa cells [121]. GO decreased the viability of A549 cells, while the same concentration and time of exposure increased the cell viability of CaCo2 colorectal carcinoma cells [122]. Another study reported that GO dramatically enhanced the differentiation of SH-SY5Y, accompanied by increasing neurite length and the expression of neuronal marker MAP2 at low concentrations but that GO suppressed the viability of SH-SY5Y cells at high doses (≥80 mg/mL) [123]. Functionalized coatings on GO, such as GO-PEG [124] and GO-chitosan [125], can profoundly attenuate the particles’ cytotoxicity by inhibiting the interactions between cells.

Concentration

Numerous results have shown that graphene materials cause dose-dependent toxicity in animals and cells, such as liver and kidney injury, lung granuloma formation, decreased cell viability and cell apoptosis [130–134]. In vivo studies, GO did not exhibit obvious toxicity in mice exposed to a low dose (0.1 mg) and middle dose (0.25 mg) but induced chronic toxicity at a high dose (0.4 mg). The high content of GO mainly deposited in the lungs, liver, spleen, and kidneys and was difficult to be cleaned by the kidneys via a single tail vein injection [135]. Intriguingly, increasing the dose resulted in a dramatic decrease in the hepatic uptake but an increase in the pulmonary uptake of s-GO by intravenous injection [31], because the high dose of GO potentially surpassed the uptake saturation or depleted the mass of plasma opsonins, which consequently suppressed the hepatic uptake. Moreover, an in vitro study reported that 20 μg/mL GO nanosheets exhibited no cytotoxicity in A549 within 2 h of incubation, but a higher concentration (85 μg/mL) decreased the cell viability to 50 % within 24 h [136, 137]. Lü et al. also demonstrated that GO had no obvious cytotoxicity at low concentrations for 96 h in a human neuroblastoma SH-SY5Y cell line, but the viability of cells sharply decreased to 20 % after treatment with 100 mg/mL GO for 96 h of incubation [123]. The results in HeLa cells, NIH-3 T3 cells, and breast cancer cells (SKBR3, MCF7) treated with graphene nanoribbons also showed a dose- (10–400 mg/ml) and time-dependent (12–48 h) decrease in cell viability [138]. Increasing concentrations of GO entered the lysosomes, mitochondria, endoplasm, and cell nucleus [119]. Several data indicated that rGO caused apoptosis-mediated cell death at a lower dose and early time point but that necrosis was prevalent with the increase in time/dose [110, 135].

Lateral dimension

Nanoparticles with sizes <100 nm can enter the cell, <40 nm can enter nucleus, and smaller than <35 nm can cross the blood brain barrier [85]. One study showed that GO (588, 556, 148 nm) did not enter A549 cells and had no obvious cytotoxicity [112]. When the diameter of graphene is between 100 ~ 500 nm, the smallest size may cause the most severe toxicity, and when the diameter is below 40 nm, the smallest sizes may be the safest. For instance, rGO with a diameter of 11 ± 4 nm could enter into the nucleus of the hMSCs and cause chromosomal aberrations and DNA fragmentation at very low concentrations of 0.1 and 1.0 mg/mL in 1 h. However, rGO sheets with diameters of 3.8 ± 0.4 nm exhibited no notable genotoxicity in hMSCs even at a high dose of 100 mg/mL after 24 h [118].

In an in vivo study, s-GO (100–500 nm) preferentially accumulated in the liver, whereas l-GO (1–5 μm) was mainly located in the lungs because l-GO formed larger GO-protein complexes that were filtered out by the pulmonary capillary vessels after intravenously injection [31]. Given the relative lateral sizes (205.8 nm, 146.8 nm and 33.78 nm) of the three GO nanosheets at the same concentration, smaller GO experiences much greater uptake than larger GO in Hela cells [139]. The high uptake of s-GO changed in the microenvironment of cells and consequently induced the greatest viability loss and most serious oxidative stress among three sizes of GO samples [119]. As a result, one study delineated that GO size-dependently induced the M1 polarization of macrophages and pro-inflammatory responses in vitro and in vivo. Larger GO showed stronger adsorption onto the plasma membrane with less phagocytosis, eliciting robust interactions with TLRs and activating NF-κB pathways, compared to smaller GO sheets, which were more likely taken up by cells [94]. To further uncover the detailed mechanism underlying these effects, more studies are needed to illustrate the vital mechanism of the lateral size of graphene materials.

Surface structure

GFNs possess widely varying surface chemistries. For example, the pristine graphene surface is hydrophobic, GO surface is partially hydrophobic with carboxylate groups [140–142], and rGO has intermediate hydrophilicity [143]. GFNs were observed to disrupt the function and structure of cell membranes and proteins probably by exceptionally strong molecular interactions with cells [2, 91]. For instance, rGO bonded to cell membranes, stimulated receptors and activated mitochondrial pathways, inducing apoptosis [110, 111, 144]. Limited evidence showed that GO is smaller and less toxic than rGO because of the high oxygen content, smoother edges, and hydrophilic properties of the former species [104, 145, 146]. Because of the different surface oxidation states of GO and rGO, GO possessing distinct hydrophilicity might be internalized and taken up by HepG2 cells easily. On the contrary, rGO with evident hydrophobicity, could be adsorbed and aggregated at cell surfaces without (or with lower) uptake [110]. Due to strong π-π stacking interactions, graphene is highly capability of breaking many residues of the protein, particularly the aromatic ones, such as the villin headpiece (HP), F10, W23, and F35. The protein’s secondary and tertiary structures are largely lying on the graphene surface, disrupting the structure and function of the protein [41] (Fig. 2). In addition, GO can insert between the base pairs of double-stranded DNA and disturb the flow of genetic information at the molecular level, which might be one of the main causes of the mutagenic effect of GO [7, 112, 146, 147].

Charge

A number of studies have highlighted the importance of the GO surface charge because of its ability to affect the internalization and uptake mechanism of cells [148–150]. GO internalization was negligible in non-phagocytes, which was likely due to the strong electrostatic repulsion between the negatively charged GO and the cell surface [34]. However, others have suggested that negatively charged nanoparticles can be internalized into non-phagocytic cells by binding to available cationic sites on the cell surface and be taken up by scavenger receptors [110, 146, 150]. GO/GS particles reportedly cause morphological changes and significant lysis, leading to high haemolysis in red blood cells (RBCs). RBC membrane disruption is probably attributed to the strong electrostatic interactions between the negatively charged oxygen groups on the GO/GS surface and positively charged phosphatidylcholine lipids on the RBC outer membrane [106].

Functionalization

Studies confirmed that functionalization with PEG [52], PEGylated poly-L-lysine (PLL) [151], poly(ε-caprolactone) [152], polyvinyl alcohol [3], Pluronic [153], amine [98], carboxyl, and dextran [79] groups largely decreases the toxicity and improves the biocompatibility of graphene. In vivo results revealed that only mild chronic inflammation emerged after the subcutaneous injection of GO-Pluronic hydrogel and no noticeable short-term toxicity was tested after the intravenous injection of GO-DEX [79, 154]. PEGylated GS did not induce appreciable toxicity in mice exposed to 20 mg/kg for 3 months, as evaluated by blood biochemistry and histological examinations, and showed relatively low retention in the RES [52, 155]. Coating GO with chitosan almost eliminated the haemolytic activity in blood [39]. Moreover, the PEG coating effectively alleviated GO-induced acute tissue injuries; decreased GO aggregation and retention in the liver, lungs, and spleen; and promoted the clearance of GO [81], GO-DEX [79], and fluorinated graphene oxide (FGO) [156].

In vitro, several cell function assays showed clear evidence that the surface functionalization of pristine graphene or GO was critical for reducing the strong toxicity effects [91]. PEG-GO, PEI-GO and LA-PEG-GO damaged human lung fibroblast cells less than GO [148]. PEG-GO exhibited no cytotoxicity toward several cell cultures, such as glioblastoma cells (U87MG), breast cancer cells (MCF-7), human ovarian carcinoma cells (OVCAR-3), colon cancer cells (HCT-116), and lymphoblastoid cells (RAJI), at concentrations up to 100 μg/mL [119, 157, 158]. GQDs-PEG exhibited very low or no toxicity against lung and cervical cancer cells even at very high concentrations (200 μg/mL) [159]. However, as a non-biodegradable material with great potential for cellular internalization, further investigation is needed to assess the possible long-term adverse effects of functionalized graphene.

Aggregations and sedimentation

Reportedly, nanomaterials have a propensity to form aggregates rather than individual units, particularly under physiological conditions. GS surfaces allowed fewer RBCs attach comparing to GO, and GS had the lower haemolytic activity for more aqueous aggregations formation. In contrast, the fast sedimentation and aggregate formation of GS greatly inhibited the nutrient availability of human skin fibroblast cells that were grown on the bottom of wells [106]. Therefore, the aggregations and sedimentation of graphene particles exert varying effects on different cells.

Impurities

Nanomaterial purity is an important consideration because residual, contaminating metals may be responsible for the observed toxicity, rather than the nanomaterial itself, which has resulted in conflicting data on GFNs cytotoxicity [35, 160]. Traditionally prepared GO often contains high levels of Mn²⁺ and Fe²⁺, which are highly mutagenic to cells. The nonspecific release of these ions from traditionally prepared GO might lead to unusually high levels of cytotoxicity and DNA fracturing [39]. In particular, Peng et al. [161] produced high-purity GO containing only 0.025 ppm Mn²⁺ and 0.13 ppm Fe²⁺, and Hanene et al. [162] invented a new method to prepare high-purity, single-layer GO sheets with good aqueous dispersibility and colloidal stability. GO produced by these new methods did not induce significant cytotoxic responses (at exposure doses up to 100 μg/mL) in vitro, and no obvious inflammatory response or granuloma formation (exposure doses up to 50 μg/animal) were observed in vivo. Therefore, the purity of GFNs deserves attention and is a vital step towards the determination of GFNs involved in bioapplications.

Protein corona effect

Because of the high free surface charge, nanomaterials can easily form “coronas” with proteins in biological systems [163, 164]. The protein corona is suggested to affect the circulation, distribution, clearance and toxicity of nanoparticles. Several papers reported that GO forms GO-protein coronas with adsorbed plasma proteins in serum and these GO-protein coronas play an important role in deciding the fate of the GO biokinetic behaviour in vivo. Such GO-protein coronas can regulate the adhesion of GO to endothelial and immune cells through both specific and nonspecific interactions [165]. Basically, immunoglobulin G and complement proteins in the protein corona help to reorganize nanoparticles in immune cells, causing the particles to be engulfed by the RES, and IgG-coated GO was taken up by either specific or nonspecific interactions with cell membrane receptors [31, 165]. However, another study found that GO could not adhere to mucosal epithelial cells directly in the intestinal tract after the filial mice drank an aqueous GO solution because abundant proteins in the milk had adsorbed on the surface of the GO and thus inhibited their direct interaction with the mucosal epithelial cells [53]. Protein corona mitigated the cytotoxicity of GO by limiting its physical interaction with the cell membrane and reducing the cellular morphological damage in HeLa, THP-1 and A549 cells [166–168]. The cytotoxic effect was largely reduced when GO was pre-coated with FBS and incubated with cells; nearly ∼ 90 % survival was observed with 100 μg/mL FBS-coated GO and 100 % survival with 20 μg/mL FBS-coated GO. Similar trends were observed for GO covered by BSA [166, 169]. Consistently, additional serum could neutralize the toxicity of pristine GO in J774.A1 cells at a dose of 4 μg/mL, which lead to a decrease in cell number of 52.5 % compared to untreated cells [89].

After reviewing many studies, it can be concluded that the toxicity of graphene is influenced by multiple factors. Those factors combined to largely change the toxicity of GFNs in many cases. Scientific studies often need the clear identification of cause and effect, which should keep only one factor different at a time, so that the effect of that single factor can be determined. But in some papers, several factors influencing GFNs toxicity were studied at the same time, which led to confused results.

Physical destruction

Graphene is a unique nanomaterial compared with other spherical or one-dimensional nanoparticles due to its two-dimensional structure with sp2-carbons. The physical interaction of graphene nanoparticles with cell membranes is one of the major causes of graphene cytotoxicity [7, 170, 171]. Graphene has high capability to bind with the α-helical structures of peptides because of its favourable surface curvature [172]. At concentration above 75 μg/mL, pristine graphene largely adhered to the surfaces of RAW 264.7 cells and resulted in abnormal stretching of the cell membrane [104]. The strong hydrophobic interactions of GFNs with the cell membrane lead to the morphological extension of F-actin filopodial and cytoskeletal dysfunction. Furthermore, the sharpened edges of GNS may act as ‘blades’, inserting and cutting through bacterial cell membranes [173]. Moreover, GO also damaged the outer membrane of E. coli bacteria directly, resulting in the release of intracellular components [173]. However, TEM imaging revealed that pre-coating GO with FBS eliminated the destruction of cell membranes [166].

ROS production leading to oxidative stress

Oxidative stress arises when increasing levels of ROS overwhelm the activity of antioxidant enzymes, including catalase, SOD, or glutathione peroxidase (GSH-PX) [174]. ROS act as second messengers in many intracellular signalling cascades and lead to cellular macromolecular damage, such as membrane lipid breakdown, DNA fragmentation, protein denaturation and mitochondrial dysfunction, which greatly influence cell metabolism and signalling [175–177]. The interactions of GO with cells can lead to excessive ROS generation, which is the first step in the mechanisms of carcinogenesis, ageing, and mutagenesis [83, 122]. Oxidative stress had a significant role in GO-induced acute lung injury [30], and the inflammatory responses caused by oxidative stress often emerged upon exposure to GFNs [133, 177, 178]. The activity of SOD and GSH-PX decreased after exposed to GO in a time- and dosage-dependent manner [82, 106, 119]. Similarly, oxidative stress was the key cause of apoptosis and DNA damage after HLF cells were exposed to GO [148]. Both the mitogen-activated protein kinase (MAPK) (JNK, ERK and p38) and TGF-beta-related signalling pathways were triggered by ROS generation in pristine graphene-treated cells, accompanied by the activation of Bim and Bax, which are two pro-apoptotic members of the Bcl-2 protein family. As a result, caspase-3 and its downstream effector proteins such as PARP were activated, and apoptosis was initiated [83, 179]. Detailed information regarding the MAPK-, TGF-β- and TNF-α-related signalling pathways, which induce inflammation, apoptosis and necrosis, are summarized in Fig. 4.

Mitochondrial damage

Mitochondria are energy production centres involved in various signalling pathways in cells and are also a key point of apoptotic regulation [83]. After exposure to GO and carboxyl graphene (GXYG), the mitochondrial membrane was depolarized, and the amount of mitochondria decreased in HepG2 cells [180]. Exposure to GFNs resulted in significantly increased coupled and uncoupled mitochondrial oxygen consumption, dissipation of the mitochondrial membrane potential, and eventual triggering of apoptosis by activating the mitochondrial pathway [181]. For instance, GO increased the activity of mitochondrial electron transport complexes I/III and the supply of electrons to site I/II of the electron transport chain, accelerating the generation of ROS during mitochondrial respiration in MHS cells [99]. The formation of •OH mediated by GO and the cytochrome-c/H₂O₂ electron-transfer system could enhance oxidative and thermal stress to impair the mitochondrial respiration system and eventually result in dramatic toxicity [151]. Additionally, the oxygen moieties on GO might accept electrons from cellular redox proteins, supporting the redox cycling of cytochrome c and electron transport proteins, and cytochromes MtrA, MtrB, and MtrC/OmcA might be involved in transferring electrons to GO [182]. Therefore, except for the plasma membrane damage and oxidative stress induction, GFNs can cause apoptosis and/or cell necrosis by direct influencing cell mitochondrial activity [183, 184].

DNA damage

Due to its small size, high surface area and surface charge, GO may possess significant genotoxic properties and cause severe DNA damage, for example, chromosomal fragmentation, DNA strand breakages, point mutations, and oxidative DNA adducts and alterations [87, 122, 185, 186]. Mutagenesis was observed in mice after intravenous injection of GO at a dose of 20 mg/kg compared with cyclophosphamide (50 mg/kg), a classic mutagen [112]. Even if GO cannot enter into the nucleus of a cell, it may still interact with DNA during mitosis when the nuclear membrane breaks down, which increases the opportunity for DNA aberrations [87, 147, 187, 188]. The π stacking interaction between the graphene carbon rings and the hydrophobic DNA base pairs can make a DNA segment ‘stand up’ or ‘lay on’ the surface of graphene with its helical axis perpendicular or parallel, respectively. The intermolecular forces severely deform the end base pairs of DNA, which potentially increases the genotoxicity [189]. GO may also induce chromosomal fragmentation, DNA adducts and point mutations by promoting oxidative stress or triggering inflammation through the activation of intracellular signalling pathways such as MAPK, TGF-β and NF-κB [110, 112, 146]. Graphene and rGO can also elevate the expression of p53, Rad51, and MOGG1-1, which reflect chromosomal damage, and decrease the expression of CDK2 and CDK4 by arresting the cell cycle transition from the G1 to the S phase in various cell lines [112]. DNA damage can not only initiate cancer development but also possibly threaten the health of the next generation if the mutagenic potential of GO arises in reproductive cells, which impacts fertility and the health of offspring [112, 190].

Inflammatory response

GFNs can cause a significant inflammatory response including inflammatory cell infiltration, pulmonary edema and granuloma formation at high doses via intratracheally instillation or intravenous administration [30, 49]. Platelets are the important components in clot formation to attack pathogens and particulate matter during the inflammatory response, and GO could directly activate platelet-rich thrombi formation to occlude lung vessels after intravenous injection [98, 191]. A strong inflammatory response was induced by subcutaneously injection with GO for 21 days, along with the secretion of key cytokines, including IL-6, IL-12, TNF-α, MCP-1, and IFN-g [34, 192]. GFNs can trigger an inflammatory response and tissue injury by releasing cytokines and chemokines that lead to the recruitment of circulating monocytes and stimulating the secretion of Th1/Th2 cytokines and chemokines [124, 193]. Additionally, pristine graphene [193] and rGO [110] evoke an inflammatory response by binding to toll-like receptors (TLRs) and activating the NF-κB signalling pathway in cells. The NF-κB signalling cascade is triggered by TLRs and pro-inflammatory cytokines such as IL-1 and TNF-α. Upon activation, NF-κB shifts from the cytoplasm to the nucleus, facilitating the binding of degrading IκB and acting as a transcription factor to synthesize numerous pro-inflammatory cytokines [194]. A schematic of the signalling pathway of TLR4 and TLR9 activated by GFNs is shown in Fig. 5.

Apoptosis

Apoptosis is defined as the self-destruction of a cell regulated by genes through complicated programmes [83, 195]. GO and rGO caused apoptosis and inflammation in mice lungs after inhalation [99], and GFNs also had pro-apoptotic effects in cells [111, 113, 124, 196]. Additionally, graphene and GO physically damaged cell membranes [166], increased the permeabilization of the outer mitochondrial membrane and changed the mitochondrial membrane potential; the increased ROS triggered the MAPK and TGF-β signalling pathways and activated caspase-3 via mitochondrial-dependent apoptotic cascades, prompting the execution of apoptosis [83, 99]. Similarly, rGO caused apoptosis at a low dose and an early time point, triggered by the death-receptor and canonical mitochondrial pathway [110]. Another study showed three different apoptosis pathways by GFNs: GO led to ROS-dependent apoptosis through direct interaction with protein receptors and subsequent activation of the B-cell lymphoma-2 (Bcl-2) pathway; GO-COOH transmitted a passive apoptosis signal to nuclear DNA by binding to protein receptors and activating a ROS-independent pathway; However, GO-PEI severely damaged the membranes of T lymphocytes to trigger apoptosis [105, 197].

Autophagy

Autophagy is the process of self-degradation of cellular components and recently recognized as non-apoptotic cell death [198–200]. Autophagy activation requires autophagosome formation containing Beclin 1, multiple autophagy-related proteins (ATG), microtubule-associated protein light chain 3 (LC3) and p62 [201]. Autophagosome accumulation is associated with exposure to various nanoparticles [202–205], and autophagy can remove extracellular organisms and destruct the organisms in the cytosol [206]. GO and GQDs was shown to induce autophagosome accumulation and the conversion of LC3-I to LC3-II; inhibit the degradation of the autophagic substrate p62 protein [207, 208]. Furthermore, GO can simultaneously trigger TLR4 and TLR9 responses in macrophages [34, 192] and colon cancer cells CT26 [206]. The autophagy pathway is linked to phagocytosis by TLR signalling in macrophages [206, 209].

Necrosis

Necrosis is an alternate form of cell death induced by inflammatory responses or cellular injury. The exposure of cells to pristine graphene causes apoptosis and necrosis at high doses (50 mg/mL) [83]. Reportedly, LDH leakage and the opening of the mitochondrial permeability transition pore, induced by elevated level of cytoplasmic Ca²⁺, lead to apoptosis/necrosis [210]. GO treatment was revealed to induce macrophagic necrosis by activating TLR4 signalling and subsequently partly triggering autocrine TNF-α production [93]. GO combined with CDDP (GO/CDDP) triggered necrosis by decreasing RIP1 and increasing RIP3 proteins, accompanied with the release of high mobility group B1 (HMGB1) into the cytosol from the nucleus and out of CT26 cells [205, 211, 212].

Epigenetic changes

Epigenetics involve DNA methylation, genomic imprinting, maternal effects, gene silencing, and RNA editing [213–215]. DNA methylation, which is one of the best-studied epigenetic modifications, includes phosphorylation, ubiquitination, and ATP-ribosylation and can lead to chromatin remodelling [197, 216, 217]. A recently paper reported that SL-GO/FL-GO exposure resulted in global DNA hypermethylation through upregulating DNMT3B and MBD1 genes; GNP treatment caused hypomethylation by decreasing the expression of DNMT3B and MBD1 genes [216]. GO could activate the miRNA-360 regulation pathway to suppress the DNA damage-apoptosis signalling cascade by affecting the component of CEP-1 [218]. Taken together, these data suggest that GFNs could cause subtle changes in gene expression programming by modulating epigenetic changes. However, studies of GFNs-induced epigenetic changes are few, and the epigenetic mechanism caused by GFNs exposure is not fully understood.

To conclude, many studies have discussed representative mechanisms of GFNs toxicity involving four signalling pathways: TLRs, TGF-β, TNF-α and MAPKs. These four signalling pathways are correlative and cross-modulatory, making the inflammatory response, autophagy, apoptosis and other mechanisms independent and yet connected to each other. Additionally, oxidative stress appears to play the most important role in activating these signalling pathways. It has been reported that there are intersections of apoptosis, autophagy and necrosis in the studies of other nanomaterials toxicity, they inhibit or promote mutually in some conditions. However, the signalling pathways of GFNs toxicity investigated in papers to date are only a small part of an intricate web, and the network of signalling pathways needs to be explored in detail in the future.