Graphene Oxide Nanosheets Stimulate Ruffling and Shedding of Mammalian Cell Plasma Membranes

Summary

Graphene oxide (GO) has attracted intense interest for use in living systems and environmental applications. GO’s compatibility with mammalian cells is sometimes inferred from its low cytotoxicity, but such conclusions ignore non-lethal effects that will influence GO’s utility. Here we demonstrate, with rat basophilic leukemia (RBL) cells, profound plasma membrane (PM) ruffling and shedding induced by GO using confocal and live cell fluorescence microscopy, as well as scanning electron microscopy. These membrane structures contain immunoglobulin E receptors, are resistant to detergents, and lack detectable fluorescence labeling of F-actin and fibronectin. The formation of these membrane structures correlates with a loss of contact inhibition between RBL cells. We observe similar cellular responses towards GO for NIH-3T3 fibroblast cells and MDA-MB-231 human breast cancer cells. These findings reveal a previously unreported cellular response towards foreign nanomaterials. Membrane ruffling and shedding raise fundamental questions about how GO interacts with the PM, as well as its potential to modulate cellular mechanosensing for tissue engineering, stem cell differentiation, and other biomedical applications.

Graphical abstract

eTOC

Graphene oxide (GO) induces significant ruffling and shedding of mammalian cell plasma membranes, in addition to a loss of contact inhibition by rat basophilic leukemia (RBL) cells. These findings demonstrate profound non-lethal effects of GO and challenges assumptions, based on its low toxicity, that GO has only benign effects on cells.

Introduction

Two-dimensional nanomaterials offer a distinct combination of properties, including high surface-area-to-volume ratios, mechanical flexibility and toughness, and the potential for multivalent interactions, all of which are of interest for biomedical applications.^1–6 For example, graphene oxide (GO) nanosheets promote cardiac repair⁷ and stem cell chondrogenic differentiation⁸ and comprise proposed tissue engineering scaffolds.⁹ Evaluating the safety and biological activity of 2D nanomaterials is an essential yet complex prerequisite for realizing these applications.^10–13 Many reports assume that GO is benign because it generally exhibits low cytotoxicity, which varies somewhat as a function of its dimensions and aggregation state,^14–18 the nature of its functional groups,¹⁹ as well as the specific assay and cell type under study.^19,20 Long-term incubation of mammalian cells with GO has been found to induce inflammatory responses and macrophage necrosis,^21,22 but reports of acute cellular responses after GO treatment are rare.²³ Nevertheless, the nature of the interface between dispersed GO and cells is poorly understood,^18,24 and the possibility of these nanomaterials inducing non-lethal cellular responses is not considered or evaluated by most studies.

Here we show that GO stimulates membrane ruffling and shedding in three mammalian cell lines and characterize this behavior in detail using rat basophilic leukemia (RBL) mast cells. After incubation with GO, RBL cells lose their characteristic contact inhibition and large amounts of peripheral membrane structures are generated, most of which detach from the cells and cover the substrate. The separated membrane fragments contain IgE receptors (FcεRI), are resistant to detergent solubilization, and do not show detectable F-actin or fibronectin by fluorescence labeling. In the presence of GO, most RBL cells remain alive but proliferate into uncharacteristic multilayer clusters, even before reaching confluence. We also observe similar membrane shedding for NIH-3T3 mouse fibroblast cells and MDA-MB-231 human breast cancer cells. These findings raise questions about how GO affects cellular mechanosensing as it relates to adhesion,⁷ polarity, motility,²⁵ and receptor-mediated chemical signaling.^21–23,26 These previously uncharacterized effects of GO on the PM and on cellular behavior are particularly noteworthy given interest in functionalized GO for tissue repair,^7,9 stem cell research,⁸ and cancer therapy.²⁷

Results and Discussion

GO Causes Loss of Contact Inhibition in RBL Cells

The interaction of GO with RBL mast cells was evaluated using commercial samples of graphene oxide (Sigma) monolayer dispersion that were prepared through the common Hummers oxidation procedure (see Supporting Information for characterization details).²⁸ Minimal trypan blue staining of nuclei in these GO-treated RBL cell samples (Fig. S1A) revealed low cytotoxicity, consistent with previous studies.^14,20,23 This impermeability to trypan blue also suggested that GO treatment did not cause severe leakiness of the plasma membrane (PM). However, elevated GO concentrations (100 µg/mL) caused RBL cells to lose their characteristic contact inhibition.²⁹ Adherent RBL cells at either a low cell density (20% confluence) or a high cell density (80% confluence) were incubated in the presence or absence of GO (100 µg/mL) in buffered salt solution (BSS, pH 7.4) for 4 h at 37 °C, washed and cultured overnigh t in full medium, and finally labeled with fluorescent AlexaFluor-555 (A555) cholera toxin B subunit (CTxB) in BSS. GO exposure for 4 h did not change the initial RBL cell density, indicating that the incubation procedure does not cause a loss of cell adhesion or cell death. No significant aggregation or change of the solution GO concentration was observed during these experiments or others described below. (Fig. S5G & Supplemental Experimental Procedures). In the absence of GO, RBL cells stopped proliferating after forming a confluent monolayer (compare Fig. 1C with Fig. 1D). However, RBL cells plated at a high initial cell density and incubated with GO (followed by washing) formed a second layer of cells that was 80% as dense as the bottom layer (Fig. 1A–B). Experiments performed at a lower initial cell density indicate that these multilayers can develop even before the cells form a confluent surface-attached layer (Fig. 1E–F; See also Fig. S1B–C). As expected, only surface-attached RBL cell monolayers were observed in low-density samples that were not incubated with GO (Fig. 1G–H). The multilayer growth of GO-treated RBL cells was further visualized with XZ and YZ sections of 3D reconstructed multilayer cell clusters fluorescently labeled on the PMs and nuclei. (Fig. S1D–E) This uncharacteristic formation of cellular multilayers and loss of contact inhibition suggests that GO can perturb the RBL cell plasma membrane (PM), thereby affecting cell-cell communication and the recognition of specific environmental cues. These observations motivated further experiments to characterize the extent of GO’s effects on RBL cells.

Figure 1. Confocal micrographs of A555-CTxB-labeled RBL cells (left: A555-CTxB channel; middle: bright field (BF); right: Overlay).

A, B) Top and bottom layer, respectively, of RBL cells at an initially high density that were incubated with GO (100 µg/mL) prior to washing and proliferation. C, D) Top and bottom layer, respectively, of RBL cells initially at a high density that were not treated with GO prior to washing and proliferation. E, F) Top and bottom layer, respectively, of RBL cells at an initially low density that were incubated with GO (100 µg/mL) prior to washing and proliferation. G, H) Top and bottom layer, respectively, of RBL cells initially at a low density that were not treated with GO prior to washing and proliferation. Scale bars: 50 µm. See also Figure S1.

GO Distribution after Incubation with RBL Cells

Labeling GO sheets with a fluorescent probe enabled their imaging and unambiguous identification after incubation with RBL cells. Because the graphitic regions of GO quench many fluorophores,³⁰ we attached the charge-neutral fluorophore BODIPY to a tripodal graphene binding motif (Fig. S2A–B).^31–33 The tripod binds to graphitic surfaces through three pyrene groups, which projects the fluorophore away from the surface and maintains its bright fluorescence.³¹ The GO sheets functionalized non-covalently with tripodal fluorophores remained well dispersed in BSS and during incubation with cells. RBL cells incubated with BODIPY-labeled GO (GO-BODIPY) also formed cellular multilayers (Fig. S1B–C), indicating that GO and GO-BODIPY have similar effects. Adherent RBL cells were incubated with GO-BODIPY (100 µg/mL in BSS, 4 h), after which the cells were washed with BSS to remove unbound GO-BODIPY. Confocal microscopy of RBL cells revealed the presence of GO-BODIPY as immobilized fluorescent particles colocalized with dark particles in the bright-field images (Fig. S2C). All visible and in-focus GO-BODIPY particles in the bright-field channel were fluorescently labeled in the BODIPY channel. This suggests that GO-BODIPY nanoparticles did not lose their fluorescence in the biological environment, consistent with our previous studies (Figs. S2C, S5 D–E).³¹ The degree of GO endocytosis was judged qualitatively by labeling the RBL cell membrane with A555-CTxB, which binds to the ganglioside GM₁, prior to incubation with GO-BODIPY (Fig. 2A, left path). Some internalization of GO-BODIPY is observed (Fig. S2D), which increases with GO-BODIPY incubation time (Fig. S2E–G), but most of the GO-BODIPY sheets appear colocalized with the PM in top-down images (Fig. S2D, white arrowheads). XZ and YZ sections of GO-BODIPY treated RBL cells also revealed the PM localization of GO-BODIPYs (Fig. S2H). However, 3D reconstructions of the RBL cells depicted in Fig. 2B reveal that most of these GO-BODIPY particles are not directly associated with the A555-CTxB-labeled PM (Fig. 2B–C). They instead appear as fixed particles that are displaced above and around the PM, and the amount of these displaced GOBODIPY particles increases after longer incubation times (Fig. S3A–D). Notably, GO itself is not labeled by A555-CTxB in the absence of RBL cells under similar conditions (Fig. S5A), confirming that any A555-CTxB fluorescence observed is confined to the RBL cell. These unexpected observations led us to hypothesize that the displaced GO-BODIPY particles were immobilized on new, non-A555-CTxB-labeled cell-derived structures.Reversing the order of the A555-CTxB labeling and GO-BODIPY incubation procedures provided distinct micrographs that were consistent with this hypothesis. When the RBL cells were labeled with A555-CTxB after GO incubation (Fig. 2A, right path), similar colocalized GO-BODIPY particles are observed, but these appear directly associated with, and not displaced from, the labeled cellular structures (Fig. 2D–E). These observations suggest that GO-BODIPY is immobilized on new cellular structures that are formed during incubation, and that these structures are not labeled effectively by the A555-CTxB that is already present on the PM when GO is introduced. However, the new structures are readily labeled by A555-CTxB after GO-BODIPY incubation. Although we do not yet have a conclusive explanation, this difference in A55CTxB labeling before and after GO-BODOPY treatment may be related to previous observations that stimulated PM ruffles are populated largely by recycling endosomes in the process of exocytosis.³⁴

Figure 2. 3D reconstructions of A555-CTxB-labeled RBL cells incubated with GO-BODIPY.

A) Two experimental protocols for RBL cell labeling and GO incubation. Left path: RBL cells were first labeled using A555-CTxB and then incubated with GO-BODIPY for 4 h. Right path: RBL cells were first incubated with GO-BODIPY for 4 h, after which they were labeled with A555-CTxB. B, C) 3D reconstructions of RBL cells incubated with GO-BODIPY after A555-CTxB labeling (left path in A). Blue, green, and white arrowheads mark the same cells shown with the different 3D perspectives in B and C, respectively. D, E) 3D reconstructions of RBL cells incubated with GO-BODIPY before A555-CTxB labeling (right path in A). Purple and gray arrowheads point to the same cells shown with the different 3D perspectives in D and E, respectively. These are the only two cells (or part of cells) within the field of view. Other A555-CTxB labeled structures have been identified to be non-cell structures from the bright-field image and F-actin labeling (see below). See also Figures S2 & S3.

SEM Characterization of Membrane Ruffles and Fragments

We employed scanning electron microscopy (SEM) as a complementary technique to visualize the cellular structures induced by incubating RBL cells with GO. Adherent RBL cells were treated with GO suspensions (100 µg/mL in BSS, 4 h) before washing, fixation, and supercritical drying. The RBL cell samples were imaged on an indium tin oxide (ITO) coated coverslip without a conductive coating to reveal sample details, such that charging effects were inevitably present in the images. Control samples that were not incubated with GO show RBL cells with a uniform morphology typical of this cell type with clearly visible microvilli (Fig. 3A–C).³⁵ In contrast, cells incubated with GO appear to be encased in web-like structures (Fig. 3D–F), which are more prevalent with increased GO dosage (Fig. S3E–G). The web-like structures appear to emanate from cell bodies and sometimes extend tens of microns to cover neighboring cells, which are reminiscent of membrane ruffles (Fig. 3E).^35–37 Some of these structures are detached from the cells and appear as either thin fragments with 3D features or wrinkled and absorbed to the substrate. Many cells are completely covered by these structures (Fig. 3F). Their prevalence, extended size, and distinct morphology, as well as their limited punctate fluorescence when generated as a result of GO-BODIPY treatment (Fig. S2C & S5D), indicate that the structures are not simply GO or GO aggregates. They most likely originate from the RBL cells and are responsible for the unexpected immobilization of GO-BODIPY depicted in Figure 2. We hypothesized that these cell-derived structures comprise a mixture of membrane ruffles and shed membrane fragments, and we designed additional imaging experiments with GO-treated RBL cells for more detailed characterization. Further discussion of the origin and identity of the apparent membrane fragments follows a description of these results.

Figure 3. SEM micrographs of RBL cells incubated in the presence and absence of GO.

A–C) RBL cells incubated in BSS buffer for 4 h, then fixed. Scale bars: A) 50 µm; B, C) 10 µm. D-F) RBL cells incubated with GO (100 µg/mL) in BSS buffer for 4 h, then fixed. Scale bars: D) 50 µm; E, F) 10 µm. See also Figure S3.

Confocal Microscopy of Membrane Ruffles and Fragments

Adherent RBL cells were incubated with GO-BODIPY (100 µg/mL, BSS, 4 h), washed, and labeled with A555-CTxB. Following this standard procedure, the samples were fixed, permeabilized, and treated with A647-phalloidin, which labels F-actin for visualization of cell boundaries. A555-CTxB labeling performed after GO-BODIPY incubation reveals apparent membrane fragments that exhibit almost no F-actin labeling. This difference distinguished the separated membrane fragments from cell-associated membranes (Fig. 4A, white arrowheads). Specifically, membrane fragments are faintly visible in the bright-field image and are labeled by A555-CTxB (Fig. 4A). Phalloidin labeling of these structures is below the observable threshold, whereas cells can still be clearly identified in the phalloidin channel. GO-BODIPY particles are associated with the shed membrane fragments (Fig. 4A) and with cell-associated PM ruffles (Fig. 4B–D), similar to those observed in Fig. 2D. 3D reconstructions (Fig. 4C–D) of the cell shown in Fig. 4B reveal that GO-BODIPY particles associate with the PM ruffles. The reconstructed image with overlaid phalloidin and BODIPY channels (Fig. 4D) mimics the displaced GO particles observed in Fig. 2C, and the GO-BODIPY particles are clearly associated with the PM ruffles when the A555-CTxB channel is overlaid with the other images (Fig. 4C). Finally, structured illumination microscopy (SIM) imaging provided enhanced details as to the texture, morphology, and 3D features of the cell-associated PM ruffles and shed membrane fragments (Fig. S4A, S5D–E), which are consistent with our interpretation of confocal images (Figs. 2 and 4).

Figure 4. Confocal microscopy of membrane fragments, GO-BODIPY, and RBL cells after A555-CTxB and phalloidin labeling.

A) Confocal microscopy images of A555-CTxB, GO-BODIPY, and A647-phalloidin channels, as well as the corresponding bright-field image of RBL cells incubated with GO-BODIPY. White arrowheads in the A555-CTxB channel image indicate membrane fragments with associated GO-BODIPY particles. Scale bar: 20 µm. B) Confocal microscopy images of A555-CTxB-labeled membrane ruffles still attached to an underlying RBL cell. Scale bar: 10 µm. C-D) 3D reconstructions of the RBL cell shown in B. Overlaid A555-CTxB, GO-BODIPY, and A647-phalloidin channels are shown in C. The 3D reconstruction shown in D is the same as C but now with the A555-CTxB channel removed. See also Figure S5.

In addition to A555-CTxB labeling, membrane fragments shed from RBL cells can also be labeled by AlexaFluor 488 immunoglobulin E (A488-IgE), indicating that these structures contain IgE receptors. Adherent RBL cells were incubated with GO (100 µg/mL, BSS, 4 h), washed, labeled with A488-IgE, fixed, and finally labeled with phalloidin for F-actin. Similar to treatments described previously (Fig. S4D–E), RBL cells also show noticeable morphological changes after this procedure, including additional spreading and variable surface topography (Fig. 5A). A488-IgE-labeled membrane fragments shed during GO incubation were also co-labeled by both A488-IgE and A555-CTxB (Fig. S5C) to confirm their similarity to those observed in other experiments. These shed membrane fragments prominently decorate the entire substrate.

Figure 5. Confocal micrographs of RBL cells and membrane fragments after A488-IgE and phalloidin labeling.

A) RBL cells incubated with GO (100 µg/mL, 4 h), washed with BSS, and labeled with A488-IgE are fixed and labeled for F-actin with A647-phalloidin. B) Membrane fragments observed elsewhere in the same sample. C) A representative confocal micrograph at lower magnification reveals the abundance of membrane fragments (green) around RBL cells (blue). Confocal and bright-field microscopy images of D) membrane fragments from RBL cells treated with 100 µg/mL GO for 4 h and washed before incubation with unlabeled IgE, followed by addition of A488-IgE. E) Membrane fragments from RBL cells treated with 100 µg/mL GO for 4 h and washed before A488-IgE labeling. Scale bars: 20 µm. See also Figure S5.

Notably, the membrane fragments labeled with A488-IgE are significantly more fluorescent than the RBL cell PM (Fig. 5B). Membrane ruffles are observed still attached to some RBL cells (Fig. S5F). Much like the A555-CTxB labeling experiments described above (Fig. 2), these structures show almost no labeling when A488-IgE is introduced before GO incubation. We considered whether the significantly enhanced A488-IgE labeling efficiency of the membrane ruffles and fragments could be explained by nonspecific binding. To address this question, we extracted the membrane fragments by CHCl₃/MeOH (1:1 v/v, see Supplemental Movie S5). The dispersal of the A488-IgE labeled membrane fragments in organic solvents suggests that the association of A488-IgE to the membrane fragments is mediated by lipids, not through electrostatic attraction to GO’s anionic functional groups. In addition, GO-incubated RBL cells were treated with unlabeled IgE for 1 h, followed by a 5:1 mixture of unlabeled IgE and A488-IgE for an additional 3 h. Under these conditions, the non-fluorescent IgE blocked A488-IgE labeling on both the PM and membrane fragments (Fig. 5D), compared to a control experiment that used only A488-IgE (Fig. 5E). These experiments suggest that IgE binds specifically to receptors that are present on the membrane fragments. They appear to be at a higher density than on the RBL cell surface as evidenced by the increased A488-IgE brightness of the fragments. Our A488-IgE labeling results are consistent with A555-CTxB labeling experiments (Fig. 2), further suggesting that the structures formed as a result of GO incubation come from a membrane pool that does not incorporate fluorescent labels present on the PM before GO is introduced.

Live cell microscopy was used to observe membrane ruffling and fragmentation directly. RBL cells cultured on imaging dishes were labeled with A488-IgE and placed on the microscope, kept at 37 °C. Approximately half of the RBL cells exhibited membrane ruffling and fragmentation within 2 h after the addition of GO. (See Supplemental Movies S1 & S2) Real time imaging confirmed that the observed membrane ruffles and fragments are new cellular structures that form upon interaction with GO. A significant loss of PM fluorescence occurred when cells shed membranes, consistent with internalization of the A488-IgE-receptors apparent in the movie and possible loss to the peripheral membrane structures. This may explain why the cells are dimmer than the membrane fragments when A488-IgE labeling is performed after GO treatment. (Fig. 5B) When A488-IgE labeling was performed prior to GO treatment (See Supplemental Movies_S1 & S2), the shed membranes could not be observed in A488 channel. The membrane structures (Fig. 4 & 5) do not appear as membrane vesicles [microvesicles³⁸ or giant plasma membrane vesicles (GPMVs or blebs)³⁹], but it is possible that such vesicles quickly rupture in the presence of GO and lose their characteristic morphology.⁴⁰ Significant membrane ruffling and fragmentation from the PM might also be related to PM permeabilization reported previously,²³ which occurs on a similar time scale and as a consequence of PM penetration by multilayer GO or graphene aggregates.^17,18 However, the monolayer GO dispersions in our study did not aggregate or precipitate significantly in BSS after incubation with cells. (Fig. S5G)

An anti-fibronectin (FN) antibody was used to evaluate the presence of this extracellular matrix (ECM) protein with the membrane fragments. RBL cells were incubated with GO (100 µg/mL, 4 h), washed, A488-IgE labeled, fixed, and labeled with anti-FN antibody and A647-tagged secondary antibody. Cells, clearly identifiable in bright-field images, exhibit patches of FN at their periphery, consistent with spontaneous RBL cell adhesion. The membrane fragments are poorly labeled, which suggests that these structures do not contain or associate with this ECM protein (Fig. S6E). Although FN was not clearly detected, membrane fragments might incorporate other ECM proteins or rely on other PM components to adhere to the imaging dish surface.

Resistance to Detergent Solubilization and Trypsinization of Membrane Fragments

The absence of observable fibronectin labeling prompted further study of the stability of substrate-adherent membrane fragments. The GO-induced membrane fragments resist solubilization by Triton X-100 (TX-100). Membrane solubilization was characterized by time-lapse microscopy of A488-IgE-labeled cells and membrane fragments after the addition of BSS containing 1% TX-100. A488-IgE-labeled cells lose their integrity within minutes under these conditions, but nearby membrane fragments appear unaffected (see Supplemental Movie S3). A555-CTxB-labeled membrane fragments behave similarly, indicating that the response is not specific to the A488-IgE label (see Supplemental Movie S4). Previous work associates the detergent resistance of membranes to altered interactions with the cytoskeleton, suggesting the possible presence of F-actin on GO-stimulated membrane structures.⁴¹ If so, the detergent resistance of membrane fragments may explain the lack of phalloidin labeling of these structures, because phalloidin labeling generally requires membrane permeabilization by the TX-100 surfactant.

Here we consider the possibility of GO aggregating with lipids extracted from the PM into the observed membrane structures, because the detergent resistance of the membrane fragments raised questions about the nature of these new membranes, and previous studies revealed the interactions between GO and various model lipid structures.^24,40 We think this possibility is unlikely. Firstly, GO is unlikely to form electrostatic aggregates with shed PM lipids due to limited amounts of cationic lipid species in the outer leaflet of mammalian cell PM. The hydrophobic nature of the A488-IgE labeled membrane fragments is also evident by the dissociation of A488-IgE in CHCl₃/MeOH (1:1 v/v, See Supplemental Movie_S5) Secondly, dispersed GO is stabilized by electrostatic repulsions and does not aggregate laterally with physiosorbed lipids to form micron-scale thin films, which has been noted in previous studies.⁴⁰ The insignificant change in [GO] after incubation with cells also suggests minimal loss of GO through aggregation and precipitation (Fig. S5G). In addition, the membrane fragments do not label similarly as GO: The membrane fragments lacked BODIPY labeling in GO-BODIPY treated RBL cells. In the unlikely event that BODIPY-tripods were displaced from GO, unmodified GO quenches the fluorescence of A555-CTxB and A488-IgE (Fig S5A–B). We also collected the membrane fragments and extracted them into yellow oil for liquid chromatography-tandem mass spectrometry (LC/MS-MS). Characteristic PM lipid species, such as phosphatidyl cholines and sphingomyelins, were abundant in the fragments (Fig. S5I). Lastly, live cell imaging of the shedding process revealed the formation of the membrane fragments from the PM by cells. Taken together, these observations cast doubt that the observed structures are GO/lipid aggregates.

The membrane fragments also resist trypsinization procedures routinely used to harvest cells. Incubation with trypsin (see Supporting Information for experimental details) causes a control sample of RBL cells to become more rounded (Fig. S6D), after which shear forces can detach them from the substrate. In GO-treated cell samples, membrane fragments appear unchanged during this procedure (Fig. S6B). This insensitivity to trypsinization is consistent with our observation that fibronectin was not detected in the membrane fragments and suggests that their substrate adhesion occurs through other interactions. Although the molecular basis of the increased stability of membrane fragments to detergents and trypsin is not clear, it is possible that membrane fragments are chemically cross-linked. GO has been characterized as a potent oxidant for preparative organic chemistry,⁴² and treatment of RBL cells may result in the cross-linking of various PM components. However, we are unaware of other examples of such reactivity in biological systems. The trypsin-resistant adhesion of membrane fragments raises questions of how GO-PM interactions would affect cell mechanosensing related to polarity, motility and adhesion, and whether this property provides an explanation for previous anecdotal applications of GO as a cell adhesive and stem cell stimulant.^7,8

We propose that the interface between GO and the PM is complex and ill defined based on the complex nature of the PM. Previous studies of GO and model lipid bilayers revealed electrostatic interactions between negatively charged GO and positively charged lipids.⁴³ Computational studies also suggested that GO is capable of extracting lipids from the model lipid bilayer through a combination of electrostatic and hydrophobic interactions.²⁴ These model studies are informative, but limited when considering the interface between the complex plasma membranes and GO. First, unlike model lipid bilayers of artificially constituted charges, the PM outer leaflet does not contain net positively charged lipids. It is unlikely that GO would be electrostatically bound to the PM lipids; secondly, the PM lipids are layered underneath the abundant membrane-bound proteins and glycans, and therefore are not as readily accessible as they are in the simple lipid bilayer models; furthermore, the lateral dimensions of GO are orders-of-magnitude larger than the PM-bound molecular species. Therefore, GO nanosheets are likely to adhere to compositionally complex patches of the PM rather than to specific PM species. These nonspecific interactions might induce intricate disruptions to the PM, such as altered permeability to molecules. For example, we have made preliminary observations of rapid intracellular Ca²⁺ mobilization in RBL cells upon the addition of GO at few µg/mL concentrations, as well as an efflux of intracellular Ca²⁺ at several tens of µg/mL GO concentration. (D. A. Holowka, and C. Sun, unpublished data) Previous studies also reported activation of aquaporin and water influx in GO treated cells.²³ Yet no permeability to trypan blue was observed in GO-treated RBL cells. For a functioning cell, such changes in PM permeability to different molecules likely reflect perturbations to transmembrane ion channels and other proteins. The PM shedding induced by GO likely results from complex biophysical processes occurring in the PM. These findings warrant follow-up studies to understand PM shedding and its function in mammalian cells, particularly tumor cells.⁴⁴

GO-Induced Membrane Ruffling and Shedding in Other Cell Lines

The generality of the anomalous effects of GO incubation on RBL cells was evaluated in other mammalian cell types. Very similar membrane ruffling and shedding are observed in both NIH-3T3 fibroblast cells (3T3 cells) and MDA-MB-231 human breast cancer cells (231 cells) after GO incubation (Fig. 6). As was observed with the RBL cells, membrane ruffles and fragments formed by 3T3 cells or 231 cells are unlabeled if A555-CTxB labeling is performed prior to GO incubation (Fig. 6B&E). However, when GO incubation is performed prior to A555-CTxB labeling, A555-CTxB-labeled membrane fragments cover the substrate (Fig. 6C&F). The formation of membrane fragments in three different mammalian cell lines demonstrates that these anomalous cellular responses are not limited to RBL cells. Further studies are underway to determine how these cellular membrane responses relate to disrupted chemical signaling and mechanosensing.^25,26,45

Figure 6. GO-induced membrane ruffling and shedding in other mammalian cell types.

Confocal microscopy images of A) Control MDA-MB-231 human breast cancer cells labeled with A555-CTxB. B) MDA-MB-231 cells that were first labeled with A555-CTxB and then incubated with GO for 4 h. C) MDA-MB-231 human breast cancer cells incubated with GO for 4 h, then labeled with A555-CTxB. D) Control NIH-3T3 fibroblast cells labeled with A555-CTxB. E) NIH-3T3 fibroblast cells first labeled with A555-CTxB and then incubated with GO for 4 h. F) NIH-3T3 cells first incubated with GO for 4 h, then labeled with A555-CTxB. Scale bars: 20 µm.

Conclusions

We have identified a profound and previously uncharacterized response from RBL cells to GO derived from the ubiquitous Hummers oxidation process. The graphitic regions of GO samples can be further functionalized with a dye-labeled molecular tripod to fluorescently visualize the distribution of GO with treated cells. GO causes the PM of RBL cells to ruffle extensively, resulting in the shedding of a significant quantity of membrane fragments. SEM and confocal microscopy of fixed samples were used to identify membrane ruffles and fragments formed during GO incubation, and real time imaging of this process revealed the dynamic fragmentation of membrane-derived structures. The membrane fragments do not receive fluorescent labels already bound to the cell PM when GO is introduced but are labeled by a variety of membrane-specific probes after they are formed. Membrane fragments contain IgE receptors, resist detergent solubilization, and are not detectably labeled by fluorescent phalloidin or anti-fibronectin. GO incubation also causes the loss of contact inhibition between RBL cells, which proliferate into multilayered clusters even before they form a confluent monolayer. Very similar membrane fragmentation and shedding processes are observed for NIH-3T3 mouse fibroblast cells and MDA-MB-231 human breast cancer cells.

We and others have observed multiple GO-stimulated cellular responses occurring on the temporal scale of a few hours, such as vacuolization,²³ PM permeabilization,²³ PM ruffling and fragmentation, and PM receptor endocytosis. GO-induced Toll-like receptor-based inflammatory responses and necrosis occur at a much longer incubation time and are likely linked to aforementioned short-term responses of PM disruption.^21,22 The temporal progression of GO-stimulated cellular responses provides a clue to uncover the molecular basis of these phenomena. Future studies will focus on understanding the cellular mechanisms underlying this membrane fragmentation, and how the anomalous membrane behaviors relate to altered membrane signaling, cell motility, adhesion, polarity, and other cellular processes.^7,25,26,45 Our findings reveal previously unknown mammalian cellular responses to this foreign nanomaterial, which, while not cytotoxic, likely reflect environmental stress. These collective experiments question the biocompatibility of GO and caution against assuming its safety in biomedical or environmental applications based on its low cytotoxicity.

Experimental Procedures

Materials

All cell culture reagents, as well as A647-phalloidin, A555-cholera toxin subunit B (A555-CTxB), and A647-labeled goat anti-mouse IgGγ1 were purchased from Invitrogen. Mouse anti-fibronectin was purchased from BD Biosciences. A488-IgE was prepared by modification of purified mouse monoclonal anti-DNP IgE⁴⁶ with AlexaFluor 488 (A488, Invitrogen) as previously described.⁴⁷ BSA and grapheme oxide (GO) were obtained from Sigma-Aldrich. The 35-mm imaging dishes with glass bottoms (14 mm well) were obtained from MatTek Corp. (Ashland, MA).

Cell culture and transfection

RBL-2H3 cells were maintained as monolayer cultures in RBL media: MEM containing 20% (v/v) fetal bovine serum (FBS) and 50 µg/ml gentamicin. Cells were harvested using trypsin-EDTA (Invitrogen) 3–5 days after passage as previously described.⁴⁸ NIH-3T3 cells were cultured as monolayers in DMEM containing 10% (v/v) calf serum and harvested every 3 days after passage. MDA-MB-231 cells were maintained as monolayer cultures in RPMI containing 10% (v/v) FBS and harvested every 3 days after passage.

Cell immunolabeling and fluorescence microscopy

RBL cells were incubated with GO in buffered salt solution (BSS: 135 mM NaCl, 5.0 mM KCl, 1.8 mM CaCl₂, 1.0 mM MgCl₂, 5.6 mM glucose, and 20 mM HEPES, pH 7.4). Cell fixation was carried out with 4% paraformaldehyde in PBS for 10 minutes followed by multiple washes with 1 mg/ml BSA in phosphate-buffered saline (PBS). For F-actin labeling, fixed cells were incubated with 5 µg/mL A647-phalloidin in PBS (10 mg/mL BSA) using 0.1% v/v Triton-X100 for 20 minutes at room temperature before washing with PBS. For FN immunofluorescence, fixed cells were labeled with primary antibody (5–10 µg/mL) at room temperature for 1 h in PBS with 10 mg/mL BSA. After additional washing, fluorescent secondary antibody (5–10 µg/mL) in PBS with 10 mg/mL BSA was added to the samples at room temperature for another 1 h. A final series of PBS washes were performed before storing samples in 2 mL of PBS with 0.01% sodium azide at 4°C until ready for imaging on the microscope. Unless otherwise indicated, samples were imaged on a Zeiss LSM 710 inverted microscope with a motorized stage using a 63× Oil Plan-Apochromat objective lens. A DF 488/561/647 filter set was used to perform sequential 1/2/3 color imaging of the samples. The area of the focal plane was adjusted for optimal image quality.

Scanning Electron Microscopy

RBL cells were plated on ITO coated coverslips overnight before 4 h incubation with 0, 20, and 100 µg/mL GO in BSS buffer. GO-treated RBL cells were washed and fixed with 4% paraformaldehyde in PBS for 10 minutes followed by multiple washes with 1 mg/mL BSA in PBS. Dehydration followed in 25%, 50%, 70%, 95%, and 100% ethanol, twice each, and 100% acetone, also twice. Finally, for imaging (7.0 kV electron beam energy) on samples without gold coating.

Bigger Picture statement.

It is essential to understand the biological effects of nanomaterials to safely and effectively deploy them in living cells. Graphene oxide (GO) is a leading candidate material to use within biological systems because of its low cost, water solubility, and low toxicity. Here we identify previously unrecognized changes in mammalian cell behavior and the plasma membranes in the presence of GO. GO induces dramatic ruffling and shedding of the plasma membranes, which correlates with the cells forming uncharacteristic multilayers. These findings should encourage others to look for nonlethal cellular responses by GO or GO composites in other contexts. Once the biochemical basis of these effects is better understood, GO s ability to present multiple functional chemical handles and modulate cellular responses might be exploited for tissue engineering, stem cell differentiation, or reducing the invasiveness of cancer cells.

Highlights.

Graphene oxide induces ruffling and shedding of mammalian cell plasma membranes

GO-induced plasma membrane responses induce loss of contact inhibition in RBL cells

GO-treated plasma membranes undergo nuanced structural and functional changes