Quantification of C60-induced membrane disruption using a quartz crystal microbalance

Direct contact between fullerene C60 nanoparticles (NPs) and cell membranes is one of mechanisms for its cytotoxicity. In this study, the influence of C60 NPs on lipid membranes was investigated. Giant unilamellar vesicles (GUVs) were used as model cell membranes to observe the membrane disruption after C60 exposure. C60 NPs disrupted the positively charged GUVs but not the negatively charged vesicles, confirming the role of electrostatic forces. To quantify the C60 adhesion on membrane and the induced membrane disruption, a supported lipid bilayer (SLB) and a layer of small unilamellar vesicles (SUVs) were used to cover the sensor of a quartz crystal microbalance (QCM). The mass change on the SLB (ΔmSLB) was caused by the C60 adhesion on the membrane, while the mass change on the SUV layer (ΔmSUV) was the combined result of C60 adhesion (mass increase) and SUV disruption (mass loss). The surface area of SLB (ASLB) was much smaller than the surface area of SUV (ASUV), but ΔmSLB was larger than ΔmSUV after C60 deposition, indicating that C60 NPs caused remarkable membrane disruption. Therefore a new method was built to quantify the degree of NP-induced membrane disruption using the values of ΔmSUV/ΔmSLB and ASUV/ASLB. In this way, C60 can be compared with other types of NPs to know which one causes more serious membrane disruption. In addition, C60 NPs caused negligible change in the membrane phase, indicating that membrane gelation was not the mechanism of cytotoxicity for C60 NPs. This study provides important information to predict the environmental hazard presented by fullerene NPs and to evaluate the degree of membrane damage caused by different NPs.


Introduction
Since its discovery in 1985, fullerene C 60 has attracted great attention because of its unique spherical cage-like molecule structure. 1 The extraordinary properties of C 60 and its derivatives possibly cause them be employed in optical, biological and electronic engineering elds. [2][3][4] Large amounts of C 60 its derivatives are possibly released into the environment, hence their biological toxicities and environmental impact must be considered. C 60 and its derivatives have been reported to cause toxicities on mammalian cell lines, bacteria, invertebrates and sh. [5][6][7][8][9] The direct contact between nanoparticles (NPs) and cell membranes is suggested to cause cytotoxicity. 10 Therefore both computer simulations [11][12][13][14][15][16] and experimental studies 15,[17][18][19] have been conducted to investigate the interaction between C 60 and cell membranes. C 60 possibly disturbs normal cellular functions via lipid peroxidation, production of reactive oxygen species, 17,18 membrane conformational changes, 11 and membrane disruption. 19 The adhesion of NPs on membrane and the subsequent inuence on the structure of lipid membranes is a possible cytotoxic mechanism. Hence, the membrane damage caused by different types of NPs should be compared, which is important to select or design safe nanomaterials. A reliable experimental method is required to quantify the C 60 adhesion on membrane and the induced membrane damage. The living cell is a complicated and dynamic system. For such a purpose, articial lipid membranes are more convenient and reproducible models to avoid the uncertainty from living cells. 20 The limitation of articial membranes is that they do not contain all the lipid and protein components as the plasma cell membrane. Those complicated contaminants of lipids and proteins perform various functions and interact differently with nanoparticles, 21 which cannot be mimicked by any single type of articial membranes. However, articial membranes are exible to be modied according to the research purpose. Giant unilamellar vesicles (GUVs) are good models to observe the membrane disruption and morphologic change directly. 22,23 Membrane charge and components are convenient to be adjusted in GUVs to gure out the interaction mechanism. The membrane disruption is possibly quantied via intact vesicle counting, 24 but it is time-consuming and the accuracy depends on the number of counted vesicles and the observer's experiences. Moreover, the mass of adhered NPs on membrane cannot be monitored by the above-mentioned methods. New techniques are still needed to meet the quantication requirements.
Quartz crystal microbalance with dissipation monitoring (QCM-D) has been used to investigate the interaction of NPs with bio-surface. [25][26][27] By the real-time monitoring of the crystal frequency shis, the mass of adhered NPs on membrane-coated crystal sensor is measured. 26,28 The adsorption amount of C 60 on the lipid membranes has not been reported in previous studies. Furthermore, the degree of membrane damage caused by C 60 possibly be quantied using QCM-D and specially designed model membranes at the rst time. Then we can know fullerene NPs cause more or less serious membrane damage compared with another type of NPs.
The surface charge of NPs is an important factor in the NPinduced membrane damage because charged NPs usually interact with the oppositely charged groups in the membrane. 29 The role of electrostatic force in C 60 -membrane interaction is worth investigations. Fluid-phase cell membranes are essential to keep the normal cellular activities. 30 Nanoparticles exposure has been reported either to cause membrane gelation, 24,31 or increase membrane uidity. 32 C 60 may cause unique effects on membrane uidity, which needs to be evaluated in this study. Therefore, this study aims to evaluate the inuences of C 60 on membrane integrity, morphology and uidity; and aim to build up a method to quantify the degree of membrane damage. The results will provide better understanding to the fullerenemembrane interaction and the related cytotoxic mechanisms.

Preparation of C 60 suspension and characterization
Fullerene is extremely hydrophobic and insoluble. Fullerene stock suspension (100 mg L À1 ) was prepared by weighting 10 mg C 60 powder, and bath sonication (40 kHz) in 100 mL deionized (DI) water or in 0.1 M glucose for 4 hours at room temperature. The stock suspensions were adjusted to pH 6.5 and stored at 4 C, which were used to interact with GUVs and were sonicated again for 30 min before experiments. The suspensions used in QCM-D measurements were obtained from the stock fullerene suspension in DI water. The supernatant was collected from the stock suspension aer standing 24 hours in darkness, which was approximately 60-70 mg L À1 quantied by a UV-visible spectrophotometer. The supernatant was diluted by DI water to 30 mg L À1 , and was sonicated for 30 min to obtain a yellow suspension (Fig. 1d). It was sonicated again for 30 min before QCM-D experiments.
The morphology of C 60 aggregates was imaged by a transmission electron microscopy (TEM, JEM-1011, JEOL Ltd., Japan). To prepare TEM specimens, a tiny drop of C 60 suspension was put on a copper mesh grid, and dried under an infrared lamp. The imaging was performed at an electron emission of 100 kV. The hydrodynamic diameter (d H ) and zeta potentials of C 60 NPs were measured by a Malvern Zetasizer (Nano ZS90, Malvern Instruments Ltd., UK). The Raman spectrum of pristine and sonicated C 60 (4 hours) was obtained at the excitation wavelength of 532 nm via a Raman spectrometer (inVia Reex, Renishaw plc, UK). Fourier transform infrared (FTIR) spectra were obtained using a FTIR spectroscopy (ALPHA-T, Bruker Co., Germany) by collecting more than 50 scans at a resolution of 4 cm À1 . Three C 60 samples were prepared for IR scans: C 60 powder as purchased, C 60 stock suspension in DI water sonicated for 4 hours and for 10 hours and then freeze dried. The collected C 60 samples were mixed with spectrum pure potassium bromide and pulverized, then squashed seconds into a chip under 10 MPa pressure for infrared transmission testing.

Preparation of GUVs
Gaint unilamellar vesicles were prepared by the gentle hydration method using DOPC and charged lipids. 33 10% (w/w) 16:0 TAP (positively charged) or DOPG (negatively charged) was added into DOPC solution in 2 : 1 (v/v) chloroform : methanol to obtain positively/negatively charged GUVs. To prepare RhBlabeled GUVs, 0.1% (w/w) of RhB-PE was added to the lipid solution. The lipid solution (50 mL, 10 mg mL À1 ) was dried under N 2 gas in a 4 mL glass vial to form a lm inside the vial. Then the vial was kept in a desiccator in vacuum for 2 hours to remove the residual organic solvent. Finally the vial was lled with 0.1 M sucrose and incubated at 40 C for 24 hours to obtain the stock solution of GUVs.

Fullerene exposure to GUVs and the microscopic observation
Fullerene suspension (100 mg L À1 ) in 0.1 M glucose was prepared before experiments. GUV stock solution of 10 mL was mixed with 390 mL C 60 suspension in a Petri dish with a glass bottom (30 mm diameter, 0.15 mm thickness). The GUV concentration in the Petri dish was 100 mg L À1 . Morphologies of GUVs were recorded from 30 min to 24 h under bright eld by an inverted microscope with a 40Â objective lens. During the exposure experiment, 0.1 M sucrose inside GUVs and 0.1 M glucose outside GUVs had different refractive indices, which made GUVs visible under bright eld. Moreover, RhB-labeled GUVs were applied to observe the broken GUVs and lipid fragments caused by C 60 exposure. Fluorescent images were taken by a confocal laser microscopy (LSM 700, Zeiss, Germany), and RhB-PE probes were excited by 555 nm wavelength laser.

QCM-D measurement to evaluate C 60 -membrane interaction
The C 60 adhesion on model membrane was monitored using a QCM-D at 25 C (E4, Q-Sense, Sweden). Before experiments, supported lipid bilayer (SLB) or a layer of small unilamellar vesicles (SUVs) was prepared on QCM sensor as the model membrane. 34 SUVs (50 nm) were produced though extruding method, 35 which was introduced in detail in the ESI. † DI water was rst injected into the measurement chamber mounted with SiO 2 crystal sensor (5 MHz, work area: 0.78 cm 2 ) to create a baseline, and then Tris/NaCl buffer (10 mM Tris, 150 mM NaCl, pH 7.0 AE 0.2) was injected as the background solution. Aer the baseline of Tris/NaCl buffer became stable, 100 mg L À1 SUV suspension was injected into the chamber, nally Tris/ NaCl buffer was injected again to remove the residual vesicles in the ow module, and SLB was formed on the silica-coated sensor. The ow rate was 70 mL min À1 during the whole process. To prepare SUV-coated sensor, SUVs of 100 nm in diameter were prepared through the same extruding method. Similar to the process of SLB preparation, DI water, Tris/NaCl buffer, 100 mg L À1 SUV suspension and Tris/NaCl buffer were successively injected into the chamber, but the chamber was mounted with a gold crystal sensor (5 MHz, work area: 0.78 cm 2 ) and the ow rate was 100 mL min À1 . A layer of SUVs was formed on the gold sensor.
Aer the SLB or SUV layer was prepared on the crystal sensor, DI water was injected again to obtain a baseline, and then 30 mg L À1 C 60 suspension in DI water (pH 6.5) was injected into the chamber at the rate of 100 mL min À1 for 1 h. Resonant frequency (f) and dissipation (D) responses of crystal sensors at nth overtones (n ¼ 3, 5, 7, 9, 11) were monitored throughout the experiment, and frequency and dissipation shis (Df and DD) at the 3rd overtone were used for the data processing.
To calculate the mass of C 60 NPs deposited on model membranes, an appropriate model was needed. For thin and rigid lms like SLBs, the DD is nearly zero, and the Df is proportional to the mass change on the sensor surface, as described by the Sauerbrey equation: 36,37 Dm ¼ ÀC Df n n Dm is the adsorbed mass per unit surface, C is the mass sensitivity constant (17.7 ng Hz À1 cm À2 when the quartz crystal frequency is 5 MHz), n is the overtone number and Df n is the frequency shi at the nth overtone. For the viscoelastic and so layer deposited on the crystal sensor, the Sauerbrey equation underestimates the mass change. Instead the Voigt model was applied to quantify the mass change on a viscoelastic layer, [38][39][40] which was tted by the QTools soware (Q-Sense, Sweden). The Voigt model is adapted to analyze the mass, viscosity, thickness of viscoelastic layers. 41,42 Df and DD values were input into model, and uid density, viscosity and the density of the deposited layer were needed as tting constants. The uid density and viscosity were 1.00 Â 10 3 kg m À3 and 1.00 Â 10 À3 kg (m s) À1 . The density of the deposited lipid layer was 1.10 Â 10 3 kg m À3 , 43 and the density of C 60 layer was 1.05 Â 10 3 kg m À3 . 40 More details of QCM-D experiments and data processing are provided in the ESI. † 2.6. Membrane phase quantication aer C 60 exposure In order to study the inuence of C 60 NPs on the membrane uidity, 0.4% (w/w) of laurdan probe was added into the lipid bilayers during the GUV preparation. 24 Laurdan-labeled GUVs of 50 mg L À1 were then exposed to 100 mg L À1 C 60 suspension. The uorescence spectra of GUVs were measured by a uorescence spectrophotometer (F7000, HITACHI, Japan) at 1 h and 24 h, respectively. The sample was excited at 380 nm, and the emission spectrum was collected from 400 nm to 600 nm. Generalized polarization (GP) was used to quantify the phase of phospholipid membranes, 24,44,45 which was dened as:

Characteristics of C 60 powder and suspension
Fullerene C 60 forms clusters of tens of nanometers revealed by TEM images (Fig. 1a). The C 60 clusters are stable crystalline form with reported diameters of 25-500 nm in water, 6 which is much larger than C 60 molecule (0.71 nm in diameter). 47,48 The measured average d H of 100 mg L À1 C 60 stock suspension is 453.5 AE 5.1 nm in 0.1 M glucose and 409.6 AE 13.6 nm in DI water with broad size distribution (Fig. 1b). The two C 60 stock suspensions are negatively charged in the whole measured pH range (pH 3-10) (Fig. 1c). The diluted C 60 supernatant in water (30 mg L À1 ) has smaller average d H (256.9 AE 1.4 nm) with narrower size distribution, as well as more negative zeta potentials, suggesting a better dispersion. Although pristine C 60 molecules with perfect structure do not contain functional groups, they can extend to water via sonication or long-term mixing in water for a period of days to weeks. 49,50 C 60 molecules are strong electron accepters, and interact with the potential electron donors like water molecules to become negatively charged. 51,52 The adsorption of water molecules on pristine C 60 clusters forms hydrated layer and improve its solubility. 52,53 The diluted C 60 supernatant (30 mg L À1 ) shows negligible sedimentation in initial 2 hours and stable hydrodynamic diameter aer 30 min sonication (ESI Fig. S1c and d †), suggesting C 60 aggregates can be stably suspended in supernatant.
Furthermore, the Raman spectra of pristine C 60 powder and sonicated C 60 are presented in Fig. 1e. Both pristine and sonicated C 60 show Raman-active A g (495 and 1469 cm À1 ) and H g vibrational modes (268, 769, 1423 and 1570 cm À1 ), 54,55 indicating no distinct structural defects. The FTIR spectra (Fig. 1f) show sharp peaks at 525, 575, 1185 and 1428 cm À1 , which derives from the fundamental IR-active vibrational modes of C 60 skeleton with icosahedral symmetry. 56,57 Aer 4 hour sonication, new distinguishable absorb peaks emerge at 3421 and 1628 cm À1 corresponding to O-H groups, 58 which enhances with longer sonication time (10 hours). Another band appears from 930 to 1200 cm À1 aer 10 hour sonication, corresponding to C-O groups with various chemical environments. 58 It indicates that 10 hour sonication may induce more serious structural changes compared to 4 hour sonication. A slight absorbance increase is detected at the range of 2800-3000 cm À1 , which is caused by the formation of C-H bonds. 59 Therefore, sonication increases the hydroxyl groups on C 60 surface, which makes C 60 more hydrophilic and contributes to the dispersion of C 60 . The form of surface hydroxyl is attributed to the hydrogen radicals that generated from water by sonication. 57 Compare to 10 hour sonication, 4 hour sonication effectively stabilizes C 60 suspension and does not cause serious C 60 structure damage (conrmed by Raman spectrum), which is more suitable for pristine C 60 study.

Effects of C 60 on GUVs revealed by microscopic imaging
Giant unilamellar vesicles are exposed to C 60 suspension in 0.1 M glucose at pH 6.5, and their morphology is imaged under bright eld (Fig. 2). Both GUVs + and GUVs À are 20-100 mm spheres under microscopy. However, there is big difference between GUVs + and GUVs À aer C 60 exposure. C 60 NPs form obvious aggregates and adhere on the GUVs + at 2 hour (Fig. 2a). Serious membrane disruption appears at 8 hour, and few GUVs exist when it comes to 24 hour. GUVs are broken into lipid fragments and aggregate with C 60 NPs, which appears as the black spots on the images. The broken GUVs are not visible under bright eld due to the leakage of the enclosed sucrose. Therefore RhB-labeled GUVs are used to show the morphology of disrupted GUVs. The formation of lipid agglomerates from disrupted GUVs is revealed by the uorescence imaging (Fig. 3), indicating that C 60 adhesion causes lipid fragments to attach to each other. However, GUVs À are not disrupted by C 60 up to 24 hours, and no C 60 aggregates or lipid fragments are found in the images (Fig. 2b and ESI Fig. S2 †).
Cell membranes contain both positively and negatively charged domains, the positively charged domains are relatively scarcer than negatively charged ones. 29 In the GUV exposure experiment, the zeta potentials of 100 mg L À1 C 60 are À17.5 mV in DI water and À13.7 mV in glucose at pH 6.5 (Fig. 1c). Negatively charged C 60 NPs only adhere on and damage the membranes containing positively charged groups. The positively charged sites on the membrane seem to be necessary for C 60 NPs approaching to the membrane and for the consequent membrane damage. It indicates that negatively charged NPs can approach to the cationic sites on the plasma membrane through nonspecic binding, which can explain the strong and nonspecic interaction of anionic NPs with the plasma membrane and their subsequent endocytosis reported in previous studies. 60,61

C 60 adhesion on model membrane and induced membrane disruption monitored by QCM-D
Supported lipid bilayer on SiO 2 sensor is designed to show the mass of adhered C 60 on membrane, while a SUV layer on Au sensor is designed to reveal the combined mass change of C 60 adhesion and vesicle disruption. To prepare SLBs on QCM-D sensor, SUVs are injected into QCM chamber and then disrupt quickly to form a continuous SLB on the SiO 2 sensor. Df and DD provide the information of mass change and viscoelastic properties of the deposited layer, respectively. 41 During SLB formation process, Df decreases because of the vesicle deposition, and then becomes stable aer SLB formation ( Fig. 4a and b). It indicates that no more phospholipids adsorb  Paper on the sensor aer the sensor has been fully covered by the lipid bilayer. The vesicle disruption induces a peak on DD curve because of the different viscoelasticity between the adhered intact vesicles and the consequently formed phospholipid bilayer. DD just slightly increases aer SLB formation and DD/Df is less than 1 Â 10 À8 Hz À1 (Table S1 †), indicating that SLB is a rigid lm. 34 When 30 mg L À1 C 60 suspension is injected to the SLB + -coated sensor, Df decreases by 27.5 Hz and DD increases by 12.5 Â 10 À6 aer one hour (Fig. 5a), indicating the fast deposition of C 60 NPs on the SLB + . To prepare the layer of SUVs, intact SUVs deposit on the Au sensor, leading to the decrease of Df and the increase of DD (Fig. 4c and d). Different from SLB, the layer of SUVs is a viscoelastic lm (Table S1, † DD/Df > 1 Â 10 À8 Hz À1 ). 34 Aer C 60 injection, a decrease of Df and an increase of DD are also observed on SUV + -coated sensor (Fig. 5c), indicating the deposition of C 60 NPs on SUVs + .
Fullerene suspension is also injected to uncoated SiO 2 and Au sensor to compare with the membrane-coated sensor (Fig. S3 †). No deposition is observed on the uncoated sensor, which conrms that changes of Df and DD on SLBs/SUVs are due to the interaction between C 60 NPs and the membrane. Based on the relatively high DD/Df values of C 60 deposition on SLB and SUV layer (Table S1 †), the deposited C 60 NPs form a viscoelastic lm on both SLBs + and SUVs + . Therefore the Voigt model is employed to calculate the mass of adhered C 60 NPs on the SLB/SUV-coated crystal sensor. 41,42 Mass changes on the SLB-or SUV-coated crystal sensors during C 60 injection are presented in Fig. 6a. Aer 1 hour injection, the masses of adhered C 60 on SLBs + (Dm SLB ) and on SUVs + (Dm SUV ) are calculated to be 0.9226 mg and 0.4466 mg on sensor, respectively. The mass increase on SLB + is remarkably higher than on SUV + layer. The surface area of SUV layer (A SUV ) is calculated to be 1.42 cm 2 on QCM sensor (Fig. S4 †), which is larger than the surface area of SLB (A SLB ¼ 0.78 cm 2 ). Hence the ratio of adhesion mass on SUV + to SLB + (Dm SUV /Dm SLB ) should be proportional to the ratio of surface area (A SUV /A SLB ¼ 1.81).
If C 60 deposition on sensor only cause particle adhesion on the two types of model membranes, the constant k should be 1 at a given injection time. However, Dm SUV /Dm SLB of C 60 is between 0.33 and 0.50 (Fig. 6b), and the constant k is between 0.18 and 0.28 (much less than 1). It indicates that partial SUVs are disrupted by C 60 , therefore the removal of lipid fragments and the release of the enclosed uid from SUVs cause mass loss. Imaging on GUV morphology conrms that C 60 NPs disrupt positively charged vesicles ( Fig. 2 and 3), however a quantitative way is required to evaluate how serious the membrane damage is and to compare the biological risks between different NPs. The constant k can be the used to quantify the membrane disruption, and the smaller k value means more serious membrane disruption. We also calculated Dm SUV /Dm SLB and constant k values of CdTe quantum dots (QDs) using the QCM monitored mass data in a previous study. 62 The Dm SUV /Dm SLB of amino-coated QDs is between 1.44 and 2.21, and the constant k is between 0.79 and 1.22 ( Fig. S5a and b †). Dm SUV /Dm SLB of carboxyl-coated QDs is between 1.16 and 1.34, and the constant k is between 0.64 and 0.74 ( Fig. S5c and d †). Therefore C 60 cause more serious membrane disruption than CdTe QDs. However, the mass loss induced by SUV disruption does not exceed the mass increase caused by C 60 adhesion, hence only a small part of SUVs are disrupted and Df is governed by C 60 adhesion.
On SLB À and SUV À layer, C 60 injection does not cause obvious Df and DD change ( Fig. 5b and d), and results in almost zero mass increase in Fig. 6. It suggests that C 60 NPs do not adhere to the membrane containing negatively and neutrally charged lipids. C 60 NPs only adhere on SLB + or SUV + layer containing positively charged lipids, which is consistent with the microscopic observation using GUVs.
Experiments using GUV, SUV and SLB model membranes have conrmed that negatively charged C 60 NPs only adhere on and disrupt membranes containing positively charged lipids, indicating the importance of the electrostatic force in the interaction between C 60 NPs and lipid membranes. Most biological membranes are negatively charged. However, there are still a small amount of cationic sites on the cell membranes, 29,47 hence negatively charged NPs can bind to the cationic sites and cause the subsequent cell membrane damage. Studies on the interaction of lipid membranes with other carbon nanomaterials (single-walled and multi-walled carbon nanotubes, graphene oxides) also demonstrate the role of the electrostatic force. 26,63,64 Therefore, the electrostatic force is a crucial mechanism in the interaction of carbon nanomaterials with lipid membranes.

Inuence of C 60 NPs on membrane uidity
Fluid phase cell membrane is essential to support membrane proteins and to maintain normal molecular transport into and out of cells. 10 Laurdan is a uorescent probe which is sensitive to the polarity of the surrounding environment. The uorescent spectra of laurdan emission and GP values are presented in Fig. S6. † In gel-phase membrane few water molecules exist inside phospholipid bilayers, therefore the maximum of laurdan emission is about 440 nm. 45 In uid-phase membrane, more water molecules exist between phospholipid molecules and cause the dipolar relaxation of laurdan. The maximum laurdan emission shis to 490 nm. 45 Negatively charged NPs usually cause membrane phase gelation because they attract the -N + terminus of the phospholipid head groups and increase the angle of P À -N + electric dipole. 65 However, the prepared GUVs + and GUVs À are both in uid phase (GP < À0.05), and the exposure of 100 mg L À1 C 60 NPs does not change the GP values signicantly. Such particles inducing membrane phase gelation have been reported for carboxyl-modied polystyrene latex and silica NPs ($20 nm in diameter). 24,65 In contrast to those NPs, C 60 molecules reside on the border between molecular level chemicals and nanomaterials (Fig. 1c). They penetrate into the interior of membranes and then remain encased in the hydrophobic lipid tails revealed by both molecular dynamic simulations and experimental conrmation. 13,14,66,67 C 60 forms stable aggregates in water. Transferring one C 60 from the aggregate into bulk water is highly unfavorable, but removing one C 60 from a large aggregate and placing it into the lipid bilayer interior is favorable. 12 Large C 60 aggregates adsorb on the lipid membranes, and partially disaggregate into small aggregates or single C 60 molecules which can penetrate into lipid bilayers. 68 This special process differs to the adhesion of common charged NPs, possibly prevents the gap narrowing between phospholipid chains and membrane gelation. In summary, the particle attachment induced membrane gelation is not a concern for the cytotoxicity of C 60 NPs.

Conclusions
In summary, fullerene C 60 disrupts the positively charged membranes, but not the negatively charged membranes, suggesting the interaction between C 60 and the cationic moieties in the membrane. The degree of the induced membrane disruption is evaluated by the mass increase on SLB and on SUV layer. SUV layer and SLB have same lipid components, however the value of Dm SUV /Dm SLB aer C 60 deposition is much smaller than the value of A SUV /A SLB (constant k: 0.18-0.28). It indicates some of SUVs have been disrupted by C 60 . The smaller constant k suggests more serious membrane disruption. In this way, the membrane disruption induced by C 60 can be quantied and compared with other types of NPs. It is important to know which type of NPs causes more serious membrane disruption, which is helpful for the biological risk evaluation and the safe application of nanomaterials.

Conflicts of interest
There are no conicts to declare.