Letter pubs.acs.org/NanoLett
Graphene Oxide and Lipid Membranes: Interactions and Nanocomposite Structures Rickard Frost, Gustav Edman Jönsson, Dinko Chakarov, Sofia Svedhem,* and Bengt Kasemo Department of Applied Physics, Chalmers University of Technology, SE-412 96 Göteborg, Sweden S Supporting Information *
ABSTRACT: We have investigated the interaction between graphene oxide and lipid membranes, using both supported lipid membranes and supported liposomes. Also, the reverse situation, where a surface coated with graphene oxide was exposed to liposomes in solution, was studied. We discovered graphene oxide-induced rupture of preadsorbed liposomes and the formation of a nanocomposite, bio-nonbio multilayer structure, consisting of alternating graphene oxide monolayers and lipid membranes. The assembly process was monitored in real time by two complementary surface analytical techniques (the quartz crystal microbalance with dissipation monitoring technique (QCM-D) and dual polarization interferometry (DPI)), and the formed structures were imaged with atomic force microscopy (AFM). From a basic science point of view, the results point toward the importance of electrostatic interactions between graphene oxide and lipid headgroups. Implications from a more practical point of view concern structure−activity relationship for biological health/safety aspects of graphene oxide and the potential of the nanocomposite, multilayer structure as scaffolds for advanced biomolecular functions and sensing applications. KEYWORDS: Graphene oxide, lipid membrane, QCM-D, DPI, AFM, layer-by-layer
G
and inexpensive LbL deposition methodology,5,6 graphene derivatives as well as lipid membranes are still rarely used in LbL-assemblies. Hitherto, the most common materials used in these systems are synthetic polymers,7 biopolymers,8 and inorganic substances.9 A recent study reports on the LbLassembly of graphene oxide and block copolymer micelles.10 In this Letter, we report for the first time on the assembly of a multilayered structure of lipid membranes, which are biologically relevant supramolecular structures, and graphene oxide. These novel nanocomposites might, for example, pave the way for intercalation of (conducting) graphene-derived sheets between (insulating) lipid membranes and related structures. Another aspect relating to the present study is the potential health/safety risk with graphene and graphene oxide in biological systems, and the development of structure− activity relationships to predict nanotoxicity. Although far from the complexity of biological systems, we believe that studies of simple model systems are an important complement to standard toxicity assays. We, and others, have earlier demonstrated the usefulness of early stage screening of nanoparticles, by using lipid membranes as mimics of cell membranes.11,12 Interaction of Graphene Oxide with Positively and Negatively Charged Lipid Membranes. Differently charged supported lipid membranes were prepared on SiO2 surfaces
raphene-based materials are intensely explored for various applications,1 and a novel direction in this field is represented by nanocomposite materials of graphene or carbon nanotubes with lipid membranes. Two significant examples of this development are the use of graphene flakes in miniaturized bioelectronic devices,2 and of carbon nanotubes as scaffolds for photoelectrochemical processes intended for light harvesting.3 The advancement of these and similar applications will be dependent on control over the assembly processes of graphene, or graphene derivatives, and lipid membranes. Toward this end, the aim of the present study was to explore the interaction of graphene oxide (graphene with OH and COOH functional groups)4 with lipid membranes of different compositions. It was of particular interest to investigate if the expected electrostatic attraction between graphene oxide and oppositely charged membranes would dominate over other interactions and lead to (i) flat deposition of graphene oxide onto membranes and (ii) very different interactions with positively and negatively charged lipid membranes. Possible interference from hydrophobic interactions between graphite-like areas of graphene oxide and the interior (hydrophobic) lipid tails of the membrane, or van der Waals interactions, which might be sizable in view of the large number of atoms in a large graphene oxide flake, were hypothesized to potentially lead to more disordered structures, than expected on purely electrostatic grounds. These ideas were tested experimentally and the results were applied to assemble multilayer structures of alternating graphene oxide sheets and lipid membranes in a layer-by-layer (LbL) fashion. Despite the broad application of the versatile © 2012 American Chemical Society
Received: September 7, 2011 Revised: May 2, 2012 Published: June 1, 2012 3356
dx.doi.org/10.1021/nl203107k | Nano Lett. 2012, 12, 3356−3362
Nano Letters
Letter
separate control experiments it is shown that the graphene oxide did not adsorb to the bare SiO2 substrate, which is an expected result since it is negatively charged under the present experimental conditions (see Supporting Information). Because of the seemingly rigid attachment of the graphene oxide to the lipid membrane (low dissipation and similar QCMD responses for all harmonics), the Sauerbrey equation was applied to estimate the amount of deposited mass (macoustic = C*Δfz where C = −17.7 ng/(cm2Hz) and Δfz is the normalized frequency shift at the zth harmonic (here z = 7)).17 The adsorbed mass of graphene oxide at saturation (Figure 3) was found to be 103 ± 11 ng/cm2. This number is close to the theoretical mass of a flat monolayer of graphene oxide, which is 101 ng/cm2. Nevertheless, a certain degree of nonideality in the graphene oxide layers can be expected due to, for example, incomplete coverage, or partly overlapping of graphene oxide flakes prior to (according to specifications >80% of the graphene oxide are single monolayers in suspension) or caused upon adsorption to the surface, despite the electrostatic repulsion between them. Given that the oxygen content of graphene oxide is about 20%, hydration effects may also significantly alter the mass detected by QCM-D, which includes both the actual graphene oxide mass and any liquid acoustically coupled to adsorbed flakes (for example, at the rim of the flakes, or trapped in between the flakes and the membrane). The interpretation of the QCM-D results as discussed above was strengthened by data acquired by two complementary techniques. Atomic force microscopy (AFM) results (see text and Figure 6 below) showed that large parts of the POPC/ POEPC (3:1) lipid membrane surface was covered with a monolayer of graphene oxide but it also revealed gaps in between adjacent graphene oxide flakes and overlapping flakes. The properties of this system were further studied by an optical surface analytical technique to obtain fundamentally different information about the material compared to the viscoelastic (nanomechanical) characterization by QCM-D. Common optical techniques, which are often used for such studies, include surface plasmon resonance (SPR)-based techniques, reflectometry, and, more recently, the dual polarization interferometry (DPI). Here DPI experiments were performed, as this technique offers the possibility to explore the anisotropic properties of the studied materials. In particular DPI is useful for studies of lipid membranes,18 and an attractive approach to address the nonrandom orientation of both the lipid molecules and the graphene oxide flakes in our study. The DPI technique is based on two parallel waveguides, where the upper side of the top waveguide is the sensor surface (silicon oxinitride). Changes in refractive index (due to the polarizability of adsorbed molecules) close to this surface are measured as changes in the position of the fringes in the interference pattern generated beyond the end of the waveguides, using two orthogonal polarizations of the incident light (transverse electric (TE) and transverse magnetic (TM)). DPI has previously been used to study optical anisotropy of lipid membranes.18 For isotropic materials, the TM response is always larger than the TE response due to the differences in field intensity profiles. In Figure 2, DPI data for the formation of a POPC/POEPC (3:1) membrane and subsequent addition of graphene oxide are shown (see Supporting Information for experimental details). In the figure, the anisotropic properties of both the lipid membrane and the layer of graphene oxide are evident through the difference between the TM and TE
using extruded liposomes (mean diameter of 80−90 nm) composed of different lipids. 1-Palmitoyl-2-oleyl-sn-glycero-3phosphocholine (POPC), 1-palmitoyl-2-oleyl-sn-glycero-3phospho-L-serine (POPS), and 1-palmitoyl-2-oleyl-sn-glycero3-ethylphosphocholine (POEPC) were used to prepare negatively charged POPC/POPS (3:1) liposomes (ζ-potential: −26 ± 1.2 mV) and positively charged POPC/POEPC (3:1) liposomes (ζ-potential: 22 ± 0.8 mV), respectively.13 Under appropriate conditions, these liposomes adsorb to the SiO2 surface and spontaneously rupture to form fluid supported lipid membranes containing about 25% of the charged lipid, according to a well-established protocol and mechanistic scenario.14,15 The formation of supported lipid membranes, as well as the subsequent exposure of these model membranes to an aqueous suspension of graphene oxide (0.5−5 μm sheets with an oxygen content of 20%, according to specifications), were monitored by the quartz crystal microbalance with dissipation monitoring technique (QCM-D) (experimental details are provided in the Supporting Information). In a QCM-D experiment, the frequency shift, Δf, efficiently measures the accumulation of mass onto the sensor surface and the D factor (energy dissipation or damping of the shear oscillation of the QCM-D sensor), represented as ΔD, reflects the structural/viscoelastic properties of the adlayer. The QCM-D signals, resulting from exposure of the lipid membranes to graphene oxide, are shown in Figure 1. When
Figure 1. QCM-D data recorded during exposure of supported lipid membranes, POPC/POPS (3:1) and POPC/POEPC (3:1), to a graphene oxide suspension. Blue lines represent frequency shifts, Δf, (mass uptake) and red lines represent dissipation shifts, ΔD.
the negatively charged POPC/POPS (3:1) membrane surface is exposed to graphene oxide, the frequency and the dissipation responses (Δf and ΔD) show that no adsorption of graphene oxide occurs, which we attribute to the electrostatic repulsion between the lipid membrane headgroups and graphene oxide. This is in accordance with the measured ζ-potential of the aqueous dispersion of graphene oxide (−56 ± 1.1 mV, pH 4), which is similar to previously reported values.16 In contrast, on the positively charged POPC/POEPC (3:1) membrane, there is a clear mass uptake signaled by a monotonic decrease of the frequency, which saturates after about 10 min at a frequency shift of around -6 Hz. The magnitude of this shift and the fact that no dissipation shift is observed suggest that the graphene oxide flakes adsorbs flat on the membrane and covers most of the surface. The absence of any ΔD response indicates that the flakes are sufficiently rigidly attached not to slip on the lipid surface during the oscillatory motion of the sensor surface. The interaction is irreversible upon rinsing with buffer, and in 3357
dx.doi.org/10.1021/nl203107k | Nano Lett. 2012, 12, 3356−3362
Nano Letters
Letter
refractive index (n) and the refractive index increment (dn/dc) of graphene oxide were estimated from literature data (see Supporting Information for details), or (ii) the thickness of the graphene oxide layer was determined from the QCM-D data using Voight-based modeling.19 The two approaches yield similar results. In Figure 3A, it is evident that the calculated masses of a POPC/POEPC (3:1) membrane from DPI and QCM-D data are very similar. However, the mass of the graphene oxide layer as determined by QCM-D (103 ± 11 ng/ cm2) is larger than the corresponding mass extracted from the DPI data (70 ± 7 ng/cm2), leading to an estimation of the water content of the graphene oxide layer of around 32%. Such a difference in mass is commonly seen when comparing QCMD data with optical data, for example, for biomolecules (the water content of a dense protein layer is typically 55%20) due to the difference in sensing principle. In contrast to QCM-D, which is an acoustic method sensitive not only to the adsorbed material but also to acoustically coupled solvent (water), DPI is an optical technique where the solvent associated with the sensor surface does not generate a response (the same holds for SPR, ellipsometry, and reflectrometry). The present result shows that the layer of graphene oxide is more hydrated than the lipid membrane. Since the optical mass is smaller than the theoretical mass of a monolayer of graphene oxide, this result supports the above interpretation of the QCM-D results, that there are gaps between adjacent flakes of graphene oxide. The birefringence of the lipid membrane and the layer of graphene oxide are shown in Figure 3B. It is evident that the lipid membrane has a larger polarizability perpendicular to the sensor surface as opposed to the layer of graphene oxide. This data is interpreted as that the graphene oxide is ordered in plane with (parallel to) the sensor surface. LbL Structures of Graphene Oxide and Lipid Membranes. The previous section demonstrated preparation of a structure consisting of a positively charged POPC/POEPC (3:1) lipid membrane on the SiO2 substrate, covered on the liquid side by a saturation layer of graphene oxide, as shown in Figures 1 and 2. Further sequential additions of POPC/POEPC (3:1) liposomes and graphene oxide to the initial layers of lipid membrane and graphene oxide generates an interesting multilayered structure, as described in detail below and in Figure 4. QCM-D Observations. The QCM-D data for the sequential build-up of the multilayered graphene oxide−lipid membrane structure are presented in Figure 5. At the second injection of liposomes, starting at step IV in Figure 5, there is a large negative shift in the frequency, signaling a quite large mass uptake, and a large increase in dissipation, corresponding to the formation of a softer layer on the QCM-D sensor surface. We attribute these results to the formation of about one monolayer of intact liposomes on the previously formed graphene oxide layer. This is similar to what has been observed previously for intact liposome adsorption on inorganic surfaces like TiO221 and Au,22 where the surfaces interact strongly enough with the liposomes to adsorb them irreversibly but not strongly enough to induce rupture. In this respect, the liposome interaction with graphene oxide is qualitatively similar to TiO2 and Au. The reason for the large changes in Δf (−50 ± 2 Hz) and ΔD (9 ± 2 × 10−6) is that the intact liposomes carry water in their interior, causing a large mass adsorption, which is the sum of the lipid and water masses. The “soft” structure of adsorbed liposomes, much softer than a lipid bilayer, is responsible for
Figure 2. DPI data obtained (I−II) during the formation of a supported POPC/POEPC (3:1) membrane onto the waveguide sensor surface, (*) the subsequent liquid exchange from PBS to water, and (III) the addition of the graphene oxide suspension.
responses. The lipid membrane generates an enhanced TM response relative to the TE response compared to the expected results for an isotropic lipid film due to the larger polarizability of the lipid molecules in the membrane in the direction normal to the sensor surface, similarly to results obtained in previous studies.18 In contrast, the graphene oxide generates an enhanced TE response relative to the TM response, which is a quite uncommon situation. This result supports the conclusion that the graphene oxide adsorbs flat onto the lipid membrane. From the DPI data, the mass and the birefringence (that is, the difference between nTM and nTE) of the POPC/POEPC (3:1) membrane and the layer of graphene oxide were calculated (Figure 3). Two different approaches were used to calculate the mass and birefringence of the latter layer, (i) the
Figure 3. (A) Calculated masses of a POPC/POEPC (3:1) membrane and the subsequently adsorbed layer of graphene oxide based on QCM-D and DPI data. Calculations were performed according to approach (i) as discussed above. Similar results were obtained by assuming a layer thicknesses based on modeling of the QCM-D data (approach (ii)). (B) Birefringence of the lipid membrane and the layer of graphene oxide, determined by DPI. Experiments were performed in three (QCM-D) or four (DPI) replicates. 3358
dx.doi.org/10.1021/nl203107k | Nano Lett. 2012, 12, 3356−3362
Nano Letters
Letter
liquid (causing a mass loss that by far exceeds the added mass of the rupture-inducing adsorbed graphene oxide) and a second lipid membrane is formed. The high dissipation value returns to a low value, since there are no longer soft, intact liposomes on the surface, after they have ruptured and fused to a membrane. A similar process occurs for some liposomes (for example, POPC) when they reach a critical coverage on, for example, SiO2 surfaces, but in this case a one-sided surface interaction is enough to induce rupture.14,15 A question that arises is where the graphene oxide flakes that induce the liposome rupture go after the ruptures have happened. Most likely they form a second adsorbed graphene oxide layer on the second lipid membrane (originating from the rupturing liposomes), because the situation is very similar to the first adsorption step of graphene oxide on the first lipid membrane. However, in this case liposome rupture and graphene oxide adsorption occur simultaneously rather than sequentially. The frequency and dissipation responses due to the sum of the second lipid membrane and the second layer of graphene oxide (Δf and ΔD between steps I and IV in Figure 5) are larger compared to the corresponding values for the first two layers (the Δf and ΔD values between steps IV and VI in Figure 5). There are several possible explanations for these larger values. The three most likely reasons are the following: (i) The second lipid membrane plus its adsorbed layer of graphene oxide is a more hydrated and softer structure. If this were the case the larger Δf and ΔD responses would be due to more associated/trapped solvent than in the first two layers (a first lipid membrane and a first graphene oxide layer). (ii) It is not unlikely that the second lipid membrane is less perfect than the first one, since the rupture of liposomes, induced by the graphene oxide, may be incomplete. Furthermore, we suspect that there are areas of the first membrane that are not covered by graphene oxide, as discussed above. In this case, the larger Δf and ΔD responses would be due to some intact liposomes coexisting with the second membrane. (iii) When graphene oxide flakes land on the preadsorbed monolayer of intact liposomes and induce their rupture, excess lipid material will return to the bulk phase or stay around as excess lipid membrane patches. This lipid material can adsorb on top of the second layer of graphene oxide, that is, on its side facing the bulk liquid, yielding a larger mass uptake than for just one sequence of lipid membrane and graphene oxide. The latter would actually correspond to the beginning of the buildup of the third lipid membrane, which in turn would enable some further adsorption of graphene oxide and so on. The net effect would be a mass uptake, which is larger than for formation of just a second lipid membrane + graphene oxide layer, as was also observed in the experiment. Note that this process was not possible upon the first addition of graphene oxide, since the starting point in that case was a completed lipid membrane, with no intact liposomes and no surplus lipid mass. The relative importance of these three possible mechanisms was elucidated by AFM analyses. AFM data showed that there are multiple layers on the surface and that the number of layers exceeds the sum of the number of cycles of added liposomes and graphene oxide, that is, scenario (iii) is supported by the AFM data as at least one contributing mechanism. These data are presented in Figure 6C and described in more detail in the related text. When continuing the experiment by again adding liposomes of the same lipid composition as before, followed by addition of
Figure 4. Schematic representation of the formation of a POPC/ POEPC (3:1) lipid membrane and subsequent adsorption of graphene oxide. Further addition of POPC/POEPC (3:1) liposomes and graphene oxide results in a multilayered structure of these materials, as deduced from the results shown in Figure 5. The figure is not drawn to scale.
Figure 5. QCM-D data showing the sequential build-up of a multilayered structure of POPC/POEPC (3:1) membranes and graphene oxide, starting from (I) the adsorption of liposomes to a silica surface and (II) the spontaneous formation of a supported lipid membrane. Next, (III) graphene oxide is added, followed by (IV and VI) additional injections of liposomes and (V and VII) of graphene oxide. The additions of graphene oxide and liposomes are preceded by a (*) liquid exchange to water and PBS respectively. These events generate responses in both Δf and ΔD due to the change in liquid bulk composition. (These liquid exchanges (marked *) generated shifts in both Δf and ΔD due to the change of bulk composition. However, when replacing the PBS with water after adsorption of liposomes, an osmotic gradient is formed across the membrane of the liposomes. This gradient may cause transient variations in the amount of coupled water, processes that also generate responses in Δf and ΔD.)
the large increase in dissipation, as previously shown by combined QCM-D and AFM studies.22 The subsequent addition of graphene oxide to the liposome covered graphene oxide surface, step V in Figure 5, generates a surprising result, namely an increase in frequency, that is, a decrease in Δf and thus a net mass loss, and a simultaneous large decrease of the dissipation (negative ΔD change). These two observations constitute strong evidence that the graphene oxide induces rupture of the preadsorbed liposomes. It should be noted that the average size of the graphene oxide flakes is much larger than the liposome diameter and results obtained with smaller flakes might be different. In the liposome rupture process, the water inside the liposomes is released to the bulk 3359
dx.doi.org/10.1021/nl203107k | Nano Lett. 2012, 12, 3356−3362
Nano Letters
Letter
electron microscopy (not shown). Interestingly, a topography investigation of the image in Figure 6B reveals the formation of lamellar structures with steps of different heights. This is visualized in the histogram in Figure 6C that shows that some heights on the surface occur more frequently than others. The histogram shows at least six peaks, that is, six different heights that are frequent and well separated from each other. We attribute these specific heights to multiples of the thickness of a lipid membrane with an adsorbed graphene oxide layer. For comparison, a lipid membrane is around 5 nm thick, while the thickness of a mono-, bi-, and trilayer of graphene oxide is 1.6, 2.6, and 3.6 nm, respectively.23 In contrast to graphite intercalation compounds24 in which the planes of carbon atoms are separated by a few angstroms, the graphene flakes in our multilamellar structures are separated at the nanometer scale. The fact that there are more layers detected in the histogram than cycles of added liposomes and graphene oxide in the experiment was explained above as an overspill of lipids to adjacent sheets of graphene oxide upon liposome rupture (mechanism (iii)), a process that enables further adsorption of graphene oxide. In this way, each cycle of added liposomes and graphene oxide gives on average rise to more than a single pair of layers of these materials. Additional Comments on the Interaction between Graphene Oxide and Lipid Membranes. The present study provides insight into the nature of the interaction between lipid membranes and graphene oxide. Under the given experimental conditions, the driving force for adsorption of negatively charged flakes of graphene oxide to positively charged lipid membranes seems to be governed primarily by electrostatics. We cannot exclude a significant contribution from other factors like H-bonding to functional groups (for example, OH or COOH groups) on graphene oxide, or van der Waals interactions. However, the electrostatic interactions are likely dominant, based on the absence of adsorption of graphene oxide on negatively charged lipid membranes. An interesting extension of the present experiments would be to gradually decrease the positive charge of the lipid membrane to investigate if this would alter the way the graphene oxide flakes bind to the membrane, for example, if there is some low level of positively charged lipids at which the adsorption becomes reversible and/or where the multilayer LbL structure would be built by alternating layers of adsorbed intact vesicles and graphene oxide rather than with lipid membranes in between the graphene oxide flakes. An experiment in this direction, using liposomes containing only 10% POEPC, is presented as Supporting Information. Briefly, the results indicate that too low POEPC-content in the liposomes lead to incomplete rupture when exposed to graphene oxide. There are two additional features in the present work worth highlighting. First, it was observed that graphene oxide induces rupture of preadsorbed liposomes on a graphene surface. This observation can most likely be generalized to the possibility of preparation of supported lipid membranes from liposomes on other surfaces than those that do not spontaneously promote such formation. Indeed, in a separate experiment, graphene oxide was seen to also induce rupture of intact POPC/POEPC (3:1) liposomes adsorbed on a gold substrate (see Supporting Information). In addition, graphene oxide may open up a way to form lipid membranes of other lipid compositions, than those where the liposomes spontaneously rupture and fuse into a supported lipid membrane. Today, research within this area is usually focused on identifying fusogenic agents such as peptides
Figure 6. (A) AFM image of graphene oxide adsorbed on a supported POPC/POEPC (3:1) membrane in water (z-range: 29 nm). (B) Multilayered structure of graphene oxide and POPC/POEPC (3:1) membranes on a QCM-D crystal that was dried and imaged after step VII in Figure 3 (z-range: 81 nm). (C) Histogram of the height profile in image (B). The zero level in the histogram corresponds to the lowest point in the image. The z-range in image A and B refers to the total range in the z-dimension, that is, the height difference between the lowest (darkest) and the highest (brightest) point in the image.
graphene oxide, the result is qualitatively similar to the second cycle. Liposomes adsorb intact to the surface and rupture upon the subsequent addition of graphene oxide. However, in this cycle the QCM-D responses are even larger than before, both after adsorption of liposomes and after addition of graphene oxide. Again, we attribute these large responses to scenario (iii) as described above. These results prove that that the formed multilayered structure is terminated by graphene oxide, that is, graphene oxide stays adsorbed to the lipid membrane after liposome rupture. In separate experiments, the sequential assembly of alternating lipid membrane and graphene oxide structures was studied by DPI. The analysis of these data supports the interpretation of the QCM-D data (DPI data are given as Supporting Information). AFM Observations. As described above, graphene oxide selectively adsorbs to the positively charged POPC/POEPC (3:1) membrane. The adsorbed material on this type of membrane, that is, after stage III in Figure 5, was visualized by AFM. In such an AFM image, a partial overlap of some graphene oxide sheets is visible (Figure 6A). For the graphene oxide terminated structures, it was suggested by the surface wettability properties that the multilayers remained on the surface as the samples were dried (see further below). The AFM image of a sample obtained after three cycles of liposomes and graphene oxide had been added, that is, after stage VII in Figure 5, followed by evaporation of the solvent (water) at room temperature, is shown in Figure 6B. This image was recorded on a QCM-D sensor that was demounted after the build-up of the nanocomposite structure on the sensor surface. Similar images were obtained in liquid (requiring a manual sample preparation procedure) by AFM and in air by scanning 3360
dx.doi.org/10.1021/nl203107k | Nano Lett. 2012, 12, 3356−3362
Nano Letters
■
that can induce rupture and fusion.25 The present observations identify graphene oxide as an interesting “fusogenic” alternative. Mechanistically, it must be the additional deformation of the liposomes, when they have a graphene oxide surface on two rather than on only one side, which induces the rupture. In some applications, removal of the graphene oxide after the membrane has been formed would be desirable. However, we have found that increasing the ionic strength or lowering the pH does not easily remove the graphene oxide. Optimization of the interaction forces is likely needed in order to establish a protocol for the removal of the adsorbed graphene oxide. Second, it was noted that the prepared samples did not dewet when exposed to air, which is otherwise commonly observed for plain lipid membranes. Accordingly, AFM analyses revealed a well organized multilayered structure both in liquid and on a dried QCM-D crystal (Figure 6), that is, the multilayered structure was largely maintained even after drying. These results indicate that the presence of graphene oxide preserves the lipid membrane upon drying. Previous work in this area has suggested that covering a lipid membrane with polyethylene glycol26 or proteins27 will induce resistance to the air−water interface. In a different approach, the lipids in the membrane have been cross-linked to enhance its resistance to detergent.28 It should be noted that the LbL growth of alternating supported lipid membrane and graphene oxide layers was not perfect and that some intact liposomes and some more than double layer structures are formed for each pair of liposome and graphene oxide exposures. However, the process was also not optimized. Improvements approaching a more ideal LbL growth seem probable. One such step would involve adsorbing less than a saturation layer of intact liposomes on the previously formed graphene oxide layer, since this would reduce the excess amount of lipid material, beyond what is needed for just a lipid membrane. We also speculate that it might be possible to chemically reduce the graphene oxide in order to restore the conductive properties of graphene, after adsorption on the membrane. If this were possible one could produce conducting graphene layers intercalated between insulating lipid membranes. In this context, we note that a recent theoretical study suggests that graphene flakes would prefer to be incorporated inside the membrane, rather than adsorbed to the membrane surface.29 A second possible extension of the current work is to first cover graphene with lipids before suspending it in an aqueous medium and depositing the lipid-graphene assemblies on a surface. Concluding Remarks. In conclusion, this letter presents sequential build-up of multilayered lipid membranes separated by graphene oxide. It is shown that anionic graphene oxide adsorbs selectively to positively charged lipid membranes, which subsequently enables further adsorption of intact liposomes. Furthermore, graphene oxide induces rupture of adsorbed liposomes on graphene oxide (and on gold) and forms additional stacked lipid membranes on the former surface. The final surface assembly is composed of several layers of graphene oxide with lipid membranes on each side. After drying this assembly, where the last exposure was made to graphene oxide, it was shown that a well-organized structure remained at the surface. The presented results open up for new preparation procedures to intercalate structures of graphene oxide, and potentially of graphene, between lipid membranes.
Letter
ASSOCIATED CONTENT
S Supporting Information *
Experimental methods and additional QCM-D and DPI data are available. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: sofi
[email protected]. Telephone: +46 (0)31 772 34 28. Fax: +46 (0)31 772 31 34. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work was financially supported by the Swedish Environmental Research Council FORMAS (NanoSphere, Project Number 2009-1696), the Swedish Foundation for Strategic Research (SSF, RAM08-0109, methamaterials), and the Swedish Research Council (VR, Project Number 621-20074375). The Farfield Group Ltd. (Manchester, U.K.) is gratefully acknowledged for the access to a DPI instrument and especially Dr. Marcus Swann for supporting the DPI data analyses.
■
REFERENCES
(1) Compton, O. C.; Nguyen, S. T. Small 2010, 6 (6), 711−723. (2) Ang, P. K.; Jaiswal, M.; Lim, C.; Wang, Y.; Sankaran, J.; Li, A.; Lim, C. T.; Wohland, T.; Barbaros, O.; Loh, K. P. ACS Nano 2010, 4 (12), 7387−7394. (3) Ham, M. H.; Choi, J. H.; Boghossian, A. A.; Jeng, E. S.; Graff, R. A.; Heller, D. A.; Chang, A. C.; Mattis, A.; Bayburt, T. H.; Grinkova, Y. V.; Zeiger, A. S.; Van Vliet, K. J.; Hobbie, E. K.; Sligar, S. G.; Wraight, C. A.; Strano, M. S. Nature Chem. 2010, 2 (11), 929−936. (4) Dreyer, D. R.; Park, S.; Bielawski, C. W.; Ruoff, R. S. Chem. Soc. Rev. 2010, 39 (1), 228−240. (5) Ariga, K.; Hill, J. P.; Ji, Q. Phys. Chem. Chem. Phys. 2007, 9 (19), 2319−2340. (6) Tang, Z.; Wang, Y.; Podsiadlo, P.; Kotov, N. A. Adv. Mater. 2006, 18 (24), 3203−3224. (7) Bertrand, P.; Jonas, A.; Laschewsky, A.; Legras, R. Macromol. Rapid Commun. 2000, 21 (7), 319−348. (8) Crouzier, T.; Boudou, T.; Picart, C. Curr. Opin. Colloid Interface Sci. 2010, 15 (6), 417−426. (9) Srivastava, S.; Kotov, N. A. Acc. Chem. Res. 2008, 41 (12), 1831− 1841. (10) Hong, J.; Kang, Y. S.; Kang, S. W. Ind. Eng. Chem. Res. 2011, 50 (6), 3095−3099. (11) Frost, R.; Grandfils, C.; Cerda, B.; Kasemo, B.; Svedhem, S. J. Biomater. Nanobiotechnol. 2011, 2 (2), 180−192. (12) Nel, A. E.; Madler, L.; Velegol, D.; Xia, T.; Hoek, E. M. V.; Somasundaran, P.; Klaessig, F.; Castranova, V.; Thompson, M. Nat. Mater. 2009, 8 (7), 543−557. (13) Kunze, A.; Svedhem, S.; Kasemo, B. Langmuir 2009, 25 (9), 5146−5158. (14) Reimhult, E.; Höök, F.; Kasemo, B. Langmuir 2003, 19 (5), 1681−1691. (15) Richter, R. P.; Bérat, R.; Brisson, A. R. Langmuir 2006, 22 (8), 3497−3505. (16) Si, Y.; Samulski, E. T. Nano Lett. 2008, 8 (6), 1679−1682. (17) Sauerbrey, G. Z. Phys. 1959, 155 (2), 206−222. (18) Mashaghi, A.; Swann, M.; Popplewell, J.; Textor, M.; Reimhult, E. Anal. Chem. 2008, 80 (10), 3666−3676. (19) Voinova, M. V.; Jonson, M.; Kasemo, B. Biosens. Bioelectron. 2002, 17 (10), 835−841. (20) Edvardsson, M.; Svedhem, S.; Wang, G.; Richter, R.; Rodahl, M.; Kasemo, B. Anal. Chem. 2009, 81 (1), 349−361. 3361
dx.doi.org/10.1021/nl203107k | Nano Lett. 2012, 12, 3356−3362
Nano Letters
Letter
(21) Reimhult, E.; Höök, F.; Kasemo, B. J. Chem. Phys. 2002, 117 (16), 7401−4. (22) Reimhult, E.; Zäch, M.; Höök, F.; Kasemo, B. Langmuir 2006, 22 (7), 3313−3319. (23) Mkhoyan, K. A.; Contryman, A. W.; Silcox, J.; Stewart, D. A.; Eda, G.; Mattevi, C.; Miller, S.; Chhowalla, M. Nano Lett. 2009, 9 (3), 1058−1063. (24) Dresselhaus, M. S.; Dresselhaus, G. Adv. Phys. 1981, 30 (2), 139−326. (25) Cho, N. J.; Cho, S. J.; Kwang, H. C.; Glenn, J. S.; Frank, C. W. J. Am. Chem. Soc. 2007, 129 (33), 10050−10051. (26) Albertorio, F.; Diaz, A. J.; Yang, T.; Chapa, V. A.; Kataoka, S.; Castellana, E. T.; Cremer, P. S. Langmuir 2005, 21 (16), 7476−7482. (27) Holden, M. A.; Jung, S. Y.; Yang, T.; Castellana, E. T.; Cremer, P. S. J. Am. Chem. Soc. 2004, 126 (21), 6512−6513. (28) Subramaniam, V.; Alves, I. D.; Salgado, G. F. J.; Lau, P. W.; Wysocki, R. J., Jr; Salamon, Z.; Tollin, G.; Hruby, V. J.; Brown, M. F.; Saavedra, S. S. J. Am. Chem. Soc. 2005, 127 (15), 5320−5321. (29) Titov, A. V.; Král, P.; Pearson, R. ACS Nano 2010, 4 (1), 229− 234.
3362
dx.doi.org/10.1021/nl203107k | Nano Lett. 2012, 12, 3356−3362