Controlling fluorescence quenching efficiency by graphene oxide in supported lipid bilayers using SiO₂ layer fabricated by atomic layer deposition

Lipid bilayers are the fundamental structure of biomembranes such as cell membranes, which are reaction fields for transportation of materials, information, and energy into and out of cells. These reactions deeply relate to neurological, metabolic and infection diseases, thus they are regarded as emphasized research targets in the fields of medicine and drug discovery. ¹⁾ Lateral and vertical distribution of lipids and proteins in lipid bilayers are essential information to understand the membrane reactions on the molecular level. ^2–4) For this purpose, artificial lipid bilayer systems such as vesicles and supported lipid bilayers (SLBs) have been adopted. ^5–11) The SLB is an artificial lipid bilayer formed at interfaces between aqueous solutions and hydrophilic solid substrates, existing in the vicinity of approximately 1 nm to the substrate surface. Therefore the SLB system has a high technical affinity with functionalized surfaces and solid sensors. ^10–18)

Graphene oxide (GO) is a chemical derivative of graphene, which is a two-dimensional carbon nanomaterial, modified with hydrophilic functional groups e.g. hydroxy, carboxy, and carbonyl groups. ^19–21) These oxygen functional groups make the hydrophobic pristine graphene so hydrophilic as to be available in aqueous systems. Therefore GO has been utilized in various biological applications. ^21–25) GO works as a unique quencher for fluorophores working efficiently and independently of the emission wavelength of the fluorophores. ²⁰⁾ The quenching by GO occurs through the Förster resonance energy transfer (FRET), and its efficiency (E) decreases rapidly with increasing distance between fluorophore and GO, as in the Förster equation with a two-dimensional acceptor. ^26,27) Fluorescence lifetime study on DNA molecules that are labeled with a fluorescence dye on the different positions demonstrated that the quenching efficiency of GO is expressed as

$$\begin{eqnarray}&&E=\displaystyle \frac{1}{1+{\left(\displaystyle \frac{r}{{R}_{0}}\right)}^{4}},\end{eqnarray} \tag{ 1 }$$

where r is the distance between GO and the fluorescent dye, and R₀ is the Förster distance, at which E becomes 0.5, and that estimated R₀ is 7.5 nm. ²⁸⁾ The quenching by GO has been applied for sensing hybridization of nucleic acids and conjugation of proteins. ^29–33)

SLBs have been formed on graphene and GO applying their functions to lipids and proteins in lipid bilayers. ^24,34–36) The thickness of a lipid bilayer is 4–5 nm, ³⁷⁾ and thus SLBs on GO are in the effective range of the quenching by GO. The fluorescence from dye-labeled lipids in SLBs on GO are quenched effectively, resulting in the dark shapes of the GO flakes under the fluorescence-labeled SLB in wide-field fluorescence observation. ^38,39) The efficiency is so effective that fluoresce from each dye-labeled lipid is not detected in fluoresce single molecule observation. ⁴⁰⁾ Single particle observation of a quantum dot (Qdot) that is conjugated on SLB on GO demonstrated that each Qdot is visible even in the presence of GO under SLB and that the fluorescence intensity of the Qdots on single and double SLBs varies depending on the layer number. ⁴¹⁾

We aim to control the efficiency of GO quenching to SLB by manipulating the distance between GO and dye-labeled lipids for developing a new method to determine the position of molecules within SLBs in the z-direction from the fluorescence intensity. In this study, we deposited a SiO₂ layer on GO by the atomic-layer deposition (ALD) method as a spacer between GO and SiO₂. SiO₂ films are fabricated with the precision of 0.08 nm by the ALD method using tris(dimethylamino)silane (TDMAS) as a precursor and plasma-excited water vapor. ⁴²⁾ Recently, we presented that the 5–20 nm thick ALD-SiO₂ layers on GO retained sufficiently flat for SLB formation, and the fluorescence intensity of SLB varied depending on the thickness of the ALD-SiO₂ layer. ⁴³⁾ We investigated the dependence of the fluorescence intensity on the distance between the SLB and GO that was varied with the ALD-SiO₂ thickness in detail. The ALD-SiO₂ layer attenuated the quenching efficiency of GO and achieved the fluorescence single molecule observation of dye-labeled lipids in SLBs that were still in the effective range of the fluorescence quenching by GO. The difference in the brightness of the dye-labeled lipid in the upper and lower leaflets of SLB was also demonstrated.

A GO suspension was prepared through chemical exfoliation by the modified Hummer's method. ^44,45) Graphite particles (Ito Graphite Co., Ltd., Kuwana, Japan) were oxidized in two steps with peroxydisulfuric acid and potassium permanganate in sulfuric acid. Single-layered GO flakes were obtained after the oxidized graphite particles were dispersed into pure water. ^38,39) Residual graphite particles and multi-layered GO flakes were removed by centrifugation. The GO prepared in the present protocol gave typical Raman and XPS spectra as shown in the previous studies. ^17,18)

Thermally oxidized SiO₂/Si substrates with a 90 nm thick SiO₂ layer were cleaned in piranha solution (1:3 v/v solution of 30% H₂O₂ and sulfuric acid) at 180 °C for 30 min followed by sonication in 0.02 M KOH aqueous solution for 10 min. Spin coating of GO was performed at the maximum rotation speed of 6500 rpm. For chemically modifying the substrate surface, we immersed the piranha-cleaned SiO₂/Si substrate in an ethanol solution of (3-aminopropyl)triethoxysilane (APTES) (Tokyo Chemical Industry Co., Ltd., Tokyo, Japan) (5:95 v/v) at 25 °C for 60 min and rinsed in ethanol and pure water for 5 min each.

The fabrication of the SiO₂ layer by the ALD method was performed with the protocols in the previous study. ⁴²⁾ Briefly, the cycle of TDMAS injection for 20 s and the introduction of plasma-excited water vapor for 10 min were repeatedly performed in the ALD chamber. The process gas in the ALD chamber was exchanged by evacuation for 30 s before injecting TDMAS or plasma-excited water–vapor. During the process, the substrate temperature was maintained at 25 °C. The growth rate the ALD-SiO₂ layer was 0.08 nm cycle⁻¹. Details about the ALD system are described elsewhere. ⁴²⁾

The reagents 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), and 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (Rb-DPPE, Ex/Em: 560/583 nm) were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL, USA), and DPPE-N-ATTO532 (ATTO-DPPE, Ex/Em: 532/553 nm) was purchased from Thermo Fisher Scientific Inc. (Waltham, MA, USA). These lipid reagents were used as received, without further purification. The chloroform solution of DOPC or DMPC was mixed with that of a dye-labeled lipid (0.2 mol% Rb-DPPE for wide-field fluorescence observation, or 10⁻⁵ to 10⁻⁷ mol% ATTO-DPPE for single molecule observation) and dried with a nitrogen flow. The vacuum-dried lipid film was suspended in a buffer solution (100 mM KCl, 25 mM HEPES/NaOH [pH 7.4]) through agitation and sonication processes to prepare the suspensions of unilamellar vesicles. The ALD-deposited SiO₂ surface was cleaned and hydrophilized with a UV ozone cleaner with a Mercury lamp (ProCleaner Plus, BioForce Nanosciences, Inc., Ames, IA, USA) in air for 15 min, followed by incubation in the vesicle suspension for 60 min to prepare SLB by the vesicle fusion method. ^10,11) The buffer solution containing CoCl₂·6H₂O was added to SLB for quenching ATTO at the final Co²⁺ concentration of 1 mM.

Wide-field fluorescence observation was performed using an epi-fluorescence microscope (epi-FM) (BX51WI, Olympus Inc., Tokyo, Japan) equipped with a 60 × water immersion objective (N.A. = 1.00) and a mirror unit U-MWIG3 (Ex: 530–550 nm, Em > 575 nm, Olympus). The epi-FM images were recorded using a CMOS camera (DS-Qi2, Nikon Solutions Co., Ltd., Tokyo, Japan) and the fluorescence intensities were analyzed with NIS Elements software (Nikon). Fluorescence single molecule observation was performed using an inverted fluorescence microscope (IX-71, Olympus) equipped with a 100 × oil immersion objective (N.A. = 1.45). Excitation light from 532 nm DPSS laser was introduced to the Si substrate with a diagonal illumination setup, ^41,46,47) and the fluorescence images were recorded with an EM-CCD camera (iXon DU-897, Andor Technology, Ltd., Belfast, UK) at the frame rate of 30 frames s⁻¹ (fps). The trajectory coordinates and fluorescence intensity of each molecule were obtained from the movies using an ImageJ (NIH, http://imagej.nih.gov/ij/) and the Particle Tracker plug-in on the basis of the theory and protocol developed by Sbalzarini and Koumoutsakos. ⁴⁸⁾ Morphologies of the sample surfaces were observed with an atomic force microscope (AFM) (SPM9700-HT, Shimadzu Corp., Kyoto, Japan) in the intermittent contact mode. Epi-FM, fluorescence single molecule, and AFM observations were performed at 25 °C unless otherwise stated.

Figure 1 shows the preparation processes of the ALD-SiO₂ layers on SiO₂/Si substrates covered with GO flakes. Figure 2 shows fluorescence images of thermally oxidized SiO₂/Si substrates covered with submonolayer GO flakes (subGO/SiO₂/Si) [Fig. 2(a)] and full coverage GO flakes (fullGO/SiO₂/Si) [Fig. 2(b)]. Bright regions correspond to GO flakes, because single-layered GO shows broad fluorescence emission around 600 nm. ^20,22) To prepare subGO/SiO₂/Si, the GO suspension was deposited by spin coating on a piranha-cleaned SiO₂/Si substrate [Fig. 1(a)]. GO flakes randomly distributed and some of the flakes partly overlapped, and the bare SiO₂ surface was exposed between the GO flakes [Fig. 2(a)]. A piranha-cleaned SiO₂/Si substrate was modified with APTES prior to the deposition of GO [Fig. 2(b)] to prepare fullGO/SiO₂/Si. ⁴⁹⁾ The GO suspension was drop-cast on the APTES-modified SiO₂/Si substrate, and the substrate was sonicated in pure water for 15 min [Fig. 1(c)]. The GO flakes covered the substrate surface almost completely [Fig. 2(b)] because of the electrostatic attraction between the negatively charged GO flakes and the positively charged amino group of APTES, whereas the excess GO flakes overlapping as the second layer were removed by sonication. ⁴⁹⁾ Additional SiO₂ layers of 5–20 nm thickness were deposited on the subGO/SiO₂/Si and fullGO/SiO₂/Si substrates by the ALD method (Fig. 1).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figures 3(a)–3(c) show the AFM topographies of the surfaces of the ALD-SiO₂/subGO/SiO₂/Si substrate with the ALD-SiO₂ thickness (d) of 5, 10, and 20 nm, respectively. We recognized the steps of approximately 1 nm between the regions with and without GO flakes on the ALD-SiO₂ surface at d = 5 and 10 nm [Figs. 3(a) and 3(b)], while we did not find the step on the ALD-SiO₂ surface at d = 20 nm [Fig. 3(c)]. A single layered GO flake on the thermally oxidized SiO₂/Si gives a thickness of approximately 1.5 nm that includes the thickness of a confined water layer between the GO flake and substrate ^38,50) The thickness of the ALD-SiO₂ layer on GO was close to, but not identical to that on the bare SiO₂ surface. In the ALD process, TDMAS adsorbs on the substrate surface via surface hydroxy groups. ⁵¹⁾ It is reasonably proposed that TDMAS molecules also adsorbed on GO via the hydroxy group in its hydrophilic part. The hydroxyl radical for the oxidation step of TDMAS provides additional hydroxy groups on GO. ⁵²⁾ The 1 nm high step on the ALD-SiO₂ layer due to GO flakes became blurred after repeated ALD cycles up to d = 20 nm. We assume that the ALD-SiO₂ thickness was reproduced on the SiO₂ surface, but not necessarily on GO. The roughness of the ALD-SiO₂ surface was almost identical between the regions with and without GO flakes in the present range of d (Fig. 3). Other morphologies due to the GO flakes were also recognized, e.g. the second GO flake stacking on the other [marked with * in Figs. 3(a) and 3(b)] and the narrow ridges corresponding to wrinkles of the flakes that were generally observed on GO flakes on solid substrates. ^36,39)

Zoom In Zoom Out Reset image size

The DOPC-SLB was formed on the ALD-SiO₂/subGO/SiO₂/Si surfaces and observed with the epi-fluorescence microscope. Figures 4(a)–4(c) show the fluorescence images of the DOPC-SLB on the ALD-SiO₂/subGO/SiO₂/Si surfaces of d = 5, 10, and 20 nm, respectively. The SLB covered the entire surface of the ALD-SiO₂ layer in the regions with and without GO. The shape of each GO flake was recognized as the SLB in the GO region was darker than the surrounding region because of the fluorescence quenching by GO. ^20,38) Much darker regions in SLB [* in Figs. 4(a) and 4(b)] probably had doubly stacked GO under them as observed in the AFM topographies (* in Fig. 3). The fluorescence intensity in the GO region increased with the thickness of the ALD-SiO₂ layer, whereas the intensity in the surrounding region was independent of the ALD-SiO₂ thickness [Figs. 4(a)–4(c)]. The result indicates that the quenching efficiency of GO decreased with the increase in the thickness of the ALD-deposited SiO₂ layer. The quenching efficiency of GO was varied by the ALD-SiO₂ layer.

Zoom In Zoom Out Reset image size

The quenching efficiency (E) in the SLB on ALD-SiO₂/subGO/SiO₂/Si was calculated using the intensities in the fluorescence images at the regions with and without GO by

$$\begin{eqnarray}&&E=1-\frac{I(d)-{I}_{{\rm{b}}}}{{I}_{0}-{I}_{{\rm{b}}}}=\frac{{I}_{0}-I(d)}{{I}_{0}-{I}_{{\rm{b}}}},\end{eqnarray} \tag{ 2 }$$

where I₀ and I(d) represent the fluorescence intensities of the regions without GO and with GO, respectively. The background intensity, I_b, was obtained from a substrate without SLB. We obtained E of 0. 53 ± 0.07, 0.21 ± 0.05, and 0.14 ± 0.02 at d = 5, 10, and 20 nm, respectively. We evaluated the distance between GO and SLB (r) from d and the step height (h) in AFM topographies (Fig. 3), assuming the typical thickness of a DOPC bilayer (4 nm) ³⁷⁾ and the water layer between the SLB and the substrate (1 nm) ^53,54) as illustrated in Fig. 5: r = d + h + 1 + 4/2. The values of h were 1 nm at d = 5 and 10 nm, and 0 nm at d = 20 nm that were obtained from the step heights in the AFM topographies (Fig. 3). The values of r and E at each d are listed in Table I. The dependence of E on r is plotted in Fig. 6. By fitting E with Eq. (1), we found that experimentally obtain E reasonably depended on r. We obtained R₀ = 8.1 nm that is comparable with the result in the previous study, R₀ = 7.5 nm. ²⁸⁾ These results show that inserting a ALD-SiO₂ layer between SLB and GO is available to control the fluorescence quenching efficiency in GO. A rather large discrepancy between the plot and the fitting curve at d = 20 nm possibly came from the interference effect. On a SiO₂/Si substrate, fluorescence intensity oscillates with the thickness of the SiO₂ layer, and its constructive interference peak exists at the thickness of approximately 100 nm. ⁵³⁾ At the off-peak position, the fluorescence intensity is sensitive to the SiO₂ thickness and also the refractive index of the upper layer. The ALD-SiO₂ layer was deposited on a thermally oxidized SiO₂/Si substrate with a 90 nm thick SiO₂ layer, and thus the total SiO₂ thickness was 110 nm at d = 20 nm. Note that for FRET experiments the r range with a large gradient in E is practically valuable, and it is around d = 5 nm in Fig. 6.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We performed fluorescence single molecule observation in SLB on ALD-SiO₂/subGO/SiO₂/Si with d = 5 nm. A fluorescence image of DOPC-SLB containing 10⁻⁷ mol% of ATTO-DPPE is shown in Fig. 7. Laterally diffusing bright spots were observed in the GO region (white arrows in Fig. 7), whereas their intensities were lower than those in the surrounding region. Homogeneous intensity and the single step photobleaching indicated that each bright spot corresponded to the fluorescence from a single dye molecule. The transition temperature between the gel and liquid crystalline phases (T_m), which correspond to solid-like and fluid states, respectively, of DOPC is −19 °C. The DOPC-SLB is in the liquid crystalline phase at 25 °C, and thus the dye-labeled lipids in it laterally diffuse. In the DOPC-SLB directly supported on GO/SiO₂/Si, fluorescence from each dye-labeled lipid molecule was quenched too effectively to be detected. ⁴⁰⁾ As shown in Table I, E = 0.53 at d = 5 nm in Fig. 7. Inserting ALD-SiO₂ layer achieved the fluorescence single molecule observation within the effective range of the fluorescence quenching by GO.

Zoom In Zoom Out Reset image size

We investigated fluorescence intensity of single the ATTO-DPPE molecules in the GO region in detail using SLB of DMPC, whose T_m is 24 °C, in the gel phase on the ALD-SiO₂/fullGO/SiO₂/Si substrate. Randomly diffusing fluorescence probes (Fig. 7) give a larger variation in the maximum intensity at each frame of a movie compared to immobile probes. On the subGO/SiO₂/Si substrate, the fluorescence from significantly brighter dyes in the SiO₂ region interfered the evaluation of the darker dyes in the GO region. Therefore, we adopted the fullGO/SiO₂/Si substrate [Fig. 2(a)] instead of the subGO/SiO₂/Si substrate. Figure 8(a) shows a fluorescence image of DMPC-SLB containing 5 × 10⁻⁶ mol% of ATTO-DPPE at 15 °C, which is lower than T_m of DMPC. The bright spots were immobile because the DMPC is in the gel phase. Figure 8(b) shows DMPC-SLB after addition of 1 mM Co²⁺. Cobalt (II) ion in a bulk aqueous phase above SLB selectively quenches fluorescence dyes labeled on the upper leaflet of a lipid bilayer. ^54,55) Density of the bright spots decreased from 0.451 μm⁻² without Co²⁺ [Fig. 8(a)] to 0.222 μm⁻² with Co²⁺ [Fig. 8(b)]. Approximately half of bright spots disappeared indicating that the dye-labeled lipid molecules in the upper leaflet in DMPC-SLB were quenched, while those in the lower leaflet retained fluorescence emission. Preliminary we had performed epi-FM observation of SLB in the presence of 0–10 mM Co²⁺. The fluorescence intensity was lower in the presence of Co²⁺ compared to that without Co²⁺ but was constant in the concentration range of 1–10 mM (data not shown). Therefore, most of the dye-labeled lipids in the upper leaflet was quenched in the presence of 1 mM Co²⁺.

Zoom In Zoom Out Reset image size

The histograms of the fluorescence intensity in the single molecule observation before and after the addition of Co²⁺ are shown in Figs. 9(a) and 9(b), respectively. The former showed broad distribution in the intensity region (arbitrary unit) of 0.5–4.5, whereas the latter had a rather sharp distribution at 1.0–1.5 accompanied with a tail up to ∼4.5. The peak intensities were found to be 1.38 and 3.00, respectively, by fitting the histogram in Fig. 9(b) with Gaussian curves. We attributed the former to the intensity of a single dye molecule, and the latter nearly double of the former to that of two dye molecules overlapping. The spatial resolution of the current experiment was approximately 240 nm based on the Rayleigh criterion with the emission window around 580 nm and the objective of N.A. = 1.45. The fluorescence intensities of two dye-labeled lipids within this region were overlaid. The histogram in Fig. 9(a) was fitted with three Gaussian components assuming that the dyes in the lower leaflet had the same intensity as in the presence of Co²⁺ [Fig. 9(b)] and that the number of the dye labeled lipid is identical between the two leaflets considering the number density in Fig. 8 as mentioned above. The histogram in Fig. 9(a) was assigned to three components with their peaks at 1.38, 2.31, and 3.50 [Fig. 9(a)]. The component at 2.31 that disappeared after the addition of Co²⁺ is attributed to the dye-labeled lipids in the upper leaflet of SLB. The upper leaflet that was further from GO than the lower leaflet was brighter and quenched by Co²⁺ in the bulk solution. The lower leaflet closer to GO than the upper one was darker and was not affected by Co²⁺. The ALD-SiO₂ layer effectively reduced E so that single molecule observation in the lower leaflet of SLB was achieved. The results in Fig. 9 also revealed that the upper and lower leaflets had different fluorescence intensity because of their distances from GO corresponding to the lipid bilayer thickness approximately 5 nm.

Zoom In Zoom Out Reset image size

Conventional observation techniques are not capable to determine the vertical position of molecules in a ∼5 nm thick lipid bilayer membrane even though they have high spatial resolutions. AFM has a vertical spatial resolution of subnanometer, but its observation is limited to the topmost surface of samples. In SLB, a cantilever tip cannot access the lower leaflet without destructing the bilayer membrane. Super-resolution optical microscopies such as localization microscopies [e.g. the photo-activated localization microscopy and stochastic optical reconstruction microscopy] and stimulated emission depletion microscopy show vertical resolutions in the order of a hundred nanometer, which are not comparable to their lateral resolution. ⁵⁶⁾ The results in the present study demonstrated a proof of concept recognizing molecules in the vertical region of 5 nm.

In this study, we demonstrated that the effectiveness of the ALD-SiO₂ layer in controlling the fluorescence quenching efficiency by GO in the SLB systems. The SiO₂ layers were fabricated on GO as well as on a thermally oxidized SiO₂ surface by the ALD method. Epi-fluorescence microscope images of SLB with 0.2 mol% dye-labeled lipid revealed that separating SLB from GO with the ALD-SiO₂ reasonably varied the quenching efficiency as expected from the FRET equation. Single molecule observation using a dye-labeled lipid was achieved in SLBs existing in the effective range of the fluorescence quenching by GO. The dye-labeled lipids in the upper and lower leaflets of SLB gave brighter and darker fluorescence depending on the distance from GO.

This work was supported by JST-CREST JPMJCR14F3, JSPS KAKENHI Grant Nos. JP20H02690 and JP20K21125, and the Nitto Foundation, Japan. We acknowledge supports from the Cooperation Research Project of the Research Institute of Electrical Communication (RIEC), Tohoku University, and the Electronics-Inspired Interdisciplinary Research Institute (EIIRIS) Project of Toyohashi University of Technology.