High-Speed Microscopy of Diffusion in Pore ... - ACS Publications

Letter Cite This: Nano Lett. XXXX, XXX, XXX−XXX

pubs.acs.org/NanoLett

High-Speed Microscopy of Diffusion in Pore-Spanning Lipid Membranes Susann Spindler,†,‡ Jeremias Sibold,§ Reza Gholami Mahmoodabadi,†,‡ Claudia Steinem,*,§,∥ and Vahid Sandoghdar*,†,‡ †

Max Planck Institute for the Science of Light, Staudtstraße 2, 91058 Erlangen, Germany Department of Physics, Friedrich Alexander University Erlangen-Nuremberg, Staudtstraße 5, 91058 Erlangen, Germany § Institute for Organic and Biomolecular Chemistry, Tammannstraße 2. 37077 Göttingen, Germany ∥ Max Planck Institute for Dynamics and Self-Organization, Am Faßberg 17, 37077 Göttingen, Germany

Downloaded via UNIV OF SOUTH DAKOTA on July 26, 2018 at 22:21:58 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

‡

S Supporting Information *

ABSTRACT: Pore-spanning membranes (PSMs) provide a highly attractive model system for investigating fundamental processes in lipid bilayers. We measure and compare lipid diffusion in the supported and suspended regions of PSMs prepared on a microfabricated porous substrate. Although some properties of the suspended regions in PSMs have been characterized using fluorescence studies, it has not been possible to examine the mobility of membrane components on the supported membrane parts. Here, we resolve this issue by employing interferometric scattering microscopy (iSCAT). We study the location-dependent diffusion of DOPE 1,2-dioleoylsn-glycero-3-phosphoethanolamine) lipids (DOPE) labeled with gold nanoparticles in (1,2-dioleoyl-sn-glycero-3-phosphocholine) (DOPC) bilayers prepared on holey silicon nitride substrates that were either (i) oxygen-plasma-treated or (ii) functionalized with gold and 6-mercapto-1-hexanol. For both substrate treatments, diffusion in regions suspended on pores with diameters of 5 μm is found to be free. In the case of functionalization with gold and 6-mercapto-1-hexanol, similar diffusion coefficients are obtained for both the suspended and the supported regions, whereas for oxygen-plasma-treated surfaces, diffusion is almost 4 times slower in the supported parts of the membranes. We attribute this reduced diffusion on the supported parts in the case of oxygen-plasma-treated surfaces to larger membrane−substrate interactions, which lead to a higher membrane tension in the freestanding membrane parts. Furthermore, we find clear indications for a decrease of the diffusion constant in the freestanding regions away from the pore center. We provide a detailed characterization of the diffusion behavior in these membrane systems and discuss future directions. KEYWORDS: Pore-spanning membrane, model membrane system, lipid diffusion, interferometric scattering microscopy, single-particle tracking

M

membranes to the micrometer regime allows for sufficient stability without the need for a solvent. Various micro- and nanofabrication technologies can be employed to structure an array of micrometer-sized pores in thin dielectric substrates. Such substrates are commercially available and have been used as a platform for preparing porespanning lipid membranes (PSMs)9−11 to study the mechanical properties of lipid bilayers,12 phase behavior of multicomponent lipid mixtures,9 transport through ion channels,13,14 and fusion of protein-containing vesicles with SNARE-doped lipid membranes.15 These studies all relied on fluorescent labels, which have several limitations. First,

odel lipid membranes are widely used to investigate various physical and biological properties of lipid bilayers in a well-controlled environment. Among them, supported lipid bilayers (SLBs) are popular due to their ease of fabrication, stability, and flat geometry.1 However, the proximity to the substrate can affect the mobility of the membrane components and prohibit the reconstitution of proteins.2−5 Thus, the fabrication and characterization methods for freestanding membranes are highly desirable. A well-known system is the so-called black lipid membrane,6 which is generally formed over millimeter-sized apertures using an apolar solvent. Transmembrane protein diffusion has been monitored and quantified in these black lipid membranes with high accuracy,7 but residuals of the organic solvent used in the fabrication process can change the lipid diffusion behavior and influence protein function.8 Reducing the extent of suspended © XXXX American Chemical Society

Received: June 2, 2018 Revised: July 20, 2018

A

DOI: 10.1021/acs.nanolett.8b02240 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 1. Preparation of pore-spanning membranes. (a) Schematic illustration of the formation process. Fluorescence-labeled GUVs composed of DOPC doped with 1 mol % cb-DOPE and a trace amount of Atto532-DOPE are added to the functionalized Si3N4 substrate in PBS. The GUVs settle down and rupture, creating membrane patches that span the 5 μm size holes. After the addition of GNPs and 30 min of incubation time, the substrate is transferred into another chamber filled with PBS, where it is flipped and placed on a spacer (sputtered gold strips). (b) Fluorescence image of PSMs on a Si3N4 substrate functionalized with gold and MCH (Au-SiN) in pseudocolor. (c) Fluorescence image of PSMs on a oxygenplasma-treated Si3N4 substrate with PEGylated inner pore walls (p-SiN) in pseudocolor. A membrane patch created by a burst GUV is clearly visible, although in most cases, the membrane has been ruptured.

Here, a 30 nm gold layer is sputtered selectively on the top part of the porous Si3N4 substrates, which is then functionalized with 6-mercapto-1-hexanol providing a hydrophilic OHterminated surface that allows the spreading of the GUVs and formation of PSMs. Alternatively, the silicon nitride substrates were first hydrophilized by oxygen plasma, resulting in an SiO2 surface layer on the whole silicon nitride substrate.18 Afterwards, the inner pore walls were specifically functionalized with methoxy PEG succinimidyl carbonate (see the Methods section for details) to prevent attachment of the membranes inside the pores, while the top surface remained hydrophilic to induce spreading of the GUVs. In both surface-treatment methods, each substrate was rinsed with ethanol followed by PBS before PSM preparation. Immediately after rinsing, it was placed into a chamber filled with PBS, and GUVs prepared in 300 mM sucrose were added (Figure 1a). The GUVs contained a trace amount of fluorescently labeled lipids (Atto532-DOPE) for membrane visualization using a CCD camera. GUVs burst after settling and formed PSMs with a homogeneous fluorescence intensity distribution as visible in Figure 1b, where we show an image of PSMs on Au-SiN. On p-SiN, most of the PSMs rupture, and only a few PSMs remain (Figure 1c). For SPT experiments, GNPs were added. After 30 min of incubation, the substrate was transferred into another chamber on a cover glass filled with PBS. It was flipped so that its flat side with the lipid membrane faced the cover glass and placed onto two sputtered gold strips as spacer (about 200 nm thick). iSCAT. GNPs were detected using interferometric scattering microscopy (iSCAT). The principle of iSCAT is based on the homodyne detection of the light scattered from the nanoparticle with a reference light beam, usually the reflection from the cover glass−sample interface. The signal I at the detector reads:

photobleaching prevents one from following the behavior of the same nano-object for times much longer than 1 min. Second, fluorescence signals are strongly quenched in the vicinity of metallic coatings, which are advantageous for preparing high-quality PSMs. Hence, fluorescence recovery after photobleaching (FRAP) or fluorescence correlation spectroscopy (FCS) experiments cannot deliver diffusion coefficients for the supported regions of PSMs, and instead, only indirect FRAP experiments have been used for arriving at estimated quantities.15 Here, we employ a single-particle tracking (SPT) approach by labeling lipids with gold nanoparticles (GNPs) and detecting their scattering signal using interferometric scattering microscopy (iSCAT).16 We have previously used iSCAT to study lipid diffusion with ultrahigh temporal and spatial precision on supported membranes and on giant unilamellar vesicles.4,17 Applying iSCAT-SPT to PSMs allows us to compare the diffusion behavior of the same lipid molecules in a freestanding part of a PSM (f-PSM) with its supported section (s-PSM). In this work, we employ silicon nitride (Si3N4) substrates with 5 μm pores with two different substrate functionalization techniques of (i) oxygen-plasma treatment resulting in an SiO2 surface layer18 and (ii) functionalization with gold and 6-mercapto-1-hexanol (MCH).9 Plasma treatment of the substrates (p-SiN) results in a strong membrane adhesion, which slows down diffusion on the supported areas, while it induces a high membrane tension and, thus, a faster diffusion in the freestanding regions of the PSM. The use of MCH on gold (Au-SiN) diminishes membrane adhesion and renders diffusion on s-PSM and f-PSM similar. Results and Discussion. Preparation of Pore-Spanning Membranes. In one approach, PSMs were prepared by spreading giant unilamellar vesicles (GUVs) on the silicon nitride substrates according to the protocol of refs 9 and 15. B


Letter

Nano Letters I = |Eref + Escat|2 = |Einc|2 (r 2 + |s|2 + 2r |s|cos(Δϕ))

interference between the reflections at the lower and upper membrane leaflets. However, ring-shaped patterns are observed in the f-PSM and s-PSM parts of the sample due to diffraction at the circular rims. Figure 3b displays a temporal median background correction4 of a single pore region to account for static background contributions. We see that the background pattern could not be completely removed in this way due to nanoscopic vibrations in the setup. In some cases, further complication arises if the particle signal becomes lower than the residual background variations from the median correction. To address these issues, we applied a more-sophisticated algorithm based on image registration.19 Briefly, a stack of images is aligned to the frame of interest for each frame by translation. An optimization routine is used to find the coefficients of the registered group of images to optimally describe the background of the frame of interest. Finally, each pixel is divided by the estimated background at the pixel position and multiplied by the mean value of the whole background image. To avoid eliminating the particle signal in the background correction procedure, we introduce a delay between the frames used for creating the background to ensure that the particle moved a substantial distance from the initial position. As shown in Figure 3c, the rings could be almost fully removed by this background correction procedure, revealing the location of a GNP. Figure 3d shows a typical histogram of the x and y fit errors obtained from fitting a Gaussian function to the point-spread function of a 40 nm GNP on an f-PSM for a background-corrected video based on image registration. Lipid Tracking on PSMs. We recorded trajectories of GNPs bound to lipids on s-PSMs and f-PSMs with a high temporal resolution of 1 ms, typically for 5 s. In Figure 4, we show example trajectories of three particles for a p-SiN support: (i) on an s-PSM (Figure 4a), (ii) moving between f-PSM and sPSM (Figure 4b), and (iii) trapped in an f-PSM (Figure 4c). In the latter case, the particle was possibly bound to the inner membrane leaflet (see Figure 1) or was trapped in the f-PSM due to strong membrane bending, as will be discussed in the next section. Particle tracking on Au-SiN substrates is more challenging due to the high reflectivity of the gold layer, which results in a 5 times larger reflection than for p-SiN substrates and, correspondingly, a lower iSCAT contrast.16 Nevertheless, our efficient background correction procedure allows us to detect 40 nm GNPs with a contrast of 1−2% on Au-SiN substrates. We point out that although the signal drops below the shot noise fluctuations of 0.22% for 20 nm GNPs on these samples, detection of these particles would also be possible if one

(1)

where Escat and Eref denote the scattered and reference fields, respectively. The quantity Einc is the incoming field, while r and s signify the reflectivity of the substrate−sample interface and the particle’s polarizability, respectively. The last term includes the phase difference Δϕ between the scattered and reflected fields. The iSCAT signal of the particle can be quantified by a normalized contrast c, given by c = (IP − IBG)/IBG, where IP is the signal measured from the particle, and IBG is the background signal in the absence of the particle. Setup. The microscopic setup combining iSCAT and fluorescence microscopy is shown in Figure 2. The scattered

Figure 2. Schematics of the experimental iSCAT and fluorescence setup.

and reflected lights were collected by the objective, passed the λ/4 plate, were thus reflected by the PBS plate, and imaged onto a CMOS camera (MV-D1024E-CL, Photonfocus AG) using a singlet (f = 50 cm) lens. For fluorescence imaging, a beam splitter plate (T/R = 90:10) in front of the objective reflected a portion of the collected light onto a highly sensitive CCD camera (Sensicam qe, PCO). A long-pass filter (540 nm LP) was used for filtering the fluorescence light from the excitation, and an f = 40 cm lens was used for imaging. Image Background Correction. In Figure 3a, we show a raw iSCAT image of PSMs formed on a p-SiN substrate. The lipid membrane itself is not visible due to destructive

Figure 3. Image background correction for SPT on PSMs. (a) Raw iSCAT image of a lipid membrane on a substrate with 5 μm sized pores. (b) Temporal median background-corrected image of a pore region with a 40 nm GNP. (c) Improved background-corrected image of the same region as in panel b using image registration. (d) Histogram of fit errors from localized particle positions in the pore region. C


Letter

Nano Letters

Figure 4. Raw iSCAT images from videos recorded at 1 kHz for 5 s overlaid with trajectories of 40 nm GNPs attached to cb-DOPE lipids diffusing in a DOPC membrane on a p-SiN substrate. (a) Trajectory of a GNP on the s-PSM. (b) Trajectory of a GNP moving on the f-PSM, crossing over to the s-PSM and coming back to the f-PSM. (c) Trajectory of a GNP trapped in the f-PSM. The color code shows the time evolution in each case.

Figure 5. Particle contrast on different membrane regions. (a) Schematic illustration of the reflections at a PSM. When the particle is located on the supported membrane (s-PSM), its scattered light interferes with the light reflected at the cover glass−buffer interface (3) as well as the reflected light at the Si3N4 substrate (1), which acquires a π-phase shift. When the particle is on the freestanding membrane part (f-PSM), the reference beam is the reflected light from the glass−buffer interface (3). The reflection at the membrane itself (2) does not contribute because the reflected light at both leaflets interferes destructively due to the π-phase shift of one of the beams. (b−d) Variations in particle contrast on different membrane locations obtained from the same time series: (b) 40 nm particle (indicated by the white arrow) displaying a positive contrast on an fPSM, (c) a negative contrast on the s-PSM on the Au-SiN substrate, and (d) no contrast inversion on a neighboring f-PSM.

averaged over many frames N to reduce the noise by a factor of √N (i.e., at the cost of lower temporal resolution). Membrane Located at Different Heights inside the Pores. The interferometric approach used in our studies has the advantage of detecting the encoded height information in the particle signal Ip.17,20 As eq 1 indicates, the particle signal depends on the phase difference Δϕ between the scattered and reflected fields, which, in turn, depends on the height difference between the particle and the cover glass. As a result, we can make observations on membrane bending inside the pore. On an f-PSM, Escat from the particle interferes with Eref from the cover glass, as illustrated in Figure 5a. On the sPSM, Escat interferes with Eref from both the Si3N4 support and the cover glass. Hence, the scattered light from a particle on the f-PSM interferes with a reference beam of lower intensity (IBG), so that one expects a larger particle contrast on the fPSM than on the s-PSM. Furthermore, a contrast inversion

should occur because the reflected light at the Si3N4 substrate acquires a π-phase shift, while the reflected light at the glass− water interface does not. This behavior was, indeed, often observed as demonstrated in Figure 5b,c. Here, the contrast of the same particle is changed from 20.5% on an f-PSM to a negative value of −0.64% on the s-PSM of the Au-SiN substrate. Interestingly, we also found particles displaying much lower contrast on the f-PSM and no contrast inversion. In some cases, the same particle featured a positive contrast on one f-PSM, while it stayed dark on another one. Such a situation is demonstrated in Figure 5b−d. The particle contrast on an f-PSM depends on the distance between the glass surface and the membrane-bound particle (eq 1). Considering that variations in the substrate height are negligible over small distances between neighboring pores, we attribute the contrast change to differences in the membrane height, implying that the membrane is bent at the pore rim. In D


Letter

Nano Letters fact, this effect was also observed in scanning ion-conductance microscopy (SICM) studies of PSMs on pores of 1.2 μm diameter, where height differences of 45−95 nm between the s-PSM and f-PSM were measured.21 A similar height difference of Δh = 135 ± 55 nm (n = 12) was also found for 5.5 μm pores,9 supporting our observations. The exquisite sensitivity of iSCAT to axial displacements is one of its important advantages over fluorescence measurements. Lipid Diffusion in Freestanding vs Supported Membrane Regions. We now discuss quantitative analyses of the recorded trajectories according to two different methods. Squared Displacement Analysis. We computed the cumulative probability distribution of squared displacements P(r2, Δt = 1 ms) for all particle trajectories4 to test for multiple mobilities. Displacements were only considered if the particle’s PSF could be fitted well in two consecutive frames. For analyzing lipid diffusion within f-PSMs, we first only consider trajectory segments within a radius of 2 μm about a pore center to exclude potential systematic effects close to the pore rim. Figure 6a,b summarizes the results of our analysis. The trajectories of particles diffusing on f-PSMs could be well described by a one-component fit for both substrate functionalization methods. However, the diffusion coefficient D obtained from a P(r2) analysis is distributed around larger values for p-SiN substrates compared with Au-SiN substrates. We attribute this to a larger tension in the freestanding membranes on p-SiN substrates. We also remark that neither transient MSD analyses4 nor direct scrutiny of the trajectories revealed any confinements. We shall return to this point again. A different diffusion behavior was observed on s-PSMs. Here, for many of the trajectories good fits to the diffusion behavior required two mobility components D1 and D2 with weighting factors ε1 and ε2, respectively. This observation agrees with our previous findings on supported lipid bilayers.4 In addition, we note that a faster diffusion is obtained for the Au-SiN substrates: Figure 6c,d shows that D1 peaks at around 0.8 μm2 s−1 for p-SiN and at 2.0 μm2 s−1 for Au-SiN. Moreover, as Figure 6h indicates, the fraction of the trajectories displaying only one mobility (i.e., ε1 = 1) is larger for Au-SiN substrates. The minor mobility D2 for s-PSMs is significantly lower than D1 for both substrate types (see Figure 6e,f). This is signified by D2 having a peak at 0.3 μm2 s−1 on p-SiN and at 0.5 μm2 s−1 on Au-SiN. The difference is likely related to membrane defects. In fact, it was frequently observed that particles became stuck on s-PSMs for a few tens of milliseconds up to several seconds. We will discuss this more intensively below. When applying the P(r2) analysis, one has to keep in mind that blurring of the point-spread function (PSF) during the camera exposure time can lead to deviations from the true diffusion coefficient. Although it has been pointed out that blurring leads to an underestimation of the diffusion coefficient,22,23our Monte Carlo simulations show that a fit error can also result in overestimation of the diffusion coefficient. In Table S1, we show the outcome for simulating freely diffusing particles with three different diffusion coefficients with and without taking PSF blurring into account as well as for two different noise levels. Due to large differences in particle contrast observed in various measurements and, thus, different fit error distributions, a fair comparison of the derived diffusion coefficients is not possible for s-PSMs and f-PSMs. For example, it is difficult to make a clear statement about the observation that D

Figure 6. Histograms of the major mobility D1, minor mobility D2, and the weighting factor ε1 obtained from the analysis of the cumulative probability distribution of squared displacements. ε1 = 1 indicates the fraction of trajectories with only one mobility. (a) D for particles diffusing on freestanding (f-PSMs) on plasma-treated silicon nitride substrates (p-SiN) obtained from a total of n = 28 trajectories. (b) Same as in panel a but on gold-mercaptohexanol-coated substrates (Au-SiN) obtained from n = 89 trajectories. All trajectories on f-PSMs could be fitted with a one-component fit. (c, e, g) D1, D2, and ε1 for particles diffusing on s-PSMs on p-SiN substrates (n = 820). (d, f, h) D1, D2, and ε1 for particles diffusing on s-PSMs on AuSiN substrates (n = 322).

gravitates toward larger values in trajectories on the substrate than in the freestanding part of Au-SiN samples (see Figure 6b,d). Nevertheless, this analysis still provides useful information on the presence of multiple mobilities and, thus, deviations from normal free diffusion. Covariance-Based Estimation. An analysis method that accounts for motion blur and limited localization precision is provided by the covariance-based estimation (CVE).24 To benchmark this approach, we estimated the diffusion coefficient DCVE (eq 14 in ref 24) for the simulated trajectories E


Letter

Nano Letters

Figure 7. Results from covariance-based estimation (CVE) analysis. (a) Histograms of the diffusion coefficients for p-SiN (blue striped bars, n = 28) and Au-SiN (red bars, n = 89) substrates for particles diffusing on f-PSMs. (b) Histograms of the diffusion coefficients for p-SiN substrates (n = 416) and Au-SiN (n = 284) for particles diffusing on s-PSMs. (c) Ratio between the diffusion coefficient on the f-PSM and s-PSM for individual particles with trajectory parts in both regions.

± 0.1 mN/m (SD), which is three times smaller than the value of 1.2 ± 0.4 mN/m (SD) found on p-SiN substrates. MD simulations show that an increase in membrane tension is accompanied by an increase in the diffusion coefficient of lipids. For example, raising the membrane tension from 0 to 15 mN/m resulted in a 52% increase in the diffusion coefficient.25 This behavior was also reported in GUV experiments, where a 44% faster diffusion was found for an increase in membrane tension from 0.02 to 0.1 mN/m.27 In addition, a curvature-induced effect might play a role. Considering the large coverage of the GNPs with streptavidin, it is possible that more than one lipid is bound to a GNP. The chance of multiple binding might increase with lower membrane tension due to the ability of the membrane to curve around the GNP, which, in turn, can influence the diffusion coefficient. However, this effect is probably very small because the membranes are pre-stressed. We note that, compared with many studies in the literature (e.g., refs 10 and 28), our obtained diffusion coefficients on the freestanding parts might appear low. Due to differences in the accessible length and time scales, however, the measured diffusion coefficient of lipids can depend on the measurement technique. In 2008, a study was published3 in which the diffusion coefficients measured by FRAP, different implementations of FCS (conventional FCS, Z-scan FCS, and Imaging Total Internal Reflection (ITIR) FCS) and SPT were compared. For POPC-SLBs prepared on glass, variations in the diffusion coefficient up to a factor of 3 were found. Furthermore, the effect of the localization precision on the diffusion coefficient has to be considered in SPT: a larger fit error leads to an over-estimation of the diffusion coefficient when it is not accounted for.29,30 The localization precision of iSCAT is much better than what can be typically achieved using fluorescence single-molecule tracking and can, thus, contribute to the smaller diffusion constants compared with previously published values. Nanoscale Confinements on Supported Membrane Regions. Besides the differences in the diffusion coefficients on s-PSMs and f-PSMs, another interesting feature of our observations is lack of any confinement on f-PSMs, while frequent trapping in nanoscopic domains is found on s-PMS. Such immobilization events were observed both on p-SiN samples (Figure 8a and inset) and on Au-SiN substrates (Figure 8b and inset). By the application of a transient MSD analysis,4 these immobilization events can be clearly detected. Figure 8c illustrates this for the trajectory shown in Figure 8b. Here, the sliding window T was chosen to be 30 frames long,

used in Table S1. In all of the simulated cases, the estimated diffusion coefficient DCVE matches the true one. The shortcoming of this formalism is, however, that it only applies to freely diffusing particles. Keeping this in mind, we employed the CVE method for a direct comparison of the diffusion coefficients between s-PSMs and f-PSMs as well as the two substrate types. Here, we excluded all trajectories that showed a deviation from free diffusion, i.e., trajectories with two mobilities. We further applied a free diffusion test and compared the normalized periodigram to a χ2 distribution by using a Pearson’s χ2 test for each trajectory.24 Only data with p values above 5% were analyzed. The results from the CVE analysis are summarized in Figure 7. The diffusion coefficient DCVE for f-PSMs on p-SiN substrates is shifted to larger values with a broad distribution around 2.7 μm2 s−1 (Figure 7a) in comparison with Au-SiN substrates with a distribution centered at about 1.8 μm2 s−1. The contrary behavior is observed on s-PSMs, shown in Figure 7b. On p-SiN substrates, the diffusion is much slower with a peak at 0.7 μm2 s−1 (here, data from 20 and 40 nm GNPs were plotted together because no obvious difference could be observed between the two particle sizes) compared with AuSiN substrates in which a distribution with a peak at about 2.0 μm2 s−1 is observed. These trends conform with the results from P(r2) analysis; however, it became clear now that the diffusion on the supported parts of Au-SiN substrates is very similar to the one found in f-PSMs. In Figure 7c, the ratio of the diffusion coefficients on f-PSMs and s-PSMs is shown for several individual particles that crossed the two regions. In both cases, diffusion is typically faster on the freestanding part, although this effect is much less pronounced for the Au-SiN samples. These findings clearly indicate that the p-SiN substrate slows diffusion, indicating a strong interaction of the membrane with the support that also affects the top leaflet due to interleaflet coupling. By functionalization with gold and MCH, lipid− substrate interaction is reduced. The functionalization of the pore rims strongly influences the lateral membrane tension of the f-PSMs. Stronger adhesion of the membrane on the pore rims leads to a larger membrane tension in the f-PSMs,25 as confirmed by atomic force microscopy (AFM) indentation experiments.26 The difference in lateral membrane tension might explain the difference in diffusion coefficients on f-PSMs between the two substrate types. We also employed AFM indentation experiments and determined the lateral membrane tension of PSMs on Au-SiN and p-SiN surfaces (Figure S1). Membranes on Au-SiN samples exhibit a lateral tension of 0.4 F


Letter

Nano Letters

Figure 9. Results from the transient MSD analysis for confinement detection. (a) Confinement time distribution and (b) confinement radius distribution for trajectories recorded on p-SiN substrates (blue curves) and on Au-SiN substrates (red curves).

In a previous study on lipid diffusion in supported lipid bilayers, we also detected confinement events.4 In the current experiments, the membrane was prepared as a continuous bilayer from a single GUV. The ubiquitous appearance of nanoscale transient traps is consistent with the hypothesis that they stem from small pinning points caused by the underlying surface roughness or chemical heterogeneity. Uniformity of Diffusion within a Pore. For the results presented above, we excluded the regions of the membrane close to the pore rim. However, the high speed and long measurement duration of iSCAT recordings allow us to investigate the diffusion behavior as a function of the particle location within the pore. Here, we examined three different regions of 0 < r < 1.3 μm, 1.3 μm < r < 1.9 μm, and 1.9 μm < r < 2.4 μm. The region between 2.4 and 2.5 μm is not accessible due to distortions of the point-spread function at the edge of the pore. The estimated diffusion coefficients obtained for all particles with enough data points in all three regions for robust statistics are plotted in Figure 10 (note that the number of trajectories analyzed here is lower than in Figure 7). Figure 10a,b shows the cases for particles on p-SiN and Au-SiN substrates, respectively. In almost all cases, the diffusion coefficient decreases when going from the inner to the outer part, while this trend is particularly pronounced when comparing the middle to the outer region. The mean diffusion coefficient from all particles on the freestanding part of the pSiN substrates amounts to 2.8 μm2 s−1 in the inner region (R1), 2.5 μm2 s−1 in the middle (R2), and 2.0 μm2 s−1 in the outer region (R3). In f-PSMs prepared on Au-SiN, these values decrease to 2.5 μm2 s−1 in the center, 2.2 μm2 s−1 in the middle part, and 1.6 μm2 s−1 in the region close to the pore rim (see Figure 10c). To evaluate whether the differences are of statistical significance, we performed Welch’s t test as described in the Supporting Information. Although the values for R1, R2,

Figure 8. Particle confinement on the substrate. (a) Particle trajectory with two confinements on a p-SiN substrate. (b) Particle trajectory with two confinements on the Au-SiN substrate. (c) Transient MSD for a sliding window of 30 frames and Δt of 15 frames for calculating the MSD. The threshold is set to 70 nm2 (indicated by the red line), revealing two confinements for 734 ms (I, upper inset in panel b) and 1535 ms (II, lower inset in panel b).

and the threshold MSDtresh was set to 70 nm2. We find two examples in which the particle was immobilized for 734 ms in a region of radius r = 24 nm (confinement I) and for 1535 ms in a region of r = 27 nm (confinement II). In addition to a transient immobilization, particles were also observed to stay confined in a small area as long as the whole length of the video (5 s) for both types of substrates. The results of the transient MSD analysis are summarized in Figure 9. Here, single events for times longer than 200 ms or areas with radii larger than 200 nm are not included. Particles stay predominantly confined for a few tens of milliseconds in regions of a few tens of nanometers. No notable difference in the confinement time and size distribution is visible between the two types of substrates. However, the number of trajectories displaying confinements is almost twice (87%) for p-SiN compared with the value for Au-SiN substrates (44.3%). G


Letter

Nano Letters

Figure 10. Distance-dependent diffusion coefficient in f-PSMs for (a) p-SiN substrates and for (b) Au-SiN substrates. For each particle, the diffusion coefficient in three regions within the pore was calculated using CVE analysis. Gray, red, and blue bars correspond to the diffusion constant for the regions R1 (0 < r < 1.3 μm), R2 (1.3 μm < r < 1.9 μm), and R3 (1.9 μm < r < 2.4 μm), respectively. The data are arranged, for convenience’s sake, according to the descending order of the diffusion coefficients for the innermost region. (c) Mean diffusion coefficient with standard deviation for the three regions and for both p-SiN and Au-SiN cases.

Materials and Methods. Materials. 1,2-Dioleoyl-snglycero-3-phosphocholine (DOPC) and 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-cap-biotinyl (cb-DOPE) were purchased from Avanti Polar Lipids (Alabaster, AL; USA). Atto532−1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (Atto532-DOPE) was from ATTO-TEC GmbH (Siegen, Germany). Lipids were stored in chloroform at a concentration of 10 mg/mL at −20 °C. Streptavidin-conjugated GNPs were purchased from BBI (Cardiff). Phosphate-buffered saline (PBS), sucrose, 6-mercapto-1-hexanol, 3-aminopropyltrimethoxysilane, and methanol were purchased from Sigma-Aldrich (Taufkirchen, Germany). Ethanol and isopropanol were obtained from VWR International (Ismaning, Germany). NPropanol was purchased from Carl Roth (Karlsruhe, Germany). Methoxypolyethylene glycol succinimidyl carbonate (methoxy PEG succinimidyl carbonate, 5 kDa) was from Nanocs Inc. (Boston, MA). PDMS was prepared from Sylgard 184 silicon elastomer base and Sylgard 184 elastomer curing agent (10:1 weight ratio) from Dow Corning (Wiesbaden, Germany). Silicon nitride substrates with 5 μm diameter holes and a thickness of 500 nm were obtained from Aquamarijn Micro Filtration BV (Zutphen, Netherlands). Indium tin oxide (ITO) cover glasses with 15−30 ohms per square were purchased from Diamond Coatings. All chemicals and solvents were of analytical grade and used without further purification. Substrate Functionalization. For surface functionalization with gold and mercaptohexanol, silicon nitride substrates were first rinsed with ethanol and dried in a stream of nitrogen. After plasma cleaning for 30 s, a thin titanium layer was sputtered onto the surface as an adhesive followed by the evaporation of 30 nm of gold. The gold layer was then functionalized with 6mercapto1-hexanol by incubating the substrates overnight in a 1 mM solution in n-propanol. For PEGylating the inner pore walls of oxygen plasmatreated silicon nitride substrates, a gold layer was first evaporated on top of the porous substrates without an adhesive titanium layer to allow the removal of the gold using adhesive tape after PEGylation. The binding of methoxy PEG succinimidyl carbonate to the inner pore walls was realized by first silanizing the oxygen-plasma treated porous substrates with a 0.5% 3-aminopropyltrimethoxysilane methanolic solution for 5 min followed by incubation with a methoxy PEG succinimidyl carbonate solution (20 mg in 1.5 mL of ethanol). Giant Unilamellar Vesicle Preparation. PSMs are typically prepared by spreading giant unilamellar vesicles (GUV) on silicon nitride holey substrates. We prepared GUVs with a lipid composition of DOPC and 1 mol % cb-DOPE via electro-

and R3 differ not significantly among p-SiN and Au-SiN for this subset of trajectories, the distributions from all trajectories displayed in Figure 7 show a significant difference. For both substrate types, we could not detect a significant difference between R1 and R2. However, the diffusion coefficients are different when comparing R2 and R3 for Au-SiN as well as R1 and R3 for both substrate types (see Table S3), strongly indicating a slow-down in vicinity of the pore rim. Possible origins of this intriguing observation would be curvatureinduced effects due to membrane bending in this region9 but also hydrodynamic effects31 caused by the pore rim might play a role. Varying the pore diameter in future studies can provide valuable information on the details of the membrane mechanics and its effect on diffusion behavior. Conclusions. Using iSCAT single-particle tracking, we investigated lipid diffusion in pore spanning membranes (PSMs). Because PSMs are composed of both supported (sPSMs) and freestanding bilayer regions (f-PSMs), the influence of the underlying substrate on lipid diffusion can be studied within one measurement. In a nutshell, the diffusion behavior reveals two diffusion constants and transient nanoscopic confinements on solid substrates, while we find free diffusion with a single diffusion constant for suspended membrane parts. Interestingly, however, our data also indicate a gradual variation of the diffusion behavior toward the pore rims within the suspended region. Our investigations show that the surface functionalization greatly influences the diffusion parameters. In the case of oxygen-plasma-treated substrates, we obtain much-slower diffusion in the support than in the freestanding region. Considering that plasma treatment of Si3N4 leads to the formation of a silicon dioxide layer, s-PSMs on these substrates are expected to behave similarly to supported lipid bilayers prepared on standard cover glass,4 which is indeed the case. For PSMs prepared on gold and mercaptohexanol-functionalized substrates, diffusion is faster and similar to what we observe on the freestanding parts of the membrane. In the future, it would be interesting to increase the pore diameter and study membrane dynamics in a completely unperturbed system over a larger area. Compared with GUVs, PSMs provide a flat geometry, large membrane stability, and reduced membrane fluctuations, which render them highly desirable for sensitive iSCAT measurements such as the detection and tracking of much smaller objects down to single unlabeled proteins. Another attractive direction would be to reduce the pore diameters and the wall thickness to mimic the interaction of a lipid membrane with the cytoskeleton in a biological cell. H


Letter

Nano Letters formation on indium tin oxide (ITO)32 in 300 mM sucrose solution. Briefly, two ITO-coated cover glasses separated with a sticky-tape spacer made from PDMS were connected via copper tape to a function generator. Before assembly, the ITO glasses were rinsed with ethanol and isopropanol and dried with a nitrogen stream. A total of 2.5 μL of 1 mg/mL lipid solution was spread on the conductive site of one of the ITO glasses. After evaporation of the solvent for at least 2 h in the desiccator, the 3 mm thick PDMS spacer was placed on the glass slide with the dried lipids and filled with the swelling solution. The chamber was closed by carefully placing the second ITO-coated glass slide on top. For GUV formation, a peak-to-peak voltage (Vpp) of 3.5 V at a frequency of 10 Hz was applied for 2 h, followed by a detachment step at 2 Hz for 30 min. After the formation process, the upper glass slide was carefully removed, and the GUVs were harvested using a pipet. Experimental Setup. For illumination, a linearly polarized continuous-wave laser at a wavelength of 532 nm was used (Verdi G2, Coherent, Santa Clara, CA) to excite gold nanoparticles near their plasmon resonance. The laser beam was scanned by two acousto-optical deflectors (AODs DTSXY-400, AA Optoelectronic, Orsay, France) to create a homogeneous illumination over the whole field of view (FOV) of 256 pixels × 256 pixels (16.28 μm2). Using a singlet lens with 40 cm focal length, the scanned beam was focused onto the back focal plane of an oil-immersion objective (Olympus ApoN, NA = 1.49) after passing a polarizing beam splitter plate (PBSW-532, Thorlabs) and a λ/4 wave plate. A 45° mirror (not drawn in Figure 2) sent the light into the objective. The sample was mounted on a 3D piezoelectric stage (P-621.2CL and PI-621.ZCL, PI), which allowed for precise movement of the sample and focusing. Manual translation stages in the x, y, and z planes enabled coarse positioning. Particle Tracking. For lipid labeling, we used predominantly GNPs with a diameter of 40 nm. For some SPT experiments on PSMs prepared on p-SiN substrates, 20 nm GNPs were also used. For automated particle tracking, we developed a Matlab code. First, a temporal median background correction was applied. For analyzing trajectories inside pores, the pore region was background-corrected using image registration of a number of images separated by a time lag of δt. The number of excluded consecutive frames is a function of the typical particle diffusion coefficient, the size of the PSF, and the frame rate. To achieve a sufficiently low probability for the overlap of the particle’s PSF with the one from the frame of interest, we allowed the particle to move a distance 5 times larger than a PSF diameter. For our typical experimental situation, δt = 30 frames around the frame of interest had to be excluded to fulfill this criterion. We considered a group of frames to ensure that frame-specific features are averaged out. The spacing between these frames was again 30 frames. After background correction, the particle position was determined in each frame by fitting a two-dimensional (2D) Gaussian function (with positive or negative amplitude) to the PSF. If the fit error for the Gaussian width was above a given threshold (e.g., because the contrast of the particle was low), the fit was discarded. For such frames, no particle position was assigned. Monte Carlo Simulations. We estimated diffusion coefficients from single particle trajectories using two methods: fitting the cumulative probability distribution of squared displacements,4 P(r2), and covariance-based estimation of the diffusion coefficient.24 To estimate the diffusion coefficient derived by these two methods as a function of the experimental

parameters such as the camera exposure time and localization precision, we simulated 2D random walks. Step sizes were assigned on the basis of random, normal-distributed numbers with the standard deviation according to a given diffusion coefficient (D). We then created videos of particles undergoing free diffusion by calculating the point-spread function (PSF) of a scatterer with a given contrast on top of a background noise. The PSF was obtained by a numerical Fourier transformation of a disc. The videos were analyzed using the same tracking routine employed for experimentally obtained iSCAT movies. The background amplitude was set to 2 × 105, which is the well depth of the camera used. A total of four videos were simulated for three different diffusion coefficients, two different noise levels, and with or without considering PSF blurring during the entire camera exposure time: (1) (2) (3) (4)

no blurring, shot noise (n = √2 × 105 = 450); blurred PSF, shot noise; no blurring, elevated noise level (n = 2400); and blurred PSF, elevated noise level (n = 2400).

The elevated noise level, which is given by the parameter noise amplitude n, should account for the higher localization error present on PSMs due to residual background noise or low particle contrast. In the shot noise limit (n ≈ 450), the fit error for a particle with a contrast of 7% is distributed around 2 nm,4 while a noise amplitude of 2400 results in a fit error distribution around 7 nm, which is comparable with our experimental observations (see Figure 3d). For the videos in which blurring was considered, the PSF in each frame was calculated as the normalized sum of PSFs from 200 simulation substeps for each frame.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.8b02240.

■

A figure showing lateral membrane tension in pores. Tables showing estimated diffusion coefficient for simulated trajectories from Monte Carlo simulations, mean diffusion coefficients in the three different pore regions, and T-values and degrees of freedom from a comparison of the distributions in the different pore regions. (PDF)

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. ORCID

Susann Spindler: 0000-0002-2143-417X Claudia Steinem: 0000-0001-8778-9283 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the Max Planck Society, the Alexander von Humboldt Professorship (V.S.), and the Deutsche Forschungsgemeinschaft (RTG 1962, SFB 803, project A05). I


Letter

Nano Letters

■

(20) Krishnan, M.; Mojarad, N.; Kukura, P.; Sandoghdar, V. Geometry-Induced Electrostatic Trapping of Nanometric Objects in a Fluid. Nature 2010, 467, 692−695. (21) Höfer, I.; Steinem, C. A Membrane Fusion Assay Based on Pore-Spanning Lipid Bilayers. Soft Matter 2011, 7, 1644. (22) Goulian, M.; Simon, S. M. Tracking Single Proteins within Cells. Biophys. J. 2000, 79, 2188−2198. (23) Montiel, D.; Cang, H.; Yang, H. Quantitative Characterization of Changes in Dynamical Behavior for Single-Particle Tracking Studies. J. Phys. Chem. B 2006, 110, 19763−19770. (24) Vestergaard, C. L.; Blainey, P. C.; Flyvbjerg, H. Optimal Estimation of Diffusion Coefficients from Single-Particle Trajectories. Phys. Rev. E 2014, 89, 022726. (25) Reddy, A. S.; Warshaviak, D. T.; Chachisvilis, M. Effect of Membrane Tension on the Physical Properties of DOPC Lipid Bilayer Membrane. Biochim. Biophys. Acta, Biomembr. 2012, 1818, 2271− 2281. (26) Mey, I.; Stephan, M.; Schmitt, E. K.; Müller, M. M.; Ben Amar, M.; Steinem, C.; Janshoff, A. Local Membrane Mechanics of PoreSpanning Bilayers. J. Am. Chem. Soc. 2009, 131, 7031−7039. (27) Muddana, H. S.; Gullapalli, R. R.; Tabouillot, T.; Butler, P. J. Physiological Membrane Tension Causes an Increase in Lipid Diffusion: A Single Molecule Fluorescence Study. Biophys. J. 2009, 96, 197a−198a. (28) Sonnleitner, A.; Schuetz, G. J.; Schmidt, T. Free Brownian Motion of Individual Lipid Molecules in Biomembranes. Biophys. J. 1999, 77, 2638−2642. (29) Weber, S. C.; Thompson, M. A.; Moerner, W. E.; Spakowitz, A. J.; Theriot, J. A. Analytical Tools To Distinguish the Effects of Localization Error, Confinement, and Medium Elasticity on the Velocity Autocorrelation Function. Biophys. J. 2012, 102, 2443−2450. (30) Michalet, X.; Berglund, A. J. Optimal Diffusion Coefficient Estimation in Single-Particle Tracking. Phys. Rev. E Stat Nonlin Soft Matter Phys. 2012, 85, 061916. (31) Bussell, S. J.; Koch, D. L.; Hammer, D. A. Effect of Hydrodynamic Interaction on the Diffusion of Integral Membrane Proteins: Diffusion in Plasma Membranes. Biophys. J. 1995, 68, 1836− 1849. (32) Angelova, M. I.; Soléau, S.; Méléard, P.; Faucon, F.; Bothorel, P. Preparation of Giant Vesicles by External AC Electric Fields. Kinetics and Applications. Trends in Colloid and Interface Science VI. 1986, 89, 127−131.

REFERENCES

(1) Sackmann, E. Supported Membranes: Scientific and Practical Applications. Science 1996, 271, 43−48. (2) Przybylo, M.; Sýkora, J.; Humpolíckova, J.; Benda, A.; Zan, A.; Hof, M. Lipid Diffusion in Giant Unilamellar Vesicles Is More Than 2 Times Faster Than in Supported Phospholipid Bilayers Under Identical Conditions. Langmuir 2006, 22, 9096−9099. (3) Guo, L.; Har, J.; Sankaran, J.; Hong, Y.; Kannan, B.; Wohland, T. Molecular Diffusion Measurement in Lipid Bilayers Over Wide Concentration Ranges: A Comparative Study. ChemPhysChem 2008, 9, 721−728. (4) Hsieh, C.-L.; Spindler, S.; Ehrig, J.; Sandoghdar, V. Tracking Single Particles on Supported Lipid Membranes: Multimobility Diffusion and Nanoscopic Confinement. J. Phys. Chem. B 2014, 118, 1545−1554. (5) Shoaib, T.; Nalam, P. C.; He, Y.; Chen, Y.; Espinosa-Marzal, R. M. Assembly, Morphology, Diffusivity, and Indentation of HydrogelSupported Lipid Bilayers. Langmuir 2017, 33, 7105−7117. (6) Müller, P.; Rudin, D. O.; Tien, H. T.; Wescott, W. C. Reconstitution of Cell Membrane Structure in Vitro and Its Transformation Into an Excitable System. Nature 1962, 194, 979− 980. (7) Weiß, K.; Neef, A.; Van, Q.; Kramer, S.; Gregor, I.; Enderlein, J. Quantifying the Diffusion of Membrane Proteins and Peptides in Black Lipid Membranes with 2-Focus Fluorescence Correlation Spectroscopy. Biophys. J. 2013, 105, 455−462. (8) Honigmann, A.; Walter, C.; Erdmann, F.; Eggeling, C.; Wagner, R. Characterization of Horizontal Lipid Bilayers As a Model System To Study Lipid Phase Separation. Biophys. J. 2010, 98, 2886−2894. (9) Schütte, O. M.; Mey, I.; Enderlein, J.; Savic, F.; Geil, B.; Janshoff, A.; Steinem, C. Size and Mobility of Lipid Domains Tuned by Geometrical Constraints. Proc. Natl. Acad. Sci. U. S. A. 2017, 114, E6064−E6071. (10) Heinemann, F.; Schwille, P. Preparation of Micrometer-Sized Free-Standing Membranes. ChemPhysChem 2011, 12, 2568−2571. (11) Sumitomo, K.; Tamba, Y.; Shinozaki, Y.; Torimitsu, K. Confinement of Fluorescent Probes in Microwells on Si Substrates by Sealing with Lipid Bilayers. Appl. Phys. Express 2010, 3, 107001. (12) Janshoff, A.; Steinem, C. Mechanics of Lipid Bilayers: What Do We Learn From Pore-Spanning Membranes? Biochim. Biophys. Acta, Mol. Cell Res. 2015, 1853, 2977−2983. (13) Sumitomo, K.; McAllister, A.; Tamba, Y.; Kashimura, Y.; Tanaka, A.; Shinozaki, Y.; Torimitsu, K. Ca2+ Ion Transport Through Channels Formed by Alpha-Hemolysin Analyzed Using a Microwell Array on a Si Substrate. Biosens. Bioelectron. 2012, 31, 445−450. (14) Urban, M.; Kleefen, A.; Mukherjee, N.; Seelheim, P.; Windschiegl, B.; Vor der Brüggen, M.; Kocer, A.; Tampé, R. Highly Parallel Transport Recordings on a Membrane-on-Nanopore Chip at Single Molecule Resolution. Nano Lett. 2014, 14, 1674−1680. (15) Kuhlmann, J. W.; Junius, M.; Diederichsen, U.; Steinem, C. SNARE-Mediated Single-Vesicle Fusion Events with Supported and Freestanding Lipid Membranes. Biophys. J. 2017, 112, 2348−2356. (16) Lindfors, K.; Kalkbrenner, T.; Stoller, P.; Sandoghdar, V. Detection and Spectroscopy of Gold Nanoparticles Using Supercontinuum White Light Confocal Microscopy. Phys. Rev. Lett. 2004, 93, 037401. (17) Spindler, S.; Ehrig, J.; König, K.; Nowak, T.; Piliarik, M.; Stein, H. E.; Taylor, R. W.; Garanger, E.; Lecommandoux, S.; Alves, I. D.; Sandoghdar, V. Visualization of Lipids and Proteins at High Spatial and Temporal Resolution via Interferometric Scattering (iSCAT) Microscopy. J. Phys. D: Appl. Phys. 2016, 49, 274002. (18) Jiménez, C.; Perrière, J.; Vickridge, I.; Enard, J. P.; Albella, J. M. Transformation of Silicon Nitride in Oxygen Plasma. Surf. Coat. Technol. 1991, 45, 147−154. (19) Reddy, B. S.; Chatterji, B. N. An FFT-Based Technique for Translation, Rotation, and Scale-Invariant Image Registration. IEEE Transactions on Image Processing 1996, 5, 1266−1271. J


High-Speed Microscopy of Diffusion in Pore ... - ACS Publications

Recommend Documents