Transparent Nanopore Cavity Arrays Enable Highly ParallelizedOptical Studies of Single Membrane Proteins on ChipTim Diederichs,†,∇ Quoc Hung Nguyen,‡,∇ Michael Urban,† Robert Tampe,*,†,§

and Marc Tornow*,‡,∥,⊥

†Institute of Biochemistry, Biocenter, Goethe University Frankfurt, Max-von-Laue-Str. 9, 60438 Frankfurt/M., Germany‡Molecular Electronics, Technical University of Munich, Theresienstrasse 90, 80333 Munich, Germany§Cluster of Excellence Frankfurt (CEF) Macromolecular Complexes; Goethe University Frankfurt, Max-von-Laue-Strasse 9, 60438Frankfurt/M., Germany∥Fraunhofer Research Institution for Microsystems and Solid State Technologies (EMFT), Hansastrasse 27d, 80686 Munich,Germany⊥Center for NanoScience (CeNS), Ludwig-Maximilians-University, Geschwister-Scholl Platz 1, 80539 Munich, Germany

*S Supporting Information

ABSTRACT: Membrane proteins involved in transport processes are keytargets for pharmaceutical research and industry. Despite continuousimprovements and new developments in the field of electrical readouts forthe analysis of transport kinetics, a well-suited methodology for high-throughput characterization of single transporters with nonionic substratesand slow turnover rates is still lacking. Here, we report on a novel architectureof silicon chips with embedded nanopore microcavities, based on a silicon-on-insulator technology for high-throughput optical readouts. Arrays containingmore than 14 000 inverted-pyramidal cavities of 50 femtoliter volumes and 80nm circular pore openings were constructed via high-resolution electron-beamlithography in combination with reactive ion etching and anisotropic wetetching. These cavities feature both, an optically transparent bottom and topcap. Atomic force microscopy analysis reveals an overall extremely smoothchip surface, particularly in the vicinity of the nanopores, which exhibits well-defined edges. Our unprecedented transparent chip design provides parallel and independent fluorescent readout of both cavitiesand buffer reservoir for unbiased single-transporter recordings. Spreading of large unilamellar vesicles with efficiencies up to 96%created nanopore-supported lipid bilayers, which are stable for more than 1 day. A high lipid mobility in the supportedmembrane was determined by fluorescent recovery after photobleaching. Flux kinetics of α-hemolysin were characterized atsingle-pore resolution with a rate constant of 0.96 ± 0.06 × 10−3 s−1. Here, we deliver an ideal chip platform for pharmaceuticalresearch, which features high parallelism and throughput, synergistically combined with single-transporter resolution.

KEYWORDS: Silicon-on-insulator chips, transport kinetics, membrane proteins, optical readout, supported lipid bilayers, nanopores,microcavities

To date, membrane proteins (MPs) are targeted by morethan half of all developed drugs,1 underlining their

outstanding importance for pharmaceutical industry and basicresearch. They are involved in a multitude of essential cellularprocesses including bioenergetics, transport, and communica-tion. Especially, their role in trafficking substrates across lipidbilayers or in artificial nanocontainers became an importantresearch focus for the identification of new drug targets as wellas for biosensing applications.2

Multiple techniques for the characterization of MP kineticsare available. For instance, patch clamp is able to directlyanalyze ion channel activity in cell membranes, however, it islimited to single or several simultaneous channel recordings.3−6

High-throughput screening applications are confined by thecomplexity of interfaces between electronics and ionic systems,7

except for some advanced techniques using whole-cell orautomated patch-clamp recordings.8−11 For parallel analysis ofelectrogenic transporter activities, an alternative electrophysio-logical method has been based on solid-supported mem-branes.12,13 Yet, this technique is unable to achieve the single-transporter level. While these electrode-based approaches arecapable of measuring ionic fluxes or the translocation of ionicsubstances they still lack the efficiency or ability to characterizetransporters with nonelectrogenic solutes. Other tools for theanalysis of MPs have been utilized, including tetheredbilayers,14,15 native vesicle arrays,16−21 and micro-black-lipidmembranes,3,22−25 which are compatible with nonionic cargos.

Received: March 28, 2018Published: May 9, 2018

Letter

pubs.acs.org/NanoLettCite This: Nano Lett. 2018, 18, 3901−3910

© 2018 American Chemical Society 3901 DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

Dow

nloa

ded

via

UN

IV F

RA

NK

FUR

T o

n O

ctob

er 2

3, 2

018

at 1

0:48

:03

(UT

C).

Se

e ht

tps:

//pub

s.ac

s.or

g/sh

arin

ggui

delin

es f

or o

ptio

ns o

n ho

w to

legi

timat

ely

shar

e pu

blis

hed

artic

les.

So far, these applications are unsuitable for highly parallelizedkinetic studies of MPs.In the past decade, supported lipid bilayers (SLBs) have

gained attention due to their long-term stability and their abilityto mimic native membranes.26,27 However, the incorporation oftransmembrane proteins is inherently limited because of theirclose proximity to the solid support.15,28 To overcome thisdrawback, silicon-based microcavities were spanned withmembranes creating a platform of supported and free-standinglipid bilayers with techniques such as spreading of giantunilamellar vesicles (GUVs),29 large unilamellar vesicles(LUVs),30−32 or shear-driven SLB formation.33 In order totransfer the methodology from laboratory research to industrialscreening applications, it is required to have in vitro systems (i)revealing highly automated channel recordings for high-throughput applications, (ii) minimizing the sample con-sumption, and (iii) possessing long-term stability to ensurerepeatability without suffering from degradation of sensorinterfaces.34 Recent advances in semiconductor fabricationtechniques of micro- and nanostructures with well-defineddimensions as well as the development in the field of solidsupported membranes provide a promising direction forstudying MP mediated transport processes.28,35

Silicon-based chips containing arrays of nanopores withmicrocavities combined with fluorescent readout are in thefocus of particular interest, as they enable highly parallel proteinrecordings across lipid membranes. Nonelectrogenic membranetransport proteins can be incorporated via self-spreading ofliposomes and their kinetics can be investigated, which is,however, not feasible with electrical setups.34 Only a fewexamples for multiplexed analysis using silicon arrays werepublished, including the characterization of the mechanosensi-tive channel of large conductance (MscL),36 the FOF1-ATPsynthase, and α-hemolysin.37,38 Recently, SLBs were producedwith the aid of organic solvent across femtoliter- or attoliter-sized cavities,39,40 suffering from the fact that organic solvents

may be harmful to MPs. Kleefen and colleagues used nanoporeswith femtoliter compartments, fabricated by electron-beamlithography (EBL), reactive dry, and anisotropic wet etching ofa Si3N4 covered silicon chip.38 In particular, the bulk bottomhandle wafer substrate was fully opaque for visible light, hencenot accessible for common backside fluorescent microscopy butlimited to upward oriented microscopes.38 Because of thislimitation and owing to the large buffer backgroundfluorescence intensity, it was not possible to separatefluorophores in cavities and buffer reservoir for unbiasedsingle-transporter recordings. Nevertheless, fluxes of fluorescentsubstances through α-hemolysin and saponin were charac-terized at single-protein resolution. Urban and colleaguesutilized silicon-on-insulator (SOI) substrates, which wereprocessed via reactive ion etching (RIE) creating cylindricalcavities.36 Chemical vapor deposition of silicon dioxide (SiO2)was applied to reduce the cylindrical micrometer openings tonanometer scales.36 Thereby, rough surfaces with intrusions ofSiO2 were obtained, leading to diminished spreading abilities aswell as limited diffusion properties for encapsulated substances.Nevertheless, an analysis of MP kinetics was presented forMscL and α-hemolysin.36,41

Here, we present a novel architecture of silicon chipscontaining macroscopic-scale arrays of optically transparentnanopore microcavities that feature a superior homogeneity inshape and size, constituting well-defined femtoliter compart-ments in SOI substrates. Our unprecedented chip architecturecombines the advantages of both previously reported chipstructures: (i) extremely smooth, low-strain surface and poresurroundings with well-defined and tapered pore edges, and (ii)top and bottom chip transparency, allowing for full opticalreadout access for inverted microscopy. Hence automated andmultiplexed readout as well as screening become possible.Cavity volume, nanopore diameter, and the patterning of cavityarrays (size and shape) can be precisely controlled on waferscales via high-resolution EBL in combination with RIE and

Figure 1. Scheme of the fabrication process of nanopore cavity arrays in SOI wafer. (A) A large pyramidal etch pit connecting the chip bottom to theBOX layer is created by optical lithography, RIE, and KOH (aq) wet etching. (B,C) nanopores (array, 120 × 120 pores) with 80 nm diameter in theSi3N4 top layer are patterned by EBL and another RIE step. (D) A second wet chemical etch in KOH (aq) of the chip results in the formation ofinverted-pyramidal cavity arrays. Because SiO2 serves as etch stop layer, the pyramid tip is truncated at the cavity bottom after etching. Dimensionsare not drawn to scale.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3902

anisotropic wet etching. Our chips contain more than 14 000inverted-pyramidal cavities arranged in a rectangular patternwith pore distances of 10 μm and openings of 80 nm. Asvalidated by atomic force microscopy (AFM) analysis, thesurfaces are extremely smooth on pore surroundings with well-defined and tapered pore edges, hence providing optimalspreading conditions for lipid bilayers.42 The material proper-ties of the employed SiO2 and Si3N4 layers allow an opticallytransparent chip design for parallel fluorescence readout of bothcavities and buffer reservoir for unbiased single-transporterrecordings in a highly parallel fashion. We show that asanticipated, our chip platform indeed features high spreadingefficiencies of large unilamellar vesicles (LUV), high lipidmobility, and long-term stability of the free-standing, nanopore-supported lipid bilayer. As proof of concept, kinetics of themodel α-hemolysin toxin were analyzed down to single-poreresolution in high-throughput.Experimental section. Silicon-on-Insulator Substrate.

Si3N4 coated, prime-quality 4-in. diameter SOI wafers (Si-Mat, Kaufering, Germany) were composed of a top 3.0 ± 0.5μm thick undoped silicon (100) device layer, which wasseparated from the 380 ± 15 μm undoped bulk silicon (100)substrate by a 100 ± 10 nm buried oxide layer (SiO2, “BOX”).Stoichiometric Si3N4 (thickness ∼50 nm) was then depositedon both sides of the substrate by low-pressure chemical vapordeposition (LPCVD). The wafers were sawed into 13 × 13mm2 pieces for the following nanopore cavity structurefabrication.Fabrication and Characterization of Nanopore Cavity

Arrays with Transparent Bottoms. The fabrication process ofnanopore cavity arrays is schematically depicted in Figure 1.First, a large pyramidal pit was structured on the backside of thechip by a combination of optical lithography, RIE, and aqueouspotassium hydroxide (KOH (aq)) wet etching. For this step,Shipley S1818 photoresist was spun onto the wafer at a speedof 6000 rpm for 60 s followed by a soft-bake at 115 °C for 3min. We then exposed the resist to UV light (wavelength 365nm, power density ∼6.2 mW/cm2) for 15 s through a squaremask (1.8 × 1.8 mm2) at the center of the wafer using a maskaligner (SUSS MicroTec, Garching, Germany). Subsequently,the chip was developed in DEV 351 solution (diluted 1:5 vol %in DI water) for 30 s, followed by a RIE (Oxford Instruments,Oxfordshire, U.K.) step using a C4F8/O2 gas mixture for 85 s toremove the exposed Si3N4 layer. After rinsing with acetone,isopropanol, and drying with nitrogen flow, the chips wereanisotropically chemically etched in 20 wt % KOH (aq)solution at 80 °C to remove bulk silicon from the backside,resulting in a large pyramidal etch pit. The etching was stoppedafter 3 h when the etch pit reached the BOX layer which servedas an etch stop.For the fabrication of cavity arrays, the chips were extensively

rinsed with water to remove KOH residues prior to thefollowing EBL process. For EBL, an e-beam resist (AR-P 6200,Allresist, Strausberg, Germany) was spun on the chip front sideat 3000 rpm for 60 s and prebaked at 150 °C on a hot plate for1 min, resulting in a resist thickness of ∼100 nm. Arrays ofnanopores (120 × 120 pores) with diameter of 80 nm and poredistance of 10 μm were patterned by an e_LiNE systemequipped with the Elphy Quantum pattern generator (Raith,Dortmund, Germany) at an acceleration voltage of 30 kV andwith beam currents of ∼30 pA. The development of thepatterns was done in developer AR 600-546 (Allresist,Strausberg, Germany) for 1 min. Subsequently, the chips

were rinsed with isopropanol and dried under nitrogen flow.After a postbaking at 130 °C for 1 min, the nanopore patternswere transferred to the Si3N4 layer by RIE, as described above.The residual resist was then removed by immersion in acetoneand isopropanol for 10 min each, followed by drying withnitrogen. Finally, arrays of femtoliter cavities with invertedpyramidal shape were formed by an anisotropic wet etchingstep of the Si device layer in 15 wt % KOH (aq) at 50 °C, usingthe Si3N4 nanopore as an etch mask. Under these conditions,the etching of mainly Si(111) planes proceeds at a very low ratedue to the high etching anisotropy (0.9 ± 0.1 μm h−1). Thefinal cavity volume is adjusted by the etching time. For a 50femtoliter cavity volume, the etching takes about 3.5 h, resultingin cavities with free-standing Si3N4 membranes of ∼36 μm2 insize, forming the cavity tops, and transparent cavity bottomareas of ∼3 μm2.The nanopore cavity arrays were characterized by scanning

electron microscopy (SEM) using the e_LiNE system (Raith,Dortmund, Germany) at an acceleration voltage of 5 kV. Thesurface topography of the Si3N4 membranes on top of thecavities, including the single nanopores, was imaged by high-resolution AFM, utilizing a Veeco Dimension V system (VeecoInstruments Inc., New York, United States) equipped with aVeeco Nanoscope V controller in Tapping Mode. For alltopographic measurements, Si cantilever probes (model: SSS-NCLR (super sharp), Nanosensor, Neuchatel, Switzerland)were used, having nominal force constant of 48 N/m andresonance frequency of 190 kHz. These probes feature a tipradius of about 2 nm, which helps to minimize tip/sampleconvolution effects. The surface roughness was determined asroot-mean-square (rms) value.

Liposome Preparation. LUVs composed of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG) werepurchased from Avanti Polar Lipids Inc. (Hamburg, Germany)and mixed in molar ratios of 4:3:3 to a concentration of 5 mg/mL and subsequently resolved in HEPES buffer (20 mMHEPES/NaOH, 150 mM NaCl, pH 7.4). LUV formation wasconducted as described.38,43 Liposomes were extruded 21 timesthrough 100 nm polycarbonate membranes at the LiposoFast-Basic extruder (AVESTIN, Mannheim, Germany). Vesicle sizeswere analyzed using the nanoparticle tracking system Nano-Sight LM10 (Malvern, Herrenberg, Germany).

Chip Preparation. The SOI chips were glued on 8-wellsticky-slides (ibidi, Planegg/Martinsried, Germany), followedby oxygen plasma cleaning for 2 min at 0.3 mbar and at 80%power settings with a plasma cleaner (Diener Electronics,Ebhausen, Germany). After immersion in 300 μL of ethanol for5 min, successive exchange of ethanol was conducted bydilution with HEPES buffer including 5 mM CaCl2 (totalvolume 500 μL). This step was repeated 15 times to completelydilute ethanol and ensure wetting of the microcavities. Differentdyes were encapsulated via spreading (1 h incubation) of LUVs(1 mg/mL), followed by buffer exchange to remove lipid andfluorophore excess. Optical readout was performed with theconfocal laser scanning microscope (CLSM) LSM 880(AxioObserver from Zeiss, Jena, Germany) equipped with aPlan-Apochromat 20x/0.8 M27 air objective or with theautomated NyONE epifluorescence microscope (SYNENTECGmbH, Elmshorn, Germany) equipped with a 20× magnifica-tion air objective (Olympus UPlan SApo 20x/0.75). Imageswere recorded at different time intervals dependent on the

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3903

experiment. Each assay was stopped by adding 25 μL Triton-X100 10% (v/v) to induce SLB rupture that induce rapid effluxof fluorescent molecules and represented additional proof forthe absence of unspecific binding.Spreading Efficiency, Long-Term Stability, and Lipid

Fluidity of the SLB. The spreading of LUVs and the resultingSLB were visualized via fluorophore encapsulation. Tenmicromolar of ATTO655 (ATTO-TEC, Siegen, Germany)and 5 μM of Oregon Green (OG) dextran (10 kDa)(ThermoFisher Scientific, Darmstadt, Germany) were added,followed by liposome spreading and buffer exchange. Sealingefficiencies were calculated as ratio of dye encapsulated cavitiesregarding total cavities per field of view. Pearson correlationcoefficients were determined for the location of bothfluorophores in the individual cavities with ImageJ 1.51o. Thelong-term stability was monitored at different time points (1, 3,and 24 h) after vesicle spreading with the NyONE microscope.For this purpose, 5 μM of ATTO655 was encapsulated insidethe cavities.FRAP experiments were conducted at the CLSM with LUVs

composed of POPC/POPE/POPG as described and addition-ally doped with 3,3′-dioctadecyl-oxacarbocyanine perchlorate(DiO 0.5 mol %) (ThermoFisher Scientific, Darmstadt,Germany). After spreading, small circular areas of 8 to 15μm2 diameter were bleached within 4 s with high laserintensities. Fluorescence recovery was monitored for ∼1 min in∼100 ms intervals dependent on the bleached area. Thenormalized fluorescence intensity (It) was calculated accordingeq 1

τ τ= + +

−⎡⎣⎢⎢

⎛⎝⎜

⎞⎠⎟⎤⎦⎥⎥⎡⎣⎢⎢⎛⎝⎜

⎞⎠⎟⎤⎦⎥⎥I I I

t t1t o inf

1/2 1/2

1

(1)

in which Io is the initial fluorescence, Iinf is the fluorescenceintensity after recovery, and τ1/2 is the time at half-maximalfluorescence intensity. The lateral lipid diffusion coefficient (D)was determined regarding eq 244−46

γτ

=Dr

4o2

1/2 (2)

with ro as the radius of the excitation spot and γ as thebleaching parameter.47

Single-Transporter Recordings. Ten micromolar ofATTO655 (0.6 kDa) and 5 μM of OG dextran (10 kDa)were encapsulated via liposome spreading. After bufferexchange, α-hemolysin was added to final concentrations of500, 100, or 50 nM. Data were collected in 11 s intervals for 3−4 h with the CLSM. Unspecific binding of enclosed dyes wasforeclosed by addition of 25 μL Triton-X 100 10% (v/v) afterthe experiment. The resulting time traces were processed viathe Zen 2.1 black software from Zeiss and the correspondingmean gray values were calculated via ImageJ, followed byplotting and fitting of the translocation kinetics with theNanocal software (Nanospot GmbH). Two different types ofexponential decays were identified: (i) monoexponential decaysand (ii) more sigmoidal decreasing curves abruptly convertingto exponential decays. Both types of curves were fitted withmonoexponential equations, whereas the inflection point ofcurve type (ii) was used as lower fit boundary (see SI Figure 6).First-order rate constants (keff) were further processed withOrigin 9.1 Pro (OriginLab), including descriptive statistics tocount traces in individual intervals, peak analysis (Origin 9.1Pro, peak analyzer), and Gaussian fitting.38 Rate constant errorswere obtained for both the fitting error of the singleexponentials and the Gaussian distribution fit of the histograms.To provide an upper bound estimate for the total error, we listthe larger derived value of both in each case, respectively.

Figure 2. Silicon nanopore chip design for multiplexed parallel recordings. Eighty nanometer pores in a Si3N4 layer were fabricated via EBL and RIE,followed by wet etching with KOH (aq) to create more than 14 000 inverted-pyramidal cavities per chip. (A) SEM micrograph of a portion of thewhole homogeneous array with nanopore cavities appearing as dark gray squares. (B,C) SEM images presenting a single cavity, disclosing a singlenanopore in the center of the Si3N4 membrane with diameter of ∼80 nm. (D) Front side view of a single cavity after removal of most of the Si3N4membrane, disclosing the truncated, inverted pyramidal shape of the cavity with transparent bottom (black square). (E−G) Images of the chipbackside after the final KOH etch step showing an array of homogeneous but smaller squares compared to the front side. These squares correspondto the truncated, inverted pyramid tops, hence to the transparent cavity bottoms (cf. dark square in panel D). (H) AFM topography showing a 3Drendered image of a nanopore with highly rounded and smooth edges of the pore opening.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3904

Results and Discussion. Device Fabrication andCharacterization. Nanopore cavity arrays with transparentcavity bottoms were realized by a combination of differentprocesses that included optical lithography, EBL, RIE, andKOH (aq) wet etching. The fabrication flowchart is schemati-cally shown in Figure 1 and detailed in Experimental Section. ASOI wafer with 3 μm silicon (100) device layer and 100 nmburied oxide (BOX) was used as a starting substrate, which wasthen coated with 50 nm Si3N4 by LPCVD on both sides. Asquare window in the Si3N4 layer on the backside of the waferwas processed by optical lithography and RIE. Subsequently, ananisotropic wet etch in KOH solution was performed toremove the thick bulk Si through this window, resulting in alarge etch pit of pyramidal shape truncated at the BOX layer(etch stop layer; Figure 1A). Here, the shape of a largepyramidal cavity is formed because the etch rate of Siperpendicular to the (100) crystal planes is much higher thanperpendicular to the (111) planes.48 In 20% (wt/v) aqueousKOH solution at 80 °C, we observed that the ratio of etch rates(100)/(111) ranged from 45:1 to 50:1.In the second step, a square array of nanopores (120 × 120

pores) with a pore diameter of 80 nm and pore distance of 10μm was patterned in the Si3N4 top layer by high-resolution EBLand RIE (Figure 1B,C). Finally, the wafer was etched again inKOH solution to fabricate a homogeneous nanopore cavityarray (Figure 1D). The size of the cavities was well controlledby adjusting the etching time (see Experimental Section).The nanopore cavity array was characterized by SEM. The

cavities form a well-ordered array with highly homogeneouscavity sizes (Figure 2A). The Si3N4 layers on top of the cavitiesare intact and appear as dark gray squares as they are almosttransparent for the electron beam. An image of a single cavityreveals a single pore in the very center of the Si3N4 membranewith a diameter of ∼80 nm (Figure 2B,C). The magnificationof one cavity with the Si3N4 top layer almost removed confirmsthe inverted pyramidal shape of the cavity (Figure 2D). The

pyramid tip, visible as a small black square, is truncated at thecavity bottom composed of a SiO2 layer.An SEM investigation from the chip backside verifies that the

cavities are homogeneous in size (Figure 2E−G). The cavityarray was well aligned with the BOX layer window exposedfrom the backside, both in the center of the chip. Thedimensions of the cavity top and bottom squares can beaccurately measured by SEM. By further knowing that the anglebetween the cavity top (Si (100) plane) and the cavity sidewalls(Si (111) plane) is 54.74° and that the height of the cavities is3.0 ± 0.5 μm (Si device layer thickness), the volume of thecavities can be calculated. In this study, chips with cavityvolumes of 50 ± 3 femtoliters were used.The topography of the nanopores was investigated by high-

resolution AFM in tapping mode using a Si cantilever probefeaturing a tip with radius ∼2 nm, thereby minimizing tip/sample convolution effects. Figure 2H shows a 3D renderedAFM image of a nanopore, revealing that the nanopore openinghas well curved and smooth edges. The topography of theSi3N4 layer in the area of a certain nanopore cavity was alsodetailed. SI Figure S1A shows a single cavity with the areas inthe middle (freely suspended Si3N4 layer) appearing to belower in height than the rest of the Si3N4 surface. From theaverage height profiles of the marked regions on this image inboth horizontal (blue) and vertical (red) direction (SI FigureS1B), we observed that the freely suspended Si3N4 layer on topof the cavity is weakly bent downward, from the four edgestoward the center of the cavity top where the nanopore islocated (downward spikes). This is likely due to small residualstress within the deposited Si3N4 layer resulting in visible strainupon release. Notably, the overall height variance is stillextremely small with respect to the lateral dimensions of thecavities, by about a factor of 1/5000, which is much less thanwhat we have reported for freely suspended silicon layersreleased from SOI.29,49 SI Figure S1C and D present AFMheight images of Si3N4 surfaces in a small area close to the poreopening and in an area where no cavities were structured

Figure 3. Sealing efficiency. Ten micromolar of ATTO655 (0.6 kDa) and 5 μM of OG dextran (10 kDa) were incubated in HEPES buffer with 5mM CaCl2. Liposomes were added to a total concentration of 1 mg/mL followed by self-spreading, creating an SLB. After buffer exchange, thecavities were imaged by CLSM. Colored cavities (merge channel: both dyes were encapsulated) indicate an intact SLB whereas unsealed cavities orareas with membrane defect and leakage remain dark. (A) Field of view in a CLSM. Overall, a representative sealing efficiency of 94% and a Pearsoncorrelation coefficient of 0.93 for the location of both fluorophores in the individual cavities were observed. (B) The homogeneity of fluorophoreencapsulation as well as the high signal-to-background ratios are illustrated as magnifications of sections of the images in A (white squares).

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3905

showing ultraflat surfaces. The rms roughness in both cases isapproximately 0.2 nm. Hence together, practically stress-free Si-nitride layers deposited on polished, prime-quality Si wafersconstitute an ideal starting system for the preparation of bothultraflat and smooth suspended solid-state interface as supportfor lipid membranes. We would like to emphasize that both thehighly smooth surface around the pore area and the roundedpore edges are essential for the successful spreading of artificialmembranes from LUVs, as confirmed below.SLB: Spreading Efficiency, Long-Term-Stability, and

Mobility. The formation of pore-spanning lipid bilayers occursvia self-spreading of liposomes. Vesicles with an averagediameter of ∼150 ± 27 nm (SI Figure S2) were used forSLB formation induced by adding CaCl2 (5 mM final).Shielding effects of the divalent cations allow to overcome therepulsion forces and the vesicles adhere via van der Waals andelectric interactions to the solid support, similar to spreading atSiO2 surfaces.50 The ratio between nanopore opening andvesicle diameter is critical for successful spreading.38,51

Liposome intrusion into the nanopores was avoided by vesiclesizes larger than the nanopore openings.36 A systematic analysisof nanopore and lipid vesicle sizes demonstrated that pores of80 nm result in the highest spreading efficiency for LUVs of∼150 nm diameter.38

Our solvent-free method ensures a spatial compartmentaliza-tion of cavities and buffer reservoir, which can be visualized viathe strict separation of fluorescent dyes. Ten micromolar ofATTO655 (0.6 kDa) as well as 5 μM of OG dextran (10 kDa)were selectively trapped by SLB formation inside the cavities.After buffer exchange, cavities with bilayer defects did not showfluorescent signals based on rapid leakage out of the cavitiesand infinite dilution into the buffer reservoir. In contrast, areaswith nanopore-supported membranes show well-definedsquared patterns of fluorescence, separated by a center-to-center distance of 10 μm. A Pearson correlation coefficient of0.93 (Figure 3A, merge channel) was obtained for the localcorrelation of ATTO655 (0.6 kDa) and OG dextran (10 kDa)in the underlying cavities. These findings are consistent withperfect colocalization of both dye molecules in the micro-cavities. Colocalization deviations, lacking either green or redfluorescence, can be caused by variations of the encapsulationefficiencies. The latter are mainly influenced by the physical

properties of the fluorophores, for example, size and hydro-phobicity, especially of the dextran conjugate. We like to notethat regions of interest without a defined colocalization wereneglected in further analysis. The homogeneity of fluorophoreencapsulation, an optimal cavity filling, and the high signal-to-background ratio of our novel device are illustrated asmagnifications of sections of Figure 3A, cf. Figure 3B below.Optimal conditions for optical single-transporter trans-

location assays were provided by perfect wetting abilities andlow unspecific binding of fluorescent dyes (SI Figure S3).Furthermore, compartmentalization of fluorophores in cavitiesand buffer reservoir can be achieved and imaged due to thetransparency of the chip design. For this purpose, ATTO655was trapped via vesicle spreading, followed by buffer exchangeand OG dextran addition (SI Figure 4, merge channel). Thefocus area of the CLSM was adjusted horizontally to thecavities to visualize solely the chip cavities. The rounded edgeshape and the surface of the pore surroundings withsubnanometer roughness revealed perfect conditions forliposome spreading with 80−96% sealing efficiency.42

As illustrated above, advanced applications will be theinvestigations of slow, nonionic transporters. Therefore, longobservation periods with suspended lipid bilayers stable formultiple hours are required. The long-term stability wasanalyzed for 24 h and visualized via selectively entrappedATTO655 molecules. Previous studies revealed high SLBstabilities for nanopores in the range of 100 nm.42 Our SLBs arestable for 3 h without any leakage of fluorophores (see Figure4A). Even after 24 h, more than 67% of the cavities are stillsealed. Notably, fluorophore bleaching as well as membranepermeability can be neglected due to millisecond illuminationtimes and the low membrane interaction factor of ATTO655.52

Unspecific fluorophore occlusion was ruled out by addingTriton-X 100 10% (v/v) after 24 h to solubilize the SLB.Continuous homogeneous coverage as well as the dynamic

properties of the SLB, which are important to preserve thephysiological environment of MPs, were examined via FRAP onthe SLBs doped with DiO (0.5 mol %). Fluorescence recoveryof 97.1 ± 0.3% and a lateral diffusion coefficient of 0.85 ± 0.20μm2 s−1 are consistent with an intact homogeneous, fluidsuspended lipid bilayer determined in nine independentmeasurements. These values are in agreement with the lateral

Figure 4. Long-term stability and high lipid mobility of the SLB. (A) Five micromolar of ATTO655 was encapsulated inside the cavities via self-spreading of liposomes, followed by buffer exchange and imaging with epifluorescence microscopy. Images were recorded at different time pointsafter liposome speading (1, 3, and 24 h). Adding Triton-X 100 10% (v/v) induces bilayer rupture and the release of the compartmentalizedfluorophore molecules. (B) LUVs doped with DiO (0.5 mol %) were spread on the chip surface and a representing FRAP was recorded with theCLSM.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3906

diffusion coefficient of lipids measured on glass or silicon basedsurfaces, which range from 1 to 3 μm2 s−1 (Figure 4B).53−55

In summary, all results reveal a successful spreading process,resulting in a stable impermeable and extremely fluidicsuspended lipid bilayer, which is able to spatially andtemporally compartmentalize molecules in cavities and bufferreservoir. The barrier was used to investigate the transportprocesses via optical readout and ensures a nature-likeenvironment for integral proteins.Single-Transporter Recording of the Membrane Protein

Complex α-Hemolysin. α-Hemolysin was chosen as modelmembrane protein complex to investigate MPs at single-

transporter resolution and in high-throughput. This bacterialpore-forming toxin binds to the choline groups of phospho-lipids, followed by oligomerization and spontaneous insertioninto lipid bilayers.41 The crystal structure of this heptamericprotein revealed a pore with its narrowest opening of 1.4nm.56−59 Nevertheless, apertures up to 2.9 nm were reported,dependent on lipid composition and buffer/salt conditions,allowing the diffusion of molecules up to 2 kDa in size.56 Priorto SLB formation, 10 μM of ATTO655 (0.6 kDa) and 5 μM ofOG dextran (10 kDa) were trapped inside the cavities followedby buffer exchange and protein addition. α-Hemolysindiscriminates fluorophores based on their molecular sizes.

Figure 5. Flux kinetics of single α-hemolysin pores. Ten micromolar of ATTO655 (0.6 kDa) and 5 μM of OG dextran (10 kDa) werecompartmentalized inside the cavities, followed by buffer exchange and imaging via CLSM. The keff values were determined for single and doubleincorporations of α-hemolysin after separation in 35 classes for each histogram. (A) Schematic illustration of the transport experiments with possiblescenarios including α-hemolysin-mediated transport, rupture, and cavities without fluorescence changes. (B) Exemplary efflux traces of ATTO655 atdifferent pores at the same chip, mediated by α-hemolysin. (C−E) Histograms of the ATTO655 efflux rates for 500, 100, and 50 nM of α-hemolysin,respectively.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3907

Here, ATTO655 acts as a flux reporter whereas the dextran-coupled fluorophore serves as membrane impermeable control.Hence, SLB rupture is indicated by a rapid efflux of bothfluorophores, while α-hemolysin-mediated translocation eventsare classified by an exponential decline of the ATTO655 signaland a constant OG signal (Figure 5A). In total, four differentsituations were observed: (i) α-hemolysin-mediated transport,(ii) membrane rupture, (iii) complex kinetics, and (iv)undefined events. The latter mainly comprised events withoutcontrol signals of the OG channel, whereas complex kineticsinclude processes with two declining steps or spontaneoussignal increases in one of the two channels. Only tracesdisplaying single-exponential decays in the ATTO655 channel(red) together with a constant signal in the OG channel(green) were accepted for final analysis. Exemplary efflux tracesof ATTO655 are shown in Figure 5B. On the basis of thespontaneous insertion and oligomerization of the toxin, thesingle-exponential decline occurs at different time points, whichis similar to previous studies.38

Single-transporter recordings were successively investigatedby decreasing the protein concentration. α-Hemolysin coalescesto heptameric pores before those insert into the membrane.60

The likelihood for multiple protein insertion per silicon nitridenanopore drastically diminishes with declining proteinconcentration. We started by applying α-hemolysin (500 nMfinal), which has been reported to address the range of single-transporter recordings.38,41 The efflux rates (keff) of 275 α-hemolysin traces are summarized in a histogram composed of35 classes. A broad Gaussian distribution was observed with anaverage efflux value of 3.29 ± 0.23 × 10−3 s−1 (Figure 5C).Contrary to our observation, single-translocation processes areexpected to consist of narrow distributions, as shown for theprotein MscL.36 To identify single-transporter events theconcentration of α-hemolysin was further decreased succes-sively down to 100 and 50 nM (Figure 5D,E, respectively).Two hundred and four traces of α-hemolysin-mediated

translocation events were analyzed after adding α-hemolysin(100 nM final). The classification as well as the intervals werechosen as described (Figure 5D). Two populations wereidentified by peak analysis and fitted with Gaussiandistributions. The average efflux values of the first Gaussianpeak (red) at 1.23 ± 0.25 × 10−3 s−1 clearly suggest that at 500nM of α-hemolysin (see Figure 5C) single-transporterrecordings were superimposed by the elevated statistics ofmultiple protein insertions per silicon nanopore. This statementis corroborated by similar average efflux values of 3.25 ± 0.44 ×10−3 s−1 at 100 nM of α-hemolysin (Figure 5D, blue curve) inrespect to the average efflux value of 3.29 ± 0.23 × 10−3 s−1 at500 nM of α-hemolysin (Figure 5C).The single-transporter regime was reached by finally

decreasing the concentration of α-hemolysin down to 50 nM.Here, 117 efflux traces were analyzed and distributed asdescribed above, and two distinct Gaussian peaks wereidentified. The first population (red) raises at the expense ofthe second one (blue) (Figure 5E). The 10-fold lowerconcentration than previously reported for single α-hemolysinrecordings,38,41 in combination with only 4.8% of sealed cavitiesshowing monoexponential decays, indicates that the single-transporter regime has been reached because the likelihood ofprotein insertion per silicon nitride nanopore has decreaseddrastically. We interpreted our findings as incorporation of oneor two α-hemolysin molecules inside a single silicon nitridenanopore. Thereby, we identified an average efflux value of 0.96

± 0.06 × 10−3 s−1 for single- and an average efflux constant of2.89 ± 0.12 × 10−3 s−1 for double-protein recordings (Figure5E). By decreasing the α-hemolysin concentration, the mainpeak shifted from two protein insertions per nanopore in panelC to one protein per nanopore in panel E.Slight shifts in the exact average efflux values at all three

concentrations can be explained by statistic variations,especially for the final concentration of 50 nM α-hemolysin.However, all corresponding values are comparable in the rangeof error. Nevertheless, the concentration series of α-hemolysinclearly demonstrates that our silicon chips are suitable forsingle-transporter recordings. Additionally, the efflux values forsingle α-hemolysin recordings are in the range of previous datafor MscL.36 Efflux rates of α-hemolysin in the order of 10−5 s−1

were not reproduced.41 However, a later study showed single-protein recordings of α-hemolysin with rates of 5.5 × 10−4

s−1.37 Their results are in good agreement with our findings andsmall deviations occur based on different fluorophores, lipids,buffer/salt conditions, and variations in chip architecture andvolumes. These parameters make it difficult to compare theexact α-hemolysin.efflux kinetics. Nevertheless, our values are inthe range of reported α-hemolysin efflux rates, whichdemonstrates that our novel chip design is suitable for single-protein recordings.37

We finally tested if the efflux kinetics are altered byencapsulation of two dyes in one cavity. Therefore, 10 μM ofATTO655 was encapsulated inside the cavity and 5 μM of OGdextran (10 kDa) was applied after SLB formation (SI FigureS4). α-Hemolysin was added (500 nM final) and 186 traceswere analyzed resulting in an average efflux value of 3.40 ± 0.13× 10−3 s−1 (SI Figures S5A and S6), consistent with the resultsabove (Figure 5C). Furthermore, rupture events of the SLBwere analyzed revealing more than 10-fold faster efflux rates forthe diffusion of 10 μM ATTO655, indicating that our single-protein recordings are not biased by lipid rupture processes (SIFigure S5B). In short, single-pore efflux kinetics of α-hemolysinat different concentrations were presented that demonstrate theviability of the transparent nanopore-chips for high-throughputsingle-transporter studies.

Conclusion. We presented a novel SOI-based nanoporecavity platform for highly parallelized single-transporter studiesusing optical readout. The fabrication of the chips, whichcomprise large, homogeneous arrays of cavities with both topand bottom optical transparency, were controlled precisely onwafer-scale, delivering ideal conditions for screening ap-proaches. A homogeneous coverage of the chip surface withan SLB was verified by fluorescent dye encapsulationexperiments. On the basis of the transparent nature of SiO2and Si3N4 it is possible to track influx as well as efflux processesof fluorescent solutes with high contrast. Single-proteinresolution was clearly demonstrated with the bacterial toxinα-hemolysin. The self-spreading process allows the use ofproteoliposomes. Thereby, reconstitution of MPs in liposomescan be used to make the presented chip design suitable formedically relevant transporters, like the transporter associatedwith antigen processing (TAP),61 the transporter associatedwith antigen processing like (TAPL),62 or Thermus thermophi-lus multidrug resistance proteins A and B (TmrAB), afunctional homologue of human TAP.63 The high lipid mobilityas well as the long-term stability of the SLB create a perfectplatform for the characterization of nonionic transporters insingle-protein resolution, which we intend to elaborate on inthe future. By reducing the cavity volume and increasing the

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3908

number of microcavities per field of view, higher parallelism canbe accomplished, based on thinner top silicon layerarchitectures and smaller pore-to-pore distances. Further,vector-based cluster analysis may become a future alternativeapproach to data analysis for extracting pore populations withtransporters.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.nano-lett.8b01252.

Figures S1−S6 (PDF)

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: (R.T.) tampe@em.uni-frankfurt.de.*E-mail: (M.T.) tornow@tum.de.ORCIDRobert Tampe: 0000-0002-0403-2160Author Contributions∇T.D. and Q.H.N. contributed equally to this work.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe German Research Foundation (SFB 807 to R.T.), LOEWE(DynaMem to R.T.), and a German-Israeli Project Cooperation(DIP, TO 266/8-1 to R.T. and M.T.) supported this work. Wethank C. Kutter (Fraunhofer EMFT, Munich) for usefuldiscussions. We are grateful to all members of the Institute ofBiochemistry (Goethe University Frankfurt) and Dr. K. Fendler(MPI of Biophysics, Frankfurt a.M.) for helpful comments onthe manuscript.

■ REFERENCES(1) Yildirim, M. A.; Goh, K. I.; Cusick, M. E.; Barabasi, A. L.; Vidal,M. Nat. Biotechnol. 2007, 25, 1119−1126.(2) Osaki, T.; Takeuchi, S. Anal. Chem. 2017, 89, 216−231.(3) Osaki, T.; Suzuki, H.; Le Pioufle, B.; Takeuchi, S. Anal. Chem.2009, 81, 9866−9870.(4) Carrillo, L.; Cucu, B.; Bandmann, V.; Homann, U.; Hertel, B.;Hillmer, S.; Thiel, G.; Bertl, A. Traffic 2015, 16, 760−772.(5) Oshima, A.; Hirano-Iwata, A.; Mozumi, H.; Ishinari, Y.; Kimura,Y.; Niwano, M. Anal. Chem. 2013, 85, 4363−4369.(6) Baaken, G.; Ankri, N.; Schuler, A. K.; Ruhe, J.; Behrends, J. C.ACS Nano 2011, 5, 8080−8088.(7) Lemay, S. G. ACS Nano 2009, 3, 775−779.(8) Fertig, N.; Blick, R. H.; Behrends, J. C. Biophys. J. 2002, 82,3056−3062.(9) Dunlop, J.; Bowlby, M.; Peri, R.; Vasilyev, D.; Arias, R. Nat. Rev.Drug Discovery 2008, 7, 358−368.(10) Stoelzle-Feix, S. State-of-the-art automated patch clamp: heatactivation, action potentials, and high throughput in ion channelscreening. In Patch-Clamp Methods and Protocols; Martina, M.,Taverna, S., Eds.; Springer: New York, 2014; Vol. 1183, pp 65−80.(11) Schroeder, K.; Neagle, B.; Trezise, D. J.; Worley, J. J. Biomol.Screening 2003, 8, 50−64.(12) Schulz, P.; Garcia-Celma, J. J.; Fendler, K. Methods 2008, 46,97−103.(13) Bazzone, A.; Barthmes, M.; Fendler, K. SSM-based electro-physiology for transporter research. In Methods in Enzymology; Ziegler,C., Ed.; Academic Press: 2017; Vol. 594, pp 31−83.

(14) Naumann, C. A.; Prucker, O.; Lehmann, T.; Ruhe, J.; Knoll, W.;Frank, C. W. Biomacromolecules 2002, 3, 27−35.(15) Castellana, E. T.; Cremer, P. S. Surf. Sci. Rep. 2006, 61, 429−444.(16) Stamou, D.; Duschl, C.; Delamarche, E.; Vogel, H. Angew.Chem., Int. Ed. 2003, 42, 5580−5583.(17) Christensen, S. M.; Stamou, D. Soft Matter 2007, 3, 828−836.(18) Christensen, A. L.; Lohr, C.; Christensen, S. M.; Stamou, D. LabChip 2013, 13, 3613−3625.(19) Christensen, S. M.; Stamou, D. G. Sensors 2010, 10, 11352−11368.(20) Christensen, S. M.; Triplet, M. G.; Rhodes, C.; Iwig, J. S.; Tu, H.L.; Stamou, D.; Groves, J. T. Nano Lett. 2016, 16, 2890−2895.(21) Mathiasen, S.; Christensen, S. M.; Fung, J. J.; Rasmussen, S. G.;Fay, J. F.; Jorgensen, S. K.; Veshaguri, S.; Farrens, D. L.; Kiskowski, M.;Kobilka, B.; Stamou, D. Nat. Methods 2014, 11, 931−934.(22) Weiskopf, D.; Schmitt, E. K.; Kluhr, M. H.; Dertinger, S. K.;Steinem, C. Langmuir 2007, 23, 9134−9139.(23) Kullman, L.; Winterhalter, M.; Bezrukov, S. M. Biophys. J. 2002,82, 803−812.(24) Horn, C.; Steinem, C. Biophys. J. 2005, 89, 1046−1054.(25) Kang, X. F.; Cheley, S.; Rice-Ficht, A. C.; Bayley, H. J. Am.Chem. Soc. 2007, 129, 4701−4705.(26) McConnell, H. M.; Watts, T. H.; Weis, R. M.; Brian, A. A.Biochim. Biophys. Acta, Rev. Biomembr. 1986, 864, 95−106.(27) Sackmann, E. Science 1996, 271, 43−48.(28) Reimhult, E.; Kumar, K. Trends Biotechnol. 2008, 26, 82−89.(29) Buchholz, K.; Tinazli, A.; Kleefen, A.; Dorfner, D.; Pedone, D.;Rant, U.; Tampe, R.; Abstreiter, G.; Tornow, M. Nanotechnology 2008,19, 445305.(30) Richter, R. P.; Berat, R.; Brisson, A. R. Langmuir 2006, 22,3497−3505.(31) Cremer, P. S.; Boxer, S. G. J. Phys. Chem. B 1999, 103, 2554−2559.(32) Hennesthal, C.; Drexler, J.; Steinem, C. ChemPhysChem 2002, 3,885−889.(33) Jonsson, P.; Jonsson, M. P.; Hook, F. Nano Lett. 2010, 10,1900−1906.(34) Urban, M.; Tampe, R. Microchim. Acta 2016, 183, 965−971.(35) Basit, H.; Gaul, V.; Maher, S.; Forster, R. J.; Keyes, T. E. Analyst2015, 140, 3012−3018.(36) Urban, M.; Kleefen, A.; Mukherjee, N.; Seelheim, P.;Windschiegl, B.; vor der Bruggen, M.; Kocer, A.; Tampe, R. NanoLett. 2014, 14, 1674−1680.(37) Watanabe, R.; Soga, N.; Fujita, D.; Tabata, K. V.; Yamauchi, L.;Hyeon Kim, S.; Asanuma, D.; Kamiya, M.; Urano, Y.; Suga, H.; Noji,H. Nat. Commun. 2014, 5, 4519.(38) Kleefen, A.; Pedone, D.; Grunwald, C.; Wei, R.; Firnkes, M.;Abstreiter, G.; Rant, U.; Tampe, R. Nano Lett. 2010, 10, 5080−5087.(39) Soga, N.; Watanabe, R.; Noji, H. Sci. Rep. 2015, 5, 11025.(40) Watanabe, R.; Soga, N.; Hara, M.; Noji, H. Lab Chip 2016, 16,3043−3048.(41) Kusters, I.; van Oijen, A. M.; Driessen, A. J. ACS Nano 2014, 8,3380−3392.(42) Mager, M. D.; Melosh, N. A. Adv. Mater. 2008, 20, 4423−4427.(43) Urban, M.; vor der Bruggen, M.; Tampe, R. MembraneTransport Processes Analyzed by a Highly Parallel Nanopore ChipSystem at Single Protein Resolution. J. Visualized Exp. 2016,DOI: 10.3791/53373.(44) Phair, R. D.; Gorski, S. A.; Misteli, T., Measurement of DynamicProtein Binding to Chromatin In Vivo, Using PhotobleachingMicroscopy. In Methods in Enzymology; Academic Press, 2003; Vol.375, pp 393−414.(45) Feder, T. J.; Brust-Mascher, I.; Slattery, J. P.; Baird, B.; Webb,W. W. Biophys. J. 1996, 70, 2767−2773.(46) Axelrod, D.; Koppel, D. E.; Schlessinger, J.; Elson, E.; Webb, W.W. Biophys. J. 1976, 16, 1055−1069.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3909

(47) Subburaj, Y.; Gallego, R. S.; García-Saez, A. J., MembraneDynamics: Fluorescence Spectroscopy. In Encyclopedia of AnalyticalChemistry; John Wiley & Sons, Ltd, 2006.(48) Seidel, H.; Csepregi, L.; Heuberger, A.; Baumgartel, H. J.Electrochem. Soc. 1990, 137, 3612−3626.(49) Liebes-Peer, Y.; Bandalo, V.; Sokmen, U.; Tornow, M.;Ashkenasy, N. Microchim. Acta 2016, 183, 987−994.(50) Anderson, T. H.; Min, Y.; Weirich, K. L.; Zeng, H.; Fygenson,D.; Israelachvili, J. N. Langmuir 2009, 25, 6997−7005.(51) Janshoff, A.; Steinem, C. Biochim. Biophys. Acta, Mol. Cell Res.2015, 1853, 2977−2983.(52) Hughes, L. D.; Rawle, R. J.; Boxer, S. G. PLoS One 2014, 9,e87649.(53) Venkatesan, B. M.; Polans, J.; Comer, J.; Sridhar, S.; Wendell,D.; Aksimentiev, A.; Bashir, R. Biomed. Microdevices 2011, 13, 671−682.(54) Mager, M. D.; Almquist, B.; Melosh, N. A. Langmuir 2008, 24,12734−12737.(55) Schmidt, T.; Schutz, G. J.; Baumgartner, W.; Gruber, H. J.;Schindler, H. Proc. Natl. Acad. Sci. U. S. A. 1996, 93, 2926−2929.(56) Song, L.; Hobaugh, M. R.; Shustak, C.; Cheley, S.; Bayley, H.;Gouaux, J. E. Science 1996, 274, 1859−1866.(57) Deamer, D. W.; Branton, D. Acc. Chem. Res. 2002, 35, 817−825.(58) Rosen, C. B.; Rodriguez-Larrea, D.; Bayley, H. Nat. Biotechnol.2014, 32, 179−181.(59) Di Marino, D.; Bonome, E. L.; Tramontano, A.; Chinappi, M. J.Phys. Chem. Lett. 2015, 6, 2963−2968.(60) Watanabe, H.; Gubbiotti, A.; Chinappi, M.; Takai, N.; Tanaka,K.; Tsumoto, K.; Kawano, R. Anal. Chem. 2017, 89, 11269−11277.(61) Mayerhofer, P. U.; Tampe, R. J. Mol. Biol. 2015, 427, 1102−1118.(62) Zollmann, T.; Moiset, G.; Tumulka, F.; Tampe, R.; Poolman, B.;Abele, R. Proc. Natl. Acad. Sci. U. S. A. 2015, 112, 2046−2051.(63) Noll, A.; Thomas, C.; Herbring, V.; Zollmann, T.; Barth, K.;Mehdipour, A. R.; Tomasiak, T. M.; Bruchert, S.; Joseph, B.; Abele, R.;Olieric, V.; Wang, M.; Diederichs, K.; Hummer, G.; Stroud, R. M.;Pos, K. M.; Tampe, R. Proc. Natl. Acad. Sci. U. S. A. 2017, 114, E438−E447.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.8b01252Nano Lett. 2018, 18, 3901−3910

3910