Light-Controlled Selective Collection-and-Release of Biomoleculesby an On-Chip Nanostructured DeviceElla Borberg,†,# Marina Zverzhinetsky,†,# Adva Krivitsky,‡ Alon Kosloff,† Omri Heifler,∥ Gal Degabli,†

Hagit Peretz Soroka,† Ronit Satchi Fainaro,‡ Larisa Burstein,§ Shlomi Reuveni,† Haim Diamant,†

Vadim Krivitsky,*,† and Fernando Patolsky*,†,∥

†School of Chemistry, Faculty of Exact Sciences, Tel Aviv University, Tel Aviv, 69978, Israel‡Department of Physiology and Pharmacology, Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv 69978, Israel§The Wolfson Applied Materials Research Centre, Tel-Aviv University, Tel-Aviv 69978, Israel∥Department of Materials Science and Engineering, The Iby and Aladar Fleischman Faculty of Engineering, Tel Aviv University, TelAviv 69978, Israel

*S Supporting Information

ABSTRACT: The analysis of biosamples, e.g., blood, is a ubiquitous taskof proteomics, genomics, and biosensing fields; yet, it still faces multiplechallenges, one of the greatest being the selective separation and detectionof target proteins from these complex biosamples. Here, we demonstratethe development of an on-chip light-triggered reusable nanostructuredselective and quantitative protein separation and preconcentrationplatform for the direct analysis of complex biosamples. The on-chipselective separation of required protein analytes from raw biosamples isperformed using antibody−photoacid-modified Si nanopillars verticalarrays (SiNPs) of ultralarge binding surface area and enormously highbinding affinity, followed by the light-controlled rapid release of the tightlybound target proteins in a controlled liquid media. Two importantexperimental observations are presented: (1) the first demonstration onthe control of biological reaction binding affinity by the nanostructuring ofthe capturing surface, leading to highly efficient protein collection capabilities, and (2) the light-triggered switching of the highlysticky binding surfaces into highly reflective nonbinding surfaces, leading to the rapid and quantitative release of the originallytightly bound protein species. Both of these two novel behaviors were theoretically and experimentally investigated.Importantly, this is the first demonstration of a three-dimensional (3D) SiNPs on-chip filter with ultralarge binding surface areaand reversible light-controlled quantitative release of adsorbed biomolecules for direct purification of blood samples, able toselectively collect and separate specific low abundant proteins, while easily removing unwanted blood components (proteins,cells) and achieving desalting results, without the requirement of time-consuming centrifugation steps, the use of desaltingmembranes, or affinity columns.

KEYWORDS: silicon nanopillars, photoacid, filtering, biomolecules, light control, preconcentration

■ INTRODUCTION

Analyzing biosamples is a ubiquitous task of proteomics,genomics, and biosensing fields; yet it still faces multiplechallenges, one of the greatest being the separation anddetection of target proteins from complex biosamples.1−4

Biosamples, such as blood and urine, are complex mixturesprimarily comprising numerous biomolecules and chemicalspecies, which represent a great practical challenge to anyanalytical technique being applied. Blood samples typicallycontain more than 10 000 different proteins at a concentrationrange varying over 10 orders of magnitude.5 This great diversity,along with the concentration range, poses an enormouschallenge during sample analysis in proteomics. Thus, therequirement to detect and differentiate proteins in this wide

range of concentrations in an efficientmanner is not yet fully metby conventional protein analysis techniques.Conventional techniques for protein separation include

chromatography,6−8 electrophoresis9−12 centrifugation13,14

and filtration.15,16 Chromatographic methods efficiently sepa-rate pure biomolecules but require long pretreatment processesand large sample volumes. Electrophoresis could be performedwith smaller samples and in a timely fashion; however, it is ableto separate biomolecules only according to their charge-to-sizeratio.

Received: March 31, 2019Revised: July 15, 2019Published: August 5, 2019

Letter

pubs.acs.org/NanoLettCite This: Nano Lett. 2019, 19, 5868−5878

© 2019 American Chemical Society 5868 DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

Dow

nloa

ded

via

TE

L A

VIV

UN

IV o

n O

ctob

er 3

1, 2

019

at 0

9:11

:05

(UT

C).

See

http

s://p

ubs.

acs.

org/

shar

ingg

uide

lines

for

opt

ions

on

how

to le

gitim

atel

y sh

are

publ

ishe

d ar

ticle

s.

The use of filtration methods, such as nanoporous materials,requires controllable pore sizes, lengths, and surface chemistry.To achieve highly selective separation and obtaining sufficientanalyte flux, the pores must be of uniform diameter distributionand with high abundance. Most of the commercially availablenanoporous materials are characterized by large size distribu-tions and relatively large thicknesses.17−20 Moreover, thesetechniques fall short in their integration on a single platformwithdownstream technologies, such as sensing devices, high-throughput systems, lab-on-chip microfluidic devices, andcomplementary metal−oxide−semiconductor fabricationroutes (CMOS).Another well-established method is immunoprecipitation, in

which antibodies are immobilized to the surface of polystyrenemicrobeads of high capacity and surface area that allow forquantitative separation; however, this method requires arepetitive multistep process of centrifugation and wash that isvery time-consuming. This also leads to precipitant loss, whichreduces the chance to separate low-abundant proteins inphysiological concentrations (pM-fM).Improved protocols that include the use of magnetic beads

accelerate the separation process, reducing the time required to asingle hour. Still, this does not solve the need for a multistepprocess, and it is not without faults, including a high percentageof nonspecific adsorption and the final use of a low-pH orsodium dodecyl sulfate (SDS)-containing elution buffer, thatusually causes nonreversible denaturation of the antibodycapturing units and the eluted proteins.Microfluidic-based separation devices have recently attracted

extensive research efforts. In contrast to nanoporous materials,these devices are mechanically and chemically more robust, arereadily integrated within microfabrication processes, can bepredesigned to yield improved separation performances, andwere already demonstrated to efficiently separate DNAmolecules, particles, and proteins in controlled fluidic environ-ments.21−24 These microfluidic devices have become increas-ingly popular in biological analyses by virtue of their minimalreagent use, cost-effectiveness, and automation.25,26 Mostmicrofluidic-based separation approaches commonly analyzeonly a tiny fraction of a tested biosample, a fact limiting theoverall analysis sensitivity, especially when dealing with lowlyabundant protein species. In this context, several strategies weredeveloped for the preconcentration of liquid biosamples,including field-amplified sample stacking,27,28 isotachophore-sis,29−31 micellar electrokinetic sweeping,32,33 isoelectric focus-ing,34−37 and dielectrophoresis.38,39 Nevertheless, these techni-ques are limited by buffer handling challenges, long preconcen-tration times, fabrication complexities making them difficult tointegrate with lab-on-chip systems, and their incapability toseparate and concentrate specific target molecules from acomplex biosample selectively.40,41

In a previous work from our group,42 we showed the potentialof vertical Si nanopillars vertical arrays (SiNPs) as an efficientdevice to quickly separate target proteins from complexbiosamples, and their integration on one microchip to a sensingdevice, to compose a lab-on-a-chip multitask system. The siliconnanowire-based field effect transistor (SiNW-FET) proved to bea reliable tool for label-free detection of various biomole-cules.43,44 However, the intrinsic limitations of SiNW-FETdevices require an aqueous medium of a controlled low ionicstrength (due to the Debye screening length limitations of lessthan ∼1 mM).45,46 Integrating SiNW-FET sensing devicesdownstream to this separation array allowed for the quick and

label-free detection of analytes directly from complex bio-samples.42

Here, we demonstrate the development of an on-chip light-controlled reusable nanostructured selective and quantitativeprotein separation, desalting and preconcentration platform forthe direct analysis of complex biosamples. The on-chip selectiveseparation of required protein analytes from raw biosamples isfirst performed using antibody∼photoacid-modified SiNPs ofultralarge binding surface area and enormously high bindingaffinity, followed by the light-controlled rapid release of thetightly bound target proteins in a controlled liquid media.Shortly, nanostructured SiNPs arrays create cavities in which

proteins experience limited diffusion, with a significantly lowermean-free path. Proteins delay inside the cavities while beingadsorbed to, and desorbed from, antibodies on the surface of thepillars into the confined interpillars space repeatedly, untilultimately being released to the above medium. This repeatedlyoccurring rebinding behavior of captured proteins inside theinterpillars aqueous space leads to extremely prolonged effectivedissociation rates. When accounting for the high protein-antibody affinity, the complete desorption process could takedays or longer.42 In order to control the dissociation rate, adrastic pH change near the SiNPs surfaces, caused by acovalently bonded photoacid-molecular monolayer, is applied.Photoacids are aromatic organic molecules that displayproperties of weak acids in their ground electronic state, butexhibit acidity that is many orders of magnitude greater in theirfirst excited electronic state. These materials have been widelystudied in several applications,47−51 including the light-triggered“pH drop” caused by photoactivation of photoacid moleculesapplied in SiNW-FET devices for the on-surface modulation ofprotein affinity to an antibody.52 The resulting on-surfacecontrolled pH drop immediately converts the highly “receptive”SiNPs binding surface into a highly “reflective” antibindingsurface, because of the dissociation of antigen−antibody pairsand the complete binding destabilization of the antibody units atthe light-triggered low “surface” pH. This light-triggered steptransforms the SiNPs protein collection surface into a fastreleasing unit, thus exploiting the best advantages from thesetwo worlds. Most notably, the pH change caused by thephotoactivation of photoacid molecules is triggerable, reversible,and does not involve buffer handling limitations. Also, thecontrolled surface density of photoacid molecules, along withthe applied light intensity, leads to the achievement of acontrolled “surface pH” and the subsequent rapid seconds-longrelease of the tightly captured protein species from the inter-SiNPs cavity into the bulk solution. Importantly, this is the firstdemonstration of a three-dimensional (3D) SiNPs on-chip filterwith ultralarge binding surface area and reversible light-controlled quantitative release of adsorbed biomolecules fordirect purification of blood samples, able to selectively collectand separate specific low-abundant proteins, while easilyremoving unwanted blood components (proteins, cells) andachieving desalting results, without the requirement of time-consuming centrifugation steps, the use of desalting membranes,or affinity columns.Furthermore, this 3D SiNPs light-controlled separation

device can be easily integrated into a single platform withdownstream sensing technologies, which can be successfullyapplied for the multiplex, real-time, and ultrasensitive detectionof biomarkers.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5869

Figure 1. SiNP array filter fabrication and modification analysis using scanning electron microscopy (SEM) and X-ray photoelectron spectroscopy(XPS). (a) Schematic representation of the process. (b) SEM image of the polystyrene beadmonolayer deposited on a silicon wafer. (c) SEM image ofthe polystyrene bead monolayer following oxygen plasma treatment. (d) SEM image of the porous SiNP array demonstrating high surface area formedby the wet etching process. (e) Close-up view of the porous SiNP array presented in panel (d). (f) Schematic representation, and XPS-analyzed atomicconcentrations during each step of the HPTS and antibodies immobilization on the SiNPs surface.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5870

■ RESULTS AND DISCUSSION

SiNPs Array Preparation and Chemical Modification.SiNPs were fabricated by the deposition of a densely packedmonolayer of polystyrene beads, 500 nm in diameter, through aspin coating step (see the Materials and Methods section in theSupporting Information for details) on a silicon wafer53 (seeFigures 1a and 1b). The beads monolayer is later subjected toplasma etching in order to reduce the diameter of the beads,down to 250 nm, and to create a beads array interdistance of ca.250 nm. Modulation of the plasma conditions allows thediameter, morphology and surface roughness of the resultingbeads to be controlled. Then, beads are used as etching masksfor a metal-assisted wet etching step, leading to the formation ofthe vertical SiNPs array, with a SiNP height of 3−20 μm, using ametal silver film as a catalyst and an HF/H2O2 mixture as theetchant and oxidant, respectively.54 The rough morphology ofthe etched bead mask elements dictates the final shape androughness of resulting SiNPs (see Figures 1c and 1d). Clearly,the use of bead masks of highly rough morphology is preferredwhen a larger active surface for the binding of biomolecules isrequired. Using this procedure, porous SiNPs arrays of tunablesurface areas up to ∼500 m2 g−1 and pore sizes up to 15 nm canbe simply achieved.42 TEM and AFM measurements demon-strated that SiNPs with roughness factors of up to 20 can beobtained bymodulation of the beadmorphology and applicationof the silicon etching conditions.42 Thus, a planar biomolecular-capturing device with a geometrical surface area of 1 cm2 can betransformed to an improved capturing element of 1200 cm2 afterthe etching of a SiNPs array consisting of SiNPs 20 μm in heightand 250 nm in diameter, with an inter-NPs distance of 250 and aroughness factor of 16. This represents a dramatic increase ofmore than 1200-fold in active surface area, in comparison to aplanar device of an identical geometrical area.Higher SiNPs arrays displaying increased roughness will

further increase the active capturing area, and thus the finalextent of captured biomolecules and preconcentration levelachievable.Chemical modification of the SiNPs array with 8-hydrox-

ypyrene-1,3,6-trisulfonic acid (HPTS) and antibody units wasperformed and XPS-analyzed following the steps outlined inFigure 1e.52 First, the silicon substrate is modified with the silanederivative (3-aminopropyl)-dimethyl-ethoxysilane (APDMES).Then, a derivative of the HPTS molecule, 8-acetoxy-pyrene-1,3,6-trisulfonyl chloride,51 is chemically linked to the surface ofamino-terminated SiNPs (see the Experimental Methodssection in the Supporting Information for detailed information).HPTS is a water-soluble pH indicator with a pKa of ∼7.3 at theground state. Remarkably, the first excited state of HPTS isextremely more acidic than its ground state, with a pKa of ∼0.4,and thus it is frequently applied as a light-triggered source ofprotons in various studies.50−52,55−57 The modification of theSiNPs array with HPTS was verified by the use of fluorescencemicroscopy measurements, as well as by XPS spectroscopy (seeFigures S4 and S1 in the Supporting Information). In addition,an APDMES surface coverage of ∼1.6 × 1014 molecules/cm2,and an HPTS surface coverage in the range of (0.1−3) × 1013

molecules/cm2, depending on the period of time allowed for itscoupling to the surface-confined amino groups, were deter-mined by quartz crystal microbalance (QCM).50 Only 3%−15%of the APDMES amino groups are linked to HPTS molecules;thus, the resulting SiNPs contain free amino groups available forthe further chemical coupling of antibody receptors. Next, SiNPs

elements are chemically modified with a layer of IgGmonoclonal antibodies specific against a single protein markeror multiple protein markers. Assuming SiNPs with a height of 5μm, with a roughness factor of 10, a maximum density of 7.8 ×1014 IgG molecules/cm2 (a 1 cm2 geometric flat area converts to∼200 cm2) can be theoretically expected (60% surface coverageof 8 nm footprint IgG molecules). Experimentally, a surfacecoverage of 2.6 × 1014 IgG molecules/cm2 was measured (usingthe modified Bradford method for the determination of surface-bound protein).42 Higher antibody densities (per cm2 ofgeometrical area) could be increased by using higher androugher SiNPs arrays. These 3D-SiNPs arrays will ultimatelyboost the capturing capabilities of the resulting separationdevice.

SiNPs Array Protein Collection and Release Efficiency.The protein capturing efficiency of our antibody−photoacid-modified SiNPs arrays was tested through the separation ofseveral serum-like biosamples containing different targetproteins, e.g., eGFP, hemoglobin, and CA-15.3. eGFP proteinis used as a model protein due to its strong emission at 509 nm,which can be sensitively followed by simply measuring theemission of the SiNPs array as a function of time. Proteincocktails in a concentration range of 5 pM to 1 μMwere appliedfor testing the binding/collection capability of SiNPs arrays (seeFigures 2a−c). Notably, all proteins can be directly andeffectively collected, and their binding to the antibody-modifiedcapturing array is observed to occur very fast, achieving a plateauafter only ca. 10−120 min of incubation for all tested proteins, ata density of ca. 430 pmol/cm2 eGFP, close to 82% of the totaleGFP incubated amount (ca. 525 pmol eGFP in the 1 mL 0.525μM eGFP sample). Importantly, incubation of protein samplesof lower concentrations, in the range of nanomolar to sub-picomolar, leads to complete depletion of proteins from thetreated biosamples, since the antibody surface concentration isultrahigh in comparison to the protein sample concentration.These results demonstrate the effectiveness of our protein

collection device for the rapid and quantitative capturing andpreconcentration of specific proteins from complex biosamples.In addition, the extent of nonspecific adsorption was also testedby the interaction of an eGFP-containing sample with anantihemoglobin IgG antibody, nonspecific to eGFP (Figure 2a,red curve), showing that nonspecific adsorption is negligible inthe absence of specific capturing antibodies. This quantitativecapturing behavior can be directly explained by the large relativeexcess of antibody capturing units on the large effective surfaceof the SiNPs array. For instance, 50 μL of 1 nM eGFP samplecontains only ca. 3 × 1010 eGFP molecules, while a SiNPscapturing an array of 1 cm2 displays a 2000-fold higher numberof 5.1 × 1013 IgG molecules. This protein/antibody molar ratiois responsible for the observed rapid and quantitative depletionof proteins from the tested biosamples, even at high sub-micromolar concentrations.Furthermore, we examined the temporal protein release

behavior of our SiNPs arrays following complete capture of thethree proteins eGFP, hemoglobin, and CA-15.3 (see Figure 3).Notably, unlike the observed minutes-long rapid and

quantitative protein collection, only up to a ca.12% of thecollected protein molecules desorb from the antibody-modifiedSiNPs arrays after a period of 2 h (with a dramatically longeffective dissociation rate of Koff ≈ 2 × 10−3 %/min, thus thetime for quantitative release is 76 000 min or ∼53 days),meaning that it would require more than an entire month for allprotein molecules to desorb from the capturing surface. We have

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5871

experimentally followed the dissociation of protein species froma SiNPs array and found that all protein species would bereleased from the platform after almost 50 days (Figure 3b).This observation is unexpected, considering an antibody

maximum dissociation rate of Koff = 1 × 10−3−1 × 10−4 s−1,thus expecting a full protein dissociation after tens of minutes.Notably, we believe that the first ca. 12% of fast-releasingproteins originate from the tip sections of the NPs elements. Atthese tip sections, protein molecules may release almost directlyto the bulk solution, without undergoing numerous reassocia-tion events on the NP walls. Also, as shown previously, this slow-release behavior cannot be accounted for by the potentialnonspecific adsorption of protein species to the surface of theSiNPs array, as negligible nonspecific adsorption was previouslyobserved. Thus, another capturing mechanism must beresponsible for this observed days-long slow-release behavior.In this context, we hypothesized that the nanostructured SiNPsarrays create physically confined nanocavities in which proteinmolecules experience limited diffusion, with a significantly lowermean-free path. Proteins delay inside the cavities while beingadsorbed to, and desorbed from, antibodies on the surface of thepillars into the confined interpillars space repeatedly andnumerously, until ultimately being released to the abovemedium. This behavior of captured proteins, repeatedlyrebinding inside the interpillars aqueous space, leads toextremely prolonged effective dissociation rates. When account-ing for the high protein-antibody affinity, the completedesorption process could take days or longer as observedexperimentally. Thus, our highly receptive SiNPs arrays caneffectively and very efficiently collect protein molecules frombulk solutions, further preventing their release back to the bulkby this novel nanoconfinement-related capturing mechanism.Once a protein molecule binds to the walls inside the cavities ofthe SiNPs arrays, it practically experiences numerous dissocia-tion/association events inside the SiNPs cavity, thus preventingits release into the bulk solution and dramatically prolonging itsobserved effective dissociation rate.To examine this effect theoretically, let us consider the

following problem. A particle diffuses along a channel of length Land square cross-section b × b. The diffusion coefficients alongand perpendicular to the channel axis are given by D∥ and D⊥,respectively. The channel is closed on one end and open on theother. Given the particle’s initial position, we are interested inN,the mean number of collisions with the side walls before theparticle reaches the open end (the bulk solution). Unlike thetotal time of the process, the number of collisions is notdependent on the time that the particle spends bound to thewall. Thus, we can consider the collisions to be instantaneous.We assume that L ≫ b, such that the time scales of transverseand axial motions (τ⊥ and τ∥, respectively) are well-separated(τ⊥ ≪ τ∥).The typical time of axial diffusion to the open end is τ∥ ≈ L2/

D∥. The typical time between collisions with the wall, however, isnot b2/D⊥, since these events are strongly correlated−if theparticle hits the wall, it remains in its vicinity and is likely to hit itagain. Therefore, the mean time between binding events isgoverned by the diffusive mean first passage time of a particle,starting a distance δx away from a wall (having escaped thebinding site) and ending on a wall again, in the presence of asecond wall a distance b≫ δx away. This mean first-passage timeis defined as τ⊥

−1 ≈ bδx/D⊥. The mean number of collisionsuntil the particle reaches the open end is given by∼τ∥/τ⊥. Thus,

N ADD

Lb x

2i

kjjjjjj

y

{zzzzzzikjjjj

y{zzzzδ

= ⊥

(1)

Figure 2. Protein capturing capabilities of the multi-antibodies-modified SiNPs light-controlled array. The nanoarray is modifiedwith anti-eGFP, antihemoglobin, and anti-CA 15.3 antibodies. (a)Specific adsorption of eGFP to a SiNP light-controlled array modifiedwith anti-eGFP antibodies (black line) versus nonspecific adsorption ofeGFP to a SiNPs light-controlled array not modified with anti-eGFPantibodies (red line). (b) Specific adsorption of hemoglobin to HPTS-modified SiNPs light-controlled array (black line) versus SiNP arraynot modified with HPTS (red line). (c) Adsorption kinetics of low-abundant proteins, as demonstrated on a sample spiked with 5 pM ofCA-15.3.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5872

where A is a numerical prefactor of order 1, dependent on theinitial conditions. We have numerically confirmed the scaling ofNwith L, b, and δx, as predicted by eq 1, using a simple random-walk simulation. If the particle’s initial position is takenrandomly anywhere inside the channel, we get A ≃ 0.50, withina 10% error.For b = 250 nm, L = 5 μm,D∥

GFP = 87 μm2 s−1, δx = 1 nm, andA = 0.50, a protein particle is expected to collide/rebind to thecavity’s walls ca. 5 × 104 times before finally escaping the cavityinto the bulk solution. This should theoretically delay theprotein from escaping the interior of the cavity by a factor of 104,compared to its bare dissociation rate. Even if only a smallfraction of these collisions leads to rebinding of the antigen tothe walls of the NPs array, then a dramatic influence on thedissociation kinetics should be observed in accord to ourmeasurements. This mechanism accounts for the observedsuperior collecting capabilities displayed by our antibody-modified SiNPs arrays. Notably, in the absence of high-affinityantibody units on the SiNPs surface, and based on basicdiffusion assumptions, the protein eGFP would escape from theinterpillar void volume in only 0.14 s.Furthermore, we performed a series of experiments with

SiNPs arrays of controlled interpillar distance. SiNPs arrays withan interpillars distance of 250, 450, 850, and 6000 nm, as well asa control planar surface of the same geometrical area, were

experimentally tested for their protein dissociation behavior (seeFigure 4). These arrays were fabricated using a differentapproach, shown in Figure S3 in the Supporting Information, inorder to have full control over both the dimensions of the pillarsand the interpillar distance simultaneously.Notably, increasing the interpillar distance leads to a clear

decrease in the observed protein dissociation rate, reaching aplateau at an interpillars distance of >6000 nm. This clearlyindicates that cavities with smaller dimensions lead to prolongeddissociation times, because of a larger number of potentialcollisions inside the nanocavity, as predicted by eq 1. Theseresults demonstrate, for the first time, the effect of surfacenanostructuring on the effective dissociation rate of biomole-cules from the corresponding collecting device surface, and it isnot limited to SiNPs arrays, but may as well be achieved byadditional types of surface nanostructuring.Also, for a 1 mm × 5 mm SiNPs array device, with an NP

height of 5 μm, we can expect an interpillars void volume of only5 nL. The incubation of a 10 μL biosample will lead to the fulldepletion of protein species and their preconcentration in theinterpillar void volume of 5 nL. This represents a preconcentra-tion factor of 1000. The use of smaller capturing surface andlarger biosample volumes will further increase the preconcentra-tion factors achievable.

Figure 3.Multi-antibodies-modified SiNP light-controlled array release performance in a controlled media. Desorption kinetics of (a) eGFP and (b)hemoglobin under illumination at a wavelength of 400 nm light (black line) and under “dark” conditions (red line). (c) Desorption kinetics of the low-abundant protein CA-15.3 described in Figure 2c under illumination at a wavelength of 400 nm. (d) Long-term release profile of eGFP proteinfrom antibody-modified SiNPs arrays in the absence of HPTS.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5873

Generally, this protein “delayed” release mechanism fromSiNPs arrays greatly contributes to the collection capabilities ofthese devices, although it also strongly handicaps its subsequentdeployment in protein analysis and detection applications wherethe release of the captured proteins is required. Thus, amechanism for the effective switching from a highly stickynonreleasing state into a highly reflective fast-releasing state ishighly desirable.Light-Triggered Rapid and Quantitative Release of

Proteins by SiNPs Arrays. Antibody−antigen bindinginteractions are optimal in aqueous buffers at physiological pHand ionic strength, such as in phosphate-buffered saline (PBS)(pH 7). The light-triggered pH jump caused by photoactivationof the surface-confined photoacid molecules may be applied forthe on-surface activation of pH-dependent chemical andbiological processes, such as the dissociation of antigen−antibody pairs.52 Irradiation at a wavelength of 400 nm bringsHPTS molecules to their excited state, where they display adramatic change in their pKa value, from 7.4 to ∼0.4, becomingsuperacid molecules, which rapidly expel their phenolic protons.The HPTS-modified SiNPs arrays are expected to experience adramatic “switching” effect after light irradiation, as a result of

the fast deprotonation of protonatedHPTS species, thus causingdissociation of the surface-attached antibody−antigen pairs, bydecreasing the pH around and in the close vicinity of the SiNPssurfaces.51,52 In conventional affinity methods, proteins elutioncan be accomplished by lowering the pH, or altering the ionicstate, to disrupt the antigen−antibody binding interaction.58 Inpractice, elution buffers occasionally cause some loss of antibodyor antigen function, limiting the number of times affinity supportcan be reused. The most widely used elution buffer for affinitypurification of proteins is 0.1 M glycine HCl, pH 2.5−3.0. Thisbuffer is known to effectively dissociate most protein−proteinand antibody−antigen binding interactions, preserving proteinstructure and binding affinity in most cases. Thus, based on ourhypothesis, on-surface pH jump caused by the photoexcitationof surface-confined HPTS molecules, are expected to decreasethe antibody−antigen binding interaction strength and bring totheir fast dissociation, along with impediment of theirreassociation under light irradiation.Figure 3 demonstrates the effect of light irradiation on the

protein dissociation from antibody∼photoacid-modified SiNPsarrays. In contrast to the already described slow dissociationbehavior displayed by the SiNPs array under “dark” conditions,

Figure 4. Effect of the distance between SiNPs on the antigen release kinetics. (a−d) SEM images of SiNPs arrays fabricated using nanoimprintlithography with controlled distances between the pillars: (a) 250 nm, (b) 450 nm, (c) 850 nm, and (d) 6000 nm. (e) Table summarizing thedesorption of eGFP from the various anti-eGFP modified SiNPs array presented in panels (a)−(d). The final row corresponds to a flat wafer surface,representing the infinite distance between the SiNPs elements.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5874

irradiation of light (50 mW/cm2) at a wavelength of 400 nm(maximum HPTS absorbance) leads to the immediate fastdissociation of protein molecules from the SiNPs capturingsurface. For most proteins, a rapid and complete quantitativerelease into the bulk solution can be observed after only 2−5min, in contrast to several days required for complete releaseunder “dark” conditions (see Movies 1 and 2 in the SupportingInformation, which demonstrate the effect of light illuminationon the protein release kinetics). Light illumination at wave-lengths other than those required for excitation of HPTSmolecules does not cause any change in the effective dissociationrates of proteins from the SiNPs array. This further corroboratesthe key task of surface-confined photoacid HPTS molecules inthe light-triggered surface-controlled pH modulation process.Notably, the SiNPs collecting device can be switched from ahighly sticky state, in darkness, to a highly reflective state afterillumination at 400 nm for multiple cycles, without anyobservable degradation of the collecting−releasing deviceperformance. These results directly imply that continuousirradiation of the multifunctional antibody∼HPTS-modifiedSiNPs devices at 400 nm causes a sharp “pH jump” in the closesurroundings of the NPs surfaces, because of deprotonation ofphotoexcited photoacid molecules from a baseline pH of 7.0 toan “effective pH” value of probably <3.5, which is suitable for thedissociation of antibody-protein pairs.Next, optimization of the HPTS photoacid molecules surface

density was performed (see Figure 5). SiNPs arrays comprising

different concentrations of HPTS, from ∼5 × 1012 molecules/cm2 to 4 × 1013 molecules/cm2, were prepared and tested inprotein collection−releasing experiments. Low surface densitiesof the photoacid HPTS (5 × 1012 molecules/cm2) did not leadto improved protein dissociation rates, while a density of 2 ×1013 molecules/cm2 led to the optimal fastest dissociation rate(see Figure 5, black curve) with complete dissociation of theantibody-protein pairs after ca. 5 min of light illumination.Higher HPTS surface densities of ca. 4 × 1013 molecules/cm2

led to the fast dissociation of protein molecules as well as

surface-anchored HPTS molecules, plausibly due to extremelylow pH values and the hydrolysis of surface-bound photoacidmolecules. Thus, the optimal surface density for achieving thefastest protein dissociation rates, along with the best devicereusability, is 2 × 1013 molecules/cm2.Furthermore, a calculation taking into consideration the

surface coverage of HPTS photoacid units and the intensity ofilluminated light, further validates that pH jumps of thenecessary extent, down to an effective pH of ∼2.5, can beachieved in the volume in close vicinity to the nanowire surface,assuming that photoreleased protons are limited to diffusewithin a thin water layer with a thickness of 10 nm (for a detailedcalculation, please see the Materials and Methods section in theSupporting Information). Furthermore, this approximatecalculation shows the capability to directly control the surface“effective pH” via a simple modulation of the surface density ofHPTS photoacid species (or alternatively by modulating theintensity of the irradiated light source) (see Figure 5). Inaddition, previous experiments showed that irradiation of HPTSmolecules brings a sharp decrease of pH near the surface ofnanowire elements and that the corresponding pH decrease isdependent on the intensity of the irradiating light source.52

Thus, we can confidently assume that a surface pH jump downto pH ∼2.5 is experienced in close vicinity to the surface of theSiNPs.Notably, attempts to cause complete dissociation of protein−

antibody pairs through the incubation of the SiNPs arrays with aglycine−HCl buffer (pH 2.5), and additional acidic solutions,did not lead to the complete rapid removal of proteins from thecollecting device surface (see Figure S2 in the SupportingInformation), while causing fast degradation of both the releasedprotein as well as the binding capacity of the antibody-modifiedSiNPs device. Taken together, these results pinpoint thecapability of modulating the binding affinity of surface-confinedpH-dependent biomolecular receptors, such as protein, anti-bodies, and DNA,59,60 by a light-triggered switching of pHconditions caused by surface-confined photoacid species.

Real-World Applications of SiNPs Biomolecular Cap-turing Devices. The selective capture of multiple proteinssimultaneously by a single separating device is a task of greatimportance in the field of proteomics and medical diagnostics.Figure 6 demonstrates the selective multiple protein

separation and preconcentration capabilities of the proposedantibody−photoacid-modified SiNPs arrays. Figure 6c showsthe results of SDS-PAGE performed on samples released fromour SiNPs array that is premodified with eGFP, GAPDH, andHSC-70 antibodies HeLa whole cell lysate was spiked witheGFP (25 μg/mL), while GAPDH and HSC-70 proteins werenative to the lysate. The sample was incubated with ourdevice, then washed, and the proteins eGFP, GAPDH, andHSC-70 were allowed to release under illumination. Lysateremains that were collected immediately after adsorption, andsamples of buffer that were collected following 1 min ofillumination, were analyzed on acrylamide gel. The gel imageindicates that while eGFP, GAPDH, and HSC-70 were releasedfrom the SiNPs array following 1 min of ilumination, β-actin,which is a protein native to the lysate that did not have acorresponding antibody modification, was present only in theoriginal lysate and was not released from the device. Notably,cell lysate samples contain large amounts of foreign proteinspecies and represent a challenge for most separation techniquesin the analytical field. These results demonstrate the excellent

Figure 5. Optimization of the SiNP array light-controlled releaseperformance by tuning the concentrations of the photoacid (HPTS)modified on the SiNPs surface. The desorption kinetics of eGFP fromthe SiNPs array modified with 36 μMHPTS (shown in black), 30 μMHPTS (shown in green), 24 μM HPTS (shown in blue), and 18 μMHPTS (shown in red) under illumination at a wavelength of 400 nmlight is presented.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5875

separation skills of our SiNPs arrays in the separation of specific

protein biomarkers from highly complex biosamples.Also, we attempted the separation of protein biomarkers of

interest directly from whole blood samples (see Figures 6a and

6b). For this task, we designed and fabricated a 3D-printed

separation device that consisted of the light source, inlet, and

outlet fluid valves and the antibody−photoacid-SiNPs capturingarray. This prototype device allows the easy and rapid separation

of specific protein markers from whole blood samples to be

performed within <10 min. Also, the separation platform is

shown to be reusable for tens of cycles without degradation in its

separation performance (see Figure S7 in the Supporting

Information).

■ CONCLUSIONS

We demonstrated, for the first time, the development of an on-chip light-controlled reusable nanostructured selective andquantitative protein separation and preconcentration platformfor the direct analysis of complex biosamples. The on-chipselective separation of required protein analytes from rawbiosamples is performed using antibody−photoacid-modifiedSiNPs of ultralarge binding surface area and enormously highbinding affinity, followed by the light-controlled rapid release ofthe tightly bound target proteins in a controlled liquid media.We found that SiNPs arrays create cavities in which proteinsexperience limited diffusion, and thus, delay inside the cavitieswhile being adsorbed to, and desorbed from, antibodies on thesurface of the pillars into the confined interpillars spacerepeatedly. This repeatedly occurring rebinding behavior ofcaptured proteins inside the aqueous space of the interpillars

Figure 6. Microfluidics-integrated SiNPs light-controlled device for fast biosample analysis. (a) The integrated system, including a 3D printedmicrofluidic channel and a light source of 400 nm, 50 mW, at action. (Inset shows eGFP desorption from the SiNPs light-controlled array duringillumination.) (b) Capturing and release of 1.58 nmol eGFP from a blood sample by 2 cm × 2 cm SiNPs light-controlled wafer. (c) Estimation of thepurification performance of the SiNPs light-controlled device by SDS-PAGE analysis. The cell lysate was spiked with eGFP (25 μg/mL) and incubatedwith the SiNPs light-controlled array device, which was previously modified with anti-eGFP, anti- GAPDH and anti-HSC-70. GAPDH, HSC-70, andβ-actin are endogenic proteins in the tested lysate. Desorption fraction was diluted by a factor of ∼20, in comparison to original lysate solution.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5876

leads to extremely prolonged effective dissociation rates. Whenaccounting for the high protein−antibody affinity, the completedesorption process could take days or longer. In order todramatically accelerate the protein desorption rate, a light-triggered drastic pH change near the surfaces of the Si pillarscaused by a covalently bonded photoacid-molecular monolayeris applied. The resulting on-surface controlled pH dropimmediately converts the highly “receptive” SiNPs bindingsurface into a highly “reflective” antibinding surface, due to thedissociation of antigen−antibody pairs and the completebinding destabilization of the antibody units at the light-triggered low “surface” pH. This light triggered step transformsthe SiNPs protein collection surface into a fast releasing element,thus exploiting the best advantages from these two worlds.Importantly, this is the first demonstration of a 3D SiNPs on-chip filter with an ultralarge binding surface area and reversiblelight-controlled quantitative release of adsorbed biomoleculesfor direct purification of blood samples, that is able to selectivelycollect and separate specific low-abundant proteins, while easilyremoving unwanted blood components (proteins, cells) andachieving desalting results, without the requirement of time-consuming centrifugation steps, the use of desalting membranes,or affinity columns.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.nano-lett.9b01323.

Figures S1−S7; sections discussing materials andchemicals, XPS analysis, HPTS molecules surface densitycalculation, comparison with other methods for fastprotein−antibody pairs dissociation, fabrication of SiNParray using metal-assisted etching, fabrication of SiNParray using nanoimprint lithography, synthesis of 8-acetoxy-pyrene-1,3,6-trisulfonyl chloride, antibodies andHPTS immobilization on SiNP array, fabrication of SiNWFETs sensing system, capturing and sensing, andfluorescence microscopy analysis (PDF)Movie 1: Comparison of non-modified SiNPs array with aanti-eGFP-HPTS-modified SiNPs array after collection ofeGFP (MP4)Movie 2: Release after collection of eGFP by a anti-eGFP-HPTS-modified SiNPs array (MP4)

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: vadimkri@mail.tau.ac.il (V. Krivitsky).*E-mail: fernando@post.tau.ac.il (F. Patolsky).ORCIDFernando Patolsky: 0000-0002-1382-5357Author Contributions#E.B. and M.Z. contributed equally to this work.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work was in part financially supported by the Legacy Fund,Israel Science Foundation (ISF). Authors wish to thank OdedFarago and Anton Zilman for helpful discussions.

■ REFERENCES(1) Issaq, H. J.; Conrads, T. P.; Janini, G. M.; Veenstra, T. D.Electrophoresis 2002, 23 (17), 3048−3061.(2) Elnathan, R.; Kwiat, M.; Pevzner, A.; Engel, Y.; Burstein, L.;Khatchtourints, A.; Lichtenstein, A.; Kantaev, R.; Patolsky, F. NanoLett. 2012, 12 (10), 5245−5254.(3) Krivitsky, V.; Zverzhinetsky, M.; Patolsky, F. Nano Lett. 2016, 16(10), 6272−6281.(4) Long, Y.; Gooding, J. J. ACS Sensors 2018, 3 (12), 2471−2472.(5) Gorg, A.; Weiss, W.; Dunn, M. J. Proteomics 2004, 4 (12), 3665−3685.(6) Katsuhara, T.; Matsumoto, M.; Ishihara, K., Biomicromolecule-separating, monolithic silica column and its production process, andseparation method of biomicromolecules. U.S. Patent Application No.US 2010/0210458 A1, 2010.(7) Miyazaki, S.; Morisato, K.; Ishizuka, N.; Minakuchi, H.; Shintani,Y.; Furuno, M.; Nakanishi, K. J. Chromatogr. A 2004, 1043 (1), 19−25.(8) Muller, C.; Fietz, C.; Koehler, P.; Sibley, G.; Nforbugwe, A. C. A.;Streichert, T.; Wiesen, M. H. Antimicrob. Agents Chemother. 2018, 62,01356-18.(9) Lao, A. I.; Trau, D.; Hsing, I.-M. Anal. Chem. 2002, 74 (20),5364−5369.(10) Zhao, L.; Zhou, J.; Xie, H.; Huang, D.; Zhou, P. Electrophoresis2012, 33 (12), 1703−1708.(11) Vahey, M. D.; Voldman, J. Anal. Chem. 2009, 81 (7), 2446−55.(12) Volkmuth, W. D.; Austin, R. H. Nature 1992, 358, 600.(13) Redgrave, T.; Roberts, D.; West, C. Anal. Biochem. 1975, 65 (1−2), 42−49.(14) Ang, B. J.; Nickerson, K. W. Appl. Environ. Microbiol. 1978, 36(4), 625−626.(15) Wang, Y.-C.; Singh, A. K.; Hatch, A. V. Ultra-Rapid SamplePreconcentration under Slant Field Using High-Aspect-Ratio NanoporousMembranes; Sandia National Laboratories (SNL-CA): Livermore, CA,USA, 2008.(16) Chen, C.-C.; Chen, Y.-A.; Liu, Y.-J.; Yao, D.-J. Lab Chip 2014, 14(8), 1459−1468.(17) van Rijn, C. J. M. Nano and Micro Engineered MembraneTechnology; Membrane Science and Technology Series, Vol. 10;Elsevier: 2004.(18) Han, J.; Fu, J.; Schoch, R. B. Lab Chip 2008, 8 (1), 23−33.(19) Fissell, W. H.; Dubnisheva, A.; Eldridge, A. N.; Fleischman, A. J.;Zydney, A. L.; Roy, S. J. Membr. Sci. 2009, 326 (1), 58−63.(20)Warkiani, M. E.; Bhagat, A. A. S.; Khoo, B. L.; Han, J.; Lim, C. T.;Gong, H. Q.; Fane, A. G. ACS Nano 2013, 7 (3), 1882−1904.(21) Hagan, K. A.; Meier, W. L.; Ferrance, J. P.; Landers, J. P. Anal.Chem. 2009, 81 (13), 5249−56.(22) Stern, E.; Vacic, A.; Rajan, N. K.; Criscione, J. M.; Park, J.; Ilic, B.R.;Mooney, D. J.; Reed,M. A.; Fahmy, T.M.Nat. Nanotechnol. 2010, 5,138.(23) Devendra, R.; Drazer, G. Anal. Chem. 2012, 84 (24), 10621−7.(24) Chen, C. L.; Chen, K. C.; Pan, Y. C.; Lee, T. P.; Hsiung, L. C.;Lin, C. M.; Chen, C. Y.; Lin, C. H.; Chiang, B. L.; Wo, A. M. Lab Chip2011, 11 (3), 474−83.(25) Ohno, K. i.; Tachikawa, K.; Manz, A. Electrophoresis 2008, 29(22), 4443−4453.(26) Arora, A.; Simone, G.; Salieb-Beugelaar, G. B.; Kim, J. T.; Manz,A. Anal. Chem. 2010, 82 (12), 4830−4847.(27) Keithley, R. B.; Metzinger, M. P.; Rosado, A. M.; Dovichi, N. J.Talanta 2013, 111, 206−214.(28)Michalska, K.; Pajchel, G.; Tyski, S. J. Sep. Sci. 2011, 34 (4), 475−482.(29) Jung, B.; Zhu, Y.; Santiago, J. G. Anal. Chem. 2007, 79 (1), 345−349.(30) Liu, D.; Ou, Z.; Xu, M.; Wang, L. J. Chromatogr. A 2008, 1214(1−2), 165−170.(31) Wang, J.; Zhang, Y.; Mohamadi, M. R.; Kaji, N.; Tokeshi, M.;Baba, Y. Electrophoresis 2009, 30 (18), 3250−3256.(32) Quirino, J. P.; Terabe, S. Science 1998, 282 (5388), 465−468.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5877

(33) Sera, Y.; Matsubara, N.; Otsuka, K.; Terabe, S. Electrophoresis2001, 22 (16), 3509−3513.(34) Cabrera, C. R.; Yager, P. Electrophoresis 2001, 22 (2), 355−362.(35) O'Neill, R. A.; Bhamidipati, A.; Bi, X.; Deb-Basu, D.; Cahill, L.;Ferrante, J.; Gentalen, E.; Glazer, M.; Gossett, J.; Hacker, K.; Kirby, C.;Knittle, J.; Loder, R.; Mastroieni, C.; MacLaren, M.; Mills, T.; Nguyen,U.; Parker, N.; Rice, A.; Roach, D.; Suich, D.; Voehringer, D.; Voss, K.;Yang, J.; Yang, T.; Vander Horn, P. B. Proc. Natl. Acad. Sci. U. S. A. 2006,103 (44), 16153−16158.(36) Tan, W.; Fan, Z. H.; Qiu, C. X.; Ricco, A. J.; Gibbons, I.Electrophoresis 2002, 23 (20), 3638−3645.(37) Li, Y.; DeVoe, D. L.; Lee, C. S. Electrophoresis 2003, 24 (1−2),193−199.(38) Moncada-Hernandez, H.; Lapizco-Encinas, B. H. Anal. Bioanal.Chem. 2010, 396 (5), 1805−1816.(39) Chen, D.; Du, H. Microfluid. Nanofluid. 2010, 9 (2), 281−291.(40) Kim, S. M.; Burns, M. A.; Hasselbrink, E. F. Anal. Chem. 2006, 78(14), 4779−4785.(41) Foote, R. S.; Khandurina, J.; Jacobson, S. C.; Ramsey, J. M. Anal.Chem. 2005, 77 (1), 57−63.(42) Krivitsky, V.; Hsiung, L.-C.; Lichtenstein, A.; Brudnik, B.;Kantaev, R.; Elnathan, R.; Pevzner, A.; Khatchtourints, A.; Patolsky, F.Nano Lett. 2012, 12 (9), 4748−4756.(43) Cui, Y.; Wei, Q.; Park, H.; Lieber, C. M. Science 2001, 293(5533), 1289−92.(44) Zheng, G.; Patolsky, F.; Cui, Y.; Wang, W. U.; Lieber, C. M.Nat.Biotechnol. 2005, 23, 1294.(45) Stern, E.; Vacic, A.; Reed, M. A. IEEE Trans. Electron Devices2008, 55 (11), 3119−3130.(46) Stern, E.; Wagner, R.; Sigworth, F. J.; Breaker, R.; Fahmy, T. M.;Reed, M. A. Nano Lett. 2007, 7 (11), 3405−9.(47) Luo, Y.; Wang, C.; Peng, P.; Hossain, M.; Jiang, T.; Fu, W.; Liao,Y.; Su, M. J. Mater. Chem. B 2013, 1 (7), 997−1001.(48) Yue, X.; Yanez, C. O.; Yao, S.; Belfield, K. D. J. Am. Chem. Soc.2013, 135 (6), 2112−2115.(49) Guardado-Alvarez, T. M.; Russell, M. M.; Zink, J. I. Chem.Commun. 2014, 50 (61), 8388−8390.(50) Soroka, H. P.; Simkovitch, R.; Kosloff, A.; Shomer, S.; Pevzner,A.; Tzang, O.; Tirosh, R.; Patolsky, F.; Huppert, D. J. Phys. Chem. C2013, 117 (48), 25786−25797.(51) Peretz-Soroka, H.; Pevzner, A.; Davidi, G.; Naddaka, V.; Tirosh,R.; Flaxer, E.; Patolsky, F. Nano Lett. 2013, 13 (7), 3157−3168.(52) Peretz-Soroka, H.; Pevzner, A.; Davidi, G.; Naddaka, V.; Kwiat,M.; Huppert, D.; Patolsky, F. Nano Lett. 2015, 15 (7), 4758−4768.(53) Wang, D.; Mohwald, H. Adv. Mater. 2004, 16 (3), 244−247.(54) Huang, Z.; Geyer, N.; Werner, P.; de Boor, J.; Gosele, U. Adv.Mater. 2011, 23 (2), 285−308.(55) Timmer, R. L. A.; Cox, M. J.; Bakker, H. J. J. Phys. Chem. A 2010,114 (5), 2091−2101.(56) Kelson, M. M. A.; Bhosale, R. S.; Ohkubo, K.; Jones, L. A.;Bhosale, S. V.; Gupta, A.; Fukuzumi, S.; Bhosale, S. V. Dyes Pigm. 2015,120, 340−346.(57) Simkovitch, R.; Pines, D.; Agmon, N.; Pines, E.; Huppert, D. J.Phys. Chem. B 2016, 120 (49), 12615−12632.(58) Antonsen, K. P.; Colton, C. K.; Yarmush, M. L. Biotechnol. Prog.1991, 7 (2), 159−172.(59) Tempestini, A.; Cassina, V.; Brogioli, D.; Ziano, R.; Erba, S.;Giovannoni, R.; Cerrito, M. G.; Salerno, D.; Mantegazza, F. NucleicAcids Res. 2013, 41 (3), 2009−2019.(60) Dubey, R. K.; Tripathi, D. Indian J. Biochem. Biophys. 2005, 42,301−307.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b01323Nano Lett. 2019, 19, 5868−5878

5878