rXXXX American Chemical Society A dx.doi.org/10.1021/la2015868 | Langmuir XXXX, XXX, 000–000

ARTICLE

pubs.acs.org/Langmuir

Effect of Surface Nanotopography on Immunoaffinity Cell Capturein Microfluidic DevicesBu Wang,†,‡ Alex L. Weldon,‡ Pisist Kumnorkaew,‡ Bu Xu,‡ James F. Gilchrist,‡ and Xuanhong Cheng*,†,§

†Department of Materials Science and Engineering, ‡Department of Chemical Engineering, and §Bioengineering Program,Lehigh University, 5 E. Packer Avenue, Bethlehem, Pennsylvania 18015, United States

bS Supporting Information

’ INTRODUCTION

Microfluidic cell isolation devices have recently found applica-tions in tissue engineering,1,2 clinical diagnosis,3,4 and funda-mental cell biology.5,6 Immunoaffinity isolation based on specificcell surface markers attracts special attention due to its capabilityto distinguish cell types of similar physical properties in complexsamples, for example, human blood.4,7,8 Its ease of use makes it agood candidate for point of care settings.

Immunoaffinity cell capture generally takes one of two forms:target cells attach to either suspendedmagnetic beads9 or interiorsurfaces of a microfluidic chip.10 Cell�antibody interactions aremore accurately controlled in the latter strategy, and as such, ityields higher capture efficiency and purity of target cells.4 Toner’sgroup has designed immunoaffinity microfluidics for reliable iso-lation of circulating cells at concentrations as low as 5 cells/mLfrom whole blood with average purity higher than 50%,4,8 whichexceeds the performance of commercially available magnetic bead-based assays. These microchips have been shown to diagnosecancer,4 infection,11 and inflammatory7 diseases with minimalsample handling.

To enhance cell interactions with antibody-functionalizeddevice walls, several groups have employed capture beds withnanotopography. Wang et al. fabricated nanopillars to enhancecontact frequency and duration of circulating tumor cells withfunctionalized substrates and increased capture yield by greaterthan 40%.12 TheKing group deposited nanobeads and nanofibers in

capillary channels to isolate stem cells and tumor cells.13,14

Capture yield increased by up to 100% in deposited versussmooth capillaries. While these studies suggest a positive effect ofnanotopography on cell capture, nanopatterns studied containeither a single geometry or random structures that do notelucidate detailed relations between the two. In a separate study,Tuttle et al. immobilized antibodies on surface-bound nanobeads40 to 860 nm in diameter and tested their functionality bycapturing antigen-coated fluorescent microspheres.15 Specificcapture increased continuously with the nanobead size. Theresults were suggested to stem from greater amounts of immo-bilized antibodies and an enhancement of collective antibodyfunctionality on rough surfaces. Additionally, the Santore groupinvestigated affinity between microbeads and patched chargecapture surfaces with varying overall charge density and patchsize.16�18 Although these studies explored a variety of nanos-tructured geometries, the interaction of microbeads with roughor patterned surfaces may not translate fully to that of the cells.Built on these previous findings, in this study we carefullycontrolled surface nanotopography over a greater range thanthat previously explored and studied its influence on immunoaf-finity capture of lymphocytes.

Received: April 28, 2011Revised: June 30, 2011

ABSTRACT: Immunoaffinity microfluidic devices have re-cently become a popular choice to isolate specific cells for manyapplications. To increase cell capture efficiency, several groupshave employed capture beds with nanotopography. However,no systematic study has been performed to quantitativelycorrelate surface nanopatterns with immunoaffinity cell immo-bilization. In this work, we controlled substrate topography bydepositing close-packed arrays of silica nanobeads with uniformdiameters ranging from 100 to 1150 nm onto flat glass. Thesesurfaces were functionalized with a specific antibody andassembled as the base in microfluidic channels, which were then used to capture CD4+ T cells under continuous flow. It isobserved that capture efficiency generally increases with nanoparticle size under low flow rate. At higher flow rates, cell captureefficiency becomes increasingly complex; it initially increases with the bead size then gradually decreases. Surprisingly, capture yieldplummets atop depositions of some particle diameters. These dips likely stem from dynamic interactions between nanostructures onthe substrate and cell membrane as indicated by roughness-insensitive cell capture after glutaraldehyde fixing. This systematic studyof surface nanotopography and cell capture efficiency will help optimize the physical properties of microfluidic capture beds for cellisolation from biological fluids.

B dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

To create surfaces with reproducible and uniform nanotopo-graphy we modified smooth glass substrates by the addition of amonolayer of close-packed, uniform-sized, silica nanobeads 100to 1150 nm in diameter. These surfaces were incorporated as thebase of microfluidic channels then functionalized with CD4antibodies in order to capture CD4+ cells from human lympho-cyte culture under continuous flow. In addition to depositingnanobeads with varying diameters, numerous flow rates andaltered cell mechanical properties were evaluated in order toexplore the mechanisms behind cell interactions with roughfunctionalized substrates.

’EXPERIMENTAL SECTION

Materials. Plain glass microslides (76 mm � 25 mm � 1 mm),ammonium hydroxide (NH4OH, 28�30% and 29%), parafilm,Rhodamine B, and fetal bovine serum (FBS) were all purchased fromFisher Scientific (Pittsburgh, PA). Polydimethylsiloxane (PDMS)Sylgard 184 silicone elastomer kit, was purchased from Dow CorningCorporation (Midland, MI). 3-mercaptopropyl trimethoxysilane wasobtained from Gelest Inc. (Morrisville, Pa). NeutrAvidin and N-[g-maleimidobutyryloxy]sulfosuccinimide ester (GMBS) were purchasedfrom Pierce Protein Inc. (Rockford, IL). Biotinylated antihuman CD4monoclonal antibody was obtained from Beckman Coulter (Brea, CA).Phosphate buffered saline (PBS), lyophilized bovine serum albumin(BSA), and RPMI-1640 cell media were obtained from Sigma AldrichChemical Co. (Milwaukee, WI). Penicillin�streptomycin (pen-strep)and Hoechst 33258 cell stain were received from Invitrogen (Carlsbad,CA). Ethyl alcohol (190 proof or 95%) was purchased from Pharmco-AAPER (Brookfield, CT). Ethanol (200 proof) was purchased fromKoptec (King of Prussia, PA) and Pharmco-AAPER (Brookfield, CT).Tetraethyl orthosilicate (TEOS) (98%) was purchased from SigmaAldrich (Milwaukee, WI) and Acros Organics (Pittsburgh, PA). SU8photoresist was purchased from MicroChem (Newton, MA). Three-inch silicon wafers were purchased from Silicon Inc. (Boise, Idaho). Fivehundred nanometer SiO2 particles used to seed the core�shell semi-batch synthesis were purchased from Fuso Chemical Co. (Japan).Preparation of Silica Nanobeads. The particles used in this

study have diameters between 100 and 1150 nm to cover the size rangereported in the literature for cell capture enhancement. Silica particleswithin this size range also form stable suspensions compatible with theconvective deposition process. We refer to the spheres used in this studyas nanoparticles in this paper, and they are sometimes also referred to ascolloidal particles or microspheres in the literature. Nanospheres100 nm, 460 nm, and 490 nm in diameter were prepared throughhydrolysis of TEOS in 200 proof ethanol with 0.2 M TEOS, 17.0 M DIwater, and 0.2, 1.2, and 1.6 MNH3 respectively.

19 309 nm particles wereprepared through similar hydrolysis of TEOS in 190 proof ethanol with0.04 M TEOS and 2.45 M NH3.

20 319 nm particles were prepared with200 proof EtOHwith 0.13MTEOS, 18.9 MDIH2O, and 1.6 MNH3.

20

Particles 407 nm in diameter were prepared in 190 proof ethanol with0.29 M TEOS, 6 MDI water, and 1.9 MNH3

21 In all cases, NH4OH, DIwater (when applicable) and ethanol were mixed initially. TEOS andethanol, also premixed, were added to the ammonia solution to start theparticle synthesis reaction. Note that DI H2O was not a uniquecomponent in some recipes—NH4OH is NH3 in H2O (29% NH3 onaverage and used as received).

While the upper limit of synthesizable particles through batchreactions is around 500 nm, larger particles were made by a core�shelltype approach. The first method was a continuous core�shell synthesisprocess in which seeds were first prepared by the addition of 1mLTEOSand 4 mL 200 proof ethanol to 46 mL 200 proof ethanol and 9.0 mLNH4OH (29%). After two hours an additional 5 mL TEOS/20 mLethanol solution was fed continuously at 0.1 mL/min to yield 700 nm

SiO2 microspheres.20 In a separate semibatch process, 500 nm seedswere grown to 928 and 1150 nm in diameters. Here, hydrolysis similar tothe batch processes presented earlier was carried out as 4.25 g TEOS([TEOS] = 0.2 M) and 35.88 g 200 proof EtOH were added to 13.63 gEtOH, 24.88 gDIH2O ([DIH2O] = 17.0M), 7.03 gNH4OH ([NH3] =1.2 M), and 300 μL 500 nm SiO2/EtOH (46% v/v).19 This yielded928 nm SiO2microspheres. An additional 4.25 g TEOS in 35.88 g EtOHadded one day later formed an additional shell atop the 928 nm SiO2

microspheres and brought the final diameter to 1150 nm.All experiments were carried out at room temperature in sealed

vessels under moderate stirring. Postsynthesis, the particles werecentrifuged and washed with ethanol and DI H2O a minimum of threetimes each. Centrifugation times were based on the terminal particlesettling velocity in order to remove both excess reagents and any smallersecondary constituents formed as byproducts in our reactions. Fordeposition, particles were suspended in DI H2O and brought to anappropriate volume fraction based on their size (see Table 1). Note thatchanges in chemical suppliers and product lots from those suppliersyielded small size variations in particles synthesized. However, withconsistent reagents particle growth was highly linear and scalable(Supporting Information S1).Preparation of Close-Packed Silica-Nanobead Substrates.

Close-packed silica-nanobead substrates were prepared through con-vective deposition following the method reported by Kumnorkaew.22

Plain glass microslides were used as deposition blades and substrates.Deposition substrates were cleaned in piranha solution (sulfuric acid andhydrogen peroxide at 3:1 volume ratio) for over 1 h, thoroughly rinsedwith deionized (DI) water, and then stored in DI water until needed.

Table 1. Concentration and Optimal Deposition Speed ofNanobeads Used in This Study

size of silica beads

((1 SD)

concentration

(v/v in DI-water) optimal speed

100( 10 nm 12.84% 3μm/s

309( 30 nm 12.84% 6μm/s

319( 17 nm 11.80% 33 μm/s

407( 23 nm 10.84% 10 μm/s

460( 20 nm 26.50% 21 μm/s

490( 20 nm 25.90% 58 μm/s

700( 24 nm 32.40% 17 μm/s

928( 16 nm 15.20% 42 μm/s

1150( 24 nm 12.50% 67 μm/s

Figure 1. Experimental setup for nanobead deposition. Silica nano-beads from 100 to 1150 nm in diameter, synthesized through TEOShydrolysis, were deposited into close-packed monolayers to controlsubstrate nanotopography. A linear motor was used to translate thesubstrate. The hydrophobically treated glass blade was positioned at a45� angle with respect to the substrate and 10 μL SiO2 nanobead/DIH2O suspension was pipetted into the space between. The beads packedinto a 2D crystal as the bulk suspension dragged across the glasssubstrate and the solvent (DI H2O) evaporated.

C dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

The ends of the deposition blades were hydrophobically treated bycoating with parafilm.

A schematic of the convective deposition setup is shown in Figure 1.The deposition blade angle was positioned at 45� above the substrate ascalibrated with a digital camera. Ten microliters of SiO2/DI H2Osuspension was injected between the substrate and blade with amicropipet. The substrate was then linearly translated at the appropriatespeed, as governed by specific suspension evaporative flux, for mono-layer deposition.

Substrates covered with a smooth layer of TEOS were prepared as acontrol with identical surface chemistry but minimal roughness (root-mean-squared surface roughness <2.0 nm, Supporting Information S2).Clean glass sides (root-mean-squared surface roughness <3.0 nm,Supporting Information S2). were soaked in an ethanol solution withTEOS for 15min. Subsequently, the slides were thoroughly washed withethanol and blown dry with nitrogen.Device Preparation. The nanoparticle-coated substrates were

patterned in a 3 mm � 3 mm square region for cell capture. Beadsoutside the square capture region were removed with a clean roomadhesive tape. A microchannel was then bonded atop the substrate toenclose both the nanosphere pattern and a bare glass control region.

Polydimethylsiloxane (PDMS) microchannels were prepared bypouring a 10:1 mixture of silicone elastomer base and silicone elastomercuring agent onto an SU8 mold patterned on a silicon wafer. Simplestraight channels with dimensions 14mm� 4mm� 50μm(L�W�H)were fabricated. PDMS was degassed and cured at 70 �C and themicrodevices were cut out. Fluid inlets and outlets were drilled usinga syringe needle. Nanobead-deposited substrates and microchannelswere activated by oxygen plasma, then carefully aligned and heated for5�10 min at 70 �C to produce permanent bonding. An image of a typicalassembled device is presented in Figure 2.

After assembly, the interior walls of the device, including thenanobead-deposited area and a bare glass region, were functionalizedusing a previously published method.11 A 4% (v/v) solution of 3-mer-captopropyl trimethoxysilane in ethanol was injected into the device andallowed to incubate for 1 h at room temperature. Next, the device wasincubated with 0.01 μmol/mL GMBS in ethanol for 1 h at roomtemperature. NeutrAvidin was then immobilized to GMBS by incubat-ing chamber surfaces with a 10 μg/mL NeutrAvidin solution in PBS(phosphate buffered saline) for at least 1 h at 4 �C. Finally, 10 μg/mL of

biotinylated antibody solution in PBS containing 1% (w/v) BSA wasinjected to react at room temperature for at least 15 min before cellcapture experiments were performed.Cell Culture and Live Cell Capture. Jurkat cells (human

lymphocytes) were maintained following the ATCC protocol at acell concentration of 105 cells/mL to 106 cells/mL using RPMI-1640media supplemented with 10% (v/v) FBS and 1% (v/v) pen-strep.Cells were centrifuged and resuspended at 106 cells/mL in culturemedia prior to cell capture experiments. Cell suspension was deliv-ered into the devices using a syringe pump at constant flow rates of2 μL/min, 5 μL/min, or 8 μL/min for 15 min. These flow rates arewithin the optimal window for CD4+ T cell capture, as found in priorwork.11 The sample injection time was chosen to ensure that targetcells were not depleted from the suspension prior to reachingthe downstream capture area. The numbers of cells captured inthe patterned and bare glass regions in a given channel (Figure 2,Device 1) were counted using a bright field microscope. The countson bare glass served as an internal control to account for slight batchto batch variations in testing conditions.

As bare glass slides have slightly different chemistry versus silica beadssynthesized by TEOS hydrolysis,23 we also prepared smooth TEOScoatings on glass substrates as samples withminimal roughness (Device 2).The normalized cell capture efficiency, η, were obtained by taking the ratioof cell counts on silica beads versus that on smooth TEOS coatings usingthe following formula

ηðRÞ ¼ nnbðDÞnbg1

� nbg2nTEOS

where nnb(D) is the number of cells captured on nanobeads in thepatterned region in Device 1, nbg1 and nbg2 are counts of cells capturedon bare glass region within Device 1 and Device 2, respectively, and nTEOSis the cell count from smooth TEOS-treated regions in Device 2.Capture of Fixed Cells. In order to test cells with altered

mechanical properties, cells were centrifuged and resuspended in 1%glutaraldehyde in PBS (v/v) for 30 min. This glutaraldehyde fixa-tion process cross-links proteins in the cell membrane and thus increasesits modulus. The fixed cells were centrifuged, washed in PBS, andthen resuspended in culture media at a concentration of 106 cells/mLfor capture. In a separate experiment, Jurkat cells were fixed with

Figure 2. Cell capture microchip used in this study. (A) Photograph of a microchannel formed by oxygen plasma activation and bonding of PDMS, withan imprinted microchannel, to the glass substrate, with nanoparticle-patterned and bare glass regions. The deposited nanobead pattern (cloudy region inthe black square) and a control region of bare glass (red dashed square) were enclosed in the same microchannel. The entire substrate was furtherfunctionalized with a CD4 antibody for CD4+ T cell capture from Jurkat cell cultures. (B) Schematic showing nanobead-covered and control glassregions within the same channel.

D dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

1% glutaraldehyde in PBS for 30 min and then stained with Hoechst33258 (0.5 mg/mL in PBS) for 10 min. After rinsing, the fixedand stained cells were mixed with untreated live Jurkat cells at a 1:1ratio in culture media with a total cell concentration of 106 cells/mL.This cell mixture was also used for cell capture at selected bead size andflow rates. To test if glutaraldehyde fixation affected antibody�antigenbinding, the same cell mixture was used for capture experimentsin microchannels bonded to antibody-functionalized bare glass.

The number of each cell type captured within the same imaging areaswas counted and the two quantities compared.COMSOL Simulation. Flow above a layer of close-packed beads

was simulated inCOMSOLMultiphysics 4.0 (COMSOL Inc., Burlington,MA). Assuming a 2D crystal structure, we carry out the simulationwithin a periodic cell of a hexagonal array of 1 μm spheres. A constantvelocity is imposed on the layer three particle diameters from thesubstrate in order to investigate the near-field disturbance. The fluid is

Figure 3. Scanning electronmicroscope images of close-packed particle arrays with diameters of (B) 100 nm, (C) 309 nm, (D) 319 nm, (E) 407 nm, (F)460 nm, (G) 490 nm, (H) 700 nm, (I) 928 nm, and (J) 1150 nm. Image (A) shows a glass surface coated with a smooth layer of TEOS. All scale bars =500 nm. Standard deviations of the particle sizes are summarized in Table 1.

Figure 4. (A) Bright field image showing CD4+ cells captured on a surface patterned with 928 nm beads under 5 μL/min sample flow (scale bar =50 μm). (B�D)Normalized capture efficiency on nanobead-covered substrates. Cell suspension flow rates were (B) 2 μL/min, (C) 5 μL/min, and (D)8 μL/min. The numbers of cells captured on the silica nanobead surfaces were normalized to those on smooth TEOS-coated surfaces (approximating aparticle layer of infinitely small particles) for direct comparison of capture efficiencies. Error bars show one standard deviation in 5�10 trials. Data pointslabeled with * have a statistically different cell capture compared with the capture on TEOS with p < 0.1, while data points in red circles have a staticallydifferent cell capture compared to its preceding neighbor (neighboring data set of a smaller nanobead size) with p < 0.1.

E dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

Newtonian and a volumetric flow rate of 2 μL/min renders a very smallReynolds number (Re, 1). Therefore, flow near the walls is self-similarand scales with bead size.Cell Imaging by Confocal Microscopy. Antibody immobilized

cells on smooth glass substrate were imaged using a confocal laserscanning microscope (Visitech international, model vteye). The cellswere stained by 0.16 mM Rhodamine B in the normal culture media for10 min and imaged without further washing. The z-stack images weretaken using 100� magnification and expressed as a 3D image withImageJ.Surface Characterization by Scanning Electron Micro-

scopy (SEM) and Atomic Force Microscopy (AFM). AFMmeasurements were carried out on a Nanoscale V AFM in tappingmode to characterize 100, 490, and 1150 nm individual particle rough-nesses as well as the roughness of TEOS-coated and bare glass.Gwyddion was used for data analysis and visualization. Individualparticle roughness values are documented as root mean squared (rms)roughness. SEM measurements were performed on all nanobeaddepositions with iridium coating using a Hitachi 4300 instrument(Japan). The diameter and packing of nanobeads are documented inthe images.Statistical Analysis. All cell captured experiments were repeated in

at least 5 independent devices, except for Figure 7, where the experi-ments were repeated in at least 3 independent devices. The data wereanalyzed using two-tailed t tests. Two p values were calculated for thedata points presented in Figure 4, Figure 6, and Figure 7: one fromcomparing the cell capture on nanobead surfaces to that on TEOS underthe same flow rate; the other by comparing the cell capture on aparticularly sized nanobead surface to its neighboring data point of asmaller nanobead diameter. When the p value between two data pointswas less than 0.1, their difference was considered statistically significant.

’RESULTS AND DISCUSSION

Preparation of Silica Bead Depositions. In this work, close-packed silica�nanobead depositions were used to accuratelycontrol surface nanotopography in microchannels over a broadrange of experimental conditions. The influences of these layerson immunoaffinity cell isolation were studied. Since the particlesurfaces, glass, and TEOS directly treated glass slides are allvery smooth with an rms roughness less than a few nanometersby AFMmeasurements (Supporting Information S2), the surfaceroughness is mainly controlled by the radius of the particles.The arithmetic surface roughness is defined by the followingequation:24

Ra ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1n ∑i¼ n

i¼ 1Yi2

s

where Ra is the arithmetic surface roughness, n is the number ofpoints along the top surfaces, and Yi is the vertical distance of theith point from the mean surface height. For close-packednanobead depositions, Ra can be further expressed as:

Ra ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1n ∑i¼ n

i¼ 1

Z θ¼π=2

θ¼ 0R cos θ

!2vuut ¼ radius of nanobeads

We used convective deposition to create ordered nanobeadmonolayers atop glass substrates. While humidity, particle size,suspension volume fraction, substrate speed, and blade anglewere all found to influence the packing quality of nanoparticles,22

preliminary work showed that solely varying deposition speedwas sufficient to obtain hexagonally close-packed nanobead

arrays. The beads were deposited from suspensions at10.84�32.40% volume fractions in water. Close-packed mor-phologies of the deposited layers were verified with scanningelectron microscopy (SEM). Optimized deposition speeds foreach particle size are summarized in Table 1. The correspondingnanoparticle layer morphologies are presented in Figure 3. Thesize range of nanobeads was chosen because they are on the sameorder of magnitude as the size of nanostructures on cellmembrane, which was suggested to enhance cell capture inprevious work.12

Even though the size of the nanobeads used in our studycovers a broad range, the increased surface area for cell captureactually remains the same for all sizes and is independent ofnanobead diameter as long as the beads have similar packingstructure. This can be shown by the following calculation:

Total area for cell capture after nanobead depositon

¼ AfπR2

2πR2 ¼ Aπffiffiffi3

p

where A is the surface area on the substrate covered by beads, f isthe fraction of the substrate covered by beads, which is π/(2

√3)

for 2D hexagonal close-packing, and R is the radius of thenanobead. The capture surface area on the nanobead-coveredsubstrates is∼1.8 times of that on a flat substrate, such as TEOS.Live Cell Capture on Silica-Nanobead Substrates.Next, the

silica nanobead deposited substrates were enclosed withinPDMS microchannels and functionalized with CD4 antibodiesto capture CD4+ T cells under continuous flow. Three differentflow rates, 2, 5, and 8 μL/min, covering the optimal range forCD4+ T cell capture, were evaluated.11 Jurkat cells, a human Tlymphocyte cell line, were used for capture experiments, and23.8% of the cells were found to be CD4+ as by flow cytometryanalysis (Supporting Information S3). The total number ofcaptured cells was enumerated as specific capture since thesurface chemistry has been shown to yield greater than 95%purity.11 The number of cells captured on the silica nanobeaddeposition was normalized to that on a functionalized smoothTEOS surface (approximating a continuous array of infinitelysmall silica beads) using the formula shown in the Experimental

Figure 5. (A) Fluid velocity profile at a cross section parallel to the flowdirection simulated by COMSOL. Velocities and shear rates correspondto 2 μL/min. Not far from the beads, the flow profile nears Poiseuilleconditions and is essentially undisturbed by the bead patterns. The shearrate along the z-direction above the bead center and above the two beadcontact point at various height points were computed and shown in (B).Greater than two bead diameters from the glass surface, velocitydifferences between the two points are effectively zero and the flow isno longer disturbed by the imposed roughness.

F dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

Section. This normalization minimizes the effects of any run-to-run nonuniformity and allows direct comparison of captureefficiencies with different surface nanotopography.Figure 4A shows a typical image of cells captured on the

nanopatterned surface. Typically,∼50 cells/mm2 were capturedwith 15 min sample delivery. Reversing the flow direction did notaffect capture results, indicating that the capture difference in theup- and downstream regions was not due to target cell depletion.Figure 4B�D shows that nanopatterned surfaces generallyimprove capture efficiency versus smooth substrates; however,the two do not have a monotonic relationship. Statisticallysignificant data sets are highlighted. Data points with * have astatistically different cell capture compared with the capture onTEOS under the same flow rate with p < 0.1. Data points in redcircles have statistically different cell capture as compared to theirpreceding neighbor (neighboring data set of a smaller nanobeadsize) with p < 0.1. The cell capture efficiency profile stronglydepends on flow rate. At 2 μL/min (wall shear stress of 5.5 �10�5 N/m2), the capture efficiency increases nearly continuouslyfrom 100%, the normalization value on smooth TEOS (control)to 160% on roughened surfaces with the largest beads(1150 nm). Interestingly, the capture efficiency dips on surfacescoated with 700 nm particles to the level of the control, smoothTEOS, samples. At 5 μL/min (wall shear stress of 1.385� 10�4

N/m2), the cell capture efficiency first climbs with increasedsurface roughness, and gradually decreases as bead size furtherincreases. An outlier to this trend is a significant decrease incapture efficiency with a 407 nm bead-patterned substrate—thisyielded only 60% capture as compared with the smooth TEOSsurface. The 8 μL/min (wall shear stress of 2.215� 10�4 N/m2)trend follows the 5 μL/min capture profile. The cell captureefficiency nearly doubles for the smallest deposited particles,then decreases with nanobeads greater than 700 nm in diameter.The capture profile shows two significant dips at 407 and 490 nmwhere the capture efficiencies are comparable to those on smoothsurfaces. Thus, cell capture on rough surfaces is dependent notonly on surface topography, but also on wall shear stress.

The complex capture profiles are not expected to result fromincreased surface area after nanobead deposition. The differencesbetween TEOS and nanobead surfaces are likely due to surface areaincrease (an 80% increase in surface area with nanobead depos-ition), but differences between the various beads are not (surfacearea is independent of nanobead size under hexagonal close-packing). They are also likely not a result of antibody surfacechemistry or antibody density as different-sized beads were synthe-sized under the same TEOS hydrolysis reaction with only reactiontime and reagent concentration varying. In addition, treating amonolayer of nanoparticle deposition with a TEOS solutionfollowed with antibody conjugation does not change cell capture(Supporting Information S4); this also suggests comparable surfacechemistry among the different roughness conditions.Cell capture on rough surfaces under continuous flow could be

influenced by cell movement near substrates, shear induceddeformation, and increased capture surface area as proposedpreviously. Inertial migration of flowing cells, however, shouldnot be a main contributor to the complex capture patterns as theReynolds number Re is much smaller than 1 under all flowconditions. Numerical simulation (Figure 5) confirms that theflow is essentially undisturbed at heights roughly two bead-diameters from the bottom substrate. Even the largest particlesdeposited here (1150 nm in diameter) are much smaller thanJurkat cells (7�20 μm in diameter25) captured. As a result, cellmovement under these near-Stokes flow conditions should notsignificantly undulate with surface topography or lead to anyabrupt changes to cell migration at the dipping points in thecapture efficiency curves. In addition, in these dilute cellularsuspensions, multibody hydrodynamic interactions are alsominimal. The only nonlinear behavior in this near-wall interac-tion is the elasticity of the cellular membrane. We thus hypothe-size that mechanical deformation of cells on the cellular andsubcellular level is the key factor contributing to the complexcapture profiles observed.Capture of Glutaraldehyde Fixed Cells on Silica-Nano-

bead Substrates. To evaluate the hypothesis that cell mechan-ical properties influence their capture, Jurkat cells were fixedwith 1% glutaraldehyde and used in capture experiments under

Figure 6. Capture of glutaraldehyde-fixed Jurkat cells from culturemedia onto nanobead-patterned substrates at 5 μL/min. Numbers ofcells captured on nanobeads were normalized to those captured on flatTEOS-coated surfaces. The error bars represent 1 standard deviation ascalculated from greater than four repeats of each condition. p value wasfound to be greater than 0.126 for every data point, either comparingwith capture efficiency on TEOS or its preceding neighbor, indicatingthat capture on the various surfaces was not significantly different.

Figure 7. Capture of glutaraldehyde-fixed and live Jurkat cell mixture(1:1 ratio) from culture media onto 490 nm nanobead-patternedsubstrates at 5 μL/min and 8 μL/min. Numbers of cells captured onnanobeads were normalized to those captured on flat TEOS-coatedsurfaces. The error bars represent 1 standard deviation as calculatedfrom greater than three repeats under each condition. Data pointslabeled with * have a statistically different cell capture compared with thecapture on TEOS with p < 0.1.

G dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

5 μL/min flow. Glutaraldehyde cross-links amine groups andthus can increase the Young’s modulus of cells by greater than anorder of magnitude.26 Figure 6 shows the normalized captureefficiency of fixed cells: capture efficiency increases slightly forthe smallest nanobead layers then gradually decreases. With fixedcells, no dip in cell capture efficiency is seen, and the surfacetopography effects are greatly dampened, as no statically sig-nificant difference was found no matter when neighboring datapoints were compared or when the capture on nanobeads werecompared to that on flat TEOS.To further verify the different capture between live and fixed

cells, live (triangles in Figure 7) and fixed (dots) Jurkat cells weremixed at a 1:1 ratio for capture on 490 nm beads under 5 μL/min(empty symbols) and 8 μL/min flow (solid symbols). As shownin Figure 7, the normalized capture efficient is consistent with theresults in Figure 4 and Figure 6 using live or fixed cells alone

under the respective conditions. In addition, capture of glutar-aldehyde-fixed cells is much less sensitive to change in surfaceroughness: under 5 μL/min flow, the p-value is 0.7597 whencomparing capture of fixed cells on 490 nm bead monolayers tothat on smooth TEOS, while there is a significant difference forlive cell capture on the two surfaces (p = 0.0002). Live cellcapture is also highly sensitive to the flow rate, showing by thesignificant difference of capture efficiency at the two tested flowrates (p = 0.005 18).To test if glutaraldehyde fixation interferes with antibody�

antigen binding chemistry, we fixed Jurkat cells with glutaralde-hyde and then stained the nuclei with Hoechst 33258. Afterward,fixed and stained Jurkat cells were mixed with live cells at a 1:1ratio. The cell mixture was manually injected into a PDMSmicrochannel bonded to functionalized bare glass and incubatedunder static conditions for 15 min. The device was rinsed andcaptured cells were counted—Figure 8 shows a representativeimage. Average cell counts from multiple devices and imagingareas show that identical numbers of fixed and untreated cellswere captured. This suggests that the different capture profiles oflive and fixed cells are not due to altered antibody�antigeninteractions.A direct outcome of glutaradehyde fixation is significant

enhancement of cell moduli.26 Live CD4+ cells have a flexiblemembrane undulating even in static culture medium (SupportingInformation S5). The varying levels of shear deformation underdifferent flow rates are expected to influence the interaction areabetween cells and antibody-coated substrates. Although sheardeformation on the cellular level may contribute to the flow ratedependent capture, it does not fully explain cell interaction withnanotopography, which is an order of magnitude smaller thancell dimension in our study.Instead, a more plausible explanation of the capture depen-

dence on surface nanofeatures and wall shear stress may arisefrom protruding nanostructures on the cell membrane. Thepredominant features on the surface of human lymphocytes aremicrovilli27 that consist of bundles of cross-linked actinfilaments.27 CD4 receptors, the target antigen in this study, arepreferentially located on the microvilli ends.28 Indeed, weobserved cell extensions a few hundred nanometers in diametertethered to antibody functionalized substrates (Figure 9), in-dicating interactions between membrane nanostructures and thecapture bed. With a dimension of 300 to 400 nm in length,∼100 nm in width, and a spatial distribution of ∼4/μm2,27

microvilli are comparable in size to the nanobeads used in thisstudy. This comparable dimension scale suggests a physicalcomplementarity between certain substrate nanotopographiesand the cell surface in addition to specific antibody�antigeninteractions. Complementary interactions have been reportedwith nanopillars to enhance the capture yield of circulating tumorcells.12 Optimal capture yield occurs with 6 μm or longernanopillars which are on the same length scale as the protrusionlengths of tumor cells.12 In our system, while the amount ofaccessible antibody may increase continuously with the nano-bead diameter,15 the physical complementarity has a nonmono-tonic dependence on the nanoparticle size. The interplay of thetwo factors may strongly contribute to the complex captureprofiles on the nanostructured surfaces. Interestingly, the tethersobserved in Figure 9 aremuch longer than the length of microvillimeasured on unattached lymphocytes.27 These long extensionsmay be a result of microvilli being stretched after initial tethering.The diameter of these extensions is also slightly thicker than

Figure 8. Representative bright field image showing the capture ofglutaraldehyde-fixed (blue) and live (gray) Jurkat cells mixed 1:1 inculture media on antibody-functionalized bare glass substrates understatic incubation. Average cell counts frommultiple devices and imagingareas show that identical numbers of fixed and untreated cells werecaptured. This indicates that the fixation procedure does not interferewith antibody�antigen recognition. Scale bar = 50 μm.

Figure 9. A confocal Z-stack image showing tethers from a RhodamineB stained Jurkat cell to an antibody functionalized smooth glass substrate(scale bar = 2 μm). After cell capture from continuous flow on anantibody treated smooth glass slide, the flow was stopped and confocalimages were acquired. The fluid flowed from the right to the left for cellcapture, and the substrate is located at the bottom of the image asindicated by the dashed line. The fuzziness of the membrane away fromthe substrate is a result of spontaneous membrane fluctuation.

H dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

those measured by AFM or SEM,27 likely due to optical diffrac-tion. Since microvilli are susceptible to shear forces, highlydynamic and deformable to help cells contact and bind topatterned substrates,29,30 they can experience different sheardeformation and contribute to the flow dependence of the cellcapture. While fixed cells also present microvilli,27 their lack ofdynamics and weak response to shear forces makes capture offixed cells much less sensitive to the surface topography.

’CONCLUSION

Close-packed silica bead layers provide a powerful tool tosystematically control surface roughness and enhance immu-noaffinity cell capture under continuous flow. Interestingly,complicated cell capture efficiency profiles were observed acrossseveral flow rates and surface nanotopographies. Hardening cellsthrough glutaraldehyde fixation significantly dampened thisdependence. The nonmonotonic relationships between thecapture efficiency and surface roughness under various flow ratesare likely the results of complementarity and shear deformationof nanoprotrusions on the cell membrane. In future studies, weplan to extend the length scale and investigate the influence ofsubstrate microstructures on cell capture. This relationshipbetween surface topography and cell capture efficiency, onceestablished, will enable more rational design of cell capturedevices.

’ASSOCIATED CONTENT

bS Supporting Information. Preparation of SilicaMicro andNanoparticles and Control Experiments, with figures and a video.This material is available free of charge via the Internet at http://pubs.acs.org.

’AUTHOR INFORMATION

Corresponding Author*Email: xuc207@lehigh.edu. Address: Dr. Xuanhong Cheng, 5 E.Packer Ave, Bethlehem, PA, 18015. Phone: 610-758-2002. Fax:610-758-4244.

Author ContributionsB.W. and A.W. contributed equally to the work presented in themanuscript.

’ACKNOWLEDGMENT

We are grateful for Makenzie Wolfe, Christopher Tibaldi, andJonathan Longley’s help in nanoparticle deposition, COMSOLsimulation, and AFM imaging, respectively. Funding for theresearch is provided by National Science Foundation underGrant No. CBET-0828426 and by the Biosystems DynamicsSummer Institute 2009, 2010 sponsored by the Howard HughesMedical Institute.

’REFERENCES

(1) Plouffe, B. D.; Radisic, M.; Murthy, S. K. Microfluidic depletionof endothelial cells, smooth muscle cells, and fibroblasts from hetero-geneous suspensions. Lab Chip 2008, 8 (3), 462–472.(2) Gomez-Sjoberg, R.; Morisette, D. T.; Bashir, R. Impedance

microbiology-on-a-chip: Microfluidic bioprocessor for rapid detectionof bacterial metabolism. J. Microelectromech. Syst. 2005, 14 (4), 829–838.

(3) Du, Z.; Colls, N.; Cheng, K. H.; Vaughn, M. W.; Gollahon, L.Microfluidic-based diagnostics for cervical cancer cells. Biosens. Bioelec-tron. 2006, 21 (10), 1991–1995.

(4) Nagrath, S.; Sequist, L. V.; Maheswaran, S.; Bell, D. W.; Irimia,D.; Ulkus, L.; Smith, M. R.; Kwak, E. L.; Digumarthy, S.; Muzikansky, A.;Ryan, P.; Balis, U. J.; Tompkins, R. G.; Haber, D. A.; Toner, M. Isolationof rare circulating tumour cells in cancer patients by microchip technol-ogy. Nature 2007, 450 (7173), 1235–U10.

(5) Liu, Y. J.; Guo, S. S.; Zhang, Z. L.; Huang, W. H.; Baigl, D.; Xie,M.; Chen, Y.; Pang, D. W. A micropillar-integrated smart microfluidicdevice for specific capture and sorting of cells. Electrophoresis 2007,28 (24), 4713–4722.

(6) Faley, S.; Seale, K.; Hughey, J.; Schaffer, D. K.; VanCornpernolle,S.; McKinney, B.; Baudenbacher, F.; Unutmaz, D.; Wikswo, J. P.Microfluidic platform for real-time signaling analysis of multiple singleT cells in parallel. Lab Chip 2008, 8 (10), 1700–1712.

(7) Kotz, K. T.; Xiao, W.; Miller-Graziano, C.; Qian, W. J.; Russom,A.; Warner, E. A.; Moldawer, L. L.; De, A.; Bankey, P. E.; Petritis, B. O.;Camp, D. G.; Rosenbach, A. E.; Goverman, J.; Fagan, S. P.; Brownstein,B. H.; Irimia, D.; Xu, W. H.; Wilhelmy, J.; Mindrinos, M. N.; Smith,R. D.; Davis, R. W.; Tompkins, R. G.; Toner, M. Inflammation HostResponse, I., Clinical microfluidics for neutrophil genomics and pro-teomics. Nat. Med. 2010, 16 (9), 1042–U142.

(8) Stott, S. L.; Hsu, C. H.; Tsukrov, D. I.; Yu, M.; Miyamoto, D. T.;Waltman, B. A.; Rothenberg, S. M.; Shah, A. M.; Smas, M. E.; Korir,G. K.; Floyd, F. P.; Gilman, A. J.; Lord, J. B.; Winokur, D.; Springer, S.;Irimia, D.; Nagrath, S.; Sequist, L. V.; Lee, R. J.; Isselbacher, K. J.;Maheswaran, S.; Haber, D. A.; Toner, M. Isolation of circulating tumorcells using a microvortex-generating herringbone-chip. Proc. Natl. Acad.Sci. U. S. A. 2010, 107 (43), 18392–18397.

(9) Neurauter, A. A.; Bonyhadi, M.; Lien, E.; Nokleby, L.; Ruud, E.;Camacho, S.; Aarvak, T., Cell isolation and expansion using Dynabeads-((R)). In Cell Separation: Fundamentals, Analytical and PreparativeMethods; Springer-Verlag Berlin: Berlin, 2007; Vol. 106, pp 41�73.

(10) Hashimoto, M.; Kaji, H.; Nishizawa, M. Selective capture of aspecific cell type from mixed leucocytes in an electrode-integratedmicrofluidic device. Biosens. Bioelectron. 2009, 24 (9), 2892–2897.

(11) Cheng, X. H.; Irimia, D.; Dixon, M.; Sekine, K.; Demirci, U.;Zamir, L.; Tompkins, R. G.; Rodriguez, W.; Toner, M. A microfluidicdevice for practical label-free CD4+T cell counting of HIV-infectedsubjects. Lab Chip 2007, 7 (2), 170–178.

(12) Wang, S. T.; Wang, H.; Jiao, J.; Chen, K. J.; Owens, G. E.; Kamei,K. I.; Sun, J.; Sherman, D. J.; Behrenbruch, C. P.; Wu, H.; Tseng, H. R.Three-dimensional nanostructured substrates toward efficient capture ofcirculating tumor cells. Angew. Chem., Int. Ed. 2009, 48 (47), 8970–8973.

(13) Hughes, A. D.; King, M. R. Use of naturally occurring halloysitenanotubes for enhanced capture of flowing cells. Langmuir 2010,26 (14), 12155–12164.

(14) Han, W. J.; Allio, B. A.; Foster, D. G.; King, M. R. Nanoparticlecoatings for enhanced capture of flowing cells in microtubes. ACS Nano2009, 4 (1), 174–180.

(15) Tuttle, P. V.; Rundell, A. E.; Webster, T. J. Influence ofbiologically inspired nanometer surface roughness on antigen-anti bodyinteractions for immunoassay-biosensor applications. Int. J. Nanomed.2006, 1 (4), 497–505.

(16) Kalasin, S.; Martwiset, S.; Coughlin, E. B.; Santore, M. M.Particle capture via discrete binding elements: systematic variations inbinding energy for randomly distributed nanoscale surface features.Langmuir 2010, 26 (22), 16865–16870.

(17) Kalasin, S.; Santore, M. M. Sustained rolling of microparticlesin shear flow over an electrostatically patchy surface. Langmuir 2010,26 (4), 2317–2324.

(18) Santore, M. M.; Kozlova, N. Micrometer scale adhesion onnanometer-scale patchy surfaces: Adhesion rates, adhesion thresholds,and curvature-based selectivity. Langmuir 2007, 23 (9), 4782–4791.

(19) Wang, W.; Gu, B. H.; Liang, L. Y.; Hamilton, W. Fabrication oftwo- and three-dimensional silica nanocolloidal particle arrays. J. Phys.Chem. B 2003, 107 (15), 3400–3404.

I dx.doi.org/10.1021/la2015868 |Langmuir XXXX, XXX, 000–000

Langmuir ARTICLE

(20) Zhang, J. H.; Zhan, P.; Wang, Z. L.; Zhang, W. Y.; Ming, N. B.Preparation of monodisperse silica particles with controllable size andshape. J. Mater. Res. 2003, 18 (3), 649–653.(21) Jiang, S. Recipes for St€ober Silica Particles. http://mit.edu/

sjiang2/www/Resources/stober_recipes.pdf.(22) Kumnorkaew, P.; Ee, Y. K.; Tansu, N.; Gilchrist, J. F. Investiga-

tion of the deposition of microsphere monolayers for fabrication ofmicrolens arrays. Langmuir 2008, 24 (21), 12150–12157.(23) Tabatabaei, S; Shukohfar, A; Aghababazadeh, R; Mirhabibi, A

Experimental study of the synthesis and characterisation of silicananoparticles via the sol-gel method. J. Phys.: Conf. Ser. 2006, 26 (1),371–374.(24) Huang, J. Q.; Hu, G. S.; Orkoulas, G.; Christofides, P. D.

Dependence of film surface roughness and slope on surface migrationand lattice size in thin film deposition processes. Chem. Eng. Sci. 2010,65 (23), 6101–6111.(25) Deng, J. D.; Schoenbach, K. H.; Buescher, E. S.; Hair, P. S.; Fox,

P.M.; Beebe, S. J. The effects of intense submicrosecond electrical pulseson cells. Biophys. J. 2003, 84 (4), 2709–2714.(26) Lulevich, V.; Zink, T.; Chen, H. Y.; Liu, F. T.; Liu, G. Y. Cell

mechanics using atomic force microscopy-based single-cell compres-sion. Langmuir 2006, 22 (19), 8151–8155.(27) Majstoravich, S.; Zhang, J.; Nicholson-Dykstra, S.; Linder, S.;

Friedrich, W.; Siminovitch, K. A.; Higgs, H. N. Lymphocyte microvilliare dynamic, actin-dependent structures that do not require Wiskott-Aldrich syndrome protein (WASp) for their morphology. Blood 2004,104 (5), 1396–1403.(28) Singer, I. I.; Scott, S.; Kawka, D. W.; Chin, J.; Daugherty, B. L.;

DeMartino, J. A.; DiSalvo, J.; Gould, S. L.; Lineberger, J. E.; Malkowitz,L.; Miller, M. D.; Mitnaul, L.; Siciliano, S. J.; Staruch, M. J.; Williams,H. R.; Zweerink, H. J.; Springer, M. S. CCR5, CXCR4, and CD4 areclustered and closely apposed on microvilli of human macrophages andT cells. J. Virol. 2001, 75 (8), 3779–3790.(29) Amieva, M. R.; Furthmayr, H. Subcellular-localization of moe-

sin in dynamic filopodia, retraction fibers, and other structures involvedin substrate exploration, attachment, and cell-cell contacts. Exp. Cell Res.1995, 219 (1), 180–196.(30) Pospieszalska, M. K.; Ley, K. Dynamics of microvillus extension

and tether formation in rolling leukocytes. Cell. Mol. Bioeng. 2009, 2 (2),207–217.