Lab on a Chip

PAPER

Cite this: Lab Chip, 2017, 17, 702

Received 6th December 2016,Accepted 23rd January 2017

DOI: 10.1039/c6lc01495j

rsc.li/loc

Pipetting-driven microfluidicimmunohistochemistry to facilitate enhancedimmunoreaction and effective use of antibodies†

Segi Kim,‡a Seyong Kwon,‡a Chang Hyun Choa and Je-Kyun Park*ab

Immunohistochemistry (IHC), which has been used to detect antigens in cells of a tissue section using an

immunoreaction between an antibody and an antigen, is a practical tool for identifying the type and stage

of diseases in cancer diagnosis and scientific research. However, conventional IHC requires long, laborious

process times and high costs. Although microfluidic IHC platforms have been developed to overcome

these limitations, the application of microfluidic IHC in real-world environments is still limited due to the

additional equipment needed to operate the microfluidic systems. In addition, continuous flow in a micro-

fluidic channel leads to a waste of unbound antibodies. In this study, we demonstrate a novel and easy-to-

use microfluidic IHC platform operated only using a manual pipette that is commonly available in research

laboratories or hospitals. No other device such as a pump or a controller is required to operate our system.

Bidirectional flows of the antibody solution in a microfluidic device are induced by repetitive manual

pipetting which facilitates the enhanced antigen–antibody reaction and enables the effective use of a lim-

ited amount of antibody. When breast cancer cell and tissue sections are reacted with antibodies using our

platform, pipetting for less than 2 min is sufficient to obtain immunostaining results without damaging the

sample. The staining intensity by our method is similar to that of the sample stained for 1 h by a conven-

tional batch process. We believe that this pipetting-based approach to the operation of a microfluidic sys-

tem allows end users to use microfluidic IHC more conveniently and easily in real-world environments.

Introduction

Immunohistochemistry (IHC) is a diagnostic technique usedto detect biomarkers in cells or tissues by antigen–antibodyreaction used for identifying the type and stage of diseases.1

IHC is the only technique capable of providing quantitativeinformation on the expression, localization, and morphologyof specific proteins2 among various kinds of cancer diagnostictechniques such as polymerase chain reaction,3 western blot-ting,4 enzyme-linked immunosorbent assay5 and massspectroscopy.6 Because of these characteristics, IHC has beenused as the gold standard for pathologists to diagnose cancer7

and plays an important role in many situations, includingsubtyping and characterization of breast cancer,8 discrimina-tion between benign and malignant tumors,9 histogenetic di-

agnosis of morphologically non-differentiated tumors,10,11

and primary site tracking of metastatic tumors.12

In conventional IHC, a drop of antibody solution is ap-plied to a tissue sample and incubated for a certain period oftime for a sufficient immunoreaction of the antibody. Thisstatic method is inevitably faced with the limitation of reac-tion kinetics; there should be a depletion zone where theantibody is rare around the reaction surface and reduces theefficiency of the antigen–antibody reaction. This is due to theinsufficient diffusion of the antibody from the bulk solutionto the surface-immobilized targets.13 Considering the diffu-sivity of immunoglobulin G, a depletion zone is formedwithin 1 min of incubation, which results in the antigen–antibody reaction taking more than an hour or even over-night to reach equilibrium.14

To address this problem, some microfluidic platformshave been developed for the IHC process. In general, micro-fluidic technology has the advantage of allowing precise oper-ation and manipulation of fluid and biological particles suchas DNA, RNA, proteins and cells.15 The ultra-low dimensionsof the microfluidic platform provide a high surface-to-volumeratio that induces effective antigen–antibody reactions andreduces the consumption of expensive reagents and valuablesamples.16 These benefits can potentially reduce the cost,

702 | Lab Chip, 2017, 17, 702–709 This journal is © The Royal Society of Chemistry 2017

aDepartment of Bio and Brain Engineering, Korea Advanced Institute of Science

and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Republic of

Korea. E-mail: jekyun@kaist.ac.kr; Fax: +82 42 350 4310; Tel: +82 42 350 4315bKAIST Institute for Health Science and Technology, 291 Daehak-ro, Yuseong-gu,

Daejeon 34141, Republic of Korea

† Electronic supplementary information (ESI) available: Fig. S1 and supplemen-tary video. See DOI: 10.1039/c6lc01495j‡ The first two authors contributed equally to this work.

Lab Chip, 2017, 17, 702–709 | 703This journal is © The Royal Society of Chemistry 2017

time, and labor associated with the experimental process.Previously, we developed a multiplexed microfluidic IHC plat-form that could target a variety of biomarkers.2,17 This plat-form enables the use of diluted antibodies, much fewer thanthose in conventional IHC, allowing the simultaneous investi-gation of multiple quantitative information for a single sam-ple section. Ciftlik et al. have also developed a rapid micro-fluidic tissue processor for IHC.18 They were able toimmunostain within 5 min and discriminate the ambiguousHER2 expression of breast cancer with their device.

Although these microfluidic approaches can overcome sev-eral limitations of conventional IHC, the microfluidic IHChas not yet been applied in clinical laboratories. This is dueto the external equipment required to operate the micro-fluidic system, such as a micropump, a microvalve and a con-troller. Therefore, a simple and easy-to-use operating ap-proach is required to operate the microfluidic system.Moreover, less than 10% of the total mass fraction of bulk so-lution can participate in the reaction and the remaining anti-bodies above the reaction surface are washed out.19 Thismeans that unreacted antibodies are wasted. The maximumflow rate is also limited in the range of μL h−1 to μL min−1.This is because the higher the flow rate, the more antibodiesare consumed.

In this study, we demonstrate a simple and easy-to-usepipetting-driven microfluidic immunohistochemistry (PMI)platform. Only a conventional micropipette was used to oper-ate the microfluidic IHC system without external pumps. Re-petitive manual pipetting induced bidirectional (pulling andpushing) flows of the antibody solution in the microfluidicdevice, resulting in an enhanced antigen–antibody reactionand efficient use of the unreacted antibody in a fixed amountof reagents. In addition, the flow rate of the reagents can beadjusted by changing the volume of the pipette to μL s−1.Based on the PMI, we investigated the effect of the micropi-pette volume on the immunostaining results and comparedthem to the results from breast cancer cell and tissuesections.

Materials and methodsPDMS device fabrication

SU-8 2050 was purchased from MicroChem Corp. (Newtown,MA, USA). PolyIJdimethylsiloxane) (PDMS) precursors, Sylgard184, were obtained from Dow Corning (Midland, MA, USA).Trimethylchlorosilane was purchased from Sigma-Aldrich (St.Louis, MO, USA). The mold for the PDMS device was fabri-cated by a standard soft lithography process. To construct amicrochannel and a rectangular reaction microchamber, thephotoresist SU-8 2050 was spin-coated to obtain a 60 μmthick layer on a bare silicon wafer. After pre-baking, itwas exposed to ultraviolet (UV) light, post baked, anddeveloped. To fabricate the PDMS device, the mold was ex-posed to trimethylchlorosilane for 20 min. As for the PDMSprecursors, a monomer and a curing agent were mixed wellin a ratio of 10 : 1 and poured on the SU-8 patterned master

mold. After baking for 60 min at 90 °C on a hot plate, thePDMS device was peeled off from the master mold.

Preparation of cell and tissue samples

MCF-7 and SKBR3 cells were purchased from the Korean CellLine Bank (Seoul, Korea). Roswell Park Memorial Institutemedium 1640 (RPMI 1640), fetal bovine serum (FBS), penicil-lin and streptomycin were purchased from Dow Corning(Midland). For the demonstration of our platform, aformalin-fixed and paraffin-embedded breast cancer cellblock was used. Two breast cancer cell lines, MCF-7 andSKBR-3, were cultured in RPMI 1640 supplemented with 10%FBS, 100 IU mL−1 penicillin and 100 mg mL−1 streptomycin.All cell lines were maintained at 37 °C and 5% CO2. For thefabrication of the formalin-fixed and paraffin-embedded sec-tion, the confluent cells were trypsinized, harvested, andcentrifuged. Precipitated cells were fixed in 4% formalin for10 min, dehydrated with graded alcohol, and embedded withparaffin. Paraffin embedded cell blocks were then sectionedwith a thickness of 4 μm using a microtome (Leica Bio-systems, Wetzlar, Germany). After being mounted on adhe-sive microscope slides (Marienfeld-Superior, Lauda-Königshofen, Germany), the samples were dried overnight at40 °C.

Human breast cancer tissue samples were kindly providedby the National Cancer Center Hospital (Goyang, Korea) witha corresponding written agreement with the patients or theirrelatives. This study was approved by the Korea Advanced In-stitute of Science and Technology (KAIST) and the Institu-tional Review Board (IRB) at the National Cancer Center Hos-pital. The human tissue samples were fixed for 48 h in 4%neutral-buffered formalin. Paraffin embedded tissue blockswere then sectioned at 4 μm thicknesses using a microtome.After being mounted on an adhesive microscope slide, thesamples were dried for 1 h at room temperature and for 1 hin a convection incubator at 60 °C.

Immunohistochemical staining

The HRP-DAB staining kit (ab80436) was purchased fromAbcam (Cambridge, UK). The target retrieval solution,cytokeratin antibody, and anti-human epidermal growth factorreceptor 2 (HER2) antibody were purchased from Dako (Troy,MI, USA). The antibody diluent was purchased from Life Tech-nologies (Carlsbad, CA, USA). Tris-buffered saline with Tween(TBS-T; 0.1% Tween 20) was purchased from ScyTek (Logan,UT, USA). Before the reaction with staining solution, the celland tissue sections were deparaffinized and rehydrated withxylene, 100% ethyl alcohol, and 95% ethyl alcohol for 20, 6,and 2 min, respectively. To enhance the antigenicity of thecell and tissue sections, a microwave antigen-retrieval tech-nique was used. The deparaffinized samples were treated for20 min at 750 W in a target retrieval solution (pH 9). Aftercooling for 20 min at room temperature, immunostainingwas conducted using a commercially available HRP/DAB IHCkit. Two biological markers were investigated. The cytokeratin

Lab on a Chip Paper

704 | Lab Chip, 2017, 17, 702–709 This journal is © The Royal Society of Chemistry 2017

antibody was diluted to 1 : 1000 with the diluent solution andthe HER2 antibody was diluted to 1 : 500 with the diluent so-lution before use. For primary antibody incubation, the pre-diluted antibody was incubated for a few cycles in the PMIplatform and for 1 h in the conventional batch process. TheTBS-T buffer was used for every washing step.

Operation of the pipetting-driven microfluidic immunohisto-chemistry (PMI) platform

After blocking the endogenous peroxidase and the non-specific binding site of the antibody using a blocking solu-tion (component of ab80436; Abcam), immunostaining wasperformed by reacting the deparaffinized sample with a pri-mary antibody in the PMI platform. Fig. 1a shows a sche-matic representation of the operation of the PMI device forantibody incubation. First, the PMI device was placed on thesample slide by pressure-based reversible bonding, whichwas conducted using a previously reported method.17 Briefly,a sample slide pre-soaked with washing buffer was placed onthe bottom plate. After the plasma-treated PDMS device cov-ered with a transparent upper plate was placed on the sam-ple, a weight (1 kg) was placed on the upper plate, wherebythe sample was pressed by the walls of the microchannels.Second, while pressing the button of the micropipette (Re-search®; Eppendorf, Hamburg, Germany), the antibody solu-tion was loaded into one inlet and the yellow pipette tip wasinserted into the other inlet. A PDMS adaptor, made of sim-ple PDMS pieces having a through-hole, and a Tygon tubewere used to tightly link the pipette tip and the inlet. Oncethe pipette is released, the loaded antibody solution at the in-let is aspirated into the microchannel by the negative pres-

sure of the pipette inducing a pulling flow. After a while,pressing the pipette again causes the antibody to flow in theopposite direction, leading to a pushing flow. During repeti-tive pipetting, the antibody can effectively bind to the antigenon the sample (Fig. 1b). While a portion of the antibodybinds to the antigen sites on the tissue by the pulling flow,the unreacted antibodies can bind to the unbound antigensites by the pushing flow.

Image analysis

ImageJ software, which is a public domain software releasedby the National Institutes of Health in the USA, was used toquantitatively analyze and compare the stained samples. Af-ter staining, the images of the stained samples were capturedusing a microscope (IX51; Olympus, Japan) and a CCDcamera (DP72; Olympus). The color of the stained cells wasextracted from the captured image using the colordeconvolution plugin of ImageJ. After the threshold was setto eliminate the background signal, the staining intensityand the staining area of the stained parts of the cells weremeasured. The staining quality was obtained by multiplyingthe staining intensity and the staining area.

Results and discussionDevice design for an even flow rate in the microchamber

A PDMS microfluidic device for PMI consists of three parts;two distribution microchannels and a reaction microchamber(Fig. 2a). The distribution microchannels are designed sym-metrically because the flow runs bidirectionally (along thepulling and pushing directions). The microchamber in whichthe flowing antibody binds to the antigen on the sample is

Fig. 1 Schematic representation of the antibody incubation process using pipetting-driven microfluidic immunohistochemistry (PMI). A PDMS de-vice is reversibly bonded with a tissue section by pressure. After loading the antibody solution into one inlet, a pipette tip is inserted into the otherinlet while the pipette button is being pressed. A pulling flow is induced by the negative pressure of pipette releasing. After a while, pipette press-ing causes the antibody to flow in the opposite direction, leading to a pushing flow. (b) Side view of the PMI device (i–i′) when a pulling (left) or apushing flow (right) occurs.

Lab on a ChipPaper

Lab Chip, 2017, 17, 702–709 | 705This journal is © The Royal Society of Chemistry 2017

designed to be in direct contact with the tissue sample. Sincethe equilibrium binding time of antigen–antibody reactiondepends on the linear flow rate in the microchannels,20 thecross-sectional flow rate in the microchamber should be uni-form so that the flowing antibody is evenly exposed to a widetissue section area. If the flow rate is different across themicrochamber, there is a possibility that a part of the tissuedoes not reach an equilibrium state and generate variationsin the staining intensity. In other words, a uniform flow ratein the microchamber can guarantee the equilibrium of theimmunoreaction and a uniform staining quality in a wide tis-sue section area. The velocity at all outlets (see point i inFig. 2a) should be the same to generate a uniform velocityacross the microchamber. We optimized the curvature,length, and width of the microfluidic distribution channelsusing commercial computational fluid dynamics software,CFD-GEOM and CFD-ACE+ (ESI group, Paris, France). As a re-sult, a parabolic velocity profile was created at the end of alldistribution channels (see Fig. S1, ESI†), which enables eachvelocity at point i to be identical. At all outlets, except for thewall region, the velocity maintained a uniform linear velocityprofile from 1 mm to 14 mm (Fig. 2b). Based on the simula-tion results, the microchamber was fabricated with dimen-sions of 14.4 mm × 15.0 mm (W × L) to cover a wide tissuesample area currently used in the field.

Actual flow rate depending on the set volume of themicropipette

Using the PMI device, the actual flow velocity was measuredindirectly in the microchamber depending on the pipette

volume (from 30 μL to 70 μL) to investigate the possibility ofvelocity control. The video of the pulling and pushing flowswas taken to indirectly measure the velocity using a stereomi-croscope (SZX16; Olympus), which is defined as the micro-chamber length divided by the time it took to pass throughthe microchamber. Fig. 3a shows a representative image ofthe pulling and pushing flows when the volume of the micro-pipette was set to 70 μL. A red dye fluid from distributionchannel 2 was spread linearly into the microchamber duringthe pulling flow induced by pipette releasing. In the pushingflow, pipette pressing causes the fluid to run linearly in theopposite direction. Repetitive pipetting also did not result inthe formation of a bubble or leakage of fluid in the micro-chamber (see Video S1, ESI†). The flow velocity of the pullingflow increased as the set volume of the pipette increased, asexpected (Fig. 3b). When the pipette button is pressed, thevolume of air equivalent to the volume of aspirated fluid isdisplaced by the moving piston, which causes a partial vac-uum inside the shaft.21 At higher set volumes, more partialFig. 2 Computational fluid dynamics simulation of the velocity profile

within the PMI device. (a) Top view of the device and the velocityprofile map. (b) The profile within the microchamber ii) 1 mm, iii) 2mm, iv) 7 mm, and v) 14 mm away from the i) 0 mm position.Throughout the whole area of the microchamber, a uniform linearvelocity profile was observed.

Fig. 3 Actual flow rate in the microchamber. (a) Time-lapse captureimage of the pulling and pushing flows using a color dye when thepipetting volume was set to 70 μL. (b) Actual flow rate in the pullingflow. The flow rate increased as the set volume increased. (c) Actualflow rate in the pushing flow. The flow rate was constant independentof the set volume. p1–p3 indicate individual persons and the data arepresented as the mean ± SD (n = 3).

Lab on a Chip Paper

706 | Lab Chip, 2017, 17, 702–709 This journal is © The Royal Society of Chemistry 2017

vacuum induces a higher pressure drop and higher velocityin the microchamber. However, in the pushing flow, thesame velocity was observed independent of the set volume ofthe pipette (Fig. 3c). One possible reason is due to the conicalgeometry of the pipette tip. The diameter of the part wherethe liquid flows out is much smaller than that of the partwhere the pressure is applied in the pipette tip. In this geom-etry, the fluidic resistance is different depending on the flowdirection.22 In the pulling flow, the resistance of the pipettetip is small, which does not affect the flow rate. However, inthe pushing flow, the resistance of the pipette tip is high,which results in a constant flow velocity independent of theset volume of the pipette. We also examined the reproducibil-ity of our device by testing the velocity with three differentpersons. As shown in Fig. 3a and b, there was a slight differ-ence in the flow velocity depending on the users in bothpulling and pushing flows. This result indicates that thestaining quality would not depend on the user since theimmunoreaction depends on the flow velocity.

Effect of pipette volume on immunostaining results

For the demonstration of PMI, the breast cancer cell sectionwas immunostained with PMI according to the set volume ofthe pipette. SKBR3 cell sections were incubated with theHER2 antibody. The volume of the antibody solution wasmore than the calculated volume (13.5 μL) of the micro-chamber because low volume loading could result in air bub-bling during the flow cycle. The pulling flow was allowed for

6 s and the pushing flow was allowed for 12 s to fully utilizeall loaded antibodies. The sample was immunostained forfive cycles of PMI at different set volumes (e.g., 30 μL, 50 μL,70 μL) of the pipette. As the set volume increased, thestaining color became stronger and the normalized stainingintensity increased (Fig. 4). Three, four, and five-fold anti-body solutions were exchanged in the microchamber by re-petitive pipetting with pipette volumes of 30 μL, 50 μL, and70 μL, respectively. The more antibody solution is exchanged,the greater is the mass transport of the antibody to the anti-gen on the surface of the sample, thereby increasing the pos-sibility of an antigen–antibody reaction. Thus, we selectedthe set volume of 70 μL as the operating volume of the PMIdevice in the following experiments.

Enhanced antigen–antibody binding in the microchamber bypipetting

After characterization of the PMI device, breast cancer cellsections were immunostained with PMI according topipetting cycles (e.g., 0 cycle, 1 cycle, 3 cycles, 5 cycles, 10 cy-cles). Two breast cancer cell sections, MCF-7 and SKBR3, werereacted with cytokeratin and HER2 antibodies, respectively.One cycle of pipetting took about 18 s and the antibody wasincubated for 1 h in a conventional batch process. As shown

Fig. 4 Immunostaining results of the breast cancer cell (SKBR3)sections with the HER2 antibody depending on the set volume of thepipette in the five cycles of pipetting. (a–c) SKBR3 cells stained with theHER2 antibody using PMI with a pipette volume of (a) 30 μL, (b) 50 μL,and (c) 70 μL. (d) Normalized staining intensities of SKBR3 dependingon the set volume of the pipette. The staining intensity increased asthe set volume increased. The data are presented as the mean ± SD (n= 3). Scale bars = 50 μm.

Fig. 5 Immunostaining results of the breast cancer cell sections usingPMI with a pipette volume of 70 μL. One cycle of pipetting took about18 s. Batch indicates that the antibody was incubated for 1 h in thebatch process. (a–c) Stained image of MCF-7 cells with the cytokeratin(Cyto) antibody. (d–f) Stained images of SKBR3 cells with the HER2antibody. (g) Normalized staining intensities of the MCF-7 and SKBR3cells. The staining intensity using PMI seems to be saturated from thefive cycles, of which the signal is comparable with the result of theconventional batch process. The data are presented as the mean ± SD(n = 3). Scale bars = 50 μm.

Lab on a ChipPaper

Lab Chip, 2017, 17, 702–709 | 707This journal is © The Royal Society of Chemistry 2017

in Fig. 5a–f, only five cycles of pipetting show the samestaining quality as the batch process for 1 h. To compare withthe batch process quantitatively, the stained sample was ana-lyzed using the ImageJ program. More stained samples meanmore staining intensity and area. The staining quality was cal-culated by multiplying the staining intensity and the stainingarea and normalized to the result of the batch process(Fig. 5g). The zero cycle means that there is no flow in themicrochamber, and the antibody has just been incubated inthe microchamber for 18 s, which corresponds to the time forthe zero cycle of pipetting. As a result of the zero cycle andone cycle of pipetting, the flow did not significantly affect thestaining quality. Three cycles of pipetting resulted in about80% staining quality compared to the batch process in bothcells. However, five cycles of pipetting showed a similar qual-ity compared to the batch process, and an approximately 10%increase was observed in ten cycles of pipetting compared tothe batch process. This is due to the increased mass transportby flow, which is induced by repetitive pipetting. Withoutflow, the antigen–antibody reaction is limited by the masstransport of the antibody to the surface region, representing adiffusion limited reaction. In other words, there is a depletionzone between the bulk solution and the surface, and moretime is required for the reaction. Under the flow situation,however, fresh antibodies are continuously replenished onthe reaction surface and the reaction rate is increased.

Theoretically, increasing the flow rate can increase masstransport, but a waste of antibody is inevitable because onlya small fraction of the total mass can bind to the surface.Thus, too many antibodies are wasted at higher flow rates,limiting the operation velocity. On this account, the flow rateof our previous microfluidic IHC was limited to below 1 μLs−1 and the time required for the saturation of theimmunoreaction was more than 10 min.2,17 Meanwhile,Dupouy et al. reported that the antibody was flowed at a highflow rate of 10 μL s−1 for 12 s in their microfluidic tissue pro-cessor and incubated for 2 min at a low flow rate of 20 nL s−1

to minimize the waste of the antibody.23 As a result, about120 μL was consumed for 2 min. However, the PMI deviceconsumes only 70 μL of antibody under 2 min by a simplemethod for inducing the flow of antibodies. The fluid in thePMI device could be operated at a relatively high flow rate.

Furthermore, since the bidirectional flows are induced byrepetitive manual pipetting, the high flow rate is maintainedcontinuously, and an unbound antibody flowing through thechamber can be available in the reverse flow. In other words,the limited amount of antibody solution can be used effec-tively without waste using bidirectional flows. However, theresults showed relatively high errors at some points. Thismay be due to the long and multiple steps of the samplepretreatment (e.g., deparaffinization and antigen retrieval)prior to staining. Although the high errors are not good forquantitative assay, this system can be useful for qualitativeassay. Currently, most IHC assays in clinical research areperformed for qualitative analysis such as negative/positiveanalysis of specific biomarkers and observation of morpho-

logical characteristics. The result that more than five cyclesof pipetting can immunostain samples with comparablestaining quality in the batch process for 1 h can verify its usefor qualitative analysis for IHC.

Immunostaining with breast cancer tissue sections

After demonstrating PMI in breast cancer cell sections, breastcancer tissue sections were immunostained using PMI. A

Fig. 6 Immunostaining results of real breast cancer tissue sectionsusing the PMI with five cycles of pipetting and the batch process for 1h. Two adjacent tissue sections were immunostained. (a) A dark fieldimage of the tissue section under the PMI device. (b) Stained images ofdifferent spots on the tissue sample with cytokeratin (Cyto) antibodies.(c) Stained images of different spots on the tissue sample with HER2antibodies. (d) Normalized staining intensities of the breast cancertissues. Based on the staining intensity measured on the same spot oftwo adjacent sections, the 5 cycle-based PMI is comparable to theconventional batch process. Scale bars = 400 μm.

Lab on a Chip Paper

708 | Lab Chip, 2017, 17, 702–709 This journal is © The Royal Society of Chemistry 2017

wide tissue section can be covered and aligned with the de-vice as expected (Fig. 6a). To compare the results with thebatch process, two adjacent tissue sections were immuno-stained with the cytokeratin and HER2 antibodies. One sec-tion was immunostained by the PMI device for five cycles ofpipetting and the other section was immunostained by theconventional batch process for 1 h. Three different adjacentspots on a wide tissue section were compared. There was lit-tle difference between the sample stained with PMI and thesample stained with the conventional batch process for bothcytokeratin and HER2 antibodies (Fig. 6b and c). Tissue de-tachment was not observed in the PMI device, indicating thatthe relatively high shear stress due to higher flow rates didnot affect the sample. In the ImageJ analysis results, the mea-sured intensity of the PMI was normalized to the measuredintensity of the batch process (Fig. 6d). Almost the samestaining quality was observed for each spot between the PMIand the batch process. The PMI with cytokeratin showed anincreased intensity by about 2% compared to the batch pro-cess. However, the PMI with HER2 showed a decreased inten-sity compared to the batch process. Since the variance is verylow (2%), this may not affect the qualitative detection of thedesired biomarkers. From these results, we were able to con-firm the applicability of PMI to real tissue samples.

Conclusions

In this study, we demonstrated a simple microfluidic deviceoperated by manual pipetting for IHC. The pressure pro-duced by the movement of the pipette piston was used to op-erate a microfluidic system as compared to a conventionalmicrofluidic system operated by a pressure pump or a syringepump. The device was designed to allow the fluid to run at auniform flow rate, ensuring a uniform staining quality of awide tissue section. With just five cycles of pipetting usingPMI, we were able to immunostain the breast cancer samplesto an extent comparable to the 1 h-based conventional batchprocess. It only took about 1 min and 30 s for five cycles ofpipetting in the PMI device, which can reduce the processtime by 2.5% compared to the 1 h-based conventional batchprocess. In addition, the amount of consumed antibody solu-tion was 70 μL, which was much less than the amount re-quired for the conventional process. These better characteris-tics than those of the conventional method are derived fromthe micropipette introduction for operating a microfluidic de-vice. Bidirectional flows induced by manual pipetting facili-tate not only enhanced immunoreaction with a high flow ratebut also the efficient use of a small, fixed amount of antibodysolution. These bidirectional flows can be applied to othersystems using microfluidic immunoassays to enhanceimmunoreaction and reduce the reagents required. In addi-tion, the performance of the existing microfluidic IHC can besignificantly improved for the quantitative analysis of proteinexpression by inducing bidirectional flows since the flowscan be controlled automatically and accurately using the pro-grammable pump system. In conclusion, we expect that this

simple and easy-to-use pipetting-based method will enablethe distribution of microfluidic IHC, which can reduce theamount of antibody, time and cost in practical applications.

Acknowledgements

This research is supported by the Mid-Career Researcher Pro-gram (NRF-2016R1A2B3015986) and the Bio & Medical Tech-nology Development Program (NRF-2015M3A9B3028685)through the National Research Foundation of Korea fundedby the Ministry of Education, Science and Technology. Theauthors also acknowledge the KAIST Systems Healthcare Pro-gram and Dr. Eun Sook Lee of the National Cancer Center,Korea.

Notes and references

1 P. Brandtzaeg, J. Immunol. Methods, 1998, 216, 49–67.2 S. Kwon, M. S. Kim, E. S. Lee, J. S. Sohn and J.-K. Park,

Integr. Biol., 2014, 6, 430–437.3 E. K. O. Ng, R. Li, V. Y. Shin, H. C. Jin, C. P. H. Leung,

E. S. K. Ma, R. Pang, D. Chua, K.-M. Chu, W. L. Law, S. Y. K.Law, R. T. P. Poon and A. Kwong, PLoS One, 2013, 8, e53141.

4 J. Dalmau, H. M. Furneaux, R. J. Gralla, M. G. Kris and J. B.Posner, Ann. Neurol., 1990, 27, 544–552.

5 R. Hassan, A. T. Remaley, M. L. Sampson, J. Zhang, D. D.Cox, J. Pingpank, R. Alexander, M. Willingham, I. Pastanand M. Onda, Clin. Cancer Res., 2006, 12, 447–453.

6 O. J. Semmes, Z. Feng, B.-L. Adam, L. L. Banez, W. L. Bigbee,D. Campos, L. H. Cazares, D. W. Chan, W. E. Grizzle, E.Izbicka, J. Kagan, G. Malik, D. McLerran, J. W. Moul, A.Partin, P. Prasanna, J. Rosenzweig, L. J. Sokoll, S. Srivastava,S. Srivastava, I. Thompson, M. J. Welsh, N. White, M.Winget, Y. Yasui, Z. Zhang and L. Zhu, Clin. Chem., 2005, 51,102–112.

7 M. Cregger, A. J. Berger and D. L. Rimm, Arch. Pathol. Lab.Med., 2006, 130, 1026–1030.

8 T. O. Nielsen, F. D. Hsu, K. Jensen, M. Cheang, G. Karaca, Z.Hu, T. Hernandez-Boussard, C. Livasy, D. Cowan, L. Dressler,L. A. Akslen, J. Ragaz, A. M. Gown, C. B. Gilks, M. van de Rijnand C. M. Perou, Clin. Cancer Res., 2004, 10, 5367–5374.

9 W. G. Cance, J. E. Harris, M. V. Iacocca, E. Roche, X. Yang, J.Chang, S. Simkins and L. Xu, Clin. Cancer Res., 2000, 6,2417–2423.

10 P. S. Loo, S. C. Thomas, M. C. Nicolson, M. N. Fyfe andK. M. Kerr, J. Thorac. Oncol., 2010, 5, 442–447.

11 W. G. Morice and J. A. Ferreiro, Hum. Pathol., 1998, 29,609–612.

12 M. Srodon and W. H. Westra, Hum. Pathol., 2002, 33,642–645.

13 M. Stenberg and H. Nygren, J. Immunol. Methods, 1988, 113,3–15.

14 J. Li, P. Zrazhevskiy and X. Gao, Small, 2016, 12, 1035–1043.15 L. Ying and Q. Wang, BMC Biotechnol., 2013, 13, 76.16 A. H. C. Ng, U. Uddayasankar and A. R. Wheeler, Anal.

Bioanal. Chem., 2010, 397, 991–1007.

Lab on a ChipPaper

Lab Chip, 2017, 17, 702–709 | 709This journal is © The Royal Society of Chemistry 2017

17 M. S. Kim, T. Kim, S.-Y. Kong, S. Kwon, C. Y. Bae, J. Choi,C. H. Kim, E. S. Lee and J.-K. Park, PLoS One, 2010, 5,e10441.

18 A. T. Ciftlik, H.-A. Lehr and M. A. M. Gijs, Proc. Natl. Acad.Sci. U. S. A., 2013, 110, 5363–5368.

19 T. Gervais and K. F. Jensen, Chem. Eng. Sci., 2006, 61,1102–1121.

20 G. Hu, Y. Gao and D. Li, Biosens. Bioelectron., 2007, 22,1403–1409.

21 R. Feldmann and K. H. Lochner, Accredit. Qual. Assur.,2016, 21, 69–82.

22 T. Gerlach, Sens. Actuators, A, 1998, 69, 181–191.23 D. G. Dupouy, A. T. Ciftlik, M. Fiche, D. Heintze, B. Bisig, L.

de Leval and M. A. M. Gijs, Sci. Rep., 2016, 6, 20277.

Lab on a Chip Paper