rsc ib c3ib40224j 3.nanobio.kaist.ac.kr/papers/c3ib40224j_author_reprint.pdfa quantum dot-based...

430 | Integr. Biol., 2014, 6, 430--437 This journal is©The Royal Society of Chemistry 2014

Cite this: Integr. Biol., 2014,6, 430

A quantum dot-based microfluidic multi-windowplatform for quantifying the biomarkers of breastcancer cells†

Seyong Kwon,a Minseok S. Kim,b Eun Sook Lee,c Jang Sihn Sohnd andJe-Kyun Park*a

Conventional molecular profiling methods using immunochemical assays have limits in terms of multiplexity

and the quantification of biomarkers in investigation of cancer cells. In this paper, we demonstrate a

quantum dot (QD)-based microfluidic multiple biomarker quantification (QD-MMBQ) method that enables

labeling of more than eight proteins immunochemically on cell blocks within 1 h, in a quantitative manner.

An internal reference, b-actin, was used as a loading control to compensate for differences in not only the

cell number but also in staining quality among specimens. Furthermore, the microfluidic blocking method

exhibited less nonspecific binding of QDs than the conventional static blocking method.

Insight, innovation, integrationImmunochemical assay using an antibody-based molecular detection technology can provide information on both cellular morphology and the quantities ofmolecules within cells (immunocytochemistry) or tissues (immunohistochemistry). In this field, multiplexed protein quantification remains difficult usingconventional methods. Here, we demonstrate a new analytical concept that integrates a microfluidic multiplexing platform and a QD double-staining method.The microfluidic double-staining method enabled accurate quantification by normalization of biomarker levels to that of b-actin as an internal reference. Thisnovel molecular profiling method will accelerate cancer cell studies and the development of diagnostic tools for personalized medicine.

Introduction

The development of an accurate, quantitative molecular profilingtechnology for cellular materials remains a challenge for investiga-tion of cell physiology, drug responses, and fatal diseases such asvarious types of cancer.1 In particular, a molecular profiling-basedpersonalized cancer diagnosis system is desirable because a con-siderable number of cancer patients do not benefit from existingtherapeutics, including molecular-targeted drugs.2–5 Indeed, breastcancer has a high recurrence rate due to its heterogeneity.6,7 Thus,there is a need for tailored treatment strategies for breast cancer.

Understanding the complex molecular mechanisms of individualcancer cells requires molecular detection of multiple proteins atthe single-cell level. Among the existing molecular detectiontechnologies (e.g., polymerase-chain reaction, western blotting,mass spectrometry, and enzyme-linked immunosorbent assay),only immunocytochemistry (ICC) or immunohistochemistry(IHC) can provide information on both morphology and thequantity of molecules, which is not possible with other techniquesthat involve destruction of cellular structures.

Quantum dot (QD) nanoparticles enable multiplexed proteindetection in immunochemical assays due to their superioroptical properties, such as brightness, narrow emission peak,photobleaching resistance, large Stokes shift, and simultaneousexcitation of multiple fluorescence colors.8–11 In recent reports,however, the number of proteins that can be analyzed usingmultiplexed QD-immunochemical assay techniques has beenlimited to four or five.12–14 To increase the number of proteinsin QD multiplexing methods, several approaches have beenattempted; to date, all were unsuccessful or inefficient. Forexample, for the multiplexed use of primary antibodies (Abs),QDs were directly conjugated to primary Abs. This makes themultiplexed assay process easier and faster. Nevertheless, this

a Department of Bio and Brain Engineering, Korea Advanced Institute of Science and

Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon 305-701,

Republic of Korea. E-mail: [email protected]; Tel: +82-42-350-4315b Samsung Advanced Institute of Technology, 97 Samsung 2-ro, Giheung-gu,

Yongin-si, Gyeonggi-do 446-712, Republic of Koreac Research Institute and Hospital, National Cancer Center, 323 Ilsan-ro,

Ilsandong-gu, Goyang-si, Gyeonggi-do 410-769, Republic of Koread Department of Pathology, College of Medicine, Konyang University,

158 Gwanjeodong-ro, Seo-gu, Daejeon 302-718, Republic of Korea

† Electronic supplementary information (ESI) available: Fig. S1 and S2. See DOI:10.1039/c3ib40224j

Received 27th October 2013,Accepted 11th February 2014

DOI: 10.1039/c3ib40224j

www.rsc.org/ibiology

Integrative Biology

PAPER

This journal is©The Royal Society of Chemistry 2014 Integr. Biol., 2014, 6, 430--437 | 431

method has disadvantages in its preparation process. Due tothe harsh conditions of Ab conjugation to QDs, it consumes aconsiderable amount of Abs and time that is not efficient forthe use of large numbers of biomarkers. A sequential QDmultiplexing method is more flexible to overcome these problemsand could be used to identify a number of proteins by labeling withdifferent Abs from the same species. However, this method requiresmultiple blocking steps between sequential reactions. Due to theprocessing time required per reaction, this also requires more than3 h using conventional methods; thus labeling large numbers ofproteins is laborious and time-consuming. Moreover, recentmultiplexed QD-based IHC studies showed that the sequentialQD multiplexing method caused distorted results.15–17

Previously, we reported a microfluidic immunochemicalassay platform that enabled simultaneous detection of multipleproteins by dividing a cell block or tissue section into severalisolated areas by means of reversible bonding of a PDMSmicrofluidic layer with multiple parallel channels on the cellor tissue slide directly.18,19 A multiple-channel layout can beused with simultaneous incubation of multiple Abs on thesample slide. Although this method can be used to increasethe maximum multiplexing number in immunochemicalassays, protein quantification was difficult due to the lack ofa robust quantification method. In this report, we developed anaccurate QD-based microfluidic multiple biomarker quantification(QD-MMBQ) method that facilitated determination of the levels ofproteins of interest at the single-cell level.

For the accurate quantitative analysis of several proteins inthe same biological specimen, we exploited the normalizationconcept using a microfluidic immunochemical assay platformby co-labeling the loading control (LC) protein with a differenttarget protein in each microchannel. Expression levels of eightproteins in various cell lines were compared by co-labeling adifferent target protein and the same LC in each microchannel.Our microfluidic method represents a flexible and rapid multiplexedQD-immunochemical assay technique, facilitating analysis ofthe data and more accurate quantification of biomarkers.

Results and discussionCrosstalk of QD-conjugated secondary Abs

As noted above, crosstalk of QD-conjugated secondary Absis regarded as a key technical problem in QD-multiplexedimmunostaining. Prior to the development of microchannelQD-multiplexing methods, validation of several QD-multiplexingtechniques was performed to avoid nonspecific binding of QDparticles. A formalin-fixed paraffin-embedded cell block wasused to address QD multiplexing problems systematically. Thebreast cancer cell line SK-BR-3 strongly expresses HER2 in themembrane and Ki-67 in the nucleus. To investigate the stainingtendency of a sequential method, a single primary Ab species anddifferent-color-QD-conjugated secondary Abs (QD525-conjugatedanti-rabbit Ab and QD605-conjugated anti-rabbit Ab) were used tosequentially label HER2 (rabbit) and Ki-67 (rabbit). Primary Absfrom different species and different-color-QD-conjugated secondary

Abs (QD525-conjugated anti-mouse Ab and QD605-conjugatedanti-rabbit Ab) were also used to label HER2 (mouse) and Ki-67(rabbit) using a cocktail method (a mixture of primary Absproduced by different species), which is widely used forQD-based detection of two or three biomarkers. Control datawere obtained by QD staining for each protein separately. Typicalsequential and cocktail staining results are shown in Fig. 1.

Notably, the sequential QD staining results showed that thelast-in-sequence QD-conjugated secondary Abs bound not onlyto the final primary Abs but also to previous primary Absdespite the blocking process that is supposed to prevent suchunintended binding. This was demonstrated by changingthe multiplexing order and observing the localization of theQD-conjugated secondary Abs (Fig. 1). Increasing the incubationtime of the QD-conjugated secondary Abs so that the bindingsites of the primary Abs were fully occupied for 2 h, had no effecton this phenomenon (see ESI,† Fig. S1). This problem is criticalbecause unintended QD labeling of primary Abs will result inboth IHC and ICC being compromised by inaccurate proteinlocalization information and incorrect measurements of proteinquantities (Fig. 1E and G). This hinders the study of cell andprotein functions, such as protein translocalization, using IHCor ICC. Using a cocktail method, multiple QD conjugates werelabelled on the desired proteins (Fig. 1J and K). HER2 proteins(cell membrane) were labelled specifically with QD525, and Ki-67proteins (nucleus) with QD605. The same staining tendencieswere also observed with microfluidic staining (data not shown).

Fig. 1 Staining of HER2 with QD525, and Ki-67 with QD605 on SK-BR-3cell-block sections. (A) HER2 was labeled with a rabbit anti-HER2 Ab andQD525-conjugated goat anti-rabbit Ab. (B) Ki-67 was labeled with a rabbitanti-Ki-67 Ab and QD605-conjugated goat anti-rabbit Ab. (C–E) HER2 andKi-67 were stained in the order, rabbit anti-HER2, QD525-conjugated goatanti-rabbit, rabbit anti-Ki-67, and QD605-conjugated goat anti-rabbit Abs.The fluorescence image obtained using a long-pass fluorescence filter(C) to receive the total QD525 and QD605, and the individual QD525(D) and QD605 (E) fluorescence signals; the latter were obtained using aband-pass fluorescence filter. (F–H) HER2 and Ki-67 were stained in theorder, rabbit anti-Ki-67, QD605-conjugated goat anti-rabbit, rabbit anti-HER2, and QD525-conjugated goat anti-rabbit Abs. The total fluorescenceimage obtained using a long-pass fluorescence filter (F), and individualQD525 (G) and QD605 (H) fluorescence signals. (I–K) HER2 and Ki-67were labeled using a cocktail method with a mixture of mouse anti-HER2and rabbit anti-Ki-67 Abs, and a mixture of QD525-conjugated goat anti-mouse and QD605-conjugated goat anti-rabbit Abs. The total fluores-cence image obtained using a long-pass fluorescence filter (I), andindividual QD525 (J) and QD605 (K) fluorescence signals.

Paper Integrative Biology


Although the cocktail method has benefits in QD multiplexing,such as high selectivity due to use of primary Abs that originatefrom different species, the QD multiplexing number is limitedto three or four due to the number of commercially availableAb-producing species (mouse, rabbit, rat, goat, etc.).

Microfluidic biomarker screening

For the accurate quantitative analysis of large numbers ofproteins in the same biological specimen, we exploited thenormalization concept using a microfluidic immunochemicalassay platform by co-labeling the LC with a different targetprotein in each microchannel (Fig. 2). Since the cocktailmethod is useful for multiplexing of small numbers of proteinswith QDs, such as by reducing the total process time andinvolving high selectivity, we applied a cocktail method todevelop our QD-MMBQ technique. Blocking is important inQD-based immunochemical assay to control nonspecific bindingof QDs. To investigate the efficiency of the blocking process inassociation with microfluidics, two cell-block sections weretreated with blocking solution; one was incubated statically(conventional method), while the other was incubated with flow.After blocking for 30 min, the sections were incubated withQD605–secondary Ab conjugates for 1 h. The flow-based blockingsection showed a significant decrease in the nonspecific bindingof QDs at a flow rate of 6 mL h�1 (Fig. 3A). Furthermore, the dataindicate that the conventional static incubation method cannot

block completely the nonspecific binding of QD–secondaryAb conjugates. A longer incubation period had little effect onblocking. In contrast, the flow-based blocking process wasefficient and could be used to control the blocking time by varyingthe flow rate (Fig. 3B). This is due to the effective deposition ofBSA on the section surface by the continuous flow within themicrochannels. The brightness of a QD can be regarded as aweakness in any immunochemical assay if nonspecific bindingcannot be controlled. Therefore, efficient, controllable blockingis an advantage of microfluidic immunochemical assay overconventional immunochemical assay techniques.

The left panel in Fig. 4A shows our immunostaining systemplatform. The sample slide was covered with a microfluidicdevice and specialized metal frames were used to hold thedevice in the desired place on the sample. The center panel ofFig. 4A shows a dark-field image of microfluidic channelsplaced and held on the sample. After the flow-based blockingprocess, the selected rabbit Abs were injected into each inletmixed with mouse b-actin, which is used as a LC. b-actin hasbeen known as a stable LC for breast cancer cells.20 Afterincubation of the Ab mixtures under continuous flow conditions,unbound Abs were washed away with Tris-buffered saline containingTween (TBS-T). Then, a mixture of QD605-conjugated anti-rabbit Aband QD525-conjugated anti-mouse Ab was injected into each inlet,followed by a further incubation to finish the QD labeling. Allsamples were mounted with QD mounting medium (Invitrogen).Due to the microfluidic immunostaining, the shape of the channelwas visible on the sample (Fig. 4A, right panel). Consequently, redQD-labeling of the eight proteins can be seen in the form of stripeson the cell-block surface: no non-specific binding was observed(Fig. 4B and C). Note that more proteins can be multiplexed as the

Fig. 2 Schematic of the QD-based microfluidic multiple biomarkerquantification (QD-MMBQ) method. In step 1, the reference marker Aband each biomarker Ab, originating from different species, were reactedwith the cell-block section within the microchannel. Here, the referencemarker stained the cytoplasm, while each biomarker stained the nuclearregion. In step 2, two types of QDs conjugated with the various secondaryAbs (goat anti-rabbit Ab and goat anti-mouse Ab) were reacted with thecell-block section on the microchannel. Inset a and b in each step indicatethe area of the circle in each microchannel.

Fig. 3 Result of QD-secondary Ab incubation with blocking solutionunder flow and static conditions. (A) MCF-7-cell-block sections wereincubated with blocking solution under microfluidic flow for 30 min at aflow rate of 6 mL h�1 (left), or static incubation for 30 (center) or 60 (right)min. After the blocking process, the QD605-conjugated anti-rabbit goat Abwas added and incubated for 30 or 60 min. (B) Results of QD-secondary Abincubation after the flow-based blocking at various flow rates.

Integrative Biology Paper


number of microchannels increases.19 Moreover, this methodrequires only 1 h for incubation of nine Abs (including areference marker). A recent report indicated that 46 h isrequired to label four proteins with QDs.14 Since our methodtakes less time than conventional QD-based IHC, it reduces thetotal process time markedly. This is due to the parallel micro-fluidic multiplexing and efficient protein–Ab reactions in multi-ple microchannels resulting from the enhanced mass transferof Abs to the surface proteins of the biospecimen.18

Recently, Ciftlik et al. reported that Ab incubation and high-Reynolds-number flow in a microchannel can reduce Ab incubationtimes to 5 min.21 Because the QD-MMBQ method can be usedreadily in microfluidics-based immunochemical assay, it is possibleto reduce the total processing time further than those describedhere by changing the material of the microchannel or the devicegeometry. Additionally, to enhance the multiplexity of the method,repeated Ab staining is possible, using, for example, the Abde-staining method reported by Zrazhevskiy et al.22

Protein quantification using multi-windows and LC

High flexibility is one of the strongpoints of the QD-MMBQmethod. As previously stated, a cocktail method, which doesnot allow unintended reactions of QD–Ab conjugates, facilitatesexploitation of Abs produced by different species to enhancemultiplexing efficiency and avoid inappropriate QD binding.Nevertheless, more Ab producing species are needed to allowlabeling of more proteins. Until now, since few animal specieshave been developed to produce Abs, the use of existing QD-IHCmethods in pathology is problematic. In contrast, our QD-MMBQmethod requires only two Abs from different animal species.All of the anti-protein-target Abs can be produced from thesame species; only the LC Ab needs to be come from a distinctspecies. Accordingly, only two QD-conjugated reagents arerequired. Therefore, our method ensures high flexibility of

antibody selection for multiplexed assay development. In addition,because the emission peaks of the QDs were sufficiently separated,no spectral deconvolution was necessary for plotting of the proteinexpression profiles, which was important in previous quantitativeQD-immunochemical assay studies.12,13 Furthermore, our methodallows use of QDs of only two colors, and no adjustment offluorescence intensities is required. These characteristics enablemore accurate quantitative comparisons of multiple biomarkers.Red or far-red QDs are brighter than other QDs.23 The QD-MMBQmethod labels all target biomarkers with the same QDs such as redQDs. Thus, our method is easier, more accurate, and moresensitive than previous multiplexed QD detection techniques.

Note that although a cell-block section comprises homogeneouscells and it is regarded as a uniformly dispersed cell layer, the signalof a particular QD-labelled protein from randomly selected areascan differ considerably before LC normalization. This is due tovariation in cell numbers and cell volume difference in a confinedarea (Fig. 5). The variation in cell number and volume in a cell-blocksection may be limited, but a significant signal difference arises dueto the high sensitivity of QDs. However, QD signal variation wasreduced by normalization of protein expression to that of theLC at the same location. Thus the immunochemical normalizingapproach can compensate for the cell number and volume variation,guaranteeing a more accurate single-cell level quantification ofprotein expression in ICC (Fig. 4B). Specifically, a microfluidicimmunochemical assay uses the microchannels as individualAb-incubation chambers, and each protein is stained for in onlya small region of the cell-block section. Therefore, microfluidicimmunochemical assays would result in quantification errors ifprotein expression was measured without a LC in each channeldue to the small number of sampling locations. Therefore, LC

Fig. 4 The fluorescence image of an MCF-7-cell-block section,multiplex-stained using the QD-MMBQ system. (A) The microfluidic deviceis integrated with the cell-block section using an integration system (left).Multi-microchannels are placed on the cell-block section (center). Afterthe immunoreactions, the multi-microchannels are detached from thecell-block section (right). (B) The cell-block section was labeled with eightAbs using a microfluidic immunostaining system. (C) Magnified images ofAb-labeled areas. Each image indicates the upper biomarker area in (B).Both the QD525 and QD605 signals are shown in the upper images andthose of each biomarker with QD605 in the lower images.

Fig. 5 Fluorescence intensity of QD-labeled mTOR and b-actin. TheSK-BR-2 cell-block section was stained with anti-mTOR Ab-QD605 andanti-b-actin Ab-QD525. (A) Fluorescence QD-stained images of randomlyselected areas of a cell-block section. To emphasize the effect of LCnormalization, sampling was performed from locations with different celldensities. Scale bar, 50 mm. (B) Fluorescence intensities of QD525 andQD605 in randomly selected areas of cell-block sections. mTOR*, themTOR signal normalized to that of b-actin.



normalization ensures the accuracy of protein quantification bymicrofluidic immunochemical assays.

Comparison of protein expression levels between cell lines

Furthermore, we were able to quantitate and compare theexpression of various proteins in not only one cell-block sectionbut also among cell-block sections. Eight proteins in SK-BR-3,MCF-7, and HCC-70 cells were compared quantitatively. Asshown in Fig. 6, we compared the expression of the eightproteins with and without normalization. Importantly, withoutnormalization, the ER expression of SK-BR-3 and MCF-7 cellswas similar, although SK-BR-3 cells are ER-negative and MCF-7cells ER-positive.24 In comparison, the normalized ER signal ofSK-BR-3 and MCF-7 cells showed a considerable difference,being identified as negative and positive, respectively (Fig. 6B).Also, SK-BR-3 cells are PR-negative and MCF-7 cells are PR-positive.25

However, the PR expression levels of these cell lines becomedistinct after normalization. Thus we confirmed that normalizedprotein expression signals are more credible than non-normalizedsignals comparing previous proteomic studies of breast cancercell lines.24,25 This is due to the overall elevation of proteinexpression in MCF-7 cells with normalization. Note that there issignificant variation in the protein staining quality betweentissues or cell sections with IHC or ICC due to the complexsteps involved in sample preparation and staining steps (e.g., gelloading, fixation, paraffin embedding, microtome sectioning,deparaffinization, re-hydration, antigen retrieval process, etc.).26–31

Likewise, proteins in the MCF-7 cell-block section were generallystained more lightly than those in other sections; comparison ofprotein expression among different sections accurately is difficultwithout a calibration process. This is a critical issue in all kinds ofimmunochemical assays because conventional immunochemicalassay is associated with technical difficulties in calibratingdifferences in staining among specimens. The b-actin-basednormalization strategy compensates for differences in not only

cell number but also in staining quality among specimenswhich is not possible using conventional methods.

Multiple biomarker quantification in tissues

Due to the use of micro-sized channels, a limited biomarkerstaining area can be problematic considering the unexpectedlydistributed cancer cells in tissues. For this reason, determinationof the region where tumor cells are populated in high densityis critical to achieve successive biomarker screening using theQD-MMBQ method. To address this issue, we mounted a bundleof multiple channels over a densely populated cancer cell areawhich was determined by hematoxylin and eosin staining.

Fig. 7A shows a staining result of eight biomarkers on abreast cancer tissue. Like the cell specimen staining using ourmethod, biomarkers were labeled with QD605 and b-actin wasattached with QD525. QDs were stained correctly at eachlocation as shown in right images of Fig. 7A. Using our b-actin-based normalization strategy, eight biomarkers were quantifiedusing spectroscopy (Fig. 7B).

Although we demonstrated our method using clinical samples,tissue heterogeneity could be a problem using our biomarkermultiplexing system, which uses tissue dimensions of about400 mm � 5000 mm for each biomarker, separately. This wouldnot be adequate for the cases such as rare cell discovery usingIHC.17 Our system could be effective for use in multiplexedbiomarker screening of less heterogeneous cancer tissues. Wepreviously demonstrated that for cases in which breast cancer

Fig. 6 Protein expression by MCF-7, HCC-70, and SK-BR-3 cells. (A) Eighttarget proteins and b-actin were labeled with QDs on MCF-7, HCC-70,and SK-BR-3 cell blocks. QD605 was used to visualize the expression ofeach protein, while the QD525 intensity represented the reference markersignal. The lower images show the gray signal of biomarker expression(QD605). (B) Fluorescence intensity is expressed as arbitrary units (a.u.) ofthe QD605 signal (left panel) and the QD605 divided by QD525 signal(right panel) intensities for each strip (generated by the microchannels). Inthis way, the expression levels of the eight proteins were quantitated at thesingle-cell level (right panel).

Fig. 7 Quantitative analysis of eight proteins on a breast cancer tissuesection. (A) Eight target proteins and b-actin were labeled with QDs on abreast cancer tissue. QD605 was used to visualize the expression of eachprotein, while the QD525 intensity represented the reference markersignal. (B) The fluorescence intensity in arbitrary units (a.u.) of each areawas compared quantitatively using the b-actin signal. (C) The values ofcoefficient of variation (CV) were compared between the b-actin normalizedsignal and the original signal of QD605. The b-actin normalized signal tendsto be less variable than that was not normalized.



tissues were analyzed, the parallel biomarker multiplexingapproach showed strong consensus with conventional whole-section analysis.18 Also, because our system can be readily appliedto various clinical specimens, assay accuracy can be improved bycombining our system with a tissue microarray consisting of smallreplicate specimens. On the other hand, the use of an appropriateLC for the particular cancer can be a better solution to improve ourquantification system in tissues.

Notably, even when a tissue consisted of heterogeneouscells, normalized biomarker intensity showed less variance overthe signal acquisition areas than biomarker intensity whichwas not normalized (Fig. 7C). This shows that the b-actin-basednormalization concept has a possibility to be used for theaccurate biomarker quantification of tissue samples.

Materials and methodsPreparation of specimens

A formalin-fixed paraffin-embedded breast cancer cell blockwas used for the demonstration of our protein quantificationsystem. Breast cell lines, including MCF-7, SK-BR-3, and HCC-70 cells, were purchased from the American Type CultureCollection (ATCC; Manassas, VA). MCF-7 and HCC-70 cellswere cultured in RPMI-1640. SK-BR-3 cells were cultured inDulbecco’s modified Eagle’s medium (DMEM) supplementedwith 10% fetal bovine serum (FBS), 100 IU mL�1 penicillin, and100 mg mL�1 streptomycin. All cells were maintained at 37 1Cand 5% CO2. For the fabrication of engineered tissues, aftertrypsinization, harvested cells were centrifuged. Formalin fixation,agar suspension, and paraffin embedding was accomplished tomake cell blocks. Paraffin embedded blocks were then sectionedat 4 mm, the sections were mounted and baked onto positivelycharged glass slides. These samples were dried for 1 h at roomtemperature followed by 1 h in an incubator at 60 1C.

Human breast cancer tissue samples were obtained fromthe Konyang University Hospital (Daejeon, Korea), with thecorresponding written consents given by the patients or theirrelatives. This study was approved by the Institutional ReviewBoard (IRB) at the Konyang University Hospital and the KoreaAdvanced Institute of Science and Technology (KAIST). Humantissue samples were fixed in 4% neutral-buffered formalin,Bouin’s fixative, acetic formalin alcohol (AFA), or 4% or 10%unbuffered formalin; 4 h in PreFer (Anatech, Battle Creek, MI)or Pen-Fix (Richard Allen Scientific; Kalamazoo, MI); or 48 h in4% neutral-buffered formalin. After the paraffin embeddingprocess, tumor specimens were sectioned into 4 mm and driedfor 1 h at room temperature, followed by 1 h in a convectionincubator at 60 1C.

Immunocytochemical staining

For the investigation of the multiplexed staining tendency ofthe sequential method, rabbit anti-human epidermal growthfactor receptor 2 (HER2) antibody (Ab) (1 : 500, Dako, Denmark),rabbit anti-Ki-67 Ab (1 : 200, Novus Biologicals), and QD605-conjugated goat anti-rabbit Ab (1 : 250, Invitrogen, Carlsbad, CA)

and QD525-conjugated goat anti-rabbit Ab (1 : 100, Invitrogen) wereused in sequential multiplexing. Cell blocks were deparaffinized inxylene and hydrated through a graded series of ethanol (70%, 80%,95%, and 100% ethanol). The microwave antigen-retrievaltechnique was conducted in target retrieval solution, pH 9(Dako), for 20 min at 750 W. The slide surface was blockedusing blocking solution (2% bovine serum albumin with 5%goat serum in 1� phosphate buffered saline) or commerciallyavailable blocking solution (Zymed, San Francisco, CA) for30 min at room temperature, the slides were incubated withprimary Ab at room temperature for 30 min or 1 h. Afterwashing with 1� Tris buffered saline-Tween (TBS-T; 0.1%Tween 20), QD-conjugated secondary Ab was applied to theslides at room temperature for 30 min or 2 h.

Using this protocol, SK-BR-3 cells were labeled with rabbit anti-HER2 Ab-QD525-conjugated goat anti-rabbit Ab and rabbit anti-Ki-67Ab-QD605-conjugated goat anti-rabbit Ab. One slide was labeled withanti-HER2–QD525 first, the other was labeled with anti-Ki-67–QD605first. For the demonstration of the cocktail method, mouse anti-HER2 Ab (1 : 500, Dako) and rabbit anti-Ki-67 (1 : 200, Dako) weremixed prior to incubation. The Ab mixture was incubated on theSK-BR-3 cell-block section at room temperature for 30 min. After thewashing step with TBS-T, QD605-conjugated goat anti-rabbit Ab(1 : 250, Invitrogen) and QD525-conjugated goat anti-mouse Ab(1 : 100, Invitrogen) were mixed and the QD-labeled Ab mixturewas applied to the slides at room temperature for 30 min.

Fabrication of a microfluidic device

A conventional soft lithography technique was conducted to fabricatethe devices. To make a mold of the microchannels, SU-8 3025 wasspincoated on a bare silicon wafer to make rectangular reactionchannels. The wafer was exposed to ultraviolet light with a mask andsubsequently developed using an SU-8 developer. Then, the micro-fluidic device was fabricated by using a poly(dimethylsiloxane)(PDMS; Sylgard 184; Dow Corning, MA) replica molding process.18,19

Microfluidic immunochemical staining

Estrogen receptor (ER) Ab (Ventana, Tucson, AZ), progesteronereceptor (PR) Ab (Ventana), HER2 Ab (1 : 1000, Dako), humangrowth factor receptor 3 (HER3) Ab (1 : 50, Abcam, Cambridge, MA)were used at 1� concentrations or ready-to-use concentrations.Ki-67 (1 : 200, Dako) Ab, the mammalian target of rapamycin(mTOR) Ab (Abcam), transforming growth factor alpha (TGF-a)Ab (1 : 50, Abcam), and betacellulin (BTC) Ab (1 : 100, ProteintechGroup, USA) were also used at 1� ready-to-use concentrations.All these antibodies were produced from rabbits, 2.5 mL of eachbiomarker was mixed with 2.5 mL of b-actin Ab (1 : 1000, Abcam),produced from mice prior to channel injection. To apply theQD-MMBQ to tissue models, SK-BR-3, MCF-7, and HCC-70 breastcancer cell-block sections were deparaffinized. After antigenretrieval treatment, a microfluidic device was mounted on thecell block slide, and the microfluidic channel layer was pressure-sealed to prevent leakage between the channels formed onthe sample slide. The mounted PDMS layer provided multipleisolated areas on the specimen. Then, the deparaffinized cellblocks were blocked with a blocking solution containing bovine



serum albumin (BSA) and goat serum in phosphate-bufferedsaline (PBS; pH 7.4). 5 mL of each mixture was injected to eachinlet, and withdrawal pumping from the outlet was accomplishedto incubate biomarkers at a flow rate of 80 mL h�1 for 30 min.All inlets were washed with TBS-T after the incubation. Then,QD605-conjugated goat anti-rabbit Ab (1 : 250, Invitrogen) andQD525-conjugated goat anti-mouse Ab (1 : 100, Invitrogen) weremixed and 5 mL of the mixture was injected into every inlet. QDmixtures were incubated at a flow rate of 80 mL h�1 for 30 min.

Prior to tissue staining, device alignment to the specific region oftissue section is required to collect protein signals from the regionwhere the cancer cells are densely populated due to the morpholo-gical heterogeneity of breast cancer tissues. The alignment processwas described in our previous report.18 Briefly, one of the tissuesections was stained with hematoxylin and eosin. After the determi-nation of the region where tumor cells are populated in high density,the device was aligned to that region by comparing adjacent tissuesections to the hematoxylin and eosin stained one. All remainingprocesses for multiplexed QD staining were the same as above.

Quantification of proteins

For biomarker quantification, optical spectroscopy was performedusing a fluorescence microscope (IX72; Olympus) with a spectro-meter (QE65000; Ocean Optics) and a charge-coupled device. Todetermine the staining intensity of each protein, the fluorescencesignal was collected from each strip individually. As mentionedabove, we used two QDs (QD525 and QD605) with emission peaksof 525 nm for the reference marker and 605 nm for the targetproteins. After elimination of autofluorescence, the QD605 signalwas normalized to that of QD525 (see ESI,† Fig. S2).

Conclusions

We have demonstrated a microfluidic multiple biomarker quanti-fication method using QD nanoparticles and microfluidics.Particularly, we verified the QD sequential staining methodstaining-sequence problem using cell-block sections. The QDlabeling method based on sequential staining induces unintendedbinding of QD–secondary-Ab conjugates, making qualitative andquantitative analyses problematic. In addition, our novel methodcan quantify simultaneously more than eight proteins in the samebiological specimen. Expression levels of eight proteins in variouscell lines were compared by co-labeling each target protein and thesame LC in each microchannel. Our microfluidic method providesthe most flexible multiplexed QD-immunochemical assay techni-que, allowing facile data analysis and more accurate biomarkerquantification. LC-based quantification also shows the effect ofcell number and staining quality variation among specimens.We also showed a possibility of using this system in multiplebiomarker quantification of breast cancer tissues. Furthermore,microfluidic blocking is superior and controllable compared to theconventional static method. Based on its applicability in breasttissue from cancer patients, this facile method for quantificationof multiple biomarkers will contribute to realization of the use ofmultiplexed QD-immunochemical assay in many fields.

Acknowledgements

This research was supported by a National Leading ResearchLaboratory Program (NRF-2013R1A2A1A05006378), a Nano/BioScience and Technology Program (NRF-2005-2001291), and aConverging Research Center Program (2011K000864) throughthe National Research Foundation of Korea funded by theMinistry of Science, ICT and Future Planning. The authors alsoacknowledge a Cooperative Research Program for AgricultureScience and Technology Development (Grant PJ009842) supportedby the Rural Development Administration of Korea.

Notes and references

1 D. K. Nomura, M. M. Dix and B. F. Cravatt, Nat. Rev. Cancer,2010, 10, 630–638.

2 N. Turner, M. B. Lambros, H. M. Horlings, A. Pearson,R. Sharpe, R. Natrajan, F. C. Geyer, M. van Kouwenhove,B. Kreike and A. Mackay, Oncogene, 2010, 29, 2013–2023.

3 P. Zrazhevskiy and X. Gao, Nano Today, 2009, 4, 414–428.4 B. Weigelt, A. Mackay, R. A’hern, R. Natrajan, D. S. Tan,

M. Dowsett, A. Ashworth and J. S. Reis-Filho, Lancet Oncol.,2010, 11, 339–349.

5 M. C. Cheang, D. Voduc, C. Bajdik, S. Leung, S. McKinney,S. K. Chia, C. M. Perou and T. O. Nielsen, Clin. Cancer Res.,2008, 14, 1368–1376.

6 M. Scaltriti, P. J. Eichhorn, J. Cortes, L. Prudkin, C. Aura,J. Jimenez, S. Chandarlapaty, V. Serra, A. Prat and Y. H.Ibrahim, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 3761–3766.

7 S. Zhang, W.-C. Huang, P. Li, H. Guo, S.-B. Poh, S. W. Brady,Y. Xiong, L.-M. Tseng, S.-H. Li and Z. Ding, Nat. Med., 2011,17, 461–469.

8 U. Resch-Genger, M. Grabolle, S. Cavaliere-Jaricot,R. Nitschke and T. Nann, Nat. Methods, 2008, 5, 763–775.

9 I. L. Medintz, H. T. Uyeda, E. R. Goldman and H. Mattoussi,Nat. Mater., 2005, 4, 435–446.

10 X. Gao, Y. Cui, R. M. Levenson, L. W. Chung and S. Nie,Nat. Biotechnol., 2004, 22, 969–976.

11 X. Michalet, F. Pinaud, L. Bentolila, J. Tsay, S. Doose, J. Li,G. Sundaresan, A. Wu, S. Gambhir and S. Weiss, Science,2005, 307, 538–544.

12 Y. Xing, Q. Chaudry, C. Shen, K. Y. Kong, H. E. Zhau,L. W. Chung, J. A. Petros, R. M. O’Regan, M. V. Yezhelyevand J. W. Simons, Nat. Protoc., 2007, 2, 1152–1165.

13 M. V. Yezhelyev, A. Al-Hajj, C. Morris, A. I. Marcus, T. Liu,M. Lewis, C. Cohen, P. Zrazhevskiy, J. W. Simons andA. Rogatko, Adv. Mater., 2007, 19, 3146–3151.

14 J. Liu, S. K. Lau, V. A. Varma, R. A. Moffitt, M. Caldwell,T. Liu, A. N. Young, J. A. Petros, A. O. Osunkoya andT. Krogstad, ACS Nano, 2010, 4, 2755–2765.

15 E. Sweeney, T. H. Ward, N. Gray, C. Womack, G. Jayson,A. Hughes, C. Dive and R. Byers, Biochem. Biophys. Res.Commun., 2008, 374, 181–186.

16 D. Huang, X. Peng, L. Su, D. Wang, F. R. Khuri andD. M. Shin, Nano Res., 2010, 3, 61–68.



17 J. Liu, S. K. Lau, V. A. Varma, B. A. Kairdolf and S. Nie,Anal. Chem., 2010, 82, 6237–6243.

18 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.

19 M. S. Kim, S. Kwon, T. Kim, E. S. Lee and J.-K. Park,Biomaterials, 2011, 32, 1396–1403.

20 J. B. de Kok, R. W. Roelofs, B. A. Giesendorf, J. L. Pennings,E. T. Waas, T. Feuth, D. W. Swinkels and P. N. Span,Lab. Invest., 2004, 85, 154–159.

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

22 P. Zrazhevskiy and X. Gao, Nat. Commun., 2013, 4, 1619.23 P. K. Chattopadhyay, D. A. Price, T. F. Harper, M. R. Betts,

J. Yu, E. Gostick, S. P. Perfetto, P. Goepfert, R. A. Koup andS. C. De Rosa, Nat. Med., 2006, 12, 972–977.

24 J. Kao, K. Salari, M. Bocanegra, Y.-L. Choi, L. Girard,J. Gandhi, K. A. Kwei, T. Hernandez-Boussard, P. Wangand A. F. Gazdar, PLoS One, 2009, 4, e6146.

25 P. A. Kenny, G. Y. Lee, C. A. Myers, R. M. Neve, J. R. Semeiks,P. T. Spellman, K. Lorenz, E. H. Lee, M. H. Barcellos-Hoffand O. W. Petersen, Mol. Oncol., 2007, 1, 84–96.

26 L. C. Tong, N. Nelson, J. Tsourigiannis and A. M. Mulligan,Am. J. Surg. Pathol., 2011, 35, 545–552.

27 I. Z. Yildiz-Aktas, D. J. Dabbs and R. Bhargava, Mod. Pathol.,2012, 25, 1098–1105.

28 P. Maxwell and W. McCluggage, J. Clin. Pathol., 2000, 53,929–932.

29 T. J. O’Leary, Appl. Immunohistochem. Mol. Morphol., 2001, 9, 3–8.30 T. Seidal, A. J. Balaton and H. Battifora, Am. J. Surg. Pathol.,

2001, 25, 1204–1207.31 C. R. Taylor, Hum. Pathol., 1994, 25, 2–11.


rsc ib c3ib40224j 3.nanobio.kaist.ac.kr/papers/c3ib40224j_author_reprint.pdfa quantum dot-based...

Documents