Elasticity-Modulated Microbeads for Classification of FloatingNormal and Cancer Cells Using Confining MicrochannelsJifeng Ren,† Jiyu Li,† Yongshu Li,‡ Peng Xiao,† Yi Liu,† Chi Man Tsang,‡ Sai Wah Tsao,‡ Denvid Lau,§
Kannie W. Y. Chan,*,† and Raymond H. W. Lam*,†,∥,⊥
†Department of Biomedical Engineering, City University of Hong Kong, Kowloon, Hong Kong, Special Administrative Region of thePeople’s Republic of China‡School of Biomedical Sciences, The University of Hong Kong, Pok Fu Lam, Hong Kong, Special Administrative Region of thePeople’s Republic of China§Department of Architecture and Civil Engineering, City University of Hong Kong, Kowloon, Hong Kong, Special AdministrativeRegion of the People’s Republic of China∥City University of Hong Kong, Shenzhen Research Institute, Shenzhen 518057, China⊥Centre for Biosystems, Neuroscience, and Nanotechnology, City University of Hong Kong, Kowloon, Hong Kong, SpecialAdministrative Region of the People’s Republic of China
ABSTRACT: Engineered microbeads have a wide range ofapplications in cancer research including identification,characterization, and sorting of cancer cells. In particular, themicrobead-based cancer identification techniques are mainlybased on the known genetic or biochemical biomarkers; anddetection specificity is yet to be improved. On the other hand, ithas been discovered that biomechanical properties of cancercells such as cell-body elasticity can be considered as cancerbiomarkers. Here, we report a straightforward microfluidicclassification scheme for floating/dissociated normal and cancerepithelial cells using a confining microchannel device togetherwith calcium-alginate hydrogel microbeads. The hydrogelmicrobeads are generated based on the microfluidic emulsionprocess, with characterization on the process parameters (e.g., liquid driving pressure and cross-linking duration) in order tospecify the resultant bead diameter and elasticity. These engineered microbeads are first mixed with a cell mixture of dissociatedhuman nasopharyngeal epithelial cells (NP460) and nasopharyngeal carcinoma cells (NPC43). The cell elasticity can then bereflected from the locations of captured cells in the device. Experiments further demonstrate that the cell classification has asuccess rate of >95%. Furthermore, we performed the microbead-based cell classification on a whole blood sample containingfloating human breast epithelial cells (MCF-10A) and breast cancer epithelial cells (MDA-MB-231) with a success rate of>75%, revealing its directly applicability to identification of circulating tumor cells in human blood. Together, this researchdemonstrates a new application of engineered hydrogel microbeads for classification of cells based on their mechanicalproperties.
KEYWORDS: microfluidic, microbead, elasticity, cell classification, cancer
■ INTRODUCTION
Engineered microbeads have been widely adopted in cancerdiagnosis1 and other biomedical applications such as drugdelivery2 and tissue regeneration.3 They can function as micro/nanoparticles as affinity-based cell carriers4 or cell-positionindicators.5 Their biomaterials are often engineered tointegrate with biochemical biomarkers such as antibodies6
and aptamers.7 The microbeads are typically applied as animaging accessory to enhance the biomarker signals,8 toobserve locations of target cancer cells,9 and to isolate cancercells.10 While very effective, there are still great demands onfurther development of functional microbeads for extendedcancer cell characterization capability.
Considering the heterogeneous genetic and phenotypicproperties of cancer cells, cancer diagnosis can be achieved inthe manner of single-cell analysis, in which the single-cells areobtained from chemical-dissociation of a resected tumorportion or directly from human blood. Tissue dissectionshave been widely used in cancer diagnosis,11 treatment,12 aswell as the mechanistic study of cancer pathology,13 with
Special Issue: Biomaterials for Mechanobiology
Received: October 16, 2018Accepted: May 29, 2019Published: May 29, 2019
various techniques recently developed such as arginine-glycine-aspartic acid magneto-optical nanoparticles14 and indocyaninegreen fluorescence.15 On the other hand, cancer cell propertiesincluding their metastatic tendency can be examined using thecirculating tumor cells (CTCs).16 The presence of CTCsindicates the higher potential of establishing new colonies atother body sites through extravasation.17 Further, the densityof CTCs in cancer patients’ blood is considered a prognosticmarker, especially after surgical resection.9
Although there has been a wide range of genetic andbiochemical biomarkers for cancer cells, the specificity forcancer identification is yet to be improved.17 Importantly, ithas also been discovered that the biophysical cell propertiesrelated closely to the invasiveness of cancer cells;18 and thecancer metastatic potential depends also on the mechanicalproperties of the tumor cells.19 It has been shown that theelasticity of cancer cells is often significantly lower compared tothe normal cells.20 In particular, elasticity can also beconsidered an indicator for fluidization of malignant cancercells transited from tumor cells.21 Cell-body elasticity candirectly reflect deformation of a cell when it squeezes through anarrow region. For instance, breast cancer cells require ashorter time than normal breast cells to deform and enter amicrochannel.22 To date, elasticity has already been consideredas a biomarker for cancer types such as ovarian cancer23 andprostate cancer.24 Recently, it has been demonstrated thathydrogel microbeads can be applied for cell-stress sensing,25
which has a close correlation with the cell elasticity. Thoughvery effective, the throughput is largely limited by thenanoindentation-based cell-by-cell probing manner and themeasurement is constrained to only adherent cells. Together,development of high-throughput microbead-based techniquesclassifying cancer cells for their distinct elasticity, as anadditional step of cancer cell identification, would help tofurther enhance specificity of cancer detection and character-ization.Microfluidic techniques have demonstrated its outstanding
microflow control and microdroplet/microbead generationcapability.26 Emulsion of two immiscible liquids can generatedispersing microdroplets.27 For example, oil and water are twoimmiscible phases and water droplets can be formed by theemulsion process.28 Microfluidics can offer precise flow controland generate a very consistent size of the emulsifiedmicrodroplets,29 which is a significant improvement from theconventional macroscale emulsion technique.26 The micro-droplets can include defined compositions or be embeddedwith nanoparticles (e.g., quantum dots).30 It has also beenreported that microbeads with regulated porosity can then besolidified from the microdroplets by either solvent evapo-ration.31 Alternatively, microbead solidification can beachieved via internal gelation of polymers by chemicalreactions, such as ionic cross-linking of sodium alginate withthe liberated Ca2+ ions,32 cross-linking of ethylene glycoldimethacrylate-glycidyl methacrylate,33 and gelation of aque-ous pregel mixture of N-isopropylacrylamide, bis-acrylamide,and ammonium persulfate.34
On the other hand, there are a number of microfluidicplatforms developed for biomechanical phenotyping of floatingcells.35 Generally, microfluidics offer rapid and high-through-put cell characterization comparing to the traditional methodssuch as micropipet aspiration and atomic force microscopy.36
For example, an acoustic trapping force has been applied tomeasure the drag viscous effect and estimate the cell stiffness.37
The design of microchannels can help characterizing andsorting cells based on their deformability.38 Recently, we havereported on a confining microchannel device for characterizingnormal and cancerous human epithelial floating cells of theirmechanical properties such as elasticity39 and cytoplasmicviscosity.40 Further cell characteristics such as surface proteinexpression39 and migration along constrictions can also bequantified.41 Despite the characterization, specificity can beimproved by utilizing a more representative model to convertthe measured parameter to cell elasticity,7,11 and the elasticityestimate is sensitive to any variations in the experimentalconfiguration. Hence, addition of microparticles with areference elasticity level as the cutoff element to reveal thesofter cancer cells would provide a high level of applicability inidentification and classification of the cancer cells in anunknown cell extract from cancer patients.In this work, we, for the first time, utilize engineered
hydrogel microbeads as an indicator for elasticity-basedclassification of floating human normal and cancer cells. Weexamine the operation parameters of microfluidic emulsion inorder to produce the microbeads with desired properties.These microbeads are then mixed with the cell samples,following by injecting the mixture into a confining micro-channel device for cell classification. The lower elasticity of thecancer cells can then be reflected from locations of thecaptured cells. We investigate micromanufacturing of themicrobeads with defined size and elasticity for classification ofnormal and cancerous nasopharyngeal epithelial cells. Wefurther apply this technique for classifying normal andcancerous breast epithelial cells in whole-blood, in order todemonstration its applicability in the direct cell classification ofblood, including identification of CTCs.
■ METHODS AND MATERIALSFabrication. The confining microchannel device and the micro-
fluidic microbead generator used in this work were fabricated basedon soft photolithography. The mold for the emulsion device wasfabricated by patterning a 20 μm-thick layer of negative photoresist(SU-8 2010, Microchem) on a silicon wafer by photolithography. Forthe confining microchannel mold, a layer of 50 μm thick positivephotoresist (AZ50XT, AZ Electronic Materials) was patterning on asilicon wafer by photolithography. Then, deep reactive ion etchingwas applied to etch the uncovered silicon regions for a depth of 50μm, followed by removing all the residual photoresist using acetone.Both molds were then deposited with a molecular layer of trichloro(1H,1H,2H,2H-perfluoro-octyl) silane (Sigma-Aldrich) to facilitatesubstrate release from them in the later process.
In the later fabrication steps, both devices basically shared the samefabrication process. Polydimethylsiloxane (PDMS) prepolymer(Sylgard-184, Dow Corning) was prepared by mixing the PDMSmonomer with a 10% volumetric ratio of the curing agent. Afterdegassing in a vacuum chamber and pouring the prepolymer onto themold masters with a thickness of 3 mm, the masters were baked in anoven at 80 °C overnight for thorough cross-linking of PDMS. Thecured PDMS substrate was torn from the mold, and the unwantedPDMS outside the device boundaries was chopped away. We thenpunched holes at the liquid inlets and outlets. The diameters of theseholes were typically 1 mm, except that the outlet of the microbeadgenerator was 5 mm, acting as the microbead collection chamber.Afterward, the PDMS substrate was bonded onto a glass slide usingoxygen plasma (Plasma Prep II, SPI Supplies) treatment (energy, 10kJ). Finally, each fabricated device was baked in an oven for another 6h to enhance the plasma bonding strength. Stainless steel adaptors(New England Small Tube) and Tygon tubing (Cole-Parmer) werethen inserted into the device inlet/outlet for fluidic connections.Before experiments, the device with tubing was exposure under
ultraviolet light for >2 h in a tissue culture hood for sterilization.Surface treatment was then applied to avoid unwanted cell adhesionby injecting pluronics F-127 into fabricated device for 30 min beforeexperiments.Cell Culture. An immortal human nasopharyngeal epithelial cell
line (NP460) and a nasopharyngeal carcinoma cell line (NPC43)were developed by the research team of S. W. Tsao, from cell extractsof nasopharyngeal cancer patients. NP460 cells were maintained inthe 50% of complete Eplife medium (Thermo Fisher Scientific), 50%of complete Defined Keratinocyte-SFM (Thermo Fisher Scientific),100 units/mL penicillin and 100 μg/mL streptomycin. NPC43 cellswere maintained in RPMI-1640 (Sigma) added with 10% fetal bovineserum, 4 μM Y27632 dihydrochloride (Alexis), 100 unit/mLpenicillin, and 100 μg/mL streptomycin.A human breast cancer cell line MDA-MB-231 was cultured with
high-glucose Dulbecco’s modified Eagle’s medium (DMEM; Invi-trogen, Carlsbad, CA), 10% fetal bovine serum (Atlanta Biological,Atlanta, GA), and 100 units/mL penicillin. A noncancerous breastepithelium cell line MCF-10A was cultured with the MammaryEpithelial Cell Growth Medium SingleQuots Kit (MEGM; CC-4136,Lonza, New York City, NY). All the cell types were cultured in anincubator at 37 °C, saturated humidity and 5% CO2 in air. Cellpassage was performed once the cell population reaches ∼80%confluence. In the experiments, 0.25% trypsin-EDTA was applied toextract the cells. Each cell type was diluted to a sufficiently low density(104 cell/mL) for avoiding any cell aggregation in the microfluidicdevice in the following experiments.Cell Staining. Vybrant CFDA SE dye (Thermo Fisher Scientific)
with a concentration of 1 μL/mL was added in cell culture media for10 min to stain NPC43 cells or MDA-MB-231 cells, followed byreplacing the dye solution with fresh media. Likewise, Hoechst 33342(Thermo Fisher Scientific) with a concentration of 0.1 μg/mL inphosphate buffered saline (PBS) was applied to stain NP460 cells orMCF-10A cells for 10 min, followed by replacement of fresh media.Whole-Blood Sample. Anticoagulated bovine whole blood
samples were purchased from Hongquan Bio, Guangzhou, China.Image Capture. Microscopic images of the cells and microbeads
were captured under an inverted fluorescence microscope (TE300,Nikon) equipped with an sCMOS microscope camera (Zyla 4.2,Andor).Statistics. All error bars in the plots represent standard errors. p-
values are obtained using the Student’s t test. Asterisks represent asignificant statistical difference between two groups of data in a plot.
■ RESULTS AND DISCUSSION
Microbead-Based Cell Classification Scheme. In thiswork, we consider deformation of microbeads in confiningmicrochannel structures as a cell elasticity indicator for cancercell classification applications. We should first engineer themicrobeads with a comparable size with the cell samples andan elasticity level between the cell types we plan to classify(details are provided in the later sections). On the other hand,the confining microchannel device (Figure 1a) has beendeveloped and configured for human cancer cells based on ourprevious works.42 Briefly, its flow region (height, 50 μm)contains two series of 300 μm long confining microchannels,whose width is 30 μm at the channel entrance and reduces to 4μm at the exit. As the microbeads and cells are smaller than thechannel entrance and larger than the channel exit, they can betrapped in the microchannels under adequate pressures.The cell elasticity classification procedures are straightfor-
ward. The microbeads with defined elasticity are first mixedwith a cell sample, which might include normal and cancercells with different elasticity levels. We can then inject thecell−bead mixture into the confining microchannel device witha steady pressure from the inlet. The pressure regulation can beachieved by filling the sample in a syringe without a piston,
connecting the syringe to the device tubing, and applying acontrolled gage pressure from the rear opening of the syringe.When the sample is driven into the microchannels, some of thecells/microbeads flowing along the microchannels woulddeform upon compression by the channel walls and stay inthe channel with a “penetration distance” from the channelentrance. Such penetration distance depends on both the cellsize and elasticity, e.g., a larger or stiffer microparticle has ashorter penetration distance. We have previously derived atheoretical relation that elasticity of a cell can be calculateddirectly from its corresponding size and penetration distance.39
It should be mentioned that the estimated elasticity level is alsosensitive to the operation parameters such as the inlet pressure;and therefore, our previous theoretical approach has a hightechnical barrier on the precise pressure control for practicalimplementation. Here, the added microbeads can eliminatesuch a technical barrier by offering a reference penetrationdistance for distinguishing harder and softer cells. That is, aharder or softer cell can be directly observed in the confiningchannels on whether it has a shorter or longer penetrationdistance than the microbeads, respectively (Figure 1b). Thisoperation-friendly microbead-based approach provides a directcell classification result, the population proportion of “softer”cells in a biosample. This approach is highly applicable incancer diagnosis as the softer cells are likely the malignant cellsof many cancer types, such as breast cancer43 and nasopharynxcancer.44 Together, the key factor for success of thismicrobead-based cell classification is on generating microbeadswith defined physical properties, functioning as the cutofflevels in size and elasticity for distinguishing normal and cancercells.
Microbead Formation. We have adopted a microfluidicemulsion device to generator hydrogel microbeads withdefined physical properties. This device (Figure 2a) includesa layer of flow microchannels with a consistent height of 20μm. There was an inlet for the oil phase sample and another
Figure 1. (a) Fabricated confining microchannel device. Inset:microscopic image of the flow region containing two series ofconfining microchannels. Red dye is injected into the device for bettervisualization. The expected liquid flow is indicated by the arrows.Scale bar: 500 μm. (b) Concept of the elasticity-based cellclassification. Positions of microbeads inside the microchannelsreflect the microbead elasticity; and we can consider them as cutoffpositions for distinguishing harder and softer cells in a cell mixture.
inlet for water phase sample. As shown in Figure 2b, the oilsample flowing through the oil inlet would split into twomicrochannels with the same width of 20 μm, whereas theaqueous sample flowed along the center microchannel (width,20 μm), merged with, and was sandwiched by the two oilflows. The shear stresses over the water−oil interface thenbroke the water flow into microdroplets, as emulsions, alongthe downstream microchannel with a width of 30 μm. Themicrodroplets then flowed to a collection chamber. Sub-sequently, the polymers in the aqueous droplet are furtherprocessed and cross-linked as a hydrogel.
In the experiments, mineral oil (M5904, Sigma-Aldrich) wasfirst injected from both inlets under a positive gauge pressurealong in order to remove all air bubbles in the microchannels.We replaced the center microchannel with a solution of 1%(w/v) aqueous alginate (Novamatrix) and 50 mM/mL calciumethylenediaminetetraacetic acid (Ca-EDTA; SinopharmChemical Reagent Co., Ltd.) at another gauge pressure. Afterthe two liquid phases met at the merging channel junction, theemulsified microdroplets were collected at the outlet chamber.Apparently, the droplets were harvested over the chamber basebecause mineral oil has a lower density (0.8 g/cm3) than water(Figure 2c). We then stopped the flows until the collectionchamber had a liquid volume of ∼30 μL.Afterward, we prepared a solution of 0.1% (v/v) acetic acid
(Sigma-Aldrich) in mineral oil with vortex mixing and mildagitation (>10 min). It has been reported that hydrophilicacetic acid dissolves in oils.45 Its dielectric constant (= 6.2) ismuch lower than that of water (= 80.4) at room temperature,implicating that acetic acid can dissolve in nonpolarcompounds including oils.46 Ca-EDTA was initially stable inthe collection chamber because EDTA completely chelated theCa2+ at the neutral pH state. Yet, after pipetting such “acidic-oil” with a volume of ∼100 μL into the chamber, acetic acidgradually diffused from the oil phase and reduced the pH ofthe water phase, triggering the release of Ca2+ from the Ca-EDTA compounds. It has been reported that the G-blocks ofalginate should then cross-link with Ca2+ to form hydrogelmicrobeads.47 We kept the microbeads in the collectionchamber for target duration to achieve a sufficient level ofcross-linking. We then pipetted the microbeads into a 15 mLsyringe tube containing >5 mL of phosphate buffered saline(PBS), followed by a brief centrifuge (500 rpm, 3 min) totransfer the microbead into PBS and prevent any further cross-linking of the hydrogel microbeads.
Bead Size Modulation. Diameter of the hydrogelmicrobeads can be modulated by varying the operationparameters with the microbead generator. In particular, wehave characterized the microbead diameter as a function of thedriving pressures during the microfluidic emulsion. Werepeatedly generated the microbeads with the water-phasepressure ranging from 6 to 15 kPa under a fixed oil phasepressure as 17.2 kPa, followed by quantifying them after theacidic oil treatment of 2 h. We have examined that a smallportion of the microdroplets might combine together duringthe acidic oil treatment. This could induce the resultantmicrobead diameter of <5%, compared to the microdropletdiameter before treatment. Results (Figure 2d) indicate thatthe microbead size did not significantly vary under the water-phase pressure of 6−8 kPa, but it increased with the pressurerange of 8−15 kPa. A water-phase pressure >15 kPa could notinduce any microdroplet formation.We have quantified for diameters of NP460 cells and
NPC43 cells. We took microscopic images of the trypsinizedcells and measured the cell diameters as summarized in Figure2e, indicating that there is no significant difference on cell sizebetween the two cell types. Hence, we applied a water-phasepressure of 10.3 kPa to prepare the microdroplets (diameter,13.62 ± SE 0.205 ± SE 0.2 μm) and then the hydrogelmicrobeads (diameter, 13.93 ± SE 0.205 μm). Furthermore,we have also measured the diameters of MCF-10A (13.16 ±SE 0.35 μm) and MDA-MB-231 (16.05 ± SE 0.46 μm).According to Figure 2e, we configure the water-phase pressure
Figure 2. (a) Photograph of a microfluidic emulsion device. (b)Microscopic image of the microchannel junction at which microfluidicemulsion occurs. Arrows indicate the flow direction. Scale bar: 80 μm.(c) Generated microbeads in the collection chamber. Scale bar: 40μm. (d) Diameter of microbeads as a function of inlet pressure of thewater-phase sample. (N > 30). (e) Average diameters of NP460 cells(N = 90), NPC43 (N = 103) cells, and the microbeads (N = 65)generated with a water-phase pressure of 10.3 kPa. (f) Averagediameter of MCF-10A cells (N = 20), MDA-MB-231 (N = 21), andthe microbeads (N = 20) generated using a water-phase pressure of11 kPa. All error bars in the plots are the standard errors.
of 11 kPa to generate microbeads with a diameter (14.11 ± SE0.3 μm)) between the two cell sizes, as described in Figure 2f.Bead Elasticity Modulation. The microbead elasticity is
determined by the cross-linking between alginate and Ca2+.More specifically, the proportion of cross-linked G-blocks inalginate increases with the material elasticity.48 Consideringthe microbead formation process in this work, the alginate-Ca2+ hydrogel only forms at an adequately acidic environment(pH < 4) in the collection microchamber; and the rate ofalginate-Ca2+ cross-linking increases with the acidity.40 Thus,the rate of hydrogel cross-linking should be largely constrainedby the concentration of acetic acid in the microdroplets.Molecular transport of the acetic acid is driven by diffusion ofacetic acid from the surrounding mineral oil. Considering thatmolecular diffusivity in mineral oil49 can be a few orders ofmagnitude smaller than that in water, the pH value of theemulsified microdroplets decreases slowly over time, simulta-neously increasing the level of hydrogel cross-linking, i.e., theamount of cross-linked G-blocks in alginate. Therefore, we canmodulate the microbead elasticity by changing the duration ofthe acidic-oil application.We performed repeated runs of the microbead generation
using different acidic-oil treatment durations from 1 to 5 h. Foreach group of the treatment duration, we quantified for thecorresponding microbead elasticity using the confining micro-channel device as we previously reported.39 Briefly, we injectedthe microbeads into the confining microchannels under asteady driving pressure, measured the bead diameters and thepenetration distance, and converted these measured values asthe bead elasticity based on the Hertz and Tatara model.Results (Figure 3a) indicate a clear trend that the beadelasticity increases with the treatment duration. On the otherhand, we applied the confining microchannels with the sameprocedures to quantify the elasticity values of NPC43 (6.43 ±SE 0.11 kPa) and NP460 cells (7.48 ± SE 0.28 kPa). We thenchose the oil-acid treatment duration of 2 h such that thehydrogel microbeads would have elasticity (6.76 ± SE 0.033kPa) between those two cell types (Figure 3b). Likewise, wemeasured the elastic moduli of MCF-10A (3.62 ± SE 0.10kPa) and MDA-MB-231 (3.10 ± SE 0.03 kPa) using the sameapproach (Figure 3c), suggesting that an oil-acid treatmentduration of 0.5 h should be configured for distinguishing thesetwo cell types, as shown in Figure 3a.In addition, we conducted further experiments to examine
variation of the microbead elasticity after the microbeads weretransferred in the culture media with cells. We investigatedelasticity of the microbeads in the culture media for differenttime periods, as shown in Figure 3d. Typically, the preparationprocedures (e.g., centrifuge, mixing with cells, and injectioninto the confining microchannel) could be done within 10 min.The results indicate that the microbead elasticity was stable inthe media for at least 15 min, which should be an ideal timeframe for the cell classification test. For the bead suspension of>1 h, we have observed that microbeads softened and could nolonger be captured in the confining microchannels. Themicrobeads then became smaller gradually and most of themwould degrade thoroughly after 3 days. In fact, it has been alsoreported that Ca-alginate microbeads would degrade graduallyin aqueous solutions such as PBS.21 The Ca-alginate cross-linking can be lost through calcium exchange with thesurrounding solution.50 A higher pH environment wouldallow a faster calcium exchange and Ca-alginate degradation;41
and this may explain why the microbeads would soften in the
culture media with a natural pH in our experiments. Therefore,it is suggested that the cross-linked microbeads should bemaintained in pure mineral oil for storage; and they can betransferred to the biosample right before the cell classificationoperation.
Classification of Dissociated Cells. We further demon-strated classification of nasopharyngeal normal (NP460) andcancer (NPC43) cells using the Ca-alginate microbeads,prepared with a water-phase pressure of 10.3 kPa and an oil-acid treatment period of 2 h. In practice, these cells aretypically obtained from dissected normal/tumorous tissues.For distinguishing the two different cell types, we first stainedthe NP460 and NPC43 cells with different fluorescence. Wethen mixed the designed microbeads (density, 4 × 103 bead/mL) and the NP460 cells (density, 104 cell/ml) and NPC43cells (density, 104 cell/ml) in a solution containing the sameratio of both culture media. We injected the cell−bead mixtureinto the confining microchannel device with a driving hydraulicpressure of 0.3 kPa. Next, we took microscopic images at theconfining microchannels and measured the penetrationdistance of the captured cells or microbeads. A representativeimage of the confining microchannels is shown in Figure 4a.Briefly, it can be observed that the penetration distance ofcaptured microbeads was generally longer than that of theNP460 cells and shorter than that of the NPC43 cells (Figure4b). Since the beads and both cell types have similar ranges ofcell diameter, these penetration distances reflect that themicrobeads have an elasticity level between the two cell types.
Figure 3. (a) Elasticity of microbeads with a diameter of ∼14 μm as afunction of acidic-oil treatment duration (N > 17). (b) Averageelasticity levels of NP460 cells (N = 21), NPC43 cells (N = 21), andthe microbeads (N = 20) generated with acid treatment for 2 h. (c)Average elasticity levels of MCF-10A (N = 20), MDA-MB-231 (N =21), and microbeads (N = 20) generated with acid treatment for 0.5h. (d) Elasticity modulus of the microbeads (water-phase pressure,10.3 kPa; oi-aid treatment, 2 h) for different suspension duration inculture media (N = 10 for each point). Error bars are the standarderrors in all the above plots.
Although NP460 and NPC43 adopted in this work belong todifferent patients, it has been well proven that cancerousnasopharyngeal cells are generally softer than normalnasopharyngeal cells.11b In principle, the cancer cells canthen be identified by directly observing whether the penetratedposition of cells is longer that of the microbeads.Additionally, distributions of penetration distance against
undeformed cell diameter for the microbeads and the two cellstypes are analyzed as a scattered plot as shown in Figure 4c, inorder to reveal the cell classification performance. The celldiameter D can be further taken into the consideration; and itcan be revealed by a simple relation51 based on the volumeconservation: D = [3W(L2W2/3)]1/3, where L is the cell lengthalong channel and W is the lateral width of the cell center inthe channel. A linear regression line was applied to illustratebasic tendency for the microbeads (hidden line) as reference.Furthermore, we wrote customized scripts with MATLABR2017a (MathWorks) to implement the Quadratic Discrim-inate Analysis52 based on the measured data for cellclassification. The computed boundaries of classified regionsare shown in Figure 4c, with a separating curve between theregions of NP460 and NPC43 cells. From our results, 95.24%of NP460 cells and 100% of NPC43 cells can be classified
correctly, implicating that this cell classification techniqueshould have a success rate of >95%. If necessary, a higherspecificity could then be achieved by recollecting all the cellsand performing other cell identification tests for thebiochemical biomarkers. Here, the objective of adding theengineered microbeads with the cells is to provide a referencecutoff characteristics curve in the penetration distance−diameter plot. Both the fitting line of the microbead dataand the separating curve between the two cell types can helpthe future cell classification test with an unknown nasophar-yngeal epithelial cell sample on whether there are presentcancer cells.For the later classification test on an unknown cell sample,
the engineered microbeads can be added to the sample,followed by injecting the cell−bead mixture into the confiningmicrochannels. We should then quantify the properties of boththe cells and microbeads as a scattered plot of penetrationdistance against cell diameter. An updated fitting line can becalculated by on linear regression using the new microbeaddata (R2 = 0.58). By comparing the new regression line with areference line (hidden line in Figure 4c), we can then obtainproper scaling parameters to rescaling the cell separating curvedescribed in Figure 4c for the new measurement. Notably,
Figure 4. (a) Micrograph of a confining microchannel region with captured microbeads, NP460 cells, and NPC43 cells. It is combined with thebright-field and fluorescence (one for blue emission light and one for green emission light) images. White arrows indicate NP460 cells whereasblack arrows indicate NPC43 cells. A captured microbead is labeled with its deformed length L and width W. Scale bar: 50 μm. (b) Penetrationdistances of cells and microbeads in the cell classification experiment (N > 15). Error bars are standard errors. (c) Penetration distance againstdiameter of the cells and beads. A fitting line of the microbead distribution is shown as the hidden line. A separating curve between the two cellregions is in red.
these scaling parameters can eliminate minor variations ofexperimental configurations (e.g., driving pressure) betweendifferent classification tests. The possible existence of cancercells can then be directly identified by whether there exists anycell with a location above the separating curve in the scatteredplot.Future applications of this cell classification scheme is not
limited to identifying nasopharyngeal epithelial cells dis-sociated from the corresponding resected tissue for presenceof nasopharyngeal carcinoma, which is relatively morecommon among southern Chinese areas50 with to date adeath of tens of thousands of people or even more.53
Dimensions of microbeads and confining microchannels aswell as the bead elasticity can be reconfigured to identify othercancer types. Importantly, the captured cells in the device canbe recollected using a higher driving pressure (>1 kPa) and thedegradable Ca-alginate hydrogel microbeads would allow awide range of follow-up cell analyses and engineeringprocesses;54 and the device can be reused after the cellremoval by flushing PBS with either a higher forward drivingpressure or a reversed flow.
Classification of Floating Cells in Whole Blood. Wefurther examined the feasibility of the microbead-based cellclassification for whole-blood samples containing multipleepithelial cell types. Importantly, cancer patients may have onlythe cancer cells found in their blood rather than thenoninvasive normal cells, therefore mixing the microbeadswith the blood sample can then provide the reference normal-cell properties to identify the softer cells, which are potentiallythe CTCs. As the whole-blood consists of multiple blood celltypes55 including red blood cells with a relatively high density,direct injection of the whole-blood can lead to clogging of cells(e.g., red blood cells and platelets) along the microchannels.We have tested that the direct blood injection inducesaggregation of the blood cells and blocking all the micro-channels in the microfluidic device. Besides, such high celldensity may interfere the flow characteristics and thecorresponding cell penetration lengths along the confiningmicrochannels. We first examined a dilution rate of the bloodsuch that cells flowing with the diluted blood would induce thesame penetration distance as the case without mixing with theblood. We conducted the experiments with cancerous (MDA-MB-231, at a density of 104 cell/mL) breast cells and
Figure 5. (a) Comparison of penetration distances of MDA-MB-231 cells mixed with different diluted bovine blood concentration, under a drivingpressure of 0.15 kPa. (b) Microscopic images of captured MCF10A cells, MDA-MB-231 cells, and microbeads. Pathlines of flowing red blood cells(RBCs) have also been observed in the confining channels. Scale bar, 30 μm (inset, fluorescence micrographs). (c) Penetration distance againstdiameter of the cells and beads mixed in bovine blood with a dilution ratio of 1:8. A fitting line of the microbead distribution is shown as the hiddenline. A separating curve between the two cell regions is in red.
noncancerous epithelial cells in bovine whole-blood. Asindicated in Figure 3b,c that breast cells are generally softerthan nasopharyngeal cells, we applied a lower pressure of 0.15kPa such that these cells can be trapped in the confiningmicrochannels with measurable penetration distances.39 Werepeated the penetration distance measurement of MDA-MB-231 for different dilution ratios of the bovine whole-blood inPBS. The results (Figure 5a) indicate that a blood dilutionratio of at least 1:8 should be adopted for the morerepresentative measured penetration lengths.In the classification experiments on cells in the whole-blood,
we prepared the bovine-whole blood with a dilution ratio of1:8 in PBS with addition of MCF-10A and MDA-MB-231prestained with fluorescence, both at a density of 104 cell/mL.We also prepared the Ca-alginate microbeads with a water-phase pressure of 11 kPa and an oil-acid treatment period of0.5 h to configure the microbead with comparable size andelasticity as the MCF-10A cells. We took both bright-field andfluorescence micrographs of the microchannel regions tocapture the positions and dimensions of the cells and themicrobeads as shown in Figure 5b, which indicates also thepathlines of flowing blood cells (mainly the red blood cells).The length (L) and lateral width (W) of the cells/beads wereconverted to the undeformed diameters as described in theprevious section. The cell positions are considered as thepenetration distance from the channel entrance. Similar to theprevious section, a scattered plot of penetration distances anddiameters of both the breast cells and the microbeads is shownin Figure 5c. A fitting line is added for the microbead data asreference. The percentages of correct cell classification are100% for MDA-MB-231 and 77.5% for MCF-10A. It should benoted that Figure 5c is presented for demonstrating thecapability of cell classification using the engineered microbe-ads. The blood samples from cancer patients may include onlythe cancer cells but not the normal noninvasive normal cells;and therefore, we have to add the microbeads to set thereference biophysical conditions, reflected by the measuredpenetration distance in the microfluidic device such that wecan then consider the cells with a longer penetration distancethan the microbeads that is comparable to the potentialcandidates of CTCs.Together, these results support that the microbead-based
cell classification strategy can be applied to identifying CTCsusing the cancer patients’ blood directly, with a reasonablesuccess rate of >75%. Nevertheless, it should be mentionedthat although significant differences on the cell properties canstill be observed between the normal and cancerous cells fromthe cancer patients,20,56 there may exist higher heterogeneityon cell properties (i.e., size and elasticity vary among differentcells of the same cell type) in the patients’ body, meaning thatperformance of the cell classification on a particular patient is,to a certain extent, undetermined. Another technical challengeshould be related to different levels of the physical propertiesof cells in different individuals’ body; and hence, theengineered microbeads may not have the physical propertiesmatching very closely to the targeted normal cells for the bestsuccess rate in cell classification and cancer cell identificationapplications.
■ CONCLUSIONWe report for the first the application of engineered hydrogelmicrobeads in elasticity-based classification of floating normaland cancer cells. Fabrication of Ca-alginate microbeads with
defined diameter and elasticity is achieved by regulatingprocess parameters in the microfluidic emulsion of micro-droplets (driving pressures) and the subsequently solidificationprocess (acid treatment duration). These engineered microbe-ads are then mixed together with the cell samples and injectedinto a confining microchannel device for quantifying the cellelasticity. In this work, we have demonstrated configuration ofthe microbeads for matching the physical properties (diameterand elasticity) close to the target cell types. In essence, we havedemonstrated the classification of dissociated normal andcancerous nasopharyngeal epithelial cells with a high successrate of >95% by quantifying the captured cell and beadlocations in the confining microchannels. The further cellclassification of normal and cancer breast cells in whole bloodshows a success rate of >75%, which is lower compared to thecase without the whole blood. This microbead-based techniquecan be further applied not only in cancer cell identification formore cancer types but also as a universal elasticity character-ization and classification scheme for human blood andcontribute to general cell analysis.
■ AUTHOR INFORMATIONCorresponding Authors*E-mail: [email protected]. Phone: +852-3442-9141. Fax +852-3442-8577.*E-mail: [email protected]. Phone: +852-3442-8577. Fax:+852-3442-0172.ORCIDRaymond H. W. Lam: 0000-0002-5188-3830NotesThe authors declare no competing financial interest.
■ ACKNOWLEDGMENTSWe acknowledge financial supports from the National NaturalScience Foundation of China (NSFC Grant 31770920),General Research Grant (Project No. 11206014), andCollaborative Research Fund (Project No. C1013-15GF) ofHong Kong.
■ REFERENCES(1) Paterlini-Brechot, P.; Benali, N. L. Circulating tumor cells(CTC) detection: clinical impact and future directions. Cancer Lett.2007, 253 (2), 180−204.(2) Calo, E.; Khutoryanskiy, V. V. Biomedical applications ofhydrogels: A review of patents and commerical products. Eur. Polym.J. 2015, 65, 252−267.(3) Bhattarai, N.; Gunn, J.; Zhang, M. Chitosan-based hydrogels forcontrolled, localized drug delivery. Adv. Drug Delivery Rev. 2010, 62(1), 83−99.(4) Huh, Y. M.; Jun, Y. W.; Song, H. T.; Kim, S.; Choi, J. S.; Lee, J.H.; Yoon, S.; Kim, K. S.; Shin, J. S.; Sun, J. S.; Cheon, J. In vivomagnetic resonance detection of cancer by using multifunctionalmagnetic nanocrystals. J. Am. Chem. Soc. 2005, 127 (35), 12387−12391.(5) Luo, S.; Zhang, E.; Su, Y.; Cheng, T.; Shi, C. A review of NIRdyes in cancer targeting and imaging. Biomaterials 2011, 32 (29),7127−7138.(6) Feng, S.; Chen, R.; Lin, J.; Pan, J.; Chen, G.; Li, Y.; Cheng, M.;Huang, Z.; Chen, J.; Zeng, H. Nasopharyngeal cancer detection basedon blood plasma surface-enhanced Raman spectroscopy and multi-variate analysis. Biosens. Bioelectron. 2010, 25 (11), 2414−2419.(7) Hu, S.; Wang, R.; Tsang, C. M.; Tsao, S. W.; Sun, D.; Lam, R. H.Revealing elasticity of largely deformed cells flowing along confiningmicrochannels. RSC Adv. 2018, 8 (2), 1030−1038.
(8) Shotelersuk, K.; Khorprasert, C.; Sakdikul, S.; Pornthanakasem,W.; Voravud, N.; Mutirangura, A. Epstein-Barr virus DNA in serum/plasma as a tumor marker for nasopharyngeal cancer. Clin. Cancer Res.2000, 6 (3), 1046−1051.(9) Peach, G.; Kim, C.; Zacharakis, E.; Purkayastha, S.; Ziprin, P.Prognostic significance of circulating tumour cells following surgicalresection of colorectal cancers: a systematic review. Br. J. Cancer 2010,102 (9), 1327.(10) Lin, J.; Chen, R.; Feng, S.; Pan, J.; Li, B.; Chen, G.; Lin, S.; Li,C.; Sun, L.; Huang, Z.; Zeng, H. Surface-enhanced Raman scatteringspectroscopy for potential noninvasive nasopharyngeal cancerdetection. J. Raman Spectrosc. 2012, 43 (4), 497−502.(11) (a) Oyama, T.; Tomori, A.; Hotta, K.; Morita, S.; Kominato,K.; Tanaka, M.; Miyata, Y. Endoscopic submucosal dissection of earlyesophageal cancer. Clin. Gastroenterol. Hepatol. 2005, 3 (7), S67−S70.(b) Hui, T.; Zhou, Z.; Fong, H.; Ngan, R. K.; Lee, T.; Au, J. S.; Ngan,A.; Yip, T. T.; Lin, Y. Characterizing the malignancy and drugresistance of cancer cells from their membrane resealing response. Sci.Rep. 2016, 6, 26692.(12) (a) Alberts, S. R.; Cervantes, A.; Van de Velde, C. J. H. Gastriccancer: epidemiology, pathology and treatment. Ann. Oncol. 2003, 14(Suppl 2), ii31−ii36. (b) Liauw, S. L.; Mancuso, A. A.; Amdur, R. J.;Morris, C. G.; Villaret, D. B.; Werning, J. W.; Mendenhall, W. M.Postradiotherapy neck dissection for lymph node-positive head andneck cancer: the use of computed tomography to manage the neck. J.Clin. Oncol. 2006, 24 (9), 1421−2427.(13) Ali, M. Y.; Anand, S. V.; Tangella, K.; Ramkumar, D.; Saif, T. A.Isolation of Primary Human Colon Tumor Cells from SurgicalTissues and Culturing Them Directly on Soft Elastic Substrates forTraction Cytometry. J. Visualized Exp. 2015, 100, No. e52532.(14) Montet, X.; Montet-Abou, K.; Reynolds, F.; Weissleder, R.;Josephson, L. Nanoparticle imaging of integrins on tumor cells.Neoplasia 2006, 8 (3), 214−222.(15) Kim, T. H.; Chen, Y.; Mount, C. W.; Gombotz, W. R.; Li, X.;Pun, S. H. Evaluation of temperature-sensitive, indocyanine green-encapsulating micelles for noninvasive near-infrared tumor imaging.Pharm. Res. 2010, 27 (9), 1900−1913.(16) Ashworth, T. R. A Case of Cancer in Which Cells Similar toThose in the Tumors Were Seen in the Blood after Death.Australasian Med. J. 1869, 14, 146−149.(17) Thiery, J. P. Epithelial−mesenchymal transitions in tumourprogression. Nat. Rev. Cancer 2002, 2 (6), 442.(18) Skalak, R.; Branemark, P. I. Deformation of red blood cells incapillaries. Science 1969, 164 (3880), 717−719.(19) Krause, M.; Wolf, K. Cancer cell migration in 3D tissue:Negotiating space by proteolysis and nuclear deformability. CellAdhes. Migr. 2015, 9 (5), 357−366.(20) Cross, S. E.; Jin, Y. S.; Rao, J.; Gimzewski, J. K.Nanomechanical analysis of cells from cancer patients. Nat.Nanotechnol. 2007, 2 (12), 780.(21) Mierke, C. T. Initiation of a neoplasm or tumor. In Physics ofCancer; Mierke, C. T., Ed.; IOP Publishing: Bristol, U.K., 2015.(22) Hou, H. W.; Li, Q. S.; Lee, G. Y. H.; Kumar, A. P.; Ong, C. N.;Lim, C. T. Deformability study of breast cancer cells usingmicrofluidics. Biomed. Microdevices 2009, 11 (3), 557−564.(23) Castle, J.; Shaker, H.; Morris, K.; Tugwood, J. D.; Kirwan, C. C.The significance of circulating tumour cells in breast cancer: a review.Breast 2014, 23 (5), 552−560.(24) Vendrell, M.; Maiti, K. K.; Dhaliwal, K.; Chang, Y. T. Surface-enhanced Raman scattering in cancer detection and imaging. TrendsBiotechnol. 2013, 31 (4), 249−257.(25) Girardo, S.; Traeber, N.; Wagner, K.; Cojoc, G.; Herold, C.;Goswami, R.; Schlußler, R.; Abuhattum, S.; Taubenberger, A.;Reichel, F.; Mokbel, D.; Herbig, M.; Schurmann, M.; Muller, P.;Heida, T.; Jacobi, A.; Thiele, J.; Werner, C.; Guck, J. Standardizedmicrogel beads as elastic cell mechanical probes. J. Mater. Chem. B2018, 6 (39), 6245−6261.
(26) Su, Y. F.; Kim, H.; Kovenklioglu, S.; Lee, W. Y. Continuousnanoparticle production by microfluidic-based emulsion, mixing andcrystallization. J. Solid State Chem. 2007, 180 (9), 2625−2629.(27) Kokal, S. L. Crude oil emulsions: A state-of-the-art review. SPEProd. Facil. 2005, 20 (01), 5−13.(28) Nadler, M.; Mewes, D. Flow induced emulsification in the flowof two immiscible liquids in horizontal pipes. Int. J. Multiphase Flow1997, 23 (1), 55−68.(29) Anna, S. L.; Bontoux, N.; Stone, H. A. Formation of dispersionsusing “flow focusing” in microchannels. Appl. Phys. Lett. 2003, 82 (3),364−366.(30) (a) Clausell-Tormos, J.; Lieber, D.; Baret, J. C.; El-Harrak, A.;Miller, O. J.; Frenz, L.; Blouwolff, J.; Humphry, K. J.; Koster, S.; Duan,H.; Holtze, C.; Weitz, D. A.; Griffiths, A. D.; Merten, C. A. Droplet-based microfluidic platforms for the encapsulation and screening ofmammalian cells and multicellular organisms. Chem. Biol. 2008, 15(5), 427−437. (b) Mazutis, L.; Araghi, A. F.; Miller, O. J.; Baret, J. C.;Frenz, L.; Janoshazi, A.; Taly, V.; Miller, B. J.; Hutchison, J. B.; Link,D.; Griffiths, A. D.; Ryckelynck, M. Droplet-based microfluidicsystems for high-throughput single DNA molecule isothermalamplification and analysis. Anal. Chem. 2009, 81 (12), 4813−4821.(31) Carroll, N. J.; Rathod, S. B.; Derbins, E.; Mendez, S.; Weitz, D.A.; Petsev, D. N. Droplet-based microfluidics for emulsion and solventevaporation synthesis of monodisperse mesoporous silica micro-spheres. Langmuir 2008, 24 (3), 658−661.(32) Zhang, H.; Tumarkin, E.; Sullan, R. M. A.; Walker, G. C.;Kumacheva, E. Exploring microfluidic routes to microgels ofbiological polymers. Macromol. Rapid Commun. 2007, 28 (5), 527−538.(33) Gokmen, M. T.; Van Camp, W.; Colver, P. J.; Bon, S. A.; DuPrez, F. E. Fabrication of porous “clickable” polymer beads and rodsthrough generation of high internal phase emulsion (HIPE) dropletsin a simple microfluidic device. Macromolecules 2009, 42 (23), 9289−9294.(34) Duncanson, W. J.; Lin, T.; Abate, A. R.; Seiffert, S.; Shah, R. K.;Weitz, D. A. Microfluidic synthesis of advanced microparticles forencapsulation and controlled release. Lab Chip 2012, 12 (12), 2135−2145.(35) Ahmad, I. L.; Ahmad, M. R. Trends in characterizing singlecell’s stiffness properties. Micro Nano Syst. Lett. 2014, 2 (1), 8.(36) Sato, M.; Levesque, M. J.; Nerem, R. M. Micropipetteaspiration of cultured bovine aortic endothelial cells exposed to shearstress. Arteriosclerosis 1987, 7 (3), 276−286.(37) Lee, J.; Jeong, J. S.; Shung, K. K. Microfluidic acoustic trappingforce and stiffness measurement using viscous drag effect. Ultrasonics2013, 53 (1), 249−254.(38) Sajeesh, P.; Raj, A.; Doble, M.; Sen, A. K. Characterization andsorting of cells based on stiffness contrast in a microfluidic channel.RSC Adv. 2016, 6 (78), 74704−74714.(39) Hu, S.; Liu, G.; Chen, W.; Li, X.; Lu, W.; Lam, R. H.; Fu, J.Multiparametric biomechanical and biochemical phenotypic profilingof single cancer cells using an elasticity microcytometer. Small 2016,12 (17), 2300−2311.(40) Hu, S.; Lam, R. H. Characterization of Viscoelastic Propertiesof Normal and Cancerous Human Breast Cells Using a ConfiningMicrochannel. Microfluid. Nanofluid. 2017, 21, 68.(41) Irimia, D.; Toner, M. Spontaneous migration of cancer cellsunder conditions of mechanical confinement. Integr. Biol. 2009, 1,506−512.(42) Plaks, V.; Koopman, C. D.; Werb, Z. Circulating tumor cells.Science 2013, 341 (6151), 1186−1188.(43) Li, Q.; Lee, G. Y.; Ong, C. N.; Lim, C. T. AFM indentationstudy of breast cancer cells. Biochem. Biophys. Res. Commun. 2008, 374(4), 609−613.(44) Yang, Y.; Li, M.; Sun, X.; Zhou, C.; Wang, Y.; Wang, L.; Chen,L.; Liang, Z.; Zhu, L.; Yang, H. The selective cytotoxicity of DSF-cuattributes to the biomechanical properties and cytoskeleton rearrange-ments in the normal and cancerous nasopharyngeal epithelial cells.Int. J. Biochem. Cell Biol. 2017, 84, 96−108.
(45) Hu, F. Q.; Wei, L.; Zhou, Z.; Ran, Y. L.; Li, Z.; Gao, M. Y.Preparation of biocompatible magnetite nanocrystals for in vivomagnetic reasonance detection of cancer. Adv. Mater. 2006, 18 (19),2553−2556.(46) Wang, H.; Zheng, L.; Peng, C.; Guo, R.; Shen, M.; Shi, X.;Zhang, G. Computed tomography imaging of cancer cells usingacetylated dendrimer-entrapped gold nanoparticles. Biomaterials2011, 32 (11), 2979−2988.(47) Artemov, D.; Mori, N.; Okollie, B.; Bhujwalla, Z. M. MRmolecular imaging of the Her-2/neu receptor in breast cancer cellsusing targeted iron oxide nanoparticles. Magn. Reson. Med. 2003, 49(3), 403−408.(48) Binks, B. P.; Clint, J. H.; Whitby, C. P. Rheological behavior ofwater-in-oil emulsions stabilized by hydrophobic bentonite particles.Langmuir 2005, 21, 5307−5316.(49) Huang, X.; Jain, P. K.; El-Sayed, I. H.; El-Sayed, M. A. Goldnanoparticles: interesting optical properties and recent applications incancer diagnostics and therapy. Nanomedicine 2007, 2 (5), 681−693.(50) Wee, J. T. S.; Ha, T. C.; Loong, S.; Qian, C. N. Isnasopharyngeal cancer really a” Cantonese cancer”? Aizheng 2010, 29(5), 517−526.(51) Liu, K.-K. Deformation behaviour of soft particles; a review. J.Phys. D: Appl. Phys. 2006, 39, R189.(52) Friedman, J. H. Regularized discriminant analysis. J. Am. Stat.Assoc. 1989, 84, 165−175.(53) Wei, K. R.; Zheng, R. S.; Zhang, S. W.; Liang, Z. H.; Ou, Z. X.;Chen, W. Q. Nasopharyngeal carcinoma incidence and mortality inChina in 2010. Chin. J. Cancer 2014, 33 (8), 381.(54) Ignatiadis, M.; Rothe, F.; Chaboteaux, C.; Durbecq, V.; Rouas,G.; Criscitiello, C.; Metallo, J.; Kheddoumi, N.; Singhal, S. K.;Michiels, S.; Veys, I.; Rossari, J.; Larsimont, D.; Carly, B.; Pestrin, M.;Bessi, S.; Buxant, F.; Liebens, F.; Piccart, M.; Sotiriou, C. HER2-positive circulating tumor cells in breast cancer. PLoS One 2011, 6(1), No. e15624.(55) Andersson, H.; van den Berg, A. Microfluidic devices forcellomics: a review. Sens. Actuators, B 2003, 92 (3), 315−325.(56) Lekka, M. Discrimination between normal and cancerous cellsusing AFM. Bionanoscience 2016, 6 (1), 65−80.
