Microfluidic sorting and multimodal typing of cancercells in self-assembled magnetic arraysAntoine-Emmanuel Salibaa,2, Laure Saiasa, Eleni Psycharia,3, Nicolas Minca, Damien Simonb,4, François-Clément Bidardc,Claire Mathiotd, Jean-Yves Piergac,e, Vincent Fraisierf, Jean Salamerof, Véronique Saadag, Françoise Faraceg,h,Philippe Vielhg, Laurent Malaquina, and Jean-Louis Viovya,1

aInstitut Curie, Centre National de la Recherche Scientifique, Université Pierre et Marie Curie, Unité Mixte de Recherche 168, 75005 Paris, France;bLaboratoire de Physique Statistique, Ecole Normale Supérieure, 75005 Paris, France; cInstitut Curie, Département d'Oncologie Médicale, 75005 Paris,France; dInstitut Curie, Département de Biologie des Tumeurs, 75005 Paris, France; eUniversité Paris Descartes, 75006 Paris, France; fInstitut Curie, Cell andTissue Imaging platform, Nikon Imaging Centre, Centre National de la Recherche Scientifique, Unité Mixte de Recherche 144, 75005 Paris, France;gInstitut de Cancérologie Gustave Roussy, Département de Biologie et de Pathologie Médicales et Laboratoire de Recherche Translationnelle, 94805Villejuif, France; and hInstitut National de la Santé et de la Recherche Médicale, Unité 981, 94805 Villejuif, France

Edited* by Robert H. Austin, Princeton University, Princeton, NJ, and approved June 11, 2010 (received for review February 12, 2010)

We propose a unique method for cell sorting, “Ephesia,” usingcolumns of biofunctionalized superparamagnetic beads self-assembled in amicrofluidic channel onto an array of magnetic trapsprepared by microcontact printing. It combines the advantages ofmicrofluidic cell sorting, notably the applicationof awell controlled,flow-activated interaction between cells and beads, and those ofimmunomagnetic sorting, notably the use of batch-prepared, wellcharacterized antibody-bearing beads. On cell lines mixtures, wedemonstrated a capture yield better than 94%, and the possibilityto cultivate in situ the captured cells. A second series of experimentsinvolved clinical samples—blood, pleural effusion, and fine needleaspirates— issued from healthy donors and patients with B-cellhematologicalmalignant tumors (leukemiaand lymphoma). The im-munophenotype andmorphology of B-lymphocytes were analyzeddirectly in the microfluidic chamber, and compared with conven-tional flow cytometry and visual cytology data, in a blind test. Im-munophenotyping results using Ephesia were fully consistent withthose obtained by flow cytometry. We obtained in situ high resolu-tion confocal three-dimensional images of the cell nuclei, showingintranuclear details consistent with conventional cytological stain-ing. Ephesia thus provides a powerful approach to cell capture andtyping allowing fully automated high resolution and quantitativeimmunophenotyping and morphological analysis. It requires atleast 10 times smaller sample volume and cell numbers thancytometry, potentially increasing the range of indications and thesuccess rate of microbiopsy-based diagnosis, and reducing analysistime and cost.

lab-on-a-chip ∣ magnetic beads ∣ cell sorting ∣ cancer diagnosis

Cell-based screening is a major tool in medicine and pharma-ceutical research. In oncology and haematology, the morpho-logical and phenotypic typing of cancer cells is already usedroutinely for diagnostic and therapeutic purposes. This typingis made all the more relevant by the development of “persona-lized medicine” approaches, that of new anticancer drugs target-ing specific mutations, such as trastuzumab or rituximab, whichrequire specific tumor cell typing regarding HER2 and CD20expression, respectively (1), and by the discovery of “cancer stemcells” (2). Unraveling the detailed molecular characteristics ofcancer cells from clinical samples will thus play a paramount rolefor the progress of cancer research, diagnosis and treatment.Regarding clinical sampling, the current trend towards minimallyinvasive diagnostic procedures follows two distinct tracks: a firstone aiming at the identification and molecular typing of circulat-ing tumor cells (CTC) in peripheral blood (3), using notably theCell-Search® (4) system; and a second one using microbiopsiessuch as fine needle aspirates (FNA) (5, 6). FNA alleviate manyproblems of conventional surgical biopsies, involving lighteranesthesia and clinical procedures. They offer better patient

comfort, can be more easily repeated if necessary, and are alsomore time and cost effective.

The methods mostly used for the characterization of cancercells in FNA are direct optical cytology (7), Laser ScanningCytometry (8), and Flow Cytometry (FC) (9, 10). Most oftencytology and cytometry must be combined. FC is routinely usedin haematology, in which sample size is not an issue (11), butits use on FNA remains challenging. Detailed classification mayrequire sample volumes up to 100 μL and with a high cellularity(from 100,000 to 1,000,000 cells, typically) and highly skilledpathologists, preferably under the guidance of computer-assistedultrasound tomography. Even in such optimal (but expensiveand not always available) environments, typically 10% to 30% ofFNA-cytometry analyses are inconclusive andmust be followed bysurgical biopsies (8, 9, 11). Failures can be due to samplingdifficulties, a low intrinsic cellularity of the sampled lymph node,or a specific fragility of the cancer cells to the strong shear stressinvolved in FC, e.g., in the case of large B-cells lymphomas (10).There is thus a strong need for methods less demanding regardingsample size and cellularity, in order to generalize the use of FNAand reduce the need for macroscopic biopsies.

Here, we propose a unique microfluidic strategy to overcomethe above limitations. Microfluidic techniques are rather newchallengers for cell sorting (12, 13), but they offer interestingperspectives. These techniques are particularly powerful whencombined with mAb developed against membrane proteins, inorder to capture cells with a high specificity (14). Using biofunc-tionalized arrays of pillars, Nagrath et al. (15) captured CTC fromthe blood of patients with various cancers, with a yield apparentlymuch higher than that of current cell sorting approaches. Thishigher yield was claimed to be a consequence of the hydrody-namic interactions between the cells and the surface bearingthe antibodies, which is stronger and more long lived in micro-fluidic post arrays than in batch sorting. This system, however,requires a direct modification of the pillar’s surface in each chip

Author contributions: A.-E.S. and J.-L.V. designed research; A.-E.S., L.S., E.P., and N.M.performed research; D.S., F.-C.B., C.M., J.-Y.P., V.F., J.S., V.S., F.F., and P.V. contributednew reagents/analytic tools; A.-E.S., V.S., P.V., and L.M. analyzed data; and A.-E.S. andJ.-L.V. wrote the paper.

Conflict of interest statement: Part of the methodology described is dependent from CNRSpatent WO9823379 and Curie Institute patent application PCT/FR2009/051942.

*This Direct Submission article had a prearranged editor.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: Jean-Louis.Viovy@curie.fr.2Present address: EMBL, Meyerhofstraße 1, 69117 Heidelberg, Germany.3Present address: INSERM, 102 Rue Didot, F75014 Paris, France.4Present address: Laboratoire de Probabilités et Modèles Aléatoires, Université Paris VI,4 Place Jussieu, 75252 Paris Cedex 05, France.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental.

14524–14529 ∣ PNAS ∣ August 17, 2010 ∣ vol. 107 ∣ no. 33 www.pnas.org/cgi/doi/10.1073/pnas.1001515107

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental

by antibodies, and the presence of thick and opaque pillarsrequires specific software development and limits the potentialfor optical characterization of the captured cells (16). A micro-fluidic diagnostic magnetic resonance device with low sampleconsumption was also recently proposed as an interesting alter-native to FC, but it does not allow morphological analysis (17).

The system proposed here combines the advantages of micro-fluidic and magnetic cell sorting, with minimal sample require-ments. At its heart it is an array of biofunctionalized super-paramagnetic particles, self-organized in a microchannel (Fig. 1).Performances were first characterized by positive sorting ofB-lymphocytes among a mixture of B and T lymphocyte cell lines.The captured cells were viable, and could be cultured in situ. Wethen performed the sorting, morphological characterization, andimmunophenotyping of B-cell malignancies from clinical sam-ples. The results showed 100% agreement in a blind-test protocolwith a combination of FC and cytological morphological analysis.

Design Principles, Operation, and Modeling“Ephesia” Technology. Superparamagnetic beads are extensivelyused in conventional immunomagnetic sorting. Upon application

of an external magnetic field, each bead develops a strong mag-netic moment parallel to the field. In previous magnetic sortingstrategies (e.g., 5, 18, 19), the cells to be captured were mixed witha suspension of beads, and a highmagnetic field gradient was usedto collect the beads either statically or in a flow.Here,we transposethis paradigm, and use the dipole-dipole interactions betweenbeads to create a posts array. Under moderate external fields(typically 10–30 mTesla), these interactions overcome Brownianmotions and induce the formation of chains oriented in the fielddirection (20). When the suspension is enclosed in a slit-likemicrochannel (Hele-Shaw cell) with its main walls perpendicularto a uniform field, an array of columns is formed with a hexagonalorder on short distances and glass-like order on long distances(21). The average center-to-center distance between two columns,and the distribution of pore sizes and column diameters, resultfrom a complex interplay between the thermodynamics of bead-bead interactions, the kinetics of bead chains formation, andthe interactions of the chains with the microchannel’s walls.The stability of this array decreases when the microchannelthickness and/or the column spacing increase. In practice, arrayswith a spacing of 2–5 μm are easy to achieve with a good reprodu-cibility, and were previously used in our laboratory for the separa-tion of DNA (22, 23). In contrast, arrays with column spacinglarger than 10 μm are irregular and fragile. Thus, spontaneouslyself-assembled magnetic arrays are not suitable for separatingeukaryote cells with a typical diameter ranging from 5 μm–20 μm.

Templated Self-Assembly Using Microcontact Printing. To overcomethe above limitation, we propose here to use a permanent mag-netic pattern with the desired organization, deposited at thebottom of the microchannel, to direct beads self-assembly(Fig. 1A). Several methods for microfabricatingmagnetic patternswere previously proposed (24, 25), but they require advancedfacilities. We propose here a simpler alternative, based on themicrocontact printing (26) of a water-based ferrofluid (“magneticink”) onto glass (SI Text and Fig. S1). These magnetic ink dotshave amagnetic volume susceptibility, 2.33, about 105 times higherthan that of glass or buffer. When an external uniform magneticfield is applied, they concentrate field lines and create localgradients that act as potential wells for the magnetic beadscontained in the supernatant liquid.

Microfluidic Control and Imaging. An integrated system was pre-pared in PDMS (Polydimethylsiloxane) following a processinspired from ref. 27. This system comprises a lower layer withmicrochannels for the transport of fluids and cells, and an upperlayer containing microchannels crossing the fluid transportchannels and acting as pinch valves (Fig. 1B). These valves openor close the inlet channels by application of an external pressure.To avoid flow pulses detrimental to the stability of the magneticcolumns, the valves and the flow of samples were controlled bya customized high resolution pressure controller (28) (Fluigent,Paris, see SI Text).

Conceptual and Practical Advantages. The cell-antibodies interac-tions in our system occur in hydrodynamic conditions very similarto those at play in micropillar arrays, so we expect to recover herethe high capture efficiency emphasized e.g., ref. 15, 29. Ephesiahas four further advantages: (i) Functionalized beads can beprepared in large quantities in a batch process, yielding a lowerproduction cost and easier quality control. Using modified beadsalso opens access to more elaborated antibody immobilizationprotocols, and to hundreds of commercially available microparti-cles, capitalizing the know-how of about 20 yr of research anddevelopment and clinical validation; (ii) The self-assembly processcan yield convenient microstructures with an aspect ratio out ofreach of the most sophisticated nanofabrication tools (30), offer-ing new possibilities for imaging. Notably, we could achieve post

Fig. 1. Principle and practical implementation of the Ephesia system.(A) Principle of magnetic self-assembly. A hexagonal array of magnetic ink ispatterned at the bottom of amicrofluidic channel. Beads coated with an anti-body are injected in the channel. Beads are submitted to Brownian motion.The application of an external vertical magnetic field induces the formationof a regular array of bead columns localized on top of the ink dots. (B) Twolevels PDMS integrated microchip. Channels were filled with colored water.Delivery and separation channels for the cells appear in yellow. Inlets portsappear in orange. The separation channel is the longer vertical branch. Thearea bearing magnetic posts is marked by the dotted white box. Channelsin the upper PDMS layer, controlling the opening and closing of the inletchannels, appear in blue. The green wire is a thermocouple for in situ controlof the temperature in the system. (Scale bar: 0.5 cm.) (C) Magnetically as-sembled array of columns of 4.5 μm beads coated with anti-CD19 mAb (spe-cifically retaining Raji B-Lymphocytes) (Movie S1). Typical column shapes areshown in the insets. (Scale bar: 80 μm.) (D) Optical micrograph of the columnsafter the passage of 1,000 Jurkat cells. No cell can be seen (Movie S2). (Scalebar: 80 μm.) (E) After the passage of 400 Raji cells, numerous ones arecaptured and rosetted on the columns (Movie S3). (Scale bar: 80 μm.)

Saliba et al. PNAS ∣ August 17, 2010 ∣ vol. 107 ∣ no. 33 ∣ 14525

MED

ICALSC

IENCE

SAPP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SM1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SM2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SM3

arrays with 50 μm height and 4.5 μm diameter (aspect ratio 12),allowing imaging with 100X, 1.4 NA objectives, and analysis withcommercial softwares, whereas the silicon devices in (15) onlyused 10X objectives, and even with such low aperture objectives,they required specific software developments to deal with itsimaging complexity (16); (iii) Our devices can be prepared by acombination of low resolution molding or embossing and micro-contact stamping, for which both low cost laboratory and highthroughput industrial machines exist. Thus, the chips can be massproduced at low cost (in contrast e.g., with the devices in ref. 15),which require for each chip a large area monocrystalline siliconand deep wet etching. (iv) Finally our system can accommodatesmaller volumes: our cell’s volume is around 0.5 μL and can beoperated with less than 10 μL total sample volume, whereas thecell in ref. 15 has an internal volume of 60 μL. These volumesare consistent with our different objectives, CTC for ref. 15,and FNA for us.

As compared to our earlier work on self-assembly (21–23), theuse of magnetic ink templates has two advantages: (i) Firstthe magnetic potential wells impose onto the magnetic columnsa perfectly predefined order, instead of the glass-like orderobtained at long range with spontaneous self-assembly (21);(ii) Second, the magnetic wells stabilize the array, and allow muchhigher flow rates.

Strategy Used for Validation on Model Cell Lines and Clinical Samples.Two types of experiments were designed to validate the potentialof this system for cell sorting. First, yield, specificity, and in situpostcapture culture were evaluated using mixtures of cells fromculture cell lines: T (Jurkat cell line, ATCC TIB-152) and B (Rajicell line, ATCC CCL-86) lymphoid cells.

Second, to assess the potential of Ephesia on clinical samples,we focused on the classification of B-cell malignancies, a majortype of hematological cancers. B-cell malignant tumors generallydevelop in lymph nodes. Cancerous B-cells may circulate in blood,populate serosal membranes, and/or give rise to effusions such aspleural effusion. The classification of B-cell cancers is complex,and diagnosis is secured by a combination of clinical observation,morphology, immunophenotyping and genetic characterization(World Health Organization (WHO) classification) (31). Briefly,diagnostic steps starts with a morphological characterizationaimed at determining the presence of cancer cells. Upon positiveresponse, a further immunophenotypic characterization is per-formed (most often by FC), to identify specific antigen expressionat the cell surface. This procedure may require several sequentialimmunophenotyping experiments, and in the most unfavorablesituations several hundred μL of sample, exceeding the volumeof a FNA. It thus constitutes a good challenge for evaluating anEphesia system for multimodal, high-content morphotyping andphenotyping of cancer cells from minute samples.

Theoretical Modeling of the Capture Process. In order to extract fromexperimental data microscopic parameters useful for further op-timization, a theoretical model of cell capture in our device wasdeveloped. The array is subdivided into a series of rows. A masterequation predicting the time-dependent occupancy of each row ofobstacles is expressed as a function of cells flow rate, a geometricalfactor (collision probability) and a microscopic, biological factor(capture efficiency) is derived (see SI Text for the full model). Inthe limit of low occupancy (short times or samples with only a fewpositive cells), themodel simplifies greatly, and the number of cellsin row k after the entrance of the Nth cell, mk;N , per row can beapproximated by a simple exponentialmk;N≅c expð−αukÞ where cis a constant depending on the total number of cells, u is thecollision probability per row (a hydrodynamic parameter) and αthe capture efficiency. A comparison between the theory andexperimental data was used to determine the microscopic captureefficiency, in the general case and in the “rare cells” limit.

Results and DiscussionCharacterization of the Magnetic Template. The array of magneticdots was prepared on a 0.5 × 10 mm2 area. Optical imagingshows that a regular, uniform, and essentially defectless array isachieved over the whole surface (Fig. 1C). Electronic microscopyshows that dots adopt a reproducible cone-like shape (Fig. S2).Their diameter, 5� 1 μm, is significantly smaller than the initialsize of the pins in the stamp. This size reduction and cone-likeshape are consequences of hydrodynamics flows in the ink andpartial dewetting during the lift-off process. They allow for a bet-ter alignment of the magnetic column on the spot’s center. Atom-ic Force Microscopy (AFM) measurements indicate that the topof the spot is 475� 25 nm high. After a baking step overnight at150 °C the arrays could be washed and used repetitively for days.

Formation and Characterization of the Three-Dimensional Array ofMagnetic Beads. The magnetic beads, in suspension in PBS con-taining 0.1% BSA at a concentration 3 mg∕mL are introducedinto the separation channel under microfluidic control. Flow isthen arrested, and the magnetic field is applied (Fig. S3 andMovie S1). The beads self-organize into columns over the ferro-fluid dots, as expected (Fig. 1C). Surprisingly, when a moderatevelocity (less than 20 μm∕s) was applied to the array and themagnetic field was turned off, the columns kept their cohesionand remained attached to the magnetic dots. We attribute thisattachment to field-mediated bead-bead adhesion promoted bythe interpenetration of polymer or protein layers at the surfaceof the beads, under the pressure induced by dipole-dipole inter-actions (30). In the present microchannel configuration, undermoderate flow these cohesive columns align in the flow at thebottom of the channel. We could then characterize the columnsin side-view (Fig. 1C, insert). We studied the dependence of thestructure of the magnetic columns, upon the initial concentrationof the beads suspension. For the beads and arrays used here(4.5 μm Dynal beads on a hexagonal array with a 40 μm cen-ter-to center spacing) the optimal beads concentration in the sus-pension is 3 mg∕mL. For this concentration, columns have aheight equal to the channel’s thickness, and are made of singlealigned beads with only a few defects. At higher concentration,thicker columns are obtained (Fig. S4). When magnetic columnsare assembled under optimal conditions (see SI Text), they startto detach for average flow velocities between 800 μm∕s and1 mm∕s. This resistance is a considerable improvement withregards to nontemplated magnetic arrays, which are destabilizedfor fluid velocities around 20 μm∕s. For flow velocities above1 mm∕s, no particle remains in the channel, but the templateof magnetic dots is undamaged.

Quantification of Cell Sorting. Magnetic beads with a diameter of4.5 μm coated with anti-CD19 mAb (Dynabeads, Dynal) wereused and self-assembled as described above. Cell suspensionsof mixtures of T (Jurkat cell line) and B (Raji cell line) lymphoidcells (see SI Text) were flown in the array at a concentrationof 2.106 cells∕mL.

Cell capture was first studied as a function of flow rate. After aprescribed capture time, the array was washed with free buffer toeliminate nonattached cells. The fraction of captured cells wasquantified by visual counting. No significant change in the captureefficiency was observed for flow velocities from 50 μm∕s to700 μm∕s (see SI Text and Fig. S5). This result is consistent withthe optimal range of flow velocities reported earlier (15, 29). Wecould not investigate capture efficiency at higher flow rates, be-cause then the magnetic array is dragged away. The experimentsdescribed belowwere performed at flow velocities 100� 10 μm∕s.At this flow rate, visual observation is easy, and we did not observeany column detachment, even when the columns are saturatedin cells (in this situation, columns start to detach from the postsat flows around 500 μm∕s instead of 1;000 μm∕s in the absence

14526 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1001515107 Saliba et al.

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SM1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SF5

of cells, but cells remain attached to the beads). In our device, witha channel width 500 μm and a channel height 50 μm, a 100 μm∕sflow corresponds to a flow rate of a few μL∕min and a throughputin the range of ten to hundred cells∕s.

Two micrographs taken at the entrance of the magnetic arrayafter the passage of pure populations of Jurkat and Raji cells,respectively, are shown in Fig. 1 D, E. Introduced Raji cells wereobserved as “rosettes” on the columns whereas no cell was seen inthe array after passage of a population of 1,000 Jurkat cells,qualitatively demonstrating the specificity of the system (visuali-zation in Movies S2 and S3).

Cell sorting from a mixture was also evaluated regarding purityand yield (see SI Text). Purity is defined here as: (number ofpositive cell captured)/ (total number of cells captured). The yieldrepresents the total number of cells of a given type captureddivided by the total incoming number of cells of that type.Table S1 summarizes the results for Raji and Jurkat cell popula-tions, starting from initial ratios of Raji to total: 33%, 10%, and1%, respectively. As expected it decreased by increasing the ratioof Jurkat cells (which play here the role of “contaminant” nega-tive cells) in the sample, from typically 97� 1% for 33% positivecells, to a typical 85� 5% at one positive cell per 100. For bothpositive and negative sorting, yields were extremely high. Rajicells were captured with a yield of 97� 2% whereas Jurkat cellsexited the array with a yield >98%. Fig. 1E also shows that eachtrapped cell is visible individually, contrary to routine bulk experi-ments where cells are often entrapped in aggregates (32).

For comparison with earlier work, we also performed experi-ments using anti-EpCAM Dynal beads and MCF7 breast cancercells spiked in a large excess of endothelial cells (SI Text). Thecapture yield was 80� 20%, comparable with those previouslyreported for the CellSearch system® (80%) (4) and the Nagrathsystem (>65% in (15), and 42%–64% on clinical samples in (16)).

Comparison with Theory and Evaluation of Microscopic Parameters.The model was fitted to experimental data using a unique valuefor r (capture efficiency) and L (number of capture sites perrow) for all data. Theoretical and experimental results agree well(Fig. 2), allowing an evaluation of the coefficients L and r. Thefirst cells entering the channel are preferentially captured on thefirst rows of the array. While increasing the number of enteringcells, the reduction in the number of free capture sites becomessignificant, and more and more cells cross first rows and are cap-tured downstream. The saturation of the first rows of columns

was just reached in our experiments, and simulations indicate thata continuing flow of the sample through the array should yield apropagating sigmoidal-like front, the amplitude of which is givenby the total number of capture sites. Comparing the best-fit valueL ¼ 25 with the actual number of columns in a row, 10, one candeduce that each column is capable of capturing 2–3 cells in aver-age, in consistency with visual observation in saturated conditions(Fig. 1E). The best-fit parameter r, 0.2� 0.05 for all datasets, isequal in this model to the product of the probability of collision ina row by the probability of attachment upon collision. The accu-rate calculation of the collision probability per row is delicate (seeSI Text), but knowing the average size of lymphocytes, 10 μm, thebeads size, 4.5 μm, and the beads spacing in a row (40 μm center-to-center) it should lie between 0.28 and 0.36. This yields esti-mated values of the probability of attachment per collision (be-tween 0.55 and 0.8). This excellent capture efficiency is consistentwith visual observations, in which the first collision most of thetimes leads to attachment. This efficiency, and the fact thatthe sample is completely depleted from positive cells long beforethe microchannel’s end, suggest that the system will be robust tolimited changes in the array geometry, or to defects in the col-umns (see SI Text for further discussion).

In situ Culture. The cells captured on the column could be keptin physiological conditions over long time periods. To that aim,complete and buffered (CO2-independent) growth medium (Invi-trogen) were continuously infused in the channel after capture.Temperature was maintained between 36 °C and 37 °C. Raji cellsattached to the columns were viable as demonstrated by theirability to move and divide (Fig. 3). We could observe mitosis after12 h of culture, consistent with the population doubling time of thiscell line. The mother and daughter cells remain attached to thecapture column, but they eventually displace individual beadsand modify the structure of the magnetic columns. The cell on theleft (Fig. 3) has extracted one bead from the column, and drives itaround between different frames. This behavior is consistent withthe forces involved in cell adhesion and cytoskeleton motions,in the nN range (33), since the magnetic cohesive force of ourcolumns is rather in the upper pN range. Ephesia is thus a straight-forward microfluidic method to cultivate nonadherent cell imme-diately after their sorting, opening the route to fundamental cellbiology research directly on cancer cells from patients.

Characterization of B-Cell Malignancies from Clinical Samples. Themethod was validated regarding different types of lymphomasand leukemia: chronic lymphocytic leukemia (CLL) (n ¼number of patients ¼ 4), mantle cell lymphoma (n ¼ 1) and fol-licular lymphoma (n ¼ 2), and two healthy volunteers (Table S2

Fig. 2. Cell capture profile in the channel. x-axis: magnetic columns row,numbered from the capture microchannel entrance. y-axis: Number of cellscaptured in each row (sum over all the columns in the row). Data from threeexperiments with different total numbers (N) of cells injected are plotted astriangles (N ¼ 214), circles (N ¼ 320) and squares (N ¼ 400). Theoreticalbest-fit values for eachnumber of cells are represented by full lines (all derivedwith a single set of parameters r ¼ 0.18 and L ¼ 25). The last two curves forN ¼ 700andN ¼ 1;000 are extrapolations for larger numbers of cells, showinga saturation plateau progressing with time from the entrance to the outlet.

Fig. 3. In-situ culture and observation of Raji cells in the magnetic array.Numbers under the frames represent the time elapsed since capture, inhours:minute. The cell with the white arrow undergoes cell division duringthis time lapse. The cell on the left has extracted a bead from the magneticcolumn (black arrow), and is moving it around. (Scale bar: 40 μm.)

Saliba et al. PNAS ∣ August 17, 2010 ∣ vol. 107 ∣ no. 33 ∣ 14527

MED

ICALSC

IENCE

SAPP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SM2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SM3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=ST2

and Fig. 4). Samples were analyzed in parallel in a blind protocol,by a combination of cytology and FC on the one hand, and by asingle step capture and image analysis using the Ephesia systemon theother hand,without prior knowledgeof the diagnosis. Threedifferent types of clinical samples were involved: blood (n ¼ 7),pleural effusion (n ¼ 1), and FNA (n ¼ 1) from lymph nodes.Capture of B-cells in Ephesia was performed using anti-CD194.5 μm Dynabeads. 10 μL of sample were used for each analysis.The whole array of cells-bearing beads can be imaged in situ byepifluorescence, or immobilized by agarose, allowing for the trans-port and ex-situ imaging of the whole chip by confocal fluores-cence. A library of images was prepared and analyzed for eachsample (see SI Text). Typical “vignette” images from differentclinical samples, acquired ex-situ on a A1RNikon confocal micro-scope, are provided in Fig. 4. Ephesia provides the same majornucleus morphological characteristics as classical cytologicalstaining. Normal lymphocytes and B-cell malignant cells fromCLL (Fig. 4A,B) are similar in cytological examination. This resultis also observed in Ephesia, with bright field images showing asimilar cell diameter, andHoechst staining showing a nucleus witha chromatin mainly located at its periphery and absent from thecenter in both normal and CLL cells. Typical morphological

features of nuclei observed in mantle and follicular lymphoma(nucleoli and fragmented-like nucleus respectively) are also visibleby confocal microscopy (Fig. 4 C, D). Movie S4 emphasizes thesenuclear characteristics by presenting the stack images of trappedDoHH2cells (a cell line of a follicular lymphoma (seeSIText)) thatexhibit large and multiple nucleolus and nucleus fragmentation.Antigen expression analysis shows a strong correlation with FCdata. Healthy B-lymphocytes are CD10 negative and present aminor population of CD5 positive cells. This CD5 expression islinked to the B1 subset of B-lymphocytes (34). Malignant B-cellshomogeneously express CD5 in CLL and mantle lymphoma butare CD10 negative. We observe the localization of the antigenat the membrane in these two latter conditions. As in FC data,follicular lymphoma malignant cells express CD10 but not CD5.In this case, fluorescence appears localized at the membraneand in the cytoplasm at the center of the cell. This peculiar topo-logical distribution for CD10 might be due either to a plasmamembrane invagination or to labeling of cytoplasmic CD10 afterits internalization from the membrane. Because of the labelingprocedures (cells are labeled first and then fixed) these twooptionsare plausible, and will require further investigation, since this kindof information is not provided by FC.

Fig. 4. Examples of normal and malignant B-cell analysis by cytological staining, flow cytometry, and Ephesia system. Each line represents one subject. Col-umns represent, from left to right: (i) patient status; (ii) cytological observation (May Grunwald Giensa (MGG) coloration); (iii and iv) flow cytometry gated onlymphoid population data representing the intensity of CD19 versus CD5 and CD10. The color code distinguishes non B (red) and B CD19þ (green) lymphocytes;(v to viii): Ephesia representative confocal images selected from an image library for each patient. Subsequent images are, respectively: vi: equatorial z-planebright field image, with indication of cell diameter; (vii): equatorial z-plane nucleus image (Hoechst staining); (vii): integration of three z confocal planescentered on the cell’s center, for fluorescent anti-CD10 (yellow); (viii) same as (vii) for anti-CD5 (red); in each image, the dashed circle represents the magneticbead localization; (ix) a quantification of CD10 versus CD5 fluorescence intensity extracted from confocal images of trapped CD19 population; a red circlepoints to the data from the cells presented in vignettes (v) to (viii). Note that these datasets are to be compared with the green dots of flow cytometry data(columns (iii) and (iv)) only, because CD19- cells are not captured in Ephesia. A, first line: Healthy subject (identified as n°1 in Table S2) shows, among the CD19þcells in FC, a majority of CD5-/CD10- cells, and minor populations identified as CD5þ or CD10þ. Ephesia reproduces this trend with minor populations as CD10þand CD5þ. The nucleus has a regular outline, both in MGG (col. ii) and Hoechst staining in Ephesia, with a preferred chromatin localization at the periphery ofthe nucleus. B, line 2: Patient with chronic lymphocytic leukemia (n°3 in Table S2) shows a vast majority of CD5þ and CD10- cells, according to both flowcytometry and Ephesia fluorescence quantification. Ephesia additionally shows CD5 located at the membrane. C, line 3: Cells from patient with Mantle-celllymphoma (n°7 in Table S2) show a larger diameter than normal B-cell and nucleolus on cytological observation (arrows); CD19þ cells are CD5þ and CD10- in FCdata. Ephesia reproduces these results showing intranuclear nucleolus (arrow), no expression of CD10 and a bright membrane expression of CD5. D, line 4: Cellsfrom patient with follicular lymphoma (n°9 in Table S2) show a fragmented nucleus on cytological observation (arrow). FC shows that B-cells even if CD19 areexpressed at low levels, present a high expression level of CD10 but not CD5. Ephesia reproduces these results showing cleaved nucleus (arrow), no expressionof CD5 and an expression of CD10, localized at the membrane and in the cytoplasm. (Scale bar: 5 μm.)

14528 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1001515107 Saliba et al.

Dow

nloa

ded

by g

uest

on

June

22,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=SM4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001515107/-/DCSupplemental/pnas.1001515107_SI.pdf?targetid=ST2

ConclusionWe proposed a unique approach to sort cells according to surfacebiomarkers, Ephesia, which inverts the usual paradigm of immu-nomagnetic sorting, in which magnetic particles capture cells inbatch, and are then driven to a capture zone across the sampleliquid. Here, the magnetic beads are immobile, and the sampleflows across them. Ephesia yields a capture efficiency above 94%on cell lines. Its cell characterization capacity was evaluated onclinical samples in blind tests. Ephesia allows confocal trans-mission and fluorescence microscopy, and a combination of mor-phological analysis and immunophenoptyping on the same cells.Quantitative fluorescence analysis after labeling in situ yieldedan excellent agreement between the distributions of intensitiesobtained by this method and by FC. Morphological analysis inthe Ephesia system was also fully consistent with the analysisof conventional cytological spreads by pathologists. Clinicalstudies clearly emphasize the need to combine FC with cytologyfor diagnosis (9). These two levels of information are provided ina single run by Ephesia, and thanks to microfluidics the wholeprocess can be automated into a compact and reasonably pricedinstrument. These advantages may be critical for acute diseasessuch as Burkitt lymphoma (35), for which diagnosis is a matter ofhours, and addressing patients to highly specialized centers maynot be an option. In addition, the power of Ephesia technologyfor high-content imaging opens the route to more elaborate phe-notyping approaches, using for instance differences in the topo-logical distribution of biomarkers or their colocalization, and alsoto in situ genetic analysis, using e.g., FISH. Ephesia also offersa flexible platform to perform detailed cell biology studies on can-cer cells directly captured from clinical samples, and thus over-come some of the limitations of studies on cell lines. Last but notleast, this system requires an initial sample volume typically 10

times smaller than FC, and a number of cells analyzed typicallybetween 10–100 times smaller, thanks to the low dispersion andhigh-content information of the images. Ephesia should thusmake FNA-based diagnosis more robust to suboptimal or lowvolume samplings, and expand its use towards samples with asmall cellularity or a large excess of normal cells with reducedreagents consumption.

Here, antigen quantification was performed in a semimanualprocedure. This analysis procedure limited the number of samplesand cells studied. However, this limitation will be easily alleviatedin the future with automated image acquisition and analysissoftware. Thanks to the small number of cells to be analyzed,computing time will be short, and Ephesia should considerablyreduce the expert’s time needed for final validation of the data.In a longer term perspective, finally, the fluidics of the systemcould be redesigned to permit CTC screening from large volumesof peripheral blood. In that application, we believe that the Ephe-sia technology could bring significant advantages over earliermethods using microfabricated posts, notably a reduced produc-tion cost, a smaller footprint, the possibility to perform biofunca-tionalization in large batches, and better imaging capacities.

ACKNOWLEDGMENTS. We thank J. Goulpeau (FLUIGENT) V. Studer (EcoleSupérieure de Physique et Chimie Industrielle de Paris), C. Gosse (LaboratoirePhotonique et Nanostructures), C. Hivroz and S. Coscoy (Institut Curie), andL. Sengmanivong (Nikon Imaging Centre, Institut Curie) for assistance in flowcontrol, electron microscopy, microfabrication, cell culture, and imaging,respectively. Work supported in part by ANR (MICAD 08-BIOT-015-02project) and EU (CAMINEMS Project—NMP4-SL-2009-228980). A-E.S. andL.S. acknowledge fellowships from Direction Générale de l’Armement andInstitut National de Recherche contre le Cancer, respectively.Microfabricationand imaging were performed in the Laboratoire Physicochimie Curie cleanroom and in the Nikon-Curie Imaging Centre.

1. Hortobagyi GN (2005) Trastuzumab in the treatment of breast cancer. N Engl J Med353:1734–1736.

2. Bonnet D, Dick JE (1997) Human acute myeloid leukemia is organized as a hierarchythat originates from a primitive hematopoietic cell. Nat Med 3:730–737.

3. Pachmann K, et al. (2008) Monitoring the response of circulating epithelial tumorcells to adjuvant chemotherapy in breast cancer allows detection of patients at riskof early relapse. J Clin Oncol 26:1208–1215.

4. Allard WJ, et al. (2004) Tumor cells circulate in the peripheral blood of all majorcarcinomas but not in healthy subjects or patients with nonmalignant diseases. ClinCancer Res 10:6897–6904.

5. Jorgensen JL (2005) State of the art symposium: flow cytometry in the diagnosis oflymphoproliferative disorders by fine-needle aspiration. Cancer 105:443–451.

6. DeMay RM (2007) Practical principles of cytopathology (American Society for ClinicalPathology, Chicago, IL), Rev Ed, pp 131–139.

7. Pierga JY, et al. (2004) Clinical significance of immunocytochemical detection of tumorcells using digital microscopy in peripheral blood and bone marrow of breast cancerpatients. Clin Cancer Res 10:1392–1400.

8. Clatch RJ, Walloch JL (1997) Multiparameter immunophenotypic analysis of fineneedle aspiration biopsies and other hematologic specimens by laser scanningcytometry. Acta Cytol 41:109–122.

9. Meda BA, et al. (2000) Diagnosis and subclassification of primary and recurrentlymphoma. The usefulness and limitations of combined fine-needle aspirationcytomorphology and flow cytometry. Am J Clin Pathol 113:688–699.

10. Verstovsek G, Chakraborty S, Ramzy I, Jorgensen JL (2002) Large B-cell lymphomas:fine-needle aspiration plays an important role in initial diagnosis of cases whichare falsely negative by flow cytometry. Diagn Cytopathol 27:282–285.

11. Racila E, et al. (1998) Detection and characterization of carcinoma cells in the blood.Proc Natl Acad Sci USA 95:4589–4594.

12. Toner M, Irimia D (2005) Blood-on-a-chip. Annu Rev Biomed Eng 7:77–103.13. El-Ali J, Sorger PK, Jensen KF (2006) Cells on chips. Nature 442:403–411.14. Cheng X, et al. (2007) A microfluidic device for practical label-free CD4ðþÞ T cell

counting of HIV-infected subjects. Lab Chip 7:170–178.15. Nagrath S, et al. (2007) Isolation of rare circulating tumor cells in cancer patients by

microchip technology. Nature 450:1235–1239.16. Stott SL, et al. (2010) Isolation and characterization of circulating tumor cells from

patients with localized andmetastatic prostate cancer. Science Translational Medecine2:25ra23.

17. Lee H, Yoon TJ, Figueiredo JL, Swirski FK, Weissleder R (2009) Rapid detectionand profiling of cancer cells in fine-needle aspirates. Proc Natl Acad Sci USA106:12459–12464.

18. Miltenyi S, Muller W, Weichel W, Radbruch A (1990) High gradient magnetic cellseparation with MACS. Cytometry 11:231–238.

19. Adams JD, Kim U, Soh HT (2008) Multitarget magnetic activated cell sorter. Proc NatlAcad Sci USA 105:18165–18170.

20. Liu J, et al. (1995) Field-induced structures in ferrofluid emulsions. Phys Rev Lett74:2828–2831.

21. Minc N, et al. (2005) Motion of single long DNAmolecules through arrays of magneticcolumns. Electrophoresis 26:362–375.

22. Doyle PS, Bibette J, Bancaud A, Viovy JL (2002) Self-assembled magnetic matrices forDNA separation chips. Science 295:2237.

23. Minc N, et al. (2004) Quantitative microfluidic separation of DNA in self-assembledmagnetic matrixes. Anal Chem 76:3770–3776.

24. Inglis DW, Riehn R, Austin RH, Sturm JC (2004) Continuous microfluidic immunomag-netic cell separation. Appl Phys Lett 85:5093–5095.

25. Guo SS, Zuo CC, Huang WH, Peroz C, Chen Y (2006) Response of super-paramagneticbeads in microfluidic devices with integrated magnetic microcolumns. MicroelectronEng 83:1655–1659.

26. Xia Y, Whitesides GM (1998) Soft lithography. Annu Rev Mater Sci 28:153–184.27. UngerMA, Chou HP, Thorsen T, Scherer A, Quake SR (2000)Monolithic microfabricated

valves and pumps by multilayer soft lithography. Science 288:113–116.28. Futterer C, et al. (2004) Injection and flow control system for microchannels. Lab Chip

4:351–356.29. Chang WC, Lee LP, Liepmann D (2005) Biomimetic technique for adhesion-based

collection and separation of cells in a microfluidic channel. Lab Chip 5:64–73.30. Goubault C, Leal-Calderon F, Viovy JL, Bibette J (2005) Self-assembled magnetic

nanowires made irreversible by polymer bridging. Langmuir 21:3725–3729.31. Swerdlow SH, et al. (2008) WHO classification of tumors of haematopoietic and

lymphoid tissues (IARC, Lyon), 4th Ed.32. Choesmel V, Anract P, Hoifodt H, Thiery JP, Blin N (2004) A relevant immunomagnetic

assay to detect and characterize epithelial cell adhesionmolecule-positive cells in bonemarrow from patients with breast carcinoma: immunomagnetic purification ofmicrometastases. Cancer 101:693–703.

33. du Roure O, et al. (2005) Force mapping in epithelial cell migration. Proc Natl Acad SciUSA 102:2390–2395.

34. Martin F, Kearney JF (2001) B1 cells: similarities and differences with other B cellsubsets. Curr Opin Immunol 13:195–201.

35. Das DK (1999) Value and limitations of fine-needle aspiration cytology in diagnosisand classification of lymphomas: a review. Diagn Cytopathol 21:240–249.

Saliba et al. PNAS ∣ August 17, 2010 ∣ vol. 107 ∣ no. 33 ∣ 14529

MED

ICALSC

IENCE

SAPP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

June

22,

202

1