Experimental Hematology 32 (2004) 891–904

Enrichment of rare cancer cells through depletion of normalcells using density and flow-through, immunomagnetic cell separation

Oscar Laraa, Xiaodong Tonga, Maciej Zborowskib, and Jeffrey J. Chalmersa,c

aDepartment of Chemical and Biomolecular Engineering and cUniversity Cell Sorting Core and Analysis Laboratory, Heart and Lung Research Institute,The Ohio State University, Columbus, Ohio, USA; bDepartment of Biomedical Engineering, The Cleveland Clinic Foundation, Cleveland, Ohio, USA

(Received 6 November 2003; revised 6 July 2004; accepted 12 July 2004)

Objective. To develop a reliable technique to enrich for rare cells in blood suspensions usingonly negative selection steps including a flow-through immunomagnetic cell separations systemand by optimizing variables normally encountered during such enrichment processes.

Methods. A human breast cancer cell line was cultivated and spiked at a ratio of 1 cancercell to 105 total leukocytes in buffy coat or 1 cancer cell to 108 total cells in whole bloodsamples. The final, optimized process consisted of: a red cell lysis step, immunomagneticallystaining leukocytes with an anti-CD45 PE, anti- MACS sandwich, immunomagnetic sortingusing a flow-through system (QMS), and a final cell analysis step using either an automatedcell counter, filtration, and visual counting or a cytospin analysis.

Results. The final, optimized process produced a final enrichment of the rare cancer cells of5.17 log10 and an average, final recovery of 46%. It should be noted that a negative depletionprotocol was used (i.e., no labeling of the rare cancer cells was used).

Conclusions. To the authors’ knowledge, no examples in the literature exist of a 5.17 log10

enrichment of cancer cells in human blood using a negative depletion protocol. The closest ex-ample is a 4 log10 enrichment in which two positive magnetic cell separation steps were used(none were used in this study). Ongoing studies are investigating further modifications of theprecommercial, prototype flow-through immunmagnetic separation system to increase boththe enrichment and recovery rate. However, even at current performance levels, the present-ed process could significantly improve visual and molecular analysis of rare cells in blood. �2004 International Society for Experimental Hematology. Published by Elsevier Inc.

Detection of cancer cells present in body fluids is regardedas a potential indicator of prognosis and diagnosis in oncol-ogy [1]. Presence of cancer cells in the blood stream or bonemarrow has been correlated with clinical stage [2–4], patientsurvival after therapy [5], and tumor metastasis [6–9].

Immunocytochemistry, with minimum pre-enrichment,has been routinely used to assess the presence of circulatingcancer cells. However, due to its relatively high labor re-quirement, it requires several slides to be analyzed to accu-rately detect the cancer cells [10–12]; consequently, othertechniques, typically molecular biology based, have beenadvocated for this purpose. Analytic methods such as flowcytometry [6], RT-PCR and its variations [13–19], andMutant Allele-Specific Amplification, MASA [20], have allbeen successfully applied to detect cancer cells, in some

Offprint requests to: Jeffrey Chalmers, Ph.D., Koffolt Laboratories, 140W. 19th Avenue, Columbus, OH, 43210; E-mail: chalmers.1@osu.edu

0301-472X/04 $–see front matter. Copyright � 2004 International Society fordoi: 10.1016/ j .exphem.2004.07.007

cases to a sensitivity limit of at least one cancer cell permillion total cells.

To improve sensitivity for all of the above-listed tech-niques, it is highly desirable to apply an enrichment stepprior to the actual detection procedure. In all the reportslisted above, immunomagnetic cell separation has been oneof the techniques used. This technique is capable of at least1000-fold (3 log10) enrichment of the original sample, thusreducing the number of slides required for analysis or thenumber of false-positive outcomes from RT-PCR analysis.Table 1 summarizes some of these examples.

Batch commercial systems to enrich suspensions con-taining rare cells are readily available; however, concernsabout their limited load capacity and potential irreversibleentrapment of cells inside the column [35,36] have led usto develop and optimize “flow-through” cell separationdevices.

A flow-through, high-throughput quadrupole magneticcell sorter (QMS) has been developed whose operating

Experimental Hematology. Published by Elsevier Inc.

O.

Lara

etal./E

xperimental

Hem

atology32

(2004)891–904

892

Ta

ecovery ofTar arget cells Enrichment Sensitivity Reference

ImM 11.4% — — 2R 73.6% — — 12B 24.4% 6.8 — 21B 57.7% 10477 — 22

(Two IMS)R 84.3% — — 12G — 9 — 20C — — 1/106 MMC 20

ImC — — 1/108 MMC 10U 61.2% 15.3 — 11C — — 1/105 MMC 15C — — 10/1 mL PB 18H 22–32% — — 23C 9–45% — — 24U 57.3% 13 — 11H 36–62% — — 23

ImB 76% — 1/1 mL Blood 25P 73.7% — — 26B 60–90% — — 27P — — 1/7.5 mL Blood 28

ImB 20–60% — 1/106 MMC 29

ImB 66–89% 5–50 1/106 MMC 30

ImB — 200 — 31

FilT — — 1/1 mL Blood 32B 50–60% — 2/106 MMC 33

DeM 55.2% 303 — 34

ble 1. Some examples of rare Cancer cell enrichment in blood

Initial purityAntigen Detection (# target cells / R

get cell Cell sample Selection for IMS method #total cells t

munomagnetic: MACS (Miltenyi Biotec, Germany)elanoma CL SK-Mel-30 in PB Positive MCSP ICC 50–1000 per 50 mL PBenal carcinoma CL Caki-1 in PBL Positive CK ICC 1/10–1/107

reast cancer patients PB, BM, PBPC Positive HEA-125 ICC 6/106

reast cancer CL BT474 in PBL Positive CK-8 ICC, FC 4/108–4/105

enal carcinoma CL Caki-1 in PBL Negative CD45 ICC 1/10–1/107

astric cancer CL MKN-45 in PBL Negative CD45 FC 1/10–1/106

olorectal cancer CL SW-480 in PBL Negative CD45 RT-PCR 1/10–1/106

munomagnetic: Dynabeads (Dynal, Norway)ancer CL T47D in PBL Positive MUC-1 RT-PCR 1/103–1/108

rological malignant CL 5 CLs in PBL Positive Ber-EP4 ICC 4/105–4/108

olorectal cancer CL SW-480 in PBL Positive Ber-EP4 RT-PCR 1/104–1/106

olon cancer CL SNUC4 in PB Positive CEA RT-PCR 1–106/1 mL PBead and neck SCC PB, BM Positive Ber-EP4 ICC 1–25/106

ancer CL 4 CLs in PBL Positive Ber-EP4 ICC —

rological malignant CL 5 CLs in PBL Negative CD45 ICC 4/105–4/108

ead and neck SCC PB, BM Negative CD45 ICC 1–25/106

munomagnetic: Immunicon (USA)reast cancer CL SKBr3 in PB Positive EpCAM ICC, FC 50–4500 per 5 mL PBrostate cancer CL PC3 in PB Positive EpCAM FC 25–800 per 7 mL PBreast cancer CL Colo-205 in PB Positive EpCAM ICC, FC 1–200 per 5 mL PBrostate cancer PB Positive EpCAM FC —

munomagnetic: Magnetic deposition (CCF, USA)reast cancer CL MCF-7 in PBL Positive EMA ICC, FC 1/106–1/108

munomagnetic: QMS (OSU, USA)reast cancer CL HCC1954 in PBL Positive HER-2/neu FC 1/102

munoaffinity Column (CellPro, USA)reast cancer patients PB, BM Positive ICC, FC —

trationumor CL 5 CLs in PB Positive — ICC 1–3/1 mL PBreast cancer CL MCF-7 in PB Positive Ber-EP4 FC 2/104–1/106

nsity-gradient centrifugationDA-361 PB Positive — ICC, FC 0.73/103

O. Lara et al. /Experimental Hematology 32 (2004) 891–904 893

principle is based on the deflection, under the influence of amagnetic energy gradient, of immunomagnetically labeledcells into one stream and the unlabeled cells into a secondstream [37]. Results showing 99% purity and 86% recoveryhave been achieved using QMS in progenitor cell isolation[38] and CD4� and CD8� cell sorting. Additionally, Naka-mura et al. [30] reported the positive separation of a spikedbreast cancer cell line from human blood with a throughputof 3.29 × 105 cells/s and a recovery of 89% of the cancer cells.

A photograph of the complete QMS system, includingsyringe pumps, separation element, and magnet, is shownin Figure 1. The schematic drawing of the QMS separa-tion element is also shown in Figure 1. Cell suspension isfed to inlet a’ while carrier or sheath fluid is fed to inlet b’.Ideally, immunomagnetically labeled cells deflected due tothe magnetic energy gradient are collected from outlet b andnonmagnetic cells are collected from outlet a.

The QMS has several advantages over fluorescence-acti-vated cell sorting (FACS) and batch immunomagnetic cellsorting systems in applications to rare cell separations. Inaddition to the high throughput, demonstrated to be as highas 106 to 107 cells/s (positive selection mode of operation)[38], performance of this sorter can be tuned. Specifically, theperformance of the QMS can be predicted using a computermodel, experimental determination of the magnetophoreticmobility of labeled cells, and QMS operating conditions.These model predictions of the QMS performance have beenexperimentally verified for several systems including mag-netic microbeads, human hematopoietic stem cells, and im-munomagnetically labeled cancer cells spiked into humanblood [30,38–40]. To measure the magnetophoretic mobilityof cells, our laboratory has developed an instrument, referredto as a cell tracking velocimeter (CTV), which is capable ofmeasuring the magnetically induced velocity on a particle-to-particle basis [41].

While positive selection has proved effective for enrich-ment and isolation of rare, or not so rare, cells in a cellsuspension of various types, there are significant limitations.One of these major limitations is the lack of informationabout the phenotype of the target cell. In the case of arare, circulating cancer cell, one is assuming that the label,typically an antibody conjugate, can tag the cancer cell suf-ficiently, and specifically, to allow an acceptable separation.A study has recently been published which addresses thelabeling conditions and performance of a number of com-mercial cancer cell immunolabels on two human cancer celllines [42]. A second limitation of a positive selection modeof operation is the fact that one is labeling the very cellthat one wishes to analyze further. While in many cases ithas been demonstrated that binding of antibody magneticnanoparticle-conjugates to target cells does not have effectson cell function, it is more desirable to have an enrichmentstep that does not modify and/or bind the target cell. Thiswould then allow further molecular analysis of a nonmanipu-lated cell. Third, for some immunomagnetic cell separation

systems, a negative depletion of the nontargeted cells po-tentially provides overall superior operational performance.

In addition to a magnetic separation technique, many ofthe reported processes to enrich for a rare cancer cell includean initial depletion of the red blood cells, RBC. Two com-monly used techniques are reported: depletion of RBCusing a Ficoll-Hypaque cushion and the use of a red cell lysisbuffer. Despite its popularity, the Ficoll-Hypaque cushionperformance has been linked to the cell load and differencesin individual harvesting techniques. The Becton-Dickinsonweb page (www.bd.com/vacutainer/products/molecular/heparin/results.asp) states that when Ficoll-Hypaque separa-tion gradient is used, the expected recovery of mononuclearcells is around 68% of the original number of cells present.

Partridge et al. [23] reported that epithelial tumor cellssediment preferentially with the mononuclear cell fraction,although the authors also state that in some cases it waspossible to find more cancer cells in the granulocyte fractionthan in the mononuclear fraction. Kruger et al. [21] statedthat the detection of epithelial tumor cells is influenced by theerythrocyte depletion method. The authors report further thatmore cells are detected when red cell lysis is used insteadof Ficoll-Hypaque centrifugation.

Based on the research and potential clinical need to enrichfor very rare cells, such as cancer cells in blood, and thelimitations of the approaches outlined above, we are devel-oping and optimizing a process, based solely on negativeselection, to significantly enrich a blood sample for a rare cellof interest. The process presented in this paper has threeseparation stages: a density-gradient separation or lysis toremove erythrocytes, an immunomagnetic separation stepto remove leukocytes, and a filtration step to concentrate,collect, and evaluate the final cell suspension. Alternatively,the final step can be modified and replaced with a centrifuga-tion and/or direct cell lysis step for further molecular evalu-ations of the final, enriched cancer cell suspension.

Material and methods

Cell sourcesA breast carcinoma cell line (MCF-7) was grown to confluenceon Eagle medium (Invitrogen Corporation, Carlsbad, CA, USA)supplemented with 10% fetal bovine serum (FBS; JRH Biosci-ences, Lenexa, KS, USA) in a 75 cm2 tissue culture flask (Corning,Corning, NY, USA) and incubated at 37�C in 5% CO2. MCF-7cells were harvested using Accutase (Innovative Cell Technologies,USA) as per manufacturer instructions. Cell number was assessedusing hemocytometer, and cell viability was measured using a dyeexclusion method (Invitrogen Corporation, USA). Viability resultsshowed at least 90% healthy cells after detachment from cultureflask. Before the enrichment experiments, MCF-7 cells were filteredthrough a 20-µm nylon mesh in order to obtain a single cellsuspension.

Fresh blood (whole blood) was provided by healthy donors be-fore each enrichment experiment. Buffy coats of peripheral blood

O. Lara et al. /Experimental Hematology 32 (2004) 891–904894

a’ b’

ab

v

Outlet flow splitter

Inlet flow splitter

Core rod

Mag

net

Input syringe pump

Outlet Syringe Pump

Figure 1. Left, photograph of the quadrupole magnetic sorter (QMS) in operation. The syringe pumps used for injection and retrieval, as well as tubingconnecting the system, can be clearly seen. Right, diagram of the channel within the magnetic energy gradient.

from healthy donors were obtained from American Red Cross,Central Ohio Region. The concentration of leukocytes in bloodwas determined by hemocytometer using Unopette �Microcollec-tion system (BD Biosciences, San Jose, CA, USA). The total cellconcentration in blood, including erythrocytes and white blood cells,was measured using an automated particle counter (Multisizer II,Beckmann Coulter, Fullerton, CA, USA).

Density separationFor some experiments, density separation was conducted to re-move erythrocytes using Ficoll-Hypaque (Accurate Chemical and

Scientific Corporation, Westbury, NY, USA) with a gradient densityof 1.077 g/mL. Prior determination of MCF-7 cell density, 1.066g/mL, using a continuous Percoll gradient, indicates that cancercells will preferentially sediment with the leukocytes at the interfaceseparating plasma and the neutrophile fraction. Accordingly, isolat-ing peripheral blood leukocytes (PBL) from the interface will theo-retically yield the spiked MCF-7 cells in the leukocyte fraction.

Lysis of erythrocytesFollowing the basic protocol shown on the manufacturer’s website(http://www.zellbiophysik.fh-aachen.de/methoden/protocol1.html),

O. Lara et al. /Experimental Hematology 32 (2004) 891–904 895

typically 2 mL of fresh blood at the concentration of 1 MCF-7cell per 108 total cells (except for studies when a high concentrationof MCF-7 cells was used) was rapidly added to 40 mL fresh lysingbuffer (154 mM NH4Cl, 10 mM KHCO3, 0.1 mM EDTA), andincubated for 5 minutes at room temperature with occasional shak-ing. After 5 minutes of centrifugation at 300g, the cell pellets werewashed twice using the labeling buffer, and resuspended in the bufferto obtain the PBL suspension. In several runs, the lysis protocolwas repeated in an attempt to remove more RBC.

Immunological labelingCell preparation. Blood suspensions spiked with MCF-7 cells,from either density separation or red cell lysis, were washed twiceusing phosphate-buffered saline (PBS; JRH Biosciences, Lenexa,KS, USA) supplemented with 2 mM EDTA (Invitrogen Corpo-ration, Carlsbad, CA, USA) and 0.5% bovine serum albumin (BSA;Invitrogen Corporation, Carlsbad, CA, USA). After the final wash-ing, cells were resuspended to a final concentration of 40 × 106

cells/mL.

Antibodies used and labeling optimization studies. A double-stepimmunomagnetic staining technique was used to target the PBLfor the immunomagnetic separation step. The primary antibodyused was a mouse anti CD45 PE (Cat# IM 2078, Immunotech,Marseilles, France) and the secondary antibody was an anti-PEMACS microbeads (Miltenyi Biotec, Auburn, CA, USA). A thirdreagent, FcR blocking reagent (Miltenyi Biotec), was also tested.Flow cytometry measurements were made on a FACSCalibur flowcytometer (Becton Dickinson, San Jose, CA, USA) and magne-tophoretic mobility measurements were made on our previouslydeveloped, and reported, cell tracking velocimetry instrument [42].

Final cancer cell staining. To visually assist in the detection ofcancer cells after the final filtration step for some of the studiesconducted, an antibody specific to a human epithelial marker, mouseanti-human HEA FITC (Miltenyi Biotec, Auburn, CA, USA) wasused. The labeling of this marker, at a concentration of 10 µL ofantibody per 106 cells, occurred when cells were directly on thefiltration membrane. While cells stained with the anti-HEA FITCcan be easily identified under a fluorescence microscope, an anti-body-enzyme-dye procedure was used to further facilitate visuali-zation with a light microscope. Specifically, a rabbit-anti-FITCphosphatase alkaline conjugate was used as the secondary antibody(Cat # A4843, Sigma, St. Louis, MO, USA) and Fast Red TR-Naphtol AS-MX (Cat # F – 71, Sigma, USA) as the substrate forthe enzyme. The action of the alkaline phosphatase and Fast Redcreates an intense red-color dye that is easily detected using alight microscope.

Immunomagnetic cell separation stepThe design and construction of the QMS system has been describedpreviously [40]. The characteristics of the magnetic system include:a maximum magnetic field strength of the quadrupole (B0) of 1.37T and a mean force field strength (Sm) of 2.382 × 108 (T-A)/m2.In contrast to the previous studies, however, a precommercialseparation column (prototype # 10597) manufactured by SHOTINC. (Greenville, IN, USA) was used. This separation column hada length of 15.24 cm, an internal diameter of 0.6922 cm, and amaximum diameter of the core rod of 0.4661 cm. The inlet splitterouter diameter was 0.5128 cm and outlet splitter outer diameterwas 0.6058 cm.

Two modes of operation were evaluated: continuous mode, inwhich both the cell suspension (a’) and sheath flow (b’) flowedcontinuously through the channel during the separation; and semi-continuous mode, in which the cell suspension was continuous andthe sheath flow was zero. To operate the QMS in the continuousmode, total flow rates at the inlet and outlet remained constantusing three syringe pumps. Two pumps were used at the inlet: thefirst pump injected the sample and sheath carrier at the beginningof the operation until the sample has been injected into the column(a’ inlet). Upon completion of sample injection, flow entering thea’ inlet was switched to the second pump using a three-wayvalve which injected sheath fluid into the a’ inlet to flush remainingsample from the system and to maintain a constant flow rate atthe inlet. A third pump was connected to the outlets a and b tocollect sorted fractions. This third pump creates the transport laminathickness required for sorting while maintaining the total flow rateconstant. For the semi-continuous mode of operation, the systemwas prepared as above, and when the cell suspension sample wasinjected, both b and b’ flow was stopped.

Before performing an actual separation, the QMS column isfilled with degassed PBS buffer (PBS, 0.5 mM EDTA, and 0.5%BSA) so that no bubbles are trapped inside the device. Labeledcells are diluted to a final concentration in the range of 1 to 3 ×106 cells/mL using PBS buffer before injection. Syringes withsample, carrier, and empty syringes to collect sorted fractions aremounted and separation performed. After the sample has beencompletely injected, the three-way valve is switched to allow carrierfrom the second pump to enter the system and push the remainingsample within the tubing through the column.

Operating parameters of the QMSCell separations in the continuous mode in the QMS are governedby three adjustable parameters: total flow rate (Qt), inlet flow rateratio (Qa’/Qt), and outlet flow rate ratio (Qa/Qt). The first parameter,total flow ratio (Qt), is responsible for the residence time of thecell inside the column; therefore, (Qt) is directly responsible forthe radial displacement of an immunomagnetically labeled cell.The larger the flow rate, the less time the cell is subjected to themagnetic field and the less distance traveled in the radial directionas a result of the magnetic force. The remaining two parameters,inlet flow ratio and outlet flow ratio (Qa’/Qt, Qa/Qt, respectively),are responsible for creating the transport lamina that an immuno-magnetically labeled cell must cross radially to reach elution portb [43].

Final processing stepThe enriched fraction from the nonmagnetic outlet, stream a(Fig. 1), was processed in one of three ways: a) total cell recoveryand distribution with Coulter Multisizer II measurements and to-tal recovered volume, b) total cell recovery and analysis by filtra-tion, and c) cytospin. Figure 2 presents the size distribution of theperipheral blood lymphocytes and MCF-7 cells, as measured byCoulter Multisizer II, clearly indicating significant difference incell size. Cell samples analyzed by filtration were filtered using asyringe pump at a flow rate of 1 mL/min through a 5-µm pore sizediameter polycarbonate membrane (Whatman, Maidstone, Kent,UK). These retained cells were subsequently stained with anti-FITC alkaline phosphatase antibody and Fast Red substrate to assistin visually counting through a light microscope (Model BX60F,Olympus, Tokyo, Japan). Cytospin analysis was accomplished fol-lowing a typical protocol. Specifically, the enriched fraction from

O. Lara et al. /Experimental Hematology 32 (2004) 891–904896

0100020003000400050006000700080009000

Cell diameter (µm)

# E

ven

ts (

-)

0

20

40

60

80

100

120

10 15 20 25 30 35 40 45MCF-7 Cell diameter (µm)

# E

ven

ts (

-)

A

B

84 6 10 12 14

Figure 2. Cell size distribution of PBL and MCF-7. (A) represents PBL(including monocytes and neutrophils); (B) is the size distribution forMCF-7.

the QMS a outlet was centrifuged for 5 minutes at 300g, the pelletresuspended in 50 µL labeling buffer, and added to the cytospinfunnel for 2 minutes at 2000 rpm. The cells were then layeredevenly onto the glass slide, stained with Diff-Quik (DADE BehringAG, Dudingen, Switzerland) and air-dried for subsequent micro-scopic evaluation.

ResultsA commonly understood principle in chemical synthesis orseparation processes is the negative, multiplicative effect ofnonperfect recoveries and/or yields. For example, a five-stepseparation process in which each step has a 90% recovery ofthe targeted cell will result in an overall recovery at the endof the process of only 59%. Consequently, considerable effortwas placed on minimizing the number of steps and maximiz-ing the recovery of the spiked cancer cells in each step used.

Since centrifugation is a processing step that is used anumber of times to prepare samples for labeling and removalof specific agents (such as lysis buffer), a set of experimentswere conducted to determine the recovery of a typical cen-trifugation step. Specifically, nine experiments were con-ducted in which 2.00 × 106 MCF-7 cells were placed in a50-mL conical tube, resuspended in 40 mL fresh labelingbuffer (PBS, 2 mM EDTA, and 0.5% BSA), and centrifugedfor 5 minutes at 300g. The supernatant was then removed,and the cell pellet was resuspended in 2 mL labeling buffer.The cell concentration of this resuspended cell suspensionwas then determined with the Coulter Multisizer II. Theaverage recovery of these nine experiments was 93.4% witha relative standard deviation of 0.04.

Density-gradient separation optimization studiesTo assess the performance of the Ficoll-Hypaque density-gradient separation step, DGS, MCF-7 cells were seeded intoaliquots of blood from buffy coats, DGS was performed, andthe number of MCF-7 cells recovered was counted with theCoulter Multisizer II using a gate setting of 5.64 to 12.43µm for PBL and 12.95 to 40.18 µm for MCF-7 cells.

Figure 3 presents the recovery of MCF-7 cells, seededat a fractional concentration of 0.13%, as a function ofpercent cell volume (hematocrit). This curve clearly showsthat there is a maximum in MCF-7 cell recovery, whichwas used as the optimum point to perform density-gradientseparation. MCF-7 cell recovery peaks at 73% when acell suspension with percent cell volume concentration of22% is layered on top of the high-density cushion.

Red cell lysis studiesTo replace the DGS step which, as shown previously, pro-duces suboptimal recovery of the spiked cancer cells, studieswere conducted to measure the recovery of spiked MCF-7cells in human blood and to determine the effect, if any, thatthe red cell lysis step has on the viability of the desiredcells as well as the binding of antibodies to specific cell-surface markers.

To test the performance of the red cell lysis step withrespect to the recovery of MCF-7 cells, aliquots of whole,human blood were seeded at a concentration of 1 MCF-7cell per 1.00 × 104 total cells. After lysis, the number ofMCF-7 cells recovered was counted using the CoulterMultisizer II. The results of 10 experiments demonstratedan average recovery of MCF-7 cells of 89.4%, with a relativestandard deviation of 0.04.

In general, no adverse effect was detected with respectto antibody binding to lymphocytes enriched with Ficoll

Hematocrit (% total solids)

0 5 10 15 20 25 30 35

MC

F-7

rec

ove

ry (

%)

0

20

40

60

80

100

Figure 3. Recovery of MCF-7 cells seeded at a concentration of 0.13%in buffy coat samples as a function of hematocrit. Samples were loadedon a Ficoll-Hypaque gradient at the concentration specified by the experi-mental point.

O. Lara et al. /Experimental Hematology 32 (2004) 891–904 897

DGS versus red cell lysis. Specifically, 99% of the leukocyteswere observed to bind anti-CD45–PE antibodies after Redcell lysis, vs 96% after Ficoll separation (relative to totalleukocytes identified by forward and side scatter). Based ontrypan blue staining, 90% of the MCF-7 and 95% of theleukocytes were observed to be viable after the lysis step.However, a slight decrease in absolute fluorescence intensity,FI, between labeled leukocytes obtained from red cell lysiswhen compared to labeled leukocytes obtained from Ficollseparation was observed. (It should be noted that these com-parisons were made sequentially on the flow cytometer sodirect comparisons of the FI could be made.)

Labeling saturation studiesPrevious studies in our laboratory indicate that the magne-tophoretic mobility of an immunomagnetically labeled cellsignificantly affects the performance of the magnetic sepa-ration process [30,36,44]. We have also demonstrated thatthe magnetophoretic mobility is a function of the numberof antibody binding sites on a cell, the degree to which thebinding sites are saturated, the field interaction parameterof the magnetic nanoparticle, and the number of magneticnanoparticles per site [42,45] (the amplification).

Of these four variables, once a cell type and antibodyare chosen, the remaining variable to optimize is the labelingconcentration, which has been shown to have a significanteffect on the saturation of the binding sites, and correspond-ingly, the magnetophoretic mobility of the labeled cells [42].

Figure 4A is a saturation-style plot of the normalizedfluorescence intensity as a function of the concentration ofthe primary, anti-CD45 PE antibody, in units of µg of anti-body per mL of cell suspension with a total of 1.0 × 106 cellspresent. The normalization process consisted of dividingthe fluorescence intensity at a given condition to the maxi-mum value obtained in the given set of experimental flowcytometry analysis experiments. Figure 4B is a saturationplot of the magnetophoretic mobility of the PBL cells, pre-viously labeled with anti-CD45 PE (15 µg/mL), as a functionof the number of µL of MACS reagent per total µL ofsuspension (note Miltenyi Biotec does not provide concen-tration of their reagents in units of µg). Finally, Figure 4Cis a bar graph demonstrating the effect, or lack thereof, ofthe use of FcR blocking reagent on PBL cells labeled with2.82 µg/mL of anti-CD45 PE and 50 µL of MACS reagent.

Since the cell suspension consists of RBC, leukocytes,and rare MCF-7 cells, and the overall goal is to recover allof the MCF-7 cells in the desired final fraction, nonspecificbinding of the anti-CD45 PE to the MCF-7 cells is to beavoided. Figure 5 presents flow cytometry plots of PBLand MCF-7 cell samples stained with anti-CD45 PE andanti-HEAFITCantibodies. Plotsshowthat whenPBLisstain-ed with anti-HEA FITC or MCF-7 cells are stained withanti-CD45 PE antibody, no nonspecific binding is detected(Fig. 5C and E). However, when a sample containing PBLand MCF-7 cells in a 1:1 ratio is used, flow cytometry detects

some cross binding (Fig. 5F). It is suggested that this effectis an analysis artifact due to the gating technique of theanalysis region (i.e., the gating region for the MCF-7 cellsis more ambiguous than for the PBL); however, such ambi-guities in flow cytometry analysis contributed toward themotivation to use a direct staining and visual counting pro-cedure of the final, separated product.

Figure 6A and B present representative histograms, inlinear and log format, of the magnetophoretic mobility ofPBL labeled with antiCD45 PE antibodies (0.225 µg/106

cells) and 15 µL of secondary antibody/106 cells. For com-parison, a histogram for the magnetophoretic mobility distri-bution of unlabeled MCF-7 and MCF-7 cells only labeledwith the secondary antibody is presented in Figure 6C.

Results of overall processusing optimized operating parametersBased on the previously presented optimization studies, threesets of experiments were conducted using slightly differentmethodologies to enrich spiked cancer cells in human blood.These three sets of experiments are presented in Tables 2,3, and 4.

In the three separation methodologies, MCF-7 cells wereadded to fresh, whole human blood and subsequently sub-jected to the red cell lysis protocol (an exception was inthree runs in Table 2, buffy coat of human blood was used andin several experiments the lysis step was preformed twice).The cell suspension was subsequently labeled with opti-mized amounts of antibodies (2.82 µg/mL of anti-CD3 PE,and 100 µL of MACS beads) and pumped through the QMSseparation system using the flow rates presented in the foot-notes of the tables. For the first two sets of experiments, theQMS was operated in the complete flow-through mode (twoexit streams a and b) while in the last set of experiments(Table 4) the sheath flow (b’ and b) was turned off duringthe actual experiment and the system was operated in adeposition mode of operation. A final difference was in thefirst set of experiments (Table 2), a significantly highercancer cell concentration was used (namely, 1 in 105 or 106

total cells) while in the last two sets of experiments (Tables3 and 4) the concentration of cancer cells was significantlylower, on the order of 1 in 108 total cells. This higher levelof spiked cancer cells allowed the final concentration(number) of cancer cells obtained after the QMS separation tobe evaluated using the Coulter Multisizer II; therefore, thefinal number of cells recovered, and final purity, was deter-mined from knowing the final volume collected and thefinal concentration from the Coulter Multisizer II. The finalnumber of cancer cells recovered in the last two sets ofexperiments (Tables 3 and 4) was determined by visuallycounting the cells under a microscope. In all of these experi-mental runs, the nonmagnetic flow stream from the QMSsorter (stream a) was split into two samples, one of whichwasfiltered, stained, andvisually counted, while the other was

O. Lara et al. /Experimental Hematology 32 (2004) 891–904898

0.0E+00

5.0E-05

1.0E-04

1.5E-04

2.0E-04

2.5E-04

3.0E-04

3.5E-04

0 0.2 0.4 0.6 0.8MACS concentration (µL/µL in total)

Mea

n m

agnt

opho

retic

mob

ility

(mm

3 /T

.A.s

)

0.0E+00

5.0E-05

1.0E-04

1.5E-04

2.0E-04

2.5E-04

3.0E-04

3.5E-04

0 0.0625 0.125 0.1875 0.3125 0.625

FcR initial concentration (µL/µL)

m (

mm

3 /T

.A.s

)

0

0.2

0.4

0.6

0.8

1

1.2

Anti-CD45 PE concentration (µg/mL)

Nor

mal

ized

FI (

-)

0 2 4 6 8 10

A B

C

Figure 4. Saturation curve of CD45 surface receptor on PBL (A) with the independent variable the concentration of antibody (µg of Ab/mL-106 cells)and dependent variable is normalized fluorescence intensity. (B) is saturation plot of the magnetophoretic mobility of the PBL cells, previously labeledwith anti-CD45 PE (15 µg/mL), as a function of the number of µL of MACS reagent per total µL of suspension. (C) is a bar graph demonstrating theeffect of the use of FcR blocking reagent on PBL cells labeled with 2.82 µg/mL of anti-CD45 PE and 50 µL of MACS reagent. The abbreviations in the y-axis of panel B refer to the SI units of Tesla (T), ampere (A), seconds (s).

centrifuged and then the cells were collected using the cyto-spin protocol.

The results of these three sets of experiments are pre-sented in Tables 2, 3, and 4. Specifically, for the set of experi-ments presented in Table 2, the average purity increased froman initial purity of 3.15 × 10�6 to 1.87 × 10�2, which repre-sents an average 3.38 log10 enrichment with an average

recovery of the spiked cancer cells of 46%. At a lower initialpurity, an average of 9.7 × 10�9 (Table 3), an increase ofpurity to 5.9 × 10�5 was obtained, a similar 3.75 log10 enrich-ment. The average recovery of cancer cells using this method-ology was 47.5%. However, when the mode of operation ofthe QMS is changed from completely flow-through to apartial deposition mode of operation (sheath fluid flow is

O. Lara et al. /Experimental Hematology 32 (2004) 891–904 899

HEA+

HEA+HEA-

CD45+CD45+

CD45-

HEA FITC

HEA FITC

HEA FITC

HEA FITC

100

100

101

102

103

104

100

101

102

103

104

100

101

102

103

104

100

101

102

103

104

100 101 102 103 104

100 101 102 103 104100 101 102 103 104

101 102 103 104

CD

45P

E

CD

45P

EC

D 4

5PE

CD

45P

ESSC

SSC

R1

0

0

0

1023

1023

1023

1023

FS

C

FS

C

A B C

FE

D

Figure 5. Flow cytometry results of PBL labeled with anti-CD45 PE antibody and MCF-7 stained with anti-HEA-FITC. (A) shows the quadrants todiscriminate positive and negatively stained cells; (B) displays PBL population after Ficoll-Hypaque isolation; (C) shows PBL stained with anti-CD45-PEantibody, approximately 99.96% of the cells are stained; (D) represents unstained MCF-7 cells; (E) shows anti-HEA-FITC stained cells (98.39% labeled);and (F) is a 1:1 mixture of MCF-7/PBL cells.

stopped), the average log10 enrichment increased to 5.17,which translates into taking an average initial purity of 1.43× 10�8 and increasing it to an average of 1.95 × 10�3; yetthe average recovery was maintained at 46%. Photographsof final, separated samples of stained cells on the filtrationmembrane and after cytospin are presented in Figures 7 and8, respectively.

A final point with respect to this separation methodologyis time. The current process takes approximately two hoursfrom the beginning of the lysis of RBC to final analysis,this time including the incubation of antibodies with thecells. Further optimization should shorten this time.

DiscussionObviously, a 7 to 8 log10 enrichment, with 100% recovery,is ideal. Nevertheless, an average log10 enrichment of 5.17and average recovery of 46% of the spiked cancer cells inwhole blood using a completely negative selection processis encouraging. This result is especially encouraging when

one considers that the best reported enrichment we have beenable to find (either positive or negative mode of operation)was a 4 log10 positive enrichment using two positive selec-tion steps [22].

While it is always important to increase the overall recov-ery of the target cell or molecule in a separation process, a100% recovery of the spiked cancer cells in the experimentspresented in Table 4 would only increase the enrichmentfrom 5.17 to 5.47 log10. The most obvious increase in enrich-ment performance would result from an increase in theremoval of unwanted cells from the final sample. This isclearly demonstrated by comparing the results in Tables 2and 3. Specifically, despite lowering the initial purity from3.2 × 10�6 to 9.7 × 10�9, the removal of RBCs in the finalstep through filtration allowed the log enrichment to staythe same. Finally, the higher level of PBL removal usingthe partial deposition mode of operation further reduced thenumber of PBLs in the final product, resulting in a furthertwo-log increase in enrichment (Table 3 vs Table 4).

O. Lara et al. /Experimental Hematology 32 (2004) 891–904900

0.35

0.06

0.05

0.04

0.03

0.02

0.01

0.00

0.30

0.25

0.20

0.15

0.10

0.05

0.00-2e-4 -1e-4 1e-4 2e-4 3e-4

1e-61e-6-2e-4 -1e-4 1e-40 2e-4 3e-4 1e-5 1e-4 1e-3

0

Fre

qu

ency

(-)

Fre

qu

ency

(-)

0.35

0.30

0.25

0.20

0.15

0.10

0.05

0.00

Fre

qu

ency

(-)

m (mm3 /T-A-s)m (mm3 /T-A-s)

m (mm3 /T-A-s)

Unlabeled PBLLabeled PBL

UnlabeledLabeled

Unlabeled MCF-72nd Ab only

A

C

B

Figure 6. (A): Linear plot of the magnetophoretic mobility of unlabeled PBL and labeled PBL with anti-CD45-PE as primary antibody (0.225 µg/106 cells)and 15 µL/106 cells of secondary antibody (anti-PE MACS). (B): A log plot of the same data as in A. (C): Magnetophoretic mobility of unlabeled cancercells (MCF-7), cancer cells labeled with 15 µL/106 cell of secondary antibody (anti-PE MACS).

As with all separation and analysis technology, improve-ment of the resolving power improves the overall perfor-mance of the technology. This same concept has beendemonstrated theoretically and experimentally with immu-nomagnetic cell separation in the QMS system [42,45,46].In the current application, an increase in resolving powerwould theoretically allow a further improvement in the re-moval of unwanted cells in the final product. Figure 6A andB illustrates this resolving power as the difference in meanmagnetophoretic mobility between labeled and unlabeledPBL, when compared to dispersion of the mobility distribu-tion [47]. As can be observed, despite a significant differencein the mean magnetophoretic mobility between the labeledand unlabeled cells (a factor of 104) there still is someoverlap in the magnetophoretic mobility between the labeledand unlabeled cells. It is possible that it is this overlapthat accounts for some of the PBL in the final product thatresults in a lower, final enrichment. Previous and ongoingresearch in our laboratories has focused on developing meth-ods to increase the magnetophoretic mobility of target cells

without increasing nonspecific binding of nontargetedcells [36,45].

With respect to recovery, the average of recovery of 46%,approximately the same using the three separation methodol-ogies, is respectable considering that in our laboratory, weonly obtain an average recovery of approximately 93% percentrifuge step. Thus, with six centrifuge steps (typicallyused in this study) at 93% yield and one lysis step at 89% yield,one would expect an overall recovery of 57%, surprisinglyclose to the actual recoveries obtained. This is especially sur-prising since this simple calculation assumes 100% recoveryin the QMS step.

The desire to obtain the highest recovery per step led tothe replacement of the Ficoll density sedimentation stepwith the red cell lysis step. Figure 3 suggests that the recoveryof cancer cells sedimenting with the mononuclear fraction isa function of the percentage of cell volume concentrationinitially present in the suspension to be separated. The lackof increase in recovery at high initial concentrations (�30%)suggests a potential overloading of the density gradient.

O. Lara et al. /Experimental Hematology 32 (2004) 891–904 901

Table 2. Experimental data of rare cancer cell enrichment using a MCF-7 concentration of 10�5–10�6 (cells/total cells) and thecomplete flow-through mode of operation

Run 1a 2b 3c 4d 5e 6f Average

FeedstockNumber of total blood cell added 3.90 × 1010 1.39 × 1010 2.88 × 1010 4.01 × 1010 2.17 × 1010 1.76 × 1010

Number of PBL added — — — — 1.75 × 107

Number of total MCF-7 added 552,000 13,851 26,400 40,100 21,700 17,570MCF-7 concentration (cells/total cells) 1.4 × 10�5 1.0 × 10�6 0.92 × 10�6 1.0 × 10�6 1.0 × 10�6 1.0 × 10�6 3.15 × 10�6

MCF-7 concentration (cells/PBL) — — — — — 1.0 × 10�3

Cell numbers after lysisNumber of total PBL obtained 1.17 × 107 3.5 × 107 2.60 × 106 3.00 × 107 8.40 × 107 7.20 × 106

Number of RBC obtained 3.60 × 107 — 3.04 × 107 3.35 × 107 8.47 × 107 2.12 × 107

Cell numbers after QMSNumber of PBL in the outlet a 0.81 × 106 7.30 × 106 1.26 × 106 1.00 × 106 0.40 × 106 2.80 × 106

Number of RBC in the outlet a 4.14 × 106 1.81 × 107 7.80 × 105 1.49 × 106 2.64 × 106 6.96 × 106

Number of MCF-7 in the outlet a 317,458 9,081 15,660 6,140 4,148 10,500Final purity, number MCF-7 6.41 × 10�2 3.58 × 10�4 7.62 × 10�3 2.46 × 10�3 1.36 × 10�3 1.08 × 10�3 1.87 × 10�2

concentration (cells/total cells)MCF-7 recovery in the outlet a (%) 57.4 65.6 59.3 15.3 19.2 59.8 46.1Log enrichment (-) 3.66 2.55 3.92 3.39 3.13 3.03 3.38

a, c–f Magnetophoretic mobility is about 1.12 × 10�4 mm3/T.A.s. Qt � 10 mL/min, Qa’/Qt � 0.15, Qa/Qt � 0.5, transport lamina thickness � 586 µm. Cellfeed volume is 15 mL. The number of MCF-7 after enrichment was determined using Coulter Counter and microscope.

b Magnetophoretic mobility � 1.24 × 10�4 mm3/T.A.s. Qt � 8 mL/min, Qa’/Qt � 0.2, Qa/Qt � 0.5, transport lamina thickness � 486 µm. Cell feedvolume is 15 mL. The number of MCF-7 after enrichment was determined using Coulter Counter and microscope.

a, d, f Fresh blood was used in the experiments.b, c, e Buffy coat was used in the experiments.

A concern can be raised with respect to how representa-tive these presented studies are to actually separating cancercells from the blood of cancer patients. Several observationscan be made with respect to this concern. First, in all but afew of the experimental results presented in Tables 2–4, fresh

human peripheral blood from normal donors was used forthe spiking studies (in the few other studies, buffy coatpurchased from the American Red Cross was used).

Second, how representative are MCF-7 cells of actualcancer cells circulating in a cancer patient’s blood? In the

Table 3. Experimental data of rare cancer cell enrichment using a MCF-7 concentration of 10�8 (cells/total cells) and the completeflow-through mode of operation

Run 1 2 3 4 5 Average

FeedstockNumber of total blood cell added 3.84 × 1010 3.93 × 1010 1.76 × 1010 1.58 × 1010 1.21 × 1010

Number of PBL added 12.23 × 107 14.22 × 107 1.75 × 107 1.25 × 107 1.75 × 107

Number of total MCF-7 added 389 335 176 158 120MCF-7 concentration (cells/total cells) 1.0 × 10�8 0.85 × 10�8 1.0 × 10�8 1.0 × 10�8 1.0 × 10�8 9.7 × 10�9

MCF-7 concentration (cells/PBL) 0.32 × 10�5 0.25 × 10�5 1.0 × 10�5 1.26 × 10�5 0.69 × 10�5

Cell numbers after lysisNumber of total PBL obtained 5.69 × 107 5.13 × 107 7.2 × 106 8.60 × 106 6.99 × 106

Number of RBC obtained 2.35 × 107 1.79 × 107 2.12 × 107 1.45 × 107 3.83 × 106

Cell numbers after QMSNumber of PBL recovered in the outlet a 3.55 × 106 5.9 × 106 2.77 × 106 0.74 × 106 0.35 × 106

Number of RBC recovered in the outlet a 1.06 × 107 1.81 × 107 7.00 × 106 6.60 × 106 4.29 × 105

Final recovery after filtrationNumber of PBL recovered in the outlet a 3.55 × 106 5.9 × 106 2.77 × 106 7.40 × 105 3.50 × 105

Number of RBC recovered in the outlet a 0 0 0 0 0Number of MCF-7 recovered in the outlet a 164 208 120 66 28Final purity, number MCF-7 4.62 × 10�5 3.53 × 10�5 4.33 × 10�5 8.92 × 10�5 8.00 × 10�5 5.88 × 10�5

concentration (cells/total cells)MCF-7 recovery in the outlet a (%) 42.2 62.1 68.2 41.8 23.3 47.5Log enrichment (-) 3.66 3.62 3.64 3.95 3.90 3.75

aMagnetophoretic mobility was approximately 1.12 × 10�4 mm3/T.A.s. Qt � 10 mL/min, Qa’/Qt � 0.15, Qa/Qt � 0.5, transport lamina thickness � 586µm. Cell feed volume was 10 mL. The number of MCF-7 after enrichment was determined using Cytospin or membrane stain. Fresh blood was used inthe experiments.

O.

Lara

etal./E

xperimental

Hem

atology32

(2004)891–904

902

Table 4. Experimenta l cells) and the partial deposition mode of operationa

Cycle 5 6b Average 7b 8

FeedstockNumber of total blo 1.28 × 1010 1.74 × 1010 1.28 × 1010 1.16 × 1010

Number of PBL add 0.78 × 107 3.77 × 107 0.78 × 107 0.25 × 107

Number of total MC 128 150 1,160,000 500,000MCF-7 concentratio 1.0 × 10�8 0.86 × 10�8 1.43 × 10�8 9.09 × 10�5 4.31 × 10�5

MCF-7 concentratio 1.64 × 10�5 0.40 × 10�5 1.49 × 10�1 1.99×10�1

Cell numbers after lysNumber of total PB 12.77 × 106 18.99 × 106 12.77 × 106 25.00 × 106

Number of RBC ob 9.07 × 107 8.3 × 106 9.07 × 107 14.28 × 106

Cell numbers after QMNumber of PBL rec 45,964 18,286 69,669 38,897Number of RBC rec 13.88 × 106 0.73 × 106 31.78 × 106 0.54 × 106

Final recovery after fiNumber of PBL rec 45,964 18,286 69,669 38,897Number of RBC rec 0 0 0 0Number of MCF-7 65 69 500,000 186,224Final purity, numbe 1.41 × 10�3 3.77 × 10�3 1.95 × 10�3 0.877 0.827

concentration (ceMCF-7 recovery in 50.8 46.0 46 43.1 37.2Log enrichment (-) 5.15 5.64 5.17 4.90 5.05

aMagnetophoretic mob ation is in the range of 1–3 × 106 cells/mL. The number of MCF-7 after QMS separationwas determined using

bLysis twice.

l data of rare cancer cell enrichment using using a MCF-7 concentration of 10�8 (cells/tota

1 2 3 4

od cell added 1.58 × 1010 1.21 × 1010 3.57 × 1010 5.62 × 109

ed 1.25 × 107 1.75 × 107 3.84 × 107 4.24 × 107

F-7 added 158 120 360 212n (cells/total cells) 1.0 × 10�8 1.0 × 10�8 1.0 × 10�8 3.77 × 10�8

n (cells/PBL) 1.26 × 10�5 0.69 × 10�5 0.94 × 10�5 0.5 × 10�5

isL obtained 8.60 × 106 6.99 × 106 2.70 × 107 14.70 × 106

tained 4.40 × 107 1.16 × 107 5.79 × 107 1.40 × 107

Sovered in the outlet a 45,454 32,567 130,000 46,165overed in the outlet a 1.58 × 107 0.91 × 106 0.80 × 106 1.10 × 106

ltrationovered in the outlet a 45,454 32,567 130,000 46,165overed in the outlet a 0 0 0 0

recovered in the outlet a 76 61 160 80r MCF-7 1.67 × 10�3 1.87 × 10�3 1.23 × 10�3 1.73 × 10�3

lls/total cells)the outlet a (%) 48.1 50.8 44.4 37.7

5.22 5.27 5.09 4.66

ility is in the range of 1.8–3.0 × 10�4 mm3/T.A.s Qa � Qa’ � 15 mL/min. Cell feed concentrcytospin or membrance stain.

O. Lara et al. /Experimental Hematology 32 (2004) 891–904 903

Figure 7. Photographs of MCF-7 cells recovered in the nonmagnetic frac-tion stained with anti-HEA FITC as primary antibody, anti-FITC phospha-tase alkaline as secondary antibody, and enzyme substrate (FastRed). Cellsappear as a bright red color body in the microscope field.

overall process used in this study, no specific markers orassumptions on the cancer cell density (w/v) were usedfor the separation/isolation of the cancer cell, except to makeit appear red on the filter to facilitate counting in the finalquantification step. (This dye was not used in the cytospinprocedure.) With respect to the size of the cell, informaldiscussions with cancer researchers at the Ohio State Univer-sity Comprehensive Cancer Center indicated that all thehuman cancer cells currently being grown in the researchers’labs are significantly larger than PBL, and typically on theorder of the MCF-7 used in this study. The only exceptionare blood-based cancer cell lines which, in many cases,probably express CD45, which would then defeat the nega-tive selection used in this study. While filtration was usedfor half of the cell suspension from the QMS separationstep, this filtration was not necessary as reported since theother half was centrifuged and subjected to a cytospin. Inaddition, the filter medium used allowed red blood cells topass, but not PBL; hence the pore size of the membrane

Figure 8. Representative photograph of cytospin of pure MCF-7 cells,pure PBL, and two samples from the QMS after magnetic separation (A,B, C, and D, respectively).

was smaller than six microns; considerably smaller than anyother cancer cells (of which the authors are aware).

In conclusion, it has been experimentally demonstratedthat it is possible to obtain an average 5.17 log10 enrichmentof rare cells using a negative cell selection process. Ongoingstudies are currently focused on further improvement ofboth the enrichment and recovery of the target cells, andshortening the time of the overall separation. Also, studiesusing more clinically relevant cell lines with the ultimategoal of testing and use on human cancer patients is currentlybeing conducted.

AcknowledgmentsThis work has been supported by the National Science Foundation(BES-9731059 and BES-0124897 to J.J.C.; NSF SBIR 02-056awarded to SHOT INC.) and the National Cancer Institute (R01CA62349 to M.Z., R33 CA81662-01, R01 CA97391-01A1 toJ.J.C.), and P30CA16058.

References1. Pantel K, Cote RJ, Fodstat Ø. Detection and clinical importance of

micrometastatic disease. J Natl Cancer Inst. 1999;91(13):1113–1124.2. Benez A, Geiselhart A, Handgretinger R, et al. Detection of circulating

melanoma cells by immunomagnetic cell sorting. J Clin Lab Anal.1999;13:229–233.

3. Taback B, Chan AD, Kuo CT, et al. Detection of occult metastatic breastcancer cells in blood by multimolecular marker assay: Correlationwith clinical stage of disease. J Cancer Res. 2001;61:8845–8850.

4. Weihrauch MR, Skibowski E, Koslowski TC, et al. Immunomag-netic enrichment and detection of micrometastases in colorectal cancer:Correlation with established clinical parameters. J Clin Oncol. 2002;20(21):4338–4343.

5. Braun S, Pantel K, Muller P, et al. Cytokeratin-positive cells in thebone marrow and survival of patients with stage I, II, or III breastcancer. N Engl J Med. 2000;342:525–533.

6. Beitsch P, Clifford E. Detection of carcinoma cells in the blood ofbreast cancer patients. Am J Surg. 2000;180:446–449.

7. Naume B, Borgen E, Nesland J, et al. Increased sensitivity for detectionof micrometastases in bone-marrow/peripheral-blood stem-cell prod-ucts from breast-cancer patients by negative immunomagnetic separa-tion. Int J Cancer. 1998;78:556–560.

8. Braun S, Pantel K. Clinical significance of occult metastatic cells inbone marrow of breast cancer patients. Oncologist. 2001;6:125–132.

9. Maguire D, O’Sullivan GC, Collins JK, et al. Bone marrow micrometas-tases and gastrointestinal cancer detection and significance. AJG. 2000;95(7):1644–1651.

10. Zhong XY, Kaul S, Lin YS, et al. Sensitive detection of micrometastasesin bone marrow from patients with breast cancer using immunomag-netic isolation of tumor cells in combination with reverse transcriptase/polymerase chain reaction for cytokeratin-19. J Cancer Res Clin Oncol.2000;126:212–218.

11. Zigeuner RE, Riesenberg R, Pohla H, et al. Immunomagnetic cellenrichment detects more disseminated cancer cells than immunocyto-chemistry in vitro. J Urol. 2000;164:1834–1837.

12. Bilkenroth U, Taubert H, Riemann D, et al. Detection and enrichmentof disseminated renal carcinoma cells from peripheral blood by immu-nomagnetic cell separation. Int J Cancer. 2001;92:577–582.

13. Makarovskiy AN, Ackerley W III, Wojcik L, et al. Application ofimmunomagnetic beads on combination with RT-PCR for the detectionof circulating prostate cancer cells. J Clin Lab Anal. 1997;11:346–350.

O. Lara et al. /Experimental Hematology 32 (2004) 891–904904

14. Fishman A, Klein A, Zemer R, et al. Detection of micrometastasis bycytokeratin-20 (reverse transcription polymerase chain reaction) inlymph nodes of patients with endometrial cancer. Gynecol Oncol.2000;77:399–404.

15. Hardingham JE, Kotasek D, Farmer B, et al. Immunobead-PCR: Atechnique for the detection of circulating tumor cells using immuno-magnetic beads and the polymerase chain reaction. Cancer Res. 1993;53:3455–3458.

16. Hardingham JE, Hewett PJ, Sage RE, et al. Molecular detection ofblood-borne epithelial cells in colorectal cancer patients and in patientswith benign bowel disease. Int J Cancer. 2000;89:8–13.

17. Molnar B, Ladayni A, Tanko L, et al. Circulating tumor cells in theperipheral blood of colorectal cancer patients. Clin Cancer Res. 2001;7:4080–4085.

18. Park S, Lee B, Kin I, et al. Immunobead RT-PCR versus regular RT-PCRamplification of CEA mRNA in peripheral blood. J Cancer Res ClinOncol. 2001;127:489–494.

19. Georgieva J, Milling A, Orfanos CE, et al. Magnetic bead-PCR: estab-lishment of a new method for detecting circulating melanoma cells. Mel-anoma Res. 2002;12:309–317.

20. Iinuma H, Okinaga K, Adachi M, et al. Detection of tumor cells in bloodusing CD45 magnetic cell separation followed by nested mutant allele-specific amplification of p53 and K-ras genes in patients with colorectalcancer. Int J Cancer. 2000;89:337–344.

21. Kruger W, Datta C, Badbaran A, et al. Immunomagnetic tumor cell selec-tion—implications for the detection of disseminated cancer cells. Trans-fusion. 2000;40:1489–1493.

22. Martin VM, Siewert C, Scharl A, et al. Immunomagnetic enrichment ofdisseminated epithelial tumor cells fromperipheralblood byMACS. ExpHematol. 1998;26:252–264.

23. Partridge M, Phillips E, Francis R, et al. Immunomagnetic separationfor enrichment and sensitive detection of disseminated tumor cells inpatients with head and neck SCC. J Pathol. 1999;189:368–377.

24. Sieben S, Bergemann C, Lubbe A, et al. Comparison of different particlesand methods for magnetic isolation of circulating tumor cells. Journalof Magnetism and Magnetic Materials. 2001;225:175–179.

25. Racila E, Euhus D, Weiss AJ, et al. Detection and characterization ofcarcinoma cells in the blood. Proc Natl Acad Sci U S A. 1998;95:4589–4594.

26. Moreno JG, O’Hara SM, Gross S, et al. Changes in circulating carcinomacells in patients with metastasis prostate cancer correlate with diseasestatus. Urology. 2001;58(3):386–392.

27. Liberti PA, Rao CG, Terstappen LWMM. Optimization of ferrofluids andprotocols for the enrichment of breast tumors cells in blood. Journal ofMagnetism and Magnetic Materials. 2001;225:301–307.

28. O’Hara SM, Moreno JG, Zweitzic DR, et al. Multigene reverse transcrip-tion-PCR profiling of circulating tumor cells in hormone-refractorypros-tate cancer. Clin Chem. 2004;50(5):826–835.

29. Fang B,Zborowski M, Moore RL.Detection of rare MCF-7breast cancercells from mixtures of peripheral blood leucocytes by magnetic deposi-tion analysis. Cytometry. 1999;36:294–302.

30. Nakamura M, Decker K, Chosy J, et al. The separation of a breast cancercell line from human blood using a quadrupole magnetic flow sorter.Biotechnol Prog. 2001;17:1145–1155.

31. Shammo JL, Smith SL, Bennet MV, et al. Use of tumor-cell enrichmentcolumn for the enhanced detection of minimal residual disease in theBM or apheresis peripheral blood transplant products of breast-cancerpatients. Cytotherapy. 1999;1(5):367–376.

32. Vona G, Sabile A, Louha M, et al. Isolation by size of epithelial tumorcells: A new method for the immunomorphological and molecularcharacterization of circulating tumor cells. AJP. 2000;156:57–63.

33. Rostagno P, Moll JL, Bisconte JC, et al. Detection of rare circulatingbreast cancer cells by filtration cytometry and identification by DNA

content: Sensitivity in an experimental model. Anticancer Res. 1997;17:2481–2486.

34. Baker MK, Mikhitarian K, Osta W, et al. Molecular detection of breastcancer cells in the peripheral blood of advanced-stage breast cancer pa-tients using multimarker real-time reverse transcription–polymerasechain reaction and a novel porous barrier density gradient centrifugationtechnology. Clin Cancer Res. 2003;9:4865–4871.

35. Melnik K, Nakamura M, Comella K, et al. Evaluation of eluents fromseparations of CD34� cells from human cord blood using a commer-cial, immunomagnetic cell separation system. Biotechnol Prog. 2001;17:907–916.

36. Comella K, Nakamura M, Melnik K, et al. The effect of antibody concen-tration on the separation of human natural killer cells in a commercialimmunomagnetic separation system. Cytometry. 2001;45:285–293.

37. Sun L, Zborowski M, Moore LR, et al. Continuous, flow-through immu-nomagnetic cell separation in a quadrupole field. Cytometry. 1998;33:469–475.

38. Moore LR, Rodriguez AR, Williams PS, et al. Progenitor cell isolationwith a high-capacity quadrupole magnetic flow sorter. Journal of Magne-tism and Magnetic Materials. 2001;225:277–284.

39. Hoyos M, Moore LR, McCloskey KE, et al. Study of magnetic particlespulsed-injected into an annular SPLITT-like channel inside a quadrupolemagnetic field. J Chromatogr A. 2000;903(1–2):99–116.

40. Hoyos M, McCloskey K, Moore L, et al. Pulse-injection studies of bloodprogenitor cells in a quadrupole magnetic flow sorter. Sep Sci Technol.2002;37(4):1–23.

41. Chalmers JJ, Zhao Y, Nakamura M, et al. An instrument to determinethe magnetophoretic mobility of paramagnetic particles and labeled, bio-logical cells. Journal of Magnetism and Magnetic Materials. 1999;194:231–241.

42. Chosy J, Melnik K, Comella K, et al. Antibody saturation studies ofthree antigens proposed for use in rare cancer cell detection. BiotechnolBioeng. 2003;82:340–351.

43. Williams PS, Zborowski M, Chalmers JJ. Flow rate optimization for thequadrupole magnetic cell sorter. Anal. Chem. 1999;71:3799–3807.

44. McCloskey K, Moore L, Hoyos M, et al. Magnetic cell separation is afunction of antibody binding capacity (ABC). Biotechnol Prog. 2003;19(3):899–907.

45. McCloskey KE, Chalmers JJ, Zborowski M. Magnetic cell separa-tion: Characterization of magnetophoretic mobility. Anal Chem. 2003;75(4):6868–6874.

46. Zborowski M, Moore LR, Williams PS, et al. Separations based on mag-netophoretic mobility. Sep Sci Technol. 2002;37:3611–3633.

47. Giddings J. A system based on split-flow lateral transport thin (SPLITT)separation cells for rapid and continuous particle fractionation. Sep SciTechnol. 1985;20(9&10):749–768.

48. Cristofanilli M, Budd GT, Ellis MJ, et al. Circulating Tumor Cells, Dis-ease Progression, and Survival in Metastatic Breast Cancer. N Engl JMed. 2004;351:781–791.

49. Braun S, Marth C. Circulating tumor cells in metastatic breast cancer—toward individualized treatment? N Engl J Med. 2004;351:824–826.

AddendumAfter final acceptance of this manuscript, another significantpublication of a clinincal study, and corresponding editorial,appeared further underscoring the link between circulatingcancer cells and patient outcomes as well as suggesting futureneeded studies (48, 49).