Investigating the Feasibility of Stem Cell Enrichment Mediated by Immobilized Selectins

Investigating the Feasibility of Stem Cell Enrichment Mediated by ImmobilizedSelectins

Nichola Charles,† Jane L. Liesveld,‡ and Michael R. King* ,†

Department of Biomedical Engineering, University of Rochester, Rochester, New York, and Department of Medicine and theJames P. Wilmot Cancer Center, University of Rochester, Rochester, New York

Hematopoietic stem cell therapy is used to treat both malignant and non-malignant diseases,and enrichment of the hematopoietic stem and progenitor cells (HSPCs) has the potential toreduce the likelihood of graft vs host disease or relapse, potentially fatal complications associatedwith the therapy. Current commercial HSPC isolation technologies rely solely on the CD34surface marker, and while they have proven to be invaluable, they can be time-consuming withvariable recoveries reported. We propose that selectin-mediated enrichment could prove to be aquick and effective method for recovering HSPCs from adult bone marrow (ABM) on the basisof differences in rolling velocities and independently of CD34 expression. Purified CD34+ ABMcells and the unselected CD34- ABM cells were perfused over immobilized P-, E-, andL-selectin-IgG at physiologic wall shear stresses, and rolling velocities and cell retention datawere collected. CD34+ ABM cells generally exhibited lower rolling velocities and higherretention than the unselected CD34- ABM cells on all three selectins. For initial CD34+ ABMcell concentrations ranging from 1% to 5%, we predict an increase in purity ranging from 5.2%to 36.1%, depending on the selectin used. Additionally, selectin-mediated cell enrichment is notlimited to subsets of cells with inherent differences in rolling velocities. CD34+ KG1a cellsand CD34- HL60 cells exhibited nearly identical rolling velocities on immobilized P-selectin-IgG over the entire range of shear stresses studied. However, when anti-CD34 antibody wasco-immobilized with the P-selectin-IgG, the rolling velocity of the CD34+ KG1a cells wassignificantly reduced, making selectin-mediated cell enrichment a feasible option. Optimal cellenrichment in immobilized selectin surfaces can be achieved within 10 min, much faster thanmost current commercially available systems.

Introduction

Hematopoietic stem cell therapy is used to treat a number ofmalignant and benign diseases, and the source of stem cellscan be from either a matched donor or the patient, dependingon the nature of the disease. Hematopoietic stem and progenitorcells (HSPCs) can be harvested by direct bone marrow aspirationor mobilized using growth factors and collected from peripheralblood. Mononuclear cells (MNCs) are isolated from the bonemarrow or peripheral blood extract, and the transplant can beperformed using the entire fraction or it can be further processedto isolate the HSPCs. Extracts obtained from unmatched donorsundergo HSPC enrichment to reduce the T cell load and lessenthe possibility of graft vs host disease, a potentially life-threatening condition in which T cells from the donor initiatean immune response against the host tissue (1, 2). HSPCenrichment of extracts obtained from the patient is performedto reduce the possibility of reintroducing the abnormal, diseasedcells into the patient, which might result in relapse (3).

Current clinical and benchtop methods for isolating HSPCsare entirely immunologic, utilizing antibodies against CD34, aconfirmed cell surface antigen expressed by HSPCs that is lostupon maturation to terminally differentiated cells (4). These

CD34+ cells usually account for 1-5% of the mononuclearfraction of adult bone marrow (ABM) (1, 5). In order to isolatethe CD34+ fraction, the MNCs can be treated with fluorescentanti-CD34 conjugates and purified using a high-speed flowcytometer (fluorescence activated cell sorting or FACS) or theantibody can be conjugated to paramagnetic beads (Dynabeads,Isolex) or iron-dextran nanoparticles (MiniMACS, CliniMACS)and then purified using a strong magnetic field. Although FACSwill consistently yield a very pure population of CD34+ HSPCs,routinely>90%, it can be prohibitively time-consuming (6, 7).Immunomagnetic cell enrichment is much faster than FACS,requiring ∼1 h to process up to 4× 108 cells/mL, but finalpurities are less consistent, giving anywhere from 18% to 99%CD34+ HSPC purity after enrichment(1, 3, 8-10). Immuno-magnetic enrichment methods may also fail to select theCD34low lineage-committed progenitor cells that are responsiblefor early engraftment post transplant (11), and there is sufficientevidence to suggest that a small portion of HSPCs do not expressCD34 and hence cannot be selected using these immunologicmethods (12-15). CD34+ HSPC recovery for all methods islow, generally<50% (1, 3, 8-10).

CD34+ HSPCs isolated using immunomagnetic cell enrich-ment methods have exhibited lower rolling velocities and higherrolling fluxes on immobilized E-, L-, and P-selectin than moredifferentiated CD34- hematopoietic cells (16, 17). Rolling isa typical event in leukocyte recruitment to sites of inflammation,as well as lymphocyte and stem cell homing, and is mediated

* To whom correspondence should be addressed. Tel: (585) 275-3285.Fax: (585) 276-1999. E-mail: [email protected].

† Department of Biomedical Engineering.‡ Department of Medicine and the James P. Wilmot Cancer Center.

1463Biotechnol. Prog. 2007, 23, 1463−1472

10.1021/bp0702222 CCC: $37.00 © 2007 American Chemical Society and American Institute of Chemical EngineersPublished on Web 10/31/2007

by transient interactions between selectins and their ligands, inthe presence of calcium ion (18, 19). The differences in rollingvelocities and fluxes suggest that selectin-mediated enrichmentmay be possible (17), so we investigated the feasibility of usingimmobilized selectins to enrich CD34+ cells from ABM MNCs,independent of CD34 expression.

Hematologic cancers and certain metastatic cancer cells (20-24) are also responsive to immobilized selectins, and a long-term goal is to develop devices to isolate the rare cancer andstem cells from normal leukocytes in peripheral blood. Whileintrinsic differences in rolling velocities may exist betweendifferent groups of rolling cells (16, 17), subsets of these cellstend to have very similar rolling velocities that make selectin-mediated enrichment difficult. Eniola et al. (25) reported thatthe rolling velocity of antigen-coated beads could be reducedif an appropriate antibody is co-immobilized with the selectin,but often times the behavior of rigid beads under the influenceof a flow field does not accurately predict the behavior of cellsunder the same conditions. We investigated whether co-immobilization of a suitable antibody together with the selectinwas sufficient to introduce a difference in rolling velocitiesbetween two subsets of acute myeloid leukemia (AML) celllines of similar size and rolling velocities on selectins alonebut with different surface antigen profiles.

Materials and Methods

Reagents and Antibodies.Stemline II hematopoietic stemcell expansion media, human serum albumin (HSA), bovineserum albumin (BSA), HEPES, EDTA, and CaCO3 were allobtained from Sigma-Aldrich (St Louis, MO). The Dynabeadscell selection system, Hanks balanced salt solution (HBSS), andphosphate-buffered saline (PBS) were all obtained from Invit-rogen (Grand Island, NY). Recombinant human P-selectin-IgG,E-selectin-IgG, and L-selectin-IgG chimeras were purchasedfrom R&D Systems (Minneapolis, MN). Antibodies againsthuman CD34 were obtained from BD Pharmingen (FranklinLakes, NJ), and Ficoll-Paque PLUS came from GE Healthcare(Piscataway, NJ). RPMI andL-glutamine came from Mediatech(Herndon, VA), and fetal bovine serum (FBS) was obtainedfrom Atlanta biologicals (Lawrenceville GA).

Mononuclear Cell and CD34+ ABM Cell Isolation. Allhuman subject protocols have been approved by the ResearchSubjects Review Board of the University of Rochester. ABMwas collected from willing donors after informed consent intovacutainer tubes containing heparin and allowed to equilibrateat room temperature before use. Ten to fifteen milliliters ofABM was diluted with PBS to a final volume of 35 mL, thencarefully layered over 10 mL of Ficoll-Paque PLUS, andcentrifuged at 1600 rpm for 30 min. MNCs were collected andwashed twice with PBS before being resuspended to between4 × 107 and 4× 108 cells/mL. The Dynabeads cell selectionsystem was used for isolation of CD34+ ABM cells as permanufacturer’s instructions. Briefly, anti-CD34 antibody con-jugated to 4.5-µm diameter paramagnetic beads was added tothe MNC suspension at 4× 107 to 4× 108 cells/mL and allowedto incubate at 4°C for 30 min before being placed in a magneticfield. The paramagnetic beads, attached to CD34+ ABM cells,precipitated within the magnetic field so that the unselectedCD34- ABM cells could easily be removed and collected. TheCD34+ ABM cells were released from the antibody-beadcomplex using a competitive peptide, washed, and resuspendedin HBSS+ (HBSS without Ca2+ and Mg2+ supplemented with0.5% w/v HSA, 10 mM HEPES, and 2 mM CaCO3) to 1× 106

cells/mL.

CD34+ ABM Cell Purity. Samples of the isolated CD34+ABM cells were incubated with anti-human CD34-phycoerythrinor the corresponding isotype control for 15 min at 4°C andthen washed with PBS to remove unbound immunoglobulins.Samples and controls were analyzed by flow cytometry usingthe BD FACSCalibur (Becton-Dickinson) to determine thepercent of cells that are positive for CD34. Mean CD34+ ABMcell purity used for flow experiments was determined to be 73%.

Acute Myeloid Leukemia Cell Maintenance and Prepara-tion. CD34+ KG1a and CD34- HL60 cells were maintainedin culture in RPMI media supplemented with 1%L-glutamineand 10% FBS. On the day of experimentation, cells were washedwith PBS and resuspended in HBSS+ to a final concentrationof 1 × 106 cells/mL.

Preparation of Immobilized Protein Surfaces.Recombinanthuman P-selectin-IgG, E-selectin-IgG, or L-selectin-IgG wasdissolved in PBS to give a final concentration of 100µg/mL.Aliquots of the stock solutions were stored at-20 °C and usedas needed within 30 days. A rectangular flexiPERM gasket(E&K Scientific, Santa Clara, CA) was placed on the bottomof a 35-mm culture dish (Corning Inc, Corning, NY) to createa leak-proof well, and 0.5µg/mL P-selectin-IgG, 0.1µg/mLE-selectin-IgG, or 10µg/mL L-selectin-IgG solution was addedto the well and incubated at room temperature for 2 h. Theconcentration of each selectin used was optimized so that asignificant number of CD34+ ABM cells would roll 4 celldiameters within 1 min. The selectin solution was aspirated,and the culture dish was washed three times with PBS, incubatedwith blocking buffer (2%w/v BSA in PBS) for 1 h, and thenwashed three more times with PBS before experimentation.

In instances where anti-human CD34 antibody was co-immobilized with the P-selectin-IgG, the antibody was addedto the selectin solution to a final concentration of 40µg/mLprior to incubation. We chose to use only P-selectin-IgG forthe antibody-selectin experiments because of the intermediateeffects on rolling velocity of ABM cells. We can qualitativelypredict the effects of co-immobilizing E- or L-selectin with theantibody based on the results obtained with the ABM cells onthose selectins. Control surfaces consisting of a petri dish treatedwith blocking buffer but no selectin or immobilized anti-CD34antibody alone did not support rolling.

Rolling Experiments. Rolling experiments were performedusing a parallel plate flow chamber (Glycotech, Rockville, MD),assembled as shown in Figure 1A. The flow chamber wassecured to the stage of the Olympus IX81 motorized invertedresearch microscope (Olympus America Inc, Melville, NY)equipped with a CCD camera (model KP-M1AN, Hitachi,Japan) connected to an S-VHS videocassette recorder (modelSVO-9500MD2, Sony Electronics, Park Ridge, NJ) to facilitateimage capture. A syringe pump (New Era Pump Systems,Wantagh, NY) was used to control the flow rate of the cellsuspension. Under the influence of a shear stress, cells express-ing the correct cell surface markers are captured by the selectinor selectin-antibody surface and appear to slowly roll in thedirection of flow (Figure 1B).

Data Analysis.“Rolling” cells were defined as those observedto translate in the direction of flow with an average velocityless than 50% of the calculated hydrodynamic free streamvelocity (26). Cells that remained stationary for more than10 s or rolled less than 4 cell diameters were not considered tobe rolling. Instances where only one rolling cell was visible in1 min were not included in the flux analysis. The rolling fluxwas calculated by counting the number of rolling cells in thearea of interest over a 1 min time interval at each shear stress.

1464 Biotechnol. Prog., 2007, Vol. 23, No. 6

Velocities of single cells were determined using a MATLABprogram designed to capture the change in position of the cellcentroid in a given time period.

To determine the theoretical effectiveness of the enrichmentprocess, an exponentially modified Gaussian (EMG) was usedto describe the velocity distribution of the cells at each shearstress tested (17, 27). A velocity bin-size of 3µm/s wasdetermined to be adequate for P- and E-selectin-IgG, whereas10 µm/s was best for L-selectin-IgG. The addition of theexponential term introduces a distortion to the Gaussiandistribution to account for asymmetry.

wherea0 ) area under curve (constant for all curves),a1 )peak of curve,a2 ) standard deviation of Gaussian,a3 ) timeconstant of exponential,y(x) ) frequency, andx ) velocity.

The potential to enrich the CD34+ KG1a cells or CD34+ABM cells was quantified using the peak-to-peak resolution;higher resolutions suggest better separation of the peaks andhence, better potential for enrichment (17, 27).

wherep1 ) velocity at peak of curve 1,p2 ) velocity at peakof curve 2,w1 ) width at 10% of height of curve 1, andw2 )width at 10% of height of curve 2.

Cell retention measures the number of cells remaining boundto the immobilized selectin or selectin-antibody surfaces afterthe start of flow:

where Ci is the initial number of cells within the region ofinterest, andCb is the number of cells remaining bound onceflow has begun.

Statistics.All bone marrow experiments were performed intriplicate (unless otherwise noted in the figure legend) to accountfor donor variability, while experiments using the CD34+ KG1a

Figure 1. Illustration of equipment and principles involved in obtainingrolling data. (A) Flow chamber assembly for performing cell rollingexperiments. The gasket provides a channel (0.127 mm× 2.5 mm×20 mm) for observing the rolling cells. (B) Cells interact with P-selectin-IgG and anti-human CD34 antibody and slowly roll in the direction ofan applied shear stress.

Table 1. R-Squared Values from Fit of Experimental RollingVelocities to the Exponentially Modified Gaussian Equation forABM Cells on All Selectins

dyn/cm2

1 3 5 8 10

P-selectin-IgG CD34- ABM cells 0.97 0.92 0.73 0.92 0.72CD34+ ABM cells 1 0.99 0.99 0.98 0.98

E-selectin-IgG CD34- ABM cells 0.97 0.99 0.87 0.97 0.98CD34+ ABM cells 0.95 0.98 0.97 0.97 0.98

L-selectin-IgG CD34- ABM cells 0.99 0.82 0.73 0.85 0.81CD34+ ABM cells 0.99 0.96 0.95 0.94 0.96

y(x) )a0

2a3EXP( a2

2

2a32

+a1 - x

a3 )(ERF( 1

x2 (a1

a2-

a2

a3)) +

ERF( 1

x2 (x - a1

a2-

a2

a3))) (1)

Figure 2. Effect of shear stress on selectin-mediated rolling of ABMcells. CD34+ ABM cells and CD34- ABM cells were perfused overimmobilized P-selectin-IgG (A, D), E-selectin-IgG (B, E), or L-selectin-IgG (C, F) at wall shear stresses ranging from 1 to 10 dyn/cm2. (A-C)CD34+ ABM cells roll significantly slower than CD34- ABM cellsat all shear stresses tested (p < 0.005 in all cases). (B-F) Increasingthe wall shear stress increased the rolling flux, but statistically significantdifferences in flux between CD34+ ABM cells and CD34- ABM cellsoccur only on E-selectin-IgG (p ) 0.011) and L-selectin-IgG (p )0.034).

resolution)2(p1 - p2)

w1 + w2(2)

retention (%))Cb

Ci× 100 (3)

Biotechnol. Prog., 2007, Vol. 23, No. 6 1465

and CD34- HL60 cell lines were performed in duplicate. Dataare reported as mean( SEM, except on velocity distributionand prediction graphs. Student pairedt-tests were performed todetermine significance (p-value<0.05) when necessary.

Results and Discussion

Current stem cell separation methods use antibodies againstCD34 to identify and isolate HSPCs. Although FACS andimmunomagnetic cell enrichment have been invaluable, FACSis very time-consuming, and immunomagnetic cell enrichmentmethods may fail to select some CD34low cells (11). Bothmethods generally have low cell recoveries, usually less than50%, and there is evidence to suggest that the most primitivehematopoietic stem cells do not express CD34 (12-15).

We propose a new functional method for isolating cells basedon their tendency to roll on selectins. The rolling event occursas a result of the transient interactions between immobilizedselectins and their ligands on the cell surface when the cellsare subjected to a flow field. The velocity of rolling cells iseasily determined using image analysis software.

CD34+ ABM cells have lower rolling velocities thanCD34- ABM MNCs on all selectins. We investigated thefeasibility of a selectin-mediated CD34+ ABM cell enrichmentdevice using a commercially available flow chamber function-alized with immobilized selectin chimeras. CD34+ cells isolatedfrom ABM using Dynabeads and the unselected CD34- fractionwere perfused over immobilized P-selectin-IgG (Figure 2A,B),E-selectin-IgG (Figure 2C,D), or L-selectin-IgG (Figure 2E,F)at wall shear stresses ranging from 1 to 10 dyn/cm2. We havefound that CD34+ ABM cells roll more slowly than their

unselected CD34- counterparts (p < 0.05) on P-selectin-IgG(Figure 2A), E-selectin-IgG (Figure 2C), and L-selectin-IgG(Figure 2E) over the range of physiologically relevant wall shearstresses tested. Cells on L-selectin-IgG rolled 3-4 times faster

Figure 3. Velocity distributions of CD34+ and CD34- ABM cells. Velocity data was further analyzed for ABM cells rolling on P-, E-, andL-selectin-IgG. The exponentially modified Gaussian (EMG) equation was used, and a velocity bin-size of 3µm/s for P- and E-selectin-IgG and10 µm/s for L-selectin-IgG were adequate to describe the experimental data (EXP). The velocity distribution for ABM cells on (A) P-selectin-IgG,(B) E-selectin-IgG, and (C) L-selectin-IgG show significant overlap at all shear stresses tested indicating that significant cross-contamination canbe expected.

Figure 4. Effect of shear stress on rolling behavior of AML cells onP-selectin-IgG in the absence and presence of anti-human CD34antibody. CD34+ KG1a cells were perfused over immobilized P-selectin-IgG or P-selectin-IgG with anti-human CD34 antibody. In theabsence of anti-CD34 antibody, CD34- HL60 and CD34+ KG1a cellshave (A) similar rolling velocities (p ) 0.155) and (B) similar rollingfluxes (p ) 0.28). In the presence of the antibody, there is a significantdifference between (C) the rolling velocities (p ) 0.0009) but (D) nosignificant difference between rolling fluxes (p ) 0.062).


than cells on P-selectin-IgG or E-selectin-IgG. A control surfaceconsisting of a petri dish treated with blocking buffer but noselectin did not support adhesion or rolling of the cell suspen-sions.

Rolling flux, defined as the number of cells crossing a regionof interest in a given time, generally increased as the wall shearstress increased on all selectins. The flux of CD34+ ABM cellstended to be higher than that of CD34- ABM cells onE-selectin-IgG (p ) 0.01) and lower than CD34- ABM cellson L-selectin-IgG (p ) 0.03) and showed no significantdifference from CD34- ABM cells on P-selectin-IgG (p )0.091) (Figure 2B,D,F). It is not clear whether these trends inrolling velocities and fluxes are due to differences in the numberof and type of ligands available for each selectin to bind orreceptor-ligand binding kinetics or a combination of all thesefactors, but the data suggest that it may be possible to isolateCD34+ ABM cells from an initial mixture with CD34- ABMcells on the basis of their rolling velocity on selectins.

There is significant overlap of velocity distributions ofCD34+ ABM cells and CD34- bone marrow MNCs on allselectins.The efficiency of a selectin-mediated CD34+ ABMcell enrichment can be predicted from the experimental databy analyzing the rolling velocity data to reveal the velocitydistribution for both types of cells at each shear stress tested.The exponentially modified Gaussian equation (eq 1) was usedto fit the velocity data in order to account for distortions in thedistributions away from a Gaussian. A velocity bin-size of3 µm/s was used for P- and E-selectin-IgG and 10µm/s forL-selectin. The equation quantifies the fraction of cells pos-sessing velocities within the limits defined by the bin-size andwas adequate to describe these distributions, withR2 > 0.85 inmost cases (Table 1).

The efficiency of a velocity-dependent separation can bedetermined by calculating the peak-to-peak resolution, deter-mined from the width of and distance between the two peaks(eq 2). A higher resolution implies better cell enrichment,whereas lower resolutions imply that more cross-contaminationcan be expected. There was significant overlap between thevelocity distributions of CD34+ ABM cells and CD34- ABMcells on all selectins resulting in low peak-to-peak resolu-tions (Figure 3), so significant cross-contamination can beexpected in all cases. Optimum cell separation, the best balanceof purity and recovery, occurs at the intersection of the twodistributions and the velocity at which this optimum occursgenerally increased as the wall shear stress increased for eachselectin.

Table 2. R-Squared Values from Fit of Experimental RollingVelocities to the Exponentially Modified Gaussian Equation forCD34+ KG1a and CD34- HL60 Cells on P-selectin-IgG and theCombination of P-selectin-IgG and Anti-human CD34 Antibody

dyn/cm2

1 3 5 8

P-selectin-IgG CD34- HL60 cells 0.96 0.99 0.94 0.97 0.92CD34+ KG1a cells 0.90 0.73 0.89 0.95 0.63

P-selectin-IgG+anti-CD34

CD34- HL60 cells 0.94 0.97 0.96 0.91 0.66

CD34+ KG1a cells 0.97 0.91 0.98 0.95 0.93

Figure 5. Velocity distributions of AML cells on P-selectin-IgG in the absence and presence of anti-human CD34 antibody. The exponentiallymodified Gaussian equation (EMG) was used, and a velocity bin-size of 3µm/s was adequate to describe the experimental rolling velocities (EXP)of CD34+ KG1a and CD34- HL60 cells on P-selectin-IgG in the absence and presence of anti-CD34 antibody. (A) There was significant overlapof the velocity distribution of CD34- HL60 and CD34+ KG1a cells rolling on immobilized P-selectin-IgG alone, resulting in very low peak-to-peak resolutions and suggesting that CD34+ KG1a enrichment is unlikely under these conditions. (B) When anti-CD34 antibody is co-immobilizedwith P-selectin-IgG, peak separation becomes evident, resulting in higher peak-to-peak resolutions and making CD34+ KG1a cell enrichment amore feasible option.


Although CD34+ ABM cells and CD34- ABM cells exhibitan inherent difference in rolling velocities, this may not be truefor all cell subsets. We investigated whether co-immobilizedselectin and a suitable antibody was sufficient to create adifference in rolling velocity between two cells subsets thatwould otherwise have similar rolling velocities on immobilizedselectins alone.

CD34+ KG1a cells have lower rolling velocities thanCD34- HL60 cells on immobilized P-selectin and anti-human CD34 antibody.CD34+ KG1a and CD34- HL60 cellsuspensions were also perfused over immobilized P-selectin-IgG at wall shear stresses ranging from 1 to 10 dyn/cm2 (Figure4A,B) and exhibited no difference between rolling velocities(p ) 0.155) or rolling fluxes (p ) 0.280), suggesting that cellenrichment with immobilized P-selectin-IgG alone would beunsuccessful for these two cell lines. Eniola et al. (25) foundthat beads coated with sialyl Lewisx and anti-ICAM-1 antibodygenerally rolled more slowly over immobilized P-selectin andICAM-1 than if they rolled on P-selectin alone. We appliedthis principle to CD34+ KG1a cells and CD34- HL60 cellsby co-immobilizing anti-human CD34 antibody with the P-selectin-IgG.

When anti-human CD34 antibody was co-immobilized withthe P-selectin-IgG, there was a decrease in rolling velocity ofthe CD34+ KG1a cells and an increase in the rolling velocity

of CD34- HL60 cells, resulting in a significant difference (p) 0.0009) at all shear stresses tested (Figure 4C). The increasein CD34- HL60 rolling velocity on the combined selectin-antibody surface was most likely due to a competitive decreasein selectin surface density. On the selectin-antibody surface,CD34+ KG1a rolling flux appears to have been reduced (Figure4D), but this is only a consequence of the definition of a rollingcell. Many CD34+ KG1a cells tended to roll less than fourcell diameters before stopping due to the effects of the antibodyand hence did not contribute to the rolling velocity or rolling

Figure 6. Time-dependent depletion of ABM cells on P-, E-, andL-selectin-IgG. The number of cells retained by the immobilized selectinsurfaces generally decreases with time, although the rate of this decreaseis highly selectin-dependent. At the start of flow, att ) 0, 60% ofCD34+ ABM cells is retained on immobilized (A) P-selectin- IgG at3 dyn/cm2, (B) E-selectin- IgG at 3 dyn/cm2, and (C) L-selectin-IgG at5 dyn/cm2, compared to only 20-30% for CD34- ABM cells. Therate of cell loss is highest on the L-selectin-IgG surface and lowest onthe E-selectin-IgG surface.

Figure 7. Time-dependent depletion of AML cells on P-selectin-IgGin the presence of anti-CD34 antibody. Rolling cells lose contact withthe surface when subjected to shear flow for extended periods of time.CD34+ KG1a cells are preferentially retained by the combined selectin-antibody surface at 3 dyn/cm2. CD34- HL60 cells are depleted fromthe surface at a faster rate than CD34+ KG1a cells and become entirelydepleted within 24 min, whereas it would take 56 min to fully depleteCD34+ KG1a cells.

Figure 8. Predicting CD34+ ABM cell purity and recovery fromvelocity distributions and cell retention data for P-selectin-IgG. (A)When physiologic concentrations of 1-5% CD34+ ABM cells areincorporated into the velocity distributions to show the effect ofconcentration, the distribution of the CD34+ ABM cells are almostentirely contained within the distribution of the CD34- ABM cellsand there is no obvious point of optimum separation. (B) Using a deviceworking length of 1 mm, optimum cell purities/recoveries ranging from24.5% to 36.1% are possible in less than 5 min. (C) A longer workinglength of 5 mm does not increase the optimum enrichment (26.9-33.1%), but very pure populations of CD34+ ABM cells (>70%) canbe obtained within 15 min, with∼25% recovery.


flux. Statistically, the difference in flux between CD34+ KG1acells and CD34- HL60 cells was not significant (p ) 0.062).Immobilized anti-CD34 antibody alone did not support rolling.Cells either remained firmly adherent or were released into thefree stream when subjected to flow.

Peak-to-peak resolution between CD34+ KG1a andCD34- HL60 cells velocity distributions increases on theimmobilized selectin-antibody surface.There was significantoverlap of the velocity distributions for CD34+ KG1a andCD34- HL60 cells rolling on immobilized P-selectin-IgG alone(Figure 5A). The overlap was so extensive that, in many cases,the peak-to-peak resolution was close to zero (<0.05), so thatCD34+ KG1a cell enrichment from a mixture with CD34-HL60 cells using immobilized P-selectin-IgG alone would bedifficult. When anti-human CD34 antibody was co-immobilizedwith P-selectin-IgG, however, the peak-to-peak resolutionincreased by more than 5-fold, to between 0.22 and 0.32 (Figure5B), indicating that some cell enrichment should be achievableon the basis of rolling velocity alone. While these values areimproved over those obtained on P-selectin-IgG alone, itsuggests that there would still be considerable contaminationof CD34+ KG1a cells with CD34- HL60 cells for separationsbased solely on differences in rolling velocity. The rollingvelocities at which optimum cell separation occurs increasesas the shear stress increases for both the P-selectin-IgG aloneand the combination selectin-antibody surface.

During the course of experimentation, we also observed thatall cells did not remain bound to the immobilized selectinsurfaces, even at the lowest wall shear stress and irrespective

of whether they were allowed to settle and interact with theimmobilized selectins under static conditions for significantperiods of time. Since the fraction of cells retained has aconsiderable impact on the resulting cell purity and recovery,cell retention experiments were also performed to determinethe true potential of selectin-mediated cell enrichment.

A Coulter counter was used to analyze the cell sizes forcalculation of the Stokes sedimentation velocity (28). CD34+ABM cell diameter was 6µm while both CD34- HL60 andCD34+ KG1a cells had a median diameter of 10.5µm. On thebasis of the particle size and calculated Stokes velocities, wedetermined that CD34+ ABM cells require at least 2 min tosettle completely from a static suspension under the influenceof gravity in our system, whereas the AML cell lines onlyrequired 30 s.

CD34+ ABM cells are retained to a greater extent thanCD34- ABM cells on all selectins, and CD34+ KG1a cellsare retained to a higher extent than CD34- HL60 cells onthe co-immobilized P-selectin-IgG and anti-human CD34antibody surface.We quantified the number of cells retainedby the immobilized selectin surfaces by allowing isolatedCD34+ cells and CD34- cells to settle out of the suspensionand interact with the immobilized selectins for 2 min prior tothe start of flow. A wall shear stress was applied to the cellsfor up to 10 min, and the number of each cell type remainingon the surface was enumerated at 2 min intervals, allowing thecell retention to be calculated according to eq 3. The analysiswas restricted to a wall shear stress of 3 dyn/cm2 on P-selectin-IgG and E- selectin-IgG, and 5 dyn/cm2 on L-selectin-IgG for

Figure 9. Predicting CD34+ ABM cell purity and recovery fromvelocity distributions and cell retention data for E-selectin-IgG. (A)Using physiologic concentrations of 1-5% CD34+ ABM cells altersthe velocity distribution curves so that the distribution of the CD34+ABM cells are entirely contained within the distribution of the CD34-ABM cells. (B) A working length of 1 mm results in optimum cellpurities/recoveries ranging from 6.2% to 21.6% obtainable within 5min. (C) A longer working length of 5 mm does not increase theoptimum enrichment (5.2-21.3%), and longer perfusion times resultin higher purities but very low recoveries.

Figure 10. Predicting CD34+ ABM cell purity and recovery fromvelocity distributions and cell retention data for L-selectin-IgG. (A)Physiologic concentrations of 1-5% CD34+ ABM cells cause thevelocity distribution curves of CD34+ ABM cells to become entirelyenclosed within the distribution of the CD34- ABM cells, with noobvious point of peak separation. (B) A device working length of1 mm results in optimum cell purities/recoveries ranging from 13.3%to 32.2%. (C) A longer working length of 5 mm results in optimumpurities/recoveries ranging from 16.3% to 30.5%, and although longerperfusion times can give higher purities (>70%), the recovery wouldbe very low (<10%).


the ABM cells, and 3 dyn/cm2 on the co-immobilized P-selectin-IgG and anti-CD34 antibody surface for the AML cell lines.We chose not to continue the AML cell line experiments onimmobilized P-selectin-IgG alone since we had already deter-mined this condition to be unsuitable for the requirements ofthe enrichment device. The antibody alone did not supportrolling, so this condition was excluded from the retentionexperiments. Generally, the retention of all cell subsets decreasedwith time (Figures 6and 7).

CD34+ ABM cells are retained to a greater extent thanCD34- ABM cells on all selectins. The initial cell retentionwas very similar for all selectins, about 60% for CD34+ ABMcells and between 20% and 30% for CD34- ABM cells;however, the rate of cell loss from the surface was highlyselectin-dependent. Cell loss was highest on the immobilizedL-selectin-IgG (Figure 6C), lowest on the E-selectin-IgG (Figure6B), and median on P-selectin-IgG (Figure 6A), anotherindication of the relative strengths with which each selectin bindsits ligands. It is interesting to note that, for each immobilizedselectin surface, cell loss for both CD34+ and CD34- ABMcells is very similar. The results of the cell retention experimentswere particularly important for determining the time limitationsof selectin-mediated cell enrichment. The time-dependent natureof the enrichment process means that the duration of perfusioncannot be longer than the maximum residence time for theCD34+ cells within the device. CD34+ ABM cells will bedepleted from an immobilized P-selectin-IgG surface within 33min, while it would take 73 min on E-selectin-IgG and only 19min on L-selectin-IgG.

Similarly, CD34+ KG1a cells are retained to a greater extentthan CD34- HL60 cells on the combined selectin-antibodysurface, with the rate of cell loss being twice as high as forCD34- HL60 cells than CD34+ KG1a cells (Figure 7). At theend of 10 min, 71% of CD34+ KG1a cells and 39% of CD34-HL60 cells remained on the surface, and the entire CD34+KG1a cell population would be depleted from the combinedselectin-antibody surface within 56 min. To obtain close to 100%purity for CD34+ KG1a cells, the device must run for 24 minin order to release all CD34- HL60 cells, at which point 52%of CD34+ KG1a cells will be retained.

Final CD34+ cell purity is highly dependent on startingpurity and the length of the device. The device is achromatography-type technology, with the rolling cells as themobile phase and the immobilized protein surfaces representingthe stationary phases. The optimum enrichment point representsthe best balance of cross-contamination and recovery and is theintersection of the velocity distributions of the two cell subsetsto be separated. The enrichment process depends on the workinglength of the device, which defines the relative contributionsof differential rolling velocity and cell retention to the overallprocess.

Velocity distribution curves can be used to illustrate therelative concentrations of the two cell subsets to be separated.In Figures 3 and 5, the area under both CD34+ and CD34-ABM cell velocity distribution curves were equal, representinga 1:1 mixture of the two populations. However, physiologicalconcentrations of CD34+ ABM cells tend to be low, usuallyranging from 1% to 5% (1, 5), and this must be incorporatedinto the model before accurate predictions of the enrichmentpotential of a selectin-mediated device can be developed. Bymodifying the areas under the curves to represent 1-5% CD34+ABM cells, a larger fraction of the CD34+ ABM cell velocitydistributions become contained within the distributions of theCD34- ABM cells and the optimum separation point is not

easily identified (Figures 8A, 9A, 10A). Calculation of peak-to-peak resolution is impossible in these cases.

By defining a working length for the device, we can extractpurity and recovery data from these modified velocity distribu-tion curves and the cell retention data presented in Figures 6and 7. A short working length of 1 mm means that optimumcell enrichment (intersection of purity and recovery curves) takesplace in a relatively short time. For ABM MNCs on P-selectin-IgG, 1-5% CD34+ ABM cells can be enriched to optimumpurities/recoveries ranging from 24.5% to 36.1% within 5 min(Figure 8B). Likewise, on E-selectin-IgG (Figure 9B) andL-selectin-IgG (Figure 10B), enrichment can range from 6.2%to 21.6% and 13.3% to 32.2%, respectively, in the same timerange. Increasing the working length to 5 mm increases the timeto reach the optimum but has minimum effect on the optimumpurities, with 1-5% CD34+ ABM cells being enriched to26.9-33.1%, 5.2-21.3%, and 16.3-30.5% on P-selectin-IgG(Figure 8C), E-selectin-IgG (Figure 9C), and L-selectin-IgG(Figure 10C), respectively. The major benefit of a longerworking length is the ability to achieve higher purities yet stillrecover a portion of CD34+ ABM cells as time elapses. Forexample, 22.4% of CD34+ ABM cells can be recovered on a5 mm working length of immobilized P-selectin-IgG within 15min when the purity jumps to 100% (Figure 8C) due to depletionof CD34- ABM cells. From the conditions tested, P-selectin-IgG appears to be the selectin best suited for selectin-mediatedenrichment of CD34+ ABM cells from ABM MNCs since itgives the highest optimum cell recoveries and purities.

The separation of the velocity distributions is more definedwith the AML cell lines (Figure 11A) so that 1-5% starting

Figure 11. Predicting CD34+ KG1a cell purity and recovery fromvelocity distributions and cell retention data for the combined P-selectin-IgG and anti-CD34 antibody surface. (A) The velocity distributions of1-5% CD34+ KG1a cells in a mixture with CD34- HL60 cells stillshow defined points of peak separation, suggesting that significantenrichment should still be possible at these low starting concentrations.(B) Using a working length of 1 mm, optimum CD34+ KG1a cellpurities/recoveries ranging from 42.3% to 51.5% are possible in lessthan 5 min. (C) A longer working length of 5 mm reduces the optimumenrichment to 31.5-40.9%.


purity of CD34+ KG1a cells gives much higher optimumpurities/recoveries than the ABM cells. On a 1 mmworkingsurface, optimum purities range from 42.3% to 51.5% (Figure11B), whereas a longer working length of 5 mm decreases thatrange to 31.5-40.9% (Figure 11C).

The data suggest that a shorter device that takes advantageof the differences in rolling velocity and where the optimum isachieved in a relatively short time, hence reducing the time-dependent cell loss that was illustrated by Figures 6 and 7, islikely the best option for selectin-mediated separations. In mostcases, optimum separation can be achieved in less than 10 min,with significant increases in purity from the starting cellconcentrations of 1-5%. A longer device might be suitable incases where the rate of cell loss is low and hence very purepopulations of cells can be recovered with longer exposure toflow (Figure 9C). Changing the working length of the devicehas dramatic effects on the time course of the enrichmentprocess, and it would ultimately be up to the user to define anoptimum based on the intended application of the enriched cellfraction.

ConclusionsWe have shown that cell enrichment mediated by immobilized

selectins can be a feasible technology. Suspensions of 1-5%

CD34+ cells can be enriched by more than 5-fold in most cases,with optimum enrichment possible within 10 min. The proposeddevice could be designed to take advantage of both a differentialrolling velocity and cell retention to facilitate enrichment ofone subset of cells from a mixture. Figure 12 summarizes theparameters that influence selectin-mediated cell enrichment.

The fact that cells do not remain in a permanently rollingstate in the presence of the selectin is also advantageous forapplications in which the surface is meant to capture cells froma flowing suspension, deliver a signal, and then subsequentlyrelease the targeted cells back into the flowing suspension. Insuch a case, there would be no need to develop a separatemechanism to release cells once they are captured by the surface.A multiple-stage system may also be useful for improvingpurities and yields. The technology has great potential and canbe optimized for each type of suspension that it is required toprocess.

Acknowledgment

The authors gratefully acknowledge technical assistance fromMrs. Karen Rosell. This work was funded by a James D. WatsonInvestigator Award from the New York State Office of Science,Technology, and Academic Research (NYSTAR), to M.R.K.

Figure 12. Summary of the parameters involved in selectin-mediated cell enrichment. The success of selectin-mediated enrichment of a sampledepends on the cell retention and differential rolling velocities on the immobilized selectins. These key parameters are influenced by the type ofselectin used and whether an antibody is co-immobilized with the selectin, the concentrations used, the shear stress employed, and the length of thedevice. The third key parameter, the starting purity, is dependent on the donor and directly determines the extent of enrichment that is possible.


Michael R. King serves on the scientific advisory board ofCellTraffix, a company in which he holds financial interest.

References and Notes(1) Cornetta, K.; Gharpure, V.; Mills, B.; Hromas, R.; Abonour, R.;

Broun, E. R.; Traycoff, C. M.; Hanna, M.; Wyman, N.; Danielson,C.; Gonin, R.; Kunkel, L. K.; Oldham, F.; Srour, E. F. Rapidengraftment after allogeneic transplantation using CD34-enrichedmarrow cells.Bone Marrow Transplant1998, 21 (1), 65-71.

(2) Farag, S. S. Chronic graft-versus-host disease: where do we gofrom here?Bone Marrow Transplant2004, 33 (6), 569-577.

(3) Abonour, R.; Scott, K. M.; Kunkel, L. A.; Robertson, M. J.;Hromas, R.; Graves, V.; Lazaridis, E. N.; Cripe, L.; Gharpure, V.;Traycoff, C. M.; Mills, B.; Srour, E. F.; Cornetta, K. Autologoustransplantation of mobilized peripheral blood CD34+ cells selectedby immunomagnetic procedures in patients with multiple myeloma.Bone Marrow Transplant1998, 22 (10), 957-963.

(4) Baum, C. M.; Weissman, I. L.; Tsukamoto, A. S.; Buckle, A. M.;Peault, B. Isolation of a candidate human hematopoietic stem-cellpopulation.Proc. Natl. Acad. Sci. U.S.A.1992, 89 (7), 2804-2808.

(5) Ema, H.; Suda, T.; Miura, Y.; Nakauchi, H. Colony formation ofclone-sorted human hematopoietic progenitors.Blood1990, 75 (10),1941-1946.

(6) Ibrahim, S. F.; van den Engh, G. High-speed cell sorting:fundamentals and recent advances.Curr. Opin. Biotechnol.2003,14 (1), 5-12.

(7) Chalmers, J. J.; Zborowski, M.; Sun, L.; Moore, L. Flow through,immunomagnetic cell separation.Biotechnol. Prog.1998, 14 (1),141-148.

(8) de Wynter, E. A.; Ryder, D.; Lanza, F.; Nadali, G.; Johnsen, H.;Denning-Kendall, P.; Thing-Mortensen, B.; Silvestri, F.; Testa, N.G. Multicentre European study comparing selection techniques forthe isolation of CD34+ cells. Bone Marrow Transplant1999, 23(11), 1191-1196.

(9) Firat, H.; Giarratana, M. C.; Kobari, L.; Poloni, A.; Bouchet, S.;Labopin, M.; Gorin, N. C.; Douay, L. Comparison of CD34+ bonemarrow cells purified by immunomagnetic and immunoadsorptioncell separation techniques.Bone Marrow Transplant1998, 21 (9),933-938.

(10) Cancelas, J. A.; Querol, S.; Martin-Henao, G.; Canals, C.; Azqueta,C.; Petriz, J.; Ingles-Esteve, J.; Amill, B.; Garcia, J. Isolation ofhematopoietic progenitors. An appraoch to two different immuno-magnetic methods at the lab scale.Pure Appl. Chem.1996, 68 (10),1897-1901.

(11) Johnsen, H. E.; Hutchings, M.; Taaning, E.; Rasmussen, T.;Knudsen, L. M.; Hansen, S. W.; Andersen, H.; Gaarsdal, E.; Jensen,L.; Nikolajsen, K.; Kjaesgard, E.; Hansen, N. E. Selective loss ofprogenitor subsets following clinical CD34+ cell enrichment bymagnetic field, magnetic beads or chromatography separation.BoneMarrow Transplant1999, 24 (12), 1329-1336.

(12) Gallacher, L.; Murdoch, B.; Wu, D. M.; Karanu, F. N.; Keeney,M.; Bhatia, M. Isolation and characterization of human CD34(-)-Lin(-) and CD34(+)Lin(-) hematopoietic stem cells using cell surfacemarkers AC133 and CD7.Blood 2000, 95 (9), 2813-2820.

(13) Handgretinger, R.; Gordon, P. R.; Leimig, T.; Chen, X.; Buhring,H. J.; Niethammer, D.; Kuci, S. Biology and plasticity of CD133+hematopoietic stem cells.Ann. N. Y. Acad. Sci.2003, 996, 141-51.

(14) Kuci, S.; Wessels, J. T.; Buhring, H. J.; Schilbach, K.; Schumm,M.; Seitz, G.; Loffler, J.; Bader, P.; Schlegel, P. G.; Niethammer,D.; Handgretinger, R. Identification of a novel class of humanadherent CD34- stem cells that give rise to SCID-repopulating cells.Blood 2003, 101 (3), 869-876.

(15) Storms, R. W.; Goodell, M. A.; Fisher, A.; Mulligan, R. C.; Smith,C. Hoechst dye efflux reveals a novel CD7(+)CD34(-) lymphoidprogenitor in human umbilical cord blood.Blood 2000, 96 (6),2125-2133.

(16) Greenberg, A. W.; Kerr, W. G.; Hammer, D. A. Relationshipbetween selectin-mediated rolling of hematopoietic stem and pro-genitor cells and progression in hematopoietic development.Blood2000, 95 (2), 478-486.

(17) Greenberg, A. W.; Hammer, D. A. Cell separation mediated bydifferential rolling adhesion.Biotechnol. Bioeng.2001, 73 (2), 111-124.

(18) Alon, R.; Feizi, T.; Yuen, C. T.; Fuhlbrigge, R. C.; Springer, T.A. Glycolipid ligands for selectins support leukocyte tethering androlling under physiologic flow conditions.J. Immunol.1995, 154(10), 5356-5366.

(19) Geng, J. G.; Heavner, G. A.; McEver, R. P. Lectin domain peptidesfrom selectins interact with both cell surface ligands and Ca2+ ions.J. Biol. Chem.1992, 267 (28), 19846-19853.

(20) Kansas, G. S. Selectins and their ligands: current concepts andcontroversies.Blood 1996, 88 (9), 3259-3287.

(21) Borsig, L. Selectins facilitate carcinoma metastasis and heparincan prevent them.News Physiol. Sci.2004, 19, 16-21.

(22) Borsig, L.; Wong, R.; Feramisco, J.; Nadeau, D. R.; Varki, N.M.; Varki, A. Heparin and cancer revisited: mechanistic connectionsinvolving platelets, P-selectin, carcinoma mucins, and tumor me-tastasis.Proc. Natl. Acad. Sci. U.S.A.2001, 98 (6), 3352-3357.

(23) Borsig, L.; Wong, R.; Hynes, R. O.; Varki, N. M.; Varki, A.Synergistic effects of L- and P-selectin in facilitating tumormetastasis can involve non-mucin ligands and implicate leukocytesas enhancers of metastasis.Proc. Natl. Acad. Sci. U.S.A.2002, 99(4), 2193-2198.

(24) Kim, Y. J.; Borsig, L.; Varki, N. M.; Varki, A. P-selectindeficiency attenuates tumor growth and metastasis.Proc. Natl. Acad.Sci. U.S.A.1998, 95 (16), 9325-9330.

(25) Eniola, A. O.; Willcox, P. J.; Hammer, D. A. Interplay betweenrolling and firm adhesion elucidated with a cell-free systemengineered with two distinct receptor-ligand pairs.Biophys. J.2003,85 (4), 2720-2731.

(26) King, M. R.; Hammer, D. A. Multiparticle, adhesive dynamics.Interactions between stably rolling cells.Biophys. J.2001, 81 (2),799-813.

(27) Chang, W. C.; Lee, L. P.; Liepmann, D. Biomimetic techniquefor adhesion-based collection and separation of cells in a microfluidicchannel.Lab Chip2005, 5 (1), 64-73.

(28) Leal, L. Laminar Flow and ConVectiVe Transport Processes-Scaling Principles and Asymptotic Analysis; Butterworth-Heine-man: Woburn, MA, 1992.

Received July 11, 2007. Accepted September 14, 2007.

BP0702222


Investigating the Feasibility of Stem Cell Enrichment Mediated by Immobilized Selectins

Documents

Transcript of Investigating the Feasibility of Stem Cell Enrichment Mediated by Immobilized Selectins