Proceedings of the 2nd IEEE International

Conference on Nano/Micro Engineered and Molecular Systems

January 16 - 19, 2007, Bangkok, Thailand

On-Chip Continuous Blood Cell Subtype Separation by

Deterministic Lateral Displacement
Nan Li', Daniel T. Kamei2, and Chih-Ming Ho'
'Department ofMechanical & Aerospace Engineering, University ofCalifornia, Los Angeles, CA, USA
2
DepartmentofBioengineering, University of California, Los Angeles, CA, USA
Abstract-This paper presents a microfluidic device for blood cell subtype, since that would yield information
continuous human blood cell subtype separation using the regarding the human body infectious status and immune
deterministic lateral displacement principle. Based on their responses. For example, the diagnosis of HIV infection
significant size and shape differences, three major cell types of depends on isolation of human CD4+ lymphocytes from whole
human whole blood - platelets, red blood cells and white blood blood [2]. Besides RBCs and WBCs, another major component
cells - were demonstrated to be directly separated using a two- of whole blood is the platelets with a diameter between 2pim to
stage separation strategy. Even though all white blood cells are 3ptm. The ratio of platelets to RBCs to WBCs in normal human
spherical and have diameters within a narrow range (8 - 20pm), whole blood is about 50:500:1. The large amount and apparent
the initial limitation for using this principle to separate white small size of platelets makes it possible to directly separate
blood cell subtypes was conquered by attaching larger
polystyrene microbeads to one of the subtypes to amplify the size them from RBCs and WBCs using the deterministic lateral
differences. Specifically, continuous separation of human CD4+ T displacement technique. Furthermore, WBCs include five
helper lymphocytes from other white blood cell subtypes was subtypes (monocytes, lymphocytes, neutrophils, basophils and
achieved with high purity and recovery due to the underlying eosinophils) and their diameters are all within a narrow range
high affinity and high specificity of the antigen-antibody between 8pm and 20pm which poses limitations for using the
interaction used to attach the microbeads to the lymphocytes. deterministic lateral displacement technique to separate them
With our novel approach, the pure population of one blood cell from each other. However, all of these WBC subtypes express
subtype can be effectively isolated by exploring the deterministic their unique surface antigens, which can be specifically
lateral displacement principle, which has the advantages of the targeted by antibodies. This property provides the possibility
simplicity, high speed and high resolution. Because many cells for attaching microbeads to each WBC subtype through
express unique surface markers, this method can theoretically be specific antigen-antibody interaction so that size differences
applied to separate any target cell type from a heterogeneous can be "amplified". Although attaching microbeads to cells has
mixture for downstream analysis. been previously demonstrated to effectively separate one
specific cell type from others by both electrical and optical
Keywords-microfluidics; blood cell separation; deterministic methods [15, 16], the novelty of this approach is combining the
lateral displacement specificity of microbeads attachment with the simplicity of the
deterministic lateral displacement technique. Because of the
I. INTRODUCTION high affinity and specificity associated with antigen-antibody
Isolation of a pure biological cell population from a interaction, high recovery ratio and purity may be achieved.
heterogeneous mixture of different cell types plays a very In this paper, we first demonstrate successful separation of
important role in both clinical diagnosis and academic research. human whole blood cells into three major groups using a two-
Microfluidic cell separation devices provide more advantages stage deterministic lateral separation strategy. By attaching
compared with their conventional counterparts in smaller 25pm diameter polystyrene microbeads, CD4+ T helper
sample volume consumption, lower cost and easier for lymphocytes are successfully isolated from other WBC
integration with downstream analytical components [1, 2]. subtypes using deterministic lateral displacement technique.
Integrated microchips for separating and isolating cells Disposable microfluidic devices using this method are capable
utilizing hydrodynamic, dielectrophoretic (DEP) and magnetic of achieving continuous WBC subtype separation with high
forces have been demonstrated within recent years [3-10]. throughput, recovery and purity.
Relative to these strategies, however, cell separation based
purely on size differences is the simplest approach and II. PRINCIPLE
possesses the advantages of low cost and high throughput.
Deterministic lateral displacement separation method
Deterministic lateral displacement is a novel size-based utilizes the asymmetric bifurcation of laminar flow
separation mechanism with high resolution (up to 0.1ILm) [11, phenomenon around an array of micro obstacles. These micro
12]. This principle has been explored for separating human red obstacles are shifted a specific distance AX from the previous
blood cells (RBCs, 6-7pm) from white blood cells (WBCs, 8- row. This shifted distance AX, together with the gap g and the
20jim) based on their distinct size differences [13, 14]. center-to-center distance between two adjacent obstacles i,
However, most clinical diagnosis and academic research start determines the critical separating particle size. When particles
with whole blood separation and are focused on specific white with different sizes flow through this micro obstacles array,

This project was funded by NASA National Space Biomedical Research

Institute (NSBRI). The co-operative agreement number is NCC 9-58-317.
* Contacting Author: Engineering IV 18-111, 420 Westwood Plaza, Los
Angeles, CA 90095, USA (Email: [email protected])

1-4244-0610-2/07/$20.00 C)2007 IEEE 932

they will follow specific pathways "deterministically". Microbeads and Reagent
Particles smaller than the critical separating size will change Polybead® microbeads used in this study were obtained
the path lane periodically and thus follow a "zigzag" mode, from Polysciences Inc. (Warrington, PA). Monoclonal
whereas particles larger than the critical separating size will antibodies (mAbs) (FITC-conjugated anti-human CD4) and red
keep getting "bumped" when they travel to the next row of the blood cell lysis buffer were obtained from eBioscience (San
obstacles and thus be displaced to one direction. Assuming the Diego, CA). Human whole blood with heparin as anticoagulant
flow profile between the micro obstacles is parabolic which was obtained from Innovative Research, Inc. (Southfield,
holds true at low Reynolds number, the critical separation Michigan) and used while fresh. Phosphate buffered saline
diameter D can be numerically determined by [12]: (PBS), bovine serum albumin (BSA), heparin sodium and
I
glycerol were obtained from Sigma (St. Louis, MO).
D=g 1+2w+
2w] (1) Antibody Coating on Microbeads
where Coating of the microbeads with mAbs was achieved by
pure protein adsorption on the polystyrene surface. 0.5mL of a
W=
I
-+ 168 (e 1) ]1 (113)(- 21 i 2
8
-.3) 25ptm bead suspension at a concentration of 5.68X 106 beads/ml
-8 4
were mixed with 100pg FITC conjugated anti-human CD4
and £ is defined as the row shift fraction which equals AA / A . antibody in lmL PBS and incubated overnight with gentle
c is not linearly proportional to the product of £ and g is vortexing at room temperature. This corresponded to 100pg
mAbs per 2.2 x 10'° Rm2 of bead surface area following the
because of the non-uniform flow flux between the micro protocol of Fortin and Hugo [16]. The beads were then washed
obstacles. twice with 1XPBS buffer. The excess hydrophobic binding
sites on the beads were blocked by incubating with 1XPBS
III. MATERIALS AND METHODS containing 2% bovine serum albumin (PBS/BSA) for 30min at
room temperature. Finally, these mAbs-coated beads were
Device Fabrication washed and suspended in 1 XPBS to a concentration of 1 x 106
Microfluidic separation devices are fabricated using soft beads/mL. The degree of mAbs adsorption was visualized with
lithography with polydimethylsiloxane (PDMS, Sylgard 184, a fluorescence microscope.
Dow Coming, MI) polymer from deep reactive ion etched
(DRIE) silicon mold. The effective separation area is 6.8mm Attaching mAbs-Coated Microbeads to Cells
by 1.6mm. At the inlet, samples are hydrodynamically focused Human blood cells were first harvested by centrifuging
by two streams of sheath buffer (Figure 1(A)). A separation lmL of human whole blood at 1,500 x g for 5min and the
distance measurement ruler near the outlet is used to quantify upper plasma layer was disposed. Red blood cells were lysed
the separation (Figure 1(B)). The important device parameters two times using lysis buffer following manufacture-supplied
for separation, row-to-row shift distance AX, gap g and center- protocol. The remaining white blood cells were then suspended
to-center distance between two adjacent obstacles i, are
illustrated in Figure 1(C). in IXPBS to a final concentration of 1 X 106 cells/mL. 50[tL of
this processed white blood cell suspension was then mixed
with 500[tL of the mAbs-coated 25ptm microbeads solution and
incubated for 60min with gentle vortexing at room temperature.
The samples were then washed twice and resuspended in lmL
PBS.

Cell Staining with Fluorochrome-Conjugated mAbs

To measure the microbeads-cell attachment efficiency and
specificity, a second round cell staining with FITC conjugated
k mAbs was performed. 10Og FITC conjugated anti-human
CD4 were added to lmL cell-beads mixture and incubated for
30min at room temperature. Then bead-cell mixture was
washed twice with PBS. Staining effect was visualized with
fluorescence microscope. Quantitation of the cell-beads
binding efficiency and specificity was achieved by counting
how many percent of fluorescently labeled cells were bonded
to microbeads and how many percent of beads attached cells
Flow were fluorescently labeled using a hemocytometer.

Separation Experiments Setup

Figure 1. (A) Device inlet. (B) Device outlet. (C) High magnification of the For the separation experiments, Tygon tubings with inner
micro obstacles with critical parameters labeled. diameter of 0.02 inches (Fisher Scientific, Tustin, CA) were
inserted into the PDMS device at the inlet and outlet. The

933
device was placed onto a Leica inverted microscope equipped
with a MotionPro MP1 000 high speed camera (Redlake MASD 200
LLC, Tucson, AZ) for visualization and image capture. All I +GOullm s

videos were captured at a speed of 60 frames/second. Two 150

syringe pumps (Harvard Apparatus, Holliston, MA) were used I 1.Oins11lW lI
to drive the cell suspension and sheath buffer into the device at
a constant flow rate. For the microbeads separation
experiments, microbeads were first incubated with PBS/BSA
100
- - - ------------------- - - -1 - -t -------------
for 30min to prevent non-specific adhesion on device surface.
For cell separation experiments, 1 XPBS containing 1% heparin
sodium (PBS/heparin) was first pumped into the device from
the sample inlet and sheath inlets to drive out the air bubbles. 0
Then this solution was kept flowing at l1tL/min for 10min to -20 40 100 160 220 280 340 400

allow heparin to adsorb on the device surface in order to separatio n dlista nce at th e ouitlet ({ ni)
prevent cell adhesion. To prevent microbeads and cells from
settling during the separation experiment, 20% glycerol was Figure 3. Separation profiles of 8ptm and 9ptm polystyrene beads at two
added into each sample solution to adjust the solution density. different flow velocities.

IV. RESULTS preferred because higher throughput can be achieved as long as

the device is still operated at the laminar flow regime.
Separation of Microbeads
Polystyrene microbeads with diameter of 8, 9 and 10~Lm Separation ofHuman Blood Cells
were separated from each other. The device parameters for Fresh human whole blood was directly diluted 50 times
8pm and 9pm beads separation were AX =6Rm, X=80Rm and with PBS/heparin and separated immediately to prevent
g=25jm with a theoretical critical separation particle diameter coagulation. A two-stage separation device was used to
of 8.4pm. Experimental results show that the 8pm beads follow separate whole human blood cells into three major groups:
the zigzag mode whereas the 9pm beads follow the platelets, RBCs and WBCs. The first stage of the separation
displacement mode ((Figure 2(A)). The device parameters for device includes 36 rows of micro obstacles with AX =2Rm
9pm and IO0m beads separation were AX =8Rm, X=90%m and distance shifted for each row. The second stage includes 49
g=25jm with a theoretical critical separation particle diameter rows of micro obstacles with AX=5Rm distance shifted for each
of 9.2pm. Experimental results show that, in this case, the 9pm row. The center-to-center distance i (80pm) and the gap g
beads follow the zigzag mode whereas the IO0m beads follow (20pm) are the same for both stages. Theoretically, the critical
the displacement mode. ((Figure 2(B)). The experimental separation diameters for the first and the second stages are
results are in well agreement with the theoretical predictions. 3.8pm and 6.1 pm respectively. Therefore, separation of
At the outlet of the device, each group can be effectively platelets from RBCs and WBCs are expected in the first stage,
separated by at least a 20Opm distance. Flow velocity was also whereas further separation of the platelets, RBCs and WBCs
investigated and found to have negligible influence on the from each other are expected in the second stage.
separation results (Figure 3). Accordingly, the high velocity is
Figure 4 shows the separation results of human whole
blood cells. In the first stage, platelets follow the zigzag mode
due to their smaller size compared with the critical separation
diameter, whereas both RBCs and WBCs were kept displaced
as they passed each row. At the transition from the first stage to
the second stage, platelets were separated by about 10Oim
distance from RBCs and WBCs. When they entered the second
stage, RBCs changed their pathways from displacement mode
to zigzag mode whereas WBCs still proceeded in displacement
mode.
Statistical analysis of the separation effects is shown in
Figure 5. The effective separation distance between platelets
and RBCs was about 200pm and that between RBCs and
WBCs was about 250pm. However, the distribution profile
associated with the platelets was larger than that associated
with the RBCs and WBCs. This resulted from their higher
diffusion rates throughout the separationg process due to their
smaller sizes (2-3pm). The ratio of separated RBCs to platelets
Figure 2. Separation of polystyrene microbeads. (A) Trace of 8pm and 9pim to WBCs was found about 470:36:1, which is similar to the
diameter beads. (B) Trace of 91im and 1Oim diameter beads. Sample and ratio of 500:50:1 in normal whole blood. A lower proportion of
sheath flow rates were 0.24L min and 1.8iL min respectively. Picture was platelets were quantified, and this was most probably attributed
generated by superimposing multiple video frames.

934
to the fact that some of the platelets were out of focus and diameter was about 23pim. Because the 25p.m beads are larger
therefore were not counted accordingly. Note that the total than almost all of WBCs, both beads and bead-cell complexes
separation throughput is about 1,000 cells per second. were expected to be isolated from the other WBC subtypes.
After separation, the isolated cells can be detached from the
microbeads by adding a second antibody as a competitor and
pure T helper lymphocytes can be harvested. Meanwhile, the
other WBC subtypes can also be harvested with this specific
lymphocytes depleted.
Since the deterministic lateral displacement principle
allows all of the 25pim beads and bead-cell complexes to be
isolated, the purity and recovery ratio for the target cells are
mainly determined by the beads attachment procedure. By
performing a second round cell staining using FITC conjugated
CD4 mAbs, we quantitatively determined that about 91% of T
helper lymphocytes were attached to the 25ptm beads and thus
could be separated afterwards. Also, about 98% of cells
__1E.2.2.m.E. * iS.g..1.
attached to the microbeads were stained fluorescently, which
indicated that almost all of the cells that attached to the beads
were T helper lymphocytes. The non-specific binding of other
cell types to the beads was therefore negligible. This high
purity is due to the especially high affinity and specificity
associated with the antigen-antibody interaction. Figure 6
shows a CD4+ T lymphocyte attached to a 25ptm bead and its
associated fluorescence.

Figure 4. Trace of platelets, RBCs, amd WBCs through the two stages of the
whole blood separation device. The sample and sheath flow rates wuie
0.1IL min and 1 iLtmin respectively. Picture was generated by superimposing
multiple video frames. (A) First stage. Separation of platelets (top pathway-
zigzag mode) from RBCs and WBCs (bottom pathway-displacement mode).
20X maginification. (B) Second stage. Separation of platelets (top pathway-
zigzag mode), RBCs (middle pathway-zigzag mode), and WBCs (bottom Figure 6. FITC stained CD4+ T helper lymphocyte attached to a 25pim
pathway-displacement mode) from each other. I OX magnification. polystyrene beads.

Once target cells were attached to microbeads, they were

run through the separation device. Figure 7 is the trace of a
WBC without bead bonded and a T helper lymphocyte bonded
to a 25pim bead. Clearly, since the gap is much larger than the
size of a WBC, the non-bead-bonded WBCs flowed through
the micro obstacles in a straight line with no direction changes
at all, whereas the bead-lymphocyte complex was kept
displaced by the micro obstacles.

Figure 5. Statistical analysis of the separation of three major human whole

blood cells: platelets, red blood cells (RBCs) and white blood cells (WBCs).

Separation of White Blood Cell Subtypes

To demonstrate the ability to separate WBC subtypes using
the deterministic lateral displacement technique, we attached
25ptm diameter polystyrene beads to the CD4+ T helper Figure 7. Separation of a T helper lymphocyte bonded to a 25pm bead from
lymphocytes through their unique surface marker. The device other WBC subtypes. Sample and sheath flow rates were 0.2iL min and
parameters for this separation were designed as AX=9pm, 1.2iL min respctively. Picture was generated by superimposing multiple
X=6Qim and g=47Rm so that the theoretical critical separation video frames.

935
Summary rate, however, may also pose a problem, since it can cause
Deterministic lateral displacement technique has been some cells to deform to some extent and not follow the
demonstrated to successfully separate polystyrene microbeads predicted pathway.
with different sizes, three major components of human whole In the future, the WBC subtype separating efficiency will
blood, and WBC subtypes with surface marker bound to be further characterized using fluorescence activated cell
microbeads. Figure 8 is the summary of all our experiments. sorting (FACS). Also, a micro cell Coulter counter, which has
Particles that followed the displacement mode are plotted as already been demonstrated in our lab [17], will be incorporated
filled squares, while particles that followed the zigzag mode into current device to count the separated cells.
are plotted as filled circles. All experiments were performed
under laminar flow conditions so that the flow profile between REFERENCES
the micro obstacles was parabolic. All "zigzaged" particles
were below the critical separation line, while all "displaced" [1] M. Toner and D. Irimia, "Blood-on-a-chip," Annual Review of
particles were above the critical separation line. The Biomedical Engineering, vol. 7, pp. 77-103, 2005.
experimental results are therefore consistent with the theory. [2] M. Radisic, R. K. Lyer, and S. K. Murthy, "Micro- and nanotechnology
in cell separation," International Journal of Nanomedicine, vol. 1(1), pp.
3-14, 2006.
[3] J. Takagi, M. Yamada, M. Yasuda, and M Seki, "Continuous particle
separation in a microchannel having asymmetrically arranged multiple
branches," Lab on a Chip, vol. 7(7), pp. 778-784, 2006.
ci_ [4] M. Yamada and M. Seki, "Hydrodynamic filtration for on-chip particle
concentration and classification utilizing microfluidics," Lab on a Chip,
vol. 5(11), pp. 1233-1239, 2005.
[5] M. Yamada, M. Nakashima, and M. Seki, "Pinched flow fractionation:
.__ __. .__

Continuous size separation of particles utilizing a laminar flow profile in

..q
El iFn 0
a pinched microchannel," Analytical Chemistry, vol. 76(18), pp. 5465-
Ilfiffll. 5471, 2004.
0
4A '01
[6] J.G. Kralj, M. T. W. Lis, M. A. Schmidt, K. F. Jensen, "Continuous
dielectrophoretic size-based particle sorting," Analytical Chemistry, vol.
78(14), pp. 5019-5025, 2006.
oilS p'Ar fat
- ofi 1icfCp r l x [7] S. Choi and J. K. Park, "Microfluidic system for dielectrophoretic
urifor fiot rofle separation based on a trapezoidal electrode array," Lab on a Chip, vol.
0 6 101i 20O 2 30 35 40 4 0 5(10), pp. 1161-1167, 2005.
Rw Shft Fractioni
[8] J. Park, B. Kim, S. K. Choi, S. Hong, S. H. Lee, and K. Lee, "An
Figure 8. Separation summary for microbeads and human blood cells using efficient cell separation system using 3D-asymmetric microelectrodes,"
the deterministic lateral displacement technique. Square: particles follow Lab on a Chip, vol. 5(11), pp. 1264-1270, 2005.
displacement mode; Circle: particles follow zigzag mode. [9] D. W. Inglis, R. Riehn, J. C. Sturm, and R. H. Austin, "Microfluidic
high gradient magnetic cell separation," Journal of Applied Physics, vol.
99(8), 2006.
CONCLUSION AND DISCUSSION
V. [10] D. W. Inglis, R. Riehn, R. H. Austin, and J. C. Sturm,, "Continuous
microfluidic immunomagnetic cell separation," Applied Physics Letters,
A disposable MEMS device for continuous separation of 85(21), pp. 5093-5095, 2004
human whole blood cells (platelets, RBCs and WBCs) and [11] L. R. Huang, E. C. Cox, R. H. Austin, and J. C. Sturm, "Continuous
white blood cell subtypes was demonstrated. The deterministic particle separation through deterministic lateral displacement," Science,
lateral displacement principle was explored for successful vol. 304(5673), pp. 987-990, 2004.
separation with high throughput and resolution. For whole [12] D. W. Inglis, J. A. Davis, R. H. Austin, and J. C. Sturm, "Critical
blood cells separation, the ratio of each separated group particle size for fractionation by deterministic lateral displacement," Lab
collected at the outlet matches very well with the ratio in on a Chip, vol. 6(5), pp. 655-658, 2006.
normal whole blood. Furthermore, by attaching microbeads to [13] S. Zheng, Y. Tai, and H. Kasdan, "Deterministic lateral displacement
a specific cell group, this technique can be extended to separate MEMS device for continuous blood cell separation," in Proc. of IEEE
MEMS 2005, pp. 851-854.
target cells from other groups with minimal size differences [14] S. Zheng, Y. Tai, and H. Kasdan, "A micro device for separation of
between them. Due to the high affinity and specificity erythrocytes and leucocytes in human blood," in Proc. of the 2005 IEEE
associated with the antigen-antibody interaction, high recovery Engineering in Medicine and Biology 27th Annual Conference, pp.
and purity can be achieved using this method. In principle, this 1024-1027.
strategy can be utilized for other cell separation applications [15] X. Hu, P. H. Bessette, J. Qian, C. D. Meinhart, P. S. Dargherty, and H. T.
because many cells express unique surface markers. Soh, "Marker-specific sorting of rare cells using dielectrophoresis,"
Proceedings of the National Academy of Sciences of the United States
Coating the device surface with heparin before experiments of America, vol. 102(44), pp. 15757-15761, 2005
was found to significantly decrease the cell adhesion and [16] M. Fortin and P. Hugo, "Surface antigen detection with non-fluorescent,
clogging problem. However, clogging became severe when the antibody-coated microbeads: An alternative method compatible with
blood was not fresh, since coagulant factors were released due conventional fluorochrome-based labeling," Cytometry, vol. 36(1), pp.
27-35, 1999.
to platelets activation. In addition, the flow rate was found to [17] C. Kwong, N. Li, and C. M. Ho, "Studies of deionization and impedance
have minor effects on the separation effects as long as the flow spectroscopy for blood analyzer," in Proc. SPIE 2005, Vol. 6003,
was in the laminar flow regime. Too high of an operating flow 60030N.

936