Sensors and Actuators B 245 (2017) 1050–1061

Contents lists available at ScienceDirect

Sensors and Actuators B: Chemical

journal homepage: www.elsevier.com/locate/snb

Centrifugal microfluidics for sorting immune cells from whole blood

Zeta Tak For Yu a,1 , Jophin George Joseph a,b,1 , Shirley Xiaosu Liu a , Mei Ki Cheung a ,
Parker James Haffey a , Katsuo Kurabayashi f , Jianping Fu a,c,d,e,∗
a
Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, 48109, USA
b
Department of Chemical Engineering, Indian Institute of Technology Hyderabad, Kandi, Sangareddy, Telangana, 502285, India
c
Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI, 48109, USA
d
Department of Cell and Developmental Biology, University of Michigan Medical School, Ann Arbor, MI, 48109, USA
e
Michigan Center for Integrative Research in Critical Care, University of Michigan, Ann Arbor, MI, 48109, USA
f
Department of Mechanical Engineering Electrical Engineering and Computer Science, University of Michigan, USA

a r t i c l e i n f o a b s t r a c t

Article history: Sorting and enumeration of immune cells from blood are critical operations involved in many clinical
Received 18 August 2016 applications. Conventional methods for sorting and counting immune cells from blood, such as flow
Received in revised form 5 December 2016 cytometry and hemocytometers, are tedious, inaccurate, and difficult for implementation for point-
Accepted 20 January 2017
of-care (POC) testing. Herein we developed a microscale centrifugal technology termed Centrifugal
Available online 23 January 2017
Microfluidic Chip (CMC) capable of sorting immune cells from blood and in situ cellular analysis in a lab-
oratory setting. Operation of the CMC entailed a blood specimen layered on a density gradient medium
Keywords:
and centrifuged in microfluidic channels where immune cell subpopulations could rapidly be sorted
Centrifugation
Microfluidics into distinct layers according to their density differentials. We systematically studied effects of differ-
Blood cell sorting ent blocking molecules for surface passivation of the CMC. We further demonstrated the applicability of
Immune cells CMCs for rapid separation of minimally processed human whole blood without affecting immune cell
viability. Multi-color imaging and analysis of immune cell distributions and enrichment such as recovery
and purity rates of peripheral blood mononuclear cells (PBMCs) were demonstrated using CMCs. Given
its design and operation simplicity, portability, blood cell sorting efficiency, and in situ cellular analysis
capability, the CMC holds promise for blood-based diagnosis and disease monitoring in POC applications.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction mononuclear cells (PBMCs). Importantly, the number of different

immune cells in blood circulation has both prognostic and diagnos-
Status of human immune system can reflect health, aid disease tic values for diseases such as cancer and HIV [4].
diagnosis, and facilitate drug treatment efficacy monitoring [1–3]. In the clinic, complete blood count (CBC) is a common blood
Each type of immune cells (or white blood cells, WBCs) mounts test that reports the count of immune cell subpopulations. Auto-
a specific defense mechanism against diverse invasions. The five matic CBC can be achieved by flow cytometric instruments, which,
main immune cells in circulating blood: neutrophils (CD15+), however, are expensive due to sophisticated electrical and optical
eosinophils (CD193+), basophils (CD123+), lymphocytes (CD3+ for methods used for characterizing cell types. Moreover, flow cytom-
T cells, CD19+ for B cells, and CD56+ for natural killer cells or NK etry cannot distinguish abnormal cells by morphology, nor can it
cells), and monocytes (CD14+) are related to bacterial infection, confirm the existence of pathogens. Manual CBC can be conducted
parasite infection, allergic reaction, adaptive immunity, and phago- on a blood smear. Although manual counting allows identification
cytosis, respectively. Neutrophils, eosinophils, and basophils are of abnormal cells and pathogens, it requires an experienced techni-
collectively called polymorphonuclear (PMN) cells, whereas lym- cian to distinguish subtle differences among chromatically stained
phocytes and monocytes are collectively termed peripheral blood WBC subpopulations, thus making it error prone. Both conventional
automatic and manual CBC methods process a mixture of target
immune cells over abundant red blood cell (RBC) background, mak-
ing such methods laborious and time-consuming [1].
∗ Corresponding author at: Department of Mechanical Engineering, University of Blood centrifugation over a density gradient medium (DGM)
Michigan, Ann Arbor, MI, 48109, USA. such as the Ficoll or Percoll solution is an efficient technique to
E-mail address: [email protected] (J. Fu).
1
These authors contributed equally to this work.
separate and sort blood cells into different layers according to their

http://dx.doi.org/10.1016/j.snb.2017.01.113
0925-4005/© 2017 Elsevier B.V. All rights reserved.
 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061 1051

holds promise for blood-based diagnosis and disease monitoring in

point-of-care (POC) applications.

2. Materials and methods

2.1. Chip fabrication

CMCs were fabricated using soft lithography [11,13–15]. CAD

designs were printed out as transparency masks (CAD/Art Services,
Inc., Bandon, OR). Channel patterns on Si wafers were fabricated
by patterning a layer of negative photoresist, SU8 (Microchem,
Westborough, MA), with a thickness of about 100 ␮m, using pho-
tolithography. The Si wafer was further coated conformably with a
0.5–1 ␮m thick Parylene C dielectric layer at 690 ◦ C and 30 mTorr
(Specialty Coating Systems, Indianapolis, IN).
Based on the same planar design, four types of CMCs, with
varying poly-dimethylsiloxane (PDMS; Dow Corning, Ellsworth,
Germantown, WI) channel-layer thickness and number of glass
slides bounding PDMS layers, were fabricated using the Si mold
by soft lithography (see Fig. 5 a for illustrations of different CMC
configurations) [11,13–15]. For no-glass, bottom-glass, and glass-
sandwich CMCs with a 2 mm thick PDMS channel-layer, PDMS
prepolymer with a 10:1 monomer to curing agent ratio was poured
over the Si mold and degassed. The PDMS channel-layer thickness
was controlled by squeezing PDMS prepolymer against microscope
slides. For fabrication of glass-sandwich CMCs with a 400 ␮m thin
channel-layer membrane, PDMS prepolymer was spin-coated on
the Si mold at a spin speed of 500 rpm. After baking at 80 ◦ C for
30 min, PDMS layers were peeled off from the Si mold, cut into
Fig. 1. Blood cell densities and sorting using density gradient centrifugation. (a)
cubes, and punched with 0.5 mm diameter holes before cleaning
Scheme showing density distributions of human blood cells [32]. (b) Movement of
blood cells upon centrifugation for human blood placed on top of the Ficoll solution. and treating with air plasma (Femto Science, Gyeonggi-Do, Korea).
White arrows indicate cell movement directions. For no-glass CMCs, 2 mm thick PDMS channel-layers were bonded
to another 2 mm thick blank PDMS. For bottom-glass and glass-
sandwich CMCs, PDMS channel-layers were first bonded to glass
inherent density differences (Fig. 1a) [1]. In blood centrifugation slides (Thermo Fisher Scientific, Waltham, MA) precoated with a
over a DGM, different blood cells in a bulk fluid migrate to dif- 50 ␮m thick PDMS layer. A thin glass coverslip (Thermo Fisher Sci-
ferent equilibrium positions according to density differentials in entific) was then bonded on top of the PDMS channel-layer to
a rotational motion (Fig. 1b). Through miniaturization, microflu- transform a bottom-glass to a glass-sandwich CMC. CMCs were
idic centrifugation systems that incorporate a DGM have been further baked at 80 ◦ C overnight to complete PDMS polymeriza-
successfully developed, which have been shown to be capable of tion and strengthen irreversible bonding between PDMS and glass
separating whole blood into multiple cell layers. However, existing surfaces.
microfluidic centrifugation systems involving a DGM still fall short
of integrating both blood cell sorting and in situ cellular imaging
and enumeration [5–8]. Specifically, designs of existing microflu-
idic centrifugation systems, with complex microfluidic networks 2.2. Blood collection and treatment
to handle blood and the DGM, are suboptimal for integration
with optical detection due to large device thickness, bulky setups, Blood collection was approved by the Institutional Review Board
intricate units, and complicated operations that prohibit optical (IRB) of the University of Michigan, Ann Arbor. Informed con-
accessibility [9,10]. Thus, for existing microfluidic centrifugation sent was obtained from healthy donors before blood collection.
systems using the DGM, cellular analysis is commonly achieved by Briefly, 50–100 ␮L capillary blood from donor fingers was sam-
an external platform to collect blood cells after centrifugation for pled by lancets (Owen Mumford Ltd., Marietta, GA) according to
downstream cellular analysis. the manufacturer manual, and collected into blood collection tubes
Herein, we report the development of a single-layer, simple- pre-coated with EDTA (BD Biosciences, San Jose, CA). Blood was
configuration Centrifugal Microfluidic Chip (CMC) that is com- diluted with phosphate buffered saline (PBS) solution at a 1:1
patible with in situ multi-color fluorescence microscopy [11,12]. ratio. Diluted blood was stained with Hoechst 33342 (0.1%, v/v; Life
The CMC enables efficient and convenient sample manipulation, Technologies, Grand Island, NY), calcein AM (1 ␮g mL−1 ), fluores-
blood separation operation, and quantitative cellular analysis of cein isothiocyanate (FITC)-labeled antibody specific to human CD14
minimally processed whole blood. We investigated and optimized (15 ␮g mL−1 ; Life Technologies, Carlsbad, CA), phycoerythrin (PE)-
various aspects of the CMC, including surface passivation, struc- labeled antibody specific to human CD15 (10 ␮g mL−1 ; BioLegend,
tural design, operational control, cell viability, separation stability, San Diego, CA), or Alexa Fluor (AF) 647-labeled antibody specific
and multi-color fluorescence imaging, in order to achieve efficient to human CD3 (10 ␮g mL−1 ; BioLegend), for 5 min at room tem-
blood cell sorting and on-chip cellular analysis and enumeration. perature before being assayed. Blood samples were vortexed to
Given its design and operation simplicity, portability, blood cell avoid sedimentation before being used within 4 h of blood collec-
sorting efficiency, and in situ cellular analysis capability, the CMC tion.
1052 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061

Fig. 2. Blood cell sorting using density gradient centrifugation in a microfluidic device. (a) Photos of the centrifuge microfluidic chip (CMC) filled with dyes (red for blood
and blue for the Ficoll solution) for visualization. An array of microposts inside the main branch served as barriers for stabilizing flow. Dashed lines highlight microposts. (b)
Operation procedure of the CMC for sorting blood cells. Initially, the CMC was empty (i). After passivation with a blocking solution (dark grey) (ii), the CMC was air dried (iii)
before loading with blood specimens (red) from the main branch down to the barrier array (iv) and the Ficoll solution (blue) from the side branch to fill the remaining channels
(v and vi). (c) the CMC loaded with blood and the Ficoll solution was inserted into a centrifuge tube before spinning in a benchtop centrifuge machine. Foam was placed
inside the centrifuge tube underneath the CMC to absorb uneven pressure under high spin-speed. (d) Cartoon showing blood cell distribution expected after separation in
the CMC. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

2.3. Surface treatment and centrifugation operation All centrifugation experiments were conducted at room tem-
perature.
The CMC channel surface was passivated with one of the fol-
lowing blocking solutions for 30 min at room temperature before it 2.4. Imaging and analysis
was completely air-dried: fetal bovine serum (FBS; Thermo Fisher
Scientific), trypsin (0.25% or 2.5 g L−1 in EDTA; Thermo Fisher Sci- Blood separation was recorded using an imaging setup com-
entific), bovine serum albumin with Tween-20 (BSA-T20, 10% BSA prising an inverted microscope (Benz Microscope Optics Center,
and 1% (v/v) T20 in PBS; Sigma-Aldrich, St. Louis, MO), Pluronic Ann Arbor, MI) and a digital color camera (BigCatchUSA, Torrance,
F127 (50% in PBS; Sigma–Aldrich), polyethylene glycol (PEG, molec- CA). Photo collages of entire channels were stitched from individ-
ular weight of 8000, 50% in PBS; Sigma-Aldrich), Teflon (DuPont, ual images or video frames. Bright field and fluorescence images
Wilmington, DE), and poly-l-lysine conjugated with PEG (PLL- showing fluorescently labeled blood cells were recorded using a
PEG, 0.1 mg mL−1 ; Susos AG, Dübendorf, Switzerland). After surface fluorescence microscope (AXIO Observer.Z1) attached with a CCD
passivation, the CMC was filled with blood and Ficoll solution camera (AxioCam MRm), before the images were processed and
(1.077 g mL−1 ; GE Healthcare Bio-Sciences, Pittsburgh, PA) before stitched into a mosaic by software AxioVision (all from Carl Zeiss
it was placed inside a 50 mL centrifuge tube that was partially filled Microscopy, LLC, Thornwood, NY).
with foam to stabilize the CMC under centrifugation. The centrifuge We used fluorescence signals for identifying blood cells. Specifi-
tube was spun in a benchtop centrifuge (Eppendorf, Hauppauge, cally, cells without any staining indicated RBCs, whereas cells with
NY) with brakes turned off to minimize liquid agitation in the CMC. Hoechst staining indicated WBCs. Since populations of eosinophils
To examine cell adhesion in the CMC, motions of RBCs and WBCs (CD193+) and basophils (CD123+) are relatively small compared
under flow were induced by slightly tapping tubing connection with neutrophil population, CD15+ and Hoechst-stained blood cells
to the CMC while observing cells in the CMC under fluorescence were referred to directly as PMNs. PBMC population was estimated
microscopy. by differential counts between WBCs and PMNs. Platelets were
As a reference cell sorting technique, 100 ␮L blood layered on ruled out during cell enumeration based on their small size relative
top of 100 ␮L Ficoll solution in a 200 ␮L vial was spun at 400g for to typical WBCs.
30 min as advised by the manufacturer instruction. To examine Blood cell counts were analyzed along the CMC’s vertical main
viability of unprocessed blood cells, 2 ␮L intact whole blood was branch. The channel distance, d, was defined as the length mea-
covered by a glass slide before being scanned and imaged. sured from the bottom of the main branch. In normal situations,
 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061 1053

Fig. 3. Improper blood sorting in the CMC. (a) Phenomena leading to blood cell separation failure in the CMC. Please refer to Fig. S4a for occurrence locations inside the main
branch. (i) Blood cells became adherent. (ii) Blood cell agglutination. (iii) Air bubble generation, anchoring, and congestion. (iv) Multiphase formation between blood and
Teflon residue inside a Teflon-treated main branch. (v) Blood and bubble clogging due to the arching effect at the barrier array. (vi) Loosely packed RBC patches scattered over
the main branch. (b) Spatial distributions of RBCs and WBCs along the main branch after centrifugation. The CMC was passivated with either (i) BSA or (ii) F127 as indicated.
(c) Time-lapse images showing motions of RBCs (black, under bright field) and Hoechst-stained WBCs (blue, under fluorescence) under flow. No-glass CMCs were centrifuged
at an acceleration of 1000g for 10 min. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

RBCs would sediment to the bottom of the main branch after cen- separation measured the distance between the two regions defined
trifugation of the CMC. Formation of RBC sediments prohibited by PMN spread distance and PBMC spread distance.
WBC enumeration within the pellet; therefore, cell counts started Statistical analysis was conducted with n ≥ 3 independent
at different distances, depending on pellet size. We arbitrarily set experiments. Error bars represent one standard deviation (s.d.).
an “infinite” value for RBC counts to show a steep increment in
RBC numbers, indicating RBC pellet when accumulation started to
appear. Recovery rate of PBMCs at any given position along the
CMC main branch was defined as the PBMC count above the posi- 3. Results and discussion
tion divided by all PBMCs counted in the main branch. Purity rate of
PBMCs was defined as the PBMC count above a position along the 3.1. Design and operation of the CMC
CMC main branch divided by the sum of PBMCs and PMNs above the
position. Three additional terms were further quantified to examine Leveraging the intrinsic optical transparency of PDMS and the
immune cell sorting in the CMC (Fig. S1). PMN spread distance mea- capability to create microscale features in PDMS using soft lithog-
sured the distance for 70% of all PMNs from the bottom of the main raphy, herein we developed a simple biomedical device termed
channel. PBMC spread distance measured the distance for 70% of Centrifugal Microfluidic Chip (CMC) to integrate blood cell separa-
all PBMCs from the top position of PBMC distributions. PMN-PBMC tion and in situ cellular analysis and enumeration. Using PDMS and
glass as building materials, the CMC was optically transparent with
1054 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061

Fig. 4. Blood cell sorting in the CMCs passivated with PEG, PLL-PEG, and trypsin. (a) Photo collages showing blood separation along the main branch of the no-glass CMC
treated with either (i) PEG, (ii) PLL-PEG, or (iii) trypsin as indicated, after 10 min centrifugation. (b) Magnified views showing blood cell distribution along the main branch
of the no-glass CMC treated with either (i) PEG, (ii) PLL-PEG, or (iii) trypsin, after centrifugation for 10, 20, or 30 min as indicated. Please refer to Fig. S4b for locations where
the images were taken. (c & d) RBC (c) and WBC (d) counts above RBC pellets as functions of spin time and blocker. (e) Spatial distributions of RBCs and WBCs along the main
branch. Results were obtained from (a and b). (f) Time-lapse images showing motions of RBCs (black, under bright field) and Hoechst-stained WBCs (blue, under fluorescence)
under flow. A centrifuge acceleration of 1000g was used. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)
 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061 1055

Fig. 5. Blood cell sorting performance by the CMC with different structural supports. (a) Cross-sectional schemes of the CMC with no-glass, bottom-glass, and glass-sandwich
configurations at rest or deformed states under spinning. (b) Merged bright-field and fluorescence images showing Hoechst (blue, for staining WBCs) and CD15 (orange, for
staining PMNs) staining. Images were taken below, at, and above the barrier array along the CMC main branch as indicated. (c–e) PMN (c) and PBMC (d) spreads, and PMN-
PBMC separation (e) distances as a function of structural support. (f) Spatial distributions of PMNs and PBMCs along the main branch of the CMC with no-glass, bottom-glass,
or glass-sandwich configuration as indicated. (g & h) Recovery (g) and purity rates (h) of PBMCs as a function of channel distance d. (i) Correlation of PBMC purity versus
recovery rates. Two reference lines (grey) were drawn to represent recovery and purity rates at 90%. CMCs were centrifuged at an acceleration of 1000g for 10 min. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
1056 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061

low autofluorescence (Fig. S2), beneficial for fluorescence-based solution from hydrophobic PDMS channels. Teflon residue could
cellular analysis and enumeration. lead to the formation of multiphase bubbles and droplets (Fig. 3a
The CMC was a single-layer microfluidic chip containing a sim- iv). Due to an arching effect, the micropost barrier array designed
ple channel network. Specifically, to facilitate fluid exchange and for stabilizing flow became a physical obstacle to resist movement
sample recovery, the CMC contained a 2.5 cm long, vertical main of cells and air bubbles (Fig. 3a v). Loosely packed RBC patches were
branch, a middle side branch connected to the center of the main also observed along the main branch, which became hurdles to RBC
branch, and a bottom side branch connected to the main branch sedimentation (Fig. 3a vi).
bottom (Fig. 2a; please see Fig. S3 for detailed description of the Analyses of spatial distributions of blood cells along the main
CMC dimensions). Design of the CMC had several unique features branch for CMCs treated with BSA and Pluronic F127 showed that
in relation to its functionalities. The CMC main branch was a long both RBCs and WBCs remained widely spread out along the entire
straight channel built with location markers for easy alignment CMC main branch (Fig. 3b). We further examined which types
during imaging and analysis. The CMC main branch was 0.5 mm of blood cells were adherent to PDMS surfaces in the CMC, with
wide, and thus it was within the field of view (FOV) for a single linear results showing that RBCs, but not WBCs, adhered to PDMS surfaces
scanning using a 10× objective in typical microscopes. In conjunc- blocked with BSA-T20, and Pluronic F127 (Fig. 3c).
tion to a thin CMC channel, using 10× objectives also provided a Three blockers were identified in this work capable of passi-
large depth of focus (DOF) to image cells in the main branch, with- vating PDMS surfaces in the CMC. When the CMC was passivated
out missing cells due to out-of-focus or a need for resorting multiple with PEG, PLL-PEG, or trypsin, blood separation after centrifuga-
scanning at different focal planes. Besides the main branch, the mid- tion at 1000g for 10 min resembled to what is typically seen in
dle and bottom side branches allowed for extracting PBMCs and conventional large centrifugation tubes (Fig. 4a). Both PEG and
PMNs for subsequent examination after cell sorting. These two side PLL-PEG have been commonly used as blockers, given the multiva-
branches were narrower (0.1 mm wide) than the main branch to lent hydrophobic interactions from its repeating methylene units
minimize fluid mixing with the main branch during CMC operation. in PEG [22,23], and the polycationic PEG grafted copolymer with
In addition, to minimize contamination at the blood-DGM interface a PLL backbone which strongly adsorbs onto negatively charged
during liquid exchange and centrifugation [16], two micropost bar- surfaces in aqueous solution [24]. Detailed explanation as how
rier arrays were placed inside the main branch on top and below exactly trypsin, which is commonly used for detaching adherent
the middle side branch to stabilize flow (Figs. 2a and S3). Inlet and cells, interact with PDMS to prevent blood cell adhesion will require
outlet holes of the CMC allowed operators to conveniently load an in-depth investigation and is beyond the scope of this work.
solutions using pipettors or syringes. When centrifugation was increased from 10 min to 20 min and
The entire surface of the CMC was first passivated with a block- 30 min at 1000g for the CMCs treated with PEG, PLL-PEG, and
ing solution to prevent non-specific adhesion of blood cells (Fig. 2b trypsin, more RBCs sank to the main branch bottom (Fig. 4b). The
i&ii). After the CMC was air-dried (Fig. 2b iii), about 0.25 ␮L blood counts of contaminant RBCs above RBC pellets were plotted in
(diluted in PBS with a 1:1 v/v ratio) was injected into the CMC main Fig. 4c, which showed reduced RBC contaminant counts as the CMC
branch through the main branch inlet, and stopped at the barrier were spun longer. Trypsin-coated CMCs not only had the least RBC
array (Fig. 2b iv). About 1 ␮L Ficoll solution was then injected into contaminant counts even after 10 min spin, but also had RBC con-
the CMC main branch through the middle side branch inlet and taminant counts sharply dropped to only a few for a total of 20 min
exited via the bottom side branch outlet (Fig. 2b v&vi). Thus, the spin. Although the countable WBCs above RBC pellets dropped as
upper portion of the main branch above the middle side branch spin time increased (Fig. 4d), PMNs were separated further from
(about 1 cm long) contained blood samples to be processed, while PBMCs and immersed in RBC pellets, thus leading to improved sep-
the bottom portion of the main branch below the middle side aration of both WBC subpopulations. The final amounts of WBCs
branch (about 1.5 cm long) contained the Ficoll solution (Fig. S3). after 30 min spin of CMCs treated with PEG, PLL-PEG, and trypsin
The CMC was then packaged into a 50 mL centrifuge tube embed- were almost the same, suggesting an equilibrium amount of PBMCs
ded with a foam stabilizer before centrifugation using a benchtop has reached. Quantitative analysis of spatial distributions of RBCs
centrifuge (Fig. 2c). After centrifugation, immune cells in the blood and WBCs clearly revealed their dynamics in the CMC main branch
redistributed along the CMC main branch according to their relative (Fig. 4e). Most RBCs sank and formed dense pellets at the main
densities (Fig. 2d). Owing to its simple design and error-proof oper- branch bottom after the first 10 min of centrifugation with PLL-PEG
ation, the CMC could be operated without using a microscope. Blood and trypsin passivation, whereas RBCs still scattered over the entire
cell sorting in the CMC main branch was analyzed immediately after main branch with PEG passivation. Moreover, two WBC regions,
centrifugation to minimize blood evaporation. one residing on top of RBC pellets and another residing near the
micropost area, emerged after 30 min spin, resembling the typical
3.2. Surface passivation layer distribution in conventional centrifugal tubes. Therefore, both
PLL-PEG and trypsin had a comparable surface passivation effect
As PDMS has a dangling molecular structure and dynamic sur- and were better blockers than PEG for the CMC. We further vali-
face properties, a list of potential blockers [17–21] was examined dated that neither RBCs nor WBCs adhered to PDMS surfaces of the
for surface passivation of PDMS surfaces in the CMC to prevent CMC treated with PEG, PLL-PEG, or trypsin (Fig. 4f).
blood cell adhesion (Supplementary Table 1). We used no-glass For the remaining study, we chose to use trypsin as the blocker
CMCs for surface passivation tests because of the ease to fabricate for surface passivation of the CMC, as it is a common biological solu-
them. The majority of blockers examined in this work, including tion, and notably, significantly smaller populations of RBCs were
FBS, BSA-T20, Pluronic F127, and Teflon, did not lead to success- left above the RBC pellet region even after a 10 min centrifuga-
ful blood cell sorting in the CMC after centrifugation at 1000g for tion, clearly demonstrating well sorted and enriched immune cell
10 min. Specifically which blockers were involved in different oper- subpopulations. Moreover, PLL-PEG is a costly synthetic chemical,
ation failures were listed in Supplementary Table 1. Besides blood limiting its accessibility to most biological laboratories.
cells adhered to the CMC main branch surface (Fig. 3a i), cell aggluti-
nation as large clumps occurred at different main branch locations 3.3. Blood centrifugation and cell distribution
(Fig. 3a ii). As air bubbles could inevitably form during sample load-
ing, these bubbles had led to cavitation in the main branch (Fig. 3a Simple fluid mechanics calculations revealed that the maximum
iii). It was difficult to completely remove highly hydrophobic Teflon rotational acceleration used in this study, 1500g, could exert a pres-
 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061 1057

Fig. 6. Blood cell sorting performance of the CMC with glass-sandwich configuration under different centrifugal accelerations. (a–c) PMN (a) and PBMC (b) spreads, and
PMN-PBMC separation (c) distances as a function of centrifugal acceleration. (d) Spatial distributions of PBMCs and PMNs along the CMC main branch after centrifugation at
500g, 1000g, or 1500g for 10 min as indicated. (e & f) Recovery rate (e) and purity rate (f) of PBMCs as a function of channel distance d. (g) Correlation of PBMC purity rate
versus recovery rate. Two reference lines (grey) were drawn to represent recovery and purity rates at 90%.

sure as high as 55 psi to the CMC. Contrary to centrifugal apparatus surfaces of the CMC (Fig. 5a). After centrifugation at 1000g for
made of hard plastics and glass, the structural material of the CMC, 10 min, PBMCs became separated from PMNs in all configurations
PDMS, was soft enough to deform and distort significantly at such (Fig. 5b). We further analyzed the quality of cell separation by PMN
a high spin speed [25,26]. Although blood components could sep- and PBMC spreads and PMN-PBMC separation distance (see Meth-
arate and distribute quickly during centrifugation, resilient PDMS ods for their definitions; Fig. S1). While there was no apparent
channels experienced significant volume and shape change when difference in PMN spread distance among different configurations
the CMC returned to a stationary state, which would substan- (Fig. 5c), providing more structural supports for CMC showed a
tially and adversely distort blood cell distribution, leading to severe clear trend of reducing PBMC spread distance (Fig. 5d). With consis-
cross-contamination among separated blood cell layers. To inves- tently the farthest separation between PMN and PBMC populations
tigate how the structural and mechanical factors of the CMC would (Fig. 5e & f), the glass-sandwich configuration led to the best PBMC
play a role in affecting its blood cell sorting performance, we sys- purity from CMC operations. Together, these observations sup-
tematically modulated the extent of PDMS channel deformation ported that bounding the CMC with rigid glass slides had reduced
during centrifugation by bounding glass slides tightly against the PDMS channel deformation during high-speed centrifugation and
2 mm thick PDMS channel layer. Three CMC configurations, termed consequently reduced cell redistribution and cross-contamination
no-glass, bottom-glass, and glass-sandwich, were assembled, with [25,26].
no-glass CMC without any bounded glass slide, bottom-glass CMC Since RBCs can easily be removed from blood samples by RBC
with a glass slide bounded to the CMC bottom surface, and glass- lysis buffer, herein we focused on analyzing spatial distributions
sandwich CMC with two glass slides bounded to the top and bottom of WBCs in the CMC after centrifugation at 1000g for 10 min. Tak-
1058 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061

Fig. 7. Validation of cell health, CMC operations and multi-color fluorescence imaging. (a) Bright field and fluorescence images of dead and live cells showing Hoechst (blue,
for staining WBCs) and calcein AM (green, for staining live cells). (b) Comparison of live cell populations between the CMC and vial-based operations. A red region representing
the range of live cell populations from intact whole blood was included for comparison. Vial centrifugation was performed at an acceleration of 400g for 30 min. (c) Changes
of contents in CMCs at different time points. Hoechst staining (pseudo color: green) was imaged. (d) Redistributions of Hoechst-stained cell in (c). (e) Identification of WBC
subpopulations through Hoechst (blue), and antibodies against CD14 (green; a monocyte marker), CD15 (orange; a PMN marker), and CD3 (red; a T-cell marker). CMCs were
centrifuged at an acceleration of 1000g for 10 min. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

ing PBMCs as target cells and PMNs as contaminants, recovery and after centrifugation at 1000g for 10 min. Based on the recovery-
purity rates of PBMCs were analyzed as a function of channel dis- purity correlation curve, generally speaking, no-glass CMCs were
tance d in no-glass, bottom-glass, and glass-sandwich CMCs (see the worst and glass-sandwich CMCs were the best for attaining
Methods for definitions of recovery and purity rates; Fig. 5g & h), concurrent high recovery and purity values for PBMCs.
with results showing that in all three device types, PBMC recovery We next examined how different centrifugation acceleration
and purity rates decreased and increased with channel distance rates (500g, 1000g, and 1500g) could affect blood cell separation
d, respectively. The glass-sandwich CMC showed the best PBMC in the CMC. To this end, we chose to use glass-sandwich CMCs
recovery performance, as it allowed collection of more PBMCs at and further reduce the PDMS channel-layer thickness to about
a shorter distance from the main channel top. On the other hand, 400 ␮m, such that the effect of glass slides as structural supports
both bottom-glass and glass-sandwich CMCs showed better sorting for the CMC became more prominent. Compared with 500 and
performance for PBMCs than no-glass CMCs, owing to purer PBMC 1500g spin speeds, 1000g acceleration consistently showed the
populations at a shorter distance from the main channel top. To least PMN (Fig. 6a) and PBMC (Fig. 6b) spread distances and the
further evaluate their trade-off, PBMC purity rate was correlated farthest separation between the two cell populations (Fig. 6c),
with its recovery rate (Fig. 5i). A data point closer to the top-right allowing better PBMC sorting with concurrent high recovery and
corner in such a correlation plot would mean a desirable condi- purity values. On the other hand, in all experiments except those
tion to attain concurrent high recovery and purity values. Thus, with 500g acceleration, peaks of PBMCs and PMNs were well sep-
Fig. 5i revealed that centrifugation of the glass-sandwich CMC could arated from each other along the main branch of the CMC after
achieve both recovery and purity rates of above 90% for PBMCs 10 min of centrifugation (Fig. 6d). The cell distribution plot also
 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061 1059

showed that centrifugation at 1000g yielded the highest number on cell membrane, these results indicated that cells undergone
of PBMCs and minimum overlap between PMN and PBMC popula- CMC operations maintained good cell surface integrity. As these
tions compared with the other two accelerations (Fig. 6d). While three major important immune cell subpopulations are related to
varying the acceleration condition had little effect on PBMC recov- diverse diseases, and a large pool of antibody markers to identify
ery rate (Fig. 6e), it did show significantly improved PBMC purity hematopoietic stem cells (HSCs) or CD4 and CD8T cell subpopu-
rate from 500g to 1000g but compromised purity rates at 1500 g lations are available, CMCs can be further developed into a blood
(Fig. 6f). The recovery-purity correlation clearly showed a wide diagnostic device or a CBC platform for blood examination or cyto-
range of recovery rates or high purity rates closed to 100% at 1000g informatics, respectively, in healthcare or critical care applications.
(Fig. 6g). In particular, a purity of 97% and recovery of 95% could be It would be ideal to achieve baseline separation between PMNs
achieved at 1000g for PBMCs. Interestingly, centrifugation of the and PBMCs by centrifugation of the CMC. However, as the densities
CMC at 1500g did not further improve separation between PBMCs of PMNs and PBMCs overlap with each other (Fig. 1a), complete
and PMNs when compared with centrifugation at 1000g; this obser- separation between PBMCs and PMNs is, in principle, not possible
vation was further confirmed when the recovery and purity rates by centrifugation. In addition, operations to load solutions into the
of PBMCs were analyzed under different centrifugation accelera- CMC might cause contaminations of the DGM and blood samples at
tion rates (Fig. 6e–g). We suspect that centrifugation of the CMC their interface, further compromising separation efficiency of the
at 1500g might be too forceful to cause unexpected cell dynamic CMC for isolating PBMCs from other blood cells including PMNs.
behaviors or too much material deformation in the CMC as investi- Lastly, the cell distribution was the same with or without the bar-
gated in Fig. 5, leading to compromised cell sorting. All together, an rier array. Although the microposts are optional components, and
acceleration of 1000g for CMC operation was the optimal condition removing them can avoid channel clogging, we found them useful
for isolating PBMCs from blood. when monitoring solution exchange without using a microscope.

3.4. Blood cell health, CMC post-spin stability and multi-color

imaging
4. Conclusion
From a design perspective, although the thin CMC channel
allowed for single scanning under the limited DOF (see Design and In this work, we developed a centrifugal microfluidic device
operation of the CMC for detailed discussion), it created a larger termed the CMC as a promising, portable microfluidic platform
shear stress when cells migrated in the channel. To validate any for rapid and efficient blood cell sorting and cellular imaging and
adverse effects that might incur to cells during cell sorting using analysis. The CMC, based on a 1-layer microfluidic design, could
the CMC, we compared cell viability using cells undergone a vial- achieve immune cell sorting and enumeration by one single cen-
scale Ficoll centrifugation and cells from intact whole blood. Using trifugation step using common technical periphery available in
Hoechst and a viability marker calcein AM, WBCs could be identi- a biomedical laboratory. We further identified different surface
fied live or dead according to the calcein AM fluorescence (Fig. 7a). passivation methods suitable for the CMC to prevent non-specific
Regardless CMC or vial operations, cell viability was almost 100% binding of blood cells to PDMS surfaces; this is important as
within 40 min after spinning (Fig. 7b). Since Ficoll solution is not blood separation in confining microfluidic environments is very
a cell medium, unhealthy cells emerged after 1 h of culture in the sensitive to cell-surface interactions due to a high surface-to-
Ficoll solution. The number of unhealthy cells using CMC opera- volume ratio in microfluidics. We also devised and validated a
tions was less than that of the vial method, possibly attributable glass-sandwich channel-layer assembly to strengthen the CMC
to shorter Ficoll exposure and spinning times in CMC operations mechanical property to minimize PDMS channel deformation dur-
(10 min for CMC vs. 30 min in the vial method). In all cases, cell ing high-speed centrifugation. A single step centrifugation of the
viability was above 97% within 1.5 h after centrifugation processes CMC for 10 min could achieve both recovery and purity rates >95%
or within 4 h after blood collection without centrifugation. Such for PBMCs, a WBC viability rate >98%, and stable separation within
high cell viability rates confirmed a minimum cell damage by CMC the channel from minimally processed whole blood. Importantly,
operations. we further demonstrated 4-color immunostaining and fluores-
Besides separating and extracting cells, continuous scanning of cence microscopy for in situ cellular analysis in the CMC, further
in situ assays is also an important consideration for CMCs. To ver- supporting the CMC as a versatile platform for blood-based diag-
ify the stability for short-term handling and storage, CMCs were nosis and disease monitoring.
repeatedly operated, scanned, and stored. Using Hoechst-stained While there are numerous sophisticated microfluidic cell sort-
blood, CMC main channels were imaged 4 times (Fig. 7c). Formation ing and imaging platforms reported in the literature [1,9,10,27–29],
of RBC pellets was stable and did not scatter towards the chan- however, very few of them can be used directly for whole blood
nel top even though the chips were laid horizontally and moved samples due to intricate design, bulky setup, and complicated
around places. However, an air-liquid interface formed and steadily control and operations. The CMC can process a few microliters
grew towards the channel bottom due to evaporation. With the per- of blood, reducing blood consumption to a thousandth of con-
fectly collapsed Hoechst-stained cell distribution curves, neither a ventional methods, which is especially advantageous for pediatric
global shift nor a redistribution of the cell location was observed, patients. In this work, we have further devised an optimal device
confirming such a moving interface had negligible effect on cell design and operation parameters to shorten the centrifugation time
sorting within 1.5 h after centrifugation (Fig. 7d). Therefore, short- to 10 min, shorter compared to 30 min as advised by the manufac-
term in situ examination of evolving cell behaviors or dynamics is turer instruction. The compact and low-cost CMC that can achieve
feasible. efficient and rapid separation of immune cells from complex back-
Multi-color imaging is another critical aspect for the delivery of ground of minimally processed whole blood for on-chip cytometric
a versatile CMC platform capable of comprehensive cellular charac- examination will hold promise for blood-based diagnosis and dis-
terization [11,12]. Besides CD15 antibody, antibodies for CD14 and ease monitoring in POC applications [30]. We envision that our
CD3 were added to blood before centrifugation. Cells stained specif- newly invented Vacuum–Pressure Accelerated Movement (V-PAM)
ically against these antibody markers, with CD15+ PMNs below technology [31] can be further incorporated into CMCs in the future
microposts and CD14+ monocytes or CD3+ T cells above microposts, for improved operations such as automatic fluid filling and trans-
were detected (Fig. 7e). Since cells retained fluorescence signals ferring to an extraction or concentration chamber.
1060 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061

Acknowledgements [23] B. Liu, P.J. Huang, X. Zhang, F. Wang, R. Pautler, A.C. Ip, et al., Parts-per-million
of polyethylene glycol as a non-interfering blocking agent for homogeneous
biosensor development, Anal. Chem. 85 (2013) 10045–10050.
We acknowledge financial support from the National Science [24] I. Wong, C.M. Ho, Surface molecular property modifications for
Foundation (CBET 1263889 and CMMI 1536087), the National Insti- poly(dimethylsiloxane) (PDMS) based microfluidic devices, Microfluid
tutes of Health (R01 HL119542), and the Michigan Center for Nanofluidics 7 (2009) 291–306.
[25] Q. Chen, G. Li, Y. Nie, S.H. Yao, J.L. Zhao, Investigation and improvement of
Integrative Research in Critical Care (M-CIRCC). The Lurie Nanofab- reversible microfluidic devices based on glass-PDMS-glass sandwich
rication Facility at the University of Michigan, a member of the configuration, Microfluid. Nanofluid. 16 (2014) 83–90.
National Nanotechnology Infrastructure Network (NNIN) funded [26] D.W. Inglis, A method for reducing pressure-induced deformation in silicone
microfluidics, Biomicrofluidics 4 (2010).
by the National Science Foundation, is acknowledged for support
[27] N.T. Huang, W. Chen, B.R. Oh, T.T. Cornell, T.P. Shanley, J. Fu, et al., An
in microfabrication. integrated microfluidic platform for in situ cellular cytokine secretion
immunophenotyping, Lab Chip 12 (2012) 4093–4101.
[28] X. Li, W. Chen, G. Liu, W. Lu, J. Fu, Continuous-flow microfluidic blood cell
Appendix A. Supplementary data sorting for unprocessed whole blood using surface-micromachined
microfiltration membranes, Lab Chip 14 (2014) 2565–2575.
[29] S. Wang, K. Liu, J. Liu, Z.T. Yu, X. Xu, L. Zhao, et al., Highly efficient capture of
Supplementary data associated with this article can be found, in
circulating tumor cells by using nanostructured silicon substrates with
the online version, at http://dx.doi.org/10.1016/j.snb.2017.01.113. integrated chaotic micromixers, Angew. Chem. Int. Ed. Engl. 50 (2011)
3084–3088.
[30] M.N. Gulari, A. Tripathi, M. Ghannad-Rezaie, N. Chronis, An optofluidic lens
References array microchip for high resolution stereo microscopy, Micromachines (Basel)
5 (2014) 607–621.
[1] Z.T. Yu, K.M. Aw Yong, J. Fu, Microfluidic blood cell sorting: now and beyond, [31] Z.T. Yu, M.K. Cheung, S.X. Liu, J. Fu, Accelerated biofluid filling in complex
Small 10 (2014) 1687–1703. microfluidic networks by vacuum-pressure accelerated movement (V-PAM),
[2] L. Sompayrac, How the Immune System Works, 4th ed., Wiley-Blackwell, Small 12 (2016) 4521–4530.
2012. [32] Y. Takamura, K. Idegami, M. Kogi, H. Takabayashi, in: USPTO (Ed.), Method for
[3] K.M. Aw Yong, Z.T.F. Yu, K.H. Guan, J. Fu, Types of clinical samples and cellular Collecting Nucleated Red Blood Cells via Density-gradient Centrifugation
enrichment strategies, in: J.C. Love (Ed.), Micro- and Nanosystems for Utilizing Changes in Blood Cell Density, Japan Advanced Institute Of Science
Biotechnology, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany, And Technology, Kanazawa Medical University, & Kanazawa Institute Of
2016, pp. 1–28. Technology, USA, 2013.
[4] A.A. Okoye, L.J. Picker, CD4(+) T-cell depletion in HIV infection: mechanisms
of immunological failure, Immunol. Rev. 254 (2013) 54–64.
[5] T.H. Kim, H. Hwang, R. Gorkin, M. Madou, Y.K. Cho, Geometry effects on blood
Biographies
separation rate on a rotating disc, Sens. Actuators B: Chem. 178 (2013)
648–655.
[6] D.J. Kinahan, S.M. Kearney, M.T. Glynn, J. Ducree, Spira mirabilis enhanced
Zeta Tak For Yu is a post-doctoral fellow of Mechanical
whole blood processing in a lab-on-a-disk, Sens. Actuators A: Phys. 215
Engineering at the University of Michigan, Ann Arbor since
(2014) 71–76.
2012. Dr. Yu received his Ph. D. degree from the University
[7] J.M. Park, M.S. Kim, H.S. Moon, C.E. Yoo, D. Park, Y.J. Kim, et al., Fully
of California, Los Angeles (UCLA) and master and bache-
automated circulating tumor cell isolation platform with large-volume
lor degrees from the Hong Kong University of Science and
capacity based on lab-on-a-disc, Anal. Chem. 86 (2014) 3735–3742.
Technology (HKUST) in 2009, 2003 and 2001 respectively.
[8] H. Shiono, H.M. Chen, T. Okada, Y. Ito, Colony-forming cell assay for human
Dr. Yu’s research focuses on applying microfluidics or lab-
hematopoietic progenitor cells harvested by a novel continuous-flow cell
on-a-chip technology to perform cellular phenotyping and
separation method, J. Chromatogr. A 1151 (2007) 153–157.
proteomics in immunology, cancer biology and stem cell
[9] O. Strohmeier, M. Keller, F. Schwemmer, S. Zehnle, D. Mark, F. von Stetten,
biology.
et al., Centrifugal microfluidic platforms: advanced unit operations and
applications, Chem. Soc. Rev. 44 (2015) 6187–6229.
[10] D.J. Kinahan, S.M. Kearney, N.A. Kilcawley, P.L. Early, M.T. Glynn, J. Ducree,
Density-gradient mediated band extraction of leukocytes from whole blood
using centrifugo-pneumatic siphon valving on centrifugal microfluidic discs,
PLoS One 11 (2016).
[11] K.I. Kamei, M. Ohashi, E. Gschweng, Q. Ho, J. Suh, J.H. Tang, et al., Microfluidic
image cytometry for quantitative single-cell profiling of human pluripotent Jophin George Joseph completed his bachelor of Tech-
stem cells in chemically defined conditions, Lab Chip 10 (2010) 1113–1119. nology in Chemical Engineering (Honors) from Indian
[12] J. Sun, M.D. Masterman-Smith, N.A. Graham, J. Jiao, J. Mottahedeh, D.R. Laks, Institute of Technology Hyderabad in 2016. He was an
et al., A microfluidic platform for systems pathology: multiparameter SN Bose Scholar in 2014 and awarded the ‘Institute
single-cell signaling measurements of clinical brain tumor specimens, Cancer Gold medal for excellence in Academics and Co-curicullar
Res. 70 (2010) 6128–6138. Activities’ from IIT Hyderabad of class 2016. He will join
[13] Z.T. Yu, H. Guan, M.K. Cheung, W.M. McHugh, T.T. Cornell, T.P. Shanley, et al., as a pre PhD candidate in Department of Mechanical Engi-
Rapid, automated, parallel quantitative immunoassays using highly neering University of Michigan in fall 2016. His research
integrated microfluidics and AlphaLISA, Sci. Rep. 5 (2015) 11339. is mainly focused on microfluidics and molecular systems
[14] Z.T. Yu, K. Kamei, H. Takahashi, C.J. Shu, X. Wang, G.W. He, et al., Integrated biology
microfluidic devices for combinatorial cell-based assays, Biomed.
Microdevices 11 (2009) 547–555.
[15] K. Kamei, S. Guo, Z.T. Yu, H. Takahashi, E. Gschweng, C. Suh, et al., An
integrated microfluidic culture device for quantitative analysis of human
embryonic stem cells, Lab Chip 9 (2009) 555–563.
[16] T.J. Li, L.M. Zhang, K.M. Leung, J. Yang, Out-of-plane microvalves for whole
blood separation on lab-on-a-CD, J. Micromech. Microeng. 20 (2010). Shirley Xiaosu Liu is currently a fourth year Chemical
[17] H. Vaisocherova, E. Brynda, J. Homola, Functionalizable low-fouling coatings Engineering undergraduate student working in Integrated
for label-free biosensing in complex biological media: advances and Biosystems and Biomechanics Laboratory (IBBL) at the
applications, Anal. Bioanal. Chem. 407 (2015) 3927–3953. University of Michigan, Ann Arbor.
[18] D. Kim, A.E. Herr, Protein immobilization techniques for microfluidic assays,
Biomicrofluidics 7 (2013).
[19] J.W. Zhou, A.V. Ellis, N.H. Voelcker, Recent developments in PDMS surface
modification for microfluidic devices, Electrophoresis 31 (2010) 2–16.
[20] J.W. Zhou, D.A. Khodakov, A.V. Ellis, N.H. Voelcker, Surface modification for
PDMS-based microfluidic devices, Electrophoresis 33 (2012) 89–104.
[21] Q.L. Yu, Q.H. Wang, B.M. Li, Q.Y. Lin, Y.X. Duan, Technological development of
antibody immobilization for optical immunoassays: progress and prospects,
Crit. Rev. Anal. Chem. 45 (2015) 62–75.
[22] G.C. Shao, J. Wang, Z.H. Li, L. Saraf, W.J. Wang, Y.H. Lin, Poly(dimethylsiloxane)
microchip-based immunoassay with multiple reaction zones: toward on-chip
multiplex detection platform, Sens. Actuators B: Chem. 159 (2011) 44–50.
 Z.T.F. Yu et al. / Sensors and Actuators B 245 (2017) 1050–1061 1061

Meiki Cheung is a graduate of the University of Michigan Jianping Fu is an Associate Professor at the University
with a Bachelors in Chemical Engineering. During under- of Michigan, Ann Arbor, with a primary appoint-
graduate career, she was involved in UROP, WISE, and is ment in the Mechanical Engineering Department and
a sister of Kappa Phi Lambda Sorority Incorporated. She courtesy appointments in the Biomedical Engineering
enjoys finding new running routes, making new discover- Department and the Cell and Developmental Biology
ies, and exploring new cities. Department. He also serves as the Associate Director for
the Michigan Center for Integrative Research in Criti-
cal Care (MCIRCC) and is a Core Faculty Member for
the UM Center for Organogenesis, the UM Comprehen-
sive Cancer Center, and the UM Center for Systems
Biology. Dr. Fu’s current research focuses on mechanobi-
ology, stem cell biology, Bio-Microelectromechanical
and −Nanoelectromechanical Systems (BioMEMS/NEMS),
Lab-on-Chip (LOC), and applying microfabrication technology to illuminate biolog-
Parker Haffey is currently a third year Chemical Engi- ical systems at both the molecular and cellular levels. Dr. Fu is the recipient of
neering undergraduate student with interests in inorganic the American Heart Association Scientist Development Award (2012), the National
chemistry, fluid mechanics, and engineering. He plans on Science Foundation CAREER Award (2012), the Mechanical Engineering Outstand-
pursuing a career in process engineering after he gradu- ing Faculty Achievement Award (2014), the Robert M. Caddell Memorial Award
ates. for Research (2014), and the Ted Kennedy Family Team Excellence Award (2015).
Dr. Fu’s research group is currently supported by the National Science Foundation,
the National Institutes of Health, the American Heart Association, and some other
foundations and agencies.