Vo

lum

e 1

1

| N

um

be

r 17

|

20

11

Lab

on a C

hip

Pa

ge

s 2

79

7–

30

16 1473-0197(2011)11:17;1-#

ISSN 1473-0197

Lab on a ChipMiniaturisation for chemistry, physics, biology and bioengineering

www.rsc.org/loc Volume 11 | Number 17 | 7 September 2011 | Pages 2797–3016

PAPERDino Di Carlo et al.Automated cellular sample preparation using a Centrifuge-on-a-Chip

www.rsc.org/locRegistered Charity Number 207890

Featuring work from the group of Prof. Yanyi Huang

at the College of Engineering and the Biodynamic

Optical Imaging Center (BIOPIC), Peking University,

Beijing, China.

Title: Discretely tunable optofl uidic compound microlenses

Optofl uidic compound microlenses, made from PDMS through

multilayer soft lithography, can tune the focal length and zooming

power discretely and accurately by actuating monolithically

integrated pneumatic valves.

As featured in:

See Yanyi Huang et al.,

Lab Chip, 2011, 11, 2835.

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online / Journal Homepage / Table of Contents for this issue

Dynamic Article LinksC<Lab on a Chip

Cite this: Lab Chip, 2011, 11, 2827

www.rsc.org/loc PAPER

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

Automated cellular sample preparation using a Centrifuge-on-a-Chip†

Albert J. Mach,ab Jae Hyun Kim,a Armin Arshi,a Soojung Claire Hurab and Dino Di Carlo*ab

Received 19th April 2011, Accepted 14th July 2011

DOI: 10.1039/c1lc20330d

The standard centrifuge is a laboratory instrument widely used by biologists and medical technicians

for preparing cell samples. Efforts to automate the operations of concentration, cell separation, and

solution exchange that a centrifuge performs in a simpler and smaller platform have had limited

success. Here, we present a microfluidic chip that replicates the functions of a centrifuge without

moving parts or external forces. The device operates using a purely fluid dynamic phenomenon in which

cells selectively enter and are maintained in microscale vortices. Continuous and sequential operation

allows enrichment of cancer cells from spiked blood samples at the mL min�1 scale, followed by

fluorescent labeling of intra- and extra-cellular antigens on the cells without the need for manual

pipetting and washing steps. A versatile centrifuge-analogue may open opportunities in automated,

low-cost and high-throughput sample preparation as an alternative to the standard benchtop centrifuge

in standardized clinical diagnostics or resource poor settings.

Introduction

The standard benchtop centrifuge is one of the most common

instruments in the life science laboratory used ubiquitously for

sample preparation in cell biology research and medical diag-

nostics. Typical sample preparation procedures require multiple

centrifugation steps for cell labeling and washing, which can be

a time consuming, laborious, and costly process for diagnostics

and research. In fact, while assays themselves have widely been

miniaturized and automated, sample preparation required for

these assays has been identified as a key target for future auto-

mation.1 Centrifuges perform three critical sample preparation

steps that make them so widely used: (i) separation of cells by

size/density, (ii) concentration of cells, and (iii) solution

exchange. Because centrifuges can perform such disparate

functions, realizing these functions in a miniaturized platform

has been challenging.

Miniaturized microfluidic approaches often successfully

implement one or two of these functions. For example, cell

separation by size and density has been accomplished by using

physical obstacles, external forces, or fluidic forces to guide

particles to defined locations in a microchannel for collection at

different outlets.2–6 While these methods may offer high resolu-

tion cell separation, the typical collected liquid volume is similar

to the injected liquid volume – that is, no significant concentra-

tion is achieved. This large output volume can hinder

aDepartment of Bioengineering, University of California, Los Angeles, CA,90095, USA. E-mail: dicarlo@seas.ucla.edubCalifornia NanoSystems Institute, Los Angeles, CA, 90095, USA

† Electronic supplementary information (ESI) available. Detaileddescriptions of device fabrication and experimental setups. See DOI:10.1039/c1lc20330d

This journal is ª The Royal Society of Chemistry 2011

downstream cell detection platforms that may require scanning

large fields of view to observe the cells of interest or leads to

dilution of biomolecules of interest if collected cells must be

lysed. Thus, a method of concentration must be used in-line with

the separation system to reduce the liquid volume for rapid

detection and analysis. There are a variety of techniques for

concentrating particles and cells in localized regions with

microfluidic systems.7,8 Of these, mechanical traps are the most

commonly used method that anchors particles and cells to

a physical structure and enables multistep perfusion of reagents

to perform cell assays on-chip via solution exchange.9,10 Often,

however, it may be important to release particles and cells

on-demand for further downstream analysis.11,12 Although

successful at concentration and release, cells immobilized in these

trap-and-release systems can squeeze through traps and become

damaged when operated at higher volumetric throughput,

thereby limiting concentration factors to below what is necessary

for concentration of rare cells or dilute cell solutions. Thus,

a general purpose miniaturized tool that recapitulates all of the

functions and flexibility of a traditional centrifuge has yet to be

achieved.

Here, we demonstrate a microfluidic chip that can perform all

of the operations attributed to a benchtop centrifuge. This

‘‘Centrifuge-on-a-Chip’’ can perform high-throughput cell

concentration, size-based cell sorting, and solution exchange.

The system operates without moving parts but instead employs

fluid vortices to trap cells passively using only hydrodynamic

forces. We investigate the behavior of particle entry and trapping

within fluid vortices and identify the hydrodynamic forces

responsible for the particle trapping phenomenon. Unlike

traditional centrifugation, trapping in vortices is an all-or-none

process in which a balance of forces controls trapping, and no

Lab Chip, 2011, 11, 2827–2834 | 2827

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

trapping occurs below a threshold. Additionally, we demonstrate

biological sample preparation applications in concentrating and

enriching rare cells from heterogeneous solutions as well as

performing cell labeling assays on-chip. We envision this simple

method may open opportunities in automated, low-cost and

high-throughput sample preparation: (i) as an alternative to the

standard benchtop centrifuge in resource poor settings, (ii) for

streamlining standard cell-based diagnostics to reduce human

error, or (iii) for large volume size-based cell separations.

Fig. 1 Particle entry mechanism in laminar microvortices. (a) For

a polydisperse particle solution injected into a device with a straight high-

aspect ratio channel leading into an expansion-contraction chamber we

expect size-dependent entry into the laminar vortices created. (b, c)

Particles are subjected to a shear gradient lift force, which directs particles

toward the channel wall, and a wall effect lift force, directed toward the

channel center, which leads to entrainment of particles at dynamic

equilibrium positions, Xeq. (d) As focused particles enter the vortex

chamber, the lift forces are decoupled due to the absence of a nearby wall,

resulting in a dominate shear gradient lift force. Larger particles (red)

experience larger lift forces and are able to migrate across fluid stream-

lines into the vortices while smaller particles (blue) follow fluid stream-

lines and flow out of the system.

Results

Theoretical background

The Centrifuge-on-a-Chip device employs microscale fluid

vortices to ‘trap’ and ‘release’ particles and cells in suspension.

While microscale vortices are less well-known, mainly since flow

in microchannels has been regarded to lack appreciable fluid

inertia,13 recent investigations have demonstrated the use of

microvortices for cell manipulation,14–18 plasma extraction,19

particle focusing,20 and fluid mixing.21 Previously, we have shown

particle trapping in fluid vortices using a similar platform,22 but

did not show a theoretical understanding of this phenomenon.

Microvortices can arise when a microchannel is quickly

expanded in width leading to jetting of the narrow entering

stream of fluid, detachment of the boundary layer, and recircu-

lation in the expansion region. Vortex formation relies on fluid

inertia: that is increasing Reynolds number of the flow leads to

increasing vortex size until the full expansion region is occupied

(Supplementary Fig. 1). Here, the Reynolds number is defined as

Rc ¼ rUW/m, where U is the maximum fluid velocity, W is the

channel dimension prior to the expansion, m is the fluid viscosity,

and r is the fluid density. It is important to note that the

microvortices created in this system are different from vortices

created in the streamwise direction such as Dean vortices created

in curved channel flows with inertia4,23 or vortices created due to

asymmetrically structured microchannels.24

In order to use trapping of particles in vortices for sample

preparation, there are two aspects of particle interaction with

laminar microvortices that are important to investigate: (i) under

what conditions do particles migrate across streamlines to enter

vortices, and (ii) what leads to maintenance of particles within

vortices once they have entered.

We first explored the mechanism of particle migration and

entry into two symmetric vortices created in a microfluidic chip

consisting of a single microchannel with an expansion-contrac-

tion chamber (Fig. 1a). When a polydisperse solution of particles

is introduced into the system, the particles begin randomly

distributed throughout the channel cross-section at the inlet

(Fig. 1b). As particles travel downstream, they are subjected to

a shear gradient lift force, directing particles toward the channel

wall, and an opposing wall effect lift force, that is due to the

presence of the wall, leading to migration of particles toward the

channel centerline. A combination of these forces leads to

entrainment of particles at dynamic equilibrium positions about

halfway between the channel centerline and wall, Xeq

(Fig. 1c).4,5,25 For the high-aspect ratio microchannels used (70

mm height, 50 mm width), this leads to two focusing positions

along the long face of the channel as previously reported.25–28

2828 | Lab Chip, 2011, 11, 2827–2834

There is also a slight particle size-dependence to focusing posi-

tion, where larger particles are focused closer to the channel

center.25 Importantly, when a focused particle reaches the

expansion region (Fig. 1d) the neighboring microchannel wall is

no longer within its vicinity. Therefore, the hydrodynamic

interaction with the wall disappears leading to a loss of a signif-

icant wall-effect lift to balance the remaining shear-gradient lift

directed down the shear gradient towards the vortex center. The

gradient in shear rate responsible for shear gradient lift decays

only slowly as the particle moves downstream through the

expansion (Supplementary Fig. 5a,b).

The end result of these effects is that transverse particle

migration due to shear gradient lift controls entry of particles

into vortices. In previous work25 we demonstrated that when the

ratio of the particle diameter, a, to channel width,W, approaches

one, the shear gradient lift (FL) in a straight channel would scale

as: FL ¼ fLrU2a3

WwhereU is the maximum fluid velocity, and r is

the fluid density. The non-dimensional lift coefficient (fL) is

dependent on the local shape of the velocity profile, and the

channel Reynolds number (Rc). For small a/W, however, FL was

analytically determined to scale with a4.29 Taking these two

possible scalings into account, and assuming that the velocity

profile remains similar upon entering the expansion, we can

determine a transverse migration velocity, vt, into the vortex by

balancing shear gradient lift with Stokes drag Fstk ¼ 3pmavt. We

find that the transverse migration velocity, vt, will be a function

of particle diameter - between a2 to a3. Following this scaling,

larger particles (red) are expected to migrate laterally through

fluid streamlines and into the vortex while smaller particles (blue)

below a critical size will not migrate fast enough to cross into the

vortex before passing out of the expansion region (Fig. 1d).

This journal is ª The Royal Society of Chemistry 2011

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

It should be noted that entry into the vortex is controlled solely

by shear gradient lift forces in the case of a dilute solution

neglecting particle–particle interactions. In the case of highly

concentrated particle solutions, like blood used in this work,

interparticle hydrodynamic interactions are common, and cross-

stream migration and entry can be assisted by particle collisions

or disturbance flows from neighboring particles.30 Although we

have not explored it in detail, it is apparent that trapping is

robust to changes in the vortex chamber geometry – as long as

the shear gradient lift is present – including whether vortex

chambers expand in one or both directions simultaneously

during an expansion.22 This design flexibility can aid in devel-

oping parallel arrays with a small footprint for higher

throughput operation.

Once particle entry into a microvortex has occurred, the

problem of maintaining particles in vortices becomes important.

Shear gradient lift should still operate in the vortex, given that

a gradient in shear rate is present (Supplementary Fig. 4), leading

to a force towards the vortex center. All particles would even-

tually migrate to the vortex center without the presence of

a balancing force directed outwards. The key candidate is the

inertia of the particles themselves, which leads to an outward

motion with respect to the occupied fluid streamline when the

streamline changes direction. For the simplest case in which

particles are traveling in a circular path and the pressure gradient

is constant across the particle, this leads to an effective centrif-

ugal force directed outwards:31 Fcfg ¼ (rp � r)pa3vp2/6r, where vp

is the tangential particle velocity, rp is the particle density, and r

is the radius of the orbit. For this case, if centrifugal force is

larger than the lift force over the entire vortex region (e.g. Fcfg/FL

> 1) then the particle would leave the vortex, while if Fcfg/FL is#

1 within any particular orbit then the particle would remain

trapped. Therefore, it is instructive to look at this ratio of forces.

Looking at how this ratio depends on particle properties, we find

that Fcfg/FL � (rp � r)/ran, where n ranges between 0 and 1. This

analysis indicates that higher density particles that may have

entered the vortex would leave the vortex due to centrifugal

effects, and would be separated from lower density particles. This

analysis also predicts dependence on particle size for mainte-

nance within a vortex, with larger particles more likely to remain

trapped. Admittedly, the presented model is a simplification of

the problem and assumptions of a circular path and constant

pressure gradients across particles should be modified to better

describe the phenomenon. In fact, the real system has additional

complexities from: (i) variable velocity throughout the orbit

(faster velocities near the channel centerline and slower near the

outer wall), (ii) an eccentric orbit that is not circular or elliptical

(Fig. 2a, Supplementary Video 1), and (iii) a complex shear

gradient (not a parabolic velocity profile) changing with position

along the orbit (Supplementary Fig. 5).

Size-based particle entry and trapping in microvortices

To explore the behavior of particles in fluid vortices, polymer

particles were injected into the expansion-contraction system

(Fig. 2a and Supplementary Video 1). Particles above a critical

size migrated into the vortex and maintained a stable orbital

position within the vortex (Fig. 2b,c). Particle trapping was

observed for a range of particle sizes between a/W ¼ 0.3–0.4 for

This journal is ª The Royal Society of Chemistry 2011

W ¼ 50 mm and a/W ¼ 0.4–0.45 for W ¼ 40 mm, where a is the

particle diameter and W is the channel width (Supplementary

Fig. 2). Particles smaller than the critical size range were never

observed to be trapped and flowed past the trapping chamber

and out of the system (Supplementary Fig. 3a). The two datasets

from Supplementary Figure 2 together suggest that the critical

size leading to entry into the vortex does not depend strongly on

the inlet channel width and should be robust to small variations

in this value.

With increasing particle size, migration occurred over

a shorter downstream distance and resulted in final positions

closer to the vortex center (Fig. 2b). In agreement with predic-

tions, this is a result of larger particles experiencing higher shear

gradient lift forces leading to faster migration. Note that in the

region in which particles enter the vortex the fluid does not

change direction appreciably, minimizing local centrifugal effects

that might push these particles outwards. Therefore, measured

particle trajectories and dynamics calculated from these trajec-

tories (Supplementary Fig. 4) are expected to yield a scaling

similar to shear gradient lift force alone. In fact, the maximum

lift force in this entry region for individual particles over a range

of sizes was found to scale with a best fit of a3.2, which closely

matches previously reported predictions of between a3 25 to a4 29

(Fig. 2e).

Once particles migrated into the vortex they traveled in

continuous closed paths around the vortex center (Fig. 2c). While

particles were orbiting, we observed an increase in particle

velocity within the orbit while traveling close to the main flow as

it experiences higher shear rates (Supplementary Fig. 5a–c,

Supplementary Video 1). We also found that larger particles

occupied the inner orbits closer to the vortex center while smaller

particles traveled in outer orbits (Fig. 2c). We observed a similar

phenomenon with multiple particles in the vortex traps, where

larger particles orbited closer to the vortex center (Fig. 2d,

Supplementary Video 2). These orbital patterns confirm that

particles experience a balance of inertial and centrifugal forces

within the vortex region that depends on the particle size.

Trapped orbiting particles could be released by changing the

flow conditions such that the vortex dissipated (Supplementary

Fig. 3b, Supplementary Video 3). This was accomplished by

decreasing the input flow rate, which simultaneously reduced the

vortex size, allowing particles to escape the vortices into the main

channel flow.

Enrichment of rare cancer cells from blood

The Centrifuge-on-a-Chip was applied to separating and

concentrating cancer cells (diameter of 20 mm) from normal

human blood cells (diameters range from 2 to 15 mm) to

demonstrate utility for size-based enrichment and concentration

in a high-throughput manner. Enriching and concentrating

cancer cells from blood is particularly important for clinical

diagnostics as circulating tumor cells (CTCs) can provide real-

time information on patient status and monitoring of cancer

therapies.32 Isolating viable CTCs from blood in a quick, effec-

tive and label-free approach remains a significant technical

challenge – CTCs are rare events at rates as low as one cell per

one billion blood cells.32,33 While current strategies focus on

enumeration of CTCs for diagnostics,34 there is a critical need for

Lab Chip, 2011, 11, 2827–2834 | 2829

Fig. 2 Size-based entry and maintenance of particles in microvortices. (a) Time-lapse high-speed image of a PDMS particle becoming ‘trapped’ in the

fluid vortex and orbiting around the vortex center at a stable position. Elapsed time is �8 ms. (b) Average trajectories of polydisperse PDMS particles

overlaid on fluid direction vectors obtained from a COMSOL model. Particle sizes larger than the ratio of particle diameter to channel width of 0.3

entered in vortex traps. As the particle size increases, the particle migrates closer to the vortex center. (c) Once the particle migrates into the vortex, it

occupies an orbit around the vortex center that is dependent on its size. (d) Time-lapse high-speed image of two PDMS particles interacting and traveling

in two separate orbits inside a vortex. The larger particle occupies an orbit closer to the vortex center. Elapsed time is�7 ms. (e) The maximum lift force

upon entry into a vortex was found to scale with the particle diameter to the 3.2 power.

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

gathering larger sample sizes of viable CTCs for research

purposes.35 This requires processing large blood volumes with

higher throughputs and enriching target cells without the

attachment to modified substrates or magnetic beads, providing

an advantage for individually selecting captured cells for further

analysis or culture.

The Centrifuge-on-a-Chip system addresses the need for rare

cell enrichment with a massively parallel device that processes

liquid volumes in the mL min�1 range, enriches target cells

through size and density-based separation, and releases captured

cells into a smaller concentrated volume (Fig. 3a,b). To

demonstrate rare cell enrichment, fluorescently-labeled breast

cancer cells (MCF-7) spiked into diluted human blood was

injected into the Centrifuge-on-a-Chip device at 4.4 mL min�1.

At these high flow rates channel deformation is observed in the

upstream vortex reservoirs,39 however trapping is not signifi-

cantly impacted given that downstream vortex chambers oper-

ating closer to ambient pressure remain undeformed. Higher

operational flow rates are instead limited by bond strength.

Spiked MCF-7 cells included single cells and 2–4 cell clusters, as

clustered cells have been shown to be present at significant levels

in clinical samples.24 Blood and cancer cells were observed to

enter and orbit in the vortices during the injection step (Fig. 3c,

Supplementary Video 4). Red blood cells were observed to enter

vortices even though particles of similar size did not migrate into

vortices in experiments with dilute samples. Likely, the high cell

concentration induces collisions and hydrodynamic disturbances

between cells that lead to cross-stream migration and entrance

into vortices. Additionally, there is a maximum capacity of cells

2830 | Lab Chip, 2011, 11, 2827–2834

each vortex chamber can maintain. After the vortex occupies the

entire reservoir a maximum of �40 single MCF7 cells can be

maintained over a range of higher flow rates. For most spiking

experiments we operated at conditions well below this maximum.

Once the solution was completely processed, the vortex-trapped

cells were ‘‘washed’’ with PBS without disrupting the vortices

(Fig. 3d, Supplementary Video 4). Interestingly, we observed

that blood cells that initially entered the vortex were not stably

trapped and quickly exited from the traps and out of the system

leaving only the larger stably trapped cancer cells orbiting

(Fig. 3d, Supplementary Video 4). In agreement with our model,

red and white blood cells have both higher density and/or smaller

size, and therefore cannot form stable orbits. Washed cells were

released into one well of a 96-well-plate for characterization and

enumeration.

The Centrifuge-on-a-Chip system performs well when quan-

tifying key metrics for target cell concentration, enrichment, and

purity. 10 mL volume blood samples (n $ 6 samples) of 5% v/v

blood (i.e. 0.5 mL whole blood or�2.5 billion blood cells) spiked

with�500 cancer cells were concentrated to a final volume of less

than 200 mL (20-fold volumetric concentration) with relatively

little blood cell contamination in < 3 min (Fig. 3e, Supplemen-

tary Table 1). This corresponds to an enrichment ratio (the ratio

of target cancer cells to contaminant blood cells in the output

divided by the same ratio in the input solution) of 3.4 million

(Fig. 3f). This high level of enrichment leads to high purity of the

cancer cells in the 200 mL final volume:�40% (Fig. 3g, an average

of 102 � 21 cancer cells, and 221 � 155 blood cells). Blood

samples without spiked cancer cells (n ¼ 3) that were processed

This journal is ª The Royal Society of Chemistry 2011

Fig. 3 High purity rare cell enrichment from whole human blood. (a) A schematic of the experimental setup is shown with two syringe pumps con-

taining a blood sample spiked with cancer cells and a PBS wash solution. (b) A photograph demonstrates the small footprint of two Centrifuge-on-a-

Chip devices. (c) A spiked blood sample is injected into the microchip causing blood and cancer cells to enter the fluid vortices. (d) Once the blood

solution has been flowed through, a wash step is performed to remove smaller and denser blood cells while maintaining the vortices and trapped cancer

cells. (e) Wide-field fluorescent and brightfield images of cancer cells and blood cells are shown before and after processing of blood through the

Centrifuge-on-a-Chip. (f) Capture efficiency, (g) enrichment, and (h) purity data forMCF7 cells spiked into 5–20% diluted blood (�500 cells mL�1 whole

blood) and processed on the Centrifuge-on-a-Chip indicates the ability to concentrate cancer cells while rejecting other blood cells at clinically useful

levels. (i) Characterization of a cell cluster with anti-Cytokeratin-PE and DAPI labeled in a 96-well-plate indicates the ability to perform molecular

analysis on the captured cells.Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

with Centrifuge-on-a-Chip and collected in the well were found

to have 772 � 283 red blood cells and 4 � 1 CD45+ white blood

cells, which is similar to the amount of blood cell contaminants

found in the microwells using spiked blood samples. The level of

enrichment achieved is comparable to molecular affinity-based

and filter-based approaches for target cell separation which have

reported enrichments from 1 million36 to 10 million.37 The purity

of our processed sample is high when compared to affinity-based

approaches which report purities of spiked cancer cells of 9.2–

14.0%.24 Reducing the dilution of blood in processed samples

leads to increases in cell-processing throughput, but also results

in reduced capture efficiency of spiked cells. We recovered 10–

20% of the spiked cancer cells, with decreasing capture efficiency

with increasing blood concentrations (Fig. 3h). Higher blood

concentrations lead to higher fluid viscosities which modify the

fluid vortex size and position, resulting in lower trapping effi-

ciency. This relatively low capture efficiency suggests that in

order for this technique to be useful in isolating ultra-rare cells

occurring at 1–10 cells mL�1, a large volume of blood must be

processed (10 mL or more). However, the high throughput of our

approach (�5 mL min�1 of diluted blood for a 2 cm2 chip)

indicates that operation on large volumes in a reasonable time

period (< 30 min) is achievable.

This journal is ª The Royal Society of Chemistry 2011

Cells captured in the device maintained high levels of viability.

We observed no significant changes in cell viability (90.1% vs.

90.3% initial) after injecting cells through the Centrifuge-on-a-

Chip as determined by a fluorescent live/dead assay. Cells

exposed to similar flow in a microfluidic device were also not

shown to have significant changes in gene expression.38 Viable

cells may be important for some sample preparation applications

of a Centrifuge-on-a-Chip.24

Cells captured and released from the Centrifuge-on-a-Chip

system are available for standard molecular assays such as

immunostaining. As a proof of concept, unlabeled spiked blood

samples were enriched with the chip. Cancer cells were then

released and labeled in a microwell. Cancer cells stained positive

for Cytokeratin-PE and DAPI and negative for CD45 (Fig. 3i).

This ability to enrich on one platform but transfer cells in a small

volume for further processing offers significant advantages for

rare single cell analysis.

On-chip cell labeling and solution exchange

The Centrifuge-on-a-Chip was also used to effectively label cells

for specific molecular markers. In traditional centrifugation, cell

samples are labeled for specific markers through a series of

Lab Chip, 2011, 11, 2827–2834 | 2831

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

labeling and washing steps. This includes incubating the cells

with labeling reagents in a centrifuge tube, concentrating the cells

into a pellet with a benchtop centrifuge, removing the superna-

tant layer containing unbound labeling reagents through manual

aspiration, and manually resuspending the cells in a new

medium. These operations were performed within the micro-

fluidic chip by trapping the cells within fluid vortices and

sequentially exposing trapped orbiting cells to labeling reagents,

followed by a PBS wash solution (Fig. 4b–e, Supplementary

Fig. 6a). Labeled cells were then released within a small volume

into a collection vial by reducing flow.

Cells prepared using standard protocols with a benchtop

centrifuge and cells prepared with the Centrifuge-on-a-Chip were

observed to have similar fluorescent labeling (Fig. 4a). Labeling

with antibodies to intracellular (cytokeratin) and cell surface

proteins (EpCAM), fluorogenic enzyme substrates (Calcein

AM), and direct labeling of DNA (DAPI) yielded similar results

when compared to macroscale centrifugation (Fig. 4a). However,

Fig. 4 On-chip cell labeling of small molecules, antibodies, and microspheres

with standard centrifuge-based protocols and the Centrifuge-on-a-Chip. La

DAPI), cell surface proteins (anti-EpCAM FITC), and fluorogenic enzym

sequentially with primary anti-EpCAM followed by a secondary antibody

examining the binding of streptavidin-coated microspheres to biotinylated an

biotinylated EpCAMwere injected into the device and trapped in the vortex, u

with streptavidin-coated microspheres and (d) continuous 3D reaction. (e)

microspheres (yellow arrow) and release cells from vortex traps into a 96-wel

point to particles that are increasingly bound to the cell over 2 min.

2832 | Lab Chip, 2011, 11, 2827–2834

unlike the standard protocols, none of the multiple 5 min

centrifugation and manual resuspension steps were required with

the microchip device. For a valid comparison, cells were exposed

to the same amount (mass) of labeling reagents and incubated for

the same amount of time in both methods. The ability to hold

cells stably in place within fluid vortices allowed for multiple

solution exchanges with labeling agents and wash solutions

(Supplementary Fig. 6b–f) in a format that can be automated.

Each addition of a new solution took approximately 100 ms for

complete exchange. For the same labeling reaction a traditional

centrifuge-based process requires 6 centrifugation steps that

includes 3 washing steps and requires > 30 min of sample prep-

aration time (this excludes the incubation time with labeling

reagents). Each centrifugation and wash step can potentially

result in a loss of a small proportion of cells and requires between

5–10 min.40

Fast labeling is aided by cells that rotate and orbit in the

fluid vortex such that they are exposed to a constantly

via solution exchange. (a) Side-by-side comparison of MCF7 cells labeled

beling of intracellular proteins and DNA (anti-Cytokeratin-FITC and

e substrates (Calcein AM). Additionally, cells were able to be labeled

conjugated to AlexaFluor647. We observed the process of labeling by

ti-EpCAM antibodies on the cell surface. (b) MCF7 cells incubated with

ndergoing a constant rotating and orbiting motion. (c) Solution exchange

Once complete, a ‘wash’ is performed with PBS to remove unbound

l-plate for characterization. Scale bars correspond to 50 mm. Red arrows

This journal is ª The Royal Society of Chemistry 2011

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

refreshed milieu of molecular labels. In other words, strong

convection of labeling reagents in the vortex leads to a very

small depleted region of reagents near the cell surface and

a strong gradient driving more reagents to the cell surface.21,41

We observed this process of fast labeling by examining the

binding of streptavidin-coated microspheres to biotinylated

anti-EpCAM antibodies on the cell surface. We found that the

cells in the Centrifuge-on-a-Chip system accumulated the same

number of microbeads in 5 min that cells prepared with the

standard protocol accumulated in 30 min (Supplementary

Fig. 6h). Further, after 30 min, cells labeled with Centrifuge-

on-a-Chip on average had twice the number of microbeads

bound per cell compared to standard methods (Supplementary

Fig. 6i).

Sequential operations: rare cell enrichment followed by

fluorescent labeling

Multiple sequential sample preparation steps enabled by

a centrifuge were successfully conducted using Centrifuge-on-a-

Chip. Size-based trapping of cancer cells from blood, sequential

fluorescent labeling, and analysis of released cells were conducted

in < 1 h. Diluted human blood (10 mL) spiked with cancer cells

was injected into the device for �3 min to enrich the cancer cells.

Trapped cells were sequentially prepared with fixation and per-

meabilization agents and stained with fluorescent antibodies for

20 min. Cells were then washed with PBS for < 1 min, and

collected into a 96-well-plate for characterization. Collected cells

labeled positive for cytokeratin and DAPI, indicating the success

of sequential sample preparation (Supplementary Fig. 6g). This

demonstrates a complete route to automation of all of the sample

preparation processes required for cell analysis in a single simple

platform.

Discussion

In this work, we demonstrate a flexible microfluidic chip that can

perform several functions: (i) separating cell samples in a high-

throughput, size and density-based manner, (ii) concentrating

target cells from tens of millilitres of volume into smaller mL

volumes, and (iii) labeling cells with fluorescent markers and

washing unbound dye through solution exchange. This was

achieved through a unique technology that allows trapping of

particles and cells in massively parallel laminar microscale

vortices. Using this phenomenon we go beyond the capabilities

of a traditional centrifuge and demonstrate two sample prepa-

ration applications in which we isolate cancer cells from human

blood and fluorescently label rare cells.

The Centrifuge-on-a-Chip harbors a distinct advantage for

preparing rare cell samples compared to the standard benchtop

centrifuge. Traditional centrifugation is especially difficult to

work with small volumes and rare cell samples since cell pellets

cannot be observed with the naked eye, resulting in manual

pipetting of blood and plasma layers that limits enrichment and

purity in traditional systems. In contrast, the Centrifuge-on-a-

Chip system works well with rare and dilute cell samples as it

can concentrate these samples into a smaller trapped volume

and perform sequential assays without cell loss with each

solution exchange. This may be useful when working with

This journal is ª The Royal Society of Chemistry 2011

clinical samples containing very small numbers of cells such as

fine needle aspirates, small tumor biopsies, and cells in dilute

solutions such as pleural fluid or urine. Additionally, the

continuous filtration of mL blood volumes is similar to elutri-

ation approaches,42 but is superior in that target cells are

concentrated and accumulated into a smaller volume. This

ability to enrich on one platform but transfer cells in a small

volume for further processing offers significant advantages for

rare cell analysis. First, concentrating the captured cells within

a microwell with a field-of-view spanning 0.3 cm2 allows for

rapid imaging and detection in minutes compared to other

separation approaches which can employ wide field areas larger

than 6.25 cm2 and require extensive imaging time. Furthermore,

target cells collected in the microwell can be selected and

manually transferred to other volumes as the cells are not

bound to any physical structure. Thus, the cells can be inves-

tigated using molecular analysis approaches or various down-

stream assays which are simply and rapidly integrated into

current clinical or research practice. Additionally, the ability to

concentrate and sequentially treat cells on-chip with chemical

and biological agents provides a powerful and simple tool for

sample preparation that can be used in-line with downstream

cell analysis platforms such as flow cytometry.

More generally, the Centrifuge-on-a-Chip can continuously

filter particles by size and density at high flow rates, without

external forces, and in a small footprint, which can be applied to

applications in water filtration and concentration systems. The

capability to release and refresh the vortex traps allows for long-

term operation without clog-prone filters. More immediately, we

envision this simple method may open opportunities in auto-

mated, low-cost and high-throughput sample preparation, as an

alternative to the standard bench top centrifuge in resource poor

settings, for streamlining standard cell-based diagnostics to

reduce human error, or for large volume size-based cell

separations.

Experimental methods

Materials

PDMS (Polydimethylsiloxane) beads were prepared from

a mixture of 10% w/v PDMS (Dow Corning; Sylgard 184),

deionized water and 0.1% Tween 20. The mixture was shaken

vigorously with a vortexer and placed at 65 �C overnight to allow

hardening into solid PDMS beads. After curing, PDMS beads

smaller than 50 mm were extracted from the bead solution via

centrifugation. MCF7 breast cancer cells cultured in media

containing DMEM supplemented with 10% FBS, 1% bovine

insulin, and 1% penicillin/streptomycin were trypsinized and

resuspended before use. Blood was collected from healthy human

volunteers by a trained physician and diluted in PBS to 5–20%

for experiments.

Author contributions

A.J.M. and D.D. proposed the concept of the work, carried out

theoretical analysis, and wrote the paper. A.J.M., A.A., J.H.K.,

and S.C.H. carried out the experiments. A.J.M., A.A., J.H.K.,

and D.D. carried out experimental analysis.

Lab Chip, 2011, 11, 2827–2834 | 2833

Publ

ishe

d on

29

July

201

1. D

ownl

oade

d by

Lom

onos

ov M

osco

w S

tate

Uni

vers

ity o

n 16

/08/

2013

09:

26:5

1.

View Article Online

Acknowledgements

We thank Dr Sean McGhee, Dr Edward R. B. McCabe and Dr

Elodie Sollier for providing de-identified blood samples and Dr

Wonhee Lee for helpful discussions. This work was supported by

aWallace H. Coulter Foundation Translational Research Award

and grant N66001-10-1-4072 from the Defense Advanced

Research Projects Agency. We also thank Marc Lim for the

artistic rendering of the Centrifuge-on-a-Chip.

References

1 M. Toner and D. Irimia, Annu. Rev. Biomed. Eng., 2005, 7, 77–103.2 N. Pamme, Lab Chip, 2007, 7, 1644–1659.3 L. R. Huang, E. C. Cox, R. H. Austin and J. C. Sturm, Science, 2004,304, 987–990.

4 D. Di Carlo, D. Irimia, R. G. Tompkins and M. Toner, Proc. Natl.Acad. Sci. U. S. A., 2007, 104, 18892–18897.

5 D. Di Carlo, Lab Chip, 2009, 9, 3038–3046.6 D. R. Gossett, W. M. Weaver, A. J. Mach, S. C. Hur, H. T. K. Tse,W. Lee, H. Amini and D. Di Carlo, Anal. Bioanal. Chem., 2010, 397,3249–3267.

7 J. Nilsson,M. Evander, B. Hammarstr€om and T. Laurell,Anal. Chim.Acta, 2009, 649, 141–157.

8 J. Warrick, B. Casavant, M. Frisk and D. Beebe, Anal. Chem., 2010,82, 8320–8326.

9 A. M. Skelley, O. Kirak, H. Suh, R. Jaenisch and J. Voldman, Nat.Methods, 2009, 6, 147–152.

10 J. El-Ali, P. K. Sorger and K. F. Jensen, Nature, 2006, 442, 403–411.11 W.-H. Tan and S. Takeuchi, Proc. Natl. Acad. Sci. U. S. A., 2007, 104,

1146–1151.12 D. Irimia and M. Toner, Lab Chip, 2006, 6, 345.13 C. M. Lin, Y. S. Lai, H. P. Liu, C. Y. Chen and A. M. Wo, Anal.

Chem., 2008, 80, 8937–8945.14 Marcos and R. Stocker, Limnol. Oceanogr.: Methods, 2006, 4, 392–

398.15 D. Chiu, Anal. Bioanal. Chem., 2007, 387, 17–20.16 J. P. Shelby and D. T. Chiu, Lab Chip, 2004, 4, 168–170.17 J. P. Shelby, S. A. Mutch and D. T. Chiu, Anal. Chem., 2004, 76,

2492–2497.18 J. P. Shelby, D. S. W. Lim, J. S. Kuo and D. T. Chiu, Nature, 2003,

425, 38.19 E. Sollier, M. Cubizolles, Y. Fouillet and J.-L. Achard, Biomed.

Microdevices, 2010, 12, 485–497.20 J.-S. Park, S.-H. Song and H.-I. Jung, Lab Chip, 2009, 9, 939–948.

2834 | Lab Chip, 2011, 11, 2827–2834

21 A. P. Sudarsan and V. M. Ugaz, 2006, 103, pp. 7228–7233.22 S. C. Hur, A. J. Mach and D. Di Carlo, Biomicrofluidics, 2011, 5.23 J.Wang, Y. Zhan, V.M.Ugaz andC. Lu,LabChip, 2010, 10, 2057–2061.24 S. L. Stott, C.-H. Hsu, D. I. Tsukrov, M. Yu, D. T. Miyamoto,

B. A. Waltman, S. M. Rothenberg, A. M. Shah, M. E. Smas,G. K. Korir, F. P. Floyd, A. J. Gilman, J. B. Lord, D. Winokur,S. Springer, D. Irimia, S. Nagrath, L. V. Sequist, R. J. Lee,K. J. Isselbacher, S. Maheswaran, D. A. Haber and M. Toner,Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 18392–18397.

25 D. Di Carlo, J. F. Edd, K. J. Humphry, H. A. Stone and M. Toner,Phys. Rev. Lett., 2009, 102, 094503-1–094503-4.

26 S. C. Hur, H. T. K. Tse and D. Di Carlo, Lab Chip, 2010, 10, 274–280.27 J. F. Edd, D. Di Carlo, K. J. Humphry, S. Koster, D. Irimia,

D. A. Weitz and M. Toner, Lab Chip, 2008, 8, 1262–1264.28 K. J. Humphry, P. M. Kulkarni, D. A. Weitz, J. F. Morris and

H. A. Stone, Phys. Fluids, 2010, 22, 081703.29 E. S. Asmolov, J. Fluid Mech., 1999, 381, 63–87.30 W. Lee, H. Amini, H. A. Stone and D. Di Carlo, Proc. Natl. Acad.

Sci. U. S. A., 2010, 107, 22413–22418.31 D. S. W. Lim, J. P. Shelby, J. S. Kuo and D. T. Chiu, Appl. Phys.

Lett., 2003, 83, 1145–1147.32 M. Cristofanilli, G. T. Budd, M. J. Ellis, A. Stopeck, J. Matera,

M. C. Miller, J. M. Reuben, G. V. Doyle, W. J. Allard,L. W. M. M. Terstappen and D. F. Hayes, N. Engl. J. Med., 2004,351, 781–791.

33 J.-M. Hou, M. Krebs, T. Ward, K. Morris, R. Sloane, F. Blackhalland C. Dive, Cancers, 2010, 2, 1236–1250.

34 B. Mostert, S. Sleijfer, J. A. Foekens and J. W. Gratama, CancerTreat. Rev., 2009, 35, 463–474.

35 Z. S. Lalmahomed, J. Kraan, J. W. Gratama, B. Mostert, S. Sleijferand C. Verhoef, J. Clin. Oncol., 2010, 28, e288–e289.

36 S. Nagrath, L. V. Sequist, S. Maheswaran, D. W. Bell, D. Irimia,L. Ulkus, M. R. Smith, E. L. Kwak, S. Digumarthy,A. Muzikansky, P. Ryan, U. J. Balis, R. G. Tompkins,D. A. Haber and M. Toner, Nature, 2007, 450, 1235–1239.

37 H. K. Lin, S. Zheng, A. J. Williams, M. Balic, S. Groshen, H. I. Scher,M. Fleisher, W. Stadler, R. H. Datar, Y.-C. Tai and R. J. Cote, Clin.Cancer Res., 2010, 16, 5011–5018.

38 S. C. Hur, N. K. Henderson-MacLennan, E. R. B.McCabe and D. DiCarlo, Lab Chip, 2011, 11, 912.

39 E. Sollier, C. Murray, P. Maoddi, and D. Di Carlo, Lab Chip, inrevision.

40 O. Lara, X. Tong, M. Zborowski and J. J. Chalmers, Exp. Hematol.,2004, 32, 891–904.

41 H. Lee, J. Kim, H. Kim, J. Kim and S. Kwon, Nat. Mater., 2010, 9,745–749.

42 R. L. Eifler, J. Lind, D. Falkenhagen, V. Weber, M. B. Fischer andR. Zeillinger, Cytometry B Clin Cytom, 2011, 80, 100–111.

This journal is ª The Royal Society of Chemistry 2011