untitled - c2lc40392g.pdf

A magnetic cell-based sensor

{

Hua Wang,

{

Alborz Mahdavi,

{

David A. Tirrell

and Ali Hajimiri

Received 23rd April 2012, Accepted 13th August 2012

DOI: 10.1039/c2lc40392g

Cell-based sensing represents a new paradigm for performing direct and accurate detection of cell- or

tissue-specific responses by incorporating living cells or tissues as an integral part of a sensor. Here we

report a new magnetic cell-based sensing platform by combining magnetic sensors implemented in the

complementary metal-oxide-semiconductor (CMOS) integrated microelectronics process with cardiac

progenitor cells that are differentiated directly on-chip. We show that the pulsatile movements of on-

chip cardiac progenitor cells can be monitored in a real-time manner. Our work provides a new low-

cost approach to enable high-throughput screening systems as used in drug development and hand-

held devices for point-of-care (PoC) biomedical diagnostic applications.

Introduction

There is an unmet need for high-throughput and cost-effective

detection platforms for chemical and biological agents. These

platforms can be utilized for a plethora of analytical and

diagnostic applications including screening chemical libraries for

drug development, toxicity studies, point-of-care (PoC) diag-

nostics, and environmental monitoring. Conventional sensors

generally use chemical, optical, spectroscopic, electrical impe-

dance- or mass-based detection to interpret biochemical phe-

nomena. Cell-based sensing on the other hand makes use of

living cells or tissues as an integral part of the sensor, and utilizes

inherent cellular mechanisms to perform accurate detection of

cell- or tissue-specific responses.

1–6

By providing both high

sensitivity and specificity, cell-based sensing has tremendous

potential for screening biochemical agents, particularly in the

context of individualized medicine where effects often vary from

patient to patient.

7,8

However, the applications of existing cell-

based sensing platforms are often limited by detection mechan-

isms. For instance, optical detection, though useful in many

applications, requires visible changes in cell morphology, a

transparent medium, and analysis of cell images.

Moreover,

existing cell-based sensors typically need dedicated device

fabrication processes, which have low scalability and suffer

from low yields in manufacturing. On the other hand, integrated

electronic technology is becoming a powerful and low-cost

platform for implementing advanced sensors. In particular, the

complementary metal-oxide semiconductor (CMOS) process,

which is widely employed in manufacturing consumer electro-

nics, such as microprocessors and memory chips, can be directly

used to realize large biosensor arrays with high scalability at very

low cost.

9–27

More importantly, standard CMOS supports

magnetic sensing; it uses magnetic particles as sensing tags,

26,27

whose functionality is independent of the optical or electrical

properties of the cells or the medium.

28,29

This strategy provides

a robust platform for implementing cell-based sensors.

Here we seek to combine the benefits of CMOS technology

with the advantages of cell-based sensing by developing a new

detection platform based on a CMOS magnetic sensor with

cardiac progenitor cells tagged by magnetic particles. We show

that the cardiac progenitor cells can be differentiated directly on

a CMOS integrated sensor chip to form a cell-based sensor, and

that the periodic and autonomous beating of the progenitors can

be detected in real-time by CMOS magnetic sensing. We further

demonstrate that the sensor enables accurate detection of

chemical agents. The approach described here can be readily

utilized to construct low-cost and field-deployable biochemical

sensors.

Methods and materials

We used cardiac progenitors, differentiated from mouse

embryonic stem cells (ESCs), as the sensing cells. Cardiac

progenitors beat autonomously and undergo displacements of

tens of microns.

We designed and implemented a fully

integrated inductive frequency-shift magnetic sensor array in a

standard CMOS process.

With magnetic particles coated on

the cells, the beating movements of the cardiac progenitors can

be accurately captured by the CMOS magnetic sensors without

any post-processing of the data (Fig. 1). If the analyte of interest

alters the physiology of the cells and subsequently their pulsatile

Department of Electrical Engineering, California Institute of Technology,

Pasadena, CA 91125, USA. E-mail: hajimiri@caltech.edu

School of Electrical and Computer Engineering, Georgia Institute of

Technology, Atlanta, GA 30332, USA. E-mail: hua.wang@ece.gatech.edu

Department of Bioengineering, California Institute of Technology,

Pasadena, CA 91125, USA

Division of Chemistry and Chemical Engineering, California Institute of

Technology, Pasadena, CA 91125, USA

{

Electronic Supplementary Information (ESI) available: Supplementary

figures for sensor noise cancellations, SEM images, mouse embryonic

stem cells preparations, and sensor system diagram with the measure-

ment setup. See DOI: 10.1039/c2lc40392g

{

Both authors contributed equally to this paper.

Lab on a Chip

Dynamic Article Links

Cite this:

Lab Chip

, 2012,

12

, 4465–4471

www.rsc.org/loc

PAPER

This journal is

ß

The Royal Society of Chemistry 2012

Lab Chip

, 2012,

, 4465–4471

4465

Downloaded by California Institute of Technology on 13 December 2012

Published on 14 August 2012 on http://pubs.rsc.org | doi:10.1039/C2LC40392G

View Article Online

/ Journal Homepage

/ Table of Contents for this issue

movements, the changes are readily recorded by the sensor in

real-time.

Preparation of the cardiac progenitor cells

Culturing cardiac progenitors from mouse ESCs.

R1 mouse

embryonic stem cells (ESCs) were obtained from ATCC and

propagated in ESC maintenance medium supplemented with

leukemia inhibitory factor (LIF) to maintain their undifferen-

tiated state. The ESC maintenance medium consists of

Dulbecco’s Modified Eagle’s Medium containing 15% of ES

qualified fetal bovine serum (FBS) (16141-061; Gibco), 100

-mercaptoethanol (M6250; Sigma-Aldrich), 2 mM

-glutamine

(25030-081; Invitrogen Inc.), 0.1 mM nonessential amino acids

(11140-050; Invitrogen Inc.), 1 mM sodium pyruvate (11360-070;

Invitrogen Inc.), 50 U mL

2

penicillin and 50

gmL

2

streptomycin (15140-122; Invitrogen Inc.). We supplemented

the ESC maintenance medium with 500 pM murine leukemia

inhibitory factor (LIF) (ESG1107; Millipore). Differentiation

medium had an equivalent composition to ESC maintenance

medium but lacked LIF. We cultured mouse ESCs routinely on

0.2% gelatin-coated tissue-culture flasks and trypsinized them

with 0.05% Trypsin-EDTA (25300062; Gibco) for passaging

purposes. Embryoid bodies (EBs) were made from these cells by

the hanging drop technique.

Mouse ESCs of passage number

15–20 were used. Differentiation was initiated by resuspension in

differentiation medium lacking LIF. Cells were resuspended at a

density of 500 cells per 20

L of medium for the hanging drops.

Each hanging drop contained 20

L of cell suspension.

Direct EB differentiation on the CMOS sensor surface.

In order

to yield pulsatile cardiac progenitor cells with high viability and

reliable attachment to the magnetic sensor, we directly differ-

entiated ES cells on the CMOS sensor surface. To minimize

contamination, we first rinsed the chip surface as well as the

PDMS reservoir with 100% ethanol and PBS before surface

modification. Since standard CMOS processes use silicon nitride

as the top passivation layer, to enhance cell attachment to the

silicon nitride surface, we coated the active sensing area (the 64

sensing sites) with 12.5

gmL

2

fibronectin (F1141; Sigma) in

0.02% gelatin at 37

C and 5% CO

for 24 h.

31,32

The fibronectin

coating allows integrin receptors on the cell membrane to bind to

the surface, localizes the cells to the active sensing area, and

Fig. 1

Magnetic cell-based sensor. (a) Magnetic particles result in an increased effective inductance

eff

in the LC resonator, which causes a downshift

in the resonant frequency. This frequency shift can be detected by the CMOS circuits and serves as the readout for the sensing unit. (b) The normalized

transducer gain (sensitivity to magnetic particles) of the CMOS magnetic sensor is plotted with respect to the sensor inductor geometry. (c) A

spontaneous cardiac cell beating. Inset at top shows the cell potential changes (in milli-volts) during the periodic beating motion. SR = sarcoplasm

reticulum. (d) Autonomous beating of the cardiac progenitor cells leads to displacements of the magnetic particles and is detected by the CMOS

magnetic sensor as periodic shifts in resonant frequency. (e) The sensor chip contains 64 (8

8) independent sensing sites as a sensor array. A zoom-in

view shows the individual sensing site. (f) The sensor module fits in a petri dish. The module includes a PDMS reservoir to hold the cardiac progenitor

cells and the medium. (g) An image of the PDMS sample reservoir shows a cell cluster on the CMOS sensor surface. (h) Diagram for the CMOS

magnetic sensor chip microphotograph, chip system architecture, and the measurement setup.

4466

Lab Chip

, 2012,

, 4465–4471

This journal is

ß

The Royal Society of Chemistry 2012

Downloaded by California Institute of Technology on 13 December 2012

Published on 14 August 2012 on http://pubs.rsc.org | doi:10.1039/C2LC40392G

View Article Online

increases their adhesion to the sensor. Although the sensor can

be used with cells in suspension, surface attachment of cells to

the sensor area provides for a robust signal by making sure that

small movements of the entire assembly are not registered as cell

movements and that the cells are optimally positioned on the

sensor coils for maximum signal. We seeded the EBs onto the

sensor and incubated the entire assembly at 37

C and 5% CO

for ten days in differentiation medium lacking LIF. The medium

was replaced every 24 h. During this time, the EBs proliferated,

attached to the sensor surface, and formed micro-tissues

consisting of cardiac progenitors. The presence of cardiac

progenitors, which formed spontaneously in differentiation

medium, was visually verified.

Coating the cells with magnetic particles.

We coated the cardiac

progenitor cells with micron-sized magnetic particles (D =

2.4

m) to track their pulsatile movements. We used epoxy

functionalized magnetic beads (Dynabead M-270; Invitrogen

Inc). In order to enhance bead attachment to cells, the magnetic

beads were functionalized with fibronectin (F1141; Sigma) by

using an epoxy functionalization kit according to the manufac-

turer’s recommendations (Dynabead M-270; Invitrogen Inc).

After 30 min of attachment, excess beads were removed with

three sequential medium changes. Bead attachment to the

surface of the cells was observed visually by using a stereo

microscope; subsequent medium changes did not alter bead

positions. Visual inspection of the cells showed that on average

10 beads were attached to each cell.

CMOS magnetic sensor for real-time monitoring of cell beating

CMOS magnetic sensor and its implementation.

The core of the

CMOS magnetic sensor is an on-chip inductor-capacitor (LC)

resonator.

When magnetic particles are present, they increase

the total magnetic energy in the space, lead to a change in the

effective inductance in the resonator, and result in a resonant

frequency down-shift (Fig. 1a). The frequency-shift per magnetic

particle can be modelled as the sensor transducer gain and is

designed to present high spatial non-uniformity in our LC

sensing resonator (Fig. 1b). Changes in the cell shape or position

during the sensing process directly cause redistribution of the

attached particles. This spatial distribution leads to variation in

the total resonant frequency-shift and is readily detected by the

CMOS sensor in real-time. Thus, for the autonomous beating of

the cardiac cells in this study, the resulting magnetic sensor

output is a series of periodic pulses (Fig. 1c and 1d). To achieve a

sub-part-per-million (10

2

) relative frequency-shift sensitivity,

we implemented a low noise on-chip oscillator with Correlated-

Double-Counting (CDC) noise suppression technique for each

LC resonator as the read-out circuits on the CMOS chip.

The

sensor’s magnetic sensitivity is limited by the noise in the CMOS

electronics. The sensor is capable of detecting a single micron-

size magnetic bead;

therefore even with very small numbers of

beads on cells, the system is able to detect movement of the cells.

We implemented a magnetic sensor array with 64 independent

sensing sites by paralleling 16 quad-core sensor units each with

4 independent on-chip magnetic sensor sites

(Fig. 1e). Each

sensor site occupies an area of 120

m by 120

m, which is

conducive to cell- or tissue-level detection. By adding more

sensing sites, this architecture enables straightforward extension

to very-large-scale sensor arrays for high throughput applica-

tions, where massively paralleled testing can be used for

characterizing different cell types and testing different chemicals

simultaneously.

The sensor inductor optimization was achieved by using

Ansoft HFSS

V11 and Ansoft Maxwell

V10. The CMOS

sensor chip integrated circuits were designed using the Cadence

Design System as the Electronic Design Automation (EDA)

software. The system’s magnetic detection functionalities and the

digital programmability and real-time data readout were verified

by using simulators of SpectreRF

and Ultrasim

, respectively.

Finally, we customized the sensor chip for a standard 65 nm

CMOS process fabricated by United Microelectronics Corporation

(UMC).

Assembly of the sensing module.

We used a brass substrate as

the base of the sensor module to provide good mechanical

stability, low resistance thermal path, and electrical ground

reference of the sensor chip. The Printed-Circuit-Board (PCB)

was implemented using RT/Duroid

6010 laminate material

(Rogers Corporation) and was fabricated by DVH Circuit. We

completed the attachment of the CMOS sensor chip, the PCB

board, and the brass substrate by using silver epoxy (H20E,

Epoxy Technology, Inc), which was cured after a 2 h baking at

120

C. We then used gold wire-bonds to complete the sensor

chip packaging and electrical connections.

In order to hold the sensing cells and the test chemicals, we

constructed a polydimethylsiloxane (PDMS) reservoir on top of

the CMOS sensor chip (Fig. 1f and 1g). GE Silicones

RTV

615 kit, as the PDMS raw material, was obtained from Applied

Material Tech. We first mixed the PDMS base material and its

curing agent at a 10 : 1 ratio to form the sealing material for the

reservoir. Then, we constructed a cylindrical reservoir from a

short piece of plastic tubing (

= 0.5 cm). The plastic reservoir

was placed over the sensor chip and aligned with the sensing

area. We used the PDMS as adhesive around the reservoir and

thermally cured the entire assembly for one hour at 50

Preparation of the real-time data acquisition interface.

control the sensor chip operation and perform data acquisition,

we used an Altera

DE2 board with Cyclone

II 2C35 FPGA

core, which directly interfaced with the personal computer as the

measurement console. This setting provides fully reprogram-

mable digital control signals, which were sent to the CMOS

sensor chip to enable the operation of the desired magnetic

sensor sites. The corresponding measured data were directly

acquired also by the FPGA board and subsequently saved on the

personal computer (Fig. 1h). Customized Verilog

codes were

developed to operate the FPGA board.

Toxicity test of chemical agents

Two sets of chemical sensing experiments were performed to

demonstrate the detection functionality of the magnetic cell-

based sensor in this study. In the first set of experiments,

extracellular ion concentrations were varied independently to

test the corresponding beating frequency changes of the cardiac

progenitor cells. In the second experiment, lidocaine (Sigma) was

This journal is

ß

The Royal Society of Chemistry 2012

Lab Chip

, 2012,

, 4465–4471

4467

Downloaded by California Institute of Technology on 13 December 2012

Published on 14 August 2012 on http://pubs.rsc.org | doi:10.1039/C2LC40392G

View Article Online

used as a small molecule drug, whose effect on the cell beating

amplitude was monitored. The test chemicals were prepared in

the culture medium and directly introduced into the PDMS

chambers on the CMOS magnetic sensors. Before each chemical

was applied, the sensors registered the beating frequency or

amplitude of the cardiac cells under the normal physiological

condition for at least 5 min as the control measurements. These

data were later used as the baseline values for normalization.

Results and discussion

Real-time chemical detection

The magnetic beads were attached to the cardiac progenitor cells

as a one-step preparation procedure before the test chemicals

were introduced and remained bound throughout the measure-

ments. Since no further labelling is needed during the sensing

process for either the sensor or the test chemicals, in this respect

our sensor performs direct detection, similar to the previously

invented method of optical detection of cardiac toxicity on chip.

Since the CMOS magnetic sensor directly outputs the beating

frequency and amplitude of the cardiac cells, no post-processing

of data is required by our approach, which enables real-time

monitoring during detection. These features of the method

significantly simplify the sample preparation and sensing

procedures, and facilitate highly parallel and low-cost biochem-

ical screening.

Due to the autonomous beating of the cardiac progenitor cells,

the output of the magnetic sensor is a series of periodic pulses,

which we verified to match the beating of the cell cluster. The

signal amplitude of the sensor output is proportional to the

displacement in the pulsatile cell beatings, and the averaged

period between pulses determines the beating frequency of the

cell cluster. Since the cardiac cells beat with a frequency below

1 Hz in our experiments, we set the sensor sampling rate at 20 Hz

to satisfy the Nyquist minimum sampling rate and ensure

accurate amplitude peak recording for the pulsatile waveforms.

As a typical control measurement on the cardiac progenitor cells’

beating rate, we observed a frequency of 0.48 Hz under normal

physiological conditions,

i.e.

with the extracellular potassium ion

concentration of 3.4 mM (Fig. 2a and 2b).

In addition, we found that differentiation of the EBs directly

on the CMOS sensor surface was more effective in yielding

pulsatile cell clusters compared to seeding the clusters after off-

chip differentiation. Physical disruption in the latter approach,

e.g.

with cell scrapers, could abrogate the clusters’ synchronous

motions. We maintained the cells on the sensor surface for up to

15 days without loss of viability or significant morphological

changes. The sensor life-time can potentially be extended by

frequent medium changes and accurate temperature control,

which is important for in-field diagnostic sensors.

Toxicity test of extracellular potassium and calcium ions

Extracellular potassium and calcium ion concentrations have

well-known and significant effects on action potential generation

and thus on the autonomous beating activities of the cardiac

cells. We used different concentrations of these ions for proof-of-

concept testing of the magnetic cell-based sensor functionality.

We dissolved potassium and calcium in the culture media and

prepared the potassium test samples with concentrations of

3.4 mM (normal physiological concentration), 10 mM, and

20 mM and the calcium test samples with concentrations of

2 mM (normal physiological concentration), 4 mM, and 8 mM.

The test chemicals were delivered into the PDMS reservoir by

pipettes.

We first increased the potassium concentration from normal

physiological conditions (3.4 mM) to 10 mM and observed a

corresponding decrease in the recorded pulse frequency from

0.48 Hz to 0.19 Hz (Fig. 2a and 2c). This result verifies that the

recorded frequency shifts were indeed due to the pulsatile

motions of the cardiac progenitors, and is in agreement with

previously reported responses of spontaneously pulsating cardiac

cells.

33,34

Since no labelling step was required, the measurement

time was governed by the inherent response time of the cardiac

progenitor cells. We observed a three minute response time to a

shift in potassium ion concentration from 3.4 mM to 10 mM,

which shows fast sample-in-answer-out detection capability.

Increases in the extracellular potassium concentration from

3.4 mM to 10 mM and then to 20 mM resulted in monotonically

reduced pulse frequency until the cells stopped beating (Fig. 3a

and 3b). In addition, we increased the extracellular calcium

concentration from 2 mM to 4 mM and 8 mM, and measured

increases in the pulsatile frequency of the cells by 27% and 46%,

respectively (Fig. 3c and 3d). These observed trends were consistent

with previously reported responses of cardiomyocytes.

33,34

Finally, we tested the reversibility of the sensor to show that it

can be restored to its nominal state and reused. We first

increased the extracellular potassium concentration to 20 mM to

stop the autonomous cell movement, and were able to recover

the pulsatile motion by washing the sensor with the cell medium

to return the potassium concentration to its normal physiological

value of 3.4 mM (Fig. 4). In our measurement, two consecutive

washes were needed to adjust the potassium concentration. This

result suggests that the sensor can potentially be reused without

Fig. 2

Real-time sensor response when the extracellular potassium (K

)

level is increased. (a) The recorded sensor response and the correspond-

ing change in the extracellular potassium level. The averaged cell beating

rate stabilizes within 3 min after the K

level increase, which

demonstrates a fast response time of the cell-based sensor. (b) Zoom-in

plot of the stabilized sensor output in (a) at an extracellular K

level of

3.4 mM. (c) Zoom-in plot of the stabilized sensor output in (a) at an

extracellular K

level of 10 mM. (b) and (c) show the steady state sensor

outputs in (a).

4468

Lab Chip

, 2012,

, 4465–4471

This journal is

ß

The Royal Society of Chemistry 2012

Downloaded by California Institute of Technology on 13 December 2012

Published on 14 August 2012 on http://pubs.rsc.org | doi:10.1039/C2LC40392G

View Article Online

loss of functionality, and further shows the robustness and

potentially extended life-time of the sensor.

Small molecule detection

To test the functionality of the magnetic cell-based sensor in

detecting small molecules, we used lidocaine as a test chemical.

Lidocaine is a small molecule anti-arrhythmic drug that blocks

fast-gated sodium ion channels, impedes the cells from depolar-

izing, and decreases the amplitude of the pulsatile motions.

Due to its rapid onset of action, lidocaine is commonly used

intravenously as an amino amide-type local anesthetic for the

treatment of ventricular arrhythmias.

We first dissolved lidocaine in cell culture medium at

concentrations of 50 pM and 50 nM. The lidocaine test samples

were then delivered into the sensor PDMS reservoir through

pipettes. With increases in lidocaine concentration to 50 pM and

then to 50 nM, the recorded cell beating amplitudes were reduced

to 89% and 68% of the values observed without lidocaine (Fig. 3e

and 3f). These observations were consistent with the mechanistic

effect of lidocaine.

The real-time recorded sensor output data in Fig. 5 shows the

cell beating amplitudes with respect to time. We observed that

the beating amplitudes were initially increased after the lidocaine

samples were applied, potentially due to the temperature changes

in the cell medium. The beating amplitudes then gradually

slowed down and stabilized with increasing lidocaine concentra-

tion. The recorded response times were 10 min for 50 pM

lidocaine and 13.8 min for 50 nM lidocaine. These short response

times manifest the fast onset of the lidocaine toxicity effect and

also demonstrate the fast detection capability of the magnetic

cell-based sensor.

Fig. 3

Measurement of the magnetic cell-based sensor steady state outputs. (a) and (b) The real-time measured sensor output and the beating

frequency summary when the cell clusters are subjected to changes in extracellular potassium concentration. Beating rate decreases with an increas

ein

extracellular potassium level. (c) and (d) The real-time measured sensor output and the beating frequency summary when the cell clusters are subject

to extracellular calcium concentration changes. The cell pulsatile beating rate is accelerated when the extracellular calcium level is increased.

(e) and (f)

The real-time measured sensor output and the beating amplitude summary when lidocaine is added to the cell medium. The amplitude of the cell

beating is reduced at a higher extracellular lidocaine concentration. Error bars represent standard errors.

Fig. 4

Recovery test of the magnetic cell-based sensor. A real-time measured sensor output shows that the autonomous beating of the sensor cells can

be restored. Cell beating stops when the extracellular potassium level is raised to 20 mM, and is then recovered after two washes with the medium

containing 3.4 mM potassium ion. The steady state sensor outputs are highlighted in the boxed regions.

This journal is

ß

The Royal Society of Chemistry 2012

Lab Chip

, 2012,

, 4465–4471

4469

Downloaded by California Institute of Technology on 13 December 2012

Published on 14 August 2012 on http://pubs.rsc.org | doi:10.1039/C2LC40392G

View Article Online

Conclusions

We have demonstrated a magnetic cell-based sensor that

combines the advantages of a low-cost CMOS microelectronics

manufacturing process and a cell-based sensing scheme with cell-

or tissue-specific responses. Existing biochemical sensing mod-

alities are either incompatible with the CMOS manufacturing

processes or are limited due to their detection requirements. For

example, optical detection methods require visible changes in cell

morphology, can be limited by photo-bleaching, and are difficult

to integrate into compact system.

6,9,11

Electrochemical sensing is

sensitive to electrical charge variations at the electrode-electro-

lyte interfaces, and thus experience significant signal drift and

offset due to the measurement background.

18–20

In contrast,

magnetic sensing offers unique advantages as a non-optical and

non-contact detection scheme, avoiding all the aforementioned

issues.

21–27

Moreover, since most biological samples and media

are magnetically transparent, a low background noise level and

reduced off-target effects are obtained.

28,29

More importantly,

the CMOS magnetic sensor can be manufactured at low cost and

for large production volumes.

Depending on the application of interest, our magnetic cell-

based sensor can be outfitted with a variety of cell types with

diverse functionalities. Mouse ESCs are particularly suitable for

building high-throughput cell-based sensors in large volume,

since they can be readily expanded to large numbers and

differentiated to various cell types. The direct on-chip differ-

entiation of mouse ESCs substantially simplifies the required

preparation and handling steps, and further benefits mass

production of the magnetic cell-based sensor.

In this study, the physical movement of the cardiac progenitor

cells,

i.e.

pulsatile motion, was used as a robust detection signal;

this movement could also take other forms, such as cell

migration, for implementing similar cell-based sensing platforms.

Moreover, sensor cells can be genetically modified to increase

sensitivity and selectivity to certain chemical stimuli, thereby

enhancing the functionality of the cell-based sensor.

Since this magnetic cell-based sensor does not require labelling

during the measurement, it enables fast sample-in-answer-out

chemical detection. Measurement times are limited only by the

response time of the cells and can be as fast as three minutes as

shown here. The long sensor life-time and high level of robustness

should make these systems useful for in-field diagnostic applica-

tions. Cell proliferation may potentially be measured by this

system by tracking long-term changes in positions of magnetic

beads. Since cell proliferation occurs on time scales much longer

than those of pulsatile motions, the sensor makes accurate

measurements in the presence of cell proliferation. The most

important limitations of this sensor technology arise from the

need to maintain the sensing unit under physiological conditions.

Several implementation refinements can be made to the sensor

to achieve a complete lab-on-chip system. A microfluidic

structure can be easily included to standardize sample delivery

and medium exchange, and to provide the parallel detection

chambers required for high-throughput chemical screening.

26,36,37

Temperature regulators can be integrated into the CMOS circuit

to accurately control the local or global thermal environment on

the sensor,

26,38

and to extend the long-term viability of the sensor

cells with no need for any external heating devices.

In conclusion, our magnetic cell-based sensing platform

provides a compact solution for rapid and real-time detection

of biochemical agents. Based on this method, it is possible to

construct fully portable chemical sensors. This approach may

greatly reduce the cost of high-throughput screening systems in

drug development and also enable hand-held devices for point-

of-care (PoC) diagnostic applications, such as epidemic disease

control and environmental monitoring.

Acknowledgements

H. W. was supported by a California Institute of Technology

Innovation Initiative (CI2) Research Grant. A.M. was supported

by a National Science and Engineering Research Council of

Canada (NSERC) Scholarship, a post-graduate scholarship by

Caltech

Donna

and

Benjamin

Rosen

Center

for

Bioengineering and the NSF Center for the Science and

Engineering of Materials at Caltech (NSF DMR 0520565). The

authors would like to acknowledge Dr Shouhei Kousai for his

support on the CMOS magnetic sensor chip development;

Constantine Sideris for his help on developing the FPGA

Verilog programs; Alex Pai for his support on building the

sensor modules; and United Microelectronics Corporation

(UMC) for providing CMOS sensor chip foundry service.

References

1 S.V. Sharma, D.A. Haber and J. Settleman,

Nat. Rev. Cancer

, 2010,

, 241–253.

2 Y. Meng, A. Kasai, N. Hiramatsu, K. Hayakawa, K. Yamauchi, M.

Takeda, H. Kawachi, F. Shimizu, J. Yao and M. Kitamura1,

Lab.

Invest.

, 2005,

, 1429–1439.

3 P. Banerjee, D. Lenz, J. P. Robinson, J. L. Rickus and A. K. Bhunia,

Lab. Invest.

, 2007,

, 196–206.

4 T. H. Rider, M. S. Petrovick, F. E. Nargi, J. D. Harper, E. D.

Schwoebel, R. H. Mathews, D. J. Blanchard, L. T. Bortolin, A. M.

Young, J. Chen and M. A. Hollis,

Science

, 2003,

301

, 213–215.

Fig. 5

Real-time recorded sensor response when lidocaine is introduced.

A measured sensor output shows that the beating amplitude of the sensor

cells decreases when increasing the extracellular lidocaine level. The cell

beating amplitude at physiological condition is used as the baseline

reference. Medium containing a lidocaine concentration of 50 pM is first

added at

= 510 s. The sensor output stabilizes after 10 min with its

normalized amplitude decreases to 89% of the baseline value. Then,

medium containing 50 nM lidocaine is added at

= 1280 s. The sensor

output stabilizes after 14 min, and the normalized beating amplitude

decreases to 68% of the baseline value. The steady state sensor outputs

are highlighted in the boxed regions.

4470

Lab Chip

, 2012,

, 4465–4471

This journal is

ß

The Royal Society of Chemistry 2012

Downloaded by California Institute of Technology on 13 December 2012

Published on 14 August 2012 on http://pubs.rsc.org | doi:10.1039/C2LC40392G

View Article Online

5 J. J. Pancrazio, J. P. Whelan, D. A. Borkholder, W. Ma and D. A.

Stenger,

Ann. Biomed. Eng.

, 1999,

, 697–711.

6 S. B. Kim, H. Bae, J. M. Cha, S. J. Moon, M. R. Dokmeci, D. M.

Cropek and A. Khademhosseini,

Lab Chip

, 2011,

, 1801–1807.

7 T. Y. Chang, C. Pardo-Martin, A. Allalou, C. Wa

̈ hlby and M. F.

Yanik,

Lab Chip

, 2012,

, 711–716.

8 C. Pardo-Martin, T. Y. Chang, B. K. Koo, C. L. Gilleland, S. C.

Wasserman and M. F. Yanik,

Nat. Methods

, 2010,

, 634–636.

9 A. El Gamal and H. Eltoukhy,

IEEE Circuits Devices Mag.

, 2005,

6–20.

10 O. Graydon,

Nat. Photonics

, 2010,

, 668.

11 T. Huang, S. Sorgenfrei, K. L. Shepard, P. Gong and R. Levicky,

IEEE J. Solid-State Circuits

, 2009,

, 1644–1654.

12 K. K. Ghosh,

Nat. Methods

, 2011,

, 871–878.

13 D. Moser and H. Bales,

Sens. Actuators, A

, 1993,

37-38

, 33–34.

14 P-Y. Wang and M. S. Lu,

IEEE Sens. J.

, 2011,

, 3469–3475.

15 E. Forsen, G. Abadal, S. Ghatnekar-Nilsson, J. Teva, J. Verd, R.

Sandberg, W. Svendsen, F. Perez-Murano, J. Esteve, E. Figueras, F.

Campabadal, L. Montelius, N. Barniol and A. Boisen,

Appl. Phys.

Lett.

, 2005,

, 043507.

16 M. L. Johnston, I. Kymissis and K. L. Shepard,

IEEE Sens. J.

, 2010,

, 1042–1047.

17 C. Hagleitner, A. Hierlemann, D. Lange, A. Kummer, N. Kerness, O.

Brand and H. Baltes,

Nature

, 2011,

414

, 293–296.

18 P. M. Levine, P. Gong, R. Levicky and K. L. Shepard,

IEEE J. Solid-

State Circuits

, 2008,

, 1859–1871.

19 S. Joo and R. B. Brown,

Chem. Rev.

, 2008,

108

, 638–651.

20 M. Schienle, C. Paulus, A. Frey, F. Hofmann, B. Holzapfl, P.

Schindler-Bauer and R. Thewes,

IEEE J. Solid-State Circuits

, 2004,

, 2438–2445.

21 G. Li, V. Joshi, R. L. White, S. X. Wang, J. T. Kemp, C. Webb,

R. W. Davis and S. Sun,

J. Appl. Phys.

, 2003,

, 7557–7559.

22 P. Besse, G. Boero, M. Demierre, V. Pott and R. Popovic,

Appl.

Phys. Lett.

, 2002,

, 4199–4201.

23 T. Aytur, J. Foley, M. Anwar, B. Boser, E. Harris and P. R. Beatty,

J. Immunol. Methods

, 2006,

314

, 21–29.

24 P. Liu, K. Skucha, M. Megens and B. Boser,

IEEE Trans. Magn.

2011,

, 3449–3451.

25 H. Lee, E. Sun, D. Ham and R. Weissleder,

Nat. Med.

, 2008,

869–874.

26 H. Wang, Y. Chen, A. Hassibi, A. Scherer and A. Hajimiri,

IEEE

ISSCC Dig. Tech.

, 2009, 438–439.

27 H. Wang, S. Kosai, S. Sideris and A. Hajimiri,

IEEE MTT-S Int.

Microwave Symp. Dig.

, 2010, 616–619.

28 R. Matijevi,

Fine Particles in Medicine and Pharmacy

, Springer, 2011.

29 J. Llandro, J. J. Palfreyman, A. Ionescu and C. H. W. Barnes,

Med.

Biol. Eng. Comput.

, 2010,

, 977–998.

30 K. R. Boheler, J. Czyz, D. Tweedie, H.-T. Yang, S. V. Anisimov and

A. M. Wobus,

Circ. Res.

, 2002,

, 189–201.

31 A. Mahdavi, R. E. Davey, P. Bhola, T. Yin and P. W. Zandstra,

PLoS Comput. Biol.

, 2007,

, 1257–1267.

32 J. M. Karp, J. Yeh, G. Eng, J. Fukuda, J. Blumling, K.-Y. Suh,

J. Cheng, A. Mahdavi, J. Borenstein, R. Langer and A.

Khademhosseini,

Lab Chip

, 2007,

, 786–794.

33 R. E. Klabunde,

Cardiovascular Physiology Concepts

, Lippincott

Williams & Wilkins, 2005.

34 D. Mohrman and L. Heller,

Cardiovascular Physiology

, McGraw-Hill

Professional, 2010.

35 R. Gianelly, J. O. von der Groeben, A. P. Spivack and D. C.

Harrison,

N. Engl. J. Med.

, 1967,

277

, 1215–1219.

36 J. B. Christen and A. G. Andreou,

IEEE Transactions on Biomedical

Circuits and Systems

, 2007,

, 3–18.

37 J. El-Ali, P. K. Sorger and K. F. Jensen,

Nature

, 2006,

442

, 403–411.

38 E. Lauwers, J. Suls, W. Gumbrecht, D. Maes, G. Gielen and W.

Sansen,

IEEE J. Solid-State Circuits

, 2001,

, 2030–2038.

This journal is

ß

The Royal Society of Chemistry 2012

Lab Chip

, 2012,

, 4465–4471

4471

Downloaded by California Institute of Technology on 13 December 2012

Published on 14 August 2012 on http://pubs.rsc.org | doi:10.1039/C2LC40392G

View Article Online