Biosensors and Bioelectronics 26 (2010) 1565–1570

Contents lists available at ScienceDirect

Biosensors and Bioelectronics

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

Integration of microfluidic and cantilever technology for biosensing application

in liquid environment夽
Carlo Ricciardi a,∗ , Giancarlo Canavese b , Riccardo Castagna a , Ivan Ferrante a ,
Alessandro Ricci a , Simone Luigi Marasso a , Lucia Napione c , Federico Bussolino c
a
LATEMAR – Politecnico di Torino, Dipartimento di Scienza dei Materiali ed Ingegneria Chimica, Corso Duca degli Abruzzi 24, I-10129 Torino, Italy
b
IIT – Italian Institute of Technology @ POLITO Center for Human Space Robotics, Corso Trieste 21, 10129 Torino, Italy
c
Department of Oncological Sciences, Institute for Cancer Research and Treatment, University of Torino, 10060 Candiolo (TO), Italy

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

Article history: Microcantilever based oscillators have shown the possibility of highly sensitive label-free detection by
Received 14 April 2010 allowing the transduction of a target mass into a resonant frequency shift. Most of such measurements
Received in revised form 9 July 2010 were performed in air or vacuum environment, since immersion in liquid dramatically deteriorates the
Accepted 29 July 2010
mechanical response of the sensor. Besides, the integration of microcantilever detection in a microfluidic
Available online 5 August 2010
platform appears a highly performing technological solution to exploit real time monitoring of biomolec-
ular interactions, while limiting sample handling and promoting portability and automation of routine
Keywords:
diagnostic tests (Point-Of-Care devices). In the present paper, we report on the realization and opti-
Lab-on-Chip
Microcantilever
mization of a microcantilever-based Lab-on-Chip, showing that microplates rather than microbeams
Real-time exhibit largest mass sensitivity in liquid, while pirex rather than polymers represents the best choice for
Protein detection microfluidic channels. Maximum Q factor achieved was 140 (for fifth resonance mode of Pirex prototype),
Angiogenesis as our knowledge the highest value reported in literature for cantilever biosensors resonating in liquid
environment without electronic feedback. Then, we proved the successfully detection of Angiopoietin-1
(a putative marker in tumor progression), showing that the related frequency shifts coming from non-
specific interactions (negative controls) are roughly one order of magnitude lower than typical variations
due to specific protein binding. Furthermore, we monitored the formation of antibody–antigen com-
plex on MC surface in real-time. The proposed tool could be extremely useful for the comprehension of
complex biological systems such as angiogenic machinery and cancer progression.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction nanoscale have recently shown evident performance limits in

Resonance operation method, which aims to quantify the
adsorbed target mass thanks to the oscillator frequency shift, seems
to be the most successful application of microcantilever (MC) based
biosensor. Despite of static operation method (where surface stress
generated in binding of the target molecules to the receptors on one
MC side cause the beam to deflect), this technique is less affected
by the thermal drift of beam deflection and stabilization problems
(Shen et al., 2001; Lochon et al., 2006).
The progress of microcantilever technology and the need for
increasing device sensitivity have favored the reduction of sen-
sor dimensions up to the nanoscale. It has been shown that
nanocantilevers are able to detect few biomolecules or single
viruses (Ilic et al., 2004; Gupta et al., 2004), thus displaying
very high mass sensitivity. On the other hand, biosensors at the
夽 Finalist for the World Congress on Biosensors 2010 selected Keynote Paper.
∗ Corresponding author. Tel.: +39 011 0907383.
E-mail address: carlo.ricciardi@polito.it (C. Ricciardi).
terms of analyte density, response time (Nair and Alam, 2006)
and statistical variability (Gupta et al., 2006) due to their intrinsic
diffusion-limited regime. Furthermore, such highly sensitive can
be achieved in air or vacuum environment, since immersion in
liquid would dramatically deteriorate the response of the sensor.
Indeed, minimum detectable mass can be defined as mmin ∝ (m/Q)
(Waggoner and Craighead, 2007), where m and Q are the oscilla-
tor mass and quality factor, respectively. The latter is defined as
Q = fr (f−3 dB )−1 , where fr is the resonance frequency and f−3 dB
is the width of the resonance curve at −3 dB from the maxi-
mum.
Operation in liquid in fact would be desirable in order to retain
biomolecule physiological structure and function, since it has been
shown that proteins and cell membrane can change their con-
formation passing from liquid environment to vacuum condition
(Sharma and Kalonia, 2004). Liquid environment is then funda-
mental in those applications where the observation of binding
and unbinding kinetics are needed as well as for in situ and real-
time measurement. An on-line measurement in liquid can also
reduce false positive and false negative responses, which are an

0956-5663/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.bios.2010.07.114
1566 C. Ricciardi et al. / Biosensors and Bioelectronics 26 (2010) 1565–1570
important drawback in MC-based biosensing (Waggoner and
Craighead, 2007). Moreover, operation in liquid assures the reduc-
tion of the risk of breaking the MCs in transporting them between
different environments as well as during the drying procedure.
Finally, in vacuum conditions, residual salt precipitated from the
buffer solution, employed in cleaning or functionalization steps,
could remain on cantilever surfaces thus invalidating target mass
estimation (Liu et al., 2009).
Recent literature contains some interesting studies about can-
tilevers vibrating in liquid performed under static fluid conditions
(Campbell and Mutharasan, 2005a; Park et al., 2005). Campbell and
Mutharasan (2005b) performed measurements using millimeter-
sized cantilever partially immersed in a static liquid; this approach
has the disadvantage to register a decrease of volume during the
experiment due to the evaporation and consequently it requires a
more complex calibration. It is worth to note that almost all the
experiments concerning cantilevers immersed in liquid are carried
out in commercial or home-made fluid cell (Arntz et al., 2003; Braun
et al., 2005; Ghatkesar et al., 2008; Moulin et al., 2000; Vančura et
al., 2005) with volume from tens of microliters to 2 mL, in which
the device is placed and sealed through o-ring or rubber membrane.
Maraldo et al. (2007a, 2007b) and Campbell et al. (2007) obtained
great results in biomolecules mass detection at different concentra-
tions by piezoelectric-excited millimeter-sized cantilever (PEMC).
Since the cantilevers are made individually and manually, a low
reproducibility and repeatability in fabrication process is obtained.
In addition their design is difficult to be integrated in standard low
cost production both for what concerns the materials and the pro-
cesses. The large dimension of the PEMC forces to design a wide
housing flow cell and the fabrication processes are not easy to
integrate in Lab-on-Chip (LOC) technology.
Very few works report a fluidic circuit integrated on microcan-
tilever sensor device (Park et al., 2008; Thaysen et al., 2001). Aubin
et al. (2005) presented a design of integration between resonator
sensor array and microfluidic channel using micromachining pro-
cesses. Since the dimensions of their resonators are about 10 ␮m,
during the measurements, the microfluidic channels hosting the
sensors are pumped down to a pressure where damping effects
become negligible. This last solution appears to be expensive and
does not promote integration and device portability.
It is clear, therefore, that MC biosensors would benefit in
case of integration with microfluidic technology; advantages
such as reduction of volume reagents and time assay, in situ
measurements and high level of automation are evident. More-
over, diagnostic biosensing protocols, as well as single-molecule
detection, could express their practical potentiality only with
microfluidic assembly, where total fluidic volumes is minimized,
recirculation procedure can be created to achieve enhanced
binding probabilities, and the environment to which the sensor is
exposed can be actively controlled.
In this paper, we report on the realization and optimization
of a MC-based LOC to perform in liquid real-time detection of
specific proteins. We chose Angiopoietin-1 (Ang-1) as our target
molecule representative of the wide range of angiogenic factors,
whose expression level is largely investigated in different tumors.
Angiopoietin-1 and -2 are oligomeric-secreted glycoprotein ligands
of the receptor tyrosine kinase Tie-2 (Davis et al., 1996). Even if
it is well established that the Ang–Tie-2 pathway is involved in
tumor angiogenesis, the exact effects of angiopoietins on tumor
angiogenesis are under debate (Metheny-Barlow and Li, 2003).
Nevertheless experimental and clinical studies have demonstrated
that increased expression of Ang-1 and -2 promotes or inhibits
tumor angiogenesis (Yu, 2005), suggesting that Ang-1 is a pro-
angiogenic factor that promotes endothelial cell survival and tumor
angiogenesis, especially in the presence of vascular endothelial
growth factor.
Four main objectives can be identified in the following work: (I)
optimization of sensor geometry (in particular, MC aspect ratio); (II)
optimization of microfluidic platform (in particular, best choice for
materials of microchannels and interconnections); (III) successfully
in liquid detection of Ang-1 (with respect to negative controls); (IV)
observation of antibody–antigen real-time kinetics.
2. Materials and methods
2.1. Reagents
3-Aminopropyltriehoxysilane (APTES, anhydrous, 99% Aldrich),
glutaraldehyde (GA, 25%, v/v, water solution) and toluene
(anhydrous, 99.8% Aldrich) were used without any further purifi-
cation. Sulphuric acid (95–97%, w/w) and hydrogen peroxide
(30%, w/w) were also purchased from Sigma–Aldrich. Ortho-
boric acid and sodium chloride used to prepare borate buffer
were ACS reagents (assay ≥ 99.5%) and were obtained from
Sigma–Aldrich. Recombinant Protein G, purified from Streptococ-
cus, was from PIERCE. Ang-1 was obtained from R&D Systems,
while the anti-Ang-1 (A0604) mAb was from SIGMA. Dulbecco’s
Phosphate-Buffered Saline (PBS; BE17-512F) was procured from
BioWhittaker/Cambrex.
2.2. LOC fabrication
The cantilever chips were fabricated starting from Silicon On
Insulator (SOI) wafers with the following layers and thicknesses:
polished silicon device layer of 7 ± 0.5 ␮m, silicon handle layer of
450 ± 5 ␮m, buried and backside oxide layers 1.5 ± 0.1 ␮m. The pro-
cess flow is composed of the following steps: photolithography on
the back side, wet etching in Buffered Oxide Etch (BOE) solution for
the patterning of the mask layer on the back side, front side pro-
tection with a polymeric coating (Protek B2 from Brewer Science),
wet etching in KOH solution, sample cleaning, photolithography
on the front side of the membrane, Reactive Ion Etching (RIE) of
silicon, removal of the buried oxide layer in BOE and cleaning
by piranha solution (Canavese et al., 2007). The complete process
flow, together with additional MC fabrication details, is reported in
Supplementary material.
Three different materials (SU-8, PDMS, Pirex) and relative
process fabrication were chosen for three different microfluidic
prototypes of LOC for the integration with microcantilever sen-
sors. All layouts have at least two wells (roughly 2 ␮L in volume,
each), in order to perform different biodetection experiments
simultaneously. Polydimethylsiloxane (PDMS) was selected for the
fabrication of the microfluidic interconnection for its peculiar char-
acteristics as biocompatibility, transparency, cost-efficiency, etc.
A particular interconnection layout requiring no external clamp-
ing or gluing was studied. For all the prototypes were adopted the
same PDMS microfluidic reversible interconnections designed and
fabricated as described in a previous study (Quaglio et al., 2008).
A picture and a 3D sketch of the device, together with impor-
tant parts of the experimental set-up are reported in Fig. 1a and b,
respectively.
Processes flow description as well as fabrication details are
reported in Supplementary material.
2.3. Experimental set-up and data analysis
Cantilever vibrational characteristics were measured using an
apparatus in which a function generator (HP 33120A) produced
a sinusoidal signal that was amplified and sent to a piezoelectric
actuator (PI Ceramic). Cantilever resonance curves were monitored
using an optical lever read-out (laser diode and PSD by Hama-
matsu): the current output of the PSD was amplified and converted
 C. Ricciardi et al. / Biosensors and Bioelectronics 26 (2010) 1565–1570
Fig. 1. Microcantilever-based LOC: (a) picture and (b) 3D sketch. Most important
parts of the device are labeled as: (1) PDMS interconnections; (2) cantilever chip;
(3) microfluidic platform (Pirex, SU-8 or PDMS); (4) piezo disk; (5) Peltier cell; (6)
heat sink.
into a voltage output, sent to a lock-in amplifier (EG&G 7260) for
signal extraction and filtering, and stored to a personal computer.
To properly mechanically lock the chip to the piezo disc and to
improve the heat transfer between Peltier element (PF-127-14-
15, 40 mm × 40 mm × 3.9 mm, 58 W, SuperCool) and chip, thermal
glue was employed (Thermal Bonding System by Electrolube Der-
byshire, UK). The temperature was set by a temperature controller
(LDT-5948 by ILX Lightwave) and the feedback signal necessary
to regulate the temperature was obtained interfacing the Propor-
tional Integral Derivative (PID) controller to a thermistor placed in
contact to the upper surface of the piezo element: PID coefficients
were obtained using the AutoTune function of the temperature con-
troller. The temperature was stabilized in the range of 0.1 ◦ C. Liquids
are injected inside the LOC by means of a syringe pump (by Harvard
Apparatus) that allows for a fine setting of pumping rate.
The measurement procedure as well as the fitting of data with a
Lorentzian curve were performed by a software in LABVIEW® envi-
ronment. The resonant frequency f and the quality factor Q of each
single curve were extracted from fitting the measured amplitude
of vibration as function of the excitation frequency, A(f) as follows:
 tors and resonance frequencies, in the case in which the magnitude
Amax 4Q 2 − 1
A(f ) = b1 + b2 f +
2Q 2 2 2
(1 − (f 2 /fr2 )) + (f /Qfr )
where Amax is the maximum amplitude of vibration and b1 and b2
are baseline fitting coefficients.
Since we used MC with slightly different geometrical dimen-
sions (due to inherent tolerances in the fabrication process) and
three resonance modes are usually detected, bioexperiments are
compared in terms of relative frequency deviation f/f rather than
absolute frequency shift. For each MC, we calculate the arithmetic
mean of relative frequency deviation f/f over the modes and the
uncertainty as half-deviation of the modes (using standard devia-
1567
tion for few data can lead to an underestimation of the uncertainty).
Finally, we use the “weighted average” method to have the best esti-
mation of the true value f/f of all the MCs, as reported elsewhere
(Ricciardi et al., 2010).
2.4. MC functionalization and binding assays
Silicon MCs were soaked into piranha solution (70% H2 SO4 :30%
H2 O2 ) for 15 min, rinsed with doubly distilled water and dried in
a stream of nitrogen; after that, a silicon oxide film was grown
on them by thermal oxidation at 1100 ◦ C in O2 atmosphere for
3 h. Freshly cleaned MC were then incubated with 1% (v/v) solu-
tion of APTES in toluene at 60 ◦ C for 10 min. Silane-coated MCs
were rinsed several times with toluene and dried with nitrogen.
Freshly silanized-MCs were then incubated in a 1% GA solution
(in borate buffer 0.1 M, pH 8.8) for 1 h using an orbital shaker at
100 rpm. After 15 min of incubation, 300 ␮L of a solution of sodi-
umcyanoboro hydride (5 M) in NaOH have been added to reduce
the imine (–N CH–) formed from the reaction between –NH2 and
–CHO moieties. After incubation, MCs have been rinsed several
times with doubly distilled water and dried in a stream of nitro-
gen (Ricciardi et al., 2010). Then, 60 ␮L of protein G solution (PtG,
0.05 mg/mL) in PBS was pumped at 0.5 ␮L/min through microfluidic
channels on functionalized cantilever for 2 h at room tempera-
ture. Afterward, Tween 20 (0.05% in PBS) was pumped for 30 min
to remove non-specifically bound proteins. After protein G bind-
ing, 30 ␮L of anti-Ang-1 mAb (A0604) solution in PBS (0.05 mg/mL)
were pumped on cantilever surface and the incubation performed
for 2 h at room temperature. After this, channels and cantilevers
were washed with PBS. Antigen recognition was carried out by
flowing 30 ␮L of Ang-1 solution in PBS (25 ␮g/mL) at controlled
temperature 23 ± 0.2 ◦ C at a flow rate of 0.5 ␮L/min.
3. Theory
According to Euler–Bernoulli theory, the nth mode resonance
frequency of a rectangular cantilever beam in vacuum fnv is given
by:

2 E t
fnv = n (2)
2 12 l2
where E and  are respectively the elastic modulus and the
density of the cantilever material, while t and l are the beam
thickness and length. The constants n are dimensionless parame-
ters depending on boundary conditions. For rectangular cantilever
beam, the numerical values are: 1 = 1.875, 2 = 4.694, 3 = 7.85, . . .,
n = (n − 0.5) when n > 5 (Lochon et al., 2005).
When such a structure oscillates in a viscous fluid, inertial and
viscous forces act against its motion so that resonant frequency
and quality factor Q are significantly affected. In the case of an
unbounded fluid environment, an analytical model, holding for low
mode numbers, is available for estimating the fluid induced Q fac-
of the dissipative effects is small, i.e. Q  1 (Sader, 1998). Defining
(1)
f ,  and w respectively the fluid density, the fluid viscosity and
the beam width, the nth mode resonance frequency in fluid can be
written as (Sader, 1998):
fnv
f
fn =  (3)
1 + (f /4t)(c1 ı + c2 w)
being:

2
ı= (4)
f
2f fn
1568 C. Ricciardi et al. / Biosensors and Bioelectronics 26 (2010) 1565–1570
Fig. 2. Histogram of Q factor of the first three normal modes of microcantilevers
with different aspect ratios (3-2-1.5) resonating in water.
The parameter ı represents the unsteady viscous layer thickness,
i.e. thickness of the fluid layer, sticking to the cantilever surfaces,
where viscous forces are significant. In the frequency range of
1–200 kHz, values of viscous layer in water environment are in the
range of 1–10 ␮m.
For what concerns mode Q factors, one has:
(4wt/f ) + c1 ıw + c2 w2
Qn =
c3 ı2 + c4 ıw
Values of coefficients c1–4 can be found in Maali et al. (2005) as:
c1 = 3.7997, c2 = 1.0553, c3 = 2.7374, c4 = 3.8018
Since ı only slightly depends on w (as one can verify by simply iter-
atively solving Eqs. (3) and (4) for a given value of fnv ), a nearly linear
increase of Q with w is obtained from Eq. (5). Therefore, wider struc-
tures are expected to exhibit narrower curves (and thus enhanced
mass detection limits) in fluid, a behavior that commonly seems
a little surprising and counterintuitive. In other words, this means
that a reduction of microcantilever planar aspect ratio (AR) = l/w,
while keeping constant l and t, is in favor of an increase of Q, a fact
that also finds confirmation in experiments (Vančura et al., 2008).
It is worth to point out anyway that the above treatment is
strictly valid for beams, i.e. structures for which the condition
l  w  t holds (typical values are l ≈ 10w ≈ 100t). Therefore, in
case of low AR structures such as cantilever microplates (i.e. when
w∼l) an experimental validation is still needed.
4. Results and discussion
4.1. Design optimization
First objective of the work is the optimization of device geom-
etry, in particular MC aspect ratio (AR). According to above
guidelines, we designed and realized rather wide microstruc-
tures to achieve relative high Q resonators. While thickness and
length were respectively fixed to 6 ␮m and 900 ␮m, three val-
ues of MC width were tested: 300 ␮m (i.e. AR = 3), 450 ␮m (i.e.
AR = 2), 600 ␮m (i.e. AR = 1.5). Experimentally, first five flexural res-
onance modes of such structures when vibrating in water are in the
following ranges: m1 ∼ = 1–2 kHz, m2 ∼= 8–12 kHz, m3 ∼= 24–35 kHz,
m4 ∼= 50–75 kHz, m5 ∼ = 90–150 kHz (differences in resonance fre-
quencies are due to unavoidable fabrication tolerances and SOI
substrates thickness uncertainties). Fig. 2 shows the experimental
Fig. 3. Histogram of Q factor of the first five normal modes of microcantilevers
(AR = 1.5) with different microfluidic channel materials: Pirex, SU-8, PDMS.
quality factors (extrapolated from fit introduced in Eq. (1)) of the
first three normal modes (m1 , m2 and m3 ) in water environment as
a function of the three different aspect ratios. It arises that Q factor
can be easily tuned both employing higher modes and reducing the
cantilever aspect ratio, from a minimum of 5 to a maximum of 24,
a value comparable with most of similar literature works. There-
(5) fore, although being quite counterintuitive, minimum detectable
mass of microplates (structures with l∼w) increments with oscil-
lator width, as theoretically demonstrated for beams (structures
(6) with l  w) in Section 3. On the other hand, the increment of Q
with higher modes is simply due to the reduction of viscous layer
(Eq. (4)), and thus dumping effects, with frequency increment.
The second objective of the work mainly concerns with find-
ing the best material to be employed for the microfluidic platform.
Such material should generally be compatible with biological pro-
tocols and LOC technology, but, in particular, it should also exhibit
suitable mechanical and vibrational characteristics. Due to our
peculiar design, the microfluidic platform is also responsible for
the propagation of mechanical vibration from the piezo actuator
to MC (Fig. 1b). Therefore, we fabricated three different prototypes
in order to investigate the influence on the sensor performance
of the following materials: Pirex, SU-8, PDMS (for details con-
cerning device fabrication, please refer to Supplementary text
and figures). The Si MCs have nominally the same dimensions
(900 ␮m × 600 ␮m × 6 ␮m), except for unavoidable fabrication tol-
erances and SOI substrates thickness uncertainties. Fig. 3 shows
the experimental quality factors (extrapolated from fit introduced
in Eq. (1)) of the first five normal modes (m1 , m2 , m3 , m4 and m5 )
in water environment for the three prototypes. It can be clearly
seen that the presence of a soft polymeric layer (SU-8 or PDMS)
between the piezo actuator and the cantilever seriously affects
the oscillator response. While at low frequency (first three modes,
f ≤ 35 kHz) Q factors are similar, at high frequencies (fourth and fifth
mode, f ≥ 75 kHz) the damping effects of the polymeric platform are
so relevant that the resonance peaks completely flatten for PDMS
chips, drastically weaken for SU-8 chips (which has a larger stiff-
ness respect to PDMS). Raw data and fittings related to Fig. 3 are
reported in Supplementary Material.
Maximum Q factor achieved was 140 (for m5 of Pirex proto-
type), as our knowledge the highest value reported in literature for
MC biosensors resonating in liquid environment without electronic
feedback.
 C. Ricciardi et al. / Biosensors and Bioelectronics 26 (2010) 1565–1570 1569

Table 1
Measured frequency shifts and relative frequency changes for Ang-1 at 25 ␮g/mL,
negative control 1 and 2. See text for further details.

Chip Bioexperiment Relative frequency

change, f/f (×10−2 )

FC1 Ang-1 25 ␮g/mL −1.12 ± 0.72

Negative control 1 0.17 ± 0.20

FC2 Ang-1 25 ␮g/mL −1.52 ± 0.94

Negative control 2 −0.10 ± 1.05

FC3 Ang-1 25 ␮g/mL −2.80 ± 1.54

Negative control 2 0.47 ± 0.53

FC4 Ang-1 25 ␮g/mL −6.10 ± 3.77

N/A N/A

Fig. 4. Histogram of average relative frequency deviation for Ang-1 specific binding,
negative control 1 (PBS on Ab-coated MC) and negative control 2 (Ang-1 on Ab-
uncoated MC).

4.2. Biosensing experiments

The high quality factor in a viscous liquid environment obtained

thanks to the integration of new cantilever design with microflu-
idic technology and the use of proper materials entail to perform in
situ real time monitoring of bio-molecular interactions by charac-
terizing resonance frequency shift induced by the specific binding
occurred on the Ab-immobilized cantilever surface. Since highest
Q values were obtained with the pirex-silicon chip, the third device
design has been chosen for biosensing experiments of Ang-1 detec-
tion. MC dimensions were increased to 1200 ␮m × 800 ␮m × 7 ␮m,
while keeping AR = 1.5, to further enhance resonator quality fac-
tor. As reported in Section 2, the biodesign is composed by 2
functionalization steps (APTES and GA), and 3 biomolecule bind-
ing steps (Protein G, anti-Ang-1 Ab, Ang-1); previous experiments
in vacuum environment have shown that such biodesign exhibits
optimal specificity and fine precision, with a ratio of Ang-1/Ab sur-
face density of 2.06 ± 0.05 (Ricciardi et al., 2010). We fabricated
and functionalized five chips (FC1-5) that present two indepen-
dent cantilevers and wells on the same LOC platform (Fig. 1): the
first resonator is used for the Ang-1 specific recognition, while the
second is used as a negative control. We considered two type of neg-
ative control (also referred as “blank”): (1) PBS solution without the
target analyte flowing on Ab-coated MC, and (2) Ang-1 solution in
PBS (25 ␮g/mL) flowing on Ab-uncoated MC (coated with Protein
G). Such negative controls are fundamental to determine the influ-
ence of non-specific adsorption/desorption on resonance curves,
therefore setting up the experimental limit of detection of the sys-
tem. All the solutions were delivered at a flow rate of 0.5 ␮L/min at
controlled temperature 23 ± 0.2 ◦ C.
Since we used MC with slightly different geometrical dimen-
sions (due to inherent tolerances in the fabrication process) and
three resonance modes are usually detected, bioexperiments are
compared in terms of relative frequency deviation f/f rather than
absolute frequency shift. For each MC, we calculate the arith-
metic mean of relative frequency deviation f/f over the modes
and the uncertainty as half-deviation of the modes. Finally, the
“weighted average” method is applied to have the best estimation
of the true value f /f of all the MCs (Ricciardi et al., 2010). Fig. 4
shows the histograms of average relative frequency deviation for
Ang-1 specific binding, negative control 1 (PBS on Ab-coated MC)
and negative control 2 (Ang-1 on Ab-uncoated MC), while data
for each MC are reported in Table 1. The calculated values are,
respectively: (f /f )Ang-1 = (−1.59 ± 0.52) × 10−2 , (f /f )blank1 =
Fig. 5. Real-time monitoring of antigen–antibody hybridization compared to nega-
tive control experiment (PBS on Ab-coated MC).
(−0.17 ± 0.20) × 10−2 , (f /f )blank2 = (−0.36 ± 0.470) × 10−2 . As
can be noticed, the average relative frequency shift of negative con-
trol experiments resulted roughly one order of magnitude lower
than successfully Ang-1 detection. As could be expected the non-
specific protein–protein interaction between Ang-1 and protein G
coated MC (negative control 2, Ang-1 on Ab-uncoated MC) is a
larger interferon respect to standard blank experiments (negative
control 1, PBS on Ab-coated MC). Calculating the ratio of Ang-1/Ab
surface density, we obtain 5.2 ± 1.4, a value characterized by a con-
siderably larger relative uncertainty, as expected for measurements
in liquid respect to previously reported results in vacuum.
Last fabricated and functionalized chip (FC5) was used for the
real-time monitoring of antibody–antigen biomolecular interac-
tions, as reported in Fig. 5. After few minutes, probably needed
for the antigen to reach a minimal detectable concentration on
MC surface, the (normalized) resonant frequency of the Ab-coated
MC clearly decreases and quickly reaches saturation. On the con-
trary, the signal coming from negative control (PBS on Ab-coated
MC) just exhibits negligible fluctuations. Similar measurements
at different concentrations are planned to deeply investigate the
Ab/Ang-1 binding model and kinetics, as well as to better compare
the stoichiometry of such interaction in liquid environment with
previously reported results in vacuum.
5. Conclusions
We report on the realization and optimization of a
microcantilever-based LOC, showing that microplates rather
1570 C. Ricciardi et al. / Biosensors and Bioelectronics 26 (2010) 1565–1570
than microbeams exhibit largest mass sensitivity in liquid, while
pirex rather than polymers represents the best choice for microflu-
idic channels. Maximum Q factor achieved was 140 (for fifth
resonance mode of Pirex prototype), as our knowledge the highest
value reported in literature for cantilever biosensors resonating in
liquid environment without electronic feedback. Then, we proved
the successfully detection of Angiopoietin-1 (a putative marker
in tumor progression), showing that the related frequency shifts
coming from non-specific interactions (negative controls) are
roughly one order of magnitude lower than typical variations due
to specific protein binding. Furthermore, we monitored the forma-
tion of antibody–antigen complex on MC surface in real-time. The
proposed tool could be extremely useful for the comprehension
of complex biological systems such as angiogenic machinery and
cancer progression.
Acknowledgements
The authors are thankful for the financial support of MIUR
(FIRB2003-LATEMAR and PRIN2007 grant) and Regione Piemonte
(Converging Technologies, NAMATECH grant).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bios.2010.07.114.
References
Arntz, Y., Seelig, J.D., Lang, H.P., Zhang, J., Hunziker, P., Ramseyer, J.P., Meyer, E.,
Hegner, M., Gerber, C., 2003. Nanotechnology 14, 86–90.
Aubin, K.L., Park, S.M., Huang, J., Craighead, H.G., Ilic, B.R., 2005. Sensors (2005 IEEE).
Braun, T., Barwich, V., Ghatkesar, M.K., Bredekamp, A.H., Gerber, C., Hegner, M., Lang,
H.P., 2005. Phys. Rev. E 72, 031907.
Campbell, G.A., Mutharasan, R., 2005a. Sens. Actuators A 122, 326–334.
Campbell, G.A., Mutharasan, R., 2005b. Biosens. Bioelectron. 21, 462–473.
Campbell, G.A., Uknalis, J., Tu, S.I., Mutharasan, R., 2007. Biosens. Bioelectron. 22,
1296–1302.
Canavese, G., Marasso, S.L., Quaglio, M., Cocuzza, M., Ricciardi, C., Pirri, C.F., 2007. J.
Micromech. Microeng. 17, 1387–1393.
Davis, S., Aldrich, T.H., Jones, P.F., Acheson, A., Compton, D.L., Jain, V., Ryan, T.E.,
Bruno, J., Radziejewski, C., Maisonpierre, P.C., Yancopoulos, G.D., 1996. Cell 87,
1161–1169.
Ghatkesar, M.K., Braun, T., Barwich, V., Ramseyer, J.P., Gerber, C., Hegner, M., Lang,
H.P., 2008. Appl. Phys. Lett. 92, 043106.
Gupta, A., Akin, D., Bashir, R., 2004. Appl. Phys. Lett. 84, 1976–1978.
Gupta, A.K., Nair, P.R., Akin, D., Ladisch, M.R., Broyles, S., Alam, M.A., Bashir, R., 2006.
Proc. Natl. Acad. Sci. U.S.A. 103, 13362–13367.
Ilic, B., Craighead, H.G., Krylov, S., Senaratne, W., Ober, C., Neuzil, P., 2004. J. Appl.
Phys. 95, 3694–3703.
Liu, Y., Li, X., Zhang, Z., Zuo, G., Cheng, Z., Yu, H., 2009. Biomed. Microdev. 11, 183–
191.
Lochon, F., Dufour, I., Rebière, D., 2005. Sens. Actuators B 108, 979–985.
Lochon, F., Dufour, I., Rebière, D., 2006. Sens. Actuators B 118, 292–296.
Maali, A., Hurth, C., Boisgard, R., Jai, C., Cohen-Bouhacina, T., Aimé, J.P., 2005. J. Appl.
Phys. 97, 074907.
Maraldo, D., Rijal, K., Campbell, G., Mutharasan, R., 2007a. Anal. Chem. 79,
2762–2770.
Maraldo, D., Garcia, F.U., Mutharasan, R., 2007b. Anal. Chem. 79, 7683–7690.
Metheny-Barlow, L.J., Li, L.Y., 2003. Cell Res. 13, 309–317.
Moulin, A.M., O’Shea, S.J., Welland, M.E., 2000. Ultramicroscopy 82, 23–31.
Nair, P.R., Alam, M.A., 2006. Appl. Phys. Lett. 88, 233120.
Park, J.H., Know, T.Y., Yoon, D.S., Kim, H, Kim, T.S., 2005. Adv. Funct. Mater. 15,
2021–2028.
Park, K., Jang, J., Irimia, D., Sturgis, J., Lee, J., Robinson, P., Tonerd, M., Bashir, R., 2008.
Lab Chip 8, 1034–1041.
Quaglio, M., Canavese, G., Giuri, E., Marasso, L.S., Perrone, D., Cocuzza, M., Pirri, C.F.,
2008. J. Micromech. Microeng. 18, 055012.
Ricciardi, C., Fiorilli, S., Bianco, S., Canavese, G., astagna, R., Ferrante, I., Digrego-
rio, G., Marasso, L.S., Napione, L., Bussolino, F., 2010. Biosens. Bioelectron. 25,
1193–1198.
Sader, J.E., 1998. J. Appl. Phys. 84, 64–76.
Sharma, V.K., Kalonia, D.S., 2004. AAPS PharmSciTech 17 5 (1), E10.
Shen, F., Lu, P., O’Shea, S.J., Lee, K.H., Ng, T.Y., 2001. Sens. Actuators A 15, 17–23.
Thaysen, J., Marie, R., Boisen, A., 2001. Proc. IEEE, 401–404 (MEMS, 2001).
Vančura, C., Lichtenberg, J., Hierlemann, A., Josse, F., 2005. Appl. Phys. Lett. 87,
162510.
Vančura, C., Dufour, I., Heinrich, S.M., Josse, F., Hierlemann, A., 2008. Sens. Actuators
A 141, 43–51.
Waggoner, P.S., Craighead, H.G., 2007. Lab Chip 7, 1238–1255.
Yu, Q., 2005. Future Oncol. 1, 475–484.