A multiplexed optofluidic biomolecular sensor for low mass detection†
Sudeep Mandal,‡
a
Julie M. Goddard‡
b
and David Erickson
*
b
Received 20th April 2009, Accepted 29th June 2009
First published as an Advance Article on the web 11th July 2009
DOI: 10.1039/b907826f
Optical techniques have proven to be well suited for in situ biomolecular sensing because they enable
high fidelity measurements in aqueous environments, are minimally affected by background
solution pH or ionic strength, and facilitate label-free detection. Recently, there has been significant
interest in developing new classes of optically resonant biosensors possessing very high quality-factors.
This high quality-factor enables them to resolve the presence of very small amounts of bound mass
and leads to very low limits of detection. A drawback of these devices is that the majority of the
resonant electromagnetic energy is confined within the solid light-guiding structure thus limiting the
degree to which it overlaps with the bound matter. This in turn lowers the ultimate device sensitivity, or
the change in output signal in response to changes in bound mass. Here we present a novel
optofluidic biosensor platform that incorporates a unique one-dimensional photonic crystal resonator
array which enables significantly stronger light-matter interaction. We show here how this, coupled
with the ability of planar photonic crystals to spatially localize the optical field to mode volumes on the
order of a wavelength cubed, enables a limit of detection on the order of 63 ag total bound mass
(estimated using a polyelectrolyte growth model) and a device sensitivity an order of magnitude better
than similar devices. The multiplexing capabilities of our sensor are demonstrated by the individual and
concurrent detection of interleukins 4, 6 and 8 using a sandwich assay.
Introduction
Biosensors that exploit optical,
1–3
electrical,
4
and mechanical
5
methods of signal transduction bypass the need for fluorescent,
radio, or enzymatic labels. Label-free approaches can exhibit
enhanced sensitivity and specificity over traditional sensors
because such labels can: interfere with the binding event, non-
specifically adsorb to the surface, and complicate the chemistry
of the detection reaction.
6
Of these methods, optical biosensors
are particularly promising because of their low limits of detec-
tion, high sensitivities and capacity for multiplexed detection.
7–9
While recent nano-mechanical
10
and nano-electrical
11,12
devices
have similarly proven successful, they tend to be limited in that
they require measurements to be made outside of the liquid
environment, or exhibit sensitivity to background electrical
conditions such as solution pH and ionic strength. The perfor-
mance of optical sensors is far less dependent on favourable
environmental conditions making them appropriate for a much
broader range of applications.
This advantage, amongst others, has led to the development
of a large number of label-free optical sensor technologies,
including interferometric,
13,14
resonant-cavity,
3,15,16
photonic
crystal
17–19
and surface plasmon resonance (SPR)
20–22
devices.
The performance of a biosensor is characterized by two
parameters: its limit of detection (the smallest amount or target
that can be detected) and the device sensitivity (or the amount
by which the detection signal responds to a change in input). In
the case of optical biosensors the limit of detection is largely
determined by how precisely a change in the output signal can
be resolved and sensitivity by how strong the overlap is between
the electromagnetic energy
23
and the target molecules. As
a result of this, there has been significant recent interest in the
development of whispering gallery mode (WGM) type biosen-
sors, exploiting toroidal,
24
microsphere,
25
microring,
26
and
microdisk
27
resonator structures. The advantage of these devices
is the high degree to which electromagnetic energy can be
localized resulting in very high quality factors (Q-factor) and
very narrow line-widths. These narrow line widths make it
relatively simple to resolve subtle changes in the resonant
frequency, yielding very low limits of detection.
3
In general
however to achieve these high Q-factors the electromagnetic
energy must be largely confined within a solid structure and
thus the extent to which it can interact with the bound molec-
ular targets is limited to a small portion of the evanescent field,
negatively impacting device sensitivity.
We present here a label-free multiplexed immunosensor
based on our nanoscale optofluidic sensor array (NOSA)
architecture.
28
Illustrated in Fig. 1, our device consists of
arrays of evanescently coupled one-dimensional photonic
crystal resonators. As we demonstrate below, the accessible
optical field inside the holes of photonic crystal is significantly
a
School of Applied and Engineering Physics, Cornell University, Ithaca,
NY, 14853, USA
b
Sibley School of Mechanical and Aerospace Engineering, Cornell
University, Ithaca, NY, 14853, USA. E-mail: de54@cornell.edu; Fax: +1
607-255-1222; Tel: +1 607-255-4861
† Electronic supplementary information (ESI) available: Schematic of
surface functionalization chemistry (Fig. S1), plot showing the resulting
spectra from multiplexed assays after introduction of 10 mg/ml of (a)
interleukin 6, (b) interleukin 8, and (c) interleukin 4, followed by
association with secondary antibody (Fig. S2), NOSA chip integrated
with PDMS microfluidics and secured in a plexiglass housing (Fig. S3),
effect of blocking buffer on device performance (Fig. S4). See DOI:
10.1039/b907826f
‡ S.M. and J.M.G. contributed equally to this work.
2924 | Lab Chip, 2009, 9, 2924–2932 This journal is ª The Royal Society of Chemistry 2009
PAPER www.rsc.org/loc | Lab on a Chip
Downloaded by University of Massachusetts - Amherst on 10 December 2012
Published on 11 July 2009 on http://pubs.rsc.org | doi:10.1039/B907826F
View Article Online
/ Journal Homepage
/ Table of Contents for this issue
stronger than the evanescent fields of WGM sensors, thereby
strengthening the light-matter interactions that enhance sensi-
tivity. We demonstrate here how this coupled with the ability
of planar photonic crystals to spatially localize the optical field
to mode volumes on the order of a wavelength cubed,
29,30
enables a limit of detection on the order of 10s of attograms
total bound mass (compared with femtograms
17,31,32
) and
a device sensitivity an order of magnitude better than similar
devices. In this paper we characterize the sensitivity of our
device to bound mass by monitoring the growth of a poly-
electrolyte multilayer and the resulting change in resonant
wavelength. Real-time binding kinetics and dynamic range
are determined by associating anti-streptavidin with surface-
immobilized streptavidin. While many optically resonant
sensors face design challenges when moving towards perform-
ing highly multiplexed detection, we demonstrate the facile
multiplexability our approach affords through the concurrent
detection of multiple interleukins (IL-4, IL-6, and IL-8) in
a single integrated optofluidic
33,34
device.
Results
Device operation and detection principle
As shown in Fig. 1, the NOSA platform consists of multiple
evanescently coupled 1-D photonic crystal resonators situated
along a single bus waveguide. A central cavity in the 1-D
photonic crystal structure of each resonator gives rise to a defect
state
35
in the photonic bandgap. This results in a resonant dip in
the output spectrum of the bus waveguide. By tailoring the cavity
length, each of the evanescently coupled 1-D resonators is
designed to possess a unique resonant wavelength. Fig. 1d
illustrates the steady-state electric field intensity distribution
within the 1-D resonator at the resonant wavelength. The
binding of target biomolecules to the surface of the resonator
induces a slight increase in the local refractive index around it.
The interaction of the resonant optical field with the bound
target biomolecules at the sensor surface and within the photonic
crystal holes results in a red-shift in the corresponding resonant
Fig. 1 Nanoscale Optofluidic Sensor Arrays. (a) 3D rendering of the NOSA device showing two 1-D photonic crystal resonators evanescently coupled
to a silicon bus waveguide. The first resonator is immobilized with an antigen whereas the second resonator acts as a control. (b) 3D rendering
illustrating the association of the corresponding antibody to the antigen immobilized resonator (not drawn to scale). (c) Experimental data illustrating
the successful detection of 45 mg/ml of anti-streptavidin antibody. The blue trace shows the initial baseline spectrum corresponding to Figure 1a where
the first resonator is immobilized with streptavidin. The red trace shows the test spectra after the association of anti-streptavidin as shown in
Figure 1b. The resonant wavelength of the control is unchanged while that of the streptavidin immobilized resonator red-shifts appreciably indicating
successful detection of anti-streptavidin. (d) Finite difference time domain (FDTD) simulation of the steady state electric field distribution within the
1-D photonic crystal resonator at the resonant wavelength. (e) SEM image demonstrating the 2-dimensional multiplexing capability of the NOSA
architecture.
This journal is ª The Royal Society of Chemistry 2009 Lab Chip, 2009, 9, 2924–2932 | 2925
Downloaded by University of Massachusetts - Amherst on 10 December 2012
Published on 11 July 2009 on http://pubs.rsc.org | doi:10.1039/B907826F
View Article Online
wavelength of the resonator. Put simply, the increase in the
refractive index of the optical cavity caused by the presence of
bound mass increases the effective optical length of the cavity
and thereby the wavelength of light that will resonate within it.
The output optical spectrum from the bus waveguide can be
constantly monitored and binding of target biomolecules to the
resonator is inferred when a red-shift is observed. Fig. 1c illus-
trates a typical experiment demonstrating the detection of anti-
streptavidin antibody binding to a streptavidin functionalized
resonator. Note that no observable shift is detected in the case of
the control resonator which is not streptavidin functionalized.
Since each evanescently coupled 1-D resonator possesses
a unique resonance in the output spectrum, multiplexed detection
along a single waveguide is possible. In this case, each of the 1-D
resonators is initially functionalized with a unique capture mole-
cule. The sample containing target bio-molecules is made to flow
over the 1-D resonators while the output spectrum is monitored. By
observing the combination of resonances that red-shift it is possible
to determine which target molecules were present in the detection
sample. Analysis of the degree of red-shift provides quantitative
information regarding the amount of bound mass which can be
correlated to concentration of target molecule present in a sample.
Fig. 1e is an SEM image of a NOSA array, illustrating the
2-dimensional multiplexing capabilities of the platform.
Device characterization
To quantitatively determine the device sensitivity to bound mass, we
use a polyelectrolyte multilayer ‘‘layer-by-layer’’ deposition
technique. This technique enables an accurate determination of the
sensing range of a biosensor, i.e., how far above the sensor surface
can biomolecules bind and still produce a shift in the resonant
wavelength. This is important to characterize as it helps determine
what kinds of surface conjugation techniques can be used success-
fully to immobilize the required capture biomolecules without
pushing the bound target biomolecules too far above the sensor
surface and away from the sensing region. Briefly, multilayers of
polyethyleneimine and polyacrylic acid were deposited on the
glutaraldehyde functionalized NOSA device, and on similarly
functionalized silicon wafers in parallel. The scheme for surface
functionalization is illustrated in Figure S1 (see ESI†) and is fully
described in the methods section. After deposition of each layer,
output spectra were recorded to quantify shift in resonant wave-
lengths and polyelectrolyte multilayer film thickness was determined
on silicon wafers using ellipsometry. Output spectra were compared
to the initial baseline spectra to determine resonance shift (Dl,in
nm), and were plotted against film thickness as plotted in Fig. 2.
Baseline spectra (Dl ¼ 0 nm) were taken on 5 NOSA reso-
nators and parallel silicon wafers after surface functionalization.
A film thickness of 3.11 nm corresponds to the molecular
thickness of native oxide, amine-terminated silane monolayer,
and glutaraldehyde functionalization, each of which contributes
1 nm to total film thickness. Since the field at the sensor surface
exhibits an exponential decay, the growth of the polyelectrolyte
multilayer and the resulting effect on resonance shift were fit to
an exponential model as shown in Fig. 2. Although the device
continues to respond past 30 nm of deposited polyelectrolyte
multilayers, it exhibits the greatest sensitivity in the first 20 nm of
multilayer growth, with an apparent sensitivity of 0.35 nm
resonance shift per nanometer multilayer growth. This suggests
that the NOSA device is well suited for small molecule detection,
as the Stokes radii of proteins relevant in medical diagnostics and
nucleic acids relevant in pathogen detection tend to be below
15 nm, in the region of greatest device sensitivity. Fig. 3 is a finite
difference time domain (FDTD) simulation that illustrates the
field decay within the innermost hole of the 1-D resonator which
reaches a 1/e value 80 nm away from the hole surface. Thus the
NOSA architecture can probe regions even 80–90 nm away from
Fig. 2 Optical response to polyelectrolyte layer growth. Effect of poly-
electrolyte multilayer thickness on resonance shift. Data have been fit to
an exponential model; error bars represent standard deviation.
Fig. 3 Electric field distribution inside resonator hole. (a) FDTD simu-
lation showing the steady state electric field distribution within the central
resonator cavity for the resonant wavelength. (b) Plot of the Electric field
intensity along the red line shown in Figure 3a. The gray boxes indicate
the device silicon whereas the white area indicates the electric field
intensity within the innermost hole of the 1-D resonator. It is clear that
a significant portion of the resonant field extends within the holes of the
resonator thus allowing for strong light-matter interactions within the
holes. The field decays to its 1/e value approximately 80 nm away from
the wall of the hole.
2926 | Lab Chip, 2009, 9, 2924–2932 This journal is ª The Royal Society of Chemistry 2009
Downloaded by University of Massachusetts - Amherst on 10 December 2012
Published on 11 July 2009 on http://pubs.rsc.org | doi:10.1039/B907826F
View Article Online
the substrate, which is consistent with our experimental obser-
vation (Fig. 2) that the red-shift did not saturate for
polyelectrolyte layers up to 35 nm thick.
Label-free multiplexed immunoassay
Cytokines represent a unique class of serum signaling proteins
that have proven to be a useful tool in diagnostics. Specifically,
interleukins have been demonstrated to be excellent candidates
as cancer biomarkers.
36
The potential for monitoring of in-vivo
concentrations of various interleukins for prognosis in cancer
patients
37,38
has generated significant interest in the biosensor
community.
To demonstrate the multiplexability of the NOSA device,
monoclonal antibodies to interleukin 4, 6, and 8 were immobi-
lized on adjacent resonators and tested for cross-reactivity
(see methods). Glutaraldehyde functionalized resonators and
streptavidin immobilized resonators served as controls for non-
specific analyte adsorption. As an additional control, cross-
reactivity of secondary antibodies with complementary as well as
non-complementary capture antibodies was assessed. Resonance
shifts due to cross-reactivity were less than 0.02 nm in all
experiments and did not therefore significantly affect the detec-
tion resonance shift. The Q-factor of 3000 for our resonators sets
our detection limit at 0.01 nm. Reported data are representative
of experiments repeated on at least two separate days. To assess
multiplex capability, we tested concurrent detection of multiple
interleukins after their association with secondary interleukin
antibodies. Fig. 4a shows the resulting spectra after introducing
1 mg/ml of interleukin 8 along with 10 mg/ml of interleukin 6,
followed by sequential association of secondary antibodies cor-
responding to each of these interleukins. In the figure, the
resonant wavelengths numbered 1 through 5 correspond to
control (glutaraldehyde functionalized), streptavidin-function-
alized control, anti-interleukin 6, anti-interleukin 4, and anti-
interleukin 8, respectively (Fig. 4b). The test spectrum (red) is
superimposed over the baseline spectrum (blue) to illustrate the
lack of significant non-specific binding. We observe shifts in the
resonance corresponding to immobilized monoclonal anti-
interleukin 8 (0.58 nm) and 6 (0.68 nm), but no significant shift in
the resonance corresponding to immobilized monoclonal anti-
interleukin 4. This further supports the ability of the NOSA
device to function as a multiplexed biosensor with little cross-
reactivity or non-specific binding. Figure S2 (see ESI†) shows the
resulting spectra from other multiplexed assays after introduc-
tion of 10 mg/ml of interleukin 6, 8, or 4 (respectively), followed
by association with secondary antibody. In each case, there is an
average shift of 0.72 nm (mean of 6 determinations) with
a standard deviation of 0.1 nm in the resonant wavelength cor-
responding to the target analyte used in a given experiment.
Physiologically relevant concentrations of serum interleukins
for in-vivo monitoring are on the order of 1–10 pg/ml, which is
within the detection limits of available ELISA techniques.
39,40
While the demonstrated LOD of the NOSA prevents the detec-
tion of interleukins at these concentrations, as outlined in the
subsequent section, improvements in the sensor design can
enable extremely high-Q microcavity resonators with a signifi-
cantly improved detection limit. In addition, by tagging the
secondary antibodies with nanoparticles possessing a high
refractive (such as titanium dioxide) the induced red-shift can be
amplified thus offering another means for enhancing the sensi-
tivity of this biosensor platform. In its current design, our device
can detect antibodies in a concentration range of 1 mg/ml to
1 mg/ml, which is of clinical significance in medical diagnostics
(i.e. HIV detection), and drug screening.
41
Determination of dynamic range
A dose-response curve was generated by associating varying
concentrations of anti-streptavidin antibody to immobilized
streptavidin. Since all antibodies possess roughly the same size
and molecular weight, we utilize an antibody-antigen system for
determining and comparing the dynamic range of our
NOSA sensor. Antibody concentration was varied from 0.010 to
Fig. 4 Multipexed detection of interleukins. (a) Spectra for resonators labelled 1 through 5 that correspond to control (glutaraldehyde functionalized),
streptavidin-functionalized control, anti-interleukin 6, anti-interleukin 4, and anti-interleukin 8, respectively. The trace in blue shows the initial baseline
spectrum. The red trace corresponds to the test spectrum after introducing 10 mg/ml of interleukin 6 along with 1 mg/ml of interleukin 8, followed by the
sequential association of secondary antibodies corresponding to each of these interleukins. We clearly see shifts corresponding to the resonators
functionalized with anti-interleukin 6 and 8 (Resonance 3 and 5 respectively) while the other resonances do not shift appreciably thus indicating the lack
of non-specific binding. Fabry-Perot resonances were filtered out in both spectra by performing a fast Fourier transform. (b) Reaction stages at each of
5 resonators.
This journal is ª The Royal Society of Chemistry 2009 Lab Chip, 2009, 9, 2924–2932 | 2927
Downloaded by University of Massachusetts - Amherst on 10 December 2012
Published on 11 July 2009 on http://pubs.rsc.org | doi:10.1039/B907826F
View Article Online
