Diffractive micro bar codes for encoding of biomolecules in multiplexed assays.
ABSTRACT Microparticles incorporating micrometer-sized diffractive bar codes have been modified with oligonucleotides and immunoglobulin Gs to enable DNA hybridization and immunoassays. The bar codes are manufactured using photolithography of a chemically functional commercial epoxy photoresist (SU-8). When attached by suitable linkers, immobilized probe molecules exhibit high affinity for analytes and fast reaction kinetics, allowing detection of single nucleotide differences in DNA sequences and multiplexed immunoassays in <45 min. Analysis of raw data from assays carried out on the diffractive microparticles indicates that the reproducibility and sensitivity approach those of commercial encoding platforms. Micrometer-sized particles, imprinted with several superimposed diffraction gratings, can encode many million unique codes. The high encoding capacity of this technology along with the applicability of the manufactured bar codes to multiplexed assays will allow accurate measurement of a wide variety of molecular interactions, leading to new opportunities in diverse areas of biotechnology such as genomics, proteomics, high-throughput screening, and medical diagnostics.
- SourceAvailable from: onlinelibrary.wiley.com[Show abstract] [Hide abstract]
ABSTRACT: Advances in lab-on-a-chip technologies enabled programmable, reconfigurable, and scalable manipulation of a variety of laboratory procedures. Samples, reagents, and fluids can be precisely controlled; buffer temperature, pH, and concentration control systems as well as a variety of detection systems can be integrated on a small chip. These advantages have attracted attention in various fields of clinical application including leukemia diagnosis and research. A lot of research on lab-on-a-chip based diagnosis has been reported and the field is rapidly expanding. This review describes recent developments of lab-on-a-chip technologies as solutions to challenges for high-throughput leukemia diagnosis.International journal of laboratory hematology 02/2013; · 1.30 Impact Factor
- Advanced Materials 06/2009; 21(40):4020 - 4024. · 14.83 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: A multiplexed suspension array platform, based on SU8 disks patterned with machine-readable binary identification codes is presented. Multiple probe molecules, each attached to individual disks with different unique codes, provide multiplexed detection of targets in a small sample volume. The experimental system consists of a microfluidic chamber for arraying the particles in a manner suitable for high throughput imaging using a simple fluorescent microscope, together with custom software for automated code readout and analysis of assay response. The platform is demonstrated with a multiplexed antibody assay targeting 3 different human inflammatory cytokines. The suitability of the platform for other bio-analytical applications is discussed.Biomedical Microdevices 03/2012; 14(4):651-7. · 2.72 Impact Factor
Diffractive Micro Bar Codes for Encoding of
Biomolecules in Multiplexed Assays
Graham R. Broder,†Rohan T. Ranasinghe,†Joseph K. She,†Shahanara Banu,†,‡Sam W. Birtwell,§,|
Gabriel Cavalli,†Gerasim S. Galitonov,§David Holmes,‡Hugo F. P. Martins,†Kevin F. MacDonald,§
Cameron Neylon,†,⊥Nikolay Zheludev,§Peter L. Roach,*,†and Hywel Morgan*,‡
School of Chemistry, School of Electronics and Computer Science, Optoelectronics Research Centre, and School of Physics
and Astronomy, University of Southampton, Highfield, Southampton, SO17 1BJ, UK, and STFC, Rutherford Appleton
Laboratory, Didcot, OX11 0QX, UK
Microparticles incorporating micrometer-sized diffractive
bar codes have been modified with oligonucleotides and
immunoglobulin Gs to enable DNA hybridization and
immunoassays. The bar codes are manufactured using
photolithography of a chemically functional commercial
epoxy photoresist (SU-8). When attached by suitable
linkers, immobilized probe molecules exhibit high affinity
for analytes and fast reaction kinetics, allowing detection
of single nucleotide differences in DNA sequences and
multiplexed immunoassays in <45 min. Analysis of raw
data from assays carried out on the diffractive micropar-
ticles indicates that the reproducibility and sensitivity
approach those of commercial encoding platforms. Mi-
crometer-sized particles, imprinted with several super-
imposed diffraction gratings, can encode many million
unique codes. The high encoding capacity of this technol-
ogy along with the applicability of the manufactured bar
codes to multiplexed assays will allow accurate measure-
ment of a wide variety of molecular interactions, leading
to new opportunities in diverse areas of biotechnology
such as genomics, proteomics, high-throughput screen-
ing, and medical diagnostics.
Rapid and multiplexed molecular detection with high sensitivity
and specificity is of great importance in gene profiling, clinical
diagnostics, and environmental monitoring. A key requirement
for multiplexing is the identification of each molecule set within
the library of probes used in an assay. In this context, microarrays
have revolutionized DNA expression profiling1,2and show great
promise for proteomics3-6and DNA sequencing.7In these
systems, a large number of different biomolecules are identified
by their position on a two-dimensional grid. However, microarrays
have certain disadvantages, including slow diffusion of molecules
to their binding sites and the inability to perform large numbers
of reactions simultaneously over a wide dynamic range.
Bead or suspension arrays, in which probes are attached to
the surface of microparticles, are an attractive alternative for
multiplexed analysis and have found applications in profiling DNA
or proteins from complex biological samples.8,9Such arrays offer
a number of advantages, including fast reaction kinetics and
flexibility of library content. High-throughput analysis by flow
cytometry makes it possible to interrogate vast numbers of probes
or to perform large numbers of replicate measurements.
In suspension arrays, sets of beads carry a unique bar code.10,11
To fully exploit the potential for multiplexed bead-based arrays, a
robust and reliable method of manufacturing encoded beads is
required. For high-throughput analysis, the decoding system must
identify each and every code quickly and reliably. A number of
encoding strategies have been proposed and demonstrated includ-
ing the use of nanoscale metallic bar codes that are read by
reflectance,12quantum dot-encoded mesoporous beads,13spatially
resolved photobleaching of fluorescent particles,14direct optical
reading of the shape of micromachined particles,15incorporation
of Raman-active organic compounds,16or optical resonance
(whispering gallery mode) in micrometer-sized silica spheres.17
* To whom correspondence should be addressed. E-mail: email@example.com
†School of Chemistry, University of Southampton.
‡School of Electronics and Computer Science, University of Southampton.
§Optoelectronics Research Centre, University of Southampton.
|School of Physics and Astronomy, University of Southampton.
⊥Rutherford Appleton Laboratory.
(1) Gresham, D.; Ruderfer, D. M.; Pratt, S. C.; Schacherer, J.; Dunham, M. J.;
Botstein, D.; Kruglyak, L. Science 2006, 311, 1932-1936.
(2) Abdullah-Sayani, A.; Bueno-de-Mesquita, J. M.; van de Vijver, M. J. Nat.
Clin. Pract. Oncol. 2006, 3, 501-516.
(3) Becker, K. F.; Metzger, V.; Hipp, S.; Hofler, H. Curr. Med. Chem. 2006,
(4) Balboni, I.; Chan, S. M.; Kattah, M.; Tenenbaum, J. D.; Butte, A. J.; Utz, P.
J. Annu. Rev. Immunol. 2006, 24, 391-418.
(5) Barbulovic-Nad, I.; Lucente, M.; Sun, Y.; Zhang, M. J.; Wheeler, A. R.;
Bussmann, M. Crit. Rev. Biotechnol. 2006, 26, 237-259.
(6) Wingren, C.; Borrebaeck, C. A. K. Omics 2006, 10, 411-427.
(7) Bentley, D. R. Curr. Opin. Genet. Dev. 2006, 16, 545-552.
(8) Meza, M. B. Drug Discovery Today 2000, 38-41.
(9) Nolan, J. P.; Mandy, F. Cytometry, Part A 2006, 69A, 318-325.
(10) Braeckmans, K.; De Smedt, S. C.; Leblans, M.; Pauwels, R.; Demeester, J.
Nat. Rev. Drug Discovery 2002, 1, 447-456.
(11) Wilson, R.; Cossins, A. R.; Spiller, D. G. Angew. Chem., Int. Ed. 2006, 45,
(12) Nicewarner-Pena, S. R.; Freeman, R. G.; Reiss, B. D.; He, L.; Pena, D. J.;
Walton, I. D.; Cromer, R.; Keating, C. D.; Natan, M. J. Science 2001, 294,
(13) Han, M. Y.; Gao, X. H.; Su, J. Z.; Nie, S. Nat. Biotechnol. 2001, 19, 631-
(14) Braeckmans, K.; De Smedt, S. C.; Roelant, C.; Leblans, M.; Pauwels, R.;
Demeester, J. Nat. Mater. 2003, 2, 169-173.
(15) Zhi, Z. L.; Morita, Y.; Yamamura, S.; Tamiya, E. Chem. Commun. 2005,
(16) Jun, B. H.; Kim, J. H.; Park, H.; Kim, J. S.; Yu, K. N.; Lee, S. M.; Choi, H.;
Kwak, S. Y.; Kim, Y. K.; Jeong, D. H.; Cho, M. H.; Lee, Y. S. J. Comb. Chem.
2007, 9, 237-244.
Anal. Chem. 2008, 80, 1902-1909
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008
10.1021/ac7018574 CCC: $40.75© 2008 American Chemical Society
Published on Web 02/14/2008
Nucleic acids have also been used as coding elements; these are
read in situ by sequential hybridizations of labeled oligonucle-
otides18,19or via hybridization to a conventional DNA chip after
removal of the tag sequence from the beads.20
Among these, spectral encoding, based on fluorescence
wavelength and intensity, appears to be the most promising.
However, this approach limits the number of bead-based assays
that can be performed to the number of spectrally resolvable
fluorophores that can be incorporated into a bead. Current spectral
encoding technology (e.g., Luminex xMAP) provides ∼100 unique
codes. In principle, the possibilities are far greater; for example,
6 colors at 10 intensity levels gives ∼1 million codes. However,
practical encoding capabilities are much lower (due to spectral
overlap, fluorescence intensity variations, signal-to-noise ratio, etc.)
and realistically five to six colors with six intensity levels could
yield ∼10 000-40 000 codes.10A major drawback of this method
is that most bead-based biochemical assays rely on fluorescence
readout, so that the additional use of fluorescence for encoding,
particularly where multiple fluorophores are used, places limits
on the spectral bandwidth available for fluorescent-based assays
or the number of individual codes that can be identified. Recently,
a method for fabricating graphical bar codes within microfluidic
channels was reported. Using laminar flow, separate coding and
probing regions were simultaneously assembled during a polym-
erization step.21The potential for encoding ∼106unique codes in
particles of length 180-270 µm was demonstrated.
We have developed a new and robust encoding system,
orthogonal to fluorescence-based assays, based on microparticles
that incorporate diffractive elements as identifiers.22Noninvasive
and noncontact reading of codes can be accomplished by analyzing
the spatial distribution of the diffracted light. This is shown
schematically in Figure 1, which demonstrates the principle of
encoding using microdiffraction gratings. A single grating diffracts
the incident light into a unique diffraction pattern as shown in
Figure 1a. The “code” is read by measuring the spatial distribution
of this diffracted light. In the simplest implementation, information
is encoded in the pitch of the grating, a. When the grating is
illuminated with light at wavelength λ, a series of diffracted beams
is created at angles θ, according to eq 1, where m is the order of
the diffracted beam. Measurement of the first order (m ) (1)
diffracted beam position (with respect to the zero order) gives
direct information about the pitch a. This is shown in Figure 1a
a sinθ ) mλ
We have previously demonstrated that ∼109unique codes can
be created from a library of five times superimposed gratings
fabricated on a 100-µm-long particle.22In fact, the encoding has
recently been extended to two dimensions, greatly amplifying the
encoding capacity to huge numbers; up to 1018unique codes.23
Shrinking the gratings reduces the encoding capacity; the value
quoted above is for a 100-µm particle read with a laser light of
wavelength 633 nm. The lower limit of grating size is ∼3 times
the wavelength of the reading laser (in our case, ∼1.9 µm);
however, gratings of this size only have one possible code. The
encoding capacity increases rapidly as the grating size is in-
creased, already reaching ∼104codes for 10-µm gratings.
In this paper, we assess the functionality of probe biomolecules
immobilized onto these diffractive micro bar codes, namely,
immunoglobulin Gs (IgGs) and DNA. We characterize the
thermodynamics and kinetics of their molecular interactions and
go on to demonstrate their use in multiplexed assays.
Optical Hardware. Simultaneous diffraction and fluorescence
measurements were made using a custom-made optical system
with an infinity corrected high numerical aperture objective lens
and appropriate optical band-pass filters and dichroics (Supporting
Information Figure S-2). Diffraction patterns were generated using
plane incident light. The diffracted light was collected by a CCD
camera and the diffraction pattern analyzed by dedicated computer
Particle Fabrication. Diffractive bar codes were fabricated
from photoactive epoxy SU-8-5 (Chestech Ltd. Rugby, UK). The
resist was spin-coated onto Si wafers (525 ( 25 µm thick) onto
which a thin layer of Al had been evaporated. After spinning onto
the primed wafers, the SU-8 was soft baked, exposed (using an
EVG 620 mask aligner with photomask from Compugraphic, UK),
and postexposure baked. The wafers were then developed in EC
solvent poly(propylene glycol) methyl ether acetate (Chestech
Ltd., Rugby, UK) for 2 min with agitation. The wafers were
thoroughly rinsed with isopropyl alcohol and blow dried. The Al
sacrificial layer was removed by sonicating the wafers in Microp-
osit MF-319 (tetramethylammonium hydroxide (2.2% w/v solution
in water), Chestech Ltd., Rugby, UK) at room temperature for 10
min. The microparticles were collected by centrifugation (12500g,
1 min), washed in methanol (1 mL × 8), and dried under vacuum
at room temperature for 4 h.
Reagents. Oligonucleotides (P1, 5′-biotin-AAAAAGTTGGATCC-
3′; P2, 5′-biotin-AAAAACTTGGATCC-3′ C1; 5′-Cy5-GGGATC-
CAAGTTTTTT-3′; FP1, 5′-Cy5-CTAGTTACTCTTGTTC-biotin-3′)
were purchased from ATDBio (Southampton, UK) or synthesized
using a MerMade 192 DNA synthesizer (Bioautomation Inc.), for
which all reagents were obtained from Link Technologies (Bellshill,
UK). Cy3-labeled goat polyclonal anti-human IgG and Cy5-labeled
goat polyclonal anti-rabbit IgG, anti-mouse IgG, and anti-guinea
pig IgG were purchased from Abcam plc. (Cambridge, UK).
Conjugation of Oligonucleotides and IgGs to Encoded
Microparticles. SU-8 microparticles for use in both hybridization
assays and immunoassays were functionalized by reaction of
avidin-DN (Vector laboratories, Burlingame, CA) or protein A
(Cambridge Bioscience, Cambridge, UK), respectively, with car-
boxyl-functionalized particles according to standard amide cou-
pling protocols.24Carboxyl groups were introduced by succiny-
lation of primary amine groups resulting from ring opening of
(17) Vollmer, F.; Arnold, S.; Braun, D.; Teraoka, I.; Libchaber, A. Biophys. J. 2003,
(18) Gunderson, K. L.; Kruglyak, S.; Graige, M. S.; Garcia, F.; Kermani, B. G.;
Zhao, C. F.; Che, D. P.; Dickinson, T.; Wickham, E.; Bierle, J.; Doucet, D.;
Milewski, M.; Yang, R.; Siegmund, C.; Haas, J.; Zhou, L. X.; Oliphant, A.;
Fan, J. B.; Barnard, S.; Chee, M. S. Genome Res. 2004, 14, 870-877.
(19) Li, Y. G.; Cu, Y. T. H.; Luo, D. Nat. Biotechnol. 2005, 23, 885-889.
(20) Nam, J. M.; Thaxton, C. S.; Mirkin, C. A. Science 2003, 301, 1884-1886.
(21) Pregibon, D. C.; Toner, M.; Doyle, P. S. Science 2007, 315, 1393-1396.
(22) Galitonov, G. S.; Birtwell, S. W.; Zheludev, N. I.; Morgan, H. Opt. Express
2006, 14, 1382-1387.
(23) Birtwell, S. W.; Galitonov, G. S.; Morgan, H.; Zheludev, N. I. Opt. Commun.
In press. Doi:10.1016/j.optcom.2007.04.066.
(24) Hermanson, G. T. Bioconjugate Techniques; Academic Press: London, 1996.
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008
residual surface epoxide groups of SU-8 with Jeffamine, as
described previously.25Detailed protocols for all synthetic steps
are available as Supporting Information.
Determination of Equilibrium Dissociation Constants.
Thermodynamics of molecular interactions were determined using
titration of analytes and flow cytometry. Human IgG-coated SU-8
microparticles (1 mg), suspended in storage buffer (10 mM NaH2-
PO4, 150 mM NaCl, 0.1% Tween-20, pH 7.8, 100 µL, giving a 0.75
µM concentration of immobilized human IgG) were incubated with
Cy5-labeled detector antibody (3 pM) at room temperature for
15-30 min. An aliquot of the sample (5 µL) was withdrawn,
immediately washed with PBS (200 µL), and stored at 4 °C in the
dark. The microparticles were separated from the reaction mixture
by centrifugation (12500g, 1 min), and an aliquot of the supernatant
(5 µL) was removed. A further aliquot of Cy5-labeled anti-IgG was
added, and the above procedures were repeated. The fluorescence
intensity of the samples was measured by fluorescence-activated
(25) Cavalli, G.; Banu, S.; Ranasinghe, R. T.; Broder, G. R.; Martins, H. F. P.;
Neylon, C.; Morgan, H.; Roach, P. L. J. Comb. Chem. 2007, 462-472.
Figure 1. (a) First-order (m ) (1) diffraction from a grating with pitch a1. (b) First-order diffraction from a grating with pitch a2, where a1> a2.
The two codes are distinguishable if the beam separation is greater than the beam width. (c) Four different SU-8 microparticles with their
corresponding diffraction patterns immersed in water. Insets show micrographs of the particles manufactured by optical lithography. The separations
between the first-order diffraction lines, which vary with the inverse of the grating period, are indicated. (d) Histogram showing the distribution
of observed diffraction angles from particles used in multiplexed assays.
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008
Solution-phase data for the human IgG/anti-human IgG inter-
action were obtained by measuring fluorescence polarization. To
a solution of Cy5-labeled goat anti-human IgG (0.67 nM) in buffer
(10 mM NaH2PO4, 150 mM NaCl, 0.01% NaN3, pH 7.8) was added
human-IgG (final concentration 0-666 nM), and the reaction
mixture was incubated at room temperature in the dark for 16 h.
The fluorescence polarization of each of the samples was recorded
using a microplate reader (Safire2, Tecan, Switzerland, λex) 635
nm, λem) 666 nm).
Avidin-DN coated SU-8 microparticles (0.4 mg), suspended in
SSPE (5×) buffer (0.75 M NaCl, 50 mM NaH2PO4, 5 mM EDTA,
0.02% Tween-20, pH 7.0, 250 µL, giving a 0.1 µM concentration of
immobilized avidin-DN), were incubated with Cy5-labeled bioti-
nylated oligonucleotide, FP1 (2.5 nM) at room temperature for
15-30 min. An aliquot of the sample (10 µL) was withdrawn and
stored at 4 °C in the dark. The microparticles were separated from
the reaction mixture by centrifugation (12500g, 1 min), and an
aliquot of the supernatant (5 µL) was removed. A further aliquot
of FP1 was added, and the above procedures were repeated. The
fluorescence intensity of the samples was measured by fluorescence-
activated flow cytometry (FACS).
The data (Figure 2) were fitted to a ligand-binding function
(eq 2, where [Bound], [Free], and [Nonspecific] refer to concen-
trations of specifically bound, free, and nonspecifically bound Cy5-
labeled analyte) using commercially available software (SigmaPlot
9, Systat Inc., San Jose, CA), to determine the total number of
binding sites (Bmax) and Kd.
Kinetic Analysis of Avidin/Biotin Binding. Avidin-coated
SU-8 microparticles, (0.08 mg, bearing 5.7 pmol of avidin-DN) were
suspended in SSPE (5×) buffer (45.5 µL). Biotinylated Cy-5
labeled oligonucleotide (FP1) was added (0.3 nmol, final concen-
tration 6.5 nM). The suspension was agitated and aliquots (5 µL)
were removed periodically for immediate analysis, without wash-
ing. The fluorescence intensity of the samples was measured by
Solution-phase data were obtained by measuring fluorescence
polarization. A solution containing biotinylated Cy-5-labeled oli-
gonucleotide (FP1, 0.2 pmol, final concentration 1 nM), biotin (1.8
pmol, final concentration 9 nM), and avidin-DN (2 pmol, final
concentration 10 nM) in SSPE (5×) buffer (200 µL) was incubated
at room temperature. (Note: Solutions containing only FP1 and
avidin-DN were found to undergo self-quenching upon binding,
presumably due to the proximity of four fluorophores to each other
in the protein tetramer.) The fluorescence polarization was
recorded periodically using a microplate reader (Safire2, Tecan,
Switzerland, λex) 635 nm, λem) 666 nm).
Both data sets were fitted to eq 3 using commercially available
software (SigmaPlot 9, Systat Inc., San Jose, CA) in order to
determine the rate constant, kon.
Kinetic Analysis of Human IgG/Anti-Human IgG Binding
on Encoded Microparticles. Human IgG immobilized SU-8
microparticles (1 mg), suspended in storage buffer (100 µL), were
incubated with Cy5-labeled detector antibody (6.67 pmol, final
concentration 67 nM) at room temperature. After fixed time
intervals, samples (5 µL) were withdrawn, immediately washed
with storage buffer (200 µL), and stored at 4 °C in the dark. The
Figure 2. Saturation binding curves for Cy5-anti human/human IgG interaction in solution (a) on SU-8 microparticles, (b) and for avidin-DN/
Cy5-labeled biotinylated oligonucleotide FP1 on SU-8 particles (c). Time-course measurements of avidin/biotin binding in solution (d) and with
avidin-DN immobilized on SU-8 microparticles (e), with insets showing linear plots of the integrated rate equation. Panel f shows fluorescence
time-course measurement of human IgG/anti-human IgG binding on encoded microparticles.
Kd+ [Free]+ [Nonspecific][Free](2)
∂[Avidin - Biotin Complex]
kon[Free Avidin][Free Biotin] (3)
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008
fluorescence intensity of the samples was measured by fluorescence-
Multiplexed DNA Hybridization Assay. Biotinylated oligo-
nucleotides P1 and P2 were captured onto encoded particles by
immobilized avidin in 5× SSPE buffer in separate reactions. A
mixture of the encoded particles underwent hybridization to C1
(1 pmol, 3 nM) in 5× SSPE buffer at room temperature for 40
min, followed by centrifugation and resuspension in 2 M aqueous
urea for 2 min. The microparticles were then washed twice with
5× SSPE buffer before analysis.
Mutliplexed Immunoassays. Polyclonal IgGs from human,
rabbit, guinea pig, or mouse IgG were captured onto encoded
particles by immobilized protein A in PBS in separate reactions.
Labeled anti-IgGs (Cy3- or Cy5-labeled, supplied by Abcam,
Cambridge, UK, 13 pmol, 260 nM) were added to the appropriate
mixture of the encoded particles in PBS at room temperature for
1 h. with occasional gentle agitation. The microparticles were then
washed twice with PBS before analysis.
Statistical Analysis of Fluorescence Measurements and
Estimate of Sensitivity of Multiplexed Assays. Student’s t-test
was used to calculate values of P for measured fluorescence
intensities for positive and negative data sets (using Microsoft
Excel). The sensitivity of assay for anti-human IgG was estimated
according to Kd, Bmax, and the number of required replicate
measurements as follows. The ratio of complexed to free protein
concentration is determined through eq 4:
[Free Analyte] ) Kd[Complex]/[Free Probe]
Assuming 20 particles are measured (P <0.0003 for all data sets
measured), then [Free Probe] + [Complex] ) 0.37 nM, and at
saturation [Complex] ) 0.37 nM. Hence [Complex] required for
a signal/noise ratio of five ) (5 ÷ 7.7) × 0.37 nM ) 0.24 nM, and
[Free Probe] ) 0.37 nM - 0.24 nM ) 0.13 nM. Substituting into
eq 4 gives [Free Analyte] ) 16.6 nM. Therefore, the total
concentration of analyte in the sample ) [Free Analyte] +
[Complex] ) 16.6 + 0.37 ) 16.97 nM.
RESULTS AND DISCUSSION
Particle Fabrication. The multiplexed encoding principle was
demonstrated using bar-shaped microparticles fabricated using
conventional photolithography from a photoactive epoxy resin (SU-
8).26This material has been previously shown to be suitable for
multistep solid-phase peptide and oligonucleotide synthesis.25
Micrometer-sized diffraction gratings were created using particles
with serrated edges, as show in Figure 1c. The advantage of this
method is that it involves a single photolithographic step. Large
numbers of particles (2.5 × 106particles/silicon wafer) can be
manufactured with identical chemical and physical properties
using this method. More complex and higher density gratings
can be made by nanoembossing.26,27The diffracted first-order
beams from the four codes used for multiplexed assays are
sufficiently separated in angular space to permit unambiguous
decoding, with a small overlap between codes 2 and 3 (Figure
1d). Data from particles with diffraction angles within the overlap
region were discarded.
Thermodynamic and Kinetic Characterization of Probes
Attached to Encoded Microparticles. One of the advantages
of particle-based suspension arrays over planar array formats is
that reaction kinetics and thermodynamics resemble those ob-
served in solution phase, enabling higher throughput and sensitiv-
ity.28In order to demonstrate these principles, we measured
binding thermodynamics and kinetics on our encoded micropar-
ticles via titrations of analyte binding and fluorescence time-course
We determined the equilibrium dissociation constants for key
molecular interactions by carrying out binding titrations (Figure
2a-c). These data confirm that both anti-IgG/IgG (Kdin solution
) 15 ( 6 nM; Kdon SU-8 microparticles ) 9 ( 3 nM) and avidin/
biotin (Kdin solution ) 1 × 10-6nM;29Kdon SU-8 microparticles
) 1 ( 0.3 nM) interactions retained high affinity when transferred
to the solid phase, though in the case of avidin/biotin, a significant
reduction in affinity was observed, which may be due to the use
of recombinant avidin-DN or the bulky oligonucleotide-biotin
conjugate. From Bmax, the number of analyte molecules bound to
each microparticle at saturation was calculated. For Cy5-labeled
anti-IgG, this value was ∼107molecules (or 17 amol) per particle,
while ∼0.8 × 107molecules (or 13 amol) of Cy5-labeled biotiny-
lated oligonucleotide were bound to each microparticle.
The kinetics of binding of Cy5-labeled biotinylated oligonucle-
otide to immobilized avidin-DN, and Cy5-labeled anti-human IgG
to human IgG were analyzed by carrying out fluorescence time-
course experiments. For the avidin-DN-biotinylated oligonucle-
otide interaction (Figure 2d), the data were fitted to eq 3 in order
to calculate kon, with the linear plots of the integrated rate equation
used to confirm the second-order nature of the reaction. The value
obtained on SU-8 particles (4.4 × 106M-1s-1) was very close to
that observed in solution for avidin-DN/biotinylated oligonucle-
otide FP1 (7.4 × 106M-1s-1) and within the range previously
reported for the native biotin-avidin interaction (kon) 5 × 105-
8.2 × 107M-1s-1).29For the kinetically more complex sandwich
system used for binding of labeled detection IgGs, we did not
attempt to parse individual rate constants but simply measured
the time taken for a system containing a 67 nM concentration of
analyte to reach equilibrium (Figure 2f), which was found to be
<10 min. We were therefore able to confirm that binding assays
on our encoded microparticles can be carried out in short time
Multiplexed Biomolecular Assays. Multiplexed analysis was
demonstrated using the four different photolithographically fab-
ricated bar codes (designated codes 1-4) shown in Figure 1c.
These codes can easily be differentiated with the aid of a light
microscope, but to increase throughput, the bar codes were read
using an automated recognition and analysis system that captured
the diffraction image on a CCD camera. Image processing software
identifies a particle from the captured image, assigns a direction
vector to the principal axis, and then performs a Fourier analysis
of the first-order spots to determine the particle code. This process
is performed at high speed, taking less than 1 ms.
A multiplexed DNA hybridization assay was performed using
sequences corresponding to a polymorphic locus (N1303K) on
the gene encoding the cystic fibrosis transmembrane conductance
(26) Banu, S.; Birtwell, S. W.; Galitonov, G. S.; Chen, Y.-F.; Zheludev, N. I.;
Morgan, H. J. Micromech. Microeng. 2007, 17, S116-S121.
(27) Guo, L. J. J. Phys. D 2004, 37, R123-R141.
(28) Finkel, N. H.; Lou, X.; Wang, C.; He, L. Anal. Chem. 2004, 76, 352A-
(29) Green, N. M. Biochem. J. 1963, 89, 585.
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008
regulator protein (Figure 3a).30Two sets of encoded particles were
functionalized with avidin and used to capture biotinylated probe
oligonucleotides. Wild type sequence P2 was bound to code 3 and
the mutant sequence, P1, differing by a singe nucleotide, was
attached to code 2. Figure 3b shows a microscope image and
fluorescence image of the two particle types, demonstrating that
the perfectly matched sequence hybridizes but the mismatched
sequence does not. The observed fluorescence shows a degree
of inhomogeneity across the particle although the physical basis
for this observation has not yet been determined. The particles
were assayed using conventional flow cytometry, and the scatter
plot for the two different populations is shown in Figure 3c. Sorting
on the basis of the fluorescence signal, followed by decoding of
a representative sample of particles (50 from each fraction) showed
that the hybridized population consisted of 96% code 3, with the
unhybridized population B consisting of 96% code 2. These results
demonstrate that the use of diffractively encoded microparticles
permits discrimination of a single mismatch from a Watson-Crick
base pair within a short DNA duplex, confirming their applicability
to SNP genotyping applications.
To demonstrate the applicability of the technology to immu-
nodiagnostics, a multiplexed immunoassay was implemented for
the detection of IgGs. Two sets of encoded particles were
functionalized with two different IgGs using protein A capture.
The particles were incubated with complementary antibodies as
shown schematically in Figure 4a. The fluorescence images
(Figure 4b) show the selective binding of the differentially labeled
detection antibodies to the correct encoded particle. The labeled
particles were again analyzed using flow cytometry and sorted
into two populations on the basis of relative fluorescence (Figure
4c). A total of 50 particles from each of these two sorted
populations were then decoded, showing that particles from the
Cy5-labeled population (A) were correctly identified as code 2,
and the Cy3-labeled population (B) as code 4 with greater than
A prerequisite for reliable multiplexed immunoassays is the
demonstration of selective antigen-antibody binding against a
background of other potentially competing interactions. Therefore,
a small library consisting of four sets of different encoded particles
were each functionalized with a different antigen. Mixtures of
fluorescently labeled detection antibodies were added to aliquots
of the library (Figure 4d). For the experiments described above,
a two-step sorting/decoding process was used to read the code
and measure the fluorescence from each particle. Since this
approach is not amenable to high-throughput analysis, the mi-
croparticles from these quadruplexed experiments were dispersed
on a microscope slide and analyzed using a modified optical
microscope setup capable of reading diffraction and fluorescence
intensity. Manual reading of each particle using this apparatus
took ∼10-30 s. Binding of the complementary antibody gave
fluorescence intensities in the range 250 000-430 000 (au, cor-
rected for the quantum yield of the detection antibodies). Control
particles, whose corresponding anti-IgG was not present in the
sample, had intensities in the range 13 000-58 000 (au), demon-
strating very little cross-reaction.
Replicate Measurements within Multiplexed Assays and
Analyte Sensitivity. The throughput and sensitivity of micropar-
ticle-based assays is determined in practice by the detectable
amount of analyte per particle and the number of replicate samples
that must be measured to satisfy a set confidence level in the
results. The latter in turn is determined by the variation of signal
(in this case fluorescence) between particles of the same type that
have been subject to the same assay conditions. We analyzed data
(30) Osborne, L.; Santis, G.; Schwarz, M.; Klinger, K.; Dork, T.; McIntosh, I.;
Schwartz, M.; Nunes, V.; Macek, M.; Reiss, J.; Highsmith, W. E.; McMahon,
R.; Novelli, G.; Malik, N.; Burger, J.; Anvret, M.; Wallace, A.; Williams, C.;
Mathew, C.; Rozen, R.; Graham, C.; Gasparini, P.; Bal, J.; Cassiman, J. J.;
Balassopoulou, A.; Davidow, L.; Raskin, S.; Kalaydjieva, L.; Kerem, B.;
Richards, S.; Simonbouy, B.; Super, M.; Wulbrand, U.; Keston, M.; Estivill,
X.; Vavrova, V.; Friedman, K. J.; Barton, D.; Dallapiccola, B.; Stuhrmann,
M.; Beards, F.; Hill, A. J. M.; Pignatti, P. F.; Cuppens, H.; Angelicheva, D.;
Tummler, B.; Brock, D. J. H.; Casals, T.; Macek, M.; Schmidtke, J.; Magee,
A. C.; Bonizzato, A.; Deboeck, C.; Kuffardjieva, A.; Hodson, M.; Knight, R.
A. Hum. Genet. 1992, 89, 653-658.
Figure 3. Assay used to differentiate oligonucleotides differing by a single nucleotide. (a) Two different oligonucleotides (P1 and P2) that vary
only at a single position are bound to encoded microparticles. After hybridization to a third oligonucleotide (C1) that is a perfect match for one
sequence, the particles were washed in 2 M urea and then imaged. (b) White light microscope image (left panel) and fluorescence image (right
panel). The particles bearing the perfectly matched sequence fluoresce and those with the single mismatch do not. (c) Scatter plot of the two
populations (A and B). These were sorted by FACS on the basis of Cy5 fluorescence and the codes of 50 particles in each population read.
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008
from the quadruplex immunoassay and single base mismatch
DNA hybridization assay, since these assays are the most complex
and have the highest probability of cross-reactions and high
background fluorescence, allowing the most realistic estimation
of these parameters. Mean fluorescence intensities with standard
deviations for representative positive (complementary anti-IgG or
oligonucleotide present) and negative (complementary anti-IgG
or oligonucleotide absent) data sets are shown in Figure 5a and
c. Though the standard deviations of positive and negative
populations are well separated, the coefficients of variance (47-
60%) are larger than those typical of optimized ELISAs (5-10%).31
A possible contribution to these relatively large coefficients of
variance may be the observed inhomogeneity of the fluorescence
(Figures 3b and 4b). There is therefore a nonzero probability of
any given particle within either positive or negative population
giving a fluorescent signal that falls within the range attributed
to the other population, i.e., generating a false positive or false
Student’s t-test was used to compare the difference between
“negative” and “positive” data sets from the quadruplexed immu-
noassay and DNA hybridization assay. Cumulative P values from
three positive data sets are plotted against number of particles
read, Figure 5b and d. As expected, as more particles are read,
the value of P falls, indicating greater confidence in the difference
between data sets. After reading 11 particles, the probability of
any of the positive data sets being indistinguishable from the
negative data set is <0.01, and after reading 16 particles, the
probability is <0.0003. Therefore, for multiplexed platforms where
thousands of molecular interactions are assayed, measurement
of 15-20 replicates is necessary to satisfy a confidence level of
>99.97%. This level of degeneracy is greater than that necessary
for ELISAs, but is comparable to that used in Illumina BeadArrays
where ∼30 replicate beads are used for each oligonucleotide to
enable averaging and thorough statistical analysis of results.32
From Bmax, the number of analyte molecules bound to each
microparticle at saturation can be calculated. For Cy5-labeled anti-
IgG, this was ∼107molecules (or 17 amol) per particle; a similar
number (∼0.8 × 107molecules, or 13 amol) of Cy5-labeled
biotinylated oligonucleotide were bound to each microparticle.
Given the number of replicate particles required for analysis, we
can estimate the amount of bound analyte required to generate a
positive signal. The maximum binding capacity of a single particle
is ∼107molecules (17 amol); therefore, 20 particles will be
saturated by 340 amol of bound (labeled) IgG or DNA analyte.
The sensitivity of individual assays will be determined by the
signal/noise ratio (SNR) obtained for the analyte and its affinity
for the immobilized complementary probe. For example, Figure
5a shows that the human IgG/anti-human IgG interaction, has a
SNR (at saturation) of 7.7, with Kd) 9 ( 3 nM. Assuming that a
SNR of 5 is sufficient for detection, and that the assay is carried
out in a volume of 1 µL, eq 4 can be used to determine the total
concentration of anti-human IgG required to produce a signal. This
predicts that the total concentration of analyte in the sample
is16.97 nM, giving an estimate of sensitivity of approximately 17
fmol, or 1010molecules. This is a relatively conservative estimate
of sensitivity; if a cutoff value equal to twice background fluores-
cence is used, as for other bead-based assays,33the sensitivity
increases to a detection limit of 5.35 fmol (in a 1-µL volume), or
3.2 × 109molecules. To demonstrate this experimentally, we
(31) de Jager, W.; Rijkers, G. T. Methods 2006, 38, 294-303.
(32) Novak, J. P.; Miller, M. C.; Bell, D. A. Biol. Direct In press. DOI 10.1186/
(33) Dunbar, S. A.; Vander Zee, C. A.; Oliver, K. G.; Karem, K. L.; Jacobson, J.
W. J. Microbiol. Methods 2003, 53, 245-252.
(34) Marie, R.; Schmid, S.; Johansson, A.; Ejsing, L. E.; Nordstrom, M.; Hafliger,
D.; Christensen, C. B. V.; Boisen, A.; Dufva, M. Biosens. Bioelectron. 2006,
Figure 4. Multiplexed immunoassays with encoded microparticles.
(a) Rabbit and human antibodies are attached to encoded micropar-
ticles, which are then mixed with detection antibodies in a single
reaction. Detection antibodies were Cy3-labeled anti-human and Cy5-
labeled anti-rabbit IgG, respectively. (b) Microscope images of two
particles from the multiplex reaction, imaged with white light (top),
Cy3 fluorescence (middle) and Cy5 fluorescence (bottom). The faint
fluorescence from code 2 in the second image (Cy3 filter) is due to
the intrinsic fluorescence of the SU-8 polymer34and not to cross-
reaction of the Cy3-labeled antibody. (c) Scatter plot for two popula-
tions of particles, which were then sorted on the basis of Cy3 and
Cy5 fluorescence and the codes of 50 particles in each population
read. Population A: (bar code 2, 49 (98%); bar code 4, 0 (0%);
unreadable/incorrectly identified, 1 (2%). Population B: (bar code 2,
0 (0%); bar code 4, 50 (100%); unreadable/incorrectly identified, 0
(0%). (d) Results from the quadruplex immunoassay showing mean
particle fluorescence intensity (n ) 15-25) for a small library of four
different IgG functionalized bar codes after addition of Cy5-labeled
anti-IgGs to an aliquot of the library. Sample A: only anti-mouse IgG
was added, binding selectively to code 3. The mean fluorescence
following binding (and washing) of ∼100 particles is shown (with
standard error). Sample B: anti-guinea pig, anti-mouse, and anti-
rabbit IgG were all added to an aliquot of the codes, giving a
fluorescent “readout” as for sample A. Sample C: all four comple-
mentary fluorescently labeled antibodies were added to the sample
of four codes.
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008
attempted to detect 5.35 fmol of IgG in an immunoassay. The IgG
was successfully detected with an average SNR of 4 (for a sample
of 20 particles, data shown in Supporting Information Figure S-1).
We have demonstrated a new method for encoding biomo-
lecular probes, based on fabricating micrometer-sized diffraction
gratings. The technology has the potential to provide vast numbers
of unique codes, up to 1018for a 50-µm particle using 5 times
superimposed 2-D gratings.23The particles can be rapidly decoded
(in <1 ms) by automated software. The encoded particles are
fabricated using basic photolithography and are made from a
functional polymer (SU-8) that is compatible with multistep
chemical synthesis25and multiplexed assays. Statistical analysis
of the signal variation from particle to particle in assays indicates
that 15-20 replicate measurements are required to achieve a
>99.97% confidence in the discrimination between positive and
negative samples. Biochemical assays carried out on these
microparticles exhibit favorable binding kinetics and thermody-
namics, allowing small amounts (<20 fmol in 1-µL volume) of
target analyte to be detected, thus enabling massively parallel
determination of molecular interactions with high sensitivity and
accuracy. While manual operation of the combined diffraction/
fluorescence microscope described herein requires 10-30 s to
analyze each particle, the throughput could be greatly increased
by coupling decoding optical hardware to a fluidic system. Since
analysis of diffraction can be accomplished in an automated
manner on the millisecond time scale, we anticipate throughput
comparable to conventional flow cytometers. We envisage using
the same devices for directed sorting of particles during multistep
chemical synthesis to provide an integrated system for the high-
speed manufacture of encoded libraries of molecules (oligonucle-
otides, peptides, etc.) and analysis of their molecular interactions.
This project was supported by Research Councils UK through
the Basic Technology Programme. We also thank the cleanroom
staff at EPFL, Switzerland, for their assistance with fabrication
and Andrew Whitton for technical assistance. G.R.B., R.T.R. and
J.K.S. contributed equally to this work.
SUPPORTING INFORMATION AVAILABLE
Detailed experimental protocols for functionalization of SU-8
microparticles with oligonucleotides/antibodies and detection of
femtomole quantities of anti-human IgG, as well as schematic
representation of the modified microscope setup for simultaneous
measurement of diffraction and fluorescence from encoded mi-
croparticles. This material is available free of charge via the
Internet at http://pubs.acs.org.
November 21, 2007.
for review September 4,2007. Accepted
Figure 5. Mean fluorescence intensities with standard deviations for representative positive (complementary anti-IgG or labeled oligonucleotide
present) and negative (complementary anti-IgG or labeled oligonucleotide absent) particle sets functionalized with different probes, (a) and (c),
and cumulative P values from positive data sets plotted against read number, (b) and (d). Data from particles functionalized with human IgG
(blue squares), guinea pig IgG (green triangles), and rabbit IgG (pink circles) are compared in (b), while data from P2- and P1-functionalized
particles (differing by a single nucleotide) are compared in (d). The 0.03% probability that the difference in mean fluorescence of positive and
negative populations could arise by chance is shown by the horizontal dotted line in (b) and (d).
Analytical Chemistry, Vol. 80, No. 6, March 15, 2008