pubs.acs.

org/Langmuir
© XXXX American Chemical Society

Improved Method for Counting DNA Molecules on

Biofunctionalized Nanoparticles
Filip Delport,† Ania Deres,‡ Jun-ichi Hotta,‡ Jeroen Pollet,† Bert Verbruggen,† Bert Sels,§
Johan Hofkens,‡ and Jeroen Lammertyn*,†
†
Department of Biosystems, Division Mechatronics, Biostatistics and Sensors, ‡Department of Chemistry,
Molecular and Nanomaterials, and §Department of Microbial and Molecular Systems, Centre for Surface
Chemistry and Catalysis, KULeuven, Leuven, Belgium

Received December 14, 2009

In order to accurately determine low numbers (1-100) of immobilized ssDNA molecules at a single, silica 250 nm
nanoparticle surface, we hereby propose an integrated approach combining classic single molecule confocal microscopy
(SMCM), that is, stepwise photobleaching of labeled ssDNA, with modified total internal reflection fluorescence
microscopy (mTIRF). We postulate that SMCM alone is unable to exactly account for all labeled ssDNA because of
inherent laser polarization effects; that is, perpendicularly oriented molecules to the sample surface are not (or are only
slightly) susceptible to laser excitation and thus are invisible in a classic photobleaching experiment. The SMCM method
accounts for at best two-thirds (68%) of the present ssDNA molecules. The principle of the mTIRF technique, which
relies on the creation of highly inclined illumination combined with part of the laser remaining in normal K€ohler
illumination, enables accurate counting of SMCM invisble molecules. The combined approach proposed here
circumvents the polarization issue and allows a complete single molecule counting on individual nanoparticles, fully
in line with bulk measurements, as will be demonstrated.

New developments in biofunctionalized nanomaterials today
are the driving force for innovative future applications in the
electronic, chemical, biotechnology, and medical industries. Par-
ticularly, functionalized nanoparticles (NPs) attract a lot of
attention in the field of life sciences for their diagnostic and
therapeutic properties, for example, as miniaturized biosensors or
as drug delivery vehicles.1-4 NPs have for instance been con-
jugated with a variety of biomolecules such as proteins, enzymes,
and antibodies.5-9 Recently, DNA functionalized NPs have been
recognized as attractive nanotools in medical and food biomole-
cular diagnostics because of their inherent specific properties
including DNA detection through hybridization, aptamer based
ligand detection,10,11 and DNA mediated NP aggregation and
disaggregation.12
While accurate quantification of low numbers of anchored
DNA biomolecules on individual NPs is of utmost importance, it
is a challenging task. Common direct detection techniques based
*To whom correspondence should be addressed. E-mail: Jeroen.Lammertyn@
biw.kuleuven.be.
(1) Quarta, A.; Ragusa, A.; Deka, S.; Tortiglione, C.; Tino, A.; Cingolani, R.;
Pellegrino, T. Langmuir 2009, 25(21), 12614–12622.
(2) Kim, J. S.; Rieter, W. J.; Taylor, K. M. L.; An, H.; Lin, W. L.; Lin, W. B.
J. Am. Chem. Soc. 2007, 129(29), 8962–8963.
(3) Fu, Y.; Zhang, J.; Lakowicz, J. R. Langmuir 2008, 24(7), 3429–3433.
(4) Senarath-Yapa, M. D.; Phimphivong, S.; Coym, J. W.; Wirth, M. J.;
Aspinwall, C. A.; Saavedra, S. S. Langmuir 2007, 23(25), 12624–12633.
(5) Tan, J. S.; Butterfield, D. E.; Voycheck, C. L.; Caldwell, K. D.; Li, J. T.
Biomaterials 1993, 14(11), 823–833.
(6) Tansil, N. C.; Gao, Z. Q. Nano Today 2006, 1(1), 28–37.
(7) Wang, J. Small 2005, 1(11), 1036–1043.
(8) Wang, L.; Wang, K. M.; Santra, S.; Zhao, X. J.; Hilliard, L. R.; Smith, J. E.;
Wu, J. R.; Tan, W. H. Anal. Chem. 2006, 78(3), 646–654.
(9) You, C. C.; De, M.; Rotello, V. M. Curr. Opin. Chem. Biol. 2005, 9(6), 639–
646.
(10) Ellington, A. D.; Szostak, J. W. Nature 2002, 355(6363), 850-852.
(11) Jhaveri, S.; Rajendran, M.; Ellington, A. D. Nat. Biotechnol. 2000, 18(12),
1293–1297.
(12) Hurst, S. J.; Hill, H. D.; Mirkin, C. A. J. Am. Chem. Soc. 2008, 130(36),
12192–12200.
on thermogravimetric analysis, atom emission spectroscopy,
infrared spectroscopy, and liquid chromatography13-16 are some-
times labor intensive and require relatively large amounts of
sample, and hence, they are less suitable for the correct analysis
of ultralow quantities of DNA, when attached on nanomaterials.
Instead, measuring directly on the single NP is more accurate but
also more complex. Recently, single molecule detection strategies,
such as single molecule counting and single molecule bleaching,
have been proposed to offer a solution for the problem of
counting single proteins17 and quantum dots18,19 when coupled
to silica, polystyrene, gold, silver, lipid, NPs, and so on.20,21
Although clear molecule-by-molecule counting has been demon-
strated, a deviation from bulk measurements was always encoun-
tered.17
In order to accurately determine low numbers of immobilized
ssDNA molecules at a single NP surface, we hereby propose an
integrated approach combining classic single molecule confocal
microscopy (SMCM) techniques, that is, stepwise photobleaching
of labeled ssDNA, and modified total internal reflection fluores-
cence (mTIRF) illumination (vide infra and Supporting Informa-
tion, Figure S4). We postulate that SMCM alone is unable to
exactly account for all labeled tags because of laser polarization
(13) Chen, S. H.; Kimura, K. Langmuir 1999, 15(4), 1075–1082.
(14) Chen, Y.; Lu, Z. H. Anal. Chim. Acta 2007, 587(2), 180–186.
(15) Li, Z. Z.; Wen, L. X.; Shao, L.; Chen, J. F. J. Controlled Release 2004, 98(2),
245–254.
(16) Mao, Y.; Daniel, L. N.; Whittaker, N.; Saffiotti, U. Environ. Health
Perspect. 1994, 102, 165–171.
(17) Casanova, D.; Giaume, D.; Moreau, M.; Martin, J. L.; Gacoin, T.; Boilot,
J. P.; Alexandrou, A. J. Am. Chem. Soc. 2007, 129(42), 12592–12593.
(18) Stavis, S. M.; Edel, J. B.; Samiee, K. T.; Craighead, H. G. Lab Chip 2005,
5(3), 337–343.
(19) Zhang, C. Y.; Johnson, L. W. J. Am. Chem. Soc. 2008, 130(12), 3750–3751.
(20) Sonnichsen, C.; Reinhard, B. M.; Liphardt, J.; Alivisatos, A. P. Nat.
Biotechnol. 2005, 23(6), 741–745.
(21) Kunding, A. H.; Mortensen, M. W.; Christensen, S. M.; Stamou, D.
Biophys. J. 2008, 95(3), 1176–1188.

Langmuir XXXX, XXX(XX), XXX–XXX DOI: 10.1021/la904702j A

Letter Delport et al.

effects; that is, perpendicularly oriented molecules to the sample

surface are not (or are only slightly) susceptible to laser excitation
and thus are invisible in a classic photobleaching experiment. The
mTIRF approach circumvents the polarization issue and allows a
complete single molecule counting on individual NPs, fully in line
with bulk measurements, as will be demonstrated.
Immobilization of the ssDNA on the NPs was based on
synthesis procedures reported by Hermanson:22 15-mer ssDNA
tagged with both an amine group and a fluorescent ATTO 647N
dye was coupled to 249 ((27) nm silica carboxyl functionalized
nanoparticles using carbodiimide chemistry (Supporting
Information, Scheme S1). A linear relationship was observed
between the average number of ssDNA immobilized on the NP
and the initial concentration of ssDNA in the reaction mixture
in the range from 0.5 to 500 nM, as determined with bulk
fluorescence measurement (Supporting Information, Figure S2).
Early attempts to determine coupling efficiency revealed a dis-
crepancy between direct measurements on the NP and indirect
measurements on the supernatant. This discrepancy originates Figure 1. (A) 10
10 μm confocal scan of fluo-labeled ssDNA
functionalized NPs. (B) Stepwise photobleaching of 3 fluorescent
from handling mistakes and losses which influence the ssDNA molecules on 1 NP until a baseline signal is reached.
content of the supernatant and thus the calculated bound
fraction.
As a reference, classic single molecule photobleaching was
applied according to Casanova et al.17 to quantify the immobi-
lized ssDNA molecules on a single NP with the aim of compa-
ring these results to bulk fluorescence measurements. The
SMCM setup is described in detail in the Supporting Information
(Figure S3). Previous experiments have indicated that the max-
imum fluorophore labeled ssDNA on a single NP, that can be
differentiated by discrete energy levels, is limited to six molecules.
To match the immobilization conditions of the bulk measure-
ment, samples for the SMCM experiment were prepared by Figure 2. (A) Histogram of the SMCM measurements showing
immobilizing different mixtures of fluorescent and nonfluorescent the distribution of the number of NPs bioconjugated with 0-4
anchored fluorescent ssDNA molecules and corresponding fitted
ssDNA with the following ratios: 1/20, 1/37.5, 1/75, and 1/150 Poisson distribution (illustrated here for the dilution 1/37.5). (B)
with each time a total added ssDNA concentration of 25 nM. Average conjugated ssDNA per NP for the different fluorescence
No significant effect of diluting fluorescent ssDNA with non- dilution ratios.
fluorescent ssDNA on the bioconjugation was observed for
the bulk measurements, with respect to the dilution factor images were observed. A Poisson distribution23 was fitted to the
(Supporting Information, Figure S6). histogram data, after correction for the nonfluorescent NPs, to
In a typical bleaching experiment, 20 μL of the ssDNA-NP estimate the average number (c) of immobilized fluorescent
sample was then dispensed on ozone treated cover glass and spin- ssDNA molecules per NP using MATLAB software (The
coated. The sample was scanned in the confocal setup to locate the Mathworks).
NP in a 10
10 μm area, after which the laser (633 nm) was Using the ssDNA immobilization protocols mentioned above,
focused on a single NP (Figure 1A). For each sample with a the bulk fluorescence measurements reveal a total average num-
different ratio labeled/unlabeled ssDNA, around 100 NPs were ber of 80 immobilized ssDNA molecules per NP (Supporting
measured individually by stochastically photobleaching the fluor- Information, Figure S2). However, the average number of fluor-
ophores. The number of different energy levels in the stepwise escent ssDNA molecules counted in the 1/20 ratio of fluorescent/
decrease of the fluorescence intensity as a function of time are a nonfluorescent ssDNA mixture with the SMCM setup was equal
measure for the number of fluorescent ssDNA molecules bound to 2.54, giving, accounting for a dilution factor of 20, a total of 51
to the NP surface (Figure 1B). The fluorescence intensity steps are DNA molecules on 1 NP. Similar experiments were repeated for
not equidistant and suggest a xyz component dependency of the other dilution factors (Supporting Information, Table S5).
differently oriented dyes and hence a polarization effect. Based on The data are summarized in Figure 2B. As was suggested in the
these counting data, a histogram was created relating the number introduction, SMCM measurements with focused illumination
of NPs with their respective number of bound fluorescent ssDNA are indeed unable to accurately account for the presence of all
molecules (Figure 2A). Since the NPs without fluorescently attached ssDNA. Instead, at best, two-thirds (68%) of the
labeled ssDNA are not visible on the confocal scan and, hence, immobilized ssDNA could be counted as compared to the bulk
not accounted for in the histogram, the number of empty NPs was fluorescence measurement.
determined by comparing the density of NPs on the fluorescence We believe the difference can be accounted for by polarization
image to a regular optical transmission image of the same cover effects of the excitation laser light. Indeed, with an incoming light
glass. No interfering impurities in the transmission or fluorescent source perpendicularly oriented to the surface (along the z-axis),
only ssDNA molecules, more specifically the orientation of the
disklike fluorescent dye, oriented with a certain angle to the z-axis
(22) Hermanson, G. T. Bioconjugate Techniques, 2nd ed.; Elsevier: London, 2008.
(23) Agresti, A. An Introduction to Categorical Data Analysis, 2nd ed.; Wiley- are appropriately susceptible to excitation (Figure 3A, gray cone)
Interscience: New York, 2007. and are thus visible for counting. Because of the high N.A. (1.3) of

B DOI: 10.1021/la904702j Langmuir XXXX, XXX(XX), XXX–XXX

Delport et al.
Figure 3. (A) 3D artist impression and fluorescent microscopy
image of NPs with K€ ohler illumination (gray cone represents
unexcitable orientation, blue labels are fluorescing, red labels in
the gray cone are not fluorescing). (B) 3D artist impression and
fluorescent microscopy image of the same area as in (A) now
excited with mTIRF after complete photobleaching with K€ ohler
illumination. Size of the microscopy image is 24.6
24.6 μm2.
the objective, some z-axis polarization (up to 12.8%) might occur
in the confocal setup.24-26 Nevertheless, the fluorescence intensity
caused by excited z-axis oriented fluorophores remains small
compared to the noise on the signal of xy-oriented fluorophores.
If this hypothesis is valid, the missing fluorescence should be
visible using the wide field fluorescence imaging technique with
mTIRF. The principle of this technique is based on the creation of
a highly inclined illumination with a substantial z-component
combined with part of the laser beam remaining in K€ohler
illumination. This total excitation contains all directions of
polarization components. As a result, every fluorophore is excited
irrespective of their orientation toward the laser.27 This mTIRF
setup (Supporting Information, Figure S4) allows a depth of field
excitation of several micrometers.28,29 Figure 3 illustrates the
envisioned polarization effect. First, a laser with a high power
beam (50 mW) was employed to illuminate the sample in K€ohler
illumination modus (Figure 3A) to image the molecules with a
substantial x-axis or y-axis component orientation. Imaging was
continued until complete photobleaching of the dye molecules
(24) Dedecker, P.; Muls, B.; Hofkens, J.; Enderlein, J.; Hotta, J. I. Opt. Express
2007, 15(6), 3372–3383.
(25) Richards, B.; Wolf, E. Proc. R. Soc. London, Ser. A 1959, 253(1274), 358–
379.
(26) Wolf, E. Proc. R. Soc. London, Ser. A 1959, 253(1274), 349–357.
(27) Schneckenburger, H. Curr. Opin. Biotechnol. 2005, 16(1), 13–18.
(28) Tokunaga, M.; Imamoto, N.; Sakata-Sogawa, K. Nat. Methods 2008, 5(2),
159–161.
(29) Rocha, S.; Hutchison, J. A.; Peneva, K.; Herrmann, A.; Muellen, K.; Skjot,
M.; Jorgensen, C. I.; Svendsen, A.; de Schryver, F. C.; Hofkens, J.; Uji, I.
ChemPhysChem 2009, 10(1), 151–161.
Langmuir XXXX, XXX(XX), XXX–XXX
Letter
Figure 4. (A) Stepwise photobleaching with mTIRF illumination
of 5 fluorescent groups on 1 NP; fluoresence intensity expressed in
kRFU per 0.28 μm2. (B) Histogram of K€ ohler photobleached NPs.
(C) Histogram of mTIRF photobleached NPs.
was achieved, leaving no detectable signal from the screened
region above the background level. Then the angle of laser
incidence was increased above the critical angle to establish a
partial objective based mTIRF with the creation of the highly
inclined illumination as a result. As can clearly be seen from
Figure 3B, under the mTIRF conditions, some fluorescence
indeed reappears, confirming our hypothesis. Thus, by counting
the SMCM nonbleached fluorescent NPs in the mTIRF mode, we
account for between 13% and 30% of the fluorescent ssDNA NPs
with respect to the bulk measurements, respectively, for the 1/20
and 1/10 dilution factors.
To further quantify the exact number of SMCM invisible
fluorophores, a “counting-by-photobleaching experiment” was
performed with confocal and mTIRF microscopy on the same
sample with a dilution ratio of 1/25 (Figure 4A). The fluorescence
intensity of multiple NPs on a 24.6
24.6 μm2 image was
recorded simultaneously with a highly sensitive cooled CCD
camera and integrated over a square region of 0.28 μm2, compris-
ing all emitted light of one NP.
By fitting a Poisson distribution on the histogram of the
counted ssDNA molecules per NP in the combined confocal
and mTIRF (Figure 4B, C) photobleaching measurements, we
determined the mean of the distribution (c) to be 2.33 and 3.37,
respectively. This amounts to a 31% difference in counted
ssDNA. Since an average of 68% of fluorescent ssDNA was
counted using the SMCM experiment compared to the bulk
measurement, the combined experiment fairly adds up the missing
ssDNA. A simple calculation reveals that 99% of the ssDNA is
traced when combining both single molecule techniques.
In conclusion, we have demonstrated a combined, general
approach for accurate quantification of low numbers of biocon-
jugated ssDNA molecules on single nanoparticles. The single
molecule microscopy approach was compared to bulk fluores-
cence measurements which are commonly applied to determine
the density of bioconjugated molecules on the nanoparticles. The
presented improved counting method is based on a simple step-
by-step photobleaching of single molecules both in confocal and
wide field mTIRF mode. The use of the latter technique accounts
for the missing number of fluorescent molecules that are not
excited and bleached in the confocal technique as a result of
polarization effects. Adding up the measured ssDNA numbers
resulting from both bleaching techniques fairly accounts for the
total number of attached target biomolecules such as ssDNA.
This counting method allows quantification, through Poisson
modeling, of the number of bioconjugated molecules per NP in
DOI: 10.1021/la904702j C
Letter
contrast to the bulk measurements that only give an average
estimate for a complete batch of NPs.
As the precise detection of small numbers of bioconjugated
molecules is a prerequisite for many particle based theranostic
applications, including cancer detection and treatment, single
virus detection, and nucleic acid based biosensors, we believe the
presented improved counting methodology is of value for many
biorelated fields. Our combined counting approach has been
demonstrated here for labeled ssDNA, but it certainly contributes
to the characterization of a broader range of bioconjugated
nanomaterials (e.g., gold, silver, silica, polysterene) and nanos-
tructures (e.g., nanorods, nanoshells, and branched nano-
particles) potentially having a large impact on the development
of new technologies with applications in medical and food
diagnostics, drug delivery systems, and biopolymer mediated
catalysis.
Materials and Methods
Sample Preparation. Before functionalization of the silica
NPs (Micromod, Germany) with ssDNA (Eurogentec, Belgium),
the carboxyl modified NPs were washed in a 25 mM MES buffer
(pH = 5, Sigma Aldrich), centrifuged, and decanted. A common
method in biomolecule immobilization was used to covalently
link ssDNA molecules to the cleaned NPs. Clean NPs were
contacted with 12.5 mg/mL 1-ethyl-3-[3-dimethylaminopro-
pyl]carbodiimide hydrochloride (EDC, Pierce Biotechnology)
and 12.5 mg/mL N-hydroxysuccinimide (NHS, Sigma Aldrich),
yielding active N-hydroxy succinimide pendant groups on the
surface of the NPs, which spontaneously react with the amine of
ssDNA to form covalent links in the next step. Variable ssDNA
loadings were prepared simply by varying the concentration (in
the nM range) of 50 -amine functionalized, 30 -Atto647N dye
labeled ssDNA (15 nucleotides). The ssDNA modified NPs were
collected and washed thoroughly in several solutions (once in a
SSC-SDS buffer (pH = 7.0) supplemented with 0.5% SDS and
D DOI: 10.1021/la904702j
Delport et al.
twice in the MES buffer) to remove unbound ssDNA. As the
ssDNA does not form a dense layer around the NP, nonspecific
binding sites at the NP surface were blocked by incubating the
ssDNA NPs in a new mixture of EDC/NHS together with 25 mM
ethanolamine, resulting in a hydrophilic uncharged surface of
pendant alcohol groups. A final washing step (once in SSC-SDS
buffer and twice in PBS buffer) was applied before storage of the
final ssDNA NPs in PBS buffer until measurement.
Measurement. The supernatant was collected after each
washing step, diluted to 1 mL, and stored for further analysis.
Bulk fluorescence measurements were carried out to quantify the
amount of unbound ssDNA in the supernatant after the different
washing steps and of the bound DNA on the NPs. Samples were
transferred to a microtiterplate and measured on a spectrophot-
ometer (Spectramax M2e, molecular Probes) at an excitation
wavelength of 600 nm, cutoff wavelength of 630 nm, and emission
wavelength of 664 nm. Calibration curves linking the Atto-647N
labeled ssDNA concentration to the fluorescence intensity in
solutions were carefully set up taking into account possible buffer
effects. A detailed description of the SMCM and mTIRF setup
can be found in the Supporting Information.
Acknowledgment. This work was financially supported by
the Institute for the Promotion and Innovation by Science
and Technology (IWT)-Flanders (63437-63384), the Fund for
Scientific Research (FWO)-Flanders (Grants G.0337.08 and
G.0298.06), the Industrial Research Fund (IOF) KULeuven,
and the Research Fund KULeuven (Centre of Excellence
CECAT). Long term structural funding “Methusalem” by the
Flemmish government is gratefully acknowledged.
Supporting Information Available: List of abbreviations,
experimental details, bulk fluorescence data, microscopy setups,
and widefield fluorescence data summarizing table and bulk
dilution data. This material is available free of charge via the
Internet at http://pubs.acs.org.
Langmuir XXXX, XXX(XX), XXX–XXX