DNA conformation on surfaces measured
by fluorescence self-interference
Lev Moiseev*, M. Selim ¨ Unlu ¨†‡§, Anna K. Swan†, Bennett B. Goldberg†‡§, and Charles R. Cantor*¶
*Center for Advanced Biotechnology and Departments of†Electrical and Computer Engineering,‡Physics, and§Biomedical Engineering, Boston University,
Boston, MA 02215
Contributed by Charles R. Cantor, December 27, 2005
The conformation of DNA molecules tethered to the surface of a
microarray may significantly affect the efficiency of hybridization.
Although a number of methods have been applied to determine
the structure of the DNA layer, they are not very sensitive to
variations in the shape of DNA molecules. Here we describe the
application of an interferometric technique called spectral self-
interference fluorescence microscopy to the precise measurement
to the surface and thus determine specific information on the
conformation of the surface-bound DNA molecules. Using spectral
self-interference fluorescence microscopy, we have estimated the
shape of coiled single-stranded DNA, the average tilt of double-
stranded DNA of different lengths, and the amount of hybridiza-
tion. The data provide important proofs of concept for the capa-
bilities of novel optical surface analytical methods of the molecular
disposition of DNA on surfaces. The determination of DNA con-
formations on surfaces and hybridization behavior provide infor-
mation required to move DNA interfacial applications forward and
thus impact emerging clinical and biotechnological fields.
hybridization ? microarray ? spectroscopy
sequencing, and drug discovery, all benefiting greatly by the highly
paralleled detection of the technique. One of the defining charac-
teristics of a DNA array is the availability of the single-stranded
probes for hybridization with the target. Immobilized molecules
located farther away from the solid support are closer to the
solution state and are more accessible for contact with dissolved
analytes. The surface, especially a hydrophobic one, acts as a shield
for probes positioned close to it because of the associated steric
factors and lack of diffusion of the bound molecules (1–5). Thus,
only in the future development and fabrication of microarrays, but
also in designing new applications (6).
Recently, advances have been made to characterize the structure
of surface-bound DNA probes using such optical or contact meth-
ods as ellipsometry, optical reflectivity (7, 8), neutron reflectivity
the structure of the surface-bound DNA depending on its density
and surface charge. However, most experimental techniques char-
acterize the DNA layer as a single entity (for instance, parameter-
izing its thickness or density) without examining the specific posi-
tions of internal elements of the DNA chain. Spectral self-
interference fluorescence microscopy (SSFM) can measure the
vertical position of a fluorescent label above an optically structured
silicon chip. We show that locating the label attached to a certain
position within a DNA chain provides insight into the shape of
DNA molecules bound to the surface.
NA array technology has become a widespread tool in bio-
logical research with applications in expression screening,
We study the conformation of single-stranded DNA (ssDNA) and
double-stranded DNA (dsDNA) on glass surfaces by using 50- and
bound to either the first strand at its distal 3? end, or the second
strand at its 3? or 5? end. Below, we describe the detection principle
along with results on ssDNA and dsDNA conformation. A sum-
mary of all of the data is presented in Table 1.
Detection Principle. We use a complementary combination of a
traditional reflection technique and an interferometric fluores-
cence spectroscopy technique to determine the average optical
thickness of biological layers as well as the height of sparse
fluorescent markers, both with subnanometer accuracy. Samples
are prepared on layered dielectric films, specifically an oxide-
coated silicon substrate. The first technique, white light (WL)
reflection spectroscopy, is based on spectral variations of re-
flection from thin transparent films. Interference of light re-
flected from the top surface and a buried reference surface
results in periodic oscillations in the reflection spectrum as
shown in Fig. 1. The principle is similar to the interference-based
detection technique using color variations due to increased path
length as a consequence of surface binding on optically coated
silicon (19). Our spectral measurements obtain very high accu-
racy (?0.2 nm), comparable to ellipsometry. The result is an
average optical density measurement of the biological film and
thus provides a precise relative measure of the additional
‘‘optical’’ mass on the substrate, after, e.g., hybridization.
The second technique, SSFM (20), is an interferometric
technique in fluorescent imaging that analyzes the spectral
oscillations emitted by a fluorophore located on a layered
reflecting surface and yields a precise position determination
(Fig. 1). The spectral oscillations are due to the self-interference
from the direct and reflected emission and thus encode the
vertical position of that fluorophore with subnanometer accu-
racy. These markers can be sparse or buried under a biological
film. In contrast to earlier fluorescence interference microscopy
techniques that rely on intensity variation of total fluorescent
emission (21, 22), SSFM utilizes spectral information and pro-
vides higher precision with a single measurement. The following
discussion of measurements on oligonucleotides illustrates the
WL Reflectivity Measurements. As schematically shown in Fig. 2,
oligonucleotides carrying a 5?-amino tag are covalently bound to
an aminated surface via a homobifunctional crosslinker. Using
WL reflectivity, we determine progressive growth of the surface-
bound thin films during DNA immobilization steps. The thick-
ness of the silane layer is 0.8–1.0 nm, which roughly corresponds
to a monolayer; phenylene isothiocyanate adds another 0.5–0.6
nm. Immobilization of DNA leads to a further increase in the
Conflict of interest statement: No conflicts declared.
Abbreviations: dsDNA, double-stranded DNA; ssDNA, single-stranded DNA; SSFM, spectral
self-interference fluorescence microscopy; WL, white light.
¶To whom correspondence should be sent at the present address: SEQUENOM, Inc., 3595
Johns Hopkins Court, San Diego, CA 92121. E-mail: email@example.com.
© 2006 by The National Academy of Sciences of the USA
February 21, 2006 ?
vol. 103 ?
no. 8 ?
optical thickness. In this particular experiment, oligonucleotides
are not fluorescently labeled, and the average film thickness
determined by WL reflection spectroscopy is depicted as trans-
parent box above the oxide in Fig. 2a.
The physical thickness of the DNA layer can be established
from WL reflectivity or ellipsometry only if the value of the
index of refraction is known. In these experiments, we assumed
the index of refraction of DNA to be 1.46 as was measured for
dense layers (23). We measured the optical film thickness for
ssDNA of 21 and 50 nt as 1.0–1.5 and 2.0–2.5 nm, respectively.
Although the precision of absolute thickness will depend on the
index, WL reflectivity provides an accurate relative measure of
additional mass on the surface and can thus monitor the
efficiency of hybridization. For example, adding complementary
second strands to 50-mers results in an increase in the film
thickness by ?1.0 nm (compared with ?2 nm for the first strand)
corresponding to a hybridization efficiency of ?50%. Precise
determination of the hybridization efficiency is not crucial for
the significant conclusions drawn about the conformation of
ssDNA and dsDNA on surfaces. An approximate value is
estimated to check for consistency of independent measure-
ments. Incomplete hybridization is schematically illustrated by
converting half of the oligonucleotides from Fig. 2a to dsDNA
in Fig. 2 b and c.
Determination of the Position of a Fluorescent Label. Once the
optical thicknesses of the layers are established, they can be used
in SSFM measurements to determine the position of the flu-
orophores relative to the surface.
The first experiment studied the elevation of 21- and 50-bp
dsDNA fragments. In principle, the maximum elevation of the
label is limited by the length of the double helix, which is ?7 nm
for 21-bp and ?17 nm for 50-bp fragments. dsDNA has a
persistence length of ?50 nm (24), so the short fragments in our
experiments can be viewed as rigid rods on hinges. A simple rod,
hinged to the surface, would have an average height of the distal
label, assuming free rotation of one half of the length, with an
average tilt angle of 60° from normal. However, free rotation
may be limited by steric constraints from nearby DNA molecules
and interactions with the surface. Using SSFM, we measured the
elevation of the label on top of DNA double helices to be 5.5 nm
for 21-bp fragments and 10.5 nm for 50-bp fragments, which
represents tilt angles from the normal of 50° and 40°, respec-
tively. These values are a measure of the average distribution of
heights within the microscope focal spot. Whereas some DNA
helices may be standing straight up, others may even be lying flat
on the surface. In Fig. 2b, we also show statistical distribution of
many measurement results illustrating a variance of the average
label position on the order of 1 nm.
If the label on the second strand is at the 3? end, its location
in the double helix will be at the bottom, close to the surface.
Although we did not expect to see a significant variation, the
proximal, 3?-end label on 50-bp DNA is elevated by ?2–3 nm
compared with only 0.5–1 nm in the case of the 21-bp fragment.
Theoretically, the position of the proximal label on stand-alone
double-stranded fragments, especially sharing the sequence of
the first 21 bp, should not be higher just because the double-
stranded fragment above it is longer. However, this may be due
to the fact that a 50-mer is long enough to have stable partial
hybridization that could free the proximal end and yield an
elevated label position. Steric hindrance is also a possibility.
We also studied the conformation of ssDNA by measuring the
height of fluorescent tags attached to the free end of surface-
bound DNA oligonucleotides (Fig. 3.). Unlike dsDNA, ssDNA
is flexible and little is known about the shape or size of ssDNAs
on the surface. AFM measurements suggest that ssDNA immo-
bilized on a surface exists in a globular conformation (16).
However, there are reports that because of steric hindrance from
nearby molecules, ssDNA may change its conformation from a
random coil to more extended forms (23, 25). The fluorescent
label attached to the distal end of surface-bound single-stranded
21-mer is found to be close to the surface: within 1 nm. In a
5.5 nm above the surface as illustrated in Fig. 3b. It is difficult
to calculate the average expected position of an end-label for
random coils of this size, yet the disproportionally high location
of the end-label on 50-mers points out to a considerably more
extended conformation compared with 21-mers. This extended
conformation may be caused by the steric effect from closely
located grafts in the DNA layer. The surface density of immo-
bilized ssDNA measured by using a radiolabel is ?35 fmol?mm2
for both 21- and 50-mers, which translates into 11-nm distances
between adjacent molecules or a 5.5-nm radius of free space
around each. At the same time, the length of a fully extended
50-mer is 27.5 nm, enough to interact with its neighbors, at least
intermittently. It appears that there is a pronounced effect from
When a second, unlabeled strand is hybridized, the label at the
distal end of the newly formed duplex extends out as well. Unlike
the dsDNA and the unhybridized strands, both of which are
carrying a fluorescent marker but at different heights. The
average position of the marker should be somewhere in between
depending on the degree of hybridization. The binding efficiency
can be calculated from comparing the average height of the
the first strand or the second strand is labeled. In our experi-
ments, an estimate of the extent of hybridization is between 30%
and 50%. This rough estimate is close to the result obtained by
WL reflectivity: 50% hybridization for both 21- and 50-mers,
demonstrating self-consistency. As a further check, we have
performed density measurements with radiolabeled DNA and
found it consistent with 50% hybridization. Because of intrinsic
limitations such as substrate related quenching of radiative
emission, we do not consider the radiolabeling for absolute
determination of DNA densities, but rather only for relative
estimation of DNA densities.
Interesting results were obtained when we assessed how the
conformation of a surface-bound 50-nt labeled oligonucleotide
changes when it is annealed with a 21-mer complementary to
figurations are shown along with examples of measured spectra for each
technique. (Left) WL reflection spectroscopy is based on spectral variations of
top surface and a buried reference surface results in periodic oscillations.
(Right) The SSFM technique maps the spectral oscillations emitted by a flu-
orophore located on a layered reflecting surface into a precise position
determination. Analysis of the spectral oscillations (using a grating spectrom-
the vertical position of that fluorophore with subnanometer accuracy.
Schematic representation of the interferometric experimental con-
www.pnas.org?cgi?doi?10.1073?pnas.0511214103 Moiseev et al.
either its top or bottom part. When a 21-mer complementary to
the top section was annealed, the position of the distal end
increased from 5.5 to 6.5 nm. However, a different situation is
observed when an ssDNA–dsDNA construct has the double-
stranded part proximal to the surface. In this case, the position
of the label is lower than that of an unhybridized oligo, decreas-
ing in average height from 5.5 to 3.5 nm. Recalling that mea-
surements such as these of pretagged oligos always average over
the unhybridized single strands, our data suggests that the distal
bound 21-mer construct is nearly vertical, and the proximal
bound construct has formed a rotation point allowing the
flexible distal end to approach closer to the surface. The
sequence of our oligonucleotides rules out the possibility of
intramolecular DNA structures.
Table 1 summarizes the entire range of experiments, both with
single oligonucleotides and complementary pairs of 21- and
50-bp units, and the final results of hybridization with segments
of different length.
We have demonstrated that SSFM is a powerful tool for studying
the conformation of DNA molecules immobilized on a surface.
The method is noncontact, and there are no limitations to
applying it to DNA arrays submerged in an aqueous buffer. The
technique specifically determines the axial position of the la-
beled nucleotide only and is therefore complementary to WL
reflectivity and ellipsometry techniques. Because the fluoro-
phore height is encoded only through the phase and frequency
of the oscillations from the spectral interference, the measure-
ment is insensitive to the intensity of fluorophore emission.
Thus, photobleaching, surface density-dependent dye–dye non-
radiate transfer effects, or spectral modifications will not affect
bound on the surface (a, top) after hybridization with complementary strand labeled at the 5? (distal) end (b) and 3? (proximal) end (c, bottom).
Moiseev et al.
February 21, 2006 ?
vol. 103 ?
no. 8 ?
the results. They may result in an overall intensity variation but
will yield similar oscillatory behavior of the spectra; hence,
misinterpretation of the height measurement is unlikely. These
processes merely alter the total number of photons, and inte-
gration times can be easily modified to provide whatever signal-
to-noise is required for the necessary precision of the measure-
ment. Details of the methods used for processing spectral data
are described below.
grafted DNA segments on surfaces with partially hybridized
molecules, and yet the majority of experimental techniques are
not able to determine ssDNA and dsDNA conformation. The
SSFM technique will allow the study of a variety of surface-
bound DNA structures, for example, hairpin loops formed with
DNA synthesis techniques. The use of hairpin loops offers the
potential of a cost-effective method for producing high-quality
synthesized dsDNA arrays for protein arrays and transcription
factor measurements. The idea is to synthesize arrays of long
single strands on the surface whose end sequences include their
own complement such that the completed strand is likely to
hybridize itself, thereby eliminating the need for batch-
synthesized oligonucleotides and on-chip polymerization. Then,
by simply tagging the distal end, SSFM can be used to confirm
the array completion because the fluorescent label for the fully
self hybridized DNA will go down to the surface.
Little is known about the conformation of ssDNAs, and there
is no reliable model for predicting the shape and diameter of the
random coil form (12). In the absence of base-pairing, long
ssDNA has been considered as a flexible polymeric molecule
Using SSFM, we estimated the shape of single-stranded pieces
of DNA 21 and 50 nt long. In our experiments, energy transfer
and quenching have no effect due to intensity insensitivity.
Therefore, SSFM can be extended to multiple tags in different
Materials and Methods
Materials. All oligonucleotides were custom-synthesized by Inte-
grated DNA Technologies (Coralville, IA). The following chemi-
cals were used: acetone (HPLC grade; Merck); 3-aminopropyltri-
ethoxysilane (98%; Aldrich), 1,4-phenylene diisothiocyanate
(?98%; Fluka), and dimethyl sulfoxide (anhydrous 99.9%; Al-
drich). TE buffer was composed of 10 mM Tris?HCl and 1 mM
of silicon oxide grown by plasma-enhanced chemical vapor depo-
sition and polished chemo-mechanically to a roughness of ?2 nm.
The relatively thick layer of silicon oxide serves as a transparent
end and covalently bound on the surface (a, top), and after hybridization with complementary (nonlabeled) strand (b).
www.pnas.org?cgi?doi?10.1073?pnas.0511214103Moiseev et al.
silicon. At larger distances, even a small shift in wavelength leads
to sharp changes in the intensity of emitted light. The result was a
shorter distance between interference peaks, about four to five
peaks in the typical emission spectrum of an organic dye located
10–15 wavelengths above the mirror (20).
Covalent Attachment of DNA to the Surface. Wafers were cut into
15 ? 5-mm chips. The chips were washed with acetone, sonicated
in water for 10 min, and cleaned with 10% NaOH for 10 min. The
treatment with alkali etched ?1 nm of silicon oxide without
deionized water, the chips were blown dry, treated with 5%
aminopropyltriethoxysilane in acetone for 2 min, washed a few
Longer silanation times cause polymerization of silane and depo-
sition of multiple silane layers, which creates a nonuniform surface
and is not suitable for our experiments (27). Aminated chips were
functionalized with a homobifunctional crosslinker (1 mg?ml phe-
nylene diisothiocyanate in DMSO, 1-h reaction under argon while
stirring), washed a few times with DMSO, and rinsed with water.
The chips were then immediately covered with a 10 ?M solution of
amino-labeled oligonucleotide in 1 M potassium phosphate (pH
8.0) and left for 1 h on a shaker. After washing three times with TE
measurements or hybridized with a complementary oligo (10 ?M
in 1 M NaCl?TE, pH 7.0). After hybridization, the chips were
washed again three times in the same buffer. The 21-mer oligonu-
cleotide sequence was 5?-GAA TTC GAG CTC GGT ACC
CGG-3?; the 50-mer oligonucleotide sequence was 5?-GAA TTC
GAG CTC GGT ACC CGG GGA TCC TCT AGA GTC GAC
CTG CAG GCA TG-3?. We designed and checked the oligos to
have no self-complementarity. The sequences were taken from a
polylinker region of a popular cloning vector. For experiments with
ssDNA, the oligos were carrying an amino group with a C6 spacer
on the 5? end and an Oregon green 488 label on the 3? end. For
experiments with dsDNA, these had only the amino group. The
the 5? or the 3? end.
Acquisition and Processing of Spectra. The spectra were obtained
with a Renishaw 1000B micro-Raman spectrometer coupled
with a Leica DM?LM upright microscope. A low-numerical-
aperture objective (?5, 0.12 numerical aperture) was used to
minimize the collection cone. An 1,800-grooves-per-mm grating
was used with a spectral resolution of 2 cm?1at 500 nm. For WL
measurements, normal Koehler illumination with a standard
halogen lamp was used. The light source for fluorescent mea-
surements was the 488-nm line of an argon ion laser. Both WL
reflectivity and fluorescence self-interference spectra were fit-
ted by using a custom-built MATLAB application that separates
the oscillatory component from the envelope function. This
program automatically calculates the parameters of the system
such as the thickness of thin films or position of the emitters
above the mirror. There is a noticeable variation in the index of
refraction of silicon oxide within the wavelength span we used.
This variation was taken into account in the fitting algorithm.
The classical model of fluorescence self-interference is de-
scribed in detail in ref. 28. The essence of the method is that
fluorescent emission near reflecting surfaces is modified by the
interference between the direct and the reflected waves. The
position of the emitter above the mirror has a direct effect on
the phase of the resulting oscillatory component and can be
deduced from the emission spectra. The model also takes into
account the complex reflectivity of the underlying stack of
dielectrics and the orientation of the dipole moments of the
emitters. Because the curve-fitting algorithm extracts the oscil-
latory term to determine the label height, our measurements are
immune not only to potential quenching of the entire spectrum,
but also to any spectral modifications or nonradiative transfer
effects, because such effects would result in a similar oscillatory
behavior of the spectra.
WL reflectivity measures the optical thickness of the trans-
parent material on top of a mirror. It is a process similar to
ellipsometry, except that it works by fitting the spectra of WL
reflected from the surface and modified by thin-film interfer-
ence (20). Measurements were taken from 10 to 20 spots on a
chip, usually in a linear pattern. The spot size was determined to
be on the order of 10–20 ?m.
Quantitation of the Density of Immobilized DNA. DNA oligonucle-
otides were radiolabeled at the 3? end with [?-32P]ddATP (Amer-
sham Pharmacia) by using terminal transferase (Roche). Labeled
oligonucleotides were purified on a gel-filtration column packed
with Sephadex G-50 gel-filtration resin. DNA immobilization was
carried out the same way as for fluorescently labeled oligos. The
chips were dried, and the amount of radioactive material on the
surface was measured in a scintillation counter.
Nonspecific binding was checked by using radiolabeled oligo-
small, ? ?1 fmol?mm2.
We thank A. Yo ¨net for help with graphical visualization. This work was
supported by National Science Foundation Grant DBI0138425, Air
Force Office of Scientific Research Grant MURI F-49620-03-1-0379,
and National Institutes of Health–National Institute of Biomedical
Imaging and BioEngineering Grant 5R01 EB00 756-03.
Table 1. Average height of the fluorescent marker above the
surface as measured by fluorescence interference (SSFM) and
the optical thickness of the DNA layer on the surface (measured
by WL reflectance)
In all experiments we start with a ssDNA covalently bound to the surface
(shown on the left in red). In one set of experiments, the first oligonucleotide
(either 50- or 21-mer) is covalently bound to the surface and does not have a
fluorescent label. In this case, only WL measurements can be performed prior
to hybridization. After hybridization with a fluorescently labeled matching
strand (shown on the right in blue), SSFM yields the average fluorophore
second set of experiments, covalently bound oligonucleotide (either 50-or
21-mer) is fluorescently labeled at the distal end and SSFM yields the average
the DNA layer now consists of two species: the dsDNA and the unhybridized
strands both of which are carrying a fluorescent marker, but expected at
different heights. SSFM measures the average height of the fluorophores.
Experiments with 50-mers also include when the first strand is annealed with
a 21-mer complementary to either its top or bottom part as schematically
the accuracy of a single height measurement is ?0.2 nm, the variation be-
tween different measurements is ?10% of the nominal values.
Moiseev et al.
February 21, 2006 ?
vol. 103 ?
no. 8 ?
1. Shchepinov, M. S., Case-Green, S. C. & Southern, E. M. (1997) Nucleic Acids Download full-text
Res. 25, 1155–1161.
2. Southern, E., Mir, K. & Shchepinov, M. (1999) Nat. Genet. 21, 5–9.
3. Mir, K. U. & Southern, E. M. (1999) Nat. Biotechnol. 17, 788–792.
4. Vainrub, A. & Pettitt, B. M. (2002) Phys. Rev. E 66, 041905.
5. Vainrub, A. & Pettitt, B. M. (2003) Biopolymers 68, 265–270.
6. Wong, K.-Y. & Pettitt, B. M. (2001) Theor. Chem. Acc. 106, 233–235.
7. Chrisey, L. A., Lee, G. U. & O’Ferrall, C. E. (1996) Nucleic Acids Res. 24,
8. Gray, D. E., Case-Green, S. C., Fell, T. S., Dobson, P. J. & Southern, E. M.
(1997) Langmuir 13, 2833–2842.
9. Levicky, R., Herne, T. M., Tarlov, M. J. & Satija, S. K. (1998) J. Am. Chem. Soc.
10. Herne, T. M. & Tarlov, M. J. (1997) J. Am. Chem. Soc. 119, 8916–8920.
11. Charreyre, M.-T., Tcherkasskaya, O. & Winnik, M. A. (1997) Langmuir 13,
12. Murphy, M. C., Rasnik, I., Cheng, W., Lohman, T. M. & Ha, T. (2004) Biophys.
J. 86, 2530–2537.
13. Peterlinz, K. A., Georgiadis, R. M., Herne, T. M. & Tarlov, M. J. (1997) J. Am.
Chem. Soc. 119, 3401–3402.
14. Wolf, L. K., Gao, Y. & Georgiadis, R. M. (2004) Langmuir 20, 3357–3361.
15. Kelley, S. O., Barton, J. K., Jackson, N. M., McPherson, L. D., Potter, A. B.,
Spain, E. M., Allen, M. J. & Hill, M. G. (1998) Langmuir 14, 6781–6784.
16. Shlyakhtenko, L. S., Gall, A. A., Weimer, J. J., Hawn, D. D. & Lyubchenko,
Y. L. (1999) Biophys. J. 77, 568–576.
18. Mo ¨ller, R., Csa ´ki, A., Ko ¨hler, J. M. & Fritzsche, W. (2000) Nucleic Acids Res.
19. Jenison, R., Yang, S., Haeberli, A. & Polisky, B. (2001) Nat. Biotechnol. 19,
20. Swan, A. K., Moiseev, L., Cantor, C. R., Davis, B., Ippolito, S. B., Karl, W. C.,
Goldberg, B. B. & U ¨nlu ¨, M. S. (2003) IEEE J. Select. Top. Quantum Electron.
21. Braun, D. & Fromherz, P. (1998) Phys. Rev. Lett. 81, 5241–5244.
22. Parthasarathy, R. & Groves, J. T. (2004) Cell Biochem. Biophys. 41,
23. Du, Q., Vologodskaia, M., Kuhn, H., Frank-Kamenetskii, M. & Volodogskii,
A. (2005) Biophys. J. 88, 4137–4145.
24. Elhadj, S., Singh, G. & Saraf, R. F. (2004) Langmuir 20, 5539–5543.
25. Steel, A. B., Levicky, R. L., Herne, T. M. & Tarlov, M. J. (2000) Biophys. J. 79,
26. Zhang, Y., Zhou, H. & Ou-Yang, Z. C. (2001) Biophys. J. 81, 1133–1143.
27. Yoshida, W., Castro, R. P., Jou, J.-D. & Cohen, Y. (2001) Langmuir 17,
28. Moiseev, L., Cantor, C. R., Aksun, I., Dogan, M., Goldberg, B. B., Swan, A. K.
& U ¨nlu ¨, M. S. (2004) J. Appl. Phys. 96, 5311–5315.
www.pnas.org?cgi?doi?10.1073?pnas.0511214103Moiseev et al.