Dynamic Article LinksC<Nanoscale

Cite this: Nanoscale, 2012, 4, 1734

www.rsc.org/nanoscale PAPER

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2FView Article Online / Journal Homepage / Table of Contents for this issue

Hybridization in nanostructured DNA monolayers probed by AFM: theoryversus experiment†

Alessandro Bosco,a Fouzia Bano,ab Pietro Parisse,c Loredana Casalis,*c Antonio DeSimonead

and Cristian Michelettiad

Received 4th November 2011, Accepted 16th December 2011

DOI: 10.1039/c2nr11662f

Nanografted monolayers (NAMs) of DNA show novel physico-chemical properties that make them

ideally suited for advanced biosensing applications. In comparison with alternative solid-phase

techniques for diagnostic DNA detection, NAMs have the advantage of combining a small size with

a high homogeneity of the DNA surface coverage. These two properties favour the extreme

miniaturization and ultrasensitivity in high-throughput biosensing devices. The systematic use of

NAMs for quantitative DNA (and protein) detection has so far suffered from the lack of a control on

key fabrication parameters, such as the ss- or ds-DNA surface coverage. Here we report on a combined

experimental–computational study that allows us to estimate the surface density of the grafted DNA by

analyzing the sample mechanical response, that is the DNA patch height vs. applied tip load curves. It is

shown that the same analysis scheme can be used to detect the occurrence of hybridization with

complementary strands in solution and estimate its efficiency. Thanks to these quantitative

relationships it is possible to use a single AFM-based setup to: (i) fabricate a DNA NAM, (ii) control

the DNA surface coverage, and (iii) characterize its level of hybridization helping the design of NAMs

with pre-determined fabrication parameters.

Introduction

Gene expression microarrays and DNA-barcode protein micro-

arrays are powerful tools for disease identification.1–3 The unique

biomolecular recognition properties of DNA, based on Watson–

Crick base-pairing, have been exploited to immobilize structures

of mono- or poly-dispersed DNA filaments on solid surfaces.4,5

Such molecular structures can be used to detect complementary

DNA/RNA strands in solution and/or as building blocks to bind

antibody (proteins or enzymes)–DNA conjugates in bio-affinity

assays for novel applications, ranging from the study of protein

networks, protein correlations and pathways to monitor the

progress of diseases, to novel biomarker discovery and to the

identification of new targets for therapeutics. The need for early

and timely medical diagnosis and the quest for multiplexing has

aScuola Internazionale Superiore di Studi Avanzati (SISSA), ViaBonomea 265, I-34136 Trieste, ItalybNanoChemistry and Molecular Systems Department of Chemistry,University of Li�ege, B6a Sart-Tilman, B-4000 Li�ege, BelgiumcSincrotrone Trieste S.C.p.A., Strada Statale 14—km 163,5 in AREAScience Park, I-34149 Basovizza, Trieste, Italy. E-mail: loredana.casalis@elettra.trieste.itdCNR-IOM Democritos and Italian Institute of Technology (SISSA unit)

† Electronic supplementary information (ESI) available: Experimentalprocedures, computational details and comparison betweenexperimental and computational data for the H–L response of ssDNANAMs after hybridization. See DOI: 10.1039/c2nr11662f

1734 | Nanoscale, 2012, 4, 1734–1741

further spurred the development of miniaturized, DNA nano-

scale arrays with minimal sample volume and high sensitivity

requirements. Early applications of these techniques range from

immunological nano-assays6 to the addressing of multi-enzy-

matic catalytic reactions.7 In all these DNA-based devices, it is

crucial to optimize, at each step of the biosensor development,

the immobilization protocol, to control probe density and, to

a lesser extent, the physiological conditions (i.e. ionic strength of

the working solution and screening of surface charges) in order

to achieve the highest reproducibility and to optimize the sensi-

tivity of the device. Low DNA densities, corresponding to few

hundreds of molecules per 0.1 micron square, in fact, can result

in poor signal-to-noise ratio, while a too high density can steri-

cally hinder the complementary strand or the DNA–protein

conjugates from entering the nano-structure and hybridizing

with the complementary probes.

These observations motivate the present study where we

investigate surface-confined, nanografted assembled monolayers

(NAMs) of DNA to exploit their unique properties to establish

reliable protocols of higher reproducibility. Nanografting is an

AFM-assisted lithography technique where the AFM tip, oper-

ated at relatively high load, catalyzes, for instance, the replace-

ment of the molecules of a given self-assembled monolayer

(SAM) with thiolated DNA molecules present in solution. By

tuning nanografting parameters such as the concentration of the

molecules in solution, DNA nanostructures with different

This journal is ª The Royal Society of Chemistry 2012

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2F

View Article Online

surface densities can be obtained. It is well known that by using

nanografting the fabricated DNA nanostructures are endowed

with novel physicochemical properties.8,9 Mirmomtaz et al.,11

based on AFM topography height measurements have shown

that in contrast with typical SAMs10 DNANAMs show a higher

homogeneity surface coverage, due to the reduced number of

entanglements between different ssDNA chains in the patch.11–13

Consequently, DNA NAMs are ideally suited for DNA sensing

at the nanoscale.

The reduced size of NAM samples, while having a great

potential for use in high sensitivity–high throughput contexts,

adds the challenge of measuring and controlling the surface

density of the grafted DNA and its hybridization efficiency, since

a number of well-established techniques, such as surface-plas-

mon resonance or electrochemical techniques,14–18 are best suited

for DNA samples of much larger size. Mirmomtaz et al.11 have

demonstrated, through AFM differential height measurements,

the possibility to determine whether ssDNAmolecules in a NAM

undergo (at least partial) hybridization, since the flexibility (i.e.

the persistent length) of DNA strands changes dramatically from

the single to double strand configuration. By means of the sole

topographic height, however, Mirmomtaz et al. could not esti-

mate the surface coverage of DNA patches before and after

hybridization, and therefore the efficiency of the hybridization

process remained uncharacterized.

In this study we demonstrate that these difficulties can be

overcome by suitably comparing the experimentally measured

height vs. applied load relationship against the one established

theoretically and computationally.

Prior to the application to the challenging case of DNA

NAMs, our approach is validated on the well-characterized case

of DNA SAMs, demonstrating that with the sole input of the

model response (with no externally calibrated parameters) we

can infer the correct density with an uncertainty of less than 20%.

The same approach is used to obtain bounds for the hybridiza-

tion efficiency of ssDNA patches with different surface densities.

The strategy devised here allows for optimizing by design the

experimental conditions for optimal biorecognition detection.

We can in fact profitably use our theoretical/computational

model to single out the range of patch densities and forces where

the discriminatory power of AFM loading of ss and dsDNA

patches is highest.

Our approach can be adopted also to estimate and optimize

the density of DNAmatrices for optimal confinement of protein/

enzyme diffusion in nanofluidic channels19 or for the optimiza-

tion of densities in DNA origami for enzyme immobilization,

drug delivery and, more generally, for the nanotechnological self-

assembly of novel materials.

Experimental section

DNA NAMs of different densities were obtained by nano-

grafting thiol-modified ssDNA with 24 base-pairs (for DNA

sequence, see ESI†) within a monolayer of ethylene-glycol-

terminated alkylthiols (HS–(CH2)11–(OCH2CH2)3–OH) on a flat

gold substrate. The scanning of the SAM area by the AFM tip

has an ordering and disentangling effect on the grafted molecules

through a ‘‘combing action’’. By operating on two key fabrica-

tion parameters, primarily the number of times that the AFM tip

This journal is ª The Royal Society of Chemistry 2012

overwrites the same area, a parameter later on referred to as S/A

(Scanned area/Actual area, see ESI† for the proper definition),

the concentration of DNA and monovalent counterions in

solution, it is possible to modulate the ssDNA surface density in

a consistent manner across various samples. Consequently, the

repeated overwriting of a NAM during grafting results in an

increased density and an improved resistance of the patch

molecules to the tip load. As the patch height increases it

becomes increasingly difficult to discriminate the height of the

patch before and after hybridization.11

Theoretical and computational section

Schematic models for NAMs of short (<100 bp) dsDNA

segments have recently been used for an order-of-magnitude

estimate of the probe density.20,21 In these models, the configu-

ration space accessible to each dsDNA molecule consists of

a cone with the apex anchored to the surface and the axis

perpendicular to the surface. The model captures basic steric

hindrance effects between the dsDNA chains and is amenable to

a straightforward analytical treatment. However, its highly

simplified nature makes it inappropriate to capture the patch

height vs. applied load (hereafter indicated as H–L) response of

a dsDNA patch and is unsuitable for application to flexible

molecules, such as ssDNA.

We accordingly chose to employ a more realistic coarse-

grained model of DNA with the purpose of characterizing

numerically the equilibrium response to the mechanical H–L

relationship of both ssDNA and dsDNA, and to use it for

a quantitative characterization of how the surface density of

a DNA patch impacts on the patch height and H–L response.

The comparison of the mechanical response of the model with

the experimentally measured one can therefore be used to esti-

mate two quantities that are typically difficult to measure directly

in experiments, namely the patch surface density and the sample

hybridization efficiency.

To gain the maximum possible insight from the theoretical/

experimental comparison, we opted for a simplified coarse-

grained description of our system. As is typically the case in

dense biopolymeric systems,22,23 the complexity of the problem

stems from the presence of several length scales: the contour

length of the grafted molecules, their persistence length, the

typical distance between the anchors at the surface, the Debye

screening length and the range of dehydration interactions. To

maintain the system at a minimal level of complexity, we per-

formed experiments in a regime where hydration forces24 and

electrostatic effects are negligible. This is achieved by working at

very low surface densities of DNA NAMs, and high salt

concentration during AFM imaging.

The ssDNA and dsDNA molecules were modeled as discrete

chains with appropriate thickness and bending rigidity. A chain

was represented through its backbone consisting of n segments of

equal length, each segment corresponding to a nucleotide (for

ssDNA) or a base-pair (for dsDNA). In both cases, the segment

length was set to 0.34 nm, which corresponds to the canonical

base-pair separation in B-form dsDNA,25 as well as the axial rise

of ssDNA between two consecutive nucleotides.26 The excluded

volume effects were taken into account by considering the

backbone as being the centerline of a discrete thick chain having

Nanoscale, 2012, 4, 1734–1741 | 1735

Fig. 1 Schematic representation of a 3 � 3 patch with a periodic

boundary condition with a focus on the constraints applied to the chains.

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2F

View Article Online

a cross-sectional radius D (see Fig. 1). The value of D was set

equal to 0.45 nm for ssDNA and to 1.25 nm for dsDNA.27–29

Following the scheme of Gonzalez and Maddocks,30 the finite

chain thickness,D, impacts on the smallest attainable local radius

of curvature of the chain (which has to be greater than D) and on

the distance of the minimum approach of any two non-consec-

utive backbone segments, which cannot be smaller than 2D. The

restrictions imposed by the steric hindrance on the local radius of

curvature are sufficient to endow the ssDNA with a persistence

length31 of about 2 nm (see ESI†), which is compatible with the

experimental value.32 Accordingly, no explicit bending rigidity

term was included in the model ssDNA chains. Conversely, chain

thickness alone is not sufficient to account for the full persistence

length of dsDNA33 and therefore a bending rigidity term,

appropriate to reproduce the persistence length of 50 nm in

a discrete piece-wise linear chain (i.e. a Kratky–Porod chain),

was introduced for this second type of molecule.33

In the experimental setup, the ssDNA and dsDNA molecules

are attached to the surface by means of 1.2 nm long flexible

alkanethiol spacers. Consistent with this fact, the backbones of

the model DNA chains were prolonged at one of the ends by

a second piece-wise linear chain consisting of 3 segments of

length 0.4 nm and thickness equal to 0.45 nm. Given the high

flexibility of the alkanethiol spacer, no bending rigidity term was

introduced for the model linker. The DNA + linker constructs

were anchored to the surface in a regular hexagonal lattice

pattern, which is the natural reference arrangement in the

absence of disorder. To keep the computational expenditure to

a manageable size, only a ‘‘tile’’ of the lattice comprising N � N

chains (where N is between 4 and 6) was considered and the

behaviour of an extended sample was mimicked by introducing

periodic boundary conditions.

An order-of-magnitude estimate of the surface densities that

are expected to be observed in DNA NAMs is provided by the

maximum surface density reported for dsDNA and ssDNA

SAMs,14,34,35 which are of about 3� 1012 and 12� 1012 molecules

per cm2, respectively. This density corresponds to a hexagonal

lattice spacing,36 a, equal to 6.2 nm for dsDNA and 3.1 nm for

ssDNA, respectively. Based on these observations, the simula-

tions (see later) were performed for values of a ranging from

3.4 nm to 20 nm for the dsDNA and from 1.7 nm to 20 nm for the

ssDNA. Note that at the shortest considered lattice spacing, the

DNA interstrand separation in a patch of upstanding parallel

DNA molecules is larger than the distance at which dehydration

effects are expected to significantly affect the DNA-self-interac-

tion energy. The separation is also larger than the Debye–

Hueckel screening length appropriate for the high salt conditions

(1 M NaCl) at which our experiments are carried out. For these

1736 | Nanoscale, 2012, 4, 1734–1741

reasons we decided to neglect these interactions in our model. We

notice, however, that at interhelical distances less than �5 nm,

cholesteric self-interaction between dsDNA segments may start

playing a role.37 We have neglected them in the present study for

the sake of simplicity and we plan to address these effects in the

future, using more refined models.

Finally, the surface was treated as an impenetrable plane and

an attractive interaction was introduced between the surface and

each vertex of the chain segments at a distance smaller than 2 nm

from the surface. The interaction strength was set equal to

0.2 kBT per bead in the case of ssDNA and equal to 0.4 kBT for

a dsDNA bead and the range is set to 2 nm. These values of the

range and strength of the attractive potential are chosen empir-

ically and are consistent with the Gouy–Chapman potentials for

a unit electric charge with a moderately polarized gold surface at

high concentrations of counterions.38 Concerning the spacer, the

interaction strength was set equal to 0.2 kBT for each bead.

Formally, the total potential energy of a patch of N chains,

subject to a compressive force, f, consists of the following terms:

H ¼XI

Hbr

I þHsaI þH surface

I þHextI þ

XJ.I

HsaI ;J

!(1)

where I and J are indices running over the chains in patch so that

single and pairwise chain contributions are denoted by subscripts

I and I,J, respectively. The superscripts br, sa and surface refer

respectively to the bending-rigidity, the excluded-volume (self-

avoidance) interaction and the attraction of the impenetrable

surface.Hext contains the effect of the load of the tip on DNA (see

ESI† for details). This last compressive term is modeled through

the action of an external force:

Hext ¼ fext max(ri) (2)

where fext is the force per chain, applied to the bead at the largest

height in each chain of the patch. The equilibrium properties of

the model were computed by means of Monte Carlo simulations.

An elementary step of the Monte Carlo evolution consisted of

modifying the configuration of each DNA chain in the patch by

means of unrestricted crankshaft or pivot moves (with equal

probability for the two types of moves). The newly generated

configurations were accepted/rejected with the standard

Metropolis criterion. For each considered value of the hexagonal

lattice spacing and applied compressive force, the Monte Carlo

simulations were started from the configurations with all the

DNA strands in the upright position. From each simulation we

collected at least 50 000 configurations picked at time intervals

greater than the autocorrelation time of the mean sample height.

Results and discussion

The possibility to rely on nano-mechanical probes to detect and

characterize the hybridization of ssDNA NAMs builds on the

very different elastic properties of ss- and ds-DNA. Because of

their different thicknesses and charge densities, the persistence

lengths (a measure of the bending rigidity) of these two types of

molecules differ by more than one order of magnitude. For

ssDNA, the persistence length is of the order of 2 nm, corre-

sponding to about 7 nucleotides, while for dsDNA it is of 50 nm,

corresponding to 150 base-pairs.39Consequently, a pristine patch

This journal is ª The Royal Society of Chemistry 2012

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2F

View Article Online

of surface-immobilized ssDNA chains, each consisting of several

tens of nucleotides, will yield under mechanical compression

much more than after hybridization with complementary ssDNA

strands present in solution.

These effects, recently surveyed by Mirmomtaz et al.,11 are

investigated here in detail and as a function of the key fabri-

cation parameters: the S/A number and the concentration of

DNA and counterions in solution. The general phenomeno-

logical aspects are aptly illustrated in Fig. 2. In particular, in

panel (a) a 1 mm � 1 mm patch fabricated at a DNA concen-

tration of 1 mM and S/A ¼ 2.5 is shown. The corresponding

height histogram (Fig. 2d, black open circles) revealed a relative

height of about 1.8 � 0.3 nm with respect to the surrounding

carpet of TOEG3 molecules. The absolute height of the DNA

from the gold surface can be obtained by adding the height of

the TOEG3 carpet (1.8 � 0.1 nm) to the relative height

measured by AFM, resulting in 3.6 � 0.3 nm for the patch

shown in Fig. 2 panel (a). Hereafter we will always refer to the

absolute heights of DNA from the gold surface. A ssDNA

NAM grafted using the same conditions as the one in Fig. 2a,

Fig. 2 AFM topographic images of 24 bp ssDNA (a and e), hybridized

ssDNA (b and f) and dsDNA (c and g) NAMs that were grafted within a

monolayer of OEG-terminated alkylthiols on gold films, at the conditions

described above. Height histograms, corresponding to NAMs grafted at

1 mM (d) and 2 mM (h) DNA concentrations, show the clear shift in height

values as a result of variable grafting densities of DNA molecules and

hybridization reactions with complementary strand (black circles corre-

spond to ssDNA, red circles to dsDNA and blue circles to hybridized

ssDNA). The absolute height of the DNAmatrices (HDNA) is determined

by AFM imaging at low forces and can be derived from the measured

relative height (reported in the histograms) of DNA (DH) adding the

thickness of the TOEG3 layer (HTOEG3): HDNA ¼ DH + HTOEG3.

This journal is ª The Royal Society of Chemistry 2012

and afterwards hybridized with the fully complementary strand,

is shown in Fig. 2b. The resulting height histogram (Fig. 2d,

blue open circles), shows an increase in absolute height from

about 3.6 � 0.3 nm (ssDNA NAM) to about 4.3 � 0.3 nm: this

is associated with the increased rigidity of the system, and is

therefore an indicator of successful hybridization. In Fig. 2c,

a patch of grafted dsDNA duplexes, hybridized in solution, is

reported. The patch height is of 4.3 � 0.3 nm, as for the case of

surface-hybridized ssDNA NAMs. The associated height

histogram is plotted in Fig. 2d (red open circles). When NAMs

are grafted at higher DNA concentration in solution (2 mM),

keeping all the other fabrication parameters constant (e.g.

S/A ¼ 2.5), the height of the patches increases. In fact, the

higher is the DNA concentration in solution, the higher is the

efficiency of the nanografting process and hence a higher surface

density of the NAM is expected. This will result in an increase

of the height in the dense patch since the steric hindrance of

neighboring chains provides an enhanced support to stand up

compared to a less dense patch. Fig. 2e–g show NAMs of

ssDNA, surface-hybridized ssDNA and solution-hybridized,

grafted dsDNA, respectively. The absolute heights calculated

from the height histogram (shown in Fig. 2h) are 4.7 � 0.2 nm,

7.3 � 0.2 nm and 7.0 � 0.4 nm, respectively.

The ‘‘zero-force’’ height measurements in Fig. 2 are clearly

compatible with the expected higher mechanical rigidity of

dsNAMs compared to ssNAMs. The different response in

compression of the two types of patches is, however, best char-

acterized by means of AFM height versus applied load

measurements. The latter entail the recording of AFM topog-

raphy images (from which differential height profiles are

extracted) at different loads applied by the AFM tip so as to

establish the height versus applied load curves. The absolute

heights are plotted in Fig. 3 as a function of tip load. The main

findings are the following: (i) the different mechanical response

of ssDNA NAMs (black circles) with respect to surface-hybrid-

ized ssDNA (blue circles) and dsDNA NAMs (red circles), at

both grafting concentrations and (ii) the lack of any significant

difference in the behavior of solution- and surface-hybridized

grafted strands.

We emphasize that the height measurements shown in Fig. 3

are averaged over repeated experiments carried out with different

DNA patches (clearly prepared with the same protocol) and

AFM tips. The reproducibility and robustness of the height

measurements collected with different tips reflect the fact that,

independently of the initial sharpness, the patch scanning

rounds AFM tips to an average apical radius of curvature of

about Rc ¼ 25 nm (see ESI†).

The experimental findings were next compared with the

numerical ones. From the extensively sampled configurations of

model DNA patches we computed the mean patch height (i.e. the

mean height of all the highest beads of each chain in the patch) as

a function of the applied force and of the surface density of

ssDNA or dsDNA. The data are summarized in Fig. 4.

Throughout the explored range of force and density values,

both the ssDNA and dsDNA data are well interpolated by the

following expression:

h ¼ a0 þ a1

ða2 þ fextÞ2(3)

Nanoscale, 2012, 4, 1734–1741 | 1737

Fig. 3 Absolute heights of DNA NAMs recorded as a function of the

applied load by AFM tip. (a) A comparison between ssDNA (open

symbols) and dsDNA (filled symbols) NAMs that were prepared at DNA

concentrations of 1 mM (circles) and 2 mM (triangles). (b) Height vs.

applied load curves of hybridized-ssDNA (half filled symbols) NAMs are

compared to the same curves of ssDNA NAMs that are reported in (a).

Fig. 4 Representation of the manifolds of mean patch height vs. surface

density and force per chain, obtained by the fitting of simulation results for ss-

(a) and ds-DNA (b). The parameters obtained by the fit with eqn (3) are a00¼0.940, a01 ¼ 0.045, a10 ¼ 263, a11 ¼ 0.997, a20 ¼ 11.0, and a21 ¼ 0.005 for

ssDNAand a00¼ 0.658, a01¼ 0.544, a10¼ 75.9, a11¼ 6.87, a20¼ 4.36, and a21¼ 0.111 fordsDNA.Theunitsof thecoefficientsare implicitlydefined toreturn

h in nm if fext and s are expressed in pN and 1012 molecules cm�2, respectively.

1738 | Nanoscale, 2012, 4, 1734–1741

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2F

View Article Online

yielding h, the height of the patch as a function of the applied

force fext and the surface density s through the following para-

metric dependence of a0, a1 and a2:

a0 ¼ a00 þ a01s;a1 ¼ a10 þ a11s

2;a2 ¼ a20 þ a21s

2:(4)

The numerical values of the parametric coefficients are

provided in the caption of Fig. 4 along with an illustration of the

manifolds describing the mean patch height versus the patch

density and applied force per chain in ssDNA and dsDNA

NAMs. Notice that, at fixed surface density, the functional form

in eqn (3) ensures that the patch height has a Hookean (linear)

relationship with force at small loads and reaches a plateau at

high force.

The height vs. force results obtained from the model cannot be

directly compared with experimental measurements because they

pertain to a uniformly pressed DNA patch and hence do not

account for the tip geometry. To include the latter effect, the

model height–force curves were suitably reweighted so as to

reproduce the compressive action of a paraboloid tip40–42 with an

apical radius of curvature, Rc ¼ 25 nm, as per SEM measure-

ments. The sought force–height relationship for the paraboloid

tip is (see ESI† for details):

Ftip¼pRcs2a1

�a0 � htipþ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�htip�a0

�ðh0�a0Þq �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia1�htip � a0

�q þ a2�htip�h0

�0B@

1CA(5)

where htip is the height of the apex of the paraboloid tip and h0 is

the height of the patch obtained from eqn (3) at fext ¼ 0. Notice

that eqn (5) requires that the tip height, htip, is larger than a0which is the asymptotic sample height for a very large applied

load. The previous expression can be used to fit a set of experi-

mental force–height measurements (see ESI†) so as to recover the

patch density, s. We stress that the latter is the only free

parameter entering eqn (5).

To validate our procedure, we compared the theoretical results

with AFM height vs. applied load measurements carried out on

DNA SAMs. Several experimental determinations of the SAMs

surface density have been carried out under a variety of condi-

tions. In particular, for saturated SAMs consisting of thiolated

DNA molecules of 24 nucleotides or base-pairs, the surface

density of ssDNA and dsDNA patches probed with surface-

plasmon resonance (SPR) was found to be equal to 12 � 1012

molecules cm�2 and 3 � 1012 molecules cm�2, respectively.14

We accordingly performed AFM height vs. applied load

measurements on a saturated, 24 bp-long thiolated ssDNA SAM

prepared according to the protocol of ref. 14; based on the

above-mentioned experiments, the expected surface-density of

this patch is 12 � 1012 molecules cm�2. The height of the DNA

SAM was measured relative to a OEG NAM grafted into it and

the curve was obtained by recording the height at different tip

loads, see Fig. 5. As is shown in the same figure, the experimental

data are well fitted by the force–height curve of eqn (5) derived

from the coarse-grained model after an optimal choice (fit) of the

model patch density. The surface density returned by the fit

procedure was equal to 13 � 2.5 � 1012 molecules cm�2 which is

This journal is ª The Royal Society of Chemistry 2012

Fig. 5 Upper panel: AFM micrograph of a TOEG3 patch nanografted

into a high density ssDNA SAM with corresponding height histogram

and schematics (scale bar ¼ 200 nm). Lower panel: comparison between

experimental (symbols) and computational data (lines) for height vs.

applied load for a high density ssDNA SAM.

Fig. 6 Comparison between the experimental data (symbols) and

theoretical fits (lines) of patch height vs. applied load. Black open

symbols stand for ssDNA data, while red filled symbols stand for

dsDNA. The heights of the patches nanografted at 1 mM and 2 mM

concentrations of ss- and ds-DNA in solution are represented in (a) and

(b), respectively.

Table 1 Estimated surface density values for the ssDNA and dsDNAgrafted at 1 and 2 mM concentrations

NAM [DNA] ¼ 1 mM [DNA] ¼ 2 mM

ssDNA 13.1 � 1.1 � 1012

molecules cm�2

17.7 � 1.7 � 1012

molecules cm�2

dsDNA 3.8 � 0.3 � 1012

molecules cm�2

6.2 � 0.3 � 1012

molecules cm�2

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2F

View Article Online

in very good agreement with the nominal SPR values14 and hence

provides a successful validation of the computational approach.

The model parametric curve of eqn (5) was next used to fit the

height vs. applied load measurements collected for ssDNA and

dsDNANAMs. The experimental data were all well-fitted by the

model curve, as visible in Fig. 6. The optimal values of patch

surface density resulting from this fitting procedure are shown in

Table 1. The data pertain to ssDNA and dsDNA patches grafted

at 1 and 2 mM concentrations.

It is interesting to notice that the fitted values for the surface

densities of the NAM patches are higher compared to those

found in the SAM counterparts.14,43,44

This notable result can be tentatively rationalised in terms of

the ‘‘combing’’ action of the AFM tip during the nanografting

process. Because of high grafting force values (tens of nN) the

tip action expectedly favors the local annealing of the surface

thus modifying the kinetics of SAM formation and reducing

molecular entanglement. The latter promotes, in turn, the

ordering of the molecular layer and hence its higher surface

density.11 It is worth noting here that the maximum theoretical

surface density that is achievable by ssDNA, if modeled as

a chain with diameter equal to 0.9 nm, is of about 1.4 � 1014

molecules cm�2, which is well above the estimated density of the

DNA NAMs considered in this study. Therefore it is plausible

that with NAMs one could span a wide range of surface

This journal is ª The Royal Society of Chemistry 2012

densities even beyond the limiting values reported for SAMs,

both for ssDNA and dsDNA.

Furthermore, we point out that the modeled mechanical

response of ssDNA and dsDNA patches can be used to estimate

the hybridization efficiency of ssDNA NAMs.

A proper understanding of the hybridization mechanism in

such dense systems is obviously crucial to the use of our nano-

grafting-based nano-arrays for quantitative assays.45–47

Nanoscale, 2012, 4, 1734–1741 | 1739

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2F

View Article Online

It can be argued that in a partially hybridized patch, the

molecules that bear the compression exerted by a mild AFM load

are essentially only the hybridized ones. The double stranded

nature of these molecules, in fact, causes them to stand out of the

background height of the non-hybridised DNA strands.

Neglecting the presence of single-stranded molecules, the density

of the hybridized patch can be obtained by fitting the experi-

mental height vs. applied load curve with the one of a pure

dsDNA patch. This value clearly overestimates the true

density of the double-stranded hybridized molecules because it

effectively ascribes to the latter also the contribution to the

mechanical resistance due to the ssDNA background. Conse-

quently, the ratio between the effective dsDNA density of the

hybridized patch and the non-hybridized one provides an upper

bound on the hybridization efficiency. For the case reported in

Fig. 3b we accordingly found that the upper bound for the

hybridization efficiency is about 30% (see Fig. S1 in the ESI†).

This bound is well above the 10% value found from SPR data for

saturated DNA SAMs.14 While we stress that the value refers to

an upper bound and hence all lower values could be compatible

with the result, the significant difference of the NAM value with

the SAM one is probably genuine (it is not far from a previous

estimate of 50% hybridization efficiency for NAMs given by

Mirmomtaz et al.11 and with the recent observation of 40%

hybridization efficiency obtained on DNA functionalized

micromechanical pillars48) and is consistent with the expectedly

higher order of NAM monolayers.

Conclusions

In conclusion, we have reported on the use of an integrated

experimental/theoretical approach to provide a phenomenolog-

ical, quantitative description and prediction of the response to

mechanical compression of ssDNA and dsDNA NAMs with

varying levels of surface density.

The computationally derived quantitative relation that ties the

surface density of ssDNA or dsDNA NAMs to the average

height of a compressed patch can be profitably used for a twofold

purpose. On the one hand, by fitting the AFM measurements of

patch height versus applied load, it is possible to infer the surface

density of the patch. On the other hand, by comparing the

different response of non-hybridized and hybridized patches it is

possible to obtain an estimate, and in particular, a stringent

bound on the hybridization efficiency of the patch. Both the

patch density and hybridization efficiency are crucial fabrication

parameters for the patches, but difficult to probe on the nano-

scale by typical experimental setups. This study indicates that

they can be more straightforwardly obtained through compar-

ison with coarse-grained models amenable to extensive compu-

tational characterization.

The computational approach brings the additional benefit of

controlling a priori several salient properties of the systems and

hence provides a powerful guiding tool to design special-purpose

nano-patches. One such notable application is the optimization

of the fabrication parameters of a ssDNA patch to yield the

highest resolution and sensitivity in detecting the binding of

complementary DNA strands in solution. In fact, the proposed

phenomenological model can be used to choose the appropriate

patch surface density to ensure that the change in height upon

1740 | Nanoscale, 2012, 4, 1734–1741

(possibly partial) ssDNA hybridization is easily and unambigu-

ously ascertained by AFM probing.

The present investigation can be extended to specific diag-

nostic directions. In particular, it would be worth covering

a wider range of lengths of the grafted DNA sequences, say,

spanning from 18 to 60 nucleotides, as these lengths represent the

most relevant ones from the biomedical and nano-technological

points of view.

Acknowledgements

This work was supported by Regione Autonoma Friuli Venezia

Giulia (L.R. 30/84), the Italian Institute of Technology—SISSA

unit and CNR-IOM Democritos. The authors are grateful to

Giacinto Scoles for illuminating discussions on data interpreta-

tion and analysis, Denis Scaini and Mauro Prasciolu for Scan-

ning Electron Microscopy measurements, and to Barbara

Sanavio, Matteo Castronovo, Allen W. Nicholson and Felix

Ritort for valuable discussions and critical reading of the

manuscript.

Notes and references

1 L. S. Wong, F. Khan and J. Micklefield, Chem. Rev., 2009, 109, 4025–4053.

2 C. Wingren and C. A. K. Borrebaeck, Curr. Opin. Biotechnol., 2008,19, 55–61.

3 C. J. Johnson, Proteomics, 2008, 8, 715–730.4 C. M. Niemeyer, Anal. Biochem., 1999, 268, 54–63.5 C. M. Niemeyer, Angew. Chem., Int. Ed., 2010, 49, 1200–1216.6 H. Schroeder, M. Adler, K. Gerigk, B. Muller-Chorus, F. Gotz andC. M. Niemeyer, Anal. Chem., 2009, 9, 1275–1279.

7 J. Muller and C. M. Niemeyer, Biochem. Biophys. Res. Commun.,2008, 377, 62–67.

8 G.-Y. Liu and M. B. Salmeron, Langmuir, 1994, 10, 367–370.9 P. S. Cremer, J. Am. Chem. Soc., 2011, 133, 167–169.10 J. N. Murphy, A. K. H. Cheng, H.-Z. Yu and D. Bizzotto, J. Am.

Chem. Soc., 2009, 131, 4042–4050.11 E. Mirmomtaz, M. Castronovo, C. Grunwald, F. Bano, D. Scaini,

A. A. Ensafi, G. Scoles and L. Casalis,Nano Lett., 2008, 8, 4134–4139.12 F. Bano, L. Fruk, B. Sanavio, M. Glettenberg, L. Casalis,

C. M. Niemeyer and G. Scoles, Nano Lett., 2009, 9, 2614–2618.13 I. Kopf, C. Grunwald, E. Brundermann, L. Casalis, G. Scoles and

M. Havenith, J. Phys. Chem. C, 2010, 114, 1306–1311.14 A. W. Peterson, R. J. Heaton and R. M. Georgiadis, Nucleic Acids

Res., 2001, 29, 5163–5168.15 C. E. Immoos, S. J. Lee and M. W. Grinstaff, J. Am. Chem. Soc.,

2004, 126, 10814.16 L. K. Gifford, I. E. Sendroiu, R. J. Corn and A. Luptak, J. Am. Chem.

Soc., 2010, 132, 9265–9267.17 T. Uzawa, R. R. Cheng, R. J. White, D. E. Makarov and

K. W. Plaxco, J. Am. Chem. Soc., 2010, 132, 16120–16126.18 Z. Zhang, J. Zhou, A. Tang, Z. Wu, G. Shen and R. Yu, Biosens.

Bioelectron., 2010, 25, 1953–1957.19 M. Castronovo, A. Lucesoli, P. Parisse, A. Kurnikova, A. Malhotra,

M. Grassi, G. Grassi, B. Scaggiante, L. Casalis and G. Scoles, Nat.Commun., 2011, 2, 297.

20 M. Castronovo, S. Radovic, C. Grunwald, L. Casalis, M. Morganteand G. Scoles, Nano Lett., 2008, 8, 4140–4145.

21 U. Rant, K. Arinaga, S. Fujita, N. Yokoyama, G. Abstreiter andM. Tornow, Langmuir, 2004, 20, 10086.

22 D. Marenduzzo, C. Micheletti and E. Orlandini, J. Phys.: Condens.Matter, 2010, 22, 283102.

23 L. Tubiana, E. Orlandini and C. Micheletti, Phys. Rev. Lett., 2011,107, 188302.

24 S. Leikin, V. A. Parsegian, D. C. Rau and R. P. Rand, Annu. Rev.Phys. Chem., 1993, 44, 369–395.

25 N. L. Goddard, G. Bonnet, O. Krichevsky and A. Libchaber, Phys.Rev. Lett., 2000, 85, 2400–2403.

This journal is ª The Royal Society of Chemistry 2012

Dow

nloa

ded

by L

aure

ntia

n U

nive

rsity

on

01 M

arch

201

3Pu

blis

hed

on 2

0 D

ecem

ber

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2NR

1166

2F

View Article Online

26 W. Saenger, Principles of Nucleic Acid Structure, Springer-Verlag,New York, 1984.

27 A. V. Rybenkov, N. R. Cozzarelli and A. V. Vologodskii, Proc. Natl.Acad. Sci. U. S. A., 1993, 90, 5307–5311.

28 N. M. Toan and C. Micheletti, J. Phys.: Condens. Matter, 2006, 18,S269–S281.

29 D. Stigter, Biopolymers, 1977, 16, 1435–1448.30 O. Gonzalez and J. H. Maddocks, Proc. Natl. Acad. Sci. U. S. A.,

1999, 96, 4769–4773.31 D. Marenduzzo and C. Micheletti, J. Mol. Biol., 2003, 330, 485–

492.32 M. C. Murphy, I. Rasnik, W. Cheng, T. M. Lohman and T. Ha,

Biophys. J., 2004, 86, 2530–2537.33 C. Micheletti, D. Marenduzzo, E. Orlandini and D. W. Sumners,

Biophys. J., 2008, 95, 3591–3599.34 A. B. Steel, R. L. Levicky, T. M. Herne and M. J. Tarlov, Biophys. J.,

2000, 79, 975–981.35 C. Howell, J. Zhao, P. Koelsch and M. Zharnikov, Phys. Chem.

Chem. Phys., 2011, 13, 15512–15522.36 F. Livolant and F. Leforestier, Prog. Polym. Sci., 1996, 21, 1115–

1164.

This journal is ª The Royal Society of Chemistry 2012

37 D.Marenduzzo, E. Orlandini, A. Stasiak, D.W. Sumners, L. Tubianaand C. Micheletti, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 22269–22274.

38 W. Kaiser and U. Rant, J. Am. Chem. Soc., 2010, 132, 7935–7945.39 J. F. Marko and E. D. Siggia, Macromolecules, 1995, 28, 8759–8770.40 J. N. Sneddon, Int. J. Eng. Sci., 1965, 3, 47–57.41 S. Steltenkamp, M. M. Muller, M. Deserno, C. Hennesthal,

C. Steinem and A. Janshoff, Biophys. J., 2006, 91, 217–226.42 M. Heuberger, G. Dietler and L. Schlapbach, J. Vac. Sci. Technol., B,

1996, 14, 1250–1254.43 D. Y. Petrovykh, H. Kimura-Suda, M. J. Tarlov and L. J. Whitman,

Langmuir, 2004, 20, 429–440.44 L. K. Wolf, Y. Gao and R. M. Georgiadis, Langmuir, 2004, 20, 3357–

3361.45 I. Y. Wong and N. A. Melosh, Biophys. J., 2010, 98, 2954–2963.46 D. Irving, P. Gong and R. Levicky, J. Phys. Chem. B, 2010, 114,

7631–7640.47 P. Gong and R. Levicky, Proc. Natl. Acad. Sci. U. S. A., 2008, 105,

5301–5306.48 M. Melli, G. Scoles and M. Lazzarino, ACS Nano, 2011, 5, 7928–

7935.

Nanoscale, 2012, 4, 1734–1741 | 1741