GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12030
November 20, 2014
C
2014 American Chemical Society
Optimized Assembly and Covalent
Coupling of Single-Molecule DNA
Origami Nanoarrays
Ashwin Gopinath
*
,†
and Paul W. K. Rothemund
*
,†,‡,§
Departments of
†
Bioengineering,
‡
Computer Science, and
§
Computation & Neural Systems, California Institute of Technology, Pasadena, California 91125,
United States
S
tructural DNA nanotechnology
1
allows
an experimenter to specify complex
nanoscale geometries and to decorate
those geometries with arbitrary patterns
of functional materials to create devices.
In particular, two-dimensional
2
and three-
dimensional
3
5
DNA origami provide
modular
“
molecular breadboards
”
upon
which 200 di
ff
erent components can be
self-assembled with a resolution of
∼
6nm.
DNA origami have already been used to
assemble prototype electronic
6,7
devices,
optical devices,
8
12
biosensing assays,
13
15
and custom single-molecule instruments for
answering biological questions.
16
18
Further,
origami are being developed as templates
19,20
and as masks
21,22
for new nanolithographies
to pattern gold,
7,19
graphene,
22
silicon,
21
and
other materials.
20
Such applications demonstrate the po-
tential of DNA nanostructures for bottom-
up fabrication, but characterization and
integration of DNA-organized devices have
proven di
ffi
cult, primarily because DNA
nanostructures are synthesized in solution.
Whether devices are assembled on origami
in solution or created after origami deposi-
tion, simple surface deposition results in
random arrangements of devices with
random orientations. Prior to any analysis,
device locations must be mapped with an
ultramicroscopy such as scanning electron
microscopy (SEM) or atomic force micro-
scopy (AFM). For applications which require
addressing or integrating devices, such
as nanoelectronics,
6
custom patterns of
electrodes must be written using e-beam
lithography. Optical applications
23
requir-
ing periodic arrangements of devices are
impossible with simple surface deposition.
Thus, it is crucial to develop reliable meth-
ods for positioning and orienting individual
DNA origami (and hence associated devices)
on planar substrates.
A variety of techniques now allow
the directed assembly of DNA origami on
lithographically patterned substrates.
24
33
Strong gold
thiol interactions have been
successful in directing one-dimensional
DNA origami tubes to create point-to-point
connections between gold islands.
29,30
In
general, however, methods based on strong
* Address correspondence to
ashwing@caltech.edu,
pwkr@dna.caltech.edu.
Received for review October 21, 2014
and accepted November 20, 2014.
Published online
10.1021/nn506014s
ABSTRACT
Arti
fi
cial DNA nanostructures, such as DNA origami,
have great potential as templates for the bottom-up fabrication of
both biological and nonbiological nanodevices at a resolution
unachievable by conventional top-down approaches. However,
because origami are synthesized in solution, origami-templated
devices cannot easily be studied or integrated into larger on-chip
architectures. Electrostatic self-assembly of origami onto lithogra-
phically de
fi
ned binding sites on Si/SiO
2
substrates has been
achieved, but conditions for optimal assembly have not been characterized, and the method requires high Mg
2
þ
concentrations at which most devices
aggregate. We present a quantitative study of parameters a
ff
ecting origami placement, reproducibly achieving single-origami binding at 94
(
4% of sites,
with 90% of these origami having an orientation within
(
10
°
of their target orientation. Further, we introduce two techniques for converting electrostatic
DNA
surface bonds to covalent bonds, allowing origami arrays to be used under a wide variety of Mg
2
þ
-free solution conditions.
KEYWORDS:
DNA nanotechnology
.
directed self-assembly
.
single molecule
.
nanoarray
.
surface di
ff
usion
ARTICLE
This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the
article or any adaptations for non-commercial purposes.
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12031
electrostatic
28
or strong covalent
32
interactions be-
tween DNA and substrate binding sites yield poor
orientational precision for two-dimensional origami
because origami are
fi
xed quickly, without an oppor-
tunity to realign. Weaker electrostatic interactions
were used by Kershner
et al.
25
to position individual
DNA origami at binding sites on diamond-like carbon
(DLC) and SiO
2
substrates with reasonable yield and
orientational precision. However, this
“
origami place-
ment technique
”
has been rarely reproduced,
27
or
extended,
31
because (i) appropriate DLC substrates
are not widely available, (ii) the binding mechanism
and experimental conditions for high-quality position-
ing on SiO
2
substrates have not been well-understood,
and (iii) small details of substrate fabrication can have
a large e
ff
ect on placement. Furthermore, the tech-
nique required a high (125 mM) Mg
2
þ
concentration
which causes most components of interest, including
carbon nanotubes, metal nanoparticles, and proteins,
to aggregate.
RESULTS AND DISCUSSION
Here we build on the basic method of Kershner
et al
.
25
Binding sites of the same shape and size as
triangular origami are patterned on an SiO
2
substrate
by O
2
plasma etching through a trimethylsilyl (TMS)
passivation layer (Figure 1a, Steps 1
5, Supporting
Information Figures S1 and S2); this creates silanols
which ionize at an appropriate pH to become
negatively charged. A solution of origami is applied
to the substrate, and Mg
2
þ
in the bu
ff
er provides
an electrostatic bridge between ionized silanols and
negatively charged origami (Figure 1b, Method 1).
We study, optimize, and extend this placement tech-
nique with the goal of making it accessible for a wide
variety of applications under diverse experimental
conditions. For the basic technique, we achieve
both higher orientational precision and a lower work-
ing Mg
2
þ
concentration than previous studies. We
fi
nd that the placement yield and quality are highly
nonlinear in a number of global (origami concentra-
tion, pH, Mg
2
þ
concentration, incubation time) and
spatial (binding site size and spacing) parameters.
Most surprising were the results of varying binding
site spacing, which lead to the discovery that the
mechanism of binding is not limited to direct di
ff
usion
from solution: it also involves an indirect pathway in
which origami
fi
rst bind to unpatterned regions and
then undergo 2D di
ff
usion to reach binding sites.
Finally, we introduce protocols for placement on
positively charged substrates (Figure 1b, Method 2),
microcontact printing (Figure 1b, Method 3) and post-
placement covalent coupling (Figure 1b, Method 4),
which allow the creation and use of DNA origami
arrays over a large range of Mg
2
þ
-free conditions. A
variety of technical details, important to the reprodu-
cibility of the technique, are given in the Supporting
Information.
Figure 1. Fabrication of substrates and methods for origami placement and immobilization. (a) Arrays of triangular binding
sites are patterned on a SiO
2
substrate
via
e-beam lithography
25
(top, Steps 1
5). An O
2
plasma (Step 4) is used to etch
through the trimethysilyl templatelayer to create silanolgroupsat eachsite. AFM orSEM documents how substratesor molds
should look at crucial steps (bottom). AFM before lithography (Step 1) shows that substrates are
∼
3
as rough as the mica
typically used as DNA origami substrates, with an rms roughness of 3 Å. SEM shows the quality of a silicon mold for an
alternative patterning method (Step 2b, nanoimprinting
31
) as well as pattern quality in the resist after development (Step 3).
AFM after the resist strip (Step 5) shows that binding sites can occasionally be observed by phase imaging, but naked binding
sites are di
ffi
cult to resolve. AFM after placement (Method 1) shows mostly well-oriented single origami. Scale bars, 400 nm.
(b) Four variations of placement. Method 1: At an appropriate pH, surface silanols become negatively charged and divalent
Mg
2
þ
ions can be used as a bridge to immobilize negatively charged DNA origami (black circles). Method 2: Placement
without Mg
2
þ
can be achieved by functionalizing sites with an amino-terminated silane (Step 6), resulting in lower quality
arrays (Figure 4d,e). Method 3: Printing from a substrate made using Method 1 onto an unpatterned amino-terminated
surface allows construction of high-qualityorigami arrays without Mg
2
þ
(Figure 4g,g). Method 4: Covalentbonding is another
route to Mg
2
þ
-free retention of origami: amino-functionalized origami are placed using Method 1, and then the surface is
treated with cross-linkers which can form either an amide bond (shown) or an isourea bond (see Figure 4h
k).
ARTICLE
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12032
Effects of Global Parameters.
After placement (Supporting
Information Figure S3), each binding site can exhibit
one of several different states (Figure 2a). A fraction of
sites are occupied with single origami in the
“
correct
orientation
”
in which the origami has maximal over-
lap with the binding site
;
the remaining sites may
contain a single origami with an incorrect orientation
(measured relative to the binding site, Figure 2b), they
may contain multiple origami, or they may be empty.
Quantitative measurements of the fraction of sites in
each state, as well as distribution of orientations, were
used to optimize global parameters.
Our initial model of how these states arise during
placement is captured in Figure 2c, which depicts a
Figure2. Optimization of origamiplacement.(a)Schematicshows correctplacement,misalignments, vacancies,andmultiple
binding events. (b) Measurement of origami orientation (
θ
) relative to its binding site. (c) Model of transitions between the
states in (a), assuming that origami arrive from or depart to solution. (d) AFM data and (j) plot showing the nonlinear
dependence of placement quality on origami concentration. (f,k) Same for Mg
2
þ
concentration, (h,l) for pH, and (i,m) for
incubation time. Nonvarying parameters were 110 pM origami, 35 mM Mg
2
þ
, pH 8.3, and 60 min incubation (in 5 mM Tris
bu
ff
er) as applicable. For plots, orange squares show the desired state, a binding site with a single DNA origami. Black
triangles show empty sites, and red triangles indicate total occupied sites. (e,g) Histograms show the quality of origami
orientation for origami and Mg
2
þ
concentration based on (d,f). Error bars are SEM for
N
= 3 independent replications (of
placement, washing,
etc.
) using di
ff
erent chips from the same wafer. 800
1000 binding sites were scored for each replicate.
Scale bars, 400 nm.
ARTICLE
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12033
single binding site interacting with origami that bind
directly from solution: a single origami may bind with
imperfect orientation and realign to the correct orien-
tation, or additional origami may bind to free area on
the site before realignment can occur. The mininum
free energy state for a single origami is assumed to be
the correct orientation, which should maximize the
number of silanol
Mg
2
þ
origami bridges. This model
is only intended to be a guide to understanding the
sign of the e
ff
ect of changes made to global para-
meters, and thus before describing results, we make
a few observations regarding the model and place-
ment experiments. Several things prevented us from
attempting to extract equilibrium constants or rate
constants associated with this model: (i) while binding
is weak enough that origami can reorient, it is strong
enough that we have been unable to observe a
signi
fi
cant dissociation rate, and thus, our experiments
are not at equilibrium; (ii) our primary method of
observation is
fl
uid-mode AFM, which is highly pertur-
bative, requires strong origami-substrate binding, and
does not allow accurate statistics in a weaker binding
regime where origami can be knocked o
ff
the surface
by the AFM tip; and (iii) we will later present evidence
that a signi
fi
cant number of origami do not bind sites
directly from solution but
via
an indirect pathway
which involves them
fi
rst binding the surface.
Because our experiments were not equilibrium
experiments, we chose to measure the state of binding
sites after 1 h, a period of time over which initial experi-
ments showed that binding site state evolved from
mostly empty to mostly occupied. At 1 h, a series of
washes was performed to transfer the sample into an
imaging bu
ff
er, taking care not to dewet the sample
(Supporting Information Figure S4). In particular,
washes with the detergent Tween 20 served to prevent
further binding and removed origami which were
weakly bound to the background or to binding sites
(Supporting Information Figure S5). Thus, our measure-
ments likely underestimate the occurrence of single
and multiply bound origami during placement, but
instead re
fl
ect the quality of the complete fabrica-
tion process (before drying, Supporting Information
Figure S6). AFM movies (Supporting Information
movies 1
4, 3 frames/s) taken during placement with
a less perturbative, fast-scan AFM show multiple bind-
ing, unbinding, and realignment, providing direct
observation of the processes proposed in Figure 2c;
however, it is possible that the observed unbinding
and realignment were induced by interaction between
the origami and the tip.
We expected that the
fi
rst parameter we varied,
origami concentration (Figure 2d, Supporting Informa-
tion Figures S7
S17), would e
ff
ect the quality of
placement
via
the rate of binding at a site. As origami
concentration increases, the rate of origami binding
site encounters increases, and second or third origami
may bind (essentially irreversibly) before a single
origami has a chance to realign and fully occupy the
site. This would predict an increase in multiple bind-
ings with increasing concentration, which is re
fl
ected
in the data by a sharp decrease in single-origami
bindings (indicated by orange squares in Figure 2j) as
a percent of the number of occupied sites (indicated by
red triangles) above 100 pM origami. An unexpected
decrease in the quality of alignment of single origami
was also observed with increasing concentration
(above 100 pM, Figure 2e). Our hypothesis is that the
poor alignment of single origami at high concentration
actually re
fl
ects
“
cryptic
”
multiple bindings which
have been reduced to single bindings by the Tween
20 wash, which may remove loosely bound second and
third origami.
In general, changing a parameter to increase bind-
ing strength will decrease
k
o
ff
;
this provides a second
mechanism to increase multiple bindings since second
or third origami will unbind at a low rate. A concomi-
tant increase in cryptic multiple bindings may lead to
a decrease in the quality of single-origami alignment,
as proposed above. Further, realignment may have an
associated activation energy, and increasing binding
strength may increase this activation energy and de-
crease
k
align
, thus providing a second and distinct
mechanism for decreasing alignment quality. For the
second parameter varied, [Mg
2
þ
] (Figure 2f and Sup-
porting Information Figures S18
S30), increasing
[Mg
2
þ
] was expected to increase binding strength
by providing more bridges between the origami and
negative charges on ionized silanols; thus increased
multiple bindings and worse alignment were ex-
pected. Both e
ff
ects were observed: above 35 mM
Mg
2
þ
multiple bindings increased (Figure 2k) and
by 80 mM Mg
2
þ
the quality of alignment decreased
(Figure 2g).
Similarly, for the third parameter studied, pH
(Figure 2h and Supporting Information Figures
S31
S38), increased [OH
] was expected to increase
binding strength by increasing the number of ionized
silanols available for Mg
2
þ
binding. Indeed, origami
binding increased dramatically (Figure 2l) from pH 7.0
to pH 8.4, and multiple bindings increased above
pH 8.4. However, at pH 9.1, the TMS groups used to
prevent binding of orgami to the background hydro-
lyzed and origami bound everywhere on the surface.
Surface silanols exist in multiple geometries and have
varying p
K
a
:
34
19% have a p
K
a
of 4.9, while 81% have
ap
K
a
of 8.5. Thus, the observed increase in origami
binding centered at pH 8 is consistent both with the pH
at which 50% of silanols are ionized (7.9) and with the
pH with the highest rate of change in silanol ionization
(8.3), but surface charge, metal
surface binding, and
eventually DNA
surface binding are not solely deter-
mined by silanol p
K
a
: they are also in
fl
uenced by the
speci
fi
ca
ffi
nity of a particular metal ion for silanols
35
ARTICLE
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12034
and DNA, the density of the silanols, and the concen-
tration of competing monovalent and divalent cations
in solution.
36,37
Hence we do not conclude that the
p
K
a
of surface silanols will necessarily predict origami
behavior for other metal ions or other surfaces.
Lastly, to optimize incubation time, we studied the
kinetics of nanoarray formation (Figure 2i and Support-
ing Information Figures S39
S47). With all other para-
meters optimized (110 pM origami, 35 mM Mg
2
þ
,
pH 8.35), we observed that single-origami binding
consistently rose to more than 90% at an incubation
time of 60 min (Figure 2m), and multiple bindings
remained infrequent, at just 5
10% of sites. Further
incubation, however, led to a dramatic increase in
multiple binding events. By 120 min,
∼
64% of sites
were multiply bound, and by 480 min, the passivating
background had undergone signi
fi
cant hydrolyis, or-
igami bound indiscriminantly, and binding sites could
not be di
ff
erentiated from the background. Thus, we
chose 60 min as an optimal incubation time. If multiple
bindings are more problematic for a downstream
application than empty sites, incubation can be ar-
rested early
;
within 5 min, 40% of sites have single
origami and only 2% of sites are multiply bound. The
observed kinetics re
fl
ect only the binding of origami to
sites since dissociation of origami from binding sites
has proven unmeasurable: when origami arrays were
placed under optimal conditions, washed to remove
excess origami, and left under clean bu
ff
er for several
days, no loss of origami was observed.
Two observations hold for all four parameters
studied. First, a fast rise in site occupancy saturates
quickly, and in this regime, placement quality is high
since most sites hold single origami (Figure 2j
m,
orange single-origami traces and red occupied-site
traces overlap). For Mg
2
þ
and pH, the initial rise was
surprisingly fast: from 25 to 35 mM, Mg
2
þ
site occu-
pancy jumped from 8 to 96% (Supporting Information
Figure S29 shows similar behavior on unpatterned
substrates), and from pH 7.5 to pH 8.1, site occupancy
jumped from 2 to 92%. Second, after an optimum has
been reached, placement quality degrades more
slowly, through an increase in multiple binding events
(Figure 2j
m, orange traces fall below red). While our
optimization highlights the sensitivity of placement
quality to all four parameters, careful and precise
maintenance of all four parameters enables highly
reproducible results. Over the course of 90 indepen-
dent replications of placement (using chips from four
di
ff
erent wafers under optimal conditions, performed
over 16 months, for which at least 100 binding sites
were measured), we have achieved single-origami
binding at 94
(
4% of sites, with 90% of these origami
having an orientation within
(
10
°
of the correct
orientation. Further, the similar behavior of placement
under each of the four parameters suggests that
placement can be reoptimized by adjusting whichever
parameter is most convenient
;
increasing the para-
meter's value if occupancy is too low, or decreasing the
parameter's value if multiple bindings are too high.
We note that the orientational precision achieved
here is much better than that previously achieved
25
on
Si/SiO
2
substrates (68% of origami within
(
20
°
of the
correct orientation) and better than the best precision
previously achieved, which was performed on di
ffi
cult
to source DLC substrates (68% of origami within
(
10
°
of the correct orientation). We further note that the
optimum concentration we report for Mg
2
þ
, 35 mM, is
much lower than the concentration previously used
25
for origami placement, 125 mM. We associate this
improvement with the addition of a new cleaning step,
involving HF/NH
4
F treatment,
38
which reduces SiO
2
surface roughness to 3 Å. Placement has not previously
been characterized with respect to surface roughness,
but we have observed (Supporting Information Figure
S30) that thermally grown SiO
2
with 9
10 Å rms
roughness requires 90 mM Mg
2
þ
to achieve good
placement, under otherwise optimized conditions.
Further, a surface of intermediate 5 Å roughness re-
quired 55 mM Mg
2
þ
. Our hypothesis is that a smoother
surface may allow a greater number of Mg
2
þ
bridges
between origami and ionized silanols, but we have not
yet ruled out changes to surface chemistry due to the
cleaning step.
Placement on thermally grown SiO
2
on Si substrates
should enable many electronic device applications, but
optical applications such as single-molecule biophysics
require transparent substrates. Quartz wafers are one
such optically transparent substrate, which is
fl
at (4 Å;
glass slides are too rough) and o
ff
ers surface chemistry
similar to thermally grown SiO
2
. Unfortunately, quartz
is nonconductive, which complicates e-beam lithogra-
phy. Previously, nanoimprint lithography
31
has been
used for origami placement onto SiO
2
on Si substrates
under high Mg
2
þ
(125 mM) conditions. Initial experi-
ments on quartz using our optimized, lower Mg
2
þ
protocol (Step 2b in Figure 1a and Supporting Informa-
tion Figure S48) suggests that nanoimprint on quartz
can achieve placement of equal quality to that created
by e-beam on thermally grown SiO
2
.
Effects of Spatial Parameters.
After optimizing global para-
meters, we investigated how two spatial parameters
;
the size and the periodicity of binding sites
;
affected
placement quality. Binding-site size was expected to
affect site occupancy, single-origami binding, and
alignment; we have only measured the first two. We
created a chip with 21 sizes of binding sites, with edge
lengths ranging from 15% of the 127 nm origami edge
length up to 200%, and performed placement over
a range of Mg
2
þ
concentrations from 25 to 60 mM
(Figure 3a and Supporting Information Figures S49
S55).
By varying [Mg
2
þ
], our goal was to gain information
about the interplay of binding area, the amount
of Mg
2
þ
available for binding, and binding energy.
ARTICLE
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12035
Under optimized (35 mM Mg
2
þ
, 110 pM origami, pH
8.35, 60 min) conditions, oversized sites were expected
to bind multiple origami, and undersized sites were
expected to have a lower site occupancy than standard
127 nm sites. This trend was observed (Figure 3b, red
trace), but site occupancy saturated (at 100%) for
binding sites having an edge length only 80% of the
standard size, encompassing only 64% of the total
binding area
;
thus site occupancy is relatively insensi-
tive to a significant undersizing of binding sites. In
contrast, single-origami binding events (Supporting
Information Figure S55) were maximized (90%) for
binding sites that were slightly undersized (85
90%
origami edge length). Surprisingly, changing [Mg
2
þ
]
from 35 to 60 mM had no significant effect: site
occupancy saturated at essentially the same binding-
site size. Little binding (<20% for all binding-site sizes)
was observed at 25 mM Mg
2
þ
.
In Figure 2k, site occupancy as a function of [Mg
2
þ
]
exhibits a Hill coe
ffi
cient of
∼
15, suggesting that Mg
2
þ
binding to origami and the surface is highly coopera-
tive. In Figure 3b, the large jump and saturation of site
occupancy with increasing [Mg
2
þ
] supports coopera-
tive Mg
2
þ
binding for site sizes down to 25% of the
standard area. The binding-site size variation data
also support a picture in which Mg
2
þ
forms a layer of
de
fi
ned density and contributes a characteristic bind-
ing energy per unit area. Even though equilibrium
binding constants cannot be measured, as we have
discussed above, binding energy may be re
fl
ected in
site occupancies observed by AFM because more
strongly bound origami can be more stably imaged
and better resist being detached by the AFM tip.
Assuming that, for a
fi
xed area of overlap between an
origami and a site, more Mg
2
þ
between the origami
and site will translate
via
higher binding energy to a
higher site occupancy, then one would predict that
higher [Mg
2
þ
] should increase site occupancy for un-
dersized sites. Instead, constancy of site occupancy is
observedacross all site sizes, which suggeststhat above
a threshold concentration of 35 mM Mg
2
þ
, the amount
of Mg
2
þ
per unit area between the origami and the
binding site is constant and has reached a maximum
value. Such a limit might be set by the density of
Figure 3. E
ff
ects of binding site size and spacing. (a) Binding to triangular sites with 13 di
ff
erent edge lengths ranging from
55 to 200% of the edge length of an origami (127 nm) for two di
ff
erent [Mg
2
þ
]. For select sizes, both size as a percentage of
edge length and size as a percentage of area (that enclosed by the outer edge of an origami) are noted. (b) Plot of site
occupancy (both single and multiple bindings) at
fi
ve di
ff
erent Mg
2
þ
concentrations for 21 di
ff
erent edge lengths from 15 to
200% (error bars are SEM from
N
= 4 independent replicates). (c) AFM under optimized conditions (110 pM origami, 35 mM
Mg
2
þ
, pH 8.35 and 60 min) but with di
ff
erent binding-site spacings. For 800 and 2000 nm periods, small regions have been
combined into composite images. (d) Plot of binding-site state
vs
period (error bars are SEM from
N
= 3 independent
replicates). (e) AFM of an unpatterned TMS-passivated substrate incubated with 1 nM origami under otherwise optimized
conditions and washedwith bu
ff
er ten times. (f) Chip imaged in (e) after the passivation layer has been removed by increasing
the pH to 11. (g) AFM of inhomogeneous binding to a 5
μ
m
5
μ
m activated window, with vacant areas near the center.
(h) Probability map calculated from AFM of 25 di
ff
erent 5
μ
m
5
μ
m windows gives the probability of an origami being
observed at each position. (i) Model shows direct binding
via
3D solution di
ff
usion (black arrow) or indirect binding
via
background binding and 2D di
ff
usion (red arrows). Corner (
c
), edge (
e
), and interior (
i
) sites are equivalent for 3D di
ff
usion, but
we expect binding rates for corners and edges to higher for 2D di
ff
usion. Scale bars, 400 nm.
ARTICLE
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12036
available ionized silanols, the density of available
DNA backbone phosphates, or the interplay between
Mg
2
þ
Mg
2
þ
repulsion and the geometry of both
surface silanols and the DNA backbone, as has been
proposed in the
“
ion correlation model
”
for DNA
mica
binding interactions.
39
The second spatial parameter studied, the spacing
between binding sites, revealed that the mechanism
of origami-site binding is much more complex than
proposed in our initial model. Under optimized condi-
tions, we examined binding to sites in square arrays
with periods from 200 nm up to 2000 nm (Figure 3c
and Supporting Information Figures S56
S61). The
number of binding sites (
∼
2 million) was held con-
stant. Additionally, we created a
“
0nm
”
array, in which
a 117
μ
m
117
μ
m square patch having the same area
as
∼
2 million binding sites was fabricated and site
occupancy was measured as the number of origami
per unit binding-site area. As period increased from
400 to 2000 nm, site occupancy (red) decreased some-
what, from roughly 95 to 80%. This is not a signi
fi
cant
trend, and it might be explained because data for
larger periods were collected over multiple AFM
images, each requiring a separate engage and optimi-
zation of conditions to minimize tip
origami interac-
tions
;
weakly bound origami can be removed from
the surface in this process. Single-origami binding
(orange) ranged between 75 and 80%. These experi-
ments (included in our overall statistics for single-
origami binding) represent the lowest quality place-
ment observed, perhaps because the chips used derive
from a di
ff
erent wafer and e-beam write. Below 400 nm,
site occupancy dropped quickly
;
70% for 200 nm and
25% for 0 nm spacings. This suggested that origami-site
binding was being negatively a
ff
ected by the presence
of origami at adjacent sites.
We hypothesized that this crowding e
ff
ect might
be mediated by AFM-invisible, weakly bound origami
undergoing surface di
ff
usiononthepassivatedback-
ground. To test this, we incubated origami over a com-
pletely passivated chip at 10
the standard
concentration (1 nM) under otherwise optimized condi-
tions. After 10 bu
ff
er washes to remove all free origami
from the chip, we imaged the chip by solution AFM and
found that no origami could be observed (Figure 3e).
Next we hydrolyzed the TMS passivation layer by increas-
ing the pH of the bu
ff
er to 11
via
addition of 0.2 M NaOH.
Reimaging revealed a high density of immobilized or-
igami (Figure 3f), which we infer remained weakly bound
to the passivation layer during earlier bu
ff
er washes;
these origami represent
∼
42% of the total origami
standardly applied to a chip. While surprising, this result
is not unprecedented: adsorption and 2D di
ff
usion of
DNA hairpins and other molecules on TMS-functionalized
surfaces have recently been observed optically.
40,41
Modeling of 2D di
ff
usion has shown that adsorption
on background areas can greatly accelerate the rate of
adsorption at speci
fi
c binding sites.
42
To understand
the potential role of surface di
ff
usion here, we simu-
lated placement in two limits (Supporting Information
Figure S62):
fi
rst, conditions under which origami ex-
clusively di
ff
use in solution in 3D; second, conditions
under which all origami immediately condense on the
surface and then exclusively di
ff
use in 2D. Unsurpris-
ingly, in the 3D case, the rate of site occupation is
independent of period: 80% site occupancy is reached
by 800 time steps for all periods ranging from 0 to 9
lattice sites. However, on a 2D surface, occupied sites
block di
ff
usion of origami to unoccupied sites, and
so the occupation rate decreases as the period de-
creases
;
period 8 reaches 80% site occupancy by
800 time steps, but period 3 requires almost 5000 time
steps to reach 80%, and period 0 reaches only 20%
occupancy at 5000 time steps. Further, site occupancy
is spatially inhomogeneous, with the highest occupan-
cies at the edges of the array. This e
ff
ect is most
extreme in the case of close-packed sites (period 0), for
which a narrow band of origami bound atthe edge of the
array prevents origami from reaching the interior. This
result provided a testable prediction for large squares of
activated surface (0 nm arrays), and thus we fabricated
chipswithactivatedsquareswhoseedgesrangedfrom
500 nm to 5
μ
m in size. These chips were incubated with
origami for 30 min with other
wise optimized conditions
(Supporting Information Figure S63). As the size of the
squares increased, the size and number of origami-free
patches on the interior of squares grew; Figure 3g shows
a5
μ
m square with several large (>400 nm diameter)
vacancies. A probability map (Figure 3h) calculated from
the interaction between Mg
2
þ
AFM images shows that
the probability of an origami being present at a particular
position varies from greater than 0.9 at the edge of the
square to less than 0.25 at the center. An optical experi-
ment for 20
μ
m squares gives a similar but nonquanti-
tative result (Supporting Information Figure S63).
Experimental spatial inhomog
eneities are not as striking
as those predicted by the 2D simulation, probably
because the experimental results re
fl
ect a combination
of 3D and 2D di
ff
usion. The fact that 2D di
ff
usion
enhances edge and corner binding for large squares
suggests that binding rates fo
r corner, edge, and interior
sites of nanoarrays should be di
ff
erent (Figure 3i) for
closely spaced arrays. Further, it suggests why binding is
a function of period for interior sites: as the spacing
increases to 400 nm (
∼
3 origami in size), a site no longer
competes with its neighbors for origami di
ff
using on
the background when empty nor blocks origami dif-
fusing to neighbors when occupied. We note that back-
ground binding and 2D di
ff
usion cannot be inferred
from binding kinetics alone: the Langmuir adsorption
model, which assumes exclusively 3D di
ff
usion and
independent single-origami binding,
fi
ts our incuba-
tion time data reasonably well (Supporting Information
Figure S64).
ARTICLE
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12037
Maintaining Placement without Mg
2
þ
.
Origami placed
using the normal high Mg
2
þ
(35 mM) protocol
(Figure 4a,b) simply fall off the surface when washed
with Mg
2
þ
-free phophate-buffered saline (PBS) at pH
8.3 (Figure 4c). To some extent, lower Mg
2
þ
concentra-
tion can be compensated by increasing pH after
placement if this is compatible with downstream
applications. For example, origami are stably attached
to SiO
2
in lower Mg
2
þ
formation buffer (12.5 mM) at
pH
g
9. One approach to achieving truly Mg
2
þ
-free
placement has been to create positively charged bind-
ing sites,
28
achieved through the conversion of surface
silanols into positively charged amino groups (
NH
2
,
occurring predominantly as
NH
3
þ
when significantly
below its p
K
a
of
∼
10) using a silanization agent such as
aminopropyl silatrane
43
(APS, Figure 4d and Support-
ing Information Figure S65; this compound is much less
prone to aggregation than the common amino func-
tionalization agent, (3-aminopropyl)triethoxysilane
[APTES]). So far, with such amino-functionalized bind-
ing sites (Figure 4e), we have observed (i) a high rate of
multiple binding, (ii) poor orientation, and (iii) distor-
tions of origami, such as folding. These observations
are consistent with the hypothesis that origami bind-
ing to the sites is too strong, and that origami are
getting kinetically trapped before they have a chance
realign; (i) and (iii) make scoring sites difficult and so we
do not quantitatively compare this method to others.
Another approach combines the placement quality
a
ff
orded by Mg
2
þ
binding with the bond stability
a
ff
orded by a positively charged surface: microcontact
printing
44
46
allows origami placed using Mg
2
þ
to be
transferred to an amino-functionalized surface by
bringing the original placement substrate (the stamp)
into contact with an amino-functionalized surface and
then reducing pH and [Mg
2
þ
] to facilitate release from
the stamp (Figure 4f). When compared, the resultant
nanoarrays (Figure 4g) typically have a slightly smaller
site occupancy than that of the stamp (
e.g.
,91
vs
96%,
Supporting Information Figure S66) and mild degrada-
tion of the alignment (70% within
(
10
°
,
vs
87%). The
periodic grid of original binding sites is well preserved
by transfer, and no origami are found o
ff
the grid,
indicating that origami do not move signi
fi
cantly.
Printed nanoarrays are robust to changes in bu
ff
er
conditions (Figure 4g), as desired. However, the ami-
nated background between origami is highly sticky
and may have high nonspeci
fi
c binding for any devices
which might be added downstream
;
thus devices
should be coupled to origami prior to placement
and stamping or device-appropriate blocking agents
should be used (
e.g.
, noninterfering DNA for DNA-
linked objects, or BSA for proteins).
Covalent immobilization of origami after placement
o
ff
ers the most simple route for the creation of nano-
arrays that are robust to bu
ff
er changes since, unlike
stamping, it requires no mechanical manipulation of
the substrates. In such an approach, origami are
fi
rst
bound using Mg
2
þ
to achieve high-quality placement
as usual, and then cross-linking reagents are added
to covalently link the origami to the surface. Here, we
describe two methods (Figure 4h) which utilize trian-
gular origami bearing an average of 10 primary amines
at sites along their inner edge (purple dots Figure 4h,
Figure 4. Approaches to Mg
2
þ
-free placement. (a
c) Bu
ff
er exchange (a) that removes Mg
2
þ
and replaces it with Na
þ
causes
placed origami (b, 5 mM Tris, 35 mM Mg
2
þ
, pH 8.3) to be released from the surface (c, 1
PBS, pH 8.3). (d) Amino
functionalization
via
aminopropyl silatrane (APS) treatment. (e) Placement on the resulting positively charged surface results
in a greater incidence of multple bindings, misalignments, and folding, but is stable in 1
PBS. (f,g) Contact printing (f) from a
parent substrate to an amino-functionalized substrate (substrates are clamped together in pure water) preserves the quality
of placement and gives stability in 1
PBS (g). (h) Two methods for covalently immobilizing amino-functionalized origami.
Origami are placed as usual, and the surface is activated for cross-linking by either a cyanylating (CDAP) or carboxylating
(CTES) agent. Cross-linking is either spontaneous (CDAP) or requires a second catalytic step (CTES) using a carbodiimide (EDC)
and
N
-hydroxysulfosuccinimide (sulfo-NHS). Inset diagrams surface functionalization reagents. (i) CDAP renders origami
stable to 1
PBS and pure water, pH 6.0 (Supporting Information Figure S69). (j) Further addition of a Tris-sodium bu
ff
er
(5 mM Tris, 150 mM Na
þ
) aminolyzes the isourea bonds and releases origami. (k) CTES/EDC renders origami stable to pure
water, pH 6.0, as well as 1
PBS and Tris (Supporting Information Figure S70).
ARTICLE
GOPINATH AND ROTHEMUND
VOL. 8
’
NO. 12
’
12030
–
12040
’
2014
www.acsnano.org
12038
red dots Supporting Information Figure S67). In the
fi
rst method (Figure 4h, orange box), surface silanols
are converted into cyano groups using 1-cyano(4-
dimethylamino)pyridinium tetra
fl
uoroborate (CDAP)
in a coupling solution created by mixing acetonitrile
50% (v/v) with 5 mM MOPS bu
ff
er containing 250 mM
Mg
2
þ
at pH 7.0. Coupling solution is applied for just
10 min to a chip cooled on ice; a pH of 7.0 is used to
increase CDAP activity, and elevated [Mg
2
þ
] (150 mM
after mixing onto the chip) is used to
fi
x origami stably
on the surface at this lower pH. As soon as the cyano
group is created, it can spontaneously react with an
amino-functionalized origami to form an isourea bond.
The isourea derivative so formed is susceptible amino-
lysis by primary amines.
47
This is why MOPS bu
ff
er
is used for the coupling reaction rather than bu
ff
er
containing the amino-bearing Tris base, and this sus-
ceptibility enables us to con
fi
rm the isourea bond's
role in origami immobilization: postcoupling, we
fi
rst
imaged origami in PBS bu
ff
er (150 mM Na
þ
, pH 8.3) to
demonstrate that they are stable (for at least an hour)
under Mg
2
þ
-free conditions (Figure 4i and Supporting
Information Figure 68), consistent with the formation
of covalent isourea bonds. We next imaged the sample
under similar conditions in a bu
ff
er that contains 5 mM
Tris (150 mM Na
þ
, pH 8.3) and observed that the
origami were completely removed from the surface
(Figure 4j).
In a second covalent approach (Figure 4h, black
box), carboxyl groups are used to cross-link amino-
functionalized origami to the surface using EDC
[1-ethyl-3-(3-(dimethylamino)propyl)carbodiimide]
catalysis.
47
After standard Mg
2
þ
-mediated placement,
surface silanols were converted to carboxyl groups
by incubating the substrate in a 0.01% (v/v) CTES
(carboxyethylsilanetriol) solution. In a second step,
carboxyl groups were cross-linked to amino-functiona-
lized origami by incubating the substrate in a coupling
bu
ff
er (50 mM EDC, 25 mM sulfo-NHS, in 10 mM MOPS,
pH 8.1, 125 mM Mg
2
þ
). The resulting amide bond is
stable (for at least 24 h) in PBS bu
ff
er, and unlike the
isourea bond, it is stable in Tris-containing bu
ff
er
(Supporting Information Figure 70). In general, the
amidebondshouldbestabletoa wide varietyofbu
ff
ers
having a pH between 5 and 9.
Both covalent approaches provide for total removal
of salts: the CDAP and CTES/EDC procedures both
achieve the stable
fi
xation of origami on substrates
under pure water (freshly prepared Milli-Q, nominal pH
of 6.0) for at least an hour (Supporting Information
Figure 69 and Figure 4k), a condition under which
origami would otherwise be completely removed from
the surface. Despite the susceptibility of the isourea
bond to aminolysis, the CDAP procedure is much
simpler, involving a single chemical step and a sig-
ni
fi
cant reduction in the number of bu
ff
er washes
(16 fewer). The CTES/EDC procedure, on the other
hand, should be used in cases where the substrate will
later be subjected to primary amines (as in the case of
Tris bu
ff
er), other strong nucleophiles, or high pH.
CONCLUSIONS
Here, we have provided experimental conditions for
achieving reproducible, high-quality placement on
Si/SiO
2
substrates, along with protocols for noncova-
lently and covalently stabilized origami nanoarrays un-
der Mg
2
þ
-free conditions. By examining the spatial
dependence of origami bin
ding, we have revealed that
surface di
ff
usion plays an important role in the binding
mechanism, one that is crucial to understand for creating
nanoarrays of di
ff
erent spacings. With these advances,
origami placement shoul
d now be widely available
for nanotechnological and biophysical applications
in a variety of microfabricated devices
;
deterministic
positioninganindividualreceptorintheheartofa
photonic biosensor or a single ligand at the exact end
of an AFM tip.
The ability to position individual molecules may
also open new doors in chemistry. Origami have been
previously used to organize individual reactive chemi-
cal groups with
∼
5 nm precision
13
;
each origami
serves as an independent nanoscopic reference frame
and provides a 100 nm
70 nm
fi
eld in which single-
molecule reactions can be positioned. Uniquely, our
work has used directed self-assembly to deliver indivi-
dual reactive chemical groups
;
amines along the
inner triangle of the origami
;
to precise positions
on a macroscopic surface, in the laboratory frame
of reference. Here the amines were positioned in the
service of covalently coupling origami to the surface,
but this need not be the case. Looking forward, one can
envision uses of placement wherein origami are merely
a shuttle, escorting molecular cargo into position for
covalent coupling and then departing, leaving behind
single molecules precisely located with respect to the
laboratory frame
;
iterative applications of placement
might underlie a new type of synthesis with exquisite
positional control.
MATERIALS AND METHODS
Origami Formation and Purification.
A variation of the
“
sharp
triangle
”
design described previously
2
was chosen because
of its rigidity and low tendency to aggregate. Staple strands
(Integrated DNA Technologies, 100
μ
M each in water) and the
scaffold strand (single-stranded M13mp18, Bayou Biolabs,
P-107) were mixed together to target concentrations of 100
and 40 nM, respectively (a 2.5:1 staple/scaffold ratio) in 10 mM
Tris base, 1 mM EDTA buffer (adjusted to to pH 8.35 with HCl)
with 12.5 mM MgCl
2
. Next, 50
μ
L volumes of staple/scaffold
mixture were heated to 90
°
C for 5 min and annealed from
ARTICLE