As demonstrated by means of DNA nanoconstructs1, as well as DNA functionalization of nanoparticles2,3,4 and micrometre-scale colloids5,6,7,8, complex self-assembly processes require components to associate with particular partners in a programmable fashion. In many cases the reversibility of the interactions between complementary DNA sequences is an advantage9. However, permanently bonding some or all of the complementary pairs may allow for flexibility in design and construction10. Here, we show that the substitution of a cinnamate group for a pair of complementary bases provides an efficient, addressable, ultraviolet light-based method to bond complementary DNA covalently. To show the potential of this approach, we wrote micrometre-scale patterns on a surface using ultraviolet light and demonstrated the reversible attachment of conjugated DNA and DNA-coated colloids. Our strategy enables both functional DNA photolithography and multistep, specific binding in self-assembly processes.
The highly specific, thermoreversible base-pair interactions that are formed by complementary DNA strands play an important role in biology and have recently been exploited to selectively bind and create a host of potentially important nanostructures such as branched junction motifs containing crossovers11. Using this strategy, functional systems including crystals4,10,12,13,14,15,16, motors/robots15,16, computers17,18, replicators9,19 and machines20 have been fabricated. For nanotechnology applications as well as biological assays, however, crosslinking some or all of these bonds permanently can provide stability to higher- order constructs, thereby introducing a robust static self-assembly strategy. For instance, this strategy has been demonstrated to be powerful in providing structural information on biomacromolecules when standard techniques were not applicable21.
Several different crosslinking methods have been investigated. The most common methods employ the crosslinking agent psoralen that forms covalent links to thymines on exposure to ultraviolet light22. Unfortunately, the crosslinking efficiency for psoralen in solution with DNA is relatively low (effective cross-section σ = 5 × 10−6 nm2, percentage of crosslinking ~σΦt, where Φ is photon flux in photons nm −2 s−1, and t is time in seconds) in comparison with other crosslinking agents (which will be further discussed below); otherwise it requires the attachment of the psoralen group at the 5′-TpA site (T–A sequence) of a DNA sticky end23. Alternative ultraviolet crosslinking agents for DNA that allow for higher efficiency, increased specificity and flexibility in sequence placement are being pursued. The use of p-carbamoylvinyl phenol nucleoside (p-CVP; ref. 24) and 3-cyanovinylcarbazole nucleoside (CNVK; ref. 25) as ultraviolet crosslinking agents has been popularized as a result of higher crosslinking efficiencies than psoralen (~3 × 10−5 nm2 and 1.3 × 10−4 nm2 for p-CVP and CNVK, respectively). For crosslinking to take place, in these cases, p-CVP requires specific placement next to an adjacent adenine base, and CNVK must be adjacent to a pyrimidine.
Here we report a highly selective and efficient interstrand crosslinking methodology using the p-CVP group, herein referred to as cinnamate, where two cinnamate groups are placed directly across from each other on complementary strands. We demonstrate that this strategy increases crosslinking specificity and efficiency, and that introducing these cinnamate-modified DNA strands onto surfaces allows for the realization of a photolithography application.
Figure 1 shows the cinnamate-containing nucleoside and the two cycloaddition products that can occur on exposure of the cinnamate to 360–390 nm ultraviolet light. Synthetic procedures are described in Supplementary Section S1. The size of the cinnamate-based artificial base is comparable to that of natural purines or pyrimidines; that is, its structural perturbation is comparable to the effect of one mismatched base pair (see Supplementary Section S2). The cinnamate nucleoside was incorporated into DNA strands resulting in the key building blocks for the guided self-assembly and ultraviolet crosslinking units of DNA-functionalized colloids.
Initially, we incorporated the cinnamate groups into DNA sequences and investigated whether or not having two cinnamate groups directly across from each other inhibits hybridization of complementary strands. We demonstrate that the hybridization was not affected (see below). Next, we determined whether this new strategy allows for the successful crosslinking of complementary strands. Finally, we introduced the functionalized DNA onto a gold surface as well as onto colloids to determine whether this strategy allows for the realization of a photolithographic process.
Two of the DNA strands used in this study (strands A and B) consist of two major parts: an inert section and a sticky end. The longest sections of strands A and B are identical inert sequences, which hybridize with the same complementary strand (CS) to form a double-strand duplex. A short single strand of DNA, a sticky end, was placed at the 3′ end of the inert section. The sticky end was used as the directing and recognition unit to bind to a complementary sticky end of the desired particles5,6,7,8,26 and surfaces27 used in our studies. The sticky ends of the two strands contain complementary cinnamate-modified sequences. The sequences used are: (A) 5′-ATC GCT ACC CTT CGC ACA GTC AAT CCA GAG AGC CCT GCC TTT CAT TAC GAC CAA GT-Crosslinker-T ATG A - 3′ (B) 5′-ATC GCT ACC CTT CGC ACA GTC AAT CCA GAG AGC CCT GCC TTT CAT TAC GAT CAT A-Crosslinker-AC TTG G- 3′ (CS) 5′-TCG TAA TGA AAG GCA GGG CTC TCT GGA TTG ACT GTG CGA AGG GTA GCG AT- 3′. We inserted the cinnamate nucleoside at the seventh nucleoside position (5′–3′) within the 11-base sticky end of the A strand, as well as at the associated sixth nucleoside position along the B strand. Such 11-base sticky ends were used in earlier studies5,6,7,8,26,27 and have been well characterized. Synthetically, the cinnamate group could be placed at any position; however, the crosslinking efficiency is increased when the two cinnamate groups are held in close proximity, and placing the cinnamates towards the centre of the complementary sticky ends gives a slight advantage. The A and B DNA strands were annealed with an equal number of CS strands to hybridize and form a construct with a 50 nucleotide-pair rigid duplex followed by an 11-nucleotide sticky end.
We first investigated whether the cinnamate-functionalized DNA strand binds selectively with its complementary strand. A denaturing gel (see Methods; Fig. 2a) de-hybridizes non-covalently-bound DNA strands without disruption of the crosslinked strands. Lanes 1 to 3 were loaded with the DNA strands without ultraviolet exposure. Lanes 4 to 6 were loaded with DNA strands that were exposed to 350 nm ultraviolet light for five minutes (refer to the caption of Fig. 2a for complete experimental details). As shown in lanes 1 to 3, after running all of these strand combinations on the denaturing gel, only single-stranded DNA was obtained, demonstrating that these strands are held together by reversible hydrogen bonds. On ultraviolet exposure (10 photons s−1 nm−2; see Methods for ultraviolet calibration) for 5 min, however, we observed substantial crosslinking between the complementary strands A and B (lane 6) that increased after 15 min of ultraviolet light exposure (lane 9). As a control, strand A with CS (lane 4) and strand B with CS (lane 5) were also exposed to ultraviolet light but did not exhibit any new band corresponding to crosslinked product. Furthermore, although A and B are complementary strands, no crosslinking is observed without ultraviolet exposure, as shown in lane 3, because no new band for the hybridized product is visible. These results demonstrate that two cinnamate-modified DNA strands allow for crosslinking only when the two cinnamate nucleosides are held in juxtaposition by flanking complementary sequences in the sticky ends. The efficiency and time dependence of cinnamate crosslinking is shown in Supplementary Section S3. The effective linking cross-section was found to be 3 × 10−5 nm2.
Next, we modified the aforementioned strands to be able to functionalize colloidal particles (see Methods for DNA-coated colloids). We used 1 μm streptavidin-coated poly(styrene) particles for these experiments. We added biotin (and tetraethylene glycol as the terminating group at the 5′ end) to sequences A and B to attach the DNA to the streptavidin-coated particles. A similar construct was fabricated where the sticky ends were self-complementary palindromic sequences (P: 5′-AATCATGATT- 3′) lacking cinnamate. The P DNA sequences were designed so that the particles have a lower melting temperature (~40 °C; ref. 26, normal buffer solution, defined in Methods) than the A and B complementary pair (~45 °C, in normal buffer). We then fabricated five sets of particles functionalized with different combinations of the cinnamate-functionalized DNA strands: A, B, P, A+P and B+P, where the A+P and B+P functionalized particles have half of their surface randomly coated with P strands and the other half with A or B strands26. We carried out four sets of experiments using these particles to investigate whether it is possible to bind them permanently using a combination of DNA hybridization and ultraviolet-light-triggered crosslinking.
In the experiments described below, the sample was first annealed for one hour by cooling it slowly from 55 °C to room temperature, followed by exposure to ultraviolet light (16 s at 60 photons s−1 nm−2) at room temperature. The sample was then heated to 55 °C for 10 min. The P-coated particles completely dissociated on heating (Supplementary Fig. S5), indicating that ultraviolet exposure did not link these DNA-functionalized particles. The mixture of A-functionalized particles and B-functionalized particles, and the mixture of A+P functionalized particles and B+P functionalized particles showed minimal dissociation (Fig. 2b, bottom left and Supplementary Fig. S6). Ultraviolet irradiation successfully crosslinked the cinnamate-modified paired strands. The palindrome strands (P) did not interfere with the cinnamate crosslinking. The most revealing experiment had only particles A+P (Fig. 2b, bottom right) that nearly completely dissociated after heating. Although the cinnamate groups on the non-complementary strands of two particles were held together within a range of ~ 20 nm, ultraviolet exposure was not effective for crosslinking in this case.
Figure 2c compares the fraction of unbound particles for the equal mixture of complementary cinnamate-modified A- and B-functionalized particles to the A+P functionalized particles. This is a particularly sensitive measure because these particles are typically joined by approximately 150 sticky-end pairs containing cinnamate26. If any pair of them were to be crosslinked, the complementary strands would remain stuck together. As demonstrated in Fig. 2c, the singlet fraction for the A and B particle mixture is minimal in comparison with the A+P particles. These data suggest that the selectivity of the cinnamate to itself in complementary versus non-complementary strands is ~ 300:1.
After we demonstrated the efficiency and specificity of the cinnamate-modified DNA strands on particles, we targeted photolithography as an application for these crosslinkable particles, so as to demonstrate the utility of our methodology. The crosslinking can be selectively activated by ultraviolet exposure on a substrate,allowing access to multi-functionalized DNA surfaces that are highly desirable for bio-medical applications28. We modified our design strategy for the DNA sequences to include a three-component structure: a surface strand (SS), a particle strand (PS) and a linker strand (LS; sequences included in the caption of Fig. 3). The SS was end-functionalized with a disulphide group to allow for attachment to a gold surface27. The SS consists of a 50-base-pair DNA backbone (Fig. 3a, black) and a 22-base-pair sticky end (Fig. 3a, purple) containing the cinnamate nucleoside (red). The LS strand has two parts: a cinnamate-containing sequence complementary to SS (Fig. 3a, purple) and a functional DNA sequence that is designed to bind specifically to PS (Fig. 3a, blue). Finally, the target material, in our case fluorescently labelled PS or colloidal particles, contains PS strands that are complementary to LS (Fig. 3a, blue). The LS–SS DNA was designed to have a melting temperature without ultraviolet radiation of ~45 °C (0.1 μM strand concentration, in normal buffer, see Supplementary Section S2), which ensures efficient crosslinking at room temperature.
Figure 3b shows schematically the procedures used for the preparation of the samples for photolithography. First, a gilded surface was coated with SS (see Methods for DNA-coated gold surface). Next, LS was hybridized to SS by annealing from 55 to 25 °C over a 30 min period (steps 1 and 2). Then, the samples were exposed to ultraviolet light to permanently crosslink LS to SS in the exposed regions (step 3). Finally, the sample was heated to 55 °C and washed three times with normal buffer to de-hybridize any unlinked strands (step 4). The functionalized DNA surface was then ready to use for the reversible binding of colloids or fluorescently labelled DNA to the patterned regions (steps 5 and 6).
Figure 4a shows images of the surfaces decorated by 1 μm poly(styrene) particles (left) and fluorescently labelled DNA (middle), both within the region exposed with ultraviolet light through a Y-shaped photomask (right). Similar experiments were carried out using an NYU photomask of different feature sizes (Fig. 4b, right). Both fluorescent and colloid images show a minimum feature size of ~2 μm (Fig. 4b). (See Methods for light microscopy and confocal microscopy.)
Our particles have a gravitational height, kBT/mg, of approximately 1 μm and hence sediment towards the bottom. To determine whether the particles attach to the surface by DNA hybridization, after allowing the particles to hybridize with the complementary LS (Supplementary Movie S1) we inverted the sample. The particles remain on the surface even after inversion of the sample, verifying that the particles were attached to the surface by DNA hybridization between PS and LS (Fig. 4a,b, left). As the binding between LS and the colloids is thermoreversible, the colloids diffuse away from the patterned regions of the surface on heating to 55 °C (see Methods for temperature controlling stage and Supplementary Movie S2). In addition, cooling the sample leads to re-association (Supplementary Movie S3). We confirmed the effective cross-section, 3 × 10−5 nm2, of the cinnamate by colloid-surface melting temperatures (Supplementary Section S4).
To make a spatially dependent multi-functionalized surface and demonstrate the versatility of our method, we first carried out the photolithography as described above (steps 1 to 6), and then repeated steps 3 and 4 on a different region of the gold surface using modified linker strands (steps 7 and 8). The new linker strand 2 (LS2) contains a functional molecule (biotin in this case) at its 5′ end, unlike the single-stranded sticky-end of LS. The newly functionalized region of the gold surface with LS2 strands was visualized by the conjugation of red fluorescent streptavidin (see Methods for fluorescent imaging). The sticky LS in the first exposure region remained functional as tested by the addition of fluorescently labelled PS strands (step 9).
Figure 4c shows the result of our multi-functionalization experiment, after step 9 of Fig. 3b. The green fluorescently labelled PS strand sticks to the LS patterned region (letters of ‘materials’ in Fig. 4c, left) by complementary DNA hybridization. Red streptavidin binds to the LS2-patterned biotin region (letters of ‘nature’ in Fig. 4c, left). The distinct colours in separate filtering channels confirmed the successful patterning of two different functions without interference (see Methods for confocal imaging).
Finally, we investigated the pattern resolution (Fig. 5a). We exposed surfaces that were decorated with fluorescently labelled DNA and colloids to ultraviolet irradiation for different durations of time through an inverted U-shaped photomask. The light intensity and particle concentration were analysed through a cut across the two straight legs of the inverted U. Looking at the intensity profile (bottom row), the black curves represent the relative particle concentration, the green curves represent the intensity of green fluorescent light from the confocal images, and the red curves are a linear rescaling of the exposed peaks based on estimated coverage (from left to right: 8%, 15%, 28% and 48%; see Supplementary Section S4; the last panel is the exposure profile (bottom row)). The images and curves show that shorter exposure time (5 s) led to weak fluorescent light and non-uniformity in colloidal patterns (Fig. 5a, left) due to inhomogeneous crosslinking. Increasing exposure time yielded a stronger fluorescent signal and complete colloidal patterning, but it also lowered the pattern resolution by producing undesired crosslinking in the peripheral regions (Fig. 5a, right). An exposure time of 10 s gave a good balance between uniform functionalization and high resolution.
Figure 5b shows the pattern image superimposed on the colloid and fluorescent images. The line width is ~1.5 μm in the exposure pattern and ~2 μm in the fluorescent image, which suggests at least a 1 μm resolution on the edges. Although this resolution is slightly lower than standard ultraviolet photolithography in vacuum (resolution of ~0.8 μm; refs 29, 30) or direct laser writing31, the resolution obtained using this technique is competitive with other solution-based methods using DNA for photolithography32. This multi-functionalized DNA technique with micrometre-scale resolution might prove useful in DNA chips for gene identification and for directing colloidal self-assembly.
We have introduced a simple, highly selective and efficient method to covalently bind complementary DNA strands in solution and on surfaces using ultraviolet-irradiation-triggered crosslinking as an optional separate step. As DNA is a functional material, these techniques can also be used to bind reversibly and/or permanently, through DNA linkers, an assortment of molecules, proteins and nanostructures. Potential applications range from the simple permanent joining of particles, to advanced self-assembly protocols that assemble support scaffolds, and also a multi-dimensional system in which a support scaffold is first assembled, a structure formed and crosslinked on the scaffold followed by reversible disassembly of the scaffold. Using this methodology, we have developed a new photolithography method capable of functionally and chemically patterning surfaces down to a 1 μm resolution. Our photolithographic technique opens the door for the possibility to write chemical and physical patterns, or to make high-resolution multi-functionalized DNA surfaces for genetic detection33 or DNA computing34. Future studies using total internal reflection microscopy may allow for the exposure of only a few hundred nanometres of a particle to ultraviolet light. This may enable chemically different patches to be printed on micrometre-sized colloidal particles for medical and soft-matter research uses in a more controlled manner35,36.
Synthesis of p-carbamoylvinyl phenol nucleoside.
The cinnamate-functionalized phosphoramidite was obtained in four high-yielding steps (Supplementary Section S1). Briefly, the glycosylation of the cinnamate derivative with Hoffer’s α-chloro saccharide was achieved using caesium carbonate. Sodium methoxide was employed to remove the p-toluyl protecting groups, which afforded free hydroxyl groups at the 3′ and 5′ positions of the saccharide. Finally, the 5′-OH was protected with dimethoxytritylchloride followed by the conversion of the 3′-OH to a phosphoramidite. The cinnamate-containing phosphoramidite was incorporated into a DNA oligonucleotide using published procedures37.
Ultraviolet power output was measured with a SiC photodiode with a peak sensitivity at 340 nm. The power output 1 cm from the ultraviolet lamp was ~0.6 mW cm−2 or ~10 photons s−1 nm−2. For the Hg lamp (Leica 106z lamp housing with a 50 W mercury burner) with a × 100 Leica air objective on a Leica DMRXA microscope with a Leica Type A filter cube, the power output is ~60 photons s−1 nm−2 as in the crosslinking experiment on the particles. For the external Hg lamp that was used for the photolithography experiments (Leica EL6000) with a × 63 Leica air objective, the power output is ~30 mW cm−2 or ~500 photons s−1 nm−2.
The gels contained 8.3 M urea and were run at 55 °C. The running buffer consisted of 89 mM Tris–HCl, at pH 8.0, 89 mM boric acid and 2 mM EDTA (TBE). The sample buffer consisted of 10 mM NaOH, 1 mM EDTA, a trace amount of xylene cyanol FF, and bromophenol blue tracking dye. Gels were run on a Hoefer SE 600 electrophoresis unit at 55 °C (31 V cm−1, constant voltage). When further denaturation was required, a 6% acrylamide gel solution containing 7.0 M urea and 41% formamide, or a 4% acrylamide gel solution containing 6.8 M urea and 47% formamide, was substituted for a regular denaturing gel.
DNA-coated gold surface.
A coverslip (2.5 cm × 2.5 cm) was cleaned with acetone, plasma etched using a SPI supplies Plasma Prep II, and coated with 5 nm of chromium and 40 nm of gold (SIGMA-ALDRICH 99.999%) using a BAL-TEC MCS 010 Multi Control System evaporator. An end-functionalized thiol surface strand (72 base pairs) was annealed to its complementary strand (49 base pair) by heating it to 95 °C and then cooling it to 25 °C, resulting in the formation of a rigid double-stranded backbone. The sample was then incubated on the gold surface for 12 h under the following conditions: 25 °C normal buffer (10 mM PB, 50 mM NaCl, 1% w/w F127, pH 7.4) with a thiol DNA concentration of 40 μM. The small incubation chamber was sealed with vacuum grease to avoid evaporation. After incubation, excess strands were washed away using normal buffer at 55 °C. This procedure was repeated three times. The resulting surface had a DNA density of ~3,000 μm−2 as determined by radioactive measurement27.
Preparation of DNA-coated colloids.
Streptavidin-covered poly(styrene) particles (1 μm) were purchased from Invitrogen (Dynabeads MyOne Streptavidin C1). Biotinylated particle strands with a double-stranded backbone were incubated with these particles for one hour under the following conditions: 25 °C, light shaking, normal buffer, particle volume fraction 0.05% and a biotinylated strand concentration of 0.5 μM. Particles were washed by centrifugation and re-suspensed in normal buffer. This process was repeated three times to wash away excess strands. The resulting surface had a DNA density of ~6,400 μm−2 as determined by radioactive measurement26.
A Leica DMRXA microscope with a Qimaging Retiga 1300 camera and a Leica external Hg lamp was used to observe colloidal assembly and to perform the photolithography experiments. A Leica Type A filter cube was used to provide 365 nm wavelength ultraviolet light for the cinnamate crosslinking. Black field masks were purchased from Fine Line Imaging and placed at the aperture conjugated to the sample image.
A temperature stage was built on a light microscope to provide fast in situ temperature control. Briefly, 1,000 Ω indium tin oxide glass was placed on a 3-mm-thick copper plate, two ends of which were connected to peltiers (2.5 cm × 2.5 cm) and then to a thermal sink with constant temperature. We were able to control and detect the temperature with < 0.5 °C relative error using a LakeShore DRC 93C temperature controller and a LakeShore PT-111 temperature sensor.
Confocal microscope and fluorescent imaging.
A Leica DM6000 CS confocal microscope (TCS SP5 II) was used to take all fluorescent images. Excitations with 488 nm and 631 nm were used. Streptavidin with Alexa Fluor 488 (Invitrogen) was used to dye the biotinylated complementary DNA (particle strand) that was conjugated to the DNA sticky end in the LS pattern. Streptavidin with Alexa Fluor 633 (Invitrogen) was used to directly conjugate and visualize the biotin-patterned region. A high-salt-concentration buffer (500 mM NaCl) was used to stabilize the DNA interaction.
This research has been partially supported by the MRSEC Program of the National Science Foundation under Award Number DMR-0820341 for the cinnamate-functionalized phosphoramidite, NASA NNX08AK04G for microscopy, and DOE-BES-DE-SC0007991 to P.C. for data acquisition and analysis, as well as by the following grants to N.C.S. for DNA synthesis and characterization: GM-29554 from the National Institute of General Medical Sciences, CTS-0608889 and CCF-0726378 from the National Science Foundation, 48681-EL and W911NF-07-1-0439 from the Army Research Office, and N000140910181 and N000140911118 from the Office of Naval Research. J. Romulus acknowledges support through the Margaret Strauss Kramer Graduate Student Fellowship in Chemistry.
About this article
Nature Communications (2014)