Reversible Covalent Stabilization of Stacking Contacts in DNA Assemblies

Abstract

Stacking bonds formed between two blunt-ended DNA double helices can be used for reversibly stabilizing higher-order complexes that are assembled from rigid DNA components. Typically, at low concentrations of cations stacking bonds break and thus higher-order complexes disassemble. Here, we present a site-specific photochemical mechanism for the reversible covalent stabilization of stacking bonds in DNA assemblies. To this end we modified one blunt end with the 3-cyanovinylcarbazole (^cnvK) moiety and positioned a thymine at the other blunt end. In the bound state, the two blunt-ended helices are stacked on one another resulting in a co-localization of the ^cnvK and the thymine. We show that such a configuration is sufficient to efficiently induce the formation of a covalent bond across the stacking contact upon irradiation with 365 nm light. The thus created covalent bond can be split again upon irradiation with 310 nm light. Moreover, our mechanism also allows to repeatedly form and break the same covalent bond on the timescale of seconds. Using our system will expand the range of conditions under which stacking-bond stabilized objects may be utilized.

The sequence-programmable self-assembly of DNA single-strands enables the bottom-up fabrication of arbitrarily shaped, rigid, three-dimensional DNA components^[1]. These components can be used as building blocks for the self-assembly of yet larger objects that can reach molecular weights up to the gigadalton-scale with dimensions comparable to those of natural viruses^[2]. Such higher-order complexes can be stabilized through non-covalent base pair stacking bonds between helical blunt ends^{[2a, 3]}. Depending on the number and type of blunt-end stacking contacts used in the system, forming bound states may require concentrations of cations much beyond those found e.g. in physiological fluids. The strong dependence on cation concentration therefore limits the applicability of the higher-order objects formed via shape-complementary stacking bond schemes. Here, we solve the stability problem and expand the range of conditions under which stacking-bond stabilized objects may be utilized. To this end we present a photochemical strategy for site-specifically introducing and breaking covalent bonds across stacking contacts within higher-order DNA complexes.

In 2008, Yoshimura and Fujimoto reported that a nucleobase modified with 3-cyanovinylcarbazole (^cnvK) may be covalently linked to an adjacent pyrimidine base on the complementary strand^[4]. Both the ^cnvK and the pyrimidine were located within a continuous DNA double helix and crosslinking proceeds through a [2+2]-cycloaddition reaction induced by UVA irradiation^[5] (Figure 1a). Our mechanism builds on these findings, but in our case, the covalent bond will be formed between the blunt ends of two separate DNA double helices when they form a stacking bond (Figure 1b). To this end, we replaced a terminal base with the ^cnvK moiety in one blunt end and positioned a thymine at the other corresponding blunt end. To test the possibility for photo-crosslinking and photo-splitting of covalent bonds across stacking contacts, we used our previously described DNA origami switch object^{[2a, 6]} (Figure 1b). The object is built from two rigid bundles of DNA double helices that represent the two arms of the switch. The arms are connected in the middle by a single Holliday junction that acts as the pivot point. The closed state is stabilized by up to 32 stacking interactions between blunt-ended double helical protrusions on one arm and recessions on the other arm. At low ionic strength, switch particles assume open x-shaped conformations, while at high ionic strength switch particles populate the closed brick-shaped conformation.

Figure 1.

Schematics of photo-crosslinking and photo-splitting of ^cnvK-modified nucleosides across stacking contacts. a) Covalent photo-crosslinking within the context of a continuous DNA double helix. The cyanovinyl group of the ^cnvK-modified nucleobase (highlighted in light green) can be covalently linked to the thymine on the ‘-1’ position (from 5’ to 3’) of the complementary strand. The black ellipsoid represents the covalent bond. b) Covalent photo-crosslinking within the context of a stacking contact in DNA assemblies. The stacking contact consists of double helical protrusions (highlighted in red) and double helical recessions (highlighted in blue) that form discontinuous DNA double helices in the bound state. (1) – (4): Scheme of photo-crosslinking and photo-splitting using the DNA origami switch object. (1) Open conformation of the switch object under low ionic strength conditions. The ^cnvK moiety and the thymine (T) are separated in space. The black arrow indicates the position of the second ^cnvK-modification (Figure S1). (2) Closed conformation of the switch object under high ionic strength conditions. The ^cnvK moiety and the thymine are co-localized in space. (3) Formation of the covalent bond between the ^cnvK and T upon 365 nm irradiation. (4) The stacking contact is under tension when decreasing the cation concentration. Splitting the ^cnvK-T-covalent bond can be triggered upon 310 nm irradiation.

To implement our mechanism, we introduced two ^cnvK moieties at stacking contacts next to the pivot point of the switch object (Figure S1). At low ionic strength conditions switch particles sample open states (Figure 1b-(1)). In these states, the ^cnvK moiety and the thymine are far apart from each other and we expect that the covalent bond cannot be formed across the stacking contact. Increasing the concentration of cations in solution shifts the conformational equilibrium of the switch towards the closed state. In this state, we expect that the ^cnvK moiety and the thymine will be properly positioned (Figure 1b-(2)) to enable the formation of a covalent bond upon 365 nm light irradiation (Figure 1b-(3)). Decreasing the concentration of cations in solution after irradiation will drive the two arms of the switch again away from each other towards the open state. However, the object will remain in the closed state due to the newly created covalent bond across the stacking contact (Figure 1b-(4)). The conformational transition to the open state may then be remotely triggered through irradiation with light of shorter wavelength (e.g. 310 nm), which will split the covalent bond between the ^cnvK moiety and the thymine.

We performed experiments to test for the possibility of forming covalent bonds across stacking contacts and to elucidate the kinetics of photo-crosslinking. We incubated switch particles at the high cation concentrations (30 mM MgCl₂) that stabilize the closed state and exposed samples to 365 nm light. Subsequently, we gel-electrophoresed the irradiated samples at the low cation concentrations that induce the transition to the open state. As can be seen by evaluating the band intensities in the gel image shown in Figure 2a (left), the fraction of high-mobility, closed particles increases with increasing irradiation time, even though all samples are exposed to object-opening conditions. Therefore, we conclude that the high-mobility particles were covalently crosslinked and thus trapped in the closed state. The fraction of crosslinked particles saturates after approximately 10 seconds of exposure to 365 nm light (Figure 2b (left)). The yield of crosslinked particles not only depends on the irradiation time, it also increases with increasing MgCl₂ concentrations when comparing fixed irradiation times (Figure S3). The crosslinking reaction may be considered as a first-order rate equation. For constant illumination intensity and constant cation concentration, we then expect the concentration of crosslinked particles to increase exponentially with illumination time, which is what we observe (Figure 2b (left, solid line)). However, in this model the yield of crosslinked particles should always go to 100% for sufficiently long illumination time. Instead, we observe that the maximally attainable yield is around 75%. We attribute the incomplete crosslinking to structurally defective switch particles that cannot adopt the closed state on the one hand and to particles missing the ^cnvK modifications due to synthesis defects on the other hand. The first factor could be addressed by purifying particles capable of forming the closed state as we previously described^[2a]. The second factor could be addressed by introducing more ^cnvK moieties. In support, we note that reducing the redundancy from two to one ^cnvK modifications per switch particle resulted in a reduced maximally attainable crosslinking yield (Figure S4). Direct imaging with negative-stain transmission electron microscopy (nsTEM) provided complementary evidence for successful photo-induced crosslinking. As expected, non-irradiated switch particles dissolved in low ionic strength buffer adopted the open x-shaped conformation (Figure 2c (left)). By contrast, switch particles that we previously irradiated at high ionic strength conditions and then dissolved in low ionic strength buffer to induce opening appear as closed brick-shaped particles (Figure 2c; right), which thus shows successful crosslinking.

Figure 2.

Kinetic characterization of photo-crosslinking and photo-splitting using the DNA origami switch object. a) Fluorescent images of 1.5% agarose gels laser-scanned in the Cy5-channel (635 nm). Time series of photo-crosslinking at 365 nm irradiation (left) and photo-splitting at 310 nm irradiation (right). p: pocket, o: band corresponding to open (non-crosslinked) switch particles, c: band corresponding to closed (crosslinked) switch particles. The images of the gels are globally autoleveled. b) Evaluation of the gel data shown in a). We performed the experiment in triplicate (Figure S2); data points represent the mean and error bars the standard deviation. The fraction of crosslinked particles is computed as described in the methods section (4. UV-irradiation). We fitted the photo-crosslinking data by using a single exponential with rate constant k_c and the photo-splitting data by using a double exponential with rate constants k_s1 and k_s2 (solid lines). We obtained for the rate constants of photo-crosslinking and -splitting k_c = 0.7 s^-1, k_s1 = 0.02 s^-1, and k_s2 = 0.3 s^-1, respectively. c) Negative-stain TEM data of non-irradiated (non-crosslinked) switch particles (left) and irradiated (crosslinked) switch particles (right). Scale bars: 100 nm. See Figure S6 for field-of-views. Average 2D-particle micrograph of the ^cnvK-crosslinked switch object (rightmost). Scale bar: 25 nm.

Next, we tested for the possibility of breaking the covalent bonds again and evaluated the kinetics of photo-splitting (Figure 2a; right). We incubated previously crosslinked switch particles at low, object-opening cation concentrations (5 mM MgCl₂) and exposed the samples to 310 nm light irradiation. Gel electrophoresis under object-opening, low-ionic strength conditions (5 mM MgCl₂) reveals that the fraction of low-mobility, open particles increases with increasing irradiation time (Figure 2a (right)). Hence, the covalent bonds across stacking contacts can also be split again. The photo-splitting reaction goes to completion after about 100 seconds of exposure to 310 nm light (Figure 2b (right)). Moreover, we observe that the yield of photo-splitting increases with decreasing MgCl₂ concentrations (Figure S5). This finding suggests that a destabilizing force is needed to drive the ^cnvK moiety and the thymine away from each other. In support, Yoshimura and Fujimoto previously achieved photo-splitting of ^cnvK bonds within continuous double helices only in the presence of denaturing urea^[4]. The time-dependence of the yield of photo-splitting (Figure 2b (right)) may be understood by considering that the closed species can contain particles having two or one intact ^cnvK bonds, which cannot be discriminated by their electrophoretic mobilities. Breaking the bond in the species having only one bond should be a simple single-exponential decay reaction. By contrast, the conversion of the species with two bonds to the open state with zero bonds proceeds via an obligate intermediate state with one bond that is hidden to the observer resulting in overall slower decay kinetics. Our data obtained from gel electrophoresis can be described favorably with the superposition of two single exponentials having a fast and a slow time constant (Figure 2b (right, solid line)).

We also tested whether multiple iterations through the photo-crosslinking and photo-splitting cycle put forward in Figure 1 can be experimentally realized. For each cycle, we irradiated the sample for 5 minutes with 365 nm light for crosslinking, exchanged the buffer, and irradiated the sample for 5 minutes with 310 nm light to split the covalent bonds again. We followed one cycle in detail by directly imaging switch particles with nsTEM (Figure 3a) and monitored nine successive cycles via gel electrophoresis mobility analysis (Figure 3b). Evaluation of the gel data reveals that the fraction of crosslinked switch particles gradually decreases from cycle to cycle (Figure 3c), which we attribute to photochemical side reactions that lead to non-functional ^cnvK modifications. However, this effect can be mitigated by reducing the exposure time to 310 nm light (Figure S9).

Figure 3.

Reversibility and cycling. a) Negative-stain TEM data of one cycle of photo-crosslinking and photo-splitting that is illustrated in Figure1. Scale bars: 100 nm. See Figure S7 for field-of-views. b) Fluorescent image of a 1.5% agarose gel laser-scanned in the Cy5-channel (635 nm). Samples of nine successive cycles of photo-crosslinking and photo-splitting are loaded on the gel. c: crosslinking with 365 nm light for 5 minutes, s: splitting with 310 nm light for 5 minutes (top labels). p: pocket, o: band corresponding to open (non-crosslinked) switch particles, c: band corresponding to closed (crosslinked) switch particles (left labels). The image of the gel is globally autoleveled. c) Evaluation of the gel data shown in b). We performed the experiment in duplicate (Figure S8); data points represent the mean and error bars the standard deviation. The fraction of crosslinked particles is computed as described in the methods section (4. UV-irradiation). c: crosslinking at 365 nm, s: splitting at 310 nm.

In summary, we have presented a mechanism that allows to reversibly form and split covalent bonds between separate blunt-ended DNA double helices when they form a stacking bond. To this end, the terminal base of one blunt-ended helix of the stacking contact is modified with the ^cnvK moiety and a thymine is positioned at the other corresponding blunt-end. Structural context provided by the DNA origami object enables positioning the two blunt-ends of the separate DNA double helices in a base pair stacking configuration. As a result, the ^cnvK moiety and the thymine are co-localized in such a manner that 365 nm light irradiation can induce efficiently the formation of a covalent bond across the stacking contact. Hence, the DNA origami object acts as a templating device for the photochemical reaction. Moreover, the system that we implemented allows to repeatedly form and break the same covalent bond in multiple cycles. We also found that a destabilizing force is necessary to efficiently split the thus formed bond through irradiation with 310 nm light. In our experiments we generated this force by increasing the electrostatic repulsion between the two blunt-end carrying interfaces of the switch object, thereby pulling the ^cnvK moiety and its thymine binding partner apart.

The strategy described herein for the reversible covalent stabilization of stacking contacts in DNA assemblies has direct practical utility. Our mechanism opens access to prepare pure and dense solutions of stacking-bond stabilized higher-order DNA complexes, as needed for potential in vitro and in vivo applications, for crystallization assays, or for cryo-EM imaging. With ^cnvK-crosslinking across stacking contacts, higher-order complexes may be covalently linked in their fully assembled states. Once covalently stabilized, higher-order assemblies can be subjected to purification and enrichment procedures, which typically require low ionic strength buffers to obtain satisfying recovery yields. Under these conditions, higher-order assemblies would normally break down into their individual building blocks unless further stabilized. Furthermore, ^cnvK-crosslinking of stacking contacts may be also used for photo-caging of mechanically interlocked DNA assemblies such as rotaxanes^[7] and other dynamic mechanisms in which movable parts need to be encapsulated^[8]. Taken together, our mechanism opens up new possibilities for sample handling and photo-caging applications in the context of structural and dynamic DNA nanotechnology.