A nanoscale Jitterbug transformer from DNA Authors: Seongmin Seo, Alexander A Swett, Mallikarjuna Reddy Kesama, Anirudh S Madhvacharyula, Ruixin Li, Yancheng Du, Markus Eder, Friedrich C Simmel, Jong Hyun Choi Published: 2026 ABSTRACT AbstractMany viruses have evolved remarkably intricate polyhedral shells capable of undergoing symmetric transformations in response to external stimuli to initiate payload release. So far, such deployable auxetic nanostructures are not available in the synthetic realm. Here we present a nanoscale Jitterbug transformer realized by a DNA origami structure that can reconfigure its conformation upon chemical and optical signals while maintaining a Poisson’s ratio of −1. By combining mechanical design principles with molecular dynamics simulations, we design the DNA Jitterbug to form a compact octahedron that stores elastic energy and spontaneously transitions into an expanded cuboctahedron by releasing it. DNA transformers are demonstrated to act similar to viruses that can create nanopores on lipid membranes and regulate payload release into vesicles. Integrating programmable DNA self-assembly with free-energy-guided mechanical design, this work provides a pathway toward adaptive nanomaterials with potential in synthetic organelles and stimuli-responsive nanodevices.Subject terms: DNA nanostructures, Mechanical engineering Introduction Deployable auxetic materials can change geometry, switching from compact to expanded configurations, and vice versa, in response to external stimuli1–3. They exhibit unique mechanical behaviors with negative Poisson’s ratios4 (ν) which are advantageous for applications in aerospace engineering5, soft robotics6, and biomedical devices7. Especially, materials with ν = −1 show isotropic deformation, preserving geometric similarity and enabling force exertion and predictable changes in dimensions with the ability to stow into compact volumes. Similarly in nature, viruses such as cowpea chlorotic mottle virus8,9, brome mosaic virus10, and bacteriophage HK9711 undergo dynamic, reversible capsid reconfiguration in response to environmental cues such as pH change and ionic strength. These transformations scale pore size and permeability of lipid membranes while retaining symmetry, allowing for selective molecular exchange. Such naturally occurring deployable architectures highlight how intricate geometric transformations can regulate transport and mechanochemical interactions at the nanoscale. Despite their considerable utility, the development of nanoscale 3D deployable auxetics has been hindered by the difficulties of constructing precise geometries and actuating at such small length scales. Buckminster Fuller’s Jitterbug transformer12 is a classical example of a deployable mechanism13. It demonstrates a single degree-of-freedom transformation where a polyhedral framework reconfigures through coordinated rotations of rigid struts connected by flexible joints. The structure expands isotropically from octahedral to cuboctahedral states, and vice versa, during which six square faces emerge or disappear. The collective motion of its components with kinematic constraints results in a five-fold volume change14. These characteristics make the Jitterbug transformer a simple yet compelling model for construction at the nanoscale. DNA origami enables nanometer-scale assembly with exceptional precision and programmability15–17. This powerful approach has been used to construct a variety of systems such as nanoscale turbines18, ratchet motors19 and vesicle pincers20 rendering it as a versatile tool to construct stimuli-responsive reconfigurable devices. Static DNA polyhedra21 like icosahedron and dodecahedron were also constructed by strand routing and crossover placement using design tools like Cadnano22, Scadnano23 or automated workflows like Athena24 and Perdix25. However, deployable auxetics must be designed to ensure isotropic responses with structural symmetry. Proper design choices need to be made to bias their equilibrium configurations toward the deployed states based on desired functionalities. Their mechanical behavior may be tuned further by engineering structural components26, defects27, and chemical adducts28. Here we show a nanoscale, deployable Jitterbug transformer made of wireframe DNA origami via free-energy-guided mechanical design. Leveraging molecular dynamics (MD)29 and virtual-move Monte Carlo (VMMC) simulations30, we tune its free energy landscape and mechanical response such that transformations are driven by storing and releasing elastic energy. Reversible reconfiguration is demonstrated in response to external stimuli like toehold-mediated strand displacement (TMSD)31,32 and photocleavage. Furthermore, we utilize the Jitterbug DNA as a bioinspired actuator that converts the stored energy into useful mechanical work and a nanoscale platform that can regulate membrane permeability in synthetic vesicles, creating nanopores for programmed delivery of payloads. Results Free-energy-guided mechanical design A Jitterbug structure consists of eight identical equilateral triangles connected at shared vertices and switches between a regular octahedron and a cuboctahedron (an Archimedean solid with eight triangles and six squares) via a continuous, kinematic motion as shown in Fig. 1a and Supplementary Fig. 1. Each triangle undergoes a radial displacement combined with a rotation around its 3-fold symmetry axis, resulting in a screw-like motion and an overall structural response with ν = −1 (Supplementary Fig. 2)12,33. Free-energy-guided mechanical design of a DNA transformer. a Schematic of a Jitterbug undergoing symmetric transformation between octahedral and cuboctahedral states. b and c Design scheme and representation of DNA origami made of 28-nm-long 6HB edges on a honeycomb lattice. d Average free energy landscape as a function of opening distance. The cuboctahedral profile (right, including the dashed line) from umbrella sampling reveals a difference of ~40 kBT per square between closed and open conformations (marked by vertical lines in gray). VMMC-simulated energy profile (left) accounts for stabilization from DNA base-pairs in the octahedral state. Five independent simulations were performed to estimate the statistical uncertainties. The shaded region (red) indicates the uncertainty ( ± standard deviation, σ). e Opening distance and Poisson’s ratio during spontaneous transition after releasing the stabilization energy. Eight independent simulation replicates were conducted to ensure statistical robustness. Shaded regions represent ±σ. f, Transformation kinematics analysis between radial displacement and rotation angle. Portions of the conformation space are inaccessible due to the finite edge thickness. g Schematic and averaged equilibrium MD conformation of octahedral state stabilized by short jack edges. h Front view and averaged conformation of cuboctahedron with unpaired scaffold segments in the square face. Vertices A and B are four and three 10 ssnt connections, respectively. i Relative binding kinetics from DNAfold simulations for (i) one, (ii) two, (iii) three, and (iv) four staples per jack edge. In each of these cases, blue indicates the staple segments binding first, while those binding later are colored red. j Elastic energy of a single square during transformation. Three independent simulations were performed to confirm statistical reliability. Shaded band indicates ±σ. k Force distribution as a function of opening distance (with red region representing ±σ). Five independent simulation runs were performed to support statistical robustness. The area shaded in gray is the overall mechanical work. All simulation data for d, e, f, j, and k are provided in the Source Data file. The DNA Jitterbug is realized using wireframe DNA origami made of 6-helix bundles (6HB) arranged on a honeycomb lattice (Fig. 1b, c and Supplementary Figs. 3–5). The thick edges provide adequate rigidity during transformation, preventing unintended edge curvatures and enforcing similar conditions as that of the ideal mechanism. The structure is composed of 24 edges (from 8 constituent triangles), each approximately 28 nm long with the overall structure size of ~40 nm. All edges have similar staple routing, with deliberate differences in single-stranded nucleotide (ssnt) connections at the edges (Supplementary Fig. 3c, d) to bias the conformations toward the open, expanded state. The nanostructure was examined via coarse-grained MD simulations performed using oxDNA29 (see Methods). Principal component analysis (PCA) during structural transformation identifies the primary mode of deformation along the square opening direction (Supplementary Fig. 7). Due to symmetry, the extent of transformation can be projected onto a 1D reaction coordinate, which is set to the opening distance (Fig. 1a(iii)). The MD simulations provide useful design guidelines to (i) control reconfiguration via modulating the energy landscape of the structure, (ii) compare the behavior of the DNA construct with its theoretical counterpart, (iii) correlate overall structure responses with a simple model of its subunits subject to topological constraints, and finally (iv) quantify elastic energy storage and maximum possible mechanical forces that can be extracted from the transformation. The deployability of the Jitterbug was designed by tuning two main components that affect its free energy; namely, the inherent structural response of the design and external chemical stabilization that may be provided via DNA base-pairing. The structural free energy was estimated using umbrella sampling simulations34,35 as a function of the average opening distance of the six squares. As shown in Fig. 1d, the structure has a minimum at approximately 33 nm, revealing a natural bias towards the open state. A closer look at the dotted curve indicates that the difference in energy between the open and closed states (marked by light gray lines) is approximately 40 kBT per square. This is the minimum amount of energy needed to stabilize the octahedral configuration. Considering this energy, the minimum number of DNA base-pairs (bp) required for a stable octahedral state was estimated using NUPACK36. The calculation suggests that at least 20 bp are necessary. The free energy of the closed state is a summation of structural and stabilization energies. The latter was characterized as a function of opening distance using VMMC simulations in oxDNA. Stabilization via base-pairing enables a minimum in the closed state (the left well in Fig. 1d), negating the effect of increased strain in the structure. Both wells are nearly equal in depth since we chose the theoretical minimum number of base-pairs required for comparison. Upon removal of stabilization (i.e., with sufficient activation energy to overcome the barrier), the structure is expected to shift to its natural open state at the free energy minimum. We observed spontaneous transformation from octahedral to cuboctahedral configurations by removing the stabilizing base-pairs (Supplementary Movie 1). The average opening distance increased from approximately 12 to 33 nm (Fig. 1e and Supplementary Figs. 8, 9). Poisson’s ratio, averaged across three orthogonal planes (Fig. 1a(iv) and Supplementary Fig. 9b), is −1 ± 0.1 during the process, agreeing with the theoretical estimates (see Methods). We also analyzed the rotation of the triangular faces for their radial displacement from the origin. The rotation angle vs. the radial distance from the MD simulation closely matches theoretical predictions14 (Fig. 1f). However, portions of the initial and final transformation phases are inaccessible due to the finite edge thickness of the DNA construct. The required energy for stabilizing the octahedral configuration was supplied via adjustable DNA “jack” mechanisms, conceptually analogous to mechanical jacks that regulate structure through controlled length modulation26. We included jack edges in our design by placing them diagonally in an alternating fashion across the six square faces (Fig. 1g, h). This arrangement enables switching of the Jitterbug between two configurations by inserting a set of staples for short and long jacks (Supplementary Fig. 4). A key aspect here is that since the cuboctahedron is designed as the free energy minimum, the configuration is achieved regardless of jack staples. Our scaffold routing between 6HB edges in the square faces has two topological configurations: vertices A and B having four and three 10 ssnt connections, respectively (as shown in Fig. 1h inset). During reconfiguration, vertex A undergoes conformational changes while vertex B remains nearly unchanged. Mesoscopic simulations using a simplified jack design reconciled theoretical energy estimates with experimental feasibility. We estimated the normalized binding times between scaffold and jack staples for varying numbers of staples using DNAfold simulations37 (Fig. 1i). As more staples were added, the staples closest to the middle of the scaffold bound first, while those located at the ends of the scaffold bound later due to the high entropic cost of bringing the scaffold segments together. This is confirmed in Fig. 1i(iv), where staples are seen transitioning from blue to red (going right to left). For the Jitterbug, this cost is high due to natural preference towards the open state. Since this binding pattern correlates with experimental folding yields, we implemented four short jack staples to reliably fold each scaffold segment and thus stabilize the overall octahedral configuration. In our Jitterbug design, joints at vertices A and B behave differently but together contribute to the overall response. Due to its symmetry, the behavior of the overall structure may be understood as that of individual squares, subject to geometric restrictions enforced by the rest of the wireframe (i.e., restriction of vertex movement along orthogonal planes). We illustrate this using a simple model consisting of a square unit with triangles that contain both connection types. Planar confinement of vertices was enforced using repulsion planes in the oxDNA platform (Supplementary Fig. 12). The structural free-energy difference from this simple model (Fig. 1j) aligns with our estimate from Fig. 1d, which implies that each square can function independently. Notably, the elastic energy is stored due to careful vertex design, highlighting how small changes to the orientation and number of ssnt connections between neighboring edges dictates overall structural response. We also calculated the force required to maintain different configurations during the octahedron-to-cuboctahedron transition as a function of opening distance to estimate the mechanical work38 required throughout the transformation. For example, about 15 pN is needed to stabilize the octahedral state, whereas the cuboctahedral state requires negligible force due to its equilibrium configuration, as shown in Fig. 1k. Integration of the force profile over the distance yields the mechanical work that can be used for actuation (gray shaded area). This estimate is consistent with the free energy calculations, validating our free-energy-guided mechanical design. Understanding the means of energy storage within the structure provides further insights into the underlying mechanisms driving reconfiguration towards the open state. We estimate the average values of various energy components during equilibrium simulations of both states (Supplementary Fig. 13a). Notably, the majority of the energy difference is caused by disrupted hydrogen bonding and base stacking in the octahedral state (Supplementary Fig. 13b). Other contributions are moderate to negligible. This finding provides a complete picture of the overall transformation process; that is, the jack staples provide stabilizing energy to compensate for the disrupted hydrogen bonding and stacking in the octahedral state. Upon removal of this stabilization, the system naturally drifts towards the cuboctahedron where the disruption is mitigated. Jitterbug construction The size of the Jitterbug requires using two scaffolds in a single origami structure: M13mp18 (7249-nt) and p9072 (termed 9k scaffold)39 colored gold and black, respectively, in Figs. 1, 2. The origami design consists of 7k and 9k portions that are connected by linker staples (Fig. 2a). The structures were synthesized using one-pot assembly, where all staple strands and both scaffolds were mixed to form the initial cuboctahedra without jack staples (Fig. 2b(i)). Staple sequences were optimized computationally to minimize sequence overlap and reduce unintended hybridization (Supplementary Fig. 6). Construction and characterization of a DNA Jitterbug. a Design schematic of a DNA transformer assembled using two scaffolds: M13mp18 (7k scaffold, gold) and p9072 (9k scaffold, black) connected via linker staples. b Three distinct configurations characterized with AFM and TEM imaging. (i) initial cuboctahedra without jack staples, (ii) octahedra with short jacks, and (iii) cuboctahedra with long jacks. Scale bars: 100 nm. Comparison of c edge lengths and d opening distances from theoretical models, MD simulations, AFM, and TEM images. The number above each bar indicates the number of DNA structures counted. Error bars in the AFM and TEM data represent ±σ. Statistical details are provided in the Source Data file. e Agarose gel electrophoresis analysis of the three configurations. Octahedral and cuboctahedral configurations were built by introducing distinct types of jack staples to the initial cuboctahedron (Supplementary Fig. 4). Short jacks closed the squares diagonally, contracting the structure into an octahedron. Long jacks kept the squares open, preserving the cuboctahedral configuration. AFM and TEM images in Fig. 2b confirm these configurations. Compact octahedra are clearly observed with several jacks hanging loose, as in the MD structure. Cuboctahedra with and without jacks appear nearly identical, distinguished solely by the inclusion of long jacks in the square faces, confirming that the cuboctahedral state is thermodynamically most stable as indicated by the free energy profile. Consistency across different conformations was examined by quantitative comparison of edge lengths and opening distances. The edge lengths measured from MD, AFM, and TEM images are about 28 nm, agreeing with the theoretical design (Fig. 2c). Opening distances are also consistent, although they are slightly larger than the designed value in the octahedral state (Fig. 2d and Supplementary Fig. 33). Gel electrophoresis in Fig. 2e shows that both cuboctahedra migrate faster in the gel than the octahedral Jitterbug40. Jitterbug transformation We demonstrated the reversible Jitterbug transformation via adjustable jack edges using TMSD and jack staples. Figure 3a, b illustrates our transformation strategy, beginning with (i) a cuboctahedron without jack staples, where two unpaired scaffold segments are pre-positioned diagonally across each of the six squares to serve as binding sites for jack staples in subsequent steps. Staples for short jacks (red) contract the squares, transforming the structure into (ii) an octahedron. The introduction of ‘releasers’ removes short jack staples via TMSD, returning the structure to (i) a cuboctahedron. Likewise, long jack staples (blue) form extended jacks (iii), preserving the cuboctahedral geometry. TMSD using appropriate releaser strands can revert to (i) the cuboctahedron, as demonstrated in Fig. 3c. Reconfiguration via elastic energy storage and release. a, b Jack-mediated transformation mechanism. Starting from (i) initial cuboctahedron with unpaired scaffold segments across square openings, two pathways are possible: (ii) an octahedron is formed by shutting the opening with short jack staples (red), while long jacks (blue) maintain the square opening preserving (iii) cuboctahedral geometry. Reversible reconfiguration is initiated by TMSD that removes jack staples with releaser strands. Subsequent introduction of another set of jack staples completes reconfiguration, enabling transitions between all three configurations. c AFM, TEM, and MD results showing the three structural states achieved through controlled jack edge modulation. Scale bars: 100 nm. d UV-responsive design. Photocleavable jack staples (purple) are incorporated into the initial cuboctahedron, with red indicating UV-triggered cleavage sites. e AFM, TEM, and MD data showing transformation from (i) initial cuboctahedron to (ii) octahedron using photocleavable short jacks and eventually to (iii) cuboctahedron after UV exposure. f, g Experimental yields of three configurations using AFM, TEM, and agarose gel electrophoresis. The number of DNA structures used for yield estimation is presented above bars. Error bars for the AFM and TEM data indicate ±σ, with detailed statistical information provided in the Source Data file. h, i FRET measurement of reconfiguration kinetics, showing a complete transformation over 30 min. To evaluate reconfiguration performances, we conducted three consecutive TMSD cycles and analyzed the products by agarose gel electrophoresis and AFM imaging (Supplementary Fig. 34). The results consistently demonstrate that TMSD enables reliable, repeated switching between both configurations. The structural integrity of correctly reconfigured Jitterbug structures is preserved throughout the cycles (Supplementary Fig. 34d, e). Quantitative AFM analysis shows that edge lengths and square opening distances remain unchanged within experimental uncertainty even after three cycles, indicating the absence of cumulative geometric distortion. Importantly, the geometry and mechanical fidelity of successfully reconfigured structures remain intact. To achieve convenient reconfiguration without requiring additional annealing (with releasers) or generating chemical waste, we engineered jack staples with photocleavable moieties41 such that long-wavelength UV (UVA) light will trigger spontaneous, irreversible transformation from octahedral to cuboctahedral configurations (Fig. 3d and Supplementary Fig. 32). Figure 3e confirms this approach, validating elastic energy storage via DNA base-pairing and the release by external light irradiation. Figure 3f, g presents experimental yields of the Jitterbug structures. The initial cuboctahedra, octahedra and the cuboctahedra with extended jacks all show approximately 85% yield, while reconfigured structures are slightly lower. We observed a gradual decrease to ~67% during three TMSD reconfiguration cycles, due to aggregation and dimer formation (Supplementary Figs. 34, 35 and Supplementary Table 1). The yields calculated from gel electrophoresis were slightly higher since partially defective structures migrate similar to intact origami. We also measured reconfiguration kinetics upon UV irradiation using Förster resonance energy transfer (FRET). Cy3 (donor) and Cy5 (acceptor) dyes were placed on adjacent edges in a square (Fig. 3h) to observe a change in FRET signals upon transformation such that transformation will change the FRET signals. In the octahedral configuration, the proximity between the donor and the acceptor quenches the donor emission. During the transition into the cuboctahedron, the donor signal was restored to about 90% (the FRET efficiency decreased from nearly 80 to 20%) as shown in Fig. 3i and Supplementary Fig. 37c, suggesting that UV irradiation over 30 min is sufficient for complete reconfiguration. Light-induced nanopores on synthetic vesicles Next, we explored the utility of the DNA transformer as a nanoscale actuator to regulate permeability of lipid membranes. The Jitterbug is capable of releasing stored elastic energy that exceeds the minimum energy required to form nanopores in fluid-phase DMPC membranes (~30 kBT)42,43. DNA reconfiguration produces an external force against lateral tension on the membrane, opening a nanopore until the force equilibrium is reached. To achieve this, we anchored the DNA nanostructures onto the membrane using cholesterol moieties and triggered their transformation via UV irradiation. We estimate that roughly 2.5 × 10⁴ Jitterbug structures cover each vesicle in our experiments. UVA light minimizes photodamage to DNA origami41 and does not compromise lipid-bilayer integrity under the applied irradiation conditions44,45. The experimental workflow consists of four steps (Fig. 4a). (i) Giant unilamellar vesicles (GUVs) decorated with biotinylated lipids are immobilized on a biotinylated BSA-functionalized channel surface via biotin-streptavidin chemistry. (ii) The DNA nanostructures are labeled with Cy5 dyes and cholesterol groups for fluorescence tracking and stable membrane anchoring, respectively (Fig. 4b). Octahedral structures are bound to the membrane. (iii) Upon UV exposure, photocleavable jack staples are cleaved, transforming the Jitterbug and creating multiple transient nanopores on each vesicle. This reconfiguration allows dextran-rhodamine B (DR) in the surroundings to diffuse into the vesicle through the pores. (iv) With UV light off, the nanopores are resealed through elastic relaxation of the membrane43,46, retaining fluorescent DR molecules inside the vesicle. To verify that membrane-bound Jitterbug DNA retains its designed configurations during transformation, we performed FRET measurements on vesicle membranes (Supplementary Fig. 38). The FRET signals changed markedly before and after UV irradiation, confirming that the anchored octahedra remain intact on the membrane and then transition into cuboctahedra. UV-triggered nanopore formation and molecular transport into synthetic vesicles. a Schematics of a microfluidic imaging chamber and experimental workflow. (i) Immobilization of a biotin-functionalized vesicle on a BSA-decorated surface via biotin-streptavidin binding. (ii) Cy5-labeled Jitterbug binding to lipid membrane and introduction of dextran-rhodamine (DR). (iii) UV-triggered DNA transformation creating transient nanopores through which DR enters the vesicle. (iv) Membrane resealing after UV light is turned off, prohibiting molecular transport. b The DNA transformer with photocleavable jacks and cholesterol moieties undergoes a UV-triggered octahedral-to-cuboctahedral transition, creating transient membrane pores for controlled molecular passage. c Control experiment using non-photocleavable DNA. Despite UV irradiation, the vesicle with Cy5-labeled transformers (red) shows no DR influx (green), confirming that nanopore formation requires Jitterbug transformation. d During 60 min UV exposure, 10 kDa DR (green) enters the vesicle through nanopores created by Jitterbug transformation. Post-UV recovery shows complete pore closure with no DR leakage or exclusion of externally added Cy5.5 dyes (red), confirming a resealing of lipid membrane. e, f Size-selective molecular transport: 40 kDa DR penetrates the vesicle with UV exposure, while 70 kDa DR is excluded. g Jitterbug (red) transformation allows PK to enter the vesicle preloaded with casein-BODIPY substrate. Enzymatic proteolysis turns on BODIPY fluorescence (green), confirming PK entry and its function. h Control experiments for DR influx into GUVs. Influx of 10 kDa DR was monitored over time while independently varying three factors: the presence of photocleavable strands (PC), Jitterbug structures, and UV irradiation. The reference condition (black; identical to Fig. 4d) shows DR influx by UV-activated Jitterbugs. Control conditions include Jitterbugs with non-photocleavable jacks (red), GUVs without anchored Jitterbugs (blue), and photocleavable Jitterbugs without UV exposure (green). No significant influx was observed in any control condition. i Size-dependent DR influx into vesicles due to UV-responsive Jitterbugs. Time-resolved influx was monitored for GUVs exposed to DR of different molecular weights (10, 40, and 70 kDa). j Size-dependent efflux of DR from GUVs. Efflux was monitored over time for GUVs encapsulating DR with various molecular weights. DR-loaded GUVs were prepared, followed by attachment of Jitterbug structures onto the membrane and UV irradiation to induce nanopore formation (see Supplementary Figs. 43-52 for additional data). Error bars represent ±σ, with statistical details provided in the Source Data file. Iout and Iin represent DR fluorescence intensity outside and inside vesicles, respectively. Five GUVs per condition (N = 5). Scale bars in brightfield images: 10 μm. UV-triggered nanopores were studied in a series of experiments using different sizes of DR probes. Cy5 fluorescence along the vesicle membrane confirms the presence of membrane-anchored Jitterbugs (red, Fig. 4c–g). Upon UV irradiation, the green fluorescence signal gradually increases inside the vesicle, providing evidence of 10 kDa DR influx through the reconfigured nanopores (Fig. 4d and Supplementary Fig. 46). Notably, the membrane restored after ~60 min of UV irradiation. The absence of DR leakage and the exclusion of externally added Cy5.5 indicate complete pore closure. Time-dependent analysis shows that the fluorescence intensity inside the vesicle remains constant after the influx event (Supplementary Fig. 50), showing that no detectable leakage occurs once the pores close. To evaluate the transport mechanism, we performed control experiments by independently varying three factors: photocleavable strands, Jitterbug DNA, and UV irradiation (Fig. 4c, h, and Supplementary Fig. 44). Three control groups—Jitterbugs with non-photocleavable jacks, vesicles without anchored Jitterbugs, and photocleavable Jitterbugs without UV exposure—exhibit negligible fluorescence changes inside vesicles. The results confirm that molecular influx occurs only when Jitterbugs containing photocleavable strands are activated by UV light. This verifies that nanopore formation originates from the programmed reconfiguration of DNA Jitterbugs rather than unintended membrane permeation. To characterize size-dependent transport, we tracked time-dependent fluorescence intensity of individual GUVs and plotted normalized intensity ratios as shown in Fig. 4i. This plot compares transport behavior across probes of different sizes. While 10 kDa DR readily diffused into vesicles upon UV activation, 40 kDa DR showed reduced influx and 70 kDa DR was excluded entirely (Fig. 4d-f, i, and Supplementary Figs. 46–49). We further examined Jitterbug-mediated nanopores using a complementary efflux assay where DR was pre-encapsulated inside vesicles. Dye release to the external solution was quantified upon UV irradiation (Fig. 4j). Consistent with influx measurements, DNA transformers demonstrated efflux for 10 and 40 kDa DR, whereas 70 kDa DR remained confined inside the vesicles. From these observations, we estimate the pore diameter to be less than 16 nm47. This value represents an effective pore size inferred from transport behavior48 rather than a fixed geometric opening, consistent with membrane transport models that describe transient, dynamically fluctuating nanopores46–49. We also examined enzymatic assays inside vesicles preloaded with casein-BODIPY substrate (Fig. 4g and Supplementary Fig. 52). This fluorogenic substrate remains quenched until proteolytic cleavage turns on the green fluorescence of the BODIPY dye50,51 (Supplementary Fig. 40). Initially, casein-BODIPY encapsulated in the vesicle shows little to no emission signals, while Cy5 fluorescence confirms the Jitterbug conjugation on the membrane. UV activation creates nanopores via Jitterbug transformation that allows external proteinase K (PK) to enter the vesicle and digest the substrate, gradually restoring the fluorescence of the quenched BODIPY. This observation indicates molecular transport through UV-induced nanopores and enzymatic activities. We note that the fluorescence signal is enriched near the vesicle membrane rather than uniformly distributed in the lumen. This is attributed to peptide–membrane interactions that concentrate fluorescent peptides near lipid bilayers52,53. The signal remains confined within the vesicle boundary, indicating localized accumulation of cleavage products rather than membrane permeation. Our UV-responsive system, combined with membrane dynamics, presents a versatile platform for externally regulating membrane permeability. Programmable release of encapsulated payloads into vesicles Leveraging the deployability of our DNA Jitterbug, we investigated its potential as a programmable nanosystem that delivers payload enzymes to initiate proteolysis inside synthetic vesicles. We aimed to achieve DNA transformation by opening a single square face with all others remaining closed, thus enabling directional payload release into the vesicle interior while preventing unintended release into the surrounding solution. We first tested feasibility via oxDNA simulations. The square face with the stabilization removed opened while the others were completely closed as shown in Supplementary Fig. 15 and Supplementary Movie 2. Mapping the free energy of this structure as a function of opening distance shows a minimum at about 28 nm. The slight decrease in opening distance compared to Fig. 1d is due to a hindered range of motion with the other squares remaining closed. We functionalized 5-nm-diameter gold nanoparticles (AuNPs) with photocleavable oligonucleotides for dual purposes (Supplementary Fig. 36 and Supplementary Table 3): encapsulation via base-pairing54 and UV-triggered release from the Jitterbug. Either PK or trypsin (TRP) was used as enzyme payloads which were conjugated onto AuNPs (Fig. 5a). We encapsulated Cy5.5-labeled Au–enzyme complexes within initial cuboctahedra, which were then converted into octahedra using a mixture of photocleavable and regular jack staples to program a single square opening upon UV irradiation (Supplementary Fig. 30). Photocleavable jacks were incorporated on only one square face, while the remaining five faces contained non-photocleavable jacks. Cholesterol moieties were also positioned on the same photocleavable square to bias membrane attachment to that face (Fig. 5b). Opening all six squares would make the cargo delivery less efficient (Supplementary Fig. 57). Programmable delivery of enzyme payload via Jitterbug transformation. a Stepwise workflow for controlled enzyme transport and proteolysis. (i) Vesicle preloaded with self-quenched fluorescent substrates. (ii) Membrane anchoring of DNA transformers encapsulating Au-enzyme complexes via cholesterol tags. (iii) UV-induced transformation creating nanopores for Au-enzyme complex entry. (iv) Intravesicular enzymatic activity generating substrate fluorescence. b UV-activated transformation releases Cy5.5-labeled Au-enzyme complexes into the vesicle. AuNPs are conjugated to enzyme payloads (PK or TRP) and photocleavable oligonucleotides, which enables both encapsulation via base-pairing and UV-triggered release from the Jitterbug. Upon UV irradiation, only one square interfacing the membrane opens for directional release of the Au-enzyme complexes into the vesicle. c, UV-activated entry of Cy5.5-labeled AuNPs (red) into the vesicle cavity. Initially, Cy5-labeled DNA transformers (red) encapsulating AuNPs bind to vesicle surfaces. During 60 min UV exposure, AuNPs (red) enter vesicles through nanopores created by DNA transformation. d Proteolysis with several enzyme-substrate pairs (enzymes: PK and TRP, self-quenched (green) fluorescent substrates: casein-BODIPY or BSA-TMR). Red signals at 60 min indicate the entry of Au-enzyme complexes via UV-triggered transformation. Following UV exposure, enzymatic activity increases green fluorescence intensity (BODIPY and TMR) of substrates inside vesicles over time, confirming enzyme (PK and TRP) entry and function. e Heatmaps of normalized fluorescence intensity over 240 min for different enzyme-substrate combinations. Raw data are provided in the Source Data file. Scale bars: 10 μm. We tested the entry of the AuNPs by UV activation. Cy5 fluorescence along the vesicle boundary confirms the conjugation of the Jitterbug (encapsulating Cy5.5-labeled AuNPs) on the membrane (Fig. 5c). We observed a gradual increase of red fluorescence in the cavity during UV exposure, indicating successful transport of AuNPs into the vesicle through nanopores created during the DNA transformation. The red fluorescence appears spatially nonuniform, due to nanoscale clustering of DNA-functionalized AuNPs55. Such clustering often occurs in Mg²⁺-containing DNA samples, producing discrete fluorescence spots rather than a uniform signal56. We then investigated various enzyme-substrate combinations (Fig. 5d and Supplementary Figs. 53-56). We employed BODIPY-conjugated casein and TMR-tagged bovine serum albumin (BSA) as self-quenched fluorescent substrates pre-loaded within vesicles, while TRP and PK were conjugated to AuNPs. Following 60 min of UV exposure, DNA transformation enabled the entry of Au–enzyme complexes into vesicles, after which enzymatic proteolysis occurred inside the cavity. Green fluorescence gradually increased in the vesicle, indicating increased digestion of the substrates. Figure 5e presents the normalized proteolytic kinetics from fluorescence imaging. The fluorescence signal for each reaction was normalized to its maximum intensity at the final time point (240 min) to highlight the temporal evolution of proteolysis within each condition. Proteolysis within the vesicle microenvironment was influenced by several factors, including substrate distribution, enzyme loading, and enzyme-substrate affinity. These results demonstrate that the DNA Jitterbug can mediate enzymatic reactions within vesicles, which could be further developed into synthetic organelles57,58. Discussion This work introduces free-energy-informed design strategies for nanoscale metamaterials capable of deployable auxetic transformation. Transition between two topological states was achieved by stabilizing the compact octahedral configuration via insertion of DNA base-pairs into jack edges. Expanded cuboctahedra were designed as the global minimum in the free energy landscape, which formed naturally when the elastic energy was released or even when long jacks were implemented. We found that deployable auxetic DNA nanostructures behave under kinematic constraints, yet unlike macroscopic metamaterials, relevant chemistry and thermal fluctuations must be accounted for in the design process. Our framework opens avenues for designing increasingly complex and adaptive DNA materials by integrating mechanical energy, structural responses, and transformation mechanisms with computational models. For example, this design approach may be extended to engineer transformable multi-stable nanosystems with three or more topological states that can respond selectively to distinct external stimuli. Beyond reconfiguration, the Jitterbug serves as a nanoscale actuator capable of generating mechanical energy that exceeds that of natural biomolecular machines. The DNA transformer can store and release up to approximately 240 kBT of elastic energy (corresponding to a maximum force of 90 pN collectively from six squares), while, e.g., ATP-driven protein motors such as kinesin generate ~10 kBT per step (about 5 pN over 8 nm)59. The DNA nanomachine operates as an externally triggered energy reservoir that delivers mechanical output with tunability. Since its force profile can be accurately predicted, it may also function as a mechanical force probe to estimate membrane tension and permeability60, bridging biomolecular energy scales with engineering precision. Inspired by deployable viral capsid dynamics, the nanotransformers were exploited to create transient nanopores in GUVs and deliver payload enzymes into the vesicle lumen, thereby providing a synthetic platform to study virus–membrane interactions61,62. This capability could be extended to the treatment of bacterial biofilms63,64, where dense extracellular matrices and limited penetration often reduce the effectiveness of antimicrobial agents. The Jitterbug may enable localized membrane disruption and targeted delivery to weaken biofilm structures and enhance the efficacy. The modularity of DNA origami also allows for variable payloads (and substrates) and control over biochemical processes, making it useful for applications such as proteolysis, signal transduction65 and reaction cascades66. To render this platform applicable to living systems67, several mechanisms must be improved in the future, including high-affinity binding to the membrane, faster reconfiguration kinetics, and response to alternative stimuli like pH. Methods Materials All DNA oligomers were obtained from Integrated DNA Technologies. The M13mp18 (7k scaffold) was purchased from Bayou Biolabs, while the p9072 (9k scaffold)39 was provided by the Dietz group at Technical University of Munich. All other chemicals were acquired from Sigma-Aldrich. Filtration components included 100 kDa Pall Nanosep centrifugal filters (Cytiva OD100C35) and Amicon Ultra 50 kDa centrifugal filter units (UFC505024), both supplied by Fisher Scientific. DNA gel extraction was performed using Freeze ‘N Squeeze spin columns (7326165), and agarose powder was obtained from Bio-Rad Laboratories. The DNA ladder (Quick-Load Purple 2-Log DNA Ladder, 0.1–10 kb) was provided by New England Biolabs. Citrate-stabilized AuNPs with an average diameter of 5 nm, enzymes, and fluorophore-conjugated substrates were obtained from Thermo Fisher Scientific. Lipids including 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-biotinyl were purchased from Avanti Polar Lipids. Biotinylated BSA (29130) and streptavidin (PRO-791) were obtained from Fisher Scientific and Prospec, respectively. Buffer conditions The following buffers were used throughout this study. We used TAEM12 buffer (1× TAE containing 40 mM Tris–acetate, 1 mM EDTA (ethylenediaminetetraacetic acid) disodium salt, and 20 mM acetic acid, pH 8.0, supplemented with 12 mM magnesium acetate) for DNA origami assembly. For purification of the nanostructures and agarose gel electrophoresis, TAEM6 was employed, containing 1× TAE with 6 mM magnesium acetate. We stored enzyme solutions in TAEMC, which comprised 1× TAE with 6 mM magnesium acetate and 10 mM calcium acetate. For GUVs experiments, the inner solution contained 1×TAE supplemented with 6 mM Mg²⁺ and 200 mM sucrose, while the outer solution consisted of 1×TAE supplemented with 6 mM Mg²⁺ and 200 mM glucose. Under hypertonic conditions, deionized (DI) water was used as the internal solution, and the external medium was 1×TAE supplemented with 6 mM Mg²⁺. Substrate solutions were maintained in TAES, prepared from 1× TAE with 2 mM sodium azide. Gold nanoparticles (J67085.AC, Thermo Fisher Scientific) were stored in 1x PBS, consisting of 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄ at pH 7.5. DNA origami assembly DNA nanostructures were assembled by combining 8 nM 7k scaffolds and 10 nM 9k strands. This mixture included a 4x excess of staple oligonucleotides for the regular edges and a 6x excess of linker staples to connect the scaffold segments, resulting in a total sample volume of approximately 55 μL. The concentrations of DNA origami samples reported throughout this study are based on the 9k scaffold strand. The mixture was annealed in a thermal cycler (S1000, Bio-Rad Laboratories) using the following program: initial denaturation at 65 °C for 15 min, gradual cooling from 60 °C to 40 °C at a rate of 0.1 °C every 18 min, and final cooling from 40 °C to 4 °C. The complete annealing process required approximately 2.5 days. The initial state shown in Fig. 2c(i) was prepared using this thermal cycling protocol. The octahedral and cuboctahedral structures presented in Fig. 2c(ii)-(iii) were generated by incorporating short and long jack staples, respectively. Following purification of the initial DNA nanostructures using a 100 kDa centrifugal filter, jack staples were added, and the mixture was incubated at 50 °C for 12 h. DNA nanostructures designed for UV-triggered reconfiguration were synthesized using the same protocol but incorporated photocleavable jack staples. For GUV experiments, cholesterol-modified DNA strands were conjugated to the DNA nanostructures via incubation at 45 °C for 12 h. These cholesterol-modified strands served to anchor the nanostructures to the lipid membrane. A total of 32 cholesterol-modified strands were conjugated to each Jitterbug with six square openings, whereas 8 strands were conjugated for a single square opening. Reconfiguration induced by external stimuli Structural transformation was performed following a three-step process: (1) synthesis of the initial configuration (cuboctahedra without jack staples), (2) annealing with a set of jack staples for an intended conformation, and (3) TMSD using complementary releaser strands. The initial cuboctahedral structure without jack staples served as the starting point for reconfiguration into either octahedron with short jacks or cuboctahedron with long jacks (Supplementary Fig. 31). After synthesis of the initial conformation, jack staples were added at a 4x molar excess. The mixture was incubated at 50 °C for 12 h, followed by gradual cooling to 4 °C. To reverse the Jitterbug transformation, releaser strands were introduced to displace the jack staples through TMSD, and the reaction mixture was subjected to identical thermal conditions (50 °C for 12 h, followed by cooling to 4 °C). Between each step, excess staples were removed by centrifugal filtration. Approximately 55 μL of DNA solution was diluted with 400 μL TAEM6 buffer and centrifuged at 3000× g for 3 min using 100 kDa centrifugal filters. This procedure was repeated three times in total, and the purified sample was resuspended in fresh TAEM6 buffer. UV-responsive structures were prepared by incorporating photocleavable jack staples in step (2). Similar to regular short jack staples, the photocleavable jack staples enabled switching from cuboctahedral to octahedral configurations. Subsequent UV irradiation using a UV lamp (UVGL-25, UVP, Upland, CA) cleaved the jack staples, inducing reconfiguration back to the cuboctahedral form (Supplementary Fig. 32). We used UVA light (320–400 nm) to minimize potential damage to DNA41, lipids44,45, and fluorophores (Supplementary Fig. 39). Instrumentation AFM imaging AFM imaging was performed in air using PeakForce Tapping mode on a Bruker Dimension Icon equipped with SCANASYST-AIR probes. DNA origami samples were diluted to 0.5–1 nM prior to surface deposition. For sample preparation, 2 μL DNA solution was mixed with 8 μL TAEM6 buffer and deposited on freshly cleaved mica for 5 min. To enhance DNA adhesion to the mica surface, 20 μL buffer containing 2.5 mM NiCl₂ was added and incubated for an additional 2 min. The sample was then gently dried with compressed air and rinsed with 80 μL deionised water. Each mica substrate was scanned at multiple locations using scan sizes of 5 × 5 μm, 2 × 2 μm, 1 × 1 μm, and 500 × 500 nm. AFM images were processed using Bruker Nanoscope analysis software and ImageJ. The yield of DNA structures was estimated from 5 × 5 μm and 2 × 2 μm overview scans by quantifying the ratio of correctly formed structures to misfolded conformations (Supplementary Figs. 16–21 and Supplementary Table 1). TEM imaging A Tecnai G2 T20 microscope operated at 200 kV was used for TEM imaging. For sample preparation, formvar-coated copper grids (Electron Microscopy Sciences) were glow-discharged for 30 s using an EasiGlow (Pelco) to render them hydrophilic. A 2 μL aliquot of DNA origami solution was applied to the grid and incubated for 60 s before removal with filter paper. The grid was subsequently rinsed by pipetting deionized water until overflow, followed by excess water removal. Finally, 2 μL of 1% uranyl acetate was applied for 20 s and removed similarly. Structural yields were calculated as with AFM scans, and edge lengths were measured from the TEM micrographs using ImageJ (Supplementary Figs. 22–30). Dynamic light scattering (DLS) Hydrodynamic diameters of DNA nanostructures, AuNPs, DNA-AuNPs, and AuNP-enzyme complexes were determined using a Zetasizer Nano ZS (Malvern Panalytical, Malvern). All buffers were filtered immediately prior to use. Samples were prepared at effective nanostructure concentrations ranging from 0.5 to 2 nM and analyzed in microcuvettes. For each sample, three independent measurements were performed, and the final size distribution was determined by averaging across all measurements (Supplementary Fig. 36 and Supplementary Tables 2, 3). Spectrophotometry Absorption spectra were recorded using an Agilent Cary 6000i UV–Vis–NIR spectrophotometer. Sample volumes of ~100 μL were loaded into microcuvettes, and spectra were acquired over the wavelength range from 200 to 800 nm. Fluorescence measurements Steady-state and time-resolved fluorescence signals were recorded using a Fluorolog-3 fluorometer (HORIBA JOBIN YVON) equipped with an FL-1039 single-housing unit and a 450 W Xenon lamp. GUVs with fluorophore-conjugated DNA nanostructures were imaged using a custom-built inverted fluorescence microscope (Zeiss Axio Observer D1). Excitation was provided by diode lasers operating at 561 and 658 nm (Laserglow Technologies). Fluorescence emission was collected through a 63× oil-immersion objective (Zeiss) and detected using an Andor iXon3 electron-multiplying charge-coupled device (EMCCD) camera (Supplementary Fig. 42). Images were acquired with full fields of view (135 μm × 135 μm) under both bright-field and total internal reflection fluorescence (TIRF) modes. To minimize photobleaching, GUV imaging was performed using minimal excitation power with 200 ms exposure times. Image analysis and fluorescence intensity quantification were carried out using ImageJ software. Agarose gel electrophoresis Samples were analyzed using 1.0–1.5% agarose gels in TAEM6 buffer for 90 min at 75 V. Gels were stained with 0.5 μg/mL ethidium bromide and visualized using a UV lamp (Spectroline TE-312S, Spectronics). Target bands were excised with a razor blade and transferred to Freeze’N Squeeze tubes. Gel fragments were mechanically disrupted by repeated compression, frozen at −20 °C for 10 min, and centrifuged at 13,000 × g for 3 min to extract DNA. The resulting 16-bit images were analyzed in ImageJ. Cross-sectional intensity profiles were generated for each lane by averaging grayscale values within defined regions of interest. Peak intensities of monomer bands and higher-order assemblies were quantified to determine sample yields. DNA-coated AuNPs DNA-AuNP conjugation was performed using a freezing method developed by a previous report68. Thiolated DNA oligonucleotides were used without prior reduction treatments such as dithiothreitol (DTT) or Tris-(2-Carboxyethyl) phosphine (TCEP) reduction. In a typical experiment, 5-nm citrate-stabilized AuNPs were mixed with the thiolated DNA strands at a 1:100 molar ratio in PBS buffer. The mixture was placed in a freezer (−20 °C) for 3 h, then thawed at room temperature. After thawing, excess DNA was removed by three consecutive washing cycles using 50 kDa centrifugal filters. The purified DNA-functionalized AuNPs were added to DNA origami structures at 5× molar excess relative to the available binding sites. The mixture was incubated overnight at room temperature to facilitate hybridization between DNA-coated nanoparticles and complementary single-stranded extensions on the DNA nanostructures. Enzyme conjugation to AuNPs Negatively charged citrate-capped AuNPs were used to promote electrostatic adsorption of enzymes, including PK and TRP. This interaction reduced the surface potential through electrostatic shielding and increased the hydrodynamic diameter. AuNPs and enzymes were mixed at a 1:100 molar ratio in PBS buffer, sealed in PCR tubes, and incubated at room temperature for 24 h with constant agitation. After incubation, enzyme-functionalized nanoparticles were purified three times using 50 kDa centrifugal filters. The purified conjugates were resuspended in TAEMC buffer for subsequent hybridization with DNA nanostructures (Supplementary Fig. 41). For dual-functionalized nanoparticles containing both DNA strands and enzymes, AuNPs were first conjugated with DNA using the freezing method described above. After purification to remove excess DNA, the DNA-coated AuNPs were functionalized with enzymes following the identical incubation and purification protocol. Microfluidic channel assembly and surface passivation Microfluidic channels were constructed by sealing a thin glass coverslip (Fisher Scientific) to a quartz slide (Ted Pella) using medical-grade acrylic double-sided adhesive tape (Adhesive Research). Channel geometry was defined by slots cut into the adhesive layer. The coverslip was treated with piranha solution (H2SO4 [3%]: H2O2 [1%]) to generate a hydrophilic surface suitable for biotinylated BSA binding. Inlet and outlet ports were created by drilling the quartz slide and securing CapTite bonded port connectors with epoxy. Fluid flow was controlled using CapTite fittings (Lab Smith) connected to Tygon microbore tubing (Cole-Parmer) and driven by syringe-applied pressure. The total channel volume was approximately 20 μL. Surface passivation was performed to minimize nonspecific binding of vesicles, DNA nanostructures, and fluorescent molecules. The procedure involved sequential incubations: (1) 100 μL of biotinylated BSA (~0.5 mg/mL) for 10 min, (2) 100 μL of 0.2% (v/v) Tween 20 in PBS for 10 min to repair surface defects, followed by (3) rinsing with 100 μL PBS buffer (pH 7.5) and (4) final treatment with 100 μL streptavidin solution (10 μg/mL) for 10 min to complete biotin-streptavidin surface functionalization. GUVs were introduced and immobilized via biotin-streptavidin interactions between their biotinylated lipids and the surface-bound streptavidin. Preparation of GUVs Pristine GUVs were synthesized using the inverted emulsion method69. DMPC and biotinylated lipids were combined at a 1000:1 molar ratio in chloroform within a glass vial. The solvent was removed under vacuum for 30 min, and the dried lipid film was resuspended in 600 μL mineral oil (M3516, Sigma-Aldrich). Mineral oil is a light nonpolar hydrocarbon commonly used as the continuous phase in inverse emulsion GUV formation. The oil–lipid mixture was sonicated at 50 °C for approximately 3 h to achieve uniform dispersion. Aqueous droplets were formed by adding 20 μL inner buffer followed by 30 s of vortexing, producing a cloudy emulsion with lipid-coated droplets. A total of 600 μL emulsion was carefully layered over 300 μL TAEM buffer to create a hydrophilic interface. The assembly was centrifuged at 8000 × g for 15 min, during which monolayer-coated droplets traversed the oil-water interface to form bilayered GUVs that settled in the aqueous buffer phase. The mineral oil supernatant was discarded, and vesicles were collected from the buffer phase. All experiments were conducted in a temperature-controlled environment (26–27 °C), above the phase-transition temperature of DMPC (Tm ≈ 23–24 °C)70. The resulting vesicles exhibited smooth spherical morphologies consistent with membranes in the fluid phase. To prepare GUVs containing Casein-BODIPY or BSA-TMR conjugates, the inner buffer was prepared as described above, with the addition of 10 μM substrate. The high labeling density in both substrates leads to strong intramolecular quenching, and fluorescence is restored only after proteolytic cleavage releases smaller dye-labeled peptides (Supplementary Fig. 40). Synthesized vesicles were introduced into microfluidic channels and incubated for 45 min to establish biotin-streptavidin binding. Minimizing sequence interference to improve structural yields Due to their large size, the DNA origami structures were designed with two scaffolds: a 7249-nt M13mp18 scaffold (7k scaffold) and a custom-designed 9072-nt strand (9k scaffold). Since both scaffolds share similar sequences with considerable overlap, experimental yields varied significantly depending on the specific staple sequences used. Sequence overlaps among staples led to increased interference, resulting in a higher occurrence of misformed or partially assembled structures71. This problem was addressed by reducing staple interference. Staple interference was defined as the maximum sequence overlap between any two staples within a given set. To minimize interference, multiple staple permutations were generated by shifting the starting index along the circular 7k scaffold sequence while keeping the staple sequences for the 9k scaffold portion constant. Permutations exhibiting the lowest sequence overlap (approximately 11 nts) were selected and further filtered to eliminate staples prone to unintended hybridization with other scaffold regions or unpaired single-stranded portions. This process significantly improved the yield of Jitterbug DNA structures. The overall workflow for this staple interference minimization is illustrated in Supplementary Fig. 6. The python script used for optimization is available in the github repository. DNA origami design and equilibrium MD structures The origami design was performed using Cadnano22 and Scadnano23. In preparation for computational analysis, the DNA origami design was first exported from Scadnano into the oxDNA platform, preserving the experimental DNA sequences. The generated files were imported into oxView72 and subjected to rigid-body relaxation. For generating reference structures depicted in Fig. 1b, c, the rigid-body relaxed structure was converted into PDB format and visualized using Chimera, where the jacks were hidden in this depiction for direct comparison with ideal Jitterbug. For MD simulations, rigid-body relaxed structures were subjected to minimization and relaxation in oxDNA after which they were equilibrated for 108 simulation steps. The salt concentration was set to 1 M Na+ for all simulations to be comparable with the high salt conditions used in experiments (12 mM Mg+2). The mean structures shown in Fig. 1g-h were generated by averaging 100 configurations from the equilibrium trajectories. The only difference between the octahedral and cuboctahedral structures is the presence of stabilizing jack staples in the former. Reconfiguration analysis using oxDNA simulations Edges and jack segments were identified using DBSCAN algorithm. The structure contains 24 edges and 12 jack segments (two segments per square face). For each edge, the positions of the terminal hybridized nucleotides (i.e., 12 nts from 6 terminal bps at each end of the 6HB edge) were recorded. The opening distance was defined as the distance between the centers of mass of these terminal nucleotides at jack vertices. The reconfiguration was analyzed by monitoring the evolution of the average opening distance of all six squares (across eight independent runs) upon removal of jack staples from the octahedral state (Fig. 1g). Individual opening distances are detailed in Supplementary Fig. 9a. The opening distances ranged from approximately ~12 nm for the octahedral state to ~33 nm in the cuboctahedral configuration. The structure completed the spontaneous transition within approximately 5 × 106 simulation steps. During the transformation, radial displacements of constituent triangles and their rotation angles about the radial axis (depicted in Fig. 1a and Fig. S2) were tracked (Fig. 1f). Upon staple removal, deformation occurred nearly symmetrically (Supplementary Movie 1). The symmetry was confirmed by measuring the angles between the fitted planes of three pairs of opening squares (Supplementary Fig. 9b), which remained at ~90° throughout the reconfiguration process. Poisson’s ratio during Jitterbug transformation The three-fold symmetry of the Jitterbug transformer demonstrates a theoretical Poisson’s ratio of ν = −1 in all three directions. According to elasticity theory73, Poisson’s ratio is defined as: where εaxial and εtrans are axial and transverse strains, respectively. For the Jitterbug transformer, having three orthogonal axes, the effective Poisson’s ratio is calculated as the mean of the ratios among strains in these directions: where directions 1, 2, and 3 correspond to the three normal axes of the six squares in cuboctahedral configuration, with the origin located at the center of the structure. The effective Poisson’s ratio presented in Fig. 1e (νjitterbug=−1±0.1) is obtained by averaging results from eight parallel simulations. Principal component analysis (PCA) PCA was performed to elucidate the primary deformation modes during reconfiguration. To minimize computational load, a subset trajectory was first created from the transformation trajectory (i.e., reconfiguration from closed to open states) by retaining nucleotides from a single square of the Jitterbug as shown in Supplementary Fig. 7a. PCA was performed on this trajectory (using the oxDNA_analysis_tools package74) to identify the ten primary deformation modes. The first primary mode (Supplementary Fig. 7b), pointing along the square opening direction accounts for the majority of the deformation (scree plot, Supplementary Fig. 7c). As the first principal component in a symmetric system, the square distance was chosen as the primary reaction coordinate for the free energy landscape. Free energy simulations Free energy landscapes provide insights into the energy stored within the DNA Jitterbug, elucidating its spontaneous transformation from the octahedral to the cuboctahedral states upon jack removal. To better understand the free energy, we decompose it into two components: (i) structural free energy inherent to the cuboctahedron and (ii) hybridization energy gained from jack staples, which hold the structure in the closed state (chemical stabilization). We estimated the structural free energy of the Jitterbug (without jack staples) using umbrella sampling simulations29,35 in oxDNA performed on an NVIDIA 4070 GPU. Symmetric deformations were considered for computation with the average opening distance across all six squares as the order parameter. MD simulations employed a harmonic potential applied simultaneously to all squares (via 6 pairs of mutual traps, 2 per square face), with identical stiffness (11.4 pN/nm; 0.2 simulation units in oxDNA). For each square, 2 nts on each side of the jack are chosen (i.e., 6 terminal bps per 6HB edge, two edges intersect at each jack vertex) for applying the biasing force. A total of 100 simulation windows were generated, varying opening distance from 3.6 to 64 nm (4–75 simulation units). Each window was equilibrated over 3 × 106 simulation steps, followed by a production run of 9 × 106 steps to derive the free energy profile. The resulting distributions of opening distances were averaged across squares and the free energy was computed using the weighted histogram analysis method (WHAM). To ensure convergence, we performed 5 replica runs and estimated the error by calculating the standard deviation among the obtained estimates (depicted as the shaded region in Fig. 1d). Figure S10 presents the energy profiles for individual squares obtained by considering opening distances of each square separately for estimating the free energy. Figure 1d presents their average profile, computed by first averaging all squares and then performing WHAM. The total energy (~240 kBT) is obtained by summing the contributions from all six squares. The free energy profile shown in Fig. 1d was obtained by unbiasing the average order parameter across all the squares. Stabilization energy for octahedral configuration The amount of energy needed to stabilize octahedral configuration is provided via base-pairing. From the free energy profile of cuboctahedral configuration (Fig. 1d), NUPACK calculations suggest that at least ~20 base-pairs (~40 kBT) are required for a stable octahedral state. Note that this estimate arises from enthalpic contributions of base pair stabilization and does not account for scaffold entropy. To further analyze this process, we investigated the dissociation of the first 10 nucleotides in each jack segment. In the octahedral state, the jack segments experience non-zero force due to strain energy stored in the structure, which was estimated to be ~6 pN per jack segment within the 12–20 nm deformation range. To model this, a constant planar force with stiffness 0.57 pN/nm (10−2 simulation units) was applied using repulsion planes. VMMC simulations using oxDNA were then performed on a single jack duplex, tracking the dissociation of the first 10 bp (5 bp on each helix shown in Supplementary Fig. 11). The order parameter was defined as the number of dissociated base-pairs, ranging from 0 to 10 bp. Ideal weights were identified through iterative updates over 10 iterations. VMMC simulations were conducted in two overlapping windows (0–6 bp and 4–10 bp) for 6 × 108 steps, with seven replicas each. The simulations also printed the opening distance of the jack segment (distance between the first set of scaffold nucleotides, indicated by arrows in Supplementary Fig. 11a). Free energy profiles were plotted as a function of distance with samples placed into 1 nm wide windows. The resulting free energy was doubled to account for contributions from both jack segments (Supplementary Fig. 11b). As illustrated in the octahedral part of the free-energy plot in Fig. 1b, the total free energy of the system can thus be expressed as: The VMMC profile was integrated with the structural component by aligning them along the x-coordinate. Since base-pairing provides stabilization energy, this energy term carries a negative sign. Given that the fully hybridized octahedron represents an energy minimum, we aligned the minimum from the VMMC simulations with the opening distance of the octahedron obtained from MD simulations. Uncertainty was estimated using the standard deviations across replica runs. In the octahedral portion of the free energy, the total error is estimated by adding contributions from VMMC and structural components, while the cuboctahedral portion of the curve has uncertainty from the structural portion alone. Elastic energy stored in Jitterbug To emulate the behavior of the ideal Jitterbug, the triangular panels were designed to be perfectly rigid, and each opening square was constructed with two alternating vertex types as illustrated in Fig. 1h. To analyze the root cause for storage of elastic energy, the geometry was decomposed into three components: (i) the rigid triangular edges, (ii) vertex A, featuring four 10 ssnt connections that mediate reconfiguration, and (iii) vertex B, with three 10 ssnt connections that primarily rotates without significant deformation. For a detailed understanding, the Jitterbug structure was reduced to its simplest building block: a pair of triangles connected by vertices A and B. As described in Supplementary Fig. 12, vertices are constrained to normal planes. This confinement was mimicked using repulsion planes with a stiffness of 0.28 pN/m (0.005 simulation units), restricting vertices along two orthogonal planes. Umbrella sampling combined with WHAM was then used to calculate the free energy of this building block as a function of the opening distance. Harmonic biasing potentials (stiffness of 11.4 pN/nm, 0.2 simulation units in oxDNA) were applied over 100 windows, spanning opening distances from 3.6 to 64 nm (4–75 simulation units, same parameters used for free-energy simulations in Fig. 1d). Each window was equilibrated for 2 × 106 steps and then simulated for additional 9 × 106 steps. Errors were estimated by calculating deviations among three replicas. To compare this energy to that of a full opening square (Fig. 1j), the result was multiplied by a factor of two because one opening square consists of two triangular pairs. The resulting free energy profile for this triangular pair (denoted vertex AB) is shown in Supplementary Fig. 12. Analyzing molecular interactions To estimate the role of various interactions, such as hydrogen bonding and base stacking, we printed the pairwise interaction energy of all interacting pairs within the system during equilibrium simulations. This was carried out by adding the “pair_energy” key to the input file available in oxDNA. For proper comparison we exclude the jack sections of our scaffold and staples since jack staples were absent in the cuboctahedral state. The average energy per particle (for the chosen nucleotides) was estimated from equilibrium simulations over 108 steps (Supplementary Fig. 13a). For visualization, the average energy components for every nucleotide were recorded, including stacking (STCK), hydrogen bonding (HB), cross stacking (CRSTCK), coaxial stacking (CXSTCK), excluded volume (BEXC and NEXC), Debye-Hückel (DH), and backbone (FENE). The differences between these components in octahedral and cuboctahedral states were simplified into an oxView overlay to identify energy localization (Supplementary Fig. 13b, inset). The estimated energies and overlay files are available in the github repository. Comparison of energy estimations We compared estimated energies involved in the Jitterbug transformation from several methods: (i) umbrella sampling, (ii) NUPACK (for hybridization energy), (iii) VMMC simulations, and (iv) mechanical work. Umbrella sampling: We extracted the free energy landscape from the umbrella sampling simulations using WHAM and measured the free energy difference between the closed (~12 nm) and open (~33 nm) states: NUPACK: We used NUPACK to estimate the number of base pairs required to provide a minimum energy for stabilization of the octahedral state under salt and temperature conditions similar to experiments. Approximately 10 bp per jack segment (20 bp per opening square) were sufficient: This estimate accounts only for hybridization energy; more staples were used experimentally to offset the entropic cost of bringing scaffold ends together. VMMC: The free energy difference between the fully hybridized and 10 bp dissociated states was calculated: Mechanical work: The elastic energy was calculated by averaging the force required to maintain the opening distance during the square opening process (Fig. 1e) and integrating over the deformation range (12–33 nm) under the assumption of quasi-equilibrium (Fig. 1k): Overall, these parallel approaches converge on energy per square within a few kBT. Analysis of the jack staple binding process using DNAfold simulations To reconcile the difference between the theoretical hybridization energy estimate and the number of staples required experimentally, we performed coarse-grained simulations of a simplified jack design using DNAfold37. Supplementary Fig. 14 presents the layout of the strands for simulations. We varied the number of staples from 1 to 4 and measured the relative binding times of different segments across 10 simulations for each case. Since DNAfold software supports only square lattices and coarse-grains the system into 8-nt segments, the design was adapted to use 48-nt staples. This adjustment does not significantly affect the results. The relative binding time for each segment was defined as: where trel is relative binding time, tsegment is binding time of the specific segment, and tmean is mean binding time across all segments. The normalization procedure described above was performed for each of the individual simulations before averaging. The simulations (Fig. 1i) revealed that staples closest to the looped end of the scaffold (i.e., the rightmost staples) bind first, while the leftmost strands hybridize last. This means that the initial binding reduces the entropic cost for bringing the separated scaffold ends closer together, thereby facilitating the binding of subsequent staples. Statistics and reproducibility No statistical method was used to predetermine sample size. No data were excluded from the analyses. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Description of additional supplementary files Supplementary Movie 1 Supplementary Movie 2 Transparent Peer Review file Source data Source Data Acknowledgements The authors thank Drs. H. Dietz, M. Honemann, and M. Pinner at the Technical University of Munich for providing the p9072 scaffold. This work was financially supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences under award no. DE-SC0020673. J.H.C. acknowledges a Friedrich Wilhelm Bessel Research Award from the Alexander von Humboldt Foundation. Author contributions R.L., F.C.S. and J.H.C. conceived the idea; S.S., A.A.S., M.R.K., Y.D. and M.D. performed the experiments; A.S.M. and R.L. conducted the numerical simulations; S.S., A.A.S., M.R.K., A.S.M., F.C.S. and J.H.C. analyzed the data. All authors participated in writing the manuscript. Peer review Peer review information Nature Communications thanks Ahanjit Bhattacharya and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Data availability The main data generated in this study have been deposited within the Article and its Supplementary Information. Supplementary movies of simulations are provided with this paper for download. This work did not utilize third-party or publicly available datasets. Source data are provided with this paper. Code availability The Python source code and simulation files are available on GitHub (https://github.com/choidnalab/DNA-Jitterbug-Transformer-data) or from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Seongmin Seo, Alexander A. Swett, Mallikarjuna Reddy Kesama, Anirudh S. Madhvacharyula. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74070-4. References Associated Data Supplementary Materials Supplementary Information Reporting Summary Description of additional supplementary files Supplementary Movie 1 Supplementary Movie 2 Transparent Peer Review file Source Data Data Availability Statement The main data generated in this study have been deposited within the Article and its Supplementary Information. Supplementary movies of simulations are provided with this paper for download. This work did not utilize third-party or publicly available datasets. Source data are provided with this paper. The Python source code and simulation files are available on GitHub (https://github.com/choidnalab/DNA-Jitterbug-Transformer-data) or from the corresponding author upon reasonable request.