Periodic Operation of a Dynamic DNA Origami Structure Utilizing the Hydrophilicâ•fiHydrophobic Phaseâ•’Transition of Stimulusâ•’Sensitive Polypeptides

Repository Citation Goetzfried, Marisa A.; Vogele, Kilian; Mückl, Andrea; Kaiser, Marcus; Holland, Nolan B.; Simmel, Friedrich C.; and Pirzer, Tobias, "Periodic Operation of a Dynamic DNA Origami Structure Utilizing the Hydrophilic–Hydrophobic Phase‐Transition of Stimulus‐Sensitive Polypeptides" (2019). Chemical & Biomedical Engineering Faculty Publications. 166. https://engagedscholarship.csuohio.edu/encbe_facpub/166


Introduction
DNA nanotechnology utilizes DNA mole cules both for the creation of static nano architectures as well as for the realization of dynamic nanodevices or even macro molecular machines. Defining the latter as molecularscale equivalents of macroscopic machines, they have been considered as a particular type of mole cular device "in which the relative posi tions of the component parts can change as a result of some external stimulus." [1] In particular the DNA origami tech nique has been shown to be an excellent tool to build welldefined macromolecular structures. [2][3][4][5][6] For this, hundreds of oligo meric DNA strands (the "staple strands") are hybridized to a long DNA "scaffold strand" to enforce folding of the scaffold into a desired shape.
One approach toward the implementa tion of machinelike behaviors in DNA ori gami structures [7] is based on the creation of molecular assemblies, in which multiple origami subunits are connected with each other via simple machine elements such as pivots or hinges. [8,9] When the subunits are further made to bind with each other through sufficiently weak interactions, which can be reversibly switched by external stimuli, [10][11][12][13] the structural components can change their relative positions, resulting in an overall conformational change.
The stimuli and energy sources for the operation of such devices can be quite diverse, for instance, DNA strand displacement processes, [14,15] binding of biomole cules through aptamers, [16] environmental changes (e.g., changes in salt concentrations [9,12] or physical stimuli such as temperature [13,17] ), or electric [18,19] and magnetic fields. [20,21] In many cases, "switchability" has been introduced by appro priate chemical modifications of the devices. For instance, photoswitchable devices have been realized by utilizing the wellknown cis-trans photoisomerization of azobenzene derivatives. [22,23] Other organic modifications were used to switch the conformation of DNA nanodevices by exploiting their hydrophobic interactions. [10,11,[24][25][26][27] Recently, Vogele et al. [17] utilized thermoresponsive peptides to actuate an active plasmonic waveguide defined on a DNA origami structure and Turek et al. [13] used a similar approach to actuate a DNA origami flexor.
These switchable devices could be utilized in nanotech nology for the development of more sophisticated machinery or even microscaled factories, but also for biomedical applications, responsive macromolecular assemblies are highly desirable. In the last years, DNA origami structures have been examined and developed for combined delivery and release systems. [16,[28][29][30][31][32][33] In order to achieve this, several control mechanisms and stimuli need to be developed to fulfill the desired needs, for instance, for diseaserelated markers (e.g., microRNA) or for the specific pH in cancer tissue.
In the present work, we utilize the stimuliresponsive proper ties of elastinlike polypeptides (ELPs) to periodically switch the conformation of a rectangular DNA origami structure through two external stimuli, namely salt concentration and tempera ture. The DNA structure serves as the mechanical framework, while the ELPs are used as the decisive control elements that allow us to change the relative positions of the mechanical components of the structure on demand. When a specific stimulus is applied, ELPs can change from a hydrophilic to a hydrophobic state, as determined by whether it is below or above its transition temperature, T t . We also demonstrate how the location and arrangement of the ELP control elements (fur ther denoted as configuration) influences the performance, the conformational states, and the thermodynamics of our dynamic origami devices. Similar to proteins, our devices exhibit several conformational states which can be engineered by changing the configuration o f t he E LP m odifications. Th e de sign al so in flu ences the thermodynamic behavior which becomes apparent through changes in T t and cooperativity of the transition.
ELPs are synthetic peptides [34][35][36] consisting of repeti tive sequences derived from the naturally occurring protein tropoelastin which is the precursor of elastin. [28] The canonical amino acid motif of ELPs is (GαGVP) n , where α can be any amino acid but proline and n denotes the number of penta peptide repeats. In this work, canonical (GVGVP) 40 peptides, abbreviated as V40, but also peptides with a mixture of the GVGVP motif and the noncanonical GVGVAGVP motif were used. Both types of peptides undergo a fully reversible phase transition from a watersoluble hydrophilic state to a col lapsed hydrophobic state when the sample is heated above the transition temperature T t . The reported conformation of ELPs in their hydrophobic state is a βspiral. [34] The T t depends on various parameters such as the concentration and the type of electrolytes in the solute, ELP concentration and length, and the amino acid used for the variable position α. More hydro phobic amino acids result in a lower T t , less hydrophobic α generate a higher T t . This feature allows the specific design of T t by using one specific amino acid or a combination of polar and nonpolar amino acids for α along the peptide. The incorpo ration of noncanonical motifs changes T t in the same way. An isothermal phase transition can be triggered by changing, for instance, the environmental conditions through the addition of salt depending on the specific T t of the peptides and is called the ΔT t mechanism. Fundamentally, ELPs can be used as trans ducer to convert one form of energy into a second energy form.
Some decades ago, Urry classified ELPbased molecular machines in three categories. [37,38] A device that operates by thermally driven contraction (folding) on raising the temperature is called a zeroorder molecular machine of the T t type. Any other device, where mechanical motion (folding) is involved, is called firstorder molecular machine of the T t type. If a device converts energy not involving mechanical force, it is called a secondorder molecular machine of the T t type; an example would be the reduction of nicotinamide moieties, which are bound to ELP amino acid side chains, resulting in the protonation of ELP carboxylate groups. In principle, only one ELP sequence would be sufficient to perform all three kinds of energy conversions. Because of these properties and the reversibility of the phase transition, ELPs represent an attractive alternative for nonpolar molecules such as cholesterol or longchained alkanes [26,27] for the modification of DNA ori gami structures, as it is not necessary to use surfactants to alter the hydrophobic effect or to reconfigure a structure. [10,11] How ever, because of the contour length of commonly used ELPs and the high water content of ELP coacervates diverse hydro phobic interactions and effects might be evolved; therefore devices utilizing ELPs require a thorough design and analysis.
The term "hydrophobicity" describes various phenomena and interactions involving nonpolar molecules and interfaces in an aqueous environment. The range of the hydrophobic effect has been reported to be up to several micrometers, whereby only the shortrange part of this interaction (<10 nm) has been called the "true" hydrophobic effect. [39] Instead of a defined bond, the interaction is dominated by the hydration of nonpolar molecules and interfaces. For the longrange attraction, various mecha nisms such as bubble coalescence play a role. Because of this nonlocal nature of the hydrophobic effect, the fabrication and folding of a DNA nanodevice using hydrophobic molecules is not just a straightforward mechanical task. In fact, actuation of origami structures using intraorigami hydrophobic interactions is somewhat reminiscent of proteinfolding processes, and we therefore utilized principles taken from polymer physics and protein folding to guide the design of the ELPmodified DNA rectangle. Early models of protein folding stated that hydrogen bonding is responsible for secondary structure formation in proteins. However, experiments showed that hydrogen bonding and other local interactions only are strong determinants for the folding of short peptides, but weaker determinants for the struc ture of globular proteins. A popular and more advanced model to describe protein folding is the HP lattice model proposed by Dill and coworkers. [40] In this model, the primary sequence is modeled as a chain of hydrophobic (H) and polar (P) monomers and folding occurs by embedding these monomers into a 2D or 3D lattice. The more H monomers are adjacent on this lattice the lower is the free energy of the chain. In particular, for pep tides the model states that rather the side chains and their solva tion instead of the backbone and its torsion angles determine the structure. From a general point of view, the folding code thus is more of a solvation code and less an encoding of torsion angles along the peptide polymer. This automatically implies that multiple interactions will contribute to the final structure, its thermodynamics, and kinetics of folding. [41,42] In the present study we designed the configurations of the ELP control ele ments by following these established principles for polymer and peptide folding. The DNA rectangle stands for the "backbone," whereas the peptides represent the "side chains."

Device Configuration and Periodic Thermal Actuation
The underlying mechanical framework used for our peptide actuated nanodevice was provided by a flexible t wistcorrected DNA origami rectangle (90 nm × 60 nm), [10,11] which com prised a central hinge region to promote rotational motion of the two leaves of the structure around the hinge axis (Figure 1a; Figure S9, Supporting Information).
For the attachment of ELPs on a grid defined on the origami structure, specific staple strands were modified and elongated with sequences complementary to those of the DNAlinker molecules L DNA covalently linked to the ELPs (for illustration, see Figure 1b,c and also Figure S3 and Table S1 in the Sup porting Information). In order to be able to specify individual ELP configurations, we labeled the rows of the grid with cap ital letters from A to E, and the columns with numbers 1 to 7. The hinge was constructed in row C via the insertion of two thymidines into the local staples. The nominal separation between neighboring rows is ≈10 nm, while adjacent columns have a distance of ≈11 nm. Due to constraints by the scaffold routing of the DNA origami structure, column 4 had to be shifted by ≈6 nm (see Figures S8-S11 in the Supporting Infor mation), which was not observable, however, in atomic force microscopy (AFM) imaging (see Figure 1d). The number and arrangement of the binding sites were varied in clusters of up to 14 per leaf consisting of two rows with up to seven sites each (see Figure S1 in the Supporting Information). Example configurations are shown in Figure 1, such as the ABDE345 with six ELPs attached to the center of each leaf.
In order to determine the binding yield of L DNA -V40 to the binding sites on the hinged rectangles, we used AFM imaging and counted the visible peptides for the configurations ABDE345 (Figure 1d), ABDE1267, and BD345 (Section S2.3, Supporting Information). Due to the mirror symmetry of these configurations, we determined the binding yield per leaf. For ABDE345 comprising six binding sites per leaf, we found that in about 37% of the structures, all six binding sites were occupied by L DNA -V40, while 23% of the structures carried five L DNA -V40. Figure 1d shows the binomial distribution of the determined binding yield, with a total probability of 76% that one leaf is functionalized with at least four peptides (see Section S2.3, Supporting Information). Due to the flexibility of the rectanglebound peptides, it was only possible to determine the number of peptides per leaf, but not the exact position of each V40. We therefore cannot exclude a potential position dependence of the binding yield per "binding site." [43,44] We first studied reversible opening and closing of ABDE345 and the other ELPorigami constructs in bulk solution using Förster resonance energy transfer (FRET) between the fluo rescent dyes Atto 532 and Atto 647N (Figure 1a; Section S3.1, Supporting Information), which were incorporated into the structure using dyemodified staples. Origami motion was stimulated by the cyclic variation of the sample tempera ture between 10 and 55 °C in folding buffer (FB, 40 × 10 −3 m Trisacetate, 10 × 10 −3 m ethylenediaminetetraacetic acid (EDTA) and 20 × 10 −3 m MgCl 2 , with additional 1.0 m NaCl). As can be  Figure 2a, the FRET signal reversibly follows the tem perature variation, indicating successful and repeated confor mational switching of the devices.

seen in
The temperature transition for the ABDE345 folding can then be determined by plotting the FRET signal versus temper ature (Figure 2b,c). We found that the temperature transition is independent of the rectangle concentration (which only affects the signaltonoise ratio, Figure 2b), while T t can be systemati cally altered through a change in the NaCl concentration, which is shown in Figure 2c for FB with additional 0.5 m NaCl, 1.0 m NaCl, or 1.5 m NaCl.
The apparent signal decrease at high temperatures originates from the temperature dependence of the acceptor dye (see the Data overview file in the Supporting Information) and was fully reversible. For [NaCl] ≤ 0.5 m, no FRET signal was observed, which indicates a transition temperature above 55 °C, whereas for [NaCl] ≥ 2 m, T t was below 10 °C, resulting in a high FRET signal for all temperatures applied in the experiment (see Data overview file in the Supporting Information).

Intrastructure versus Interstructure Interaction
In a previous study performed on origami rectangles function alized with cholesterol moieties, it was found that hydrophobic interactions can also promote dimerization of the structures, [10] which would also lead to a measurable FRET signal. In order to assess the propensity of ELP rectangles for intrastructure versus interstructure assembly (i.e., dimerization), we performed transmission electron microscopy (TEM) imaging. Instead of using temperature to control the state of the ELPs, we utilized NaCl concentration to permanently switch the V40 peptides either into their hydrophilic ([NaCl] = 0 m) or their hydrophobic state ([NaCl] = 3 m) before imaging the structures at room tem perature. TEM image analysis was automated using MATLAB (see Section S2.4, Supporting Information). In Figure 3a, the results for the different ELP configurations are shown, with the monomer fraction in dark colors and the dimeric assembly fraction in light colors, at concentrations of 0 m NaCl (blue) and 3 m NaCl (red). We found that, irrespective of the salt concen tration, roughly 20% of the ABDE345 configuration, but also ≈30% of the negative control without any V40, appeared to form dimeric assemblies. This points toward a measurement artifact most likely originating from the adsorption process and an overlap of the structures on the TEM grid (Figure 3b, right). We therefore assumed that samples with "measured" dimeriza tion fractions on the order of or below ≈20-30% in reality are dimerization free (an exhaustive overview of the dimer fraction determined for all samples investigated can be found in the Data overview file of the Supporting Information). As a positive control for dimerization, we designed the "selfcomplemen tary" configuration AB1267DE345 ( Figure 3, center), which was unable to fold intrastructurally and was therefore expected to preferentially dimerize at 3 m NaCl. Indeed, experimentally configuration AB1267DE345 displayed a measured dimeric assembly fraction of about 90%. In contrast to previous work utilizing other nonpolar modifications, we did not find any evi dence for interstructure assembly that outcompeted intrastruc ture folding (if the latter was possible). This indicates that with our experimental conditions and sample concentrations used, intrastructure folding was strongly favored over dimerization.

Characterization of Rectangle Conformations
We went on to investigate the influence of the V40 configura tion on "intrastructural" folding in greater detail. Peptides were either arranged in clusters (e.g., ABDE345) or in nonclustered configurations (e.g., AE17), which also included variations in the total number of V40 modifications. In order to classify the different folding states of the structures we used the aspect ratio r of the observed rectangles, which varied between r = 0.3 (for maximum overlap of the two leaves of the rectangle) and r = 0.7 (for a completely unfolded rectangle, further denoted as −F). The aspect ratio may therefore also be interpreted as a reaction coordinate for the folding process.
We quantified intrastructural folding for 14 different con figurations by analyzing the probability distributions along the reaction coordinate r (Figure 4b; Data overview file, Supporting Information) using TEM imaging. Some of the configurations, e.g., BD17, did not fold at all and were only found in the −F state at 0 and 3 m NaCl. A second group of configurations pre dominantly populated states with mean aspect ratios of r = 0.3 and r = 0.45 (further denoted as +F0.3 and +F0.45, respectively). Only in a few cases aspect ratios between 0.45 and 0.7 were identified (further denoted as +Fx). Figure 4b shows population densities P(r) for six representa tive configurations along the reaction coordinate r for 0 m NaCl (blue) and 3 m NaCl (red). Even though the hinge of the rec tangle was designed to promote folding of the structures into state +F0.3, the population densities of configurations ABDE17 and ABDE1267 show a global maximum for +F0.45 and only a local maximum for +F0.3. Furthermore, ABDE17 displayed folding even under nominally nonfolding conditions (0 m NaCl) and also populated the extreme +Fx states. This is consistent with the observation that in our fluorescence measurements ABDE17 showed the lowest T t near room temperature (see Figure 5). In contrast, configuration ABDE345 populated mainly +F0.3 for 3 m NaCl with a smaller population at +F0. 45. AE17 is the only nonclustered structure which we found capable of folding, and it mainly populated +F0.3 and less +F0. 45. For its counterpart BD17-which has the same number of V40, but located directly next to the hinge-no folding was observed.
However, BD17 showed a large propensity for dimerization. A potential explanation for unfavorable intrastructural folding of BD17 could be the effect of steric hindrance between lines B and D, which would keep the hinge from completely closing. This would leave the ELPs partially exposed, which in turn would increase the relative energetic benefit of dimerization. The same rationale can be applied to AE1267 and its nonfolding counterpart BD1267 (see the Data overview file in the Supporting Information).

Hydrophobic Interactions and Folding Pathways
Quite generally, the creation of hydrophobic interfaces in water is energetically unfa vorable, and such interfaces always "attempt" to minimize their interfacial energy. In the HP model, hydrophobic residues are located in the interior of a folded protein, where they are less exposed to polar mole cules and where, most importantly, no bulk water exists. Accordingly, polar residues at the outside interact with the surrounding water molecules. Coacervates formed from collapsed ELPs usually have a high water content, but this "water of hydrophobic hydration" does not have the degrees of freedom of bulk water due to constraining Hbond interactions between the involved water molecules. [45] In fact, this water can be seen as part of the hydrophobic interface.
Although the twist of the structure was minimized in the design, we still expect a slight global twist. Since the origami rectangle structure is flexible and contains relatively thick adhe sive peptide patches, it is conceivable that it will not exclusively fold through a rotation around its hinge. Therefore, we consid ered two plausible kinetic paths for folding of the structures (Figure 4c). First, the rectangle moves along the reaction coordi nate and "searches" for its minimal energy state. Second, the rec tangle closes around its hinge, reaches the +F0.3 state, and then starts searching for its minimal energy state. From the perspec tive of energy minimization by simply decreasing the exposed hydrophobic area and ideal ELP binding yield, +F0.3 should be the preferred state, because in the other states hydrophobic V40 would be still exposed to the surrounding aqueous solution.

Energetic Considerations
For a discussion of the overall conformation of the ELP rectangle, we may consider a planar molecular sheet modified with closely grafted peptides. Above T t , where water is a poor solvent for the ELP, attractive interactions between the ELP chains due to van der Waals forces, solvation forces, etc., domi nate over osmotic repulsion. Below T t , with water as a good sol vent, osmotic repulsion wins out due to the unfavorable entropy penalty from chain compression. [46] Due to the complexity of the interactions it is difficult to estimate the dimensions of the peptides in their collapsed or noncollapsed phase, in par ticular when the conformation of an ELP in its collapsed form is a βspiral. In order to gain an idea of the dimensions of the structure, we can roughly picture a single hydrophilic peptide as a random coil with a Kuhn segment length of 2.1 nm. [47] For V40, this results in a mean endtoend distance of ≈ 13 nm and a radius of gyration of ≈ 5 nm (see Figure S8 in the Supporting Information). However, the distance between neighboring binding sites on the rectangle is ≈10-11 nm and the length of the doublestranded DNA (dsDNA) linker additionally adds ≈7.5 nm. Thus, the exploration volume of single L DNA -V40 is larger than the distance between the binding sites. This alone could be a rationale for the measured rectangle states, where even a stretched conformation of ELPs is imaginable without being exposed to the surrounding water (see Figure 4d). However, it has been shown [42] that the hydrophobic effect at the single peptide level can be well described by discrimi nating several contributions to the internal energy such as peptide-peptide, water-water, and peptide-water interactions. Depending on the system, these contributions can almost cancel each other, and thus the final internal energy becomes finally much smaller than the individual contributions. The more relevant free energy is much more puzzling since the entropy contributions cannot be discriminated so easily. Furthermore, the entropic nature of the ELP phase transition makes the collapse free energy strongly temperature sensitive. To get a better understanding of the system, approximations of effective interfacial energies or solvent-solute interactions could be used. [48] At the transition temperature, there is a balance between the collapsed and noncollapsed states. As the temperature increases, the entropic contribution becomes more important and results in the complete collapse of the ELP. For the ABDE345 configuration with six ELPs bound per leaf this would result in a total free energy change of about ≈800-1600 k B T. It is surprising that almost identical values can be estimated only by looking at the interfacial energy change, which we estimated as ≈900-1500 k B T (see Section S2.7 in the Supporting Information). However, the latter picture would result in a nonnegligible energy penalty of ≈150-250 k B T per single ELP (see Section S2.7 in the Supporting Information) when it stays exposed to the surrounding water, which could be the case for the +F0.45 conformations. On the other hand, because of the relatively large exploration space of L DNA -V40, in particular due to the relatively long linker, the hydrophobic ELPs could still be "buried" between the leaves and not exposed to water. Therefore, a simplified picture is not always useful, when multiple interactions are involved. Similar as in the dis cussion of the internal energy above, it is conceivable that small variations in one or two local parameters change the individual energy contributions to the hydrophobic effect quite drastically, but the total free energy does not. [42,49] When the number of V40 and thus the size of the hydro phobic interface were reduced, for instance from ABDE17 to ABDE345, the state +F0.3 was more populated. Configurations ABDE1267 and AE17 vary only by one V40 per leaf in the case of perfect ELP binding yields. But due to the measured binding yields, a number difference of 1 should be negligible. Neverthe less, these two structures show quite different folding behavior. Furthermore, AE17 and BD17 exhibit the same interfacial area, but differ significantly in terms of intrastructure dimerization. For that reason, a simple change of the peptide number or interfacial area cannot explain the measured dis tributions alone, pointing toward an intricate interplay of local concentration, ELP binding configuration, and conformation inside the peptide layer.

Influence of ELP Length
We also investigated the influence of ELP length on rectangle folding. To this end, we used four different ELPs with different contour lengths L C and radii of gyration (see Figure 4e). The ELPs with the motif (V4A′) are noncanonical ELPs with the canonical segment V4 = (GVGVP) 4 and A′ as an alaninerich peptide sequence (see Section S3 in the Supporting Informa tion). Because of their length difference, they exhibit different bulk T t , which scales with k/N, where k is a fit parameter (with Kelvin units) and N is the number of peptide motifs; [50,51] the latter can be canonical or noncanonical. Although V40 has a different amino acid composition than the noncanonical ELPs, its T t also scales well with k/N (solid line in Figure 5a, bottom). Each of the different lengths was attached to the ABDE345 configuration, and the population distributions were deter mined for 0 and 3 m NaCl using TEM imaging (Figure 4e). As a result, we found that the longest ELP (V4A′)12 only populates the +F0.45 state, but state +F0.3 gets more and more populated with decreasing radius of gyration. As expected, the rectangle with (V4A′)4 did not exhibit any folding since the peptide's bulk T t was too high and 3 m NaCl was not sufficient to decrease the T t to temperatures below room temperature. The observed tran sition from a favored +F0.45 to a favored +F0.3 due to a size change indicates that the peptides' exploration volume might be an important factor. With smaller ELPs, the rectangle is "forced" into the +F0.3 to prevent the exposure of ELPs to the surrounding water. But this would also imply a different ELP packing between the rectangle leaves and contributions from multiple interactions in the collapsed ELP phase.

Thermal Folding
In order to investigate the thermal behavior of the rectangles, we used ensemble fluorescence resonance energy transfer spectroscopy. As described above, the sample temperature was periodically changed, and the FRET signal of the incorporated dyes was used to report the conformational change of the rec tangle. In this way, we were able to assess how the type of ELP used and their onorigami configuration influences the T t of the ELP rectangle system. For +F0.3 the designed distance between the dyes is about 2.5 nm, and for +F0.45 this distance is in the range of about 17.5-22.5 nm and thus beyond the Förster radius. We therefore expected that the FRET signal mainly rep resents the occupation of the +F0.3 state.
The top graph in Figure 5a shows the normalized FRET signal versus sample temperature for the configuration ABDE345 for different ELPs, where the solid lines are sig moidal fits to the data. It clearly shows that the specific T t increases with decreasing contour length. Furthermore, the transition temperatures of the rectangle still scale with k/N (Figure 5a, bottom). It can also be seen that the slopes of the sigmoidal fits vary from peptide to peptide (Figure 5a, top). This effect is also found when the peptide is always V40, but its binding configuration is changed (Figure 5b). ABDE17 exhibits the lowest T t , whereas ABDE345, ABDE1267, and AE17 exhibit increasing T t values in that order, but with different slopes at T t . Interestingly, a clear dependence on the "number" of V40 modifications was not found.
In bulk experiments, a change in the slope at transition indicates a cooperativity effect and usually decreases with increasing T t . [51] As already discussed in more detail above, due to the random motion of the two leaves and their flexibility, it is plausible that during folding the two leaves initially contact each other only at one point, while other parts of the rectangle are still separated. For instance, configuration ABDE1267 consists of two clusters and during folding one cluster prob ably contacts its opposite cluster before the second does. This spatially confines the second cluster and decreases the entropy penalty for its collapse, which in turn gives rise to cooperative folding. Finally, the FRET measurements also hint toward an intimate coupling of multiple effects.

Conclusion
In this work, we generated a peptidefunctionalized DNA ori gamibased macromolecular machine whose conformation can be reconfigured on demand through external stimuli. As the control mechanism, we used the fully reversible phase transi tion of elastinlike polypeptides, which allowed us to repeatedly switch the DNA structure's conformation by an alternating temperature cycle. Using Urry's classification this would be a zeroorder molecular machine of the T t type. The utilization of salt as stimulus for our device would classify it as a firstorder molecular machine.
We also showed that the 2D arrangement of these ELPs on the rectangular structure used had a strong impact on its conformational states and thermodynamics. The results can be rationalized by adopting established principles for polymer and peptide folding, which point toward a balance between side chain and backbone interactions. In analogy to proteins, in our nanodevice, the DNA structure represents the backbone and the ELPs represent the side chains. The basic question in this context, and which is notoriously difficult to answer, is "what is the true balance between the different energetic con tributions?." Our findings suggest that for the specific example of the ELPmodified rectangles a sidechaincentric rather than a backbonecentric view describes their overall behavior better, since the ELPs strongly determine the folding transi tion temperature. Structural details of the rectangular structure simply finetune the kinetics and thermodynamics of the folding process.
For future studies and potential applications of dynamic DNA devices, a variety of orthogonal tools and principles for design and fabrication are needed. In particular, other mechan ical frameworks than DNA origami could be interesting, for instance proteinbased filaments. The respond times of the structures might be reduced to the millisecond [52] or even microsecond range. A good stimulus candidate for the latter would be light. [53] The latter would also be a more sophisticated stimulus than salt for a firstorder molecular machine of the T t type. In this context, it would also be appealing to design a secondorder molecular machine. Furthermore, applications in a biological context would make it necessary to utilize pH or biochemical stimuli, [37,38] e.g., phosphorylation and dephospho rylation, for a potential autonomous operation without external operator.

Experimental Section
ELP Expression and Purification: The genes for the different ELP variants were designed and synthesized using methods reported previously. [50,54] For expression, [55] 10 mL overnight culture in lysogenic broth-medium (LB-medium) of the BL21 pLys Escherichia coli-stock with the desired plasmid were prepared with 100 µg mL −1 final concentration kanamycin when the expression vector pColaDuet-1 was used and 50 µg mL −1 final concentration carbenicillin when pET20b was used. A 5 mL aliquot of this culture was used to inoculate 750 mL of LB-medium (same antibiotics concentration and prepared with a drop of Antifoam 240) and incubated at 37 °C and 250 rpm until the OD 600 reached a value of 0.5-0.8. The culture was induced by 1 × 10 −3 m final concentration isopropyl β-D-1thiolgalactopyranoside (IPTG) and transferred to 20 °C and 250 rpm to overexpress the peptide overnight. The bacteria were harvested by transferring the solution to centrifugation flasks, centrifuging the sample at 4 °C at 4000 rcf for 15 min, and discarding the supernatant. The pellet was then resuspended with PBS, transferred to 50 mL falcon tubes and centrifuged with the previous settings for 20 min. The supernatant was discarded and the pellet was immediately used or frozen with liquid nitrogen for storage at −80 °C. The cells were resuspended in 2 mL lysis buffer per gram pellet and sonicated twice on ice for 9 min using 8 W with an interval of 10 s pulse and 20 s pause. The lysis buffer consisted of 1 mg mL −1 lysozyme, 1 × 10 −3 m benzamidine, 1 × 10 −3 m phenylmethylsulfonyl fluoride (PMSF) and 20 µL Turbo DNAse (Ambion ThermoFisher) in phosphate buffered saline (PBS). An aliquot of 1.5 mL 10% polyethyleneimine (PEI) in H 2 O (2 mL per 1 L cell culture) was added and the lysate was distributed into 2 mL tubes. The tubes were heated to 60 °C for 10 min and then transferred back on ice. In the next step a "cold spin" was carried out, which consists of a centrifugation at 4 °C and 16 000 rcf for 10 min, after which the supernatant was transferred to new tubes. Crystalline NaCl was added and dissolved until the liquid turned turbid. The solution was centrifuged at room temperature at 16 000 rcf for 10 min and the supernatant was discarded; the preceding step was also called "hot spin." The pellets were resuspended in 500 µL PBS and a new cold spin was carried out. The series of cold and hot spins were repeated until no pellet was observed during the cold spin and the wanted volume was reached; usually, the series of cold and hot spins was repeated three times. The quality of the purification could be tested by running an 8-16% gradient sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) for 35 min at 180 V. It is important to note that ELPs usually do not contain the amino acids, which are typically the target for commonly used stains, and thus the bands in the gel can be faint. Therefore, a negative stain with 0.5 m CuCl 2 or a positive silver stain were recommended. For testing, the final sample for contamination with other proteins a Coomassie stain is sufficient.
Conjugation of ELP to DNA: For the conjugation of ELPs to DNA, copper(I)-catalyzed alkyne-azide cycloaddition (further denoted as click chemistry) was used. ELPs were first activated with an azide group by using three types of bifunctional N-hydroxysuccinimide-azide (NHS-azide) linkers, namely NHS-C3-azide, NHS-PEG4-azide, and NHS-PEG12-azide, which were dissolved in dry dimethyl sulfoxide (DMSO). If not mentioned otherwise, NHS-C3-azide was used. The linkers were added in a fivefold to tenfold molar excess to the peptides provided in water or PBS. The reaction was incubated and shook for at least 6 h or overnight. Since the ELPs used do not have any free amines except for the N-terminus, side reactions on the peptide chains were not expected. The peptides were purified from unused linker by using centrifugal filters from Amicon with 10 kDa cutoff. In the next step, the sample was incubated for 3 h or overnight at room temperature, and purified through salt precipitation. Therefore, 5 m NaCl in water was mixed with the sample in equal volumes and vortexed, which caused the solution to turn turbid. The solution was then centrifuged at 16 000 rcf at room temperature for 10 min. The supernatant was discarded and twofold distilled water or buffer was used to redissolve the pellet. The tube was cooled at 4 °C for 10 min and vortexed until the pellet vanished. The centrifugation and cooling steps were repeated twice for a sufficient purity.
DNA Origami Structure Assembly: The 7249 nucleotides (nt) long single-stranded "scaffold" strand from phage M13mp18 [56] was folded into a twist-corrected rectangular DNA origami structure using about 200 short staple strands. These staple strands were used in a fourfold molar excess compared to the scaffold strand. As folding buffer (1× FB) 1× Tris-Acetat-EDTA buffer (40 × 10 −3 m tris(hydroxymethyl)aminomethane (TRIS), 20 × 10 −3 m acetate, 1 × 10 −3 m EDTA) with 20 × 10 −3 m MgCl 2 was used. For the folding of the structure an initial temperature of 70 °C for 3 min was used. The following temperature ramp started at 65 °C and ended with 45 °C after 1 h. Afterward, the solution was stored at 4 °C until further usage. The excess of unbound staples was removed by polyethylene glycol (PEG) precipitation according to Stahl et al. [57,58] Purification was carried out by adding an equal volume of 15% (w/v) PEG8000 in 1× FB with additional 505 × 10 −3 m NaCl and vortexing for 30 s. Afterwards the sample tube was centrifuged at 16 000 rcf and 20 °C for 30 min. The supernatant was carefully removed. The origami pellet was redissolved in the desired volume of 1× FB with additional 505 × 10 −3 m NaCl and vortexed for 1 min. This purification step was repeated two more times. After the last purification the pellet was redissolved in 1× FB. Afterward the sample was incubated at room temperature over night or at 37 °C for at least 1 h. The concentration of the DNA rectangles was determined through absorption spectrometry (Nanophotometer IMPLEN vers. 7122V2.3.1, Munich, Germany) using an extinction coefficient of 1.12 × 10 8 m −1 cm −1 .
To equip the rectangles with binding sites for ELP binding, specific staples were elongated by the sequence TTATACATCTATTATCTACATA, whose complementary sequence was conjugated to the ELPs. The rectangle was functionalized with these DNA-ELP constructs by adding them in a two to threefold molar excess (relative to the number of binding sites) to the origami solution. If necessary, the buffer solution was readjusted by adding 10× FB (400 × 10 −3 m Tris, 200 × 10 −3 m acetate 100 × 10 −3 m EDTA 200 × 10 −3 m MgCl 2 ). After the sample was incubated at room temperature for at least 1 h, it was purified right before further use with centrifugal filters from Amicon with a 10 kDa cutoff at 8000 rcf, at 4 °C for 10 min. The purification was repeated for three times.
Electrophoresis: To control the success of the peptide-DNA coupling, urea-PAGE (15%) was performed. About 4.8 g of urea, 1 mL of 10× TRIS-Borate-EDTA buffer (TBE), 3.75 mL of acrylamide (29:1), and 1.25 µL of H 2 O were used in a total volume of 10 mL. To fully dissolve urea the solution was heated to 30-40 °C in an ultrasonic bath. The polymerization of the gel, before loading it in a preassembled gel cassette, was initiated by the addition of 28 µL 10% ammoniumpersulfate (APS) and 8 µL tetramethylethylenediamine. The samples were prepared with 2 µL gel loading buffer II and 0.5-1 µL peptide-DNA sample, and were denatured through heating to 95 °C for 5 min. Before the denatured peptide-DNA samples were loaded, the gel chamber was incubated with 1× TBE at 60 °C using a connected water bath for half an hour; furthermore, a voltage of 40 V was applied to the geld. Electrophoresis was carried out using 100 V at 60 °C for 60 to 75 min, followed by staining with SYBR Gold Nucleic Acid Gel stain (ThermoFisher Scientific) and visualized using UV light.
Atomic Force Microscopy: For AFM imaging, 5 µL of 5 × 10 −9 m origami sample in 1× FB was deposited onto freshly cleaved mica. After an incubation of 5 min, 60 µL of the same buffer was added. An Asylum Research Cypher Scanning Probe Microscope was used in tapping mode with the Asylum Research environmental scanner. The silicon cantilever (Olympus micro-cantilevers BL-AC40TS-C2) had a nominal force constant of 0.09 N m −1 and a resonance frequency in water of 25 kHz. The scan rate was between 1 and 4 Hz. The AFM software was implemented in Igor pro 6.34 A. Image processing was performed with ImageJ.
Transmission Electron Microscopy: The staining solution was freshly prepared by adding 1 µL of 5 m NaOH to 200 µL 2% (w/v) uranyl formate in water and vortexing it for 1 min. To sediment the large unwanted uranyl crystals, which would be seen as black spots on the grid, the solution was centrifuged for 5 min at 20 000 rcf at 4 °C. The sample was prepared in 5 × 10 −9 m concentration in 1× FB with an optional addition of 3 m NaCl. For the positive stain, 5 µL of the sample was put on a grid (FCF-400-Cu Formvar Carbon Film 400 Mesh grid from Electron Microscopy Sciences) and incubated for 30 s. The excess of liquid was removed by a filter paper and the grid was washed with 5 µL of the staining solution. An aliquot of 20 µL of staining solution was added, and the grid was incubated for 60 s, before the drop was removed. The grid was washed by dipping it three times in a drop of water and the excess was removed with a filter paper. The finished grid was dried for 20 min in air or in a desiccator. The prepared samples were imaged using a Philipps CM100 transmission electron microscope at 100 kV.
FRET Experiments: The FRET experiments were performed using the CARY Eclipse Fluorescence Spectrometer from Varian Inc. The microcuvettes were cleaned with H 2 O, 70% ethanol, and water, before drying with a stream of air. The cuvettes were passivated by incubating the sample chamber in 1% (w/v) bovine serum albumin for 10 min at room temperature. The cuvettes were then washed with pure water and dried with a stream of air. To prevent evaporation and therefore a change in solution properties, 600 µL of n-hexadecane was added on top of the samples. The sample contained 30 µL origami-ELP solution in 1× FB and 30 µL of the same buffer solution with additional 2 m NaCl resulting in a final concentration of 1 m NaCl. The measurements were carried out by repeated heating and cooling ramps from 10 to 55 °C with a rate of 0.5 °C min −1 and a lag time of 2 min between each ramp; at least five heating and cooling ramps were measured. The dyes used were ATTO 532 (excitation λ Ex = 530 nm; emission λ Em = 550 nm) and ATTO 647N (excitation λ Ex = 645 nm; emission λ Em = 665 nm). Throughout this work, the presented temperature profiles of the various rectangle variants were always averaged temperature profiles with a mean T t .
Absorption Measurements: The absorption measurements were carried out at a fixed wavelength of 280 nm using a Jasco J-815 CD-Spectrometer (a combined CD and UV-vis spectrometer) equipped with a peltier cuvette holder. All ELPs were dissolved in 1× FB and had a concentration of 25 × 10 −6 m to compare the T t of the ELP variants. To examine the influence of V40 concentration on T t , concentrations of 1 × 10 −6 , 5 × 10 −6 , and 10 × 10 −6 m, were used with the aforementioned buffer with additional 1 m NaCl. Initially, the sample temperature was changed between 10 to 80 °C with a rate of 2 °C min −1 to roughly determine the T t . Finally, the temperature range was set around the specific T t of each sample and measured with a rate of 0.5 °C min −1 . Heating and cooling curves were measured to observe the typical hysteresis as well.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.