Control of enzyme reactions by a reconfigurable DNA nanovault

Biological systems use compartmentalisation as a general strategy to control enzymatic reactions by precisely regulating enzyme–substrate interactions. With the advent of DNA nanotechnology, it has become possible to rationally design DNA-based nano-containers with programmable structural and dynamic properties. These DNA nanostructures have been used to cage enzymes, but control over enzyme–substrate interactions using a dynamic DNA nanostructure has not been achieved yet. Here we introduce a DNA origami device that functions as a nanoscale vault: an enzyme is loaded in an isolated cavity and the access to free substrate molecules is controlled by a multi-lock mechanism. The DNA vault is characterised for features such as reversible opening/closing, cargo loading and wall porosity, and is shown to control the enzymatic reaction catalysed by an encapsulated protease. The DNA vault represents a general concept to control enzyme–substrate interactions by inducing conformational changes in a rationally designed DNA nanodevice.


Supplementary Note 1. Design of hinges and locks.
The DV hinge module is composed of 5 staple strands that bridge the two halves of the DV connecting the DNA helices 37 and 39 ( Supplementary Fig. 2a,b). These hinge strands protrude from the DNA helices pointing approximately at each other in order to define a tight hinge region in DV-H1 (see main text and Supplementary Note 2). The DV locking system is designed to mimic a zip closure, where the stem-loop locks initiates the closing mechanism to move the sets of doublehelix locks closer to the hinge region (the rear locks) into closer proximity. The rear locks will then interact forcing the two halves of the DV even closer together, thus facilitating the interaction between the next sets of locks: the side locks farther from the hinge (the middle locks) and the front lock. Finally, the middle locks and the front lock will anneal stabilizing the DV in a tightly closed conformation.
The position and orientation of both stem-loop and double-helix lock staple strands were designed in order to force the two halves of the DV as close as possible. For this purpose we used two different and complementary design approaches. The stem-loop lock strands (Supplementary Fig. 2a,c) and the rear double-helix lock strands (Supplementary Fig. 2a,d) were designed to protrude from the DNA helices pointing at each other in order to favour lock-to-lock interactions during the first steps of the closing mechanism. The double-helix middle lock strands (Supplementary Fig. 2a,e) and the front lock strands (Supplementary Fig. 2a,f) instead were designed to protrude pointing in the same direction in order to force the DNA helices they protrude from closer together in the end. The obtained effect is that the DNA helices at the interface of the two halves of the DV should adopt the tightly packed hexagonal lattice.

Supplementary Note 2. Characterization of the hinge region.
When DV-H1 was assembled in the closed state it migrated as one major band in gel electrophoresis ( Supplementary Fig. 7a), which was observed in TEM images to be composed mainly of dimers ( Supplementary Fig. 7c). Opening the DV by incubation with OpK resulted in a faster mobility species in gel electrophoresis ( Supplementary Fig. 7a) and the observation of monomeric structures in TEM images ( Supplementary Fig. 7d).
To investigate the opening mechanism in real time we labelled the two halves of the DV with Cy3 and Cy5 fluorophores, respectively, to allow the monitoring of the closed/open conformation by Förster resonance energy transfer (FRET). FRET measurements revealed low efficiency values for a closed nanostructure (10%; Supplementary Fig. 7b). This is probably because DV that form dimers are not correctly closed: double-helix lock staples that interact with lock staples belonging to different structures are not available to interact with the lock staples belonging to the same structure, thus not keeping the DV closed. Nevertheless, when OpK is added FRET efficiency values decrease from 10% to 5% indicating that Cy3 and Cy5 molecules are moving apart. This could be due to a combination of opening of the DV structure, and to the disruption of the dimerization interactions that cause DV monomers to fall apart.
The FRET analysis did not allow a satisfying characterization of the opening/closing mechanism of DV-H1. Therefore two different strategies were adopted to obtain a better understanding of the nanomechanical properties of the DV: (i) we directly measured the size from TEM images of DV assembled in the closed state and from DV incubated with OpK, and (ii) we measured FRET efficiency on DV assembled in the open conformation after incubation with ClK. First, in order to understand if the DV with stiff hinge is capable of opening we measured the size of 100 monodispersed structures assembled in the closed state, and after incubation with OpK, as visualized on TEM ( Supplementary Fig. 7c). The statistical analysis conducted shows a value of 26.0 ± 2.4 nm for the closed structure (theoretical value = 26 nm), and a value of 30.5 ± 2.7 nm for the open structure (theoretical value = 31 nm) indicating that DV monomers are successfully assembled in the closed state, and that they open upon incubation with 5-fold molar excess of OpK.
Next, we assembled Cy3/Cy5-modified DV-H1 in the open conformation and analysed it for FRET efficiency. The FRET efficiency values measured are consistent with the previous analysis: 5% FRET efficiency was measured on the open DV, increasing to 10% after incubation with ClK ( Supplementary Fig. 8, blue line), thus demonstrating that DV with stiff hinge closes upon incubation with 1.3 fold molar excess of ClK.
These structural observations can be explained by a tight hinge region of DV-H1, which would lead to decreased DV closure allowing the locks to dimerize and to the tight 90-degree open conformation observed. To investigate this hypothesis we designed two more DV structures with different length of the hinge region of 4 nt (DV-H2) and 8 nt (DV-H3) to gradually increase the flexibility of the hinge. The structures were assembled in the open state and had similar gel mobility (data not shown) and low FRET efficiency as DV-H1 ( Supplementary Fig. 8). However, the three DV behave differently when the ClK was added: the FRET efficiency value rose only 5% for DV-H1, whereas it rose 12% for DV-H2 and 20% for DV-H3. This clearly shows that a flexible hinge of 8 nt is required to allow closure of the DV structure, while a tight hinge region inhibits correct closure.

Supplementary Note 3. Alpha-Chymotrypsin modification.
The peptidase aCt was first reacted with azide-NHS ester in order to obtain azide-exposing enzymes to directly conjugate to the alkyne-exposing DV. In order to determine the yield of the first reaction, azide-modified aCt was reacted with an alkyne-modified DNA strand ( Supplementary Fig. 11a).
SDS-PAGE analysis of aCt, azide-modified aCt (aCt-N3), and DNA-modified aCt (aCt-DNA) shows that about 50% of aCt-DNA did not react with alkyne-modified DNA ( Supplementary Fig.   11b). Considering the high efficiency that usually characterize "click" reactions, this is likely due to a non-complete modification of aCt with azide-NHS ester in the first place. Afterwards we measured the retained enzymatic activity of aCt-N3 and aCt-DNA in DV activity buffer (Buffer 1: TAE + 30 mM MgCl 2 + 3 mM CaCl 2 ), and in aCt storage buffer (Buffer 2: 1 mM Tris + 2 mM CaCl 2 ) using the chromogenic substrate sAAPFpNA ( Supplementary Fig. 11c).
The product turnover speed of aCt was reduced to 42% after conjugation with NHS ester-azide molecules ( Supplementary Fig. 11d). This could indicate that the portion of aCt that was successfully modified (about 50%, as noted above) completely lost activity due to the presence of azide handles. However, after reaction with alkyne-modified DNA the product turnover speed of the enzyme was further reduced to 22% suggesting that a different mechanism could cause the enzymatic activity loss. Finally, proteolytic activity was successfully detected on aCt-loaded DV indicating that aCt retained enzymatic activity after conjugation with the DNA origami structure.
We thus reasonably ascribe the activity loss measured on chemically-modified enzymes to both the presence of chemical groups on the enzyme surface, and to the enzyme handling along the conjugation procedure in presence of DMSO, copper and chelating agents, and shaking for 3 hours in total at 21 and 25°C (see Materials and methods).

Supplementary Note 4. DNA Vault purification.
The DV was purified by PEG precipitation 1 to remove unconjugated enzymes. The purification efficiency was estimated by testing unmodified aCt in the absence of DNA nanostructures, and assaying both the pellet and supernatant for enzymatic activity by using sAAPFpNA. As expected, aCt was almost exclusively detected in the supernatant ( Supplementary Fig. 13a) while the DV is recovered in the pellet. In order to understand if non-loaded aCt was co-purified with the DV due to unspecific enzyme-DNA interactions the purification protocol was applied on a sample composed of azide-modified aCt and DV lacking the alkyne handle necessary for enzyme loading. Only minor unconjugated enzyme impurities were retained upon purification ( Supplementary Fig. 13b).

Supplementary Note 5. Characterization of cargo anchoring site (CAS) region.
As described in the main text, open DV-ssAlk nanostructures were loaded with azide-exposing aCt, purified from excess enzyme, and closed. The sample was then incubated with either OpK or unspecific key. Finally, the chromogenic substrate FITC-casein was added to the samples and the fluorescent signal was measured at specific time-points. Empty open DV was used as control for non-specific degradation of substrate. It was observed that the enzymatic activity measured on open aCt-loaded DV was up to 3 times higher than on closed nanostructures and increasing linearly  Table 1). As expected the assay shows similar enzymatic activity values for both open structures. However, proteolytic activity levels measured on closed DV-dsAlk are higher than the ones measured on the DV-ssAlk counterpart. This effect can be ascribed to the presence of a higher amount of non-closed DNA nanostructures, probably because the bulkier and less flexible double-stranded anchoring site used here (about 5 nm long) interferes with the correct closing mechanism of the DV. We have thus identified an effect of the nature of the cargo-anchoring complex on the performance of the DV, indicating that smaller and more flexible anchoring sites allow tighter control over encapsulated enzymes.

Supplementary Note 6. Protocol for loading aCt enzyme in the DNA Vault.
Here we provide a protocol to follow in order to obtain an aCt-loaded DV sample, as the one used in Fig. 3d-f. This protocol can be applied to more different enzymes, but relevant parameters must be changed depending on the nature and chemo-physical properties of the chosen enzyme (e.g. number of lysine residues exposed on the molecular surface, tolerance to temperature, shaking, chemicals, etc.). Also, different enzyme-DNA conjugation strategies may be used to load different enzymes 2 , again depending on the nature of the chosen enzyme.
3. Modification of aCt with NHS ester-azide molecules. aCt (2 nmol) is reacted with NHS-C3azide linker in lysine : linker = 1:10 molar ratios, and shaken (400 rpm) at 21°C for 2 h in 50 µl final volume of aCt storage buffer. Dilute and wash the sample 2 times with aCt storage buffer by using Amicon Ultra-0.5 ml centrifugal filters (50 kDa) to discard unreacted NHS ester-azide and eventual organic solvents. Purified aCt-N3 may be reacted with alkyne- Incubate by shaking (500 rpm) at 25°C for 1 h. 5. Purification from excess enzyme through PEG-precipitation. To precipitate the enzymeloaded DV was mixed with PEG solution and centrifuge at 13,000 rfc for 30 minutes using a cooled centrifuge (10°C). Then, resuspend the pellet in 100 µl DV activity buffer and add ClK at lock : ClK = 1:1.3 molar ratio. Incubate the sample at room temperature overnight to allow complete resuspension of the DV, and subsequently at 4°C for 36 hours to allow full closure of the DV.

6.
Closed aCt-encapsulating DV is obtained. The sample may be incubated with either OpK or unspecific key at room temperature for 1 hour before the enzymatic activity assay is performed.

Supplementary Figure 1. DNA Vault blueprint.
Side-view (a) and blueprint (b), modified from CaDNAno software interface of DV-H1-7K-Cy3/5 as in Supplementary Table 1 The yellow, green, and cyan lines link the helices connected by hinge, side locks, and front lock, respectively. b, Schematics of the hinge staple strand, connecting helices 37 and 39. c, Schematics of the stem-loop locks, connecting helices 35 and 40. d, Schematics of the rear double-helix side locks staple strand, protruding from helices 31 and 44. e, Schematics of the middle double-helix side locks, protruding from helices 27 and 49. f, Schematics of the front double-helix lock, protruding from helices 24 and 49. Dark orange circles indicate DNA helices that form the DV shell, light orange circles indicate DNA helices that define the DV side-walls, grey circles indicate virtual DNA helices used for design purposes only. Spirals show orientation and direction of the DNA helices bearing lock strands (only the staple strand helix is shown for clarity), while yellow, green and cyan lines indicate the hinge and lock staple strand orientations, respectively, on the honeycomb lattice plane as in Supplementary Fig. 1. Arrowheads and squares-on-ends on spirals indicate 3' and 5' ends, respectively. Figure 3. Strand displacement mechanism of double-helix locks. a, Schematic of a double-helix lock connecting the two halves of the DV. The opening key (orange) first anneals onto the toehold exposed on the long lock strand, and then disrupts the lock-to-lock interactions by strand displacement mechanism, causing the structure to open up. The closing key (brown) works similarly: it first anneals onto the toehold exposed on the opening key, and then it disrupts the opening key-to-lock interactions leaving the long lock strand free to interact with the short lock strand, thus inducing the DV to close back. b, Sequence domain design of the lock and key strand interactions. Sequence domains are indicated by capital letters and small letters for complementary sequences. Furthermore, the domains are coloured to reference the DNA sequences shown in Supplementary Table 2 (Side-lock and Front-lock modules). Figure 4. Strand displacement mechanism of stem-loop locks. a, Schematic of a stem-loop lock connecting the two halves of the DV. The opening key (orange) first anneals onto the loop toehold exposed on the lock strand, and then disrupts the intermolecular lock-to-lock interactions by the strand displacement mechanism, causing the structure to open up. The closing key (brown) works similarly: it first anneals onto the toehold exposed on the opening key, and then it disrupts the opening key-to-lock interactions leaving the lock strand free to assume the stem-loop secondary structure, thus inducing the DV to close back. b, Sequence domain design of the lock and key strand interactions. Sequence domains are indicated by capital letters and small letters for complementary sequences. Furthermore, the domains are coloured to reference the DNA sequences shown in Supplementary Table 2 (Side-lock module).

Supplementary Figure 5. Monomer fraction quantification.
Agarose gel analysis of unpurified DV sample. DV monomers constitute roughly 50% of the sample, while dimers and higher-order multimers constitute the remaining 50%. Supplementary Figure 11. Alpha-Chymotrypsin modification. a, aCt was first reacted with azide-NHS ester in order to obtain azide-exposing enzymes to directly conjugate to alkyne-exposing DV. In order to assess if the enzyme was successfully modified with azide handles prior to covalent loading within the DV, it was reacted with an alkyne-modified DNA strand. b, SDS-PAGE of aCt, azide-modified aCt (aCt-N3), and DNA-modified aCt (aCt-DNA). c, Retained enzymatic activity was measure on aCt-N3 and aCt-DNA using the chromogenic substrate sAAPFpNA. aCt concentration used was 400 nM, substrate concentration was 400 µM; the tests were performed in DV activity buffer (Buffer 1: TAE + 30 mM MgCl 2 + 3 mM CaCl 2 ), and in aCt storage buffer (Buffer 2: 1 mM Tris + 2 mM CaCl 2 ) in triplicates, with error bars indicating one standard deviation. d, Quantification of the retained aCt activity as in C.

Supplementary Figure 12. Cargo displacement in the absence of the DV.
Cargo displacement reactions were performed on the CAS/cargo system alone in order to determine whether the chemo-physical properties of the different invader species affect the displacement reaction kinetics. Fluorescence data show the decrease of FRET peak intensity (cargo-Cy5 emission upon CAS-Cy3 excitation) upon invader strand addition. Cy3-modified CAS and Cy5-modified cargo (Eco-41[84]-41[104]Cy3 and Cargo-(EcoRV)Cy5, respectively) were pre-incubated together at RT for 30 minutes prior addition of the invader strands. Bars indicate one standard error from the mean fluorescence signal as recorded in three replicates. Figure 13. Alpha-Chymotrypsin (aCt) activity after PEG-precipitation. a, In order to determine the purification efficiency of the PEG-precipitation method free unmodified aCt was precipitated using different PEG concentrations, and both pellet and supernatant obtained were tested for retained enzymatic activity using the chromogenic substrate sAAPFpNA. aCt concentration used was 400 nM, substrate concentration was 400 µM; the tests were performed in DV activity buffer (TAE (1X) + MgCl 2 (30 mM) + CaCl 2 (3 mM)). b, In order to understand if nonloaded aCt is co-purified with the DV due to unspecific enzyme-DNA interactions the purification protocol was applied to a sample composed of free azide-exposing aCt and empty DV lacking the alkyne handle necessary for enzyme conjugation. Enzymatic activity was measured on the pellet containing the DV using the fluorogenic substrate FITC-Casein. Enzymatic activity values measured on empty open DV (as in Fig. 4e) are reported here for comparison. Tests were performed in DV activity buffer. Bars indicate one standard error from the mean fluorescence signal as recorded in three replicates.

Supplementary Figure 14. DV-aCt enzymatic activity with different protein-anchoring modules.
Alkyne-exposing DV was assembled with either a double-stranded CAS module (DV-dsAlk, A-D) or a single-stranded CAS module (DV-ssAlk, E-H). a, Schematics of the protein-anchoring region (the cavity is shown in grey), and b, schematic mechanism of the enzymatic activity modulation system in DV-dsAlk. c, aCt enzymatic activity measured on open aCt-loaded (blue), closed aCtloaded (red) and open empty DV-dsAlk (Control, grey). Bars indicate one standard error from the mean fluorescence signal, as recorded in three replicates. d, Normalization of measured enzymatic activity as in C after Control subtraction (closed -control at 30 min = 1.0). e, Schematics of the protein-anchoring region, and f, schematic mechanism of the enzymatic activity modulation system in DV-ssAlk. g,h, aCt enzymatic activity measured on DV-ssAlk as shown in Fig. 4e,f. The data relative to DV-ssAlk (bottom box) are shown here for a better comparison with the analysis conducted on DV-dsAlk (top box).

Supplementary Table 1. Module assembly instructions for DNA Vault samples.
Overview of the DNA origami structures, the opening keys (OpK), and closing keys (ClK) used in the experiments. The table lists the 9 different versions of the DV that was assembled for the experiments presented in this study. Each structure is composed of 8 different modules of staple strands, as listed in Supplementary Table 2