A Biomimetic DNA‐Based Membrane Gate for Protein‐Controlled Transport of Cytotoxic Drugs

Abstract Chemistry is ideally placed to replicate biomolecular structures with tuneable building materials. Of particular interest are molecular nanopores, which transport cargo across membranes, as in DNA sequencing. Advanced nanopores control transport in response to triggers, but this cannot be easily replicated with biogenic proteins. Here we use DNA nanotechnology to build a synthetic molecular gate that opens in response to a specific protein. The gate self‐assembles from six DNA strands to form a bilayer‐spanning pore, and a lid strand comprising a protein‐binding DNA aptamer to block the channel entrance. Addition of the trigger protein, thrombin, selectively opens the gate and enables a 330‐fold increase inw the transport rate of small‐molecule cargo. The molecular gate incorporates in delivery vesicles to controllably release enclosed cytotoxic drugs and kill eukaryotic cells. The generically designed gate may be applied in biomedicine, biosensing or for building synthetic cells.


Introduction
[7] Advanced membrane gates with valve-like function are also of considerable interest.Biological membrane gates specifically recognize bio-ligands,such as proteins,and in response open or close to control transport across lipid bilayers.This advanced function could be used in biosensing approaches, [8] drug delivery systems, [9] or synthetic cell-like entities. [10]Adapting natural gates [11] for applications outside their biological remit is, however, difficult.One hurdle is the complex conformational changes between molecular recognition and channel opening.Hence,there is demand for asimple chemical strategy to build de novo channels [12][13][14] with defined molecular recognition with an effective opening mechanism.Ideally,t he synthetic functional gates would be suitable to regulate flux of bioactive substances across membranes.
[24][25][26] Versions have also been made where aDNA strand reversibly blocks the channel lumen in response to ac omplimentary ligand strand, or temperature to enable fluorophore transport. [24,27]However, an unmet challenge is to rationally design advanced channels that are triggered by biological ligands to control flux of bioactive cargo.
Here we use DNAn anotechnology to build an artificial protein-regulated molecular gate for off-on switched transport of cytotoxic drugs.T he protein-gated nanopore,p NP,i s composed of six DNAoligonucleotides forming amembranespanning six-duplex nanobarrel with a2nm-wide inner lumen (Figure 1, light and dark blue), [24] and a7 th strand, ap roteinresponsive lid to regulate transport (Figure 1, red and orange).Theo uter dimensions of the molecular gate are approximately 13 5 5nm.

Results and Discussion
In pNPsclosed state,the lid strand was designed to span the channel entrance by binding to three docking sites and ah inge at the top of the pore (Figure 1,left panel;F igure S1,S2). [24,27]To function as at uneable protein-sensitive gate,t he lid was coded with aD NA aptamer sequence (Figure 1, orange)(Figure S1,S2).][30][31][32][33][34][35][36][37] In our protein-gated nanopore,w eu sed the well characterized thrombin binding aptamer (TBA) (Figure 1, orange). [38] TheTBA sequence and another part of the lid span the pore entrance in the closed pNP (Figure 1, left panel) ForpNP opening, thrombin binds to the lidsTBA domain and partially unzips the lid (Figure 1, right panel;t hrombin, purple;F igure S2).Following our rational design, lid dissociation is anticipated only at the lid docking regions (D) but not at the hinge region  S2, and Figure S1.Tw o variants of the protein-gated nanopore,pNP and pNP2, were designed to probe the influence of TBAp osition on gate function.In pNP,the TBAsequence is located in the channelspanning section of the lid between docking regions D 1 and D 2 ,while in variant pNP2, the TBAsequence is located in the other channel-spanning section between docking region D 3 and the hinge,H(Figure 2A,T able S1, Figure S1).All pores carry four cholesterol modifications to facilitate insertion into lipid bilayer membranes (Table S1,S2, Figure S1).
Prior to building the protein-gated nanopore,w ed etermined the affinity of TBAf or thrombin using an electrophoretic mobility shift assay.T BA contained a2 6-nt tail to enhance staining in the assay (Table S1).Adefined up-shift of the gel band upon thrombin addition implied the formation of acomplex between the aptamer and thrombin (Figure S3A).Its affinity was obtained by determining the band intensity of the complex as af unction of thrombin concentration (Figure S3). [40]Plotting the binding curve and applying aLangmuir fit yielded a K d of 152 AE 11 nm (n = 3).Ther esult confirmed stoichiometric binding between TBAa nd thrombin (Figure S3). [38] Thep rotein-gated nanopore pNP was self-assembled by annealing an equimolar mixture of the six pore oligonucleotides plus the lid strand containing the TBAs equence (Figure 1, Figure S1).Successful folding of pNP was con-firmed by the presence of asingle band in gel electrophoresis (Figure S4).Thebarrel without the lid migrated faster due to its smaller size and molecular weight (Figure S4).pNP2 was also assembled successfully and, as expected, ran similarly to pNP in gel electrophoresis (Figure S4).pNPsa ffinity for thrombin was then probed using ag el shift assay.T he assay is able to discriminate pNP from the pNP-thrombin complex due to their different migration through the gel matrix (Figure 2B,Fand B, respectively).The formation of the complex was followed by increasing the  The position and bp length of the dsDNA markers are given at the left of the gels.C) Quantitative analysis of the gel-shift data via plotting the normalised amount of the pNP-thrombin complex against the concentration of thrombin.The amount of the complex is derived from the gel band intensities via 1À(I pNP ÀI background )a nd normalisation to maximum binding.The derived K d is 662 AE 93 nm (n = 3).D) Kinetic Cy3f luorescence emission supports lid opening upon thrombin binding at different concentrations of thrombin of 0.2 mm (grey) and 2 mm (black) while buffer without protein (light grey) causes no change.For 100 %lid opening, asample of pNP was incubated for 30 min at 55 8 8C, which is higher than the melting temperature of the lid with docking regions D 1 -D 3 but below that of the hinge, which tethers the lid to the pore.

Angewandte Chemie
Forschungsartikel concentration of thrombin (Figure 2B).No other faster migrating bands were detected (Figure 2B), suggesting that pNP remains fully intact following thrombin binding (Figure 1).Complete binding was observed at ar atio of 20:1 of thrombin to pNP (Figure 2B).Thea ffinity of the thrombin-pNP interaction was determined by plotting the gel band intensities of the complex against the thrombin concentration (Figure 2C).TheL angmuir fit-derived K d was 662 AE 93 nM.The K d value is higher than for the isolated TBA-thrombin and can be attributed to reduced steric accessibility of the TBAs equence in pNP (Figure 1, Figure 2A).Thes ame analysis route revealed that pore variant pNP2 had atwo-fold weaker K d of 1.31 AE 0.26 mm (n = 3) and required a40:1ratio for stoichiometric binding (Figure S5).Thedata indicate that both lid designs yield afunctional response.
Thea bility of thrombin to unzip the lid from pNP was monitored kinetically using fluorescence emission.pNP was equipped with adonor Cy3 dye at the 3' end of the lid and an acceptor Cy5 dye on the barrel (Figure 2A,T able S1,S2).In the closed state of pNP,the Cy3 emission was expected to be low due to its close proximity to Cy5 (Figure 2A,F igure S6).In contrast, thrombin binding was anticipated to unzip the lid and increase the distance between Cy3 and Cy5 (Figure 2A, Figure S6).Thee xpected increase in Cy3 emission was confirmed by kinetic fluorescence measurements at 0.2 and 2 mm thrombin leading to 36 %and 75 %higher fluorescence, respectively (Figure 2D)(0.1 mm pNP).No unzipping was observed in the absence of thrombin (Figure 2D).pNP2 displayed similar opening kinetics (Figure S7).In addition, when the TBAs equence was absent from the lid, no gate opening was observed (Figure S8, pNP3).Thereversibility of the lid opening was not tested but could be achieved by as horter hinge region between aptamer and nanopore to allow dissociation of aptamer-thrombin complexes,o rc ompetitive displacement of aptamer-thrombin with free aptamer strand.
Confocal laser scanning microscopy confirmed pNP was successfully anchored to bilayer membranes.Aversion of the protein-gated nanopore carrying aT AMRA dye (Figure S1, pNP TAMRA )w as incubated with POPC giant unilamellar vesicles (GUVs).Microscopic analysis showed successful binding of the lipid-anchored pore to bilayer membranes by the formation of af luorescent halo around the GUV perimeter (Figure S4).
Controlled transport across membrane-inserted pNP was probed with adye flux assay (Figure 3A).Sulforhodamine B (SRB) was encapsulated inside large unilamellar vesicles (LUVs) (Figure S9) at ah igh, self-quenching concentration. [41]Upon dye transport, its concentration and quenching effect decreases giving rise to as ignificant increase in detectable fluorescence.A ddition of thrombin was expected to open pNP and thereby enable dye efflux and fluorescence emission (Figure 3A).In line with expectations,there was no change in fluorescence over 55 min when either pNP or thrombin was absent from the assay (Figure 3B).By contrast, emission was strongly increased upon addition of thrombin at 0.2 to 2 mm (Figure 3B).Thed ata demonstrate proteintriggered opening of pNP to enable transport of the dye.Analysis of the efflux rates revealed a330-fold enhancement upon protein binding (Figure 3C)f rom an average of 0.0039 AE 0.010 %/min (n = 7) at 0 mm thrombin to 1.2 AE 0.17 %/min at 2 mm thrombin(n = 3) (Figure 3C).Very similar fluorescence profiles confirmed protein-triggered opening for pNP2 (Figure S10,S11).In further support, ahigher nanopore concentration led to higher release activity (Figure S12) while no release was observed when using ap ore variant without the TBAs equence (Figure S13, pNP3).
In the final part of our study,weinvestigated the use of the protein-triggered gate to deliver therapeutically relevant cargo for controlled cell killing (Figure 4A).In the assay, pNP2 was inserted into LUVs filled with 3 mm topotecan (Figure 4B), ac linically used cytotoxic drug active against cervical cancer. [42]Thet ri-component vesicles were added to HeLa cervical cancer cells,a nd thrombin was used to trigger the delivery of the cytotoxic drug to cells (Figure 4A).Cell viability and death was monitored for 3dusing light microscopy and the WST-1 colorimetric assay.T he results established that the cytotoxic drug (D) was released to lower cell viability only when pNP and thrombin (T) were present (Figure 4C,F igure S14).After treatment, cell viability was 20 AE 2% when compared to 95 AE 5% for cells incubated with buffer (Figure 4C,L UV/D/pNP + T; term pNP is used instead of pNP2 for clarity and because both molecular gate variants have very similar functional properties).
Controls confirmed that cell viability was minimally affected with either thrombin or pNP (Figure 4C,T hrombin, pNP).Indeed, thrombin slightly increased viability in line with other studies. [35]Further controls elucidated the effect of each component, and in combination with others,o nc ell viability.F or example,neat drug without encapsulation led to ar eduction of cell viability to 42 AE 2% (Figure 4C,D rug).Surprisingly,v esicles filled with drug exhibited as imilar amount of cytotoxicity (Figure 4C,L UV/D) indicating that vesicles can fuse with the cell membrane and intracellularly deliver the cytotoxic cargo. [43,44] usion and uptake may be prevented by modifying the vesicle membrane surface.I ndeed, vesicles decorated with the negatively charged membrane-anchored pNP exhibited reduced cytotoxicity with an increase in cell viability to 55 AE 1% (Figure 4C,L UV/D/ pNP).By contrast, addition of thrombin to the pNP-LUVs led to areduction in cell viability to the previously noted 20 AE 2% (Figure 4C,L UV/D/pNP + T).Thed ata support the use of the protein-triggered valve to deliver therapeutically relevant cargo for controlled cell killing (Figure 4).Protein-triggered opening of pNP releases acytotoxicd rug for controlled killing of cells.A) Scheme of the assay to demonstratet he controlled killing of HeLa cells (pink).The cells are exposed to pNP-functionalised-membrane vesicles filled with the cytotoxic drug, topotecan (green), as well as thrombin (purple) to open pNP (blue, red), followed by incubation for 3dto attain the cytotoxic effect of released topotecan.Forvisual clarity,apore inserted in the membrane in the opposite orientation is not shown.The mixed orientationscan lower the release to ad egree of up to 50 %.B) Chemical structure of topotecan.C) Graph displaying the viability of HeLa cells after 3dincubation with either thrombin, pNP, topotecan, topotecan-filled LUVswith alipid ratio of PC:PE (7:3), pNP-functionalised topotecan-filled LUVs, and the latter in combination with thrombin.The data are the means AE SD collected from three independentexperiments.The assay was carried out with pNP2 but is referred to pNP for reasons of simplicity and because both moleculargate variants have very similar functional properties.The cell viability was determined with the WST-1 assay.

Forschungsartikel Conclusion
In summary,t his study describes the first DNA-based membrane gate capable of controllably delivering atherapeutic drug to acellular environment in response to abiologically relevant exogenous trigger.T he biomimetic nanodevice is self-assembled from just seven oligonucleotides and achieves high functional performance by increasing the transport rate 330-fold upon actuation. Previous DNAg ates were designed to respond to DNAl igand [24] or elevated temperature [27] but not proteins.F inally,t he gates controllable drug release implies compatibility with potential biomedical applications.B ased on its modular design, the nanodevice could be adapted for arange of different protein triggers with applications in biosensing, research, and biomedicine.

Figure 1 .
Figure 1.DNA-based protein-gated nanopore,p NP, opens upon thrombin binding allowing the transport of material.The pore's barrel is composed of six DNA strands (light and dark blue) that form six interconnected DNA duplexes arranged in ah exagonal fashion.T he protein-responsive lid (red, orange) features the thrombin-binding aptamer (orange),w hich is bound to the pore via two extended docking loops, including the hinge region.In the closed state, the lid blocks the channel of pNP.Binding to thrombin (dark purple) leads to the partial dissociation of the lid and opening of pNP's channel, allowing cargo transport.The lid remains attached to the hinge region of the pore.Four cholesterola nchors (pink) insert pNP into the hydrophobic lipid bilayer membrane.

Figure 2 .
Figure2.Thrombin binding actuates pNP lid-opening.A) 3D scheme of pNP's top with TBA sequence(orange) in the lid (red) which carries aC y3 fluorescence donor while the Cy5acceptor is linked to the pore.Thrombin-induced lid opening separatest he dyes and enhances Cy3 emission.B) Gel shift assay illustratingp NP-thrombin binding.Increasing concentrations of thrombin lead to ap rogressive mobility shift from the free pNP (F) to the bound pNP-thrombin complex (B).The position and bp length of the dsDNA markers are given at the left of the gels.C) Quantitative analysis of the gel-shift data via plotting the normalised amount of the pNP-thrombin complex against the concentration of thrombin.The amount of the complex is derived from the gel band intensities via 1À(I pNP ÀI background )a nd normalisation to maximum binding.The derived K d is 662 AE 93 nm (n = 3).D) Kinetic Cy3f luorescence emission supports lid opening upon thrombin binding at different concentrations of thrombin of 0.2 mm (grey) and 2 mm (black) while buffer without protein (light grey) causes no change.For 100 %lid opening, asample of pNP was incubated for 30 min at 55 8 8C, which is higher than the melting temperature of the lid with docking regions D 1 -D 3 but below that of the hinge, which tethers the lid to the pore.

Figure 3 .
Figure 3. Protein-actuated opening of pNP controls transport of molecular cargo across lipid bilayers.A) pNP is embedded in the lipid bilayer of avesicle filled with the fluorophore sulforhodamine B( SRB, green dots).The dye is contact-quenched at 50 mm inside the vesicle.In the closed state of pNP, the encapsulatedS RB cannot traverse the membrane.Thrombin-binding results in the partial unzippingoft he lid and pore opening to release SRB into the ambient.The lower dye concentration abolishes contact-quenching and increasesfluorescence.Forv isual clarity,apore inserted in the membrane in the opposite orientation is not shown.The mixed orientationsc an lower the release to adegree of up to 50 %.B) Kinetic traces of SRB fluorescence as af unction of increasing pNP concentration.100 %release is the total amount of fluorescence obtained upon rupturingvesicles with adetergent.C) Bar chart of net fluorescence increase, summarizingdata from (B).The data represent averages and standard deviations from at least 3i ndependent experiments.

Figure 4 .
Figure 4. Protein-triggered opening of pNP releases acytotoxicd rug for controlled killing of cells.A) Scheme of the assay to demonstratet he controlled killing of HeLa cells (pink).The cells are exposed to pNP-functionalised-membrane vesicles filled with the cytotoxic drug, topotecan (green), as well as thrombin (purple) to open pNP (blue, red), followed by incubation for 3dto attain the cytotoxic effect of released topotecan.Forvisual clarity,apore inserted in the membrane in the opposite orientation is not shown.The mixed orientationscan lower the release to ad egree of up to 50 %.B) Chemical structure of topotecan.C) Graph displaying the viability of HeLa cells after 3dincubation with either thrombin, pNP, topotecan, topotecan-filled LUVswith alipid ratio of PC:PE (7:3), pNP-functionalised topotecan-filled LUVs, and the latter in combination with thrombin.The data are the means AE SD collected from three independentexperiments.The assay was carried out with pNP2 but is referred to pNP for reasons of simplicity and because both moleculargate variants have very similar functional properties.The cell viability was determined with the WST-1 assay.