Triggered Assembly of a DNA-Based Membrane Channel

Chemistry is in a powerful position to synthetically replicate biomolecular structures. Adding functional complexity is key to increase the biomimetics’ value for science and technology yet is difficult to achieve with poorly controlled building materials. Here, we use defined DNA blocks to rationally design a triggerable synthetic nanopore that integrates multiple functions of biological membrane proteins. Soluble triggers bind via molecular recognition to the nanopore components changing their structure and membrane position, which controls the assembly into a defined channel for efficient transmembrane cargo transport. Using ensemble, single-molecule, and simulation analysis, our activatable pore provides insight into the kinetics and structural dynamics of DNA assembly at the membrane interface. The triggered channel advances functional DNA nanotechnology and synthetic biology and will guide the design of controlled nanodevices for sensing, cell biological research, and drug delivery.


TABLES AND FIGURES 14
. Names, modifications and sequences of DNA oligonucleotides used for folding A, B, A•B and variants. 14 Table S2. Names and strand compositions of structures used for the DNA nanopore with biomimetic triggered assembly 15  Table S3. Summarized FRET efficiency (E) data for pore assembly derived from the FRET pore assembly binding titrations and pore assembly kinetics. 27 Table S4. Names, modifications, and sequences of DNA oligonucleotides used for folding R, S and SNP. 28 Table S5. Names and strand compositions of structures used for the model system.

DNA assembly
Equimolar mixtures of DNA oligonucleotides (1 μL each, stock concentration of 100 µM) (Table  S2 for composition of DNA pores and components) were dissolved at 1 µM in a buffer solution of either buffer A (1x PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.4), buffer B (300 mM KCl, 15 mM Tris-HCl, pH 7.4) or buffer C (12 mM MgCl2 in 0.6x TAE (40 mM Tris, 20 mM acetic acid), pH 7.4) to a final volume of 100 μL. Folding was achieved on a BioRad T100 Thermocycler (UK) using a program involving heating to 95 °C and holding for 0.5 min, then cooling to 75°C within 5 min, holding for 1 min before cooling to 4 °C at a rate of 0.5 °C per 1 min. Samples were stored at 4 °C for up to 1 week.

PAGE
The assembled DNA nanostructure and component DNA oligonucleotides were analyzed with commercial 10% polyacrylamide gels (BioRad, UK) in 1x TBE buffer (100 mM Tris, 90 mM boric acid, 1 mM EDTA, pH 8.3). For gel loading, a solution of the DNA nanopores (2 μL, 1 μM) was mixed with folding buffer (13 µL, 2 mM MgCl2 in 0.6x TAE, pH 7.4) and 6x gel loading dye (5 μL, New England Biolabs, UK). Gels were run at 115 V for 90 min at 4 °C. The gel bands were visualized by staining with ethidium bromide and UV illumination. A 100 bp marker (New England Biolabs, UK) was used as a reference standard.

Agarose gel electrophoresis
The assembled DNA nanostructures and component DNA oligonucleotides were analyzed with 2-3% agarose (Invitrogen, UK) gels in 1x TAE buffer (40 mM Tris, 20 mM acetic acid, 1 mM EDTA, pH 8.3). For gel loading, a solution of the DNA nanostructure (2 μL, 1 μM) was mixed with folding buffer (13 µL) and 6x gel loading dye (5 μL, New England Biolabs, UK). The gel was run at 60 V for 60 min at 4 °C. The gel bands were visualized by staining with ethidium bromide and UV illumination. A 100 bp marker (New England Biolabs, UK) was used as a reference standard.

Preparation of small unilamellar vesicles (SUVs)
DPhPC (100 µL, 10 mM) or POPC (100 µL, 10 mM) in chloroform was added to a 5 mL round bottom flask. The solvent was removed using a rotary evaporator (Buchi, Newmarket, UK) to yield a thin film, which was further dried under high vacuum (Buchi, Newmarket, UK) for 1 h. The lipid was re-suspended in 1 mL of either buffer A or buffer B. The solution was sonicated for 20 min at 30 °C and then equilibrated for 1 h before being extruded 25 times through a 0.1 µm polycarbonate membrane (Avanti Polar Lipids, US) using the extruder kit (Avanti Polar Lipids, US). SUVs were then stored at 4 ˚C and used within 48 h.

Melting temperature (Tm) analysis using UV-vis spectroscopy
UV melting profiles were obtained using a 10 mm quartz cuvette (Hellma Analytics, Southendon-Sea, UK) in a Varian Cary 300 Bio UV-vis spectrophotometer (Agilent, Cheadle, UK) equipped with a Peltier element (Agilent, Cheadle, UK). Samples were analyzed at 200 nM and SUVs composed of DPhPC at 200 µM lipid concentration. Samples were analyzed by monitoring the change in absorbance at 260 nm as the temperature was increased from 20 to 80 °C at a rate of 1 °C/min. Melting profiles were then background corrected, and the 1 st derivative calculated to identify the Tm.

Electrophoretic mobility shift assay
For binding titrations, component A ∆C or A (5 μL, 1 μM) was mixed with component B ∆C or B (1 µM stock) in buffer A yielding concentrations of 0 to 0.5 µM in a final volume of 20 μL. In the case of A-SUV vs B ∆C or B, component A (5 µL, 1 µM) was first added to SUVs (5 µL, 100 nm, 16.7 nM). After incubation for 30 min at 30 °C, 6x gel loading dye (5 µL, New England Biolabs, Hitchin, UK) was added, the samples were mixed and loaded onto a thermally equilibrated 2-3% agarose gel. The gel was run in 1x TAE buffer, pH 8.3 at 60 V for 60 min at 4 °C. Staining and molecular markers were as described in section 1.4. Band intensities were analyzed using ImageJ and normalized as (1-(IA-Ibackground)). The normalized intensities were then fit to a Langmuir curve to determine the Kd.
Kinetic assembly titrations, component A ∆C (5 μL, 1 μM) was mixed with B ∆C (5 µL, 1 µM) in buffer A to a final volume of 20 μL. Samples were incubated at 30 °C for 0, 1, 5, 10, 15, 20, 25 and 30 min while shaking at 500 rpm. Samples were prepared in reverse time order and after all samples were prepared, samples were crashed in ice to arrest pore formation. For assembly locked components A ∆C L A vs B ∆C L B the keys, K A and K B (1 µL, pre-mixed, 5 µM) were also added to each timepoint. Samples were mixed with 6x gel loading dye (5 µL) and then loaded onto thermally equilibrated 10% PAGE. The gel was run in 1x TBE buffer at 115 V for 90 min at 4 °C. Staining and molecular markers were as described in section 1.3.

FRET assay on pore assembly
For binding titrations, the assembly of A•B was investigated using a fluorescence spectrophotometer (Cary Eclipse, Agilent, Cheadle, UK). To a plastic Eppendorf tube was added A ∆C , A or AL A (12 μL, 1 μM), B ∆C , B or BL B (0 µL, 1.2 µL, 2.4 µL, 6 µL, 12 µL, 24 µL; 1 µM), SUVs (0 µL, 6 µL; 1 mM lipid, 7.22 nM SUV) and buffer B to a final volume of 120 µL. The tube was then incubated at 30 °C for 30 min while shaking at 750 rpm. The combined solution was then added to a 10 mm quartz cuvette (Hellma Analytics, Southend-on-Sea, UK), which was placed in the fluorescence spectrophotometer and scanned (ex545 nm, em555-725 nm). Pre-folded A•B was used as a control for maximum assembly. Where SUVs were used, A or AL A and SUVs were mixed and left to bind for 10 min prior to addition of B ∆C , B or BL B . The emission intensity of the donor (Cy3) were normalized between A ∆C or A and a pre-folded control pore (A•B) ∆C , (A•B) 1C , or A•B. A 1:2 of A:B was used as an internal control. Due to the ability of the donor (Cy3) to donate to multiple acceptor (Cy5) molecules, this was set to the same binding level as the pre-assembled control and was used as an anchor point for Kd determination.
For kinetic assembly, the assembly of A•B was investigated by monitoring Cy3 emission (ex550nm, em570nm) using a fluorescence spectrophotometer (Cary Eclipse, Agilent, Cheadle, UK). To a 10 mm quartz cuvette (Hellma Analytics, Southend-on-Sea, UK), A ∆C or A (2.5 μL, 1 μM) was added to SUVs (0 µL, 1.25 µL; 1 mM lipid, 7.22 nM SUV) and buffer B (97.5, 96.25 µL) and the signal left to stabilize for 5 min. Then, B ∆C or B (50 µL, 1 µM) was rapidly added and mixed. Pore formation was monitored for 1 h. Where SUVs were used, A and SUVs were mixed and left to bind for 10 min prior to the start of the run.
FRET Efficiency (E) calculations were achieved using the equation (2): Where IDA is the donor intensity in the presence of the donor and acceptor; ID is the donor intensity in the absence of the acceptor.
The inter-fluorophore (Cy3-Cy5) distance was calculated using equation (3): E stands for FRET efficiency, r is the donor-acceptor separation distance, R0 is the Förster distance where E = 50%.

Dual-color FCCS analysis to measure the Kd of pore assembly
Dual-color fluorescence cross-correlation measurements were carried out on a commercial laser scanning microscope (ConfoCor 3, Carl Zeiss, Jena, Germany) equipped with a 40x water immersion objective. A ∆C and B ∆C were labeled with the spectrally non-overlapping fluorophores Alexa488 and Alexa647, respectively. Cross-correlation measurements were performed using a 635 nm secondary dichroic mirror with a 505-540 nm bandpass filter in the green channel, and a 650 nm longpass filter in the red channel. Laser power was adjusted such that the brightness ratio of Alexa647 to Alexa488 was roughly 3:1.
At a 1:1 binding stoichiometry, the bound fraction of either Alexa488 A ∆C (Xg) or Alexa647 B ∆C (Xr) can be calculated from the number of double labelled particles (N rg ) relative to the total number of each particle (N g for Alexa488 A ∆C or N r for Alexa647 B ∆C ) using equations 4 and 5: The number of particles of Alexa488 A ∆C or Alexa647 B ∆C detected in the green or red channel, respectively, was obtained from the fit of the respective autocorrelation (for N g and N r ) and cross-correlation (for N rg ) function G(τ) by the one-component model for 2D translational diffusion shown in equation 6: 1 where ! represents the diffusion time through the confocal volume. The fraction of bound particles, Xg and Xr, was corrected by accounting for the difference in size of the green and red detection volumes, Vg and Vr, by using the following formulas: 2 where Veff represents the effective cross-correlation volume, which is defined by the following equation: 3 where S is a structural parameter, which was set equal for both channels, Sg = Sr = 6, and ω is the respective lateral radius of the confocal volume, ωg or ωr, of the green or red confocal volume. Confocal volumes, V, in the red and green channels were calculated as: 1 = +/, * + * (10) The lateral radii, ωg and ωr, we obtained from calibration experiments measuring labels with known diffusion coefficients (D), Rhodamine 6G (Merck, Germany) and Alexa647 maleimide (Jena Bioscience, Germany), using the equation: 1 On each day that data were collected, the maximum achievable cross-correlation was determined using a DNA-duplex carrying both an Alexa488 and Alexa647 fluorophore (IDT, USA) to represent 100% binding.
For binding measurements, Alexa488 A ∆C was mixed with Alexa647 B ∆C in buffer B, to a final volume of 60 µL. In other measurements, the concentration of Alexa647 B ∆C was varied from 0.5 nM to 300 nM, while the concentration of Alexa488 A ∆C (7.5 µL, 100 nM) remained constant. The samples were then incubated at 30 °C for 30 min with shaking. After incubation, each ratio mixture was measured separately. Measurements where Alexa647 B ∆C remained constant (9.6 µL, 100 nM), while the concentration of Alexa488 A ∆C was varied.
To determine Kd, the resulting volume-corrected fractions of bound particles Xg/v and Xr/v were fitted to a Langmuir isotherm: where C represents concentration of the varied component.
By design, a dc-FCCS experiment shows the saturation value below 100%, and a slight overestimation of the cross-correlation is observed in the red channel ( Figure S8a). The former originates from the non-complete overlapping of the confocal volumes for the red and green lasers. The latter stems from the bleed-through of the green-labelled particles into the red channel at higher ratios of green-to-red particles, which is pronounced at ratios higher than 10:1. 4 To minimize the effect, we performed before every FCCS experiment a control measurement using a cross-correlation standard. This is a fluorescently double labelled DNAduplex. Within 5% accuracy, the control measurements showed the same deviation from 100% saturation level, and the same ratio of saturation levels between the red and the green channels.

Preparation of giant unilamellar vesicles (GUVs)
A solution of POPC lipids (5 μL, 10 mM in chloroform) was added to an indium tin-oxide (ITO) coated glass slide. Within 5 min the solvent evaporated, and a dried lipid film was formed. The glass slide was then inserted in a vesicle prep device (Nanion Technologies, Munich, Germany). An O-ring was added around the patch. Sucrose (300 μL, 1 M in water) was added to the lipid film patches confined by the O-ring. Finally, another ITO glass slide was applied from the top, resulting in a sealed chamber. An alternating electric field was applied between the two slides by using a voltage program of 3 V, 5 Hz for 120 min. The solution was collected and stored at 4 °C.

Confocal laser scanning microscopy
A GUV suspension (10 µL, 130 µM lipid concentration) was added to a FluoroDish (World Precision Instruments, Hitchin, UK) with buffer (500 µL, 1x TAE, 500 mM NaCl, pH 8.1). The solution gently mixed. After adding component A (10 µL, 1 µM) to the dish, the solution was mixed thoroughly and left for 10 min to ensure membrane binding. Component B ∆C (10 µL, 1 µM) was then added following by mixing of the solution. The mixture was left for 5 min to allow the GUVs sink to the bottom of the dish. The FluorDish (World Precision Instruments, Hitchin, UK) was placed under the microscope and GUVs were located by visualization using a 96x optical zoom. The sample was then viewed through the brightfield and at 570 nm for Cy3 A and 670 nm for Cy5 B and images were acquired.

Preparation of planar lipid bilayers on a glass slide and smFRET and single particle tracking
Planar lipid bilayers were formed on glow discharged glass slides provided by Oxford NanoImaging (Oxford, UK). SUVs composed of DPhPC in buffer A (15 µL, 1 mM) were placed onto the support and left for 15 min. Some solution (~ 5 µL) was then supplanted with H2O and left for 1-2 min. This was repeated 3x. After the 3 rd wash with H2O, the solution was washed with buffer A. Slides were used within 1 h and topped up with buffer A as necessary. smFRET and single particle tracking was performed using a NanoImager S (Oxford NanoImaging, Oxford, UK) by Jon Shewring from Oxford NanoImaging. Structures were added (1 µL, 1 nM in buffer A) to planar lipid bilayers composed of DPhPC on glass slides.

Linear dichroism
Solution-phase flow linear dichroism (LD) spectroscopy was performed on a Jasco-810 spectropolarimeter (Kromatec Ltd, Great Dunmow, UK) using a photo elastic modulator 1/2 wave plate. A micro-volume quartz Couette flow cell with ~0.5 mm annular gap and quartz capillaries were used (all from Kromatec Ltd, Great Dunmow, UK). Molecular alignment was achieved by applying the constant flow of the sample solution between two coaxial cylinders, a stationary quartz rod and a rotating cylindrical capillary. LD spectra were acquired with laminar flow obtained by maintaining the rotation speed at 3000 rpm and processed by subtracting non-rotating baseline spectra. DNA nanopores were assayed at 1.4 µM and SUVs composed of POPC at 500 µM lipid concentration.

Simulation preparation
DNA nanopore A•B, and component A were recreated in caDNAno, then converted to all atom models in python. 5 The poly-thymine linker regions at the pore termini were then constructed using the MolSoft ICM software suite. 6 TEG-Cholesterol lipid anchors were parametrized using cgenff 7 and attached using pyMol. 8 CHARMM36 compatible topology files were then generated using psfgen. 9 Initial structures of A•B and A were minimized in a vacuum for 10,000 steps (2 fs), then simulated for 100,000 steps (2 ns) using an elastic restraint network 10 derived from the ENRG webserver. 11 DNA nanopore A•B and component A were simulated in 1 M KCl and TIP3 water prepared in VMD. 12 Nanopore A•B was simulated in 13 x 11 x 15 nm box totaling 437k atoms and component A was simulated in a box of 16 x 14 x 19 nm totaling 6.5k atoms. A 1 ns NpT equilibration was run to equilibrate box size and pressure before a 50 ns NvT equilibration to further relax the DNA nanostructures. Production simulations were then run in in the NpT ensemble.
For simulations of membrane tethered component A and membrane-inserted A•B, VMD was used to generated membranes and orient the DNA nanostructures while maintaining favorable cholesterol oritentations. 13 The orientation of each structure was informed by experimental data derived from linear dichroism. The membrane-spanning nanopore A•B was simulated in a 12 x 12 x 12 nm box of 1 M KCl, bisected by a bilayer composed of POPC lipids for a total of 141k atoms. The membrane-tethered component A was simulated in a 15 x 15 x 16 nm box in the same conditions totaling 303k atoms. While the fixed atom restraints were placed on all atoms except those of the lipid tails, which were then thermally equilibrated over 0.5 ns of dynamics in the NvT ensemble as the temperature was increased to 301 K. 14 Fixed atom restraints were replaced with harmonic restraints, with a spring constant of 1 kcal/mol/Å 2 , on the heavy atoms of the DNA phosphate backbones. Simulation box size and pressure were equilibrated in the NpT ensemble for 3.5 ns, with harmonic restraints being lowered by 0.5 kcal/mol/Å 2 every 0.5 ns. Unrestrained dynamics in the NvT ensemble allowed the system to fully equilibrate, and production simulations were performed in the NpT ensemble.
Production simulations were performed at 301K and 1.013 bar pressure, maintained with the Langevin thermostat 15 and the Nosé-Hoover Langevin piston method. 16 Simulations were performed in NAMD, 17 a smooth switching algorithm 18 with a switch distance of 8 Å, a cut off of 10 Å and a pair list distance of 12 Å was implemented for van der Waals interactions. A 2 fs time step was used and hydrogen bond lengths were constrained using the SETTLE and SHAKE algorithms. 19 Particle Mesh Ewald electrostatics were computed over a cubic grid with a 1.0 Å spacing 20 and periodic boundary conditions. 21 Equilibration simulations were performed on a on a single GPU 1080Ti workstation and production runs were performed in parallel on 850 CPU cores of the UCL Grace HPC facility.

Simulation analysis
Analysis was performed using GROMACs 22 and VMD tools on the production simulations, after discarding the initial 10 ns, graphs were prepared using ggplot 23 and RStudio. 24

RMSF 10
gmx_covar and gmx_aneig were used to investigate the ten top quasi-harmonic modes 25 of root mean squared fluctuations (RMSF) of the DNA backbone heavy atoms, averaged perresidue, to interrogate structural dynamics of the DNA nanostructures while accounting for thermal noise and stochastic motion.

Clustering
gmx_cluster was used to prepare snapshots of the membrane spanning A•B DNA nanostructure trajectory. Clustering was performed with a cut-off of 0.35 nm using the gromos method. 26

Lumen analysis
Clustered coordinates were analyzed using HOLE, 27 with a channel-end radius of 0.8 nm and a sampling distance of 0.25 nm. To account for asymmetry of the DNA nanostructure, coordinates were then rotated and analyzed again.

Lipid analysis
gmx_gangle was used to measure the angle of phosphate and nitrogen atoms in the lipid head groups, split by lipid leaflet, compared to the bilayer normal, over the initial equilibration simulations. Production simulations were analyzed using gmx_rms, and the VMD plugins density_profile_too 28 and MEMBPLUGIM 29 to determine lipid RMSF, average lipid density and area-per-lipid, respectively.

Nanopore current recordings
Single-channel current measurements were achieved using an integrated chip-based, parallel bilayer recording setup (Orbit Mini, Nanion Technologies, Munich, Germany) with multielectrode-cavity-array (MECA) chips (IONERA, Freiburg, Germany). Bilayers were formed from DPhPC (10 mg/mL in octane). The electrolyte solution was 1 M KCl and 10 mM HEPES, pH 7.4. To achieve pore insertions, a 2:1 mixture of nanopore A•B and 0.5% OPOE (n-octyloligooxyethylene, in 1 M KCl, 10 mM HEPES, pH 7.4) was added to the cis side of the bilayer. Successful insertion was observed by detecting current steps. For triggered assembly, membranes were preincubated with AL A and BL B and 0.5% (v/v) OPOE. After 10 minutes, key strands were added and successful insertions were observed by detecting current steps. The current traces were not Bessel-filtered and were acquired at 10 kHz using Element Data Recorder software (Element s.r.l., Cesena, Italy). Single-channel analysis was performed using Clampfit software (Molecular Devices, Sunnyvale, CA, USA). were mixed with a 1% solution (v/v) of Triton X-100 (10 µL) to lyse all vesicles to identify maximum Ca 2+ influx. Ca 2+ influx was monitored as the ratio of the change in emission at each excitation wavelength as a ratio of 340/380 nm. The maximum 340/380 nm ratio following addition of Triton X-100 was used to normalize all traces.

Table S1. Names, modifications and sequences of DNA oligonucleotides used for folding A, B, A•B and variants.
Cy3 X = Cy3 fluorophore; Cy5 X = Cy5 fluorophore; Alexa488 X = Alexa488 fluorophore; Alexa647 X = Alexa647 fluorophore             Figure S12. FRET control between assembly-locked components at the membrane surface. (a) Fluorescence spectra of the assembly locked components AL A and BL B to confirm pore assembly. The control also serves to assess the effect of Cy3-Cy5 proximity on the membrane surface. (b) Normalized bar graph comparing the FRET signal observed from a 1:1 stoichiometric ratio of assembly locked components, AL A and BL B , on the membrane surface to FRET signal observed from the direct assembly of A+B in a 1:1 stoichiometric ratio. Each trace is an average of three independent repeats, the standard deviation shown as the error bars in the bar chart.
Figure S13. smFRET and single particle tracking confirm A+B assembly at the membrane surface. Single-molecule FRET (smFRET) and single particle tracking images confirm pore assembly from the monomer component A and B (top) in comparison to a control pre-mixed A•B (bottom). Bright green spots indicate the formation of A•B in the smFRET images (e.g., grey boxes). A carries a Cy3 fluorophore at the 5' end of strand A1 and B carries a Cy5 fluorophore at the 5' end of strand B1. Pore assembly from A and B was monitored on the supported lipid bilayer membrane at 62.5 pM. Single particle tracking experiments were monitored for 500 and 1000 s for A+B and pre-mixed A•B, respectively. Table S3. Summarized FRET efficiency (E) data for pore assembly derived from the FRET pore assembly binding titrations and pore assembly kinetics. The pre-annealed control for each of the three conditions was derived from the FRET binding data and was used as a benchmark. Averages and standard deviations were calculated from at least three independent repeats.