A reversibly gated protein-transporting membrane channel made of DNA

Design of a large and gated DNA channel

Before designing our large and gated channel (LGC) (Fig. 1a–c) we examined whether previous nanoarchitecture principles could be applied. In the previous DNA nanopores^{43,44,48,49,50,51,59}, duplex helices are aligned parallel in lattice fashion, as dictated by the widely used caDNAno design software⁶⁴. In those cases, the duplexes were routed “vertically”, i.e. at 90 degrees relative to the membrane plane (Supplementary Fig. 1a). This arrangement places the majority of DNA into an extramembrane cap region with multi-duplex layer-thick channel walls while restricting the amount of DNA available for forming a wide membrane-embedded pore (Supplementary Fig. 1a). The small lateral footprint of the pore also limits the number of attachment points for lipid anchors required for efficient pore insertion into bilayer membranes. To overcome these restrictions, we opted to route the component helices ‘horizontally’ to the membrane by using the free-form software Tiamat⁶⁵ (Supplementary Fig. 1b, c).

Using the horizontal routing, we rationally designed the channel featuring a square-channel lumen with 20.4 nm × 20.4 nm opening cross-section into a single-duplex layer DNA origami plate of 70 nm × 70 nm external dimensions (Fig. 1a–c). The four-duplex deep channel spans the bilayer membrane while the large base plate sits on top of the bilayer (Fig. 1a). By drastically expanding the footprint of the extramembrane cap, total 160 hydrophobic cholesterol anchors can be accommodated for efficient membrane insertion, out of which, 64 cholesterol anchors were used in the current study (Fig. 1a and Supplementary Fig. 2). As a further advantage of the design, the channel with 416 nm² opening cross-section can be reversibly closed and opened with a square lid composed of horizontally routed DNA duplexes to yield the closed channel LGC-C (Fig. 1a) and the open version LGC-O (Fig. 1b). To achieve this dynamic change, the lid is attached at one side to the channel plate by flexible hinges (Supplementary Fig. 3). The other lid side carries two single-stranded half locks which can hybridize with the complementary half locks at the base plate to form complete duplex locks (Fig. 1d). To open the lock and lid of the closed channel (LGC-C), a pair of single-stranded DNA keys dissociate the locks to form the open channel (LGC-O) (Fig. 1d and Supplementary Fig. 4). The opened lid can be switched back to close by a single-stranded reverse key pair (Fig. 1d). This externally controlled mechanism is expected to reversibly switch the lid-gated channel between an open and closed state to regulate the flow of large molecular cargo.

We first assembled non-lid version LGC-N and the nanostructures LGC-C and LGC-O carrying the lid in two states, by annealing the scaffold DNA with staple oligonucleotides (Supplementary Data 1). The structures did not yet contain the cholesterol anchors. OxDNA simulations⁶⁶ showed the desirable formation of LGC-N, LGC-C, and LGC-O in solution (Supplementary Fig. 6 and Supplementary Video 1) and gel electrophoresis confirmed that single assembly products had formed (Supplementary Fig. 5). Atomic force microscopy (AFM) (Fig. 1e–g and Supplementary Fig. 6) and transmission electron microscopy (TEM) (Supplementary Fig. 7) established the expected dimensions. For example, the AFM-derived external average side-lengths of LGC-N was 78.0 ± 2.9 nm for the square plate and 22.7 ± 5.1 nm (n = 65) (Fig. 1e and Supplementary Fig. 6) for the square opening, close to the expected values of 70 and 20.4 nm, respectively. The elevations at the nanostructure center stem from the four-duplex-high channel walls extending from LGC bound top-down to the mica substrate. By comparison, the closed-lid LGC-C featured no central opening, and the side-lengths were 77.4 ± 2.6 nm for the plate and 22.7 ± 1.9 nm for the channel wall (n = 11) (Fig. 1f and Supplementary Fig. 6). Similarly, open-lid LGC-O appeared partly closed as the lid can obstruct the channel opening and the dimensions were 78.4 ± 3.5 nm and 21.8 ± 2.5 nm (n = 13) (Fig. 1g and Supplementary Fig. 6). The slightly larger-than-nominal dimensions of all LGC variants are likely due to the flattening of the negatively charged DNA plate on mica surface in presence of Mg²⁺, compression by the AFM tip, or both factors⁶⁷. The channels were lipid-tagged by incubating with cholesterol-modified DNA oligonucleotides that bound to designed sites at the bottom side of the large membrane cap (Fig. 1a and Supplementary Fig. 2).

Nanopore interaction and insertion into lipid bilayers

Following successful formation, we tested whether the cholesterol-modified channel can bind to and insert into lipid bilayers. To probe for membrane binding, channel variant LGC-N was added to small unilamellar vesicles (SUVs) and analyzed by agarose gel electrophoresis (AGE). The AGE tests were performed under conditions with and without Mg²⁺, and we found that an Mg²⁺ free buffer condition is essential to prevent non-specific adsorption to chelating lipid head-groups while still ensuring the stability of the LGC variants (Supplementary Note 1 and Supplementary Fig. 8). The stability of LGC in Mg²⁺ free buffer was ascertained by AFM imaging (Supplementary Fig. 9). Increasing SUV concentrations led to concomitant gel electrophoretic upshift (Fig. 2a), suggesting that pores can bind efficiently to lipid membranes of slowly migrating vesicles⁶⁸. No gel-upshift was observed when LGC lacked cholesterol, underscoring its role for membrane binding. Cholesterol-mediated binding was also confirmed with direct visualization by TEM imaging (Fig. 2b, blue arrows and Supplementary Fig. 10). The extent of membrane binding was probed by incubating Cy3-labeled LGC-N with giant unilamellar vesicles (GUVs) and examination with fluorescence microscopy. Co-localization of the cholesterol labeled pore with the vesicle perimeter indicated successful membrane binding (Fig. 2c, d and Supplementary Fig. 11).

To explore whether LGC-N punctures lipid bilayers, we tested the influx of membrane-impermeable Atto633 dye into the interior of GUVs^48,50,51. GUVs were placed in a solution of Atto633, and fluorescence microscopy tracked any changes in the fluorescence content of GUVs after adding LGC-N. The channel successfully inserted into, and punctured membranes as indicated by Atto633 signals that increased within GUVs over the full incubation time (Fig. 2d, top). A total of 59% of GUVs showed dye influx in case of LGC-N with cholesterol (LGC-N + Chol) (Fig. 2e and Supplementary Videos 2, 3), which compares to solely 3.5% of GUVs incubated with LGC-N lacking cholesterol (LGC-N-Chol) (Fig. 2e, bottom); membrane insertion via cholesterol is key for puncturing the bilayer for cargo transport (see Methods for details on how data were analyzed to identify influx). We further validated the membrane-spanning of our large and gated channel with a fluorescence resonance energy transfer (FRET) assay (Supplementary Fig. 12), which detects the interaction of a donor dye at the inserted channel with acceptor dyes in the GUV lumen.

The observed 59% influx via the LGC pores is significantly higher than the 8–11% increase detected in previous reports^43,48. Moreover, the LGC mediated influx was completed within 1 h after channel addition, which is considerably faster than the 5–8 h completion time required in the case of previous DNA nanopores^43,48. The differences in influx rates may be a result of several interdependent phenomena that lead to DNA nanopore-mediated transport: (i) nanopore binding to the membrane, (ii) reorientation of the stem and puncturing of the lipid bilayer to form a channel, and (iii) the diffusion of the dye through the pore. The nanopore binding to the membrane is strongly governed by the number of cholesterol molecules, ionic concentration, and temperature. The reorientation is the rate-determining step and depends on the flexibility of cholesterol-bearing segments of the nanopore. The final step of dye diffusion is a function of the size of the dye molecules, medium viscosity, temperature, concentration gradient, and pore dimensions^43,48. We attribute the faster flux in the case of LGC-N compared to previous pores due to the placement of a large number of cholesterol anchors to accelerate membrane binding (step i). The flexibility of the cholesterol-bearing one-layer origami plate allows for rapid reorientation of the pores. Faster insertion kinetics were observed with LGC-N compared to previous origami pores that always placed the cholesterol molecules in a rigid framework, thereby slowing this reorientation step. Interestingly, the dye diffusion rate in LGC-N was approximately five times slower compared to the theoretically calculated time required for diffusion (Supplementary Fig. 13 and Supplementary Note 2). This is likely due to the large pore diameter slowing the rate-limiting reorientation step.

The membrane-spanning nature of LGC-N was confirmed by single-channel current recordings (Fig. 2f–h). Individual channels were inserted into a planar DPhPC lipid bilayer that separates two chambers filled with electrolyte (1 M KCl, 10 mM HEPES pH 7.6). To induce ion flow across an inserted channel, a transmembrane potential was applied. An ensuing steady current of 461 pA (Fig. 2f, +80 mV) indicated membrane insertion. LGC-N’s average conductance values distribute into multiple conductance states (Fig. 2g and Supplementary Fig. 16), in line with previous measurements of large diameter DNA nanopores^49,51. Two peaks are seen in the conductance histogram, a smaller peak with an average of 2.51 ± 0.35 nS (n = 6, ±error of Gaussian fit) and a larger peak with an average of 5.92 ± 0.43 nS (n = 16, ±error of Gaussian fit) (Fig. 2g). The linear relationship between current magnitude and voltage is expected for a channel lumen with vertical symmetry (Fig. 2h). Control experiments using fluorescence microscopy (Supplementary Fig. 14) and electrical current readout (Supplementary Fig. 15) of a flat origami plate without the stem established that the translocation of dye is indeed through the central pore and not through the toroidal pores known to form between the sides of DNA nanopores and the lipid membrane.

Reversibly ligand-gated lid-controlled transport of small-molecule cargo

Unlike constitutively open pores, biological ion channels usually open solely upon specific stimuli and then close by passive ligand dissociation. Attempts^48,53 have not been able to functionally replicate and transcend biology by actively controlling pore opening and closing via external triggers. Our large gated channel was designed to achieve defined transport control via a toehold-mediated strand displacement reaction^69,70 between lock and key (Fig. 1d and Supplementary Fig. 4). We tested the reversible opening and closing of the lid using FRET. A Cy3 donor and Cy5 acceptor dye were placed in the lid and plate, respectively (Fig. 3a, D and A). Due to spatial proximity between the dyes, FRET occurred in the closed state (Closed or LGC-C) as reflected by the low donor fluorescence at ({lambda }_{{max }}^{{Cy}3}=564,{{rm{nm}}}) and high acceptor fluorescence at ({lambda }_{{max }}^{{Cy}5}=670,{{rm{nm}}}) (Fig. 3b, solid pink line). Opening the lid by adding a key (Dyn. Open) increased the distance between the reporter dyes and resulted in a low FRET, a higher donor emission, and a drop in the acceptor emission (Fig. 3b, blue line). Closing the lid back from its open state using a reverse key (Dyn. Closed) restored FRET, lowered donor fluorescence, and increased acceptor fluorescence (Fig. 3b, dashed pink line). The key-controlled switch was sequence-specific as confirmed by a mismatch key (Supplementary Fig. 17). Analysis of the kinetic FRET signal revealed that lid opening followed second-order kinetics⁷¹ at a rate constant of 1940 ± 50 M⁻¹ s⁻¹ until transport completion after 2 h (Supplementary Fig. 18 and Supplementary Note 3). Cyclic opening (Fig. 3c, blue dots) and closing (Fig. 3c, pink dots) of the lid was also demonstrated by comparing the FRET efficiency at each stage.

After establishing reversible gating, we utilized the GUV-dye influx assay to demonstrate controlled transport through the channel. GUVs were immersed in a solution of Atto633, and influx was probed via fluorescence microscopy. In control analysis, cholesterol-modified LGC-C bound to the vesicle membranes whereas the non-cholesterol version did not show binding, as indicated by the presence or absence of fluorescent rings around the GUV perimeter (Supplementary Fig. 19). Probing of the fluorescence intensity within GUVs established that membrane-bound LGC-C did not lead to dye flux into the vesicles (Fig. 4a, b, top; Supplementary Video 4). This implies that the lid completely blocked transport through the LGC channel. Upon addition of keys, however, dye fluxed inside GUVs demonstrating that the opened-lid channel is transport active (Fig. 4a, b middle and Supplementary Video 5). To confirm that the channel can be shut back to cease transport function, a “dynamically closed pore” was obtained by first, dynamically opening the closed LGC by addition of key, followed by addition of the reverse key. After dynamically closing the membrane-inserted LGC in the absence of Atto633, the dye was added to the GUVs. No dye influx was seen in this case (Fig. 4a, b, bottom and Supplementary Video 6), which confirms that the dynamically closed LGC blocks transport. Negative control experiments with LGC-C without cholesterol modification (Supplementary Video 7) and the mismatch keys (Supplementary Fig. 20) did not show similar Atto633 dye influx. Moreover, a two-color confocal microscopy experiment was performed (Supplementary Fig. 21). Firstly, red dye (Atto633) was allowed to influx into the vesicle through the open lid pore LGC-O. Closure of the lid did not allow a second green dye (Atto 488) to the influx, demonstrating successful closure of the opened lid. These data clearly confirm that the lid can be reversibly opened and closed with external triggers. Although previously one-way or partial opening⁵³ and closing⁴³ has been shown in separate reports, dynamic and reversible gating has not been shown in a synthetic nanopore before.

Single-channel current recordings probed the characteristics of cholesterol-modified closed lid (LGC-C) and open-lid (LGO-O) LGCs (Fig. 4c–h), and the dynamic transitions between the two. As would be expected from the steric blockade by the closed lid, LGC-C featured a far smaller current and corresponding conductance at 0.64 ± 0.10 nS (n = 12, ±error of Gaussian fit) (Fig. 4c, e) than had been obtained with no lid LGC at 5.92 ± 0.43 nS (n = 15, ±error of Gaussian fit) (Fig. 2f, g). The presence of the lid did not affect linear voltage-current dependence (Fig. 4d). The small residual current of LGC-C might stem from ion leakage either through the periphery of the pore (at the DNA duplex/lipid interface)^51,52,53 or through the DNA duplexes of the lid, or a combination of both. By comparison, the conductance of LGO-O at 2.31 ± 0.18 nS (n = 17, ±error of Gaussian fit) (Fig. 4f, h) is almost four times the value of LGC-C (Fig. 4e), reflecting that the open lid allows more ion transport. However, it is still more than half the value of no-lid LGC in agreement with a residual blockade by an open lid (Fig. 2f, g). Control experiments were performed using a DNA plate construct without a pore-forming cap region (Supplementary Fig. 15). These constructs did not produce pore-forming currents, highlighting the importance of the cap region for lipid bilayer penetration.

The single-channel analysis provided further insight into the dynamic nature of the LGC lid. Current traces of the single LGC-O channels had more high-frequency fluctuations (Fig. 4f and Supplementary Figs. 22b, 23) than the less noisy trace of LGC-C (Fig. 4c and Supplementary Figs. 22a, 23) at all voltages from −100 to +100 mV (Supplementary Fig. 22). The current fluctuations likely reflect the dynamic movements of the lid to and from the channel base in LGC-O compared to the static lid in LGC-C. Indeed, electrophoretically driving the lid of LGC-O to its base plate via a negative potential led to an additional current level at −37.8 ± 15.6 pA (−60 mV, ±STD) next to the main conductance peak at −96.7 ± 7.8 pA (−60 mV, ±STD) (Supplementary Fig. 22), an increase in noise at negative potentials shown by power spectrum noise analysis (Supplementary Fig. 24), and an increase in event frequency at negative potentials (Supplementary Fig. 25). By contrast, moving the lid away from the plate via positive potentials only led to a single, main peak at 78.5 ± 5.8 pA (60 mV, ±STD) (Supplementary Fig. 22) and resulted in less noise (Supplementary Fig. 24) and fewer blocking events (Supplementary Fig. 25). The voltage-dependent noise strikingly reveals nanomechanical-dynamic changes of a DNA structure at the single-molecule level.

Single-channel analysis revealed insight into the closing and opening mechanism. LGC-O could be closed by adding a reverse key, as demonstrated by the transition from an open state current at 97.4 ± 5.2 pA (10 mV, ±STD) to the lower-amplitude closed state at 17.7 ± 2.9 pA (10 mV, ±STD) (Fig. 4h, inset and Supplementary Fig. 26). This transition occurred in a stepwise fashion, likely because the two lid locks are closed one after the other leading to an intermediate state of one closed and one open lock (Fig. 1d). Similarly, LGC-C was opened by adding key as indicated by a switch from the closed state at an amplitude of 8.7 ± 2.7 pA (−10mV, ±STD) to the open-state at amplitude 146.9 ± 2.8 pA (−10 mV, ±STD) (Fig. 4e, inset and Supplementary Fig. 26). This transition occurred also in a stepwise fashion. Our data reveal the opening and closing mechanism of the LGC channel in unprecedented detail and indicates ways to fine-tune the opening by varying the lock number.

Transport of folded proteins across the nanopore

We finally exemplified the power of large and gated channels by regulating the flux of folded proteins across membranes via defined lid opening and closing. GUVs were immersed in a solution of green fluorescent protein (GFP, hydrodynamic diameter = 5.6 nm⁷²). Cholesterol-modified channels bound to GUV membranes whereas unmodified versions did not bind (Supplementary Figs. 27, 28 and Supplementary Videos 8, 12). When influx was monitored by determining fluorescence intensity in GUVs, control channel LGC-N with cholesterol without lid led to transport across membranes (Supplementary Fig. 27 and Supplementary Video 8). By contrast, cholesterol-modified closed-lid LGC-C blocked protein flux (Fig. 5a–c, top and Supplementary Video 9). However, adding the key opened up the transport function (Fig. 5a–c, middle and Supplementary Video 10). Again, in the dynamically closed pore, no protein influx was observed (Fig. 5a–c, bottom and Supplementary Video 11). Negative control experiments with 500 kDa FITC-dextran featuring a hydrodynamic diameter of 31.8 nm⁷³ established that channel transport did not occur due to size-exclusion (Supplementary Fig. 29); FITC-dextran also ruled out membrane rupturing. Another control experiment using pre-encapsulated cargo in GUVs demonstrated that LGC can release nucleic acid cargo such as dye-labeled DNA (Supplementary Fig. 30) which can prove to be valuable for vaccine delivery. The successful data on the precisely timed transport of folded proteins and small organic dyes with equal ease (Fig. 4a, b and Supplementary Fig. 31) is beyond the scope of biological and any previously engineered membrane channels.

The transport of folded protein across the membrane-nanopore lumen was examined with single-channel current recordings (Fig. 5d–g). As a model protein, trypsin (hydrodynamic diameter of 4.0 nm⁷⁴) and net positive charge (pI 10.1, pH 7.6) was used. No translocation events occurred upon the addition of 6.6 μM trypsin to the cis chamber of membrane inserted LGC-C (Supplementary Fig. 32) as shown by relatively steady current flow (Supplementary Fig. 32a, b). By contrast, with LGC-N channels increasing concentrations of trypsin led to blockades (Fig. 5d–g and Supplementary Fig. 33), which clustered into two event types (Fig. 5d–g). When analyzed by their relative percentage blocking amplitude, A/I_O (amplitude of block/ open current amplitude), and dwell time, τ_off, (Fig. 5f, g), events clustered at 1.88 ± 1.5% (Type I) and at 10.8 ± 4.8% A (Type II). We suggest that Type I events occur due to the brief interaction of the protein with the nanopore at the lumen opening (Fig. 5d), while in Type II events proteins fully translocate through the nanopore (Fig. 5e). In order for translocation events to be detected, intermittent interactions of the positively charged trypsin to the negatively charged pore wall must occur^50,75. Voltage-dependent analysis (Supplementary Fig. 34) revealed that Type II blocking events have longer dwell times at lower voltages (1.65 ± 0.10 ms at −20 mV) than higher voltages (0.28 ± 0.04 ms at −50 mV) in line with the expected faster translocation of trypsin through the pore at higher voltages.

The transport of folded GFP through LGC-N was also studied. GFP is larger than trypsin (hydrodynamic diameter of 5.6 nm⁷²) and has a lower isoelectric point (pI 5.8) and will therefore have a net negative charge in standard electrophysiological buffers of pH 7.6. Little interaction of GFP with the nanopore is expected at neutral pH as intermittent electrostatic binding of a positively charged protein with the negatively charged DNA pore wall is thought to be necessary to resolve translocation events^50,75. Indeed, when 310 nM GFP was added at neutral pH to LGC-N no obvious translocation events could be seen (Supplementary Fig. 35a–c). However, when the pH of the electrophysiological buffer was dropped below the pI of GFP to pH 4.5 clear blockade events could be detected (Supplementary Fig. 35d–f) suggesting electrostatic interaction between GFP and the pore wall. GFP blockade events clustered into two types of events when each event was plotted in relation to its blocking amplitude and dwell time (Supplementary Fig. 36), similar to the blockades caused by trypsin. Type I GFP events clustered at 3.8 ± 3.7% and Type II events clustered at 26.1 ± 7.3%. The latter events are interpreted to fully translocate through the nanopore. The bigger blocking amplitude of GFP is expected for the protein’s larger size (hydrodynamic diameter = 5.6 nm) compared to trypsin (hydrodynamic diameter = 4.0 nm), in line with the previous reports⁵⁰.

Advantages and limitations of pore design

In this work, we set out to construct a reversibly gated DNA nanopore with a luminal diameter large enough to accommodate proteins for transport assays. Whilst achieving these goals, characterization of the pore has highlighted some key advantages of the DNA nanopore design over existing artificial nanopore designs, as well as revealing some limitations of the design. Discussing the advantages and limitations here could help future researchers improve the design and sensing abilities of DNA nanopores, accelerating the advancement of research in the field of DNA nanostructures.

One advantage of the pore design in this study lies in the nanomechanical control of pore activity, which enables controlled transmembrane transport of an array of cargoes ranging from small-molecule dyes to large folded protein. Fluorophore flux studies have shown high fidelity of dye transport through open lid nanopores, with no such transport seen in closed lid nanopores (Fig. 4a, b). However, single-channel current recordings revealed some inherent ionic leak through the closed lid nanopore (Fig. 4c–f), with an average conductance of 0.64 ± 0.10 nS, highlighting the high sensitivity of these single-molecule experiments. Such leaks have been shown previously with a closed lid DNA nanopore⁵³. Leak currents likely stem from toroidal pore formation of lipid bilayer spanning DNA duplexes^51,52, as well as ionic flow through the nanopore lid⁷⁶ (Supplementary Fig. 37). Dye transport studies (Fig. 4a, b) show that the closed lid nanopore is not permeable to the larger dyes (Atto633 diameter = 1.3 nm), suggesting the leak of ions only (chloride ion Stoke’s radius = 0.1 nm). Major groove binders⁷⁷ or hydrophobic coating⁷⁸ have been proposed as potential strategies to reduce leak currents of DNA nanostructures.

In this study, open lid nanopores have been successfully designed to controllably accommodate the folded proteins trypsin and GFP for fluorophore release flux studies (Fig. 5a–c) and single-molecule electrophysiological transport studies (Fig. 5d–g) which is a major advantage over current small-diameter protein-based nanopores. However, single-channel current recordings have revealed smaller experimental conductances of no lid nanopores when compared to designed pore diameters. No lid nanopores have an average conductance of 5.92 ± 0.43 nS, with protein transport studies of both trypsin and GFP attributing average current blocks of 11 and 26% respectively. Taking into consideration these measurements, it is therefore likely that the pore does not have the designed pore size of 20 nm, but is closer to 5–8 nm when measured in a planar lipid bilayer (based on estimated protein volumes). Whilst there is an apparent reduction in pore diameter, this is still of a size large enough to accommodate folded protein. Previous research has shown buckling and voltage-dependent conformational changes of DNA origami structures^79,80 and compression of lipid spanning helices of the lumen⁵⁰ that may account for low experimental conductances. Electro-osmotic effects due to the charged DNA backbone^50,81 as well as increased access resistance in nanopores with low thickness-to-diameter ratios and charged surfaces⁸² may also have integral roles in reducing experimental conductance. Future molecular dynamic studies^52,76,83 could help explain the conductance discrepancies in large diameter DNA nanostructures.

Another advantage of our nanopore design is the horizontal routing approach that allows high design flexibility, which means making small adjustments to tailor the pore to a required function is not a labor-intensive process. For example, tailoring pore diameters for size-dependent analysis of folded proteins, or adjusting lid opening/closing for real-time analysis of protein folding interactions. Our protein transport studies in planar lipid membranes have shown that relatively high concentrations of trypsin (between 66 nM and 6.6 µM) are required for adequate analysis of translocation events (Supplementary Fig. 33). Protein translocation under voltage is a rapid process, with the majority of proteins passing through the nanopore with dwell times <20 µs, too fast to detect with electrophysiological equipment. Therefore, protein translocation can only be detected in this case when trypsin translocation is slowed, such as via interaction with the negatively charged DNA of the pore lumen. This interaction has been noted previously with trypsin translocation through a DNA nanopore⁴⁹, and is in concordance with simulation data of trypsin translocation through a DNA nanopore⁷⁵. Only trypsin that has adequate interaction with the DNA pore lumen will therefore be detected, accounting for the larger concentration of protein required in these translocation experiments. Future work could look at adding protein-specific tags inside the pore lumen to increase the specificity of protein interactions, whilst also increasing protein dwell times.