The Ion Permeability of DNA Nanotube Channels

ABSTRACT

Techniques from structural DNA nanotechnology make it possible to assemble complex 3‐dimensional nanostructures with virtually arbitrary control over their sizes, shapes, and features at length scales of 3–100 nm, providing a flexible means for constructing nanoscale devices and machines. Here, we assemble micrometer‐long DNA nanotubes and assess their performance as pipes for controlled ion transport. DNA nanotubes grow via assembly of DNA tiles from a seed pore, a 12‐helix DNA origami cylinder functionalized with 18 cholesterol anchors, to generate a DNA nanotube channel. The central channel of a nanotube can be obstructed via Watson–Crick hybridization of a channel cap, a second DNA origami structure, clamped to its end. The single‐channel electrical recordings show that both nanotube seed pores and nanotube channels display ohmic ion conductance consistent with their central channels’ diameters. Binding of the channel cap reduces the conductances of both DNA nanotube channels and seed pores, indicating that ion transport can be modulated and measuring the resulting residual transport through DNA‐wall gaps or DNA–lipid interfacial pathways. Our findings help design self‐assembling nanofluidic devices and circuits in which transport may be regulated via dynamic biomolecular interactions, since these channels could be constructed into branched topologies or routed between specific molecular terminals.

Self‐assembled micrometer‐long DNA nanotube channels insert into lipid bilayers and exhibit reduced conductance upon capping, consistent with ion transport through both the channel end and residual wall or DNA–lipid interfacial pathways.

1. Introduction

Transport between membrane‐bound compartments is a biologically controlled process. Nanoscale channels regulate ion or small‐molecule flux across membranes [1, 2, 3] and mediate transport of these species between cells [4]. Such cell‐to‐cell communication is critical for growth and development as well as for the transmission of pathogens such as the human immunodeficiency virus (HIV) [5], herpes simplex virus (HSV) [6], and prions [7]. In addition to being crucial for understanding the basic principles of nanoscale transport, biological nanopores and other artificially created nanochannels are also significant because control over this transport can have multiple transformative applications in biotechnology and medicine [8, 9, 10]. For instance, biological nanopores find extensive application in third‐generation DNA sequencing [11, 12], protein and enzyme detection [13, 14, 15, 16, 17, 18, 19, 20], diagnostics [21], bioelectronics [22, 23], and synthetic tissues from synthetic compartments in which pores mediate transport [24, 25]. Additionally, chemical and biological alteration of nanopores has also increased the selectivity of transport and its ability to modulate conductance through pores [26, 27, 28]. While engineered nanopores can mediate transport between abutting membranes, longer nanochannels [29] could mediate transport between more distant compartments or end points. Such longer nanochannels, like the tunneling nanotubes between cells, are frequently observed in living systems [30, 31]. Additionally, longer nanochannels could mediate rapid transport of ions or molecules over longer distances by confining them within one‐dimensional channels. The ability to synthesize structures to control ion transport over longer distance could therefore open up new avenues for engineering transport between synthetic devices and living cells, facilitating intracellular communication, and the creation of nanofluidic circuits.

The nanoscale size of such channels suggests that self‐assembly is a practical means for their fabrication. In this context, self‐assembled biomolecular channels might also be functionalized or incorporated into cell culture or organized in 3‐dimensional environments [32]. While self‐assembled membrane proteins have been adopted for use as membrane pores in synthetic devices, techniques from DNA nanotechnology make it straightforward to construct nanostructures that emulate many of the structural and functional characteristics of naturally occurring membrane channels. Such DNA channels can enter into lipid bilayers and transport ion currents ranging from a few hundred to a few thousand picoamperes across these bilayers [33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]. DNA channels can act as biosensors [33, 36, 38], cytotoxic agents [44], transporting agents [45], and drug delivery vehicles [39, 46, 47, 48, 49], or organize enzymes to direct a reaction cascade [50]. In contrast to biological channels, DNA nanopores may be designed with precise dimensions and can be functionalized at specific sites using DNA conjugates [33, 34, 35, 36, 37, 38]. A broad array of micrometer‐long channels whose dimensions and structure can be precisely designed and have also been assembled using DNA [51]. Additionally, by combining solid‐state devices [52] or nanocapillaries [53, 54], DNA origami channels were utilized to create hybrid devices for sensing applications.

Here we design, self‐assemble, and characterize synthetic DNA nanochannels with sub‐10 nm diameters and lengths on the order of 2 µm, and study their capacity to serve as ion‐transporting channels. The designed structures can present specific chemistry at either end to allow for ion transport, for example, the insertion of the channel end into a membrane. They have a radius and exterior structure that can be precisely tuned and can grow to connect two end points connected by molecular labels [55], suggesting the capability of these structures to form nanochannels with precisely tuned properties at desired locations. In addition to their ability to self‐heal [56], these structures may also exhibit prolonged stability in an in vitro environment, such as cell culture medium. To serve as conduits for transport, such micrometer‐long nanochannels must mediate both cross‐membrane transport and along‐channel transport. Here, we study the flux of ions through these channels across a lipid bilayer and quantify the extent of ion transport through these channels versus ion loss into the bulk through the channel walls.

In our previous study, we reported this folded DNA origami seed/template (Figure 1a [51]; see Note S1 and Figures S1, and S2 for sequences) and its use to grow micrometer‐long DNA nanotubes via tile polymerization (Figure 1b [51]; see Note S2 and Figure S3 for sequences). Inspired by this ability of DNA to form micrometer‐long nanotubes, here we test whether this previously reported seed/nanotube system can function as an ion‐conducting transmembrane channel after insertion into a synthetic membrane via 18 hydrophobic cholesterol moieties attached to the DNA seed (Figures 1a,b). We then use a previously developed self‐assembled channel cap [57] to estimate the extent of conductance that persists when end‐accessible transport is restricted (Figure 1c; see Note S3 for sequences). The channel cap sharply constricts the channel by staples that directly connect the helices on opposing sides of its barrel, thereby modulating the channel conductance. Using this approach, we demonstrate that these micrometer‐long DNA nanotubes can form transmembrane channels in lipid bilayers with well‐defined conductances that are only weakly dependent on channel length. We found that the capped seeds and many capped seeded nanotubes (Figures 1c, d) exhibit conductances approximately half those of their open counterparts, suggesting that ions can enter through the channel end. At the same time, residual current can persist through DNA‐wall gaps and/or DNA–lipid interfacial pathways under an applied transmembrane potential. This observation is consistent with the structural features of DNA nanostructures, in which crossovers spaced two full turns apart can produce gaps of >1 nm because of electrostatic repulsion [58, 59]. In addition, toroidal‐pore‐like lipid rearrangements around DNA nanopores may also contribute to residual ionic conductance [60, 61]. This study extends the previously reported micrometer‐long DNA nanotube platform to electrophysiological measurements. It uses capped/uncapped conductance comparisons to assess whether ion transport is regulated by the channel end or persists via wall‐ or interface‐mediated pathways. Unlike earlier studies, which focused on DNA nanostructures shorter than 100 nanometers [19, 33, 34, 35, 36, 37, 38, 41], the present study demonstrates the possibility of pore formation in micrometer‐long structures. The pore and channel cap comparison also provides a practical assay for evaluating how nanotube modifications, such as wall coatings or lattice modifications, affect ion transport. Further, these micrometer‐long channels could potentially be used for sensing, by serving as long‐dwelling transport conduits [33, 36, 38], for constructing electric circuits [62], or for nanofluidic transport [63, 64].

FIGURE 1.

Schematic representations of a DNA origami seed channel, seeded nanotube channel, capped seed channel and capped nanotube channel. (a) The DNA origami seed consists of 72 ssDNA staple strands that fold a long M13mp18 ssDNA into a cylindrical shape. The DNA origami seed (including its adapters) has length 64.70 ± 2.12 nm (95% CI, N = 50 structures) with outer and inner diameters of 11 and 7.3 nm, respectively [29]. DNA hairpins on the outer region of the DNA origami seed are designed to direct the seed to cyclize in a specific orientation. The DNA seed is decorated with 18 cholesterol‐DNA conjugates. The cholesterol groups are desired to stabilize the structure within a lipid bilayer (For sequences see Figures S1, S2, and Notes S1, S4). A toroidal pore‐like configuration arises through the lipid rearrangement around the DNA seed is also shown in all the four cases (a–d); (b) Nanotubes are grown from two types of monomers (dark green and light green) with different sticky ends. The sticky ends hybridize to form a lattice which in turn cyclizes to form a nanotube which can bind to (or grow from) the nanotube seed (For sequences see Figure S3 and Note S2). A seeded DNA nanotube in a 1,2‐diphytanoyl‐sn‐glycero‐3‐phosphocholine (DPhPC) lipid bilayer is shown; (c) Hybridization of a DNA seed and a channel cap produces a channel cap‐DNA seed complex. A channel cap‐DNA seed in a synthetic lipid bilayer is shown (For sequences see Notes S3 and S4); and (d) The facet of a seeded tube can hybridize to a channel cap to produce a channel cap‐seeded tube complex. A channel cap‐seeded DNA tube is shown in a synthetic lipid bilayer.

2. Results and Discussion

2.1. DNA Origami Barrel/Nanotube Seed Channel Design

To characterize how micrometer‐long DNA nanostructures might mediate ion transport, we aimed to develop a system for assembling DNA nanostructure conduits such that they could mediate transport across a lipid bilayer. We initiated the design with a DNA origami seed structure we developed previously [51]. This structure is a scaffolded DNA origami [65] consisting of 12 DNA helices arranged in a cylinder (Figure 1a) and whose size has been measured previously using atomic force microscopy to be 63.1 ± 1.8 nm in length and 21.3 ± 1.4 nm in width [51]. The seed presents single‐stranded sticky ends that serve as templates for the assembly of DNA nanotubes that can grow many micrometers long (Figure 1b). The structures were labeled with fluorophores by binding DNA‐fluorophore (ATTO488 or ATTO647N) conjugates to regions of the M13mp18 scaffold not involved in seed formation (see Note S4, Labeling strand sequences) [55].

To generate a pore‐forming seed structure, we modified a subset of the seed's staples to include sequence domains that hybridize with DNA‐cholesterol conjugates, thereby incorporating 18 cholesterol anchors (see Note S1, 3ʹ nanotube seed_Chol.DNA strand sequences). This modification created domains capable of penetrating the membrane and forming a pore (Figure 1a) [33, 34, 35, 36, 37, 38, 39, 41, 42]. Transmission electron microscopic studies reveal the average length of the nanotube seeds with these modified staples was 64.70 ± 2.12 nm (95% CI, N = 50 structures) (see Materials and Methods), similar to the length of the unmodified structure, which is 63.1 ± 1.8 nm as previously reported from atomic force microscopic measurements (Figure 2a, bottom panel and Figure S4a,d) [51]. Transmission electron micrographs of nanotube seed channels with attached DNA‐cholesterol conjugates did not show discernible aggregation after assembly, which might be induced by the interaction between the cholesterol groups on the conjugates (Figure 2a, top left panel and Figure S4b,e). Additionally, they also remained dispersed in electrophysiology buffer (1 M KCl, 10 mM MgCl₂, 10 mM 4‐morpholineethanesulfonic acid (MES) monohydrate, pH 6.0) (Figure 2a, top right panel and Figure S4c,f), suggesting that single pores could be studied in electrophysiology experiments.

FIGURE 2.

Transmission electron microscopy (TEM) showing the complete assembly of DNA origami seeds, seeded nanotubes and the incorporation of seeds and seeded nanotubes into SUVs. Transmission electron micrographs of the DNA origami seed channels (a, top left panel) in TAEM buffer after the addition of cholesterol‐DNA conjugate strands and subsequent filter purification; (a, top right panel) in the presence of electrophysiology buffer (1 M KCl, 10 mM MgCl₂ buffered with 10 mM 4‐morpholineethanesulfonic acid (MES) monohydrate, pH 6.0). Scale bar:100 nm; (a, bottom panel) in TAEM buffer after filter purification (absence of cholesterol‐DNA conjugate strands). Scale bar: 100 and 20 nm; (b, top panel) Schematic representation showing the incorporation of several DNA origami seed structures into a single SUV; (b, bottom panel) Multiple nanotube seed channels insertion into SUVs. Scale bar: 20 nm. The average channel length of the DNA nanotube seed channel is 64.70 ± 2.12 nm (95% CI, N = 50 structures); and (c, left panel) Cartoon illustrating the insertion of a single seeded nanotube into a SUV; and (c, right panel) Transmission electron micrographs of the seeded DNA nanotubes and the incorporation of seeded DNA nanotubes into SUVs. Scale bar: 500, 100, and 50 nm.

Pore formation of the DNA nanobarrels spanning the lipid membrane was further confirmed through TEM studies. Consistent with earlier reports, the TEM images clearly show the incorporation of single and multiple DNA seed structures onto small unilamellar vesicles (SUVs) (see Materials and Methods, Figure 2b, and Figures S5, S6a) [33, 38]. Notably, the lower stem portion of the DNA nanotube seed on the lipid vesicles is invisible in the images. We attribute this to lipid patch coverage of this region, to protect from the staining reagent—an effect previously observed for stem and cap‐shaped DNA origami structures, where the lower DNA stem is coated similarly with lipid patches [33]. However, in a few cases, visible penetration of the lower stem portion of the DNA nanotube seed structures into the SUVs is observed (Figure S7a,b). In other instances, the DNA nanostructures appear to be fully immersed within the lipid vesicles, rather than associated with the vesicle exterior. This is likely an artifact of sample preparation, in which vesicles rupture upon deposition onto TEM grids, thereby allowing the encapsulation of DNA nanostructures (Figure S7c,d) [33].

The combination of nanotube seed channels with SUVs indicated that the seeds formed membrane pores, as their barrels spanned the lipid membrane and were consistently oriented perpendicular to it (see Materials and Methods, Figure 2b and Figures S5–S7). This is further validated through orientation plot analysis, which shows that ∼86% of the DNA nanotube seeds fall within 15° of the radial direction, with ∼78% within 10°, reflecting strong perpendicular alignment to the membrane plane (Figure S6a–c). However, lipid coverage and staining limitations prevent TEM from directly resolving the DNA nanotube seeds protruding through the distal end. The perpendicular orientation and conductance measurements support the formation of pores in artificial lipid bilayers.

2.2. Electrical Recordings of the DNA Nanostructures

We next sought to measure ion‐conductance through the nanotube seed channels by conducting electrophysiological measurements in 1 M KCl, 10 mM MgCl₂ buffered with 10 mM MES monohydrate, pH 6.0 using the droplet‐interface bilayer (DIB) technique (Figure 3a) [36, 66, 67, 68]. After 10 min of incubation with DPhPC (1,2‐diphytanoyl‐sn‐glycero‐3‐phosphocholine) lipid vesicles, a transmembrane potential of +50 mV was applied to initiate the insertion of nanotube seed channels into the membrane (see Materials and Methods). The electrical signature was a stable current of 63.30 ± 8.50 pA (N = 5) at +50 mV applied potential, which corresponds to a conductance of 1.26 ± 0.17 nS across a nanotube seed channel (Figure 3b and Figure S8a,c). Applying a negative transmembrane potential of −50 mV induced a current change of similar magnitude (62.32 ± 2.80 pA, N = 5) in the opposite direction (Figure 3c and Figure S8b,d). Multiple observed insertions of the nanotube seed channels into DPhPC bilayers produced roughly similar ion‐current values (Figures S8 and S9). As the two droplets forming a bilayer were separated from one another, we observed stepwise reductions in channel conductance of similar magnitude to the increases observed during insertion, that is, 6 channel closures yielding a total conductance of 5.97 nS (Figure S10). Current levels measured from stable, pre‐inserted single‐nanotube seed channels scaled approximately linearly with the applied transmembrane voltage, indicating ohmic conductance over the tested voltage range (Figure 3f). Similar linear I‐V relationships have been reported for other DNA‐based membrane channels using single‐channel current recordings, including DNA origami channels anchored by cholesterol [33, 37, 38, 42], porphyrin [35, 41], tocopherol [36], ethane‐phosphorothioates [34], and biotin‐streptavidin [36]. Furthermore, the rectification ratio (R) of the DNA nanotube seeds at ±50 mV was found to be 1.03 ± 0.03 (I _+50 mV/–_50 mV), indicative of Ohmic conductance. The rectification ratio remained unchanged even at higher voltages (R = I _+100 mV/I–_100 mV = 1.00 ± 0.03). The combined effect of high KCl and MgCl₂ could strongly screen the negative charges of the DNA backbone, neutralizing the asymmetric charge distribution, resulting in Ohmic behavior of the DNA seeds. The DNA seed is designed to be cylindrical [69], and the observation that the channel exhibits an ohmic conductance pattern is consistent with the ohmic conductance of other DNA nanopores exhibiting cylindrical geometries [33, 34, 35].

FIGURE 3.

Single channel electrical properties of the DNA origami barrel/nanotube seed membrane channels. (a) Schematic showing the insertion of a single DNA nanotube seed channel in a DPhPC lipid bilayer formed by the DIB technique; Insertion of a single DNA nanotube seed channel in a DPhPC lipid bilayer formed by the DIB technique at an applied transmembrane potential of (b) +50 mV; and (c) –50 mV; Representative single‐channel electrical traces recorded from a pre‐inserted DNA nanotube seed channel at an applied transmembrane potential of (d) +150 mV; (e) –150 mV. The upward (d) and downward (e) current spikes arises from the response of the existing channel to changes in the high applied voltage; (f) Current‐voltage (I‐V) relationship measured from stable pre‐inserted single channels (N = 4 independent recordings) and (g) Histogram of the channel conductance steps of the DNA origami seed channels inserted in a DPhPC lipid bilayer formed by the DIB technique at an applied transmembrane potential of +50 mV (N = 362). A Gaussian fit to the conductances is shown as an overlay (1100 ± 255 pS, ± S.D). The electrical measurements were carried out in 1 M KCl, 10 mM MgCl₂ buffered with 10 mM MES monohydrate, pH 6.0.

Additionally, the nanotube seed channels were relatively stable at voltages above 100 mV (150 mV; Figures 3d,e), and under these conditions, we did not observe any significant gating events (i.e., repeated switching between open and closed states). Gating has previously been observed for some DNA origami channels. For instance, the 6‐helical bundle with the cap structure reported by Langecker et al. exhibited significant gating behavior in planar lipid bilayer experiments, and these gating events were attributed to thermal fluctuations of the DNA helices within the channel [33]. Similar effects were also reported in other 6‐helix bundle [41] and 4‐helix duplex [42] origami structures when using the planar lipid bilayer and nanocapillary techniques. In contrast, large DNA origami structures, such as the T‐shaped structure reported by Krishnan et al. did not exhibit significant gating behavior even at higher voltages (+125 mV) in a DIB system [36]. The lack of gating may therefore reflect the large length and rigid architecture of the DNA origami seed as well as the use of DIB for characterization.

A histogram of the measured channel conductances at +50 mV showed a broad distribution of channel conductances. The peak conductance was at 1.10 nS with a standard deviation from a Gaussian fit of 0.25 nS (Figure 3g). As has been observed in other studies [29], there were occasional higher channel conductances, such as 2.07 and 2.98 nS. Such conductances are likely to arise from the simultaneous insertion of multiple nanotube seed channels (Figure S11). As expected, in the absence of DNA‐cholesterol conjugates, no characteristic ion‐current changes were observed when a transmembrane potential was applied across lipid bilayers in the presence of nanotube seeds (Figure S12a). Analogously, no current changes were observed in the presence of only 10 µM cholesterol‐DNA conjugate strand (but the absence of nanotube seeds) (Figure S12b).

The measured conductance of the nanotube seed channels is lower than what would be predicted theoretically if the pore were modeled as a nanocylinder with electrically non‐permeable walls, where the solvent flux controls the conductance (4.91 nS, Note S12) [70]. While the assumptions of this model apply to a significant number of biological protein channels, they are less applicable to DNA channels, whose walls are more likely to be electronically permeable because they are composed of hydrophilic, dynamic, negatively charged groups and have gaps large enough for water to pass through [71, 72]. The assumption that ions move at a constant rate through the channel is also unlikely to be satisfied by negatively charged DNA origami nanopores [73]. Consequently, the predictions of this model have both over‐ and under‐estimated the conductance of other DNA origami channels [34, 35, 38, 41, 42].

Having shown that DNA nanotube seeds could insert into the membrane to create pores with well‐defined conductance, we next asked whether micrometer long‐DNA nanotubes could also act as membrane channels (Figure 1b). DNA nanotubes were grown from seeds by first filter‐purifying the seeds and then adding monomers for assembly over a 34‐h incubation period (see Materials and Methods, and Notes S6, S7). To make it possible to characterize the yield and lengths of nanotubes, monomers were labeled with a small‐molecule fluorophore Cy3 on their central strands (see Note S2 and Figure S3 for sequences). DNA‐cholesterol conjugates were added after nanotubes were grown, and the resulting structures were filter‐purified again to remove excess strands from the solution of nanotube membrane channels (see Materials and Methods, Figure 4h, Note S7, and Figure S13).

FIGURE 4.

Single‐channel electrical properties of the DNA seeded nanotube membrane channels. (a) Diagrammatic representation showing the insertion of a single seeded DNA nanotube channel in a DIB set up; Single and multiple insertions of the DNA seeded nanotube channels inserted into DPhPC lipid bilayer at an applied transmembrane potential of (b) +50; and (c) −50 mV; Representative single‐channel electrical traces recorded from a pre‐inserted DNA seeded nanotube channel at an applied transmembrane potential of (d) +150 mV; (e) −150 mV. The upward (d) and downward (e) current spikes arises from the response of the existing channel to changes in the high applied voltage; (f) I‐V relationship of the DNA origami seeded nanotube channels (N = 4); (g) Histogram of the channel conductance steps of the seeded DNA nanotube channels inserted into a DPhPC lipid bilayer formed by the DIB technique at an applied transmembrane potential of +50 mV (N = 186). A Gaussian fit to the conductances is shown as an overlay (1150 ± 120 pS, ± S.D). The electrical measurements were carried out in 1 M KCl, 10 mM MgCl₂ buffered with 10 mM MES monohydrate, pH 6.0; Fluorescence micrographs of the seeded DNA nanotubes. Nanotubes are labeled with Cy3 (green), nanotube seeds with ATTO488 (blue). Seeded DNA nanotubes (h, top panel) After assembly; and (h bottom panel) After assembly, addition of cholesterol‐DNA conjugate strand and purification. Scale bar: 5 µm.

Fluorescence micrographs showed that 84 ± 3% (95% CI) (N = 678 structures) of DNA nanotube seed pores had nanotubes on them, which was the same fraction as just after the DNA‐cholesterol conjugates were added, 84 ± 1% (95% CI) (N = 467 structures) (Figures S13 and S14a). The average length of the nanotube channels was 2.43 ± 0.10 µm (95% CI) (N = 185 structures) before purification and 2.17 ± 0.10 µm (95% CI) (N = 272 structures) after purification (Figures S13 and S14b), suggesting that long nanotubes remained intact after filter purification.

DNA nanotube channels were designed to insert into membranes at the seed ends of the structure, with the channels extending microns away from the anchoring membrane. Transmission electron micrographs confirmed that nanotubes oriented on membranes with the same relative orientation as DNA nanotube seed pores when inserted into DPhPC SUVs (Figure 2c and Figure S15). The curvatures of the nanotubes appeared qualitatively consistent with the previously measured persistence lengths of DNA nanotubes of 8.7 µm [55].

The conductances of the nanotube channels were measured across a membrane formed by the DIB technique. A Gaussian fit to the distribution of conductances suggested that nanotube channels had an average conductance of 1.10 nS at an applied transmembrane potential of +50 mV, with different insertions producing different conductances clustered around this value (Figures 4b,c and Figures S16–S18). Similar to nanotube seeds, seeded nanotubes did not produce any notable gating events at voltages above 100 mV (150 mV, Figures 4d,e). Control experiments with DNA nanotubes without seeds showed no significant increase in current (see Materials and Methods, Note S8, and Figures S19 and S20). Like the nanotube seed channels, the ion conductance of nanotube channels was also ohmic (Figure 4f). In addition, the rectification ratio was comparable to that of the DNA nanotube seeds (R = I _+50 mV/I–_50 mV = 1.04 ± 0.03). The range of conductances observed for the nanotube channels was analogous to that of the nanotube seeds (Figure 4g, Gaussian fit, 1.15 ± 0.12 nS). This range represents the conductance of nanotube channels of different lengths together with a smaller fraction of seeds devoid of nanotubes, suggesting that the nanotube elongation does not significantly alter the conductance of the seed channel. The analogous conductances observed in nanotubes and nanotube seeds suggest a weak dependency of conductance on channel length across the examined range. Although the standard pore model [70], which was primarily designed for protein ion channels, predicts a reduction in conductance with increasing channel length, its assumptions do not fully apply to DNA and many other synthetic nanopores [74]. DNA nanopores are porous, dynamic, and negatively charged, allowing ions to permeate both the channel lumen and the walls of the DNA [71]. In addition, the lipids in the membrane reorganize around DNA nanopores to adopt a toroidal, pore‐like conformation [60, 61], potentially generating interfacial gaps that allow free ion transport. These additional transport pathways can effectively reduce the contribution of channel length to the overall resistance, leading to the observed deviation from classical behavior. In this interpretation, the measured resistance is dominated by the membrane‐spanning seed/channel region and the DNA–lipid interface rather than by transport through the full micrometer‐long nanotube lumen; ions can access the conducting pathway via hydrated wall gaps near the membrane rather than entering exclusively from the distal nanotube end. Apart from DNA nanotubes, carbon nanotubes have also been reported to have conductances independent of length: nanotubes of lengths between 5 and 15 nm showed a unitary channel conductance of 630 ± 120 pS [74], although here the range of channel lengths is far greater. Electron micrographs of SUVs and nanotubes showed that >500 nm‐long seeded nanotubes incorporated into membranes, suggesting that these measurements likely included seeded nanotubes among the measured pores (Figure 2c and Figure S15). Thus, DNA nanotubes of various lengths exhibit an average conductance in the range of 1.03–1.27 nS, similar to the conductances of DNA seed pores.

We next enquired whether we might ascertain the pathways of ion flow through the seed and nanotube channels that facilitate conductance across the membrane. Ions could either flow from bulk solution into the opening of the nanotube seed channel or nanotube and subsequently traverse the membrane, or move across the DNA walls of the nanotubes or seeds into the channel and then across the membrane. The large gaps between helices in DNA origami structures, resulting from the electrostatic repulsion of DNA's negatively charged backbones, could allow ion transport through the walls. This process has been previously observed in molecular dynamics simulations of other DNA nanopores [37, 71, 75, 76, 77].

The utilization of micron‐length channels to mediate ion transport across multiple membranes (as seen in Figure 1c), would significantly impede conductivity efficiency due to ion leakage through the channel walls in such an extended channel. To measure the ion‐conductance through the channel from end‐to‐end compared to that via the channel walls of the DNA seeds and the nanotubes, we used a channel cap [57] for attachment to the channel's terminus (Figures 1c, 5a,e, and see Note S3 for sequences) that can bind to and obstruct the channel's open end [29]. The cap incorporated staples that pulled together the opposite walls of the cylinder, so inducing a twist in the structure and constricting its opening (Figures 1 and 5a) [29]. The channel cap presented sticky ends capable of binding to the end of either a nanotube or a seed via DNA hybridization to create a stable bond. The sizes of the gaps between the channel cap and the seed or nanotube channel would be expected to be on the order of those in the channel walls. We hypothesized that the channel cap, by constricting the diameter of the channel's opening, would reduce the ability of ions to traverse the membrane, because fewer ions can access the channel's terminus and traverse its length to permeate the membrane. The channel cap should not, in contrast, significantly alter the rate of ion transport across the DNA walls of the channel. That is, ion flux across a capped nanotube seed or nanotube channel should primarily occur via transport through the channel walls (Figure 1).

FIGURE 5.

Design and conductivity of the channel cap‐DNA nanotube seed channels. (a, left) Staple strand map (from CaDNAno) showing the design of the DNA origami channel cap. The sticky ends presented by the adapter strands attach either to the free ends of seeds or to the seeded nanotubes via Watson–Crick hybridization. Pinched staples are those that were modified to reduce the size of the channel opening. (a, right) Map of helices showing how they are connected by pinched staples to reduce the opening size; Transmission electron micrographs of (b) channel caps. Scale bar: 100 and 50 nm. Channel caps have an average length of 62.43 ± 1.21 nm (95% CI, N = 45 structures); (c) Nanotube seeds with attached channel caps inserted into DPhPC lipid vesicles. Scale bar: 100 nm; (d, top panel) Fluorescence micrographs of seed pores containing cholesterol DNA conjugate with attached channel caps after 2 rounds of filter purification at 3000 g for 30 min. Seeds are labeled with ATTO488 (blue) and channel caps are labeled with ATTO647N (red). Scale bar: 2 µm; (d, bottom panel) Transmission electron micrograph of nanotube seeds bound to channel caps. Scale bar: 100 nm. Average length of the seed‐channel cap structure is 132.3 ± 3.2 nm (95% CI, N = 30 structures); (e) Schematic representation showing the insertion of a single DNA origami seed attached to a channel cap in a DPhPC lipid bilayer. Single insertion of the capped nanotube seed channel in a DPhPC bilayer at an applied transmembrane potential of (f) +50 mV; (g) Representative single‐channel electrical traces recorded from a pre‐inserted capped nanotube seed channel at an applied transmembrane potential of –50 mV. The downward (g) current spike arises from the voltage reversal and reflects the response of a pre‐inserted channel to changes in the applied transmembrane potential; (h) I‐V relationship of the capped DNA seed channels (N = 4); and (i) Histogram of the conductance steps of capped DNA seed channels across DPhPC lipid bilayers formed by the DIB technique at an applied transmembrane potential of +50 mV (N = 259). A Gaussian fit to the conductances is shown as an overlay (518 ± 90 pS and 762 ± 46 pS, ± S.D). The electrical measurements were carried out in 1 M KCl, 10 mM MgCl₂ buffered with 10 mM MES monohydrate, pH 6.0.

Channel caps were prepared similarly to the nanotube seeds, that is, using analogous concentrations of staples, the same method of filter purification, and the same annealing schedule (see Materials and Methods and Note S9). Transmission electron micrographs showed that the channel caps folded into cylindrical shapes akin to nanotube seeds, had an average length of 62.43 ± 1.21 nm (95% CI, N = 45 structures) (Figure 5b and Figure S21). Electron micrographs revealed a constriction at one end.

To verify the binding of the channel caps to seeds, we incubated nanotube seed channels with channel caps for 3 h at 32°C, facilitating the binding process (see Materials and Methods and Note S9), followed by sample preparation for electron microscopy [57]. The nanotube seeds possess adapter strands that bind to one facet of the nanotube, while the channel caps contain adapter strands that bind to the opposite facet of the nanotube or to the seed. The six adapter sticky end strands present on the channel caps and on the nanotube seed should facilitate strong binding of the channel cap to the nanotube seed when the two structures are incubated at 32°C (For complete sequences of adapter strands, see Mohammed et al) [55]. Consistent with this design, the electron micrographs revealed that the designed channel caps exhibited strong binding to the nanotube seeds. The mean total length of the bound seed‐cap structures was 132.3 ± 3.2 nm (95% CI, N = 30 structures) (Figure 5d and Figure S22). Like the nanotube seed and seeded nanotubes, nanotube seed‐channel complexes were also able to penetrate the SUVs (Figure 5c and Figure S23). Fluorescence micrographs showed that 91 ± 1% (95% CI, N = 230 structures) of the seeds were bound to channel caps following this incubation (Figure 5d and Figures S24a, and S25).

To characterize the conductance across the nanotube seed pores, whose ends were closed off by channel caps, the channel cap‐seed pore complexes were incubated with DNA‐cholesterol conjugate strands (final concentration 10 µM) for 1 h under ambient conditions, followed by filter purification in the same manner as nanotube seed pores. After purification, the fraction of seed channels bound to channel caps remained unchanged from before purification: 90 ± 1% (95% CI, N = 223 structures) (Figures S24b and S25).

The histogram of conductances generated by the nanotube seed channels with attached channel caps when inserted into the membrane exhibited two ranges of conductance peaks: one between 0.42 and 0.60 nS—approximately half the conductance of the nanotube seed channels—and the other between 0.71 and 0.80 nS at a positive transmembrane potential of +50 mV (Figures 5e–g and Figures S26–S30). There were also some channels with conductances ranging between 0.9 and 1.2 nS. This final population may consist of nanotube seed pores or nanotubes lacking attached channel caps (Figure 5i). Similar to the DNA seed channels and the nanotube seed channels, the seed channels with attached channel caps exhibited an Ohmic conductance pattern (Figure 5h). In addition, the rectification ratio (R) of the nanotube seed‐channel complex further confirms that these channels exhibit an almost symmetrical conductance pattern [R (I _50mV/_I‐50 mV) = ∼1.04]. Similar, ohmic behavior has also been reported for other DNA nanostructures containing lid assemblies. For instance, a short six‐helix DNA barrel with attached lid strands exhibited a strong ohmic pattern [40]. Likewise, a six‐helical light‐activated DNA nanobarrel with a nonirradiated lid also exhibits an ohmic conductance pattern [78]. The observation that all three origami constructs used in the present study display a nearly ohmic conductance pattern (Figure S31) further supports the notion that these structures are cylindrical in shape. Several cylindrical‐shaped DNA origami structures have also exhibited an ohmic current pattern [33, 34, 35]. Furthermore, the high KCl and the MgCl₂ used in the present study may effectively screen the negative surface charges, thereby reducing asymmetric charge distribution within the pore. This charge screening likely renders all three constructs nearly electrically neutral, resulting in ohmic conductance behavior. Thus, the Ohmic conductance pattern observed for these DNA origami constructs appears to depend on both the origami design and the ionic conditions employed in the present study, in which effective surface charge screening causes the pore to behave as a nearly neutral cylindrical channel.

To verify that the binding of the channel caps was responsible for the reduction in observed conductance of the seed pores, we eliminated the adapter strands that facilitate the binding of the channel caps to the seeds to make “control” channel caps (Figure S32a). These control channels were unable to bind to the seed pores, but were added to the membranes along with the seed pores (see Materials and Methods and Notes S5, S11). Fluorescence micrographs showed minimal or no interaction between seeds and control channel caps following their incubation (Figure S32a,c). The conductances measured across lipid bilayers using a mixture of nanotube seed channels and the control channel caps were equivalent to those of the seed channels alone, that is, approximately 1–1.2 nS (Figure S32b). Thus, the binding capability of the seed channel and channel caps is essential for the observed decrease in the channel conductance. These experiments further imply that the pores remain bound to the channel caps in the presence of transmembrane potential. As with seed channels, no gating of seed pore channels bound to channel caps was observed. The lack of gating and the constant conductance of the observed channels would be expected if the nanotube seed channels and channel caps bind irreversibly and do not undergo conformational changes during the timescales of their insertion into the membrane.

The conductances of 0.4 and 0.6 nS produced by seed‐channel cap complexes are comparable to those of the 6‐helix DNA bundle constructed by Mendoza et al. [78]. Their DNA helical nanopore produces a conductance of about 0.3 nS, featuring a lock strand attached to the barrel structure. The introduction of the key strands results in an increased conductance of 0.5 nS in the DNA channel, presumably due to the enlargement of the channel aperture. Qualitatively similar conductance changes are also observed in a much shorter six‐helix bundle nanopore [40] when a lock strand binds to the channel entrance, causing a reduction in ion flow. For instance, the nanopore in the open state has a total mean conductance of 1.34 nS; the addition of lock strands results in partial pore blockade with an ionic conductance of 0.66 nS. Conductance modulation was also documented in light‐activated DNA nanopore systems [79]. For instance, a 6‐strand DNA barrel without a lid has an open channel conductance on the same order as that of the DNA origami seed studied here, that is, ∼1.10 nS. However, the addition of lid strands shifts the conductance to a partially open state, with a mean conductance of ∼0.75 nS. A ∼32% reduction in the total open pore current is observed even when the channel is activated, due to the lid remaining tethered to the pore. This behavior provides qualitative precedent for the observation that adding the channel cap reduces but does not eliminate conductance, leaving residual current of ∼45% of the open‐pore value that may arise from incomplete cap sealing, DNA‐wall gaps, and/or DNA–lipid interfacial conduction [60, 61, 71].

Here we demonstrate that the binding of a channel cap to the seed channel, which decreases the channel's diameter, also results in reduced channel conductance [51]. The channel cap's reduction of the conductance by approximately a factor of two suggests the presence of ionic flow through the walls of the DNA channel [72, 77]. This is consistent with previous experimental and theoretical studies demonstrating that the lipid molecules can rearrange into a toroidal pore‐like structure around the DNA seed nanopore, enabling ion conduction along the DNA–lipid interface and preventing the complete blockage of ion flow [60, 61]. Additionally, the mean and standard deviation of the conductances of open DNA nanotube channels, which have an average length of 2.43 ± 0.10 µm (95% CI, N = 185 structures, Figure 4h and Figure S14), were similar to the mean and standard deviation of the conductance of the nanotube seeds, which are monodisperse with a length of 64.70 ± 2.12 nm (Figure 2a and Figure S4a,d). This observation contradicts the theoretical model, which predicts a significant decrease in conductance with increasing channel length due to ion flow from one end to the other. Thus, the similarity in conductances between the nanotube seed and the nanotube channel also suggests that a significant amount of ion flow occurs through the channel walls of both the seed and the nanotube.

We next used the channel cap to characterize the degree of ion flow through the walls of the nanotubes, thereby determining the extent of ion flow from one end of the nanotube through the membrane to the other end. The channel cap can bind to both a nanotube seed and the end of a DNA nanotube, as the free end of a DNA nanotube presents the same sticky ends to those of its nanotube seed templates (Figure 1). The channel‐capped nanotubes were constructed by first assembling seeded nanotubes using the same method as in previous experiments and then incubating the nanotubes with the channel caps at 32°C for 3 h (see Materials and Methods, Figure 6b). The nanotubes with the channel caps were further incubated with the DNA‐cholesterol conjugate for 1 h under ambient conditions and were then centrifuged at 3000 g for 30 min for 2 to 3 times using the 1X TAEM buffer supplemented with 12.5 mM magnesium chloride. Post filter purification, 82 ± 3% (95% CI) of the structures in fluorescence micrographs were identified as seeded nanotubes with attached channel caps, 9 ± 4% (95% CI) were DNA seed structures with attached channel caps, while the remaining 3 ± 2% (95% CI) and 4 ± 3% (95% CI) comprised seeds and seeded tubes lacking caps (Figure 6b and Figure S33a). Thus, >90% of structures contained channel caps. The average length of the capped nanotubes prior to purification was 1.87 ± 0.10 µm (95% CI, N = 108 structures), while post‐purification, it was 1.57 ± 0.20 (95% CI, N = 105 structures). This indicates that purification did not significantly alter the lengths of the DNA nanotube structures (Figure S33b).

FIGURE 6.

Conductivity across capped DNA nanotube membrane channels. (a) Schematic representation showing the insertion of a single capped DNA nanotube channel in a DPhPC lipid bilayer; Fluorescence micrographs of capped nanotubes (b, top panel) before cholesterol‐DNA conjugates were added and (b, bottom panel) after cholesterol‐DNA conjugates were added and the structures were purified by 2 rounds of centrifugation at 3000 g for 30 min. Seeds are labeled with ATTO488 (blue), channel caps are labeled with ATTO647N (red) and nanotubes are labeled with Cy3 (green). Scale bars: 5 µm. Electrical signatures showing different conductance patterns exhibited by capped seeded DNA nanotube channel in 1 M KCl, 10 mM MgCl₂, buffered with 10 mM MES monohydrate, pH 6.0 at an applied transmembrane potential of (c,d) +50 mV; (e) −50 mV. The electrical traces shown in (d) represent the pre‐inserted capped seeded DNA nanotube channel recorded at +50 mV; and (f) Histogram of the conductance steps of capped DNA nanotube channels in a DPhPC lipid bilayers formed by the DIB technique at an applied transmembrane potential of +50 mV (N = 144). A Gaussian fit to the conductances is shown as an overlay of three different forms of conductances in the range of 396 ± 45 pS, 697 ± 70 pS, and 1158 ± 172 pS.

The resulting nanotube channels with caps produced changes in conductance across DPhPC bilayers, suggesting that they could insert into lipid bilayers (Figures 6a, c–e and Figures S34, S35). The structures exhibited a significantly wider range of conductances than other structures (Figure 6f). We observed three distinct forms of conductance, namely, 396 ± 45 pS, 697 ± 70 pS and 1158 ± 172 pS, although the limited sample size complicates confidence in the precise location and quantity of these peaks. The existence of multiple peaks is consistent with the fact that a combination of structures (seed pores and nanotube channels with and without caps) that could insert into the membrane. The peak (1158 ± 172 pS) with the same conductance as those of open nanotubes accounts for at least 30% of the measurements, even though <10% of the structures being characterized are uncapped nanotube channels. This suggests that a fraction of capped nanotube channels had similar conductances to the open ones. Perhaps these nanotubes have rough end facets, which would create gaps in their binding interface with channel caps. These holes would be the size of nanotube monomers (about 13 × 4 nm) so that a channel cap would not effectively reduce the size of a channel's opening. The other peaks had conductance magnitudes comparable to those of the nanotube seed‐channel cap complexes. These results suggest that, in at least some cases, ion transport proceeds from one end of a micron‐long nanotube to the other across the membrane. Furthermore, due to the coexistence of multiple components in solution (i.e., DNA seed, seeded nanotube, DNA seed‐channel cap complex, and seeded nanotube‐channel cap complex) and the random insertion pattern of the DNA nanostructures in the DIB platform, it was not possible to reliably isolate and record I‐V characteristics for a single capped seeded nanotube in our current electrophysiology platform.

3. Conclusions

DNA nanostructures, which can be designed to have specific shapes, combined with hydrophobic molecules, can be used to form a wide range of transmembrane channels that enable specific targeting of genetic and signaling molecules to establish molecular communication [33, 34, 35, 36, 37, 38, 39, 40, 41, 42]. Since existing DNA origami channels are shorter than 100 nm [80], designing micrometer‐long channels could establish connections between nonadjacent cells and compartments for communications and chemical exchange [29]. In addition, such micrometer‐long channels have tremendous potential to facilitate internal molecular transport across neighboring compartments in a synthetic tissue setting.

Here, we explore the use of DNA nanotechnology to form nanofluidic channels that transport ions through micrometer‐long structures when an electrical potential is applied. Furthermore, understanding the channel properties of such longer nanochannels opens new avenues, potentially leading to synthetic cell–cell conduits similar to bacterial pili or gap junctions in eukaryotic cells. The constructed channels are ohmic in character and have conductances similar to analogous channels published in previous reports [33, 34, 35, 38, 40, 41]. The reduced conductance of both nanotube seed and nanotube channels when bound to channel caps as compared to their uncapped counterparts suggests that the ions transported through these channels migrated from one cylindrical end of the channel to the other and also through the channel walls. These findings support the idea that DNA nanostructures have the potential to guide ion flow over lengths of microns. Additionally, they provide methods to quantitatively evaluate factors that might reduce transport through walls, such as coating the surface of the nanotube. The ability to readily assemble DNA nanotubes under physiological conditions [56] or to target or orient the terminals at the sites of particular biomolecular labels [55] implies that these structures might be directed to form interconnects or channels at specific locations on cells or synthetic devices. The precise control of ion flow by nanotubes, as determined from conductivity measurements, implies a method for constructing regenerative or dynamic bioelectronic devices with well‐defined circuit topologies. Coating such structures with lipids [81, 82], proteins [83, 84, 85], polycations [86]/peptoids [87], block co‐polymers [88], or silica shells [89] could enable ion‐transport through channels without any loss of ions across the channel walls. Ion transport within insulated biomolecular channels is a key process of charge transport in living systems. The ability to recapitulate this transport process in engineered systems could be a valuable method for building synthetic, self‐assembled circuits that efficiently transmit information or chemical signals via charge transport.

4. Materials and Methods

4.1. Materials

4.1.1. DNA Strands

The sequences for nanotube seeds and nanotubes are provided in the Supporting Information (Notes S1, S2 and Figures S1–S3). All the DNA strands and the DNA‐cholesterol conjugate strand used in this work were purchased from Integrated DNA Technologies (IDT, Iowa, USA), except the M13mp18 scaffold strand which was purchased from Bayou Biolabs (Los Angeles, USA). The strands that comprise the DNA nanotube monomers and the DNA‐cholesterol conjugate strand were HPLC purified; the adapter strands on the nanotube seeds and channel caps that act as binding sites for monomers or nanotubes were PAGE purified. The staple strands for the nanotube seed and channel cap were not purified. The concentrations of the DNA tiles, adapters and the cholesterol strands were determined by measuring the absorbance at 260 nm using a spectrophotometer (Eppendorf Model 6131, Hamburg, Germany) and the Beer‐Lambert law using the extinction coefficients provided by IDT.

4.1.2. Reagents

Potassium chloride (Alfa Aesar, Massachusetts, USA. Product No: 10220196), 4‐morpholineethanesulfonic acid (MES) monohydrate (Alfa Aesar, Massachusetts, USA. Product No: 10191845), calcium chloride (Sigma Aldrich, USA. Product No: SLBW7125), magnesium chloride (Sigma Aldrich, USA. Product No: SLB3227V), hexadecane (Sigma Aldrich, USA. Product No: SHBK6702), TAE buffer (Ultrapure DNA typing grade, Invitrogen, USA), sodium hypochlorite (Sigma Aldrich, USA), agarose tablets (EZ Pack Agarose, Benchmark Scientific, Inc, New Jersey, USA. Product No: A2501), uranyl acetate (Electron Microscopy Sciences, Philadelphia, USA. Product No: 22400), bovine serum albumin (Sigma Aldrich, Product No: SLBT2162), and Amicon Ultra centrifugal filters 100 kDa (100 K, Merck Millipore Ltd., Darmstadt, Germany).

4.1.3. Sample Preparation

DNA nanotube seeds and nanotube seed channels: Seeds were annealed as described in Mohammed et al., 2013 [51]. Seeds used to grow nanotubes and seeds used as channels were annealed in 40 mM Tris‐acetate, 1 mM EDTA supplemented with 12.5 mM magnesium acetate, pH 8.0 (1X TAEM buffer). To make DNA seed channels, the DNA‐cholesterol conjugate strand (3ʹ nanotube seed_Chol.DNA, see Note S1 for sequences) dissolved in MilliQ water was heated to 50°C in an Eppendorf tube for 15 min to dissolve agglomerates. The DNA conjugate strand was then added to achieve a final concentration of 10 µM in 50 µL of unpurified DNA seed origami solution. The solution was then incubated for 1 h under ambient conditions. For electrical measurements, after incubation of the nanotube seeds with the DNA‐cholesterol conjugate strand, the DNA seed structures were then purified from excess staples and unhybridized cholesterol conjugates via spin filtration using 100 kDa MWCO filters (Amicon ultra centrifugal filters – 100 K) in 350 µL 1X TAEM buffer. The filtration process was repeated at 3000 g for 30 min at least 3–5 times (Notes S4, S6, and S7).

For control experiments where no DNA‐cholesterol conjugate strand was added to the DNA nanotube seeds, purified seeds were directly mixed with lipid vesicles prior to electrophysiology experiments. The conductance of the DNA‐cholesterol conjugate strand alone (10 µM) was also measured as a control, by directly mixing the cholesterol strands with the lipid vesicles and then performing the electrical measurements.

DNA nanotubes: The strands for (a) the nanotube seeds and (b) 40 nM of the nanotube tiles were annealed in 1X TAEM buffer in two separate tubes. The tubes were initially heated at 90°C for 5 min to melt all DNA, cooled from 90 to 45°C at 1°C/min and then incubated at 45°C for 1 h to allow the formation of seeds and nanotube tiles. The samples were then slowly cooled from 45 to 32°C at 0.1°C/min to a final temperature of 32°C. The annealed seed structures were then spin filtered using 100 kDa MWCO filters (Amicon ultra centrifugal filters – 100 K). The spin filtration process was repeated at 3000 g for 4 min at least 4 times (no addition of 3ʹ nanotube seed_Chol.DNA). The purified seeds were then added to a final concentration of 18 pM to a mixture of 40 nM of nanotube tiles mix at 32°C. The reaction mixture was then incubated at 32°C for 18–24 h or longer to allow seeded nanotubes to grow [51]. To make DNA nanotube channels, DNA‐cholesterol conjugate strand solution was heated as described for nanotube seed channel preparation, then added to a final concentration of 10 µM to a solution of 18 pM of seeded DNA nanotubes. DNA nanotube channels were filter purified as described for nanotube seed channels (Note S7). Before use in lipid bilayer experiments, the nanotubes were imaged using fluorescence microscopy to confirm the percentage of nanotube seeds with tubes and the approximate tube length. For control experiments with the DNA nanotubes alone (no seeds attached), 250 nM of each strand of the DNA tiles were annealed following the same protocol as that of the seeded tubes and the reaction mixture was kept at 32°C for 24 h. The DNA tubes were further filter purified at 3000 g for 4 min four times and directly mixed with DPhPC vesicles prior to electrophysiology experiments (Note S8).

Channel caps and nanotube seed channels with attached channel caps: Channel caps were annealed and filter purified following the methods for nanotube seed annealing and purification (Note S9). Nanotube seed channels with attached channel caps were prepared by mixing nanotube seeds and channel caps at a stoichiometric ratio of 1:3. The mixture was then incubated for approximately 3 h at 32°C. To the incubated sample, preheated DNA‐cholesterol conjugate strands were added to a final concentration of 10 µM and further incubated for 1 h under ambient conditions. The resulting structures were filter purified for 30 min at 3000 g two to three times.

Shuttered DNA nanotube channels: Seeded DNA nanotubes were prepared and filter purified as described above. Seeded nanotubes were mixed with channel caps at the same stoichiometric ratio of nanotube seeds:channel caps as for seed channels with channel caps and were incubated with DNA‐cholesterol conjugate strand and then filter purified in the same manner as nanotube seed channels with channel caps (Note S10).

Transmission electron microscopy (TEM): Formvar‐coated carbon grids (Electron Microscopy Sciences, Philadelphia, USA) were used for imaging DNA nanostructures. Prior to imaging the DNA nanostructures, the grids were glow discharged for 40 s. To image nanotube seeds or channel caps, 5–10 µL of a solution containing the seeds (1.2 nM) or channel caps (0.4 nM) was added directly onto the grid, left for 10 min and then washed with MilliQ water. To image the nanotube seeds bound to channel caps, the nanotube seeds and the channel caps were prepared and then preincubated together at 32°C for about 3 h prior to imaging under TEM. The grids were then stained with 50 µL aqueous uranyl acetate solution (2% v/v) (Electron Microscopy Sciences, Philadelphia, USA. Product No: 22400) for 5 min. The excess staining solution was carefully removed using filter paper. The grids were then dried in a vacuum chamber for 3 h prior to imaging using a field emission Tecnai‐12 TEM and with an operating voltage of 100 kV.

Preparation of SUVs: 1,2‐diphytanoyl‐sn‐glycero‐3‐phosphocholine (DPhPC, Avanti Polar Lipids, USA) was used to form SUVs. Briefly, the lipids were dissolved in chloroform in a glass vial at a concentration of 5 mg/mL. The glass vial was kept under nitrogen flow for 20 min to completely evaporate the chloroform. The lipid film that formed was then kept under vacuum in a vacuum chamber for 24 h. The lipid film was then hydrated by resuspending the sample mixture in 1 mL of 150 mM potassium chloride in water. SUVs were then formed by sonicating the solution in a bath sonicator (Bransonic Ultrasonic Cleaner (1510R‐DTH)) for 2 h until the solution turned completely transparent. Prior to imaging, the nanotube seed channels or the nanotube channels or the nanotube seed‐channel complexes with the SUVs, the SUVs were diluted 100 times with 150 mM of potassium chloride buffer. The diluted SUVs were then mixed either with the DNA seed pore or the seeded nanotube pore solution or the nanotube seed‐channel complex solution prepared as described above at a volumetric ratio of 1:3.

Electron microscopy imaging of the nanotube seeds and DNA nanotubes with SUVs: Formvar coated carbon grids (Electron Microscopy Sciences, Philadelphia, USA) were used for imaging the interaction of DNA nanostructures with SUVs. Prior to imaging, the grids were glow discharged for 40 s. To characterize the interaction of nanotube seeds or nanotubes or the nanotube seed‐channel complex with SUVs via TEM, 10 µL of a solution containing the DNA nanostructure samples with SUVs in the ratio 1:3 was added directly onto a grid and left for 10 min, then washed with MilliQ water. The grids were then stained with 50 µL aqueous uranyl acetate solution (2% v/v). The excess staining solution was carefully removed using filter paper. The grids were then dried in a vacuum chamber for 3 h prior to imaging using a field emission Tecnai‐12 TEM and with an operating voltage of 100 kV.

Preparation of phospholipid vesicles for electrical measurements: DPhPC was dissolved in chloroform at a concentration of 5 mg/mL in a glass vial. The glass vial was kept under vacuum for 16 h to allow the chloroform to evaporate. A stream of nitrogen was then flowed through the sample to remove the residual solvent. The lipid film that formed was then resuspended in an aqueous solution containing 1 M potassium chloride, 10 mM MgCl_2, and 10 mM MES monohydrate at pH 6.0. The solution was freeze‐thawed at least 6 times under liquid nitrogen and then extruded 21 times using an Avanti pressure extruder with Whatman filter paper of size 0.1 µm (Avanti Polar Lipids Alabama, USA. Product No: SKU 640005‐1Ea). The resulting transparent solution was then used for lipid bilayer experiments.

Single‐channel electrical recordings: All electrical measurements were performed using the DIB technique [66, 67, 68]. The aqueous phase of the droplets consisted of phospholipid vesicles prepared using 1 M KCl, 10 mM MgCl₂, 10 mM MES monohydrate, pH 6.0; hexadecane acted as the oil phase. Briefly, one end of the silver wire (diameter—125 µm, WPI, USA) was melted over a fire flame to create a ball end. The ball end of the silver wire was then bleached for 30 min using 20% sodium hypochlorite solution. This is followed by dipping the ball end of the silver wire into a solution of 1% agarose dissolved in water. To avoid dehydration, the agarose‐coated silver wire was then placed immediately in the chamber containing hexadecane. Approximately 0.2 µL droplet of the lipid vesicle solution was added to agarose‐coated region of the silver electrodes. The two lipid droplets residing in the hexadecane oil phase led to the formation of a stable monolayer. After 15–20 min, the electrodes were then brought in close proximity to allow the lipid droplets to come into contact with each other. Capacitance was measured by monitoring the square wave output current upon applying a triangular wave input voltage (10 mV/10 Hz frequency) across the bilayer. DNA nanostructures were then added to one of the droplets and a transmembrane potential was applied to induce DNA nanopore insertion. For both channel‐containing samples and the corresponding negative controls, current was monitored for approximately 20 min after DNA addition under the applied potential; control recordings were scored as showing no insertion when no conductance steps above 10 pS were observed during this interval. When insertion did occur, it generally occurred within 5 min. Electrical recording was performed using an Axopatch 200B and the Digidata 1440A (Molecular Devices, USA). The signals were low‐pass filtered at 1 kHz using the built‐in lBessel filter of an Axopatch 200B, digitized at 50 kHz and directly saved into computer memory. The conductance data was analyzed using Clampfit 10.7 software and the graphs were plotted using the Origin 2024b.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: smll74625‐sup‐0001‐SuppMat.DOC.

Supporting File 2: smll74625‐sup‐0002‐DataFile.zip.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.